diff options
Diffstat (limited to 'shared-core')
| -rw-r--r-- | shared-core/drm.h | 34 | ||||
| -rw-r--r-- | shared-core/drm_pciids.txt | 15 | ||||
| -rw-r--r-- | shared-core/i915_dma.c | 143 | ||||
| -rw-r--r-- | shared-core/i915_drm.h | 321 | ||||
| -rw-r--r-- | shared-core/i915_drv.h | 396 | ||||
| -rw-r--r-- | shared-core/i915_init.c | 320 | ||||
| -rw-r--r-- | shared-core/i915_irq.c | 366 | ||||
| -rw-r--r-- | shared-core/i915_suspend.c | 520 | ||||
| -rw-r--r-- | shared-core/nouveau_fifo.c | 2 | ||||
| -rw-r--r-- | shared-core/nouveau_mem.c | 1 | ||||
| -rw-r--r-- | shared-core/nv04_fifo.c | 4 | ||||
| -rw-r--r-- | shared-core/nv40_graph.c | 10 | ||||
| -rw-r--r-- | shared-core/r300_cmdbuf.c | 225 | ||||
| -rw-r--r-- | shared-core/r300_reg.h | 229 | ||||
| -rw-r--r-- | shared-core/radeon_cp.c | 1070 | ||||
| -rw-r--r-- | shared-core/radeon_drm.h | 117 | ||||
| -rw-r--r-- | shared-core/radeon_drv.h | 331 | ||||
| -rw-r--r-- | shared-core/radeon_irq.c | 71 | ||||
| -rw-r--r-- | shared-core/radeon_mem.c | 8 | ||||
| -rw-r--r-- | shared-core/radeon_state.c | 203 | ||||
| -rw-r--r-- | shared-core/xgi_drm.h | 4 |
21 files changed, 3450 insertions, 940 deletions
diff --git a/shared-core/drm.h b/shared-core/drm.h index 41190640..dd03bf9b 100644 --- a/shared-core/drm.h +++ b/shared-core/drm.h @@ -236,7 +236,7 @@ enum drm_map_type { _DRM_AGP = 3, /**< AGP/GART */ _DRM_SCATTER_GATHER = 4, /**< Scatter/gather memory for PCI DMA */ _DRM_CONSISTENT = 5, /**< Consistent memory for PCI DMA */ - _DRM_TTM = 6 + _DRM_TTM = 6, }; /** @@ -993,6 +993,30 @@ struct drm_mm_info_arg { uint64_t p_size; }; +struct drm_gem_close { + /** Handle of the object to be closed. */ + uint32_t handle; + uint32_t pad; +}; + +struct drm_gem_flink { + /** Handle for the object being named */ + uint32_t handle; + + /** Returned global name */ + uint32_t name; +}; + +struct drm_gem_open { + /** Name of object being opened */ + uint32_t name; + + /** Returned handle for the object */ + uint32_t handle; + + /** Returned size of the object */ + uint64_t size; +}; /* * Drm mode setting @@ -1257,7 +1281,11 @@ struct drm_mode_crtc_lut { #define DRM_IOCTL_GET_CLIENT DRM_IOWR(0x05, struct drm_client) #define DRM_IOCTL_GET_STATS DRM_IOR( 0x06, struct drm_stats) #define DRM_IOCTL_SET_VERSION DRM_IOWR(0x07, struct drm_set_version) -#define DRM_IOCTL_MODESET_CTL DRM_IOW(0x08, struct drm_modeset_ctl) +#define DRM_IOCTL_MODESET_CTL DRM_IOW(0x08, struct drm_modeset_ctl) + +#define DRM_IOCTL_GEM_CLOSE DRM_IOW (0x09, struct drm_gem_close) +#define DRM_IOCTL_GEM_FLINK DRM_IOWR(0x0a, struct drm_gem_flink) +#define DRM_IOCTL_GEM_OPEN DRM_IOWR(0x0b, struct drm_gem_open) #define DRM_IOCTL_SET_UNIQUE DRM_IOW( 0x10, struct drm_unique) #define DRM_IOCTL_AUTH_MAGIC DRM_IOW( 0x11, struct drm_auth) @@ -1302,7 +1330,7 @@ struct drm_mode_crtc_lut { #define DRM_IOCTL_AGP_BIND DRM_IOW( 0x36, struct drm_agp_binding) #define DRM_IOCTL_AGP_UNBIND DRM_IOW( 0x37, struct drm_agp_binding) -#define DRM_IOCTL_SG_ALLOC DRM_IOW( 0x38, struct drm_scatter_gather) +#define DRM_IOCTL_SG_ALLOC DRM_IOWR(0x38, struct drm_scatter_gather) #define DRM_IOCTL_SG_FREE DRM_IOW( 0x39, struct drm_scatter_gather) #define DRM_IOCTL_WAIT_VBLANK DRM_IOWR(0x3a, union drm_wait_vblank) diff --git a/shared-core/drm_pciids.txt b/shared-core/drm_pciids.txt index b95bc8f2..0c45b4c1 100644 --- a/shared-core/drm_pciids.txt +++ b/shared-core/drm_pciids.txt @@ -83,7 +83,6 @@ 0x1002 0x5460 CHIP_RV380|RADEON_IS_MOBILITY "ATI Radeon Mobility X300 M22" 0x1002 0x5462 CHIP_RV380|RADEON_IS_MOBILITY "ATI Radeon Mobility X600 SE M24C" 0x1002 0x5464 CHIP_RV380|RADEON_IS_MOBILITY "ATI FireGL M22 GL 5464" -0x1002 0x5657 CHIP_RV380|RADEON_NEW_MEMMAP "ATI Radeon RV370 X550XTX" 0x1002 0x5548 CHIP_R420|RADEON_NEW_MEMMAP "ATI Radeon R423 X800" 0x1002 0x5549 CHIP_R420|RADEON_NEW_MEMMAP "ATI Radeon R423 X800 Pro" 0x1002 0x554A CHIP_R420|RADEON_NEW_MEMMAP "ATI Radeon R423 X800 XT PE" @@ -101,9 +100,10 @@ 0x1002 0x564F CHIP_RV410|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP "ATI Radeon Mobility X700 XL M26" 0x1002 0x5652 CHIP_RV410|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP "ATI Radeon Mobility X700 M26" 0x1002 0x5653 CHIP_RV410|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP "ATI Radeon Mobility X700 M26" +0x1002 0x5657 CHIP_RV410|RADEON_NEW_MEMMAP "ATI Radeon X550XTX" 0x1002 0x5834 CHIP_RS300|RADEON_IS_IGP "ATI Radeon RS300 9100 IGP" 0x1002 0x5835 CHIP_RS300|RADEON_IS_IGP|RADEON_IS_MOBILITY "ATI Radeon RS300 Mobility IGP" -0x1002 0x5954 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI RS480 XPRESS 200G" +0x1002 0x5954 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_IGPGART "ATI RS480 XPRESS 200G" 0x1002 0x5955 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon XPRESS 200M 5955" 0x1002 0x5974 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon RS482 XPRESS 200" 0x1002 0x5975 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon RS485 XPRESS 1100 IGP" @@ -113,8 +113,10 @@ 0x1002 0x5964 CHIP_RV280 "ATI Radeon RV280 9200 SE" 0x1002 0x5965 CHIP_RV280 "ATI FireMV 2200 PCI" 0x1002 0x5969 CHIP_RV100 "ATI ES1000 RN50" -0x1002 0x5a61 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon RC410 XPRESS 200" -0x1002 0x5a62 CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon RC410 XPRESS 200M" +0x1002 0x5a41 CHIP_RS400|RADEON_IS_IGP|RADEON_IS_IGPGART "ATI Radeon XPRESS 200 5A41 (PCIE)" +0x1002 0x5a42 CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon XPRESS 200M 5A42 (PCIE)" +0x1002 0x5a61 CHIP_RS400|RADEON_IS_IGP|RADEON_IS_IGPGART "ATI Radeon RC410 XPRESS 200" +0x1002 0x5a62 CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART "ATI Radeon RC410 XPRESS 200M" 0x1002 0x5b60 CHIP_RV380|RADEON_NEW_MEMMAP "ATI Radeon RV370 X300 SE" 0x1002 0x5b62 CHIP_RV380|RADEON_NEW_MEMMAP "ATI Radeon RV370 X600 Pro" 0x1002 0x5b63 CHIP_RV380|RADEON_NEW_MEMMAP "ATI Radeon RV370 X550" @@ -396,7 +398,10 @@ 0x8086 0x29C2 CHIP_I9XX|CHIP_I915 "Intel G33" 0x8086 0x29B2 CHIP_I9XX|CHIP_I915 "Intel Q35" 0x8086 0x29D2 CHIP_I9XX|CHIP_I915 "Intel Q33" -0x8086 0x2A42 CHIP_I9XX|CHIP_I965 "Intel Integrated Graphics Device" +0x8086 0x2A42 CHIP_I9XX|CHIP_I965 "Mobile Intel® GM45 Express Chipset" +0x8086 0x2E02 CHIP_I9XX|CHIP_I965 "Intel Integrated Graphics Device" +0x8086 0x2E12 CHIP_I9XX|CHIP_I965 "Intel Q45/Q43" +0x8086 0x2E22 CHIP_I9XX|CHIP_I965 "Intel G45/G43" [imagine] 0x105d 0x2309 IMAGINE_128 "Imagine 128" diff --git a/shared-core/i915_dma.c b/shared-core/i915_dma.c index db857fbd..1fdc5e17 100644 --- a/shared-core/i915_dma.c +++ b/shared-core/i915_dma.c @@ -41,10 +41,14 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_ring_buffer *ring = &(dev_priv->ring); u32 last_head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD; + u32 last_acthd = I915_READ(acthd_reg); + u32 acthd; int i; - for (i = 0; i < 10000; i++) { + for (i = 0; i < 100000; i++) { ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + acthd = I915_READ(acthd_reg); ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) ring->space += ring->Size; @@ -54,13 +58,41 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) if (ring->head != last_head) i = 0; + if (acthd != last_acthd) + i = 0; + last_head = ring->head; - DRM_UDELAY(1); + last_acthd = acthd; + msleep_interruptible (10); } return -EBUSY; } +#if I915_RING_VALIDATE +/** + * Validate the cached ring tail value + * + * If the X server writes to the ring and DRM doesn't + * reload the head and tail pointers, it will end up writing + * data to the wrong place in the ring, causing havoc. + */ +void i915_ring_validate(struct drm_device *dev, const char *func, int line) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + drm_i915_ring_buffer_t *ring = &(dev_priv->ring); + u32 tail = I915_READ(LP_RING+RING_TAIL) & HEAD_ADDR; + u32 head = I915_READ(LP_RING+RING_HEAD) & HEAD_ADDR; + + if (tail != ring->tail) { + DRM_ERROR("%s:%d head sw %x, hw %x. tail sw %x hw %x\n", + func, line, + ring->head, head, ring->tail, tail); + BUG_ON(1); + } +} +#endif + void i915_kernel_lost_context(struct drm_device * dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -89,7 +121,7 @@ int i915_dma_cleanup(struct drm_device * dev) * may not have been called from userspace and after dev_private * is freed, it's too late. */ - if (dev->irq) + if (dev->irq_enabled) drm_irq_uninstall(dev); if (dev_priv->ring.virtual_start) { @@ -100,24 +132,10 @@ int i915_dma_cleanup(struct drm_device * dev) dev_priv->ring.Size = 0; } - if (dev_priv->status_page_dmah) { - drm_pci_free(dev, dev_priv->status_page_dmah); - dev_priv->status_page_dmah = NULL; - /* Need to rewrite hardware status page */ - I915_WRITE(0x02080, 0x1ffff000); - } - - if (dev_priv->status_gfx_addr) { - dev_priv->status_gfx_addr = 0; - drm_core_ioremapfree(&dev_priv->hws_map, dev); - I915_WRITE(0x02080, 0x1ffff000); - } - - return 0; } -#if defined(I915_HAVE_BUFFER) +#if defined(I915_HAVE_BUFFER) && defined(DRI2) #define DRI2_SAREA_BLOCK_TYPE(b) ((b) >> 16) #define DRI2_SAREA_BLOCK_SIZE(b) ((b) & 0xffff) #define DRI2_SAREA_BLOCK_NEXT(p) \ @@ -195,27 +213,22 @@ static int i915_initialize(struct drm_device * dev, } } - #ifdef I915_HAVE_BUFFER if (!drm_core_check_feature(dev, DRIVER_MODESET)) { dev_priv->max_validate_buffers = I915_MAX_VALIDATE_BUFFERS; } #endif - if (!dev_priv->ring.Size) { - dev_priv->ring.Start = init->ring_start; - dev_priv->ring.End = init->ring_end; + if (init->ring_size != 0) { dev_priv->ring.Size = init->ring_size; dev_priv->ring.tail_mask = dev_priv->ring.Size - 1; - dev_priv->ring.map.offset = init->ring_start; dev_priv->ring.map.size = init->ring_size; dev_priv->ring.map.type = 0; dev_priv->ring.map.flags = 0; dev_priv->ring.map.mtrr = 0; - drm_core_ioremap(&dev_priv->ring.map, dev); - + if (dev_priv->ring.map.handle == NULL) { i915_dma_cleanup(dev); DRM_ERROR("can not ioremap virtual address for" @@ -225,7 +238,6 @@ static int i915_initialize(struct drm_device * dev, dev_priv->ring.virtual_start = dev_priv->ring.map.handle; } - dev_priv->cpp = init->cpp; master_priv->sarea_priv->pf_current_page = 0; @@ -241,31 +253,11 @@ static int i915_initialize(struct drm_device * dev, */ dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A; - /* Program Hardware Status Page */ - if (!I915_NEED_GFX_HWS(dev)) { - dev_priv->status_page_dmah = - drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff); - - if (!dev_priv->status_page_dmah) { - i915_dma_cleanup(dev); - DRM_ERROR("Can not allocate hardware status page\n"); - return -ENOMEM; - } - dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr; - dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; - - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - - I915_WRITE(0x02080, dev_priv->dma_status_page); - } - DRM_DEBUG("Enabled hardware status page\n"); - #ifdef I915_HAVE_BUFFER if (!drm_core_check_feature(dev, DRIVER_MODESET)) { mutex_init(&dev_priv->cmdbuf_mutex); } -#endif -#if defined(I915_HAVE_BUFFER) +#ifdef DRI2 if (init->func == I915_INIT_DMA2) { int ret = setup_dri2_sarea(dev, file_priv, init); if (ret) { @@ -274,7 +266,8 @@ static int i915_initialize(struct drm_device * dev, return ret; } } -#endif +#endif /* DRI2 */ +#endif /* I915_HAVE_BUFFER */ return 0; } @@ -288,11 +281,6 @@ static int i915_dma_resume(struct drm_device * dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) return 0; - if (!dev_priv->mmio_map) { - DRM_ERROR("can not find mmio map!\n"); - return -EINVAL; - } - if (dev_priv->ring.map.handle == NULL) { DRM_ERROR("can not ioremap virtual address for" " ring buffer\n"); @@ -307,9 +295,9 @@ static int i915_dma_resume(struct drm_device * dev) DRM_DEBUG("hw status page @ %p\n", dev_priv->hw_status_page); if (dev_priv->status_gfx_addr != 0) - I915_WRITE(0x02080, dev_priv->status_gfx_addr); + I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); else - I915_WRITE(0x02080, dev_priv->dma_status_page); + I915_WRITE(HWS_PGA, dev_priv->dma_status_page); DRM_DEBUG("Enabled hardware status page\n"); return 0; @@ -456,9 +444,9 @@ static int i915_emit_cmds(struct drm_device *dev, int __user *buffer, return 0; } -static int i915_emit_box(struct drm_device * dev, - struct drm_clip_rect __user * boxes, - int i, int DR1, int DR4) +int i915_emit_box(struct drm_device * dev, + struct drm_clip_rect __user * boxes, + int i, int DR1, int DR4) { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_clip_rect box; @@ -514,7 +502,7 @@ void i915_emit_breadcrumb(struct drm_device *dev) BEGIN_LP_RING(4); OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(20); + OUT_RING(5 << MI_STORE_DWORD_INDEX_SHIFT); OUT_RING(dev_priv->counter); OUT_RING(0); ADVANCE_LP_RING(); @@ -713,9 +701,19 @@ void i915_dispatch_flip(struct drm_device * dev, int planes, int sync) int i915_quiescent(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + int ret; i915_kernel_lost_context(dev); - return i915_wait_ring(dev, dev_priv->ring.Size - 8, __FUNCTION__); + ret = i915_wait_ring(dev, dev_priv->ring.Size - 8, __FUNCTION__); + if (ret) + { + i915_kernel_lost_context (dev); + DRM_ERROR ("not quiescent head %08x tail %08x space %08x\n", + dev_priv->ring.head, + dev_priv->ring.tail, + dev_priv->ring.space); + } + return ret; } static int i915_flush_ioctl(struct drm_device *dev, void *data, @@ -855,7 +853,7 @@ static int i915_getparam(struct drm_device *dev, void *data, switch (param->param) { case I915_PARAM_IRQ_ACTIVE: - value = dev->irq ? 1 : 0; + value = dev->irq_enabled ? 1 : 0; break; case I915_PARAM_ALLOW_BATCHBUFFER: value = dev_priv->allow_batchbuffer ? 1 : 0; @@ -866,6 +864,9 @@ static int i915_getparam(struct drm_device *dev, void *data, case I915_PARAM_CHIPSET_ID: value = dev->pci_device; break; + case I915_PARAM_HAS_GEM: + value = 1; + break; default: DRM_ERROR("Unknown parameter %d\n", param->param); return -EINVAL; @@ -1019,13 +1020,13 @@ static int i915_set_status_page(struct drm_device *dev, void *data, " G33 hw status page\n"); return -ENOMEM; } + dev_priv->hw_status_page = dev_priv->hws_map.handle; memset(dev_priv->hw_status_page, 0, PAGE_SIZE); I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); - DRM_DEBUG("load hws 0x2080 with gfx mem 0x%x\n", - dev_priv->status_gfx_addr); DRM_DEBUG("load hws at %p\n", dev_priv->hw_status_page); + return 0; } @@ -1051,6 +1052,22 @@ struct drm_ioctl_desc i915_ioctls[] = { #ifdef I915_HAVE_BUFFER DRM_IOCTL_DEF(DRM_I915_EXECBUFFER, i915_execbuffer, DRM_AUTH), #endif + DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), + DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), + DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_THROTTLE, i915_gem_throttle_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_ENTERVT, i915_gem_entervt_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_LEAVEVT, i915_gem_leavevt_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_CREATE, i915_gem_create_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_PREAD, i915_gem_pread_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_PWRITE, i915_gem_pwrite_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_MMAP, i915_gem_mmap_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SET_DOMAIN, i915_gem_set_domain_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SW_FINISH, i915_gem_sw_finish_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SET_TILING, i915_gem_set_tiling, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_GET_TILING, i915_gem_get_tiling, 0), }; int i915_max_ioctl = DRM_ARRAY_SIZE(i915_ioctls); diff --git a/shared-core/i915_drm.h b/shared-core/i915_drm.h index bdcac9aa..53087b57 100644 --- a/shared-core/i915_drm.h +++ b/shared-core/i915_drm.h @@ -176,6 +176,22 @@ typedef struct drm_i915_sarea { #define DRM_I915_MMIO 0x10 #define DRM_I915_HWS_ADDR 0x11 #define DRM_I915_EXECBUFFER 0x12 +#define DRM_I915_GEM_INIT 0x13 +#define DRM_I915_GEM_EXECBUFFER 0x14 +#define DRM_I915_GEM_PIN 0x15 +#define DRM_I915_GEM_UNPIN 0x16 +#define DRM_I915_GEM_BUSY 0x17 +#define DRM_I915_GEM_THROTTLE 0x18 +#define DRM_I915_GEM_ENTERVT 0x19 +#define DRM_I915_GEM_LEAVEVT 0x1a +#define DRM_I915_GEM_CREATE 0x1b +#define DRM_I915_GEM_PREAD 0x1c +#define DRM_I915_GEM_PWRITE 0x1d +#define DRM_I915_GEM_MMAP 0x1e +#define DRM_I915_GEM_SET_DOMAIN 0x1f +#define DRM_I915_GEM_SW_FINISH 0x20 +#define DRM_I915_GEM_SET_TILING 0x21 +#define DRM_I915_GEM_GET_TILING 0x22 #define DRM_IOCTL_I915_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t) #define DRM_IOCTL_I915_FLUSH DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH) @@ -195,6 +211,22 @@ typedef struct drm_i915_sarea { #define DRM_IOCTL_I915_VBLANK_SWAP DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_VBLANK_SWAP, drm_i915_vblank_swap_t) #define DRM_IOCTL_I915_MMIO DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_MMIO, drm_i915_mmio) #define DRM_IOCTL_I915_EXECBUFFER DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_EXECBUFFER, struct drm_i915_execbuffer) +#define DRM_IOCTL_I915_GEM_INIT DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_INIT, struct drm_i915_gem_init) +#define DRM_IOCTL_I915_GEM_EXECBUFFER DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER, struct drm_i915_gem_execbuffer) +#define DRM_IOCTL_I915_GEM_PIN DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_PIN, struct drm_i915_gem_pin) +#define DRM_IOCTL_I915_GEM_UNPIN DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_UNPIN, struct drm_i915_gem_unpin) +#define DRM_IOCTL_I915_GEM_BUSY DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_BUSY, struct drm_i915_gem_busy) +#define DRM_IOCTL_I915_GEM_THROTTLE DRM_IO ( DRM_COMMAND_BASE + DRM_I915_GEM_THROTTLE) +#define DRM_IOCTL_I915_GEM_ENTERVT DRM_IO(DRM_COMMAND_BASE + DRM_I915_GEM_ENTERVT) +#define DRM_IOCTL_I915_GEM_LEAVEVT DRM_IO(DRM_COMMAND_BASE + DRM_I915_GEM_LEAVEVT) +#define DRM_IOCTL_I915_GEM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_CREATE, struct drm_i915_gem_create) +#define DRM_IOCTL_I915_GEM_PREAD DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_PREAD, struct drm_i915_gem_pread) +#define DRM_IOCTL_I915_GEM_PWRITE DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_PWRITE, struct drm_i915_gem_pwrite) +#define DRM_IOCTL_I915_GEM_MMAP DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_MMAP, struct drm_i915_gem_mmap) +#define DRM_IOCTL_I915_GEM_SET_DOMAIN DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_SET_DOMAIN, struct drm_i915_gem_set_domain) +#define DRM_IOCTL_I915_GEM_SW_FINISH DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_SW_FINISH, struct drm_i915_gem_sw_finish) +#define DRM_IOCTL_I915_GEM_SET_TILING DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_SET_TILING, struct drm_i915_gem_set_tiling) +#define DRM_IOCTL_I915_GEM_GET_TILING DRM_IOWR (DRM_COMMAND_BASE + DRM_I915_GEM_GET_TILING, struct drm_i915_gem_get_tiling) /* Asynchronous page flipping: */ @@ -248,6 +280,7 @@ typedef struct drm_i915_irq_wait { #define I915_PARAM_ALLOW_BATCHBUFFER 2 #define I915_PARAM_LAST_DISPATCH 3 #define I915_PARAM_CHIPSET_ID 4 +#define I915_PARAM_HAS_GEM 5 typedef struct drm_i915_getparam { int param; @@ -399,4 +432,292 @@ struct drm_i915_execbuffer { struct drm_fence_arg fence_arg; }; +struct drm_i915_gem_init { + /** + * Beginning offset in the GTT to be managed by the DRM memory + * manager. + */ + uint64_t gtt_start; + /** + * Ending offset in the GTT to be managed by the DRM memory + * manager. + */ + uint64_t gtt_end; +}; + +struct drm_i915_gem_create { + /** + * Requested size for the object. + * + * The (page-aligned) allocated size for the object will be returned. + */ + uint64_t size; + /** + * Returned handle for the object. + * + * Object handles are nonzero. + */ + uint32_t handle; + uint32_t pad; +}; + +struct drm_i915_gem_pread { + /** Handle for the object being read. */ + uint32_t handle; + uint32_t pad; + /** Offset into the object to read from */ + uint64_t offset; + /** Length of data to read */ + uint64_t size; + /** Pointer to write the data into. */ + uint64_t data_ptr; /* void *, but pointers are not 32/64 compatible */ +}; + +struct drm_i915_gem_pwrite { + /** Handle for the object being written to. */ + uint32_t handle; + uint32_t pad; + /** Offset into the object to write to */ + uint64_t offset; + /** Length of data to write */ + uint64_t size; + /** Pointer to read the data from. */ + uint64_t data_ptr; /* void *, but pointers are not 32/64 compatible */ +}; + +struct drm_i915_gem_mmap { + /** Handle for the object being mapped. */ + uint32_t handle; + uint32_t pad; + /** Offset in the object to map. */ + uint64_t offset; + /** + * Length of data to map. + * + * The value will be page-aligned. + */ + uint64_t size; + /** Returned pointer the data was mapped at */ + uint64_t addr_ptr; /* void *, but pointers are not 32/64 compatible */ +}; + +struct drm_i915_gem_set_domain { + /** Handle for the object */ + uint32_t handle; + + /** New read domains */ + uint32_t read_domains; + + /** New write domain */ + uint32_t write_domain; +}; + +struct drm_i915_gem_sw_finish { + /** Handle for the object */ + uint32_t handle; +}; + +struct drm_i915_gem_relocation_entry { + /** + * Handle of the buffer being pointed to by this relocation entry. + * + * It's appealing to make this be an index into the mm_validate_entry + * list to refer to the buffer, but this allows the driver to create + * a relocation list for state buffers and not re-write it per + * exec using the buffer. + */ + uint32_t target_handle; + + /** + * Value to be added to the offset of the target buffer to make up + * the relocation entry. + */ + uint32_t delta; + + /** Offset in the buffer the relocation entry will be written into */ + uint64_t offset; + + /** + * Offset value of the target buffer that the relocation entry was last + * written as. + * + * If the buffer has the same offset as last time, we can skip syncing + * and writing the relocation. This value is written back out by + * the execbuffer ioctl when the relocation is written. + */ + uint64_t presumed_offset; + + /** + * Target memory domains read by this operation. + */ + uint32_t read_domains; + + /** + * Target memory domains written by this operation. + * + * Note that only one domain may be written by the whole + * execbuffer operation, so that where there are conflicts, + * the application will get -EINVAL back. + */ + uint32_t write_domain; +}; + +/** @{ + * Intel memory domains + * + * Most of these just align with the various caches in + * the system and are used to flush and invalidate as + * objects end up cached in different domains. + */ +/** CPU cache */ +#define I915_GEM_DOMAIN_CPU 0x00000001 +/** Render cache, used by 2D and 3D drawing */ +#define I915_GEM_DOMAIN_RENDER 0x00000002 +/** Sampler cache, used by texture engine */ +#define I915_GEM_DOMAIN_SAMPLER 0x00000004 +/** Command queue, used to load batch buffers */ +#define I915_GEM_DOMAIN_COMMAND 0x00000008 +/** Instruction cache, used by shader programs */ +#define I915_GEM_DOMAIN_INSTRUCTION 0x00000010 +/** Vertex address cache */ +#define I915_GEM_DOMAIN_VERTEX 0x00000020 +/** GTT domain - aperture and scanout */ +#define I915_GEM_DOMAIN_GTT 0x00000040 +/** @} */ + +struct drm_i915_gem_exec_object { + /** + * User's handle for a buffer to be bound into the GTT for this + * operation. + */ + uint32_t handle; + + /** Number of relocations to be performed on this buffer */ + uint32_t relocation_count; + /** + * Pointer to array of struct drm_i915_gem_relocation_entry containing + * the relocations to be performed in this buffer. + */ + uint64_t relocs_ptr; + + /** Required alignment in graphics aperture */ + uint64_t alignment; + + /** + * Returned value of the updated offset of the object, for future + * presumed_offset writes. + */ + uint64_t offset; +}; + +struct drm_i915_gem_execbuffer { + /** + * List of buffers to be validated with their relocations to be + * performend on them. + * + * This is a pointer to an array of struct drm_i915_gem_validate_entry. + * + * These buffers must be listed in an order such that all relocations + * a buffer is performing refer to buffers that have already appeared + * in the validate list. + */ + uint64_t buffers_ptr; + uint32_t buffer_count; + + /** Offset in the batchbuffer to start execution from. */ + uint32_t batch_start_offset; + /** Bytes used in batchbuffer from batch_start_offset */ + uint32_t batch_len; + uint32_t DR1; + uint32_t DR4; + uint32_t num_cliprects; + uint64_t cliprects_ptr; /* struct drm_clip_rect *cliprects */ +}; + +struct drm_i915_gem_pin { + /** Handle of the buffer to be pinned. */ + uint32_t handle; + uint32_t pad; + + /** alignment required within the aperture */ + uint64_t alignment; + + /** Returned GTT offset of the buffer. */ + uint64_t offset; +}; + +struct drm_i915_gem_unpin { + /** Handle of the buffer to be unpinned. */ + uint32_t handle; + uint32_t pad; +}; + +struct drm_i915_gem_busy { + /** Handle of the buffer to check for busy */ + uint32_t handle; + + /** Return busy status (1 if busy, 0 if idle) */ + uint32_t busy; +}; + +#define I915_TILING_NONE 0 +#define I915_TILING_X 1 +#define I915_TILING_Y 2 + +#define I915_BIT_6_SWIZZLE_NONE 0 +#define I915_BIT_6_SWIZZLE_9 1 +#define I915_BIT_6_SWIZZLE_9_10 2 +#define I915_BIT_6_SWIZZLE_9_11 3 +#define I915_BIT_6_SWIZZLE_9_10_11 4 +/* Not seen by userland */ +#define I915_BIT_6_SWIZZLE_UNKNOWN 5 + +struct drm_i915_gem_set_tiling { + /** Handle of the buffer to have its tiling state updated */ + uint32_t handle; + + /** + * Tiling mode for the object (I915_TILING_NONE, I915_TILING_X, + * I915_TILING_Y). + * + * This value is to be set on request, and will be updated by the + * kernel on successful return with the actual chosen tiling layout. + * + * The tiling mode may be demoted to I915_TILING_NONE when the system + * has bit 6 swizzling that can't be managed correctly by GEM. + * + * Buffer contents become undefined when changing tiling_mode. + */ + uint32_t tiling_mode; + + /** + * Stride in bytes for the object when in I915_TILING_X or + * I915_TILING_Y. + */ + uint32_t stride; + + /** + * Returned address bit 6 swizzling required for CPU access through + * mmap mapping. + */ + uint32_t swizzle_mode; +}; + +struct drm_i915_gem_get_tiling { + /** Handle of the buffer to get tiling state for. */ + uint32_t handle; + + /** + * Current tiling mode for the object (I915_TILING_NONE, I915_TILING_X, + * I915_TILING_Y). + */ + uint32_t tiling_mode; + + /** + * Returned address bit 6 swizzling required for CPU access through + * mmap mapping. + */ + uint32_t swizzle_mode; +}; + #endif /* _I915_DRM_H_ */ diff --git a/shared-core/i915_drv.h b/shared-core/i915_drv.h index 6d72c051..087c6d64 100644 --- a/shared-core/i915_drv.h +++ b/shared-core/i915_drv.h @@ -39,7 +39,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20080312" +#define DRIVER_DATE "20080730" #if defined(__linux__) #define I915_HAVE_FENCE @@ -79,16 +79,23 @@ enum pipe { struct drm_i915_validate_buffer; #endif +#define WATCH_COHERENCY 0 +#define WATCH_BUF 0 +#define WATCH_EXEC 0 +#define WATCH_LRU 0 +#define WATCH_RELOC 0 +#define WATCH_INACTIVE 0 +#define WATCH_PWRITE 0 + struct drm_i915_ring_buffer { int tail_mask; - unsigned long Start; - unsigned long End; unsigned long Size; u8 *virtual_start; int head; int tail; int space; drm_local_map_t map; + struct drm_gem_object *ring_obj; }; struct mem_block { @@ -108,13 +115,30 @@ struct drm_i915_vbl_swap { struct drm_minor *minor; }; + +#ifdef __linux__ +struct opregion_header; +struct opregion_acpi; +struct opregion_swsci; +struct opregion_asle; + +struct intel_opregion { + struct opregion_header *header; + struct opregion_acpi *acpi; + struct opregion_swsci *swsci; + struct opregion_asle *asle; + + int enabled; +}; +#endif + struct drm_i915_master_private { drm_local_map_t *sarea; struct drm_i915_sarea *sarea_priv; }; struct drm_i915_private { - struct drm_buffer_object *ring_buffer; + struct drm_device *dev; drm_local_map_t *mmio_map; @@ -127,15 +151,14 @@ struct drm_i915_private { void *hw_status_page; dma_addr_t dma_status_page; uint32_t counter; - unsigned int status_gfx_addr; + uint32_t status_gfx_addr; drm_local_map_t hws_map; - struct drm_buffer_object *hws_bo; + struct drm_gem_object *hws_obj; unsigned int cpp; wait_queue_head_t irq_queue; atomic_t irq_received; - atomic_t irq_emitted; int tex_lru_log_granularity; int allow_batchbuffer; @@ -145,12 +168,14 @@ struct drm_i915_private { DRM_SPINTYPE user_irq_lock; int user_irq_refcount; int fence_irq_on; - uint32_t irq_enable_reg; + uint32_t irq_mask_reg; int irq_enabled; struct workqueue_struct *wq; bool cursor_needs_physical; + struct drm_mm vram; + #ifdef I915_HAVE_FENCE uint32_t flush_sequence; uint32_t flush_flags; @@ -161,7 +186,7 @@ struct drm_i915_private { void *agp_iomap; unsigned int max_validate_buffers; struct mutex cmdbuf_mutex; - size_t stolen_base; + u32 stolen_base; struct drm_i915_validate_buffer *val_bufs; #endif @@ -175,16 +200,22 @@ struct drm_i915_private { struct drm_display_mode *panel_fixed_mode; struct drm_display_mode *vbt_mode; /* if any */ -#if defined(I915_HAVE_BUFFER) +#if defined(I915_HAVE_BUFFER) && defined(DRI2) /* DRI2 sarea */ - struct drm_buffer_object *sarea_bo; - struct drm_bo_kmap_obj sarea_kmap; + struct drm_gem_object *sarea_object; + struct drm_bo_kmap_obj sarea_kmap; +#endif /* Feature bits from the VBIOS */ int int_tv_support:1; int lvds_dither:1; int lvds_vbt:1; int int_crt_support:1; + + struct work_struct user_interrupt_task; + +#ifdef __linux__ + struct intel_opregion opregion; #endif /* Register state */ @@ -275,6 +306,95 @@ struct drm_i915_private { u8 saveDACMASK; u8 saveDACDATA[256*3]; /* 256 3-byte colors */ u8 saveCR[37]; + + struct { + struct drm_mm gtt_space; + + /** + * List of objects currently involved in rendering from the + * ringbuffer. + * + * A reference is held on the buffer while on this list. + */ + struct list_head active_list; + + /** + * List of objects which are not in the ringbuffer but which + * still have a write_domain which needs to be flushed before + * unbinding. + * + * A reference is held on the buffer while on this list. + */ + struct list_head flushing_list; + + /** + * LRU list of objects which are not in the ringbuffer and + * are ready to unbind, but are still in the GTT. + * + * A reference is not held on the buffer while on this list, + * as merely being GTT-bound shouldn't prevent its being + * freed, and we'll pull it off the list in the free path. + */ + struct list_head inactive_list; + + /** + * List of breadcrumbs associated with GPU requests currently + * outstanding. + */ + struct list_head request_list; + + /** + * We leave the user IRQ off as much as possible, + * but this means that requests will finish and never + * be retired once the system goes idle. Set a timer to + * fire periodically while the ring is running. When it + * fires, go retire requests. + */ + struct delayed_work retire_work; + + uint32_t next_gem_seqno; + + /** + * Waiting sequence number, if any + */ + uint32_t waiting_gem_seqno; + + /** + * Last seq seen at irq time + */ + uint32_t irq_gem_seqno; + + /** + * Flag if the X Server, and thus DRM, is not currently in + * control of the device. + * + * This is set between LeaveVT and EnterVT. It needs to be + * replaced with a semaphore. It also needs to be + * transitioned away from for kernel modesetting. + */ + int suspended; + + /** + * Flag if the hardware appears to be wedged. + * + * This is set when attempts to idle the device timeout. + * It prevents command submission from occuring and makes + * every pending request fail + */ + int wedged; + + /** Bit 6 swizzling required for X tiling */ + uint32_t bit_6_swizzle_x; + /** Bit 6 swizzling required for Y tiling */ + uint32_t bit_6_swizzle_y; + } mm; +}; + +struct drm_i915_file_private { + struct { + uint32_t last_gem_seqno; + uint32_t last_gem_throttle_seqno; + } mm; }; enum intel_chip_family { @@ -284,6 +404,83 @@ enum intel_chip_family { CHIP_I965 = 0x08, }; +/** driver private structure attached to each drm_gem_object */ +struct drm_i915_gem_object { + struct drm_gem_object *obj; + + /** Current space allocated to this object in the GTT, if any. */ + struct drm_mm_node *gtt_space; + + /** This object's place on the active/flushing/inactive lists */ + struct list_head list; + + /** + * This is set if the object is on the active or flushing lists + * (has pending rendering), and is not set if it's on inactive (ready + * to be unbound). + */ + int active; + + /** + * This is set if the object has been written to since last bound + * to the GTT + */ + int dirty; + + /** AGP memory structure for our GTT binding. */ + DRM_AGP_MEM *agp_mem; + + struct page **page_list; + + /** + * Current offset of the object in GTT space. + * + * This is the same as gtt_space->start + */ + uint32_t gtt_offset; + + /** Boolean whether this object has a valid gtt offset. */ + int gtt_bound; + + /** How many users have pinned this object in GTT space */ + int pin_count; + + /** Breadcrumb of last rendering to the buffer. */ + uint32_t last_rendering_seqno; + + /** Current tiling mode for the object. */ + uint32_t tiling_mode; + + /** + * Flagging of which individual pages are valid in GEM_DOMAIN_CPU when + * GEM_DOMAIN_CPU is not in the object's read domain. + */ + uint8_t *page_cpu_valid; +}; + +/** + * Request queue structure. + * + * The request queue allows us to note sequence numbers that have been emitted + * and may be associated with active buffers to be retired. + * + * By keeping this list, we can avoid having to do questionable + * sequence-number comparisons on buffer last_rendering_seqnos, and associate + * an emission time with seqnos for tracking how far ahead of the GPU we are. + */ +struct drm_i915_gem_request { + /** GEM sequence number associated with this request. */ + uint32_t seqno; + + /** Time at which this request was emitted, in jiffies. */ + unsigned long emitted_jiffies; + + /** Cache domains that were flushed at the start of the request. */ + uint32_t flush_domains; + + struct list_head list; +}; + extern struct drm_ioctl_desc i915_ioctls[]; extern int i915_max_ioctl; @@ -294,8 +491,11 @@ extern void i915_kernel_lost_context(struct drm_device * dev); extern int i915_driver_load(struct drm_device *, unsigned long flags); extern int i915_driver_unload(struct drm_device *dev); extern void i915_driver_lastclose(struct drm_device * dev); +extern int i915_driver_open(struct drm_device *dev, struct drm_file *file_priv); extern void i915_driver_preclose(struct drm_device *dev, struct drm_file *file_priv); +extern void i915_driver_postclose(struct drm_device *dev, + struct drm_file *file_priv); extern int i915_driver_device_is_agp(struct drm_device * dev); extern long i915_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); @@ -309,6 +509,10 @@ extern int i915_dispatch_batchbuffer(struct drm_device * dev, drm_i915_batchbuffer_t * batch); extern int i915_quiescent(struct drm_device *dev); +int i915_emit_box(struct drm_device * dev, + struct drm_clip_rect __user * boxes, + int i, int DR1, int DR4); + /* i915_irq.c */ extern int i915_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv); @@ -325,6 +529,7 @@ extern int i915_vblank_pipe_get(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int i915_emit_irq(struct drm_device * dev); extern void i915_enable_interrupt (struct drm_device *dev); +extern int i915_wait_irq(struct drm_device * dev, int irq_nr); extern int i915_enable_vblank(struct drm_device *dev, int crtc); extern void i915_disable_vblank(struct drm_device *dev, int crtc); extern u32 i915_get_vblank_counter(struct drm_device *dev, int crtc); @@ -346,6 +551,11 @@ extern void i915_mem_takedown(struct mem_block **heap); extern void i915_mem_release(struct drm_device * dev, struct drm_file *file_priv, struct mem_block *heap); + +/* i915_suspend.c */ +extern int i915_save_state(struct drm_device *dev); +extern int i915_restore_state(struct drm_device *dev); + #ifdef I915_HAVE_FENCE /* i915_fence.c */ extern void i915_fence_handler(struct drm_device *dev); @@ -353,7 +563,7 @@ extern void i915_invalidate_reported_sequence(struct drm_device *dev); #endif -#ifdef I915_HAVE_BUFFER +#if defined(I915_HAVE_BUFFER) && defined(I915_TTM) /* i915_buffer.c */ extern struct drm_ttm_backend *i915_create_ttm_backend_entry(struct drm_device *dev); extern int i915_fence_type(struct drm_buffer_object *bo, uint32_t *fclass, @@ -365,14 +575,88 @@ extern uint64_t i915_evict_flags(struct drm_buffer_object *bo); extern int i915_move(struct drm_buffer_object *bo, int evict, int no_wait, struct drm_bo_mem_reg *new_mem); void i915_flush_ttm(struct drm_ttm *ttm); +#endif /* ttm */ +#ifdef I915_HAVE_BUFFER /* i915_execbuf.c */ int i915_execbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv); - +/* i915_gem.c */ +int i915_gem_init_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_create_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pread_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pwrite_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_mmap_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_execbuffer(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_unpin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_busy_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_throttle_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_entervt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_set_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_get_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv); +void i915_gem_load(struct drm_device *dev); +int i915_gem_proc_init(struct drm_minor *minor); +void i915_gem_proc_cleanup(struct drm_minor *minor); +int i915_gem_init_object(struct drm_gem_object *obj); +void i915_gem_free_object(struct drm_gem_object *obj); +int i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment); +void i915_gem_object_unpin(struct drm_gem_object *obj); +void i915_gem_lastclose(struct drm_device *dev); +uint32_t i915_get_gem_seqno(struct drm_device *dev); +void i915_gem_retire_requests(struct drm_device *dev); +int i915_gem_init_ringbuffer(struct drm_device *dev); +void i915_gem_cleanup_ringbuffer(struct drm_device *dev); +int i915_gem_do_init(struct drm_device *dev, unsigned long start, + unsigned long end); +void i915_gem_retire_work_handler(struct work_struct *work); +void i915_gem_clflush_object(struct drm_gem_object *obj); #endif extern unsigned int i915_fbpercrtc; +/* i915_gem_tiling.c */ +void i915_gem_detect_bit_6_swizzle(struct drm_device *dev); + +/* i915_gem_debug.c */ +void i915_gem_dump_object(struct drm_gem_object *obj, int len, + const char *where, uint32_t mark); +#if WATCH_INACTIVE +void i915_verify_inactive(struct drm_device *dev, char *file, int line); +#else +#define i915_verify_inactive(dev,file,line) +#endif +void i915_gem_object_check_coherency(struct drm_gem_object *obj, int handle); +void i915_gem_dump_object(struct drm_gem_object *obj, int len, + const char *where, uint32_t mark); +void i915_dump_lru(struct drm_device *dev, const char *where); + +#ifdef __linux__ +/* i915_opregion.c */ +extern int intel_opregion_init(struct drm_device *dev); +extern void intel_opregion_free(struct drm_device *dev); +extern void opregion_asle_intr(struct drm_device *dev); +extern void opregion_enable_asle(struct drm_device *dev); +#endif + #ifdef __linux__ #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25) extern void intel_init_chipset_flush_compat(struct drm_device *dev); @@ -390,18 +674,35 @@ extern void intel_modeset_cleanup(struct drm_device *dev); #define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val)) #define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg)) #define I915_WRITE16(reg,val) DRM_WRITE16(dev_priv->mmio_map, (reg), (val)) +#define I915_READ8(reg) DRM_READ8(dev_priv->mmio_map, (reg)) +#define I915_WRITE8(reg,val) DRM_WRITE8(dev_priv->mmio_map, (reg), (val)) + +#if defined(__FreeBSD__) +typedef boolean_t bool; +#endif #define I915_VERBOSE 0 +#define I915_RING_VALIDATE 0 + +#define PRIMARY_RINGBUFFER_SIZE (128*1024) #define PRIMARY_RINGBUFFER_SIZE (128*1024) #define RING_LOCALS unsigned int outring, ringmask, outcount; \ volatile char *virt; +#if I915_RING_VALIDATE +void i915_ring_validate(struct drm_device *dev, const char *func, int line); +#define I915_RING_DO_VALIDATE(dev) i915_ring_validate(dev, __FUNCTION__, __LINE__) +#else +#define I915_RING_DO_VALIDATE(dev) +#endif + #define BEGIN_LP_RING(n) do { \ if (I915_VERBOSE) \ DRM_DEBUG("BEGIN_LP_RING(%d)\n", \ (n)); \ + I915_RING_DO_VALIDATE(dev); \ if (dev_priv->ring.space < (n)*4) \ i915_wait_ring(dev, (n)*4, __FUNCTION__); \ outcount = 0; \ @@ -420,18 +721,46 @@ extern void intel_modeset_cleanup(struct drm_device *dev); #define ADVANCE_LP_RING() do { \ if (I915_VERBOSE) DRM_DEBUG("ADVANCE_LP_RING %x\n", outring); \ + I915_RING_DO_VALIDATE(dev); \ dev_priv->ring.tail = outring; \ dev_priv->ring.space -= outcount * 4; \ I915_WRITE(PRB0_TAIL, outring); \ } while(0) +extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); + #define BREADCRUMB_BITS 31 #define BREADCRUMB_MASK ((1U << BREADCRUMB_BITS) - 1) #define READ_BREADCRUMB(dev_priv) (((volatile u32*)(dev_priv->hw_status_page))[5]) +/** + * Reads a dword out of the status page, which is written to from the command + * queue by automatic updates, MI_REPORT_HEAD, MI_STORE_DATA_INDEX, or + * MI_STORE_DATA_IMM. + * + * The following dwords have a reserved meaning: + * 0: ISR copy, updated when an ISR bit not set in the HWSTAM changes. + * 4: ring 0 head pointer + * 5: ring 1 head pointer (915-class) + * 6: ring 2 head pointer (915-class) + * + * The area from dword 0x10 to 0x3ff is available for driver usage. + */ #define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) +#define I915_GEM_HWS_INDEX 0x10 -extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); +/* MCH MMIO space */ +/** 915-945 and GM965 MCH register controlling DRAM channel access */ +#define DCC 0x200 +#define DCC_ADDRESSING_MODE_SINGLE_CHANNEL (0 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC (1 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED (2 << 0) +#define DCC_ADDRESSING_MODE_MASK (3 << 0) +#define DCC_CHANNEL_XOR_DISABLE (1 << 10) + +/** 965 MCH register controlling DRAM channel configuration */ +#define CHDECMISC 0x111 +#define CHDECMISC_FLEXMEMORY (1 << 1) /* * The Bridge device's PCI config space has information about the @@ -530,8 +859,10 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define MI_BATCH_BUFFER_END MI_INSTR(0x0a, 0) #define MI_REPORT_HEAD MI_INSTR(0x07, 0) #define MI_LOAD_SCAN_LINES_INCL MI_INSTR(0x12, 0) -#define MI_STORE_DWORD_IMM MI_INSTR(0x20, 1) /* used to have 1<<22? */ +#define MI_STORE_DWORD_IMM MI_INSTR(0x20, 1) +#define MI_MEM_VIRTUAL (1 << 22) /* 965+ only */ #define MI_STORE_DWORD_INDEX MI_INSTR(0x21, 1) +#define MI_STORE_DWORD_INDEX_SHIFT 2 #define MI_LOAD_REGISTER_IMM MI_INSTR(0x22, 1) #define MI_BATCH_BUFFER MI_INSTR(0x30, 1) #define MI_BATCH_NON_SECURE (1) @@ -543,6 +874,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); */ #define GFX_INSTR(opcode, flags) ((0x3 << 29) | ((opcode) << 24) | (flags)) +#define GFX_OP_USER_INTERRUPT ((0<<29)|(2<<23)) #define GFX_OP_RASTER_RULES ((0x3<<29)|(0x7<<24)) #define GFX_OP_SCISSOR ((0x3<<29)|(0x1c<<24)|(0x10<<19)) #define SC_UPDATE_SCISSOR (0x1<<1) @@ -560,6 +892,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define GFX_OP_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) #define GFX_OP_MAP_INFO ((0x3<<29)|(0x1d<<24)|0x4) #define GFX_OP_DESTBUFFER_VARS ((0x3<<29)|(0x1d<<24)|(0x85<<16)|0x0) +#define GFX_OP_DESTBUFFER_INFO ((0x3<<29)|(0x1d<<24)|(0x8e<<16)|1) #define GFX_OP_DRAWRECT_INFO ((0x3<<29)|(0x1d<<24)|(0x80<<16)|(0x3)) #define GFX_OP_DRAWRECT_INFO_I965 ((0x7900<<16)|0x2) #define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|4) @@ -572,8 +905,8 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define BLT_DEPTH_16_1555 (2<<24) #define BLT_DEPTH_32 (3<<24) #define BLT_ROP_GXCOPY (0xcc<<16) -#define XY_SRC_COPY_BLT_SRC_TILED (1<<15) -#define XY_SRC_COPY_BLT_DST_TILED (1<<11) +#define XY_SRC_COPY_BLT_SRC_TILED (1<<15) /* 965+ only */ +#define XY_SRC_COPY_BLT_DST_TILED (1<<11) /* 965+ only */ #define CMD_OP_DISPLAYBUFFER_INFO ((0x0<<29)|(0x14<<23)|2) #define ASYNC_FLIP (1<<22) #define DISPLAY_PLANE_A (0<<20) @@ -603,7 +936,10 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define PRB1_HEAD 0x02044 /* 915+ only */ #define PRB1_START 0x02048 /* 915+ only */ #define PRB1_CTL 0x0204c /* 915+ only */ +#define ACTHD_I965 0x02074 #define HWS_PGA 0x02080 +#define HWS_ADDRESS_MASK 0xfffff000 +#define HWS_START_ADDRESS_SHIFT 4 #define IPEIR 0x02088 #define NOPID 0x02094 #define HWSTAM 0x02098 @@ -628,10 +964,12 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define I915_DISPLAY_PIPE_B_EVENT_INTERRUPT (1<<4) #define I915_DEBUG_INTERRUPT (1<<2) #define I915_USER_INTERRUPT (1<<1) +#define I915_ASLE_INTERRUPT (1<<0) #define EIR 0x020b0 #define EMR 0x020b4 #define ESR 0x020b8 #define INSTPM 0x020c0 +#define ACTHD 0x020c8 #define FW_BLC 0x020d8 #define FW_BLC_SELF 0x020e0 /* 915+ only */ #define MI_ARB_STATE 0x020e4 /* 915+ only */ @@ -681,7 +1019,6 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); /* * GPIO regs */ - #define GPIOA 0x5010 #define GPIOB 0x5014 #define GPIOC 0x5018 @@ -790,12 +1127,6 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define ADPA_DPMS_STANDBY (2<<10) #define ADPA_DPMS_OFF (3<<10) -#define LP_RING 0x2030 -#define HP_RING 0x2040 -/* The binner has its own ring buffer: - */ -#define HWB_RING 0x2400 - #define RING_TAIL 0x00 #define TAIL_ADDR 0x001FFFF8 #define RING_HEAD 0x04 @@ -1821,11 +2152,18 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); (dev)->pci_device == 0x29A2 || \ (dev)->pci_device == 0x2A02 || \ (dev)->pci_device == 0x2A12 || \ - (dev)->pci_device == 0x2A42) + (dev)->pci_device == 0x2A42 || \ + (dev)->pci_device == 0x2E02 || \ + (dev)->pci_device == 0x2E12 || \ + (dev)->pci_device == 0x2E22) #define IS_I965GM(dev) ((dev)->pci_device == 0x2A02) -#define IS_IGD_GM(dev) ((dev)->pci_device == 0x2A42) +#define IS_GM45(dev) ((dev)->pci_device == 0x2A42) + +#define IS_G4X(dev) ((dev)->pci_device == 0x2E02 || \ + (dev)->pci_device == 0x2E12 || \ + (dev)->pci_device == 0x2E22) #define IS_G33(dev) ((dev)->pci_device == 0x29C2 || \ (dev)->pci_device == 0x29B2 || \ @@ -1835,8 +2173,8 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); IS_I945GM(dev) || IS_I965G(dev) || IS_G33(dev)) #define IS_MOBILE(dev) (IS_I830(dev) || IS_I85X(dev) || IS_I915GM(dev) || \ - IS_I945GM(dev) || IS_I965GM(dev) || IS_IGD_GM(dev)) + IS_I945GM(dev) || IS_I965GM(dev) || IS_GM45(dev)) -#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_IGD_GM(dev)) +#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_GM45(dev) || IS_G4X(dev)) #endif diff --git a/shared-core/i915_init.c b/shared-core/i915_init.c index f2c07fc6..4f2d3a4f 100644 --- a/shared-core/i915_init.c +++ b/shared-core/i915_init.c @@ -100,56 +100,22 @@ int i915_probe_agp(struct pci_dev *pdev, unsigned long *aperture_size, return 0; } -int i915_load_modeset_init(struct drm_device *dev) +static int i915_load_modeset_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; unsigned long agp_size, prealloc_size; - int size, ret = 0; + int ret = 0; i915_probe_agp(dev->pdev, &agp_size, &prealloc_size); - printk("setting up %ld bytes of VRAM space\n", prealloc_size); - printk("setting up %ld bytes of TT space\n", (agp_size - prealloc_size)); - - drm_bo_init_mm(dev, DRM_BO_MEM_VRAM, 0, prealloc_size >> PAGE_SHIFT, 1); - drm_bo_init_mm(dev, DRM_BO_MEM_TT, prealloc_size >> PAGE_SHIFT, - (agp_size - prealloc_size) >> PAGE_SHIFT, 1); - I915_WRITE(PRB0_CTL, 0); - I915_WRITE(PRB0_HEAD, 0); - I915_WRITE(PRB0_TAIL, 0); - - size = PRIMARY_RINGBUFFER_SIZE; - ret = drm_buffer_object_create(dev, size, drm_bo_type_kernel, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_NO_EVICT, - DRM_BO_HINT_DONT_FENCE, 0x1, 0, - &dev_priv->ring_buffer); - if (ret < 0) { - DRM_ERROR("Unable to allocate or pin ring buffer\n"); - goto clean_mm; - } - /* remap the buffer object properly */ - dev_priv->ring.Start = dev_priv->ring_buffer->offset; - dev_priv->ring.End = dev_priv->ring.Start + size; - dev_priv->ring.Size = size; - dev_priv->ring.tail_mask = dev_priv->ring.Size - 1; + /* Basic memrange allocator for stolen space (aka vram) */ + drm_mm_init(&dev_priv->vram, 0, prealloc_size); + /* Let GEM Manage from end of prealloc space to end of aperture */ + i915_gem_do_init(dev, prealloc_size, agp_size); - /* FIXME: need wrapper with PCI mem checks */ - ret = drm_mem_reg_ioremap(dev, &dev_priv->ring_buffer->mem, - (void **) &dev_priv->ring.virtual_start); - if (ret) { - DRM_ERROR("error mapping ring buffer: %d\n", ret); - goto destroy_ringbuffer; - } - - DRM_DEBUG("ring start %08lX, %p, %08lX\n", dev_priv->ring.Start, - dev_priv->ring.virtual_start, dev_priv->ring.Size); - - memset((void *)(dev_priv->ring.virtual_start), 0, dev_priv->ring.Size); - I915_WRITE(PRB0_START, dev_priv->ring.Start); - I915_WRITE(PRB0_CTL, ((dev_priv->ring.Size - 4096) & RING_NR_PAGES) | - (RING_NO_REPORT | RING_VALID)); + ret = i915_gem_init_ringbuffer(dev); + if (ret) + goto out; /* Allow hardware batchbuffers unless told otherwise. */ @@ -157,65 +123,11 @@ int i915_load_modeset_init(struct drm_device *dev) dev_priv->max_validate_buffers = I915_MAX_VALIDATE_BUFFERS; mutex_init(&dev_priv->cmdbuf_mutex); - /* Program Hardware Status Page */ - if (!IS_G33(dev)) { - dev_priv->status_page_dmah = - drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff); - - if (!dev_priv->status_page_dmah) { - DRM_ERROR("Can not allocate hardware status page\n"); - ret = -ENOMEM; - goto destroy_ringbuffer; - } - dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr; - dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; - - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - - I915_WRITE(HWS_PGA, dev_priv->dma_status_page); - } else { - size = 4 * 1024; - ret = drm_buffer_object_create(dev, size, - drm_bo_type_kernel, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_NO_EVICT, - DRM_BO_HINT_DONT_FENCE, 0x1, 0, - &dev_priv->hws_bo); - if (ret < 0) { - DRM_ERROR("Unable to allocate or pin hw status page\n"); - ret = -EINVAL; - goto destroy_ringbuffer; - } - - dev_priv->status_gfx_addr = - dev_priv->hws_bo->offset & (0x1ffff << 12); - dev_priv->hws_map.offset = dev->agp->base + - dev_priv->hws_bo->offset; - dev_priv->hws_map.size = size; - dev_priv->hws_map.type= 0; - dev_priv->hws_map.flags= 0; - dev_priv->hws_map.mtrr = 0; - - drm_core_ioremap(&dev_priv->hws_map, dev); - if (dev_priv->hws_map.handle == NULL) { - dev_priv->status_gfx_addr = 0; - DRM_ERROR("can not ioremap virtual addr for" - "G33 hw status page\n"); - ret = -ENOMEM; - goto destroy_hws; - } - dev_priv->hw_status_page = dev_priv->hws_map.handle; - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); - } - DRM_DEBUG("Enabled hardware status page\n"); - dev_priv->wq = create_singlethread_workqueue("i915"); if (dev_priv->wq == 0) { DRM_DEBUG("Error\n"); ret = -EINVAL; - goto destroy_hws; + goto destroy_ringbuffer; } ret = intel_init_bios(dev); @@ -228,9 +140,6 @@ int i915_load_modeset_init(struct drm_device *dev) intel_modeset_init(dev); drm_helper_initial_config(dev, false); - drm_mm_print(&dev->bm.man[DRM_BO_MEM_VRAM].manager, "VRAM"); - drm_mm_print(&dev->bm.man[DRM_BO_MEM_TT].manager, "TT"); - dev->devname = kstrdup(DRIVER_NAME, GFP_KERNEL); if (!dev->devname) { ret = -ENOMEM; @@ -248,26 +157,9 @@ modeset_cleanup: intel_modeset_cleanup(dev); destroy_wq: destroy_workqueue(dev_priv->wq); -destroy_hws: - if (!IS_G33(dev)) { - if (dev_priv->status_page_dmah) - drm_pci_free(dev, dev_priv->status_page_dmah); - } else { - if (dev_priv->hws_map.handle) - drm_core_ioremapfree(&dev_priv->hws_map, dev); - if (dev_priv->hws_bo) - drm_bo_usage_deref_unlocked(&dev_priv->hws_bo); - } - I915_WRITE(HWS_PGA, 0x1ffff000); destroy_ringbuffer: - if (dev_priv->ring.virtual_start) - drm_mem_reg_iounmap(dev, &dev_priv->ring_buffer->mem, - dev_priv->ring.virtual_start); - if (dev_priv->ring_buffer) - drm_bo_usage_deref_unlocked(&dev_priv->ring_buffer); -clean_mm: - drm_bo_clean_mm(dev, DRM_BO_MEM_VRAM, 1); - drm_bo_clean_mm(dev, DRM_BO_MEM_TT, 1); + i915_gem_cleanup_ringbuffer(dev); +out: return ret; } @@ -285,7 +177,8 @@ clean_mm: int i915_driver_load(struct drm_device *dev, unsigned long flags) { struct drm_i915_private *dev_priv; - int ret = 0; + int ret = 0, num_pipes = 2; + u32 tmp; dev_priv = drm_alloc(sizeof(struct drm_i915_private), DRM_MEM_DRIVER); if (dev_priv == NULL) @@ -293,7 +186,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) memset(dev_priv, 0, sizeof(struct drm_i915_private)); dev->dev_private = (void *)dev_priv; -// dev_priv->flags = flags; + dev_priv->dev = dev; /* i915 has 4 more counters */ dev->counters += 4; @@ -310,10 +203,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) if (IS_I965G(dev) || IS_G33(dev)) dev_priv->cursor_needs_physical = false; - if (IS_I9XX(dev)) { + if (IS_I9XX(dev)) pci_read_config_dword(dev->pdev, 0x5C, &dev_priv->stolen_base); - DRM_DEBUG("stolen base %p\n", (void*)dev_priv->stolen_base); - } if (IS_I9XX(dev)) { dev_priv->mmiobase = drm_get_resource_start(dev, 0); @@ -341,33 +232,65 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) goto free_priv; } + i915_gem_load(dev); + #ifdef __linux__ #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25) - intel_init_chipset_flush_compat(dev); + intel_init_chipset_flush_compat(dev); +#endif +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,25) + intel_opregion_init(dev); #endif #endif - if (drm_core_check_feature(dev, DRIVER_MODESET)) { - /* - * Initialize the memory manager for local and AGP space - */ - ret = drm_bo_driver_init(dev); - if (ret) { - DRM_ERROR("fail to init memory manager for " - "local & AGP space\n"); - goto out_rmmap; - } + tmp = I915_READ(PIPEASTAT); + I915_WRITE(PIPEASTAT, tmp); + tmp = I915_READ(PIPEBSTAT); + I915_WRITE(PIPEBSTAT, tmp); + + atomic_set(&dev_priv->irq_received, 0); + I915_WRITE(HWSTAM, 0xeffe); + I915_WRITE(IMR, 0x0); + I915_WRITE(IER, 0x0); + + DRM_SPININIT(&dev_priv->swaps_lock, "swap"); + INIT_LIST_HEAD(&dev_priv->vbl_swaps.head); + dev_priv->swaps_pending = 0; + + DRM_SPININIT(&dev_priv->user_irq_lock, "userirq"); + dev_priv->user_irq_refcount = 0; + dev_priv->irq_mask_reg = ~0; + + ret = drm_vblank_init(dev, num_pipes); + if (ret) + goto out_rmmap; + + ret = drm_hotplug_init(dev); + if (ret) + goto out_rmmap; + + dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B; + dev->max_vblank_count = 0xffffff; /* only 24 bits of frame count */ + i915_enable_interrupt(dev); + DRM_INIT_WAITQUEUE(&dev_priv->irq_queue); + + /* + * Initialize the hardware status page IRQ location. + */ + + I915_WRITE(INSTPM, (1 << 5) | (1 << 21)); + + if (drm_core_check_feature(dev, DRIVER_MODESET)) { ret = i915_load_modeset_init(dev); if (ret < 0) { DRM_ERROR("failed to init modeset\n"); - goto driver_fini; + goto out_rmmap; } } + return 0; -driver_fini: - drm_bo_driver_finish(dev); out_rmmap: drm_rmmap(dev, dev_priv->mmio_map); free_priv: @@ -379,6 +302,23 @@ int i915_driver_unload(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + u32 temp; + + dev_priv->vblank_pipe = 0; + + dev_priv->irq_enabled = 0; + + I915_WRITE(HWSTAM, 0xffffffff); + I915_WRITE(IMR, 0xffffffff); + I915_WRITE(IER, 0x0); + + temp = I915_READ(PIPEASTAT); + I915_WRITE(PIPEASTAT, temp); + temp = I915_READ(PIPEBSTAT); + I915_WRITE(PIPEBSTAT, temp); + temp = I915_READ(IIR); + I915_WRITE(IIR, temp); + I915_WRITE(PRB0_CTL, 0); if (drm_core_check_feature(dev, DRIVER_MODESET)) { @@ -392,6 +332,8 @@ int i915_driver_unload(struct drm_device *dev) drm_core_ioremapfree(&dev_priv->ring.map, dev); } #endif + +#ifdef DRI2 if (dev_priv->sarea_kmap.virtual) { drm_bo_kunmap(&dev_priv->sarea_kmap); dev_priv->sarea_kmap.virtual = NULL; @@ -404,53 +346,27 @@ int i915_driver_unload(struct drm_device *dev) mutex_unlock(&dev->struct_mutex); dev_priv->sarea_bo = NULL; } - - if (dev_priv->status_page_dmah) { - drm_pci_free(dev, dev_priv->status_page_dmah); - dev_priv->status_page_dmah = NULL; - dev_priv->hw_status_page = NULL; - dev_priv->dma_status_page = 0; - /* Need to rewrite hardware status page */ - I915_WRITE(HWS_PGA, 0x1ffff000); - } - - if (dev_priv->status_gfx_addr) { - dev_priv->status_gfx_addr = 0; - drm_core_ioremapfree(&dev_priv->hws_map, dev); - drm_bo_usage_deref_unlocked(&dev_priv->hws_bo); - I915_WRITE(HWS_PGA, 0x1ffff000); - } +#endif if (drm_core_check_feature(dev, DRIVER_MODESET)) { - drm_mem_reg_iounmap(dev, &dev_priv->ring_buffer->mem, - dev_priv->ring.virtual_start); - - DRM_DEBUG("usage is %d\n", atomic_read(&dev_priv->ring_buffer->usage)); mutex_lock(&dev->struct_mutex); - drm_bo_usage_deref_locked(&dev_priv->ring_buffer); - - if (drm_bo_clean_mm(dev, DRM_BO_MEM_TT, 1)) { - DRM_ERROR("Memory manager type 3 not clean. " - "Delaying takedown\n"); - } - if (drm_bo_clean_mm(dev, DRM_BO_MEM_VRAM, 1)) { - DRM_ERROR("Memory manager type 3 not clean. " - "Delaying takedown\n"); - } + i915_gem_cleanup_ringbuffer(dev); mutex_unlock(&dev->struct_mutex); + drm_mm_takedown(&dev_priv->vram); + i915_gem_lastclose(dev); } - drm_bo_driver_finish(dev); + drm_rmmap(dev, dev_priv->mmio_map); #ifdef __linux__ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,25) + intel_opregion_free(dev); +#endif #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25) - intel_init_chipset_flush_compat(dev); + intel_fini_chipset_flush_compat(dev); #endif #endif - DRM_DEBUG("%p\n", dev_priv->mmio_map); - drm_rmmap(dev, dev_priv->mmio_map); - drm_free(dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER); dev->dev_private = NULL; @@ -497,10 +413,36 @@ void i915_master_destroy(struct drm_device *dev, struct drm_master *master) master->driver_priv = NULL; } +int i915_driver_open(struct drm_device *dev, struct drm_file *file_priv) +{ + struct drm_i915_file_private *i915_file_priv; + + DRM_DEBUG("\n"); + i915_file_priv = (struct drm_i915_file_private *) + drm_alloc(sizeof(*i915_file_priv), DRM_MEM_FILES); + + if (!i915_file_priv) + return -ENOMEM; + + file_priv->driver_priv = i915_file_priv; + + i915_file_priv->mm.last_gem_seqno = 0; + i915_file_priv->mm.last_gem_throttle_seqno = 0; + + return 0; +} + +void i915_driver_postclose(struct drm_device *dev, struct drm_file *file_priv) +{ + struct drm_i915_file_private *i915_file_priv = file_priv->driver_priv; + + drm_free(i915_file_priv, sizeof(*i915_file_priv), DRM_MEM_FILES); +} + void i915_driver_preclose(struct drm_device * dev, struct drm_file *file_priv) { struct drm_i915_private *dev_priv = dev->dev_private; - if (drm_core_check_feature(dev, DRIVER_MODESET)) + if (!drm_core_check_feature(dev, DRIVER_MODESET)) i915_mem_release(dev, file_priv, dev_priv->agp_heap); } @@ -511,8 +453,33 @@ void i915_driver_lastclose(struct drm_device * dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) return; +#ifdef I915_HAVE_BUFFER + if (dev_priv->val_bufs) { + vfree(dev_priv->val_bufs); + dev_priv->val_bufs = NULL; + } +#endif + + i915_gem_lastclose(dev); + if (dev_priv->agp_heap) i915_mem_takedown(&(dev_priv->agp_heap)); + +#if defined(DRI2) + if (dev_priv->sarea_kmap.virtual) { + drm_bo_kunmap(&dev_priv->sarea_kmap); + dev_priv->sarea_kmap.virtual = NULL; + dev->control->master->lock.hw_lock = NULL; + dev->sigdata.lock = NULL; + } + + if (dev_priv->sarea_bo) { + mutex_lock(&dev->struct_mutex); + drm_bo_usage_deref_locked(&dev_priv->sarea_bo); + mutex_unlock(&dev->struct_mutex); + dev_priv->sarea_bo = NULL; + } +#endif i915_dma_cleanup(dev); } @@ -521,7 +488,8 @@ int i915_driver_firstopen(struct drm_device *dev) { if (drm_core_check_feature(dev, DRIVER_MODESET)) return 0; - +#if defined(I915_HAVE_BUFFER) && defined(I915_TTM) drm_bo_driver_init(dev); +#endif return 0; } diff --git a/shared-core/i915_irq.c b/shared-core/i915_irq.c index 2d355688..28fa35fd 100644 --- a/shared-core/i915_irq.c +++ b/shared-core/i915_irq.c @@ -35,6 +35,33 @@ #define MAX_NOPID ((u32)~0) +/* + * These are the interrupts used by the driver + */ +#define I915_INTERRUPT_ENABLE_MASK (I915_USER_INTERRUPT | \ + I915_DISPLAY_PIPE_A_EVENT_INTERRUPT | \ + I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) + +static inline void +i915_enable_irq(struct drm_i915_private *dev_priv, uint32_t mask) +{ + if ((dev_priv->irq_mask_reg & mask) != 0) { + dev_priv->irq_mask_reg &= ~mask; + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } +} + +static inline void +i915_disable_irq(struct drm_i915_private *dev_priv, uint32_t mask) +{ + if ((dev_priv->irq_mask_reg & mask) != mask) { + dev_priv->irq_mask_reg |= mask; + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } +} + /** * i915_get_pipe - return the the pipe associated with a given plane * @dev: DRM device @@ -165,11 +192,11 @@ static void i915_vblank_tasklet(struct drm_device *dev) u32 ropcpp = (0xcc << 16) | ((cpp - 1) << 24); RING_LOCALS; - if (sarea_priv->front_tiled) { + if (IS_I965G(dev) && sarea_priv->front_tiled) { cmd |= XY_SRC_COPY_BLT_DST_TILED; dst_pitch >>= 2; } - if (sarea_priv->back_tiled) { + if (IS_I965G(dev) && sarea_priv->back_tiled) { cmd |= XY_SRC_COPY_BLT_SRC_TILED; src_pitch >>= 2; } @@ -367,28 +394,7 @@ static void i915_vblank_tasklet(struct drm_device *dev) drm_free(swap_hit, sizeof(*swap_hit), DRM_MEM_DRIVER); } } -#if 0 -static int i915_in_vblank(struct drm_device *dev, int pipe) -{ - struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; - unsigned long pipedsl, vblank, vtotal; - unsigned long vbl_start, vbl_end, cur_line; - - pipedsl = pipe ? PIPEBDSL : PIPEADSL; - vblank = pipe ? VBLANK_B : VBLANK_A; - vtotal = pipe ? VTOTAL_B : VTOTAL_A; - - vbl_start = I915_READ(vblank) & VBLANK_START_MASK; - vbl_end = (I915_READ(vblank) >> VBLANK_END_SHIFT) & VBLANK_END_MASK; - - cur_line = I915_READ(pipedsl); - if (cur_line >= vbl_start) - return 1; - - return 0; -} -#endif u32 i915_get_vblank_counter(struct drm_device *dev, int plane) { struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; @@ -422,30 +428,17 @@ u32 i915_get_vblank_counter(struct drm_device *dev, int plane) count = (high1 << 8) | low; - /* - * If we're in the middle of the vblank period, the - * above regs won't have been updated yet, so return - * an incremented count to stay accurate - */ -#if 0 - if (i915_in_vblank(dev, pipe)) - count++; -#endif - /* count may be reset by other driver(e.g. 2D driver), - we have no way to know if it is wrapped or resetted - when count is zero. do a rough guess. - */ - if (count == 0 && dev->last_vblank[pipe] < dev->max_vblank_count/2) - dev->last_vblank[pipe] = 0; - return count; } static struct drm_device *hotplug_dev; -/* - * This code is called in a more safe envirmoent to handle the hotplugs. - * Add code here for hotplug love to userspace. +/** + * Handler for user interrupts in process context (able to sleep, do VFS + * operations, etc. + * + * If another IRQ comes in while we're in this handler, it will still get put + * on the queue again to be rerun when we finish. */ #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) static void i915_hotplug_work_func(void *work) @@ -485,7 +478,6 @@ static int i915_run_hotplug_tasklet(struct drm_device *dev, uint32_t stat) if (stat & SDVOC_HOTPLUG_INT_STATUS) { DRM_DEBUG("sDVOC event\n"); } - queue_work(dev_priv->wq, &hotplug); return 0; @@ -497,30 +489,19 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) struct drm_i915_master_private *master_priv; struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; u32 iir; - u32 pipea_stats = 0, pipeb_stats, tvdac; + u32 pipea_stats = 0, pipeb_stats = 0, tvdac; int hotplug = 0; int vblank = 0; - /* On i8xx/i915 hw the IIR and IER are 16bit on i9xx its 32bit */ - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - iir = I915_READ(IIR); - else - iir = I915_READ16(IIR); - - iir &= (dev_priv->irq_enable_reg | I915_USER_INTERRUPT); - -#if 0 - DRM_DEBUG("flag=%08x\n", iir); -#endif + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, ~0); + iir = I915_READ(IIR); + atomic_inc(&dev_priv->irq_received); if (iir == 0) { -#if 0 - DRM_DEBUG ("iir 0x%08x im 0x%08x ie 0x%08x pipea 0x%08x pipeb 0x%08x\n", - iir, - I915_READ(IMR), - I915_READ(IER), - I915_READ(PIPEASTAT), - I915_READ(PIPEBSTAT)); -#endif + if (dev->pdev->msi_enabled) { + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } return IRQ_NONE; } @@ -530,46 +511,29 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) */ if (iir & I915_DISPLAY_PIPE_A_EVENT_INTERRUPT) { pipea_stats = I915_READ(PIPEASTAT); - if (pipea_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| - PIPE_VBLANK_INTERRUPT_STATUS)) - { - vblank++; - drm_handle_vblank(dev, i915_get_plane(dev, 0)); - } - - /* This is a global event, and not a pipe A event */ - if (pipea_stats & PIPE_HOTPLUG_INTERRUPT_STATUS) - hotplug = 1; - - if (pipea_stats & PIPE_HOTPLUG_TV_INTERRUPT_STATUS) { - hotplug = 1; - /* Toggle hotplug detection to clear hotplug status */ - tvdac = I915_READ(TV_DAC); - I915_WRITE(TV_DAC, tvdac & ~TVDAC_STATE_CHG_EN); - I915_WRITE(TV_DAC, tvdac | TVDAC_STATE_CHG_EN); - } - I915_WRITE(PIPEASTAT, pipea_stats); } if (iir & I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) { pipeb_stats = I915_READ(PIPEBSTAT); - if (pipeb_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| - PIPE_VBLANK_INTERRUPT_STATUS)) - { - vblank++; - drm_handle_vblank(dev, i915_get_plane(dev, 1)); - } I915_WRITE(PIPEBSTAT, pipeb_stats); } - /* Clear the generated interrupt */ - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) { - I915_WRITE(IIR, iir); - (void) I915_READ(IIR); - } else { - I915_WRITE16(IIR, iir); - (void) I915_READ16(IIR); + I915_WRITE(IIR, iir); + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IIR); /* Flush posted writes */ + + /* This is a global event, and not a pipe A event */ + if (pipea_stats & PIPE_HOTPLUG_INTERRUPT_STATUS) + hotplug = 1; + + if (pipea_stats & PIPE_HOTPLUG_TV_INTERRUPT_STATUS) { + hotplug = 1; + /* Toggle hotplug detection to clear hotplug status */ + tvdac = I915_READ(TV_DAC); + I915_WRITE(TV_DAC, tvdac & ~TVDAC_STATE_CHG_EN); + I915_WRITE(TV_DAC, tvdac | TVDAC_STATE_CHG_EN); } if (dev->primary->master) { @@ -577,13 +541,41 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) master_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); } +#ifdef __linux__ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,25) + if ((iir & I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) && + (pipeb_stats & I915_LEGACY_BLC_EVENT_ENABLE)) + opregion_asle_intr(dev); + if (iir & I915_ASLE_INTERRUPT) + opregion_asle_intr(dev); +#endif +#endif + if (iir & I915_USER_INTERRUPT) { + dev_priv->mm.irq_gem_seqno = i915_get_gem_seqno(dev); DRM_WAKEUP(&dev_priv->irq_queue); #ifdef I915_HAVE_FENCE i915_fence_handler(dev); #endif } + if (pipea_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| + PIPE_VBLANK_INTERRUPT_STATUS)) { + vblank++; + drm_handle_vblank(dev, i915_get_plane(dev, 0)); + } + + /* The vblank interrupt gets enabled even if we didn't ask for + it, so make sure it's shut down again */ + if (!(dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B)) + pipeb_stats &= ~(I915_VBLANK_INTERRUPT_ENABLE); + + if (pipeb_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| + PIPE_VBLANK_INTERRUPT_STATUS)) { + vblank++; + drm_handle_vblank(dev, i915_get_plane(dev, 1)); + } + if (vblank) { if (dev_priv->swaps_pending > 0) drm_locked_tasklet(dev, i915_vblank_tasklet); @@ -635,15 +627,9 @@ void i915_user_irq_on(struct drm_device *dev) struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; DRM_SPINLOCK(&dev_priv->user_irq_lock); - if (dev_priv->irq_enabled && (++dev_priv->user_irq_refcount == 1)){ - dev_priv->irq_enable_reg |= I915_USER_INTERRUPT; - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - I915_WRITE(IER, dev_priv->irq_enable_reg); - else - I915_WRITE16(IER, dev_priv->irq_enable_reg); - } + if (dev_priv->irq_enabled && (++dev_priv->user_irq_refcount == 1)) + i915_enable_irq(dev_priv, I915_USER_INTERRUPT); DRM_SPINUNLOCK(&dev_priv->user_irq_lock); - } void i915_user_irq_off(struct drm_device *dev) @@ -651,28 +637,38 @@ void i915_user_irq_off(struct drm_device *dev) struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; DRM_SPINLOCK(&dev_priv->user_irq_lock); - if (dev_priv->irq_enabled && (--dev_priv->user_irq_refcount == 0)) { - // dev_priv->irq_enable_reg &= ~I915_USER_INTERRUPT; - // if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - // I915_WRITE(IER, dev_priv->irq_enable_reg); - // else - // I915_WRITE16(IER, dev_priv->irq_enable_reg); - } + BUG_ON(dev_priv->irq_enabled && dev_priv->user_irq_refcount <= 0); + if (dev_priv->irq_enabled && (--dev_priv->user_irq_refcount == 0)) + i915_disable_irq(dev_priv, I915_USER_INTERRUPT); DRM_SPINUNLOCK(&dev_priv->user_irq_lock); } -static int i915_wait_irq(struct drm_device * dev, int irq_nr) +int i915_wait_irq(struct drm_device * dev, int irq_nr) { struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; struct drm_i915_master_private *master_priv; int ret = 0; + if (!dev_priv) { + DRM_ERROR("called with no initialization\n"); + return -EINVAL; + } + DRM_DEBUG("irq_nr=%d breadcrumb=%d\n", irq_nr, READ_BREADCRUMB(dev_priv)); - if (READ_BREADCRUMB(dev_priv) >= irq_nr) + master_priv = dev->primary->master->driver_priv; + + if (!master_priv) { + DRM_ERROR("no master priv?\n"); + return -EINVAL; + } + + if (READ_BREADCRUMB(dev_priv) >= irq_nr) { + master_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); return 0; + } i915_user_irq_on(dev); DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ, @@ -684,10 +680,8 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) READ_BREADCRUMB(dev_priv), (int)dev_priv->counter); } - if (dev->primary->master) { - master_priv = dev->primary->master->driver_priv; + if (READ_BREADCRUMB(dev_priv) >= irq_nr) master_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - } return ret; } @@ -739,16 +733,17 @@ int i915_enable_vblank(struct drm_device *dev, int plane) struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; int pipe = i915_get_pipe(dev, plane); u32 pipestat_reg = 0; + u32 mask_reg = 0; u32 pipestat; switch (pipe) { case 0: pipestat_reg = PIPEASTAT; - dev_priv->irq_enable_reg |= I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; + mask_reg |= I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; break; case 1: pipestat_reg = PIPEBSTAT; - dev_priv->irq_enable_reg |= I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; + mask_reg |= I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; break; default: DRM_ERROR("tried to enable vblank on non-existent pipe %d\n", @@ -774,12 +769,9 @@ int i915_enable_vblank(struct drm_device *dev, int plane) PIPE_VBLANK_INTERRUPT_STATUS); I915_WRITE(pipestat_reg, pipestat); } - - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - I915_WRITE(IER, dev_priv->irq_enable_reg); - else - I915_WRITE16(IER, dev_priv->irq_enable_reg); - + DRM_SPINLOCK(&dev_priv->user_irq_lock); + i915_enable_irq(dev_priv, mask_reg); + DRM_SPINUNLOCK(&dev_priv->user_irq_lock); return 0; } @@ -789,16 +781,17 @@ void i915_disable_vblank(struct drm_device *dev, int plane) struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; int pipe = i915_get_pipe(dev, plane); u32 pipestat_reg = 0; + u32 mask_reg = 0; u32 pipestat; switch (pipe) { case 0: pipestat_reg = PIPEASTAT; - dev_priv->irq_enable_reg &= ~I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; + mask_reg |= I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; break; case 1: pipestat_reg = PIPEBSTAT; - dev_priv->irq_enable_reg &= ~I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; + mask_reg |= I915_DISPLAY_PIPE_B_EVENT_INTERRUPT; break; default: DRM_ERROR("tried to disable vblank on non-existent pipe %d\n", @@ -806,13 +799,11 @@ void i915_disable_vblank(struct drm_device *dev, int plane) break; } - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - I915_WRITE(IER, dev_priv->irq_enable_reg); - else - I915_WRITE16(IER, dev_priv->irq_enable_reg); + DRM_SPINLOCK(&dev_priv->user_irq_lock); + i915_disable_irq(dev_priv, mask_reg); + DRM_SPINUNLOCK(&dev_priv->user_irq_lock); - if (pipestat_reg) - { + if (pipestat_reg) { pipestat = I915_READ (pipestat_reg); pipestat &= ~(PIPE_START_VBLANK_INTERRUPT_ENABLE | PIPE_VBLANK_INTERRUPT_ENABLE); @@ -822,6 +813,7 @@ void i915_disable_vblank(struct drm_device *dev, int plane) pipestat |= (PIPE_START_VBLANK_INTERRUPT_STATUS | PIPE_VBLANK_INTERRUPT_STATUS); I915_WRITE(pipestat_reg, pipestat); + (void) I915_READ(pipestat_reg); } } @@ -829,15 +821,15 @@ void i915_enable_interrupt (struct drm_device *dev) { struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; struct drm_connector *o; - - dev_priv->irq_enable_reg |= I915_USER_INTERRUPT; + + dev_priv->irq_mask_reg &= ~0; if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) { if (dev->mode_config.num_connector) - dev_priv->irq_enable_reg |= I915_DISPLAY_PORT_INTERRUPT; + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PORT_INTERRUPT; } else { if (dev->mode_config.num_connector) - dev_priv->irq_enable_reg |= I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_A_EVENT_INTERRUPT; /* Enable global interrupts for hotplug - not a pipeA event */ I915_WRITE(PIPEASTAT, I915_READ(PIPEASTAT) | @@ -847,7 +839,8 @@ void i915_enable_interrupt (struct drm_device *dev) PIPE_HOTPLUG_INTERRUPT_STATUS); } - if (dev_priv->irq_enable_reg & (I915_DISPLAY_PORT_INTERRUPT | I915_DISPLAY_PIPE_A_EVENT_INTERRUPT)) { + if (!(dev_priv->irq_mask_reg & I915_DISPLAY_PORT_INTERRUPT) || + !(dev_priv->irq_mask_reg & I915_DISPLAY_PIPE_A_EVENT_INTERRUPT)) { u32 temp = 0; if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) { @@ -891,11 +884,11 @@ void i915_enable_interrupt (struct drm_device *dev) } } - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - I915_WRITE(IER, dev_priv->irq_enable_reg); - else - I915_WRITE16(IER, dev_priv->irq_enable_reg); - +#ifdef __linux__ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,25) + opregion_enable_asle(dev); +#endif +#endif dev_priv->irq_enabled = 1; } @@ -905,20 +898,12 @@ int i915_vblank_pipe_set(struct drm_device *dev, void *data, struct drm_file *file_priv) { struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_vblank_pipe *pipe = data; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - if (pipe->pipe & ~(DRM_I915_VBLANK_PIPE_A|DRM_I915_VBLANK_PIPE_B)) { - DRM_ERROR("called with invalid pipe 0x%x\n", pipe->pipe); - return -EINVAL; - } - - dev_priv->vblank_pipe = pipe->pipe; - return 0; } @@ -927,17 +912,15 @@ int i915_vblank_pipe_get(struct drm_device *dev, void *data, { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_vblank_pipe *pipe = data; - u16 flag; + u32 flag = 0; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) - flag = I915_READ(IER); - else - flag = I915_READ16(IER); + if (dev_priv->irq_enabled) + flag = ~dev_priv->irq_mask_reg; pipe->pipe = 0; if (flag & I915_DISPLAY_PIPE_A_EVENT_INTERRUPT) @@ -1010,7 +993,13 @@ int i915_vblank_swap(struct drm_device *dev, void *data, DRM_SPINUNLOCK_IRQRESTORE(&dev->drw_lock, irqflags); - drm_update_vblank_count(dev, pipe); + /* + * We take the ref here and put it when the swap actually completes + * in the tasklet. + */ + ret = drm_vblank_get(dev, pipe); + if (ret) + return ret; curseq = drm_vblank_count(dev, pipe); if (seqtype == _DRM_VBLANK_RELATIVE) @@ -1021,6 +1010,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, swap->sequence = curseq + 1; } else { DRM_DEBUG("Missed target sequence\n"); + drm_vblank_put(dev, pipe); return -EINVAL; } } @@ -1042,6 +1032,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, irqflags); DRM_DEBUG("Invalid drawable ID %d\n", swap->drawable); + drm_vblank_put(dev, pipe); return -EINVAL; } @@ -1049,6 +1040,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, DRM_SPINUNLOCK_IRQRESTORE(&dev->drw_lock, irqflags); + drm_vblank_put(dev, pipe); return 0; } } @@ -1072,6 +1064,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, if (dev_priv->swaps_pending >= 100) { DRM_DEBUG("Too many swaps queued\n"); + drm_vblank_put(dev, pipe); return -EBUSY; } @@ -1079,17 +1072,12 @@ int i915_vblank_swap(struct drm_device *dev, void *data, if (!vbl_swap) { DRM_ERROR("Failed to allocate memory to queue swap\n"); + drm_vblank_put(dev, pipe); return -ENOMEM; } DRM_DEBUG("\n"); - ret = drm_vblank_get(dev, pipe); - if (ret) { - drm_free(vbl_swap, sizeof(*vbl_swap), DRM_MEM_DRIVER); - return ret; - } - vbl_swap->drw_id = swap->drawable; vbl_swap->plane = plane; vbl_swap->sequence = swap->sequence; @@ -1113,33 +1101,16 @@ int i915_vblank_swap(struct drm_device *dev, void *data, */ void i915_driver_irq_preinstall(struct drm_device * dev) { - struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; - u32 tmp; - - tmp = I915_READ(PIPEASTAT); - I915_WRITE(PIPEASTAT, tmp); - tmp = I915_READ(PIPEBSTAT); - I915_WRITE(PIPEBSTAT, tmp); - + struct drm_i915_private *dev_priv = dev->dev_private; I915_WRITE16(HWSTAM, 0xeffe); - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) { - I915_WRITE(IMR, 0x0); - I915_WRITE(IER, 0x0); - tmp = I915_READ(IIR); - I915_WRITE(IIR, tmp); - } else { - I915_WRITE16(IMR, 0x0); - I915_WRITE16(IER, 0x0); - tmp = I915_READ16(IIR); - I915_WRITE16(IIR, tmp); - } - + I915_WRITE16(IMR, 0x0); + I915_WRITE16(IER, 0x0); } int i915_driver_irq_postinstall(struct drm_device * dev) { - struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; + struct drm_i915_private *dev_priv = dev->dev_private; int ret, num_pipes = 2; DRM_SPININIT(&dev_priv->swaps_lock, "swap"); @@ -1148,16 +1119,13 @@ int i915_driver_irq_postinstall(struct drm_device * dev) DRM_SPININIT(&dev_priv->user_irq_lock, "userirq"); dev_priv->user_irq_refcount = 0; - dev_priv->irq_enable_reg = 0; + dev_priv->irq_mask_reg = ~0; ret = drm_vblank_init(dev, num_pipes); if (ret) return ret; - ret = drm_hotplug_init(dev); - if (ret) - return ret; - + dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B; dev->max_vblank_count = 0xffffff; /* only 24 bits of frame count */ i915_enable_interrupt(dev); @@ -1173,31 +1141,23 @@ int i915_driver_irq_postinstall(struct drm_device * dev) void i915_driver_irq_uninstall(struct drm_device * dev) { - struct drm_i915_private *dev_priv = (struct drm_i915_private *) dev->dev_private; + struct drm_i915_private *dev_priv = dev->dev_private; u32 temp; if (!dev_priv) return; + dev_priv->vblank_pipe = 0; + dev_priv->irq_enabled = 0; + I915_WRITE(HWSTAM, 0xffffffff); + I915_WRITE(IMR, 0xffffffff); + I915_WRITE(IER, 0x0); temp = I915_READ(PIPEASTAT); I915_WRITE(PIPEASTAT, temp); temp = I915_READ(PIPEBSTAT); I915_WRITE(PIPEBSTAT, temp); - if (IS_I9XX(dev) && !IS_I915G(dev) && !IS_I915GM(dev)) { - I915_WRITE(HWSTAM, 0xffffffff); - I915_WRITE(IMR, 0xffffffff); - I915_WRITE(IER, 0x0); - - temp = I915_READ(IIR); - I915_WRITE(IIR, temp); - } else { - I915_WRITE16(HWSTAM, 0xffff); - I915_WRITE16(IMR, 0xffff); - I915_WRITE16(IER, 0x0); - - temp = I915_READ16(IIR); - I915_WRITE16(IIR, temp); - } + temp = I915_READ(IIR); + I915_WRITE(IIR, temp); } diff --git a/shared-core/i915_suspend.c b/shared-core/i915_suspend.c new file mode 100644 index 00000000..5e759b15 --- /dev/null +++ b/shared-core/i915_suspend.c @@ -0,0 +1,520 @@ +/* i915_suspend.c -- i830,i845,i855,i865,i915 driver -*- linux-c -*- + */ +/* + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * 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 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. + * + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" + +static bool i915_pipe_enabled(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (pipe == PIPE_A) + return (I915_READ(DPLL_A) & DPLL_VCO_ENABLE); + else + return (I915_READ(DPLL_B) & DPLL_VCO_ENABLE); +} + +static void i915_save_palette(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); + u32 *array; + int i; + + if (!i915_pipe_enabled(dev, pipe)) + return; + + if (pipe == PIPE_A) + array = dev_priv->save_palette_a; + else + array = dev_priv->save_palette_b; + + for(i = 0; i < 256; i++) + array[i] = I915_READ(reg + (i << 2)); +} + +static void i915_restore_palette(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); + u32 *array; + int i; + + if (!i915_pipe_enabled(dev, pipe)) + return; + + if (pipe == PIPE_A) + array = dev_priv->save_palette_a; + else + array = dev_priv->save_palette_b; + + for(i = 0; i < 256; i++) + I915_WRITE(reg + (i << 2), array[i]); +} + +static u8 i915_read_indexed(struct drm_device *dev, u16 index_port, u16 