From e1f99ce6cac3b6a95551642be5ddb5d9c46bea76 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 27 Oct 2010 12:45:26 +0100 Subject: [PATCH] drm/i915: Propagate errors from writing to ringbuffer Preparing the ringbuffer for adding new commands can fail (a timeout whilst waiting for the GPU to catch up and free some space). So check for any potential error before overwriting HEAD with new commands, and propagate that error back to the user where possible. Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_dma.c | 119 +++++++-------- drivers/gpu/drm/i915/i915_drv.h | 28 +--- drivers/gpu/drm/i915/i915_gem.c | 5 +- drivers/gpu/drm/i915/i915_irq.c | 13 +- drivers/gpu/drm/i915/intel_display.c | 51 ++++--- drivers/gpu/drm/i915/intel_overlay.c | 30 +++- drivers/gpu/drm/i915/intel_ringbuffer.c | 189 ++++++++++++++---------- drivers/gpu/drm/i915/intel_ringbuffer.h | 4 +- 8 files changed, 244 insertions(+), 195 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 8a171394a9cf..02daf4e5c8e6 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -263,7 +263,7 @@ static int i915_dma_init(struct drm_device *dev, void *data, * instruction detected will be given a size of zero, which is a * signal to abort the rest of the buffer. */ -static int do_validate_cmd(int cmd) +static int validate_cmd(int cmd) { switch (((cmd >> 29) & 0x7)) { case 0x0: @@ -321,40 +321,27 @@ static int do_validate_cmd(int cmd) return 0; } -static int validate_cmd(int cmd) -{ - int ret = do_validate_cmd(cmd); - -/* printk("validate_cmd( %x ): %d\n", cmd, ret); */ - - return ret; -} - static int i915_emit_cmds(struct drm_device * dev, int *buffer, int dwords) { drm_i915_private_t *dev_priv = dev->dev_private; - int i; + int i, ret; if ((dwords+1) * sizeof(int) >= dev_priv->render_ring.size - 8) return -EINVAL; - BEGIN_LP_RING((dwords+1)&~1); - for (i = 0; i < dwords;) { - int cmd, sz; - - cmd = buffer[i]; - - if ((sz = validate_cmd(cmd)) == 0 || i + sz > dwords) + int sz = validate_cmd(buffer[i]); + if (sz == 0 || i + sz > dwords) return -EINVAL; - - OUT_RING(cmd); - - while (++i, --sz) { - OUT_RING(buffer[i]); - } + i += sz; } + ret = BEGIN_LP_RING((dwords+1)&~1); + if (ret) + return ret; + + for (i = 0; i < dwords; i++) + OUT_RING(buffer[i]); if (dwords & 1) OUT_RING(0); @@ -368,7 +355,9 @@ i915_emit_box(struct drm_device *dev, struct drm_clip_rect *boxes, int i, int DR1, int DR4) { + struct drm_i915_private *dev_priv = dev->dev_private; struct drm_clip_rect box = boxes[i]; + int ret; if (box.y2 <= box.y1 || box.x2 <= box.x1 || box.y2 <= 0 || box.x2 <= 0) { DRM_ERROR("Bad box %d,%d..%d,%d\n", @@ -377,22 +366,27 @@ i915_emit_box(struct drm_device *dev, } if (INTEL_INFO(dev)->gen >= 4) { - BEGIN_LP_RING(4); + ret = BEGIN_LP_RING(4); + if (ret) + return ret; + OUT_RING(GFX_OP_DRAWRECT_INFO_I965); OUT_RING((box.x1 & 0xffff) | (box.y1 << 16)); OUT_RING(((box.x2 - 1) & 0xffff) | ((box.y2 - 1) << 16)); OUT_RING(DR4); - ADVANCE_LP_RING(); } else { - BEGIN_LP_RING(6); + ret = BEGIN_LP_RING(6); + if (ret) + return ret; + OUT_RING(GFX_OP_DRAWRECT_INFO); OUT_RING(DR1); OUT_RING((box.x1 & 0xffff) | (box.y1 << 16)); OUT_RING(((box.x2 - 1) & 0xffff) | ((box.y2 - 1) << 16)); OUT_RING(DR4); OUT_RING(0); - ADVANCE_LP_RING(); } + ADVANCE_LP_RING(); return 0; } @@ -412,12 +406,13 @@ static void i915_emit_breadcrumb(struct drm_device *dev) if (master_priv->sarea_priv) master_priv->sarea_priv->last_enqueue = dev_priv->counter; - BEGIN_LP_RING(4); - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->counter); - OUT_RING(0); - ADVANCE_LP_RING(); + if (BEGIN_LP_RING(4) == 0) { + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + OUT_RING(dev_priv->counter); + OUT_RING(0); + ADVANCE_LP_RING(); + } } static int i915_dispatch_cmdbuffer(struct drm_device * dev, @@ -458,8 +453,9 