From ba59ced88cfba0568818f52c7a8d1b55633fe788 Mon Sep 17 00:00:00 2001 From: "Samuel J. Greear" Date: Tue, 11 Oct 2011 00:13:47 -0600 Subject: [PATCH] Sync DRM code with FreeBSD trunk rev 220979 --- sys/dev/drm/ati_pcigart.c | 11 +- sys/dev/drm/drm/Makefile | 3 + sys/dev/drm/drmP.h | 100 ++- sys/dev/drm/drm_bufs.c | 58 +- sys/dev/drm/drm_context.c | 2 +- sys/dev/drm/drm_drv.c | 14 +- sys/dev/drm/drm_fops.c | 4 +- sys/dev/drm/drm_hashtab.c | 219 +++++ sys/dev/drm/drm_hashtab.h | 67 ++ sys/dev/drm/drm_irq.c | 2 + sys/dev/drm/drm_linux_list.h | 36 +- sys/dev/drm/drm_memory.c | 6 +- sys/dev/drm/drm_mm.c | 366 ++++++++ sys/dev/drm/drm_mm.h | 99 +++ sys/dev/drm/drm_pciids.h | 5 +- sys/dev/drm/drm_scatter.c | 143 +--- sys/dev/drm/drm_sman.c | 351 ++++++++ sys/dev/drm/drm_sman.h | 180 ++++ sys/dev/drm/drm_sysctl.c | 17 +- sys/dev/drm/drm_vm.c | 49 +- sys/dev/drm/i915_dma.c | 14 +- sys/dev/drm/i915_drv.c | 1 + sys/dev/drm/i915_drv.h | 4 +- sys/dev/drm/mach64_dma.c | 8 +- sys/dev/drm/mach64_state.c | 3 +- sys/dev/drm/mga_dma.c | 22 +- sys/dev/drm/mga_warp.c | 4 +- sys/dev/drm/r128_cce.c | 24 +- sys/dev/drm/r128_state.c | 4 +- sys/dev/drm/r600_blit.c | 1998 ++++++++++++++++++++++++++++++++++++++++++ sys/dev/drm/r600_cp.c | 195 ++++- sys/dev/drm/radeon/Makefile | 4 +- sys/dev/drm/radeon_cp.c | 109 +-- sys/dev/drm/radeon_cs.c | 856 ++++++++++++++++++ sys/dev/drm/radeon_drm.h | 24 + sys/dev/drm/radeon_drv.h | 122 +++- sys/dev/drm/radeon_irq.c | 15 + sys/dev/drm/radeon_state.c | 39 +- sys/dev/drm/savage_bci.c | 28 +- sys/dev/drm/via_3d_reg.h | 1653 ++++++++++++++++++++++++++++++++++ sys/dev/drm/via_dma.c | 767 ++++++++++++++++ sys/dev/drm/via_dmablit.c | 801 +++++++++++++++++ sys/dev/drm/via_dmablit.h | 140 +++ sys/dev/drm/via_drm.h | 279 ++++++ sys/dev/drm/via_drv.c | 123 +++ sys/dev/drm/via_drv.h | 161 ++++ sys/dev/drm/via_irq.c | 392 +++++++++ sys/dev/drm/via_map.c | 137 +++ sys/dev/drm/via_mm.c | 180 ++++ sys/dev/drm/via_verifier.c | 1115 +++++++++++++++++++++++ sys/dev/drm/via_verifier.h | 65 ++ sys/dev/drm/via_video.c | 96 ++ 52 files changed, 10721 insertions(+), 394 deletions(-) create mode 100644 sys/dev/drm/drm_hashtab.c create mode 100644 sys/dev/drm/drm_hashtab.h create mode 100644 sys/dev/drm/drm_mm.c create mode 100644 sys/dev/drm/drm_mm.h create mode 100644 sys/dev/drm/drm_sman.c create mode 100644 sys/dev/drm/drm_sman.h create mode 100644 sys/dev/drm/r600_blit.c create mode 100644 sys/dev/drm/radeon_cs.c create mode 100644 sys/dev/drm/via_3d_reg.h create mode 100644 sys/dev/drm/via_dma.c create mode 100644 sys/dev/drm/via_dmablit.c create mode 100644 sys/dev/drm/via_dmablit.h create mode 100644 sys/dev/drm/via_drm.h create mode 100644 sys/dev/drm/via_drv.c create mode 100644 sys/dev/drm/via_drv.h create mode 100644 sys/dev/drm/via_irq.c create mode 100644 sys/dev/drm/via_map.c create mode 100644 sys/dev/drm/via_mm.c create mode 100644 sys/dev/drm/via_verifier.c create mode 100644 sys/dev/drm/via_verifier.h create mode 100644 sys/dev/drm/via_video.c diff --git a/sys/dev/drm/ati_pcigart.c b/sys/dev/drm/ati_pcigart.c index e75eafd..cabe700 100644 --- a/sys/dev/drm/ati_pcigart.c +++ b/sys/dev/drm/ati_pcigart.c @@ -36,8 +36,9 @@ #define ATI_PCIGART_PAGE_SIZE 4096 /* PCI GART page size */ #define ATI_PCIGART_PAGE_MASK (~(ATI_PCIGART_PAGE_SIZE-1)) -#define ATI_PCIE_WRITE 0x4 -#define ATI_PCIE_READ 0x8 +#define ATI_GART_NOSNOOP 0x1 +#define ATI_GART_WRITE 0x4 +#define ATI_GART_READ 0x8 static void drm_ati_alloc_pcigart_table_cb(void *arg, bus_dma_segment_t *segs, @@ -196,13 +197,15 @@ drm_ati_pcigart_init(struct drm_device *dev, case DRM_ATI_GART_IGP: page_base |= (upper_32_bits(entry_addr) & 0xff) << 4; - page_base |= 0xc; + page_base |= ATI_GART_READ | ATI_GART_WRITE; + page_base |= ATI_GART_NOSNOOP; break; case DRM_ATI_GART_PCIE: page_base >>= 8; page_base |= (upper_32_bits(entry_addr) & 0xff) << 24; - page_base |= ATI_PCIE_READ | ATI_PCIE_WRITE; + page_base |= ATI_GART_READ | ATI_GART_WRITE; + page_base |= ATI_GART_NOSNOOP; break; default: case DRM_ATI_GART_PCI: diff --git a/sys/dev/drm/drm/Makefile b/sys/dev/drm/drm/Makefile index fc7670b..f9052ad 100644 --- a/sys/dev/drm/drm/Makefile +++ b/sys/dev/drm/drm/Makefile @@ -10,12 +10,15 @@ SRCS = \ drm_drawable.c \ drm_drv.c \ drm_fops.c \ + drm_hashtab.c \ drm_ioctl.c \ drm_irq.c \ drm_lock.c \ drm_memory.c \ + drm_mm.c \ drm_pci.c \ drm_scatter.c \ + drm_sman.c \ drm_sysctl.c \ drm_vm.c diff --git a/sys/dev/drm/drmP.h b/sys/dev/drm/drmP.h index 086258b..dcee1f5 100644 --- a/sys/dev/drm/drmP.h +++ b/sys/dev/drm/drmP.h @@ -55,6 +55,7 @@ struct drm_file; #include #include #include +#include #include #include #include @@ -62,7 +63,11 @@ struct drm_file; #include #include #include +#include #include +#include +#include +#include #include #include #include @@ -84,9 +89,9 @@ struct drm_file; #include #include "dev/drm/drm.h" -#include "dev/drm/drm_linux_list.h" #include "dev/drm/drm_atomic.h" #include "dev/drm/drm_internal.h" +#include "dev/drm/drm_linux_list.h" #include #ifdef DRM_DEBUG @@ -138,8 +143,12 @@ MALLOC_DECLARE(DRM_MEM_MAPPINGS); MALLOC_DECLARE(DRM_MEM_BUFLISTS); MALLOC_DECLARE(DRM_MEM_AGPLISTS); MALLOC_DECLARE(DRM_MEM_CTXBITMAP); -MALLOC_DECLARE(DRM_MEM_SGLISTS); MALLOC_DECLARE(DRM_MEM_DRAWABLE); +MALLOC_DECLARE(DRM_MEM_MM); +MALLOC_DECLARE(DRM_MEM_HASHTAB); +MALLOC_DECLARE(DRM_MEM_SGLISTS); + +SYSCTL_DECL(_hw_drm); #define DRM_MAX_CTXBITMAP (PAGE_SIZE * 8) @@ -184,6 +193,11 @@ typedef void irqreturn_t; #define IRQ_HANDLED /* nothing */ #define IRQ_NONE /* nothing */ +#define unlikely(x) __builtin_expect(!!(x), 0) +#define container_of(ptr, type, member) ({ \ + __typeof( ((type *)0)->member ) *__mptr = (ptr); \ + (type *)( (char *)__mptr - offsetof(type,member) );}) + enum { DRM_IS_NOT_AGP, DRM_IS_AGP, @@ -200,7 +214,7 @@ enum { #define DRM_MTRR_WC MDF_WRITECOMBINE #define jiffies ticks -typedef unsigned long dma_addr_t; +typedef vm_paddr_t dma_addr_t; typedef u_int64_t u64; typedef u_int32_t u32; typedef u_int16_t u16; @@ -210,41 +224,38 @@ typedef u_int8_t u8; * DRM_WRITEMEMORYBARRIER() prevents reordering of writes. * DRM_MEMORYBARRIER() prevents reordering of reads and writes. */ -#if defined(__i386__) -#define DRM_READMEMORYBARRIER() __asm __volatile( \ - "lock; addl $0,0(%%esp)" : : : "memory"); -#define DRM_WRITEMEMORYBARRIER() __asm __volatile("" : : : "memory"); -#define DRM_MEMORYBARRIER() __asm __volatile( \ - "lock; addl $0,0(%%esp)" : : : "memory"); -#elif defined(__alpha__) -#define DRM_READMEMORYBARRIER() alpha_mb(); -#define DRM_WRITEMEMORYBARRIER() alpha_wmb(); -#define DRM_MEMORYBARRIER() alpha_mb(); -#elif defined(__x86_64__) -#define DRM_READMEMORYBARRIER() __asm __volatile( \ - "lock; addl $0,0(%%rsp)" : : : "memory"); -#define DRM_WRITEMEMORYBARRIER() __asm __volatile("" : : : "memory"); -#define DRM_MEMORYBARRIER() __asm __volatile( \ - "lock; addl $0,0(%%rsp)" : : : "memory"); +#if (__FreeBSD_version < 800000) || defined(__DragonFly__) +#if defined(__i386__) || defined(__amd64__) +#define mb() __asm volatile("mfence" ::: "memory") +#define wmb() __asm volatile("sfence" ::: "memory") +#define rmb() __asm volatile("lfence" ::: "memory") +#else +#define mb() +#define rmb() +#define wmb() #endif +#endif /*__FreeBSD_version < 800000 || __DragonFly__ */ +#define DRM_READMEMORYBARRIER() rmb() +#define DRM_WRITEMEMORYBARRIER() wmb() +#define DRM_MEMORYBARRIER() mb() #define DRM_READ8(map, offset) \ - *(volatile u_int8_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int8_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) #define DRM_READ16(map, offset) \ - *(volatile u_int16_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int16_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) #define DRM_READ32(map, offset) \ - *(volatile u_int32_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int32_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) #define DRM_WRITE8(map, offset, val) \ - *(volatile u_int8_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int8_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) = val #define DRM_WRITE16(map, offset, val) \ - *(volatile u_int16_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int16_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) = val #define DRM_WRITE32(map, offset, val) \ - *(volatile u_int32_t *)(((vm_offset_t)(map)->handle) + \ + *(volatile u_int32_t *)(((vm_offset_t)(map)->virtual) + \ (vm_offset_t)(offset)) = val #define DRM_VERIFYAREA_READ( uaddr, size ) \ @@ -423,7 +434,6 @@ struct drm_file { struct drm_device *dev; int authenticated; int master; - int minor; pid_t pid; uid_t uid; int refs; @@ -484,25 +494,26 @@ typedef struct drm_agp_head { } drm_agp_head_t; typedef struct drm_sg_mem { - unsigned long handle; - void *virtual; - int pages; - dma_addr_t *busaddr; - struct drm_dma_handle *dmah; /* Handle to PCI memory */ + vm_offset_t vaddr; + vm_paddr_t *busaddr; + vm_pindex_t pages; } drm_sg_mem_t; +#define DRM_MAP_HANDLE_BITS (sizeof(void *) == 4 ? 4 : 24) +#define DRM_MAP_HANDLE_SHIFT (sizeof(void *) * 8 - DRM_MAP_HANDLE_BITS) typedef TAILQ_HEAD(drm_map_list, drm_local_map) drm_map_list_t; typedef struct drm_local_map { - unsigned long offset; /* Physical address (0 for SAREA)*/ - unsigned long size; /* Physical size (bytes) */ - enum drm_map_type type; /* Type of memory mapped */ - enum drm_map_flags flags; /* Flags */ - void *handle; /* User-space: "Handle" to pass to mmap */ - /* Kernel-space: kernel-virtual address */ - int mtrr; /* Boolean: MTRR used */ - /* Private data */ - int rid; /* PCI resource ID for bus_space */ + unsigned long offset; /* Physical address (0 for SAREA) */ + unsigned long size; /* Physical size (bytes) */ + enum drm_map_type type; /* Type of memory mapped */ + enum drm_map_flags flags; /* Flags */ + void *handle; /* User-space: "Handle" to pass to mmap */ + /* Kernel-space: kernel-virtual address */ + int mtrr; /* Boolean: MTRR used */ + /* Private data */ + int rid; /* PCI resource ID for bus_space */ + void *virtual; /* Kernel-space: kernel-virtual address */ struct resource *bsr; bus_space_tag_t bst; bus_space_handle_t bsh; @@ -653,6 +664,7 @@ struct drm_device { /* Linked list of mappable regions. Protected by dev_lock */ drm_map_list_t maplist; + unsigned int map_seq; drm_local_map_t **context_sareas; int max_context; @@ -984,17 +996,17 @@ drm_free(void *pt, size_t size, struct malloc_type *area) static __inline__ void drm_core_ioremap_wc(struct drm_local_map *map, struct drm_device *dev) { - map->handle = drm_ioremap_wc(dev, map); + map->virtual = drm_ioremap_wc(dev, map); } static __inline__ void drm_core_ioremap(struct drm_local_map *map, struct drm_device *dev) { - map->handle = drm_ioremap(dev, map); + map->virtual = drm_ioremap(dev, map); } static __inline__ void drm_core_ioremapfree(struct drm_local_map *map, struct drm_device *dev) { - if ( map->handle && map->size ) + if ( map->virtual && map->size ) drm_ioremapfree(map); } @@ -1005,7 +1017,7 @@ drm_core_findmap(struct drm_device *dev, unsigned long offset) DRM_SPINLOCK_ASSERT(&dev->dev_lock); TAILQ_FOREACH(map, &dev->maplist, link) { - if (map->offset == offset) + if (offset == (unsigned long)map->handle) return map; } return NULL; diff --git a/sys/dev/drm/drm_bufs.c b/sys/dev/drm/drm_bufs.c index b36dafe..0de6f9a 100644 --- a/sys/dev/drm/drm_bufs.c +++ b/sys/dev/drm/drm_bufs.c @@ -42,27 +42,35 @@ */ static int drm_alloc_resource(struct drm_device *dev, int resource) { + struct resource *res; + int rid; + + DRM_SPINLOCK_ASSERT(&dev->dev_lock); + if (resource >= DRM_MAX_PCI_RESOURCE) { DRM_ERROR("Resource %d too large\n", resource); return 1; } - DRM_UNLOCK(); if (dev->pcir[resource] != NULL) { - DRM_LOCK(); return 0; } - dev->pcirid[resource] = PCIR_BAR(resource); - dev->pcir[resource] = bus_alloc_resource_any(dev->device, - SYS_RES_MEMORY, &dev->pcirid[resource], RF_SHAREABLE); + DRM_UNLOCK(); + rid = PCIR_BAR(resource); + res = bus_alloc_resource_any(dev->device, SYS_RES_MEMORY, &rid, + RF_SHAREABLE); DRM_LOCK(); - - if (dev->pcir[resource] == NULL) { + if (res == NULL) { DRM_ERROR("Couldn't find resource 0x%x\n", resource); return 1; } + if (dev->pcir[resource] == NULL) { + dev->pcirid[resource] = rid; + dev->pcir[resource] = res; + } + return 0; } @@ -149,10 +157,12 @@ int drm_addmap(struct drm_device * dev, unsigned long offset, map->size = size; map->type = type; map->flags = flags; + map->handle = (void *)((unsigned int)atomic_fetchadd_int(&dev->map_seq, 1) + << DRM_MAP_HANDLE_SHIFT); switch (map->type) { case _DRM_REGISTERS: - map->handle = drm_ioremap(dev, map); + map->virtual = drm_ioremap(dev, map); if (!(map->flags & _DRM_WRITE_COMBINING)) break; /* FALLTHROUGH */ @@ -161,25 +171,25 @@ int drm_addmap(struct drm_device * dev, unsigned long offset, map->mtrr = 1; break; case _DRM_SHM: - map->handle = malloc(map->size, DRM_MEM_MAPS, M_NOWAIT); + map->virtual = malloc(map->size, DRM_MEM_MAPS, M_NOWAIT); DRM_DEBUG("%lu %d %p\n", - map->size, drm_order(map->size), map->handle); - if (!map->handle) { + map->size, drm_order(map->size), map->virtual); + if (!map->virtual) { free(map, DRM_MEM_MAPS); DRM_LOCK(); return ENOMEM; } - map->offset = (unsigned long)map->handle; + map->offset = (unsigned long)map->virtual; if (map->flags & _DRM_CONTAINS_LOCK) { /* Prevent a 2nd X Server from creating a 2nd lock */ DRM_LOCK(); if (dev->lock.hw_lock != NULL) { DRM_UNLOCK(); - free(map->handle, DRM_MEM_MAPS); + free(map->virtual, DRM_MEM_MAPS); free(map, DRM_MEM_MAPS); return EBUSY; } - dev->lock.hw_lock = map->handle; /* Pointer to lock */ + dev->lock.hw_lock = map->virtual; /* Pointer to lock */ DRM_UNLOCK(); } break; @@ -217,7 +227,8 @@ int drm_addmap(struct drm_device * dev, unsigned long offset, DRM_LOCK(); return EINVAL; } - map->offset += dev->sg->handle; + map->virtual = (void *)(dev->sg->vaddr + offset); + map->offset = dev->sg->vaddr + offset; break; case _DRM_CONSISTENT: /* Unfortunately, we don't get any alignment specification from @@ -235,7 +246,7 @@ int drm_addmap(struct drm_device * dev, unsigned long offset, DRM_LOCK(); return ENOMEM; } - map->handle = map->dmah->vaddr; + map->virtual = map->dmah->vaddr; map->offset = map->dmah->busaddr; break; default: @@ -284,11 +295,7 @@ int drm_addmap_ioctl(struct drm_device *dev, void *data, request->type = map->type; request->flags = map->flags; request->mtrr = map->mtrr; - request->handle = map->handle; - - if (request->type != _DRM_SHM) { - request->handle = (void *)request->offset; - } + request->handle = (void *)map->handle; return 0; } @@ -300,6 +307,9 @@ void drm_rmmap(struct drm_device *dev, drm_local_map_t *map) if (map == NULL) return; + if (map == NULL) + return; + TAILQ_REMOVE(&dev->maplist, map, link); switch (map->type) { @@ -317,7 +327,7 @@ void drm_rmmap(struct drm_device *dev, drm_local_map_t *map) } break; case _DRM_SHM: - free(map->handle, DRM_MEM_MAPS); + free(map->virtual, DRM_MEM_MAPS); break; case _DRM_AGP: case _DRM_SCATTER_GATHER: @@ -732,7 +742,7 @@ static int drm_do_addbufs_sg(struct drm_device *dev, struct drm_buf_desc *reques buf->offset = (dma->byte_count + offset); buf->bus_address = agp_offset + offset; - buf->address = (void *)(agp_offset + offset + dev->sg->handle); + buf->address = (void *)(agp_offset + offset + dev->sg->vaddr); buf->next = NULL; buf->pending = 0; buf->file_priv = NULL; @@ -1047,7 +1057,7 @@ int drm_mapbufs(struct drm_device *dev, void *data, struct drm_file *file_priv) goto done; } size = round_page(map->size); - foff = map->offset; + foff = (unsigned long)map->handle; } else { size = round_page(dma->byte_count), foff = 0; diff --git a/sys/dev/drm/drm_context.c b/sys/dev/drm/drm_context.c index d0a33d0..7bfa976 100644 --- a/sys/dev/drm/drm_context.c +++ b/sys/dev/drm/drm_context.c @@ -144,7 +144,7 @@ int drm_getsareactx(struct drm_device *dev, void *data, map = dev->context_sareas[request->ctx_id]; DRM_UNLOCK(); - request->handle = map->handle; + request->handle = (void *)map->handle; return 0; } diff --git a/sys/dev/drm/drm_drv.c b/sys/dev/drm/drm_drv.c index 50de37b..d97ea92 100644 --- a/sys/dev/drm/drm_drv.c +++ b/sys/dev/drm/drm_drv.c @@ -49,9 +49,6 @@ static void drm_unload(struct drm_device *dev); static drm_pci_id_list_t *drm_find_description(int vendor, int device, drm_pci_id_list_t *idlist); -#define DRIVER_SOFTC(unit) \ - ((struct drm_device *)devclass_get_softc(drm_devclass, unit)) - MODULE_VERSION(drm, 1); MODULE_DEPEND(drm, agp, 1, 1, 1); MODULE_DEPEND(drm, pci, 1, 1, 1); @@ -130,6 +127,9 @@ static struct dev_ops drm_cdevsw = { static int drm_msi = 1; /* Enable by default. */ TUNABLE_INT("hw.drm.msi", &drm_msi); +SYSCTL_NODE(_hw, OID_AUTO, drm, CTLFLAG_RW, NULL, "DRM device"); +SYSCTL_INT(_hw_drm, OID_AUTO, msi, CTLFLAG_RD, &drm_msi, 1, + "Enable MSI interrupts for drm devices"); static struct drm_msi_blacklist_entry drm_msi_blacklist[] = { {0x8086, 0x2772}, /* Intel i945G */ \ @@ -193,8 +193,9 @@ int drm_attach(device_t kdev, drm_pci_id_list_t *idlist) else dev->device = kdev; - dev->devnode = make_dev(&drm_cdevsw, unit, DRM_DEV_UID, DRM_DEV_GID, + dev->devnode = make_dev(&drm_cdevsw, 0, DRM_DEV_UID, DRM_DEV_GID, DRM_DEV_MODE, "dri/card%d", unit); + dev->devnode->si_drv1 = dev; dev->pci_domain = 0; dev->pci_bus = pci_get_bus(dev->device); @@ -413,6 +414,7 @@ static int drm_load(struct drm_device *dev) DRM_DEBUG("\n"); TAILQ_INIT(&dev->maplist); + dev->map_seq = 1; drm_mem_init(); drm_sysctl_init(dev); @@ -578,7 +580,7 @@ int drm_open(struct dev_open_args *ap) struct drm_device *dev = NULL; int retcode = 0; - dev = DRIVER_SOFTC(minor(kdev)); + dev = kdev->si_drv1; DRM_DEBUG("open_count = %d\n", dev->open_count); @@ -603,7 +605,7 @@ int drm_close(struct dev_close_args *ap) struct drm_device *dev; int retcode = 0; - dev = DRIVER_SOFTC(minor(kdev)); + dev = kdev->si_drv1; file_priv = drm_find_file_by_proc(dev, curthread); DRM_DEBUG("open_count = %d\n", dev->open_count); diff --git a/sys/dev/drm/drm_fops.c b/sys/dev/drm/drm_fops.c index ae4a9e2..fe96d6f 100644 --- a/sys/dev/drm/drm_fops.c +++ b/sys/dev/drm/drm_fops.c @@ -53,14 +53,13 @@ int drm_open_helper(struct cdev *kdev, int flags, int fmt, DRM_STRUCTPROC *p, struct drm_device *dev) { struct drm_file *priv; - int m = minor(kdev); int retcode; if (flags & O_EXCL) return EBUSY; /* No exclusive opens */ dev->flags = flags; - DRM_DEBUG("pid = %d, minor = %d\n", DRM_CURRENTPID, m); + DRM_DEBUG("pid = %d, minor = %d\n", DRM_CURRENTPID, devtoname(kdev)); DRM_LOCK(); priv = drm_find_file_by_proc(dev, p); if (priv) { @@ -74,7 +73,6 @@ int drm_open_helper(struct cdev *kdev, int flags, int fmt, DRM_STRUCTPROC *p, priv->uid = p->td_proc->p_ucred->cr_svuid; priv->pid = p->td_proc->p_pid; priv->refs = 1; - priv->minor = m; priv->ioctl_count = 0; /* for compatibility root is always authenticated */ diff --git a/sys/dev/drm/drm_hashtab.c b/sys/dev/drm/drm_hashtab.c new file mode 100644 index 0000000..914292a --- /dev/null +++ b/sys/dev/drm/drm_hashtab.c @@ -0,0 +1,219 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + **************************************************************************/ + +#include + +/* + * Simple open hash tab implementation. + * + * Authors: + * Thomas Hellström + */ + +#include "dev/drm/drmP.h" +#include "dev/drm/drm_hashtab.h" + +/* + * Return a 32-bit hash of the given buffer. The init + * value should be 0, or the previous hash value to extend + * the previous hash. + * + * This function: + * + * Copyright (c) 2001 Tobias Weingartner + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ +static __inline uint32_t +hash32_buf(const void *buf, size_t len, uint32_t hash) +{ + const unsigned char *p = buf; + + while (len--) + hash = ((hash << 5) + hash) + (*p++); + + return hash; +} + +int drm_ht_create(struct drm_open_hash *ht, unsigned int order) +{ + ht->size = 1 << order; + ht->order = order; + ht->table = NULL; + + ht->table = hashinit(ht->size, DRM_MEM_HASHTAB, &ht->mask); + if (!ht->table) { + DRM_ERROR("Out of memory for hash table\n"); + return -ENOMEM; + } + return 0; +} + +void drm_ht_verbose_list(struct drm_open_hash *ht, unsigned long key) +{ + struct drm_hash_item *entry; + struct drm_hash_item_list *h_list; + unsigned int hashed_key; + int count = 0; + + hashed_key = hash32_buf(&key, sizeof(key), ht->order); + DRM_DEBUG("Key is 0x%08lx, Hashed key is 0x%08x\n", key, hashed_key); + h_list = &ht->table[hashed_key & ht->mask]; + LIST_FOREACH(entry, h_list, head) + DRM_DEBUG("count %d, key: 0x%08lx\n", count++, entry->key); +} + +static struct drm_hash_item * +drm_ht_find_key(struct drm_open_hash *ht, unsigned long key) +{ + struct drm_hash_item *entry; + struct drm_hash_item_list *h_list; + unsigned int hashed_key; + + hashed_key = hash32_buf(&key, sizeof(key), ht->order); + h_list = &ht->table[hashed_key & ht->mask]; + LIST_FOREACH(entry, h_list, head) { + if (entry->key == key) + return entry; + if (entry->key > key) + break; + } + return NULL; +} + + +int drm_ht_insert_item(struct drm_open_hash *ht, struct drm_hash_item *item) +{ + struct drm_hash_item *entry, *parent; + struct drm_hash_item_list *h_list; + unsigned int hashed_key; + unsigned long key = item->key; + + hashed_key = hash32_buf(&key, sizeof(key), ht->order); + h_list = &ht->table[hashed_key & ht->mask]; + parent = NULL; + LIST_FOREACH(entry, h_list, head) { + if (entry->key == key) + return -EINVAL; + if (entry->key > key) + break; + parent = entry; + } + if (parent) { + LIST_INSERT_AFTER(parent, item, head); + } else { + LIST_INSERT_HEAD(h_list, item, head); + } + return 0; +} + +/* + * Just insert an item and return any "bits" bit key that hasn't been + * used before. + */ +int drm_ht_just_insert_please(struct drm_open_hash *ht, struct drm_hash_item *item, + unsigned long seed, int bits, int shift, + unsigned long add) +{ + int ret; + unsigned long mask = (1 << bits) - 1; + unsigned long first, unshifted_key = 0; + + unshifted_key = hash32_buf(&seed, sizeof(seed), unshifted_key); + first = unshifted_key; + do { + item->key = (unshifted_key << shift) + add; + ret = drm_ht_insert_item(ht, item); + if (ret) + unshifted_key = (unshifted_key + 1) & mask; + } while(ret && (unshifted_key != first)); + + if (ret) { + DRM_ERROR("Available key bit space exhausted\n"); + return -EINVAL; + } + return 0; +} + +int drm_ht_find_item(struct drm_open_hash *ht, unsigned long key, + struct drm_hash_item **item) +{ + struct drm_hash_item *entry; + + entry = drm_ht_find_key(ht, key); + if (!entry) + return -EINVAL; + + *item = entry; + return 0; +} + +int drm_ht_remove_key(struct drm_open_hash *ht, unsigned long key) +{ + struct drm_hash_item *entry; + + entry = drm_ht_find_key(ht, key); + if (entry) { + LIST_REMOVE(entry, head); + return 0; + } + return -EINVAL; +} + +int drm_ht_remove_item(struct drm_open_hash *ht, struct drm_hash_item *item) +{ + LIST_REMOVE(item, head); + return 0; +} + +void drm_ht_remove(struct drm_open_hash *ht) +{ + if (ht->table) { + kfree(ht->table, DRM_MEM_HASHTAB); + ht->table = NULL; + } +} diff --git a/sys/dev/drm/drm_hashtab.h b/sys/dev/drm/drm_hashtab.h new file mode 100644 index 0000000..54ae08b --- /dev/null +++ b/sys/dev/drm/drm_hashtab.h @@ -0,0 +1,67 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismack, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + **************************************************************************/ + +#include + +/* + * Simple open hash tab implementation. + * + * Authors: + * Thomas Hellström + */ + +#ifndef DRM_HASHTAB_H +#define DRM_HASHTAB_H + +#define drm_hash_entry(_ptr, _type, _member) container_of(_ptr, _type, _member) + +struct drm_hash_item { + LIST_ENTRY(drm_hash_item) head; + unsigned long key; +}; + +struct drm_open_hash { + LIST_HEAD(drm_hash_item_list, drm_hash_item) *table; + unsigned int size; + unsigned int order; + unsigned long mask; +}; + +extern int drm_ht_create(struct drm_open_hash *ht, unsigned int order); +extern int drm_ht_insert_item(struct drm_open_hash *ht, struct drm_hash_item *item); +extern int drm_ht_just_insert_please(struct drm_open_hash *ht, struct drm_hash_item *item, + unsigned long seed, int bits, int shift, + unsigned long add); +extern int drm_ht_find_item(struct drm_open_hash *ht, unsigned long key, struct drm_hash_item **item); + +extern void drm_ht_verbose_list(struct drm_open_hash *ht, unsigned long key); +extern int drm_ht_remove_key(struct drm_open_hash *ht, unsigned long key); +extern int drm_ht_remove_item(struct drm_open_hash *ht, struct drm_hash_item *item); +extern void drm_ht_remove(struct drm_open_hash *ht); + +#endif diff --git a/sys/dev/drm/drm_irq.c b/sys/dev/drm/drm_irq.c index 47e3945..8f51002 100644 --- a/sys/dev/drm/drm_irq.c +++ b/sys/dev/drm/drm_irq.c @@ -101,7 +101,9 @@ void drm_vblank_cleanup(struct drm_device *dev) callout_stop(&dev->vblank_disable_timer); DRM_SPINUNLOCK(&dev->vbl_lock); + DRM_SPINLOCK(&dev->vbl_lock); vblank_disable_fn((void *)dev); + DRM_SPINUNLOCK(&dev->vbl_lock); free(dev->vblank, DRM_MEM_DRIVER); diff --git a/sys/dev/drm/drm_linux_list.h b/sys/dev/drm/drm_linux_list.h index 15c1b6e..ba8090f 100644 --- a/sys/dev/drm/drm_linux_list.h +++ b/sys/dev/drm/drm_linux_list.h @@ -29,12 +29,15 @@ * */ +#ifndef _DRM_LINUX_LIST_H_ +#define _DRM_LINUX_LIST_H_ + struct list_head { struct list_head *next, *prev; }; -/* Cheat, assume the list_head is at the start of the struct */ -#define list_entry(entry, type, member) (type *)(entry) +#define list_entry(ptr, type, member) container_of(ptr,type,member) +#define hlist_entry(ptr, type, member) container_of(ptr,type,member) static __inline__ void INIT_LIST_HEAD(struct list_head *head) { @@ -48,6 +51,14 @@ list_empty(struct list_head *head) { } static __inline__ void +list_add(struct list_head *new, struct list_head *head) { + (head)->next->prev = new; + (new)->next = (head)->next; + (new)->prev = head; + (head)->next = new; +} + +static __inline__ void list_add_tail(struct list_head *entry, struct list_head *head) { (entry)->prev = (head)->prev; (entry)->next = head; @@ -61,6 +72,13 @@ list_del(struct list_head *entry) { (entry)->prev->next = (entry)->next; } +static __inline__ void +list_del_init(struct list_head *entry) { + (entry)->next->prev = (entry)->prev; + (entry)->prev->next = (entry)->next; + INIT_LIST_HEAD(entry); +} + #define list_for_each(entry, head) \ for (entry = (head)->next; entry != head; entry = (entry)->next) @@ -73,3 +91,17 @@ list_del(struct list_head *entry) { entry != head; \ entry = temp, temp = entry->next) +/** + * list_for_each_entry_safe - iterate over list of given type safe against removal of list entry + * @pos: the type * to use as a loop cursor. + * @n: another type * to use as temporary storage + * @head: the head for your list. + * @member: the name of the list_struct within the struct. + */ +#define list_for_each_entry_safe(pos, n, head, member) \ + for (pos = list_entry((head)->next, __typeof(*pos), member), \ + n = list_entry(pos->member.next, __typeof(*pos), member); \ + &pos->member != (head); \ + pos = n, n = list_entry(n->member.next, __typeof(*n), member)) + +#endif /* _DRM_LINUX_LIST_H_ */ diff --git a/sys/dev/drm/drm_memory.c b/sys/dev/drm/drm_memory.c index 85c0545..d4eefc7 100644 --- a/sys/dev/drm/drm_memory.c +++ b/sys/dev/drm/drm_memory.c @@ -55,8 +55,10 @@ MALLOC_DEFINE(DRM_MEM_BUFLISTS, "drm_buflists", "DRM BUFLISTS Data Structures"); MALLOC_DEFINE(DRM_MEM_AGPLISTS, "drm_agplists", "DRM AGPLISTS Data Structures"); MALLOC_DEFINE(DRM_MEM_CTXBITMAP, "drm_ctxbitmap", "DRM CTXBITMAP Data Structures"); -MALLOC_DEFINE(DRM_MEM_SGLISTS, "drm_sglists", "DRM SGLISTS Data Structures"); MALLOC_DEFINE(DRM_MEM_DRAWABLE, "drm_drawable", "DRM DRAWABLE Data Structures"); +MALLOC_DEFINE(DRM_MEM_MM, "drm_sman", "DRM MEMORY MANAGER Data Structures"); +MALLOC_DEFINE(DRM_MEM_HASHTAB, "drm_hashtab", "DRM HASHTABLE Data Structures"); +MALLOC_DEFINE(DRM_MEM_SGLISTS, "drm_sglists", "DRM SGLISTS Data Structures"); void drm_mem_init(void) { @@ -81,7 +83,7 @@ void *drm_ioremap(struct drm_device *dev, drm_local_map_t *map) void drm_ioremapfree(drm_local_map_t *map) { - pmap_unmapdev((vm_offset_t) map->handle, map->size); + pmap_unmapdev((vm_offset_t) map->virtual, map->size); } int diff --git a/sys/dev/drm/drm_mm.c b/sys/dev/drm/drm_mm.c new file mode 100644 index 0000000..b14ffce --- /dev/null +++ b/sys/dev/drm/drm_mm.c @@ -0,0 +1,366 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + **************************************************************************/ + +#include + +/* + * Generic simple memory manager implementation. Intended to be used as a base + * class implementation for more advanced memory managers. + * + * Note that the algorithm used is quite simple and there might be substantial + * performance gains if a smarter free list is implemented. Currently it is just an + * unordered stack of free regions. This could easily be improved if an RB-tree + * is used instead. At least if we expect heavy fragmentation. + * + * Aligned allocations can also see improvement. + * + * Authors: + * Thomas Hellström + */ + +#include "dev/drm/drmP.h" +#include "dev/drm/drm_mm.h" + +#define MM_UNUSED_TARGET 4 + +unsigned long drm_mm_tail_space(struct drm_mm *mm) +{ + struct list_head *tail_node; + struct drm_mm_node *entry; + + tail_node = mm->ml_entry.prev; + entry = list_entry(tail_node, struct drm_mm_node, ml_entry); + if (!entry->free) + return 0; + + return entry->size; +} + +int drm_mm_remove_space_from_tail(struct drm_mm *mm, unsigned long size) +{ + struct list_head *tail_node; + struct drm_mm_node *entry; + + tail_node = mm->ml_entry.prev; + entry = list_entry(tail_node, struct drm_mm_node, ml_entry); + if (!entry->free) + return -ENOMEM; + + if (entry->size <= size) + return -ENOMEM; + + entry->size -= size; + return 0; +} + +static struct drm_mm_node *drm_mm_kmalloc(struct drm_mm *mm, int atomic) +{ + struct drm_mm_node *child; + + if (atomic) + child = malloc(sizeof(*child), DRM_MEM_MM, M_NOWAIT); + else + child = malloc(sizeof(*child), DRM_MEM_MM, M_WAITOK); + + if (unlikely(child == NULL)) { + spin_lock(&mm->unused_lock); + if (list_empty(&mm->unused_nodes)) + child = NULL; + else { + child = + list_entry(mm->unused_nodes.next, + struct drm_mm_node, fl_entry); + list_del(&child->fl_entry); + --mm->num_unused; + } + spin_unlock(&mm->unused_lock); + } + return child; +} + +int drm_mm_pre_get(struct drm_mm *mm) +{ + struct drm_mm_node *node; + + spin_lock(&mm->unused_lock); + while (mm->num_unused < MM_UNUSED_TARGET) { + spin_unlock(&mm->unused_lock); + node = malloc(sizeof(*node), DRM_MEM_MM, M_WAITOK); + spin_lock(&mm->unused_lock); + + if (unlikely(node == NULL)) { + int ret = (mm->num_unused < 2) ? -ENOMEM : 0; + spin_unlock(&mm->unused_lock); + return ret; + } + ++mm->num_unused; + list_add_tail(&node->fl_entry, &mm->unused_nodes); + } + spin_unlock(&mm->unused_lock); + return 0; +} + +static int drm_mm_create_tail_node(struct drm_mm *mm, + unsigned long start, + unsigned long size, int atomic) +{ + struct drm_mm_node *child; + + child = drm_mm_kmalloc(mm, atomic); + if (unlikely(child == NULL)) + return -ENOMEM; + + child->free = 1; + child->size = size; + child->start = start; + child->mm = mm; + + list_add_tail(&child->ml_entry, &mm->ml_entry); + list_add_tail(&child->fl_entry, &mm->fl_entry); + + return 0; +} + +int drm_mm_add_space_to_tail(struct drm_mm *mm, unsigned long size, int atomic) +{ + struct list_head *tail_node; + struct drm_mm_node *entry; + + tail_node = mm->ml_entry.prev; + entry = list_entry(tail_node, struct drm_mm_node, ml_entry); + if (!entry->free) { + return drm_mm_create_tail_node(mm, entry->start + entry->size, + size, atomic); + } + entry->size += size; + return 0; +} + +static struct drm_mm_node *drm_mm_split_at_start(struct drm_mm_node *parent, + unsigned long size, + int atomic) +{ + struct drm_mm_node *child; + + child = drm_mm_kmalloc(parent->mm, atomic); + if (unlikely(child == NULL)) + return NULL; + + INIT_LIST_HEAD(&child->fl_entry); + + child->free = 0; + child->size = size; + child->start = parent->start; + child->mm = parent->mm; + + list_add_tail(&child->ml_entry, &parent->ml_entry); + INIT_LIST_HEAD(&child->fl_entry); + + parent->size -= size; + parent->start += size; + return child; +} + + +struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node, + unsigned long size, + unsigned alignment, + int atomic) +{ + + struct drm_mm_node *align_splitoff = NULL; + unsigned tmp = 0; + + if (alignment) + tmp = node->start % alignment; + + if (tmp) { + align_splitoff = + drm_mm_split_at_start(node, alignment - tmp, atomic); + if (unlikely(align_splitoff == NULL)) + return NULL; + } + + if (node->size == size) { + list_del_init(&node->fl_entry); + node->free = 0; + } else { + node = drm_mm_split_at_start(node, size, atomic); + } + + if (align_splitoff) + drm_mm_put_block(align_splitoff); + + return node; +} + +/* + * Put a block. Merge with the previous and / or next block if they are free. + * Otherwise add to the free stack. + */ + +void drm_mm_put_block(struct drm_mm_node *cur) +{ + + struct drm_mm *mm = cur->mm; + struct list_head *cur_head = &cur->ml_entry; + struct list_head *root_head = &mm->ml_entry; + struct drm_mm_node *prev_node = NULL; + struct drm_mm_node *next_node; + + int merged = 0; + + if (cur_head->prev != root_head) { + prev_node = + list_entry(cur_head->prev, struct drm_mm_node, ml_entry); + if (prev_node->free) { + prev_node->size += cur->size; + merged = 1; + } + } + if (cur_head->next != root_head) { + next_node = + list_entry(cur_head->next, struct drm_mm_node, ml_entry); + if (next_node->free) { + if (merged) { + prev_node->size += next_node->size; + list_del(&next_node->ml_entry); + list_del(&next_node->fl_entry); + if (mm->num_unused < MM_UNUSED_TARGET) { + list_add(&next_node->fl_entry, + &mm->unused_nodes); + ++mm->num_unused; + } else + free(next_node, DRM_MEM_MM); + } else { + next_node->size += cur->size; + next_node->start = cur->start; + merged = 1; + } + } + } + if (!merged) { + cur->free = 1; + list_add(&cur->fl_entry, &mm->fl_entry); + } else { + list_del(&cur->ml_entry); + if (mm->num_unused < MM_UNUSED_TARGET) { + list_add(&cur->fl_entry, &mm->unused_nodes); + ++mm->num_unused; + } else + free(cur, DRM_MEM_MM); + } +} + +struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm, + unsigned long size, + unsigned alignment, int best_match) +{ + struct list_head *list; + const struct list_head *free_stack = &mm->fl_entry; + struct drm_mm_node *entry; + struct drm_mm_node *best; + unsigned long best_size; + unsigned wasted; + + best = NULL; + best_size = ~0UL; + + list_for_each(list, free_stack) { + entry = list_entry(list, struct drm_mm_node, fl_entry); + wasted = 0; + + if (entry->size < size) + continue; + + if (alignment) { + register unsigned tmp = entry->start % alignment; + if (tmp) + wasted += alignment - tmp; + } + + if (entry->size >= size + wasted) { + if (!best_match) + return entry; + if (size < best_size) { + best = entry; + best_size = entry->size; + } + } + } + + return best; +} + +int drm_mm_clean(struct drm_mm * mm) +{ + struct list_head *head = &mm->ml_entry; + + return (head->next->next == head); +} + +int drm_mm_init(struct drm_mm * mm, unsigned long start, unsigned long size) +{ + INIT_LIST_HEAD(&mm->ml_entry); + INIT_LIST_HEAD(&mm->fl_entry); + INIT_LIST_HEAD(&mm->unused_nodes); + mm->num_unused = 0; + spin_init(&mm->unused_lock); + + /* XXX This could be non-atomic but gets called from a locked path */ + return drm_mm_create_tail_node(mm, start, size, 1); +} + +void drm_mm_takedown(struct drm_mm * mm) +{ + struct list_head *bnode = mm->fl_entry.next; + struct drm_mm_node *entry; + struct drm_mm_node *next; + + entry = list_entry(bnode, struct drm_mm_node, fl_entry); + + if (entry->ml_entry.next != &mm->ml_entry || + entry->fl_entry.next != &mm->fl_entry) { + DRM_ERROR("Memory manager not clean. Delaying takedown\n"); + return; + } + + list_del(&entry->fl_entry); + list_del(&entry->ml_entry); + free(entry, DRM_MEM_MM); + + spin_lock(&mm->unused_lock); + list_for_each_entry_safe(entry, next, &mm->unused_nodes, fl_entry) { + list_del(&entry->fl_entry); + free(entry, DRM_MEM_MM); + --mm->num_unused; + } + spin_unlock(&mm->unused_lock); + + KASSERT(mm->num_unused == 0, ("num_unused != 0")); +} diff --git a/sys/dev/drm/drm_mm.h b/sys/dev/drm/drm_mm.h new file mode 100644 index 0000000..9815083 --- /dev/null +++ b/sys/dev/drm/drm_mm.h @@ -0,0 +1,99 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + **************************************************************************/ + +#include + +/* + * Authors: + * Thomas Hellstrom + */ + +#ifndef _DRM_MM_H_ +#define _DRM_MM_H_ + +#include "dev/drm/drm_linux_list.h" + +struct drm_mm_node { + struct list_head fl_entry; + struct list_head ml_entry; + int free; + unsigned long start; + unsigned long size; + struct drm_mm *mm; + void *private; +}; + +struct drm_mm { + struct list_head fl_entry; + struct list_head ml_entry; + struct list_head unused_nodes; + int num_unused; + struct spinlock unused_lock; +}; + +/* + * Basic range manager support (drm_mm.c) + */ +extern struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node, + unsigned long size, + unsigned alignment, + int atomic); +static inline struct drm_mm_node *drm_mm_get_block(struct drm_mm_node *parent, + unsigned long size, + unsigned alignment) +{ + return drm_mm_get_block_generic(parent, size, alignment, 0); +} +static inline struct drm_mm_node *drm_mm_get_block_atomic(struct drm_mm_node *parent, + unsigned long size, + unsigned alignment) +{ + return drm_mm_get_block_generic(parent, size, alignment, 1); +} +extern void drm_mm_put_block(struct drm_mm_node *cur); +extern struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm, + unsigned long size, + unsigned alignment, + int best_match); +extern int drm_mm_init(struct drm_mm *mm, unsigned long start, + unsigned long size); +extern void drm_mm_takedown(struct drm_mm *mm); +extern int drm_mm_clean(struct drm_mm *mm); +extern unsigned long drm_mm_tail_space(struct drm_mm *mm); +extern int drm_mm_remove_space_from_tail(struct drm_mm *mm, + unsigned long size); +extern int drm_mm_add_space_to_tail(struct drm_mm *mm, + unsigned long size, int atomic); +extern int drm_mm_pre_get(struct drm_mm *mm); + +static inline struct drm_mm *drm_get_mm(struct drm_mm_node *block) +{ + return block->mm; +} + +#endif diff --git a/sys/dev/drm/drm_pciids.h b/sys/dev/drm/drm_pciids.h index a2acc89..18892ae 100644 --- a/sys/dev/drm/drm_pciids.h +++ b/sys/dev/drm/drm_pciids.h @@ -335,6 +335,7 @@ {0x1002, 0x9712, CHIP_RS880|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP, "ATI Mobility Radeon HD 4200"}, \ {0x1002, 0x9713, CHIP_RS880|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP, "ATI Mobility Radeon 4100"}, \ {0x1002, 0x9714, CHIP_RS880|RADEON_NEW_MEMMAP|RADEON_IS_IGP, "ATI RS880"}, \ + {0x1002, 0x9715, CHIP_RS880|RADEON_NEW_MEMMAP|RADEON_IS_IGP, "ATI Radeon HD 4250"}, \ {0x1002, 0x9440, CHIP_RV770|RADEON_NEW_MEMMAP, "ATI Radeon 4800 Series"}, \ {0x1002, 0x9441, CHIP_RV770|RADEON_NEW_MEMMAP, "ATI Radeon 4870 X2"}, \ {0x1002, 0x9442, CHIP_RV770|RADEON_NEW_MEMMAP, "ATI Radeon 4800 Series"}, \ @@ -552,8 +553,8 @@ {0x8086, 0x2E12, CHIP_I9XX|CHIP_I965, "Intel Q45/Q43"}, \ {0x8086, 0x2E22, CHIP_I9XX|CHIP_I965, "Intel G45/G43"}, \ {0x8086, 0x2E32, CHIP_I9XX|CHIP_I965, "Intel G41"}, \ - {0x8086, 0xA001, CHIP_I9XX|CHIP_I965, "Intel IGD"}, \ - {0x8086, 0xA011, CHIP_I9XX|CHIP_I965, "Intel IGD"}, \ + {0x8086, 0xA001, CHIP_I9XX|CHIP_I965, "Intel Pineview"}, \ + {0x8086, 0xA011, CHIP_I9XX|CHIP_I965, "Intel Pineview (M)"}, \ {0, 0, 0, NULL} #define imagine_PCI_IDS \ diff --git a/sys/dev/drm/drm_scatter.c b/sys/dev/drm/drm_scatter.c index 34a2bcc..52b18f6 100644 --- a/sys/dev/drm/drm_scatter.c +++ b/sys/dev/drm/drm_scatter.c @@ -1,5 +1,5 @@ /*- - * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California. + * Copyright (c) 2009 Robert C. Noland III * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -20,127 +20,66 @@ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. - * - * Authors: - * Gareth Hughes - * Eric Anholt - * */ /** @file drm_scatter.c * Allocation of memory for scatter-gather mappings by the graphics chip. - * * The memory allocated here is then made into an aperture in the card - * by drm_ati_pcigart_init(). + * by mapping the pages into the GART. */ #include "dev/drm/drmP.h" -static void drm_sg_alloc_cb(void *arg, bus_dma_segment_t *segs, - int nsegs, int error); - int drm_sg_alloc(struct drm_device *dev, struct drm_scatter_gather *request) { struct drm_sg_mem *entry; - struct drm_dma_handle *dmah; - unsigned long pages; - int ret; + vm_size_t size; + vm_pindex_t pindex; if (dev->sg) return EINVAL; - entry = malloc(sizeof(*entry), DRM_MEM_SGLISTS, M_WAITOK | M_ZERO); - if (!entry) - return ENOMEM; - - pages = round_page(request->size) / PAGE_SIZE; - DRM_DEBUG("sg size=%ld pages=%ld\n", request->size, pages); - - entry->pages = pages; + DRM_DEBUG("request size=%ld\n", request->size); - entry->busaddr = malloc(pages * sizeof(*entry->busaddr), DRM_MEM_PAGES, - M_WAITOK | M_ZERO); - if (!entry->busaddr) { - free(entry, DRM_MEM_SGLISTS); - return ENOMEM; - } - - dmah = malloc(sizeof(struct drm_dma_handle), DRM_MEM_DMA, - M_ZERO | M_NOWAIT); - if (dmah == NULL) { - free(entry->busaddr, DRM_MEM_PAGES); - free(entry, DRM_MEM_SGLISTS); - return ENOMEM; - } + entry = malloc(sizeof(*entry), DRM_MEM_DRIVER, M_WAITOK | M_ZERO); - ret = bus_dma_tag_create(NULL, PAGE_SIZE, 0, /* tag, align, boundary */ - BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, /* lowaddr, highaddr */ - NULL, NULL, /* filtfunc, filtfuncargs */ - request->size, pages, /* maxsize, nsegs */ - PAGE_SIZE, 0, /* maxsegsize, flags */ - &dmah->tag); - if (ret != 0) { - free(dmah, DRM_MEM_DMA); - free(entry->busaddr, DRM_MEM_PAGES); - free(entry, DRM_MEM_SGLISTS); - return ENOMEM; - } + size = round_page(request->size); + entry->pages = OFF_TO_IDX(size); + entry->busaddr = malloc(entry->pages * sizeof(*entry->busaddr), + DRM_MEM_SGLISTS, M_WAITOK | M_ZERO); - /* XXX BUS_DMA_NOCACHE */ - ret = bus_dmamem_alloc(dmah->tag, &dmah->vaddr, - BUS_DMA_WAITOK | BUS_DMA_ZERO , &dmah->map); - if (ret != 0) { - bus_dma_tag_destroy(dmah->tag); - free(dmah, DRM_MEM_DMA); - free(entry->busaddr, DRM_MEM_PAGES); - free(entry, DRM_MEM_SGLISTS); - return ENOMEM; +/* + * XXX + entry->vaddr = kmem_alloc_attr(kernel_map, size, M_WAITOK | M_ZERO, + 0, BUS_SPACE_MAXADDR_32BIT, VM_MEMATTR_WRITE_COMBINING); + */ + entry->vaddr = kmem_alloc(&kernel_map, size); + if (entry->vaddr == 0) { + drm_sg_cleanup(entry); + return (ENOMEM); } - ret = bus_dmamap_load(dmah->tag, dmah->map, dmah->vaddr, - request->size, drm_sg_alloc_cb, entry, BUS_DMA_NOWAIT); - if (ret != 0) { - bus_dmamem_free(dmah->tag, dmah->vaddr, dmah->map); - bus_dma_tag_destroy(dmah->tag); - free(dmah, DRM_MEM_DMA); - free(entry->busaddr, DRM_MEM_PAGES); - free(entry, DRM_MEM_SGLISTS); - return ENOMEM; + for(pindex = 0; pindex < entry->pages; pindex++) { + entry->busaddr[pindex] = + vtophys(entry->vaddr + IDX_TO_OFF(pindex)); } - entry->dmah = dmah; - entry->handle = (unsigned long)dmah->vaddr; - - DRM_DEBUG("sg alloc handle = %08lx\n", entry->handle); - - entry->virtual = (void *)entry->handle; - request->handle = entry->handle; - DRM_LOCK(); if (dev->sg) { DRM_UNLOCK(); drm_sg_cleanup(entry); - return EINVAL; + return (EINVAL); } dev->sg = entry; DRM_UNLOCK(); - return 0; -} - -static void -drm_sg_alloc_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int error) -{ - struct drm_sg_mem *entry = arg; - int i; + request->handle = entry->vaddr; - if (error != 0) - return; + DRM_DEBUG("allocated %ju pages @ 0x%08zx, contents=%08lx\n", + entry->pages, entry->vaddr, *(unsigned long *)entry->vaddr); - for(i = 0 ; i < nsegs ; i++) { - entry->busaddr[i] = segs[i].ds_addr; - } + return (0); } int @@ -151,20 +90,22 @@ drm_sg_alloc_ioctl(struct drm_device *dev, void *data, DRM_DEBUG("\n"); - return drm_sg_alloc(dev, request); + return (drm_sg_alloc(dev, request)); } void drm_sg_cleanup(struct drm_sg_mem *entry) { - struct drm_dma_handle *dmah = entry->dmah; - - bus_dmamap_unload(dmah->tag, dmah->map); - bus_dmamem_free(dmah->tag, dmah->vaddr, dmah->map); - bus_dma_tag_destroy(dmah->tag); - free(dmah, DRM_MEM_DMA); - free(entry->busaddr, DRM_MEM_PAGES); - free(entry, DRM_MEM_SGLISTS); + if (entry == NULL) + return; + + if (entry->vaddr != 0) + kmem_free(&kernel_map, entry->vaddr, IDX_TO_OFF(entry->pages)); + + free(entry->busaddr, DRM_MEM_SGLISTS); + free(entry, DRM_MEM_DRIVER); + + return; } int @@ -178,12 +119,12 @@ drm_sg_free(struct drm_device *dev, void *data, struct drm_file *file_priv) dev->sg = NULL; DRM_UNLOCK(); - if (!entry || entry->handle != request->handle) - return EINVAL; + if (!entry || entry->vaddr != request->handle) + return (EINVAL); - DRM_DEBUG("sg free virtual = 0x%lx\n", entry->handle); + DRM_DEBUG("free 0x%zx\n", entry->vaddr); drm_sg_cleanup(entry); - return 0; + return (0); } diff --git a/sys/dev/drm/drm_sman.c b/sys/dev/drm/drm_sman.c new file mode 100644 index 0000000..a0e5d73 --- /dev/null +++ b/sys/dev/drm/drm_sman.c @@ -0,0 +1,351 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck., ND., USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +#include + +/* + * Simple memory manager interface that keeps track on allocate regions on a + * per "owner" basis. All regions associated with an "owner" can be released + * with a simple call. Typically if the "owner" exists. The owner is any + * "unsigned long" identifier. Can typically be a pointer to a file private + * struct or a context identifier. + * + * Authors: + * Thomas Hellström + */ + +#include "dev/drm/drmP.h" +#include "dev/drm/drm_sman.h" + +struct drm_owner_item { + struct drm_hash_item owner_hash; + struct list_head sman_list; + struct list_head mem_blocks; +}; + +void drm_sman_takedown(struct drm_sman * sman) +{ + drm_ht_remove(&sman->user_hash_tab); + drm_ht_remove(&sman->owner_hash_tab); + if (sman->mm) + drm_free(sman->mm, sman->num_managers * sizeof(*sman->mm), + DRM_MEM_MM); +} + +int +drm_sman_init(struct drm_sman * sman, unsigned int num_managers, + unsigned int user_order, unsigned int owner_order) +{ + int ret = 0; + + sman->mm = (struct drm_sman_mm *) drm_calloc(num_managers, + sizeof(*sman->mm), DRM_MEM_MM); + if (!sman->mm) { + ret = -ENOMEM; + goto out; + } + sman->num_managers = num_managers; + INIT_LIST_HEAD(&sman->owner_items); + ret = drm_ht_create(&sman->owner_hash_tab, owner_order); + if (ret) + goto out1; + ret = drm_ht_create(&sman->user_hash_tab, user_order); + if (!ret) + goto out; + + drm_ht_remove(&sman->owner_hash_tab); +out1: + drm_free(sman->mm, num_managers * sizeof(*sman->mm), DRM_MEM_MM); +out: + return ret; +} + +static void *drm_sman_mm_allocate(void *private, unsigned long size, + unsigned alignment) +{ + struct drm_mm *mm = (struct drm_mm *) private; + struct drm_mm_node *tmp; + + tmp = drm_mm_search_free(mm, size, alignment, 1); + if (!tmp) { + return NULL; + } + /* This could be non-atomic, but we are called from a locked path */ + tmp = drm_mm_get_block_atomic(tmp, size, alignment); + return tmp; +} + +static void drm_sman_mm_free(void *private, void *ref) +{ + struct drm_mm_node *node = (struct drm_mm_node *) ref; + + drm_mm_put_block(node); +} + +static void drm_sman_mm_destroy(void *private) +{ + struct drm_mm *mm = (struct drm_mm *) private; + drm_mm_takedown(mm); + drm_free(mm, sizeof(*mm), DRM_MEM_MM); +} + +static unsigned long drm_sman_mm_offset(void *private, void *ref) +{ + struct drm_mm_node *node = (struct drm_mm_node *) ref; + return node->start; +} + +int +drm_sman_set_range(struct drm_sman * sman, unsigned int manager, + unsigned long start, unsigned long size) +{ + struct drm_sman_mm *sman_mm; + struct drm_mm *mm; + int ret; + + KASSERT(manager < sman->num_managers, ("Invalid manager")); + + sman_mm = &sman->mm[manager]; + mm = malloc(sizeof(*mm), DRM_MEM_MM, M_NOWAIT | M_ZERO); + if (!mm) { + return -ENOMEM; + } + sman_mm->private = mm; + ret = drm_mm_init(mm, start, size); + + if (ret) { + drm_free(mm, sizeof(*mm), DRM_MEM_MM); + return ret; + } + + sman_mm->allocate = drm_sman_mm_allocate; + sman_mm->free = drm_sman_mm_free; + sman_mm->destroy = drm_sman_mm_destroy; + sman_mm->offset = drm_sman_mm_offset; + + return 0; +} + +int +drm_sman_set_manager(struct drm_sman * sman, unsigned int manager, + struct drm_sman_mm * allocator) +{ + KASSERT(manager < sman->num_managers, ("Invalid manager")); + sman->mm[manager] = *allocator; + + return 0; +} + +static struct drm_owner_item *drm_sman_get_owner_item(struct drm_sman * sman, + unsigned long owner) +{ + int ret; + struct drm_hash_item *owner_hash_item; + struct drm_owner_item *owner_item; + + ret = drm_ht_find_item(&sman->owner_hash_tab, owner, &owner_hash_item); + if (!ret) { + return drm_hash_entry(owner_hash_item, struct drm_owner_item, + owner_hash); + } + + owner_item = malloc(sizeof(*owner_item), DRM_MEM_MM, M_NOWAIT | M_ZERO); + if (!owner_item) + goto out; + + INIT_LIST_HEAD(&owner_item->mem_blocks); + owner_item->owner_hash.key = owner; + DRM_DEBUG("owner_item = %p, mem_blocks = %p\n", owner_item, &owner_item->mem_blocks); + if (drm_ht_insert_item(&sman->owner_hash_tab, &owner_item->owner_hash)) + goto out1; + + list_add_tail(&owner_item->sman_list, &sman->owner_items); + return owner_item; + +out1: + drm_free(owner_item, sizeof(*owner_item), DRM_MEM_MM); +out: + return NULL; +} + +struct drm_memblock_item *drm_sman_alloc(struct drm_sman *sman, unsigned int manager, + unsigned long size, unsigned alignment, + unsigned long owner) +{ + void *tmp; + struct drm_sman_mm *sman_mm; + struct drm_owner_item *owner_item; + struct drm_memblock_item *memblock; + + KASSERT(manager < sman->num_managers, ("Invalid manager")); + + sman_mm = &sman->mm[manager]; + tmp = sman_mm->allocate(sman_mm->private, size, alignment); + if (!tmp) { + return NULL; + } + + memblock = malloc(sizeof(*memblock), DRM_MEM_MM, M_NOWAIT | M_ZERO); + DRM_DEBUG("allocated mem_block %p\n", memblock); + if (!memblock) + goto out; + + memblock->mm_info = tmp; + memblock->mm = sman_mm; + memblock->sman = sman; + INIT_LIST_HEAD(&memblock->owner_list); + + if (drm_ht_just_insert_please + (&sman->user_hash_tab, &memblock->user_hash, + (unsigned long)memblock, 32, 0, 0)) + goto out1; + + owner_item = drm_sman_get_owner_item(sman, owner); + if (!owner_item) + goto out2; + + DRM_DEBUG("owner_item = %p, mem_blocks = %p\n", owner_item, &owner_item->mem_blocks); + DRM_DEBUG("owner_list.prev = %p, mem_blocks.prev = %p\n", memblock->owner_list.prev, owner_item->mem_blocks.prev); + DRM_DEBUG("owner_list.next = %p, mem_blocks.next = %p\n", memblock->owner_list.next, owner_item->mem_blocks.next); + list_add_tail(&memblock->owner_list, &owner_item->mem_blocks); + + DRM_DEBUG("Complete\n"); + return memblock; + +out2: + drm_ht_remove_item(&sman->user_hash_tab, &memblock->user_hash); +out1: + drm_free(memblock, sizeof(*memblock), DRM_MEM_MM); +out: + sman_mm->free(sman_mm->private, tmp); + + return NULL; +} + +static void drm_sman_free(struct drm_memblock_item *item) +{ + struct drm_sman *sman = item->sman; + + list_del(&item->owner_list); + drm_ht_remove_item(&sman->user_hash_tab, &item->user_hash); + item->mm->free(item->mm->private, item->mm_info); + drm_free(item, sizeof(*item), DRM_MEM_MM); +} + +int drm_sman_free_key(struct drm_sman *sman, unsigned int key) +{ + struct drm_hash_item *hash_item; + struct drm_memblock_item *memblock_item; + + if (drm_ht_find_item(&sman->user_hash_tab, key, &hash_item)) + return -EINVAL; + + memblock_item = drm_hash_entry(hash_item, struct drm_memblock_item, + user_hash); + drm_sman_free(memblock_item); + return 0; +} + +static void drm_sman_remove_owner(struct drm_sman *sman, + struct drm_owner_item *owner_item) +{ + list_del(&owner_item->sman_list); + drm_ht_remove_item(&sman->owner_hash_tab, &owner_item->owner_hash); + drm_free(owner_item, sizeof(*owner_item), DRM_MEM_MM); +} + +int drm_sman_owner_clean(struct drm_sman *sman, unsigned long owner) +{ + + struct drm_hash_item *hash_item; + struct drm_owner_item *owner_item; + + if (drm_ht_find_item(&sman->owner_hash_tab, owner, &hash_item)) { + return -1; + } + + owner_item = drm_hash_entry(hash_item, struct drm_owner_item, owner_hash); + DRM_DEBUG("cleaning owner_item %p\n", owner_item); + if (owner_item->mem_blocks.next == &owner_item->mem_blocks) { + drm_sman_remove_owner(sman, owner_item); + return -1; + } + + return 0; +} + +static void drm_sman_do_owner_cleanup(struct drm_sman *sman, + struct drm_owner_item *owner_item) +{ + struct drm_memblock_item *entry, *next; + + list_for_each_entry_safe(entry, next, &owner_item->mem_blocks, + owner_list) { + DRM_DEBUG("freeing mem_block %p\n", entry); + drm_sman_free(entry); + } + drm_sman_remove_owner(sman, owner_item); +} + +void drm_sman_owner_cleanup(struct drm_sman *sman, unsigned long owner) +{ + + struct drm_hash_item *hash_item; + struct drm_owner_item *owner_item; + + if (drm_ht_find_item(&sman->owner_hash_tab, owner, &hash_item)) { + + return; + } + + owner_item = drm_hash_entry(hash_item, struct drm_owner_item, owner_hash); + drm_sman_do_owner_cleanup(sman, owner_item); +} + +void drm_sman_cleanup(struct drm_sman *sman) +{ + struct drm_owner_item *entry, *next; + unsigned int i; + struct drm_sman_mm *sman_mm; + + DRM_DEBUG("sman = %p, owner_items = %p\n", + sman, &sman->owner_items); + list_for_each_entry_safe(entry, next, &sman->owner_items, sman_list) { + DRM_DEBUG("cleaning owner_item = %p\n", entry); + drm_sman_do_owner_cleanup(sman, entry); + } + if (sman->mm) { + for (i = 0; i < sman->num_managers; ++i) { + sman_mm = &sman->mm[i]; + if (sman_mm->private) { + sman_mm->destroy(sman_mm->private); + sman_mm->private = NULL; + } + } + } +} diff --git a/sys/dev/drm/drm_sman.h b/sys/dev/drm/drm_sman.h new file mode 100644 index 0000000..da49b04 --- /dev/null +++ b/sys/dev/drm/drm_sman.h @@ -0,0 +1,180 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + **************************************************************************/ + +#include + +/* + * Simple memory MANager interface that keeps track on allocate regions on a + * per "owner" basis. All regions associated with an "owner" can be released + * with a simple call. Typically if the "owner" exists. The owner is any + * "unsigned long" identifier. Can typically be a pointer to a file private + * struct or a context identifier. + * + * Authors: + * Thomas Hellström + */ + +#ifndef DRM_SMAN_H +#define DRM_SMAN_H + +#include "dev/drm/drm_hashtab.h" +#include "dev/drm/drm_linux_list.h" +#include "dev/drm/drm_mm.h" + +/* + * A class that is an abstration of a simple memory allocator. + * The sman implementation provides a default such allocator + * using the drm_mm.c implementation. But the user can replace it. + * See the SiS implementation, which may use the SiS FB kernel module + * for memory management. + */ + +struct drm_sman_mm { + /* private info. If allocated, needs to be destroyed by the destroy + function */ + void *private; + + /* Allocate a memory block with given size and alignment. + Return an opaque reference to the memory block */ + + void *(*allocate) (void *private, unsigned long size, + unsigned alignment); + + /* Free a memory block. "ref" is the opaque reference that we got from + the "alloc" function */ + + void (*free) (void *private, void *ref); + + /* Free all resources associated with this allocator */ + + void (*destroy) (void *private); + + /* Return a memory offset from the opaque reference returned from the + "alloc" function */ + + unsigned long (*offset) (void *private, void *ref); +}; + +struct drm_memblock_item { + struct list_head owner_list; + struct drm_hash_item user_hash; + void *mm_info; + struct drm_sman_mm *mm; + struct drm_sman *sman; +}; + +struct drm_sman { + struct drm_sman_mm *mm; + int num_managers; + struct drm_open_hash owner_hash_tab; + struct drm_open_hash user_hash_tab; + struct list_head owner_items; +}; + +/* + * Take down a memory manager. This function should only be called after a + * successful init and after a call to drm_sman_cleanup. + */ + +extern void drm_sman_takedown(struct drm_sman * sman); + +/* + * Allocate structures for a manager. + * num_managers are the number of memory pools to manage. (VRAM, AGP, ....) + * user_order is the log2 of the number of buckets in the user hash table. + * set this to approximately log2 of the max number of memory regions + * that will be allocated for _all_ pools together. + * owner_order is the log2 of the number of buckets in the owner hash table. + * set this to approximately log2 of + * the number of client file connections that will + * be using the manager. + * + */ + +extern int drm_sman_init(struct drm_sman * sman, unsigned int num_managers, + unsigned int user_order, unsigned int owner_order); + +/* + * Initialize a drm_mm.c allocator. Should be called only once for each + * manager unless a customized allogator is used. + */ + +extern int drm_sman_set_range(struct drm_sman * sman, unsigned int manager, + unsigned long start, unsigned long size); + +/* + * Initialize a customized allocator for one of the managers. + * (See the SiS module). The object pointed to by "allocator" is copied, + * so it can be destroyed after this call. + */ + +extern int drm_sman_set_manager(struct drm_sman * sman, unsigned int mananger, + struct drm_sman_mm * allocator); + +/* + * Allocate a memory block. Aligment is not implemented yet. + */ + +extern struct drm_memblock_item *drm_sman_alloc(struct drm_sman * sman, + unsigned int manager, + unsigned long size, + unsigned alignment, + unsigned long owner); +/* + * Free a memory block identified by its user hash key. + */ + +extern int drm_sman_free_key(struct drm_sman * sman, unsigned int key); + +/* + * returns 1 iff there are no stale memory blocks associated with this owner. + * Typically called to determine if we need to idle the hardware and call + * drm_sman_owner_cleanup. If there are no stale memory blocks, it removes all + * resources associated with owner. + */ + +extern int drm_sman_owner_clean(struct drm_sman * sman, unsigned long owner); + +/* + * Frees all stale memory blocks associated with this owner. Note that this + * requires that the hardware is finished with all blocks, so the graphics engine + * should be idled before this call is made. This function also frees + * any resources associated with "owner" and should be called when owner + * is not going to be referenced anymore. + */ + +extern void drm_sman_owner_cleanup(struct drm_sman * sman, unsigned long owner); + +/* + * Frees all stale memory blocks associated with the memory manager. + * See idling above. + */ + +extern void drm_sman_cleanup(struct drm_sman * sman); + +#endif diff --git a/sys/dev/drm/drm_sysctl.c b/sys/dev/drm/drm_sysctl.c index 14a7e7c..20f2e7c 100644 --- a/sys/dev/drm/drm_sysctl.c +++ b/sys/dev/drm/drm_sysctl.c @@ -185,7 +185,7 @@ static int drm_vm_info DRM_SYSCTL_HANDLER_ARGS DRM_UNLOCK(); DRM_SYSCTL_PRINT("\nslot offset size " - "type flags address mtrr\n"); + "type flags address handle mtrr\n"); for (i = 0; i < mapcount; i++) { map = &tempmaps[i]; @@ -201,9 +201,11 @@ static int drm_vm_info DRM_SYSCTL_HANDLER_ARGS yesno = "yes"; DRM_SYSCTL_PRINT( - "%4d 0x%016lx 0x%08lx %4.4s 0x%02x 0x%016lx %s\n", i, - map->offset, map->size, type, map->flags, - (unsigned long)map->handle, yesno); + "%4d 0x%016lx 0x%08lx %4.4s 0x%02x 0x%016lx %6d %s\n", + i, map->offset, map->size, type, map->flags, + (unsigned long)map->virtual, + (unsigned int)((unsigned long)map->handle >> + DRM_MAP_HANDLE_SHIFT), yesno); } SYSCTL_OUT(req, "", 1); @@ -295,12 +297,13 @@ static int drm_clients_info DRM_SYSCTL_HANDLER_ARGS DRM_UNLOCK(); - DRM_SYSCTL_PRINT("\na dev pid uid magic ioctls\n"); + DRM_SYSCTL_PRINT( + "\na dev pid uid magic ioctls\n"); for (i = 0; i < privcount; i++) { priv = &tempprivs[i]; - DRM_SYSCTL_PRINT("%c %3d %5d %5d %10u %10lu\n", + DRM_SYSCTL_PRINT("%c %-12s %5d %5d %10u %10lu\n", priv->authenticated ? 'y' : 'n', - priv->minor, + devtoname(priv->dev->devnode), priv->pid, priv->uid, priv->magic, diff --git a/sys/dev/drm/drm_vm.c b/sys/dev/drm/drm_vm.c index 1cbc313..13385fa 100644 --- a/sys/dev/drm/drm_vm.c +++ b/sys/dev/drm/drm_vm.c @@ -50,6 +50,7 @@ int drm_mmap(struct dev_mmap_args *ap) if (!file_priv->authenticated) return EACCES; + DRM_DEBUG("called with offset %016jx\n", offset); if (dev->dma && offset < ptoa(dev->dma->page_count)) { drm_device_dma_t *dma = dev->dma; @@ -67,31 +68,31 @@ int drm_mmap(struct dev_mmap_args *ap) } } - /* A sequential search of a linked list is - fine here because: 1) there will only be - about 5-10 entries in the list and, 2) a - DRI client only has to do this mapping - once, so it doesn't have to be optimized - for performance, even if the list was a - bit longer. */ + /* A sequential search of a linked list is + fine here because: 1) there will only be + about 5-10 entries in the list and, 2) a + DRI client only has to do this mapping + once, so it doesn't have to be optimized + for performance, even if the list was a + bit longer. + */ DRM_LOCK(); TAILQ_FOREACH(map, &dev->maplist, link) { - if (offset >= map->offset && offset < map->offset + map->size) + if (offset >> DRM_MAP_HANDLE_SHIFT == + (unsigned long)map->handle >> DRM_MAP_HANDLE_SHIFT) break; } if (map == NULL) { - DRM_DEBUG("Can't find map, requested offset = %016lx\n", - (unsigned long)offset); + DRM_DEBUG("Can't find map, request offset = %016jx\n", offset); TAILQ_FOREACH(map, &dev->maplist, link) { DRM_DEBUG("map offset = %016lx, handle = %016lx\n", - (unsigned long)map->offset, - (unsigned long)map->handle); + map->offset, (unsigned long)map->handle); } DRM_UNLOCK(); return -1; } - if (((map->flags&_DRM_RESTRICTED) && !DRM_SUSER(DRM_CURPROC))) { + if (((map->flags & _DRM_RESTRICTED) && !DRM_SUSER(DRM_CURPROC))) { DRM_UNLOCK(); DRM_DEBUG("restricted map\n"); return -1; @@ -99,18 +100,28 @@ int drm_mmap(struct dev_mmap_args *ap) type = map->type; DRM_UNLOCK(); + offset = offset & ((1ULL << DRM_MAP_HANDLE_SHIFT) - 1); + switch (type) { case _DRM_FRAME_BUFFER: - case _DRM_REGISTERS: case _DRM_AGP: - phys = offset; - break; - case _DRM_CONSISTENT: - phys = vtophys((char *)map->handle + (offset - map->offset)); +/* + * XXX + * *memattr = VM_MEMATTR_WRITE_COMBINING; + */ + /* FALLTHROUGH */ + case _DRM_REGISTERS: + phys = map->offset + offset; break; case _DRM_SCATTER_GATHER: +/* + * XXX + * *memattr = VM_MEMATTR_WRITE_COMBINING; + */ + /* FALLTHROUGH */ + case _DRM_CONSISTENT: case _DRM_SHM: - phys = vtophys(offset); + phys = vtophys((char *)map->virtual + offset); break; default: DRM_ERROR("bad map type %d\n", type); diff --git a/sys/dev/drm/i915_dma.c b/sys/dev/drm/i915_dma.c index 8d321d4..a9c76ae 100644 --- a/sys/dev/drm/i915_dma.c +++ b/sys/dev/drm/i915_dma.c @@ -148,7 +148,7 @@ static int i915_dma_cleanup(struct drm_device * dev) if (dev_priv->ring.virtual_start) { drm_core_ioremapfree(&dev_priv->ring.map, dev); dev_priv->ring.virtual_start = NULL; - dev_priv->ring.map.handle = NULL; + dev_priv->ring.map.virtual = NULL; dev_priv->ring.map.size = 0; } @@ -171,7 +171,7 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) } dev_priv->sarea_priv = (drm_i915_sarea_t *) - ((u8 *) dev_priv->sarea->handle + init->sarea_priv_offset); + ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); if (init->ring_size != 0) { if (dev_priv->ring.ring_obj != NULL) { @@ -192,7 +192,7 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) drm_core_ioremap_wc(&dev_priv->ring.map, dev); - if (dev_priv->ring.map.handle == NULL) { + if (dev_priv->ring.map.virtual == NULL) { i915_dma_cleanup(dev); DRM_ERROR("can not ioremap virtual address for" " ring buffer\n"); @@ -200,7 +200,7 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) } } - dev_priv->ring.virtual_start = dev_priv->ring.map.handle; + dev_priv->ring.virtual_start = dev_priv->ring.map.virtual; dev_priv->cpp = init->cpp; dev_priv->back_offset = init->back_offset; @@ -226,7 +226,7 @@ static int i915_dma_resume(struct drm_device * dev) return -EINVAL; } - if (dev_priv->ring.map.handle == NULL) { + if (dev_priv->ring.map.virtual == NULL) { DRM_ERROR("can not ioremap virtual address for" " ring buffer\n"); return -ENOMEM; @@ -802,14 +802,14 @@ static int i915_set_status_page(struct drm_device *dev, void *data, dev_priv->hws_map.mtrr = 0; drm_core_ioremap_wc(&dev_priv->hws_map, dev); - if (dev_priv->hws_map.handle == NULL) { + if (dev_priv->hws_map.virtual == NULL) { i915_dma_cleanup(dev); dev_priv->status_gfx_addr = 0; DRM_ERROR("can not ioremap virtual address for" " G33 hw status page\n"); return -ENOMEM; } - dev_priv->hw_status_page = dev_priv->hws_map.handle; + dev_priv->hw_status_page = dev_priv->hws_map.virtual; memset(dev_priv->hw_status_page, 0, PAGE_SIZE); I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); diff --git a/sys/dev/drm/i915_drv.c b/sys/dev/drm/i915_drv.c index c00fe71..c9e5461 100644 --- a/sys/dev/drm/i915_drv.c +++ b/sys/dev/drm/i915_drv.c @@ -31,6 +31,7 @@ #include "dev/drm/drmP.h" #include "dev/drm/drm.h" +#include "dev/drm/drm_mm.h" #include "dev/drm/i915_drm.h" #include "dev/drm/i915_drv.h" #include "dev/drm/drm_pciids.h" diff --git a/sys/dev/drm/i915_drv.h b/sys/dev/drm/i915_drv.h index 711fc68..0d40ebc 100644 --- a/sys/dev/drm/i915_drv.h +++ b/sys/dev/drm/i915_drv.h @@ -30,6 +30,7 @@ #ifndef _I915_DRV_H_ #define _I915_DRV_H_ +#include "dev/drm/drm_mm.h" #include "dev/drm/i915_reg.h" /* General customization: @@ -233,9 +234,8 @@ typedef struct drm_i915_private { u8 saveCR[37]; struct { -#ifdef __linux__ struct drm_mm gtt_space; -#endif + /** * List of objects currently involved in rendering from the * ringbuffer. diff --git a/sys/dev/drm/mach64_dma.c b/sys/dev/drm/mach64_dma.c index c7236d4..ff3f924 100644 --- a/sys/dev/drm/mach64_dma.c +++ b/sys/dev/drm/mach64_dma.c @@ -1075,11 +1075,11 @@ static int mach64_do_dma_init(struct drm_device * dev, drm_mach64_init_t * init) } dev_priv->sarea_priv = (drm_mach64_sarea_t *) - ((u8 *) dev_priv->sarea->handle + init->sarea_priv_offset); + ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); if (!dev_priv->is_pci) { drm_core_ioremap(dev_priv->ring_map, dev); - if (!dev_priv->ring_map->handle) { + if (!dev_priv->ring_map->virtual) { DRM_ERROR("can not ioremap virtual address for" " descriptor ring\n"); dev->dev_private = (void *)dev_priv; @@ -1100,7 +1100,7 @@ static int mach64_do_dma_init(struct drm_device * dev, drm_mach64_init_t * init) dev_priv->dev_buffers = dev->agp_buffer_map; drm_core_ioremap(dev->agp_buffer_map, dev); - if (!dev->agp_buffer_map->handle) { + if (!dev->agp_buffer_map->virtual) { DRM_ERROR("can not ioremap virtual address for" " dma buffer\n"); dev->dev_private = (void *)dev_priv; @@ -1144,7 +1144,7 @@ static int mach64_do_dma_init(struct drm_device * dev, drm_mach64_init_t * init) } dev_priv->ring.size = 0x4000; /* 16KB */ - dev_priv->ring.start = dev_priv->ring_map->handle; + dev_priv->ring.start = dev_priv->ring_map->virtual; dev_priv->ring.start_addr = (u32) dev_priv->ring_map->offset; memset(dev_priv->ring.start, 0, dev_priv->ring.size); diff --git a/sys/dev/drm/mach64_state.c b/sys/dev/drm/mach64_state.c index ec96709..0b4d097 100644 --- a/sys/dev/drm/mach64_state.c +++ b/sys/dev/drm/mach64_state.c @@ -823,7 +823,7 @@ int mach64_dma_vertex(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_mach64_private_t *dev_priv = dev->dev_private; - drm_mach64_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_mach64_sarea_t *sarea_priv; drm_mach64_vertex_t *vertex = data; LOCK_TEST_WITH_RETURN(dev, file_priv); @@ -832,6 +832,7 @@ int mach64_dma_vertex(struct drm_device *dev, void *data, DRM_ERROR("called with no initialization\n"); return -EINVAL; } + sarea_priv = dev_priv->sarea_priv; DRM_DEBUG("pid=%d buf=%p used=%lu discard=%d\n", DRM_CURRENTPID, diff --git a/sys/dev/drm/mga_dma.c b/sys/dev/drm/mga_dma.c index bec9cb5..6710f1f 100644 --- a/sys/dev/drm/mga_dma.c +++ b/sys/dev/drm/mga_dma.c @@ -582,11 +582,11 @@ static int mga_do_agp_dma_bootstrap(struct drm_device *dev, drm_core_ioremap(dev_priv->primary, dev); drm_core_ioremap(dev->agp_buffer_map, dev); - if (!dev_priv->warp->handle || - !dev_priv->primary->handle || !dev->agp_buffer_map->handle) { + if (!dev_priv->warp->virtual || + !dev_priv->primary->virtual || !dev->agp_buffer_map->virtual) { DRM_ERROR("failed to ioremap agp regions! (%p, %p, %p)\n", - dev_priv->warp->handle, dev_priv->primary->handle, - dev->agp_buffer_map->handle); + dev_priv->warp->virtual, dev_priv->primary->virtual, + dev->agp_buffer_map->virtual); return -ENOMEM; } @@ -875,14 +875,14 @@ static int mga_do_init_dma(struct drm_device * dev, drm_mga_init_t * init) } dev_priv->sarea_priv = - (drm_mga_sarea_t *) ((u8 *) dev_priv->sarea->handle + + (drm_mga_sarea_t *) ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); - if (!dev_priv->warp->handle || - !dev_priv->primary->handle || + if (!dev_priv->warp->virtual || + !dev_priv->primary->virtual || ((dev_priv->dma_access != 0) && ((dev->agp_buffer_map == NULL) || - (dev->agp_buffer_map->handle == NULL)))) { + (dev->agp_buffer_map->virtual == NULL)))) { DRM_ERROR("failed to ioremap agp regions!\n"); return -ENOMEM; } @@ -899,7 +899,7 @@ static int mga_do_init_dma(struct drm_device * dev, drm_mga_init_t * init) return ret; } - dev_priv->prim.status = (u32 *) dev_priv->status->handle; + dev_priv->prim.status = (u32 *) dev_priv->status->virtual; mga_do_wait_for_idle(dev_priv); @@ -907,8 +907,8 @@ static int mga_do_init_dma(struct drm_device * dev, drm_mga_init_t * init) */ MGA_WRITE(MGA_PRIMADDRESS, dev_priv->primary->offset | MGA_DMA_GENERAL); - dev_priv->prim.start = (u8 *) dev_priv->primary->handle; - dev_priv->prim.end = ((u8 *) dev_priv->primary->handle + dev_priv->prim.start = (u8 *) dev_priv->primary->virtual; + dev_priv->prim.end = ((u8 *) dev_priv->primary->virtual + dev_priv->primary->size); dev_priv->prim.size = dev_priv->primary->size; diff --git a/sys/dev/drm/mga_warp.c b/sys/dev/drm/mga_warp.c index f7bb011..60626b3 100644 --- a/sys/dev/drm/mga_warp.c +++ b/sys/dev/drm/mga_warp.c @@ -93,7 +93,7 @@ unsigned int mga_warp_microcode_size(const drm_mga_private_t * dev_priv) static int mga_warp_install_g400_microcode(drm_mga_private_t * dev_priv) { - unsigned char *vcbase = dev_priv->warp->handle; + unsigned char *vcbase = dev_priv->warp->virtual; unsigned long pcbase = dev_priv->warp->offset; memset(dev_priv->warp_pipe_phys, 0, sizeof(dev_priv->warp_pipe_phys)); @@ -121,7 +121,7 @@ static int mga_warp_install_g400_microcode(drm_mga_private_t * dev_priv) static int mga_warp_install_g200_microcode(drm_mga_private_t * dev_priv) { - unsigned char *vcbase = dev_priv->warp->handle; + unsigned char *vcbase = dev_priv->warp->virtual; unsigned long pcbase = dev_priv->warp->offset; memset(dev_priv->warp_pipe_phys, 0, sizeof(dev_priv->warp_pipe_phys)); diff --git a/sys/dev/drm/r128_cce.c b/sys/dev/drm/r128_cce.c index c8a2b31..c6012a3 100644 --- a/sys/dev/drm/r128_cce.c +++ b/sys/dev/drm/r128_cce.c @@ -324,8 +324,7 @@ static void r128_cce_init_ring_buffer(struct drm_device * dev, ring_start = dev_priv->cce_ring->offset - dev->agp->base; else #endif - ring_start = dev_priv->cce_ring->offset - - (unsigned long)dev->sg->virtual; + ring_start = dev_priv->cce_ring->offset - dev->sg->vaddr; R128_WRITE(R128_PM4_BUFFER_OFFSET, ring_start | R128_AGP_OFFSET); @@ -506,7 +505,7 @@ static int r128_do_init_cce(struct drm_device * dev, drm_r128_init_t * init) } dev_priv->sarea_priv = - (drm_r128_sarea_t *) ((u8 *) dev_priv->sarea->handle + + (drm_r128_sarea_t *) ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); #if __OS_HAS_AGP @@ -514,9 +513,9 @@ static int r128_do_init_cce(struct drm_device * dev, drm_r128_init_t * init) drm_core_ioremap(dev_priv->cce_ring, dev); drm_core_ioremap(dev_priv->ring_rptr, dev); drm_core_ioremap(dev->agp_buffer_map, dev); - if (!dev_priv->cce_ring->handle || - !dev_priv->ring_rptr->handle || - !dev->agp_buffer_map->handle) { + if (!dev_priv->cce_ring->virtual || + !dev_priv->ring_rptr->virtual || + !dev->agp_buffer_map->virtual) { DRM_ERROR("Could not ioremap agp regions!\n"); dev->dev_private = (void *)dev_priv; r128_do_cleanup_cce(dev); @@ -525,10 +524,11 @@ static int r128_do_init_cce(struct drm_device * dev, drm_r128_init_t * init) } else #endif { - dev_priv->cce_ring->handle = (void *)dev_priv->cce_ring->offset; - dev_priv->ring_rptr->handle = + dev_priv->cce_ring->virtual = + (void *)dev_priv->cce_ring->offset; + dev_priv->ring_rptr->virtual = (void *)dev_priv->ring_rptr->offset; - dev->agp_buffer_map->handle = + dev->agp_buffer_map->virtual = (void *)dev->agp_buffer_map->offset; } @@ -537,10 +537,10 @@ static int r128_do_init_cce(struct drm_device * dev, drm_r128_init_t * init) dev_priv->cce_buffers_offset = dev->agp->base; else #endif - dev_priv->cce_buffers_offset = (unsigned long)dev->sg->virtual; + dev_priv->cce_buffers_offset = dev->sg->vaddr; - dev_priv->ring.start = (u32 *) dev_priv->cce_ring->handle; - dev_priv->ring.end = ((u32 *) dev_priv->cce_ring->handle + dev_priv->ring.start = (u32 *) dev_priv->cce_ring->virtual; + dev_priv->ring.end = ((u32 *) dev_priv->cce_ring->virtual + init->ring_size / sizeof(u32)); dev_priv->ring.size = init->ring_size; dev_priv->ring.size_l2qw = drm_order(init->ring_size / 8); diff --git a/sys/dev/drm/r128_state.c b/sys/dev/drm/r128_state.c index 0eee4aa..c455354 100644 --- a/sys/dev/drm/r128_state.c +++ b/sys/dev/drm/r128_state.c @@ -654,7 +654,7 @@ static void r128_cce_dispatch_indirect(struct drm_device * dev, */ if (dwords & 1) { u32 *data = (u32 *) - ((char *)dev->agp_buffer_map->handle + ((char *)dev->agp_buffer_map->virtual + buf->offset + start); data[dwords++] = cpu_to_le32(R128_CCE_PACKET2); } @@ -719,7 +719,7 @@ static void r128_cce_dispatch_indices(struct drm_device * dev, dwords = (end - start + 3) / sizeof(u32); - data = (u32 *) ((char *)dev->agp_buffer_map->handle + data = (u32 *) ((char *)dev->agp_buffer_map->virtual + buf->offset + start); data[0] = cpu_to_le32(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM, diff --git a/sys/dev/drm/r600_blit.c b/sys/dev/drm/r600_blit.c new file mode 100644 index 0000000..80d4011 --- /dev/null +++ b/sys/dev/drm/r600_blit.c @@ -0,0 +1,1998 @@ +/*- + * Copyright 2009 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Authors: + * Alex Deucher + */ + +#include + +#include "dev/drm/drmP.h" +#include "dev/drm/drm.h" +#include "dev/drm/radeon_drm.h" +#include "dev/drm/radeon_drv.h" + +static u32 r6xx_default_state[] = +{ + 0xc0002400, + 0x00000000, + 0xc0012800, + 0x80000000, + 0x80000000, + 0xc0004600, + 0x00000016, + 0xc0016800, + 0x00000010, + 0x00028000, + 0xc0016800, + 0x00000010, + 0x00008000, + 0xc0016800, + 0x00000542, + 0x07000003, + 0xc0016800, + 0x000005c5, + 0x00000000, + 0xc0016800, + 0x00000363, + 0x00000000, + 0xc0016800, + 0x0000060c, + 0x82000000, + 0xc0016800, + 0x0000060e, + 0x01020204, + 0xc0016f00, + 0x00000000, + 0x00000000, + 0xc0016f00, + 0x00000001, + 0x00000000, + 0xc0096900, + 0x0000022a, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x00000004, + 0x00000000, + 0xc0016900, + 0x0000000a, + 0x00000000, + 0xc0016900, + 0x0000000b, + 0x00000000, + 0xc0016900, + 0x0000010c, + 0x00000000, + 0xc0016900, + 0x0000010d, + 0x00000000, + 0xc0016900, + 0x00000200, + 0x00000000, + 0xc0016900, + 0x00000343, + 0x00000060, + 0xc0016900, + 0x00000344, + 0x00000040, + 0xc0016900, + 0x00000351, + 0x0000aa00, + 0xc0016900, + 0x00000104, + 0x00000000, + 0xc0016900, + 0x0000010e, + 0x00000000, + 0xc0046900, + 0x00000105, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0036900, + 0x00000109, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0046900, + 0x0000030c, + 0x01000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0046900, + 0x00000048, + 0x3f800000, + 0x00000000, + 0x3f800000, + 0x3f800000, + 0xc0016900, + 0x0000008e, + 0x0000000f, + 0xc0016900, + 0x00000080, + 0x00000000, + 0xc0016900, + 0x00000083, + 0x0000ffff, + 0xc0016900, + 0x00000084, + 0x00000000, + 0xc0016900, + 0x00000085, + 0x20002000, + 0xc0016900, + 0x00000086, + 0x00000000, + 0xc0016900, + 0x00000087, + 0x20002000, + 0xc0016900, + 0x00000088, + 0x00000000, + 0xc0016900, + 0x00000089, + 0x20002000, + 0xc0016900, + 0x0000008a, + 0x00000000, + 0xc0016900, + 0x0000008b, + 0x20002000, + 0xc0016900, + 0x0000008c, + 0x00000000, + 0xc0016900, + 0x00000094, + 0x80000000, + 0xc0016900, + 0x00000095, + 0x20002000, + 0xc0026900, + 0x000000b4, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000096, + 0x80000000, + 0xc0016900, + 0x00000097, + 0x20002000, + 0xc0026900, + 0x000000b6, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000098, + 0x80000000, + 0xc0016900, + 0x00000099, + 0x20002000, + 0xc0026900, + 0x000000b8, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009a, + 0x80000000, + 0xc0016900, + 0x0000009b, + 0x20002000, + 0xc0026900, + 0x000000ba, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009c, + 0x80000000, + 0xc0016900, + 0x0000009d, + 0x20002000, + 0xc0026900, + 0x000000bc, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009e, + 0x80000000, + 0xc0016900, + 0x0000009f, + 0x20002000, + 0xc0026900, + 0x000000be, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a0, + 0x80000000, + 0xc0016900, + 0x000000a1, + 0x20002000, + 0xc0026900, + 0x000000c0, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a2, + 0x80000000, + 0xc0016900, + 0x000000a3, + 0x20002000, + 0xc0026900, + 0x000000c2, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a4, + 0x80000000, + 0xc0016900, + 0x000000a5, + 0x20002000, + 0xc0026900, + 0x000000c4, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a6, + 0x80000000, + 0xc0016900, + 0x000000a7, + 0x20002000, + 0xc0026900, + 0x000000c6, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a8, + 0x80000000, + 0xc0016900, + 0x000000a9, + 0x20002000, + 0xc0026900, + 0x000000c8, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000aa, + 0x80000000, + 0xc0016900, + 0x000000ab, + 0x20002000, + 0xc0026900, + 0x000000ca, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000ac, + 0x80000000, + 0xc0016900, + 0x000000ad, + 0x20002000, + 0xc0026900, + 0x000000cc, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000ae, + 0x80000000, + 0xc0016900, + 0x000000af, + 0x20002000, + 0xc0026900, + 0x000000ce, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000b0, + 0x80000000, + 0xc0016900, + 0x000000b1, + 0x20002000, + 0xc0026900, + 0x000000d0, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000b2, + 0x80000000, + 0xc0016900, + 0x000000b3, + 0x20002000, + 0xc0026900, + 0x000000d2, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000293, + 0x00004010, + 0xc0016900, + 0x00000300, + 0x00000000, + 0xc0016900, + 0x00000301, + 0x00000000, + 0xc0016900, + 0x00000312, + 0xffffffff, + 0xc0016900, + 0x00000307, + 0x00000000, + 0xc0016900, + 0x00000308, + 0x00000000, + 0xc0016900, + 0x00000283, + 0x00000000, + 0xc0016900, + 0x00000292, + 0x00000000, + 0xc0066900, + 0x0000010f, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x00000206, + 0x00000000, + 0xc0016900, + 0x00000207, + 0x00000000, + 0xc0016900, + 0x00000208, + 0x00000000, + 0xc0046900, + 0x00000303, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0xc0016900, + 0x00000205, + 0x00000004, + 0xc0016900, + 0x00000280, + 0x00000000, + 0xc0016900, + 0x00000281, + 0x00000000, + 0xc0016900, + 0x0000037e, + 0x00000000, + 0xc0016900, + 0x00000382, + 0x00000000, + 0xc0016900, + 0x00000380, + 0x00000000, + 0xc0016900, + 0x00000383, + 0x00000000, + 0xc0016900, + 0x00000381, + 0x00000000, + 0xc0016900, + 0x00000282, + 0x00000008, + 0xc0016900, + 0x00000302, + 0x0000002d, + 0xc0016900, + 0x0000037f, + 0x00000000, + 0xc0016900, + 0x000001b2, + 0x00000000, + 0xc0016900, + 0x000001b6, + 0x00000000, + 0xc0016900, + 0x000001b7, + 0x00000000, + 0xc0016900, + 0x000001b8, + 0x00000000, + 0xc0016900, + 0x000001b9, + 0x00000000, + 0xc0016900, + 0x00000225, + 0x00000000, + 0xc0016900, + 0x00000229, + 0x00000000, + 0xc0016900, + 0x00000237, + 0x00000000, + 0xc0016900, + 0x00000100, + 0x00000800, + 0xc0016900, + 0x00000101, + 0x00000000, + 0xc0016900, + 0x00000102, + 0x00000000, + 0xc0016900, + 0x000002a8, + 0x00000000, + 0xc0016900, + 0x000002a9, + 0x00000000, + 0xc0016900, + 0x00000103, + 0x00000000, + 0xc0016900, + 0x00000284, + 0x00000000, + 0xc0016900, + 0x00000290, + 0x00000000, + 0xc0016900, + 0x00000285, + 0x00000000, + 0xc0016900, + 0x00000286, + 0x00000000, + 0xc0016900, + 0x00000287, + 0x00000000, + 0xc0016900, + 0x00000288, + 0x00000000, + 0xc0016900, + 0x00000289, + 0x00000000, + 0xc0016900, + 0x0000028a, + 0x00000000, + 0xc0016900, + 0x0000028b, + 0x00000000, + 0xc0016900, + 0x0000028c, + 0x00000000, + 0xc0016900, + 0x0000028d, + 0x00000000, + 0xc0016900, + 0x0000028e, + 0x00000000, + 0xc0016900, + 0x0000028f, + 0x00000000, + 0xc0016900, + 0x000002a1, + 0x00000000, + 0xc0016900, + 0x000002a5, + 0x00000000, + 0xc0016900, + 0x000002ac, + 0x00000000, + 0xc0016900, + 0x000002ad, + 0x00000000, + 0xc0016900, + 0x000002ae, + 0x00000000, + 0xc0016900, + 0x000002c8, + 0x00000000, + 0xc0016900, + 0x00000206, + 0x00000100, + 0xc0016900, + 0x00000204, + 0x00010000, + 0xc0036e00, + 0x00000000, + 0x00000012, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x0000008f, + 0x0000000f, + 0xc0016900, + 0x000001e8, + 0x00000001, + 0xc0016900, + 0x00000202, + 0x00cc0000, + 0xc0016900, + 0x00000205, + 0x00000244, + 0xc0016900, + 0x00000203, + 0x00000210, + 0xc0016900, + 0x000001b1, + 0x00000000, + 0xc0016900, + 0x00000185, + 0x00000000, + 0xc0016900, + 0x000001b3, + 0x00000001, + 0xc0016900, + 0x000001b4, + 0x00000000, + 0xc0016900, + 0x00000191, + 0x00000b00, + 0xc0016900, + 0x000001b5, + 0x00000000, +}; + +static u32 r7xx_default_state[] = +{ + 0xc0012800, + 0x80000000, + 0x80000000, + 0xc0004600, + 0x00000016, + 0xc0016800, + 0x00000010, + 0x00028000, + 0xc0016800, + 0x00000010, + 0x00008000, + 0xc0016800, + 0x00000542, + 0x07000002, + 0xc0016800, + 0x000005c5, + 0x00000000, + 0xc0016800, + 0x00000363, + 0x00004000, + 0xc0016800, + 0x0000060c, + 0x00000000, + 0xc0016800, + 0x0000060e, + 0x00420204, + 0xc0016f00, + 0x00000000, + 0x00000000, + 0xc0016f00, + 0x00000001, + 0x00000000, + 0xc0096900, + 0x0000022a, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x00000004, + 0x00000000, + 0xc0016900, + 0x0000000a, + 0x00000000, + 0xc0016900, + 0x0000000b, + 0x00000000, + 0xc0016900, + 0x0000010c, + 0x00000000, + 0xc0016900, + 0x0000010d, + 0x00000000, + 0xc0016900, + 0x00000200, + 0x00000000, + 0xc0016900, + 0x00000343, + 0x00000060, + 0xc0016900, + 0x00000344, + 0x00000000, + 0xc0016900, + 0x00000351, + 0x0000aa00, + 0xc0016900, + 0x00000104, + 0x00000000, + 0xc0016900, + 0x0000010e, + 0x00000000, + 0xc0046900, + 0x00000105, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0046900, + 0x0000030c, + 0x01000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x0000008e, + 0x0000000f, + 0xc0016900, + 0x00000080, + 0x00000000, + 0xc0016900, + 0x00000083, + 0x0000ffff, + 0xc0016900, + 0x00000084, + 0x00000000, + 0xc0016900, + 0x00000085, + 0x20002000, + 0xc0016900, + 0x00000086, + 0x00000000, + 0xc0016900, + 0x00000087, + 0x20002000, + 0xc0016900, + 0x00000088, + 0x00000000, + 0xc0016900, + 0x00000089, + 0x20002000, + 0xc0016900, + 0x0000008a, + 0x00000000, + 0xc0016900, + 0x0000008b, + 0x20002000, + 0xc0016900, + 0x0000008c, + 0xaaaaaaaa, + 0xc0016900, + 0x00000094, + 0x80000000, + 0xc0016900, + 0x00000095, + 0x20002000, + 0xc0026900, + 0x000000b4, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000096, + 0x80000000, + 0xc0016900, + 0x00000097, + 0x20002000, + 0xc0026900, + 0x000000b6, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000098, + 0x80000000, + 0xc0016900, + 0x00000099, + 0x20002000, + 0xc0026900, + 0x000000b8, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009a, + 0x80000000, + 0xc0016900, + 0x0000009b, + 0x20002000, + 0xc0026900, + 0x000000ba, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009c, + 0x80000000, + 0xc0016900, + 0x0000009d, + 0x20002000, + 0xc0026900, + 0x000000bc, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x0000009e, + 0x80000000, + 0xc0016900, + 0x0000009f, + 0x20002000, + 0xc0026900, + 0x000000be, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a0, + 0x80000000, + 0xc0016900, + 0x000000a1, + 0x20002000, + 0xc0026900, + 0x000000c0, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a2, + 0x80000000, + 0xc0016900, + 0x000000a3, + 0x20002000, + 0xc0026900, + 0x000000c2, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a4, + 0x80000000, + 0xc0016900, + 0x000000a5, + 0x20002000, + 0xc0026900, + 0x000000c4, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a6, + 0x80000000, + 0xc0016900, + 0x000000a7, + 0x20002000, + 0xc0026900, + 0x000000c6, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000a8, + 0x80000000, + 0xc0016900, + 0x000000a9, + 