From 0f3b6849dd55943d915f4b461d7db5e8308d4083 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 15 Jan 2013 12:05:55 +0000 Subject: drm/i915: Record DERRMR, FORCEWAKE and RING_CTL in error-state These are useful for investigating hangs involving WAIT_FOR_EVENT. Signed-off-by: Chris Wilson [danvet: Apply a droplet of Future-Proof in the if-ladder.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index e6a11ca..7944d3015 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -641,6 +641,7 @@ static void i915_ring_error_state(struct seq_file *m, seq_printf(m, "%s command stream:\n", ring_str(ring)); seq_printf(m, " HEAD: 0x%08x\n", error->head[ring]); seq_printf(m, " TAIL: 0x%08x\n", error->tail[ring]); + seq_printf(m, " CTL: 0x%08x\n", error->ctl[ring]); seq_printf(m, " ACTHD: 0x%08x\n", error->acthd[ring]); seq_printf(m, " IPEIR: 0x%08x\n", error->ipeir[ring]); seq_printf(m, " IPEHR: 0x%08x\n", error->ipehr[ring]); @@ -693,6 +694,8 @@ static int i915_error_state(struct seq_file *m, void *unused) seq_printf(m, "EIR: 0x%08x\n", error->eir); seq_printf(m, "IER: 0x%08x\n", error->ier); seq_printf(m, "PGTBL_ER: 0x%08x\n", error->pgtbl_er); + seq_printf(m, "FORCEWAKE: 0x%08x\n", error->forcewake); + seq_printf(m, "DERRMR: 0x%08x\n", error->derrmr); seq_printf(m, "CCID: 0x%08x\n", error->ccid); for (i = 0; i < dev_priv->num_fence_regs; i++) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index ed30595..12ab3bd 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -188,10 +188,13 @@ struct drm_i915_error_state { u32 pgtbl_er; u32 ier; u32 ccid; + u32 derrmr; + u32 forcewake; bool waiting[I915_NUM_RINGS]; u32 pipestat[I915_MAX_PIPES]; u32 tail[I915_NUM_RINGS]; u32 head[I915_NUM_RINGS]; + u32 ctl[I915_NUM_RINGS]; u32 ipeir[I915_NUM_RINGS]; u32 ipehr[I915_NUM_RINGS]; u32 instdone[I915_NUM_RINGS]; diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 2220dec..fe84338 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1157,6 +1157,7 @@ static void i915_record_ring_state(struct drm_device *dev, error->acthd[ring->id] = intel_ring_get_active_head(ring); error->head[ring->id] = I915_READ_HEAD(ring); error->tail[ring->id] = I915_READ_TAIL(ring); + error->ctl[ring->id] = I915_READ_CTL(ring); error->cpu_ring_head[ring->id] = ring->head; error->cpu_ring_tail[ring->id] = ring->tail; @@ -1251,6 +1252,16 @@ static void i915_capture_error_state(struct drm_device *dev) else error->ier = I915_READ(IER); + if (INTEL_INFO(dev)->gen >= 6) + error->derrmr = I915_READ(DERRMR); + + if (IS_VALLEYVIEW(dev)) + error->forcewake = I915_READ(FORCEWAKE_VLV); + else if (INTEL_INFO(dev)->gen >= 7) + error->forcewake = I915_READ(FORCEWAKE_MT); + else if (INTEL_INFO(dev)->gen == 6) + error->forcewake = I915_READ(FORCEWAKE); + for_each_pipe(pipe) error->pipestat[pipe] = I915_READ(PIPESTAT(pipe)); diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 186ee5c..b401788 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -512,6 +512,8 @@ #define GEN7_ERR_INT 0x44040 #define ERR_INT_MMIO_UNCLAIMED (1<<13) +#define DERRMR 0x44050 + /* GM45+ chicken bits -- debug workaround bits that may be required * for various sorts of correct behavior. The top 16 bits of each are * the enables for writing to the corresponding low bit. -- cgit v0.10.2 From 19fc42ed9950d5fe17089c0a928121047c882092 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 14 Jan 2013 11:04:39 -0500 Subject: drm/radeon: clear reset flags if engines are idle Fixes a hard lock in the gpu reset code after the rework for DMA support (0ecebb9e0d14e9948e0b1529883a776758117d6f "drm/radeon: switch to a finer grained reset for evergreen") due to not bailing before the MC shutdown if the relevant engines are idle. Discussion: http://lists.freedesktop.org/archives/dri-devel/2013-January/032985.html Reported-by: Eldad Zack Tested-by: Eldad Zack Acked-by: Paul Menzel Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 061fa0a..4d0e60a 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2401,6 +2401,12 @@ static int evergreen_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) { struct evergreen_mc_save save; + if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) + reset_mask &= ~(RADEON_RESET_GFX | RADEON_RESET_COMPUTE); + + if (RREG32(DMA_STATUS_REG) & DMA_IDLE) + reset_mask &= ~RADEON_RESET_DMA; + if (reset_mask == 0) return 0; diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 896f1cb..59acabb 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1409,6 +1409,12 @@ static int cayman_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) { struct evergreen_mc_save save; + if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) + reset_mask &= ~(RADEON_RESET_GFX | RADEON_RESET_COMPUTE); + + if (RREG32(DMA_STATUS_REG) & DMA_IDLE) + reset_mask &= ~RADEON_RESET_DMA; + if (reset_mask == 0) return 0; diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 537e259..3cb9d60 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -1378,6 +1378,12 @@ static int r600_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) { struct rv515_mc_save save; + if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) + reset_mask &= ~(RADEON_RESET_GFX | RADEON_RESET_COMPUTE); + + if (RREG32(DMA_STATUS_REG) & DMA_IDLE) + reset_mask &= ~RADEON_RESET_DMA; + if (reset_mask == 0) return 0; diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 3240a3d..ae8b482 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -2215,6 +2215,12 @@ static int si_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) { struct evergreen_mc_save save; + if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) + reset_mask &= ~(RADEON_RESET_GFX | RADEON_RESET_COMPUTE); + + if (RREG32(DMA_STATUS_REG) & DMA_IDLE) + reset_mask &= ~RADEON_RESET_DMA; + if (reset_mask == 0) return 0; -- cgit v0.10.2 From c18b1170493de637eea58671da7df5689ce13491 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Marek=20Ol=C5=A1=C3=A1k?= Date: Sat, 12 Jan 2013 04:19:37 +0100 Subject: drm/radeon: allow FP16 color clear registers on r500 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Probably not a candidate for stable kernels because of conflicts in DRM versioning. Signed-off-by: Marek Olšák Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index dff6cf7..d9bf96e 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -69,9 +69,10 @@ * 2.26.0 - r600-eg: fix htile size computation * 2.27.0 - r600-SI: Add CS ioctl support for async DMA * 2.28.0 - r600-eg: Add MEM_WRITE packet support + * 2.29.0 - R500 FP16 color clear registers */ #define KMS_DRIVER_MAJOR 2 -#define KMS_DRIVER_MINOR 28 +#define KMS_DRIVER_MINOR 29 #define KMS_DRIVER_PATCHLEVEL 0 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags); int radeon_driver_unload_kms(struct drm_device *dev); diff --git a/drivers/gpu/drm/radeon/reg_srcs/rv515 b/drivers/gpu/drm/radeon/reg_srcs/rv515 index 911a8fb..78d5e99 100644 --- a/drivers/gpu/drm/radeon/reg_srcs/rv515 +++ b/drivers/gpu/drm/radeon/reg_srcs/rv515 @@ -324,6 +324,8 @@ rv515 0x6d40 0x46AC US_OUT_FMT_2 0x46B0 US_OUT_FMT_3 0x46B4 US_W_FMT +0x46C0 RB3D_COLOR_CLEAR_VALUE_AR +0x46C4 RB3D_COLOR_CLEAR_VALUE_GB 0x4BC0 FG_FOG_BLEND 0x4BC4 FG_FOG_FACTOR 0x4BC8 FG_FOG_COLOR_R -- cgit v0.10.2 From 5f0839c11e17d8d942ab02626752f95e30a02a4c Mon Sep 17 00:00:00 2001 From: Jerome Glisse Date: Fri, 11 Jan 2013 15:19:43 -0500 Subject: drm/radeon: improve semaphore debugging on lockup Signed-off-by: Jerome Glisse Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 34e5230..768e8fe 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -654,6 +654,8 @@ struct radeon_ring { u32 ptr_reg_mask; u32 nop; u32 idx; + u64 last_semaphore_signal_addr; + u64 last_semaphore_wait_addr; }; /* diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 141f2b6..2430d80 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -784,6 +784,8 @@ static int radeon_debugfs_ring_info(struct seq_file *m, void *data) } seq_printf(m, "driver's