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev, drm_i915_batchbuffer_t * batch, struct drm_clip_rect *cliprects) { + struct drm_i915_private *dev_priv = dev->dev_private; int nbox = batch->num_cliprects; - int i = 0, count; + int i, count, ret; if ((batch->start | batch->used) & 0x7) { DRM_ERROR("alignment"); @@ -469,17 +465,19 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev, i915_kernel_lost_context(dev); count = nbox ? nbox : 1; - for (i = 0; i < count; i++) { if (i < nbox) { - int ret = i915_emit_box(dev, cliprects, i, - batch->DR1, batch->DR4); + ret = i915_emit_box(dev, cliprects, i, + batch->DR1, batch->DR4); if (ret) return ret; } if (!IS_I830(dev) && !IS_845G(dev)) { - BEGIN_LP_RING(2); + ret = BEGIN_LP_RING(2); + if (ret) + return ret; + if (INTEL_INFO(dev)->gen >= 4) { OUT_RING(MI_BATCH_BUFFER_START | (2 << 6) | MI_BATCH_NON_SECURE_I965); OUT_RING(batch->start); @@ -487,26 +485,29 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev, OUT_RING(MI_BATCH_BUFFER_START | (2 << 6)); OUT_RING(batch->start | MI_BATCH_NON_SECURE); } - ADVANCE_LP_RING(); } else { - BEGIN_LP_RING(4); + ret = BEGIN_LP_RING(4); + if (ret) + return ret; + OUT_RING(MI_BATCH_BUFFER); OUT_RING(batch->start | MI_BATCH_NON_SECURE); OUT_RING(batch->start + batch->used - 4); OUT_RING(0); - ADVANCE_LP_RING(); } + ADVANCE_LP_RING(); } if (IS_G4X(dev) || IS_GEN5(dev)) { - BEGIN_LP_RING(2); - OUT_RING(MI_FLUSH | MI_NO_WRITE_FLUSH | MI_INVALIDATE_ISP); - OUT_RING(MI_NOOP); - ADVANCE_LP_RING(); + if (BEGIN_LP_RING(2) == 0) { + OUT_RING(MI_FLUSH | MI_NO_WRITE_FLUSH | MI_INVALIDATE_ISP); + OUT_RING(MI_NOOP); + ADVANCE_LP_RING(); + } } - i915_emit_breadcrumb(dev); + i915_emit_breadcrumb(dev); return 0; } @@ -515,6 +516,7 @@ static int i915_dispatch_flip(struct drm_device * dev) drm_i915_private_t *dev_priv = dev->dev_private; struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; + int ret; if (!master_priv->sarea_priv) return -EINVAL; @@ -526,12 +528,13 @@ static int i915_dispatch_flip(struct drm_device * dev) i915_kernel_lost_context(dev); - BEGIN_LP_RING(2); + ret = BEGIN_LP_RING(10); + if (ret) + return ret; + OUT_RING(MI_FLUSH | MI_READ_FLUSH); OUT_RING(0); - ADVANCE_LP_RING(); - BEGIN_LP_RING(6); OUT_RING(CMD_OP_DISPLAYBUFFER_INFO | ASYNC_FLIP); OUT_RING(0); if (dev_priv->current_page == 0) { @@ -542,21 +545,21 @@ static int i915_dispatch_flip(struct drm_device * dev) dev_priv->current_page = 0; } OUT_RING(0); - ADVANCE_LP_RING(); - BEGIN_LP_RING(2); OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP); OUT_RING(0); + ADVANCE_LP_RING(); master_priv->sarea_priv->last_enqueue = dev_priv->counter++; - BEGIN_LP_RING(4); - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->counter); - OUT_RING(0); - ADVANCE_LP_RING(); + if (BEGIN_LP_RING(4) == 0) { + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + OUT_RING(dev_priv->counter); + OUT_RING(0); + ADVANCE_LP_RING(); + } master_priv->sarea_priv->pf_current_page = dev_priv->current_page; return 0; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 6fb225f6b2c8..c241468c632e 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1216,30 +1216,14 @@ static inline void i915_write(struct drm_i915_private *dev_priv, u32 reg, #define I915_DEBUG_DISABLE_IO() (dev_priv->debug_flags &= ~(I915_DEBUG_READ | \ I915_DEBUG_WRITE)) -#define I915_VERBOSE 0 +#define BEGIN_LP_RING(n) \ + intel_ring_begin(&dev_priv->render_ring, (n)) -#define BEGIN_LP_RING(n) do { \ - drm_i915_private_t *dev_priv__ = dev->dev_private; \ - if (I915_VERBOSE) \ - DRM_DEBUG(" BEGIN_LP_RING %x\n", (int)(n)); \ - intel_ring_begin(&dev_priv__->render_ring, (n)); \ -} while (0) +#define OUT_RING(x) \ + intel_ring_emit(&dev_priv->render_ring, x) - -#define OUT_RING(x) do { \ - drm_i915_private_t *dev_priv__ = dev->dev_private; \ - if (I915_VERBOSE) \ - DRM_DEBUG(" OUT_RING %x\n", (int)(x)); \ - intel_ring_emit(&dev_priv__->render_ring, x); \ -} while (0) - -#define ADVANCE_LP_RING() do { \ - drm_i915_private_t *dev_priv__ = dev->dev_private; \ - if (I915_VERBOSE) \ - DRM_DEBUG("ADVANCE_LP_RING %x\n", \ - dev_priv__->render_ring.tail); \ - intel_ring_advance(&dev_priv__->render_ring); \ -} while(0) +#define ADVANCE_LP_RING() \ + intel_ring_advance(&dev_priv->render_ring) /** * Reads a dword out of the status page, which is written to from the command diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 97bf7c87d857..00e901483ba5 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3826,7 +3826,10 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data, else flip_mask = MI_WAIT_FOR_PLANE_A_FLIP; - intel_ring_begin(ring, 2); + ret = intel_ring_begin(ring, 2); + if (ret) + goto err; + intel_ring_emit(ring, MI_WAIT_FOR_EVENT | flip_mask); intel_ring_emit(ring, MI_NOOP); intel_ring_advance(ring); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 852a2d848bf4..8acdd6d857d3 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1101,12 +1101,13 @@ static int i915_emit_irq(struct drm_device * dev) if (master_priv->sarea_priv) master_priv->sarea_priv->last_enqueue = dev_priv->counter; - BEGIN_LP_RING(4); - OUT_RING(MI_STORE_DWORD_INDEX); - OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); - OUT_RING(dev_priv->counter); - OUT_RING(MI_USER_INTERRUPT); - ADVANCE_LP_RING(); + if (BEGIN_LP_RING(4) == 0) { + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(I915_BREADCRUMB_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + OUT_RING(dev_priv->counter); + OUT_RING(MI_USER_INTERRUPT); + ADVANCE_LP_RING(); + } return dev_priv->counter; } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 990f065374b2..eb4c725e3069 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -5090,22 +5090,16 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, if (ret) goto cleanup_objs; - /* Block clients from rendering to the new back buffer until - * the flip occurs and the object is no longer visible. - */ - atomic_add(1 << intel_crtc->plane, - &to_intel_bo(work->old_fb_obj)->pending_flip); - - work->pending_flip_obj = obj; - obj_priv = to_intel_bo(obj); - if (IS_GEN3(dev) || IS_GEN2(dev)) { u32 flip_mask; /* Can't queue multiple flips, so wait for the previous * one to finish before executing the next. */ - BEGIN_LP_RING(2); + ret = BEGIN_LP_RING(2); + if (ret) + goto cleanup_objs; + if (intel_crtc->plane) flip_mask = MI_WAIT_FOR_PLANE_B_FLIP; else @@ -5115,13 +5109,25 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, ADVANCE_LP_RING(); } + work->pending_flip_obj = obj; + obj_priv = to_intel_bo(obj); + work->enable_stall_check = true; /* Offset into the new buffer for cases of shared fbs between CRTCs */ offset = crtc->y * fb->pitch + crtc->x * fb->bits_per_pixel/8; - BEGIN_LP_RING(4); - switch(INTEL_INFO(dev)->gen) { + ret = BEGIN_LP_RING(4); + if (ret) + goto cleanup_objs; + + /* Block clients from rendering to the new back buffer until + * the flip occurs and the object is no longer visible. + */ + atomic_add(1 << intel_crtc->plane, + &to_intel_bo(work->old_fb_obj)->pending_flip); + + switch (INTEL_INFO(dev)->gen) { case 2: OUT_RING(MI_DISPLAY_FLIP | MI_DISPLAY_FLIP_PLANE(intel_crtc->plane)); @@ -5850,16 +5856,17 @@ void intel_init_clock_gating(struct drm_device *dev) struct drm_i915_gem_object *obj_priv; obj_priv = to_intel_bo(dev_priv->renderctx); if (obj_priv) { - BEGIN_LP_RING(4); - OUT_RING(MI_SET_CONTEXT); - OUT_RING(obj_priv->gtt_offset | - MI_MM_SPACE_GTT | - MI_SAVE_EXT_STATE_EN | - MI_RESTORE_EXT_STATE_EN | - MI_RESTORE_INHIBIT); - OUT_RING(MI_NOOP); - OUT_RING(MI_FLUSH); - ADVANCE_LP_RING(); + if (BEGIN_LP_RING(4) == 0) { + OUT_RING(MI_SET_CONTEXT); + OUT_RING(obj_priv->gtt_offset | + MI_MM_SPACE_GTT | + MI_SAVE_EXT_STATE_EN | + MI_RESTORE_EXT_STATE_EN | + MI_RESTORE_INHIBIT); + OUT_RING(MI_NOOP); + OUT_RING(MI_FLUSH); + ADVANCE_LP_RING(); + } } } else DRM_DEBUG_KMS("Failed to allocate render context." diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c index afb96d25219a..78fa6a249964 100644 --- a/drivers/gpu/drm/i915/intel_overlay.c +++ b/drivers/gpu/drm/i915/intel_overlay.c @@ -289,6 +289,7 @@ i830_deactivate_pipe_a(struct drm_device *dev) static int intel_overlay_on(struct intel_overlay *overlay) { struct drm_device *dev = overlay->dev; + struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_gem_request *request; int pipe_a_quirk = 0; int ret; @@ -308,7 +309,12 @@ static int intel_overlay_on(struct intel_overlay *overlay) goto out; } - BEGIN_LP_RING(4); + ret = BEGIN_LP_RING(4); + if (ret) { + kfree(request); + goto out; + } + OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON); OUT_RING(overlay->flip_addr | OFC_UPDATE); OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP); @@ -332,6 +338,7 @@ static int intel_overlay_continue(struct intel_overlay *overlay, struct drm_i915_gem_request *request; u32 flip_addr = overlay->flip_addr; u32 tmp; + int ret; BUG_ON(!overlay->active); @@ -347,7 +354,11 @@ static int intel_overlay_continue(struct intel_overlay *overlay, if (tmp & (1 << 17)) DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp); - BEGIN_LP_RING(2); + ret = BEGIN_LP_RING(2); + if (ret) { + kfree(request); + return ret; + } OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE); OUT_RING(flip_addr); ADVANCE_LP_RING(); @@ -389,8 +400,10 @@ static int intel_overlay_off(struct intel_overlay *overlay, bool interruptible) { struct drm_device *dev = overlay->dev; + struct drm_i915_private *dev_priv = dev->dev_private; u32 flip_addr = overlay->flip_addr; struct drm_i915_gem_request *request; + int ret; BUG_ON(!overlay->active); @@ -404,7 +417,11 @@ static int intel_overlay_off(struct intel_overlay *overlay, * of the hw. Do it in both cases */ flip_addr |= OFC_UPDATE; - BEGIN_LP_RING(6); + ret = BEGIN_LP_RING(6); + if (ret) { + kfree(request); + return ret; + } /* wait for overlay to go idle */ OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE); OUT_RING(flip_addr); @@ -467,7 +484,12 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay) if (request == NULL) return -ENOMEM; - BEGIN_LP_RING(2); + ret = BEGIN_LP_RING(2); + if (ret) { + kfree(request); + return ret; + } + OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP); OUT_RING(MI_NOOP); ADVANCE_LP_RING(); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index d6eba661105f..6fe42c1f4ea9 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -112,10 +112,11 @@ render_ring_flush(struct intel_ring_buffer *ring, #if WATCH_EXEC DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd); #endif - intel_ring_begin(ring, 2); - intel_ring_emit(ring, cmd); - intel_ring_emit(ring, MI_NOOP); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 2) == 0) { + intel_ring_emit(ring, cmd); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + } } } @@ -244,16 +245,17 @@ render_ring_add_request(struct intel_ring_buffer *ring, seqno = i915_gem_get_seqno(dev); if (IS_GEN6(dev)) { - intel_ring_begin(ring, 6); - intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | 3); - intel_ring_emit(ring, PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_IS_FLUSH | - PIPE_CONTROL_NOTIFY); - intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - intel_ring_emit(ring, seqno); - intel_ring_emit(ring, 0); - intel_ring_emit(ring, 0); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 6) == 0) { + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | 3); + intel_ring_emit(ring, PIPE_CONTROL_QW_WRITE | + PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_IS_FLUSH | + PIPE_CONTROL_NOTIFY); + intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, seqno); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + } } else if (HAS_PIPE_CONTROL(dev)) { u32 scratch_addr = dev_priv->seqno_gfx_addr + 128; @@ -262,38 +264,40 @@ render_ring_add_request(struct intel_ring_buffer *ring, * PIPE_NOTIFY buffers out to memory before requesting * an interrupt. */ - intel_ring_begin(ring, 32); - intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH); - intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - intel_ring_emit(ring, seqno); - intel_ring_emit(ring, 0); - PIPE_CONTROL_FLUSH(ring, scratch_addr); - scratch_addr += 128; /* write to separate cachelines */ - PIPE_CONTROL_FLUSH(ring, scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(ring, scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(ring, scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(ring, scratch_addr); - scratch_addr += 128; - PIPE_CONTROL_FLUSH(ring, scratch_addr); - intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH | - PIPE_CONTROL_NOTIFY); - intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); - intel_ring_emit(ring, seqno); - intel_ring_emit(ring, 0); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 32) == 0) { + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | + PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH); + intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, seqno); + intel_ring_emit(ring, 0); + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 128; /* write to separate cachelines */ + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 128; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 128; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 128; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + scratch_addr += 128; + PIPE_CONTROL_FLUSH(ring, scratch_addr); + intel_ring_emit(ring, GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | + PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH | + PIPE_CONTROL_NOTIFY); + intel_ring_emit(ring, dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, seqno); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + } } else { - intel_ring_begin(ring, 4); - 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, seqno); + if (intel_ring_begin(ring, 4) == 0) { + 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, seqno); - intel_ring_emit(ring, MI_USER_INTERRUPT); - intel_ring_advance(ring); + intel_ring_emit(ring, MI_USER_INTERRUPT); + intel_ring_advance(ring); + } } return seqno; } @@ -359,10 +363,11 @@ bsd_ring_flush(struct intel_ring_buffer *ring, u32 invalidate_domains, u32 flush_domains) { - intel_ring_begin(ring, 2); - intel_ring_emit(ring, MI_FLUSH); - intel_ring_emit(ring, MI_NOOP); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 2) == 0) { + intel_ring_emit(ring, MI_FLUSH); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + } } static u32 @@ -373,12 +378,13 @@ ring_add_request(struct intel_ring_buffer *ring, seqno = i915_gem_get_seqno(ring->dev); - intel_ring_begin(ring, 4); - 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, seqno); - intel_ring_emit(ring, MI_USER_INTERRUPT); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 4) == 0) { + 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, seqno); + intel_ring_emit(ring, MI_USER_INTERRUPT); + intel_ring_advance(ring); + } DRM_DEBUG_DRIVER("%s %d\n", ring->name, seqno); @@ -409,10 +415,14 @@ ring_dispatch_execbuffer(struct intel_ring_buffer *ring, uint64_t exec_offset) { uint32_t exec_start; + int ret; exec_start = (uint32_t) exec_offset + exec->batch_start_offset; - intel_ring_begin(ring, 2); + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + intel_ring_emit(ring, MI_BATCH_BUFFER_START | (2 << 6) | @@ -432,8 +442,8 @@ render_ring_dispatch_execbuffer(struct intel_ring_buffer *ring, struct