0x20002000, + 0xc0026900, + 0x000000c8, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000aa, + 0x80000000, + 0xc0016900, + 0x000000ab, + 0x20002000, + 0xc0026900, + 0x000000ca, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000ac, + 0x80000000, + 0xc0016900, + 0x000000ad, + 0x20002000, + 0xc0026900, + 0x000000cc, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000ae, + 0x80000000, + 0xc0016900, + 0x000000af, + 0x20002000, + 0xc0026900, + 0x000000ce, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000b0, + 0x80000000, + 0xc0016900, + 0x000000b1, + 0x20002000, + 0xc0026900, + 0x000000d0, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x000000b2, + 0x80000000, + 0xc0016900, + 0x000000b3, + 0x20002000, + 0xc0026900, + 0x000000d2, + 0x00000000, + 0x3f800000, + 0xc0016900, + 0x00000293, + 0x00514000, + 0xc0016900, + 0x00000300, + 0x00000000, + 0xc0016900, + 0x00000301, + 0x00000000, + 0xc0016900, + 0x00000312, + 0xffffffff, + 0xc0016900, + 0x00000307, + 0x00000000, + 0xc0016900, + 0x00000308, + 0x00000000, + 0xc0016900, + 0x00000283, + 0x00000000, + 0xc0016900, + 0x00000292, + 0x00000000, + 0xc0066900, + 0x0000010f, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x00000206, + 0x00000000, + 0xc0016900, + 0x00000207, + 0x00000000, + 0xc0016900, + 0x00000208, + 0x00000000, + 0xc0046900, + 0x00000303, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0x3f800000, + 0xc0016900, + 0x00000205, + 0x00000004, + 0xc0016900, + 0x00000280, + 0x00000000, + 0xc0016900, + 0x00000281, + 0x00000000, + 0xc0016900, + 0x0000037e, + 0x00000000, + 0xc0016900, + 0x00000382, + 0x00000000, + 0xc0016900, + 0x00000380, + 0x00000000, + 0xc0016900, + 0x00000383, + 0x00000000, + 0xc0016900, + 0x00000381, + 0x00000000, + 0xc0016900, + 0x00000282, + 0x00000008, + 0xc0016900, + 0x00000302, + 0x0000002d, + 0xc0016900, + 0x0000037f, + 0x00000000, + 0xc0016900, + 0x000001b2, + 0x00000001, + 0xc0016900, + 0x000001b6, + 0x00000000, + 0xc0016900, + 0x000001b7, + 0x00000000, + 0xc0016900, + 0x000001b8, + 0x00000000, + 0xc0016900, + 0x000001b9, + 0x00000000, + 0xc0016900, + 0x00000225, + 0x00000000, + 0xc0016900, + 0x00000229, + 0x00000000, + 0xc0016900, + 0x00000237, + 0x00000000, + 0xc0016900, + 0x00000100, + 0x00000800, + 0xc0016900, + 0x00000101, + 0x00000000, + 0xc0016900, + 0x00000102, + 0x00000000, + 0xc0016900, + 0x000002a8, + 0x00000000, + 0xc0016900, + 0x000002a9, + 0x00000000, + 0xc0016900, + 0x00000103, + 0x00000000, + 0xc0016900, + 0x00000284, + 0x00000000, + 0xc0016900, + 0x00000290, + 0x00000000, + 0xc0016900, + 0x00000285, + 0x00000000, + 0xc0016900, + 0x00000286, + 0x00000000, + 0xc0016900, + 0x00000287, + 0x00000000, + 0xc0016900, + 0x00000288, + 0x00000000, + 0xc0016900, + 0x00000289, + 0x00000000, + 0xc0016900, + 0x0000028a, + 0x00000000, + 0xc0016900, + 0x0000028b, + 0x00000000, + 0xc0016900, + 0x0000028c, + 0x00000000, + 0xc0016900, + 0x0000028d, + 0x00000000, + 0xc0016900, + 0x0000028e, + 0x00000000, + 0xc0016900, + 0x0000028f, + 0x00000000, + 0xc0016900, + 0x000002a1, + 0x00000000, + 0xc0016900, + 0x000002a5, + 0x00000000, + 0xc0016900, + 0x000002ac, + 0x00000000, + 0xc0016900, + 0x000002ad, + 0x00000000, + 0xc0016900, + 0x000002ae, + 0x00000000, + 0xc0016900, + 0x000002c8, + 0x00000000, + 0xc0016900, + 0x00000206, + 0x00000100, + 0xc0016900, + 0x00000204, + 0x00010000, + 0xc0036e00, + 0x00000000, + 0x00000012, + 0x00000000, + 0x00000000, + 0xc0016900, + 0x0000008f, + 0x0000000f, + 0xc0016900, + 0x000001e8, + 0x00000001, + 0xc0016900, + 0x00000202, + 0x00cc0000, + 0xc0016900, + 0x00000205, + 0x00000244, + 0xc0016900, + 0x00000203, + 0x00000210, + 0xc0016900, + 0x000001b1, + 0x00000000, + 0xc0016900, + 0x00000185, + 0x00000000, + 0xc0016900, + 0x000001b3, + 0x00000001, + 0xc0016900, + 0x000001b4, + 0x00000000, + 0xc0016900, + 0x00000191, + 0x00000b00, + 0xc0016900, + 0x000001b5, + 0x00000000, +}; + +/* same for r6xx/r7xx */ +static u32 r6xx_vs[] = +{ + 0x00000004, + 0x81000000, + 0x0000203c, + 0x94000b08, + 0x00004000, + 0x14200b1a, + 0x00000000, + 0x00000000, + 0x3c000000, + 0x68cd1000, + 0x00080000, + 0x00000000, +}; + +static u32 r6xx_ps[] = +{ + 0x00000002, + 0x80800000, + 0x00000000, + 0x94200688, + 0x00000010, + 0x000d1000, + 0xb0800000, + 0x00000000, +}; + +#define DI_PT_RECTLIST 0x11 +#define DI_INDEX_SIZE_16_BIT 0x0 +#define DI_SRC_SEL_AUTO_INDEX 0x2 + +#define FMT_8 1 +#define FMT_5_6_5 8 +#define FMT_8_8_8_8 0x1a +#define COLOR_8 1 +#define COLOR_5_6_5 8 +#define COLOR_8_8_8_8 0x1a + +#define R600_CB0_DEST_BASE_ENA (1 << 6) +#define R600_TC_ACTION_ENA (1 << 23) +#define R600_VC_ACTION_ENA (1 << 24) +#define R600_CB_ACTION_ENA (1 << 25) +#define R600_DB_ACTION_ENA (1 << 26) +#define R600_SH_ACTION_ENA (1 << 27) +#define R600_SMX_ACTION_ENA (1 << 28) + +#define R600_CB_COLOR0_SIZE 0x28060 +#define R600_CB_COLOR0_VIEW 0x28080 +#define R600_CB_COLOR0_INFO 0x280a0 +#define R600_CB_COLOR0_TILE 0x280c0 +#define R600_CB_COLOR0_FRAG 0x280e0 +#define R600_CB_COLOR0_MASK 0x28100 + +#define R600_SQ_PGM_START_VS 0x28858 +#define R600_SQ_PGM_RESOURCES_VS 0x28868 +#define R600_SQ_PGM_CF_OFFSET_VS 0x288d0 +#define R600_SQ_PGM_START_PS 0x28840 +#define R600_SQ_PGM_RESOURCES_PS 0x28850 +#define R600_SQ_PGM_EXPORTS_PS 0x28854 +#define R600_SQ_PGM_CF_OFFSET_PS 0x288cc + +#define R600_VGT_PRIMITIVE_TYPE 0x8958 + +#define R600_PA_SC_SCREEN_SCISSOR_TL 0x28030 +#define R600_PA_SC_GENERIC_SCISSOR_TL 0x28240 +#define R600_PA_SC_WINDOW_SCISSOR_TL 0x28204 + +#define R600_SQ_TEX_VTX_INVALID_TEXTURE 0x0 +#define R600_SQ_TEX_VTX_INVALID_BUFFER 0x1 +#define R600_SQ_TEX_VTX_VALID_TEXTURE 0x2 +#define R600_SQ_TEX_VTX_VALID_BUFFER 0x3 + +/* packet 3 type offsets */ +#define R600_SET_CONFIG_REG_OFFSET 0x00008000 +#define R600_SET_CONFIG_REG_END 0x0000ac00 +#define R600_SET_CONTEXT_REG_OFFSET 0x00028000 +#define R600_SET_CONTEXT_REG_END 0x00029000 +#define R600_SET_ALU_CONST_OFFSET 0x00030000 +#define R600_SET_ALU_CONST_END 0x00032000 +#define R600_SET_RESOURCE_OFFSET 0x00038000 +#define R600_SET_RESOURCE_END 0x0003c000 +#define R600_SET_SAMPLER_OFFSET 0x0003c000 +#define R600_SET_SAMPLER_END 0x0003cff0 +#define R600_SET_CTL_CONST_OFFSET 0x0003cff0 +#define R600_SET_CTL_CONST_END 0x0003e200 +#define R600_SET_LOOP_CONST_OFFSET 0x0003e200 +#define R600_SET_LOOP_CONST_END 0x0003e380 +#define R600_SET_BOOL_CONST_OFFSET 0x0003e380 +#define R600_SET_BOOL_CONST_END 0x00040000 + +/* Packet 3 types */ +#define R600_IT_INDIRECT_BUFFER_END 0x00001700 +#define R600_IT_SET_PREDICATION 0x00002000 +#define R600_IT_REG_RMW 0x00002100 +#define R600_IT_COND_EXEC 0x00002200 +#define R600_IT_PRED_EXEC 0x00002300 +#define R600_IT_START_3D_CMDBUF 0x00002400 +#define R600_IT_DRAW_INDEX_2 0x00002700 +#define R600_IT_CONTEXT_CONTROL 0x00002800 +#define R600_IT_DRAW_INDEX_IMMD_BE 0x00002900 +#define R600_IT_INDEX_TYPE 0x00002A00 +#define R600_IT_DRAW_INDEX 0x00002B00 +#define R600_IT_DRAW_INDEX_AUTO 0x00002D00 +#define R600_IT_DRAW_INDEX_IMMD 0x00002E00 +#define R600_IT_NUM_INSTANCES 0x00002F00 +#define R600_IT_STRMOUT_BUFFER_UPDATE 0x00003400 +#define R600_IT_INDIRECT_BUFFER_MP 0x00003800 +#define R600_IT_MEM_SEMAPHORE 0x00003900 +#define R600_IT_MPEG_INDEX 0x00003A00 +#define R600_IT_WAIT_REG_MEM 0x00003C00 +#define R600_IT_MEM_WRITE 0x00003D00 +#define R600_IT_INDIRECT_BUFFER 0x00003200 +#define R600_IT_CP_INTERRUPT 0x00004000 +#define R600_IT_SURFACE_SYNC 0x00004300 +#define R600_IT_ME_INITIALIZE 0x00004400 +#define R600_IT_COND_WRITE 0x00004500 +#define R600_IT_EVENT_WRITE 0x00004600 +#define R600_IT_EVENT_WRITE_EOP 0x00004700 +#define R600_IT_ONE_REG_WRITE 0x00005700 +#define R600_IT_SET_CONFIG_REG 0x00006800 +#define R600_IT_SET_CONTEXT_REG 0x00006900 +#define R600_IT_SET_ALU_CONST 0x00006A00 +#define R600_IT_SET_BOOL_CONST 0x00006B00 +#define R600_IT_SET_LOOP_CONST 0x00006C00 +#define R600_IT_SET_RESOURCE 0x00006D00 +#define R600_IT_SET_SAMPLER 0x00006E00 +#define R600_IT_SET_CTL_CONST 0x00006F00 +#define R600_IT_SURFACE_BASE_UPDATE 0x00007300 + +static inline void +set_render_target(drm_radeon_private_t *dev_priv, int format, int w, int h, u64 gpu_addr) +{ + u32 cb_color_info; + int pitch, slice; + RING_LOCALS; + DRM_DEBUG("\n"); + + h = (h + 7) & ~7; + if (h < 8) + h = 8; + + cb_color_info = ((format << 2) | (1 << 27)); + pitch = (w / 8) - 1; + slice = ((w * h) / 64) - 1; + + if (((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_R600) && + ((dev_priv->flags & RADEON_FAMILY_MASK) < CHIP_RV770)) { + BEGIN_RING(21 + 2); + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_BASE - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(gpu_addr >> 8); + OUT_RING(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0)); + OUT_RING(2 << 0); + } else { + BEGIN_RING(21); + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_BASE - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(gpu_addr >> 8); + } + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_SIZE - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING((pitch << 0) | (slice << 10)); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_VIEW - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_INFO - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(cb_color_info); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_TILE - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_FRAG - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_CB_COLOR0_MASK - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + + ADVANCE_RING(); +} + +static inline void +cp_set_surface_sync(drm_radeon_private_t *dev_priv, + u32 sync_type, u32 size, u64 mc_addr) +{ + u32 cp_coher_size; + RING_LOCALS; + DRM_DEBUG("\n"); + + if (size == 0xffffffff) + cp_coher_size = 0xffffffff; + else + cp_coher_size = ((size + 255) >> 8); + + BEGIN_RING(5); + OUT_RING(CP_PACKET3(R600_IT_SURFACE_SYNC, 3)); + OUT_RING(sync_type); + OUT_RING(cp_coher_size); + OUT_RING((mc_addr >> 8)); + OUT_RING(10); /* poll interval */ + ADVANCE_RING(); +} + +static inline void +set_shaders(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + u64 gpu_addr; + int shader_size, i; + u32 *vs, *ps; + uint32_t sq_pgm_resources; + RING_LOCALS; + DRM_DEBUG("\n"); + + /* load shaders */ + vs = (u32 *) ((char *)dev->agp_buffer_map->virtual + dev_priv->blit_vb->offset); + ps = (u32 *) ((char *)dev->agp_buffer_map->virtual + dev_priv->blit_vb->offset + 256); + + shader_size = sizeof(r6xx_vs) / 4; + for (i= 0; i < shader_size; i++) + vs[i] = r6xx_vs[i]; + shader_size = sizeof(r6xx_ps) / 4; + for (i= 0; i < shader_size; i++) + ps[i] = r6xx_ps[i]; + + dev_priv->blit_vb->used = 512; + + gpu_addr = dev_priv->gart_buffers_offset + dev_priv->blit_vb->offset; + + /* setup shader regs */ + sq_pgm_resources = (1 << 0); + + BEGIN_RING(9 + 12); + /* VS */ + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_START_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(gpu_addr >> 8); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_RESOURCES_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(sq_pgm_resources); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_CF_OFFSET_VS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + + /* PS */ + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_START_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING((gpu_addr + 256) >> 8); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_RESOURCES_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(sq_pgm_resources | (1 << 28)); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_EXPORTS_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(2); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 1)); + OUT_RING((R600_SQ_PGM_CF_OFFSET_PS - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING(0); + ADVANCE_RING(); + + cp_set_surface_sync(dev_priv, + R600_SH_ACTION_ENA, 512, gpu_addr); +} + +static inline void +set_vtx_resource(drm_radeon_private_t *dev_priv, u64 gpu_addr) +{ + uint32_t sq_vtx_constant_word2; + RING_LOCALS; + DRM_DEBUG("\n"); + + sq_vtx_constant_word2 = (((gpu_addr >> 32) & 0xff) | (16 << 8)); + + BEGIN_RING(9); + OUT_RING(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); + OUT_RING(0x460); + OUT_RING(gpu_addr & 0xffffffff); + OUT_RING(48 - 1); + OUT_RING(sq_vtx_constant_word2); + OUT_RING(1 << 0); + OUT_RING(0); + OUT_RING(0); + OUT_RING(R600_SQ_TEX_VTX_VALID_BUFFER << 30); + ADVANCE_RING(); + + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV610) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV620) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS780) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS880) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV710)) + cp_set_surface_sync(dev_priv, + R600_TC_ACTION_ENA, 48, gpu_addr); + else + cp_set_surface_sync(dev_priv, + R600_VC_ACTION_ENA, 48, gpu_addr); +} + +static inline void +set_tex_resource(drm_radeon_private_t *dev_priv, + int format, int w, int h, int pitch, u64 gpu_addr) +{ + uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4; + RING_LOCALS; + DRM_DEBUG("\n"); + + if (h < 1) + h = 1; + + sq_tex_resource_word0 = (1 << 0); + sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 8) | + ((w - 1) << 19)); + + sq_tex_resource_word1 = (format << 26); + sq_tex_resource_word1 |= ((h - 1) << 0); + + sq_tex_resource_word4 = ((1 << 14) | + (0 << 16) | + (1 << 19) | + (2 << 22) | + (3 << 25)); + + BEGIN_RING(9); + OUT_RING(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); + OUT_RING(0); + OUT_RING(sq_tex_resource_word0); + OUT_RING(sq_tex_resource_word1); + OUT_RING(gpu_addr >> 8); + OUT_RING(gpu_addr >> 8); + OUT_RING(sq_tex_resource_word4); + OUT_RING(0); + OUT_RING(R600_SQ_TEX_VTX_VALID_TEXTURE << 30); + ADVANCE_RING(); + +} + +static inline void +set_scissors(drm_radeon_private_t *dev_priv, int x1, int y1, int x2, int y2) +{ + RING_LOCALS; + DRM_DEBUG("\n"); + + BEGIN_RING(12); + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2)); + OUT_RING((R600_PA_SC_SCREEN_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING((x1 << 0) | (y1 << 16)); + OUT_RING((x2 << 0) | (y2 << 16)); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2)); + OUT_RING((R600_PA_SC_GENERIC_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING((x1 << 0) | (y1 << 16) | (1 << 31)); + OUT_RING((x2 << 0) | (y2 << 16)); + + OUT_RING(CP_PACKET3(R600_IT_SET_CONTEXT_REG, 2)); + OUT_RING((R600_PA_SC_WINDOW_SCISSOR_TL - R600_SET_CONTEXT_REG_OFFSET) >> 2); + OUT_RING((x1 << 0) | (y1 << 16) | (1 << 31)); + OUT_RING((x2 << 0) | (y2 << 16)); + ADVANCE_RING(); +} + +static inline void +draw_auto(drm_radeon_private_t *dev_priv) +{ + RING_LOCALS; + DRM_DEBUG("\n"); + + BEGIN_RING(10); + OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_RING((R600_VGT_PRIMITIVE_TYPE - R600_SET_CONFIG_REG_OFFSET) >> 2); + OUT_RING(DI_PT_RECTLIST); + + OUT_RING(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + OUT_RING(DI_INDEX_SIZE_16_BIT); + + OUT_RING(CP_PACKET3(R600_IT_NUM_INSTANCES, 0)); + OUT_RING(1); + + OUT_RING(CP_PACKET3(R600_IT_DRAW_INDEX_AUTO, 1)); + OUT_RING(3); + OUT_RING(DI_SRC_SEL_AUTO_INDEX); + + ADVANCE_RING(); + COMMIT_RING(); +} + +static inline void +set_default_state(drm_radeon_private_t *dev_priv) +{ + int default_state_dw, i; + u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2; + u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2; + int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs; + int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads; + int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries; + RING_LOCALS; + + switch ((dev_priv->flags & RADEON_FAMILY_MASK)) { + case CHIP_R600: + num_ps_gprs = 192; + num_vs_gprs = 56; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 136; + num_vs_threads = 48; + num_gs_threads = 4; + num_es_threads = 4; + num_ps_stack_entries = 128; + num_vs_stack_entries = 128; + num_gs_stack_entries = 0; + num_es_stack_entries = 0; + break; + case CHIP_RV630: + case CHIP_RV635: + num_ps_gprs = 84; + num_vs_gprs = 36; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 144; + num_vs_threads = 40; + num_gs_threads = 4; + num_es_threads = 4; + num_ps_stack_entries = 40; + num_vs_stack_entries = 40; + num_gs_stack_entries = 32; + num_es_stack_entries = 16; + break; + case CHIP_RV610: + case CHIP_RV620: + case CHIP_RS780: + case CHIP_RS880: + default: + num_ps_gprs = 84; + num_vs_gprs = 36; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 136; + num_vs_threads = 48; + num_gs_threads = 4; + num_es_threads = 4; + num_ps_stack_entries = 40; + num_vs_stack_entries = 40; + num_gs_stack_entries = 32; + num_es_stack_entries = 16; + break; + case CHIP_RV670: + num_ps_gprs = 144; + num_vs_gprs = 40; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 136; + num_vs_threads = 48; + num_gs_threads = 4; + num_es_threads = 4; + num_ps_stack_entries = 40; + num_vs_stack_entries = 40; + num_gs_stack_entries = 32; + num_es_stack_entries = 16; + break; + case CHIP_RV770: + num_ps_gprs = 192; + num_vs_gprs = 56; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 188; + num_vs_threads = 60; + num_gs_threads = 0; + num_es_threads = 0; + num_ps_stack_entries = 256; + num_vs_stack_entries = 256; + num_gs_stack_entries = 0; + num_es_stack_entries = 0; + break; + case CHIP_RV730: + case CHIP_RV740: + num_ps_gprs = 84; + num_vs_gprs = 36; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 188; + num_vs_threads = 60; + num_gs_threads = 0; + num_es_threads = 0; + num_ps_stack_entries = 128; + num_vs_stack_entries = 128; + num_gs_stack_entries = 0; + num_es_stack_entries = 0; + break; + case CHIP_RV710: + num_ps_gprs = 192; + num_vs_gprs = 56; + num_temp_gprs = 4; + num_gs_gprs = 0; + num_es_gprs = 0; + num_ps_threads = 144; + num_vs_threads = 48; + num_gs_threads = 0; + num_es_threads = 0; + num_ps_stack_entries = 128; + num_vs_stack_entries = 128; + num_gs_stack_entries = 0; + num_es_stack_entries = 0; + break; + } + + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV610) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV620) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS780) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS880) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV710)) + sq_config = 0; + else + sq_config = R600_VC_ENABLE; + + sq_config |= (R600_DX9_CONSTS | + R600_ALU_INST_PREFER_VECTOR | + R600_PS_PRIO(0) | + R600_VS_PRIO(1) | + R600_GS_PRIO(2) | + R600_ES_PRIO(3)); + + sq_gpr_resource_mgmt_1 = (R600_NUM_PS_GPRS(num_ps_gprs) | + R600_NUM_VS_GPRS(num_vs_gprs) | + R600_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs)); + sq_gpr_resource_mgmt_2 = (R600_NUM_GS_GPRS(num_gs_gprs) | + R600_NUM_ES_GPRS(num_es_gprs)); + sq_thread_resource_mgmt = (R600_NUM_PS_THREADS(num_ps_threads) | + R600_NUM_VS_THREADS(num_vs_threads) | + R600_NUM_GS_THREADS(num_gs_threads) | + R600_NUM_ES_THREADS(num_es_threads)); + sq_stack_resource_mgmt_1 = (R600_NUM_PS_STACK_ENTRIES(num_ps_stack_entries) | + R600_NUM_VS_STACK_ENTRIES(num_vs_stack_entries)); + sq_stack_resource_mgmt_2 = (R600_NUM_GS_STACK_ENTRIES(num_gs_stack_entries) | + R600_NUM_ES_STACK_ENTRIES(num_es_stack_entries)); + + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV770) { + default_state_dw = sizeof(r7xx_default_state) / 4; + BEGIN_RING(default_state_dw + 10); + for (i = 0; i < default_state_dw; i++) + OUT_RING(r7xx_default_state[i]); + } else { + default_state_dw = sizeof(r6xx_default_state) / 4; + BEGIN_RING(default_state_dw + 10); + for (i = 0; i < default_state_dw; i++) + OUT_RING(r6xx_default_state[i]); + } + OUT_RING(CP_PACKET3(R600_IT_EVENT_WRITE, 0)); + OUT_RING(R600_CACHE_FLUSH_AND_INV_EVENT); + /* SQ config */ + OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 6)); + OUT_RING((R600_SQ_CONFIG - R600_SET_CONFIG_REG_OFFSET) >> 2); + OUT_RING(sq_config); + OUT_RING(sq_gpr_resource_mgmt_1); + OUT_RING(sq_gpr_resource_mgmt_2); + OUT_RING(sq_thread_resource_mgmt); + OUT_RING(sq_stack_resource_mgmt_1); + OUT_RING(sq_stack_resource_mgmt_2); + ADVANCE_RING(); +} + +static inline uint32_t i2f(uint32_t input) +{ + u32 result, i, exponent, fraction; + + if ((input & 0x3fff) == 0) + result = 0; /* 0 is a special case */ + else { + exponent = 140; /* exponent biased by 127; */ + fraction = (input & 0x3fff) << 10; /* cheat and only + handle numbers below 2^^15 */ + for (i = 0; i < 14; i++) { + if (fraction & 0x800000) + break; + else { + fraction = fraction << 1; /* keep + shifting left until top bit = 1 */ + exponent = exponent -1; + } + } + result = exponent << 23 | (fraction & 0x7fffff); /* mask + off top bit; assumed 1 */ + } + return result; +} + +int +r600_prepare_blit_copy(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG("\n"); + + dev_priv->blit_vb = radeon_freelist_get(dev); + if (!dev_priv->blit_vb) { + DRM_ERROR("Unable to allocate vertex buffer for blit\n"); + return -EAGAIN; + } + + set_default_state(dev_priv); + set_shaders(dev); + + return 0; +} + +void +r600_done_blit_copy(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + RING_LOCALS; + DRM_DEBUG("\n"); + + BEGIN_RING(5); + OUT_RING(CP_PACKET3(R600_IT_EVENT_WRITE, 0)); + OUT_RING(R600_CACHE_FLUSH_AND_INV_EVENT); + /* wait for 3D idle clean */ + OUT_RING(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_RING((R600_WAIT_UNTIL - R600_SET_CONFIG_REG_OFFSET) >> 2); + OUT_RING(RADEON_WAIT_3D_IDLE | RADEON_WAIT_3D_IDLECLEAN); + + ADVANCE_RING(); + COMMIT_RING(); + + dev_priv->blit_vb->used = 0; + radeon_cp_discard_buffer(dev, dev_priv->blit_vb); +} + +void +r600_blit_copy(struct drm_device *dev, + uint64_t src_gpu_addr, uint64_t dst_gpu_addr, + int size_bytes) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + int max_bytes; + u64 vb_addr; + u32 *vb; + + vb = (u32 *) ((char *)dev->agp_buffer_map->virtual + + dev_priv->blit_vb->offset + dev_priv->blit_vb->used); + DRM_DEBUG("src=0x%016jx, dst=0x%016jx, size=%d\n", + src_gpu_addr, dst_gpu_addr, size_bytes); + + if ((size_bytes & 3) || (src_gpu_addr & 3) || (dst_gpu_addr & 3)) { + max_bytes = 8192; + + while (size_bytes) { + int cur_size = size_bytes; + int src_x = src_gpu_addr & 255; + int dst_x = dst_gpu_addr & 255; + int h = 1; + src_gpu_addr = src_gpu_addr & ~255; + dst_gpu_addr = dst_gpu_addr & ~255; + + if (!src_x && !dst_x) { + h = (cur_size / max_bytes); + if (h > 8192) + h = 8192; + if (h == 0) + h = 1; + else + cur_size = max_bytes; + } else { + if (cur_size > max_bytes) + cur_size = max_bytes; + if (cur_size > (max_bytes - dst_x)) + cur_size = (max_bytes - dst_x); + if (cur_size > (max_bytes - src_x)) + cur_size = (max_bytes - src_x); + } + + if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) { + dev_priv->blit_vb->used = 0; + radeon_cp_discard_buffer(dev, dev_priv->blit_vb); + dev_priv->blit_vb = radeon_freelist_get(dev); + if (!dev_priv->blit_vb) + return; + set_shaders(dev); + vb = (u32 *) ((char *)dev->agp_buffer_map->virtual + + dev_priv->blit_vb->offset + dev_priv->blit_vb->used); + } + + vb[0] = i2f(dst_x); + vb[1] = 0; + vb[2] = i2f(src_x); + vb[3] = 0; + + vb[4] = i2f(dst_x); + vb[5] = i2f(h); + vb[6] = i2f(src_x); + vb[7] = i2f(h); + + vb[8] = i2f(dst_x + cur_size); + vb[9] = i2f(h); + vb[10] = i2f(src_x + cur_size); + vb[11] = i2f(h); + + /* src */ + set_tex_resource(dev_priv, FMT_8, + src_x + cur_size, h, src_x + cur_size, + src_gpu_addr); + + cp_set_surface_sync(dev_priv, + R600_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr); + + /* dst */ + set_render_target(dev_priv, COLOR_8, + dst_x + cur_size, h, + dst_gpu_addr); + + /* scissors */ + set_scissors(dev_priv, dst_x, 0, dst_x + cur_size, h); + + /* Vertex buffer setup */ + vb_addr = dev_priv->gart_buffers_offset + + dev_priv->blit_vb->offset + + dev_priv->blit_vb->used; + set_vtx_resource(dev_priv, vb_addr); + + /* draw */ + draw_auto(dev_priv); + + cp_set_surface_sync(dev_priv, + R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA, + cur_size * h, dst_gpu_addr); + + vb += 12; + dev_priv->blit_vb->used += 12 * 4; + + src_gpu_addr += cur_size * h; + dst_gpu_addr += cur_size * h; + size_bytes -= cur_size * h; + } + } else { + max_bytes = 8192 * 4; + + while (size_bytes) { + int cur_size = size_bytes; + int src_x = (src_gpu_addr & 255); + int dst_x = (dst_gpu_addr & 255); + int h = 1; + src_gpu_addr = src_gpu_addr & ~255; + dst_gpu_addr = dst_gpu_addr & ~255; + + if (!src_x && !dst_x) { + h = (cur_size / max_bytes); + if (h > 8192) + h = 8192; + if (h == 0) + h = 1; + else + cur_size = max_bytes; + } else { + if (cur_size > max_bytes) + cur_size = max_bytes; + if (cur_size > (max_bytes - dst_x)) + cur_size = (max_bytes - dst_x); + if (cur_size > (max_bytes - src_x)) + cur_size = (max_bytes - src_x); + } + + if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) { + dev_priv->blit_vb->used = 0; + radeon_cp_discard_buffer(dev, dev_priv->blit_vb); + dev_priv->blit_vb = radeon_freelist_get(dev); + if (!dev_priv->blit_vb) + return; + set_shaders(dev); + vb = (u32 *) ((char *)dev->agp_buffer_map->virtual + + dev_priv->blit_vb->offset + dev_priv->blit_vb->used); + } + + vb[0] = i2f(dst_x / 4); + vb[1] = 0; + vb[2] = i2f(src_x / 4); + vb[3] = 0; + + vb[4] = i2f(dst_x / 4); + vb[5] = i2f(h); + vb[6] = i2f(src_x / 4); + vb[7] = i2f(h); + + vb[8] = i2f((dst_x + cur_size) / 4); + vb[9] = i2f(h); + vb[10] = i2f((src_x + cur_size) / 4); + vb[11] = i2f(h); + + /* src */ + set_tex_resource(dev_priv, FMT_8_8_8_8, + (src_x + cur_size) / 4, + h, (src_x + cur_size) / 4, + src_gpu_addr); + + cp_set_surface_sync(dev_priv, + R600_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr); + + /* dst */ + set_render_target(dev_priv, COLOR_8_8_8_8, + (dst_x + cur_size) / 4, h, + dst_gpu_addr); + + /* scissors */ + set_scissors(dev_priv, (dst_x / 4), 0, (dst_x + cur_size / 4), h); + + /* Vertex buffer setup */ + vb_addr = dev_priv->gart_buffers_offset + + dev_priv->blit_vb->offset + + dev_priv->blit_vb->used; + set_vtx_resource(dev_priv, vb_addr); + + /* draw */ + draw_auto(dev_priv); + + cp_set_surface_sync(dev_priv, + R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA, + cur_size * h, dst_gpu_addr); + + vb += 12; + dev_priv->blit_vb->used += 12 * 4; + + src_gpu_addr += cur_size * h; + dst_gpu_addr += cur_size * h; + size_bytes -= cur_size * h; + } + } +} + +void +r600_blit_swap(struct drm_device *dev, + uint64_t src_gpu_addr, uint64_t dst_gpu_addr, + int sx, int sy, int dx, int dy, + int w, int h, int src_pitch, int dst_pitch, int cpp) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + int cb_format, tex_format; + int sx2, sy2, dx2, dy2; + u64 vb_addr; + u32 *vb; + + if ((dev_priv->blit_vb->used + 48) > dev_priv->blit_vb->total) { + dev_priv->blit_vb->used = 0; + radeon_cp_discard_buffer(dev, dev_priv->blit_vb); + dev_priv->blit_vb = radeon_freelist_get(dev); + if (!dev_priv->blit_vb) + return; + set_shaders(dev); + } + vb = (u32 *) ((char *)dev->agp_buffer_map->virtual + + dev_priv->blit_vb->offset + dev_priv->blit_vb->used); + + sx2 = sx + w; + sy2 = sy + h; + dx2 = dx + w; + dy2 = dy + h; + + vb[0] = i2f(dx); + vb[1] = i2f(dy); + vb[2] = i2f(sx); + vb[3] = i2f(sy); + + vb[4] = i2f(dx); + vb[5] = i2f(dy2); + vb[6] = i2f(sx); + vb[7] = i2f(sy2); + + vb[8] = i2f(dx2); + vb[9] = i2f(dy2); + vb[10] = i2f(sx2); + vb[11] = i2f(sy2); + + switch(cpp) { + case 4: + cb_format = COLOR_8_8_8_8; + tex_format = FMT_8_8_8_8; + break; + case 2: + cb_format = COLOR_5_6_5; + tex_format = FMT_5_6_5; + break; + default: + cb_format = COLOR_8; + tex_format = FMT_8; + break; + } + + /* src */ + set_tex_resource(dev_priv, tex_format, + src_pitch / cpp, + sy2, src_pitch / cpp, + src_gpu_addr); + + cp_set_surface_sync(dev_priv, + R600_TC_ACTION_ENA, src_pitch * sy2, src_gpu_addr); + + /* dst */ + set_render_target(dev_priv, cb_format, + dst_pitch / cpp, dy2, + dst_gpu_addr); + + /* scissors */ + set_scissors(dev_priv, dx, dy, dx2, dy2); + + /* Vertex buffer setup */ + vb_addr = dev_priv->gart_buffers_offset + + dev_priv->blit_vb->offset + + dev_priv->blit_vb->used; + set_vtx_resource(dev_priv, vb_addr); + + /* draw */ + draw_auto(dev_priv); + + cp_set_surface_sync(dev_priv, + R600_CB_ACTION_ENA | R600_CB0_DEST_BASE_ENA, + dst_pitch * dy2, dst_gpu_addr); + + dev_priv->blit_vb->used += 12 * 4; +} diff --git a/sys/dev/drm/r600_cp.c b/sys/dev/drm/r600_cp.c index 9af4b71..0634c99 100644 --- a/sys/dev/drm/r600_cp.c +++ b/sys/dev/drm/r600_cp.c @@ -177,7 +177,7 @@ int r600_page_table_init(struct drm_device *dev) entry_addr = entry->busaddr[i]; for (j = 0; j < (PAGE_SIZE / ATI_PCIGART_PAGE_SIZE); j++) { page_base = (u64) entry_addr & ATI_PCIGART_PAGE_MASK; - page_base |= R600_PTE_VALID | R600_PTE_SYSTEM | R600_PTE_SNOOPED; + page_base |= R600_PTE_VALID | R600_PTE_SYSTEM; page_base |= R600_PTE_READABLE | R600_PTE_WRITEABLE; *pci_gart = page_base; @@ -1319,6 +1319,11 @@ static void r700_gfx_init(struct drm_device *dev, dev_priv->r700_sc_prim_fifo_size = 0x40; dev_priv->r700_sc_hiz_tile_fifo_size = 0x30; dev_priv->r700_sc_earlyz_tile_fifo_fize = 0x130; + + if (dev_priv->r600_sx_max_export_pos_size > 16) { + dev_priv->r600_sx_max_export_pos_size -= 16; + dev_priv->r600_sx_max_export_smx_size += 16; + } break; default: break; @@ -1667,9 +1672,8 @@ static void r600_cp_init_ring_buffer(struct drm_device *dev, } else #endif { - rptr_addr = dev_priv->ring_rptr->offset - - ((unsigned long) dev->sg->virtual) - + dev_priv->gart_vm_start; + rptr_addr = dev_priv->ring_rptr->offset - dev->sg->vaddr + + dev_priv->gart_vm_start; } RADEON_WRITE(R600_CP_RB_RPTR_ADDR, rptr_addr & 0xffffffff); @@ -1703,9 +1707,8 @@ static void r600_cp_init_ring_buffer(struct drm_device *dev, + dev_priv->gart_vm_start); } else #endif - ring_start = (dev_priv->cp_ring->offset - - (unsigned long)dev->sg->virtual - + dev_priv->gart_vm_start); + ring_start = dev_priv->cp_ring->offset - dev->sg->vaddr + + dev_priv->gart_vm_start; RADEON_WRITE(R600_CP_RB_BASE, ring_start >> 8); @@ -1840,6 +1843,7 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, */ dev_priv->vblank_crtc = DRM_RADEON_VBLANK_CRTC1; + dev_priv->do_boxes = 0; dev_priv->cp_mode = init->cp_mode; /* We don't support anything other than bus-mastering ring mode, @@ -1910,7 +1914,7 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, } dev_priv->sarea_priv = - (drm_radeon_sarea_t *) ((u8 *) dev_priv->sarea->handle + + (drm_radeon_sarea_t *) ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); #if __OS_HAS_AGP @@ -1919,9 +1923,9 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, drm_core_ioremap_wc(dev_priv->cp_ring, dev); drm_core_ioremap_wc(dev_priv->ring_rptr, dev); drm_core_ioremap_wc(dev->agp_buffer_map, dev); - if (!dev_priv->cp_ring->handle || - !dev_priv->ring_rptr->handle || - !dev->agp_buffer_map->handle) { + if (!dev_priv->cp_ring->virtual || + !dev_priv->ring_rptr->virtual || + !dev->agp_buffer_map->virtual) { DRM_ERROR("could not find ioremap agp regions!\n"); r600_do_cleanup_cp(dev); return -EINVAL; @@ -1929,18 +1933,19 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, } else #endif { - dev_priv->cp_ring->handle = (void *)dev_priv->cp_ring->offset; - dev_priv->ring_rptr->handle = + dev_priv->cp_ring->virtual = + (void *)dev_priv->cp_ring->offset; + dev_priv->ring_rptr->virtual = (void *)dev_priv->ring_rptr->offset; - dev->agp_buffer_map->handle = + dev->agp_buffer_map->virtual = (void *)dev->agp_buffer_map->offset; - DRM_DEBUG("dev_priv->cp_ring->handle %p\n", - dev_priv->cp_ring->handle); - DRM_DEBUG("dev_priv->ring_rptr->handle %p\n", - dev_priv->ring_rptr->handle); - DRM_DEBUG("dev->agp_buffer_map->handle %p\n", - dev->agp_buffer_map->handle); + DRM_DEBUG("dev_priv->cp_ring->virtual %p\n", + dev_priv->cp_ring->virtual); + DRM_DEBUG("dev_priv->ring_rptr->virtual %p\n", + dev_priv->ring_rptr->virtual); + DRM_DEBUG("dev->agp_buffer_map->virtual %p\n", + dev->agp_buffer_map->virtual); } dev_priv->fb_location = (radeon_read_fb_location(dev_priv) & 0xffff) << 24; @@ -2007,9 +2012,8 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, + dev_priv->gart_vm_start); else #endif - dev_priv->gart_buffers_offset = (dev->agp_buffer_map->offset - - (unsigned long)dev->sg->virtual - + dev_priv->gart_vm_start); + dev_priv->gart_buffers_offset = dev->agp_buffer_map->offset - + dev->sg->vaddr + dev_priv->gart_vm_start; DRM_DEBUG("fb 0x%08x size %d\n", (unsigned int) dev_priv->fb_location, @@ -2020,8 +2024,8 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, DRM_DEBUG("dev_priv->gart_buffers_offset 0x%08lx\n", dev_priv->gart_buffers_offset); - dev_priv->ring.start = (u32 *) dev_priv->cp_ring->handle; - dev_priv->ring.end = ((u32 *) dev_priv->cp_ring->handle + dev_priv->ring.start = (u32 *) dev_priv->cp_ring->virtual; + dev_priv->ring.end = ((u32 *) dev_priv->cp_ring->virtual + init->ring_size / sizeof(u32)); dev_priv->ring.size = init->ring_size; dev_priv->ring.size_l2qw = drm_order(init->ring_size / 8); @@ -2060,14 +2064,14 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, dev_priv->gart_info.table_size; drm_core_ioremap_wc(&dev_priv->gart_info.mapping, dev); - if (!dev_priv->gart_info.mapping.handle) { + if (!dev_priv->gart_info.mapping.virtual) { DRM_ERROR("ioremap failed.\n"); r600_do_cleanup_cp(dev); return -EINVAL; } dev_priv->gart_info.addr = - dev_priv->gart_info.mapping.handle; + dev_priv->gart_info.mapping.virtual; DRM_DEBUG("Setting phys_pci_gart to %p %08lX\n", dev_priv->gart_info.addr, @@ -2097,6 +2101,8 @@ int r600_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, r600_do_engine_reset(dev); r600_test_writeback(dev_priv); + r600_cs_init(dev); + return 0; } @@ -2213,7 +2219,7 @@ int r600_cp_dispatch_indirect(struct drm_device *dev, */ while (dwords & 0xf) { u32 *data = (u32 *) - ((char *)dev->agp_buffer_map->handle + ((char *)dev->agp_buffer_map->virtual + buf->offset + start); data[dwords++] = RADEON_CP_PACKET2; } @@ -2229,3 +2235,136 @@ int r600_cp_dispatch_indirect(struct drm_device *dev, return 0; } + +void r600_cp_dispatch_swap(struct drm_device * dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + int nbox = sarea_priv->nbox; + struct drm_clip_rect *pbox = sarea_priv->boxes; + int i, cpp, src_pitch, dst_pitch; + uint64_t src, dst; + RING_LOCALS; + DRM_DEBUG("\n"); + + if (dev_priv->color_fmt == RADEON_COLOR_FORMAT_ARGB8888) + cpp = 4; + else + cpp = 2; + + if (dev_priv->sarea_priv->pfCurrentPage == 0) { + src_pitch = dev_priv->back_pitch; + dst_pitch = dev_priv->front_pitch; + src = dev_priv->back_offset + dev_priv->fb_location; + dst = dev_priv->front_offset + dev_priv->fb_location; + } else { + src_pitch = dev_priv->front_pitch; + dst_pitch = dev_priv->back_pitch; + src = dev_priv->front_offset + dev_priv->fb_location; + dst = dev_priv->back_offset + dev_priv->fb_location; + } + + if (r600_prepare_blit_copy(dev)) { + DRM_ERROR("unable to allocate vertex buffer for swap buffer\n"); + return; + } + for (i = 0; i < nbox; i++) { + int x = pbox[i].x1; + int y = pbox[i].y1; + int w = pbox[i].x2 - x; + int h = pbox[i].y2 - y; + + DRM_DEBUG("%d,%d-%d,%d\n", x, y, w, h); + + r600_blit_swap(dev, + src, dst, + x, y, x, y, w, h, + src_pitch, dst_pitch, cpp); + } + r600_done_blit_copy(dev); + + /* Increment the frame counter. The client-side 3D driver must + * throttle the framerate by waiting for this value before + * performing the swapbuffer ioctl. + */ + dev_priv->sarea_priv->last_frame++; + + BEGIN_RING(3); + R600_FRAME_AGE(dev_priv->sarea_priv->last_frame); + ADVANCE_RING(); +} + +int r600_cp_dispatch_texture(struct drm_device * dev, + struct drm_file *file_priv, + drm_radeon_texture_t * tex, + drm_radeon_tex_image_t * image) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_buf *buf; + u32 *buffer; + const u8 __user *data; + int size, pass_size; + u64 src_offset, dst_offset; + + if (!radeon_check_offset(dev_priv, tex->offset)) { + DRM_ERROR("Invalid destination offset\n"); + return -EINVAL; + } + + /* this might fail for zero-sized uploads - are those illegal? */ + if (!radeon_check_offset(dev_priv, tex->offset + tex->height * tex->pitch - 1)) { + DRM_ERROR("Invalid final destination offset\n"); + return -EINVAL; + } + + size = tex->height * tex->pitch; + + if (size == 0) + return 0; + + dst_offset = tex->offset; + + r600_prepare_blit_copy(dev); + do { + data = (const u8 __user *)image->data; + pass_size = size; + + buf = radeon_freelist_get(dev); + if (!buf) { + DRM_DEBUG("EAGAIN\n"); + if (DRM_COPY_TO_USER(tex->image, image, sizeof(*image))) + return -EFAULT; + return -EAGAIN; + } + + if (pass_size > buf->total) + pass_size = buf->total; + + /* Dispatch the indirect buffer. + */ + buffer = + (u32 *) ((char *)dev->agp_buffer_map->virtual + + buf->offset); + + if (DRM_COPY_FROM_USER(buffer, data, pass_size)) { + DRM_ERROR("EFAULT on pad, %d bytes\n", pass_size); + return -EFAULT; + } + + buf->file_priv = file_priv; + buf->used = pass_size; + src_offset = dev_priv->gart_buffers_offset + buf->offset; + + r600_blit_copy(dev, src_offset, dst_offset, pass_size); + + radeon_cp_discard_buffer(dev, buf); + + /* Update the input parameters for next time */ + image->data = (const u8 __user *)image->data + pass_size; + dst_offset += pass_size; + size -= pass_size; + } while (size > 0); + r600_done_blit_copy(dev); + + return 0; +} diff --git a/sys/dev/drm/radeon/Makefile b/sys/dev/drm/radeon/Makefile index 313381d..031683e 100644 --- a/sys/dev/drm/radeon/Makefile +++ b/sys/dev/drm/radeon/Makefile @@ -1,7 +1,7 @@ .PATH: ${.CURDIR}/.. KMOD = radeon -SRCS = r300_cmdbuf.c r600_cp.c radeon_cp.c radeon_drv.c radeon_state.c \ - radeon_irq.c radeon_mem.c +SRCS = r300_cmdbuf.c r600_blit.c r600_cp.c radeon_cp.c radeon_cs.c \ + radeon_drv.c radeon_irq.c radeon_mem.c radeon_state.c SRCS += device_if.h bus_if.h pci_if.h opt_drm.h CFLAGS += ${DEBUG_FLAGS} -I. -I.. diff --git a/sys/dev/drm/radeon_cp.c b/sys/dev/drm/radeon_cp.c index db5d1fa..7ba1ff4 100644 --- a/sys/dev/drm/radeon_cp.c +++ b/sys/dev/drm/radeon_cp.c @@ -50,7 +50,7 @@ u32 radeon_read_ring_rptr(drm_radeon_private_t *dev_priv, u32 off) val = DRM_READ32(dev_priv->ring_rptr, off); } else { val = *(((volatile u32 *) - dev_priv->ring_rptr->handle) + + dev_priv->ring_rptr->virtual) + (off / sizeof(u32))); val = le32_to_cpu(val); } @@ -74,7 +74,7 @@ void radeon_write_ring_rptr(drm_radeon_private_t *dev_priv, u32 off, u32 val) if (dev_priv->flags & RADEON_IS_AGP) DRM_WRITE32(dev_priv->ring_rptr, off, val); else - *(((volatile u32 *) dev_priv->ring_rptr->handle) + + *(((volatile u32 *) dev_priv->ring_rptr->virtual) + (off / sizeof(u32))) = cpu_to_le32(val); } @@ -405,6 +405,15 @@ static void radeon_init_pipes(drm_radeon_private_t *dev_priv) { uint32_t gb_tile_config, gb_pipe_sel = 0; + if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV530) { + uint32_t z_pipe_sel = RADEON_READ(RV530_GB_PIPE_SELECT2); + if ((z_pipe_sel & 3) == 3) + dev_priv->num_z_pipes = 2; + else + dev_priv->num_z_pipes = 1; + } else + dev_priv->num_z_pipes = 1; + /* RS4xx/RS6xx/R4xx/R5xx */ if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R420) { gb_pipe_sel = RADEON_READ(R400_GB_PIPE_SELECT); @@ -708,9 +717,8 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, + dev_priv->gart_vm_start); } else #endif - ring_start = (dev_priv->cp_ring->offset - - (unsigned long)dev->sg->virtual - + dev_priv->gart_vm_start); + ring_start = (dev_priv->cp_ring->offset - dev->sg->vaddr + + dev_priv->gart_vm_start); RADEON_WRITE(RADEON_CP_RB_BASE, ring_start); @@ -732,9 +740,8 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, #endif { RADEON_WRITE(RADEON_CP_RB_RPTR_ADDR, - dev_priv->ring_rptr->offset - - ((unsigned long) dev->sg->virtual) - + dev_priv->gart_vm_start); + dev_priv->ring_rptr->offset - dev->sg->vaddr + + dev_priv->gart_vm_start); } /* Set ring buffer size */ @@ -1266,7 +1273,7 @@ static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, } dev_priv->sarea_priv = - (drm_radeon_sarea_t *) ((u8 *) dev_priv->sarea->handle + + (drm_radeon_sarea_t *) ((u8 *) dev_priv->sarea->virtual + init->sarea_priv_offset); #if __OS_HAS_AGP @@ -1274,9 +1281,9 @@ static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, drm_core_ioremap_wc(dev_priv->cp_ring, dev); drm_core_ioremap_wc(dev_priv->ring_rptr, dev); drm_core_ioremap_wc(dev->agp_buffer_map, dev); - if (!dev_priv->cp_ring->handle || - !dev_priv->ring_rptr->handle || - !dev->agp_buffer_map->handle) { + if (!dev_priv->cp_ring->virtual || + !dev_priv->ring_rptr->virtual || + !dev->agp_buffer_map->virtual) { DRM_ERROR("could not find ioremap agp regions!\n"); radeon_do_cleanup_cp(dev); return -EINVAL; @@ -1284,19 +1291,19 @@ static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, } else #endif { - dev_priv->cp_ring->handle = + dev_priv->cp_ring->virtual = (void *)(unsigned long)dev_priv->cp_ring->offset; - dev_priv->ring_rptr->handle = + dev_priv->ring_rptr->virtual = (void *)(unsigned long)dev_priv->ring_rptr->offset; - dev->agp_buffer_map->handle = + dev->agp_buffer_map->virtual = (void *)(unsigned long)dev->agp_buffer_map->offset; - DRM_DEBUG("dev_priv->cp_ring->handle %p\n", - dev_priv->cp_ring->handle); - DRM_DEBUG("dev_priv->ring_rptr->handle %p\n", - dev_priv->ring_rptr->handle); - DRM_DEBUG("dev->agp_buffer_map->handle %p\n", - dev->agp_buffer_map->handle); + DRM_DEBUG("dev_priv->cp_ring->virtual %p\n", + dev_priv->cp_ring->virtual); + DRM_DEBUG("dev_priv->ring_rptr->virtual %p\n", + dev_priv->ring_rptr->virtual); + DRM_DEBUG("dev->agp_buffer_map->virtual %p\n", + dev->agp_buffer_map->virtual); } dev_priv->fb_location = (radeon_read_fb_location(dev_priv) & 0xffff) << 16; @@ -1365,17 +1372,16 @@ static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, + dev_priv->gart_vm_start); else #endif - dev_priv->gart_buffers_offset = (dev->agp_buffer_map->offset - - (unsigned long)dev->sg->virtual - + dev_priv->gart_vm_start); + dev_priv->gart_buffers_offset = dev->agp_buffer_map->offset - + dev->sg->vaddr + dev_priv->gart_vm_start; DRM_DEBUG("dev_priv->gart_size %d\n", dev_priv->gart_size); DRM_DEBUG("dev_priv->gart_vm_start 0x%x\n", dev_priv->gart_vm_start); DRM_DEBUG("dev_priv->gart_buffers_offset 0x%lx\n", dev_priv->gart_buffers_offset); - dev_priv->ring.start = (u32 *) dev_priv->cp_ring->handle; - dev_priv->ring.end = ((u32 *) dev_priv->cp_ring->handle + dev_priv->ring.start = (u32 *) dev_priv->cp_ring->virtual; + dev_priv->ring.end = ((u32 *) dev_priv->cp_ring->virtual + init->ring_size / sizeof(u32)); dev_priv->ring.size = init->ring_size; dev_priv->ring.size_l2qw = drm_order(init->ring_size / 8); @@ -1411,7 +1417,7 @@ static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, drm_core_ioremap_wc(&dev_priv->gart_info.mapping, dev); dev_priv->gart_info.addr = - dev_priv->gart_info.mapping.handle; + dev_priv->gart_info.mapping.virtual; if (dev_priv->flags & RADEON_IS_PCIE) dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCIE; @@ -1858,8 +1864,8 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev) for (t = 0; t < dev_priv->usec_timeout; t++) { u32 done_age = GET_SCRATCH(dev_priv, 1); DRM_DEBUG("done_age = %d\n", done_age); - for (i = start; i < dma->buf_count; i++) { - buf = dma->buflist[i]; + for (i = 0; i < dma->buf_count; i++) { + buf = dma->buflist[start]; buf_priv = buf->dev_private; if (buf->file_priv == NULL || (buf->pending && buf_priv->age <= @@ -1868,7 +1874,8 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev) buf->pending = 0; return buf; } - start = 0; + if (++start >= dma->buf_count) + start = 0; } if (t) { @@ -1877,46 +1884,8 @@ struct drm_buf *radeon_freelist_get(struct drm_device * dev) } } - DRM_DEBUG("returning NULL!