copy of the wptr: 0x%08x [%5d]\n", ring->wptr, ring->wptr); seq_printf(m, "driver's copy of the rptr: 0x%08x [%5d]\n", ring->rptr, ring->rptr); + seq_printf(m, "last semaphore signal addr : 0x%016llx\n", ring->last_semaphore_signal_addr); + seq_printf(m, "last semaphore wait addr : 0x%016llx\n", ring->last_semaphore_wait_addr); seq_printf(m, "%u free dwords in ring\n", ring->ring_free_dw); seq_printf(m, "%u dwords in ring\n", count); /* print 8 dw before current rptr as often it's the last executed diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c index 97f3ece..8dcc20f 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -95,6 +95,10 @@ int radeon_semaphore_sync_rings(struct radeon_device *rdev, /* we assume caller has already allocated space on waiters ring */ radeon_semaphore_emit_wait(rdev, waiter, semaphore); + /* for debugging lockup only, used by sysfs debug files */ + rdev->ring[signaler].last_semaphore_signal_addr = semaphore->gpu_addr; + rdev->ring[waiter].last_semaphore_wait_addr = semaphore->gpu_addr; + return 0; } -- cgit v0.10.2 From f30d26e468322b50d5e376bec40658683aff8ece Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 16 Jan 2013 10:53:40 +0200 Subject: drm/i915/eDP: do not write power sequence registers for ghost eDP Some machines detect an eDP port even if it's not really there, and eDP initialization has a fail path for this. Typically such machines have an LVDS display instead. A regression introduced in commit 82ed61fa1a4e08d5f9e86fb1b715b50ed678b6ac Author: Daniel Vetter Date: Sat Oct 20 20:57:41 2012 +0200 drm/i915: make edp panel power sequence setup more robust updated the power sequence registers PCH_PP_ON_DELAYS, PCH_PP_OFF_DELAYS, and PCH_PP_DIVISOR also in the ghost eDP case, messing up the LVDS display. Split the power sequencer initialization into two, delaying the register updates until after we know the eDP is real. Note: Keep the PP_CONTROL unlocking in the first part, even if it does not update registers, per the commit message of the above mentioned commit. Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=52601 Reported-and-tested-by: Ryan Coe Signed-off-by: Jani Nikula Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 1b63d55..fb3715b 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2579,7 +2579,8 @@ intel_dp_add_properties(struct intel_dp *intel_dp, struct drm_connector *connect static void intel_dp_init_panel_power_sequencer(struct drm_device *dev, - struct intel_dp *intel_dp) + struct intel_dp *intel_dp, + struct edp_power_seq *out) { struct drm_i915_private *dev_priv = dev->dev_private; struct edp_power_seq cur, vbt, spec, final; @@ -2650,16 +2651,35 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, intel_dp->panel_power_cycle_delay = get_delay(t11_t12); #undef get_delay + DRM_DEBUG_KMS("panel power up delay %d, power down delay %d, power cycle delay %d\n", + intel_dp->panel_power_up_delay, intel_dp->panel_power_down_delay, + intel_dp->panel_power_cycle_delay); + + DRM_DEBUG_KMS("backlight on delay %d, off delay %d\n", + intel_dp->backlight_on_delay, intel_dp->backlight_off_delay); + + if (out) + *out = final; +} + +static void +intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, + struct intel_dp *intel_dp, + struct edp_power_seq *seq) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + u32 pp_on, pp_off, pp_div; + /* And finally store the new values in the power sequencer. */ - pp_on = (final.t1_t3 << PANEL_POWER_UP_DELAY_SHIFT) | - (final.t8 << PANEL_LIGHT_ON_DELAY_SHIFT); - pp_off = (final.t9 << PANEL_LIGHT_OFF_DELAY_SHIFT) | - (final.t10 << PANEL_POWER_DOWN_DELAY_SHIFT); + pp_on = (seq->t1_t3 << PANEL_POWER_UP_DELAY_SHIFT) | + (seq->t8 << PANEL_LIGHT_ON_DELAY_SHIFT); + pp_off = (seq->t9 << PANEL_LIGHT_OFF_DELAY_SHIFT) | + (seq->t10 << PANEL_POWER_DOWN_DELAY_SHIFT); /* Compute the divisor for the pp clock, simply match the Bspec * formula. */ pp_div = ((100 * intel_pch_rawclk(dev))/2 - 