data_port, u8 reg) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE8(index_port, reg); + return I915_READ8(data_port); +} + +static u8 i915_read_ar(struct drm_device *dev, u16 st01, u8 reg, u16 palette_enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, palette_enable | reg); + return I915_READ8(VGA_AR_DATA_READ); +} + +static void i915_write_ar(struct drm_device *dev, u16 st01, u8 reg, u8 val, u16 palette_enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, palette_enable | reg); + I915_WRITE8(VGA_AR_DATA_WRITE, val); +} + +static void i915_write_indexed(struct drm_device *dev, u16 index_port, u16 data_port, u8 reg, u8 val) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE8(index_port, reg); + I915_WRITE8(data_port, val); +} + +static void i915_save_vga(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + u16 cr_index, cr_data, st01; + + /* VGA color palette registers */ + dev_priv->saveDACMASK = I915_READ8(VGA_DACMASK); + /* DACCRX automatically increments during read */ + I915_WRITE8(VGA_DACRX, 0); + /* Read 3 bytes of color data from each index */ + for (i = 0; i < 256 * 3; i++) + dev_priv->saveDACDATA[i] = I915_READ8(VGA_DACDATA); + + /* MSR bits */ + dev_priv->saveMSR = I915_READ8(VGA_MSR_READ); + if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { + cr_index = VGA_CR_INDEX_CGA; + cr_data = VGA_CR_DATA_CGA; + st01 = VGA_ST01_CGA; + } else { + cr_index = VGA_CR_INDEX_MDA; + cr_data = VGA_CR_DATA_MDA; + st01 = VGA_ST01_MDA; + } + + /* CRT controller regs */ + i915_write_indexed(dev, cr_index, cr_data, 0x11, + i915_read_indexed(dev, cr_index, cr_data, 0x11) & + (~0x80)); + for (i = 0; i <= 0x24; i++) + dev_priv->saveCR[i] = + i915_read_indexed(dev, cr_index, cr_data, i); + /* Make sure we don't turn off CR group 0 writes */ + dev_priv->saveCR[0x11] &= ~0x80; + + /* Attribute controller registers */ + I915_READ8(st01); + dev_priv->saveAR_INDEX = I915_READ8(VGA_AR_INDEX); + for (i = 0; i <= 0x14; i++) + dev_priv->saveAR[i] = i915_read_ar(dev, st01, i, 0); + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, dev_priv->saveAR_INDEX); + I915_READ8(st01); + + /* Graphics controller registers */ + for (i = 0; i < 9; i++) + dev_priv->saveGR[i] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, i); + + dev_priv->saveGR[0x10] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x10); + dev_priv->saveGR[0x11] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x11); + dev_priv->saveGR[0x18] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x18); + + /* Sequencer registers */ + for (i = 0; i < 8; i++) + dev_priv->saveSR[i] = + i915_read_indexed(dev, VGA_SR_INDEX, VGA_SR_DATA, i); +} + +static void i915_restore_vga(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + u16 cr_index, cr_data, st01; + + /* MSR bits */ + I915_WRITE8(VGA_MSR_WRITE, dev_priv->saveMSR); + if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { + cr_index = VGA_CR_INDEX_CGA; + cr_data = VGA_CR_DATA_CGA; + st01 = VGA_ST01_CGA; + } else { + cr_index = VGA_CR_INDEX_MDA; + cr_data = VGA_CR_DATA_MDA; + st01 = VGA_ST01_MDA; + } + + /* Sequencer registers, don't write SR07 */ + for (i = 0; i < 7; i++) + i915_write_indexed(dev, VGA_SR_INDEX, VGA_SR_DATA, i, + dev_priv->saveSR[i]); + + /* CRT controller regs */ + /* Enable CR group 0 writes */ + i915_write_indexed(dev, cr_index, cr_data, 0x11, dev_priv->saveCR[0x11]); + for (i = 0; i <= 0x24; i++) + i915_write_indexed(dev, cr_index, cr_data, i, dev_priv->saveCR[i]); + + /* Graphics controller regs */ + for (i = 0; i < 9; i++) + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, i, + dev_priv->saveGR[i]); + + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x10, + dev_priv->saveGR[0x10]); + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x11, + dev_priv->saveGR[0x11]); + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x18, + dev_priv->saveGR[0x18]); + + /* Attribute controller registers */ + I915_READ8(st01); /* switch back to index mode */ + for (i = 0; i <= 0x14; i++) + i915_write_ar(dev, st01, i, dev_priv->saveAR[i], 0); + I915_READ8(st01); /* switch back to index mode */ + I915_WRITE8(VGA_AR_INDEX, dev_priv->saveAR_INDEX | 0x20); + I915_READ8(st01); + + /* VGA color palette registers */ + I915_WRITE8(VGA_DACMASK, dev_priv->saveDACMASK); + /* DACCRX automatically increments during read */ + I915_WRITE8(VGA_DACWX, 0); + /* Read 3 bytes of color data from each index */ + for (i = 0; i < 256 * 3; i++) + I915_WRITE8(VGA_DACDATA, dev_priv->saveDACDATA[i]); + +} + +int i915_save_state(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + +#if defined(__FreeBSD__) + dev_priv->saveLBB = (u8) pci_read_config(dev->device, LBB, 1); +#else + pci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB); +#endif + + /* Display arbitration control */ + dev_priv->saveDSPARB = I915_READ(DSPARB); + + /* Pipe & plane A info */ + dev_priv->savePIPEACONF = I915_READ(PIPEACONF); + dev_priv->savePIPEASRC = I915_READ(PIPEASRC); + dev_priv->saveFPA0 = I915_READ(FPA0); + dev_priv->saveFPA1 = I915_READ(FPA1); + dev_priv->saveDPLL_A = I915_READ(DPLL_A); + if (IS_I965G(dev)) + dev_priv->saveDPLL_A_MD = I915_READ(DPLL_A_MD); + dev_priv->saveHTOTAL_A = I915_READ(HTOTAL_A); + dev_priv->saveHBLANK_A = I915_READ(HBLANK_A); + dev_priv->saveHSYNC_A = I915_READ(HSYNC_A); + dev_priv->saveVTOTAL_A = I915_READ(VTOTAL_A); + dev_priv->saveVBLANK_A = I915_READ(VBLANK_A); + dev_priv->saveVSYNC_A = I915_READ(VSYNC_A); + dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); + + dev_priv->saveDSPACNTR = I915_READ(DSPACNTR); + dev_priv->saveDSPASTRIDE = I915_READ(DSPASTRIDE); + dev_priv->saveDSPASIZE = I915_READ(DSPASIZE); + dev_priv->saveDSPAPOS = I915_READ(DSPAPOS); + dev_priv->saveDSPAADDR = I915_READ(DSPAADDR); + if (IS_I965G(dev)) { + dev_priv->saveDSPASURF = I915_READ(DSPASURF); + dev_priv->saveDSPATILEOFF = I915_READ(DSPATILEOFF); + } + i915_save_palette(dev, PIPE_A); + dev_priv->savePIPEASTAT = I915_READ(PIPEASTAT); + + /* Pipe & plane B info */ + dev_priv->savePIPEBCONF = I915_READ(PIPEBCONF); + dev_priv->savePIPEBSRC = I915_READ(PIPEBSRC); + dev_priv->saveFPB0 = I915_READ(FPB0); + dev_priv->saveFPB1 = I915_READ(FPB1); + dev_priv->saveDPLL_B = I915_READ(DPLL_B); + if (IS_I965G(dev)) + dev_priv->saveDPLL_B_MD = I915_READ(DPLL_B_MD); + dev_priv->saveHTOTAL_B = I915_READ(HTOTAL_B); + dev_priv->saveHBLANK_B = I915_READ(HBLANK_B); + dev_priv->saveHSYNC_B = I915_READ(HSYNC_B); + dev_priv->saveVTOTAL_B = I915_READ(VTOTAL_B); + dev_priv->saveVBLANK_B = I915_READ(VBLANK_B); + dev_priv->saveVSYNC_B = I915_READ(VSYNC_B); + dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); + + dev_priv->saveDSPBCNTR = I915_READ(DSPBCNTR); + dev_priv->saveDSPBSTRIDE = I915_READ(DSPBSTRIDE); + dev_priv->saveDSPBSIZE = I915_READ(DSPBSIZE); + dev_priv->saveDSPBPOS = I915_READ(DSPBPOS); + dev_priv->saveDSPBADDR = I915_READ(DSPBADDR); + if (IS_I965GM(dev) || IS_GM45(dev)) { + dev_priv->saveDSPBSURF = I915_READ(DSPBSURF); + dev_priv->saveDSPBTILEOFF = I915_READ(DSPBTILEOFF); + } + i915_save_palette(dev, PIPE_B); + dev_priv->savePIPEBSTAT = I915_READ(PIPEBSTAT); + + /* CRT state */ + dev_priv->saveADPA = I915_READ(ADPA); + + /* LVDS state */ + dev_priv->savePP_CONTROL = I915_READ(PP_CONTROL); + dev_priv->savePFIT_PGM_RATIOS = I915_READ(PFIT_PGM_RATIOS); + dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_CTL); + if (IS_I965G(dev)) + dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_CTL2); + if (IS_MOBILE(dev) && !IS_I830(dev)) + dev_priv->saveLVDS = I915_READ(LVDS); + if (!IS_I830(dev) && !IS_845G(dev)) + dev_priv->savePFIT_CONTROL = I915_READ(PFIT_CONTROL); + dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS); + dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS); + dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR); + + /* FIXME: save TV & SDVO state */ + + /* FBC state */ + dev_priv->saveFBC_CFB_BASE = I915_READ(FBC_CFB_BASE); + dev_priv->saveFBC_LL_BASE = I915_READ(FBC_LL_BASE); + dev_priv->saveFBC_CONTROL2 = I915_READ(FBC_CONTROL2); + dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL); + + /* Interrupt state */ + dev_priv->saveIIR = I915_READ(IIR); + dev_priv->saveIER = I915_READ(IER); + dev_priv->saveIMR = I915_READ(IMR); + + /* VGA state */ + dev_priv->saveVGA0 = I915_READ(VGA0); + dev_priv->saveVGA1 = I915_READ(VGA1); + dev_priv->saveVGA_PD = I915_READ(VGA_PD); + dev_priv->saveVGACNTRL = I915_READ(VGACNTRL); + + /* Clock gating state */ + dev_priv->saveD_STATE = I915_READ(D_STATE); + dev_priv->saveCG_2D_DIS = I915_READ(CG_2D_DIS); + + /* Cache mode state */ + dev_priv->saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0); + + /* Memory Arbitration state */ + dev_priv->saveMI_ARB_STATE = I915_READ(MI_ARB_STATE); + + /* Scratch space */ + for (i = 0; i < 16; i++) { + dev_priv->saveSWF0[i] = I915_READ(SWF00 + (i << 2)); + dev_priv->saveSWF1[i] = I915_READ(SWF10 + (i << 2)); + } + for (i = 0; i < 3; i++) + dev_priv->saveSWF2[i] = I915_READ(SWF30 + (i << 2)); + + i915_save_vga(dev); + + return 0; +} + +int i915_restore_state(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + +#if defined(__FreeBSD__) + pci_write_config(dev->device, LBB, dev_priv->saveLBB, 1); +#else + pci_write_config_byte(dev->pdev, LBB, dev_priv->saveLBB); +#endif + + I915_WRITE(DSPARB, dev_priv->saveDSPARB); + + /* Pipe & plane A info */ + /* Prime the clock */ + if (dev_priv->saveDPLL_A & DPLL_VCO_ENABLE) { + I915_WRITE(DPLL_A, dev_priv->saveDPLL_A & + ~DPLL_VCO_ENABLE); + DRM_UDELAY(150); + } + I915_WRITE(FPA0, dev_priv->saveFPA0); + I915_WRITE(FPA1, dev_priv->saveFPA1); + /* Actually enable it */ + I915_WRITE(DPLL_A, dev_priv->saveDPLL_A); + DRM_UDELAY(150); + if (IS_I965G(dev)) + I915_WRITE(DPLL_A_MD, dev_priv->saveDPLL_A_MD); + DRM_UDELAY(150); + + /* Restore mode */ + I915_WRITE(HTOTAL_A, dev_priv->saveHTOTAL_A); + I915_WRITE(HBLANK_A, dev_priv->saveHBLANK_A); + I915_WRITE(HSYNC_A, dev_priv->saveHSYNC_A); + I915_WRITE(VTOTAL_A, dev_priv->saveVTOTAL_A); + I915_WRITE(VBLANK_A, dev_priv->saveVBLANK_A); + I915_WRITE(VSYNC_A, dev_priv->saveVSYNC_A); + I915_WRITE(BCLRPAT_A, dev_priv->saveBCLRPAT_A); + + /* Restore plane info */ + I915_WRITE(DSPASIZE, dev_priv->saveDSPASIZE); + I915_WRITE(DSPAPOS, dev_priv->saveDSPAPOS); + I915_WRITE(PIPEASRC, dev_priv->savePIPEASRC); + I915_WRITE(DSPAADDR, dev_priv->saveDSPAADDR); + I915_WRITE(DSPASTRIDE, dev_priv->saveDSPASTRIDE); + if (IS_I965G(dev)) { + I915_WRITE(DSPASURF, dev_priv->saveDSPASURF); + I915_WRITE(DSPATILEOFF, dev_priv->saveDSPATILEOFF); + } + + I915_WRITE(PIPEACONF, dev_priv->savePIPEACONF); + + i915_restore_palette(dev, PIPE_A); + /* Enable the plane */ + I915_WRITE(DSPACNTR, dev_priv->saveDSPACNTR); + I915_WRITE(DSPAADDR, I915_READ(DSPAADDR)); + + /* Pipe & plane B info */ + if (dev_priv->saveDPLL_B & DPLL_VCO_ENABLE) { + I915_WRITE(DPLL_B, dev_priv->saveDPLL_B & + ~DPLL_VCO_ENABLE); + DRM_UDELAY(150); + } + I915_WRITE(FPB0, dev_priv->saveFPB0); + I915_WRITE(FPB1, dev_priv->saveFPB1); + /* Actually enable it */ + I915_WRITE(DPLL_B, dev_priv->saveDPLL_B); + DRM_UDELAY(150); + if (IS_I965G(dev)) + I915_WRITE(DPLL_B_MD, dev_priv->saveDPLL_B_MD); + DRM_UDELAY(150); + + /* Restore mode */ + I915_WRITE(HTOTAL_B, dev_priv->saveHTOTAL_B); + I915_WRITE(HBLANK_B, dev_priv->saveHBLANK_B); + I915_WRITE(HSYNC_B, dev_priv->saveHSYNC_B); + I915_WRITE(VTOTAL_B, dev_priv->saveVTOTAL_B); + I915_WRITE(VBLANK_B, dev_priv->saveVBLANK_B); + I915_WRITE(VSYNC_B, dev_priv->saveVSYNC_B); + I915_WRITE(BCLRPAT_B, dev_priv->saveBCLRPAT_B); + + /* Restore plane info */ + I915_WRITE(DSPBSIZE, dev_priv->saveDSPBSIZE); + I915_WRITE(DSPBPOS, dev_priv->saveDSPBPOS); + I915_WRITE(PIPEBSRC, dev_priv->savePIPEBSRC); + I915_WRITE(DSPBADDR, dev_priv->saveDSPBADDR); + I915_WRITE(DSPBSTRIDE, dev_priv->saveDSPBSTRIDE); + if (IS_I965G(dev)) { + I915_WRITE(DSPBSURF, dev_priv->saveDSPBSURF); + I915_WRITE(DSPBTILEOFF, dev_priv->saveDSPBTILEOFF); + } + + I915_WRITE(PIPEBCONF, dev_priv->savePIPEBCONF); + + i915_restore_palette(dev, PIPE_B); + /* Enable the plane */ + I915_WRITE(DSPBCNTR, dev_priv->saveDSPBCNTR); + I915_WRITE(DSPBADDR, I915_READ(DSPBADDR)); + + /* CRT state */ + I915_WRITE(ADPA, dev_priv->saveADPA); + + /* LVDS state */ + if (IS_I965G(dev)) + I915_WRITE(BLC_PWM_CTL2, dev_priv->saveBLC_PWM_CTL2); + if (IS_MOBILE(dev) && !IS_I830(dev)) + I915_WRITE(LVDS, dev_priv->saveLVDS); + if (!IS_I830(dev) && !IS_845G(dev)) + I915_WRITE(PFIT_CONTROL, dev_priv->savePFIT_CONTROL); + + I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS); + I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL); + I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS); + I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS); + I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR); + I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL); + + /* FIXME: restore TV & SDVO state */ + + /* FBC info */ + I915_WRITE(FBC_CFB_BASE, dev_priv->saveFBC_CFB_BASE); + I915_WRITE(FBC_LL_BASE, dev_priv->saveFBC_LL_BASE); + I915_WRITE(FBC_CONTROL2, dev_priv->saveFBC_CONTROL2); + I915_WRITE(FBC_CONTROL, dev_priv->saveFBC_CONTROL); + + /* VGA state */ + I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL); + I915_WRITE(VGA0, dev_priv->saveVGA0); + I915_WRITE(VGA1, dev_priv->saveVGA1); + I915_WRITE(VGA_PD, dev_priv->saveVGA_PD); + DRM_UDELAY(150); + + /* Clock gating state */ + I915_WRITE (D_STATE, dev_priv->saveD_STATE); + I915_WRITE (CG_2D_DIS, dev_priv->saveCG_2D_DIS); + + /* Cache mode state */ + I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000); + + /* Memory arbitration state */ + I915_WRITE (MI_ARB_STATE, dev_priv->saveMI_ARB_STATE | 0xffff0000); + + for (i = 0; i < 16; i++) { + I915_WRITE(SWF00 + (i << 2), dev_priv->saveSWF0[i]); + I915_WRITE(SWF10 + (i << 2), dev_priv->saveSWF1[i+7]); + } + for (i = 0; i < 3; i++) + I915_WRITE(SWF30 + (i << 2), dev_priv->saveSWF2[i]); + + i915_restore_vga(dev); + + return 0; +} + diff --git a/shared-core/nouveau_fifo.c b/shared-core/nouveau_fifo.c index 085336af..5ec2bc3e 100644 --- a/shared-core/nouveau_fifo.c +++ b/shared-core/nouveau_fifo.c @@ -131,7 +131,7 @@ int nouveau_fifo_init(struct drm_device *dev) /* No cmdbuf object */ NV_WRITE(NV04_PFIFO_CACHE1_DMA_INSTANCE, 0x00000000); NV_WRITE(NV03_PFIFO_CACHE0_PUSH0, 0x00000000); - NV_WRITE(NV03_PFIFO_CACHE0_PULL0, 0x00000000); + NV_WRITE(NV04_PFIFO_CACHE0_PULL0, 0x00000000); NV_WRITE(NV04_PFIFO_SIZE, 0x0000FFFF); NV_WRITE(NV04_PFIFO_CACHE1_HASH, 0x0000FFFF); NV_WRITE(NV04_PFIFO_CACHE0_PULL1, 0x00000001); diff --git a/shared-core/nouveau_mem.c b/shared-core/nouveau_mem.c index 375463b4..1078a9c5 100644 --- a/shared-core/nouveau_mem.c +++ b/shared-core/nouveau_mem.c @@ -36,7 +36,6 @@ #include "nouveau_drv.h" #include "nv50_kms_wrapper.h" - static struct mem_block * split_block(struct mem_block *p, uint64_t start, uint64_t size, struct drm_file *file_priv) diff --git a/shared-core/nv04_fifo.c b/shared-core/nv04_fifo.c index 230c8e72..88186fe3 100644 --- a/shared-core/nv04_fifo.c +++ b/shared-core/nv04_fifo.c @@ -122,8 +122,8 @@ nv04_fifo_save_context(struct nouveau_channel *chan) struct drm_nouveau_private *dev_priv = dev->dev_private; uint32_t tmp; - RAMFC_WR(DMA_PUT, NV04_PFIFO_CACHE1_DMA_PUT); - RAMFC_WR(DMA_GET, NV04_PFIFO_CACHE1_DMA_GET); + RAMFC_WR(DMA_PUT, NV_READ(NV04_PFIFO_CACHE1_DMA_PUT)); + RAMFC_WR(DMA_GET, NV_READ(NV04_PFIFO_CACHE1_DMA_GET)); tmp = NV_READ(NV04_PFIFO_CACHE1_DMA_DCOUNT) << 16; tmp |= NV_READ(NV04_PFIFO_CACHE1_DMA_INSTANCE); diff --git a/shared-core/nv40_graph.c b/shared-core/nv40_graph.c index 2540fc5f..de178f57 100644 --- a/shared-core/nv40_graph.c +++ b/shared-core/nv40_graph.c @@ -1544,6 +1544,8 @@ nv40_graph_transfer_context(struct drm_device *dev, uint32_t inst, int save) tmp |= NV40_PGRAPH_CTXCTL_0304_XFER_CTX; NV_WRITE(NV40_PGRAPH_CTXCTL_0304, tmp); + nouveau_wait_for_idle(dev); + for (i = 0; i < tv; i++) { if (NV_READ(NV40_PGRAPH_CTXCTL_030C) == 0) break; @@ -1565,9 +1567,7 @@ nv40_graph_transfer_context(struct drm_device *dev, uint32_t inst, int save) return 0; } -/* Save current context (from PGRAPH) into the channel's context - *XXX: fails sometimes, not sure why.. - */ +/* Save current context (from PGRAPH) into the channel's context */ int nv40_graph_save_context(struct nouveau_channel *chan) { @@ -1581,9 +1581,7 @@ nv40_graph_save_context(struct nouveau_channel *chan) return nv40_graph_transfer_context(dev, inst, 1); } -/* Restore the context for a specific channel into PGRAPH - * XXX: fails sometimes.. not sure why - */ +/* Restore the context for a specific channel into PGRAPH */ int nv40_graph_load_context(struct nouveau_channel *chan) { diff --git a/shared-core/r300_cmdbuf.c b/shared-core/r300_cmdbuf.c index bff6378f..05a76ef1 100644 --- a/shared-core/r300_cmdbuf.c +++ b/shared-core/r300_cmdbuf.c @@ -77,6 +77,9 @@ static int r300_emit_cliprects(drm_radeon_private_t *dev_priv, return -EFAULT; } + box.x2--; /* Hardware expects inclusive bottom-right corner */ + box.y2--; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) { box.x1 = (box.x1) & R300_CLIPRECT_MASK; @@ -95,8 +98,8 @@ static int r300_emit_cliprects(drm_radeon_private_t *dev_priv, R300_CLIPRECT_MASK; box.y2 = (box.y2 + R300_CLIPRECT_OFFSET) & R300_CLIPRECT_MASK; - } + OUT_RING((box.x1 << R300_CLIPRECT_X_SHIFT) | (box.y1 << R300_CLIPRECT_Y_SHIFT)); OUT_RING((box.x2 << R300_CLIPRECT_X_SHIFT) | @@ -136,6 +139,18 @@ static int r300_emit_cliprects(drm_radeon_private_t *dev_priv, ADVANCE_RING(); } + /* flus cache and wait idle clean after cliprect change */ + BEGIN_RING(2); + OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); + OUT_RING(R300_RB3D_DC_FLUSH); + ADVANCE_RING(); + BEGIN_RING(2); + OUT_RING(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); + OUT_RING(RADEON_WAIT_3D_IDLECLEAN); + ADVANCE_RING(); + /* set flush flag */ + dev_priv->track_flush |= RADEON_FLUSH_EMITED; + return 0; } @@ -166,13 +181,13 @@ void r300_init_reg_flags(struct drm_device *dev) ADD_RANGE(0x21DC, 1); ADD_RANGE(R300_VAP_UNKNOWN_221C, 1); ADD_RANGE(R300_VAP_CLIP_X_0, 4); - ADD_RANGE(R300_VAP_PVS_WAITIDLE, 1); + ADD_RANGE(R300_VAP_PVS_STATE_FLUSH_REG, 1); ADD_RANGE(R300_VAP_UNKNOWN_2288, 1); ADD_RANGE(R300_VAP_OUTPUT_VTX_FMT_0, 2); ADD_RANGE(R300_VAP_PVS_CNTL_1, 3); ADD_RANGE(R300_GB_ENABLE, 1); ADD_RANGE(R300_GB_MSPOS0, 5); - ADD_RANGE(R300_TX_CNTL, 1); + ADD_RANGE(R300_TX_INVALTAGS, 1); ADD_RANGE(R300_TX_ENABLE, 1); ADD_RANGE(0x4200, 4); ADD_RANGE(0x4214, 1); @@ -190,7 +205,7 @@ void r300_init_reg_flags(struct drm_device *dev) ADD_RANGE(0x42C0, 2); ADD_RANGE(R300_RS_CNTL_0, 2); - ADD_RANGE(0x43A4, 2); + ADD_RANGE(R300_SC_HYPERZ, 2); ADD_RANGE(0x43E8, 1); ADD_RANGE(0x46A4, 5); @@ -209,14 +224,12 @@ void r300_init_reg_flags(struct drm_device *dev) ADD_RANGE(0x4E50, 9); ADD_RANGE(0x4E88, 1); ADD_RANGE(0x4EA0, 2); - ADD_RANGE(R300_RB3D_ZSTENCIL_CNTL_0, 3); - ADD_RANGE(R300_RB3D_ZSTENCIL_FORMAT, 4); - ADD_RANGE_MARK(R300_RB3D_DEPTHOFFSET, 1, MARK_CHECK_OFFSET); /* check offset */ - ADD_RANGE(R300_RB3D_DEPTHPITCH, 1); - ADD_RANGE(0x4F28, 1); - ADD_RANGE(0x4F30, 2); - ADD_RANGE(0x4F44, 1); - ADD_RANGE(0x4F54, 1); + ADD_RANGE(R300_ZB_CNTL, 3); + ADD_RANGE(R300_ZB_FORMAT, 4); + ADD_RANGE_MARK(R300_ZB_DEPTHOFFSET, 1, MARK_CHECK_OFFSET); /* check offset */ + ADD_RANGE(R300_ZB_DEPTHPITCH, 1); + ADD_RANGE(R300_ZB_DEPTHCLEARVALUE, 1); + ADD_RANGE(R300_ZB_ZMASK_OFFSET, 13); ADD_RANGE(R300_TX_FILTER_0, 16); ADD_RANGE(R300_TX_FILTER1_0, 16); @@ -229,7 +242,7 @@ void r300_init_reg_flags(struct drm_device *dev) ADD_RANGE(R300_TX_BORDER_COLOR_0, 16); /* Sporadic registers used as primitives are emitted */ - ADD_RANGE(R300_RB3D_ZCACHE_CTLSTAT, 1); + ADD_RANGE(R300_ZB_ZCACHE_CTLSTAT, 1); ADD_RANGE(R300_RB3D_DSTCACHE_CTLSTAT, 1); ADD_RANGE(R300_VAP_INPUT_ROUTE_0_0, 8); ADD_RANGE(R300_VAP_INPUT_ROUTE_1_0, 8); @@ -243,6 +256,7 @@ void r300_init_reg_flags(struct drm_device *dev) ADD_RANGE(R500_RS_INST_0, 16); ADD_RANGE(R500_RB3D_COLOR_CLEAR_VALUE_AR, 2); ADD_RANGE(R500_RB3D_CONSTANT_COLOR_AR, 2); + ADD_RANGE(R500_ZB_FIFO_SIZE, 2); } else { ADD_RANGE(R300_PFS_CNTL_0, 3); ADD_RANGE(R300_PFS_NODE_0, 4); @@ -390,15 +404,28 @@ static __inline__ int r300_emit_vpu(drm_radeon_private_t *dev_priv, if (sz * 16 > cmdbuf->bufsz) return -EINVAL; - BEGIN_RING(5 + sz * 4); - /* Wait for VAP to come to senses.. */ - /* there is no need to emit it multiple times, (only once before VAP is programmed, - but this optimization is for later */ - OUT_RING_REG(R300_VAP_PVS_WAITIDLE, 0); + /* VAP is very sensitive so we purge cache before we program it + * and we also flush its state before & after */ + BEGIN_RING(6); + OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); + OUT_RING(R300_RB3D_DC_FLUSH); + OUT_RING(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); + OUT_RING(RADEON_WAIT_3D_IDLECLEAN); + OUT_RING(CP_PACKET0(R300_VAP_PVS_STATE_FLUSH_REG, 0)); + OUT_RING(0); + ADVANCE_RING(); + /* set flush flag */ + dev_priv->track_flush |= RADEON_FLUSH_EMITED; + + BEGIN_RING(3 + sz * 4); OUT_RING_REG(R300_VAP_PVS_UPLOAD_ADDRESS, addr); OUT_RING(CP_PACKET0_TABLE(R300_VAP_PVS_UPLOAD_DATA, sz * 4 - 1)); OUT_RING_TABLE((int *)cmdbuf->buf, sz * 4); + ADVANCE_RING(); + BEGIN_RING(2); + OUT_RING(CP_PACKET0(R300_VAP_PVS_STATE_FLUSH_REG, 0)); + OUT_RING(0); ADVANCE_RING(); cmdbuf->buf += sz * 16; @@ -426,6 +453,15 @@ static __inline__ int r300_emit_clear(drm_radeon_private_t *dev_priv, OUT_RING_TABLE((int *)cmdbuf->buf, 8); ADVANCE_RING(); + BEGIN_RING(4); + OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); + OUT_RING(R300_RB3D_DC_FLUSH); + OUT_RING(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); + OUT_RING(RADEON_WAIT_3D_IDLECLEAN); + ADVANCE_RING(); + /* set flush flag */ + dev_priv->track_flush |= RADEON_FLUSH_EMITED; + cmdbuf->buf += 8 * 4; cmdbuf->bufsz -= 8 * 4; @@ -545,22 +581,23 @@ static __inline__ int r300_emit_bitblt_multi(drm_radeon_private_t *dev_priv, return 0; } -static __inline__ int r300_emit_indx_buffer(drm_radeon_private_t *dev_priv, - drm_radeon_kcmd_buffer_t *cmdbuf) +static __inline__ int r300_emit_draw_indx_2(drm_radeon_private_t *dev_priv, + drm_radeon_kcmd_buffer_t *cmdbuf) { - u32 *cmd = (u32 *) cmdbuf->buf; - int count, ret; + u32 *cmd; + int count; + int expected_count; RING_LOCALS; - count=(cmd[0]>>16) & 0x3fff; + cmd = (u32 *) cmdbuf->buf; + count = (cmd[0]>>16) & 0x3fff; + expected_count = cmd[1] >> 16; + if (!(cmd[1] & R300_VAP_VF_CNTL__INDEX_SIZE_32bit)) + expected_count = (expected_count+1)/2; - if ((cmd[1] & 0x8000ffff) != 0x80000810) { - DRM_ERROR("Invalid indx_buffer reg address %08X\n", cmd[1]); - return -EINVAL; - } - ret = !radeon_check_offset(dev_priv, cmd[2]); - if (ret) { - DRM_ERROR("Invalid indx_buffer offset is %08X\n", cmd[2]); + if (count && count != expected_count) { + DRM_ERROR("3D_DRAW_INDX_2: packet size %i, expected %i\n", + count, expected_count); return -EINVAL; } @@ -572,6 +609,50 @@ static __inline__ int r300_emit_indx_buffer(drm_radeon_private_t *dev_priv, cmdbuf->buf += (count+2)*4; cmdbuf->bufsz -= (count+2)*4; + if (!count) { + drm_r300_cmd_header_t header; + + if (cmdbuf->bufsz < 4*4 + sizeof(header)) { + DRM_ERROR("3D_DRAW_INDX_2: expect subsequent INDX_BUFFER, but stream is too short.\n"); + return -EINVAL; + } + + header.u = *(unsigned int *)cmdbuf->buf; + + cmdbuf->buf += sizeof(header); + cmdbuf->bufsz -= sizeof(header); + cmd = (u32 *) cmdbuf->buf; + + if (header.header.cmd_type != R300_CMD_PACKET3 || + header.packet3.packet != R300_CMD_PACKET3_RAW || + cmd[0] != CP_PACKET3(RADEON_CP_INDX_BUFFER, 2)) { + DRM_ERROR("3D_DRAW_INDX_2: expect subsequent INDX_BUFFER.\n"); + return -EINVAL; + } + + if ((cmd[1] & 0x8000ffff) != 0x80000810) { + DRM_ERROR("Invalid indx_buffer reg address %08X\n", cmd[1]); + return -EINVAL; + } + if (!radeon_check_offset(dev_priv, cmd[2])) { + DRM_ERROR("Invalid indx_buffer offset is %08X\n", cmd[2]); + return -EINVAL; + } + if (cmd[3] != expected_count) { + DRM_ERROR("INDX_BUFFER: buffer size %i, expected %i\n", + cmd[3], expected_count); + return -EINVAL; + } + + BEGIN_RING(4); + OUT_RING(cmd[0]); + OUT_RING_TABLE((int *)(cmdbuf->buf + 4), 3); + ADVANCE_RING(); + + cmdbuf->buf += 4*4; + cmdbuf->bufsz -= 4*4; + } + return 0; } @@ -615,11 +696,22 @@ static __inline__ int r300_emit_raw_packet3(drm_radeon_private_t *dev_priv, case RADEON_CNTL_BITBLT_MULTI: return r300_emit_bitblt_multi(dev_priv, cmdbuf); - case RADEON_CP_INDX_BUFFER: /* DRAW_INDX_2 without INDX_BUFFER seems to lock up the gpu */ - return r300_emit_indx_buffer(dev_priv, cmdbuf); - case RADEON_CP_3D_DRAW_IMMD_2: /* triggers drawing using in-packet vertex data */ - case RADEON_CP_3D_DRAW_VBUF_2: /* triggers drawing of vertex buffers setup elsewhere */ - case RADEON_CP_3D_DRAW_INDX_2: /* triggers drawing using indices to vertex buffer */ + case RADEON_CP_INDX_BUFFER: + DRM_ERROR("packet3 INDX_BUFFER without preceding 3D_DRAW_INDX_2 is illegal.\n"); + return -EINVAL; + case RADEON_CP_3D_DRAW_IMMD_2: + /* triggers drawing using in-packet vertex data */ + case RADEON_CP_3D_DRAW_VBUF_2: + /* triggers drawing of vertex buffers setup elsewhere */ + dev_priv->track_flush &= ~(RADEON_FLUSH_EMITED | + RADEON_PURGE_EMITED); + break; + case RADEON_CP_3D_DRAW_INDX_2: + /* triggers drawing using indices to vertex buffer */ + /* whenever we send vertex we clear flush & purge */ + dev_priv->track_flush &= ~(RADEON_FLUSH_EMITED | + RADEON_PURGE_EMITED); + return r300_emit_draw_indx_2(dev_priv, cmdbuf); case RADEON_WAIT_FOR_IDLE: case RADEON_CP_NOP: /* these packets are safe */ @@ -715,16 +807,53 @@ static __inline__ int r300_emit_packet3(drm_radeon_private_t *dev_priv, */ static __inline__ void r300_pacify(drm_radeon_private_t *dev_priv) { + uint32_t cache_z, cache_3d, cache_2d; RING_LOCALS; - BEGIN_RING(6); + cache_z = R300_ZC_FLUSH; + cache_2d = R300_DC_FLUSH_2D; + cache_3d = R300_DC_FLUSH_3D; + if (!