diff options
Diffstat (limited to 'drivers/gpu/drm/i915/intel_ringbuffer.c')
| -rw-r--r-- | drivers/gpu/drm/i915/intel_ringbuffer.c | 2833 |
1 files changed, 2105 insertions, 728 deletions
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index b83306f9244..279488addf3 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -27,1002 +27,2379 @@ * */ -#include "drmP.h" -#include "drm.h" +#include <drm/drmP.h> #include "i915_drv.h" -#include "i915_drm.h" +#include <drm/i915_drm.h> #include "i915_trace.h" #include "intel_drv.h" -static u32 i915_gem_get_seqno(struct drm_device *dev) +/* Early gen2 devices have a cacheline of just 32 bytes, using 64 is overkill, + * but keeps the logic simple. Indeed, the whole purpose of this macro is just + * to give some inclination as to some of the magic values used in the various + * workarounds! + */ +#define CACHELINE_BYTES 64 + +static inline int __ring_space(int head, int tail, int size) { - drm_i915_private_t *dev_priv = dev->dev_private; - u32 seqno; + int space = head - (tail + I915_RING_FREE_SPACE); + if (space < 0) + space += size; + return space; +} - seqno = dev_priv->next_seqno; +static inline int ring_space(struct intel_engine_cs *ring) +{ + struct intel_ringbuffer *ringbuf = ring->buffer; + return __ring_space(ringbuf->head & HEAD_ADDR, ringbuf->tail, ringbuf->size); +} - /* reserve 0 for non-seqno */ - if (++dev_priv->next_seqno == 0) - dev_priv->next_seqno = 1; +static bool intel_ring_stopped(struct intel_engine_cs *ring) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + return dev_priv->gpu_error.stop_rings & intel_ring_flag(ring); +} - return seqno; +void __intel_ring_advance(struct intel_engine_cs *ring) +{ + struct intel_ringbuffer *ringbuf = ring->buffer; + ringbuf->tail &= ringbuf->size - 1; + if (intel_ring_stopped(ring)) + return; + ring->write_tail(ring, ringbuf->tail); } -static void -render_ring_flush(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 invalidate_domains, - u32 flush_domains) +static int +gen2_render_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, + u32 flush_domains) +{ + u32 cmd; + int ret; + + cmd = MI_FLUSH; + if (((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER) == 0) + cmd |= MI_NO_WRITE_FLUSH; + + if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER) + cmd |= MI_READ_FLUSH; + + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + + intel_ring_emit(ring, cmd); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + + return 0; +} + +static int +gen4_render_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, + u32 flush_domains) { - drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_device *dev = ring->dev; u32 cmd; + int ret; + + /* + * read/write caches: + * + * I915_GEM_DOMAIN_RENDER is always invalidated, but is + * only flushed if MI_NO_WRITE_FLUSH is unset. On 965, it is + * also flushed at 2d versus 3d pipeline switches. + * + * read-only caches: + * + * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if + * MI_READ_FLUSH is set, and is always flushed on 965. + * + * I915_GEM_DOMAIN_COMMAND may not exist? + * + * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is + * invalidated when MI_EXE_FLUSH is set. + * + * I915_GEM_DOMAIN_VERTEX, which exists on 965, is + * invalidated with every MI_FLUSH. + * + * TLBs: + * + * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND + * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and + * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER + * are flushed at any MI_FLUSH. + */ + + cmd = MI_FLUSH | MI_NO_WRITE_FLUSH; + if ((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER) + cmd &= ~MI_NO_WRITE_FLUSH; + if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION) + cmd |= MI_EXE_FLUSH; + + if (invalidate_domains & I915_GEM_DOMAIN_COMMAND && + (IS_G4X(dev) || IS_GEN5(dev))) + cmd |= MI_INVALIDATE_ISP; + + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; -#if WATCH_EXEC - DRM_INFO("%s: invalidate %08x flush %08x\n", __func__, - invalidate_domains, flush_domains); -#endif + intel_ring_emit(ring, cmd); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + + return 0; +} + +/** + * Emits a PIPE_CONTROL with a non-zero post-sync operation, for + * implementing two workarounds on gen6. From section 1.4.7.1 + * "PIPE_CONTROL" of the Sandy Bridge PRM volume 2 part 1: + * + * [DevSNB-C+{W/A}] Before any depth stall flush (including those + * produced by non-pipelined state commands), software needs to first + * send a PIPE_CONTROL with no bits set except Post-Sync Operation != + * 0. + * + * [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache Flush Enable + * =1, a PIPE_CONTROL with any non-zero post-sync-op is required. + * + * And the workaround for these two requires this workaround first: + * + * [Dev-SNB{W/A}]: Pipe-control with CS-stall bit set must be sent + * BEFORE the pipe-control with a post-sync op and no write-cache + * flushes. + * + * And this last workaround is tricky because of the requirements on + * that bit. From section 1.4.7.2.3 "Stall" of the Sandy Bridge PRM + * volume 2 part 1: + * + * "1 of the following must also be set: + * - Render Target Cache Flush Enable ([12] of DW1) + * - Depth Cache Flush Enable ([0] of DW1) + * - Stall at Pixel Scoreboard ([1] of DW1) + * - Depth Stall ([13] of DW1) + * - Post-Sync Operation ([13] of DW1) + * - Notify Enable ([8] of DW1)" + * + * The cache flushes require the workaround flush that triggered this + * one, so we can't use it. Depth stall would trigger the same. + * Post-sync nonzero is what triggered this second workaround, so we + * can't use that one either. Notify enable is IRQs, which aren't + * really our business. That leaves only stall at scoreboard. + */ +static int +intel_emit_post_sync_nonzero_flush(struct intel_engine_cs *ring) +{ + u32 scratch_addr = ring->scratch.gtt_offset + 2 * CACHELINE_BYTES; + int ret; - trace_i915_gem_request_flush(dev, dev_priv->next_seqno, - invalidate_domains, flush_domains); - if ((invalidate_domains | flush_domains) & I915_GEM_GPU_DOMAINS) { + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5)); + intel_ring_emit(ring, PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_STALL_AT_SCOREBOARD); + intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */ + intel_ring_emit(ring, 0); /* low dword */ + intel_ring_emit(ring, 0); /* high dword */ + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5)); + intel_ring_emit(ring, PIPE_CONTROL_QW_WRITE); + intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */ + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + + return 0; +} + +static int +gen6_render_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, u32 flush_domains) +{ + u32 flags = 0; + u32 scratch_addr = ring->scratch.gtt_offset + 2 * CACHELINE_BYTES; + int ret; + + /* Force SNB workarounds for PIPE_CONTROL flushes */ + ret = intel_emit_post_sync_nonzero_flush(ring); + if (ret) + return ret; + + /* Just flush everything. Experiments have shown that reducing the + * number of bits based on the write domains has little performance + * impact. + */ + if (flush_domains) { + flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; + flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; + /* + * Ensure that any following seqno writes only happen + * when the render cache is indeed flushed. + */ + flags |= PIPE_CONTROL_CS_STALL; + } + if (invalidate_domains) { + flags |= PIPE_CONTROL_TLB_INVALIDATE; + flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; /* - * read/write caches: - * - * I915_GEM_DOMAIN_RENDER is always invalidated, but is - * only flushed if MI_NO_WRITE_FLUSH is unset. On 965, it is - * also flushed at 2d versus 3d pipeline switches. - * - * read-only caches: - * - * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if - * MI_READ_FLUSH is set, and is always flushed on 965. - * - * I915_GEM_DOMAIN_COMMAND may not exist? - * - * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is - * invalidated when MI_EXE_FLUSH is set. - * - * I915_GEM_DOMAIN_VERTEX, which exists on 965, is - * invalidated with every MI_FLUSH. - * - * TLBs: - * - * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND - * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and - * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER - * are flushed at any MI_FLUSH. + * TLB invalidate requires a post-sync write. */ + flags |= PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_CS_STALL; + } - cmd = MI_FLUSH | MI_NO_WRITE_FLUSH; - if ((invalidate_domains|flush_domains) & - I915_GEM_DOMAIN_RENDER) - cmd &= ~MI_NO_WRITE_FLUSH; - if (INTEL_INFO(dev)->gen < 4) { - /* - * On the 965, the sampler cache always gets flushed - * and this bit is reserved. - */ - if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER) - cmd |= MI_READ_FLUSH; - } - if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION) - cmd |= MI_EXE_FLUSH; - -#if WATCH_EXEC - DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd); -#endif - intel_ring_begin(dev, ring, 2); - intel_ring_emit(dev, ring, cmd); - intel_ring_emit(dev, ring, MI_NOOP); - intel_ring_advance(dev, ring); + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4)); + intel_ring_emit(ring, flags); + intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + + return 0; +} + +static int +gen7_render_ring_cs_stall_wa(struct intel_engine_cs *ring) +{ + int ret; + + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4)); + intel_ring_emit(ring, PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_STALL_AT_SCOREBOARD); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + + return 0; +} + +static int gen7_ring_fbc_flush(struct intel_engine_cs *ring, u32 value) +{ + int ret; + + if (!ring->fbc_dirty) + return 0; + + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; + /* WaFbcNukeOn3DBlt:ivb/hsw */ + intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); + intel_ring_emit(ring, MSG_FBC_REND_STATE); + intel_ring_emit(ring, value); + intel_ring_emit(ring, MI_STORE_REGISTER_MEM(1) | MI_SRM_LRM_GLOBAL_GTT); + intel_ring_emit(ring, MSG_FBC_REND_STATE); + intel_ring_emit(ring, ring->scratch.gtt_offset + 256); + intel_ring_advance(ring); + + ring->fbc_dirty = false; + return 0; +} + +static int +gen7_render_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, u32 flush_domains) +{ + u32 flags = 0; + u32 scratch_addr = ring->scratch.gtt_offset + 2 * CACHELINE_BYTES; + int ret; + + /* + * Ensure that any following seqno writes only happen when the render + * cache is indeed flushed. + * + * Workaround: 4th PIPE_CONTROL command (except the ones with only + * read-cache invalidate bits set) must have the CS_STALL bit set. We + * don't try to be clever and just set it unconditionally. + */ + flags |= PIPE_CONTROL_CS_STALL; + + /* Just flush everything. Experiments have shown that reducing the + * number of bits based on the write domains has little performance + * impact. + */ + if (flush_domains) { + flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; + flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; + } + if (invalidate_domains) { + flags |= PIPE_CONTROL_TLB_INVALIDATE; + flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; + /* + * TLB invalidate requires a post-sync write. + */ + flags |= PIPE_CONTROL_QW_WRITE; + flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + + /* Workaround: we must issue a pipe_control with CS-stall bit + * set before a pipe_control command that has the state cache + * invalidate bit set. */ + gen7_render_ring_cs_stall_wa(ring); } + + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4)); + intel_ring_emit(ring, flags); + intel_ring_emit(ring, scratch_addr); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + + if (!invalidate_domains && flush_domains) + return gen7_ring_fbc_flush(ring, FBC_REND_NUKE); + + return 0; } -static void ring_write_tail(struct drm_device *dev, - struct intel_ring_buffer *ring, +static int +gen8_render_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, u32 flush_domains) +{ + u32 flags = 0; + u32 scratch_addr = ring->scratch.gtt_offset + 2 * CACHELINE_BYTES; + int ret; + + flags |= PIPE_CONTROL_CS_STALL; + + if (flush_domains) { + flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; + flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; + } + if (invalidate_domains) { + flags |= PIPE_CONTROL_TLB_INVALIDATE; + flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_QW_WRITE; + flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + } + + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(6)); + intel_ring_emit(ring, flags); + intel_ring_emit(ring, scratch_addr); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + + return 0; + +} + +static void ring_write_tail(struct intel_engine_cs *ring, u32 value) { - drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_private *dev_priv = ring->dev->dev_private; I915_WRITE_TAIL(ring, value); } -u32 intel_ring_get_active_head(struct drm_device *dev, - struct intel_ring_buffer *ring) +u64 intel_ring_get_active_head(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - u32 acthd_reg = INTEL_INFO(dev)->gen >= 4 ? - RING_ACTHD(ring->mmio_base) : ACTHD; + struct drm_i915_private *dev_priv = ring->dev->dev_private; + u64 acthd; + + if (INTEL_INFO(ring->dev)->gen >= 8) + acthd = I915_READ64_2x32(RING_ACTHD(ring->mmio_base), + RING_ACTHD_UDW(ring->mmio_base)); + else if (INTEL_INFO(ring->dev)->gen >= 4) + acthd = I915_READ(RING_ACTHD(ring->mmio_base)); + else + acthd = I915_READ(ACTHD); - return I915_READ(acthd_reg); + return acthd; } -static int init_ring_common(struct drm_device *dev, - struct intel_ring_buffer *ring) +static void ring_setup_phys_status_page(struct intel_engine_cs *ring) { - u32 head; - drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_i915_gem_object *obj_priv; - obj_priv = to_intel_bo(ring->gem_object); + struct drm_i915_private *dev_priv = ring->dev->dev_private; + u32 addr; + + addr = dev_priv->status_page_dmah->busaddr; + if (INTEL_INFO(ring->dev)->gen >= 4) + addr |= (dev_priv->status_page_dmah->busaddr >> 28) & 0xf0; + I915_WRITE(HWS_PGA, addr); +} + +static bool stop_ring(struct intel_engine_cs *ring) +{ + struct drm_i915_private *dev_priv = to_i915(ring->dev); + + if (!IS_GEN2(ring->dev)) { + I915_WRITE_MODE(ring, _MASKED_BIT_ENABLE(STOP_RING)); + if (wait_for_atomic((I915_READ_MODE(ring) & MODE_IDLE) != 0, 1000)) { + DRM_ERROR("%s :timed out trying to stop ring\n", ring->name); + return false; + } + } - /* Stop the ring if it's running. */ I915_WRITE_CTL(ring, 0); I915_WRITE_HEAD(ring, 0); - ring->write_tail(dev, ring, 0); - - /* Initialize the ring. */ - I915_WRITE_START(ring, obj_priv->gtt_offset); - head = I915_READ_HEAD(ring) & HEAD_ADDR; - - /* G45 ring initialization fails to reset head to zero */ - if (head != 0) { - DRM_ERROR("%s head not reset to zero " - "ctl %08x head %08x tail %08x start %08x\n", - ring->name, - I915_READ_CTL(ring), - I915_READ_HEAD(ring), - I915_READ_TAIL(ring), - I915_READ_START(ring)); - - I915_WRITE_HEAD(ring, 0); - - DRM_ERROR("%s head forced to zero " - "ctl %08x head %08x tail %08x start %08x\n", - ring->name, - I915_READ_CTL(ring), - I915_READ_HEAD(ring), - I915_READ_TAIL(ring), - I915_READ_START(ring)); + ring->write_tail(ring, 0); + + if (!IS_GEN2(ring->dev)) { + (void)I915_READ_CTL(ring); + I915_WRITE_MODE(ring, _MASKED_BIT_DISABLE(STOP_RING)); + } + + return (I915_READ_HEAD(ring) & HEAD_ADDR) == 0; +} + +static int init_ring_common(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_ringbuffer *ringbuf = ring->buffer; + struct drm_i915_gem_object *obj = ringbuf->obj; + int ret = 0; + + gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL); + + if (!stop_ring(ring)) { + /* G45 ring initialization often fails to reset head to zero */ + DRM_DEBUG_KMS("%s head not reset to zero " + "ctl %08x head %08x tail %08x start %08x\n", + ring->name, + I915_READ_CTL(ring), + I915_READ_HEAD(ring), + I915_READ_TAIL(ring), + I915_READ_START(ring)); + + if (!stop_ring(ring)) { + DRM_ERROR("failed to set %s head to zero " + "ctl %08x head %08x tail %08x start %08x\n", + ring->name, + I915_READ_CTL(ring), + I915_READ_HEAD(ring), + I915_READ_TAIL(ring), + I915_READ_START(ring)); + ret = -EIO; + goto out; + } } + if (I915_NEED_GFX_HWS(dev)) + intel_ring_setup_status_page(ring); + else + ring_setup_phys_status_page(ring); + + /* Initialize the ring. This must happen _after_ we've cleared the ring + * registers with the above sequence (the readback of the HEAD registers + * also enforces ordering), otherwise the hw might lose the new ring + * register values. */ + I915_WRITE_START(ring, i915_gem_obj_ggtt_offset(obj)); I915_WRITE_CTL(ring, - ((ring->gem_object->size - PAGE_SIZE) & RING_NR_PAGES) - | RING_REPORT_64K | RING_VALID); + ((ringbuf->size - PAGE_SIZE) & RING_NR_PAGES) + | RING_VALID); - head = I915_READ_HEAD(ring) & HEAD_ADDR; /* If the head is still not zero, the ring is dead */ - if (head != 0) { + if (wait_for((I915_READ_CTL(ring) & RING_VALID) != 0 && + I915_READ_START(ring) == i915_gem_obj_ggtt_offset(obj) && + (I915_READ_HEAD(ring) & HEAD_ADDR) == 0, 50)) { DRM_ERROR("%s initialization failed " - "ctl %08x head %08x tail %08x start %08x\n", - ring->name, - I915_READ_CTL(ring), - I915_READ_HEAD(ring), - I915_READ_TAIL(ring), - I915_READ_START(ring)); - return -EIO; + "ctl %08x (valid? %d) head %08x tail %08x start %08x [expected %08lx]\n", + ring->name, + I915_READ_CTL(ring), I915_READ_CTL(ring) & RING_VALID, + I915_READ_HEAD(ring), I915_READ_TAIL(ring), + I915_READ_START(ring), (unsigned long)i915_gem_obj_ggtt_offset(obj)); + ret = -EIO; + goto out; } - if (!drm_core_check_feature(dev, DRIVER_MODESET)) - i915_kernel_lost_context(dev); + if (!drm_core_check_feature(ring->dev, DRIVER_MODESET)) + i915_kernel_lost_context(ring->dev); else { - ring->head = I915_READ_HEAD(ring) & HEAD_ADDR; - ring->tail = I915_READ_TAIL(ring) & TAIL_ADDR; - ring->space = ring->head - (ring->tail + 8); - if (ring->space < 0) - ring->space += ring->size; + ringbuf->head = I915_READ_HEAD(ring); + ringbuf->tail = I915_READ_TAIL(ring) & TAIL_ADDR; + ringbuf->space = ring_space(ring); + ringbuf->last_retired_head = -1; + } + + memset(&ring->hangcheck, 0, sizeof(ring->hangcheck)); + +out: + gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL); + + return ret; +} + +static int +init_pipe_control(struct intel_engine_cs *ring) +{ + int ret; + + if (ring->scratch.obj) + return 0; + + ring->scratch.obj = i915_gem_alloc_object(ring->dev, 4096); + if (ring->scratch.obj == NULL) { + DRM_ERROR("Failed to allocate seqno page\n"); + ret = -ENOMEM; + goto err; } + + ret = i915_gem_object_set_cache_level(ring->scratch.obj, I915_CACHE_LLC); + if (ret) + goto err_unref; + + ret = i915_gem_obj_ggtt_pin(ring->scratch.obj, 4096, 0); + if (ret) + goto err_unref; + + ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(ring->scratch.obj); + ring->scratch.cpu_page = kmap(sg_page(ring->scratch.obj->pages->sgl)); + if (ring->scratch.cpu_page == NULL) { + ret = -ENOMEM; + goto err_unpin; + } + + DRM_DEBUG_DRIVER("%s pipe control offset: 0x%08x\n", + ring->name, ring->scratch.gtt_offset); return 0; + +err_unpin: + i915_gem_object_ggtt_unpin(ring->scratch.obj); +err_unref: + drm_gem_object_unreference(&ring->scratch.obj->base); +err: + return ret; } -static int init_render_ring(struct drm_device *dev, - struct intel_ring_buffer *ring) +static int init_render_ring(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - int ret = init_ring_common(dev, ring); - int mode; + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + int ret = init_ring_common(ring); + + /* WaTimedSingleVertexDispatch:cl,bw,ctg,elk,ilk,snb */ + if (INTEL_INFO(dev)->gen >= 4 && INTEL_INFO(dev)->gen < 7) + I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH)); + + /* We need to disable the AsyncFlip performance optimisations in order + * to use MI_WAIT_FOR_EVENT within the CS. It should already be + * programmed to '1' on all products. + * + * WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv,bdw,chv + */ + if (INTEL_INFO(dev)->gen >= 6) + I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(ASYNC_FLIP_PERF_DISABLE)); + + /* Required for the hardware to program scanline values for waiting */ + /* WaEnableFlushTlbInvalidationMode:snb */ + if (INTEL_INFO(dev)->gen == 6) + I915_WRITE(GFX_MODE, + _MASKED_BIT_ENABLE(GFX_TLB_INVALIDATE_EXPLICIT)); + + /* WaBCSVCSTlbInvalidationMode:ivb,vlv,hsw */ + if (IS_GEN7(dev)) + I915_WRITE(GFX_MODE_GEN7, + _MASKED_BIT_ENABLE(GFX_TLB_INVALIDATE_EXPLICIT) | + _MASKED_BIT_ENABLE(GFX_REPLAY_MODE)); + + if (INTEL_INFO(dev)->gen >= 5) { + ret = init_pipe_control(ring); + if (ret) + return ret; + } - if (INTEL_INFO(dev)->gen > 3) { - mode = VS_TIMER_DISPATCH << 16 | VS_TIMER_DISPATCH; - if (IS_GEN6(dev)) - mode |= MI_FLUSH_ENABLE << 16 | MI_FLUSH_ENABLE; - I915_WRITE(MI_MODE, mode); + if (IS_GEN6(dev)) { + /* From the Sandybridge PRM, volume 1 part 3, page 24: + * "If this bit is set, STCunit will have LRA as replacement + * policy. [...] This bit must be reset. LRA replacement + * policy is not supported." + */ + I915_WRITE(CACHE_MODE_0, + _MASKED_BIT_DISABLE(CM0_STC_EVICT_DISABLE_LRA_SNB)); } + + if (INTEL_INFO(dev)->gen >= 6) + I915_WRITE(INSTPM, _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING)); + + if (HAS_L3_DPF(dev)) + I915_WRITE_IMR(ring, ~GT_PARITY_ERROR(dev)); + return ret; } -#define PIPE_CONTROL_FLUSH(addr) \ -do { \ - OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | \ - PIPE_CONTROL_DEPTH_STALL | 2); \ - OUT_RING(addr | PIPE_CONTROL_GLOBAL_GTT); \ - OUT_RING(0); \ - OUT_RING(0); \ -} while (0) +static void render_ring_cleanup(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + + if (ring->scratch.obj == NULL) + return; + + if (INTEL_INFO(dev)->gen >= 5) { + kunmap(sg_page(ring->scratch.obj->pages->sgl)); + i915_gem_object_ggtt_unpin(ring->scratch.obj); + } + + drm_gem_object_unreference(&ring->scratch.obj->base); + ring->scratch.obj = NULL; +} + +static int gen6_signal(struct intel_engine_cs *signaller, + unsigned int num_dwords) +{ + struct drm_device *dev = signaller->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *useless; + int i, ret; + + /* NB: In order to be able to do semaphore MBOX updates for varying + * number of rings, it's easiest if we round up each individual update + * to a multiple of 2 (since ring updates must always be a multiple of + * 2) even though the actual update only requires 3 dwords. + */ +#define MBOX_UPDATE_DWORDS 4 + if (i915_semaphore_is_enabled(dev)) + num_dwords += ((I915_NUM_RINGS-1) * MBOX_UPDATE_DWORDS); + else + return intel_ring_begin(signaller, num_dwords); + + ret = intel_ring_begin(signaller, num_dwords); + if (ret) + return ret; +#undef MBOX_UPDATE_DWORDS + + for_each_ring(useless, dev_priv, i) { + u32 mbox_reg = signaller->semaphore.mbox.signal[i]; + if (mbox_reg != GEN6_NOSYNC) { + intel_ring_emit(signaller, MI_LOAD_REGISTER_IMM(1)); + intel_ring_emit(signaller, mbox_reg); + intel_ring_emit(signaller, signaller->outstanding_lazy_seqno); + intel_ring_emit(signaller, MI_NOOP); + } else { + intel_ring_emit(signaller, MI_NOOP); + intel_ring_emit(signaller, MI_NOOP); + intel_ring_emit(signaller, MI_NOOP); + intel_ring_emit(signaller, MI_NOOP); + } + } + + return 0; +} /** - * Creates a new sequence number, emitting a write of it to the status page - * plus an interrupt, which will trigger i915_user_interrupt_handler. + * gen6_add_request - Update the semaphore mailbox registers + * + * @ring - ring that is adding a request + * @seqno - return seqno stuck into the ring * - * Must be called with struct_lock held. + * Update the mailbox registers in the *other* rings with the current seqno. + * This acts like a signal in the canonical semaphore. + */ +static int +gen6_add_request(struct intel_engine_cs *ring) +{ + int ret; + + ret = ring->semaphore.signal(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, MI_STORE_DWORD_INDEX); + intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + intel_ring_emit(ring, ring->outstanding_lazy_seqno); + intel_ring_emit(ring, MI_USER_INTERRUPT); + __intel_ring_advance(ring); + + return 0; +} + +static inline bool i915_gem_has_seqno_wrapped(struct drm_device *dev, + u32 seqno) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + return dev_priv->last_seqno < seqno; +} + +/** + * intel_ring_sync - sync the waiter to the signaller on seqno * - * Returned sequence numbers are nonzero on success. + * @waiter - ring that is waiting + * @signaller - ring which has, or will signal + * @seqno - seqno which the waiter will block on */ -static u32 -render_ring_add_request(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 flush_domains) +static int +gen6_ring_sync(struct intel_engine_cs *waiter, + struct intel_engine_cs *signaller, + u32 seqno) { - drm_i915_private_t *dev_priv = dev->dev_private; - u32 seqno; + u32 dw1 = MI_SEMAPHORE_MBOX | + MI_SEMAPHORE_COMPARE | + MI_SEMAPHORE_REGISTER; + u32 wait_mbox = signaller->semaphore.mbox.wait[waiter->id]; + int ret; - seqno = i915_gem_get_seqno(dev); + /* Throughout all of the GEM code, seqno passed implies our current + * seqno is >= the last seqno executed. However for hardware the + * comparison is strictly greater than. + */ + seqno -= 1; - if (IS_GEN6(dev)) { - BEGIN_LP_RING(6); - OUT_RING(GFX_OP_PIPE_CONTROL | 3); - OUT_RING(PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_IS_FLUSH | - PIPE_CONTROL_NOTIFY); - OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - OUT_RING(seqno); - OUT_RING(0); - OUT_RING(0); - ADVANCE_LP_RING(); - } else if (HAS_PIPE_CONTROL(dev)) { - u32 scratch_addr = dev_priv->seqno_gfx_addr + 128; + WARN_ON(wait_mbox == MI_SEMAPHORE_SYNC_INVALID); - /* - * Workaround qword write incoherence by flushing the - * PIPE_NOTIFY buffers out to memory before requesting - * an interrupt. - */ - BEGIN_LP_RING(32); - OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH); - OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - OUT_RING(seqno); - OUT_RING(0); - PIPE_CONTROL_FLUSH(scratch_addr); - scratch_addr += 128; /* write to separate cachelines */ - PIPE_CONTROL_FLUSH(scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(scratch_addr); - OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH | - PIPE_CONTROL_NOTIFY); - OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - OUT_RING(seqno); - OUT_RING(0); - ADVANCE_LP_RING(); + ret = intel_ring_begin(waiter, 4); + if (ret) + return ret; + + /* If seqno wrap happened, omit the wait with no-ops */ + if (likely(!i915_gem_has_seqno_wrapped(waiter->dev, seqno))) { + intel_ring_emit(waiter, dw1 | wait_mbox); + intel_ring_emit(waiter, seqno); + intel_ring_emit(waiter, 0); + intel_ring_emit(waiter, MI_NOOP); } else { - BEGIN_LP_RING(4); - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(seqno); + intel_ring_emit(waiter, MI_NOOP); + intel_ring_emit(waiter, MI_NOOP); + intel_ring_emit(waiter, MI_NOOP); + intel_ring_emit(waiter, MI_NOOP); + } + intel_ring_advance(waiter); + + return 0; +} + +#define PIPE_CONTROL_FLUSH(ring__, addr__) \ +do { \ + intel_ring_emit(ring__, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | \ + PIPE_CONTROL_DEPTH_STALL); \ + intel_ring_emit(ring__, (addr__) | PIPE_CONTROL_GLOBAL_GTT); \ + intel_ring_emit(ring__, 0); \ + intel_ring_emit(ring__, 0); \ +} while (0) + +static int +pc_render_add_request(struct intel_engine_cs *ring) +{ + u32 scratch_addr = ring->scratch.gtt_offset + 2 * CACHELINE_BYTES; + int ret; + + /* For Ironlake, MI_USER_INTERRUPT was deprecated and apparently + * incoherent with writes to memory, i.e. completely fubar, + * so we need to use PIPE_NOTIFY instead. + * + * However, we also need to workaround the qword write + * incoherence by flushing the 6 PIPE_NOTIFY buffers out to + * memory before requesting an interrupt. + */ + ret = intel_ring_begin(ring, 32); + if (ret) + return ret; + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | + PIPE_CONTROL_WRITE_FLUSH | + PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE); + intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, ring->outstanding_lazy_seqno); + intel_ring_emit(ring, 0); + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 2 * CACHELINE_BYTES; /* write to separate cachelines */ + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 2 * CACHELINE_BYTES; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 2 * CACHELINE_BYTES; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 2 * CACHELINE_BYTES; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 2 * CACHELINE_BYTES; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | + PIPE_CONTROL_WRITE_FLUSH | + PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE | + PIPE_CONTROL_NOTIFY); + intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, ring->outstanding_lazy_seqno); + intel_ring_emit(ring, 0); + __intel_ring_advance(ring); + + return 0; +} - OUT_RING(MI_USER_INTERRUPT); - ADVANCE_LP_RING(); +static u32 +gen6_ring_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency) +{ + /* Workaround to force correct ordering between irq and seqno writes on + * ivb (and maybe also on snb) by reading from a CS register (like + * ACTHD) before reading the status page. */ + if (!lazy_coherency) { + struct drm_i915_private *dev_priv = ring->dev->dev_private; + POSTING_READ(RING_ACTHD(ring->mmio_base)); } - return seqno; + + return intel_read_status_page(ring, I915_GEM_HWS_INDEX); } static u32 -render_ring_get_seqno(struct drm_device *dev, - struct intel_ring_buffer *ring) +ring_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency) { - drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - if (HAS_PIPE_CONTROL(dev)) - return ((volatile u32 *)(dev_priv->seqno_page))[0]; - else - return intel_read_status_page(ring, I915_GEM_HWS_INDEX); + return intel_read_status_page(ring, I915_GEM_HWS_INDEX); } static void -render_ring_get_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) +ring_set_seqno(struct intel_engine_cs *ring, u32 seqno) { - drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - unsigned long irqflags; + intel_write_status_page(ring, I915_GEM_HWS_INDEX, seqno); +} - spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags); - if (dev->irq_enabled && (++ring->user_irq_refcount == 1)) { - if (HAS_PCH_SPLIT(dev)) - ironlake_enable_graphics_irq(dev_priv, GT_PIPE_NOTIFY); - else - i915_enable_irq(dev_priv, I915_USER_INTERRUPT); +static u32 +pc_render_get_seqno(struct intel_engine_cs *ring, bool lazy_coherency) +{ + return ring->scratch.cpu_page[0]; +} + +static void +pc_render_set_seqno(struct intel_engine_cs *ring, u32 seqno) +{ + ring->scratch.cpu_page[0] = seqno; +} + +static bool +gen5_ring_get_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) + ilk_enable_gt_irq(dev_priv, ring->irq_enable_mask); + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return true; +} + +static void +gen5_ring_put_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) + ilk_disable_gt_irq(dev_priv, ring->irq_enable_mask); + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); +} + +static bool +i9xx_ring_get_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) { + dev_priv->irq_mask &= ~ring->irq_enable_mask; + I915_WRITE(IMR, dev_priv->irq_mask); + POSTING_READ(IMR); } - spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags); + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return true; } static void -render_ring_put_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) +i9xx_ring_put_irq(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - unsigned long irqflags; + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; - spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags); - BUG_ON(dev->irq_enabled && ring->user_irq_refcount <= 0); - if (dev->irq_enabled && (--ring->user_irq_refcount == 0)) { - if (HAS_PCH_SPLIT(dev)) - ironlake_disable_graphics_irq(dev_priv, GT_PIPE_NOTIFY); - else - i915_disable_irq(dev_priv, I915_USER_INTERRUPT); + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) { + dev_priv->irq_mask |= ring->irq_enable_mask; + I915_WRITE(IMR, dev_priv->irq_mask); + POSTING_READ(IMR); } - spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags); + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); } -void intel_ring_setup_status_page(struct drm_device *dev, - struct intel_ring_buffer *ring) +static bool +i8xx_ring_get_irq(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - if (IS_GEN6(dev)) { - I915_WRITE(RING_HWS_PGA_GEN6(ring->mmio_base), - ring->status_page.gfx_addr); - I915_READ(RING_HWS_PGA_GEN6(ring->mmio_base)); /* posting read */ - } else { - I915_WRITE(RING_HWS_PGA(ring->mmio_base), - ring->status_page.gfx_addr); - I915_READ(RING_HWS_PGA(ring->mmio_base)); /* posting read */ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) { + dev_priv->irq_mask &= ~ring->irq_enable_mask; + I915_WRITE16(IMR, dev_priv->irq_mask); + POSTING_READ16(IMR); } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + return true; } static void -bsd_ring_flush(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 invalidate_domains, - u32 flush_domains) +i8xx_ring_put_irq(struct intel_engine_cs *ring) { - intel_ring_begin(dev, ring, 2); - intel_ring_emit(dev, ring, MI_FLUSH); - intel_ring_emit(dev, ring, MI_NOOP); - intel_ring_advance(dev, ring); + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) { + dev_priv->irq_mask |= ring->irq_enable_mask; + I915_WRITE16(IMR, dev_priv->irq_mask); + POSTING_READ16(IMR); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); } -static int init_bsd_ring(struct drm_device *dev, - struct intel_ring_buffer *ring) +void intel_ring_setup_status_page(struct intel_engine_cs *ring) { - return init_ring_common(dev, ring); + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = ring->dev->dev_private; + u32 mmio = 0; + + /* The ring status page addresses are no longer next to the rest of + * the ring registers as of gen7. + */ + if (IS_GEN7(dev)) { + switch (ring->id) { + case RCS: + mmio = RENDER_HWS_PGA_GEN7; + break; + case BCS: + mmio = BLT_HWS_PGA_GEN7; + break; + /* + * VCS2 actually doesn't exist on Gen7. Only shut up + * gcc switch check warning + */ + case VCS2: + case VCS: + mmio = BSD_HWS_PGA_GEN7; + break; + case VECS: + mmio = VEBOX_HWS_PGA_GEN7; + break; + } + } else if (IS_GEN6(ring->dev)) { + mmio = RING_HWS_PGA_GEN6(ring->mmio_base); + } else { + /* XXX: gen8 returns to sanity */ + mmio = RING_HWS_PGA(ring->mmio_base); + } + + I915_WRITE(mmio, (u32)ring->status_page.gfx_addr); + POSTING_READ(mmio); + + /* + * Flush the TLB for this page + * + * FIXME: These two bits have disappeared on gen8, so a question + * arises: do we still need this and if so how should we go about + * invalidating the TLB? + */ + if (INTEL_INFO(dev)->gen >= 6 && INTEL_INFO(dev)->gen < 8) { + u32 reg = RING_INSTPM(ring->mmio_base); + + /* ring should be idle before issuing a sync flush*/ + WARN_ON((I915_READ_MODE(ring) & MODE_IDLE) == 0); + + I915_WRITE(reg, + _MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE | + INSTPM_SYNC_FLUSH)); + if (wait_for((I915_READ(reg) & INSTPM_SYNC_FLUSH) == 0, + 1000)) + DRM_ERROR("%s: wait for SyncFlush to complete for TLB invalidation timed out\n", + ring->name); + } } -static u32 -ring_add_request(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 flush_domains) +static int +bsd_ring_flush(struct intel_engine_cs *ring, + u32 invalidate_domains, + u32 flush_domains) { - u32 seqno; + int ret; - seqno = i915_gem_get_seqno(dev); + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; - intel_ring_begin(dev, ring, 4); - intel_ring_emit(dev, ring, MI_STORE_DWORD_INDEX); - intel_ring_emit(dev, ring, - I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - intel_ring_emit(dev, ring, seqno); - intel_ring_emit(dev, ring, MI_USER_INTERRUPT); - intel_ring_advance(dev, ring); + intel_ring_emit(ring, MI_FLUSH); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + return 0; +} - DRM_DEBUG_DRIVER("%s %d\n", ring->name, seqno); +static int +i9xx_add_request(struct intel_engine_cs *ring) +{ + int ret; + + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, MI_STORE_DWORD_INDEX); + intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + intel_ring_emit(ring, ring->outstanding_lazy_seqno); + intel_ring_emit(ring, MI_USER_INTERRUPT); + __intel_ring_advance(ring); - return seqno; + return 0; } -static void -bsd_ring_get_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) +static bool +gen6_ring_get_irq(struct intel_engine_cs *ring) { - /* do nothing */ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) { + if (HAS_L3_DPF(dev) && ring->id == RCS) + I915_WRITE_IMR(ring, + ~(ring->irq_enable_mask | + GT_PARITY_ERROR(dev))); + else + I915_WRITE_IMR(ring, ~ring->irq_enable_mask); + ilk_enable_gt_irq(dev_priv, ring->irq_enable_mask); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return true; } + static void -bsd_ring_put_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) +gen6_ring_put_irq(struct intel_engine_cs *ring) { - /* do nothing */ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) { + if (HAS_L3_DPF(dev) && ring->id == RCS) + I915_WRITE_IMR(ring, ~GT_PARITY_ERROR(dev)); + else + I915_WRITE_IMR(ring, ~0); + ilk_disable_gt_irq(dev_priv, ring->irq_enable_mask); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); } -static u32 -ring_status_page_get_seqno(struct drm_device *dev, - struct intel_ring_buffer *ring) +static bool +hsw_vebox_get_irq(struct intel_engine_cs *ring) { - return intel_read_status_page(ring, I915_GEM_HWS_INDEX); + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) { + I915_WRITE_IMR(ring, ~ring->irq_enable_mask); + snb_enable_pm_irq(dev_priv, ring->irq_enable_mask); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return true; } -static int -ring_dispatch_gem_execbuffer(struct drm_device *dev, - struct intel_ring_buffer *ring, - struct drm_i915_gem_execbuffer2 *exec, - struct drm_clip_rect *cliprects, - uint64_t exec_offset) -{ - uint32_t exec_start; - exec_start = (uint32_t) exec_offset + exec->batch_start_offset; - intel_ring_begin(dev, ring, 2); - intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START | - (2 << 6) | MI_BATCH_NON_SECURE_I965); - intel_ring_emit(dev, ring, exec_start); - intel_ring_advance(dev, ring); - return 0; +static void +hsw_vebox_put_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) { + I915_WRITE_IMR(ring, ~0); + snb_disable_pm_irq(dev_priv, ring->irq_enable_mask); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); } -static int -render_ring_dispatch_gem_execbuffer(struct drm_device *dev, - struct intel_ring_buffer *ring, - struct drm_i915_gem_execbuffer2 *exec, - struct drm_clip_rect *cliprects, - uint64_t exec_offset) -{ - drm_i915_private_t *dev_priv = dev->dev_private; - int nbox = exec->num_cliprects; - int i = 0, count; - uint32_t exec_start, exec_len; - exec_start = (uint32_t) exec_offset + exec->batch_start_offset; - exec_len = (uint32_t) exec->batch_len; - - trace_i915_gem_request_submit(dev, dev_priv->next_seqno + 1); - - count = nbox ? nbox : 1; - - for (i = 0; i < count; i++) { - if (i < nbox) { - int ret = i915_emit_box(dev, cliprects, i, - exec->DR1, exec->DR4); - if (ret) - return ret; +static bool +gen8_ring_get_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + if (!dev->irq_enabled) + return false; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (ring->irq_refcount++ == 0) { + if (HAS_L3_DPF(dev) && ring->id == RCS) { + I915_WRITE_IMR(ring, + ~(ring->irq_enable_mask | + GT_RENDER_L3_PARITY_ERROR_INTERRUPT)); + } else { + I915_WRITE_IMR(ring, ~ring->irq_enable_mask); } + POSTING_READ(RING_IMR(ring->mmio_base)); + } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); + + return true; +} - if (IS_I830(dev) || IS_845G(dev)) { - intel_ring_begin(dev, ring, 4); - intel_ring_emit(dev, ring, MI_BATCH_BUFFER); - intel_ring_emit(dev, ring, - exec_start | MI_BATCH_NON_SECURE); - intel_ring_emit(dev, ring, exec_start + exec_len - 4); - intel_ring_emit(dev, ring, 0); +static void +gen8_ring_put_irq(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long flags; + + spin_lock_irqsave(&dev_priv->irq_lock, flags); + if (--ring->irq_refcount == 0) { + if (HAS_L3_DPF(dev) && ring->id == RCS) { + I915_WRITE_IMR(ring, + ~GT_RENDER_L3_PARITY_ERROR_INTERRUPT); } else { - intel_ring_begin(dev, ring, 2); - if (INTEL_INFO(dev)->gen >= 4) { - intel_ring_emit(dev, ring, - MI_BATCH_BUFFER_START | (2 << 6) - | MI_BATCH_NON_SECURE_I965); - intel_ring_emit(dev, ring, exec_start); - } else { - intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START - | (2 << 6)); - intel_ring_emit(dev, ring, exec_start | - MI_BATCH_NON_SECURE); - } + I915_WRITE_IMR(ring, ~0); } - intel_ring_advance(dev, ring); + POSTING_READ(RING_IMR(ring->mmio_base)); } + spin_unlock_irqrestore(&dev_priv->irq_lock, flags); +} - if (IS_G4X(dev) || IS_GEN5(dev)) { - intel_ring_begin(dev, ring, 2); - intel_ring_emit(dev, ring, MI_FLUSH | - MI_NO_WRITE_FLUSH | - MI_INVALIDATE_ISP ); - intel_ring_emit(dev, ring, MI_NOOP); - intel_ring_advance(dev, ring); +static int +i965_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 length, + unsigned flags) +{ + int ret; + + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + + intel_ring_emit(ring, + MI_BATCH_BUFFER_START | + MI_BATCH_GTT | + (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965)); + intel_ring_emit(ring, offset); + intel_ring_advance(ring); + + return 0; +} + +/* Just userspace ABI convention to limit the wa batch bo to a resonable size */ +#define I830_BATCH_LIMIT (256*1024) +static int +i830_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 len, + unsigned flags) +{ + int ret; + + if (flags & I915_DISPATCH_PINNED) { + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, MI_BATCH_BUFFER); + intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); + intel_ring_emit(ring, offset + len - 8); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + } else { + u32 cs_offset = ring->scratch.gtt_offset; + + if (len > I830_BATCH_LIMIT) + return -ENOSPC; + + ret = intel_ring_begin(ring, 9+3); + if (ret) + return ret; + /* Blit the batch (which has now all relocs applied) to the stable batch + * scratch bo area (so that the CS never stumbles over its tlb + * invalidation bug) ... */ + intel_ring_emit(ring, XY_SRC_COPY_BLT_CMD | + XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + intel_ring_emit(ring, BLT_DEPTH_32 | BLT_ROP_GXCOPY | 4096); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, (DIV_ROUND_UP(len, 4096) << 16) | 1024); + intel_ring_emit(ring, cs_offset); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 4096); + intel_ring_emit(ring, offset); + intel_ring_emit(ring, MI_FLUSH); + + /* ... and execute it. */ + intel_ring_emit(ring, MI_BATCH_BUFFER); + intel_ring_emit(ring, cs_offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); + intel_ring_emit(ring, cs_offset + len - 8); + intel_ring_advance(ring); } - /* XXX breadcrumb */ return 0; } -static void cleanup_status_page(struct drm_device *dev, - struct intel_ring_buffer *ring) +static int +i915_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 len, + unsigned flags) { - drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_gem_object *obj; - struct drm_i915_gem_object *obj_priv; + int ret; + + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + + intel_ring_emit(ring, MI_BATCH_BUFFER_START | MI_BATCH_GTT); + intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); + intel_ring_advance(ring); + + return 0; +} + +static void cleanup_status_page(struct intel_engine_cs *ring) +{ + struct drm_i915_gem_object *obj; obj = ring->status_page.obj; if (obj == NULL) return; - obj_priv = to_intel_bo(obj); - kunmap(obj_priv->pages[0]); - i915_gem_object_unpin(obj); - drm_gem_object_unreference(obj); + kunmap(sg_page(obj->pages->sgl)); + i915_gem_object_ggtt_unpin(obj); + drm_gem_object_unreference(&obj->base); ring->status_page.obj = NULL; - - memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); } -static int init_status_page(struct drm_device *dev, - struct intel_ring_buffer *ring) +static int init_status_page(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_gem_object *obj; - struct drm_i915_gem_object *obj_priv; - int ret; + struct drm_i915_gem_object *obj; - obj = i915_gem_alloc_object(dev, 4096); - if (obj == NULL) { - DRM_ERROR("Failed to allocate status page\n"); - ret = -ENOMEM; - goto err; - } - obj_priv = to_intel_bo(obj); - obj_priv->agp_type = AGP_USER_CACHED_MEMORY; + if ((obj = ring->status_page.obj) == NULL) { + int ret; - ret = i915_gem_object_pin(obj, 4096); - if (ret != 0) { - goto err_unref; - } + obj = i915_gem_alloc_object(ring->dev, 4096); + if (obj == NULL) { + DRM_ERROR("Failed to allocate status page\n"); + return -ENOMEM; + } - ring->status_page.gfx_addr = obj_priv->gtt_offset; - ring->status_page.page_addr = kmap(obj_priv->pages[0]); - if (ring->status_page.page_addr == NULL) { - memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); - goto err_unpin; + ret = i915_gem_object_set_cache_level(obj, I915_CACHE_LLC); + if (ret) + goto err_unref; + + ret = i915_gem_obj_ggtt_pin(obj, 4096, 0); + if (ret) { +err_unref: + drm_gem_object_unreference(&obj->base); + return ret; + } + + ring->status_page.obj = obj; } - ring->status_page.obj = obj; + + ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(obj); + ring->status_page.page_addr = kmap(sg_page(obj->pages->sgl)); memset(ring->status_page.page_addr, 0, PAGE_SIZE); - intel_ring_setup_status_page(dev, ring); DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n", ring->name, ring->status_page.gfx_addr); return 0; +} + +static int init_phys_status_page(struct intel_engine_cs *ring) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + + if (!dev_priv->status_page_dmah) { + dev_priv->status_page_dmah = + drm_pci_alloc(ring->dev, PAGE_SIZE, PAGE_SIZE); + if (!dev_priv->status_page_dmah) + return -ENOMEM; + } + + ring->status_page.page_addr = dev_priv->status_page_dmah->vaddr; + memset(ring->status_page.page_addr, 0, PAGE_SIZE); + + return 0; +} + +static int allocate_ring_buffer(struct intel_engine_cs *ring) +{ + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = to_i915(dev); + struct intel_ringbuffer *ringbuf = ring->buffer; + struct drm_i915_gem_object *obj; + int ret; + + if (intel_ring_initialized(ring)) + return 0; + + obj = NULL; + if (!HAS_LLC(dev)) + obj = i915_gem_object_create_stolen(dev, ringbuf->size); + if (obj == NULL) + obj = i915_gem_alloc_object(dev, ringbuf->size); + if (obj == NULL) + return -ENOMEM; + + ret = i915_gem_obj_ggtt_pin(obj, PAGE_SIZE, PIN_MAPPABLE); + if (ret) + goto err_unref; + + ret = i915_gem_object_set_to_gtt_domain(obj, true); + if (ret) + goto err_unpin; + + ringbuf->virtual_start = + ioremap_wc(dev_priv->gtt.mappable_base + i915_gem_obj_ggtt_offset(obj), + ringbuf->size); + if (ringbuf->virtual_start == NULL) { + ret = -EINVAL; + goto err_unpin; + } + + ringbuf->obj = obj; + return 0; err_unpin: - i915_gem_object_unpin(obj); + i915_gem_object_ggtt_unpin(obj); err_unref: - drm_gem_object_unreference(obj); -err: + drm_gem_object_unreference(&obj->base); return ret; } -int intel_init_ring_buffer(struct drm_device *dev, - struct intel_ring_buffer *ring) +static int intel_init_ring_buffer(struct drm_device *dev, + struct intel_engine_cs *ring) { - struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_i915_gem_object *obj_priv; - struct drm_gem_object *obj; + struct intel_ringbuffer *ringbuf = ring->buffer; int ret; + if (ringbuf == NULL) { + ringbuf = kzalloc(sizeof(*ringbuf), GFP_KERNEL); + if (!ringbuf) + return -ENOMEM; + ring->buffer = ringbuf; + } + ring->dev = dev; INIT_LIST_HEAD(&ring->active_list); INIT_LIST_HEAD(&ring->request_list); - INIT_LIST_HEAD(&ring->gpu_write_list); + ringbuf->size = 32 * PAGE_SIZE; + memset(ring->semaphore.sync_seqno, 0, sizeof(ring->semaphore.sync_seqno)); + + init_waitqueue_head(&ring->irq_queue); if (I915_NEED_GFX_HWS(dev)) { - ret = init_status_page(dev, ring); + ret = init_status_page(ring); if (ret) - return ret; + goto error; + } else { + BUG_ON(ring->id != RCS); + ret = init_phys_status_page(ring); + if (ret) + goto error; } - obj = i915_gem_alloc_object(dev, ring->size); - if (obj == NULL) { - DRM_ERROR("Failed to allocate ringbuffer\n"); - ret = -ENOMEM; - goto err_hws; + ret = allocate_ring_buffer(ring); + if (ret) { + DRM_ERROR("Failed to allocate ringbuffer %s: %d\n", ring->name, ret); + goto error; } - ring->gem_object = obj; + /* Workaround an erratum on the i830 which causes a hang if + * the TAIL pointer points to within the last 2 cachelines + * of the buffer. + */ + ringbuf->effective_size = ringbuf->size; + if (IS_I830(dev) || IS_845G(dev)) + ringbuf->effective_size -= 2 * CACHELINE_BYTES; - ret = i915_gem_object_pin(obj, PAGE_SIZE); + ret = i915_cmd_parser_init_ring(ring); if (ret) - goto err_unref; - - obj_priv = to_intel_bo(obj); - ring->map.size = ring->size; - ring->map.offset = dev->agp->base + obj_priv->gtt_offset; - ring->map.type = 0; - ring->map.flags = 0; - ring->map.mtrr = 0; + goto error; - drm_core_ioremap_wc(&ring->map, dev); - if (ring->map.handle == NULL) { - DRM_ERROR("Failed to map ringbuffer.\n"); - ret = -EINVAL; - goto err_unpin; - } - - ring->virtual_start = ring->map.handle; - ret = ring->init(dev, ring); + ret = ring->init(ring); if (ret) - goto err_unmap; + goto error; - if (!drm_core_check_feature(dev, DRIVER_MODESET)) - i915_kernel_lost_context(dev); - else { - ring->head = I915_READ_HEAD(ring) & HEAD_ADDR; - ring->tail = I915_READ_TAIL(ring) & TAIL_ADDR; - ring->space = ring->head - (ring->tail + 8); - if (ring->space < 0) - ring->space += ring->size; - } - return ret; + return 0; -err_unmap: - drm_core_ioremapfree(&ring->map, dev); -err_unpin: - i915_gem_object_unpin(obj); -err_unref: - drm_gem_object_unreference(obj); - ring->gem_object = NULL; -err_hws: - cleanup_status_page(dev, ring); +error: + kfree(ringbuf); + ring->buffer = NULL; return ret; } -void intel_cleanup_ring_buffer(struct drm_device *dev, - struct intel_ring_buffer *ring) +void intel_cleanup_ring_buffer(struct intel_engine_cs *ring) { - if (ring->gem_object == NULL) + struct drm_i915_private *dev_priv = to_i915(ring->dev); + struct intel_ringbuffer *ringbuf = ring->buffer; + + if (!intel_ring_initialized(ring)) return; - drm_core_ioremapfree(&ring->map, dev); + intel_stop_ring_buffer(ring); + WARN_ON(!IS_GEN2(ring->dev) && (I915_READ_MODE(ring) & MODE_IDLE) == 0); - i915_gem_object_unpin(ring->gem_object); - drm_gem_object_unreference(ring->gem_object); - ring->gem_object = NULL; + iounmap(ringbuf->virtual_start); + + i915_gem_object_ggtt_unpin(ringbuf->obj); + drm_gem_object_unreference(&ringbuf->obj->base); + ringbuf->obj = NULL; + ring->preallocated_lazy_request = NULL; + ring->outstanding_lazy_seqno = 0; if (ring->cleanup) ring->cleanup(ring); - cleanup_status_page(dev, ring); + cleanup_status_page(ring); + + i915_cmd_parser_fini_ring(ring); + + kfree(ringbuf); + ring->buffer = NULL; } -static int intel_wrap_ring_buffer(struct drm_device *dev, - struct intel_ring_buffer *ring) +static int intel_ring_wait_request(struct intel_engine_cs *ring, int n) { - unsigned int *virt; - int rem; - rem = ring->size - ring->tail; + struct intel_ringbuffer *ringbuf = ring->buffer; + struct drm_i915_gem_request *request; + u32 seqno = 0; + int ret; - if (ring->space < rem) { - int ret = intel_wait_ring_buffer(dev, ring, rem); - if (ret) - return ret; + if (ringbuf->last_retired_head != -1) { + ringbuf->head = ringbuf->last_retired_head; + ringbuf->last_retired_head = -1; + + ringbuf->space = ring_space(ring); + if (ringbuf->space >= n) + return 0; } - virt = (unsigned int *)(ring->virtual_start + ring->tail); - rem /= 8; - while (rem--) { - *virt++ = MI_NOOP; - *virt++ = MI_NOOP; + list_for_each_entry(request, &ring->request_list, list) { + if (__ring_space(request->tail, ringbuf->tail, ringbuf->size) >= n) { + seqno = request->seqno; + break; + } } - ring->tail = 0; - ring->space = ring->head - 8; + if (seqno == 0) + return -ENOSPC; + + ret = i915_wait_seqno(ring, seqno); + if (ret) + return ret; + + i915_gem_retire_requests_ring(ring); + ringbuf->head = ringbuf->last_retired_head; + ringbuf->last_retired_head = -1; + ringbuf->space = ring_space(ring); return 0; } -int intel_wait_ring_buffer(struct drm_device *dev, - struct intel_ring_buffer *ring, int n) +static int ring_wait_for_space(struct intel_engine_cs *ring, int n) { + struct drm_device *dev = ring->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_ringbuffer *ringbuf = ring->buffer; unsigned long end; - drm_i915_private_t *dev_priv = dev->dev_private; - u32 head; - - head = intel_read_status_page(ring, 4); - if (head) { - ring->head = head & HEAD_ADDR; - ring->space = ring->head - (ring->tail + 8); - if (ring->space < 0) - ring->space += ring->size; - if (ring->space >= n) - return 0; - } + int ret; + + ret = intel_ring_wait_request(ring, n); + if (ret != -ENOSPC) + return ret; + + /* force the tail write in case we have been skipping them */ + __intel_ring_advance(ring); + + /* With GEM the hangcheck timer should kick us out of the loop, + * leaving it early runs the risk of corrupting GEM state (due + * to running on almost untested codepaths). But on resume + * timers don't work yet, so prevent a complete hang in that + * case by choosing an insanely large timeout. */ + end = jiffies + 60 * HZ; - trace_i915_ring_wait_begin (dev); - end = jiffies + 3 * HZ; + trace_i915_ring_wait_begin(ring); do { - ring->head = I915_READ_HEAD(ring) & HEAD_ADDR; - ring->space = ring->head - (ring->tail + 8); - if (ring->space < 0) - ring->space += ring->size; - if (ring->space >= n) { - trace_i915_ring_wait_end (dev); - return 0; + ringbuf->head = I915_READ_HEAD(ring); + ringbuf->space = ring_space(ring); + if (ringbuf->space >= n) { + ret = 0; + break; } - if (dev->primary->master) { + if (!drm_core_check_feature(dev, DRIVER_MODESET) && + dev->primary->master) { struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; if (master_priv->sarea_priv) master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; } msleep(1); - } while (!time_after(jiffies, end)); - trace_i915_ring_wait_end (dev); - return -EBUSY; -} - -void intel_ring_begin(struct drm_device *dev, - struct intel_ring_buffer *ring, - int num_dwords) -{ - int n = 4*num_dwords; - if (unlikely(ring->tail + n > ring->size)) - intel_wrap_ring_buffer(dev, ring); - if (unlikely(ring->space < n)) - intel_wait_ring_buffer(dev, ring, n); - - ring->space -= n; -} - -void intel_ring_advance(struct drm_device *dev, - struct intel_ring_buffer *ring) -{ - ring->tail &= ring->size - 1; - ring->write_tail(dev, ring, ring->tail); -} - -static const struct intel_ring_buffer render_ring = { - .name = "render ring", - .id = RING_RENDER, - .mmio_base = RENDER_RING_BASE, - .size = 32 * PAGE_SIZE, - .init = init_render_ring, - .write_tail = ring_write_tail, - .flush = render_ring_flush, - .add_request = render_ring_add_request, - .get_seqno = render_ring_get_seqno, - .user_irq_get = render_ring_get_user_irq, - .user_irq_put = render_ring_put_user_irq, - .dispatch_gem_execbuffer = render_ring_dispatch_gem_execbuffer, -}; - -/* ring buffer for bit-stream decoder */ - -static const struct intel_ring_buffer bsd_ring = { - .name = "bsd ring", - .id = RING_BSD, - .mmio_base = BSD_RING_BASE, - .size = 32 * PAGE_SIZE, - .init = init_bsd_ring, - .write_tail = ring_write_tail, - .flush = bsd_ring_flush, - .add_request = ring_add_request, - .get_seqno = ring_status_page_get_seqno, - .user_irq_get = bsd_ring_get_user_irq, - .user_irq_put = bsd_ring_put_user_irq, - .dispatch_gem_execbuffer = ring_dispatch_gem_execbuffer, -}; - - -static void gen6_bsd_ring_write_tail(struct drm_device *dev, - struct intel_ring_buffer *ring, + + if (dev_priv->mm.interruptible && signal_pending(current)) { + ret = -ERESTARTSYS; + break; + } + + ret = i915_gem_check_wedge(&dev_priv->gpu_error, + dev_priv->mm.interruptible); + if (ret) + break; + + if (time_after(jiffies, end)) { + ret = -EBUSY; + break; + } + } while (1); + trace_i915_ring_wait_end(ring); + return ret; +} + +static int intel_wrap_ring_buffer(struct intel_engine_cs *ring) +{ + uint32_t __iomem *virt; + struct intel_ringbuffer *ringbuf = ring->buffer; + int rem = ringbuf->size - ringbuf->tail; + + if (ringbuf->space < rem) { + int ret = ring_wait_for_space(ring, rem); + if (ret) + return ret; + } + + virt = ringbuf->virtual_start + ringbuf->tail; + rem /= 4; + while (rem--) + iowrite32(MI_NOOP, virt++); + + ringbuf->tail = 0; + ringbuf->space = ring_space(ring); + + return 0; +} + +int intel_ring_idle(struct intel_engine_cs *ring) +{ + u32 seqno; + int ret; + + /* We need to add any requests required to flush the objects and ring */ + if (ring->outstanding_lazy_seqno) { + ret = i915_add_request(ring, NULL); + if (ret) + return ret; + } + + /* Wait upon the last request to be completed */ + if (list_empty(&ring->request_list)) + return 0; + + seqno = list_entry(ring->request_list.prev, + struct drm_i915_gem_request, + list)->seqno; + + return i915_wait_seqno(ring, seqno); +} + +static int +intel_ring_alloc_seqno(struct intel_engine_cs *ring) +{ + if (ring->outstanding_lazy_seqno) + return 0; + + if (ring->preallocated_lazy_request == NULL) { + struct drm_i915_gem_request *request; + + request = kmalloc(sizeof(*request), GFP_KERNEL); + if (request == NULL) + return -ENOMEM; + + ring->preallocated_lazy_request = request; + } + + return i915_gem_get_seqno(ring->dev, &ring->outstanding_lazy_seqno); +} + +static int __intel_ring_prepare(struct intel_engine_cs *ring, + int bytes) +{ + struct intel_ringbuffer *ringbuf = ring->buffer; + int ret; + + if (unlikely(ringbuf->tail + bytes > ringbuf->effective_size)) { + ret = intel_wrap_ring_buffer(ring); + if (unlikely(ret)) + return ret; + } + + if (unlikely(ringbuf->space < bytes)) { + ret = ring_wait_for_space(ring, bytes); + if (unlikely(ret)) + return ret; + } + + return 0; +} + +int intel_ring_begin(struct intel_engine_cs *ring, + int num_dwords) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + int ret; + + ret = i915_gem_check_wedge(&dev_priv->gpu_error, + dev_priv->mm.interruptible); + if (ret) + return ret; + + ret = __intel_ring_prepare(ring, num_dwords * sizeof(uint32_t)); + if (ret) + return ret; + + /* Preallocate the olr before touching the ring */ + ret = intel_ring_alloc_seqno(ring); + if (ret) + return ret; + + ring->buffer->space -= num_dwords * sizeof(uint32_t); + return 0; +} + +/* Align the ring tail to a cacheline boundary */ +int intel_ring_cacheline_align(struct intel_engine_cs *ring) +{ + int num_dwords = (ring->buffer->tail & (CACHELINE_BYTES - 1)) / sizeof(uint32_t); + int ret; + + if (num_dwords == 0) + return 0; + + num_dwords = CACHELINE_BYTES / sizeof(uint32_t) - num_dwords; + ret = intel_ring_begin(ring, num_dwords); + if (ret) + return ret; + + while (num_dwords--) + intel_ring_emit(ring, MI_NOOP); + + intel_ring_advance(ring); + + return 0; +} + +void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + + BUG_ON(ring->outstanding_lazy_seqno); + + if (INTEL_INFO(ring->dev)->gen >= 6) { + I915_WRITE(RING_SYNC_0(ring->mmio_base), 0); + I915_WRITE(RING_SYNC_1(ring->mmio_base), 0); + if (HAS_VEBOX(ring->dev)) + I915_WRITE(RING_SYNC_2(ring->mmio_base), 0); + } + + ring->set_seqno(ring, seqno); + ring->hangcheck.seqno = seqno; +} + +static void gen6_bsd_ring_write_tail(struct intel_engine_cs *ring, u32 value) { - drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_private *dev_priv = ring->dev->dev_private; /* Every tail move must follow the sequence below */ - I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL, - GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_MODIFY_MASK | - GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_DISABLE); - I915_WRITE(GEN6_BSD_RNCID, 0x0); - if (wait_for((I915_READ(GEN6_BSD_SLEEP_PSMI_CONTROL) & - GEN6_BSD_SLEEP_PSMI_CONTROL_IDLE_INDICATOR) == 0, - 50)) - DRM_ERROR("timed out waiting for IDLE Indicator\n"); + /* Disable notification that the ring is IDLE. The GT + * will then assume that it is busy and bring it out of rc6. + */ + I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL, + _MASKED_BIT_ENABLE(GEN6_BSD_SLEEP_MSG_DISABLE)); + + /* Clear the context id. Here be magic! */ + I915_WRITE64(GEN6_BSD_RNCID, 0x0); - I915_WRITE_TAIL(ring, value); - I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL, - GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_MODIFY_MASK | - GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_ENABLE); + /* Wait for the ring not to be idle, i.e. for it to wake up. */ + if (wait_for((I915_READ(GEN6_BSD_SLEEP_PSMI_CONTROL) & + GEN6_BSD_SLEEP_INDICATOR) == 0, + 50)) + DRM_ERROR("timed out waiting for the BSD ring to wake up\n"); + + /* Now that the ring is fully powered up, update the tail */ + I915_WRITE_TAIL(ring, value); + POSTING_READ(RING_TAIL(ring->mmio_base)); + + /* Let the ring send IDLE messages to the GT again, + * and so let it sleep to conserve power when idle. + */ + I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL, + _MASKED_BIT_DISABLE(GEN6_BSD_SLEEP_MSG_DISABLE)); } -static void gen6_ring_flush(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 invalidate_domains, - u32 flush_domains) +static int gen6_bsd_ring_flush(struct intel_engine_cs *ring, + u32 invalidate, u32 flush) { - intel_ring_begin(dev, ring, 4); - intel_ring_emit(dev, ring, MI_FLUSH_DW); - intel_ring_emit(dev, ring, 0); - intel_ring_emit(dev, ring, 0); - intel_ring_emit(dev, ring, 0); - intel_ring_advance(dev, ring); + uint32_t cmd; + int ret; + + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + cmd = MI_FLUSH_DW; + if (INTEL_INFO(ring->dev)->gen >= 8) + cmd += 1; + /* + * Bspec vol 1c.5 - video engine command streamer: + * "If ENABLED, all TLBs will be invalidated once the flush + * operation is complete. This bit is only valid when the + * Post-Sync Operation field is a value of 1h or 3h." + */ + if (invalidate & I915_GEM_GPU_DOMAINS) + cmd |= MI_INVALIDATE_TLB | MI_INVALIDATE_BSD | + MI_FLUSH_DW_STORE_INDEX | MI_FLUSH_DW_OP_STOREDW; + intel_ring_emit(ring, cmd); + intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT); + if (INTEL_INFO(ring->dev)->gen >= 8) { + intel_ring_emit(ring, 0); /* upper addr */ + intel_ring_emit(ring, 0); /* value */ + } else { + intel_ring_emit(ring, 0); + intel_ring_emit(ring, MI_NOOP); + } + intel_ring_advance(ring); + return 0; } static int -gen6_ring_dispatch_gem_execbuffer(struct drm_device *dev, - struct intel_ring_buffer *ring, - struct drm_i915_gem_execbuffer2 *exec, - struct drm_clip_rect *cliprects, - uint64_t exec_offset) -{ - uint32_t exec_start; - - exec_start = (uint32_t) exec_offset + exec->batch_start_offset; - - intel_ring_begin(dev, ring, 2); - intel_ring_emit(dev, ring, - MI_BATCH_BUFFER_START | MI_BATCH_NON_SECURE_I965); - /* bit0-7 is the length on GEN6+ */ - intel_ring_emit(dev, ring, exec_start); - intel_ring_advance(dev, ring); - - return 0; -} - -/* ring buffer for Video Codec for Gen6+ */ -static const struct intel_ring_buffer gen6_bsd_ring = { - .name = "gen6 bsd ring", - .id = RING_BSD, - .mmio_base = GEN6_BSD_RING_BASE, - .size = 32 * PAGE_SIZE, - .init = init_bsd_ring, - .write_tail = gen6_bsd_ring_write_tail, - .flush = gen6_ring_flush, - .add_request = ring_add_request, - .get_seqno = ring_status_page_get_seqno, - .user_irq_get = bsd_ring_get_user_irq, - .user_irq_put = bsd_ring_put_user_irq, - .dispatch_gem_execbuffer = gen6_ring_dispatch_gem_execbuffer, -}; +gen8_ring_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 len, + unsigned flags) +{ + struct drm_i915_private *dev_priv = ring->dev->dev_private; + bool ppgtt = dev_priv->mm.aliasing_ppgtt != NULL && + !