\n"); - return NULL; -} - -#if 0 -struct drm_buf *radeon_freelist_get(struct drm_device * dev) -{ - struct drm_device_dma *dma = dev->dma; - drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_buf_priv_t *buf_priv; - struct drm_buf *buf; - int i, t; - int start; - u32 done_age; - - done_age = radeon_read_ring_rptr(dev_priv, RADEON_SCRATCHOFF(1)); - if (++dev_priv->last_buf >= dma->buf_count) - dev_priv->last_buf = 0; - - start = dev_priv->last_buf; - dev_priv->stats.freelist_loops++; - - for (t = 0; t < 2; t++) { - for (i = start; i < dma->buf_count; i++) { - buf = dma->buflist[i]; - buf_priv = buf->dev_private; - if (buf->file_priv == 0 || (buf->pending && - buf_priv->age <= - done_age)) { - dev_priv->stats.requested_bufs++; - buf->pending = 0; - return buf; - } - } - start = 0; - } - return NULL; } -#endif void radeon_freelist_reset(struct drm_device * dev) { @@ -2133,9 +2102,9 @@ void radeon_commit_ring(drm_radeon_private_t *dev_priv) /* check if the ring is padded out to 16-dword alignment */ - tail_aligned = dev_priv->ring.tail & 0xf; + tail_aligned = dev_priv->ring.tail & (RADEON_RING_ALIGN - 1); if (tail_aligned) { - int num_p2 = 16 - tail_aligned; + int num_p2 = RADEON_RING_ALIGN - tail_aligned; ring = dev_priv->ring.start; /* pad with some CP_PACKET2 */ diff --git a/sys/dev/drm/radeon_cs.c b/sys/dev/drm/radeon_cs.c new file mode 100644 index 0000000..ceb7e6c --- /dev/null +++ b/sys/dev/drm/radeon_cs.c @@ -0,0 +1,856 @@ +/*- + * Copyright 2008 Jerome Glisse. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Authors: + * Jerome Glisse + */ + +#include + +#include "dev/drm/drmP.h" +#include "dev/drm/radeon_drm.h" +#include "dev/drm/radeon_drv.h" + +/* regs */ +#define AVIVO_D1MODE_VLINE_START_END 0x6538 +#define AVIVO_D2MODE_VLINE_START_END 0x6d38 +#define R600_CP_COHER_BASE 0x85f8 +#define R600_DB_DEPTH_BASE 0x2800c +#define R600_CB_COLOR0_BASE 0x28040 +#define R600_CB_COLOR1_BASE 0x28044 +#define R600_CB_COLOR2_BASE 0x28048 +#define R600_CB_COLOR3_BASE 0x2804c +#define R600_CB_COLOR4_BASE 0x28050 +#define R600_CB_COLOR5_BASE 0x28054 +#define R600_CB_COLOR6_BASE 0x28058 +#define R600_CB_COLOR7_BASE 0x2805c +#define R600_SQ_PGM_START_FS 0x28894 +#define R600_SQ_PGM_START_ES 0x28880 +#define R600_SQ_PGM_START_VS 0x28858 +#define R600_SQ_PGM_START_GS 0x2886c +#define R600_SQ_PGM_START_PS 0x28840 +#define R600_VGT_DMA_BASE 0x287e8 +#define R600_VGT_DMA_BASE_HI 0x287e4 +#define R600_VGT_STRMOUT_BASE_OFFSET_0 0x28b10 +#define R600_VGT_STRMOUT_BASE_OFFSET_1 0x28b14 +#define R600_VGT_STRMOUT_BASE_OFFSET_2 0x28b18 +#define R600_VGT_STRMOUT_BASE_OFFSET_3 0x28b1c +#define R600_VGT_STRMOUT_BASE_OFFSET_HI_0 0x28b44 +#define R600_VGT_STRMOUT_BASE_OFFSET_HI_1 0x28b48 +#define R600_VGT_STRMOUT_BASE_OFFSET_HI_2 0x28b4c +#define R600_VGT_STRMOUT_BASE_OFFSET_HI_3 0x28b50 +#define R600_VGT_STRMOUT_BUFFER_BASE_0 0x28ad8 +#define R600_VGT_STRMOUT_BUFFER_BASE_1 0x28ae8 +#define R600_VGT_STRMOUT_BUFFER_BASE_2 0x28af8 +#define R600_VGT_STRMOUT_BUFFER_BASE_3 0x28b08 +#define R600_VGT_STRMOUT_BUFFER_OFFSET_0 0x28adc +#define R600_VGT_STRMOUT_BUFFER_OFFSET_1 0x28aec +#define R600_VGT_STRMOUT_BUFFER_OFFSET_2 0x28afc +#define R600_VGT_STRMOUT_BUFFER_OFFSET_3 0x28b0c + +/* resource type */ +#define R600_SQ_TEX_VTX_INVALID_TEXTURE 0x0 +#define R600_SQ_TEX_VTX_INVALID_BUFFER 0x1 +#define R600_SQ_TEX_VTX_VALID_TEXTURE 0x2 +#define R600_SQ_TEX_VTX_VALID_BUFFER 0x3 + +/* packet 3 type offsets */ +#define R600_SET_CONFIG_REG_OFFSET 0x00008000 +#define R600_SET_CONFIG_REG_END 0x0000ac00 +#define R600_SET_CONTEXT_REG_OFFSET 0x00028000 +#define R600_SET_CONTEXT_REG_END 0x00029000 +#define R600_SET_ALU_CONST_OFFSET 0x00030000 +#define R600_SET_ALU_CONST_END 0x00032000 +#define R600_SET_RESOURCE_OFFSET 0x00038000 +#define R600_SET_RESOURCE_END 0x0003c000 +#define R600_SET_SAMPLER_OFFSET 0x0003c000 +#define R600_SET_SAMPLER_END 0x0003cff0 +#define R600_SET_CTL_CONST_OFFSET 0x0003cff0 +#define R600_SET_CTL_CONST_END 0x0003e200 +#define R600_SET_LOOP_CONST_OFFSET 0x0003e200 +#define R600_SET_LOOP_CONST_END 0x0003e380 +#define R600_SET_BOOL_CONST_OFFSET 0x0003e380 +#define R600_SET_BOOL_CONST_END 0x00040000 + +/* Packet 3 types */ +#define R600_IT_INDIRECT_BUFFER_END 0x00001700 +#define R600_IT_SET_PREDICATION 0x00002000 +#define R600_IT_REG_RMW 0x00002100 +#define R600_IT_COND_EXEC 0x00002200 +#define R600_IT_PRED_EXEC 0x00002300 +#define R600_IT_START_3D_CMDBUF 0x00002400 +#define R600_IT_DRAW_INDEX_2 0x00002700 +#define R600_IT_CONTEXT_CONTROL 0x00002800 +#define R600_IT_DRAW_INDEX_IMMD_BE 0x00002900 +#define R600_IT_INDEX_TYPE 0x00002A00 +#define R600_IT_DRAW_INDEX 0x00002B00 +#define R600_IT_DRAW_INDEX_AUTO 0x00002D00 +#define R600_IT_DRAW_INDEX_IMMD 0x00002E00 +#define R600_IT_NUM_INSTANCES 0x00002F00 +#define R600_IT_STRMOUT_BUFFER_UPDATE 0x00003400 +#define R600_IT_INDIRECT_BUFFER_MP 0x00003800 +#define R600_IT_MEM_SEMAPHORE 0x00003900 +#define R600_IT_MPEG_INDEX 0x00003A00 +#define R600_IT_WAIT_REG_MEM 0x00003C00 +#define R600_IT_MEM_WRITE 0x00003D00 +#define R600_IT_INDIRECT_BUFFER 0x00003200 +#define R600_IT_CP_INTERRUPT 0x00004000 +#define R600_IT_SURFACE_SYNC 0x00004300 +#define R600_IT_ME_INITIALIZE 0x00004400 +#define R600_IT_COND_WRITE 0x00004500 +#define R600_IT_EVENT_WRITE 0x00004600 +#define R600_IT_EVENT_WRITE_EOP 0x00004700 +#define R600_IT_ONE_REG_WRITE 0x00005700 +#define R600_IT_SET_CONFIG_REG 0x00006800 +#define R600_IT_SET_CONTEXT_REG 0x00006900 +#define R600_IT_SET_ALU_CONST 0x00006A00 +#define R600_IT_SET_BOOL_CONST 0x00006B00 +#define R600_IT_SET_LOOP_CONST 0x00006C00 +#define R600_IT_SET_RESOURCE 0x00006D00 +#define R600_IT_SET_SAMPLER 0x00006E00 +#define R600_IT_SET_CTL_CONST 0x00006F00 +#define R600_IT_SURFACE_BASE_UPDATE 0x00007300 + +int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *fpriv) +{ + struct drm_radeon_cs_parser parser; + struct drm_radeon_private *dev_priv = dev->dev_private; + struct drm_radeon_cs *cs = data; + uint32_t cs_id; + struct drm_radeon_cs_chunk __user **chunk_ptr = NULL; + uint64_t *chunk_array; + uint64_t *chunk_array_ptr; + long size; + int r, i; + + spin_lock(&dev_priv->cs.cs_mutex); + /* set command stream id to 0 which is fake id */ + cs_id = 0; + cs->cs_id = cs_id; + + if (dev_priv == NULL) { + DRM_ERROR("called with no initialization\n"); + spin_unlock(&dev_priv->cs.cs_mutex); + return -EINVAL; + } + if (!cs->num_chunks) { + spin_unlock(&dev_priv->cs.cs_mutex); + return 0; + } + + + chunk_array = drm_calloc(cs->num_chunks, sizeof(uint64_t), DRM_MEM_DRIVER); + if (!chunk_array) { + spin_unlock(&dev_priv->cs.cs_mutex); + return -ENOMEM; + } + + chunk_array_ptr = (uint64_t *)(unsigned long)(cs->chunks); + + if (DRM_COPY_FROM_USER(chunk_array, chunk_array_ptr, sizeof(uint64_t)*cs->num_chunks)) { + r = -EFAULT; + goto out; + } + + parser.dev = dev; + parser.file_priv = fpriv; + parser.reloc_index = -1; + parser.ib_index = -1; + parser.num_chunks = cs->num_chunks; + /* copy out the chunk headers */ + parser.chunks = drm_calloc(parser.num_chunks, sizeof(struct drm_radeon_kernel_chunk), DRM_MEM_DRIVER); + if (!parser.chunks) { + r = -ENOMEM; + goto out; + } + + for (i = 0; i < parser.num_chunks; i++) { + struct drm_radeon_cs_chunk user_chunk; + + chunk_ptr = (void __user *)(unsigned long)chunk_array[i]; + + if (DRM_COPY_FROM_USER(&user_chunk, chunk_ptr, sizeof(struct drm_radeon_cs_chunk))){ + r = -EFAULT; + goto out; + } + parser.chunks[i].chunk_id = user_chunk.chunk_id; + + if (parser.chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) + parser.reloc_index = i; + + if (parser.chunks[i].chunk_id == RADEON_CHUNK_ID_IB) + parser.ib_index = i; + + if (parser.chunks[i].chunk_id == RADEON_CHUNK_ID_OLD) { + parser.ib_index = i; + parser.reloc_index = -1; + } + + parser.chunks[i].length_dw = user_chunk.length_dw; + parser.chunks[i].chunk_data = (uint32_t *)(unsigned long)user_chunk.chunk_data; + + parser.chunks[i].kdata = NULL; + size = parser.chunks[i].length_dw * sizeof(uint32_t); + + switch(parser.chunks[i].chunk_id) { + case RADEON_CHUNK_ID_IB: + case RADEON_CHUNK_ID_OLD: + if (size == 0) { + r = -EINVAL; + goto out; + } + case RADEON_CHUNK_ID_RELOCS: + if (size) { + parser.chunks[i].kdata = drm_alloc(size, DRM_MEM_DRIVER); + if (!parser.chunks[i].kdata) { + r = -ENOMEM; + goto out; + } + + if (DRM_COPY_FROM_USER(parser.chunks[i].kdata, parser.chunks[i].chunk_data, size)) { + r = -EFAULT; + goto out; + } + } else + parser.chunks[i].kdata = NULL; + break; + default: + break; + } + DRM_DEBUG("chunk %d %d %d %p\n", i, parser.chunks[i].chunk_id, parser.chunks[i].length_dw, + parser.chunks[i].chunk_data); + } + + if (parser.chunks[parser.ib_index].length_dw > (16 * 1024)) { + DRM_ERROR("cs->dwords too big: %d\n", parser.chunks[parser.ib_index].length_dw); + r = -EINVAL; + goto out; + } + + /* get ib */ + r = dev_priv->cs.ib_get(&parser); + if (r) { + DRM_ERROR("ib_get failed\n"); + goto out; + } + + /* now parse command stream */ + r = dev_priv->cs.parse(&parser); + if (r) { + goto out; + } + +out: + dev_priv->cs.ib_free(&parser, r); + + /* emit cs id sequence */ + dev_priv->cs.id_emit(&parser, &cs_id); + + cs->cs_id = cs_id; + + spin_unlock(&dev_priv->cs.cs_mutex); + + for (i = 0; i < parser.num_chunks; i++) { + if (parser.chunks[i].kdata) + drm_free(parser.chunks[i].kdata, parser.chunks[i].length_dw * sizeof(uint32_t), DRM_MEM_DRIVER); + } + + drm_free(parser.chunks, sizeof(struct drm_radeon_kernel_chunk)*parser.num_chunks, DRM_MEM_DRIVER); + drm_free(chunk_array, sizeof(uint64_t)*parser.num_chunks, DRM_MEM_DRIVER); + + return r; +} + +/* for non-mm */ +static int r600_nomm_relocate(struct drm_radeon_cs_parser *parser, uint32_t *reloc, uint64_t *offset) +{ + struct drm_device *dev = parser->dev; + drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_kernel_chunk *reloc_chunk = &parser->chunks[parser->reloc_index]; + uint32_t offset_dw = reloc[1]; + + //DRM_INFO("reloc: 0x%08x 0x%08x\n", reloc[0], reloc[1]); + //DRM_INFO("length: %d\n", reloc_chunk->length_dw); + + if (!reloc_chunk->kdata) + return -EINVAL; + + if (offset_dw > reloc_chunk->length_dw) { + DRM_ERROR("Offset larger than chunk 0x%x %d\n", offset_dw, reloc_chunk->length_dw); + return -EINVAL; + } + + /* 40 bit addr */ + *offset = reloc_chunk->kdata[offset_dw + 3]; + *offset <<= 32; + *offset |= reloc_chunk->kdata[offset_dw + 0]; + + //DRM_INFO("offset 0x%lx\n", *offset); + + if (!radeon_check_offset(dev_priv, *offset)) { + DRM_ERROR("bad offset! 0x%lx\n", (unsigned long)*offset); + return -EINVAL; + } + + return 0; +} + +static inline int r600_cs_packet0(struct drm_radeon_cs_parser *parser, uint32_t *offset_dw_p) +{ + uint32_t hdr, num_dw, reg; + int count_dw = 1; + int ret = 0; + uint32_t offset_dw = *offset_dw_p; + int incr = 2; + + hdr = parser->chunks[parser->ib_index].kdata[offset_dw]; + num_dw = ((hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16) + 2; + reg = (hdr & 0xffff) << 2; + + while (count_dw < num_dw) { + switch (reg) { + case AVIVO_D1MODE_VLINE_START_END: + case AVIVO_D2MODE_VLINE_START_END: + break; + default: + ret = -EINVAL; + DRM_ERROR("bad packet 0 reg: 0x%08x\n", reg); + break; + } + if (ret) + break; + count_dw++; + reg += 4; + } + *offset_dw_p += incr; + return ret; +} + +static inline int r600_cs_packet3(struct drm_radeon_cs_parser *parser, uint32_t *offset_dw_p) +{ + struct drm_device *dev = parser->dev; + drm_radeon_private_t *dev_priv = dev->dev_private; + uint32_t hdr, num_dw, start_reg, end_reg, reg; + uint32_t *reloc; + uint64_t offset; + int ret = 0; + uint32_t offset_dw = *offset_dw_p; + int incr = 2; + int i; + struct drm_radeon_kernel_chunk *ib_chunk; + + ib_chunk = &parser->chunks[parser->ib_index]; + + hdr = ib_chunk->kdata[offset_dw]; + num_dw = ((hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16) + 2; + + /* just the ones we use for now, add more later */ + switch (hdr & 0xff00) { + case R600_IT_START_3D_CMDBUF: + //DRM_INFO("R600_IT_START_3D_CMDBUF\n"); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV770) + ret = -EINVAL; + if (num_dw != 2) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad START_3D\n"); + break; + case R600_IT_CONTEXT_CONTROL: + //DRM_INFO("R600_IT_CONTEXT_CONTROL\n"); + if (num_dw != 3) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad CONTEXT_CONTROL\n"); + break; + case R600_IT_INDEX_TYPE: + case R600_IT_NUM_INSTANCES: + //DRM_INFO("R600_IT_INDEX_TYPE/R600_IT_NUM_INSTANCES\n"); + if (num_dw != 2) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad INDEX_TYPE/NUM_INSTANCES\n"); + break; + case R600_IT_DRAW_INDEX: + //DRM_INFO("R600_IT_DRAW_INDEX\n"); + if (num_dw != 5) { + ret = -EINVAL; + DRM_ERROR("bad DRAW_INDEX\n"); + break; + } + reloc = ib_chunk->kdata + offset_dw + num_dw; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad DRAW_INDEX\n"); + break; + } + ib_chunk->kdata[offset_dw + 1] += (offset & 0xffffffff); + ib_chunk->kdata[offset_dw + 2] += (upper_32_bits(offset) & 0xff); + break; + case R600_IT_DRAW_INDEX_AUTO: + //DRM_INFO("R600_IT_DRAW_INDEX_AUTO\n"); + if (num_dw != 3) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad DRAW_INDEX_AUTO\n"); + break; + case R600_IT_DRAW_INDEX_IMMD_BE: + case R600_IT_DRAW_INDEX_IMMD: + //DRM_INFO("R600_IT_DRAW_INDEX_IMMD\n"); + if (num_dw < 4) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad DRAW_INDEX_IMMD\n"); + break; + case R600_IT_WAIT_REG_MEM: + //DRM_INFO("R600_IT_WAIT_REG_MEM\n"); + if (num_dw != 7) + ret = -EINVAL; + /* bit 4 is reg (0) or mem (1) */ + if (ib_chunk->kdata[offset_dw + 1] & 0x10) { + reloc = ib_chunk->kdata + offset_dw + num_dw; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad WAIT_REG_MEM\n"); + break; + } + ib_chunk->kdata[offset_dw + 2] += (offset & 0xffffffff); + ib_chunk->kdata[offset_dw + 3] += (upper_32_bits(offset) & 0xff); + } + if (ret) + DRM_ERROR("bad WAIT_REG_MEM\n"); + break; + case R600_IT_SURFACE_SYNC: + //DRM_INFO("R600_IT_SURFACE_SYNC\n"); + if (num_dw != 5) + ret = -EINVAL; + /* 0xffffffff/0x0 is flush all cache flag */ + else if ((ib_chunk->kdata[offset_dw + 2] == 0xffffffff) && + (ib_chunk->kdata[offset_dw + 3] == 0)) + ret = 0; + else { + reloc = ib_chunk->kdata + offset_dw + num_dw; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad SURFACE_SYNC\n"); + break; + } + ib_chunk->kdata[offset_dw + 3] += ((offset >> 8) & 0xffffffff); + } + break; + case R600_IT_EVENT_WRITE: + //DRM_INFO("R600_IT_EVENT_WRITE\n"); + if ((num_dw != 4) && (num_dw != 2)) + ret = -EINVAL; + if (num_dw > 2) { + reloc = ib_chunk->kdata + offset_dw + num_dw; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad EVENT_WRITE\n"); + break; + } + ib_chunk->kdata[offset_dw + 2] += (offset & 0xffffffff); + ib_chunk->kdata[offset_dw + 3] += (upper_32_bits(offset) & 0xff); + } + if (ret) + DRM_ERROR("bad EVENT_WRITE\n"); + break; + case R600_IT_EVENT_WRITE_EOP: + //DRM_INFO("R600_IT_EVENT_WRITE_EOP\n"); + if (num_dw != 6) { + ret = -EINVAL; + DRM_ERROR("bad EVENT_WRITE_EOP\n"); + break; + } + reloc = ib_chunk->kdata + offset_dw + num_dw; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad EVENT_WRITE_EOP\n"); + break; + } + ib_chunk->kdata[offset_dw + 2] += (offset & 0xffffffff); + ib_chunk->kdata[offset_dw + 3] += (upper_32_bits(offset) & 0xff); + break; + case R600_IT_SET_CONFIG_REG: + //DRM_INFO("R600_IT_SET_CONFIG_REG\n"); + start_reg = (ib_chunk->kdata[offset_dw + 1] << 2) + R600_SET_CONFIG_REG_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_CONFIG_REG_OFFSET) || + (start_reg >= R600_SET_CONFIG_REG_END) || + (end_reg >= R600_SET_CONFIG_REG_END)) + ret = -EINVAL; + else { + for (i = 0; i < (num_dw - 2); i++) { + reg = start_reg + (4 * i); + switch (reg) { + case R600_CP_COHER_BASE: + /* use R600_IT_SURFACE_SYNC */ + ret = -EINVAL; + break; + default: + break; + } + if (ret) + break; + } + } + if (ret) + DRM_ERROR("bad SET_CONFIG_REG\n"); + break; + case R600_IT_SET_CONTEXT_REG: + //DRM_INFO("R600_IT_SET_CONTEXT_REG\n"); + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_CONTEXT_REG_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_CONTEXT_REG_OFFSET) || + (start_reg >= R600_SET_CONTEXT_REG_END) || + (end_reg >= R600_SET_CONTEXT_REG_END)) + ret = -EINVAL; + else { + for (i = 0; i < (num_dw - 2); i++) { + reg = start_reg + (4 * i); + switch (reg) { + case R600_DB_DEPTH_BASE: + case R600_CB_COLOR0_BASE: + case R600_CB_COLOR1_BASE: + case R600_CB_COLOR2_BASE: + case R600_CB_COLOR3_BASE: + case R600_CB_COLOR4_BASE: + case R600_CB_COLOR5_BASE: + case R600_CB_COLOR6_BASE: + case R600_CB_COLOR7_BASE: + case R600_SQ_PGM_START_FS: + case R600_SQ_PGM_START_ES: + case R600_SQ_PGM_START_VS: + case R600_SQ_PGM_START_GS: + case R600_SQ_PGM_START_PS: + //DRM_INFO("reg: 0x%08x\n", reg); + reloc = ib_chunk->kdata + offset_dw + num_dw + (i * 2); + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) { + DRM_ERROR("bad SET_CONTEXT_REG\n"); + break; + } + ib_chunk->kdata[offset_dw + 2 + i] += + ((offset >> 8) & 0xffffffff); + break; + case R600_VGT_DMA_BASE: + case R600_VGT_DMA_BASE_HI: + /* These should be handled by DRAW_INDEX packet 3 */ + case R600_VGT_STRMOUT_BASE_OFFSET_0: + case R600_VGT_STRMOUT_BASE_OFFSET_1: + case R600_VGT_STRMOUT_BASE_OFFSET_2: + case R600_VGT_STRMOUT_BASE_OFFSET_3: + case R600_VGT_STRMOUT_BASE_OFFSET_HI_0: + case R600_VGT_STRMOUT_BASE_OFFSET_HI_1: + case R600_VGT_STRMOUT_BASE_OFFSET_HI_2: + case R600_VGT_STRMOUT_BASE_OFFSET_HI_3: + case R600_VGT_STRMOUT_BUFFER_BASE_0: + case R600_VGT_STRMOUT_BUFFER_BASE_1: + case R600_VGT_STRMOUT_BUFFER_BASE_2: + case R600_VGT_STRMOUT_BUFFER_BASE_3: + case R600_VGT_STRMOUT_BUFFER_OFFSET_0: + case R600_VGT_STRMOUT_BUFFER_OFFSET_1: + case R600_VGT_STRMOUT_BUFFER_OFFSET_2: + case R600_VGT_STRMOUT_BUFFER_OFFSET_3: + /* These should be handled by STRMOUT_BUFFER packet 3 */ + DRM_ERROR("bad context reg: 0x%08x\n", reg); + ret = -EINVAL; + break; + default: + break; + } + if (ret) + break; + } + } + if (ret) + DRM_ERROR("bad SET_CONTEXT_REG\n"); + break; + case R600_IT_SET_RESOURCE: + //DRM_INFO("R600_IT_SET_RESOURCE\n"); + if ((num_dw - 2) % 7) + ret = -EINVAL; + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_RESOURCE_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_RESOURCE_OFFSET) || + (start_reg >= R600_SET_RESOURCE_END) || + (end_reg >= R600_SET_RESOURCE_END)) + ret = -EINVAL; + else { + for (i = 0; i < ((num_dw - 2) / 7); i++) { + switch ((ib_chunk->kdata[offset_dw + (i * 7) + 6 + 2] & 0xc0000000) >> 30) { + case R600_SQ_TEX_VTX_INVALID_TEXTURE: + case R600_SQ_TEX_VTX_INVALID_BUFFER: + default: + ret = -EINVAL; + break; + case R600_SQ_TEX_VTX_VALID_TEXTURE: + /* tex base */ + reloc = ib_chunk->kdata + offset_dw + num_dw + (i * 4); + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) + break; + ib_chunk->kdata[offset_dw + (i * 7) + 2 + 2] += + ((offset >> 8) & 0xffffffff); + /* tex mip base */ + reloc = ib_chunk->kdata + offset_dw + num_dw + (i * 4) + 2; + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) + break; + ib_chunk->kdata[offset_dw + (i * 7) + 3 + 2] += + ((offset >> 8) & 0xffffffff); + break; + case R600_SQ_TEX_VTX_VALID_BUFFER: + /* vtx base */ + reloc = ib_chunk->kdata + offset_dw + num_dw + (i * 2); + ret = dev_priv->cs.relocate(parser, reloc, &offset); + if (ret) + break; + ib_chunk->kdata[offset_dw + (i * 7) + 0 + 2] += (offset & 0xffffffff); + ib_chunk->kdata[offset_dw + (i * 7) + 2 + 2] += (upper_32_bits(offset) & 0xff); + break; + } + if (ret) + break; + } + } + if (ret) + DRM_ERROR("bad SET_RESOURCE\n"); + break; + case R600_IT_SET_ALU_CONST: + //DRM_INFO("R600_IT_SET_ALU_CONST\n"); + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_ALU_CONST_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_ALU_CONST_OFFSET) || + (start_reg >= R600_SET_ALU_CONST_END) || + (end_reg >= R600_SET_ALU_CONST_END)) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SET_ALU_CONST\n"); + break; + case R600_IT_SET_BOOL_CONST: + //DRM_INFO("R600_IT_SET_BOOL_CONST\n"); + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_BOOL_CONST_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_BOOL_CONST_OFFSET) || + (start_reg >= R600_SET_BOOL_CONST_END) || + (end_reg >= R600_SET_BOOL_CONST_END)) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SET_BOOL_CONST\n"); + break; + case R600_IT_SET_LOOP_CONST: + //DRM_INFO("R600_IT_SET_LOOP_CONST\n"); + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_LOOP_CONST_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_LOOP_CONST_OFFSET) || + (start_reg >= R600_SET_LOOP_CONST_END) || + (end_reg >= R600_SET_LOOP_CONST_END)) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SET_LOOP_CONST\n"); + break; + case R600_IT_SET_CTL_CONST: + //DRM_INFO("R600_IT_SET_CTL_CONST\n"); + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_CTL_CONST_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_CTL_CONST_OFFSET) || + (start_reg >= R600_SET_CTL_CONST_END) || + (end_reg >= R600_SET_CTL_CONST_END)) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SET_CTL_CONST\n"); + break; + case R600_IT_SET_SAMPLER: + //DRM_INFO("R600_IT_SET_SAMPLER\n"); + if ((num_dw - 2) % 3) + ret = -EINVAL; + start_reg = ib_chunk->kdata[offset_dw + 1] << 2; + start_reg += R600_SET_SAMPLER_OFFSET; + end_reg = 4 * (num_dw - 2) + start_reg - 4; + if ((start_reg < R600_SET_SAMPLER_OFFSET) || + (start_reg >= R600_SET_SAMPLER_END) || + (end_reg >= R600_SET_SAMPLER_END)) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SET_SAMPLER\n"); + break; + case R600_IT_SURFACE_BASE_UPDATE: + //DRM_INFO("R600_IT_SURFACE_BASE_UPDATE\n"); + if (((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV770) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_R600)) + ret = -EINVAL; + if (num_dw != 2) + ret = -EINVAL; + if (ret) + DRM_ERROR("bad SURFACE_BASE_UPDATE\n"); + break; + case RADEON_CP_NOP: + //DRM_INFO("NOP: %d\n", ib_chunk->kdata[offset_dw + 1]); + break; + default: + DRM_ERROR("invalid packet 3 0x%08x\n", 0xff00); + ret = -EINVAL; + break; + } + + *offset_dw_p += incr; + return ret; +} + +static int r600_cs_parse(struct drm_radeon_cs_parser *parser) +{ + volatile int rb; + struct drm_radeon_kernel_chunk *ib_chunk; + /* scan the packet for various things */ + int count_dw = 0, size_dw; + int ret = 0; + + ib_chunk = &parser->chunks[parser->ib_index]; + size_dw = ib_chunk->length_dw; + + while (count_dw < size_dw && ret == 0) { + int hdr = ib_chunk->kdata[count_dw]; + int num_dw = (hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16; + + switch (hdr & RADEON_CP_PACKET_MASK) { + case RADEON_CP_PACKET0: + ret = r600_cs_packet0(parser, &count_dw); + break; + case RADEON_CP_PACKET1: + ret = -EINVAL; + break; + case RADEON_CP_PACKET2: + DRM_DEBUG("Packet 2\n"); + num_dw += 1; + break; + case RADEON_CP_PACKET3: + ret = r600_cs_packet3(parser, &count_dw); + break; + } + + count_dw += num_dw; + } + + if (ret) + return ret; + + + /* copy the packet into the IB */ + memcpy(parser->ib, ib_chunk->kdata, ib_chunk->length_dw * sizeof(uint32_t)); + + /* read back last byte to flush WC buffers */ + rb = readl(((vm_offset_t)parser->ib + (ib_chunk->length_dw-1) * sizeof(uint32_t))); + + return 0; +} + +static uint32_t radeon_cs_id_get(struct drm_radeon_private *radeon) +{ + /* FIXME: protect with a spinlock */ + /* FIXME: check if wrap affect last reported wrap & sequence */ + radeon->cs.id_scnt = (radeon->cs.id_scnt + 1) & 0x00FFFFFF; + if (!radeon->cs.id_scnt) { + /* increment wrap counter */ + radeon->cs.id_wcnt += 0x01000000; + /* valid sequence counter start at 1 */ + radeon->cs.id_scnt = 1; + } + return (radeon->cs.id_scnt | radeon->cs.id_wcnt); +} + +static void r600_cs_id_emit(struct drm_radeon_cs_parser *parser, uint32_t *id) +{ + drm_radeon_private_t *dev_priv = parser->dev->dev_private; + RING_LOCALS; + + //dev_priv->irq_emitted = radeon_update_breadcrumb(parser->dev); + + *id = radeon_cs_id_get(dev_priv); + + /* SCRATCH 2 */ + BEGIN_RING(3); + R600_CLEAR_AGE(*id); + ADVANCE_RING(); + COMMIT_RING(); +} + +static uint32_t r600_cs_id_last_get(struct drm_device *dev) +{ + //drm_radeon_private_t *dev_priv = dev->dev_private; + + //return GET_R600_SCRATCH(dev_priv, 2); + return 0; +} + +static int r600_ib_get(struct drm_radeon_cs_parser *parser) +{ + struct drm_device *dev = parser->dev; + drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_buf *buf; + + buf = radeon_freelist_get(dev); + if (!buf) { + dev_priv->cs_buf = NULL; + return -EBUSY; + } + buf->file_priv = parser->file_priv; + dev_priv->cs_buf = buf; + parser->ib = (void *)((vm_offset_t)dev->agp_buffer_map->virtual + + buf->offset); + + return 0; +} + +static void r600_ib_free(struct drm_radeon_cs_parser *parser, int error) +{ + struct drm_device *dev = parser->dev; + drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_buf *buf = dev_priv->cs_buf; + + if (buf) { + if (!error) + r600_cp_dispatch_indirect(dev, buf, 0, + parser->chunks[parser->ib_index].length_dw * sizeof(uint32_t)); + radeon_cp_discard_buffer(dev, buf); + COMMIT_RING(); + } +} + +int r600_cs_init(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + dev_priv->cs.ib_get = r600_ib_get; + dev_priv->cs.ib_free = r600_ib_free; + dev_priv->cs.id_emit = r600_cs_id_emit; + dev_priv->cs.id_last_get = r600_cs_id_last_get; + dev_priv->cs.parse = r600_cs_parse; + dev_priv->cs.relocate = r600_nomm_relocate; + return 0; +} diff --git a/sys/dev/drm/radeon_drm.h b/sys/dev/drm/radeon_drm.h index c0d566c..345d21b 100644 --- a/sys/dev/drm/radeon_drm.h +++ b/sys/dev/drm/radeon_drm.h @@ -494,6 +494,8 @@ typedef struct { #define DRM_RADEON_SURF_ALLOC 0x1a #define DRM_RADEON_SURF_FREE 0x1b +#define DRM_RADEON_CS 0x26 + #define DRM_IOCTL_RADEON_CP_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_INIT, drm_radeon_init_t) #define DRM_IOCTL_RADEON_CP_START DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_START) #define DRM_IOCTL_RADEON_CP_STOP DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_STOP, drm_radeon_cp_stop_t) @@ -521,6 +523,7 @@ typedef struct { #define DRM_IOCTL_RADEON_SETPARAM DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SETPARAM, drm_radeon_setparam_t) #define DRM_IOCTL_RADEON_SURF_ALLOC DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SURF_ALLOC, drm_radeon_surface_alloc_t) #define DRM_IOCTL_RADEON_SURF_FREE DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SURF_FREE, drm_radeon_surface_free_t) +#define DRM_IOCTL_RADEON_CS DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_CS, struct drm_radeon_cs) typedef struct drm_radeon_init { enum { @@ -682,6 +685,8 @@ typedef struct drm_radeon_indirect { #define RADEON_PARAM_VBLANK_CRTC 13 /* VBLANK CRTC */ #define RADEON_PARAM_FB_LOCATION 14 /* FB location */ #define RADEON_PARAM_NUM_GB_PIPES 15 /* num GB pipes */ +#define RADEON_PARAM_DEVICE_ID 16 +#define RADEON_PARAM_NUM_Z_PIPES 17 /* num Z pipes */ typedef struct drm_radeon_getparam { int param; @@ -752,4 +757,23 @@ typedef struct drm_radeon_surface_free { #define DRM_RADEON_VBLANK_CRTC1 1 #define DRM_RADEON_VBLANK_CRTC2 2 +/* New interface which obsolete all previous interface. + */ +#define RADEON_CHUNK_ID_RELOCS 0x01 +#define RADEON_CHUNK_ID_IB 0x02 +#define RADEON_CHUNK_ID_OLD 0xff + +struct drm_radeon_cs_chunk { + uint32_t chunk_id; + uint32_t length_dw; + uint64_t chunk_data; +}; + +struct drm_radeon_cs { + uint32_t num_chunks; + uint32_t cs_id; + uint64_t chunks; /* this points to uint64_t * which point to + cs chunks */ +}; + #endif diff --git a/sys/dev/drm/radeon_drv.h b/sys/dev/drm/radeon_drv.h index 511e3eb..62a5094 100644 --- a/sys/dev/drm/radeon_drv.h +++ b/sys/dev/drm/radeon_drv.h @@ -38,7 +38,7 @@ #define DRIVER_NAME "radeon" #define DRIVER_DESC "ATI Radeon" -#define DRIVER_DATE "20080528" +#define DRIVER_DATE "20080613" /* Interface history: * @@ -99,9 +99,11 @@ * 1.27- Add support for IGP GART * 1.28- Add support for VBL on CRTC2 * 1.29- R500 3D cmd buffer support + * 1.30- Add support for occlusion queries + * 1.31- Add support for num Z pipes from GET_PARAM */ #define DRIVER_MAJOR 1 -#define DRIVER_MINOR 29 +#define DRIVER_MINOR 31 #define DRIVER_PATCHLEVEL 0 /* @@ -138,9 +140,9 @@ enum radeon_family { CHIP_R600, CHIP_RV610, CHIP_RV630, + CHIP_RV670, CHIP_RV620, CHIP_RV635, - CHIP_RV670, CHIP_RS780, CHIP_RS880, CHIP_RV770, @@ -233,6 +235,46 @@ struct radeon_virt_surface { #define PCIGART_FILE_PRIV ((void *) -1L) }; +struct drm_radeon_kernel_chunk { + uint32_t chunk_id; + uint32_t length_dw; + uint32_t __user *chunk_data; + uint32_t *kdata; +}; + +struct drm_radeon_cs_parser { + struct drm_device *dev; + struct drm_file *file_priv; + uint32_t num_chunks; + struct drm_radeon_kernel_chunk *chunks; + int ib_index; + int reloc_index; + uint32_t card_offset; + void *ib; +}; + +/* command submission struct */ +struct drm_radeon_cs_priv { + struct spinlock cs_mutex; + uint32_t id_wcnt; + uint32_t id_scnt; + uint32_t id_last_wcnt; + uint32_t id_last_scnt; + + int (*parse)(struct drm_radeon_cs_parser *parser); + void (*id_emit)(struct drm_radeon_cs_parser *parser, uint32_t *id); + uint32_t (*id_last_get)(struct drm_device *dev); + /* this ib handling callback are for hidding memory manager drm + * from memory manager less drm, free have to emit ib discard + * sequence into the ring */ + int (*ib_get)(struct drm_radeon_cs_parser *parser); + uint32_t (*ib_get_ptr)(struct drm_device *dev, void *ib); + void (*ib_free)(struct drm_radeon_cs_parser *parser, int error); + /* do a relocation either MM or non-MM */ + int (*relocate)(struct drm_radeon_cs_parser *parser, + uint32_t *reloc, uint64_t *offset); +}; + #define RADEON_FLUSH_EMITED (1 << 0) #define RADEON_PURGE_EMITED (1 << 1) @@ -325,6 +367,7 @@ typedef struct drm_radeon_private { unsigned long fb_aper_offset; int num_gb_pipes; + int num_z_pipes; int track_flush; drm_local_map_t *mmio; @@ -346,6 +389,12 @@ typedef struct drm_radeon_private { int r700_sc_prim_fifo_size; int r700_sc_hiz_tile_fifo_size; int r700_sc_earlyz_tile_fifo_fize; + /* r6xx/r7xx drm blit vertex buffer */ + struct drm_buf *blit_vb; + + /* CS */ + struct drm_radeon_cs_priv cs; + struct drm_buf *cs_buf; } drm_radeon_private_t; @@ -376,10 +425,10 @@ extern void radeon_set_ring_head(drm_radeon_private_t *dev_priv, u32 val); static __inline__ int radeon_check_offset(drm_radeon_private_t *dev_priv, u64 off) { - u32 fb_start = dev_priv->fb_location; - u32 fb_end = fb_start + dev_priv->fb_size - 1; - u32 gart_start = dev_priv->gart_vm_start; - u32 gart_end = gart_start + dev_priv->gart_size - 1; + u64 fb_start = dev_priv->fb_location; + u64 fb_end = fb_start + dev_priv->fb_size - 1; + u64 gart_start = dev_priv->gart_vm_start; + u64 gart_end = gart_start + dev_priv->gart_size - 1; return ((off >= fb_start && off <= fb_end) || (off >= gart_start && off <= gart_end)); @@ -473,6 +522,33 @@ extern int r600_cp_dispatch_indirect(struct drm_device *dev, struct drm_buf *buf, int start, int end); extern int r600_page_table_init(struct drm_device *dev); extern void r600_page_table_cleanup(struct drm_device *dev, struct drm_ati_pcigart_info *gart_info); +extern void r600_cp_dispatch_swap(struct drm_device * dev); +extern int r600_cp_dispatch_texture(struct drm_device * dev, + struct drm_file *file_priv, + drm_radeon_texture_t * tex, + drm_radeon_tex_image_t * image); + +/* r600_blit.c */ +extern int +r600_prepare_blit_copy(struct drm_device *dev); +extern void +r600_done_blit_copy(struct drm_device *dev); +extern void +r600_blit_copy(struct drm_device *dev, + uint64_t src_gpu_addr, uint64_t dst_gpu_addr, + int size_bytes); +extern void +r600_blit_swap(struct drm_device *dev, + uint64_t src_gpu_addr, uint64_t dst_gpu_addr, + int sx, int sy, int dx, int dy, + int w, int h, int src_pitch, int dst_pitch, int cpp); + +/* radeon_state.c */ +extern void radeon_cp_discard_buffer(struct drm_device * dev, struct drm_buf * buf); + +/* radeon_cs.c */ +extern int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *fpriv); +extern int r600_cs_init(struct drm_device *dev); /* Flags for stats.boxes */ @@ -682,6 +758,7 @@ extern void r600_page_table_cleanup(struct drm_device *dev, struct drm_ati_pciga /* pipe config regs */ #define R400_GB_PIPE_SELECT 0x402c +#define RV530_GB_PIPE_SELECT2 0x4124 #define R500_DYN_SCLK_PWMEM_PIPE 0x000d /* PLL */ #define R300_GB_TILE_CONFIG 0x4018 # define R300_ENABLE_TILING (1 << 0) @@ -1824,26 +1901,38 @@ do { \ */ #define RADEON_WAIT_UNTIL_2D_IDLE() do { \ - OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) \ + OUT_RING( CP_PACKET0( R600_WAIT_UNTIL, 0 ) ); \ + else \ + OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ OUT_RING( (RADEON_WAIT_2D_IDLECLEAN | \ RADEON_WAIT_HOST_IDLECLEAN) ); \ } while (0) #define RADEON_WAIT_UNTIL_3D_IDLE() do { \ - OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) \ + OUT_RING( CP_PACKET0( R600_WAIT_UNTIL, 0 ) ); \ + else \ + OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ OUT_RING( (RADEON_WAIT_3D_IDLECLEAN | \ RADEON_WAIT_HOST_IDLECLEAN) ); \ } while (0) #define RADEON_WAIT_UNTIL_IDLE() do { \ - OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) \ + OUT_RING( CP_PACKET0( R600_WAIT_UNTIL, 0 ) ); \ + else \ + OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ OUT_RING( (RADEON_WAIT_2D_IDLECLEAN | \ RADEON_WAIT_3D_IDLECLEAN | \ RADEON_WAIT_HOST_IDLECLEAN) ); \ } while (0) #define RADEON_WAIT_UNTIL_PAGE_FLIPPED() do { \ - OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) \ + OUT_RING( CP_PACKET0( R600_WAIT_UNTIL, 0 ) ); \ + else \ + OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) ); \ OUT_RING( RADEON_WAIT_CRTC_PFLIP ); \ } while (0) @@ -1958,14 +2047,17 @@ do { \ #define RING_LOCALS int write, _nr, _align_nr; unsigned int mask; u32 *ring; +#define RADEON_RING_ALIGN 16 + #define BEGIN_RING( n ) do { \ if ( RADEON_VERBOSE ) { \ DRM_INFO( "BEGIN_RING( %d )\n", (n)); \ } \ - _align_nr = (n + 0xf) & ~0xf; \ - if (dev_priv->ring.space <= (_align_nr * sizeof(u32))) { \ - COMMIT_RING(); \ - radeon_wait_ring( dev_priv, _align_nr * sizeof(u32)); \ + _align_nr = RADEON_RING_ALIGN - ((dev_priv->ring.tail + n) & (RADEON_RING_ALIGN - 1)); \ + _align_nr += n; \ + if ( dev_priv->ring.space <= (_align_nr) * sizeof(u32) ) { \ + COMMIT_RING(); \ + radeon_wait_ring( dev_priv, (_align_nr) * sizeof(u32) ); \ } \ _nr = n; dev_priv->ring.space -= (n) * sizeof(u32); \ ring = dev_priv->ring.start; \ diff --git a/sys/dev/drm/radeon_irq.c b/sys/dev/drm/radeon_irq.c index 9ff3bfd..82b6dc3 100644 --- a/sys/dev/drm/radeon_irq.c +++ b/sys/dev/drm/radeon_irq.c @@ -191,6 +191,9 @@ irqreturn_t radeon_driver_irq_handler(DRM_IRQ_ARGS) u32 r500_disp_int; u32 tmp; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + return IRQ_NONE; + /* Only consider the bits we're interested in - others could be used * outside the DRM */ @@ -320,6 +323,9 @@ int radeon_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_pr drm_radeon_irq_emit_t *emit = data; int result; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + return -EINVAL; + LOCK_TEST_WITH_RETURN(dev, file_priv); if (!dev_priv) { @@ -360,6 +366,9 @@ void radeon_driver_irq_preinstall(struct drm_device * dev) (drm_radeon_private_t *) dev->dev_private; u32 dummy; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + return; + /* Disable *all* interrupts */ if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS600) RADEON_WRITE(R500_DxMODE_INT_MASK, 0); @@ -377,6 +386,9 @@ int radeon_driver_irq_postinstall(struct drm_device * dev) atomic_set(&dev_priv->swi_emitted, 0); DRM_INIT_WAITQUEUE(&dev_priv->swi_queue); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + return 0; + radeon_irq_set_state(dev, RADEON_SW_INT_ENABLE, 1); return 0; @@ -391,6 +403,9 @@ void radeon_driver_irq_uninstall(struct drm_device * dev) dev_priv->irq_enabled = 0; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + return; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS600) RADEON_WRITE(R500_DxMODE_INT_MASK, 0); /* Disable *all* interrupts */ diff --git a/sys/dev/drm/radeon_state.c b/sys/dev/drm/radeon_state.c index e7c5c5e..31a31d6 100644 --- a/sys/dev/drm/radeon_state.c +++ b/sys/dev/drm/radeon_state.c @@ -1417,7 +1417,7 @@ static void radeon_cp_dispatch_swap(struct drm_device *dev) static void radeon_cp_dispatch_flip(struct drm_device *dev) { drm_radeon_private_t *dev_priv = dev->dev_private; - struct drm_sarea *sarea = (struct drm_sarea *)dev_priv->sarea->handle; + struct drm_sarea *sarea = (struct drm_sarea *)dev_priv->sarea->virtual; int offset = (dev_priv->sarea_priv->pfCurrentPage == 1) ? dev_priv->front_offset : dev_priv->back_offset; RING_LOCALS; @@ -1538,7 +1538,7 @@ static void radeon_cp_dispatch_vertex(struct drm_device * dev, } while (i < nbox); } -static void radeon_cp_discard_buffer(struct drm_device *dev, struct drm_buf *buf) +void radeon_cp_discard_buffer(struct drm_device *dev, struct drm_buf *buf) { drm_radeon_private_t *dev_priv = dev->dev_private; drm_radeon_buf_priv_t *buf_priv = buf->dev_private; @@ -1579,7 +1579,7 @@ static void radeon_cp_dispatch_indirect(struct drm_device * dev, */ if (dwords & 1) { u32 *data = (u32 *) - ((char *)dev->agp_buffer_map->handle + ((char *)dev->agp_buffer_map->virtual + buf->offset + start); data[dwords++] = RADEON_CP_PACKET2; } @@ -1626,7 +1626,7 @@ static void radeon_cp_dispatch_indices(struct drm_device *dev, dwords = (prim->finish - prim->start + 3) / sizeof(u32); - data = (u32 *) ((char *)dev->agp_buffer_map->handle + + data = (u32 *) ((char *)dev->agp_buffer_map->virtual + elt_buf->offset + prim->start); data[0] = CP_PACKET3(RADEON_3D_RNDR_GEN_INDX_PRIM, dwords - 2); @@ -1778,7 +1778,7 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev, /* Dispatch the indirect buffer. */ buffer = - (u32 *) ((char *)dev->agp_buffer_map->handle + buf->offset); + (u32 *) ((char *)dev->agp_buffer_map->virtual + buf->offset); dwords = size / 4; #define RADEON_COPY_MT(_buf, _data, _width) \ @@ -2199,7 +2199,10 @@ static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *f if (sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS) sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS; - radeon_cp_dispatch_swap(dev); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + r600_cp_dispatch_swap(dev); + else + radeon_cp_dispatch_swap(dev); sarea_priv->ctx_owner = 0; COMMIT_RING(); @@ -2396,7 +2399,10 @@ static int radeon_cp_texture(struct drm_device *dev, void *data, struct drm_file RING_SPACE_TEST_WITH_RETURN(dev_priv); VB_AGE_TEST_WITH_RETURN(dev_priv); - ret = radeon_cp_dispatch_texture(dev, file_priv, tex, &image); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + ret = r600_cp_dispatch_texture(dev, file_priv, tex, &image); + else + ret = radeon_cp_dispatch_texture(dev, file_priv, tex, &image); return ret; } @@ -3015,7 +3021,10 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil value = GET_SCRATCH(dev_priv, 2); break; case RADEON_PARAM_IRQ_NR: - value = dev->irq; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R600) + value = 0; + else + value = dev->irq; break; case RADEON_PARAM_GART_BASE: value = dev_priv->gart_vm_start; @@ -3069,6 +3078,9 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil case RADEON_PARAM_NUM_GB_PIPES: value = dev_priv->num_gb_pipes; break; + case RADEON_PARAM_NUM_Z_PIPES: + value = dev_priv->num_z_pipes; + break; default: DRM_DEBUG("Invalid parameter %d\n", param->param); return -EINVAL; @@ -3153,6 +3165,14 @@ void radeon_driver_preclose(struct drm_device *dev, struct drm_file *file_priv) void radeon_driver_lastclose(struct drm_device *dev) { radeon_surfaces_release(PCIGART_FILE_PRIV, dev->dev_private); + if (dev->dev_private) { + drm_radeon_private_t *dev_priv = dev->dev_private; + + if (dev_priv->sarea_priv && + dev_priv->sarea_priv->pfCurrentPage != 0) + radeon_cp_dispatch_flip(dev); + } + radeon_do_release(dev); } @@ -3213,7 +3233,8 @@ struct drm_ioctl_desc radeon_ioctls[] = { DRM_IOCTL_DEF(DRM_RADEON_IRQ_WAIT, radeon_irq_wait, DRM_AUTH), DRM_IOCTL_DEF(DRM_RADEON_SETPARAM, radeon_cp_setparam, DRM_AUTH), DRM_IOCTL_DEF(DRM_RADEON_SURF_ALLOC, radeon_surface_alloc, DRM_AUTH), - DRM_IOCTL_DEF(DRM_RADEON_SURF_FREE, radeon_surface_free, DRM_AUTH) + DRM_IOCTL_DEF(DRM_RADEON_SURF_FREE, radeon_surface_free, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_CS, radeon_cs_ioctl, DRM_AUTH) }; int radeon_max_ioctl = DRM_ARRAY_SIZE(radeon_ioctls); diff --git a/sys/dev/drm/savage_bci.c b/sys/dev/drm/savage_bci.c index 9dcf4f7..7a60fb9 100644 --- a/sys/dev/drm/savage_bci.c +++ b/sys/dev/drm/savage_bci.c @@ -374,7 +374,7 @@ uint32_t *savage_dma_alloc(drm_savage_private_t *dev_priv, unsigned int n) cur, dev_priv->dma_pages[cur].used, n, rest, nr_pages); if (cur + nr_pages < dev_priv->nr_dma_pages) { - dma_ptr = (uint32_t *)dev_priv->cmd_dma->handle + + dma_ptr = (uint32_t *)dev_priv->cmd_dma->virtual + cur * SAVAGE_DMA_PAGE_SIZE + dev_priv->dma_pages[cur].used; if (n < rest) rest = n; @@ -390,7 +390,7 @@ uint32_t *savage_dma_alloc(drm_savage_private_t *dev_priv, unsigned int n) dev_priv->dma_pages[i].used = 0; dev_priv->dma_pages[i].flushed = 0; } - dma_ptr = (uint32_t *)dev_priv->cmd_dma->handle; + dma_ptr = (uint32_t *)dev_priv->cmd_dma->virtual; dev_priv->first_dma_page = cur = 0; } for (i = cur; nr_pages > 0; ++i, --nr_pages) { @@ -441,7 +441,7 @@ static void savage_dma_flush(drm_savage_private_t *dev_priv) /* pad with noops */ if (pad) { - uint32_t *dma_ptr = (uint32_t *)dev_priv->cmd_dma->handle + + uint32_t *dma_ptr = (uint32_t *)dev_priv->cmd_dma->virtual + cur * SAVAGE_DMA_PAGE_SIZE + dev_priv->dma_pages[cur].used; dev_priv->dma_pages[cur].used += pad; while (pad != 0) { @@ -515,7 +515,7 @@ static void savage_fake_dma_flush(drm_savage_private_t *dev_priv) for (i = dev_priv->first_dma_page; i <= dev_priv->current_dma_page && dev_priv->dma_pages[i].used; ++i) { - uint32_t *dma_ptr = (uint32_t *)dev_priv->cmd_dma->handle + + uint32_t *dma_ptr = (uint32_t *)dev_priv->cmd_dma->virtual + i * SAVAGE_DMA_PAGE_SIZE; #if SAVAGE_DMA_DEBUG /* Sanity check: all pages except the last one must be full. */ @@ -782,7 +782,7 @@ static int savage_do_init_bci(struct drm_device *dev, drm_savage_init_t *init) return -EINVAL; } drm_core_ioremap(dev_priv->cmd_dma, dev); - if (!dev_priv->cmd_dma->handle) { + if (!dev_priv->cmd_dma->virtual) { DRM_ERROR("failed to ioremap command " "DMA region!