(dev_priv->track_flush & RADEON_PURGE_EMITED)) { + /* we can purge, primitive where draw since last purge */ + cache_z |= R300_ZC_FREE; + cache_2d |= R300_DC_FREE_2D; + cache_3d |= R300_DC_FREE_3D; + } + + /* flush & purge zbuffer */ + BEGIN_RING(2); + OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0)); + OUT_RING(cache_z); + ADVANCE_RING(); + /* flush & purge 3d */ + BEGIN_RING(2); OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); - OUT_RING(R300_RB3D_DSTCACHE_UNKNOWN_0A); - OUT_RING(CP_PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0)); - OUT_RING(R300_RB3D_ZCACHE_UNKNOWN_03); - OUT_RING(CP_PACKET3(RADEON_CP_NOP, 0)); - OUT_RING(0x0); + OUT_RING(cache_3d); + ADVANCE_RING(); + /* flush & purge texture */ + BEGIN_RING(2); + OUT_RING(CP_PACKET0(R300_TX_INVALTAGS, 0)); + OUT_RING(0); ADVANCE_RING(); + /* FIXME: is this one really needed ? */ + BEGIN_RING(2); + OUT_RING(CP_PACKET0(R300_RB3D_AARESOLVE_CTL, 0)); + OUT_RING(0); + ADVANCE_RING(); + BEGIN_RING(2); + OUT_RING(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); + OUT_RING(RADEON_WAIT_3D_IDLECLEAN); + ADVANCE_RING(); + /* flush & purge 2d through E2 as RB2D will trigger lockup */ + BEGIN_RING(4); + OUT_RING(CP_PACKET0(R300_DSTCACHE_CTLSTAT, 0)); + OUT_RING(cache_2d); + OUT_RING(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); + OUT_RING(RADEON_WAIT_2D_IDLECLEAN | + RADEON_WAIT_HOST_IDLECLEAN); + ADVANCE_RING(); + /* set flush & purge flags */ + dev_priv->track_flush |= RADEON_FLUSH_EMITED | RADEON_PURGE_EMITED; } /** @@ -732,12 +861,12 @@ static __inline__ void r300_pacify(drm_radeon_private_t *dev_priv) * The actual age emit is done by r300_do_cp_cmdbuf, which is why you must * be careful about how this function is called. */ -static void r300_discard_buffer(struct drm_device * dev, struct drm_buf * buf) +static void r300_discard_buffer(struct drm_device * dev, struct drm_master *master, struct drm_buf * buf) { - drm_radeon_private_t *dev_priv = dev->dev_private; drm_radeon_buf_priv_t *buf_priv = buf->dev_private; + struct drm_radeon_master_private *master_priv = master->driver_priv; - buf_priv->age = ++dev_priv->sarea_priv->last_dispatch; + buf_priv->age = ++master_priv->sarea_priv->last_dispatch; buf->pending = 1; buf->used = 0; } @@ -898,6 +1027,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev, drm_radeon_kcmd_buffer_t *cmdbuf) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; struct drm_device_dma *dma = dev->dma; struct drm_buf *buf = NULL; int emit_dispatch_age = 0; @@ -905,8 +1035,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev, DRM_DEBUG("\n"); - /* See the comment above r300_emit_begin3d for why this call must be here, - * and what the cleanup gotos are for. */ + /* pacify */ r300_pacify(dev_priv); if (cmdbuf->nbox <= R300_SIMULTANEOUS_CLIPRECTS) { @@ -1005,7 +1134,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev, } emit_dispatch_age = 1; - r300_discard_buffer(dev, buf); + r300_discard_buffer(dev, file_priv->master, buf); break; case R300_CMD_WAIT: @@ -1060,7 +1189,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev, /* Emit the vertex buffer age */ BEGIN_RING(2); - RADEON_DISPATCH_AGE(dev_priv->sarea_priv->last_dispatch); + RADEON_DISPATCH_AGE(master_priv->sarea_priv->last_dispatch); ADVANCE_RING(); } diff --git a/shared-core/r300_reg.h b/shared-core/r300_reg.h index 0be01fc7..d35dd39d 100644 --- a/shared-core/r300_reg.h +++ b/shared-core/r300_reg.h @@ -320,7 +320,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and * avoids bugs caused by still running shaders reading bad data from memory. */ -#define R300_VAP_PVS_WAITIDLE 0x2284 /* GUESS */ +#define R300_VAP_PVS_STATE_FLUSH_REG 0x2284 /* Absolutely no clue what this register is about. */ #define R300_VAP_UNKNOWN_2288 0x2288 @@ -516,7 +516,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. /* gap */ /* Zero to flush caches. */ -#define R300_TX_CNTL 0x4100 +#define R300_TX_INVALTAGS 0x4100 #define R300_TX_FLUSH 0x0 /* The upper enable bits are guessed, based on fglrx reported limits. */ @@ -705,6 +705,27 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_RS_ROUTE_1_UNKNOWN11 (1 << 11) /* END: Rasterization / Interpolators - many guesses */ +/* Hierarchical Z Enable */ +#define R300_SC_HYPERZ 0x43a4 +# define R300_SC_HYPERZ_DISABLE (0 << 0) +# define R300_SC_HYPERZ_ENABLE (1 << 0) +# define R300_SC_HYPERZ_MIN (0 << 1) +# define R300_SC_HYPERZ_MAX (1 << 1) +# define R300_SC_HYPERZ_ADJ_256 (0 << 2) +# define R300_SC_HYPERZ_ADJ_128 (1 << 2) +# define R300_SC_HYPERZ_ADJ_64 (2 << 2) +# define R300_SC_HYPERZ_ADJ_32 (3 << 2) +# define R300_SC_HYPERZ_ADJ_16 (4 << 2) +# define R300_SC_HYPERZ_ADJ_8 (5 << 2) +# define R300_SC_HYPERZ_ADJ_4 (6 << 2) +# define R300_SC_HYPERZ_ADJ_2 (7 << 2) +# define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5) +# define R300_SC_HYPERZ_HZ_Z0MIN (1 << 5) +# define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6) +# define R300_SC_HYPERZ_HZ_Z0MAX (1 << 6) + +#define R300_SC_EDGERULE 0x43a8 + /* BEGIN: Scissors and cliprects */ /* There are four clipping rectangles. Their corner coordinates are inclusive. @@ -1344,6 +1365,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R300_RB3D_COLORPITCH2 0x4E40 /* GUESS */ #define R300_RB3D_COLORPITCH3 0x4E44 /* GUESS */ +#define R300_RB3D_AARESOLVE_CTL 0x4E88 /* gap */ /* Guess by Vladimir. @@ -1358,19 +1380,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. * for this. * Bit (1<<8) is the "test" bit. so plain write is 6 - vd */ -#define R300_RB3D_ZSTENCIL_CNTL_0 0x4F00 -# define R300_RB3D_Z_DISABLED_1 0x00000010 -# define R300_RB3D_Z_DISABLED_2 0x00000014 -# define R300_RB3D_Z_TEST 0x00000012 -# define R300_RB3D_Z_TEST_AND_WRITE 0x00000016 -# define R300_RB3D_Z_WRITE_ONLY 0x00000006 - -# define R300_RB3D_Z_TEST 0x00000012 -# define R300_RB3D_Z_TEST_AND_WRITE 0x00000016 -# define R300_RB3D_Z_WRITE_ONLY 0x00000006 -# define R300_RB3D_STENCIL_ENABLE 0x00000001 - -#define R300_RB3D_ZSTENCIL_CNTL_1 0x4F04 +#define R300_ZB_CNTL 0x4F00 +# define R300_STENCIL_ENABLE (1 << 0) +# define R300_Z_ENABLE (1 << 1) +# define R300_Z_WRITE_ENABLE (1 << 2) +# define R300_Z_SIGNED_COMPARE (1 << 3) +# define R300_STENCIL_FRONT_BACK (1 << 4) + +#define R300_ZB_ZSTENCILCNTL 0x4f04 /* functions */ # define R300_ZS_NEVER 0 # define R300_ZS_LESS 1 @@ -1390,52 +1407,166 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ZS_INVERT 5 # define R300_ZS_INCR_WRAP 6 # define R300_ZS_DECR_WRAP 7 +# define R300_Z_FUNC_SHIFT 0 /* front and back refer to operations done for front and back faces, i.e. separate stencil function support */ -# define R300_RB3D_ZS1_DEPTH_FUNC_SHIFT 0 -# define R300_RB3D_ZS1_FRONT_FUNC_SHIFT 3 -# define R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT 6 -# define R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT 9 -# define R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT 12 -# define R300_RB3D_ZS1_BACK_FUNC_SHIFT 15 -# define R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT 18 -# define R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT 21 -# define R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT 24 - -#define R300_RB3D_ZSTENCIL_CNTL_2 0x4F08 -# define R300_RB3D_ZS2_STENCIL_REF_SHIFT 0 -# define R300_RB3D_ZS2_STENCIL_MASK 0xFF -# define R300_RB3D_ZS2_STENCIL_MASK_SHIFT 8 -# define R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT 16 +# define R300_S_FRONT_FUNC_SHIFT 3 +# define R300_S_FRONT_SFAIL_OP_SHIFT 6 +# define R300_S_FRONT_ZPASS_OP_SHIFT 9 +# define R300_S_FRONT_ZFAIL_OP_SHIFT 12 +# define R300_S_BACK_FUNC_SHIFT 15 +# define R300_S_BACK_SFAIL_OP_SHIFT 18 +# define R300_S_BACK_ZPASS_OP_SHIFT 21 +# define R300_S_BACK_ZFAIL_OP_SHIFT 24 + +#define R300_ZB_STENCILREFMASK 0x4f08 +# define R300_STENCILREF_SHIFT 0 +# define R300_STENCILREF_MASK 0x000000ff +# define R300_STENCILMASK_SHIFT 8 +# define R300_STENCILMASK_MASK 0x0000ff00 +# define R300_STENCILWRITEMASK_SHIFT 16 +# define R300_STENCILWRITEMASK_MASK 0x00ff0000 /* gap */ -#define R300_RB3D_ZSTENCIL_FORMAT 0x4F10 -# define R300_DEPTH_FORMAT_16BIT_INT_Z (0 << 0) -# define R300_DEPTH_FORMAT_24BIT_INT_Z (2 << 0) - /* 16 bit format or some aditional bit ? */ -# define R300_DEPTH_FORMAT_UNK32 (32 << 0) +#define R300_ZB_FORMAT 0x4f10 +# define R300_DEPTHFORMAT_16BIT_INT_Z (0 << 0) +# define R300_DEPTHFORMAT_16BIT_13E3 (1 << 0) +# define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL (2 << 0) +/* reserved up to (15 << 0) */ +# define R300_INVERT_13E3_LEADING_ONES (0 << 4) +# define R300_INVERT_13E3_LEADING_ZEROS (1 << 4) -#define R300_RB3D_EARLY_Z 0x4F14 -# define R300_EARLY_Z_DISABLE (0 << 0) -# define R300_EARLY_Z_ENABLE (1 << 0) +#define R300_ZB_ZTOP 0x4F14 +# define R300_ZTOP_DISABLE (0 << 0) +# define R300_ZTOP_ENABLE (1 << 0) /* gap */ -#define R300_RB3D_ZCACHE_CTLSTAT 0x4F18 /* GUESS */ -# define R300_RB3D_ZCACHE_UNKNOWN_01 0x1 -# define R300_RB3D_ZCACHE_UNKNOWN_03 0x3 +#define R300_ZB_ZCACHE_CTLSTAT 0x4f18 +# define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT (0 << 0) +# define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0) +# define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT (0 << 1) +# define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE (1 << 1) +# define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE (0 << 31) +# define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY (1 << 31) + +#define R300_ZB_BW_CNTL 0x4f1c +# define R300_HIZ_DISABLE (0 << 0) +# define R300_HIZ_ENABLE (1 << 0) +# define R300_HIZ_MIN (0 << 1) +# define R300_HIZ_MAX (1 << 1) +# define R300_FAST_FILL_DISABLE (0 << 2) +# define R300_FAST_FILL_ENABLE (1 << 2) +# define R300_RD_COMP_DISABLE (0 << 3) +# define R300_RD_COMP_ENABLE (1 << 3) +# define R300_WR_COMP_DISABLE (0 << 4) +# define R300_WR_COMP_ENABLE (1 << 4) +# define R300_ZB_CB_CLEAR_RMW (0 << 5) +# define R300_ZB_CB_CLEAR_CACHE_LINEAR (1 << 5) +# define R300_FORCE_COMPRESSED_STENCIL_VALUE_DISABLE (0 << 6) +# define R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE (1 << 6) + +# define R500_ZEQUAL_OPTIMIZE_ENABLE (0 << 7) +# define R500_ZEQUAL_OPTIMIZE_DISABLE (1 << 7) +# define R500_SEQUAL_OPTIMIZE_ENABLE (0 << 8) +# define R500_SEQUAL_OPTIMIZE_DISABLE (1 << 8) + +# define R500_BMASK_ENABLE (0 << 10) +# define R500_BMASK_DISABLE (1 << 10) +# define R500_HIZ_EQUAL_REJECT_DISABLE (0 << 11) +# define R500_HIZ_EQUAL_REJECT_ENABLE (1 << 11) +# define R500_HIZ_FP_EXP_BITS_DISABLE (0 << 12) +# define R500_HIZ_FP_EXP_BITS_1 (1 << 12) +# define R500_HIZ_FP_EXP_BITS_2 (2 << 12) +# define R500_HIZ_FP_EXP_BITS_3 (3 << 12) +# define R500_HIZ_FP_EXP_BITS_4 (4 << 12) +# define R500_HIZ_FP_EXP_BITS_5 (5 << 12) +# define R500_HIZ_FP_INVERT_LEADING_ONES (0 << 15) +# define R500_HIZ_FP_INVERT_LEADING_ZEROS (1 << 15) +# define R500_TILE_OVERWRITE_RECOMPRESSION_ENABLE (0 << 16) +# define R500_TILE_OVERWRITE_RECOMPRESSION_DISABLE (1 << 16) +# define R500_CONTIGUOUS_6XAA_SAMPLES_ENABLE (0 << 17) +# define R500_CONTIGUOUS_6XAA_SAMPLES_DISABLE (1 << 17) +# define R500_PEQ_PACKING_DISABLE (0 << 18) +# define R500_PEQ_PACKING_ENABLE (1 << 18) +# define R500_COVERED_PTR_MASKING_DISABLE (0 << 18) +# define R500_COVERED_PTR_MASKING_ENABLE (1 << 18) + /* gap */ -#define R300_RB3D_DEPTHOFFSET 0x4F20 -#define R300_RB3D_DEPTHPITCH 0x4F24 -# define R300_DEPTHPITCH_MASK 0x00001FF8 /* GUESS */ -# define R300_DEPTH_TILE_ENABLE (1 << 16) /* GUESS */ -# define R300_DEPTH_MICROTILE_ENABLE (1 << 17) /* GUESS */ -# define R300_DEPTH_ENDIAN_NO_SWAP (0 << 18) /* GUESS */ -# define R300_DEPTH_ENDIAN_WORD_SWAP (1 << 18) /* GUESS */ -# define R300_DEPTH_ENDIAN_DWORD_SWAP (2 << 18) /* GUESS */ +/* Z Buffer Address Offset. + * Bits 31 to 5 are used for aligned Z buffer address offset for macro tiles. + */ +#define R300_ZB_DEPTHOFFSET 0x4f20 + +/* Z Buffer Pitch and Endian Control */ +#define R300_ZB_DEPTHPITCH 0x4f24 +# define R300_DEPTHPITCH_MASK 0x00003FFC +# define R300_DEPTHMACROTILE_DISABLE (0 << 16) +# define R300_DEPTHMACROTILE_ENABLE (1 << 16) +# define R300_DEPTHMICROTILE_LINEAR (0 << 17) +# define R300_DEPTHMICROTILE_TILED (1 << 17) +# define R300_DEPTHMICROTILE_TILED_SQUARE (2 << 17) +# define R300_DEPTHENDIAN_NO_SWAP (0 << 18) +# define R300_DEPTHENDIAN_WORD_SWAP (1 << 18) +# define R300_DEPTHENDIAN_DWORD_SWAP (2 << 18) +# define R300_DEPTHENDIAN_HALF_DWORD_SWAP (3 << 18) + +/* Z Buffer Clear Value */ +#define R300_ZB_DEPTHCLEARVALUE 0x4f28 + +#define R300_ZB_ZMASK_OFFSET 0x4f30 +#define R300_ZB_ZMASK_PITCH 0x4f34 +#define R300_ZB_ZMASK_WRINDEX 0x4f38 +#define R300_ZB_ZMASK_DWORD 0x4f3c +#define R300_ZB_ZMASK_RDINDEX 0x4f40 + +/* Hierarchical Z Memory Offset */ +#define R300_ZB_HIZ_OFFSET 0x4f44 + +/* Hierarchical Z Write Index */ +#define R300_ZB_HIZ_WRINDEX 0x4f48 + +/* Hierarchical Z Data */ +#define R300_ZB_HIZ_DWORD 0x4f4c + +/* Hierarchical Z Read Index */ +#define R300_ZB_HIZ_RDINDEX 0x4f50 + +/* Hierarchical Z Pitch */ +#define R300_ZB_HIZ_PITCH 0x4f54 + +/* Z Buffer Z Pass Counter Data */ +#define R300_ZB_ZPASS_DATA 0x4f58 + +/* Z Buffer Z Pass Counter Address */ +#define R300_ZB_ZPASS_ADDR 0x4f5c + +/* Depth buffer X and Y coordinate offset */ +#define R300_ZB_DEPTHXY_OFFSET 0x4f60 +# define R300_DEPTHX_OFFSET_SHIFT 1 +# define R300_DEPTHX_OFFSET_MASK 0x000007FE +# define R300_DEPTHY_OFFSET_SHIFT 17 +# define R300_DEPTHY_OFFSET_MASK 0x07FE0000 + +/* Sets the fifo sizes */ +#define R500_ZB_FIFO_SIZE 0x4fd0 +# define R500_OP_FIFO_SIZE_FULL (0 << 0) +# define R500_OP_FIFO_SIZE_HALF (1 << 0) +# define R500_OP_FIFO_SIZE_QUATER (2 << 0) +# define R500_OP_FIFO_SIZE_EIGTHS (4 << 0) + +/* Stencil Reference Value and Mask for backfacing quads */ +/* R300_ZB_STENCILREFMASK handles front face */ +#define R500_ZB_STENCILREFMASK_BF 0x4fd4 +# define R500_STENCILREF_SHIFT 0 +# define R500_STENCILREF_MASK 0x000000ff +# define R500_STENCILMASK_SHIFT 8 +# define R500_STENCILMASK_MASK 0x0000ff00 +# define R500_STENCILWRITEMASK_SHIFT 16 +# define R500_STENCILWRITEMASK_MASK 0x00ff0000 /* BEGIN: Vertex program instruction set */ diff --git a/shared-core/radeon_cp.c b/shared-core/radeon_cp.c index 2e680306..6fbcfeab 100644 --- a/shared-core/radeon_cp.c +++ b/shared-core/radeon_cp.c @@ -31,6 +31,7 @@ #include "drmP.h" #include "drm.h" +#include "drm_sarea.h" #include "radeon_drm.h" #include "radeon_drv.h" #include "r300_reg.h" @@ -39,6 +40,7 @@ #define RADEON_FIFO_DEBUG 0 static int radeon_do_cleanup_cp(struct drm_device * dev); +static void radeon_do_cp_start(drm_radeon_private_t * dev_priv); static u32 R500_READ_MCIND(drm_radeon_private_t *dev_priv, int addr) { @@ -75,6 +77,23 @@ static u32 IGP_READ_MCIND(drm_radeon_private_t *dev_priv, int addr) return RS480_READ_MCIND(dev_priv, addr); } +u32 radeon_read_mc_reg(drm_radeon_private_t *dev_priv, int addr) +{ + if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + return IGP_READ_MCIND(dev_priv, addr); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) + return R500_READ_MCIND(dev_priv, addr); + return 0; +} + +void radeon_write_mc_reg(drm_radeon_private_t *dev_priv, u32 addr, u32 val) +{ + if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + IGP_WRITE_MCIND(addr, val); + else if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) + R500_WRITE_MCIND(addr, val); +} + u32 radeon_read_fb_location(drm_radeon_private_t *dev_priv) { @@ -112,12 +131,82 @@ static void radeon_write_agp_location(drm_radeon_private_t *dev_priv, u32 agp_lo RADEON_WRITE(RADEON_MC_AGP_LOCATION, agp_loc); } -static int RADEON_READ_PLL(struct drm_device * dev, int addr) +static void radeon_write_agp_base(drm_radeon_private_t *dev_priv, u64 agp_base) { - drm_radeon_private_t *dev_priv = dev->dev_private; + u32 agp_base_hi = upper_32_bits(agp_base); + u32 agp_base_lo = agp_base & 0xffffffff; + + if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV515) { + R500_WRITE_MCIND(RV515_MC_AGP_BASE, agp_base_lo); + R500_WRITE_MCIND(RV515_MC_AGP_BASE_2, agp_base_hi); + } else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) { + RS690_WRITE_MCIND(RS690_MC_AGP_BASE, agp_base_lo); + RS690_WRITE_MCIND(RS690_MC_AGP_BASE_2, agp_base_hi); + } else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) { + R500_WRITE_MCIND(R520_MC_AGP_BASE, agp_base_lo); + R500_WRITE_MCIND(R520_MC_AGP_BASE_2, agp_base_hi); + } else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS400) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS480)) { + RADEON_WRITE(RADEON_AGP_BASE, agp_base_lo); + RADEON_WRITE(RS480_AGP_BASE_2, agp_base_hi); + } else { + RADEON_WRITE(RADEON_AGP_BASE, agp_base_lo); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R200) + RADEON_WRITE(RADEON_AGP_BASE_2, agp_base_hi); + } +} + + +void radeon_pll_errata_after_index(struct drm_radeon_private *dev_priv) +{ + if (!(dev_priv->pll_errata & CHIP_ERRATA_PLL_DUMMYREADS)) + return; + + (void)RADEON_READ(RADEON_CLOCK_CNTL_DATA); + (void)RADEON_READ(RADEON_CRTC_GEN_CNTL); +} + +void radeon_pll_errata_after_data(struct drm_radeon_private *dev_priv) +{ + /* This workarounds is necessary on RV100, RS100 and RS200 chips + * or the chip could hang on a subsequent access + */ + if (dev_priv->pll_errata & CHIP_ERRATA_PLL_DELAY) + udelay(5000); + + /* This function is required to workaround a hardware bug in some (all?) + * revisions of the R300. This workaround should be called after every + * CLOCK_CNTL_INDEX register access. If not, register reads afterward + * may not be correct. + */ + if (dev_priv->pll_errata & CHIP_ERRATA_R300_CG) { + uint32_t save, tmp; + + save = RADEON_READ(RADEON_CLOCK_CNTL_INDEX); + tmp = save & ~(0x3f | RADEON_PLL_WR_EN); + RADEON_WRITE(RADEON_CLOCK_CNTL_INDEX, tmp); + tmp = RADEON_READ(RADEON_CLOCK_CNTL_DATA); + RADEON_WRITE(RADEON_CLOCK_CNTL_INDEX, save); + } +} + +int RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr) +{ + uint32_t data; + + RADEON_WRITE8(RADEON_CLOCK_CNTL_INDEX, addr & 0x3f); + radeon_pll_errata_after_index(dev_priv); + data = RADEON_READ(RADEON_CLOCK_CNTL_DATA); + radeon_pll_errata_after_data(dev_priv); + return data; +} - RADEON_WRITE8(RADEON_CLOCK_CNTL_INDEX, addr & 0x1f); - return RADEON_READ(RADEON_CLOCK_CNTL_DATA); +void RADEON_WRITE_PLL(struct drm_radeon_private *dev_priv, int addr, uint32_t data) +{ + RADEON_WRITE8(RADEON_CLOCK_CNTL_INDEX, ((addr & 0x3f) | RADEON_PLL_WR_EN)); + radeon_pll_errata_after_index(dev_priv); + RADEON_WRITE(RADEON_CLOCK_CNTL_DATA, data); + radeon_pll_errata_after_data(dev_priv); } static u32 RADEON_READ_PCIE(drm_radeon_private_t *dev_priv, int addr) @@ -126,6 +215,39 @@ static u32 RADEON_READ_PCIE(drm_radeon_private_t *dev_priv, int addr) return RADEON_READ(RADEON_PCIE_DATA); } +/* ATOM accessor methods */ +static uint32_t cail_mc_read(struct card_info *info, uint32_t reg) +{ + uint32_t ret = radeon_read_mc_reg(info->dev->dev_private, reg); + + // DRM_DEBUG("(%x) = %x\n", reg, ret); + return ret; +} + +static void cail_mc_write(struct card_info *info, uint32_t reg, uint32_t val) +{ + // DRM_DEBUG("(%x, %x)\n", reg, val); + radeon_write_mc_reg(info->dev->dev_private, reg, val); +} + +static void cail_reg_write(struct card_info *info, uint32_t reg, uint32_t val) +{ + drm_radeon_private_t *dev_priv = info->dev->dev_private; + + // DRM_DEBUG("(%x, %x)\n", reg*4, val); + RADEON_WRITE(reg*4, val); +} + +static uint32_t cail_reg_read(struct card_info *info, uint32_t reg) +{ + uint32_t ret; + drm_radeon_private_t *dev_priv = info->dev->dev_private; + + ret = RADEON_READ(reg*4); + // DRM_DEBUG("(%x) = %x\n", reg*4, ret); + return ret; +} + #if RADEON_FIFO_DEBUG static void radeon_status(drm_radeon_private_t * dev_priv) { @@ -173,23 +295,8 @@ static int radeon_do_pixcache_flush(drm_radeon_private_t * dev_priv) DRM_UDELAY(1); } } else { - /* 3D */ - tmp = RADEON_READ(R300_RB3D_DSTCACHE_CTLSTAT); - tmp |= RADEON_RB3D_DC_FLUSH_ALL; - RADEON_WRITE(R300_RB3D_DSTCACHE_CTLSTAT, tmp); - - /* 2D */ - tmp = RADEON_READ(RADEON_RB2D_DSTCACHE_CTLSTAT); - tmp |= RADEON_RB3D_DC_FLUSH_ALL; - RADEON_WRITE(RADEON_RB3D_DSTCACHE_CTLSTAT, tmp); - - for (i = 0; i < dev_priv->usec_timeout; i++) { - if (!(RADEON_READ(RADEON_RB2D_DSTCACHE_CTLSTAT) - & RADEON_RB3D_DC_BUSY)) { - return 0; - } - DRM_UDELAY(1); - } + /* don't flush or purge cache here or lockup */ + return 0; } #if RADEON_FIFO_DEBUG @@ -212,6 +319,9 @@ static int radeon_do_wait_for_fifo(drm_radeon_private_t * dev_priv, int entries) return 0; DRM_UDELAY(1); } + DRM_INFO("wait for fifo failed status : 0x%08X 0x%08X\n", + RADEON_READ(RADEON_RBBM_STATUS), + RADEON_READ(R300_VAP_CNTL_STATUS)); #if RADEON_FIFO_DEBUG DRM_ERROR("failed!\n"); @@ -238,6 +348,9 @@ static int radeon_do_wait_for_idle(drm_radeon_private_t * dev_priv) } DRM_UDELAY(1); } + DRM_INFO("wait idle failed status : 0x%08X 0x%08X\n", + RADEON_READ(RADEON_RBBM_STATUS), + RADEON_READ(R300_VAP_CNTL_STATUS)); #if RADEON_FIFO_DEBUG DRM_ERROR("failed!\n"); @@ -277,7 +390,7 @@ static void radeon_init_pipes(drm_radeon_private_t * dev_priv) } if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) { - RADEON_WRITE_PLL(R500_DYN_SCLK_PWMEM_PIPE, (1 | ((gb_pipe_sel >> 8) & 0xf) << 4)); + RADEON_WRITE_PLL(dev_priv, R500_DYN_SCLK_PWMEM_PIPE, (1 | ((gb_pipe_sel >> 8) & 0xf) << 4)); RADEON_WRITE(R500_SU_REG_DEST, ((1 << dev_priv->num_gb_pipes) - 1)); } RADEON_WRITE(R300_GB_TILE_CONFIG, gb_tile_config); @@ -331,6 +444,7 @@ static void radeon_cp_load_microcode(drm_radeon_private_t * dev_priv) ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_R350) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV350) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV380) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS400) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS480)) { DRM_INFO("Loading R300 Microcode\n"); for (i = 0; i < 256; i++) { @@ -419,14 +533,20 @@ static void radeon_do_cp_start(drm_radeon_private_t * dev_priv) dev_priv->cp_running = 1; - BEGIN_RING(6); - + BEGIN_RING(8); + /* isync can only be written through cp on r5xx write it here */ + OUT_RING(CP_PACKET0(RADEON_ISYNC_CNTL, 0)); + OUT_RING(RADEON_ISYNC_ANY2D_IDLE3D | + RADEON_ISYNC_ANY3D_IDLE2D | + RADEON_ISYNC_WAIT_IDLEGUI | + RADEON_ISYNC_CPSCRATCH_IDLEGUI); RADEON_PURGE_CACHE(); RADEON_PURGE_ZCACHE(); RADEON_WAIT_UNTIL_IDLE(); - ADVANCE_RING(); COMMIT_RING(); + + dev_priv->track_flush |= RADEON_FLUSH_EMITED | RADEON_PURGE_EMITED; } /* Reset the Command Processor. This will not flush any pending @@ -470,15 +590,15 @@ static int radeon_do_engine_reset(struct drm_device * dev) if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV410) { /* may need something similar for newer chips */ clock_cntl_index = RADEON_READ(RADEON_CLOCK_CNTL_INDEX); - mclk_cntl = RADEON_READ_PLL(dev, RADEON_MCLK_CNTL); - - RADEON_WRITE_PLL(RADEON_MCLK_CNTL, (mclk_cntl | - RADEON_FORCEON_MCLKA | - RADEON_FORCEON_MCLKB | - RADEON_FORCEON_YCLKA | - RADEON_FORCEON_YCLKB | - RADEON_FORCEON_MC | - RADEON_FORCEON_AIC)); + mclk_cntl = RADEON_READ_PLL(dev_priv, RADEON_MCLK_CNTL); + + RADEON_WRITE_PLL(dev_priv, RADEON_MCLK_CNTL, (mclk_cntl | + RADEON_FORCEON_MCLKA | + RADEON_FORCEON_MCLKB | + RADEON_FORCEON_YCLKA | + RADEON_FORCEON_YCLKB | + RADEON_FORCEON_MC | + RADEON_FORCEON_AIC)); } rbbm_soft_reset = RADEON_READ(RADEON_RBBM_SOFT_RESET); @@ -503,7 +623,7 @@ static int radeon_do_engine_reset(struct drm_device * dev) RADEON_READ(RADEON_RBBM_SOFT_RESET); if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV410) { - RADEON_WRITE_PLL(RADEON_MCLK_CNTL, mclk_cntl); + RADEON_WRITE_PLL(dev_priv, RADEON_MCLK_CNTL, mclk_cntl); RADEON_WRITE(RADEON_CLOCK_CNTL_INDEX, clock_cntl_index); RADEON_WRITE(RADEON_RBBM_SOFT_RESET, rbbm_soft_reset); } @@ -519,7 +639,8 @@ static int radeon_do_engine_reset(struct drm_device * dev) dev_priv->cp_running = 0; /* Reset any pending vertex, indirect buffers */ - radeon_freelist_reset(dev); + if (dev->dma) + radeon_freelist_reset(dev); return 0; } @@ -537,14 +658,17 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, */ if (!dev_priv->new_memmap) radeon_write_fb_location(dev_priv, - ((dev_priv->gart_vm_start - 1) & 0xffff0000) - | (dev_priv->fb_location >> 16)); - + ((dev_priv->gart_vm_start - 1) & 0xffff0000) + | (dev_priv->fb_location >> 16)); + + if (dev_priv->mm.ring) { + ring_start = dev_priv->mm.ring->offset + + dev_priv->gart_vm_start; + } else #if __OS_HAS_AGP if (dev_priv->flags & RADEON_IS_AGP) { - RADEON_WRITE(RADEON_AGP_BASE, (unsigned int)dev->agp->base); - if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R200) - RADEON_WRITE(RADEON_AGP_BASE_2, 0); + radeon_write_agp_base(dev_priv, dev->agp->base); + radeon_write_agp_location(dev_priv, (((dev_priv->gart_vm_start - 1 + dev_priv->gart_size) & 0xffff0000) | @@ -570,6 +694,12 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, SET_RING_HEAD(dev_priv, cur_read_ptr); dev_priv->ring.tail = cur_read_ptr; + + if (dev_priv->mm.ring_read_ptr) { + RADEON_WRITE(RADEON_CP_RB_RPTR_ADDR, + dev_priv->mm.ring_read_ptr->offset + + dev_priv->gart_vm_start); + } else #if __OS_HAS_AGP if (dev_priv->flags & RADEON_IS_AGP) { RADEON_WRITE(RADEON_CP_RB_RPTR_ADDR, @@ -618,9 +748,14 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, RADEON_WRITE(RADEON_SCRATCH_ADDR, RADEON_READ(RADEON_CP_RB_RPTR_ADDR) + RADEON_SCRATCH_REG_OFFSET); - dev_priv->scratch = ((__volatile__ u32 *) - dev_priv->ring_rptr->handle + - (RADEON_SCRATCH_REG_OFFSET / sizeof(u32))); + if (dev_priv->mm.ring_read_ptr) + dev_priv->scratch = ((__volatile__ u32 *) + dev_priv->mm.ring_read_ptr_map.virtual + + (RADEON_SCRATCH_REG_OFFSET / sizeof(u32))); + else + dev_priv->scratch = ((__volatile__ u32 *) + dev_priv->ring_rptr->handle + + (RADEON_SCRATCH_REG_OFFSET / sizeof(u32))); RADEON_WRITE(RADEON_SCRATCH_UMSK, 0x7); @@ -628,15 +763,14 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, tmp = RADEON_READ(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS; RADEON_WRITE(RADEON_BUS_CNTL, tmp); - dev_priv->sarea_priv->last_frame = dev_priv->scratch[0] = 0; - RADEON_WRITE(RADEON_LAST_FRAME_REG, dev_priv->sarea_priv->last_frame); + dev_priv->scratch[0] = 0; + RADEON_WRITE(RADEON_LAST_FRAME_REG, 0); - dev_priv->sarea_priv->last_dispatch = dev_priv->scratch[1] = 0; - RADEON_WRITE(RADEON_LAST_DISPATCH_REG, - dev_priv->sarea_priv->last_dispatch); + dev_priv->scratch[1] = 0; + RADEON_WRITE(RADEON_LAST_DISPATCH_REG, 0); - dev_priv->sarea_priv->last_clear = dev_priv->scratch[2] = 0; - RADEON_WRITE(RADEON_LAST_CLEAR_REG, dev_priv->sarea_priv->last_clear); + dev_priv->scratch[2] = 0; + RADEON_WRITE(RADEON_LAST_CLEAR_REG, 0); radeon_do_wait_for_idle(dev_priv); @@ -652,15 +786,21 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, static void radeon_test_writeback(drm_radeon_private_t * dev_priv) { u32 tmp; + void *ring_read_ptr; + + if (dev_priv->mm.ring_read_ptr) + ring_read_ptr = dev_priv->mm.ring_read_ptr_map.virtual; + else + ring_read_ptr = dev_priv->ring_rptr->handle; /* Writeback doesn't seem to work everywhere, test it here and possibly * enable it if it appears to work */ - DRM_WRITE32(dev_priv->ring_rptr, RADEON_SCRATCHOFF(1), 0); + writel(0, ring_read_ptr + RADEON_SCRATCHOFF(1)); RADEON_WRITE(RADEON_SCRATCH_REG1, 0xdeadbeef); for (tmp = 0; tmp < dev_priv->usec_timeout; tmp++) { - if (DRM_READ32(dev_priv->ring_rptr, RADEON_SCRATCHOFF(1)) == + if (readl(ring_read_ptr + RADEON_SCRATCHOFF(1)) == 0xdeadbeef) break; DRM_UDELAY(1); @@ -721,14 +861,7 @@ static void radeon_set_igpgart(drm_radeon_private_t * dev_priv, int on) IGP_WRITE_MCIND(RS480_AGP_MODE_CNTL, ((1 << RS480_REQ_TYPE_SNOOP_SHIFT) | RS480_REQ_TYPE_SNOOP_DIS)); - if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) { - IGP_WRITE_MCIND(RS690_MC_AGP_BASE, - (unsigned int)dev_priv->gart_vm_start); - IGP_WRITE_MCIND(RS690_MC_AGP_BASE_2, 0); - } else { - RADEON_WRITE(RADEON_AGP_BASE, (unsigned int)dev_priv->gart_vm_start); - RADEON_WRITE(RS480_AGP_BASE_2, 0); - } + radeon_write_agp_base(dev_priv, dev_priv->gart_vm_start); dev_priv->gart_size = 32*1024*1024; temp = (((dev_priv->gart_vm_start - 1 + dev_priv->gart_size) & @@ -793,7 +926,7 @@ static void radeon_set_pciegart(drm_radeon_private_t * dev_priv, int on) } /* Enable or disable PCI GART on the chip */ -static void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on) +void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on) { u32 tmp; @@ -834,9 +967,11 @@ static void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on) } } -static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) +static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init, + struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; DRM_DEBUG("\n"); @@ -877,17 +1012,6 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) */ dev_priv->vblank_crtc = DRM_RADEON_VBLANK_CRTC1; - switch(init->func) { - case RADEON_INIT_R200_CP: - dev_priv->microcode_version = UCODE_R200; - break; - case RADEON_INIT_R300_CP: - dev_priv->microcode_version = UCODE_R300; - break; - default: - dev_priv->microcode_version = UCODE_R100; - } - dev_priv->do_boxes = 0; dev_priv->cp_mode = init->cp_mode; @@ -935,8 +1059,7 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) */ dev_priv->depth_clear.rb3d_cntl = (RADEON_PLANE_MASK_ENABLE | (dev_priv->color_fmt << 10) | - (dev_priv->microcode_version == - UCODE_R100 ? RADEON_ZBLOCK16 : 0)); + (dev_priv->chip_family < CHIP_R200 ? RADEON_ZBLOCK16 : 0)); dev_priv->depth_clear.rb3d_zstencilcntl = (dev_priv->depth_fmt | @@ -964,8 +1087,8 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) dev_priv->buffers_offset = init->buffers_offset; dev_priv->gart_textures_offset = init->gart_textures_offset; - dev_priv->sarea = drm_getsarea(dev); - if (!dev_priv->sarea) { + master_priv->sarea = drm_getsarea(dev); + if (!master_priv->sarea) { DRM_ERROR("could not find sarea!