drm_device *dev = ring->dev; drm_i915_private_t *dev_priv = dev->dev_private; int nbox = exec->num_cliprects; - int i = 0, count; uint32_t exec_start, exec_len; + int i, count, ret; exec_start = (uint32_t) exec_offset + exec->batch_start_offset; exec_len = (uint32_t) exec->batch_len; @@ -441,23 +451,28 @@ render_ring_dispatch_execbuffer(struct intel_ring_buffer *ring, 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); + ret = i915_emit_box(dev, cliprects, i, + exec->DR1, exec->DR4); if (ret) return ret; } if (IS_I830(dev) || IS_845G(dev)) { - intel_ring_begin(ring, 4); + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + intel_ring_emit(ring, MI_BATCH_BUFFER); intel_ring_emit(ring, exec_start | MI_BATCH_NON_SECURE); intel_ring_emit(ring, exec_start + exec_len - 4); intel_ring_emit(ring, 0); } else { - intel_ring_begin(ring, 2); + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + if (INTEL_INFO(dev)->gen >= 4) { intel_ring_emit(ring, MI_BATCH_BUFFER_START | (2 << 6) @@ -474,12 +489,13 @@ render_ring_dispatch_execbuffer(struct intel_ring_buffer *ring, } if (IS_G4X(dev) || IS_GEN5(dev)) { - intel_ring_begin(ring, 2); - intel_ring_emit(ring, MI_FLUSH | - MI_NO_WRITE_FLUSH | - MI_INVALIDATE_ISP ); - intel_ring_emit(ring, MI_NOOP); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 2) == 0) { + intel_ring_emit(ring, MI_FLUSH | + MI_NO_WRITE_FLUSH | + MI_INVALIDATE_ISP ); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + } } /* XXX breadcrumb */ @@ -693,18 +709,26 @@ int intel_wait_ring_buffer(struct intel_ring_buffer *ring, int n) return -EBUSY; } -void intel_ring_begin(struct intel_ring_buffer *ring, - int num_dwords) +int intel_ring_begin(struct intel_ring_buffer *ring, + int num_dwords) { int n = 4*num_dwords; + int ret; - if (unlikely(ring->tail + n > ring->size)) - intel_wrap_ring_buffer(ring); + if (unlikely(ring->tail + n > ring->size)) { + ret = intel_wrap_ring_buffer(ring); + if (unlikely(ret)) + return ret; + } - if (unlikely(ring->space < n)) - intel_wait_ring_buffer(ring, n); + if (unlikely(ring->space < n)) { + ret = intel_wait_ring_buffer(ring, n); + if (unlikely(ret)) + return ret; + } ring->space -= n; + return 0; } void intel_ring_advance(struct intel_ring_buffer *ring) @@ -772,12 +796,13 @@ static void gen6_ring_flush(struct intel_ring_buffer *ring, u32 invalidate_domains, u32 flush_domains) { - intel_ring_begin(ring, 4); - intel_ring_emit(ring, MI_FLUSH_DW); - intel_ring_emit(ring, 0); - intel_ring_emit(ring, 0); - intel_ring_emit(ring, 0); - intel_ring_advance(ring); + if (intel_ring_begin(ring, 4) == 0) { + intel_ring_emit(ring, MI_FLUSH_DW); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_emit(ring, 0); + intel_ring_advance(ring); + } } static int @@ -787,10 +812,14 @@ gen6_ring_dispatch_execbuffer(struct intel_ring_buffer *ring, uint64_t exec_offset) { uint32_t exec_start; + int ret; exec_start = (uint32_t) exec_offset + exec->batch_start_offset; - intel_ring_begin(ring, 2); + ret = intel_ring_begin(ring, 2); + if (ret) + return ret; + intel_ring_emit(ring, MI_BATCH_BUFFER_START | MI_BATCH_NON_SECURE_I965); /* bit0-7 is the length on GEN6+ */ intel_ring_emit(ring, exec_start); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index ba4a393e6d16..35ece2b87b02 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -100,8 +100,8 @@ intel_read_status_page(struct intel_ring_buffer *ring, } void intel_cleanup_ring_buffer(struct intel_ring_buffer *ring); -int intel_wait_ring_buffer(struct intel_ring_buffer *ring, int n); -void intel_ring_begin(struct intel_ring_buffer *ring, int n); +int __must_check intel_wait_ring_buffer(struct intel_ring_buffer *ring, int n); +int __must_check intel_ring_begin(struct intel_ring_buffer *ring, int n); static inline void intel_ring_emit(struct intel_ring_buffer *ring, u32 data)