\n"); savage_do_cleanup_bci(dev); @@ -804,9 +804,9 @@ static int savage_do_init_bci(struct drm_device *dev, drm_savage_init_t *init) dev_priv->fake_dma.offset = 0; dev_priv->fake_dma.size = SAVAGE_FAKE_DMA_SIZE; dev_priv->fake_dma.type = _DRM_SHM; - dev_priv->fake_dma.handle = drm_alloc(SAVAGE_FAKE_DMA_SIZE, + dev_priv->fake_dma.virtual = drm_alloc(SAVAGE_FAKE_DMA_SIZE, DRM_MEM_DRIVER); - if (!dev_priv->fake_dma.handle) { + if (!dev_priv->fake_dma.virtual) { DRM_ERROR("could not allocate faked DMA buffer!\n"); savage_do_cleanup_bci(dev); return -ENOMEM; @@ -816,7 +816,7 @@ static int savage_do_init_bci(struct drm_device *dev, drm_savage_init_t *init) } dev_priv->sarea_priv = - (drm_savage_sarea_t *)((uint8_t *)dev_priv->sarea->handle + + (drm_savage_sarea_t *)((uint8_t *)dev_priv->sarea->virtual + init->sarea_priv_offset); /* setup bitmap descriptors */ @@ -855,7 +855,7 @@ static int savage_do_init_bci(struct drm_device *dev, drm_savage_init_t *init) dev_priv->event_counter = 0; dev_priv->event_wrap = 0; dev_priv->bci_ptr = (volatile uint32_t *) - ((uint8_t *)dev_priv->mmio->handle + SAVAGE_BCI_OFFSET); + ((uint8_t *)dev_priv->mmio->virtual + SAVAGE_BCI_OFFSET); if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) { dev_priv->status_used_mask = SAVAGE_FIFO_USED_MASK_S3D; } else { @@ -863,7 +863,7 @@ static int savage_do_init_bci(struct drm_device *dev, drm_savage_init_t *init) } if (dev_priv->status != NULL) { dev_priv->status_ptr = - (volatile uint32_t *)dev_priv->status->handle; + (volatile uint32_t *)dev_priv->status->virtual; dev_priv->wait_fifo = savage_bci_wait_fifo_shadow; dev_priv->wait_evnt = savage_bci_wait_event_shadow; dev_priv->status_ptr[1023] = dev_priv->event_counter; @@ -903,16 +903,16 @@ static int savage_do_cleanup_bci(struct drm_device *dev) drm_savage_private_t *dev_priv = dev->dev_private; if (dev_priv->cmd_dma == &dev_priv->fake_dma) { - if (dev_priv->fake_dma.handle) - drm_free(dev_priv->fake_dma.handle, + if (dev_priv->fake_dma.virtual) + drm_free(dev_priv->fake_dma.virtual, SAVAGE_FAKE_DMA_SIZE, DRM_MEM_DRIVER); - } else if (dev_priv->cmd_dma && dev_priv->cmd_dma->handle && + } else if (dev_priv->cmd_dma && dev_priv->cmd_dma->virtual && dev_priv->cmd_dma->type == _DRM_AGP && dev_priv->dma_type == SAVAGE_DMA_AGP) drm_core_ioremapfree(dev_priv->cmd_dma, dev); if (dev_priv->dma_type == SAVAGE_DMA_AGP && - dev->agp_buffer_map && dev->agp_buffer_map->handle) { + dev->agp_buffer_map && dev->agp_buffer_map->virtual) { drm_core_ioremapfree(dev->agp_buffer_map, dev); /* make sure the next instance (which may be running * in PCI mode) doesn't try to use an old diff --git a/sys/dev/drm/via_3d_reg.h b/sys/dev/drm/via_3d_reg.h new file mode 100644 index 0000000..7a5b3b6 --- /dev/null +++ b/sys/dev/drm/via_3d_reg.h @@ -0,0 +1,1653 @@ +/*- + * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. + * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifndef VIA_3D_REG_H +#define VIA_3D_REG_H +#define HC_REG_BASE 0x0400 + +#define HC_REG_TRANS_SPACE 0x0040 + +#define HC_ParaN_MASK 0xffffffff +#define HC_Para_MASK 0x00ffffff +#define HC_SubA_MASK 0xff000000 +#define HC_SubA_SHIFT 24 +/* Transmission Setting + */ +#define HC_REG_TRANS_SET 0x003c +#define HC_ParaSubType_MASK 0xff000000 +#define HC_ParaType_MASK 0x00ff0000 +#define HC_ParaOS_MASK 0x0000ff00 +#define HC_ParaAdr_MASK 0x000000ff +#define HC_ParaSubType_SHIFT 24 +#define HC_ParaType_SHIFT 16 +#define HC_ParaOS_SHIFT 8 +#define HC_ParaAdr_SHIFT 0 + +#define HC_ParaType_CmdVdata 0x0000 +#define HC_ParaType_NotTex 0x0001 +#define HC_ParaType_Tex 0x0002 +#define HC_ParaType_Palette 0x0003 +#define HC_ParaType_PreCR 0x0010 +#define HC_ParaType_Auto 0x00fe + +/* Transmission Space + */ +#define HC_REG_Hpara0 0x0040 +#define HC_REG_HpataAF 0x02fc + +/* Read + */ +#define HC_REG_HREngSt 0x0000 +#define HC_REG_HRFIFOempty 0x0004 +#define HC_REG_HRFIFOfull 0x0008 +#define HC_REG_HRErr 0x000c +#define HC_REG_FIFOstatus 0x0010 +/* HC_REG_HREngSt 0x0000 + */ +#define HC_HDASZC_MASK 0x00010000 +#define HC_HSGEMI_MASK 0x0000f000 +#define HC_HLGEMISt_MASK 0x00000f00 +#define HC_HCRSt_MASK 0x00000080 +#define HC_HSE0St_MASK 0x00000040 +#define HC_HSE1St_MASK 0x00000020 +#define HC_HPESt_MASK 0x00000010 +#define HC_HXESt_MASK 0x00000008 +#define HC_HBESt_MASK 0x00000004 +#define HC_HE2St_MASK 0x00000002 +#define HC_HE3St_MASK 0x00000001 +/* HC_REG_HRFIFOempty 0x0004 + */ +#define HC_HRZDempty_MASK 0x00000010 +#define HC_HRTXAempty_MASK 0x00000008 +#define HC_HRTXDempty_MASK 0x00000004 +#define HC_HWZDempty_MASK 0x00000002 +#define HC_HWCDempty_MASK 0x00000001 +/* HC_REG_HRFIFOfull 0x0008 + */ +#define HC_HRZDfull_MASK 0x00000010 +#define HC_HRTXAfull_MASK 0x00000008 +#define HC_HRTXDfull_MASK 0x00000004 +#define HC_HWZDfull_MASK 0x00000002 +#define HC_HWCDfull_MASK 0x00000001 +/* HC_REG_HRErr 0x000c + */ +#define HC_HAGPCMErr_MASK 0x80000000 +#define HC_HAGPCMErrC_MASK 0x70000000 +/* HC_REG_FIFOstatus 0x0010 + */ +#define HC_HRFIFOATall_MASK 0x80000000 +#define HC_HRFIFOATbusy_MASK 0x40000000 +#define HC_HRATFGMDo_MASK 0x00000100 +#define HC_HRATFGMDi_MASK 0x00000080 +#define HC_HRATFRZD_MASK 0x00000040 +#define HC_HRATFRTXA_MASK 0x00000020 +#define HC_HRATFRTXD_MASK 0x00000010 +#define HC_HRATFWZD_MASK 0x00000008 +#define HC_HRATFWCD_MASK 0x00000004 +#define HC_HRATTXTAG_MASK 0x00000002 +#define HC_HRATTXCH_MASK 0x00000001 + +/* AGP Command Setting + */ +#define HC_SubA_HAGPBstL 0x0060 +#define HC_SubA_HAGPBendL 0x0061 +#define HC_SubA_HAGPCMNT 0x0062 +#define HC_SubA_HAGPBpL 0x0063 +#define HC_SubA_HAGPBpH 0x0064 +/* HC_SubA_HAGPCMNT 0x0062 + */ +#define HC_HAGPCMNT_MASK 0x00800000 +#define HC_HCmdErrClr_MASK 0x00400000 +#define HC_HAGPBendH_MASK 0x0000ff00 +#define HC_HAGPBstH_MASK 0x000000ff +#define HC_HAGPBendH_SHIFT 8 +#define HC_HAGPBstH_SHIFT 0 +/* HC_SubA_HAGPBpL 0x0063 + */ +#define HC_HAGPBpL_MASK 0x00fffffc +#define HC_HAGPBpID_MASK 0x00000003 +#define HC_HAGPBpID_PAUSE 0x00000000 +#define HC_HAGPBpID_JUMP 0x00000001 +#define HC_HAGPBpID_STOP 0x00000002 +/* HC_SubA_HAGPBpH 0x0064 + */ +#define HC_HAGPBpH_MASK 0x00ffffff + +/* Miscellaneous Settings + */ +#define HC_SubA_HClipTB 0x0070 +#define HC_SubA_HClipLR 0x0071 +#define HC_SubA_HFPClipTL 0x0072 +#define HC_SubA_HFPClipBL 0x0073 +#define HC_SubA_HFPClipLL 0x0074 +#define HC_SubA_HFPClipRL 0x0075 +#define HC_SubA_HFPClipTBH 0x0076 +#define HC_SubA_HFPClipLRH 0x0077 +#define HC_SubA_HLP 0x0078 +#define HC_SubA_HLPRF 0x0079 +#define HC_SubA_HSolidCL 0x007a +#define HC_SubA_HPixGC 0x007b +#define HC_SubA_HSPXYOS 0x007c +#define HC_SubA_HVertexCNT 0x007d + +#define HC_HClipT_MASK 0x00fff000 +#define HC_HClipT_SHIFT 12 +#define HC_HClipB_MASK 0x00000fff +#define HC_HClipB_SHIFT 0 +#define HC_HClipL_MASK 0x00fff000 +#define HC_HClipL_SHIFT 12 +#define HC_HClipR_MASK 0x00000fff +#define HC_HClipR_SHIFT 0 +#define HC_HFPClipBH_MASK 0x0000ff00 +#define HC_HFPClipBH_SHIFT 8 +#define HC_HFPClipTH_MASK 0x000000ff +#define HC_HFPClipTH_SHIFT 0 +#define HC_HFPClipRH_MASK 0x0000ff00 +#define HC_HFPClipRH_SHIFT 8 +#define HC_HFPClipLH_MASK 0x000000ff +#define HC_HFPClipLH_SHIFT 0 +#define HC_HSolidCH_MASK 0x000000ff +#define HC_HPixGC_MASK 0x00800000 +#define HC_HSPXOS_MASK 0x00fff000 +#define HC_HSPXOS_SHIFT 12 +#define HC_HSPYOS_MASK 0x00000fff + +/* Command + * Command A + */ +#define HC_HCmdHeader_MASK 0xfe000000 /*0xffe00000 */ +#define HC_HE3Fire_MASK 0x00100000 +#define HC_HPMType_MASK 0x000f0000 +#define HC_HEFlag_MASK 0x0000e000 +#define HC_HShading_MASK 0x00001c00 +#define HC_HPMValidN_MASK 0x00000200 +#define HC_HPLEND_MASK 0x00000100 +#define HC_HVCycle_MASK 0x000000ff +#define HC_HVCycle_Style_MASK 0x000000c0 +#define HC_HVCycle_ChgA_MASK 0x00000030 +#define HC_HVCycle_ChgB_MASK 0x0000000c +#define HC_HVCycle_ChgC_MASK 0x00000003 +#define HC_HPMType_Point 0x00000000 +#define HC_HPMType_Line 0x00010000 +#define HC_HPMType_Tri 0x00020000 +#define HC_HPMType_TriWF 0x00040000 +#define HC_HEFlag_NoAA 0x00000000 +#define HC_HEFlag_ab 0x00008000 +#define HC_HEFlag_bc 0x00004000 +#define HC_HEFlag_ca 0x00002000 +#define HC_HShading_Solid 0x00000000 +#define HC_HShading_FlatA 0x00000400 +#define HC_HShading_FlatB 0x00000800 +#define HC_HShading_FlatC 0x00000c00 +#define HC_HShading_Gouraud 0x00001000 +#define HC_HVCycle_Full 0x00000000 +#define HC_HVCycle_AFP 0x00000040 +#define HC_HVCycle_One 0x000000c0 +#define HC_HVCycle_NewA 0x00000000 +#define HC_HVCycle_AA 0x00000010 +#define HC_HVCycle_AB 0x00000020 +#define HC_HVCycle_AC 0x00000030 +#define HC_HVCycle_NewB 0x00000000 +#define HC_HVCycle_BA 0x00000004 +#define HC_HVCycle_BB 0x00000008 +#define HC_HVCycle_BC 0x0000000c +#define HC_HVCycle_NewC 0x00000000 +#define HC_HVCycle_CA 0x00000001 +#define HC_HVCycle_CB 0x00000002 +#define HC_HVCycle_CC 0x00000003 + +/* Command B + */ +#define HC_HLPrst_MASK 0x00010000 +#define HC_HLLastP_MASK 0x00008000 +#define HC_HVPMSK_MASK 0x00007f80 +#define HC_HBFace_MASK 0x00000040 +#define HC_H2nd1VT_MASK 0x0000003f +#define HC_HVPMSK_X 0x00004000 +#define HC_HVPMSK_Y 0x00002000 +#define HC_HVPMSK_Z 0x00001000 +#define HC_HVPMSK_W 0x00000800 +#define HC_HVPMSK_Cd 0x00000400 +#define HC_HVPMSK_Cs 0x00000200 +#define HC_HVPMSK_S 0x00000100 +#define HC_HVPMSK_T 0x00000080 + +/* Enable Setting + */ +#define HC_SubA_HEnable 0x0000 +#define HC_HenTXEnvMap_MASK 0x00200000 +#define HC_HenVertexCNT_MASK 0x00100000 +#define HC_HenCPUDAZ_MASK 0x00080000 +#define HC_HenDASZWC_MASK 0x00040000 +#define HC_HenFBCull_MASK 0x00020000 +#define HC_HenCW_MASK 0x00010000 +#define HC_HenAA_MASK 0x00008000 +#define HC_HenST_MASK 0x00004000 +#define HC_HenZT_MASK 0x00002000 +#define HC_HenZW_MASK 0x00001000 +#define HC_HenAT_MASK 0x00000800 +#define HC_HenAW_MASK 0x00000400 +#define HC_HenSP_MASK 0x00000200 +#define HC_HenLP_MASK 0x00000100 +#define HC_HenTXCH_MASK 0x00000080 +#define HC_HenTXMP_MASK 0x00000040 +#define HC_HenTXPP_MASK 0x00000020 +#define HC_HenTXTR_MASK 0x00000010 +#define HC_HenCS_MASK 0x00000008 +#define HC_HenFOG_MASK 0x00000004 +#define HC_HenABL_MASK 0x00000002 +#define HC_HenDT_MASK 0x00000001 + +/* Z Setting + */ +#define HC_SubA_HZWBBasL 0x0010 +#define HC_SubA_HZWBBasH 0x0011 +#define HC_SubA_HZWBType 0x0012 +#define HC_SubA_HZBiasL 0x0013 +#define HC_SubA_HZWBend 0x0014 +#define HC_SubA_HZWTMD 0x0015 +#define HC_SubA_HZWCDL 0x0016 +#define HC_SubA_HZWCTAGnum 0x0017 +#define HC_SubA_HZCYNum 0x0018 +#define HC_SubA_HZWCFire 0x0019 +/* HC_SubA_HZWBType + */ +#define HC_HZWBType_MASK 0x00800000 +#define HC_HZBiasedWB_MASK 0x00400000 +#define HC_HZONEasFF_MASK 0x00200000 +#define HC_HZOONEasFF_MASK 0x00100000 +#define HC_HZWBFM_MASK 0x00030000 +#define HC_HZWBLoc_MASK 0x0000c000 +#define HC_HZWBPit_MASK 0x00003fff +#define HC_HZWBFM_16 0x00000000 +#define HC_HZWBFM_32 0x00020000 +#define HC_HZWBFM_24 0x00030000 +#define HC_HZWBLoc_Local 0x00000000 +#define HC_HZWBLoc_SyS 0x00004000 +/* HC_SubA_HZWBend + */ +#define HC_HZWBend_MASK 0x00ffe000 +#define HC_HZBiasH_MASK 0x000000ff +#define HC_HZWBend_SHIFT 10 +/* HC_SubA_HZWTMD + */ +#define HC_HZWTMD_MASK 0x00070000 +#define HC_HEBEBias_MASK 0x00007f00 +#define HC_HZNF_MASK 0x000000ff +#define HC_HZWTMD_NeverPass 0x00000000 +#define HC_HZWTMD_LT 0x00010000 +#define HC_HZWTMD_EQ 0x00020000 +#define HC_HZWTMD_LE 0x00030000 +#define HC_HZWTMD_GT 0x00040000 +#define HC_HZWTMD_NE 0x00050000 +#define HC_HZWTMD_GE 0x00060000 +#define HC_HZWTMD_AllPass 0x00070000 +#define HC_HEBEBias_SHIFT 8 +/* HC_SubA_HZWCDL 0x0016 + */ +#define HC_HZWCDL_MASK 0x00ffffff +/* HC_SubA_HZWCTAGnum 0x0017 + */ +#define HC_HZWCTAGnum_MASK 0x00ff0000 +#define HC_HZWCTAGnum_SHIFT 16 +#define HC_HZWCDH_MASK 0x000000ff +#define HC_HZWCDH_SHIFT 0 +/* HC_SubA_HZCYNum 0x0018 + */ +#define HC_HZCYNum_MASK 0x00030000 +#define HC_HZCYNum_SHIFT 16 +#define HC_HZWCQWnum_MASK 0x00003fff +#define HC_HZWCQWnum_SHIFT 0 +/* HC_SubA_HZWCFire 0x0019 + */ +#define HC_ZWCFire_MASK 0x00010000 +#define HC_HZWCQWnumLast_MASK 0x00003fff +#define HC_HZWCQWnumLast_SHIFT 0 + +/* Stencil Setting + */ +#define HC_SubA_HSTREF 0x0023 +#define HC_SubA_HSTMD 0x0024 +/* HC_SubA_HSBFM + */ +#define HC_HSBFM_MASK 0x00030000 +#define HC_HSBLoc_MASK 0x0000c000 +#define HC_HSBPit_MASK 0x00003fff +/* HC_SubA_HSTREF + */ +#define HC_HSTREF_MASK 0x00ff0000 +#define HC_HSTOPMSK_MASK 0x0000ff00 +#define HC_HSTBMSK_MASK 0x000000ff +#define HC_HSTREF_SHIFT 16 +#define HC_HSTOPMSK_SHIFT 8 +/* HC_SubA_HSTMD + */ +#define HC_HSTMD_MASK 0x00070000 +#define HC_HSTOPSF_MASK 0x000001c0 +#define HC_HSTOPSPZF_MASK 0x00000038 +#define HC_HSTOPSPZP_MASK 0x00000007 +#define HC_HSTMD_NeverPass 0x00000000 +#define HC_HSTMD_LT 0x00010000 +#define HC_HSTMD_EQ 0x00020000 +#define HC_HSTMD_LE 0x00030000 +#define HC_HSTMD_GT 0x00040000 +#define HC_HSTMD_NE 0x00050000 +#define HC_HSTMD_GE 0x00060000 +#define HC_HSTMD_AllPass 0x00070000 +#define HC_HSTOPSF_KEEP 0x00000000 +#define HC_HSTOPSF_ZERO 0x00000040 +#define HC_HSTOPSF_REPLACE 0x00000080 +#define HC_HSTOPSF_INCRSAT 0x000000c0 +#define HC_HSTOPSF_DECRSAT 0x00000100 +#define HC_HSTOPSF_INVERT 0x00000140 +#define HC_HSTOPSF_INCR 0x00000180 +#define HC_HSTOPSF_DECR 0x000001c0 +#define HC_HSTOPSPZF_KEEP 0x00000000 +#define HC_HSTOPSPZF_ZERO 0x00000008 +#define HC_HSTOPSPZF_REPLACE 0x00000010 +#define HC_HSTOPSPZF_INCRSAT 0x00000018 +#define HC_HSTOPSPZF_DECRSAT 0x00000020 +#define HC_HSTOPSPZF_INVERT 0x00000028 +#define HC_HSTOPSPZF_INCR 0x00000030 +#define HC_HSTOPSPZF_DECR 0x00000038 +#define HC_HSTOPSPZP_KEEP 0x00000000 +#define HC_HSTOPSPZP_ZERO 0x00000001 +#define HC_HSTOPSPZP_REPLACE 0x00000002 +#define HC_HSTOPSPZP_INCRSAT 0x00000003 +#define HC_HSTOPSPZP_DECRSAT 0x00000004 +#define HC_HSTOPSPZP_INVERT 0x00000005 +#define HC_HSTOPSPZP_INCR 0x00000006 +#define HC_HSTOPSPZP_DECR 0x00000007 + +/* Alpha Setting + */ +#define HC_SubA_HABBasL 0x0030 +#define HC_SubA_HABBasH 0x0031 +#define HC_SubA_HABFM 0x0032 +#define HC_SubA_HATMD 0x0033 +#define HC_SubA_HABLCsat 0x0034 +#define HC_SubA_HABLCop 0x0035 +#define HC_SubA_HABLAsat 0x0036 +#define HC_SubA_HABLAop 0x0037 +#define HC_SubA_HABLRCa 0x0038 +#define HC_SubA_HABLRFCa 0x0039 +#define HC_SubA_HABLRCbias 0x003a +#define HC_SubA_HABLRCb 0x003b +#define HC_SubA_HABLRFCb 0x003c +#define HC_SubA_HABLRAa 0x003d +#define HC_SubA_HABLRAb 0x003e +/* HC_SubA_HABFM + */ +#define HC_HABFM_MASK 0x00030000 +#define HC_HABLoc_MASK 0x0000c000 +#define HC_HABPit_MASK 0x000007ff +/* HC_SubA_HATMD + */ +#define HC_HATMD_MASK 0x00000700 +#define HC_HATREF_MASK 0x000000ff +#define HC_HATMD_NeverPass 0x00000000 +#define HC_HATMD_LT 0x00000100 +#define HC_HATMD_EQ 0x00000200 +#define HC_HATMD_LE 0x00000300 +#define HC_HATMD_GT 0x00000400 +#define HC_HATMD_NE 0x00000500 +#define HC_HATMD_GE 0x00000600 +#define HC_HATMD_AllPass 0x00000700 +/* HC_SubA_HABLCsat + */ +#define HC_HABLCsat_MASK 0x00010000 +#define HC_HABLCa_MASK 0x0000fc00 +#define HC_HABLCa_C_MASK 0x0000c000 +#define HC_HABLCa_OPC_MASK 0x00003c00 +#define HC_HABLFCa_MASK 0x000003f0 +#define HC_HABLFCa_C_MASK 0x00000300 +#define HC_HABLFCa_OPC_MASK 0x000000f0 +#define HC_HABLCbias_MASK 0x0000000f +#define HC_HABLCbias_C_MASK 0x00000008 +#define HC_HABLCbias_OPC_MASK 0x00000007 +/*-- Define the input color. + */ +#define HC_XC_Csrc 0x00000000 +#define HC_XC_Cdst 0x00000001 +#define HC_XC_Asrc 0x00000002 +#define HC_XC_Adst 0x00000003 +#define HC_XC_Fog 0x00000004 +#define HC_XC_HABLRC 0x00000005 +#define HC_XC_minSrcDst 0x00000006 +#define HC_XC_maxSrcDst 0x00000007 +#define HC_XC_mimAsrcInvAdst 0x00000008 +#define HC_XC_OPC 0x00000000 +#define HC_XC_InvOPC 0x00000010 +#define HC_XC_OPCp5 0x00000020 +/*-- Define the input Alpha + */ +#define HC_XA_OPA 0x00000000 +#define HC_XA_InvOPA 0x00000010 +#define HC_XA_OPAp5 0x00000020 +#define HC_XA_0 0x00000000 +#define HC_XA_Asrc 0x00000001 +#define HC_XA_Adst 0x00000002 +#define HC_XA_Fog 0x00000003 +#define HC_XA_minAsrcFog 0x00000004 +#define HC_XA_minAsrcAdst 0x00000005 +#define HC_XA_maxAsrcFog 0x00000006 +#define HC_XA_maxAsrcAdst 0x00000007 +#define HC_XA_HABLRA 0x00000008 +#define HC_XA_minAsrcInvAdst 0x00000008 +#define HC_XA_HABLFRA 0x00000009 +/*-- + */ +#define HC_HABLCa_OPC (HC_XC_OPC << 10) +#define HC_HABLCa_InvOPC (HC_XC_InvOPC << 10) +#define HC_HABLCa_OPCp5 (HC_XC_OPCp5 << 10) +#define HC_HABLCa_Csrc (HC_XC_Csrc << 10) +#define HC_HABLCa_Cdst (HC_XC_Cdst << 10) +#define HC_HABLCa_Asrc (HC_XC_Asrc << 10) +#define HC_HABLCa_Adst (HC_XC_Adst << 10) +#define HC_HABLCa_Fog (HC_XC_Fog << 10) +#define HC_HABLCa_HABLRCa (HC_XC_HABLRC << 10) +#define HC_HABLCa_minSrcDst (HC_XC_minSrcDst << 10) +#define HC_HABLCa_maxSrcDst (HC_XC_maxSrcDst << 10) +#define HC_HABLFCa_OPC (HC_XC_OPC << 4) +#define HC_HABLFCa_InvOPC (HC_XC_InvOPC << 4) +#define HC_HABLFCa_OPCp5 (HC_XC_OPCp5 << 4) +#define HC_HABLFCa_Csrc (HC_XC_Csrc << 4) +#define HC_HABLFCa_Cdst (HC_XC_Cdst << 4) +#define HC_HABLFCa_Asrc (HC_XC_Asrc << 4) +#define HC_HABLFCa_Adst (HC_XC_Adst << 4) +#define HC_HABLFCa_Fog (HC_XC_Fog << 4) +#define HC_HABLFCa_HABLRCa (HC_XC_HABLRC << 4) +#define HC_HABLFCa_minSrcDst (HC_XC_minSrcDst << 4) +#define HC_HABLFCa_maxSrcDst (HC_XC_maxSrcDst << 4) +#define HC_HABLFCa_mimAsrcInvAdst (HC_XC_mimAsrcInvAdst << 4) +#define HC_HABLCbias_HABLRCbias 0x00000000 +#define HC_HABLCbias_Asrc 0x00000001 +#define HC_HABLCbias_Adst 0x00000002 +#define HC_HABLCbias_Fog 0x00000003 +#define HC_HABLCbias_Cin 0x00000004 +/* HC_SubA_HABLCop 0x0035 + */ +#define HC_HABLdot_MASK 0x00010000 +#define HC_HABLCop_MASK 0x00004000 +#define HC_HABLCb_MASK 0x00003f00 +#define HC_HABLCb_C_MASK 0x00003000 +#define HC_HABLCb_OPC_MASK 0x00000f00 +#define HC_HABLFCb_MASK 0x000000fc +#define HC_HABLFCb_C_MASK 0x000000c0 +#define HC_HABLFCb_OPC_MASK 0x0000003c +#define HC_HABLCshift_MASK 0x00000003 +#define HC_HABLCb_OPC (HC_XC_OPC << 8) +#define HC_HABLCb_InvOPC (HC_XC_InvOPC << 8) +#define HC_HABLCb_OPCp5 (HC_XC_OPCp5 << 8) +#define HC_HABLCb_Csrc (HC_XC_Csrc << 8) +#define HC_HABLCb_Cdst (HC_XC_Cdst << 8) +#define HC_HABLCb_Asrc (HC_XC_Asrc << 8) +#define HC_HABLCb_Adst (HC_XC_Adst << 8) +#define HC_HABLCb_Fog (HC_XC_Fog << 8) +#define HC_HABLCb_HABLRCa (HC_XC_HABLRC << 8) +#define HC_HABLCb_minSrcDst (HC_XC_minSrcDst << 8) +#define HC_HABLCb_maxSrcDst (HC_XC_maxSrcDst << 8) +#define HC_HABLFCb_OPC (HC_XC_OPC << 2) +#define HC_HABLFCb_InvOPC (HC_XC_InvOPC << 2) +#define HC_HABLFCb_OPCp5 (HC_XC_OPCp5 << 2) +#define HC_HABLFCb_Csrc (HC_XC_Csrc << 2) +#define HC_HABLFCb_Cdst (HC_XC_Cdst << 2) +#define HC_HABLFCb_Asrc (HC_XC_Asrc << 2) +#define HC_HABLFCb_Adst (HC_XC_Adst << 2) +#define HC_HABLFCb_Fog (HC_XC_Fog << 2) +#define HC_HABLFCb_HABLRCb (HC_XC_HABLRC << 2) +#define HC_HABLFCb_minSrcDst (HC_XC_minSrcDst << 2) +#define HC_HABLFCb_maxSrcDst (HC_XC_maxSrcDst << 2) +#define HC_HABLFCb_mimAsrcInvAdst (HC_XC_mimAsrcInvAdst << 2) +/* HC_SubA_HABLAsat 0x0036 + */ +#define HC_HABLAsat_MASK 0x00010000 +#define HC_HABLAa_MASK 0x0000fc00 +#define HC_HABLAa_A_MASK 0x0000c000 +#define HC_HABLAa_OPA_MASK 0x00003c00 +#define HC_HABLFAa_MASK 0x000003f0 +#define HC_HABLFAa_A_MASK 0x00000300 +#define HC_HABLFAa_OPA_MASK 0x000000f0 +#define HC_HABLAbias_MASK 0x0000000f +#define HC_HABLAbias_A_MASK 0x00000008 +#define HC_HABLAbias_OPA_MASK 0x00000007 +#define HC_HABLAa_OPA (HC_XA_OPA << 10) +#define HC_HABLAa_InvOPA (HC_XA_InvOPA << 10) +#define HC_HABLAa_OPAp5 (HC_XA_OPAp5 << 10) +#define HC_HABLAa_0 (HC_XA_0 << 10) +#define HC_HABLAa_Asrc (HC_XA_Asrc << 10) +#define HC_HABLAa_Adst (HC_XA_Adst << 10) +#define HC_HABLAa_Fog (HC_XA_Fog << 10) +#define HC_HABLAa_minAsrcFog (HC_XA_minAsrcFog << 10) +#define HC_HABLAa_minAsrcAdst (HC_XA_minAsrcAdst << 10) +#define HC_HABLAa_maxAsrcFog (HC_XA_maxAsrcFog << 10) +#define HC_HABLAa_maxAsrcAdst (HC_XA_maxAsrcAdst << 10) +#define HC_HABLAa_HABLRA (HC_XA_HABLRA << 10) +#define HC_HABLFAa_OPA (HC_XA_OPA << 4) +#define HC_HABLFAa_InvOPA (HC_XA_InvOPA << 4) +#define HC_HABLFAa_OPAp5 (HC_XA_OPAp5 << 4) +#define HC_HABLFAa_0 (HC_XA_0 << 4) +#define HC_HABLFAa_Asrc (HC_XA_Asrc << 4) +#define HC_HABLFAa_Adst (HC_XA_Adst << 4) +#define HC_HABLFAa_Fog (HC_XA_Fog << 4) +#define HC_HABLFAa_minAsrcFog (HC_XA_minAsrcFog << 4) +#define HC_HABLFAa_minAsrcAdst (HC_XA_minAsrcAdst << 4) +#define HC_HABLFAa_maxAsrcFog (HC_XA_maxAsrcFog << 4) +#define HC_HABLFAa_maxAsrcAdst (HC_XA_maxAsrcAdst << 4) +#define HC_HABLFAa_minAsrcInvAdst (HC_XA_minAsrcInvAdst << 4) +#define HC_HABLFAa_HABLFRA (HC_XA_HABLFRA << 4) +#define HC_HABLAbias_HABLRAbias 0x00000000 +#define HC_HABLAbias_Asrc 0x00000001 +#define HC_HABLAbias_Adst 0x00000002 +#define HC_HABLAbias_Fog 0x00000003 +#define HC_HABLAbias_Aaa 0x00000004 +/* HC_SubA_HABLAop 0x0037 + */ +#define HC_HABLAop_MASK 0x00004000 +#define HC_HABLAb_MASK 0x00003f00 +#define HC_HABLAb_OPA_MASK 0x00000f00 +#define HC_HABLFAb_MASK 0x000000fc +#define HC_HABLFAb_OPA_MASK 0x0000003c +#define HC_HABLAshift_MASK 0x00000003 +#define HC_HABLAb_OPA (HC_XA_OPA << 8) +#define HC_HABLAb_InvOPA (HC_XA_InvOPA << 8) +#define HC_HABLAb_OPAp5 (HC_XA_OPAp5 << 8) +#define HC_HABLAb_0 (HC_XA_0 << 8) +#define HC_HABLAb_Asrc (HC_XA_Asrc << 8) +#define HC_HABLAb_Adst (HC_XA_Adst << 8) +#define HC_HABLAb_Fog (HC_XA_Fog << 8) +#define HC_HABLAb_minAsrcFog (HC_XA_minAsrcFog << 8) +#define HC_HABLAb_minAsrcAdst (HC_XA_minAsrcAdst << 8) +#define HC_HABLAb_maxAsrcFog (HC_XA_maxAsrcFog << 8) +#define HC_HABLAb_maxAsrcAdst (HC_XA_maxAsrcAdst << 8) +#define HC_HABLAb_HABLRA (HC_XA_HABLRA << 8) +#define HC_HABLFAb_OPA (HC_XA_OPA << 2) +#define HC_HABLFAb_InvOPA (HC_XA_InvOPA << 2) +#define HC_HABLFAb_OPAp5 (HC_XA_OPAp5 << 2) +#define HC_HABLFAb_0 (HC_XA_0 << 2) +#define HC_HABLFAb_Asrc (HC_XA_Asrc << 2) +#define HC_HABLFAb_Adst (HC_XA_Adst << 2) +#define HC_HABLFAb_Fog (HC_XA_Fog << 2) +#define HC_HABLFAb_minAsrcFog (HC_XA_minAsrcFog << 2) +#define HC_HABLFAb_minAsrcAdst (HC_XA_minAsrcAdst << 2) +#define HC_HABLFAb_maxAsrcFog (HC_XA_maxAsrcFog << 2) +#define HC_HABLFAb_maxAsrcAdst (HC_XA_maxAsrcAdst << 2) +#define HC_HABLFAb_minAsrcInvAdst (HC_XA_minAsrcInvAdst << 2) +#define HC_HABLFAb_HABLFRA (HC_XA_HABLFRA << 2) +/* HC_SubA_HABLRAa 0x003d + */ +#define HC_HABLRAa_MASK 0x00ff0000 +#define HC_HABLRFAa_MASK 0x0000ff00 +#define HC_HABLRAbias_MASK 0x000000ff +#define HC_HABLRAa_SHIFT 16 +#define HC_HABLRFAa_SHIFT 8 +/* HC_SubA_HABLRAb 0x003e + */ +#define HC_HABLRAb_MASK 0x0000ff00 +#define HC_HABLRFAb_MASK 0x000000ff +#define HC_HABLRAb_SHIFT 8 + +/* Destination Setting + */ +#define HC_SubA_HDBBasL 0x0040 +#define HC_SubA_HDBBasH 0x0041 +#define HC_SubA_HDBFM 0x0042 +#define HC_SubA_HFBBMSKL 0x0043 +#define HC_SubA_HROP 0x0044 +/* HC_SubA_HDBFM 0x0042 + */ +#define HC_HDBFM_MASK 0x001f0000 +#define HC_HDBLoc_MASK 0x0000c000 +#define HC_HDBPit_MASK 0x00003fff +#define HC_HDBFM_RGB555 0x00000000 +#define HC_HDBFM_RGB565 0x00010000 +#define HC_HDBFM_ARGB4444 0x00020000 +#define HC_HDBFM_ARGB1555 0x00030000 +#define HC_HDBFM_BGR555 0x00040000 +#define HC_HDBFM_BGR565 0x00050000 +#define HC_HDBFM_ABGR4444 0x00060000 +#define HC_HDBFM_ABGR1555 0x00070000 +#define HC_HDBFM_ARGB0888 0x00080000 +#define HC_HDBFM_ARGB8888 0x00090000 +#define HC_HDBFM_ABGR0888 0x000a0000 +#define HC_HDBFM_ABGR8888 0x000b0000 +#define HC_HDBLoc_Local 0x00000000 +#define HC_HDBLoc_Sys 0x00004000 +/* HC_SubA_HROP 0x0044 + */ +#define HC_HROP_MASK 0x00000f00 +#define HC_HFBBMSKH_MASK 0x000000ff +#define HC_HROP_BLACK 0x00000000 +#define HC_HROP_DPon 0x00000100 +#define HC_HROP_DPna 0x00000200 +#define HC_HROP_Pn 0x00000300 +#define HC_HROP_PDna 0x00000400 +#define HC_HROP_Dn 0x00000500 +#define HC_HROP_DPx 0x00000600 +#define HC_HROP_DPan 0x00000700 +#define HC_HROP_DPa 0x00000800 +#define HC_HROP_DPxn 0x00000900 +#define HC_HROP_D 0x00000a00 +#define HC_HROP_DPno 0x00000b00 +#define HC_HROP_P 0x00000c00 +#define HC_HROP_PDno 0x00000d00 +#define HC_HROP_DPo 0x00000e00 +#define HC_HROP_WHITE 0x00000f00 + +/* Fog Setting + */ +#define HC_SubA_HFogLF 0x0050 +#define HC_SubA_HFogCL 0x0051 +#define HC_SubA_HFogCH 0x0052 +#define HC_SubA_HFogStL 0x0053 +#define HC_SubA_HFogStH 0x0054 +#define HC_SubA_HFogOOdMF 0x0055 +#define HC_SubA_HFogOOdEF 0x0056 +#define HC_SubA_HFogEndL 0x0057 +#define HC_SubA_HFogDenst 0x0058 +/* HC_SubA_FogLF 0x0050 + */ +#define HC_FogLF_MASK 0x00000010 +#define HC_FogEq_MASK 0x00000008 +#define HC_FogMD_MASK 0x00000007 +#define HC_FogMD_LocalFog 0x00000000 +#define HC_FogMD_LinearFog 0x00000002 +#define HC_FogMD_ExponentialFog 0x00000004 +#define HC_FogMD_Exponential2Fog 0x00000005 +/* #define HC_FogMD_FogTable 0x00000003 */ + +/* HC_SubA_HFogDenst 0x0058 + */ +#define HC_FogDenst_MASK 0x001fff00 +#define HC_FogEndL_MASK 0x000000ff + +/* Texture subtype definitions + */ +#define HC_SubType_Tex0 0x00000000 +#define HC_SubType_Tex1 0x00000001 +#define HC_SubType_TexGeneral 0x000000fe + +/* Attribute of texture n + */ +#define HC_SubA_HTXnL0BasL 0x0000 +#define HC_SubA_HTXnL1BasL 0x0001 +#define HC_SubA_HTXnL2BasL 0x0002 +#define HC_SubA_HTXnL3BasL 0x0003 +#define HC_SubA_HTXnL4BasL 0x0004 +#define HC_SubA_HTXnL5BasL 0x0005 +#define HC_SubA_HTXnL6BasL 0x0006 +#define HC_SubA_HTXnL7BasL 0x0007 +#define HC_SubA_HTXnL8BasL 0x0008 +#define HC_SubA_HTXnL9BasL 0x0009 +#define HC_SubA_HTXnLaBasL 0x000a +#define HC_SubA_HTXnLbBasL 0x000b +#define HC_SubA_HTXnLcBasL 0x000c +#define HC_SubA_HTXnLdBasL 0x000d +#define HC_SubA_HTXnLeBasL 0x000e +#define HC_SubA_HTXnLfBasL 0x000f +#define HC_SubA_HTXnL10BasL 0x0010 +#define HC_SubA_HTXnL11BasL 0x0011 +#define HC_SubA_HTXnL012BasH 0x0020 +#define HC_SubA_HTXnL345BasH 0x0021 +#define HC_SubA_HTXnL678BasH 0x0022 +#define HC_SubA_HTXnL9abBasH 0x0023 +#define HC_SubA_HTXnLcdeBasH 0x0024 +#define HC_SubA_HTXnLf1011BasH 0x0025 +#define HC_SubA_HTXnL0Pit 0x002b +#define HC_SubA_HTXnL1Pit 0x002c +#define HC_SubA_HTXnL2Pit 0x002d +#define HC_SubA_HTXnL3Pit 0x002e +#define HC_SubA_HTXnL4Pit 0x002f +#define HC_SubA_HTXnL5Pit 0x0030 +#define HC_SubA_HTXnL6Pit 0x0031 +#define HC_SubA_HTXnL7Pit 0x0032 +#define HC_SubA_HTXnL8Pit 0x0033 +#define HC_SubA_HTXnL9Pit 0x0034 +#define HC_SubA_HTXnLaPit 0x0035 +#define HC_SubA_HTXnLbPit 0x0036 +#define HC_SubA_HTXnLcPit 0x0037 +#define HC_SubA_HTXnLdPit 0x0038 +#define HC_SubA_HTXnLePit 0x0039 +#define HC_SubA_HTXnLfPit 0x003a +#define HC_SubA_HTXnL10Pit 0x003b +#define HC_SubA_HTXnL11Pit 0x003c +#define HC_SubA_HTXnL0_5WE 0x004b +#define HC_SubA_HTXnL6_bWE 0x004c +#define HC_SubA_HTXnLc_11WE 0x004d +#define HC_SubA_HTXnL0_5HE 0x0051 +#define HC_SubA_HTXnL6_bHE 0x0052 +#define HC_SubA_HTXnLc_11HE 0x0053 +#define HC_SubA_HTXnL0OS 0x0077 +#define HC_SubA_HTXnTB 0x0078 +#define HC_SubA_HTXnMPMD 0x0079 +#define HC_SubA_HTXnCLODu 0x007a +#define HC_SubA_HTXnFM 0x007b +#define HC_SubA_HTXnTRCH 0x007c +#define HC_SubA_HTXnTRCL 0x007d +#define HC_SubA_HTXnTBC 0x007e +#define HC_SubA_HTXnTRAH 0x007f +#define HC_SubA_HTXnTBLCsat 0x0080 +#define HC_SubA_HTXnTBLCop 0x0081 +#define HC_SubA_HTXnTBLMPfog 0x0082 +#define HC_SubA_HTXnTBLAsat 0x0083 +#define HC_SubA_HTXnTBLRCa 0x0085 +#define HC_SubA_HTXnTBLRCb 0x0086 +#define HC_SubA_HTXnTBLRCc 0x0087 +#define HC_SubA_HTXnTBLRCbias 0x0088 +#define HC_SubA_HTXnTBLRAa 0x0089 +#define HC_SubA_HTXnTBLRFog 0x008a +#define HC_SubA_HTXnBumpM00 0x0090 +#define HC_SubA_HTXnBumpM01 0x0091 +#define HC_SubA_HTXnBumpM10 0x0092 +#define HC_SubA_HTXnBumpM11 0x0093 +#define HC_SubA_HTXnLScale 0x0094 +#define HC_SubA_HTXSMD 0x0000 +/* HC_SubA_HTXnL012BasH 0x0020 + */ +#define HC_HTXnL0BasH_MASK 0x000000ff +#define HC_HTXnL1BasH_MASK 0x0000ff00 +#define HC_HTXnL2BasH_MASK 0x00ff0000 +#define HC_HTXnL1BasH_SHIFT 8 +#define HC_HTXnL2BasH_SHIFT 16 +/* HC_SubA_HTXnL345BasH 0x0021 + */ +#define HC_HTXnL3BasH_MASK 0x000000ff +#define HC_HTXnL4BasH_MASK 0x0000ff00 +#define HC_HTXnL5BasH_MASK 0x00ff0000 +#define HC_HTXnL4BasH_SHIFT 8 +#define HC_HTXnL5BasH_SHIFT 16 +/* HC_SubA_HTXnL678BasH 0x0022 + */ +#define HC_HTXnL6BasH_MASK 0x000000ff +#define HC_HTXnL7BasH_MASK 0x0000ff00 +#define HC_HTXnL8BasH_MASK 0x00ff0000 +#define HC_HTXnL7BasH_SHIFT 8 +#define HC_HTXnL8BasH_SHIFT 16 +/* HC_SubA_HTXnL9abBasH 0x0023 + */ +#define HC_HTXnL9BasH_MASK 0x000000ff +#define HC_HTXnLaBasH_MASK 0x0000ff00 +#define HC_HTXnLbBasH_MASK 0x00ff0000 +#define HC_HTXnLaBasH_SHIFT 8 +#define HC_HTXnLbBasH_SHIFT 16 +/* HC_SubA_HTXnLcdeBasH 0x0024 + */ +#define HC_HTXnLcBasH_MASK 0x000000ff +#define HC_HTXnLdBasH_MASK 0x0000ff00 +#define HC_HTXnLeBasH_MASK 0x00ff0000 +#define HC_HTXnLdBasH_SHIFT 8 +#define HC_HTXnLeBasH_SHIFT 16 +/* HC_SubA_HTXnLcdeBasH 0x0025 + */ +#define HC_HTXnLfBasH_MASK 0x000000ff +#define HC_HTXnL10BasH_MASK 0x0000ff00 +#define HC_HTXnL11BasH_MASK 0x00ff0000 +#define HC_HTXnL10BasH_SHIFT 8 +#define HC_HTXnL11BasH_SHIFT 16 +/* HC_SubA_HTXnL0Pit 0x002b + */ +#define HC_HTXnLnPit_MASK 0x00003fff +#define HC_HTXnEnPit_MASK 0x00080000 +#define HC_HTXnLnPitE_MASK 0x00f00000 +#define HC_HTXnLnPitE_SHIFT 20 +/* HC_SubA_HTXnL0_5WE 0x004b + */ +#define HC_HTXnL0WE_MASK 0x0000000f +#define HC_HTXnL1WE_MASK 0x000000f0 +#define HC_HTXnL2WE_MASK 0x00000f00 +#define HC_HTXnL3WE_MASK 0x0000f000 +#define HC_HTXnL4WE_MASK 0x000f0000 +#define HC_HTXnL5WE_MASK 0x00f00000 +#define HC_HTXnL1WE_SHIFT 4 +#define HC_HTXnL2WE_SHIFT 8 +#define HC_HTXnL3WE_SHIFT 12 +#define HC_HTXnL4WE_SHIFT 16 +#define HC_HTXnL5WE_SHIFT 20 +/* HC_SubA_HTXnL6_bWE 0x004c + */ +#define HC_HTXnL6WE_MASK 0x0000000f +#define HC_HTXnL7WE_MASK 0x000000f0 +#define HC_HTXnL8WE_MASK 0x00000f00 +#define HC_HTXnL9WE_MASK 0x0000f000 +#define HC_HTXnLaWE_MASK 0x000f0000 +#define HC_HTXnLbWE_MASK 0x00f00000 +#define HC_HTXnL7WE_SHIFT 4 +#define HC_HTXnL8WE_SHIFT 8 +#define HC_HTXnL9WE_SHIFT 12 +#define HC_HTXnLaWE_SHIFT 16 +#define HC_HTXnLbWE_SHIFT 20 +/* HC_SubA_HTXnLc_11WE 0x004d + */ +#define HC_HTXnLcWE_MASK 0x0000000f +#define HC_HTXnLdWE_MASK 0x000000f0 +#define HC_HTXnLeWE_MASK 0x00000f00 +#define HC_HTXnLfWE_MASK 0x0000f000 +#define HC_HTXnL10WE_MASK 0x000f0000 +#define HC_HTXnL11WE_MASK 0x00f00000 +#define HC_HTXnLdWE_SHIFT 4 +#define HC_HTXnLeWE_SHIFT 8 +#define HC_HTXnLfWE_SHIFT 12 +#define HC_HTXnL10WE_SHIFT 16 +#define HC_HTXnL11WE_SHIFT 20 +/* HC_SubA_HTXnL0_5HE 0x0051 + */ +#define HC_HTXnL0HE_MASK 0x0000000f +#define HC_HTXnL1HE_MASK 0x000000f0 +#define HC_HTXnL2HE_MASK 0x00000f00 +#define HC_HTXnL3HE_MASK 0x0000f000 +#define HC_HTXnL4HE_MASK 0x000f0000 +#define HC_HTXnL5HE_MASK 0x00f00000 +#define HC_HTXnL1HE_SHIFT 4 +#define HC_HTXnL2HE_SHIFT 8 +#define HC_HTXnL3HE_SHIFT 12 +#define HC_HTXnL4HE_SHIFT 16 +#define HC_HTXnL5HE_SHIFT 20 +/* HC_SubA_HTXnL6_bHE 0x0052 + */ +#define HC_HTXnL6HE_MASK 0x0000000f +#define HC_HTXnL7HE_MASK 0x000000f0 +#define HC_HTXnL8HE_MASK 0x00000f00 +#define HC_HTXnL9HE_MASK 0x0000f000 +#define HC_HTXnLaHE_MASK 0x000f0000 +#define HC_HTXnLbHE_MASK 0x00f00000 +#define HC_HTXnL7HE_SHIFT 4 +#define HC_HTXnL8HE_SHIFT 8 +#define HC_HTXnL9HE_SHIFT 12 +#define HC_HTXnLaHE_SHIFT 16 +#define HC_HTXnLbHE_SHIFT 20 +/* HC_SubA_HTXnLc_11HE 0x0053 + */ +#define HC_HTXnLcHE_MASK 0x0000000f +#define HC_HTXnLdHE_MASK 0x000000f0 +#define HC_HTXnLeHE_MASK 0x00000f00 +#define HC_HTXnLfHE_MASK 0x0000f000 +#define HC_HTXnL10HE_MASK 0x000f0000 +#define HC_HTXnL11HE_MASK 0x00f00000 +#define HC_HTXnLdHE_SHIFT 4 +#define HC_HTXnLeHE_SHIFT 8 +#define HC_HTXnLfHE_SHIFT 12 +#define HC_HTXnL10HE_SHIFT 16 +#define HC_HTXnL11HE_SHIFT 20 +/* HC_SubA_HTXnL0OS 0x0077 + */ +#define HC_HTXnL0OS_MASK 0x003ff000 +#define HC_HTXnLVmax_MASK 0x00000fc0 +#define HC_HTXnLVmin_MASK 0x0000003f +#define HC_HTXnL0OS_SHIFT 12 +#define HC_HTXnLVmax_SHIFT 6 +/* HC_SubA_HTXnTB 0x0078 + */ +#define HC_HTXnTB_MASK 0x00f00000 +#define HC_HTXnFLSe_MASK 0x0000e000 +#define HC_HTXnFLSs_MASK 0x00001c00 +#define HC_HTXnFLTe_MASK 0x00000380 +#define HC_HTXnFLTs_MASK 0x00000070 +#define HC_HTXnFLDs_MASK 0x0000000f +#define HC_HTXnTB_NoTB 0x00000000 +#define HC_HTXnTB_TBC_S 0x00100000 +#define HC_HTXnTB_TBC_T 0x00200000 +#define HC_HTXnTB_TB_S 0x00400000 +#define HC_HTXnTB_TB_T 0x00800000 +#define HC_HTXnFLSe_Nearest 0x00000000 +#define HC_HTXnFLSe_Linear 0x00002000 +#define HC_HTXnFLSe_NonLinear 0x00004000 +#define HC_HTXnFLSe_Sharp 0x00008000 +#define HC_HTXnFLSe_Flat_Gaussian_Cubic 0x0000c000 +#define HC_HTXnFLSs_Nearest 0x00000000 +#define HC_HTXnFLSs_Linear 0x00000400 +#define HC_HTXnFLSs_NonLinear 0x00000800 +#define HC_HTXnFLSs_Flat_Gaussian_Cubic 0x00001800 +#define HC_HTXnFLTe_Nearest 0x00000000 +#define HC_HTXnFLTe_Linear 0x00000080 +#define HC_HTXnFLTe_NonLinear 0x00000100 +#define HC_HTXnFLTe_Sharp 0x00000180 +#define HC_HTXnFLTe_Flat_Gaussian_Cubic 0x00000300 +#define HC_HTXnFLTs_Nearest 0x00000000 +#define HC_HTXnFLTs_Linear 0x00000010 +#define HC_HTXnFLTs_NonLinear 0x00000020 +#define HC_HTXnFLTs_Flat_Gaussian_Cubic 0x00000060 +#define HC_HTXnFLDs_Tex0 0x00000000 +#define HC_HTXnFLDs_Nearest 0x00000001 +#define HC_HTXnFLDs_Linear 0x00000002 +#define HC_HTXnFLDs_NonLinear 0x00000003 +#define HC_HTXnFLDs_Dither 0x00000004 +#define HC_HTXnFLDs_ConstLOD 0x00000005 +#define HC_HTXnFLDs_Ani 0x00000006 +#define HC_HTXnFLDs_AniDither 0x00000007 +/* HC_SubA_HTXnMPMD 0x0079 + */ +#define HC_HTXnMPMD_SMASK 0x00070000 +#define HC_HTXnMPMD_TMASK 0x00380000 +#define HC_HTXnLODDTf_MASK 0x00000007 +#define HC_HTXnXY2ST_MASK 0x00000008 +#define HC_HTXnMPMD_Tsingle 0x00000000 +#define HC_HTXnMPMD_Tclamp 0x00080000 +#define HC_HTXnMPMD_Trepeat 0x00100000 +#define HC_HTXnMPMD_Tmirror 0x00180000 +#define HC_HTXnMPMD_Twrap 0x00200000 +#define HC_HTXnMPMD_Ssingle 0x00000000 +#define HC_HTXnMPMD_Sclamp 0x00010000 +#define HC_HTXnMPMD_Srepeat 0x00020000 +#define HC_HTXnMPMD_Smirror 0x00030000 +#define HC_HTXnMPMD_Swrap 0x00040000 +/* HC_SubA_HTXnCLODu 0x007a + */ +#define HC_HTXnCLODu_MASK 0x000ffc00 +#define HC_HTXnCLODd_MASK 0x000003ff +#define HC_HTXnCLODu_SHIFT 10 +/* HC_SubA_HTXnFM 0x007b + */ +#define HC_HTXnFM_MASK 0x00ff0000 +#define HC_HTXnLoc_MASK 0x00000003 +#define HC_HTXnFM_INDEX 0x00000000 +#define HC_HTXnFM_Intensity 0x00080000 +#define HC_HTXnFM_Lum 0x00100000 +#define HC_HTXnFM_Alpha 0x00180000 +#define HC_HTXnFM_DX 0x00280000 +#define HC_HTXnFM_ARGB16 0x00880000 +#define HC_HTXnFM_ARGB32 0x00980000 +#define HC_HTXnFM_ABGR16 0x00a80000 +#define HC_HTXnFM_ABGR32 0x00b80000 +#define HC_HTXnFM_RGBA16 0x00c80000 +#define HC_HTXnFM_RGBA32 0x00d80000 +#define HC_HTXnFM_BGRA16 0x00e80000 +#define HC_HTXnFM_BGRA32 0x00f80000 +#define HC_HTXnFM_BUMPMAP 0x00380000 +#define HC_HTXnFM_Index1 (HC_HTXnFM_INDEX | 0x00000000) +#define HC_HTXnFM_Index2 (HC_HTXnFM_INDEX | 0x00010000) +#define HC_HTXnFM_Index4 (HC_HTXnFM_INDEX | 0x00020000) +#define HC_HTXnFM_Index8 (HC_HTXnFM_INDEX | 0x00030000) +#define HC_HTXnFM_T1 (HC_HTXnFM_Intensity | 0x00000000) +#define HC_HTXnFM_T2 (HC_HTXnFM_Intensity | 0x00010000) +#define HC_HTXnFM_T4 (HC_HTXnFM_Intensity | 0x00020000) +#define HC_HTXnFM_T8 (HC_HTXnFM_Intensity | 0x00030000) +#define HC_HTXnFM_L1 (HC_HTXnFM_Lum | 0x00000000) +#define HC_HTXnFM_L2 (HC_HTXnFM_Lum | 0x00010000) +#define HC_HTXnFM_L4 (HC_HTXnFM_Lum | 0x00020000) +#define HC_HTXnFM_L8 (HC_HTXnFM_Lum | 0x00030000) +#define HC_HTXnFM_AL44 (HC_HTXnFM_Lum | 0x00040000) +#define HC_HTXnFM_AL88 (HC_HTXnFM_Lum | 0x00050000) +#define HC_HTXnFM_A1 (HC_HTXnFM_Alpha | 0x00000000) +#define HC_HTXnFM_A2 (HC_HTXnFM_Alpha | 0x00010000) +#define HC_HTXnFM_A4 (HC_HTXnFM_Alpha | 0x00020000) +#define HC_HTXnFM_A8 (HC_HTXnFM_Alpha | 0x00030000) +#define HC_HTXnFM_DX1 (HC_HTXnFM_DX | 0x00010000) +#define HC_HTXnFM_DX23 (HC_HTXnFM_DX | 0x00020000) +#define HC_HTXnFM_DX45 (HC_HTXnFM_DX | 0x00030000) +#define HC_HTXnFM_RGB555 (HC_HTXnFM_ARGB16 | 0x00000000) +#define HC_HTXnFM_RGB565 (HC_HTXnFM_ARGB16 | 0x00010000) +#define HC_HTXnFM_ARGB1555 (HC_HTXnFM_ARGB16 | 0x00020000) +#define HC_HTXnFM_ARGB4444 (HC_HTXnFM_ARGB16 | 0x00030000) +#define HC_HTXnFM_ARGB0888 (HC_HTXnFM_ARGB32 | 0x00000000) +#define HC_HTXnFM_ARGB8888 (HC_HTXnFM_ARGB32 | 0x00010000) +#define HC_HTXnFM_BGR555 (HC_HTXnFM_ABGR16 | 0x00000000) +#define HC_HTXnFM_BGR565 (HC_HTXnFM_ABGR16 | 0x00010000) +#define HC_HTXnFM_ABGR1555 (HC_HTXnFM_ABGR16 | 0x00020000) +#define HC_HTXnFM_ABGR4444 (HC_HTXnFM_ABGR16 | 0x00030000) +#define HC_HTXnFM_ABGR0888 (HC_HTXnFM_ABGR32 | 0x00000000) +#define HC_HTXnFM_ABGR8888 (HC_HTXnFM_ABGR32 | 0x00010000) +#define HC_HTXnFM_RGBA5550 (HC_HTXnFM_RGBA16 | 0x00000000) +#define HC_HTXnFM_RGBA5551 (HC_HTXnFM_RGBA16 | 0x00020000) +#define HC_HTXnFM_RGBA4444 (HC_HTXnFM_RGBA16 | 0x00030000) +#define HC_HTXnFM_RGBA8880 (HC_HTXnFM_RGBA32 | 0x00000000) +#define HC_HTXnFM_RGBA8888 (HC_HTXnFM_RGBA32 | 0x00010000) +#define HC_HTXnFM_BGRA5550 (HC_HTXnFM_BGRA16 | 0x00000000) +#define HC_HTXnFM_BGRA5551 (HC_HTXnFM_BGRA16 | 0x00020000) +#define HC_HTXnFM_BGRA4444 (HC_HTXnFM_BGRA16 | 0x00030000) +#define HC_HTXnFM_BGRA8880 (HC_HTXnFM_BGRA32 | 0x00000000) +#define HC_HTXnFM_BGRA8888 (HC_HTXnFM_BGRA32 | 0x00010000) +#define HC_HTXnFM_VU88 (HC_HTXnFM_BUMPMAP | 0x00000000) +#define HC_HTXnFM_LVU655 (HC_HTXnFM_BUMPMAP | 0x00010000) +#define HC_HTXnFM_LVU888 (HC_HTXnFM_BUMPMAP | 0x00020000) +#define HC_HTXnLoc_Local 0x00000000 +#define HC_HTXnLoc_Sys 0x00000002 +#define HC_HTXnLoc_AGP 0x00000003 +/* HC_SubA_HTXnTRAH 0x007f + */ +#define HC_HTXnTRAH_MASK 0x00ff0000 +#define HC_HTXnTRAL_MASK 0x0000ff00 +#define HC_HTXnTBA_MASK 0x000000ff +#define HC_HTXnTRAH_SHIFT 16 +#define HC_HTXnTRAL_SHIFT 8 +/* HC_SubA_HTXnTBLCsat 0x0080 + *-- Define the input texture. + */ +#define HC_XTC_TOPC 0x00000000 +#define HC_XTC_InvTOPC 0x00000010 +#define HC_XTC_TOPCp5 0x00000020 +#define HC_XTC_Cbias 0x00000000 +#define HC_XTC_InvCbias 0x00000010 +#define HC_XTC_0 0x00000000 +#define HC_XTC_Dif 0x00000001 +#define HC_XTC_Spec 0x00000002 +#define HC_XTC_Tex 0x00000003 +#define HC_XTC_Cur 0x00000004 +#define HC_XTC_Adif 0x00000005 +#define HC_XTC_Fog 0x00000006 +#define HC_XTC_Atex 0x00000007 +#define HC_XTC_Acur 0x00000008 +#define HC_XTC_HTXnTBLRC 0x00000009 +#define HC_XTC_Ctexnext 0x0000000a +/*-- + */ +#define HC_HTXnTBLCsat_MASK 0x00800000 +#define HC_HTXnTBLCa_MASK 0x000fc000 +#define HC_HTXnTBLCb_MASK 0x00001f80 +#define HC_HTXnTBLCc_MASK 0x0000003f +#define HC_HTXnTBLCa_TOPC (HC_XTC_TOPC << 14) +#define HC_HTXnTBLCa_InvTOPC (HC_XTC_InvTOPC << 14) +#define HC_HTXnTBLCa_TOPCp5 (HC_XTC_TOPCp5 << 14) +#define HC_HTXnTBLCa_0 (HC_XTC_0 << 14) +#define HC_HTXnTBLCa_Dif (HC_XTC_Dif << 14) +#define HC_HTXnTBLCa_Spec (HC_XTC_Spec << 14) +#define HC_HTXnTBLCa_Tex (HC_XTC_Tex << 14) +#define HC_HTXnTBLCa_Cur (HC_XTC_Cur << 14) +#define HC_HTXnTBLCa_Adif (HC_XTC_Adif << 14) +#define HC_HTXnTBLCa_Fog (HC_XTC_Fog << 14) +#define HC_HTXnTBLCa_Atex (HC_XTC_Atex << 14) +#define HC_HTXnTBLCa_Acur (HC_XTC_Acur << 14) +#define HC_HTXnTBLCa_HTXnTBLRC (HC_XTC_HTXnTBLRC << 14) +#define HC_HTXnTBLCa_Ctexnext (HC_XTC_Ctexnext << 14) +#define HC_HTXnTBLCb_TOPC (HC_XTC_TOPC << 7) +#define HC_HTXnTBLCb_InvTOPC (HC_XTC_InvTOPC << 7) +#define HC_HTXnTBLCb_TOPCp5 (HC_XTC_TOPCp5 << 7) +#define HC_HTXnTBLCb_0 (HC_XTC_0 << 7) +#define HC_HTXnTBLCb_Dif (HC_XTC_Dif << 7) +#define HC_HTXnTBLCb_Spec (HC_XTC_Spec << 7) +#define HC_HTXnTBLCb_Tex (HC_XTC_Tex << 7) +#define HC_HTXnTBLCb_Cur (HC_XTC_Cur << 7) +#define HC_HTXnTBLCb_Adif (HC_XTC_Adif << 7) +#define HC_HTXnTBLCb_Fog (HC_XTC_Fog << 7) +#define HC_HTXnTBLCb_Atex (HC_XTC_Atex << 7) +#define HC_HTXnTBLCb_Acur (HC_XTC_Acur << 7) +#define HC_HTXnTBLCb_HTXnTBLRC (HC_XTC_HTXnTBLRC << 7) +#define HC_HTXnTBLCb_Ctexnext (HC_XTC_Ctexnext << 7) +#define HC_HTXnTBLCc_TOPC (HC_XTC_TOPC << 0) +#define HC_HTXnTBLCc_InvTOPC (HC_XTC_InvTOPC << 0) +#define HC_HTXnTBLCc_TOPCp5 (HC_XTC_TOPCp5 << 0) +#define HC_HTXnTBLCc_0 (HC_XTC_0 << 0) +#define HC_HTXnTBLCc_Dif (HC_XTC_Dif << 0) +#define HC_HTXnTBLCc_Spec (HC_XTC_Spec << 0) +#define HC_HTXnTBLCc_Tex (HC_XTC_Tex << 0) +#define HC_HTXnTBLCc_Cur (HC_XTC_Cur << 0) +#define HC_HTXnTBLCc_Adif (HC_XTC_Adif << 0) +#define HC_HTXnTBLCc_Fog (HC_XTC_Fog << 0) +#define HC_HTXnTBLCc_Atex (HC_XTC_Atex << 0) +#define HC_HTXnTBLCc_Acur (HC_XTC_Acur << 0) +#define HC_HTXnTBLCc_HTXnTBLRC (HC_XTC_HTXnTBLRC << 0) +#define HC_HTXnTBLCc_Ctexnext (HC_XTC_Ctexnext << 0) +/* HC_SubA_HTXnTBLCop 0x0081 + */ +#define HC_HTXnTBLdot_MASK 0x00c00000 +#define HC_HTXnTBLCop_MASK 0x00380000 +#define HC_HTXnTBLCbias_MASK 0x0007c000 +#define HC_HTXnTBLCshift_MASK 0x00001800 +#define HC_HTXnTBLAop_MASK 0x00000380 +#define HC_HTXnTBLAbias_MASK 0x00000078 +#define HC_HTXnTBLAshift_MASK 0x00000003 +#define HC_HTXnTBLCop_Add 0x00000000 +#define HC_HTXnTBLCop_Sub 0x00080000 +#define HC_HTXnTBLCop_Min 0x00100000 +#define HC_HTXnTBLCop_Max 0x00180000 +#define HC_HTXnTBLCop_Mask 0x00200000 +#define HC_HTXnTBLCbias_Cbias (HC_XTC_Cbias << 14) +#define HC_HTXnTBLCbias_InvCbias (HC_XTC_InvCbias << 14) +#define HC_HTXnTBLCbias_0 (HC_XTC_0 << 14) +#define HC_HTXnTBLCbias_Dif (HC_XTC_Dif << 14) +#define HC_HTXnTBLCbias_Spec (HC_XTC_Spec << 14) +#define HC_HTXnTBLCbias_Tex (HC_XTC_Tex << 14) +#define HC_HTXnTBLCbias_Cur (HC_XTC_Cur << 14) +#define HC_HTXnTBLCbias_Adif (HC_XTC_Adif << 14) +#define HC_HTXnTBLCbias_Fog (HC_XTC_Fog << 14) +#define HC_HTXnTBLCbias_Atex (HC_XTC_Atex << 14) +#define HC_HTXnTBLCbias_Acur (HC_XTC_Acur << 14) +#define HC_HTXnTBLCbias_HTXnTBLRC (HC_XTC_HTXnTBLRC << 14) +#define HC_HTXnTBLCshift_1 0x00000000 +#define HC_HTXnTBLCshift_2 0x00000800 +#define HC_HTXnTBLCshift_No 0x00001000 +#define HC_HTXnTBLCshift_DotP 0x00001800 +/*=* John Sheng [2003.7.18] texture combine *=*/ +#define HC_HTXnTBLDOT3 0x00080000 +#define HC_HTXnTBLDOT4 0x000C0000 + +#define HC_HTXnTBLAop_Add 0x00000000 +#define HC_HTXnTBLAop_Sub 0x00000080 +#define HC_HTXnTBLAop_Min 0x00000100 +#define HC_HTXnTBLAop_Max 0x00000180 +#define HC_HTXnTBLAop_Mask 0x00000200 +#define HC_HTXnTBLAbias_Inv 0x00000040 +#define HC_HTXnTBLAbias_Adif 0x00000000 +#define HC_HTXnTBLAbias_Fog 0x00000008 +#define HC_HTXnTBLAbias_Acur 0x00000010 +#define HC_HTXnTBLAbias_HTXnTBLRAbias 