\n"); radeon_do_cleanup_cp(dev); return -EINVAL; @@ -1001,10 +1124,6 @@ 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 + - init->sarea_priv_offset); - #if __OS_HAS_AGP if (dev_priv->flags & RADEON_IS_AGP) { drm_core_ioremap(dev_priv->cp_ring, dev); @@ -1135,28 +1254,40 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) dev_priv->gart_info.table_mask = DMA_BIT_MASK(32); /* if we have an offset set from userspace */ if (dev_priv->pcigart_offset_set) { - dev_priv->gart_info.bus_addr = - dev_priv->pcigart_offset + dev_priv->fb_location; - dev_priv->gart_info.mapping.offset = - dev_priv->pcigart_offset + dev_priv->fb_aper_offset; - dev_priv->gart_info.mapping.size = - dev_priv->gart_info.table_size; - - drm_core_ioremap(&dev_priv->gart_info.mapping, dev); - dev_priv->gart_info.addr = - dev_priv->gart_info.mapping.handle; - - if (dev_priv->flags & RADEON_IS_PCIE) - dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCIE; - else - dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCI; - dev_priv->gart_info.gart_table_location = - DRM_ATI_GART_FB; - - DRM_DEBUG("Setting phys_pci_gart to %p %08lX\n", - dev_priv->gart_info.addr, - dev_priv->pcigart_offset); + /* if it came from userspace - remap it */ + if (dev_priv->pcigart_offset_set == 1) { + dev_priv->gart_info.bus_addr = + dev_priv->pcigart_offset + dev_priv->fb_location; + dev_priv->gart_info.mapping.offset = + dev_priv->pcigart_offset + dev_priv->fb_aper_offset; + dev_priv->gart_info.mapping.size = + dev_priv->gart_info.table_size; + + /* this is done by the mm now */ + drm_core_ioremap(&dev_priv->gart_info.mapping, dev); + dev_priv->gart_info.addr = + dev_priv->gart_info.mapping.handle; + + memset(dev_priv->gart_info.addr, 0, dev_priv->gart_info.table_size); + if (dev_priv->flags & RADEON_IS_PCIE) + dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCIE; + else + dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCI; + dev_priv->gart_info.gart_table_location = + DRM_ATI_GART_FB; + + DRM_DEBUG("Setting phys_pci_gart to %p %08lX\n", + dev_priv->gart_info.addr, + dev_priv->pcigart_offset); + } } else { + + if (dev_priv->flags & RADEON_IS_PCIE) { + DRM_ERROR + ("Cannot use PCI Express without GART in FB memory\n"); + radeon_do_cleanup_cp(dev); + return -EINVAL; + } if (dev_priv->flags & RADEON_IS_IGPGART) dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_IGP; else @@ -1165,12 +1296,7 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init) DRM_ATI_GART_MAIN; dev_priv->gart_info.addr = NULL; dev_priv->gart_info.bus_addr = 0; - if (dev_priv->flags & RADEON_IS_PCIE) { - DRM_ERROR - ("Cannot use PCI Express without GART in FB memory\n"); - radeon_do_cleanup_cp(dev); - return -EINVAL; - } + } if (!drm_ati_pcigart_init(dev, &dev_priv->gart_info)) { @@ -1233,8 +1359,10 @@ static int radeon_do_cleanup_cp(struct drm_device * dev) if (dev_priv->gart_info.gart_table_location == DRM_ATI_GART_FB) { - drm_core_ioremapfree(&dev_priv->gart_info.mapping, dev); - dev_priv->gart_info.addr = 0; + if (dev_priv->pcigart_offset_set == 1) { + drm_core_ioremapfree(&dev_priv->gart_info.mapping, dev); + dev_priv->gart_info.addr = NULL; + } } } /* only clear to the start of flags */ @@ -1276,6 +1404,7 @@ static int radeon_do_resume_cp(struct drm_device * dev) radeon_cp_init_ring_buffer(dev, dev_priv); radeon_do_engine_reset(dev); + radeon_irq_set_state(dev, RADEON_SW_INT_ENABLE, 1); DRM_DEBUG("radeon_do_resume_cp() complete\n"); @@ -1285,6 +1414,10 @@ static int radeon_do_resume_cp(struct drm_device * dev) int radeon_cp_init(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_init_t *init = data; + + /* on a modesetting driver ignore this stuff */ + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; LOCK_TEST_WITH_RETURN(dev, file_priv); @@ -1295,7 +1428,7 @@ int radeon_cp_init(struct drm_device *dev, void *data, struct drm_file *file_pri case RADEON_INIT_CP: case RADEON_INIT_R200_CP: case RADEON_INIT_R300_CP: - return radeon_do_init_cp(dev, init); + return radeon_do_init_cp(dev, init, file_priv); case RADEON_CLEANUP_CP: return radeon_do_cleanup_cp(dev); } @@ -1308,6 +1441,9 @@ int radeon_cp_start(struct drm_device *dev, void *data, struct drm_file *file_pr drm_radeon_private_t *dev_priv = dev->dev_private; DRM_DEBUG("\n"); + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; + LOCK_TEST_WITH_RETURN(dev, file_priv); if (dev_priv->cp_running) { @@ -1335,6 +1471,9 @@ int radeon_cp_stop(struct drm_device *dev, void *data, struct drm_file *file_pri int ret; DRM_DEBUG("\n"); + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; + LOCK_TEST_WITH_RETURN(dev, file_priv); if (!dev_priv->cp_running) @@ -1373,6 +1512,9 @@ void radeon_do_release(struct drm_device * dev) drm_radeon_private_t *dev_priv = dev->dev_private; int i, ret; + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return; + if (dev_priv) { if (dev_priv->cp_running) { /* Stop the cp */ @@ -1411,6 +1553,9 @@ void radeon_do_release(struct drm_device * dev) radeon_mem_takedown(&(dev_priv->gart_heap)); radeon_mem_takedown(&(dev_priv->fb_heap)); + + radeon_gem_mm_fini(dev); + /* deallocate kernel resources */ radeon_do_cleanup_cp(dev); } @@ -1423,6 +1568,9 @@ int radeon_cp_reset(struct drm_device *dev, void *data, struct drm_file *file_pr drm_radeon_private_t *dev_priv = dev->dev_private; DRM_DEBUG("\n"); + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; + LOCK_TEST_WITH_RETURN(dev, file_priv); if (!dev_priv) { @@ -1443,7 +1591,9 @@ int radeon_cp_idle(struct drm_device *dev, void *data, struct drm_file *file_pri drm_radeon_private_t *dev_priv = dev->dev_private; DRM_DEBUG("\n"); - LOCK_TEST_WITH_RETURN(dev, file_priv); + + if (!drm_core_check_feature(dev, DRIVER_MODESET)) + LOCK_TEST_WITH_RETURN(dev, file_priv); return radeon_do_cp_idle(dev_priv); } @@ -1453,6 +1603,9 @@ int radeon_cp_idle(struct drm_device *dev, void *data, struct drm_file *file_pri int radeon_cp_resume(struct drm_device *dev, void *data, struct drm_file *file_priv) { + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; + return radeon_do_resume_cp(dev); } @@ -1460,6 +1613,9 @@ int radeon_engine_reset(struct drm_device *dev, void *data, struct drm_file *fil { DRM_DEBUG("\n"); + if (drm_core_check_feature(dev, DRIVER_MODESET)) + return 0; + LOCK_TEST_WITH_RETURN(dev, file_priv); return radeon_do_engine_reset(dev); @@ -1682,6 +1838,541 @@ int radeon_cp_buffers(struct drm_device *dev, void *data, struct drm_file *file_ return ret; } +static void radeon_get_vram_type(struct drm_device *dev) +{ + struct drm_radeon_private *dev_priv = dev->dev_private; + uint32_t tmp; + + if (dev_priv->flags & RADEON_IS_IGP || (dev_priv->chip_family >= CHIP_R300)) + dev_priv->is_ddr = true; + else if (RADEON_READ(RADEON_MEM_SDRAM_MODE_REG) & RADEON_MEM_CFG_TYPE_DDR) + dev_priv->is_ddr = true; + else + dev_priv->is_ddr = false; + + if ((dev_priv->chip_family >= CHIP_R600) && + (dev_priv->chip_family <= CHIP_RV635)) { + int chansize; + + tmp = RADEON_READ(R600_RAMCFG); + if (tmp & R600_CHANSIZE_OVERRIDE) + chansize = 16; + else if (tmp & R600_CHANSIZE) + chansize = 64; + else + chansize = 32; + + if (dev_priv->chip_family == CHIP_R600) + dev_priv->ram_width = 8 * chansize; + else if (dev_priv->chip_family == CHIP_RV670) + dev_priv->ram_width = 4 * chansize; + else if ((dev_priv->chip_family == CHIP_RV610) || + (dev_priv->chip_family == CHIP_RV620)) + dev_priv->ram_width = chansize; + else if ((dev_priv->chip_family == CHIP_RV630) || + (dev_priv->chip_family == CHIP_RV635)) + dev_priv->ram_width = 2 * chansize; + } else if (dev_priv->chip_family == CHIP_RV515) { + tmp = radeon_read_mc_reg(dev_priv, RV515_MC_CNTL); + tmp &= RV515_MEM_NUM_CHANNELS_MASK; + switch (tmp) { + case 0: dev_priv->ram_width = 64; break; + case 1: dev_priv->ram_width = 128; break; + default: dev_priv->ram_width = 128; break; + } + } else if ((dev_priv->chip_family >= CHIP_R520) && + (dev_priv->chip_family <= CHIP_RV570)) { + tmp = radeon_read_mc_reg(dev_priv, R520_MC_CNTL0); + switch ((tmp & R520_MEM_NUM_CHANNELS_MASK) >> R520_MEM_NUM_CHANNELS_SHIFT) { + case 0: dev_priv->ram_width = 32; break; + case 1: dev_priv->ram_width = 64; break; + case 2: dev_priv->ram_width = 128; break; + case 3: dev_priv->ram_width = 256; break; + default: dev_priv->ram_width = 128; break; + } + } else if ((dev_priv->chip_family == CHIP_RV100) || + (dev_priv->chip_family == CHIP_RS100) || + (dev_priv->chip_family == CHIP_RS200)) { + tmp = RADEON_READ(RADEON_MEM_CNTL); + if (tmp & RV100_HALF_MODE) + dev_priv->ram_width = 32; + else + dev_priv->ram_width = 64; + + if (dev_priv->flags & RADEON_SINGLE_CRTC) { + dev_priv->ram_width /= 4; + dev_priv->is_ddr = true; + } + } else if (dev_priv->chip_family <= CHIP_RV280) { + tmp = RADEON_READ(RADEON_MEM_CNTL); + if (tmp & RADEON_MEM_NUM_CHANNELS_MASK) + dev_priv->ram_width = 128; + else + dev_priv->ram_width = 64; + } else { + /* newer IGPs */ + dev_priv->ram_width = 128; + } + DRM_DEBUG("RAM width %d bits %cDR\n", dev_priv->ram_width, dev_priv->is_ddr ? 'D' : 'S'); +} + +static void radeon_force_some_clocks(struct drm_device *dev) +{ + struct drm_radeon_private *dev_priv = dev->dev_private; + uint32_t tmp; + + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp |= RADEON_SCLK_FORCE_CP | RADEON_SCLK_FORCE_VIP; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); +} + +static void radeon_set_dynamic_clock(struct drm_device *dev, int mode) +{ + struct drm_radeon_private *dev_priv = dev->dev_private; + uint32_t tmp; + + switch(mode) { + case 0: + if (dev_priv->flags & RADEON_SINGLE_CRTC) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp |= (RADEON_SCLK_FORCE_CP | RADEON_SCLK_FORCE_HDP | + RADEON_SCLK_FORCE_DISP1 | RADEON_SCLK_FORCE_TOP | + RADEON_SCLK_FORCE_E2 | RADEON_SCLK_FORCE_SE | + RADEON_SCLK_FORCE_IDCT | RADEON_SCLK_FORCE_VIP | + RADEON_SCLK_FORCE_RE | RADEON_SCLK_FORCE_PB | + RADEON_SCLK_FORCE_TAM | RADEON_SCLK_FORCE_TDM | + RADEON_SCLK_FORCE_RB); + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + } else if (dev_priv->chip_family == CHIP_RV350) { + /* for RV350/M10, no delays are required. */ + tmp = RADEON_READ_PLL(dev_priv, R300_SCLK_CNTL2); + tmp |= (R300_SCLK_FORCE_TCL | + R300_SCLK_FORCE_GA | + R300_SCLK_FORCE_CBA); + RADEON_WRITE_PLL(dev_priv, R300_SCLK_CNTL2, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp &= ~(RADEON_SCLK_FORCE_DISP2 | RADEON_SCLK_FORCE_CP | + RADEON_SCLK_FORCE_HDP | RADEON_SCLK_FORCE_DISP1 | + RADEON_SCLK_FORCE_TOP | RADEON_SCLK_FORCE_E2 | + R300_SCLK_FORCE_VAP | RADEON_SCLK_FORCE_IDCT | + RADEON_SCLK_FORCE_VIP | R300_SCLK_FORCE_SR | + R300_SCLK_FORCE_PX | R300_SCLK_FORCE_TX | + R300_SCLK_FORCE_US | RADEON_SCLK_FORCE_TV_SCLK | + R300_SCLK_FORCE_SU | RADEON_SCLK_FORCE_OV0); + tmp |= RADEON_DYN_STOP_LAT_MASK; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_MORE_CNTL); + tmp &= ~RADEON_SCLK_MORE_FORCEON; + tmp |= RADEON_SCLK_MORE_MAX_DYN_STOP_LAT; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_MORE_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_VCLK_ECP_CNTL); + tmp |= (RADEON_PIXCLK_ALWAYS_ONb | + RADEON_PIXCLK_DAC_ALWAYS_ONb); + RADEON_WRITE_PLL(dev_priv, RADEON_VCLK_ECP_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_PIXCLKS_CNTL); + tmp |= (RADEON_PIX2CLK_ALWAYS_ONb | + RADEON_PIX2CLK_DAC_ALWAYS_ONb | + RADEON_DISP_TVOUT_PIXCLK_TV_ALWAYS_ONb | + R300_DVOCLK_ALWAYS_ONb | + RADEON_PIXCLK_BLEND_ALWAYS_ONb | + RADEON_PIXCLK_GV_ALWAYS_ONb | + R300_PIXCLK_DVO_ALWAYS_ONb | + RADEON_PIXCLK_LVDS_ALWAYS_ONb | + RADEON_PIXCLK_TMDS_ALWAYS_ONb | + R300_PIXCLK_TRANS_ALWAYS_ONb | + R300_PIXCLK_TVO_ALWAYS_ONb | + R300_P2G2CLK_ALWAYS_ONb | + R300_P2G2CLK_ALWAYS_ONb); + RADEON_WRITE_PLL(dev_priv, RADEON_PIXCLKS_CNTL, tmp); + } else { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp |= (RADEON_SCLK_FORCE_CP | RADEON_SCLK_FORCE_E2); + tmp |= RADEON_SCLK_FORCE_SE; + + if ( dev_priv->flags & RADEON_SINGLE_CRTC ) { + tmp |= ( RADEON_SCLK_FORCE_RB | + RADEON_SCLK_FORCE_TDM | + RADEON_SCLK_FORCE_TAM | + RADEON_SCLK_FORCE_PB | + RADEON_SCLK_FORCE_RE | + RADEON_SCLK_FORCE_VIP | + RADEON_SCLK_FORCE_IDCT | + RADEON_SCLK_FORCE_TOP | + RADEON_SCLK_FORCE_DISP1 | + RADEON_SCLK_FORCE_DISP2 | + RADEON_SCLK_FORCE_HDP ); + } else if ((dev_priv->chip_family == CHIP_R300) || + (dev_priv->chip_family == CHIP_R350)) { + tmp |= ( RADEON_SCLK_FORCE_HDP | + RADEON_SCLK_FORCE_DISP1 | + RADEON_SCLK_FORCE_DISP2 | + RADEON_SCLK_FORCE_TOP | + RADEON_SCLK_FORCE_IDCT | + RADEON_SCLK_FORCE_VIP); + } + + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + + udelay(16000); + + if ((dev_priv->chip_family == CHIP_R300) || + (dev_priv->chip_family == CHIP_R350)) { + tmp = RADEON_READ_PLL(dev_priv, R300_SCLK_CNTL2); + tmp |= ( R300_SCLK_FORCE_TCL | + R300_SCLK_FORCE_GA | + R300_SCLK_FORCE_CBA); + RADEON_WRITE_PLL(dev_priv, R300_SCLK_CNTL2, tmp); + udelay(16000); + } + + if (dev_priv->flags & RADEON_IS_IGP) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_MCLK_CNTL); + tmp &= ~(RADEON_FORCEON_MCLKA | + RADEON_FORCEON_YCLKA); + RADEON_WRITE_PLL(dev_priv, RADEON_MCLK_CNTL, tmp); + udelay(16000); + } + + if ((dev_priv->chip_family == CHIP_RV200) || + (dev_priv->chip_family == CHIP_RV250) || + (dev_priv->chip_family == CHIP_RV280)) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_MORE_CNTL); + tmp |= RADEON_SCLK_MORE_FORCEON; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_MORE_CNTL, tmp); + udelay(16000); + } + + tmp = RADEON_READ_PLL(dev_priv, RADEON_PIXCLKS_CNTL); + tmp &= ~(RADEON_PIX2CLK_ALWAYS_ONb | + RADEON_PIX2CLK_DAC_ALWAYS_ONb | + RADEON_PIXCLK_BLEND_ALWAYS_ONb | + RADEON_PIXCLK_GV_ALWAYS_ONb | + RADEON_PIXCLK_DIG_TMDS_ALWAYS_ONb | + RADEON_PIXCLK_LVDS_ALWAYS_ONb | + RADEON_PIXCLK_TMDS_ALWAYS_ONb); + + RADEON_WRITE_PLL(dev_priv, RADEON_PIXCLKS_CNTL, tmp); + udelay(16000); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_VCLK_ECP_CNTL); + tmp &= ~(RADEON_PIXCLK_ALWAYS_ONb | + RADEON_PIXCLK_DAC_ALWAYS_ONb); + RADEON_WRITE_PLL(dev_priv, RADEON_VCLK_ECP_CNTL, tmp); + } + DRM_DEBUG("Dynamic Clock Scaling Disabled\n"); + break; + case 1: + if (dev_priv->flags & RADEON_SINGLE_CRTC) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + if ((RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) > + RADEON_CFG_ATI_REV_A13) { + tmp &= ~(RADEON_SCLK_FORCE_CP | RADEON_SCLK_FORCE_RB); + } + tmp &= ~(RADEON_SCLK_FORCE_HDP | RADEON_SCLK_FORCE_DISP1 | + RADEON_SCLK_FORCE_TOP | RADEON_SCLK_FORCE_SE | + RADEON_SCLK_FORCE_IDCT | RADEON_SCLK_FORCE_RE | + RADEON_SCLK_FORCE_PB | RADEON_SCLK_FORCE_TAM | + RADEON_SCLK_FORCE_TDM); + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + } else if ((dev_priv->chip_family == CHIP_R300) || + (dev_priv->chip_family == CHIP_R350) || + (dev_priv->chip_family == CHIP_RV350)) { + if (dev_priv->chip_family == CHIP_RV350) { + tmp = RADEON_READ_PLL(dev_priv, R300_SCLK_CNTL2); + tmp &= ~(R300_SCLK_FORCE_TCL | + R300_SCLK_FORCE_GA | + R300_SCLK_FORCE_CBA); + tmp |= (R300_SCLK_TCL_MAX_DYN_STOP_LAT | + R300_SCLK_GA_MAX_DYN_STOP_LAT | + R300_SCLK_CBA_MAX_DYN_STOP_LAT); + RADEON_WRITE_PLL(dev_priv, R300_SCLK_CNTL2, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp &= ~(RADEON_SCLK_FORCE_DISP2 | RADEON_SCLK_FORCE_CP | + RADEON_SCLK_FORCE_HDP | RADEON_SCLK_FORCE_DISP1 | + RADEON_SCLK_FORCE_TOP | RADEON_SCLK_FORCE_E2 | + R300_SCLK_FORCE_VAP | RADEON_SCLK_FORCE_IDCT | + RADEON_SCLK_FORCE_VIP | R300_SCLK_FORCE_SR | + R300_SCLK_FORCE_PX | R300_SCLK_FORCE_TX | + R300_SCLK_FORCE_US | RADEON_SCLK_FORCE_TV_SCLK | + R300_SCLK_FORCE_SU | RADEON_SCLK_FORCE_OV0); + tmp |= RADEON_DYN_STOP_LAT_MASK; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_MORE_CNTL); + tmp &= ~RADEON_SCLK_MORE_FORCEON; + tmp |= RADEON_SCLK_MORE_MAX_DYN_STOP_LAT; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_MORE_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_VCLK_ECP_CNTL); + tmp |= (RADEON_PIXCLK_ALWAYS_ONb | + RADEON_PIXCLK_DAC_ALWAYS_ONb); + RADEON_WRITE_PLL(dev_priv, RADEON_VCLK_ECP_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_PIXCLKS_CNTL); + tmp |= (RADEON_PIX2CLK_ALWAYS_ONb | + RADEON_PIX2CLK_DAC_ALWAYS_ONb | + RADEON_DISP_TVOUT_PIXCLK_TV_ALWAYS_ONb | + R300_DVOCLK_ALWAYS_ONb | + RADEON_PIXCLK_BLEND_ALWAYS_ONb | + RADEON_PIXCLK_GV_ALWAYS_ONb | + R300_PIXCLK_DVO_ALWAYS_ONb | + RADEON_PIXCLK_LVDS_ALWAYS_ONb | + RADEON_PIXCLK_TMDS_ALWAYS_ONb | + R300_PIXCLK_TRANS_ALWAYS_ONb | + R300_PIXCLK_TVO_ALWAYS_ONb | + R300_P2G2CLK_ALWAYS_ONb | + R300_P2G2CLK_ALWAYS_ONb); + RADEON_WRITE_PLL(dev_priv, RADEON_PIXCLKS_CNTL, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_MCLK_MISC); + tmp |= (RADEON_MC_MCLK_DYN_ENABLE | + RADEON_IO_MCLK_DYN_ENABLE); + RADEON_WRITE_PLL(dev_priv, RADEON_MCLK_MISC, tmp); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_MCLK_CNTL); + tmp |= (RADEON_FORCEON_MCLKA | + RADEON_FORCEON_MCLKB); + + tmp &= ~(RADEON_FORCEON_YCLKA | + RADEON_FORCEON_YCLKB | + RADEON_FORCEON_MC); + + /* Some releases of vbios have set DISABLE_MC_MCLKA + and DISABLE_MC_MCLKB bits in the vbios table. Setting these + bits will cause H/W hang when reading video memory with dynamic clocking + enabled. */ + if ((tmp & R300_DISABLE_MC_MCLKA) && + (tmp & R300_DISABLE_MC_MCLKB)) { + /* If both bits are set, then check the active channels */ + tmp = RADEON_READ_PLL(dev_priv, RADEON_MCLK_CNTL); + if (dev_priv->ram_width == 64) { + if (RADEON_READ(RADEON_MEM_CNTL) & R300_MEM_USE_CD_CH_ONLY) + tmp &= ~R300_DISABLE_MC_MCLKB; + else + tmp &= ~R300_DISABLE_MC_MCLKA; + } else { + tmp &= ~(R300_DISABLE_MC_MCLKA | + R300_DISABLE_MC_MCLKB); + } + } + + RADEON_WRITE_PLL(dev_priv, RADEON_MCLK_CNTL, tmp); + } else { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + tmp &= ~(R300_SCLK_FORCE_VAP); + tmp |= RADEON_SCLK_FORCE_CP; + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + udelay(15000); + + tmp = RADEON_READ_PLL(dev_priv, R300_SCLK_CNTL2); + tmp &= ~(R300_SCLK_FORCE_TCL | + R300_SCLK_FORCE_GA | + R300_SCLK_FORCE_CBA); + RADEON_WRITE_PLL(dev_priv, R300_SCLK_CNTL2, tmp); + } + } else { + tmp = RADEON_READ_PLL(dev_priv, RADEON_CLK_PWRMGT_CNTL); + tmp &= ~(RADEON_ACTIVE_HILO_LAT_MASK | + RADEON_DISP_DYN_STOP_LAT_MASK | + RADEON_DYN_STOP_MODE_MASK); + + tmp |= (RADEON_ENGIN_DYNCLK_MODE | + (0x01 << RADEON_ACTIVE_HILO_LAT_SHIFT)); + RADEON_WRITE_PLL(dev_priv, RADEON_CLK_PWRMGT_CNTL, tmp); + udelay(15000); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_CLK_PIN_CNTL); + tmp |= RADEON_SCLK_DYN_START_CNTL; + RADEON_WRITE_PLL(dev_priv, RADEON_CLK_PIN_CNTL, tmp); + udelay(15000); + + /* When DRI is enabled, setting DYN_STOP_LAT to zero can cause some R200 + to lockup randomly, leave them as set by BIOS. + */ + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_CNTL); + /*tmp &= RADEON_SCLK_SRC_SEL_MASK;*/ + tmp &= ~RADEON_SCLK_FORCEON_MASK; + + /*RAGE_6::A11 A12 A12N1 A13, RV250::A11 A12, R300*/ + if (((dev_priv->chip_family == CHIP_RV250) && + ((RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) < + RADEON_CFG_ATI_REV_A13)) || + ((dev_priv->chip_family == CHIP_RV100) && + ((RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) <= + RADEON_CFG_ATI_REV_A13))){ + tmp |= RADEON_SCLK_FORCE_CP; + tmp |= RADEON_SCLK_FORCE_VIP; + } + + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_CNTL, tmp); + + if ((dev_priv->chip_family == CHIP_RV200) || + (dev_priv->chip_family == CHIP_RV250) || + (dev_priv->chip_family == CHIP_RV280)) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_SCLK_MORE_CNTL); + tmp &= ~RADEON_SCLK_MORE_FORCEON; + + /* RV200::A11 A12 RV250::A11 A12 */ + if (((dev_priv->chip_family == CHIP_RV200) || + (dev_priv->chip_family == CHIP_RV250)) && + ((RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) < + RADEON_CFG_ATI_REV_A13)) { + tmp |= RADEON_SCLK_MORE_FORCEON; + } + RADEON_WRITE_PLL(dev_priv, RADEON_SCLK_MORE_CNTL, tmp); + udelay(15000); + } + + /* RV200::A11 A12, RV250::A11 A12 */ + if (((dev_priv->chip_family == CHIP_RV200) || + (dev_priv->chip_family == CHIP_RV250)) && + ((RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) < + RADEON_CFG_ATI_REV_A13)) { + tmp = RADEON_READ_PLL(dev_priv, RADEON_PLL_PWRMGT_CNTL); + tmp |= RADEON_TCL_BYPASS_DISABLE; + RADEON_WRITE_PLL(dev_priv, RADEON_PLL_PWRMGT_CNTL, tmp); + } + udelay(15000); + + /*enable dynamic mode for display clocks (PIXCLK and PIX2CLK)*/ + tmp = RADEON_READ_PLL(dev_priv, RADEON_PIXCLKS_CNTL); + tmp |= (RADEON_PIX2CLK_ALWAYS_ONb | + RADEON_PIX2CLK_DAC_ALWAYS_ONb | + RADEON_PIXCLK_BLEND_ALWAYS_ONb | + RADEON_PIXCLK_GV_ALWAYS_ONb | + RADEON_PIXCLK_DIG_TMDS_ALWAYS_ONb | + RADEON_PIXCLK_LVDS_ALWAYS_ONb | + RADEON_PIXCLK_TMDS_ALWAYS_ONb); + + RADEON_WRITE_PLL(dev_priv, RADEON_PIXCLKS_CNTL, tmp); + udelay(15000); + + tmp = RADEON_READ_PLL(dev_priv, RADEON_VCLK_ECP_CNTL); + tmp |= (RADEON_PIXCLK_ALWAYS_ONb | + RADEON_PIXCLK_DAC_ALWAYS_ONb); + + RADEON_WRITE_PLL(dev_priv, RADEON_VCLK_ECP_CNTL, tmp); + udelay(15000); + } + DRM_DEBUG("Dynamic Clock Scaling Enabled\n"); + break; + default: + break; + } + +} + +int radeon_modeset_cp_init(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + /* allocate a ring and ring rptr bits from GART space */ + /* these are allocated in GEM files */ + + dev_priv->usec_timeout = RADEON_DEFAULT_CP_TIMEOUT; + dev_priv->ring.size = RADEON_DEFAULT_RING_SIZE; + dev_priv->cp_mode = RADEON_CSQ_PRIBM_INDBM; + + dev_priv->ring.start = (u32 *)(void *)(unsigned long)dev_priv->mm.ring_map.virtual; + dev_priv->ring.end = (u32 *)(void *)(unsigned long)dev_priv->mm.ring_map.virtual + + dev_priv->ring.size / sizeof(u32); + dev_priv->ring.size_l2qw = drm_order(dev_priv->ring.size / 8); + dev_priv->ring.rptr_update = 4096; + dev_priv->ring.rptr_update_l2qw = drm_order(4096 / 8); + dev_priv->ring.fetch_size = 32; + dev_priv->ring.fetch_size_l2ow = drm_order(32 / 16); + dev_priv->ring.tail_mask = (dev_priv->ring.size / sizeof(u32)) - 1; + dev_priv->ring.high_mark = RADEON_RING_HIGH_MARK; + + dev_priv->new_memmap = 1; + + radeon_cp_load_microcode(dev_priv); + + DRM_DEBUG("ring offset is %x %x\n", dev_priv->mm.ring->offset, dev_priv->mm.ring_read_ptr->offset); + + radeon_cp_init_ring_buffer(dev, dev_priv); + + radeon_do_engine_reset(dev); + radeon_test_writeback(dev_priv); + + radeon_do_cp_start(dev_priv); + return 0; +} + +static bool radeon_get_bios(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + u8 __iomem *bios; + size_t size; + uint16_t tmp; + + bios = pci_map_rom(dev->pdev, &size); + if (!bios) + return -1; + + dev_priv->bios = kmalloc(size, GFP_KERNEL); + if (!dev_priv->bios) { + pci_unmap_rom(dev->pdev, bios); + return -1; + } + + memcpy(dev_priv->bios, bios, size); + + pci_unmap_rom(dev->pdev, bios); + + if (dev_priv->bios[0] != 0x55 || dev_priv->bios[1] != 0xaa) + goto free_bios; + + dev_priv->bios_header_start = radeon_bios16(dev_priv, 0x48); + + if (!dev_priv->bios_header_start) + goto free_bios; + + tmp = dev_priv->bios_header_start + 4; + + if (!memcmp(dev_priv->bios + tmp, "ATOM", 4) || + !memcmp(dev_priv->bios + tmp, "MOTA", 4)) + dev_priv->is_atom_bios = true; + else + dev_priv->is_atom_bios = false; + + DRM_DEBUG("%sBIOS detected\n", dev_priv->is_atom_bios ? "ATOM" : "COM"); + return true; +free_bios: + kfree(dev_priv->bios); + dev_priv->bios = NULL; + return false; +} + +int radeon_modeset_preinit(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + static struct card_info card; + int ret; + + card.dev = dev; + card.reg_read = cail_reg_read; + card.reg_write = cail_reg_write; + card.mc_read = cail_mc_read; + card.mc_write = cail_mc_write; + + ret = radeon_get_bios(dev); + if (!ret) + return -1; + + if (dev_priv->is_atom_bios) { + dev_priv->mode_info.atom_context = atom_parse(&card, dev_priv->bios); + radeon_get_clock_info(dev); + } + return 0; +} + + int radeon_driver_load(struct drm_device *dev, unsigned long flags) { drm_radeon_private_t *dev_priv; @@ -1714,6 +2405,7 @@ int radeon_driver_load(struct drm_device *dev, unsigned long flags) break; } + dev_priv->chip_family = flags & RADEON_FAMILY_MASK; if (drm_device_is_agp(dev)) dev_priv->flags |= RADEON_IS_AGP; else if (drm_device_is_pcie(dev)) @@ -1721,11 +2413,122 @@ int radeon_driver_load(struct drm_device *dev, unsigned long flags) else dev_priv->flags |= RADEON_IS_PCI; + + DRM_DEBUG("%s card detected\n", ((dev_priv->flags & RADEON_IS_AGP) ? "AGP" : (((dev_priv->flags & RADEON_IS_PCIE) ? "PCIE" : "PCI")))); + + ret = drm_addmap(dev, drm_get_resource_start(dev, 2), + drm_get_resource_len(dev, 2), _DRM_REGISTERS, + _DRM_DRIVER | _DRM_READ_ONLY, &dev_priv->mmio); + if (ret != 0) + return ret; + + if (drm_core_check_feature(dev, DRIVER_MODESET)) + radeon_modeset_preinit(dev); + + + radeon_get_vram_type(dev); + + dev_priv->pll_errata = 0; + + if (dev_priv->chip_family == CHIP_R300 && + (RADEON_READ(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) == RADEON_CFG_ATI_REV_A11) + dev_priv->pll_errata |= CHIP_ERRATA_R300_CG; + + if (dev_priv->chip_family == CHIP_RV200 || + dev_priv->chip_family == CHIP_RS200) + dev_priv->pll_errata |= CHIP_ERRATA_PLL_DUMMYREADS; + + + if (dev_priv->chip_family == CHIP_RV100 || + dev_priv->chip_family == CHIP_RS100 || + dev_priv->chip_family == CHIP_RS200) + dev_priv->pll_errata |= CHIP_ERRATA_PLL_DELAY; + + + if (drm_core_check_feature(dev, DRIVER_MODESET)) { + if ((dev_priv->flags & RADEON_IS_MOBILITY) && !radeon_is_avivo(dev_priv)) { + radeon_set_dynamic_clock(dev, radeon_dynclks); + } else if (radeon_is_avivo(dev_priv)) { + if (radeon_dynclks) { + radeon_atom_static_pwrmgt_setup(dev, 1); + radeon_atom_dyn_clk_setup(dev, 1); + } + } + radeon_force_some_clocks(dev); + } + + /* init memory manager - start with all of VRAM and a 32MB GART aperture for now */ + dev_priv->fb_aper_offset = drm_get_resource_start(dev, 0); + + drm_bo_driver_init(dev); + + if (drm_core_check_feature(dev, DRIVER_MODESET)) { + + dev_priv->fb_location = (radeon_read_fb_location(dev_priv) & 0xffff) << 16; + dev_priv->fb_size = + ((radeon_read_fb_location(dev_priv) & 0xffff0000u) + 0x10000) + - dev_priv->fb_location; + radeon_gem_mm_init(dev); + radeon_modeset_init(dev); + + radeon_modeset_cp_init(dev); + dev->devname = kstrdup(DRIVER_NAME, GFP_KERNEL); + + drm_irq_install(dev); + } + + return ret; } + +int radeon_master_create(struct drm_device *dev, struct drm_master *master) +{ + struct drm_radeon_master_private *master_priv; + unsigned long sareapage; + int ret; + + master_priv = drm_calloc(1, sizeof(*master_priv), DRM_MEM_DRIVER); + if (!master_priv) + return -ENOMEM; + + /* prebuild the SAREA */ + sareapage = max(SAREA_MAX, PAGE_SIZE); + ret = drm_addmap(dev, 0, sareapage, _DRM_SHM, _DRM_CONTAINS_LOCK|_DRM_DRIVER, + &master_priv->sarea); + if (ret) { + DRM_ERROR("SAREA setup failed\n"); + return ret; + } + master_priv->sarea_priv = master_priv->sarea->handle + sizeof(struct drm_sarea); + master_priv->sarea_priv->pfCurrentPage = 0; + + master->driver_priv = master_priv; + return 0; +} + +void radeon_master_destroy(struct drm_device *dev, struct drm_master *master) +{ + struct drm_radeon_master_private *master_priv = master->driver_priv; + struct drm_radeon_private *dev_priv = dev->dev_private; + + if (!master_priv) + return; + + if (master_priv->sarea_priv && + master_priv->sarea_priv->pfCurrentPage != 0) + radeon_cp_dispatch_flip(dev, master); + + master_priv->sarea_priv = NULL; + if (master_priv->sarea) + drm_rmmap(dev, master_priv->sarea); + + drm_free(master_priv, sizeof(*master_priv), DRM_MEM_DRIVER); + + master->driver_priv = NULL; +} /* Create mappings for registers and framebuffer so userland doesn't necessarily * have to find them. */ @@ -1737,13 +2540,9 @@ int radeon_driver_firstopen(struct drm_device *dev) dev_priv->gart_info.table_size = RADEON_PCIGART_TABLE_SIZE; - ret = drm_addmap(dev, drm_get_resource_start(dev, 2), - drm_get_resource_len(dev, 2), _DRM_REGISTERS, - _DRM_READ_ONLY, &dev_priv->mmio); - if (ret != 0) - return ret; + if (!drm_core_check_feature(dev, DRIVER_MODESET)) + radeon_gem_mm_init(dev); - dev_priv->fb_aper_offset = drm_get_resource_start(dev, 0); ret = drm_addmap(dev, dev_priv->fb_aper_offset, drm_get_resource_len(dev, 0), _DRM_FRAME_BUFFER, _DRM_WRITE_COMBINING, &map); @@ -1757,9 +2556,40 @@ int radeon_driver_unload(struct drm_device *dev) { drm_radeon_private_t *dev_priv = dev->dev_private; + if (drm_core_check_feature(dev, DRIVER_MODESET)) { + drm_irq_uninstall(dev); + radeon_modeset_cleanup(dev); + radeon_gem_mm_fini(dev); + } + + drm_bo_driver_finish(dev); + drm_rmmap(dev, dev_priv->mmio); + DRM_DEBUG("\n"); drm_free(dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER); dev->dev_private = NULL; return 0; } + +void radeon_gart_flush(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + if (dev_priv->flags & RADEON_IS_IGPGART) { + IGP_READ_MCIND(dev_priv, RS480_GART_CACHE_CNTRL); + IGP_WRITE_MCIND(RS480_GART_CACHE_CNTRL, RS480_GART_CACHE_INVALIDATE); + IGP_READ_MCIND(dev_priv, RS480_GART_CACHE_CNTRL); + IGP_WRITE_MCIND(RS480_GART_CACHE_CNTRL, 0); + } else if (dev_priv->flags & RADEON_IS_PCIE) { + u32 tmp = RADEON_READ_PCIE(dev_priv, RADEON_PCIE_TX_GART_CNTL); + tmp |= RADEON_PCIE_TX_GART_INVALIDATE_TLB; + RADEON_WRITE_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp); + tmp &= ~RADEON_PCIE_TX_GART_INVALIDATE_TLB; + RADEON_WRITE_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp); + } else { + + + } + +} diff --git a/shared-core/radeon_drm.h b/shared-core/radeon_drm.h index 81944061..7fcf9305 100644 --- a/shared-core/radeon_drm.h +++ b/shared-core/radeon_drm.h @@ -457,12 +457,6 @@ typedef struct { unsigned int last_fence; } drm_radeon_sarea_t; -/* The only fence class we support */ -#define DRM_RADEON_FENCE_CLASS_ACCEL 0 -/* Fence type that guarantees read-write flush */ -#define DRM_RADEON_FENCE_TYPE_RW 2 -/* cache flushes programmed just before the fence */ -#define DRM_RADEON_FENCE_FLAG_FLUSHED 0x01000000 /* WARNING: If you change any of these defines, make sure to change the * defines in the Xserver file (xf86drmRadeon.h) @@ -502,6 +496,17 @@ typedef struct { #define DRM_RADEON_SURF_ALLOC 0x1a #define DRM_RADEON_SURF_FREE 0x1b +#define DRM_RADEON_GEM_INFO 0x1c +#define DRM_RADEON_GEM_CREATE 0x1d +#define DRM_RADEON_GEM_MMAP 0x1e +#define DRM_RADEON_GEM_PIN 0x1f +#define DRM_RADEON_GEM_UNPIN 0x20 +#define DRM_RADEON_GEM_PREAD 0x21 +#define DRM_RADEON_GEM_PWRITE 0x22 +#define DRM_RADEON_GEM_SET_DOMAIN 0x23 +#define DRM_RADEON_GEM_INDIRECT 0x24 // temporary for X server + + #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) @@ -530,6 +535,18 @@ typedef struct { #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_GEM_INFO DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_INFO, struct drm_radeon_gem_info) +#define DRM_IOCTL_RADEON_GEM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_CREATE, struct drm_radeon_gem_create) +#define DRM_IOCTL_RADEON_GEM_MMAP DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_MMAP, struct drm_radeon_gem_mmap) +#define DRM_IOCTL_RADEON_GEM_PIN DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_PIN, struct drm_radeon_gem_pin) +#define DRM_IOCTL_RADEON_GEM_UNPIN DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_UNPIN, struct drm_radeon_gem_unpin) +#define DRM_IOCTL_RADEON_GEM_PREAD DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_PREAD, struct drm_radeon_gem_pread) +#define DRM_IOCTL_RADEON_GEM_PWRITE DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_PWRITE, struct drm_radeon_gem_pwrite) +#define DRM_IOCTL_RADEON_GEM_SET_DOMAIN DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_SET_DOMAIN, struct drm_radeon_gem_set_domain) +#define DRM_IOCTL_RADEON_GEM_INDIRECT DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_INDIRECT, struct drm_radeon_gem_indirect) + + + typedef struct drm_radeon_init { enum { RADEON_INIT_CP = 0x01, @@ -756,4 +773,92 @@ typedef struct drm_radeon_surface_free { #define DRM_RADEON_VBLANK_CRTC1 1 #define DRM_RADEON_VBLANK_CRTC2 2 +#define RADEON_GEM_DOMAIN_CPU 0x1 +#define RADEON_GEM_DOMAIN_VRAM 0x2 +#define RADEON_GEM_DOMAIN_2D 0x4 +#define RADEON_GEM_DOMAIN_3D 0x8 +#define RADEON_GEM_DOMAIN_TEXTURE 0x10 +#define RADEON_GEM_DOMAIN_GPU 0x20 // for vertex buffers + +/* return to userspace start/size of gtt and vram apertures */ +struct drm_radeon_gem_info { + uint64_t gart_start; + uint64_t gart_size; + uint64_t vram_start; + uint64_t vram_size; + uint64_t vram_visible; +}; + +struct drm_radeon_gem_create { + uint64_t size; + uint64_t alignment; + uint32_t handle; + uint32_t initial_domain; // to allow VRAM to be created + uint32_t no_backing_store; // for VRAM objects - select whether they need backing store + // pretty much front/back/depth don't need it - other things do +}; + +struct drm_radeon_gem_mmap { + uint32_t handle; + uint32_t pad; + uint64_t offset; + uint64_t size; + uint64_t addr_ptr; +}; + +struct drm_radeon_gem_set_domain { + uint32_t handle; + uint32_t read_domains; + uint32_t write_domain; +}; + +struct drm_radeon_gem_exec_buffer { +}; + +struct drm_radeon_gem_pin { + uint32_t handle; + uint32_t pad; + uint64_t alignment; + uint64_t offset; +}; + +struct drm_radeon_gem_unpin { + uint32_t handle; + uint32_t pad; +}; + +struct drm_radeon_gem_busy { + uint32_t handle; + uint32_t busy; +}; + +struct drm_radeon_gem_pread { + /** Handle for the object being read. */ + uint32_t handle; + uint32_t pad; + /** Offset into the object to read from */ + uint64_t offset; + /** Length of data to read */ + uint64_t size; + /** Pointer to write the data into. */ + uint64_t data_ptr; /* void *, but pointers are not 32/64 compatible */ +}; + +struct drm_radeon_gem_pwrite { + /** Handle for the object being written to. */ + uint32_t handle; + uint32_t pad; + /** Offset into the object to write to */ + uint64_t offset; + /** Length of data to write */ + uint64_t size; + /** Pointer to read the data from. */ + uint64_t data_ptr; /* void *, but pointers are not 32/64 compatible */ +}; + +struct drm_radeon_gem_indirect { + uint32_t handle; + uint32_t used; +}; + #endif diff --git a/shared-core/radeon_drv.h b/shared-core/radeon_drv.h index 1b32b2f4..5ce97b64 100644 --- a/shared-core/radeon_drv.h +++ b/shared-core/radeon_drv.h @@ -31,6 +31,7 @@ #ifndef __RADEON_DRV_H__ #define __RADEON_DRV_H__ +#include "atom.h" /* General customization: */ @@ -38,7 +39,7 @@ #define DRIVER_NAME "radeon" #define DRIVER_DESC "ATI Radeon" -#define DRIVER_DATE "20080528" +#define DRIVER_DATE "20080613" /* Interface history: * @@ -96,13 +97,13 @@ * 1.25- Add support for r200 vertex programs (R200_EMIT_VAP_PVS_CNTL, * new packet type) * 1.26- Add support for variable size PCI(E) gart aperture - * 1.27- Add support for IGP GART + * 1.27- Add support for IGPGART * 1.28- Add support for VBL on CRTC2 * 1.29- R500 3D cmd buffer support */ #define DRIVER_MAJOR 1 -#define DRIVER_MINOR 29 +#define DRIVER_MINOR 30 #define DRIVER_PATCHLEVEL 0 /* @@ -124,23 +125,29 @@ enum radeon_family { CHIP_RV380, CHIP_R420, CHIP_RV410, + CHIP_RS400, CHIP_RS480, + CHIP_RS600, CHIP_RS690, + CHIP_RS740, CHIP_RV515, CHIP_R520, CHIP_RV530, CHIP_RV560, CHIP_RV570, CHIP_R580, + CHIP_R600, + CHIP_R630, + CHIP_RV610, + CHIP_RV630, + CHIP_RV670, + CHIP_RV620, + CHIP_RV635, + CHIP_RS780, + CHIP_RV770, CHIP_LAST, }; -enum radeon_cp_microcode_version { - UCODE_R100, - UCODE_R200, - UCODE_R300, -}; - /* * Chip flags */ @@ -158,9 +165,42 @@ enum radeon_chip_flags { RADEON_IS_IGPGART = 0x01000000UL, }; +/* + * Errata workarounds + */ +enum radeon_pll_errata { + CHIP_ERRATA_R300_CG = 0x00000001, + CHIP_ERRATA_PLL_DUMMYREADS = 0x00000002, + CHIP_ERRATA_PLL_DELAY = 0x00000004 +}; + +enum radeon_ext_tmds_chip { + RADEON_DVOCHIP_NONE, + RADEON_SIL_164, + RADEON_SIL_1178 +}; + +#if defined(__powerpc__) +enum radeon_mac_model { + RADEON_MAC_NONE, + RADEON_MAC_IBOOK, + RADEON_MAC_POWERBOOK_EXTERNAL, + RADEON_MAC_POWERBOOK_INTERNAL, + RADEON_MAC_POWERBOOK_VGA, + RADEON_MAC_MINI_EXTERNAL, + RADEON_MAC_MINI_INTERNAL, + RADEON_MAC_IMAC_G5_ISIGHT +}; +#endif + + #define GET_RING_HEAD(dev_priv) (dev_priv->writeback_works ? \ - DRM_READ32( (dev_priv)->ring_rptr, 0 ) : RADEON_READ(RADEON_CP_RB_RPTR)) -#define SET_RING_HEAD(dev_priv,val) DRM_WRITE32( (dev_priv)->ring_rptr, 0, (val) ) + (dev_priv->mm.ring_read_ptr ? readl(dev_priv->mm.ring_read_ptr_map.virtual + 0) : DRM_READ32((dev_priv)->ring_rptr, 0 )) : \ + RADEON_READ(RADEON_CP_RB_RPTR)) + +#define SET_RING_HEAD(dev_priv,val) (dev_priv->mm.ring_read_ptr ? \ + writel((val), dev_priv->mm.ring_read_ptr_map.virtual) : \ + DRM_WRITE32((dev_priv)->ring_rptr, 0, (val))) typedef struct drm_radeon_freelist { unsigned int age; @@ -221,13 +261,38 @@ struct radeon_virt_surface { struct drm_file *file_priv; }; +struct radeon_mm_info { + uint64_t vram_offset; // Offset into GPU space + uint64_t vram_size; + uint64_t vram_visible; + + uint64_t gart_start; + uint64_t gart_size; + + struct drm_buffer_object *pcie_table; + struct drm_bo_kmap_obj pcie_table_map; + + struct drm_buffer_object *ring; + struct drm_bo_kmap_obj ring_map; + + struct drm_buffer_object *ring_read_ptr; + struct drm_bo_kmap_obj ring_read_ptr_map; +}; + +#include "radeon_mode.h" + +struct drm_radeon_master_private { + drm_local_map_t *sarea; + drm_radeon_sarea_t *sarea_priv; +}; + +#define RADEON_FLUSH_EMITED (1 < 0) +#define RADEON_PURGE_EMITED (1 < 1) + typedef struct drm_radeon_private { drm_radeon_ring_buffer_t ring; - drm_radeon_sarea_t *sarea_priv; - u32 fb_location; - u32 fb_size; int new_memmap; int gart_size; @@ -245,8 +310,6 @@ typedef struct drm_radeon_private { int usec_timeout; - int microcode_version; - struct { u32 boxes; int freelist_timeouts; @@ -282,8 +345,6 @@ typedef struct drm_radeon_private { unsigned long buffers_offset; unsigned long gart_textures_offset; - drm_local_map_t *sarea; - drm_local_map_t *mmio; drm_local_map_t *cp_ring; drm_local_map_t *ring_rptr; drm_local_map_t *gart_textures; @@ -292,8 +353,8 @@ typedef struct drm_radeon_private { struct mem_block *fb_heap; /* SW interrupt */ + int counter; wait_queue_head_t swi_queue; - atomic_t swi_emitted; int vblank_crtc; uint32_t irq_enable_reg; int irq_enabled; @@ -302,10 +363,6 @@ typedef struct drm_radeon_private { struct radeon_surface surfaces[RADEON_MAX_SURFACES]; struct radeon_virt_surface virt_surfaces[2 * RADEON_MAX_SURFACES]; - unsigned long pcigart_offset; - unsigned int pcigart_offset_set; - struct drm_ati_pcigart_info gart_info; - u32 scratch_ages[5]; unsigned int crtc_last_cnt; @@ -315,7 +372,29 @@ typedef struct drm_radeon_private { uint32_t flags; /* see radeon_chip_flags */ unsigned long fb_aper_offset; + struct radeon_mm_info mm; + drm_local_map_t *mmio; + + + unsigned long pcigart_offset; + unsigned int pcigart_offset_set; + struct drm_ati_pcigart_info gart_info; + + struct radeon_mode_info mode_info; + + uint8_t *bios; /* copy of the BIOS image */ + bool is_atom_bios; + uint16_t bios_header_start; + u32 fb_location; + u32 fb_size; + bool is_ddr; + u32 ram_width; + + enum radeon_pll_errata pll_errata; + int num_gb_pipes; + int track_flush; + uint32_t chip_family; /* extract from flags */ } drm_radeon_private_t; typedef struct drm_radeon_buf_priv { @@ -330,6 +409,7 @@ typedef struct drm_radeon_kcmd_buffer { } drm_radeon_kcmd_buffer_t; extern int radeon_no_wb; +extern int radeon_dynclks; extern struct drm_ioctl_desc radeon_ioctls[]; extern int radeon_max_ioctl; @@ -375,6 +455,7 @@ extern void radeon_mem_release(struct drm_file *file_priv, struct mem_block *heap); /* radeon_irq.c */ +extern void radeon_irq_set_state(struct drm_device *dev, u32 mask, int state); extern int radeon_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int radeon_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv); @@ -417,9 +498,14 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_BOX_WAIT_IDLE 0x8 #define RADEON_BOX_TEXTURE_LOAD 0x10 +#define R600_CONFIG_MEMSIZE 0x5428 +#define R600_CONFIG_APER_SIZE 0x5430 /* Register definitions, register access macros and drmAddMap constants * for Radeon kernel driver. */ + +#include "radeon_reg.h" + #define RADEON_AGP_COMMAND 0x0f60 #define RADEON_AGP_COMMAND_PCI_CONFIG 0x0060 /* offset in PCI config */ # define RADEON_AGP_ENABLE (1<<8) @@ -522,12 +608,6 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define R520_MC_IND_WR_EN (1 << 24) #define R520_MC_IND_DATA 0x74 -#define RV515_MC_FB_LOCATION 0x01 -#define RV515_MC_AGP_LOCATION 0x02 - -#define R520_MC_FB_LOCATION 0x04 -#define R520_MC_AGP_LOCATION 0x05 - #define RADEON_MPP_TB_CONFIG 0x01c0 #define RADEON_MEM_CNTL 0x0140 #define RADEON_MEM_SDRAM_MODE_REG 0x0158 @@ -597,9 +677,11 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_SCRATCHOFF( x ) (RADEON_SCRATCH_REG_OFFSET + 4*(x)) -#define GET_SCRATCH( x ) (dev_priv->writeback_works \ - ? DRM_READ32( dev_priv->ring_rptr, RADEON_SCRATCHOFF(x) ) \ - : RADEON_READ( RADEON_SCRATCH_REG0 + 4*(x) ) ) +#define GET_SCRATCH( x ) (dev_priv->writeback_works ? \ + (dev_priv->mm.ring_read_ptr ? \ + readl(dev_priv->mm.ring_read_ptr_map.virtual + RADEON_SCRATCHOFF(0)) : \ + DRM_READ32(dev_priv->ring_rptr, RADEON_SCRATCHOFF(x))) : \ + RADEON_READ( RADEON_SCRATCH_REG0 + 4*(x))) #define RADEON_CRTC_CRNT_FRAME 0x0214 #define RADEON_CRTC2_CRNT_FRAME 0x0314 @@ -624,11 +706,11 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, # define RADEON_SW_INT_FIRE (1 << 26) # define R500_DISPLAY_INT_STATUS (1 << 0) +#define RADEON_HOST_PATH_CNTL 0x0130 +# define RADEON_HDP_SOFT_RESET (1 << 26) +# define RADEON_HDP_APER_CNTL (1 << 23) -#define RADEON_HOST_PATH_CNTL 0x0130 -# define RADEON_HDP_SOFT_RESET (1 << 26) -# define RADEON_HDP_WC_TIMEOUT_MASK (7 << 28) -# define RADEON_HDP_WC_TIMEOUT_28BCLK (7 << 28) +#define RADEON_NB_TOM 0x15c #define RADEON_ISYNC_CNTL 0x1724 # define RADEON_ISYNC_ANY2D_IDLE3D (1 << 0) @@ -667,11 +749,6 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_PP_TXFILTER_1 0x1c6c #define RADEON_PP_TXFILTER_2 0x1c84 -#define RADEON_RB2D_DSTCACHE_CTLSTAT 0x342c -# define RADEON_RB2D_DC_FLUSH (3 << 0) -# define RADEON_RB2D_DC_FREE (3 << 2) -# define RADEON_RB2D_DC_FLUSH_ALL 0xf -# define RADEON_RB2D_DC_BUSY (1 << 31) #define RADEON_RB3D_CNTL 0x1c3c # define RADEON_ALPHA_BLEND_ENABLE (1 << 0) # define RADEON_PLANE_MASK_ENABLE (1 << 1) @@ -697,14 +774,10 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define R300_ZB_ZCACHE_CTLSTAT 0x4f18 # define R300_ZC_FLUSH (1 << 0) # define R300_ZC_FREE (1 << 1) -# define R300_ZC_FLUSH_ALL 0x3 # define R300_ZC_BUSY (1 << 31) -#define RADEON_RB3D_DSTCACHE_CTLSTAT 0x325c -# define RADEON_RB3D_DC_FLUSH (3 << 0) -# define RADEON_RB3D_DC_FREE (3 << 2) -# define RADEON_RB3D_DC_FLUSH_ALL 0xf -# define RADEON_RB3D_DC_BUSY (1 << 31) #define R300_RB3D_DSTCACHE_CTLSTAT 0x4e4c +# define R300_RB3D_DC_FLUSH (2 << 0) +# define R300_RB3D_DC_FREE (2 << 2) # define R300_RB3D_DC_FINISH (1 << 4) #define RADEON_RB3D_ZSTENCILCNTL 0x1c2c # define RADEON_Z_TEST_MASK (7 << 4) @@ -977,27 +1050,6 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_NUM_VERTICES_SHIFT 16 #define RADEON_COLOR_FORMAT_CI8 2 -#define RADEON_COLOR_FORMAT_ARGB1555 3 -#define RADEON_COLOR_FORMAT_RGB565 4 -#define RADEON_COLOR_FORMAT_ARGB8888 6 -#define RADEON_COLOR_FORMAT_RGB332 7 -#define RADEON_COLOR_FORMAT_RGB8 9 -#define RADEON_COLOR_FORMAT_ARGB4444 15 - -#define RADEON_TXFORMAT_I8 0 -#define RADEON_TXFORMAT_AI88 1 -#define RADEON_TXFORMAT_RGB332 2 -#define RADEON_TXFORMAT_ARGB1555 3 -#define RADEON_TXFORMAT_RGB565 4 -#define RADEON_TXFORMAT_ARGB4444 5 -#define RADEON_TXFORMAT_ARGB8888 6 -#define RADEON_TXFORMAT_RGBA8888 7 -#define RADEON_TXFORMAT_Y8 8 -#define RADEON_TXFORMAT_VYUY422 10 -#define RADEON_TXFORMAT_YVYU422 11 -#define RADEON_TXFORMAT_DXT1 12 -#define RADEON_TXFORMAT_DXT23 14 -#define RADEON_TXFORMAT_DXT45 15 #define R200_PP_TXCBLEND_0 0x2f00 #define R200_PP_TXCBLEND_1 0x2f10 @@ -1183,18 +1235,16 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_RING_HIGH_MARK 128 #define RADEON_PCIGART_TABLE_SIZE (32*1024) +#define RADEON_DEFAULT_RING_SIZE (1024*1024) +#define RADEON_DEFAULT_CP_TIMEOUT 100000 /* usecs */ #define RADEON_READ(reg) DRM_READ32( dev_priv->mmio, (reg) ) #define RADEON_WRITE(reg,val) DRM_WRITE32( dev_priv->mmio, (reg), (val) ) #define RADEON_READ8(reg) DRM_READ8( dev_priv->mmio, (reg) ) #define RADEON_WRITE8(reg,val) DRM_WRITE8( dev_priv->mmio, (reg), (val) ) -#define RADEON_WRITE_PLL( addr, val ) \ -do { \ - RADEON_WRITE8( RADEON_CLOCK_CNTL_INDEX, \ - ((addr) & 0x1f) | RADEON_PLL_WR_EN ); \ - RADEON_WRITE( RADEON_CLOCK_CNTL_DATA, (val) ); \ -} while (0) +extern int RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr); +extern void RADEON_WRITE_PLL(struct drm_radeon_private *dev_priv, int addr, uint32_t data); #define RADEON_WRITE_PCIE( addr, val ) \ do { \ @@ -1274,21 +1324,21 @@ do { \ #define RADEON_FLUSH_CACHE() do { \ if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) { \ - OUT_RING( CP_PACKET0( RADEON_RB3D_DSTCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( RADEON_RB3D_DC_FLUSH ); \ + OUT_RING(CP_PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0)); \ + OUT_RING(RADEON_RB3D_DC_FLUSH); \ } else { \ - OUT_RING( CP_PACKET0( R300_RB3D_DSTCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( RADEON_RB3D_DC_FLUSH ); \ + OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); \ + OUT_RING(R300_RB3D_DC_FLUSH); \ } \ } while (0) #define RADEON_PURGE_CACHE() do { \ if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) { \ - OUT_RING( CP_PACKET0( RADEON_RB3D_DSTCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( RADEON_RB3D_DC_FLUSH_ALL ); \ + OUT_RING(CP_PACKET0( RADEON_RB3D_DSTCACHE_CTLSTAT, 0)); \ + OUT_RING(RADEON_RB3D_DC_FLUSH | RADEON_RB3D_DC_FREE); \ } else { \ - OUT_RING( CP_PACKET0( R300_RB3D_DSTCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( RADEON_RB3D_DC_FLUSH_ALL ); \ + OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); \ + OUT_RING(R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE ); \ } \ } while (0) @@ -1304,11 +1354,11 @@ do { \ #define RADEON_PURGE_ZCACHE() do { \ if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) { \ - OUT_RING( CP_PACKET0( RADEON_RB3D_ZCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( RADEON_RB3D_ZC_FLUSH_ALL ); \ + OUT_RING(CP_PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0)); \ + OUT_RING(RADEON_RB3D_ZC_FLUSH | RADEON_RB3D_ZC_FREE); \ } else { \ - OUT_RING( CP_PACKET0( R300_RB3D_DSTCACHE_CTLSTAT, 0 ) ); \ - OUT_RING( R300_ZC_FLUSH_ALL ); \ + OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0)); \ + OUT_RING(R300_ZC_FLUSH | R300_ZC_FREE); \ } \ } while (0) @@ -1329,7 +1379,8 @@ do { \ #define VB_AGE_TEST_WITH_RETURN( dev_priv ) \ do { \ - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; \ + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; \ + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; \ if ( sarea_priv->last_dispatch >= RADEON_MAX_VB_AGE ) { \ int __ret = radeon_do_cp_idle( dev_priv ); \ if ( __ret ) return __ret; \ @@ -1435,4 +1486,110 @@ do { \ write &= mask; \ } while (0) +/* radeon GEM->TTM munger */ +struct drm_radeon_gem_object { + /* wrap a TTM bo */ + struct drm_buffer_object *bo; + struct drm_fence_object *fence; + struct drm_gem_object *obj; + +}; + +extern int radeon_gem_info_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); + +extern int radeon_gem_create_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); + +extern int radeon_gem_pwrite_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); + +extern int radeon_gem_mmap_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +extern int radeon_gem_pread_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); + +extern void radeon_fence_handler(struct drm_device *dev); +extern int radeon_fence_emit_sequence(struct drm_device *dev, uint32_t class, + uint32_t flags, uint32_t *sequence, + uint32_t *native_type); +extern void radeon_poke_flush(struct drm_device *dev, uint32_t class); +extern int radeon_fence_has_irq(struct drm_device *dev, uint32_t class, uint32_t flags); + +/* radeon_buffer.c */ +extern struct drm_ttm_backend *radeon_create_ttm_backend_entry(struct drm_device *dev); +extern int radeon_fence_types(struct drm_buffer_object *bo, uint32_t *class, uint32_t *type); +extern int radeon_invalidate_caches(struct drm_device *dev, uint64_t buffer_flags); +extern int radeon_init_mem_type(struct drm_device * dev, uint32_t type, + struct drm_mem_type_manager * man); +extern int radeon_move(struct drm_buffer_object * bo, + int evict, int no_wait, struct drm_bo_mem_reg * new_mem); + +extern void radeon_gart_flush(struct drm_device *dev); +extern uint64_t radeon_evict_flags(struct drm_buffer_object *bo); + +#define BREADCRUMB_BITS 31 +#define BREADCRUMB_MASK ((1U << BREADCRUMB_BITS) - 1) + +/* Breadcrumb - swi irq */ +#define READ_BREADCRUMB(dev_priv) RADEON_READ(RADEON_LAST_SWI_REG) + +static inline int radeon_update_breadcrumb(struct drm_device *dev) +{ + struct drm_radeon_private *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv; + + ++dev_priv->counter; + if (dev_priv->counter > BREADCRUMB_MASK) + dev_priv->counter = 1; + + if (dev->primary->master) { + master_priv = dev->primary->master->driver_priv; + + if (master_priv->sarea_priv) + master_priv->sarea_priv->last_fence = dev_priv->counter; + } + return dev_priv->counter; +} + +#define radeon_is_avivo(dev_priv) ((dev_priv->chip_family >= CHIP_RS600)) + +#define radeon_is_dce3(dev_priv) ((dev_priv->chip_family >= CHIP_RV620)) + +#define radeon_bios8(dev_priv, v) (dev_priv->bios[v]) +#define radeon_bios16(dev_priv, v) (dev_priv->bios[v] | (dev_priv->bios[(v) + 1] << 8)) +#define radeon_bios32(dev_priv, v) ((dev_priv->bios[v]) | \ + (dev_priv->bios[(v) + 1] << 8) | \ + (dev_priv->bios[(v) + 2] << 16) | \ + (dev_priv->bios[(v) + 3] << 24)) + +extern int radeon_emit_irq(struct drm_device * dev); + +extern void radeon_gem_free_object(struct drm_gem_object *obj); +extern int radeon_gem_init_object(struct drm_gem_object *obj); +extern int radeon_gem_mm_init(struct drm_device *dev); +extern void radeon_gem_mm_fini(struct drm_device *dev); +extern int radeon_gem_pin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +extern int radeon_gem_unpin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int radeon_gem_object_pin(struct drm_gem_object *obj, + uint32_t alignment); +int radeon_gem_indirect_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int radeon_gem_set_domain_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +struct drm_gem_object *radeon_gem_object_alloc(struct drm_device *dev, int size, int alignment, + int initial_domain); +int radeon_modeset_init(struct drm_device *dev); +void radeon_modeset_cleanup(struct drm_device *dev); +extern u32 radeon_read_mc_reg(drm_radeon_private_t *dev_priv, int addr); +extern void radeon_write_mc_reg(drm_radeon_private_t *dev_priv, u32 addr, u32 val); + +extern void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on); +#define RADEONFB_CONN_LIMIT 4 + +extern int radeon_master_create(struct drm_device *dev, struct drm_master *master); +extern void radeon_master_destroy(struct drm_device *dev, struct drm_master *master); +extern void radeon_cp_dispatch_flip(struct drm_device * dev, struct drm_master *master); #endif /* __RADEON_DRV_H__ */ diff --git a/shared-core/radeon_irq.c b/shared-core/radeon_irq.c index d21761fb..dcf58e43 100644 --- a/shared-core/radeon_irq.c +++ b/shared-core/radeon_irq.c @@ -35,7 +35,7 @@ #include "radeon_drm.h" #include "radeon_drv.h" -static void radeon_irq_set_state(struct drm_device *dev, u32 mask, int state) +void radeon_irq_set_state(struct drm_device *dev, u32 mask, int state) { drm_radeon_private_t *dev_priv = dev->dev_private; @@ -198,8 +198,10 @@ irqreturn_t radeon_driver_irq_handler(DRM_IRQ_ARGS) stat &= dev_priv->irq_enable_reg; /* SW interrupt */ - if (stat & RADEON_SW_INT_TEST) + if (stat & RADEON_SW_INT_TEST) { DRM_WAKEUP(&dev_priv->swi_queue); + radeon_fence_handler(dev); + } /* VBLANK interrupt */ if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { @@ -216,14 +218,13 @@ irqreturn_t radeon_driver_irq_handler(DRM_IRQ_ARGS) return IRQ_HANDLED; } -static int radeon_emit_irq(struct