(flags & I915_DISPATCH_SECURE); + int ret; -/* Blitter support (SandyBridge+) */ + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; -static void -blt_ring_get_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) -{ - /* do nothing */ + /* FIXME(BDW): Address space and security selectors. */ + intel_ring_emit(ring, MI_BATCH_BUFFER_START_GEN8 | (ppgtt<<8)); + intel_ring_emit(ring, lower_32_bits(offset)); + intel_ring_emit(ring, upper_32_bits(offset)); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + + return 0; } -static void -blt_ring_put_user_irq(struct drm_device *dev, - struct intel_ring_buffer *ring) + +static int +hsw_ring_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 len, + unsigned flags) { - /* do nothing */ + int ret; + + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + + intel_ring_emit(ring, + MI_BATCH_BUFFER_START | MI_BATCH_PPGTT_HSW | + (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_HSW)); + /* bit0-7 is the length on GEN6+ */ + intel_ring_emit(ring, offset); + intel_ring_advance(ring); + + return 0; } +static int +gen6_ring_dispatch_execbuffer(struct intel_engine_cs *ring, + u64 offset, u32 len, + unsigned flags) +{ + int ret; -/* Workaround for some stepping of SNB, - * each time when BLT engine ring tail moved, - * the first command in the ring to be parsed - * should be MI_BATCH_BUFFER_START - */ -#define NEED_BLT_WORKAROUND(dev) \ - (IS_GEN6(dev) && (dev->pdev->revision < 8)) + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; -static inline struct drm_i915_gem_object * -to_blt_workaround(struct intel_ring_buffer *ring) + intel_ring_emit(ring, + MI_BATCH_BUFFER_START | + (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965)); + /* bit0-7 is the length on GEN6+ */ + intel_ring_emit(ring, offset); + intel_ring_advance(ring); + + return 0; +} + +/* Blitter support (SandyBridge+) */ + +static int gen6_ring_flush(struct intel_engine_cs *ring, + u32 invalidate, u32 flush) { - return ring->private; + struct drm_device *dev = ring->dev; + uint32_t cmd; + int ret; + + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + cmd = MI_FLUSH_DW; + if (INTEL_INFO(ring->dev)->gen >= 8) + cmd += 1; + /* + * Bspec vol 1c.3 - blitter engine command streamer: + * "If ENABLED, all TLBs will be invalidated once the flush + * operation is complete. This bit is only valid when the + * Post-Sync Operation field is a value of 1h or 3h." + */ + if (invalidate & I915_GEM_DOMAIN_RENDER) + cmd |= MI_INVALIDATE_TLB | MI_FLUSH_DW_STORE_INDEX | + MI_FLUSH_DW_OP_STOREDW; + intel_ring_emit(ring, cmd); + intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT); + if (INTEL_INFO(ring->dev)->gen >= 8) { + intel_ring_emit(ring, 0); /* upper addr */ + intel_ring_emit(ring, 0); /* value */ + } else { + intel_ring_emit(ring, 0); + intel_ring_emit(ring, MI_NOOP); + } + intel_ring_advance(ring); + + if (IS_GEN7(dev) && !invalidate && flush) + return gen7_ring_fbc_flush(ring, FBC_REND_CACHE_CLEAN); + + return 0; } -static int blt_ring_init(struct drm_device *dev, - struct intel_ring_buffer *ring) +int intel_init_render_ring_buffer(struct drm_device *dev) { - if (NEED_BLT_WORKAROUND(dev)) { + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[RCS]; + + ring->name = "render ring"; + ring->id = RCS; + ring->mmio_base = RENDER_RING_BASE; + + if (INTEL_INFO(dev)->gen >= 6) { + ring->add_request = gen6_add_request; + ring->flush = gen7_render_ring_flush; + if (INTEL_INFO(dev)->gen == 6) + ring->flush = gen6_render_ring_flush; + if (INTEL_INFO(dev)->gen >= 8) { + ring->flush = gen8_render_ring_flush; + ring->irq_get = gen8_ring_get_irq; + ring->irq_put = gen8_ring_put_irq; + } else { + ring->irq_get = gen6_ring_get_irq; + ring->irq_put = gen6_ring_put_irq; + } + ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT; + ring->get_seqno = gen6_ring_get_seqno; + ring->set_seqno = ring_set_seqno; + ring->semaphore.sync_to = gen6_ring_sync; + ring->semaphore.signal = gen6_signal; + /* + * The current semaphore is only applied on pre-gen8 platform. + * And there is no VCS2 ring on the pre-gen8 platform. So the + * semaphore between RCS and VCS2 is initialized as INVALID. + * Gen8 will initialize the sema between VCS2 and RCS later. + */ + ring->semaphore.mbox.wait[RCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VCS] = MI_SEMAPHORE_SYNC_RV; + ring->semaphore.mbox.wait[BCS] = MI_SEMAPHORE_SYNC_RB; + ring->semaphore.mbox.wait[VECS] = MI_SEMAPHORE_SYNC_RVE; + ring->semaphore.mbox.wait[VCS2] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.signal[RCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VCS] = GEN6_VRSYNC; + ring->semaphore.mbox.signal[BCS] = GEN6_BRSYNC; + ring->semaphore.mbox.signal[VECS] = GEN6_VERSYNC; + ring->semaphore.mbox.signal[VCS2] = GEN6_NOSYNC; + } else if (IS_GEN5(dev)) { + ring->add_request = pc_render_add_request; + ring->flush = gen4_render_ring_flush; + ring->get_seqno = pc_render_get_seqno; + ring->set_seqno = pc_render_set_seqno; + ring->irq_get = gen5_ring_get_irq; + ring->irq_put = gen5_ring_put_irq; + ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT | + GT_RENDER_PIPECTL_NOTIFY_INTERRUPT; + } else { + ring->add_request = i9xx_add_request; + if (INTEL_INFO(dev)->gen < 4) + ring->flush = gen2_render_ring_flush; + else + ring->flush = gen4_render_ring_flush; + ring->get_seqno = ring_get_seqno; + ring->set_seqno = ring_set_seqno; + if (IS_GEN2(dev)) { + ring->irq_get = i8xx_ring_get_irq; + ring->irq_put = i8xx_ring_put_irq; + } else { + ring->irq_get = i9xx_ring_get_irq; + ring->irq_put = i9xx_ring_put_irq; + } + ring->irq_enable_mask = I915_USER_INTERRUPT; + } + ring->write_tail = ring_write_tail; + if (IS_HASWELL(dev)) + ring->dispatch_execbuffer = hsw_ring_dispatch_execbuffer; + else if (IS_GEN8(dev)) + ring->dispatch_execbuffer = gen8_ring_dispatch_execbuffer; + else if (INTEL_INFO(dev)->gen >= 6) + ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer; + else if (INTEL_INFO(dev)->gen >= 4) + ring->dispatch_execbuffer = i965_dispatch_execbuffer; + else if (IS_I830(dev) || IS_845G(dev)) + ring->dispatch_execbuffer = i830_dispatch_execbuffer; + else + ring->dispatch_execbuffer = i915_dispatch_execbuffer; + ring->init = init_render_ring; + ring->cleanup = render_ring_cleanup; + + /* Workaround batchbuffer to combat CS tlb bug. */ + if (HAS_BROKEN_CS_TLB(dev)) { struct drm_i915_gem_object *obj; - u32 __iomem *ptr; int ret; - obj = to_intel_bo(i915_gem_alloc_object(dev, 4096)); - if (obj == NULL) + obj = i915_gem_alloc_object(dev, I830_BATCH_LIMIT); + if (obj == NULL) { + DRM_ERROR("Failed to allocate batch bo\n"); return -ENOMEM; - - ret = i915_gem_object_pin(&obj->base, 4096); - if (ret) { - drm_gem_object_unreference(&obj->base); - return ret; } - ptr = kmap(obj->pages[0]); - iowrite32(MI_BATCH_BUFFER_END, ptr); - iowrite32(MI_NOOP, ptr+1); - kunmap(obj->pages[0]); - - ret = i915_gem_object_set_to_gtt_domain(&obj->base, false); - if (ret) { - i915_gem_object_unpin(&obj->base); + ret = i915_gem_obj_ggtt_pin(obj, 0, 0); + if (ret != 0) { drm_gem_object_unreference(&obj->base); + DRM_ERROR("Failed to ping batch bo\n"); return ret; } - ring->private = obj; + ring->scratch.obj = obj; + ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(obj); } - return init_ring_common(dev, ring); + return intel_init_ring_buffer(dev, ring); } -static void blt_ring_begin(struct drm_device *dev, - struct intel_ring_buffer *ring, - int num_dwords) +int intel_render_ring_init_dri(struct drm_device *dev, u64 start, u32 size) { - if (ring->private) { - intel_ring_begin(dev, ring, num_dwords+2); - intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START); - intel_ring_emit(dev, ring, to_blt_workaround(ring)->gtt_offset); - } else - intel_ring_begin(dev, ring, 4); + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[RCS]; + struct intel_ringbuffer *ringbuf = ring->buffer; + int ret; + + if (ringbuf == NULL) { + ringbuf = kzalloc(sizeof(*ringbuf), GFP_KERNEL); + if (!ringbuf) + return -ENOMEM; + ring->buffer = ringbuf; + } + + ring->name = "render ring"; + ring->id = RCS; + ring->mmio_base = RENDER_RING_BASE; + + if (INTEL_INFO(dev)->gen >= 6) { + /* non-kms not supported on gen6+ */ + ret = -ENODEV; + goto err_ringbuf; + } + + /* Note: gem is not supported on gen5/ilk without kms (the corresponding + * gem_init ioctl returns with -ENODEV). Hence we do not need to set up + * the special gen5 functions. */ + ring->add_request = i9xx_add_request; + if (INTEL_INFO(dev)->gen < 4) + ring->flush = gen2_render_ring_flush; + else + ring->flush = gen4_render_ring_flush; + ring->get_seqno = ring_get_seqno; + ring->set_seqno = ring_set_seqno; + if (IS_GEN2(dev)) { + ring->irq_get = i8xx_ring_get_irq; + ring->irq_put = i8xx_ring_put_irq; + } else { + ring->irq_get = i9xx_ring_get_irq; + ring->irq_put = i9xx_ring_put_irq; + } + ring->irq_enable_mask = I915_USER_INTERRUPT; + ring->write_tail = ring_write_tail; + if (INTEL_INFO(dev)->gen >= 4) + ring->dispatch_execbuffer = i965_dispatch_execbuffer; + else if (IS_I830(dev) || IS_845G(dev)) + ring->dispatch_execbuffer = i830_dispatch_execbuffer; + else + ring->dispatch_execbuffer = i915_dispatch_execbuffer; + ring->init = init_render_ring; + ring->cleanup = render_ring_cleanup; + + ring->dev = dev; + INIT_LIST_HEAD(&ring->active_list); + INIT_LIST_HEAD(&ring->request_list); + + ringbuf->size = size; + ringbuf->effective_size = ringbuf->size; + if (IS_I830(ring->dev) || IS_845G(ring->dev)) + ringbuf->effective_size -= 2 * CACHELINE_BYTES; + + ringbuf->virtual_start = ioremap_wc(start, size); + if (ringbuf->virtual_start == NULL) { + DRM_ERROR("can not ioremap virtual address for" + " ring buffer\n"); + ret = -ENOMEM; + goto err_ringbuf; + } + + if (!I915_NEED_GFX_HWS(dev)) { + ret = init_phys_status_page(ring); + if (ret) + goto err_vstart; + } + + return 0; + +err_vstart: + iounmap(ringbuf->virtual_start); +err_ringbuf: + kfree(ringbuf); + ring->buffer = NULL; + return ret; } -static void blt_ring_flush(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 invalidate_domains, - u32 flush_domains) +int intel_init_bsd_ring_buffer(struct drm_device *dev) { - blt_ring_begin(dev, ring, 4); - intel_ring_emit(dev, ring, MI_FLUSH_DW); - intel_ring_emit(dev, ring, 0); - intel_ring_emit(dev, ring, 0); - intel_ring_emit(dev, ring, 0); - intel_ring_advance(dev, ring); + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[VCS]; + + ring->name = "bsd ring"; + ring->id = VCS; + + ring->write_tail = ring_write_tail; + if (INTEL_INFO(dev)->gen >= 6) { + ring->mmio_base = GEN6_BSD_RING_BASE; + /* gen6 bsd needs a special wa for tail updates */ + if (IS_GEN6(dev)) + ring->write_tail = gen6_bsd_ring_write_tail; + ring->flush = gen6_bsd_ring_flush; + ring->add_request = gen6_add_request; + ring->get_seqno = gen6_ring_get_seqno; + ring->set_seqno = ring_set_seqno; + if (INTEL_INFO(dev)->gen >= 8) { + ring->irq_enable_mask = + GT_RENDER_USER_INTERRUPT << GEN8_VCS1_IRQ_SHIFT; + ring->irq_get = gen8_ring_get_irq; + ring->irq_put = gen8_ring_put_irq; + ring->dispatch_execbuffer = + gen8_ring_dispatch_execbuffer; + } else { + ring->irq_enable_mask = GT_BSD_USER_INTERRUPT; + ring->irq_get = gen6_ring_get_irq; + ring->irq_put = gen6_ring_put_irq; + ring->dispatch_execbuffer = + gen6_ring_dispatch_execbuffer; + } + ring->semaphore.sync_to = gen6_ring_sync; + ring->semaphore.signal = gen6_signal; + /* + * The current semaphore is only applied on pre-gen8 platform. + * And there is no VCS2 ring on the pre-gen8 platform. So the + * semaphore between VCS and VCS2 is initialized as INVALID. + * Gen8 will initialize the sema between VCS2 and VCS later. + */ + ring->semaphore.mbox.wait[RCS] = MI_SEMAPHORE_SYNC_VR; + ring->semaphore.mbox.wait[VCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[BCS] = MI_SEMAPHORE_SYNC_VB; + ring->semaphore.mbox.wait[VECS] = MI_SEMAPHORE_SYNC_VVE; + ring->semaphore.mbox.wait[VCS2] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.signal[RCS] = GEN6_RVSYNC; + ring->semaphore.mbox.signal[VCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[BCS] = GEN6_BVSYNC; + ring->semaphore.mbox.signal[VECS] = GEN6_VEVSYNC; + ring->semaphore.mbox.signal[VCS2] = GEN6_NOSYNC; + } else { + ring->mmio_base = BSD_RING_BASE; + ring->flush = bsd_ring_flush; + ring->add_request = i9xx_add_request; + ring->get_seqno = ring_get_seqno; + ring->set_seqno = ring_set_seqno; + if (IS_GEN5(dev)) { + ring->irq_enable_mask = ILK_BSD_USER_INTERRUPT; + ring->irq_get = gen5_ring_get_irq; + ring->irq_put = gen5_ring_put_irq; + } else { + ring->irq_enable_mask = I915_BSD_USER_INTERRUPT; + ring->irq_get = i9xx_ring_get_irq; + ring->irq_put = i9xx_ring_put_irq; + } + ring->dispatch_execbuffer = i965_dispatch_execbuffer; + } + ring->init = init_ring_common; + + return intel_init_ring_buffer(dev, ring); } -static u32 -blt_ring_add_request(struct drm_device *dev, - struct intel_ring_buffer *ring, - u32 flush_domains) +/** + * Initialize the second BSD ring for Broadwell GT3. + * It is noted that this only exists on Broadwell GT3. + */ +int intel_init_bsd2_ring_buffer(struct drm_device *dev) { - u32 seqno = i915_gem_get_seqno(dev); + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[VCS2]; - blt_ring_begin(dev, ring, 4); - intel_ring_emit(dev, ring, MI_STORE_DWORD_INDEX); - intel_ring_emit(dev, ring, - I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - intel_ring_emit(dev, ring, seqno); - intel_ring_emit(dev, ring, MI_USER_INTERRUPT); - intel_ring_advance(dev, ring); + if ((INTEL_INFO(dev)->gen != 8)) { + DRM_ERROR("No dual-BSD ring on non-BDW machine\n"); + return -EINVAL; + } - DRM_DEBUG_DRIVER("%s %d\n", ring->name, seqno); - return seqno; + ring->name = "bds2_ring"; + ring->id = VCS2; + + ring->write_tail = ring_write_tail; + ring->mmio_base = GEN8_BSD2_RING_BASE; + ring->flush = gen6_bsd_ring_flush; + ring->add_request = gen6_add_request; + ring->get_seqno = gen6_ring_get_seqno; + ring->set_seqno = ring_set_seqno; + ring->irq_enable_mask = + GT_RENDER_USER_INTERRUPT << GEN8_VCS2_IRQ_SHIFT; + ring->irq_get = gen8_ring_get_irq; + ring->irq_put = gen8_ring_put_irq; + ring->dispatch_execbuffer = + gen8_ring_dispatch_execbuffer; + ring->semaphore.sync_to = gen6_ring_sync; + ring->semaphore.signal = gen6_signal; + /* + * The current semaphore is only applied on the pre-gen8. And there + * is no bsd2 ring on the pre-gen8. So now the semaphore_register + * between VCS2 and other ring is initialized as invalid. + * Gen8 will initialize the sema between VCS2 and other ring later. + */ + ring->semaphore.mbox.wait[RCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[BCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VECS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VCS2] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.signal[RCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[BCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VECS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VCS2] = GEN6_NOSYNC; + + ring->init = init_ring_common; + + return intel_init_ring_buffer(dev, ring); } -static void blt_ring_cleanup(struct intel_ring_buffer *ring) +int intel_init_blt_ring_buffer(struct drm_device *dev) { - if (!ring->private) - return; + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[BCS]; + + ring->name = "blitter ring"; + ring->id = BCS; + + ring->mmio_base = BLT_RING_BASE; + ring->write_tail = ring_write_tail; + ring->flush = gen6_ring_flush; + ring->add_request = gen6_add_request; + ring->get_seqno = gen6_ring_get_seqno; + ring->set_seqno = ring_set_seqno; + if (INTEL_INFO(dev)->gen >= 8) { + ring->irq_enable_mask = + GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT; + ring->irq_get = gen8_ring_get_irq; + ring->irq_put = gen8_ring_put_irq; + ring->dispatch_execbuffer = gen8_ring_dispatch_execbuffer; + } else { + ring->irq_enable_mask = GT_BLT_USER_INTERRUPT; + ring->irq_get = gen6_ring_get_irq; + ring->irq_put = gen6_ring_put_irq; + ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer; + } + ring->semaphore.sync_to = gen6_ring_sync; + ring->semaphore.signal = gen6_signal; + /* + * The current semaphore is only applied on pre-gen8 platform. And + * there is no VCS2 ring on the pre-gen8 platform. So the semaphore + * between BCS and VCS2 is initialized as INVALID. + * Gen8 will initialize the sema between BCS and VCS2 later. + */ + ring->semaphore.mbox.wait[RCS] = MI_SEMAPHORE_SYNC_BR; + ring->semaphore.mbox.wait[VCS] = MI_SEMAPHORE_SYNC_BV; + ring->semaphore.mbox.wait[BCS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VECS] = MI_SEMAPHORE_SYNC_BVE; + ring->semaphore.mbox.wait[VCS2] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.signal[RCS] = GEN6_RBSYNC; + ring->semaphore.mbox.signal[VCS] = GEN6_VBSYNC; + ring->semaphore.mbox.signal[BCS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VECS] = GEN6_VEBSYNC; + ring->semaphore.mbox.signal[VCS2] = GEN6_NOSYNC; + ring->init = init_ring_common; + + return intel_init_ring_buffer(dev, ring); +} - i915_gem_object_unpin(ring->private); - drm_gem_object_unreference(ring->private); - ring->private = NULL; -} - -static const struct intel_ring_buffer gen6_blt_ring = { - .name = "blt ring", - .id = RING_BLT, - .mmio_base = BLT_RING_BASE, - .size = 32 * PAGE_SIZE, - .init = blt_ring_init, - .write_tail = ring_write_tail, - .flush = blt_ring_flush, - .add_request = blt_ring_add_request, - .get_seqno = ring_status_page_get_seqno, - .user_irq_get = blt_ring_get_user_irq, - .user_irq_put = blt_ring_put_user_irq, - .dispatch_gem_execbuffer = gen6_ring_dispatch_gem_execbuffer, - .cleanup = blt_ring_cleanup, -}; +int intel_init_vebox_ring_buffer(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring = &dev_priv->ring[VECS]; + + ring->name = "video enhancement ring"; + ring->id = VECS; + + ring->mmio_base = VEBOX_RING_BASE; + ring->write_tail = ring_write_tail; + ring->flush = gen6_ring_flush; + ring->add_request = gen6_add_request; + ring->get_seqno = gen6_ring_get_seqno; + ring->set_seqno = ring_set_seqno; + + if (INTEL_INFO(dev)->gen >= 8) { + ring->irq_enable_mask = + GT_RENDER_USER_INTERRUPT << GEN8_VECS_IRQ_SHIFT; + ring->irq_get = gen8_ring_get_irq; + ring->irq_put = gen8_ring_put_irq; + ring->dispatch_execbuffer = gen8_ring_dispatch_execbuffer; + } else { + ring->irq_enable_mask = PM_VEBOX_USER_INTERRUPT; + ring->irq_get = hsw_vebox_get_irq; + ring->irq_put = hsw_vebox_put_irq; + ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer; + } + ring->semaphore.sync_to = gen6_ring_sync; + ring->semaphore.signal = gen6_signal; + ring->semaphore.mbox.wait[RCS] = MI_SEMAPHORE_SYNC_VER; + ring->semaphore.mbox.wait[VCS] = MI_SEMAPHORE_SYNC_VEV; + ring->semaphore.mbox.wait[BCS] = MI_SEMAPHORE_SYNC_VEB; + ring->semaphore.mbox.wait[VECS] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.wait[VCS2] = MI_SEMAPHORE_SYNC_INVALID; + ring->semaphore.mbox.signal[RCS] = GEN6_RVESYNC; + ring->semaphore.mbox.signal[VCS] = GEN6_VVESYNC; + ring->semaphore.mbox.signal[BCS] = GEN6_BVESYNC; + ring->semaphore.mbox.signal[VECS] = GEN6_NOSYNC; + ring->semaphore.mbox.signal[VCS2] = GEN6_NOSYNC; + ring->init = init_ring_common; + + return intel_init_ring_buffer(dev, ring); +} -int intel_init_render_ring_buffer(struct drm_device *dev) +int +intel_ring_flush_all_caches(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; + int ret; - dev_priv->render_ring = render_ring; + if (!ring->gpu_caches_dirty) + return 0; - if (!I915_NEED_GFX_HWS(dev)) { - dev_priv->render_ring.status_page.page_addr - = dev_priv->status_page_dmah->vaddr; - memset(dev_priv->render_ring.status_page.page_addr, - 0, PAGE_SIZE); - } + ret = ring->flush(ring, 0, I915_GEM_GPU_DOMAINS); + if (ret) + return ret; + + trace_i915_gem_ring_flush(ring, 0, I915_GEM_GPU_DOMAINS); - return intel_init_ring_buffer(dev, &dev_priv->render_ring); + ring->gpu_caches_dirty = false; + return 0; } -int intel_init_bsd_ring_buffer(struct drm_device *dev) +int +intel_ring_invalidate_all_caches(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t flush_domains; + int ret; - if (IS_GEN6(dev)) - dev_priv->bsd_ring = gen6_bsd_ring; - else - dev_priv->bsd_ring = bsd_ring; + flush_domains = 0; + if (ring->gpu_caches_dirty) + flush_domains = I915_GEM_GPU_DOMAINS; + + ret = ring->flush(ring, I915_GEM_GPU_DOMAINS, flush_domains); + if (ret) + return ret; + + trace_i915_gem_ring_flush(ring, I915_GEM_GPU_DOMAINS, flush_domains); - return intel_init_ring_buffer(dev, &dev_priv->bsd_ring); + ring->gpu_caches_dirty = false; + return 0; } -int intel_init_blt_ring_buffer(struct drm_device *dev) +void +intel_stop_ring_buffer(struct intel_engine_cs *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; + int ret; + + if (!intel_ring_initialized(ring)) + return; - dev_priv->blt_ring = gen6_blt_ring; + ret = intel_ring_idle(ring); + if (ret && !i915_reset_in_progress(&to_i915(ring->dev)->gpu_error)) + DRM_ERROR("failed to quiesce %s whilst cleaning up: %d\n", + ring->name, ret); - return intel_init_ring_buffer(dev, &dev_priv->blt_ring); + stop_ring(ring); } |