0x00000018 +#define HC_HTXnTBLAbias_Atex 0x00000020 +#define HC_HTXnTBLAshift_1 0x00000000 +#define HC_HTXnTBLAshift_2 0x00000001 +#define HC_HTXnTBLAshift_No 0x00000002 +/* #define HC_HTXnTBLAshift_DotP 0x00000003 */ +/* HC_SubA_HTXnTBLMPFog 0x0082 + */ +#define HC_HTXnTBLMPfog_MASK 0x00e00000 +#define HC_HTXnTBLMPfog_0 0x00000000 +#define HC_HTXnTBLMPfog_Adif 0x00200000 +#define HC_HTXnTBLMPfog_Fog 0x00400000 +#define HC_HTXnTBLMPfog_Atex 0x00600000 +#define HC_HTXnTBLMPfog_Acur 0x00800000 +#define HC_HTXnTBLMPfog_GHTXnTBLRFog 0x00a00000 +/* HC_SubA_HTXnTBLAsat 0x0083 + *-- Define the texture alpha input. + */ +#define HC_XTA_TOPA 0x00000000 +#define HC_XTA_InvTOPA 0x00000008 +#define HC_XTA_TOPAp5 0x00000010 +#define HC_XTA_Adif 0x00000000 +#define HC_XTA_Fog 0x00000001 +#define HC_XTA_Acur 0x00000002 +#define HC_XTA_HTXnTBLRA 0x00000003 +#define HC_XTA_Atex 0x00000004 +#define HC_XTA_Atexnext 0x00000005 +/*-- + */ +#define HC_HTXnTBLAsat_MASK 0x00800000 +#define HC_HTXnTBLAMB_MASK 0x00700000 +#define HC_HTXnTBLAa_MASK 0x0007c000 +#define HC_HTXnTBLAb_MASK 0x00000f80 +#define HC_HTXnTBLAc_MASK 0x0000001f +#define HC_HTXnTBLAMB_SHIFT 20 +#define HC_HTXnTBLAa_TOPA (HC_XTA_TOPA << 14) +#define HC_HTXnTBLAa_InvTOPA (HC_XTA_InvTOPA << 14) +#define HC_HTXnTBLAa_TOPAp5 (HC_XTA_TOPAp5 << 14) +#define HC_HTXnTBLAa_Adif (HC_XTA_Adif << 14) +#define HC_HTXnTBLAa_Fog (HC_XTA_Fog << 14) +#define HC_HTXnTBLAa_Acur (HC_XTA_Acur << 14) +#define HC_HTXnTBLAa_HTXnTBLRA (HC_XTA_HTXnTBLRA << 14) +#define HC_HTXnTBLAa_Atex (HC_XTA_Atex << 14) +#define HC_HTXnTBLAa_Atexnext (HC_XTA_Atexnext << 14) +#define HC_HTXnTBLAb_TOPA (HC_XTA_TOPA << 7) +#define HC_HTXnTBLAb_InvTOPA (HC_XTA_InvTOPA << 7) +#define HC_HTXnTBLAb_TOPAp5 (HC_XTA_TOPAp5 << 7) +#define HC_HTXnTBLAb_Adif (HC_XTA_Adif << 7) +#define HC_HTXnTBLAb_Fog (HC_XTA_Fog << 7) +#define HC_HTXnTBLAb_Acur (HC_XTA_Acur << 7) +#define HC_HTXnTBLAb_HTXnTBLRA (HC_XTA_HTXnTBLRA << 7) +#define HC_HTXnTBLAb_Atex (HC_XTA_Atex << 7) +#define HC_HTXnTBLAb_Atexnext (HC_XTA_Atexnext << 7) +#define HC_HTXnTBLAc_TOPA (HC_XTA_TOPA << 0) +#define HC_HTXnTBLAc_InvTOPA (HC_XTA_InvTOPA << 0) +#define HC_HTXnTBLAc_TOPAp5 (HC_XTA_TOPAp5 << 0) +#define HC_HTXnTBLAc_Adif (HC_XTA_Adif << 0) +#define HC_HTXnTBLAc_Fog (HC_XTA_Fog << 0) +#define HC_HTXnTBLAc_Acur (HC_XTA_Acur << 0) +#define HC_HTXnTBLAc_HTXnTBLRA (HC_XTA_HTXnTBLRA << 0) +#define HC_HTXnTBLAc_Atex (HC_XTA_Atex << 0) +#define HC_HTXnTBLAc_Atexnext (HC_XTA_Atexnext << 0) +/* HC_SubA_HTXnTBLRAa 0x0089 + */ +#define HC_HTXnTBLRAa_MASK 0x00ff0000 +#define HC_HTXnTBLRAb_MASK 0x0000ff00 +#define HC_HTXnTBLRAc_MASK 0x000000ff +#define HC_HTXnTBLRAa_SHIFT 16 +#define HC_HTXnTBLRAb_SHIFT 8 +#define HC_HTXnTBLRAc_SHIFT 0 +/* HC_SubA_HTXnTBLRFog 0x008a + */ +#define HC_HTXnTBLRFog_MASK 0x0000ff00 +#define HC_HTXnTBLRAbias_MASK 0x000000ff +#define HC_HTXnTBLRFog_SHIFT 8 +#define HC_HTXnTBLRAbias_SHIFT 0 +/* HC_SubA_HTXnLScale 0x0094 + */ +#define HC_HTXnLScale_MASK 0x0007fc00 +#define HC_HTXnLOff_MASK 0x000001ff +#define HC_HTXnLScale_SHIFT 10 +/* HC_SubA_HTXSMD 0x0000 + */ +#define HC_HTXSMD_MASK 0x00000080 +#define HC_HTXTMD_MASK 0x00000040 +#define HC_HTXNum_MASK 0x00000038 +#define HC_HTXTRMD_MASK 0x00000006 +#define HC_HTXCHCLR_MASK 0x00000001 +#define HC_HTXNum_SHIFT 3 + +/* Texture Palette n + */ +#define HC_SubType_TexPalette0 0x00000000 +#define HC_SubType_TexPalette1 0x00000001 +#define HC_SubType_FogTable 0x00000010 +#define HC_SubType_Stipple 0x00000014 +/* HC_SubA_TexPalette0 0x0000 + */ +#define HC_HTPnA_MASK 0xff000000 +#define HC_HTPnR_MASK 0x00ff0000 +#define HC_HTPnG_MASK 0x0000ff00 +#define HC_HTPnB_MASK 0x000000ff +/* HC_SubA_FogTable 0x0010 + */ +#define HC_HFPn3_MASK 0xff000000 +#define HC_HFPn2_MASK 0x00ff0000 +#define HC_HFPn1_MASK 0x0000ff00 +#define HC_HFPn_MASK 0x000000ff +#define HC_HFPn3_SHIFT 24 +#define HC_HFPn2_SHIFT 16 +#define HC_HFPn1_SHIFT 8 + +/* Auto Testing & Security + */ +#define HC_SubA_HenFIFOAT 0x0000 +#define HC_SubA_HFBDrawFirst 0x0004 +#define HC_SubA_HFBBasL 0x0005 +#define HC_SubA_HFBDst 0x0006 +/* HC_SubA_HenFIFOAT 0x0000 + */ +#define HC_HenFIFOAT_MASK 0x00000020 +#define HC_HenGEMILock_MASK 0x00000010 +#define HC_HenFBASwap_MASK 0x00000008 +#define HC_HenOT_MASK 0x00000004 +#define HC_HenCMDQ_MASK 0x00000002 +#define HC_HenTXCTSU_MASK 0x00000001 +/* HC_SubA_HFBDrawFirst 0x0004 + */ +#define HC_HFBDrawFirst_MASK 0x00000800 +#define HC_HFBQueue_MASK 0x00000400 +#define HC_HFBLock_MASK 0x00000200 +#define HC_HEOF_MASK 0x00000100 +#define HC_HFBBasH_MASK 0x000000ff + +/* GEMI Setting + */ +#define HC_SubA_HTArbRCM 0x0008 +#define HC_SubA_HTArbRZ 0x000a +#define HC_SubA_HTArbWZ 0x000b +#define HC_SubA_HTArbRTX 0x000c +#define HC_SubA_HTArbRCW 0x000d +#define HC_SubA_HTArbE2 0x000e +#define HC_SubA_HArbRQCM 0x0010 +#define HC_SubA_HArbWQCM 0x0011 +#define HC_SubA_HGEMITout 0x0020 +#define HC_SubA_HFthRTXD 0x0040 +#define HC_SubA_HFthRTXA 0x0044 +#define HC_SubA_HCMDQstL 0x0050 +#define HC_SubA_HCMDQendL 0x0051 +#define HC_SubA_HCMDQLen 0x0052 +/* HC_SubA_HTArbRCM 0x0008 + */ +#define HC_HTArbRCM_MASK 0x0000ffff +/* HC_SubA_HTArbRZ 0x000a + */ +#define HC_HTArbRZ_MASK 0x0000ffff +/* HC_SubA_HTArbWZ 0x000b + */ +#define HC_HTArbWZ_MASK 0x0000ffff +/* HC_SubA_HTArbRTX 0x000c + */ +#define HC_HTArbRTX_MASK 0x0000ffff +/* HC_SubA_HTArbRCW 0x000d + */ +#define HC_HTArbRCW_MASK 0x0000ffff +/* HC_SubA_HTArbE2 0x000e + */ +#define HC_HTArbE2_MASK 0x0000ffff +/* HC_SubA_HArbRQCM 0x0010 + */ +#define HC_HTArbRQCM_MASK 0x0000ffff +/* HC_SubA_HArbWQCM 0x0011 + */ +#define HC_HArbWQCM_MASK 0x0000ffff +/* HC_SubA_HGEMITout 0x0020 + */ +#define HC_HGEMITout_MASK 0x000f0000 +#define HC_HNPArbZC_MASK 0x0000ffff +#define HC_HGEMITout_SHIFT 16 +/* HC_SubA_HFthRTXD 0x0040 + */ +#define HC_HFthRTXD_MASK 0x00ff0000 +#define HC_HFthRZD_MASK 0x0000ff00 +#define HC_HFthWZD_MASK 0x000000ff +#define HC_HFthRTXD_SHIFT 16 +#define HC_HFthRZD_SHIFT 8 +/* HC_SubA_HFthRTXA 0x0044 + */ +#define HC_HFthRTXA_MASK 0x000000ff + +/****************************************************************************** +** Define the Halcyon Internal register access constants. For simulator only. +******************************************************************************/ +#define HC_SIMA_HAGPBstL 0x0000 +#define HC_SIMA_HAGPBendL 0x0001 +#define HC_SIMA_HAGPCMNT 0x0002 +#define HC_SIMA_HAGPBpL 0x0003 +#define HC_SIMA_HAGPBpH 0x0004 +#define HC_SIMA_HClipTB 0x0005 +#define HC_SIMA_HClipLR 0x0006 +#define HC_SIMA_HFPClipTL 0x0007 +#define HC_SIMA_HFPClipBL 0x0008 +#define HC_SIMA_HFPClipLL 0x0009 +#define HC_SIMA_HFPClipRL 0x000a +#define HC_SIMA_HFPClipTBH 0x000b +#define HC_SIMA_HFPClipLRH 0x000c +#define HC_SIMA_HLP 0x000d +#define HC_SIMA_HLPRF 0x000e +#define HC_SIMA_HSolidCL 0x000f +#define HC_SIMA_HPixGC 0x0010 +#define HC_SIMA_HSPXYOS 0x0011 +#define HC_SIMA_HCmdA 0x0012 +#define HC_SIMA_HCmdB 0x0013 +#define HC_SIMA_HEnable 0x0014 +#define HC_SIMA_HZWBBasL 0x0015 +#define HC_SIMA_HZWBBasH 0x0016 +#define HC_SIMA_HZWBType 0x0017 +#define HC_SIMA_HZBiasL 0x0018 +#define HC_SIMA_HZWBend 0x0019 +#define HC_SIMA_HZWTMD 0x001a +#define HC_SIMA_HZWCDL 0x001b +#define HC_SIMA_HZWCTAGnum 0x001c +#define HC_SIMA_HZCYNum 0x001d +#define HC_SIMA_HZWCFire 0x001e +/* #define HC_SIMA_HSBBasL 0x001d */ +/* #define HC_SIMA_HSBBasH 0x001e */ +/* #define HC_SIMA_HSBFM 0x001f */ +#define HC_SIMA_HSTREF 0x0020 +#define HC_SIMA_HSTMD 0x0021 +#define HC_SIMA_HABBasL 0x0022 +#define HC_SIMA_HABBasH 0x0023 +#define HC_SIMA_HABFM 0x0024 +#define HC_SIMA_HATMD 0x0025 +#define HC_SIMA_HABLCsat 0x0026 +#define HC_SIMA_HABLCop 0x0027 +#define HC_SIMA_HABLAsat 0x0028 +#define HC_SIMA_HABLAop 0x0029 +#define HC_SIMA_HABLRCa 0x002a +#define HC_SIMA_HABLRFCa 0x002b +#define HC_SIMA_HABLRCbias 0x002c +#define HC_SIMA_HABLRCb 0x002d +#define HC_SIMA_HABLRFCb 0x002e +#define HC_SIMA_HABLRAa 0x002f +#define HC_SIMA_HABLRAb 0x0030 +#define HC_SIMA_HDBBasL 0x0031 +#define HC_SIMA_HDBBasH 0x0032 +#define HC_SIMA_HDBFM 0x0033 +#define HC_SIMA_HFBBMSKL 0x0034 +#define HC_SIMA_HROP 0x0035 +#define HC_SIMA_HFogLF 0x0036 +#define HC_SIMA_HFogCL 0x0037 +#define HC_SIMA_HFogCH 0x0038 +#define HC_SIMA_HFogStL 0x0039 +#define HC_SIMA_HFogStH 0x003a +#define HC_SIMA_HFogOOdMF 0x003b +#define HC_SIMA_HFogOOdEF 0x003c +#define HC_SIMA_HFogEndL 0x003d +#define HC_SIMA_HFogDenst 0x003e +/*---- start of texture 0 setting ---- + */ +#define HC_SIMA_HTX0L0BasL 0x0040 +#define HC_SIMA_HTX0L1BasL 0x0041 +#define HC_SIMA_HTX0L2BasL 0x0042 +#define HC_SIMA_HTX0L3BasL 0x0043 +#define HC_SIMA_HTX0L4BasL 0x0044 +#define HC_SIMA_HTX0L5BasL 0x0045 +#define HC_SIMA_HTX0L6BasL 0x0046 +#define HC_SIMA_HTX0L7BasL 0x0047 +#define HC_SIMA_HTX0L8BasL 0x0048 +#define HC_SIMA_HTX0L9BasL 0x0049 +#define HC_SIMA_HTX0LaBasL 0x004a +#define HC_SIMA_HTX0LbBasL 0x004b +#define HC_SIMA_HTX0LcBasL 0x004c +#define HC_SIMA_HTX0LdBasL 0x004d +#define HC_SIMA_HTX0LeBasL 0x004e +#define HC_SIMA_HTX0LfBasL 0x004f +#define HC_SIMA_HTX0L10BasL 0x0050 +#define HC_SIMA_HTX0L11BasL 0x0051 +#define HC_SIMA_HTX0L012BasH 0x0052 +#define HC_SIMA_HTX0L345BasH 0x0053 +#define HC_SIMA_HTX0L678BasH 0x0054 +#define HC_SIMA_HTX0L9abBasH 0x0055 +#define HC_SIMA_HTX0LcdeBasH 0x0056 +#define HC_SIMA_HTX0Lf1011BasH 0x0057 +#define HC_SIMA_HTX0L0Pit 0x0058 +#define HC_SIMA_HTX0L1Pit 0x0059 +#define HC_SIMA_HTX0L2Pit 0x005a +#define HC_SIMA_HTX0L3Pit 0x005b +#define HC_SIMA_HTX0L4Pit 0x005c +#define HC_SIMA_HTX0L5Pit 0x005d +#define HC_SIMA_HTX0L6Pit 0x005e +#define HC_SIMA_HTX0L7Pit 0x005f +#define HC_SIMA_HTX0L8Pit 0x0060 +#define HC_SIMA_HTX0L9Pit 0x0061 +#define HC_SIMA_HTX0LaPit 0x0062 +#define HC_SIMA_HTX0LbPit 0x0063 +#define HC_SIMA_HTX0LcPit 0x0064 +#define HC_SIMA_HTX0LdPit 0x0065 +#define HC_SIMA_HTX0LePit 0x0066 +#define HC_SIMA_HTX0LfPit 0x0067 +#define HC_SIMA_HTX0L10Pit 0x0068 +#define HC_SIMA_HTX0L11Pit 0x0069 +#define HC_SIMA_HTX0L0_5WE 0x006a +#define HC_SIMA_HTX0L6_bWE 0x006b +#define HC_SIMA_HTX0Lc_11WE 0x006c +#define HC_SIMA_HTX0L0_5HE 0x006d +#define HC_SIMA_HTX0L6_bHE 0x006e +#define HC_SIMA_HTX0Lc_11HE 0x006f +#define HC_SIMA_HTX0L0OS 0x0070 +#define HC_SIMA_HTX0TB 0x0071 +#define HC_SIMA_HTX0MPMD 0x0072 +#define HC_SIMA_HTX0CLODu 0x0073 +#define HC_SIMA_HTX0FM 0x0074 +#define HC_SIMA_HTX0TRCH 0x0075 +#define HC_SIMA_HTX0TRCL 0x0076 +#define HC_SIMA_HTX0TBC 0x0077 +#define HC_SIMA_HTX0TRAH 0x0078 +#define HC_SIMA_HTX0TBLCsat 0x0079 +#define HC_SIMA_HTX0TBLCop 0x007a +#define HC_SIMA_HTX0TBLMPfog 0x007b +#define HC_SIMA_HTX0TBLAsat 0x007c +#define HC_SIMA_HTX0TBLRCa 0x007d +#define HC_SIMA_HTX0TBLRCb 0x007e +#define HC_SIMA_HTX0TBLRCc 0x007f +#define HC_SIMA_HTX0TBLRCbias 0x0080 +#define HC_SIMA_HTX0TBLRAa 0x0081 +#define HC_SIMA_HTX0TBLRFog 0x0082 +#define HC_SIMA_HTX0BumpM00 0x0083 +#define HC_SIMA_HTX0BumpM01 0x0084 +#define HC_SIMA_HTX0BumpM10 0x0085 +#define HC_SIMA_HTX0BumpM11 0x0086 +#define HC_SIMA_HTX0LScale 0x0087 +/*---- end of texture 0 setting ---- 0x008f + */ +#define HC_SIMA_TX0TX1_OFF 0x0050 +/*---- start of texture 1 setting ---- + */ +#define HC_SIMA_HTX1L0BasL (HC_SIMA_HTX0L0BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L1BasL (HC_SIMA_HTX0L1BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L2BasL (HC_SIMA_HTX0L2BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L3BasL (HC_SIMA_HTX0L3BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L4BasL (HC_SIMA_HTX0L4BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L5BasL (HC_SIMA_HTX0L5BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L6BasL (HC_SIMA_HTX0L6BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L7BasL (HC_SIMA_HTX0L7BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L8BasL (HC_SIMA_HTX0L8BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L9BasL (HC_SIMA_HTX0L9BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LaBasL (HC_SIMA_HTX0LaBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LbBasL (HC_SIMA_HTX0LbBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LcBasL (HC_SIMA_HTX0LcBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LdBasL (HC_SIMA_HTX0LdBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LeBasL (HC_SIMA_HTX0LeBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LfBasL (HC_SIMA_HTX0LfBasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L10BasL (HC_SIMA_HTX0L10BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L11BasL (HC_SIMA_HTX0L11BasL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L012BasH (HC_SIMA_HTX0L012BasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L345BasH (HC_SIMA_HTX0L345BasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L678BasH (HC_SIMA_HTX0L678BasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L9abBasH (HC_SIMA_HTX0L9abBasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LcdeBasH (HC_SIMA_HTX0LcdeBasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1Lf1011BasH (HC_SIMA_HTX0Lf1011BasH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L0Pit (HC_SIMA_HTX0L0Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L1Pit (HC_SIMA_HTX0L1Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L2Pit (HC_SIMA_HTX0L2Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L3Pit (HC_SIMA_HTX0L3Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L4Pit (HC_SIMA_HTX0L4Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L5Pit (HC_SIMA_HTX0L5Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L6Pit (HC_SIMA_HTX0L6Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L7Pit (HC_SIMA_HTX0L7Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L8Pit (HC_SIMA_HTX0L8Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L9Pit (HC_SIMA_HTX0L9Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LaPit (HC_SIMA_HTX0LaPit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LbPit (HC_SIMA_HTX0LbPit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LcPit (HC_SIMA_HTX0LcPit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LdPit (HC_SIMA_HTX0LdPit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LePit (HC_SIMA_HTX0LePit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LfPit (HC_SIMA_HTX0LfPit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L10Pit (HC_SIMA_HTX0L10Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L11Pit (HC_SIMA_HTX0L11Pit + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L0_5WE (HC_SIMA_HTX0L0_5WE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L6_bWE (HC_SIMA_HTX0L6_bWE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1Lc_11WE (HC_SIMA_HTX0Lc_11WE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L0_5HE (HC_SIMA_HTX0L0_5HE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L6_bHE (HC_SIMA_HTX0L6_bHE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1Lc_11HE (HC_SIMA_HTX0Lc_11HE + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1L0OS (HC_SIMA_HTX0L0OS + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TB (HC_SIMA_HTX0TB + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1MPMD (HC_SIMA_HTX0MPMD + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1CLODu (HC_SIMA_HTX0CLODu + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1FM (HC_SIMA_HTX0FM + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TRCH (HC_SIMA_HTX0TRCH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TRCL (HC_SIMA_HTX0TRCL + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBC (HC_SIMA_HTX0TBC + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TRAH (HC_SIMA_HTX0TRAH + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LTC (HC_SIMA_HTX0LTC + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LTA (HC_SIMA_HTX0LTA + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLCsat (HC_SIMA_HTX0TBLCsat + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLCop (HC_SIMA_HTX0TBLCop + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLMPfog (HC_SIMA_HTX0TBLMPfog + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLAsat (HC_SIMA_HTX0TBLAsat + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRCa (HC_SIMA_HTX0TBLRCa + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRCb (HC_SIMA_HTX0TBLRCb + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRCc (HC_SIMA_HTX0TBLRCc + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRCbias (HC_SIMA_HTX0TBLRCbias + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRAa (HC_SIMA_HTX0TBLRAa + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1TBLRFog (HC_SIMA_HTX0TBLRFog + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1BumpM00 (HC_SIMA_HTX0BumpM00 + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1BumpM01 (HC_SIMA_HTX0BumpM01 + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1BumpM10 (HC_SIMA_HTX0BumpM10 + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1BumpM11 (HC_SIMA_HTX0BumpM11 + HC_SIMA_TX0TX1_OFF) +#define HC_SIMA_HTX1LScale (HC_SIMA_HTX0LScale + HC_SIMA_TX0TX1_OFF) +/*---- end of texture 1 setting ---- 0xaf + */ +#define HC_SIMA_HTXSMD 0x00b0 +#define HC_SIMA_HenFIFOAT 0x00b1 +#define HC_SIMA_HFBDrawFirst 0x00b2 +#define HC_SIMA_HFBBasL 0x00b3 +#define HC_SIMA_HTArbRCM 0x00b4 +#define HC_SIMA_HTArbRZ 0x00b5 +#define HC_SIMA_HTArbWZ 0x00b6 +#define HC_SIMA_HTArbRTX 0x00b7 +#define HC_SIMA_HTArbRCW 0x00b8 +#define HC_SIMA_HTArbE2 0x00b9 +#define HC_SIMA_HGEMITout 0x00ba +#define HC_SIMA_HFthRTXD 0x00bb +#define HC_SIMA_HFthRTXA 0x00bc +/* Define the texture palette 0 + */ +#define HC_SIMA_HTP0 0x0100 +#define HC_SIMA_HTP1 0x0200 +#define HC_SIMA_FOGTABLE 0x0300 +#define HC_SIMA_STIPPLE 0x0400 +#define HC_SIMA_HE3Fire 0x0440 +#define HC_SIMA_TRANS_SET 0x0441 +#define HC_SIMA_HREngSt 0x0442 +#define HC_SIMA_HRFIFOempty 0x0443 +#define HC_SIMA_HRFIFOfull 0x0444 +#define HC_SIMA_HRErr 0x0445 +#define HC_SIMA_FIFOstatus 0x0446 + +/****************************************************************************** +** Define the AGP command header. +******************************************************************************/ +#define HC_ACMD_MASK 0xfe000000 +#define HC_ACMD_SUB_MASK 0x0c000000 +#define HC_ACMD_HCmdA 0xee000000 +#define HC_ACMD_HCmdB 0xec000000 +#define HC_ACMD_HCmdC 0xea000000 +#define HC_ACMD_H1 0xf0000000 +#define HC_ACMD_H2 0xf2000000 +#define HC_ACMD_H3 0xf4000000 +#define HC_ACMD_H4 0xf6000000 + +#define HC_ACMD_H1IO_MASK 0x000001ff +#define HC_ACMD_H2IO1_MASK 0x001ff000 +#define HC_ACMD_H2IO2_MASK 0x000001ff +#define HC_ACMD_H2IO1_SHIFT 12 +#define HC_ACMD_H2IO2_SHIFT 0 +#define HC_ACMD_H3IO_MASK 0x000001ff +#define HC_ACMD_H3COUNT_MASK 0x01fff000 +#define HC_ACMD_H3COUNT_SHIFT 12 +#define HC_ACMD_H4ID_MASK 0x000001ff +#define HC_ACMD_H4COUNT_MASK 0x01fffe00 +#define HC_ACMD_H4COUNT_SHIFT 9 + +/******************************************************************************** +** Define Header +********************************************************************************/ +#define HC_HEADER2 0xF210F110 + +/******************************************************************************** +** Define Dummy Value +********************************************************************************/ +#define HC_DUMMY 0xCCCCCCCC +/******************************************************************************** +** Define for DMA use +********************************************************************************/ +#define HALCYON_HEADER2 0XF210F110 +#define HALCYON_FIRECMD 0XEE100000 +#define HALCYON_FIREMASK 0XFFF00000 +#define HALCYON_CMDB 0XEC000000 +#define HALCYON_CMDBMASK 0XFFFE0000 +#define HALCYON_SUB_ADDR0 0X00000000 +#define HALCYON_HEADER1MASK 0XFFFFFC00 +#define HALCYON_HEADER1 0XF0000000 +#define HC_SubA_HAGPBstL 0x0060 +#define HC_SubA_HAGPBendL 0x0061 +#define HC_SubA_HAGPCMNT 0x0062 +#define HC_SubA_HAGPBpL 0x0063 +#define HC_SubA_HAGPBpH 0x0064 +#define HC_HAGPCMNT_MASK 0x00800000 +#define HC_HCmdErrClr_MASK 0x00400000 +#define HC_HAGPBendH_MASK 0x0000ff00 +#define HC_HAGPBstH_MASK 0x000000ff +#define HC_HAGPBendH_SHIFT 8 +#define HC_HAGPBstH_SHIFT 0 +#define HC_HAGPBpL_MASK 0x00fffffc +#define HC_HAGPBpID_MASK 0x00000003 +#define HC_HAGPBpID_PAUSE 0x00000000 +#define HC_HAGPBpID_JUMP 0x00000001 +#define HC_HAGPBpID_STOP 0x00000002 +#define HC_HAGPBpH_MASK 0x00ffffff + +#define VIA_VIDEO_HEADER5 0xFE040000 +#define VIA_VIDEO_HEADER6 0xFE050000 +#define VIA_VIDEO_HEADER7 0xFE060000 +#define VIA_VIDEOMASK 0xFFFF0000 +#endif diff --git a/sys/dev/drm/via_dma.c b/sys/dev/drm/via_dma.c new file mode 100644 index 0000000..6f435fe --- /dev/null +++ b/sys/dev/drm/via_dma.c @@ -0,0 +1,767 @@ +/* via_dma.c -- DMA support for the VIA Unichrome/Pro + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Copyright 2004 Digeo, Inc., Palo Alto, CA, U.S.A. + * All Rights Reserved. + * + * Copyright 2004 The Unichrome project. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Tungsten Graphics, + * Erdi Chen, + * Thomas Hellstrom. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/drm.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" +#include "dev/drm/via_3d_reg.h" + +#define CMDBUF_ALIGNMENT_SIZE (0x100) +#define CMDBUF_ALIGNMENT_MASK (0x0ff) + +/* defines for VIA 3D registers */ +#define VIA_REG_STATUS 0x400 +#define VIA_REG_TRANSET 0x43C +#define VIA_REG_TRANSPACE 0x440 + +/* VIA_REG_STATUS(0x400): Engine Status */ +#define VIA_CMD_RGTR_BUSY 0x00000080 /* Command Regulator is busy */ +#define VIA_2D_ENG_BUSY 0x00000001 /* 2D Engine is busy */ +#define VIA_3D_ENG_BUSY 0x00000002 /* 3D Engine is busy */ +#define VIA_VR_QUEUE_BUSY 0x00020000 /* Virtual Queue is busy */ + +#define SetReg2DAGP(nReg, nData) { \ + *((uint32_t *)(vb)) = ((nReg) >> 2) | HALCYON_HEADER1; \ + *((uint32_t *)(vb) + 1) = (nData); \ + vb = ((uint32_t *)vb) + 2; \ + dev_priv->dma_low +=8; \ +} + +#define via_flush_write_combine() DRM_MEMORYBARRIER() + +#define VIA_OUT_RING_QW(w1,w2) \ + *vb++ = (w1); \ + *vb++ = (w2); \ + dev_priv->dma_low += 8; + +static void via_cmdbuf_start(drm_via_private_t * dev_priv); +static void via_cmdbuf_pause(drm_via_private_t * dev_priv); +static void via_cmdbuf_reset(drm_via_private_t * dev_priv); +static void via_cmdbuf_rewind(drm_via_private_t * dev_priv); +static int via_wait_idle(drm_via_private_t * dev_priv); +static void via_pad_cache(drm_via_private_t * dev_priv, int qwords); + +/* + * Free space in command buffer. + */ + +static uint32_t via_cmdbuf_space(drm_via_private_t * dev_priv) +{ + uint32_t agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + uint32_t hw_addr = *(dev_priv->hw_addr_ptr) - agp_base; + + return ((hw_addr <= dev_priv->dma_low) ? + (dev_priv->dma_high + hw_addr - dev_priv->dma_low) : + (hw_addr - dev_priv->dma_low)); +} + +/* + * How much does the command regulator lag behind? + */ + +static uint32_t via_cmdbuf_lag(drm_via_private_t * dev_priv) +{ + uint32_t agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + uint32_t hw_addr = *(dev_priv->hw_addr_ptr) - agp_base; + + return ((hw_addr <= dev_priv->dma_low) ? + (dev_priv->dma_low - hw_addr) : + (dev_priv->dma_wrap + dev_priv->dma_low - hw_addr)); +} + +/* + * Check that the given size fits in the buffer, otherwise wait. + */ + +static inline int +via_cmdbuf_wait(drm_via_private_t * dev_priv, unsigned int size) +{ + uint32_t agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + uint32_t cur_addr, hw_addr, next_addr; + volatile uint32_t *hw_addr_ptr; + uint32_t count; + hw_addr_ptr = dev_priv->hw_addr_ptr; + cur_addr = dev_priv->dma_low; + next_addr = cur_addr + size + 512 * 1024; + count = 1000000; + do { + hw_addr = *hw_addr_ptr - agp_base; + if (count-- == 0) { + DRM_ERROR + ("via_cmdbuf_wait timed out hw %x cur_addr %x next_addr %x\n", + hw_addr, cur_addr, next_addr); + return -1; + } + if ((cur_addr < hw_addr) && (next_addr >= hw_addr)) + DRM_UDELAY(1000); + } while ((cur_addr < hw_addr) && (next_addr >= hw_addr)); + return 0; +} + +/* + * Checks whether buffer head has reach the end. Rewind the ring buffer + * when necessary. + * + * Returns virtual pointer to ring buffer. + */ + +static inline uint32_t *via_check_dma(drm_via_private_t * dev_priv, + unsigned int size) +{ + if ((dev_priv->dma_low + size + 4 * CMDBUF_ALIGNMENT_SIZE) > + dev_priv->dma_high) { + via_cmdbuf_rewind(dev_priv); + } + if (via_cmdbuf_wait(dev_priv, size) != 0) { + return NULL; + } + + return (uint32_t *) (dev_priv->dma_ptr + dev_priv->dma_low); +} + +int via_dma_cleanup(struct drm_device * dev) +{ + drm_via_blitq_t *blitq; + int i; + + if (dev->dev_private) { + drm_via_private_t *dev_priv = + (drm_via_private_t *) dev->dev_private; + + if (dev_priv->ring.virtual_start) { + via_cmdbuf_reset(dev_priv); + + drm_core_ioremapfree(&dev_priv->ring.map, dev); + dev_priv->ring.virtual_start = NULL; + } + + for (i=0; i< VIA_NUM_BLIT_ENGINES; ++i) { + blitq = dev_priv->blit_queues + i; + mtx_destroy(&blitq->blit_lock); + } + } + + return 0; +} + +static int via_initialize(struct drm_device * dev, + drm_via_private_t * dev_priv, + drm_via_dma_init_t * init) +{ + if (!dev_priv || !dev_priv->mmio) { + DRM_ERROR("via_dma_init called before via_map_init\n"); + return -EFAULT; + } + + if (dev_priv->ring.virtual_start != NULL) { + DRM_ERROR("called again without calling cleanup\n"); + return -EFAULT; + } + + if (!dev->agp || !dev->agp->base) { + DRM_ERROR("called with no agp memory available\n"); + return -EFAULT; + } + + if (dev_priv->chipset == VIA_DX9_0) { + DRM_ERROR("AGP DMA is not supported on this chip\n"); + return -EINVAL; + } + + dev_priv->ring.map.offset = dev->agp->base + init->offset; + dev_priv->ring.map.size = init->size; + dev_priv->ring.map.type = 0; + dev_priv->ring.map.flags = 0; + dev_priv->ring.map.mtrr = 0; + + drm_core_ioremap_wc(&dev_priv->ring.map, dev); + + if (dev_priv->ring.map.virtual == NULL) { + via_dma_cleanup(dev); + DRM_ERROR("can not ioremap virtual address for" + " ring buffer\n"); + return -ENOMEM; + } + + dev_priv->ring.virtual_start = dev_priv->ring.map.virtual; + + dev_priv->dma_ptr = dev_priv->ring.virtual_start; + dev_priv->dma_low = 0; + dev_priv->dma_high = init->size; + dev_priv->dma_wrap = init->size; + dev_priv->dma_offset = init->offset; + dev_priv->last_pause_ptr = NULL; + dev_priv->hw_addr_ptr = + (volatile uint32_t *)((char *)dev_priv->mmio->virtual + + init->reg_pause_addr); + + via_cmdbuf_start(dev_priv); + + return 0; +} + +static int via_dma_init(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + drm_via_dma_init_t *init = data; + int retcode = 0; + + switch (init->func) { + case VIA_INIT_DMA: + if (!DRM_SUSER(DRM_CURPROC)) + retcode = -EPERM; + else + retcode = via_initialize(dev, dev_priv, init); + break; + case VIA_CLEANUP_DMA: + if (!DRM_SUSER(DRM_CURPROC)) + retcode = -EPERM; + else + retcode = via_dma_cleanup(dev); + break; + case VIA_DMA_INITIALIZED: + retcode = (dev_priv->ring.virtual_start != NULL) ? + 0 : -EFAULT; + break; + default: + retcode = -EINVAL; + break; + } + + return retcode; +} + +static int via_dispatch_cmdbuffer(struct drm_device * dev, drm_via_cmdbuffer_t * cmd) +{ + drm_via_private_t *dev_priv; + uint32_t *vb; + int ret; + + dev_priv = (drm_via_private_t *) dev->dev_private; + + if (dev_priv->ring.virtual_start == NULL) { + DRM_ERROR("called without initializing AGP ring buffer.\n"); + return -EFAULT; + } + + if (cmd->size > VIA_PCI_BUF_SIZE) { + return -ENOMEM; + } + + if (DRM_COPY_FROM_USER(dev_priv->pci_buf, cmd->buf, cmd->size)) + return -EFAULT; + + /* + * Running this function on AGP memory is dead slow. Therefore + * we run it on a temporary cacheable system memory buffer and + * copy it to AGP memory when ready. + */ + + if ((ret = + via_verify_command_stream((uint32_t *) dev_priv->pci_buf, + cmd->size, dev, 1))) { + return ret; + } + + vb = via_check_dma(dev_priv, (cmd->size < 0x100) ? 0x102 : cmd->size); + if (vb == NULL) { + return -EAGAIN; + } + + memcpy(vb, dev_priv->pci_buf, cmd->size); + + dev_priv->dma_low += cmd->size; + + /* + * Small submissions somehow stalls the CPU. (AGP cache effects?) + * pad to greater size. + */ + + if (cmd->size < 0x100) + via_pad_cache(dev_priv, (0x100 - cmd->size) >> 3); + via_cmdbuf_pause(dev_priv); + + return 0; +} + +int via_driver_dma_quiescent(struct drm_device * dev) +{ + drm_via_private_t *dev_priv = dev->dev_private; + + if (!via_wait_idle(dev_priv)) { + return -EBUSY; + } + return 0; +} + +static int via_flush_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + + LOCK_TEST_WITH_RETURN(dev, file_priv); + + return via_driver_dma_quiescent(dev); +} + +static int via_cmdbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_cmdbuffer_t *cmdbuf = data; + int ret; + + LOCK_TEST_WITH_RETURN(dev, file_priv); + + DRM_DEBUG("buf %p size %lu\n", cmdbuf->buf, cmdbuf->size); + + ret = via_dispatch_cmdbuffer(dev, cmdbuf); + if (ret) { + return ret; + } + + return 0; +} + +static int via_dispatch_pci_cmdbuffer(struct drm_device * dev, + drm_via_cmdbuffer_t * cmd) +{ + drm_via_private_t *dev_priv = dev->dev_private; + int ret; + + if (cmd->size > VIA_PCI_BUF_SIZE) { + return -ENOMEM; + } + if (DRM_COPY_FROM_USER(dev_priv->pci_buf, cmd->buf, cmd->size)) + return -EFAULT; + + if ((ret = + via_verify_command_stream((uint32_t *) dev_priv->pci_buf, + cmd->size, dev, 0))) { + return ret; + } + + ret = + via_parse_command_stream(dev, (const uint32_t *)dev_priv->pci_buf, + cmd->size); + return ret; +} + +static int via_pci_cmdbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_cmdbuffer_t *cmdbuf = data; + int ret; + + LOCK_TEST_WITH_RETURN(dev, file_priv); + + DRM_DEBUG("buf %p size %lu\n", cmdbuf->buf, cmdbuf->size); + + ret = via_dispatch_pci_cmdbuffer(dev, cmdbuf); + if (ret) { + return ret; + } + + return 0; +} + +static inline uint32_t *via_align_buffer(drm_via_private_t * dev_priv, + uint32_t * vb, int qw_count) +{ + for (; qw_count > 0; --qw_count) { + VIA_OUT_RING_QW(HC_DUMMY, HC_DUMMY); + } + return vb; +} + +/* + * This function is used internally by ring buffer management code. + * + * Returns virtual pointer to ring buffer. + */ +static inline uint32_t *via_get_dma(drm_via_private_t * dev_priv) +{ + return (uint32_t *) (dev_priv->dma_ptr + dev_priv->dma_low); +} + +/* + * Hooks a segment of data into the tail of the ring-buffer by + * modifying the pause address stored in the buffer itself. If + * the regulator has already paused, restart it. + */ +static int via_hook_segment(drm_via_private_t * dev_priv, + uint32_t pause_addr_hi, uint32_t pause_addr_lo, + int no_pci_fire) +{ + int paused, count; + volatile uint32_t *paused_at = dev_priv->last_pause_ptr; + uint32_t reader,ptr; + uint32_t diff; + + paused = 0; + via_flush_write_combine(); + (void) *(volatile uint32_t *)(via_get_dma(dev_priv) -1); + + *paused_at = pause_addr_lo; + via_flush_write_combine(); + (void) *paused_at; + + reader = *(dev_priv->hw_addr_ptr); + ptr = ((volatile char *)paused_at - dev_priv->dma_ptr) + + dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr + 4; + + dev_priv->last_pause_ptr = via_get_dma(dev_priv) - 1; + + /* + * If there is a possibility that the command reader will + * miss the new pause address and pause on the old one, + * In that case we need to program the new start address + * using PCI. + */ + + diff = (uint32_t) (ptr - reader) - dev_priv->dma_diff; + count = 10000000; + while(diff == 0 && count--) { + paused = (VIA_READ(0x41c) & 0x80000000); + if (paused) + break; + reader = *(dev_priv->hw_addr_ptr); + diff = (uint32_t) (ptr - reader) - dev_priv->dma_diff; + } + + paused = VIA_READ(0x41c) & 0x80000000; + + if (paused && !no_pci_fire) { + reader = *(dev_priv->hw_addr_ptr); + diff = (uint32_t) (ptr - reader) - dev_priv->dma_diff; + diff &= (dev_priv->dma_high - 1); + if (diff != 0 && diff < (dev_priv->dma_high >> 1)) { + DRM_ERROR("Paused at incorrect address. " + "0x%08x, 0x%08x 0x%08x\n", + ptr, reader, dev_priv->dma_diff); + } else if (diff == 0) { + /* + * There is a concern that these writes may stall the PCI bus + * if the GPU is not idle. However, idling the GPU first + * doesn't make a difference. + */ + + VIA_WRITE(VIA_REG_TRANSET, (HC_ParaType_PreCR << 16)); + VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_hi); + VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_lo); + VIA_READ(VIA_REG_TRANSPACE); + } + } + return paused; +} + +static int via_wait_idle(drm_via_private_t * dev_priv) +{ + int count = 10000000; + + while (!(VIA_READ(VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY) && --count) + ; + + while (count && (VIA_READ(VIA_REG_STATUS) & + (VIA_CMD_RGTR_BUSY | VIA_2D_ENG_BUSY | + VIA_3D_ENG_BUSY))) + --count; + return count; +} + +static uint32_t *via_align_cmd(drm_via_private_t * dev_priv, uint32_t cmd_type, + uint32_t addr, uint32_t * cmd_addr_hi, + uint32_t * cmd_addr_lo, int skip_wait) +{ + uint32_t agp_base; + uint32_t cmd_addr, addr_lo, addr_hi; + uint32_t *vb; + uint32_t qw_pad_count; + + if (!skip_wait) + via_cmdbuf_wait(dev_priv, 2 * CMDBUF_ALIGNMENT_SIZE); + + vb = via_get_dma(dev_priv); + VIA_OUT_RING_QW(HC_HEADER2 | ((VIA_REG_TRANSET >> 2) << 12) | + (VIA_REG_TRANSPACE >> 2), HC_ParaType_PreCR << 16); + agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + qw_pad_count = (CMDBUF_ALIGNMENT_SIZE >> 3) - + ((dev_priv->dma_low & CMDBUF_ALIGNMENT_MASK) >> 3); + + cmd_addr = (addr) ? addr : + agp_base + dev_priv->dma_low - 8 + (qw_pad_count << 3); + addr_lo = ((HC_SubA_HAGPBpL << 24) | (cmd_type & HC_HAGPBpID_MASK) | + (cmd_addr & HC_HAGPBpL_MASK)); + addr_hi = ((HC_SubA_HAGPBpH << 24) | (cmd_addr >> 24)); + + vb = via_align_buffer(dev_priv, vb, qw_pad_count - 1); + VIA_OUT_RING_QW(*cmd_addr_hi = addr_hi, *cmd_addr_lo = addr_lo); + return vb; +} + +static void via_cmdbuf_start(drm_via_private_t * dev_priv) +{ + uint32_t pause_addr_lo, pause_addr_hi; + uint32_t start_addr, start_addr_lo; + uint32_t end_addr, end_addr_lo; + uint32_t command; + uint32_t agp_base; + uint32_t ptr; + uint32_t reader; + int count; + + dev_priv->dma_low = 0; + + agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + start_addr = agp_base; + end_addr = agp_base + dev_priv->dma_high; + + start_addr_lo = ((HC_SubA_HAGPBstL << 24) | (start_addr & 0xFFFFFF)); + end_addr_lo = ((HC_SubA_HAGPBendL << 24) | (end_addr & 0xFFFFFF)); + command = ((HC_SubA_HAGPCMNT << 24) | (start_addr >> 24) | + ((end_addr & 0xff000000) >> 16)); + + dev_priv->last_pause_ptr = + via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, + &pause_addr_hi, &pause_addr_lo, 1) - 1; + + via_flush_write_combine(); + (void) *(volatile uint32_t *)dev_priv->last_pause_ptr; + + VIA_WRITE(VIA_REG_TRANSET, (HC_ParaType_PreCR << 16)); + VIA_WRITE(VIA_REG_TRANSPACE, command); + VIA_WRITE(VIA_REG_TRANSPACE, start_addr_lo); + VIA_WRITE(VIA_REG_TRANSPACE, end_addr_lo); + + VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_hi); + VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_lo); + DRM_WRITEMEMORYBARRIER(); + VIA_WRITE(VIA_REG_TRANSPACE, command | HC_HAGPCMNT_MASK); + VIA_READ(VIA_REG_TRANSPACE); + + dev_priv->dma_diff = 0; + + count = 10000000; + while (!(VIA_READ(0x41c) & 0x80000000) && count--); + + reader = *(dev_priv->hw_addr_ptr); + ptr = ((volatile char *)dev_priv->last_pause_ptr - dev_priv->dma_ptr) + + dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr + 4; + + /* + * This is the difference between where we tell the + * command reader to pause and where it actually pauses. + * This differs between hw implementation so we need to + * detect it. + */ + + dev_priv->dma_diff = ptr - reader; +} + +static void via_pad_cache(drm_via_private_t * dev_priv, int qwords) +{ + uint32_t *vb; + + via_cmdbuf_wait(dev_priv, qwords + 2); + vb = via_get_dma(dev_priv); + VIA_OUT_RING_QW(HC_HEADER2, HC_ParaType_NotTex << 16); + via_align_buffer(dev_priv, vb, qwords); +} + +static inline void via_dummy_bitblt(drm_via_private_t * dev_priv) +{ + uint32_t *vb = via_get_dma(dev_priv); + SetReg2DAGP(0x0C, (0 | (0 << 16))); + SetReg2DAGP(0x10, 0 | (0 << 16)); + SetReg2DAGP(0x0, 0x1 | 0x2000 | 0xAA000000); +} + +static void via_cmdbuf_jump(drm_via_private_t * dev_priv) +{ + uint32_t agp_base; + uint32_t pause_addr_lo, pause_addr_hi; + uint32_t jump_addr_lo, jump_addr_hi; + volatile uint32_t *last_pause_ptr; + uint32_t dma_low_save1, dma_low_save2; + + agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr; + via_align_cmd(dev_priv, HC_HAGPBpID_JUMP, 0, &jump_addr_hi, + &jump_addr_lo, 0); + + dev_priv->dma_wrap = dev_priv->dma_low; + + /* + * Wrap command buffer to the beginning. + */ + + dev_priv->dma_low = 0; + if (via_cmdbuf_wait(dev_priv, CMDBUF_ALIGNMENT_SIZE) != 0) { + DRM_ERROR("via_cmdbuf_jump failed\n"); + } + + via_dummy_bitblt(dev_priv); + via_dummy_bitblt(dev_priv); + + last_pause_ptr = + via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi, + &pause_addr_lo, 0) - 1; + via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi, + &pause_addr_lo, 0); + + *last_pause_ptr = pause_addr_lo; + dma_low_save1 = dev_priv->dma_low; + + /* + * Now, set a trap that will pause the regulator if it tries to rerun the old + * command buffer. (Which may happen if via_hook_segment detecs a command regulator pause + * and reissues the jump command over PCI, while the regulator has already taken the jump + * and actually paused at the current buffer end). + * There appears to be no other way to detect this condition, since the hw_addr_pointer + * does not seem to get updated immediately when a jump occurs. + */ + + last_pause_ptr = + via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi, + &pause_addr_lo, 0) - 1; + via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi, + &pause_addr_lo, 0); + *last_pause_ptr = pause_addr_lo; + + dma_low_save2 = dev_priv->dma_low; + dev_priv->dma_low = dma_low_save1; + via_hook_segment(dev_priv, jump_addr_hi, jump_addr_lo, 0); + dev_priv->dma_low = dma_low_save2; + via_hook_segment(dev_priv, pause_addr_hi, pause_addr_lo, 0); +} + + +static void via_cmdbuf_rewind(drm_via_private_t * dev_priv) +{ + via_cmdbuf_jump(dev_priv); +} + +static void via_cmdbuf_flush(drm_via_private_t * dev_priv, uint32_t cmd_type) +{ + uint32_t pause_addr_lo, pause_addr_hi; + + via_align_cmd(dev_priv, cmd_type, 0, &pause_addr_hi, &pause_addr_lo, 0); + via_hook_segment(dev_priv, pause_addr_hi, pause_addr_lo, 0); +} + +static void via_cmdbuf_pause(drm_via_private_t * dev_priv) +{ + via_cmdbuf_flush(dev_priv, HC_HAGPBpID_PAUSE); +} + +static void via_cmdbuf_reset(drm_via_private_t * dev_priv) +{ + via_cmdbuf_flush(dev_priv, HC_HAGPBpID_STOP); + via_wait_idle(dev_priv); +} + +/* + * User interface to the space and lag functions. + */ + +static int via_cmdbuf_size(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_cmdbuf_size_t *d_siz = data; + int ret = 0; + uint32_t tmp_size, count; + drm_via_private_t *dev_priv; + + DRM_DEBUG("\n"); + LOCK_TEST_WITH_RETURN(dev, file_priv); + + dev_priv = (drm_via_private_t *) dev->dev_private; + + if (dev_priv->ring.virtual_start == NULL) { + DRM_ERROR("called without initializing AGP ring buffer.\n"); + return -EFAULT; + } + + count = 1000000; + tmp_size = d_siz->size; + switch (d_siz->func) { + case VIA_CMDBUF_SPACE: + while (((tmp_size = via_cmdbuf_space(dev_priv)) < d_siz->size) + && --count) { + if (!d_siz->wait) { + break; + } + } + if (!count) { + DRM_ERROR("VIA_CMDBUF_SPACE timed out.\n"); + ret = -EAGAIN; + } + break; + case VIA_CMDBUF_LAG: + while (((tmp_size = via_cmdbuf_lag(dev_priv)) > d_siz->size) + && --count) { + if (!d_siz->wait) { + break; + } + } + if (!count) { + DRM_ERROR("VIA_CMDBUF_LAG timed out.\n"); + ret = -EAGAIN; + } + break; + default: + ret = -EFAULT; + } + d_siz->size = tmp_size; + + return ret; +} + +struct drm_ioctl_desc via_ioctls[] = { + DRM_IOCTL_DEF(DRM_VIA_ALLOCMEM, via_mem_alloc, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_FREEMEM, via_mem_free, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_AGP_INIT, via_agp_init, DRM_AUTH|DRM_MASTER), + DRM_IOCTL_DEF(DRM_VIA_FB_INIT, via_fb_init, DRM_AUTH|DRM_MASTER), + DRM_IOCTL_DEF(DRM_VIA_MAP_INIT, via_map_init, DRM_AUTH|DRM_MASTER), + DRM_IOCTL_DEF(DRM_VIA_DEC_FUTEX, via_decoder_futex, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_DMA_INIT, via_dma_init, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_CMDBUFFER, via_cmdbuffer, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_FLUSH, via_flush_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_PCICMD, via_pci_cmdbuffer, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_CMDBUF_SIZE, via_cmdbuf_size, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_WAIT_IRQ, via_wait_irq, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_DMA_BLIT, via_dma_blit, DRM_AUTH), + DRM_IOCTL_DEF(DRM_VIA_BLIT_SYNC, via_dma_blit_sync, DRM_AUTH) +}; + +int via_max_ioctl = DRM_ARRAY_SIZE(via_ioctls); diff --git a/sys/dev/drm/via_dmablit.c b/sys/dev/drm/via_dmablit.c new file mode 100644 index 0000000..9be42ff --- /dev/null +++ b/sys/dev/drm/via_dmablit.c @@ -0,0 +1,801 @@ +/* via_dmablit.c -- PCI DMA BitBlt support for the VIA Unichrome/Pro + * + * Copyright (C) 2005 Thomas Hellstrom, All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Thomas Hellstrom. + * Partially based on code obtained from Digeo Inc. + */ + +#include +__FBSDID("$FreeBSD$"); + +/* + * Unmaps the DMA mappings. + * FIXME: Is this a NoOp on x86? Also + * FIXME: What happens if this one is called and a pending blit has previously done + * the same DMA mappings? + */ + +#include "dev/drm/drmP.