drm_device * dev) +int radeon_emit_irq(struct drm_device * dev) { drm_radeon_private_t *dev_priv = dev->dev_private; unsigned int ret; RING_LOCALS; - atomic_inc(&dev_priv->swi_emitted); - ret = atomic_read(&dev_priv->swi_emitted); + ret = radeon_update_breadcrumb(dev); BEGIN_RING(4); OUT_RING_REG(RADEON_LAST_SWI_REG, ret); @@ -240,13 +241,13 @@ static int radeon_wait_irq(struct drm_device * dev, int swi_nr) (drm_radeon_private_t *) dev->dev_private; int ret = 0; - if (RADEON_READ(RADEON_LAST_SWI_REG) >= swi_nr) + if (READ_BREADCRUMB(dev_priv) >= swi_nr) return 0; dev_priv->stats.boxes |= RADEON_BOX_WAIT_IDLE; DRM_WAIT_ON(ret, dev_priv->swi_queue, 3 * DRM_HZ, - RADEON_READ(RADEON_LAST_SWI_REG) >= swi_nr); + READ_BREADCRUMB(dev_priv) >= swi_nr); return ret; } @@ -254,35 +255,27 @@ static int radeon_wait_irq(struct drm_device * dev, int swi_nr) u32 radeon_get_vblank_counter(struct drm_device *dev, int crtc) { drm_radeon_private_t *dev_priv = dev->dev_private; - u32 crtc_cnt_reg, crtc_status_reg; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } + if (crtc < 0 || crtc > 1) { + DRM_ERROR("Invalid crtc %d\n", crtc); + return -EINVAL; + } + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { - if (crtc == 0) { - crtc_cnt_reg = R500_D1CRTC_FRAME_COUNT; - crtc_status_reg = R500_D1CRTC_STATUS; - } else if (crtc == 1) { - crtc_cnt_reg = R500_D2CRTC_FRAME_COUNT; - crtc_status_reg = R500_D2CRTC_STATUS; - } else - return -EINVAL; - return RADEON_READ(crtc_cnt_reg) + (RADEON_READ(crtc_status_reg) & 1); - + if (crtc == 0) + return RADEON_READ(R500_D1CRTC_FRAME_COUNT); + else + return RADEON_READ(R500_D2CRTC_FRAME_COUNT); } else { - if (crtc == 0) { - crtc_cnt_reg = RADEON_CRTC_CRNT_FRAME; - crtc_status_reg = RADEON_CRTC_STATUS; - } else if (crtc == 1) { - crtc_cnt_reg = RADEON_CRTC2_CRNT_FRAME; - crtc_status_reg = RADEON_CRTC2_STATUS; - } else { - return -EINVAL; - } - return RADEON_READ(crtc_cnt_reg) + (RADEON_READ(crtc_status_reg) & 1); + if (crtc == 0) + return RADEON_READ(RADEON_CRTC_CRNT_FRAME); + else + return RADEON_READ(RADEON_CRTC2_CRNT_FRAME); } } @@ -349,7 +342,6 @@ int radeon_driver_irq_postinstall(struct drm_device * dev) (drm_radeon_private_t *) dev->dev_private; int ret; - atomic_set(&dev_priv->swi_emitted, 0); DRM_INIT_WAITQUEUE(&dev_priv->swi_queue); ret = drm_vblank_init(dev, 2); @@ -382,27 +374,8 @@ void radeon_driver_irq_uninstall(struct drm_device * dev) int radeon_vblank_crtc_get(struct drm_device *dev) { drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; - u32 flag; - u32 value; - if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { - flag = RADEON_READ(R500_DxMODE_INT_MASK); - value = 0; - if (flag & R500_D1MODE_INT_MASK) - value |= DRM_RADEON_VBLANK_CRTC1; - - if (flag & R500_D2MODE_INT_MASK) - value |= DRM_RADEON_VBLANK_CRTC2; - } else { - flag = RADEON_READ(RADEON_GEN_INT_CNTL); - value = 0; - if (flag & RADEON_CRTC_VBLANK_MASK) - value |= DRM_RADEON_VBLANK_CRTC1; - - if (flag & RADEON_CRTC2_VBLANK_MASK) - value |= DRM_RADEON_VBLANK_CRTC2; - } - return value; + return dev_priv->vblank_crtc; } int radeon_vblank_crtc_set(struct drm_device *dev, int64_t value) diff --git a/shared-core/radeon_mem.c b/shared-core/radeon_mem.c index 1e582ee0..dbd73c50 100644 --- a/shared-core/radeon_mem.c +++ b/shared-core/radeon_mem.c @@ -88,7 +88,7 @@ static struct mem_block *alloc_block(struct mem_block *heap, int size, list_for_each(p, heap) { int start = (p->start + mask) & ~mask; - if (p->file_priv == 0 && start + size <= p->start + p->size) + if (p->file_priv == NULL && start + size <= p->start + p->size) return split_block(p, start, size, file_priv); } @@ -113,7 +113,7 @@ static void free_block(struct mem_block *p) /* Assumes a single contiguous range. Needs a special file_priv in * 'heap' to stop it being subsumed. */ - if (p->next->file_priv == 0) { + if (p->next->file_priv == NULL) { struct mem_block *q = p->next; p->size += q->size; p->next = q->next; @@ -121,7 +121,7 @@ static void free_block(struct mem_block *p) drm_free(q, sizeof(*q), DRM_MEM_BUFS); } - if (p->prev->file_priv == 0) { + if (p->prev->file_priv == NULL) { struct mem_block *q = p->prev; q->size += p->size; q->next = p->next; @@ -174,7 +174,7 @@ void radeon_mem_release(struct drm_file *file_priv, struct mem_block *heap) * 'heap' to stop it being subsumed. */ list_for_each(p, heap) { - while (p->file_priv == 0 && p->next->file_priv == 0) { + while (p->file_priv == NULL && p->next->file_priv == NULL) { struct mem_block *q = p->next; p->size += q->size; p->next = q->next; diff --git a/shared-core/radeon_state.c b/shared-core/radeon_state.c index 1d550a9f..6de4b135 100644 --- a/shared-core/radeon_state.c +++ b/shared-core/radeon_state.c @@ -305,8 +305,9 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t * case RADEON_CP_3D_DRAW_INDX_2: case RADEON_3D_CLEAR_HIZ: /* safe but r200 only */ - if (dev_priv->microcode_version != UCODE_R200) { - DRM_ERROR("Invalid 3d packet for r100-class chip\n"); + if ((dev_priv->chip_family < CHIP_R200) || + (dev_priv->chip_family > CHIP_RV280)) { + DRM_ERROR("Invalid 3d packet for non r200-class chip\n"); return -EINVAL; } break; @@ -359,8 +360,8 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t * break; case RADEON_3D_RNDR_GEN_INDX_PRIM: - if (dev_priv->microcode_version != UCODE_R100) { - DRM_ERROR("Invalid 3d packet for r200-class chip\n"); + if (dev_priv->chip_family > CHIP_RS200) { + DRM_ERROR("Invalid 3d packet for non-r100-class chip\n"); return -EINVAL; } if (radeon_check_and_fixup_offset(dev_priv, file_priv, &cmd[1])) { @@ -370,8 +371,10 @@ static __inline__ int radeon_check_and_fixup_packet3(drm_radeon_private_t * break; case RADEON_CP_INDX_BUFFER: - if (dev_priv->microcode_version != UCODE_R200) { - DRM_ERROR("Invalid 3d packet for r100-class chip\n"); + /* safe but r200 only */ + if ((dev_priv->chip_family < CHIP_R200) || + (dev_priv->chip_family > CHIP_RV280)) { + DRM_ERROR("Invalid 3d packet for non-r200-class chip\n"); return -EINVAL; } if ((cmd[1] & 0x8000ffff) != 0x80000810) { @@ -742,13 +745,14 @@ static struct { */ static void radeon_clear_box(drm_radeon_private_t * dev_priv, + struct drm_radeon_master_private *master_priv, int x, int y, int w, int h, int r, int g, int b) { u32 color; RING_LOCALS; - x += dev_priv->sarea_priv->boxes[0].x1; - y += dev_priv->sarea_priv->boxes[0].y1; + x += master_priv->sarea_priv->boxes[0].x1; + y += master_priv->sarea_priv->boxes[0].y1; switch (dev_priv->color_fmt) { case RADEON_COLOR_FORMAT_RGB565: @@ -776,7 +780,7 @@ static void radeon_clear_box(drm_radeon_private_t * dev_priv, RADEON_GMC_SRC_DATATYPE_COLOR | RADEON_ROP3_P | RADEON_GMC_CLR_CMP_CNTL_DIS); - if (dev_priv->sarea_priv->pfCurrentPage == 1) { + if (master_priv->sarea_priv->pfCurrentPage == 1) { OUT_RING(dev_priv->front_pitch_offset); } else { OUT_RING(dev_priv->back_pitch_offset); @@ -790,7 +794,7 @@ static void radeon_clear_box(drm_radeon_private_t * dev_priv, ADVANCE_RING(); } -static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv) +static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv, struct drm_radeon_master_private *master_priv) { /* Collapse various things into a wait flag -- trying to * guess if userspase slept -- better just to have them tell us. @@ -807,12 +811,12 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv) /* Purple box for page flipping */ if (dev_priv->stats.boxes & RADEON_BOX_FLIP) - radeon_clear_box(dev_priv, 4, 4, 8, 8, 255, 0, 255); + radeon_clear_box(dev_priv, master_priv, 4, 4, 8, 8, 255, 0, 255); /* Red box if we have to wait for idle at any point */ if (dev_priv->stats.boxes & RADEON_BOX_WAIT_IDLE) - radeon_clear_box(dev_priv, 16, 4, 8, 8, 255, 0, 0); + radeon_clear_box(dev_priv, master_priv, 16, 4, 8, 8, 255, 0, 0); /* Blue box: lost context? */ @@ -820,12 +824,12 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv) /* Yellow box for texture swaps */ if (dev_priv->stats.boxes & RADEON_BOX_TEXTURE_LOAD) - radeon_clear_box(dev_priv, 40, 4, 8, 8, 255, 255, 0); + radeon_clear_box(dev_priv, master_priv, 40, 4, 8, 8, 255, 255, 0); /* Green box if hardware never idles (as far as we can tell) */ if (!(dev_priv->stats.boxes & RADEON_BOX_DMA_IDLE)) - radeon_clear_box(dev_priv, 64, 4, 8, 8, 0, 255, 0); + radeon_clear_box(dev_priv, master_priv, 64, 4, 8, 8, 0, 255, 0); /* Draw bars indicating number of buffers allocated * (not a great measure, easily confused) @@ -834,7 +838,7 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv) if (dev_priv->stats.requested_bufs > 100) dev_priv->stats.requested_bufs = 100; - radeon_clear_box(dev_priv, 4, 16, + radeon_clear_box(dev_priv, master_priv, 4, 16, dev_priv->stats.requested_bufs, 4, 196, 128, 128); } @@ -848,11 +852,13 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv) */ static void radeon_cp_dispatch_clear(struct drm_device * dev, + struct drm_master *master, drm_radeon_clear_t * clear, drm_radeon_clear_rect_t * depth_boxes) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; drm_radeon_depth_clear_t *depth_clear = &dev_priv->depth_clear; int nbox = sarea_priv->nbox; struct drm_clip_rect *pbox = sarea_priv->boxes; @@ -864,7 +870,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, dev_priv->stats.clears++; - if (dev_priv->sarea_priv->pfCurrentPage == 1) { + if (sarea_priv->pfCurrentPage == 1) { unsigned int tmp = flags; flags &= ~(RADEON_FRONT | RADEON_BACK); @@ -890,7 +896,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, /* Make sure we restore the 3D state next time. */ - dev_priv->sarea_priv->ctx_owner = 0; + sarea_priv->ctx_owner = 0; for (i = 0; i < nbox; i++) { int x = pbox[i].x1; @@ -967,7 +973,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, /* Make sure we restore the 3D state next time. * we haven't touched any "normal" state - still need this? */ - dev_priv->sarea_priv->ctx_owner = 0; + sarea_priv->ctx_owner = 0; if ((dev_priv->flags & RADEON_HAS_HIERZ) && (flags & RADEON_USE_HIERZ)) { @@ -1015,7 +1021,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, int tileoffset, nrtilesx, nrtilesy, j; /* it looks like r200 needs rv-style clears, at least if hierz is not enabled? */ if ((dev_priv->flags & RADEON_HAS_HIERZ) - && !(dev_priv->microcode_version == UCODE_R200)) { + && (dev_priv->chip_family < CHIP_R200)) { /* FIXME : figure this out for r200 (when hierz is enabled). Or maybe r200 actually doesn't need to put the low-res z value into the tile cache like r100, but just needs to clear the hi-level z-buffer? @@ -1044,7 +1050,8 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, ADVANCE_RING(); tileoffset += depthpixperline >> 6; } - } else if (dev_priv->microcode_version == UCODE_R200) { + } else if ((dev_priv->chip_family >= CHIP_R200) && + (dev_priv->chip_family <= CHIP_RV280)) { /* works for rv250. */ /* find first macro tile (8x2 4x4 z-pixels on rv250) */ tileoffset = @@ -1099,7 +1106,8 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, /* TODO don't always clear all hi-level z tiles */ if ((dev_priv->flags & RADEON_HAS_HIERZ) - && (dev_priv->microcode_version == UCODE_R200) + && ((dev_priv->chip_family >= CHIP_R200) && + (dev_priv->chip_family <= CHIP_RV280)) && (flags & RADEON_USE_HIERZ)) /* r100 and cards without hierarchical z-buffer have no high-level z-buffer */ /* FIXME : the mask supposedly contains low-res z values. So can't set @@ -1119,8 +1127,9 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, * rendering a quad into just those buffers. Thus, we have to * make sure the 3D engine is configured correctly. */ - else if ((dev_priv->microcode_version == UCODE_R200) && - (flags & (RADEON_DEPTH | RADEON_STENCIL))) { + else if ((dev_priv->chip_family >= CHIP_R200) && + (dev_priv->chip_family <= CHIP_RV280) && + (flags & (RADEON_DEPTH | RADEON_STENCIL))) { int tempPP_CNTL; int tempRE_CNTL; @@ -1214,7 +1223,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, /* Make sure we restore the 3D state next time. */ - dev_priv->sarea_priv->ctx_owner = 0; + sarea_priv->ctx_owner = 0; for (i = 0; i < nbox; i++) { @@ -1285,7 +1294,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, /* Make sure we restore the 3D state next time. */ - dev_priv->sarea_priv->ctx_owner = 0; + sarea_priv->ctx_owner = 0; for (i = 0; i < nbox; i++) { @@ -1328,20 +1337,21 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev, * wait on this value before performing the clear ioctl. We * need this because the card's so damned fast... */ - dev_priv->sarea_priv->last_clear++; + sarea_priv->last_clear++; BEGIN_RING(4); - RADEON_CLEAR_AGE(dev_priv->sarea_priv->last_clear); + RADEON_CLEAR_AGE(sarea_priv->last_clear); RADEON_WAIT_UNTIL_IDLE(); ADVANCE_RING(); } -static void radeon_cp_dispatch_swap(struct drm_device * dev) +static void radeon_cp_dispatch_swap(struct drm_device * dev, struct drm_master *master) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; int nbox = sarea_priv->nbox; struct drm_clip_rect *pbox = sarea_priv->boxes; int i; @@ -1351,7 +1361,7 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev) /* Do some trivial performance monitoring... */ if (dev_priv->do_boxes) - radeon_cp_performance_boxes(dev_priv); + radeon_cp_performance_boxes(dev_priv, master_priv); /* Wait for the 3D stream to idle before dispatching the bitblt. * This will prevent data corruption between the two streams. @@ -1385,7 +1395,7 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev) /* Make this work even if front & back are flipped: */ OUT_RING(CP_PACKET0(RADEON_SRC_PITCH_OFFSET, 1)); - if (dev_priv->sarea_priv->pfCurrentPage == 0) { + if (sarea_priv->pfCurrentPage == 0) { OUT_RING(dev_priv->back_pitch_offset); OUT_RING(dev_priv->front_pitch_offset); } else { @@ -1405,31 +1415,32 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev) * throttle the framerate by waiting for this value before * performing the swapbuffer ioctl. */ - dev_priv->sarea_priv->last_frame++; + sarea_priv->last_frame++; BEGIN_RING(4); - RADEON_FRAME_AGE(dev_priv->sarea_priv->last_frame); + RADEON_FRAME_AGE(sarea_priv->last_frame); RADEON_WAIT_UNTIL_2D_IDLE(); ADVANCE_RING(); } -static void radeon_cp_dispatch_flip(struct drm_device * dev) +void radeon_cp_dispatch_flip(struct drm_device * dev, struct drm_master *master) { drm_radeon_private_t *dev_priv = dev->dev_private; - struct drm_sarea *sarea = (struct drm_sarea *) dev_priv->sarea->handle; - int offset = (dev_priv->sarea_priv->pfCurrentPage == 1) + struct drm_radeon_master_private *master_priv = master->driver_priv; + struct drm_sarea *sarea = (struct drm_sarea *) master_priv->sarea->handle; + int offset = (master_priv->sarea_priv->pfCurrentPage == 1) ? dev_priv->front_offset : dev_priv->back_offset; RING_LOCALS; DRM_DEBUG("pfCurrentPage=%d\n", - dev_priv->sarea_priv->pfCurrentPage); + master_priv->sarea_priv->pfCurrentPage); /* Do some trivial performance monitoring... */ if (dev_priv->do_boxes) { dev_priv->stats.boxes |= RADEON_BOX_FLIP; - radeon_cp_performance_boxes(dev_priv); + radeon_cp_performance_boxes(dev_priv, master_priv); } /* Update the frame offsets for both CRTCs @@ -1441,7 +1452,7 @@ static void radeon_cp_dispatch_flip(struct drm_device * dev) ((sarea->frame.y * dev_priv->front_pitch + sarea->frame.x * (dev_priv->color_fmt - 2)) & ~7) + offset); - OUT_RING_REG(RADEON_CRTC2_OFFSET, dev_priv->sarea_priv->crtc2_base + OUT_RING_REG(RADEON_CRTC2_OFFSET, master_priv->sarea_priv->crtc2_base + offset); ADVANCE_RING(); @@ -1450,13 +1461,13 @@ static void radeon_cp_dispatch_flip(struct drm_device * dev) * throttle the framerate by waiting for this value before * performing the swapbuffer ioctl. */ - dev_priv->sarea_priv->last_frame++; - dev_priv->sarea_priv->pfCurrentPage = - 1 - dev_priv->sarea_priv->pfCurrentPage; + master_priv->sarea_priv->last_frame++; + master_priv->sarea_priv->pfCurrentPage = + 1 - master_priv->sarea_priv->pfCurrentPage; BEGIN_RING(2); - RADEON_FRAME_AGE(dev_priv->sarea_priv->last_frame); + RADEON_FRAME_AGE(master_priv->sarea_priv->last_frame); ADVANCE_RING(); } @@ -1494,11 +1505,13 @@ typedef struct { } drm_radeon_tcl_prim_t; static void radeon_cp_dispatch_vertex(struct drm_device * dev, + struct drm_file *file_priv, struct drm_buf * buf, drm_radeon_tcl_prim_t * prim) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; int offset = dev_priv->gart_buffers_offset + buf->offset + prim->start; int numverts = (int)prim->numverts; int nbox = sarea_priv->nbox; @@ -1539,13 +1552,14 @@ 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) +static void radeon_cp_discard_buffer(struct drm_device * dev, struct drm_master *master, struct drm_buf * buf) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = master->driver_priv; drm_radeon_buf_priv_t *buf_priv = buf->dev_private; RING_LOCALS; - buf_priv->age = ++dev_priv->sarea_priv->last_dispatch; + buf_priv->age = ++master_priv->sarea_priv->last_dispatch; /* Emit the vertex buffer age */ BEGIN_RING(2); @@ -1590,12 +1604,14 @@ static void radeon_cp_dispatch_indirect(struct drm_device * dev, } } -static void radeon_cp_dispatch_indices(struct drm_device * dev, +static void radeon_cp_dispatch_indices(struct drm_device *dev, + struct drm_master *master, struct drm_buf * elt_buf, drm_radeon_tcl_prim_t * prim) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; int offset = dev_priv->gart_buffers_offset + prim->offset; u32 *data; int dwords; @@ -1870,7 +1886,7 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev, ADVANCE_RING(); COMMIT_RING(); - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); /* Update the input parameters for next time */ image->y += height; @@ -2120,7 +2136,8 @@ static int radeon_surface_free(struct drm_device *dev, void *data, struct drm_fi static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; drm_radeon_clear_t *clear = data; drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS]; DRM_DEBUG("\n"); @@ -2136,7 +2153,7 @@ static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file * sarea_priv->nbox * sizeof(depth_boxes[0]))) return -EFAULT; - radeon_cp_dispatch_clear(dev, clear, depth_boxes); + radeon_cp_dispatch_clear(dev, file_priv->master, clear, depth_boxes); COMMIT_RING(); return 0; @@ -2144,9 +2161,10 @@ static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file * /* Not sure why this isn't set all the time: */ -static int radeon_do_init_pageflip(struct drm_device * dev) +static int radeon_do_init_pageflip(struct drm_device * dev, struct drm_master *master) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = master->driver_priv; RING_LOCALS; DRM_DEBUG("\n"); @@ -2163,8 +2181,8 @@ static int radeon_do_init_pageflip(struct drm_device * dev) dev_priv->page_flipping = 1; - if (dev_priv->sarea_priv->pfCurrentPage != 1) - dev_priv->sarea_priv->pfCurrentPage = 0; + if (master_priv->sarea_priv->pfCurrentPage != 1) + master_priv->sarea_priv->pfCurrentPage = 0; return 0; } @@ -2182,9 +2200,9 @@ static int radeon_cp_flip(struct drm_device *dev, void *data, struct drm_file *f RING_SPACE_TEST_WITH_RETURN(dev_priv); if (!dev_priv->page_flipping) - radeon_do_init_pageflip(dev); + radeon_do_init_pageflip(dev, file_priv->master); - radeon_cp_dispatch_flip(dev); + radeon_cp_dispatch_flip(dev, file_priv->master); COMMIT_RING(); return 0; @@ -2193,7 +2211,9 @@ static int radeon_cp_flip(struct drm_device *dev, void *data, struct drm_file *f static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; - drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; + drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv; + DRM_DEBUG("\n"); LOCK_TEST_WITH_RETURN(dev, file_priv); @@ -2203,8 +2223,8 @@ 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); - dev_priv->sarea_priv->ctx_owner = 0; + radeon_cp_dispatch_swap(dev, file_priv->master); + sarea_priv->ctx_owner = 0; COMMIT_RING(); return 0; @@ -2213,6 +2233,7 @@ static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *f static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; drm_radeon_sarea_t *sarea_priv; struct drm_device_dma *dma = dev->dma; struct drm_buf *buf; @@ -2226,7 +2247,7 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file return -EINVAL; } - sarea_priv = dev_priv->sarea_priv; + sarea_priv = master_priv->sarea_priv; DRM_DEBUG("pid=%d index=%d count=%d discard=%d\n", DRM_CURRENTPID, vertex->idx, vertex->count, vertex->discard); @@ -2280,13 +2301,13 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file prim.finish = vertex->count; /* unused */ prim.prim = vertex->prim; prim.numverts = vertex->count; - prim.vc_format = dev_priv->sarea_priv->vc_format; + prim.vc_format = sarea_priv->vc_format; - radeon_cp_dispatch_vertex(dev, buf, &prim); + radeon_cp_dispatch_vertex(dev, file_priv, buf, &prim); } if (vertex->discard) { - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); } COMMIT_RING(); @@ -2296,6 +2317,7 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; drm_radeon_sarea_t *sarea_priv; struct drm_device_dma *dma = dev->dma; struct drm_buf *buf; @@ -2309,7 +2331,7 @@ static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file DRM_ERROR("called with no initialization\n"); return -EINVAL; } - sarea_priv = dev_priv->sarea_priv; + sarea_priv = master_priv->sarea_priv; DRM_DEBUG("pid=%d index=%d start=%d end=%d discard=%d\n", DRM_CURRENTPID, elts->idx, elts->start, elts->end, @@ -2376,11 +2398,11 @@ static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file prim.prim = elts->prim; prim.offset = 0; /* offset from start of dma buffers */ prim.numverts = RADEON_MAX_VB_VERTS; /* duh */ - prim.vc_format = dev_priv->sarea_priv->vc_format; + prim.vc_format = sarea_priv->vc_format; - radeon_cp_dispatch_indices(dev, buf, &prim); + radeon_cp_dispatch_indices(dev, file_priv->master, buf, &prim); if (elts->discard) { - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); } COMMIT_RING(); @@ -2496,7 +2518,7 @@ static int radeon_cp_indirect(struct drm_device *dev, void *data, struct drm_fil */ radeon_cp_dispatch_indirect(dev, buf, indirect->start, indirect->end); if (indirect->discard) { - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); } COMMIT_RING(); @@ -2506,6 +2528,7 @@ static int radeon_cp_indirect(struct drm_device *dev, void *data, struct drm_fil static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; drm_radeon_sarea_t *sarea_priv; struct drm_device_dma *dma = dev->dma; struct drm_buf *buf; @@ -2520,7 +2543,7 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file return -EINVAL; } - sarea_priv = dev_priv->sarea_priv; + sarea_priv = master_priv->sarea_priv; DRM_DEBUG("pid=%d index=%d discard=%d\n", DRM_CURRENTPID, vertex->idx, vertex->discard); @@ -2582,12 +2605,12 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file tclprim.offset = prim.numverts * 64; tclprim.numverts = RADEON_MAX_VB_VERTS; /* duh */ - radeon_cp_dispatch_indices(dev, buf, &tclprim); + radeon_cp_dispatch_indices(dev, file_priv->master, buf, &tclprim); } else { tclprim.numverts = prim.numverts; tclprim.offset = 0; /* not used */ - radeon_cp_dispatch_vertex(dev, buf, &tclprim); + radeon_cp_dispatch_vertex(dev, file_priv, buf, &tclprim); } if (sarea_priv->nbox == 1) @@ -2595,7 +2618,7 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file } if (vertex->discard) { - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); } COMMIT_RING(); @@ -2889,7 +2912,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file orig_nbox = cmdbuf->nbox; - if (dev_priv->microcode_version == UCODE_R300) { + if (dev_priv->chip_family >= CHIP_R300) { int temp; temp = r300_do_cp_cmdbuf(dev, file_priv, cmdbuf); @@ -2949,7 +2972,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file goto err; } - radeon_cp_discard_buffer(dev, buf); + radeon_cp_discard_buffer(dev, file_priv->master, buf); break; case RADEON_CMD_PACKET3: @@ -3110,6 +3133,7 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil static int radeon_cp_setparam(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_radeon_private_t *dev_priv = dev->dev_private; + struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv; drm_radeon_setparam_t *sp = data; struct drm_radeon_driver_file_fields *radeon_priv; @@ -3129,14 +3153,14 @@ static int radeon_cp_setparam(struct drm_device *dev, void *data, struct drm_fil DRM_DEBUG("color tiling disabled\n"); dev_priv->front_pitch_offset &= ~RADEON_DST_TILE_MACRO; dev_priv->back_pitch_offset &= ~RADEON_DST_TILE_MACRO; - if (dev_priv->sarea_priv) - dev_priv->sarea_priv->tiling_enabled = 0; + if (master_priv->sarea_priv) + master_priv->sarea_priv->tiling_enabled = 0; } else if (sp->value == 1) { DRM_DEBUG("color tiling enabled\n"); dev_priv->front_pitch_offset |= RADEON_DST_TILE_MACRO; dev_priv->back_pitch_offset |= RADEON_DST_TILE_MACRO; - if (dev_priv->sarea_priv) - dev_priv->sarea_priv->tiling_enabled = 1; + if (master_priv->sarea_priv) + master_priv->sarea_priv->tiling_enabled = 1; } break; case RADEON_SETPARAM_PCIGART_LOCATION: @@ -3183,14 +3207,6 @@ void radeon_driver_preclose(struct drm_device *dev, void radeon_driver_lastclose(struct drm_device *dev) { - 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); } @@ -3251,7 +3267,18 @@ 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_GEM_INFO, radeon_gem_info_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_CREATE, radeon_gem_create_ioctl, DRM_AUTH), + + DRM_IOCTL_DEF(DRM_RADEON_GEM_MMAP, radeon_gem_mmap_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_PIN, radeon_gem_pin_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_UNPIN, radeon_gem_unpin_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_PREAD, radeon_gem_pread_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_PWRITE, radeon_gem_pwrite_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_DOMAIN, radeon_gem_set_domain_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_INDIRECT, radeon_gem_indirect_ioctl, DRM_AUTH), }; int radeon_max_ioctl = DRM_ARRAY_SIZE(radeon_ioctls); diff --git a/shared-core/xgi_drm.h b/shared-core/xgi_drm.h index ce584420..7d01065e 100644 --- a/shared-core/xgi_drm.h +++ b/shared-core/xgi_drm.h @@ -123,11 +123,15 @@ struct xgi_state_info { #define DRM_XGI_FREE 2 #define DRM_XGI_SUBMIT_CMDLIST 3 #define DRM_XGI_STATE_CHANGE 4 +#define DRM_XGI_SET_FENCE 5 +#define DRM_XGI_WAIT_FENCE 6 #define XGI_IOCTL_BOOTSTRAP DRM_IOWR(DRM_COMMAND_BASE + DRM_XGI_BOOTSTRAP, struct xgi_bootstrap) #define XGI_IOCTL_ALLOC DRM_IOWR(DRM_COMMAND_BASE + DRM_XGI_ALLOC, struct xgi_mem_alloc) #define XGI_IOCTL_FREE DRM_IOW(DRM_COMMAND_BASE + DRM_XGI_FREE, __u32) #define XGI_IOCTL_SUBMIT_CMDLIST DRM_IOW(DRM_COMMAND_BASE + DRM_XGI_SUBMIT_CMDLIST, struct xgi_cmd_info) #define XGI_IOCTL_STATE_CHANGE DRM_IOW(DRM_COMMAND_BASE + DRM_XGI_STATE_CHANGE, struct xgi_state_info) +#define XGI_IOCTL_SET_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XGI_SET_FENCE, u32) +#define XGI_IOCTL_WAIT_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XGI_WAIT_FENCE, u32) #endif /* _XGI_DRM_H_ */ |