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" +#include "dev/drm/via_dmablit.h" + +#define VIA_PGDN(x) (((unsigned long)(x)) & ~PAGE_MASK) +#define VIA_PGOFF(x) (((unsigned long)(x)) & PAGE_MASK) +#define VIA_PFN(x) ((unsigned long)(x) >> PAGE_SHIFT) + +typedef struct _drm_via_descriptor { + uint32_t mem_addr; + uint32_t dev_addr; + uint32_t size; + uint32_t next; +} drm_via_descriptor_t; + +static void via_dmablit_timer(void *arg); + +/* + * Unmap a DMA mapping. + */ +static void +via_unmap_blit_from_device(drm_via_sg_info_t *vsg) +{ + int num_desc = vsg->num_desc; + unsigned cur_descriptor_page = num_desc / vsg->descriptors_per_page; + unsigned descriptor_this_page = num_desc % vsg->descriptors_per_page; + drm_via_descriptor_t *desc_ptr = vsg->desc_pages[cur_descriptor_page] + + descriptor_this_page; + dma_addr_t next = vsg->chain_start; + + while(num_desc--) { + if (descriptor_this_page-- == 0) { + cur_descriptor_page--; + descriptor_this_page = vsg->descriptors_per_page - 1; + desc_ptr = vsg->desc_pages[cur_descriptor_page] + + descriptor_this_page; + } + next = (dma_addr_t) desc_ptr->next; + desc_ptr--; + } +} + + +/* + * If mode = 0, count how many descriptors are needed. + * If mode = 1, Map the DMA pages for the device, put together and map also the descriptors. + * Descriptors are run in reverse order by the hardware because we are not allowed to update the + * 'next' field without syncing calls when the descriptor is already mapped. + */ +static void +via_map_blit_for_device(const drm_via_dmablit_t *xfer, + drm_via_sg_info_t *vsg, int mode) +{ + unsigned cur_descriptor_page = 0; + unsigned num_descriptors_this_page = 0; + unsigned char *mem_addr = xfer->mem_addr; + unsigned char *cur_mem; + unsigned char *first_addr = (unsigned char *)VIA_PGDN(mem_addr); + uint32_t fb_addr = xfer->fb_addr; + uint32_t cur_fb; + unsigned long line_len; + unsigned remaining_len; + int num_desc = 0; + int cur_line; + dma_addr_t next = 0 | VIA_DMA_DPR_EC; + drm_via_descriptor_t *desc_ptr = NULL; + + if (mode == 1) + desc_ptr = vsg->desc_pages[cur_descriptor_page]; + + for (cur_line = 0; cur_line < xfer->num_lines; ++cur_line) { + + line_len = xfer->line_length; + cur_fb = fb_addr; + cur_mem = mem_addr; + + while (line_len > 0) { + + remaining_len = min(PAGE_SIZE - VIA_PGOFF(cur_mem), + line_len); + line_len -= remaining_len; + + if (mode == 1) { + desc_ptr->mem_addr = + VM_PAGE_TO_PHYS( + vsg->pages[VIA_PFN(cur_mem) - + VIA_PFN(first_addr)]) + VIA_PGOFF(cur_mem); + desc_ptr->dev_addr = cur_fb; + + desc_ptr->size = remaining_len; + desc_ptr->next = (uint32_t) next; + + next = vtophys(desc_ptr); + + desc_ptr++; + if (++num_descriptors_this_page >= vsg->descriptors_per_page) { + num_descriptors_this_page = 0; + desc_ptr = vsg->desc_pages[++cur_descriptor_page]; + } + } + + num_desc++; + cur_mem += remaining_len; + cur_fb += remaining_len; + } + + mem_addr += xfer->mem_stride; + fb_addr += xfer->fb_stride; + } + + if (mode == 1) { + vsg->chain_start = next; + vsg->state = dr_via_device_mapped; + } + vsg->num_desc = num_desc; +} + + +/* + * Function that frees up all resources for a blit. It is usable even if the + * blit info has only been partially built as long as the status enum is consistent + * with the actual status of the used resources. + */ +static void +via_free_sg_info(drm_via_sg_info_t *vsg) +{ + vm_page_t page; + int i; + + switch(vsg->state) { + case dr_via_device_mapped: + via_unmap_blit_from_device(vsg); + case dr_via_desc_pages_alloc: + for (i=0; inum_desc_pages; ++i) { + if (vsg->desc_pages[i] != NULL) + free(vsg->desc_pages[i], DRM_MEM_PAGES); + } + free(vsg->desc_pages, DRM_MEM_DRIVER); + case dr_via_pages_locked: + for (i=0; i < vsg->num_pages; ++i) { + if ( NULL != (page = vsg->pages[i])) { + vm_page_lock(page); + vm_page_unwire(page, 0); + vm_page_unlock(page); + } + } + case dr_via_pages_alloc: + free(vsg->pages, DRM_MEM_DRIVER); + default: + vsg->state = dr_via_sg_init; + } + free(vsg->bounce_buffer, DRM_MEM_DRIVER); + vsg->bounce_buffer = NULL; + vsg->free_on_sequence = 0; +} + + +/* + * Fire a blit engine. + */ +static void +via_fire_dmablit(struct drm_device *dev, drm_via_sg_info_t *vsg, int engine) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + + VIA_WRITE(VIA_PCI_DMA_MAR0 + engine*0x10, 0); + VIA_WRITE(VIA_PCI_DMA_DAR0 + engine*0x10, 0); + VIA_WRITE(VIA_PCI_DMA_CSR0 + engine*0x04, VIA_DMA_CSR_DD | VIA_DMA_CSR_TD | + VIA_DMA_CSR_DE); + VIA_WRITE(VIA_PCI_DMA_MR0 + engine*0x04, VIA_DMA_MR_CM | VIA_DMA_MR_TDIE); + VIA_WRITE(VIA_PCI_DMA_BCR0 + engine*0x10, 0); + VIA_WRITE(VIA_PCI_DMA_DPR0 + engine*0x10, vsg->chain_start); + DRM_WRITEMEMORYBARRIER(); + VIA_WRITE(VIA_PCI_DMA_CSR0 + engine*0x04, VIA_DMA_CSR_DE | VIA_DMA_CSR_TS); + VIA_READ(VIA_PCI_DMA_CSR0 + engine*0x04); +} + + +/* + * Obtain a page pointer array and lock all pages into system memory. A segmentation violation will + * occur here if the calling user does not have access to the submitted address. + */ +static int +via_lock_all_dma_pages(drm_via_sg_info_t *vsg, drm_via_dmablit_t *xfer) +{ + unsigned long first_pfn = VIA_PFN(xfer->mem_addr); + vm_page_t m; + vm_map_t map; + int i; + + map = &curproc->p_vmspace->vm_map; + + vsg->num_pages = VIA_PFN(xfer->mem_addr + + (xfer->num_lines * xfer->mem_stride -1)) - first_pfn + 1; + + /* Make sure that the user has access to these pages */ + for(i = 0; i < vsg->num_pages; i++) { + if (vm_fault_quick((caddr_t)xfer->mem_addr + IDX_TO_OFF(i), + VM_PROT_RW) < 0) + return (-EACCES); + } + + if (NULL == (vsg->pages = malloc(sizeof(vm_page_t) * vsg->num_pages, + DRM_MEM_DRIVER, M_NOWAIT | M_ZERO))) + return -ENOMEM; + + for(i = 0; i < vsg->num_pages; i++) { + m = pmap_extract_and_hold(map->pmap, + (vm_offset_t)xfer->mem_addr + IDX_TO_OFF(i), VM_PROT_RW); + if (m == NULL) + break; + vm_page_lock(m); + vm_page_wire(m); + vm_page_unhold(m); + vm_page_unlock(m); + vsg->pages[i] = m; + } + vsg->state = dr_via_pages_locked; + + if (i != vsg->num_pages) + return -EINVAL; + + DRM_DEBUG("DMA pages locked\n"); + + return 0; +} + + +/* + * Allocate DMA capable memory for the blit descriptor chain, and an array that keeps track of the + * pages we allocate. We don't want to use kmalloc for the descriptor chain because it may be + * quite large for some blits, and pages don't need to be contingous. + */ +static int +via_alloc_desc_pages(drm_via_sg_info_t *vsg) +{ + int i; + + vsg->descriptors_per_page = PAGE_SIZE / sizeof(drm_via_descriptor_t); + vsg->num_desc_pages = (vsg->num_desc + vsg->descriptors_per_page - 1) / + vsg->descriptors_per_page; + + if (NULL == (vsg->desc_pages = malloc(vsg->num_desc_pages * + sizeof(void *), DRM_MEM_DRIVER, M_NOWAIT | M_ZERO))) + return -ENOMEM; + + vsg->state = dr_via_desc_pages_alloc; + for (i = 0; i < vsg->num_desc_pages; ++i) { + if (NULL == (vsg->desc_pages[i] = + (drm_via_descriptor_t *)malloc(PAGE_SIZE, DRM_MEM_PAGES, + M_NOWAIT | M_ZERO))) + return -ENOMEM; + } + DRM_DEBUG("Allocated %d pages for %d descriptors.\n", + vsg->num_desc_pages, vsg->num_desc); + + return 0; +} + + +static void +via_abort_dmablit(struct drm_device *dev, int engine) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + + VIA_WRITE(VIA_PCI_DMA_CSR0 + engine*0x04, VIA_DMA_CSR_TA); +} + + +static void +via_dmablit_engine_off(struct drm_device *dev, int engine) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + + VIA_WRITE(VIA_PCI_DMA_CSR0 + engine*0x04, VIA_DMA_CSR_TD | VIA_DMA_CSR_DD); +} + + +/* + * The dmablit part of the IRQ handler. Trying to do only reasonably fast things here. + * The rest, like unmapping and freeing memory for done blits is done in a separate workqueue + * task. Basically the task of the interrupt handler is to submit a new blit to the engine, while + * the workqueue task takes care of processing associated with the old blit. + */ +void +via_dmablit_handler(struct drm_device *dev, int engine, int from_irq) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + drm_via_blitq_t *blitq = dev_priv->blit_queues + engine; + int cur; + int done_transfer; + uint32_t status = 0; + + DRM_DEBUG("DMA blit handler called. engine = %d, from_irq = %d, blitq = 0x%lx\n", + engine, from_irq, (unsigned long) blitq); + + mtx_lock(&blitq->blit_lock); + + done_transfer = blitq->is_active && + (( status = VIA_READ(VIA_PCI_DMA_CSR0 + engine*0x04)) & VIA_DMA_CSR_TD); + done_transfer = done_transfer || ( blitq->aborting && !(status & VIA_DMA_CSR_DE)); + + cur = blitq->cur; + if (done_transfer) { + + blitq->blits[cur]->aborted = blitq->aborting; + blitq->done_blit_handle++; + DRM_WAKEUP(&blitq->blit_queue[cur]); + + cur++; + if (cur >= VIA_NUM_BLIT_SLOTS) + cur = 0; + blitq->cur = cur; + + /* + * Clear transfer done flag. + */ + + VIA_WRITE(VIA_PCI_DMA_CSR0 + engine*0x04, VIA_DMA_CSR_TD); + + blitq->is_active = 0; + blitq->aborting = 0; + + taskqueue_enqueue(taskqueue_swi, &blitq->wq); + + } else if (blitq->is_active && (ticks >= blitq->end)) { + + /* + * Abort transfer after one second. + */ + + via_abort_dmablit(dev, engine); + blitq->aborting = 1; + blitq->end = ticks + DRM_HZ; + } + + if (!blitq->is_active) { + if (blitq->num_outstanding) { + via_fire_dmablit(dev, blitq->blits[cur], engine); + blitq->is_active = 1; + blitq->cur = cur; + blitq->num_outstanding--; + blitq->end = ticks + DRM_HZ; + + if (!callout_pending(&blitq->poll_timer)) + callout_reset(&blitq->poll_timer, + 1, (timeout_t *)via_dmablit_timer, + (void *)blitq); + } else { + if (callout_pending(&blitq->poll_timer)) { + callout_stop(&blitq->poll_timer); + } + via_dmablit_engine_off(dev, engine); + } + } + + mtx_unlock(&blitq->blit_lock); +} + + +/* + * Check whether this blit is still active, performing necessary locking. + */ +static int +via_dmablit_active(drm_via_blitq_t *blitq, int engine, uint32_t handle, wait_queue_head_t **queue) +{ + uint32_t slot; + int active; + + mtx_lock(&blitq->blit_lock); + + /* + * Allow for handle wraparounds. + */ + active = ((blitq->done_blit_handle - handle) > (1 << 23)) && + ((blitq->cur_blit_handle - handle) <= (1 << 23)); + + if (queue && active) { + slot = handle - blitq->done_blit_handle + blitq->cur -1; + if (slot >= VIA_NUM_BLIT_SLOTS) { + slot -= VIA_NUM_BLIT_SLOTS; + } + *queue = blitq->blit_queue + slot; + } + + mtx_unlock(&blitq->blit_lock); + + return active; +} + + +/* + * Sync. Wait for at least three seconds for the blit to be performed. + */ +static int +via_dmablit_sync(struct drm_device *dev, uint32_t handle, int engine) +{ + + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + drm_via_blitq_t *blitq = dev_priv->blit_queues + engine; + wait_queue_head_t *queue; + int ret = 0; + + if (via_dmablit_active(blitq, engine, handle, &queue)) { + DRM_WAIT_ON(ret, *queue, 3 * DRM_HZ, + !via_dmablit_active(blitq, engine, handle, NULL)); + } + DRM_DEBUG("DMA blit sync handle 0x%x engine %d returned %d\n", + handle, engine, ret); + + return ret; +} + + +/* + * A timer that regularly polls the blit engine in cases where we don't have interrupts: + * a) Broken hardware (typically those that don't have any video capture facility). + * b) Blit abort. The hardware doesn't send an interrupt when a blit is aborted. + * The timer and hardware IRQ's can and do work in parallel. If the hardware has + * irqs, it will shorten the latency somewhat. + */ +static void +via_dmablit_timer(void *arg) +{ + drm_via_blitq_t *blitq = (drm_via_blitq_t *)arg; + struct drm_device *dev = blitq->dev; + int engine = (int) + (blitq - ((drm_via_private_t *)dev->dev_private)->blit_queues); + + DRM_DEBUG("Polling timer called for engine %d, jiffies %lu\n", engine, + (unsigned long) jiffies); + + via_dmablit_handler(dev, engine, 0); + + if (!callout_pending(&blitq->poll_timer)) { + callout_schedule(&blitq->poll_timer, 1); + + /* + * Rerun handler to delete timer if engines are off, and + * to shorten abort latency. This is a little nasty. + */ + + via_dmablit_handler(dev, engine, 0); + + } +} + + +/* + * Workqueue task that frees data and mappings associated with a blit. + * Also wakes up waiting processes. Each of these tasks handles one + * blit engine only and may not be called on each interrupt. + */ +static void +via_dmablit_workqueue(void *arg, int pending) +{ + drm_via_blitq_t *blitq = (drm_via_blitq_t *)arg; + struct drm_device *dev = blitq->dev; + drm_via_sg_info_t *cur_sg; + int cur_released; + + + DRM_DEBUG("task called for blit engine %ld\n",(unsigned long) + (blitq - ((drm_via_private_t *)dev->dev_private)->blit_queues)); + + mtx_lock(&blitq->blit_lock); + + while(blitq->serviced != blitq->cur) { + + cur_released = blitq->serviced++; + + DRM_DEBUG("Releasing blit slot %d\n", cur_released); + + if (blitq->serviced >= VIA_NUM_BLIT_SLOTS) + blitq->serviced = 0; + + cur_sg = blitq->blits[cur_released]; + blitq->num_free++; + + mtx_unlock(&blitq->blit_lock); + + DRM_WAKEUP(&blitq->busy_queue); + + via_free_sg_info(cur_sg); + free(cur_sg, DRM_MEM_DRIVER); + + mtx_lock(&blitq->blit_lock); + } + + mtx_unlock(&blitq->blit_lock); +} + + +/* + * Init all blit engines. Currently we use two, but some hardware have 4. + */ +void +via_init_dmablit(struct drm_device *dev) +{ + int i,j; + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + drm_via_blitq_t *blitq; + + for (i=0; i< VIA_NUM_BLIT_ENGINES; ++i) { + blitq = dev_priv->blit_queues + i; + blitq->dev = dev; + blitq->cur_blit_handle = 0; + blitq->done_blit_handle = 0; + blitq->head = 0; + blitq->cur = 0; + blitq->serviced = 0; + blitq->num_free = VIA_NUM_BLIT_SLOTS - 1; + blitq->num_outstanding = 0; + blitq->is_active = 0; + blitq->aborting = 0; + mtx_init(&blitq->blit_lock, "via_blit_lk", NULL, MTX_DEF); + for (j=0; jblit_queue + j); + } + DRM_INIT_WAITQUEUE(&blitq->busy_queue); + TASK_INIT(&blitq->wq, 0, via_dmablit_workqueue, blitq); + callout_init(&blitq->poll_timer, 0); + } +} + + +/* + * Build all info and do all mappings required for a blit. + */ +static int +via_build_sg_info(struct drm_device *dev, drm_via_sg_info_t *vsg, + drm_via_dmablit_t *xfer) +{ + int ret = 0; + + vsg->bounce_buffer = NULL; + + vsg->state = dr_via_sg_init; + + if (xfer->num_lines <= 0 || xfer->line_length <= 0) { + DRM_ERROR("Zero size bitblt.\n"); + return -EINVAL; + } + + /* + * Below check is a driver limitation, not a hardware one. We + * don't want to lock unused pages, and don't want to incoporate the + * extra logic of avoiding them. Make sure there are no. + * (Not a big limitation anyway.) + */ + if ((xfer->mem_stride - xfer->line_length) > 2 * PAGE_SIZE) { + DRM_ERROR("Too large system memory stride. Stride: %d, " + "Length: %d\n", xfer->mem_stride, xfer->line_length); + return -EINVAL; + } + + if ((xfer->mem_stride == xfer->line_length) && + (xfer->fb_stride == xfer->line_length)) { + xfer->mem_stride *= xfer->num_lines; + xfer->line_length = xfer->mem_stride; + xfer->fb_stride = xfer->mem_stride; + xfer->num_lines = 1; + } + + /* + * Don't lock an arbitrary large number of pages, since that causes a + * DOS security hole. + */ + if (xfer->num_lines > 2048 || + (xfer->num_lines*xfer->mem_stride > (2048*2048*4))) { + DRM_ERROR("Too large PCI DMA bitblt.\n"); + return -EINVAL; + } + + /* + * we allow a negative fb stride to allow flipping of images in + * transfer. + */ + if (xfer->mem_stride < xfer->line_length || + abs(xfer->fb_stride) < xfer->line_length) { + DRM_ERROR("Invalid frame-buffer / memory stride.\n"); + return -EINVAL; + } + + /* + * A hardware bug seems to be worked around if system memory addresses + * start on 16 byte boundaries. This seems a bit restrictive however. + * VIA is contacted about this. Meanwhile, impose the following + * restrictions: + */ +#ifdef VIA_BUGFREE + if ((((unsigned long)xfer->mem_addr & 3) != + ((unsigned long)xfer->fb_addr & 3)) || + ((xfer->num_lines > 1) && ((xfer->mem_stride & 3) != + (xfer->fb_stride & 3)))) { + DRM_ERROR("Invalid DRM bitblt alignment.\n"); + return -EINVAL; + } +#else + if ((((unsigned long)xfer->mem_addr & 15) || + ((unsigned long)xfer->fb_addr & 3)) || + ((xfer->num_lines > 1) && + ((xfer->mem_stride & 15) || (xfer->fb_stride & 3)))) { + DRM_ERROR("Invalid DRM bitblt alignment.\n"); + return -EINVAL; + } +#endif + + if (0 != (ret = via_lock_all_dma_pages(vsg, xfer))) { + DRM_ERROR("Could not lock DMA pages.\n"); + via_free_sg_info(vsg); + return ret; + } + + via_map_blit_for_device(xfer, vsg, 0); + if (0 != (ret = via_alloc_desc_pages(vsg))) { + DRM_ERROR("Could not allocate DMA descriptor pages.\n"); + via_free_sg_info(vsg); + return ret; + } + via_map_blit_for_device(xfer, vsg, 1); + + return 0; +} + + +/* + * Reserve one free slot in the blit queue. Will wait for one second for one + * to become available. Otherwise -EBUSY is returned. + */ +static int +via_dmablit_grab_slot(drm_via_blitq_t *blitq, int engine) +{ + struct drm_device *dev = blitq->dev; + int ret=0; + + DRM_DEBUG("Num free is %d\n", blitq->num_free); + mtx_lock(&blitq->blit_lock); + while(blitq->num_free == 0) { + mtx_unlock(&blitq->blit_lock); + + DRM_WAIT_ON(ret, blitq->busy_queue, DRM_HZ, + blitq->num_free > 0); + if (ret) { + return (-EINTR == ret) ? -EAGAIN : ret; + } + + mtx_lock(&blitq->blit_lock); + } + + blitq->num_free--; + mtx_unlock(&blitq->blit_lock); + + return 0; +} + + +/* + * Hand back a free slot if we changed our mind. + */ +static void +via_dmablit_release_slot(drm_via_blitq_t *blitq) +{ + + mtx_lock(&blitq->blit_lock); + blitq->num_free++; + mtx_unlock(&blitq->blit_lock); + DRM_WAKEUP( &blitq->busy_queue ); +} + + +/* + * Grab a free slot. Build blit info and queue a blit. + */ +static int +via_dmablit(struct drm_device *dev, drm_via_dmablit_t *xfer) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *)dev->dev_private; + drm_via_sg_info_t *vsg; + drm_via_blitq_t *blitq; + int ret; + int engine; + + if (dev_priv == NULL) { + DRM_ERROR("Called without initialization.\n"); + return -EINVAL; + } + + engine = (xfer->to_fb) ? 0 : 1; + blitq = dev_priv->blit_queues + engine; + if (0 != (ret = via_dmablit_grab_slot(blitq, engine))) { + return ret; + } + if (NULL == (vsg = malloc(sizeof(*vsg), DRM_MEM_DRIVER, + M_NOWAIT | M_ZERO))) { + via_dmablit_release_slot(blitq); + return -ENOMEM; + } + if (0 != (ret = via_build_sg_info(dev, vsg, xfer))) { + via_dmablit_release_slot(blitq); + free(vsg, DRM_MEM_DRIVER); + return ret; + } + mtx_lock(&blitq->blit_lock); + + blitq->blits[blitq->head++] = vsg; + if (blitq->head >= VIA_NUM_BLIT_SLOTS) + blitq->head = 0; + blitq->num_outstanding++; + xfer->sync.sync_handle = ++blitq->cur_blit_handle; + + mtx_unlock(&blitq->blit_lock); + xfer->sync.engine = engine; + + via_dmablit_handler(dev, engine, 0); + + return 0; +} + + +/* + * Sync on a previously submitted blit. Note that the X server use signals + * extensively, and that there is a very big probability that this IOCTL will + * be interrupted by a signal. In that case it returns with -EAGAIN for the + * signal to be delivered. The caller should then reissue the IOCTL. This is + * similar to what is being done for drmGetLock(). + */ +int +via_dma_blit_sync( struct drm_device *dev, void *data, + struct drm_file *file_priv ) +{ + drm_via_blitsync_t *sync = data; + int err; + + if (sync->engine >= VIA_NUM_BLIT_ENGINES) + return -EINVAL; + + err = via_dmablit_sync(dev, sync->sync_handle, sync->engine); + + if (-EINTR == err) + err = -EAGAIN; + + return err; +} + + +/* + * Queue a blit and hand back a handle to be used for sync. This IOCTL may be + * interrupted by a signal while waiting for a free slot in the blit queue. + * In that case it returns with -EAGAIN and should be reissued. See the above + * IOCTL code. + */ +int +via_dma_blit( struct drm_device *dev, void *data, struct drm_file *file_priv ) +{ + drm_via_dmablit_t *xfer = data; + int err; + + err = via_dmablit(dev, xfer); + + return err; +} diff --git a/sys/dev/drm/via_dmablit.h b/sys/dev/drm/via_dmablit.h new file mode 100644 index 0000000..c5463cb --- /dev/null +++ b/sys/dev/drm/via_dmablit.h @@ -0,0 +1,140 @@ +/* via_dmablit.h -- PCI DMA BitBlt support for the VIA Unichrome/Pro + * + * Copyright 2005 Thomas Hellstrom. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Thomas Hellstrom. + * Register info from Digeo Inc. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifndef _VIA_DMABLIT_H +#define _VIA_DMABLIT_H + +#define VIA_NUM_BLIT_ENGINES 2 +#define VIA_NUM_BLIT_SLOTS 8 + +struct _drm_via_descriptor; + +typedef struct _drm_via_sg_info { + vm_page_t *pages; + unsigned long num_pages; + struct _drm_via_descriptor **desc_pages; + int num_desc_pages; + int num_desc; + unsigned char *bounce_buffer; + dma_addr_t chain_start; + uint32_t free_on_sequence; + unsigned int descriptors_per_page; + int aborted; + enum { + dr_via_device_mapped, + dr_via_desc_pages_alloc, + dr_via_pages_locked, + dr_via_pages_alloc, + dr_via_sg_init + } state; +} drm_via_sg_info_t; + +typedef struct _drm_via_blitq { + struct drm_device *dev; + uint32_t cur_blit_handle; + uint32_t done_blit_handle; + unsigned serviced; + unsigned head; + unsigned cur; + unsigned num_free; + unsigned num_outstanding; + unsigned long end; + int aborting; + int is_active; + drm_via_sg_info_t *blits[VIA_NUM_BLIT_SLOTS]; + struct mtx blit_lock; + wait_queue_head_t blit_queue[VIA_NUM_BLIT_SLOTS]; + wait_queue_head_t busy_queue; + struct task wq; + struct callout poll_timer; +} drm_via_blitq_t; + + +/* + * PCI DMA Registers + * Channels 2 & 3 don't seem to be implemented in hardware. + */ + +#define VIA_PCI_DMA_MAR0 0xE40 /* Memory Address Register of Channel 0 */ +#define VIA_PCI_DMA_DAR0 0xE44 /* Device Address Register of Channel 0 */ +#define VIA_PCI_DMA_BCR0 0xE48 /* Byte Count Register of Channel 0 */ +#define VIA_PCI_DMA_DPR0 0xE4C /* Descriptor Pointer Register of Channel 0 */ + +#define VIA_PCI_DMA_MAR1 0xE50 /* Memory Address Register of Channel 1 */ +#define VIA_PCI_DMA_DAR1 0xE54 /* Device Address Register of Channel 1 */ +#define VIA_PCI_DMA_BCR1 0xE58 /* Byte Count Register of Channel 1 */ +#define VIA_PCI_DMA_DPR1 0xE5C /* Descriptor Pointer Register of Channel 1 */ + +#define VIA_PCI_DMA_MAR2 0xE60 /* Memory Address Register of Channel 2 */ +#define VIA_PCI_DMA_DAR2 0xE64 /* Device Address Register of Channel 2 */ +#define VIA_PCI_DMA_BCR2 0xE68 /* Byte Count Register of Channel 2 */ +#define VIA_PCI_DMA_DPR2 0xE6C /* Descriptor Pointer Register of Channel 2 */ + +#define VIA_PCI_DMA_MAR3 0xE70 /* Memory Address Register of Channel 3 */ +#define VIA_PCI_DMA_DAR3 0xE74 /* Device Address Register of Channel 3 */ +#define VIA_PCI_DMA_BCR3 0xE78 /* Byte Count Register of Channel 3 */ +#define VIA_PCI_DMA_DPR3 0xE7C /* Descriptor Pointer Register of Channel 3 */ + +#define VIA_PCI_DMA_MR0 0xE80 /* Mode Register of Channel 0 */ +#define VIA_PCI_DMA_MR1 0xE84 /* Mode Register of Channel 1 */ +#define VIA_PCI_DMA_MR2 0xE88 /* Mode Register of Channel 2 */ +#define VIA_PCI_DMA_MR3 0xE8C /* Mode Register of Channel 3 */ + +#define VIA_PCI_DMA_CSR0 0xE90 /* Command/Status Register of Channel 0 */ +#define VIA_PCI_DMA_CSR1 0xE94 /* Command/Status Register of Channel 1 */ +#define VIA_PCI_DMA_CSR2 0xE98 /* Command/Status Register of Channel 2 */ +#define VIA_PCI_DMA_CSR3 0xE9C /* Command/Status Register of Channel 3 */ + +#define VIA_PCI_DMA_PTR 0xEA0 /* Priority Type Register */ + +/* Define for DMA engine */ +/* DPR */ +#define VIA_DMA_DPR_EC (1<<1) /* end of chain */ +#define VIA_DMA_DPR_DDIE (1<<2) /* descriptor done interrupt enable */ +#define VIA_DMA_DPR_DT (1<<3) /* direction of transfer (RO) */ + +/* MR */ +#define VIA_DMA_MR_CM (1<<0) /* chaining mode */ +#define VIA_DMA_MR_TDIE (1<<1) /* transfer done interrupt enable */ +#define VIA_DMA_MR_HENDMACMD (1<<7) /* ? */ + +/* CSR */ +#define VIA_DMA_CSR_DE (1<<0) /* DMA enable */ +#define VIA_DMA_CSR_TS (1<<1) /* transfer start */ +#define VIA_DMA_CSR_TA (1<<2) /* transfer abort */ +#define VIA_DMA_CSR_TD (1<<3) /* transfer done */ +#define VIA_DMA_CSR_DD (1<<4) /* descriptor done */ +#define VIA_DMA_DPR_EC (1<<1) /* end of chain */ + + + +#endif diff --git a/sys/dev/drm/via_drm.h b/sys/dev/drm/via_drm.h new file mode 100644 index 0000000..2f40cab --- /dev/null +++ b/sys/dev/drm/via_drm.h @@ -0,0 +1,279 @@ +/*- + * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. + * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifndef _VIA_DRM_H_ +#define _VIA_DRM_H_ + +/* WARNING: These defines must be the same as what the Xserver uses. + * if you change them, you must change the defines in the Xserver. + */ + +#ifndef _VIA_DEFINES_ +#define _VIA_DEFINES_ + +#if !defined(__KERNEL__) && !defined(_KERNEL) +#include "via_drmclient.h" +#endif + +#define VIA_NR_SAREA_CLIPRECTS 8 +#define VIA_NR_XVMC_PORTS 10 +#define VIA_NR_XVMC_LOCKS 5 +#define VIA_MAX_CACHELINE_SIZE 64 +#define XVMCLOCKPTR(saPriv,lockNo) \ + ((volatile struct drm_hw_lock *)(((((unsigned long) (saPriv)->XvMCLockArea) + \ + (VIA_MAX_CACHELINE_SIZE - 1)) & \ + ~(VIA_MAX_CACHELINE_SIZE - 1)) + \ + VIA_MAX_CACHELINE_SIZE*(lockNo))) + +/* Each region is a minimum of 64k, and there are at most 64 of them. + */ +#define VIA_NR_TEX_REGIONS 64 +#define VIA_LOG_MIN_TEX_REGION_SIZE 16 +#endif + +#define VIA_UPLOAD_TEX0IMAGE 0x1 /* handled clientside */ +#define VIA_UPLOAD_TEX1IMAGE 0x2 /* handled clientside */ +#define VIA_UPLOAD_CTX 0x4 +#define VIA_UPLOAD_BUFFERS 0x8 +#define VIA_UPLOAD_TEX0 0x10 +#define VIA_UPLOAD_TEX1 0x20 +#define VIA_UPLOAD_CLIPRECTS 0x40 +#define VIA_UPLOAD_ALL 0xff + +/* VIA specific ioctls */ +#define DRM_VIA_ALLOCMEM 0x00 +#define DRM_VIA_FREEMEM 0x01 +#define DRM_VIA_AGP_INIT 0x02 +#define DRM_VIA_FB_INIT 0x03 +#define DRM_VIA_MAP_INIT 0x04 +#define DRM_VIA_DEC_FUTEX 0x05 +#define NOT_USED +#define DRM_VIA_DMA_INIT 0x07 +#define DRM_VIA_CMDBUFFER 0x08 +#define DRM_VIA_FLUSH 0x09 +#define DRM_VIA_PCICMD 0x0a +#define DRM_VIA_CMDBUF_SIZE 0x0b +#define NOT_USED +#define DRM_VIA_WAIT_IRQ 0x0d +#define DRM_VIA_DMA_BLIT 0x0e +#define DRM_VIA_BLIT_SYNC 0x0f + +#define DRM_IOCTL_VIA_ALLOCMEM DRM_IOWR(DRM_COMMAND_BASE + DRM_VIA_ALLOCMEM, drm_via_mem_t) +#define DRM_IOCTL_VIA_FREEMEM DRM_IOW( DRM_COMMAND_BASE + DRM_VIA_FREEMEM, drm_via_mem_t) +#define DRM_IOCTL_VIA_AGP_INIT DRM_IOWR(DRM_COMMAND_BASE + DRM_VIA_AGP_INIT, drm_via_agp_t) +#define DRM_IOCTL_VIA_FB_INIT DRM_IOWR(DRM_COMMAND_BASE + DRM_VIA_FB_INIT, drm_via_fb_t) +#define DRM_IOCTL_VIA_MAP_INIT DRM_IOWR(DRM_COMMAND_BASE + DRM_VIA_MAP_INIT, drm_via_init_t) +#define DRM_IOCTL_VIA_DEC_FUTEX DRM_IOW( DRM_COMMAND_BASE + DRM_VIA_DEC_FUTEX, drm_via_futex_t) +#define DRM_IOCTL_VIA_DMA_INIT DRM_IOWR(DRM_COMMAND_BASE + DRM_VIA_DMA_INIT, drm_via_dma_init_t) +#define DRM_IOCTL_VIA_CMDBUFFER DRM_IOW( DRM_COMMAND_BASE + DRM_VIA_CMDBUFFER, drm_via_cmdbuffer_t) +#define DRM_IOCTL_VIA_FLUSH DRM_IO( DRM_COMMAND_BASE + DRM_VIA_FLUSH) +#define DRM_IOCTL_VIA_PCICMD DRM_IOW( DRM_COMMAND_BASE + DRM_VIA_PCICMD, drm_via_cmdbuffer_t) +#define DRM_IOCTL_VIA_CMDBUF_SIZE DRM_IOWR( DRM_COMMAND_BASE + DRM_VIA_CMDBUF_SIZE, \ + drm_via_cmdbuf_size_t) +#define DRM_IOCTL_VIA_WAIT_IRQ DRM_IOWR( DRM_COMMAND_BASE + DRM_VIA_WAIT_IRQ, drm_via_irqwait_t) +#define DRM_IOCTL_VIA_DMA_BLIT DRM_IOW(DRM_COMMAND_BASE + DRM_VIA_DMA_BLIT, drm_via_dmablit_t) +#define DRM_IOCTL_VIA_BLIT_SYNC DRM_IOW(DRM_COMMAND_BASE + DRM_VIA_BLIT_SYNC, drm_via_blitsync_t) + +/* Indices into buf.Setup where various bits of state are mirrored per + * context and per buffer. These can be fired at the card as a unit, + * or in a piecewise fashion as required. + */ + +#define VIA_TEX_SETUP_SIZE 8 + +/* Flags for clear ioctl + */ +#define VIA_FRONT 0x1 +#define VIA_BACK 0x2 +#define VIA_DEPTH 0x4 +#define VIA_STENCIL 0x8 +#define VIA_MEM_VIDEO 0 /* matches drm constant */ +#define VIA_MEM_AGP 1 /* matches drm constant */ +#define VIA_MEM_SYSTEM 2 +#define VIA_MEM_MIXED 3 +#define VIA_MEM_UNKNOWN 4 + +typedef struct { + u32 offset; + u32 size; +} drm_via_agp_t; + +typedef struct { + u32 offset; + u32 size; +} drm_via_fb_t; + +typedef struct { + u32 context; + u32 type; + u32 size; + unsigned long index; + unsigned long offset; +} drm_via_mem_t; + +typedef struct _drm_via_init { + enum { + VIA_INIT_MAP = 0x01, + VIA_CLEANUP_MAP = 0x02 + } func; + + unsigned long sarea_priv_offset; + unsigned long fb_offset; + unsigned long mmio_offset; + unsigned long agpAddr; +} drm_via_init_t; + +typedef struct _drm_via_futex { + enum { + VIA_FUTEX_WAIT = 0x00, + VIA_FUTEX_WAKE = 0X01 + } func; + u32 ms; + u32 lock; + u32 val; +} drm_via_futex_t; + +typedef struct _drm_via_dma_init { + enum { + VIA_INIT_DMA = 0x01, + VIA_CLEANUP_DMA = 0x02, + VIA_DMA_INITIALIZED = 0x03 + } func; + + unsigned long offset; + unsigned long size; + unsigned long reg_pause_addr; +} drm_via_dma_init_t; + +typedef struct _drm_via_cmdbuffer { + char __user *buf; + unsigned long size; +} drm_via_cmdbuffer_t; + +/* Warning: If you change the SAREA structure you must change the Xserver + * structure as well */ + +typedef struct _drm_via_tex_region { + unsigned char next, prev; /* indices to form a circular LRU */ + unsigned char inUse; /* owned by a client, or free? */ + int age; /* tracked by clients to update local LRU's */ +} drm_via_tex_region_t; + +typedef struct _drm_via_sarea { + unsigned int dirty; + unsigned int nbox; + struct drm_clip_rect boxes[VIA_NR_SAREA_CLIPRECTS]; + drm_via_tex_region_t texList[VIA_NR_TEX_REGIONS + 1]; + int texAge; /* last time texture was uploaded */ + int ctxOwner; /* last context to upload state */ + int vertexPrim; + + /* + * Below is for XvMC. + * We want the lock integers alone on, and aligned to, a cache line. + * Therefore this somewhat strange construct. + */ + + char XvMCLockArea[VIA_MAX_CACHELINE_SIZE * (VIA_NR_XVMC_LOCKS + 1)]; + + unsigned int XvMCDisplaying[VIA_NR_XVMC_PORTS]; + unsigned int XvMCSubPicOn[VIA_NR_XVMC_PORTS]; + unsigned int XvMCCtxNoGrabbed; /* Last context to hold decoder */ + + /* Used by the 3d driver only at this point, for pageflipping: + */ + unsigned int pfCurrentOffset; +} drm_via_sarea_t; + +typedef struct _drm_via_cmdbuf_size { + enum { + VIA_CMDBUF_SPACE = 0x01, + VIA_CMDBUF_LAG = 0x02 + } func; + int wait; + u32 size; +} drm_via_cmdbuf_size_t; + +typedef enum { + VIA_IRQ_ABSOLUTE = 0x0, + VIA_IRQ_RELATIVE = 0x1, + VIA_IRQ_SIGNAL = 0x10000000, + VIA_IRQ_FORCE_SEQUENCE = 0x20000000 +} via_irq_seq_type_t; + +#define VIA_IRQ_FLAGS_MASK 0xF0000000 + +enum drm_via_irqs { + drm_via_irq_hqv0 = 0, + drm_via_irq_hqv1, + drm_via_irq_dma0_dd, + drm_via_irq_dma0_td, + drm_via_irq_dma1_dd, + drm_via_irq_dma1_td, + drm_via_irq_num +}; + +struct drm_via_wait_irq_request { + unsigned irq; + via_irq_seq_type_t type; + u32 sequence; + u32 signal; +}; + +typedef union drm_via_irqwait { + struct drm_via_wait_irq_request request; + struct drm_wait_vblank_reply reply; +} drm_via_irqwait_t; + +typedef struct drm_via_blitsync { + u32 sync_handle; + unsigned engine; +} drm_via_blitsync_t; + +/* - * Below,"flags" is currently unused but will be used for possible future + * extensions like kernel space bounce buffers for bad alignments and + * blit engine busy-wait polling for better latency in the absence of + * interrupts. + */ + +typedef struct drm_via_dmablit { + u32 num_lines; + u32 line_length; + + u32 fb_addr; + u32 fb_stride; + + unsigned char *mem_addr; + u32 mem_stride; + + u32 flags; + int to_fb; + + drm_via_blitsync_t sync; +} drm_via_dmablit_t; + +#endif /* _VIA_DRM_H_ */ diff --git a/sys/dev/drm/via_drv.c b/sys/dev/drm/via_drv.c new file mode 100644 index 0000000..520518d --- /dev/null +++ b/sys/dev/drm/via_drv.c @@ -0,0 +1,123 @@ +/* via_drv.c -- VIA unichrome driver -*- linux-c -*- + * Created: Fri Aug 12 2005 by anholt@FreeBSD.org + */ +/*- + * Copyright 2005 Eric Anholt + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * ERIC ANHOLT BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER + * IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Eric Anholt + * + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/drm.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" +#include "dev/drm/drm_pciids.h" + +/* drv_PCI_IDs comes from drm_pciids.h, generated from drm_pciids.txt. */ +static drm_pci_id_list_t via_pciidlist[] = { + viadrv_PCI_IDS +}; + +static void via_configure(struct drm_device *dev) +{ + dev->driver->driver_features = + DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_HAVE_IRQ; + + dev->driver->buf_priv_size = sizeof(drm_via_private_t); + dev->driver->load = via_driver_load; + dev->driver->unload = via_driver_unload; + dev->driver->lastclose = via_lastclose; + dev->driver->get_vblank_counter = via_get_vblank_counter; + dev->driver->enable_vblank = via_enable_vblank; + dev->driver->disable_vblank = via_disable_vblank; + dev->driver->irq_preinstall = via_driver_irq_preinstall; + dev->driver->irq_postinstall = via_driver_irq_postinstall; + dev->driver->irq_uninstall = via_driver_irq_uninstall; + dev->driver->irq_handler = via_driver_irq_handler; + dev->driver->dma_quiescent = via_driver_dma_quiescent; + + dev->driver->ioctls = via_ioctls; + dev->driver->max_ioctl = via_max_ioctl; + + dev->driver->name = DRIVER_NAME; + dev->driver->desc = DRIVER_DESC; + dev->driver->date = DRIVER_DATE; + dev->driver->major = DRIVER_MAJOR; + dev->driver->minor = DRIVER_MINOR; + dev->driver->patchlevel = DRIVER_PATCHLEVEL; +} + +static int +via_probe(device_t kdev) +{ + return drm_probe(kdev, via_pciidlist); +} + +static int +via_attach(device_t kdev) +{ + struct drm_device *dev = device_get_softc(kdev); + + dev->driver = malloc(sizeof(struct drm_driver_info), DRM_MEM_DRIVER, + M_WAITOK | M_ZERO); + + via_configure(dev); + + return drm_attach(kdev, via_pciidlist); +} + +static int +via_detach(device_t kdev) +{ + struct drm_device *dev = device_get_softc(kdev); + int ret; + + ret = drm_detach(kdev); + + free(dev->driver, DRM_MEM_DRIVER); + + return ret; +} + +static device_method_t via_methods[] = { + /* Device interface */ + DEVMETHOD(device_probe, via_probe), + DEVMETHOD(device_attach, via_attach), + DEVMETHOD(device_detach, via_detach), + + { 0, 0 } +}; + +static driver_t via_driver = { + "drm", + via_methods, + sizeof(struct drm_device) +}; + +extern devclass_t drm_devclass; +DRIVER_MODULE(via, vgapci, via_driver, drm_devclass, 0, 0); +MODULE_DEPEND(via, drm, 1, 1, 1); diff --git a/sys/dev/drm/via_drv.h b/sys/dev/drm/via_drv.h new file mode 100644 index 0000000..3b692f1 --- /dev/null +++ b/sys/dev/drm/via_drv.h @@ -0,0 +1,161 @@ +/*- + * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. + * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifndef _VIA_DRV_H_ +#define _VIA_DRV_H_ + +#include "dev/drm/drm_sman.h" +#define DRIVER_AUTHOR "Various" + +#define DRIVER_NAME "via" +#define DRIVER_DESC "VIA Unichrome / Pro" +#define DRIVER_DATE "20070202" + +#define DRIVER_MAJOR 2 +#define DRIVER_MINOR 11 +#define DRIVER_PATCHLEVEL 1 + +#include "dev/drm/via_verifier.h" + +#include "dev/drm/via_dmablit.h" + +#define VIA_PCI_BUF_SIZE 60000 +#define VIA_FIRE_BUF_SIZE 1024 +#define VIA_NUM_IRQS 4 + +typedef struct drm_via_ring_buffer { + drm_local_map_t map; + char *virtual_start; +} drm_via_ring_buffer_t; + +typedef uint32_t maskarray_t[5]; + +typedef struct drm_via_irq { + atomic_t irq_received; + uint32_t pending_mask; + uint32_t enable_mask; + wait_queue_head_t irq_queue; +} drm_via_irq_t; + +typedef struct drm_via_private { + drm_via_sarea_t *sarea_priv; + drm_local_map_t *sarea; + drm_local_map_t *fb; + drm_local_map_t *mmio; + unsigned long agpAddr; + wait_queue_head_t decoder_queue[VIA_NR_XVMC_LOCKS]; + char *dma_ptr; + unsigned int dma_low; + unsigned int dma_high; + unsigned int dma_offset; + uint32_t dma_wrap; + volatile uint32_t *last_pause_ptr; + volatile uint32_t *hw_addr_ptr; + drm_via_ring_buffer_t ring; + struct timeval last_vblank; + int last_vblank_valid; + unsigned usec_per_vblank; + atomic_t vbl_received; + drm_via_state_t hc_state; + char pci_buf[VIA_PCI_BUF_SIZE]; + const uint32_t *fire_offsets[VIA_FIRE_BUF_SIZE]; + uint32_t num_fire_offsets; + int chipset; + drm_via_irq_t via_irqs[VIA_NUM_IRQS]; + unsigned num_irqs; + maskarray_t *irq_masks; + uint32_t irq_enable_mask; + uint32_t irq_pending_mask; + int *irq_map; + unsigned int idle_fault; + struct drm_sman sman; + int vram_initialized; + int agp_initialized; + unsigned long vram_offset; + unsigned long agp_offset; + drm_via_blitq_t blit_queues[VIA_NUM_BLIT_ENGINES]; + uint32_t dma_diff; +} drm_via_private_t; + +enum via_family { + VIA_OTHER = 0, /* Baseline */ + VIA_PRO_GROUP_A, /* Another video engine and DMA commands */ + VIA_DX9_0 /* Same video as pro_group_a, but 3D is unsupported */ +}; + +/* VIA MMIO register access */ +#define VIA_BASE ((dev_priv->mmio)) + +#define VIA_READ(reg) DRM_READ32(VIA_BASE, reg) +#define VIA_WRITE(reg,val) DRM_WRITE32(VIA_BASE, reg, val) +#define VIA_READ8(reg) DRM_READ8(VIA_BASE, reg) +#define VIA_WRITE8(reg,val) DRM_WRITE8(VIA_BASE, reg, val) + +extern struct drm_ioctl_desc via_ioctls[]; +extern int via_max_ioctl; + +extern int via_fb_init(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_mem_alloc(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_mem_free(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_agp_init(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_map_init(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_decoder_futex(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int via_dma_blit_sync( struct drm_device *dev, void *data, struct drm_file *file_priv ); +extern int via_dma_blit( struct drm_device *dev, void *data, struct drm_file *file_priv ); + +extern int via_driver_load(struct drm_device *dev, unsigned long chipset); +extern int via_driver_unload(struct drm_device *dev); + +extern int via_init_context(struct drm_device * dev, int context); +extern int via_final_context(struct drm_device * dev, int context); + +extern int via_do_cleanup_map(struct drm_device * dev); +extern u32 via_get_vblank_counter(struct drm_device *dev, int crtc); +extern int via_enable_vblank(struct drm_device *dev, int crtc); +extern void via_disable_vblank(struct drm_device *dev, int crtc); + +extern irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS); +extern void via_driver_irq_preinstall(struct drm_device * dev); +extern int via_driver_irq_postinstall(struct drm_device *dev); +extern void via_driver_irq_uninstall(struct drm_device * dev); + +extern int via_dma_cleanup(struct drm_device * dev); +extern void via_init_command_verifier(void); +extern int via_driver_dma_quiescent(struct drm_device * dev); +extern void via_init_futex(drm_via_private_t *dev_priv); +extern void via_cleanup_futex(drm_via_private_t *dev_priv); +extern void via_release_futex(drm_via_private_t *dev_priv, int context); + +extern void via_reclaim_buffers_locked(struct drm_device *dev, + struct drm_file *file_priv); +extern void via_lastclose(struct drm_device *dev); + +extern void via_dmablit_handler(struct drm_device *dev, int engine, int from_irq); +extern void via_init_dmablit(struct drm_device *dev); + +#endif diff --git a/sys/dev/drm/via_irq.c b/sys/dev/drm/via_irq.c new file mode 100644 index 0000000..20ec77a --- /dev/null +++ b/sys/dev/drm/via_irq.c @@ -0,0 +1,392 @@ +/* via_irq.c + * + * Copyright 2004 BEAM Ltd. + * Copyright 2002 Tungsten Graphics, Inc. + * Copyright 2005 Thomas Hellstrom. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BEAM LTD, TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Authors: + * Terry Barnaby + * Keith Whitwell + * Thomas Hellstrom + * + * This code provides standard DRM access to the Via Unichrome / Pro Vertical blank + * interrupt, as well as an infrastructure to handle other interrupts of the chip. + * The refresh rate is also calculated for video playback sync purposes. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/drm.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" + +#define VIA_REG_INTERRUPT 0x200 + +/* VIA_REG_INTERRUPT */ +#define VIA_IRQ_GLOBAL (1 << 31) +#define VIA_IRQ_VBLANK_ENABLE (1 << 19) +#define VIA_IRQ_VBLANK_PENDING (1 << 3) +#define VIA_IRQ_HQV0_ENABLE (1 << 11) +#define VIA_IRQ_HQV1_ENABLE (1 << 25) +#define VIA_IRQ_HQV0_PENDING (1 << 9) +#define VIA_IRQ_HQV1_PENDING (1 << 10) +#define VIA_IRQ_DMA0_DD_ENABLE (1 << 20) +#define VIA_IRQ_DMA0_TD_ENABLE (1 << 21) +#define VIA_IRQ_DMA1_DD_ENABLE (1 << 22) +#define VIA_IRQ_DMA1_TD_ENABLE (1 << 23) +#define VIA_IRQ_DMA0_DD_PENDING (1 << 4) +#define VIA_IRQ_DMA0_TD_PENDING (1 << 5) +#define VIA_IRQ_DMA1_DD_PENDING (1 << 6) +#define VIA_IRQ_DMA1_TD_PENDING (1 << 7) + + +/* + * Device-specific IRQs go here. This type might need to be extended with + * the register if there are multiple IRQ control registers. + * Currently we activate the HQV interrupts of Unichrome Pro group A. + */ + +static maskarray_t via_pro_group_a_irqs[] = { + {VIA_IRQ_HQV0_ENABLE, VIA_IRQ_HQV0_PENDING, 0x000003D0, 0x00008010, + 0x00000000 }, + {VIA_IRQ_HQV1_ENABLE, VIA_IRQ_HQV1_PENDING, 0x000013D0, 0x00008010, + 0x00000000 }, + {VIA_IRQ_DMA0_TD_ENABLE, VIA_IRQ_DMA0_TD_PENDING, VIA_PCI_DMA_CSR0, + VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008}, + {VIA_IRQ_DMA1_TD_ENABLE, VIA_IRQ_DMA1_TD_PENDING, VIA_PCI_DMA_CSR1, + VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008}, +}; +static int via_num_pro_group_a = DRM_ARRAY_SIZE(via_pro_group_a_irqs); +static int via_irqmap_pro_group_a[] = {0, 1, -1, 2, -1, 3}; + +static maskarray_t via_unichrome_irqs[] = { + {VIA_IRQ_DMA0_TD_ENABLE, VIA_IRQ_DMA0_TD_PENDING, VIA_PCI_DMA_CSR0, + VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008}, + {VIA_IRQ_DMA1_TD_ENABLE, VIA_IRQ_DMA1_TD_PENDING, VIA_PCI_DMA_CSR1, + VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008} +}; +static int via_num_unichrome = DRM_ARRAY_SIZE(via_unichrome_irqs); +static int via_irqmap_unichrome[] = {-1, -1, -1, 0, -1, 1}; + + +static unsigned time_diff(struct timeval *now, struct timeval *then) +{ + return (now->tv_usec >= then->tv_usec) ? + now->tv_usec - then->tv_usec : + 1000000 - (then->tv_usec - now->tv_usec); +} + +u32 via_get_vblank_counter(struct drm_device *dev, int crtc) +{ + drm_via_private_t *dev_priv = dev->dev_private; + if (crtc != 0) + return 0; + + return atomic_read(&dev_priv->vbl_received); +} + +irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS) +{ + struct drm_device *dev = (struct drm_device *) arg; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + u32 status; + int handled = 0; + struct timeval cur_vblank; + drm_via_irq_t *cur_irq = dev_priv->via_irqs; + int i; + + status = VIA_READ(VIA_REG_INTERRUPT); + if (status & VIA_IRQ_VBLANK_PENDING) { + atomic_inc(&dev_priv->vbl_received); + if (!(atomic_read(&dev_priv->vbl_received) & 0x0F)) { + microtime(&cur_vblank); + if (dev_priv->last_vblank_valid) { + dev_priv->usec_per_vblank = + time_diff(&cur_vblank, + &dev_priv->last_vblank) >> 4; + } + dev_priv->last_vblank = cur_vblank; + dev_priv->last_vblank_valid = 1; + } + if (!(atomic_read(&dev_priv->vbl_received) & 0xFF)) { + DRM_DEBUG("US per vblank is: %u\n", + dev_priv->usec_per_vblank); + } + drm_handle_vblank(dev, 0); + handled = 1; + } + + for (i = 0; i < dev_priv->num_irqs; ++i) { + if (status & cur_irq->pending_mask) { + atomic_inc(&cur_irq->irq_received); + DRM_WAKEUP(&cur_irq->irq_queue); + handled = 1; + if (dev_priv->irq_map[drm_via_irq_dma0_td] == i) { + via_dmablit_handler(dev, 0, 1); + } else if (dev_priv->irq_map[drm_via_irq_dma1_td] == i) { + via_dmablit_handler(dev, 1, 1); + } + } + cur_irq++; + } + + /* Acknowlege interrupts */ + VIA_WRITE(VIA_REG_INTERRUPT, status); + + + if (handled) + return IRQ_HANDLED; + else + return IRQ_NONE; +} + +static __inline__ void viadrv_acknowledge_irqs(drm_via_private_t * dev_priv) +{ + u32 status; + + if (dev_priv) { + /* Acknowlege interrupts */ + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status | + dev_priv->irq_pending_mask); + } +} + +int via_enable_vblank(struct drm_device *dev, int crtc) +{ + drm_via_private_t *dev_priv = dev->dev_private; + u32 status; + + if (crtc != 0) { + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); + return -EINVAL; + } + + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status & VIA_IRQ_VBLANK_ENABLE); + + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) | 0x30); + + return 0; +} + +void via_disable_vblank(struct drm_device *dev, int crtc) +{ + drm_via_private_t *dev_priv = dev->dev_private; + + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) & ~0x30); + + if (crtc != 0) + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); +} + +static int +via_driver_irq_wait(struct drm_device * dev, unsigned int irq, int force_sequence, + unsigned int *sequence) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + unsigned int cur_irq_sequence; + drm_via_irq_t *cur_irq; + int ret = 0; + maskarray_t *masks; + int real_irq; + + DRM_DEBUG("\n"); + + if (!dev_priv) { + DRM_ERROR("called with no initialization\n"); + return -EINVAL; + } + + if (irq >= drm_via_irq_num) { + DRM_ERROR("Trying to wait on unknown irq %d\n", irq); + return -EINVAL; + } + + real_irq = dev_priv->irq_map[irq]; + + if (real_irq < 0) { + DRM_ERROR("Video IRQ %d not available on this hardware.\n", + irq); + return -EINVAL; + } + + masks = dev_priv->irq_masks; + cur_irq = dev_priv->via_irqs + real_irq; + + if (masks[real_irq][2] && !force_sequence) { + DRM_WAIT_ON(ret, cur_irq->irq_queue, 3 * DRM_HZ, + ((VIA_READ(masks[irq][2]) & masks[irq][3]) == + masks[irq][4])); + cur_irq_sequence = atomic_read(&cur_irq->irq_received); + } else { + DRM_WAIT_ON(ret, cur_irq->irq_queue, 3 * DRM_HZ, + (((cur_irq_sequence = + atomic_read(&cur_irq->irq_received)) - + *sequence) <= (1 << 23))); + } + *sequence = cur_irq_sequence; + return ret; +} + + +/* + * drm_dma.h hooks + */ + +void via_driver_irq_preinstall(struct drm_device * dev) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + u32 status; + drm_via_irq_t *cur_irq; + int i; + + DRM_DEBUG("dev_priv: %p\n", dev_priv); + if (dev_priv) { + cur_irq = dev_priv->via_irqs; + + dev_priv->irq_enable_mask = VIA_IRQ_VBLANK_ENABLE; + dev_priv->irq_pending_mask = VIA_IRQ_VBLANK_PENDING; + + if (dev_priv->chipset == VIA_PRO_GROUP_A || + dev_priv->chipset == VIA_DX9_0) { + dev_priv->irq_masks = via_pro_group_a_irqs; + dev_priv->num_irqs = via_num_pro_group_a; + dev_priv->irq_map = via_irqmap_pro_group_a; + } else { + dev_priv->irq_masks = via_unichrome_irqs; + dev_priv->num_irqs = via_num_unichrome; + dev_priv->irq_map = via_irqmap_unichrome; + } + + for (i = 0; i < dev_priv->num_irqs; ++i) { + atomic_set(&cur_irq->irq_received, 0); + cur_irq->enable_mask = dev_priv->irq_masks[i][0]; + cur_irq->pending_mask = dev_priv->irq_masks[i][1]; + DRM_INIT_WAITQUEUE(&cur_irq->irq_queue); + dev_priv->irq_enable_mask |= cur_irq->enable_mask; + dev_priv->irq_pending_mask |= cur_irq->pending_mask; + cur_irq++; + + DRM_DEBUG("Initializing IRQ %d\n", i); + } + + dev_priv->last_vblank_valid = 0; + + /* Clear VSync interrupt regs */ + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status & + ~(dev_priv->irq_enable_mask)); + + /* Clear bits if they're already high */ + viadrv_acknowledge_irqs(dev_priv); + } +} + +int via_driver_irq_postinstall(struct drm_device *dev) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + u32 status; + + DRM_DEBUG("via_driver_irq_postinstall\n"); + if (!dev_priv) + return -EINVAL; + + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status | VIA_IRQ_GLOBAL + | dev_priv->irq_enable_mask); + + /* Some magic, oh for some data sheets ! */ + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) | 0x30); + + return 0; +} + +void via_driver_irq_uninstall(struct drm_device * dev) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + u32 status; + + DRM_DEBUG("\n"); + if (dev_priv) { + + /* Some more magic, oh for some data sheets ! */ + + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) & ~0x30); + + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status & + ~(VIA_IRQ_VBLANK_ENABLE | dev_priv->irq_enable_mask)); + } +} + +int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_irqwait_t *irqwait = data; + struct timeval now; + int ret = 0; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + drm_via_irq_t *cur_irq = dev_priv->via_irqs; + int force_sequence; + + if (irqwait->request.irq >= dev_priv->num_irqs) { + DRM_ERROR("Trying to wait on unknown irq %d\n", + irqwait->request.irq); + return -EINVAL; + } + + cur_irq += irqwait->request.irq; + + switch (irqwait->request.type & ~VIA_IRQ_FLAGS_MASK) { + case VIA_IRQ_RELATIVE: + irqwait->request.sequence += + atomic_read(&cur_irq->irq_received); + irqwait->request.type &= ~_DRM_VBLANK_RELATIVE; + case VIA_IRQ_ABSOLUTE: + break; + default: + return -EINVAL; + } + + if (irqwait->request.type & VIA_IRQ_SIGNAL) { + DRM_ERROR("Signals on Via IRQs not implemented yet.\n"); + return -EINVAL; + } + + force_sequence = (irqwait->request.type & VIA_IRQ_FORCE_SEQUENCE); + + ret = via_driver_irq_wait(dev, irqwait->request.irq, force_sequence, + &irqwait->request.sequence); + microtime(&now); + irqwait->reply.tval_sec = now.tv_sec; + irqwait->reply.tval_usec = now.tv_usec; + + return ret; +} diff --git a/sys/dev/drm/via_map.c b/sys/dev/drm/via_map.c new file mode 100644 index 0000000..687b8be --- /dev/null +++ b/sys/dev/drm/via_map.c @@ -0,0 +1,137 @@ +/*- + * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. + * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" + +static int via_do_init_map(struct drm_device * dev, drm_via_init_t * init) +{ + drm_via_private_t *dev_priv = dev->dev_private; + + DRM_DEBUG("\n"); + + dev_priv->sarea = drm_getsarea(dev); + if (!dev_priv->sarea) { + DRM_ERROR("could not find sarea!\n"); + dev->dev_private = (void *)dev_priv; + via_do_cleanup_map(dev); + return -EINVAL; + } + + dev_priv->fb = drm_core_findmap(dev, init->fb_offset); + if (!dev_priv->fb) { + DRM_ERROR("could not find framebuffer!\n"); + dev->dev_private = (void *)dev_priv; + via_do_cleanup_map(dev); + return -EINVAL; + } + dev_priv->mmio = drm_core_findmap(dev, init->mmio_offset); + if (!dev_priv->mmio) { + DRM_ERROR("could not find mmio region!\n"); + dev->dev_private = (void *)dev_priv; + via_do_cleanup_map(dev); + return -EINVAL; + } + + dev_priv->sarea_priv = + (drm_via_sarea_t *) ((u8 *) dev_priv->sarea->virtual + + init->sarea_priv_offset); + + dev_priv->agpAddr = init->agpAddr; + + via_init_futex(dev_priv); + + via_init_dmablit(dev); + + dev->dev_private = (void *)dev_priv; + return 0; +} + +int via_do_cleanup_map(struct drm_device * dev) +{ + via_dma_cleanup(dev); + + return 0; +} + +int via_map_init(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_init_t *init = data; + + DRM_DEBUG("\n"); + + switch (init->func) { + case VIA_INIT_MAP: + return via_do_init_map(dev, init); + case VIA_CLEANUP_MAP: + return via_do_cleanup_map(dev); + } + + return -EINVAL; +} + +int via_driver_load(struct drm_device *dev, unsigned long chipset) +{ + drm_via_private_t *dev_priv; + int ret = 0; + + dev_priv = drm_alloc(sizeof(drm_via_private_t), DRM_MEM_DRIVER); + if (dev_priv == NULL) + return -ENOMEM; + + dev->dev_private = (void *)dev_priv; + + dev_priv->chipset = chipset; + + ret = drm_sman_init(&dev_priv->sman, 2, 12, 8); + if (ret) { + drm_free(dev_priv, sizeof(drm_via_private_t), DRM_MEM_DRIVER); + return ret; + } + + ret = drm_vblank_init(dev, 1); + if (ret) { + drm_sman_takedown(&dev_priv->sman); + drm_free(dev_priv, sizeof(drm_via_private_t), DRM_MEM_DRIVER); + return ret; + } + + return 0; +} + +int via_driver_unload(struct drm_device *dev) +{ + drm_via_private_t *dev_priv = dev->dev_private; + + drm_sman_takedown(&dev_priv->sman); + + drm_free(dev_priv, sizeof(drm_via_private_t), DRM_MEM_DRIVER); + dev->dev_private = NULL; + + return 0; +} diff --git a/sys/dev/drm/via_mm.c b/sys/dev/drm/via_mm.c new file mode 100644 index 0000000..6dc185b --- /dev/null +++ b/sys/dev/drm/via_mm.c @@ -0,0 +1,180 @@ +/*- + * Copyright 2006 Tungsten Graphics Inc., Bismarck, ND., USA. + * All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +/* + * Authors: Thomas Hellström + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" +#include "dev/drm/drm_sman.h" + +#define VIA_MM_ALIGN_SHIFT 4 +#define VIA_MM_ALIGN_MASK ( (1 << VIA_MM_ALIGN_SHIFT) - 1) + +int via_agp_init(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_agp_t *agp = data; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + int ret; + + ret = drm_sman_set_range(&dev_priv->sman, VIA_MEM_AGP, 0, + agp->size >> VIA_MM_ALIGN_SHIFT); + if (ret) { + DRM_ERROR("AGP memory manager initialisation error\n"); + return ret; + } + + dev_priv->agp_initialized = 1; + dev_priv->agp_offset = agp->offset; + + DRM_DEBUG("offset = %u, size = %u\n", agp->offset, agp->size); + return 0; +} + +int via_fb_init(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_fb_t *fb = data; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + int ret; + + ret = drm_sman_set_range(&dev_priv->sman, VIA_MEM_VIDEO, 0, + fb->size >> VIA_MM_ALIGN_SHIFT); + if (ret) { + DRM_ERROR("VRAM memory manager initialisation error\n"); + return ret; + } + + dev_priv->vram_initialized = 1; + dev_priv->vram_offset = fb->offset; + + DRM_DEBUG("offset = %u, size = %u\n", fb->offset, fb->size); + + return 0; + +} + +int via_final_context(struct drm_device *dev, int context) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + + via_release_futex(dev_priv, context); + +#ifdef __linux__ + /* Linux specific until context tracking code gets ported to BSD */ + /* Last context, perform cleanup */ + if (dev->ctx_count == 1 && dev->dev_private) { + DRM_DEBUG("Last Context\n"); + drm_irq_uninstall(dev); + via_cleanup_futex(dev_priv); + via_do_cleanup_map(dev); + } +#endif + return 1; +} + +void via_lastclose(struct drm_device *dev) +{ + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + + if (!dev_priv) + return; + + drm_sman_cleanup(&dev_priv->sman); + dev_priv->vram_initialized = 0; + dev_priv->agp_initialized = 0; +} + +int via_mem_alloc(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + drm_via_mem_t *mem = data; + int retval = 0; + struct drm_memblock_item *item; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + unsigned long tmpSize; + + if (mem->type > VIA_MEM_AGP) { + DRM_ERROR("Unknown memory type allocation\n"); + return -EINVAL; + } + if (0 == ((mem->type == VIA_MEM_VIDEO) ? dev_priv->vram_initialized : + dev_priv->agp_initialized)) { + DRM_ERROR + ("Attempt to allocate from uninitialized memory manager.\n"); + return -EINVAL; + } + + tmpSize = (mem->size + VIA_MM_ALIGN_MASK) >> VIA_MM_ALIGN_SHIFT; + item = drm_sman_alloc(&dev_priv->sman, mem->type, tmpSize, 0, + (unsigned long)file_priv); + if (item) { + mem->offset = ((mem->type == VIA_MEM_VIDEO) ? + dev_priv->vram_offset : dev_priv->agp_offset) + + (item->mm-> + offset(item->mm, item->mm_info) << VIA_MM_ALIGN_SHIFT); + mem->index = item->user_hash.key; + } else { + mem->offset = 0; + mem->size = 0; + mem->index = 0; + DRM_DEBUG("Video memory allocation failed\n"); + retval = -ENOMEM; + } + + return retval; +} + +int via_mem_free(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_private_t *dev_priv = dev->dev_private; + drm_via_mem_t *mem = data; + int ret; + + ret = drm_sman_free_key(&dev_priv->sman, mem->index); + DRM_DEBUG("free = 0x%lx\n", mem->index); + + return ret; +} + + +void via_reclaim_buffers_locked(struct drm_device * dev, + struct drm_file *file_priv) +{ + drm_via_private_t *dev_priv = dev->dev_private; + + if (drm_sman_owner_clean(&dev_priv->sman, (unsigned long)file_priv)) + return; + + if (dev->driver->dma_quiescent) + dev->driver->dma_quiescent(dev); + + drm_sman_owner_cleanup(&dev_priv->sman, (unsigned long)file_priv); + + return; +} diff --git a/sys/dev/drm/via_verifier.c b/sys/dev/drm/via_verifier.c new file mode 100644 index 0000000..7169e64 --- /dev/null +++ b/sys/dev/drm/via_verifier.c @@ -0,0 +1,1115 @@ +/*- + * Copyright 2004 The Unichrome Project. All Rights Reserved. + * Copyright 2005 Thomas Hellstrom. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S), AND/OR THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Author: Thomas Hellstrom 2004, 2005. + * This code was written using docs obtained under NDA from VIA Inc. + * + * Don't run this code directly on an AGP buffer. Due to cache problems it will + * be very slow. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/via_3d_reg.h" +#include "dev/drm/drmP.h" +#include "dev/drm/drm.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_verifier.h" +#include "dev/drm/via_drv.h" + +typedef enum { + state_command, + state_header2, + state_header1, + state_vheader5, + state_vheader6, + state_error +} verifier_state_t; + +typedef enum { + no_check = 0, + check_for_header2, + check_for_header1, + check_for_header2_err, + check_for_header1_err, + check_for_fire, + check_z_buffer_addr0, + check_z_buffer_addr1, + check_z_buffer_addr_mode, + check_destination_addr0, + check_destination_addr1, + check_destination_addr_mode, + check_for_dummy, + check_for_dd, + check_texture_addr0, + check_texture_addr1, + check_texture_addr2, + check_texture_addr3, + check_texture_addr4, + check_texture_addr5, + check_texture_addr6, + check_texture_addr7, + check_texture_addr8, + check_texture_addr_mode, + check_for_vertex_count, + check_number_texunits, + forbidden_command +} hazard_t; + +/* + * Associates each hazard above with a possible multi-command + * sequence. For example an address that is split over multiple + * commands and that needs to be checked at the first command + * that does not include any part of the address. + */ + +static drm_via_sequence_t seqs[] = { + no_sequence, + no_sequence, + no_sequence, + no_sequence, + no_sequence, + no_sequence, + z_address, + z_address, + z_address, + dest_address, + dest_address, + dest_address, + no_sequence, + no_sequence, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + tex_address, + no_sequence +}; + +typedef struct { + unsigned int code; + hazard_t hz; +} hz_init_t; + +static hz_init_t init_table1[] = { + {0xf2, check_for_header2_err}, + {0xf0, check_for_header1_err}, + {0xee, check_for_fire}, + {0xcc, check_for_dummy}, + {0xdd, check_for_dd}, + {0x00, no_check}, + {0x10, check_z_buffer_addr0}, + {0x11, check_z_buffer_addr1}, + {0x12, check_z_buffer_addr_mode}, + {0x13, no_check}, + {0x14, no_check}, + {0x15, no_check}, + {0x23, no_check}, + {0x24, no_check}, + {0x33, no_check}, + {0x34, no_check}, + {0x35, no_check}, + {0x36, no_check}, + {0x37, no_check}, + {0x38, no_check}, + {0x39, no_check}, + {0x3A, no_check}, + {0x3B, no_check}, + {0x3C, no_check}, + {0x3D, no_check}, + {0x3E, no_check}, + {0x40, check_destination_addr0}, + {0x41, check_destination_addr1}, + {0x42, check_destination_addr_mode}, + {0x43, no_check}, + {0x44, no_check}, + {0x50, no_check}, + {0x51, no_check}, + {0x52, no_check}, + {0x53, no_check}, + {0x54, no_check}, + {0x55, no_check}, + {0x56, no_check}, + {0x57, no_check}, + {0x58, no_check}, + {0x70, no_check}, + {0x71, no_check}, + {0x78, no_check}, + {0x79, no_check}, + {0x7A, no_check}, + {0x7B, no_check}, + {0x7C, no_check}, + {0x7D, check_for_vertex_count} +}; + +static hz_init_t init_table2[] = { + {0xf2, check_for_header2_err}, + {0xf0, check_for_header1_err}, + {0xee, check_for_fire}, + {0xcc, check_for_dummy}, + {0x00, check_texture_addr0}, + {0x01, check_texture_addr0}, + {0x02, check_texture_addr0}, + {0x03, check_texture_addr0}, + {0x04, check_texture_addr0}, + {0x05, check_texture_addr0}, + {0x06, check_texture_addr0}, + {0x07, check_texture_addr0}, + {0x08, check_texture_addr0}, + {0x09, check_texture_addr0}, + {0x20, check_texture_addr1}, + {0x21, check_texture_addr1}, + {0x22, check_texture_addr1}, + {0x23, check_texture_addr4}, + {0x2B, check_texture_addr3}, + {0x2C, check_texture_addr3}, + {0x2D, check_texture_addr3}, + {0x2E, check_texture_addr3}, + {0x2F, check_texture_addr3}, + {0x30, check_texture_addr3}, + {0x31, check_texture_addr3}, + {0x32, check_texture_addr3}, + {0x33, check_texture_addr3}, + {0x34, check_texture_addr3}, + {0x4B, check_texture_addr5}, + {0x4C, check_texture_addr6}, + {0x51, check_texture_addr7}, + {0x52, check_texture_addr8}, + {0x77, check_texture_addr2}, + {0x78, no_check}, + {0x79, no_check}, + {0x7A, no_check}, + {0x7B, check_texture_addr_mode}, + {0x7C, no_check}, + {0x7D, no_check}, + {0x7E, no_check}, + {0x7F, no_check}, + {0x80, no_check}, + {0x81, no_check}, + {0x82, no_check}, + {0x83, no_check}, + {0x85, no_check}, + {0x86, no_check}, + {0x87, no_check}, + {0x88, no_check}, + {0x89, no_check}, + {0x8A, no_check}, + {0x90, no_check}, + {0x91, no_check}, + {0x92, no_check}, + {0x93, no_check} +}; + +static hz_init_t init_table3[] = { + {0xf2, check_for_header2_err}, + {0xf0, check_for_header1_err}, + {0xcc, check_for_dummy}, + {0x00, check_number_texunits} +}; + +static hazard_t table1[256]; +static hazard_t table2[256]; +static hazard_t table3[256]; + +static __inline__ int +eat_words(const uint32_t ** buf, const uint32_t * buf_end, unsigned num_words) +{ + if ((buf_end - *buf) >= num_words) { + *buf += num_words; + return 0; + } + DRM_ERROR("Illegal termination of DMA command buffer\n"); + return 1; +} + +/* + * Partially stolen from drm_memory.h + */ + +static __inline__ drm_local_map_t *via_drm_lookup_agp_map(drm_via_state_t *seq, + unsigned long offset, + unsigned long size, + struct drm_device * dev) +{ + drm_local_map_t *map = seq->map_cache; + + if (map && map->offset <= offset + && (offset + size) <= (map->offset + map->size)) { + return map; + } + + TAILQ_FOREACH(map, &dev->maplist, link) { + if (map->offset <= offset + && (offset + size) <= (map->offset + map->size) + && !(map->flags & _DRM_RESTRICTED) + && (map->type == _DRM_AGP)) { + seq->map_cache = map; + return map; + } + } + return NULL; +} + +/* + * Require that all AGP texture levels reside in the same AGP map which should + * be mappable by the client. This is not a big restriction. + * FIXME: To actually enforce this security policy strictly, drm_rmmap + * would have to wait for dma quiescent before removing an AGP map. + * The via_drm_lookup_agp_map call in reality seems to take + * very little CPU time. + */ + +static __inline__ int finish_current_sequence(drm_via_state_t * cur_seq) +{ + switch (cur_seq->unfinished) { + case z_address: + DRM_DEBUG("Z Buffer start address is 0x%x\n", cur_seq->z_addr); + break; + case dest_address: + DRM_DEBUG("Destination start address is 0x%x\n", + cur_seq->d_addr); + break; + case tex_address: + if (cur_seq->agp_texture) { + unsigned start = + cur_seq->tex_level_lo[cur_seq->texture]; + unsigned end = cur_seq->tex_level_hi[cur_seq->texture]; + unsigned long lo = ~0, hi = 0, tmp; + uint32_t *addr, *pitch, *height, tex; + unsigned i; + int npot; + + if (end > 9) + end = 9; + if (start > 9) + start = 9; + + addr = + &(cur_seq->t_addr[tex = cur_seq->texture][start]); + pitch = &(cur_seq->pitch[tex][start]); + height = &(cur_seq->height[tex][start]); + npot = cur_seq->tex_npot[tex]; + for (i = start; i <= end; ++i) { + tmp = *addr++; + if (tmp < lo) + lo = tmp; + if (i == 0 && npot) + tmp += (*height++ * *pitch++); + else + tmp += (*height++ << *pitch++); + if (tmp > hi) + hi = tmp; + } + + if (!via_drm_lookup_agp_map + (cur_seq, lo, hi - lo, cur_seq->dev)) { + DRM_ERROR + ("AGP texture is not in allowed map\n"); + return 2; + } + } + break; + default: + break; + } + cur_seq->unfinished = no_sequence; + return 0; +} + +static __inline__ int +investigate_hazard(uint32_t cmd, hazard_t hz, drm_via_state_t * cur_seq) +{ + register uint32_t tmp, *tmp_addr; + + if (cur_seq->unfinished && (cur_seq->unfinished != seqs[hz])) { + int ret; + if ((ret = finish_current_sequence(cur_seq))) + return ret; + } + + switch (hz) { + case check_for_header2: + if (cmd == HALCYON_HEADER2) + return 1; + return 0; + case check_for_header1: + if ((cmd & HALCYON_HEADER1MASK) == HALCYON_HEADER1) + return 1; + return 0; + case check_for_header2_err: + if (cmd == HALCYON_HEADER2) + return 1; + DRM_ERROR("Illegal DMA HALCYON_HEADER2 command\n"); + break; + case check_for_header1_err: + if ((cmd & HALCYON_HEADER1MASK) == HALCYON_HEADER1) + return 1; + DRM_ERROR("Illegal DMA HALCYON_HEADER1 command\n"); + break; + case check_for_fire: + if ((cmd & HALCYON_FIREMASK) == HALCYON_FIRECMD) + return 1; + DRM_ERROR("Illegal DMA HALCYON_FIRECMD command\n"); + break; + case check_for_dummy: + if (HC_DUMMY == cmd) + return 0; + DRM_ERROR("Illegal DMA HC_DUMMY command\n"); + break; + case check_for_dd: + if (0xdddddddd == cmd) + return 0; + DRM_ERROR("Illegal DMA 0xdddddddd command\n"); + break; + case check_z_buffer_addr0: + cur_seq->unfinished = z_address; + cur_seq->z_addr = (cur_seq->z_addr & 0xFF000000) | + (cmd & 0x00FFFFFF); + return 0; + case check_z_buffer_addr1: + cur_seq->unfinished = z_address; + cur_seq->z_addr = (cur_seq->z_addr & 0x00FFFFFF) | + ((cmd & 0xFF) << 24); + return 0; + case check_z_buffer_addr_mode: + cur_seq->unfinished = z_address; + if ((cmd & 0x0000C000) == 0) + return 0; + DRM_ERROR("Attempt to place Z buffer in system memory\n"); + return 2; + case check_destination_addr0: + cur_seq->unfinished = dest_address; + cur_seq->d_addr = (cur_seq->d_addr & 0xFF000000) | + (cmd & 0x00FFFFFF); + return 0; + case check_destination_addr1: + cur_seq->unfinished = dest_address; + cur_seq->d_addr = (cur_seq->d_addr & 0x00FFFFFF) | + ((cmd & 0xFF) << 24); + return 0; + case check_destination_addr_mode: + cur_seq->unfinished = dest_address; + if ((cmd & 0x0000C000) == 0) + return 0; + DRM_ERROR + ("Attempt to place 3D drawing buffer in system memory\n"); + return 2; + case check_texture_addr0: + cur_seq->unfinished = tex_address; + tmp = (cmd >> 24); + tmp_addr = &cur_seq->t_addr[cur_seq->texture][tmp]; + *tmp_addr = (*tmp_addr & 0xFF000000) | (cmd & 0x00FFFFFF); + return 0; + case check_texture_addr1: + cur_seq->unfinished = tex_address; + tmp = ((cmd >> 24) - 0x20); + tmp += tmp << 1; + tmp_addr = &cur_seq->t_addr[cur_seq->texture][tmp]; + *tmp_addr = (*tmp_addr & 0x00FFFFFF) | ((cmd & 0xFF) << 24); + tmp_addr++; + *tmp_addr = (*tmp_addr & 0x00FFFFFF) | ((cmd & 0xFF00) << 16); + tmp_addr++; + *tmp_addr = (*tmp_addr & 0x00FFFFFF) | ((cmd & 0xFF0000) << 8); + return 0; + case check_texture_addr2: + cur_seq->unfinished = tex_address; + cur_seq->tex_level_lo[tmp = cur_seq->texture] = cmd & 0x3F; + cur_seq->tex_level_hi[tmp] = (cmd & 0xFC0) >> 6; + return 0; + case check_texture_addr3: + cur_seq->unfinished = tex_address; + tmp = ((cmd >> 24) - HC_SubA_HTXnL0Pit); + if (tmp == 0 && + (cmd & HC_HTXnEnPit_MASK)) { + cur_seq->pitch[cur_seq->texture][tmp] = + (cmd & HC_HTXnLnPit_MASK); + cur_seq->tex_npot[cur_seq->texture] = 1; + } else { + cur_seq->pitch[cur_seq->texture][tmp] = + (cmd & HC_HTXnLnPitE_MASK) >> HC_HTXnLnPitE_SHIFT; + cur_seq->tex_npot[cur_seq->texture] = 0; + if (cmd & 0x000FFFFF) { + DRM_ERROR + ("Unimplemented texture level 0 pitch mode.\n"); + return 2; + } + } + return 0; + case check_texture_addr4: + cur_seq->unfinished = tex_address; + tmp_addr = &cur_seq->t_addr[cur_seq->texture][9]; + *tmp_addr = (*tmp_addr & 0x00FFFFFF) | ((cmd & 0xFF) << 24); + return 0; + case check_texture_addr5: + case check_texture_addr6: + cur_seq->unfinished = tex_address; + /* + * Texture width. We don't care since we have the pitch. + */ + return 0; + case check_texture_addr7: + cur_seq->unfinished = tex_address; + tmp_addr = &(cur_seq->height[cur_seq->texture][0]); + tmp_addr[5] = 1 << ((cmd & 0x00F00000) >> 20); + tmp_addr[4] = 1 << ((cmd & 0x000F0000) >> 16); + tmp_addr[3] = 1 << ((cmd & 0x0000F000) >> 12); + tmp_addr[2] = 1 << ((cmd & 0x00000F00) >> 8); + tmp_addr[1] = 1 << ((cmd & 0x000000F0) >> 4); + tmp_addr[0] = 1 << (cmd & 0x0000000F); + return 0; + case check_texture_addr8: + cur_seq->unfinished = tex_address; + tmp_addr = &(cur_seq->height[cur_seq->texture][0]); + tmp_addr[9] = 1 << ((cmd & 0x0000F000) >> 12); + tmp_addr[8] = 1 << ((cmd & 0x00000F00) >> 8); + tmp_addr[7] = 1 << ((cmd & 0x000000F0) >> 4); + tmp_addr[6] = 1 << (cmd & 0x0000000F); + return 0; + case check_texture_addr_mode: + cur_seq->unfinished = tex_address; + if (2 == (tmp = cmd & 0x00000003)) { + DRM_ERROR + ("Attempt to fetch texture from system memory.\n"); + return 2; + } + cur_seq->agp_texture = (tmp == 3); + cur_seq->tex_palette_size[cur_seq->texture] = + (cmd >> 16) & 0x000000007; + return 0; + case check_for_vertex_count: + cur_seq->vertex_count = cmd & 0x0000FFFF; + return 0; + case check_number_texunits: + cur_seq->multitex = (cmd >> 3) & 1; + return 0; + default: + DRM_ERROR("Illegal DMA data: 0x%x\n", cmd); + return 2; + } + return 2; +} + +static __inline__ int +via_check_prim_list(uint32_t const **buffer, const uint32_t * buf_end, + drm_via_state_t * cur_seq) +{ + drm_via_private_t *dev_priv = + (drm_via_private_t *) cur_seq->dev->dev_private; + uint32_t a_fire, bcmd, dw_count; + int ret = 0; + int have_fire; + const uint32_t *buf = *buffer; + + while (buf < buf_end) { + have_fire = 0; + if ((buf_end - buf) < 2) { + DRM_ERROR + ("Unexpected termination of primitive list.\n"); + ret = 1; + break; + } + if ((*buf & HC_ACMD_MASK) != HC_ACMD_HCmdB) + break; + bcmd = *buf++; + if ((*buf & HC_ACMD_MASK) != HC_ACMD_HCmdA) { + DRM_ERROR("Expected Vertex List A command, got 0x%x\n", + *buf); + ret = 1; + break; + } + a_fire = + *buf++ | HC_HPLEND_MASK | HC_HPMValidN_MASK | + HC_HE3Fire_MASK; + + /* + * How many dwords per vertex ? + */ + + if (cur_seq->agp && ((bcmd & (0xF << 11)) == 0)) { + DRM_ERROR("Illegal B command vertex data for AGP.\n"); + ret = 1; + break; + } + + dw_count = 0; + if (bcmd & (1 << 7)) + dw_count += (cur_seq->multitex) ? 2 : 1; + if (bcmd & (1 << 8)) + dw_count += (cur_seq->multitex) ? 2 : 1; + if (bcmd & (1 << 9)) + dw_count++; + if (bcmd & (1 << 10)) + dw_count++; + if (bcmd & (1 << 11)) + dw_count++; + if (bcmd & (1 << 12)) + dw_count++; + if (bcmd & (1 << 13)) + dw_count++; + if (bcmd & (1 << 14)) + dw_count++; + + while (buf < buf_end) { + if (*buf == a_fire) { + if (dev_priv->num_fire_offsets >= + VIA_FIRE_BUF_SIZE) { + DRM_ERROR("Fire offset buffer full.\n"); + ret = 1; + break; + } + dev_priv->fire_offsets[dev_priv-> + num_fire_offsets++] = + buf; + have_fire = 1; + buf++; + if (buf < buf_end && *buf == a_fire) + buf++; + break; + } + if ((*buf == HALCYON_HEADER2) || + ((*buf & HALCYON_FIREMASK) == HALCYON_FIRECMD)) { + DRM_ERROR("Missing Vertex Fire command, " + "Stray Vertex Fire command or verifier " + "lost sync.\n"); + ret = 1; + break; + } + if ((ret = eat_words(&buf, buf_end, dw_count))) + break; + } + if (buf >= buf_end && !have_fire) { + DRM_ERROR("Missing Vertex Fire command or verifier " + "lost sync.\n"); + ret = 1; + break; + } + if (cur_seq->agp && ((buf - cur_seq->buf_start) & 0x01)) { + DRM_ERROR("AGP Primitive list end misaligned.\n"); + ret = 1; + break; + } + } + *buffer = buf; + return ret; +} + +static __inline__ verifier_state_t +via_check_header2(uint32_t const **buffer, const uint32_t * buf_end, + drm_via_state_t * hc_state) +{ + uint32_t cmd; + int hz_mode; + hazard_t hz; + const uint32_t *buf = *buffer; + const hazard_t *hz_table; + + if ((buf_end - buf) < 2) { + DRM_ERROR + ("Illegal termination of DMA HALCYON_HEADER2 sequence.\n"); + return state_error; + } + buf++; + cmd = (*buf++ & 0xFFFF0000) >> 16; + + switch (cmd) { + case HC_ParaType_CmdVdata: + if (via_check_prim_list(&buf, buf_end, hc_state)) + return state_error; + *buffer = buf; + return state_command; + case HC_ParaType_NotTex: + hz_table = table1; + break; + case HC_ParaType_Tex: + hc_state->texture = 0; + hz_table = table2; + break; + case (HC_ParaType_Tex | (HC_SubType_Tex1 << 8)): + hc_state->texture = 1; + hz_table = table2; + break; + case (HC_ParaType_Tex | (HC_SubType_TexGeneral << 8)): + hz_table = table3; + break; + case HC_ParaType_Auto: + if (eat_words(&buf, buf_end, 2)) + return state_error; + *buffer = buf; + return state_command; + case (HC_ParaType_Palette | (HC_SubType_Stipple << 8)): + if (eat_words(&buf, buf_end, 32)) + return state_error; + *buffer = buf; + return state_command; + case (HC_ParaType_Palette | (HC_SubType_TexPalette0 << 8)): + case (HC_ParaType_Palette | (HC_SubType_TexPalette1 << 8)): + DRM_ERROR("Texture palettes are rejected because of " + "lack of info how to determine their size.\n"); + return state_error; + case (HC_ParaType_Palette | (HC_SubType_FogTable << 8)): + DRM_ERROR("Fog factor palettes are rejected because of " + "lack of info how to determine their size.\n"); + return state_error; + default: + + /* + * There are some unimplemented HC_ParaTypes here, that + * need to be implemented if the Mesa driver is extended. + */ + + DRM_ERROR("Invalid or unimplemented HALCYON_HEADER2 " + "DMA subcommand: 0x%x. Previous dword: 0x%x\n", + cmd, *(buf - 2)); + *buffer = buf; + return state_error; + } + + while (buf < buf_end) { + cmd = *buf++; + if ((hz = hz_table[cmd >> 24])) { + if ((hz_mode = investigate_hazard(cmd, hz, hc_state))) { + if (hz_mode == 1) { + buf--; + break; + } + return state_error; + } + } else if (hc_state->unfinished && + finish_current_sequence(hc_state)) { + return state_error; + } + } + if (hc_state->unfinished && finish_current_sequence(hc_state)) { + return state_error; + } + *buffer = buf; + return state_command; +} + +static __inline__ verifier_state_t +via_parse_header2(drm_via_private_t * dev_priv, uint32_t const **buffer, + const uint32_t * buf_end, int *fire_count) +{ + uint32_t cmd; + const uint32_t *buf = *buffer; + const uint32_t *next_fire; + int burst = 0; + + next_fire = dev_priv->fire_offsets[*fire_count]; + buf++; + cmd = (*buf & 0xFFFF0000) >> 16; + VIA_WRITE(HC_REG_TRANS_SET + HC_REG_BASE, *buf++); + switch (cmd) { + case HC_ParaType_CmdVdata: + while ((buf < buf_end) && + (*fire_count < dev_priv->num_fire_offsets) && + (*buf & HC_ACMD_MASK) == HC_ACMD_HCmdB) { + while (buf <= next_fire) { + VIA_WRITE(HC_REG_TRANS_SPACE + HC_REG_BASE + + (burst & 63), *buf++); + burst += 4; + } + if ((buf < buf_end) + && ((*buf & HALCYON_FIREMASK) == HALCYON_FIRECMD)) + buf++; + + if (++(*fire_count) < dev_priv->num_fire_offsets) + next_fire = dev_priv->fire_offsets[*fire_count]; + } + break; + default: + while (buf < buf_end) { + + if (*buf == HC_HEADER2 || + (*buf & HALCYON_HEADER1MASK) == HALCYON_HEADER1 || + (*buf & VIA_VIDEOMASK) == VIA_VIDEO_HEADER5 || + (*buf & VIA_VIDEOMASK) == VIA_VIDEO_HEADER6) + break; + + VIA_WRITE(HC_REG_TRANS_SPACE + HC_REG_BASE + + (burst & 63), *buf++); + burst += 4; + } + } + *buffer = buf; + return state_command; +} + +static __inline__ int verify_mmio_address(uint32_t address) +{ + if ((address > 0x3FF) && (address < 0xC00)) { + DRM_ERROR("Invalid VIDEO DMA command. " + "Attempt to access 3D- or command burst area.\n"); + return 1; + } else if ((address > 0xCFF) && (address < 0x1300)) { + DRM_ERROR("Invalid VIDEO DMA command. " + "Attempt to access PCI DMA area.\n"); + return 1; + } else if (address > 0x13FF) { + DRM_ERROR("Invalid VIDEO DMA command. " + "Attempt to access VGA registers.\n"); + return 1; + } + return 0; +} + +static __inline__ int +verify_video_tail(uint32_t const **buffer, const uint32_t * buf_end, + uint32_t dwords) +{ + const uint32_t *buf = *buffer; + + if (buf_end - buf < dwords) { + DRM_ERROR("Illegal termination of video command.\n"); + return 1; + } + while (dwords--) { + if (*buf++) { + DRM_ERROR("Illegal video command tail.\n"); + return 1; + } + } + *buffer = buf; + return 0; +} + +static __inline__ verifier_state_t +via_check_header1(uint32_t const **buffer, const uint32_t * buf_end) +{ + uint32_t cmd; + const uint32_t *buf = *buffer; + verifier_state_t ret = state_command; + + while (buf < buf_end) { + cmd = *buf; + if ((cmd > ((0x3FF >> 2) | HALCYON_HEADER1)) && + (cmd < ((0xC00 >> 2) | HALCYON_HEADER1))) { + if ((cmd & HALCYON_HEADER1MASK) != HALCYON_HEADER1) + break; + DRM_ERROR("Invalid HALCYON_HEADER1 command. " + "Attempt to access 3D- or command burst area.\n"); + ret = state_error; + break; + } else if (cmd > ((0xCFF >> 2) | HALCYON_HEADER1)) { + if ((cmd & HALCYON_HEADER1MASK) != HALCYON_HEADER1) + break; + DRM_ERROR("Invalid HALCYON_HEADER1 command. " + "Attempt to access VGA registers.\n"); + ret = state_error; + break; + } else { + buf += 2; + } + } + *buffer = buf; + return ret; +} + +static __inline__ verifier_state_t +via_parse_header1(drm_via_private_t * dev_priv, uint32_t const **buffer, + const uint32_t * buf_end) +{ + register uint32_t cmd; + const uint32_t *buf = *buffer; + + while (buf < buf_end) { + cmd = *buf; + if ((cmd & HALCYON_HEADER1MASK) != HALCYON_HEADER1) + break; + VIA_WRITE((cmd & ~HALCYON_HEADER1MASK) << 2, *++buf); + buf++; + } + *buffer = buf; + return state_command; +} + +static __inline__ verifier_state_t +via_check_vheader5(uint32_t const **buffer, const uint32_t * buf_end) +{ + uint32_t data; + const uint32_t *buf = *buffer; + + if (buf_end - buf < 4) { + DRM_ERROR("Illegal termination of video header5 command\n"); + return state_error; + } + + data = *buf++ & ~VIA_VIDEOMASK; + if (verify_mmio_address(data)) + return state_error; + + data = *buf++; + if (*buf++ != 0x00F50000) { + DRM_ERROR("Illegal header5 header data\n"); + return state_error; + } + if (*buf++ != 0x00000000) { + DRM_ERROR("Illegal header5 header data\n"); + return state_error; + } + if (eat_words(&buf, buf_end, data)) + return state_error; + if ((data & 3) && verify_video_tail(&buf, buf_end, 4 - (data & 3))) + return state_error; + *buffer = buf; + return state_command; + +} + +static __inline__ verifier_state_t +via_parse_vheader5(drm_via_private_t * dev_priv, uint32_t const **buffer, + const uint32_t * buf_end) +{ + uint32_t addr, count, i; + const uint32_t *buf = *buffer; + + addr = *buf++ & ~VIA_VIDEOMASK; + i = count = *buf; + buf += 3; + while (i--) { + VIA_WRITE(addr, *buf++); + } + if (count & 3) + buf += 4 - (count & 3); + *buffer = buf; + return state_command; +} + +static __inline__ verifier_state_t +via_check_vheader6(uint32_t const **buffer, const uint32_t * buf_end) +{ + uint32_t data; + const uint32_t *buf = *buffer; + uint32_t i; + + if (buf_end - buf < 4) { + DRM_ERROR("Illegal termination of video header6 command\n"); + return state_error; + } + buf++; + data = *buf++; + if (*buf++ != 0x00F60000) { + DRM_ERROR("Illegal header6 header data\n"); + return state_error; + } + if (*buf++ != 0x00000000) { + DRM_ERROR("Illegal header6 header data\n"); + return state_error; + } + if ((buf_end - buf) < (data << 1)) { + DRM_ERROR("Illegal termination of video header6 command\n"); + return state_error; + } + for (i = 0; i < data; ++i) { + if (verify_mmio_address(*buf++)) + return state_error; + buf++; + } + data <<= 1; + if ((data & 3) && verify_video_tail(&buf, buf_end, 4 - (data & 3))) + return state_error; + *buffer = buf; + return state_command; +} + +static __inline__ verifier_state_t +via_parse_vheader6(drm_via_private_t * dev_priv, uint32_t const **buffer, + const uint32_t * buf_end) +{ + + uint32_t addr, count, i; + const uint32_t *buf = *buffer; + + i = count = *++buf; + buf += 3; + while (i--) { + addr = *buf++; + VIA_WRITE(addr, *buf++); + } + count <<= 1; + if (count & 3) + buf += 4 - (count & 3); + *buffer = buf; + return state_command; +} + +int +via_verify_command_stream(const uint32_t * buf, unsigned int size, + struct drm_device * dev, int agp) +{ + + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + drm_via_state_t *hc_state = &dev_priv->hc_state; + drm_via_state_t saved_state = *hc_state; + uint32_t cmd; + const uint32_t *buf_end = buf + (size >> 2); + verifier_state_t state = state_command; + int cme_video; + int supported_3d; + + cme_video = (dev_priv->chipset == VIA_PRO_GROUP_A || + dev_priv->chipset == VIA_DX9_0); + + supported_3d = dev_priv->chipset != VIA_DX9_0; + + hc_state->dev = dev; + hc_state->unfinished = no_sequence; + hc_state->map_cache = NULL; + hc_state->agp = agp; + hc_state->buf_start = buf; + dev_priv->num_fire_offsets = 0; + + while (buf < buf_end) { + + switch (state) { + case state_header2: + state = via_check_header2(&buf, buf_end, hc_state); + break; + case state_header1: + state = via_check_header1(&buf, buf_end); + break; + case state_vheader5: + state = via_check_vheader5(&buf, buf_end); + break; + case state_vheader6: + state = via_check_vheader6(&buf, buf_end); + break; + case state_command: + if ((HALCYON_HEADER2 == (cmd = *buf)) && + supported_3d) + state = state_header2; + else if ((cmd & HALCYON_HEADER1MASK) == HALCYON_HEADER1) + state = state_header1; + else if (cme_video + && (cmd & VIA_VIDEOMASK) == VIA_VIDEO_HEADER5) + state = state_vheader5; + else if (cme_video + && (cmd & VIA_VIDEOMASK) == VIA_VIDEO_HEADER6) + state = state_vheader6; + else if ((cmd == HALCYON_HEADER2) && !supported_3d) { + DRM_ERROR("Accelerated 3D is not supported on this chipset yet.\n"); + state = state_error; + } else { + DRM_ERROR + ("Invalid / Unimplemented DMA HEADER command. 0x%x\n", + cmd); + state = state_error; + } + break; + case state_error: + default: + *hc_state = saved_state; + return -EINVAL; + } + } + if (state == state_error) { + *hc_state = saved_state; + return -EINVAL; + } + return 0; +} + +int +via_parse_command_stream(struct drm_device * dev, const uint32_t * buf, + unsigned int size) +{ + + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + uint32_t cmd; + const uint32_t *buf_end = buf + (size >> 2); + verifier_state_t state = state_command; + int fire_count = 0; + + while (buf < buf_end) { + + switch (state) { + case state_header2: + state = + via_parse_header2(dev_priv, &buf, buf_end, + &fire_count); + break; + case state_header1: + state = via_parse_header1(dev_priv, &buf, buf_end); + break; + case state_vheader5: + state = via_parse_vheader5(dev_priv, &buf, buf_end); + break; + case state_vheader6: + state = via_parse_vheader6(dev_priv, &buf, buf_end); + break; + case state_command: + if (HALCYON_HEADER2 == (cmd = *buf)) + state = state_header2; + else if ((cmd & HALCYON_HEADER1MASK) == HALCYON_HEADER1) + state = state_header1; + else if ((cmd & VIA_VIDEOMASK) == VIA_VIDEO_HEADER5) + state = state_vheader5; + else if ((cmd & VIA_VIDEOMASK) == VIA_VIDEO_HEADER6) + state = state_vheader6; + else { + DRM_ERROR + ("Invalid / Unimplemented DMA HEADER command. 0x%x\n", + cmd); + state = state_error; + } + break; + case state_error: + default: + return -EINVAL; + } + } + if (state == state_error) { + return -EINVAL; + } + return 0; +} + +static void +setup_hazard_table(hz_init_t init_table[], hazard_t table[], int size) +{ + int i; + + for (i = 0; i < 256; ++i) { + table[i] = forbidden_command; + } + + for (i = 0; i < size; ++i) { + table[init_table[i].code] = init_table[i].hz; + } +} + +void via_init_command_verifier(void) +{ + setup_hazard_table(init_table1, table1, + sizeof(init_table1) / sizeof(hz_init_t)); + setup_hazard_table(init_table2, table2, + sizeof(init_table2) / sizeof(hz_init_t)); + setup_hazard_table(init_table3, table3, + sizeof(init_table3) / sizeof(hz_init_t)); +} diff --git a/sys/dev/drm/via_verifier.h b/sys/dev/drm/via_verifier.h new file mode 100644 index 0000000..153b472 --- /dev/null +++ b/sys/dev/drm/via_verifier.h @@ -0,0 +1,65 @@ +/*- + * Copyright 2004 The Unichrome Project. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE UNICHROME PROJECT, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Author: Thomas Hellström 2004. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifndef _VIA_VERIFIER_H_ +#define _VIA_VERIFIER_H_ + +typedef enum { + no_sequence = 0, + z_address, + dest_address, + tex_address +} drm_via_sequence_t; + +typedef struct { + unsigned texture; + uint32_t z_addr; + uint32_t d_addr; + uint32_t t_addr[2][10]; + uint32_t pitch[2][10]; + uint32_t height[2][10]; + uint32_t tex_level_lo[2]; + uint32_t tex_level_hi[2]; + uint32_t tex_palette_size[2]; + uint32_t tex_npot[2]; + drm_via_sequence_t unfinished; + int agp_texture; + int multitex; + struct drm_device *dev; + drm_local_map_t *map_cache; + uint32_t vertex_count; + int agp; + const uint32_t *buf_start; +} drm_via_state_t; + +extern int via_verify_command_stream(const uint32_t * buf, unsigned int size, + struct drm_device * dev, int agp); +extern int via_parse_command_stream(struct drm_device *dev, const uint32_t *buf, + unsigned int size); + +#endif diff --git a/sys/dev/drm/via_video.c b/sys/dev/drm/via_video.c new file mode 100644 index 0000000..3bd9605 --- /dev/null +++ b/sys/dev/drm/via_video.c @@ -0,0 +1,96 @@ +/*- + * Copyright 2005 Thomas Hellstrom. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S), AND/OR THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Author: Thomas Hellstrom 2005. + * + * Video and XvMC related functions. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include "dev/drm/drmP.h" +#include "dev/drm/via_drm.h" +#include "dev/drm/via_drv.h" + +void via_init_futex(drm_via_private_t * dev_priv) +{ + unsigned int i; + + DRM_DEBUG("\n"); + + for (i = 0; i < VIA_NR_XVMC_LOCKS; ++i) { + DRM_INIT_WAITQUEUE(&(dev_priv->decoder_queue[i])); + XVMCLOCKPTR(dev_priv->sarea_priv, i)->lock = 0; + } +} + +void via_cleanup_futex(drm_via_private_t * dev_priv) +{ +} + +void via_release_futex(drm_via_private_t * dev_priv, int context) +{ + unsigned int i; + volatile int *lock; + + if (!dev_priv->sarea_priv) + return; + + for (i = 0; i < VIA_NR_XVMC_LOCKS; ++i) { + lock = (volatile int *)XVMCLOCKPTR(dev_priv->sarea_priv, i); + if ((_DRM_LOCKING_CONTEXT(*lock) == context)) { + if (_DRM_LOCK_IS_HELD(*lock) + && (*lock & _DRM_LOCK_CONT)) { + DRM_WAKEUP(&(dev_priv->decoder_queue[i])); + } + *lock = 0; + } + } +} + +int via_decoder_futex(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + drm_via_futex_t *fx = data; + volatile int *lock; + drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; + drm_via_sarea_t *sAPriv = dev_priv->sarea_priv; + int ret = 0; + + DRM_DEBUG("\n"); + + if (fx->lock >= VIA_NR_XVMC_LOCKS) + return -EFAULT; + + lock = (volatile int *)XVMCLOCKPTR(sAPriv, fx->lock); + + switch (fx->func) { + case VIA_FUTEX_WAIT: + DRM_WAIT_ON(ret, dev_priv->decoder_queue[fx->lock], + (fx->ms / 10) * (DRM_HZ / 100), *lock != fx->val); + return ret; + case VIA_FUTEX_WAKE: + DRM_WAKEUP(&(dev_priv->decoder_queue[fx->lock])); + return 0; + } + return 0; +} -- 1.7.6.3