1) << PP_REFERENCE_DIVIDER_SHIFT; - pp_div |= (DIV_ROUND_UP(final.t11_t12, 1000) + pp_div |= (DIV_ROUND_UP(seq->t11_t12, 1000) << PANEL_POWER_CYCLE_DELAY_SHIFT); /* Haswell doesn't have any port selection bits for the panel @@ -2675,14 +2695,6 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, I915_WRITE(PCH_PP_OFF_DELAYS, pp_off); I915_WRITE(PCH_PP_DIVISOR, pp_div); - - DRM_DEBUG_KMS("panel power up delay %d, power down delay %d, power cycle delay %d\n", - intel_dp->panel_power_up_delay, intel_dp->panel_power_down_delay, - intel_dp->panel_power_cycle_delay); - - DRM_DEBUG_KMS("backlight on delay %d, off delay %d\n", - intel_dp->backlight_on_delay, intel_dp->backlight_off_delay); - DRM_DEBUG_KMS("panel power sequencer register settings: PP_ON %#x, PP_OFF %#x, PP_DIV %#x\n", I915_READ(PCH_PP_ON_DELAYS), I915_READ(PCH_PP_OFF_DELAYS), @@ -2699,6 +2711,7 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; struct drm_display_mode *fixed_mode = NULL; + struct edp_power_seq power_seq = { 0 }; enum port port = intel_dig_port->port; const char *name = NULL; int type; @@ -2771,7 +2784,7 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, } if (is_edp(intel_dp)) - intel_dp_init_panel_power_sequencer(dev, intel_dp); + intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq); intel_dp_i2c_init(intel_dp, intel_connector, name); @@ -2798,6 +2811,10 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, return; } + /* We now know it's not a ghost, init power sequence regs. */ + intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, + &power_seq); + ironlake_edp_panel_vdd_on(intel_dp); edid = drm_get_edid(connector, &intel_dp->adapter); if (edid) { -- cgit v0.10.2 From 262b6d363fcff16359c93bd58c297f961f6e6273 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 15 Jan 2013 16:17:54 +0000 Subject: drm/i915: Invalidate the relocation presumed_offsets along the slow path In the slow path, we are forced to copy the relocations prior to acquiring the struct mutex in order to handle pagefaults. We forgo copying the new offsets back into the relocation entries in order to prevent a recursive locking bug should we trigger a pagefault whilst holding the mutex for the reservations of the execbuffer. Therefore, we need to reset the presumed_offsets just in case the objects are rebound back into their old locations after relocating for this exexbuffer - if that were to happen we would assume the relocations were valid and leave the actual pointers to the kernels dangling, instant hang. Fixes regression from commit bcf50e2775bbc3101932d8e4ab8c7902aa4163b4 Author: Chris Wilson Date: Sun Nov 21 22:07:12 2010 +0000 drm/i915: Handle pagefaults in execbuffer user relocations Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=55984 Signed-off-by: Chris Wilson Cc: Daniel Vetter Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index d6a994a..26d08bb 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -539,6 +539,8 @@ i915_gem_execbuffer_relocate_slow(struct drm_device *dev, total = 0; for (i = 0; i < count; i++) { struct drm_i915_gem_relocation_entry __user *user_relocs; + u64 invalid_offset = (u64)-1; + int j; user_relocs = (void __user *)(uintptr_t)exec[i].relocs_ptr; @@ -549,6 +551,25 @@ i915_gem_execbuffer_relocate_slow(struct drm_device *dev, goto err; } + /* As we do not update the known relocation offsets after + * relocating (due to the complexities in lock handling), + * we need to mark them as invalid now so that we force the + * relocation processing next time. Just in case the target + * object is evicted and then rebound into its old + * presumed_offset before the next execbuffer - if that + * happened we would make the mistake of assuming that the + * relocations were valid. + */ + for (j = 0; j < exec[i].relocation_count; j++) { + if (copy_to_user(&user_relocs[j].presumed_offset, + &invalid_offset, + sizeof(invalid_offset))) { + ret = -EFAULT; + mutex_lock(&dev->struct_mutex); + goto err; + } + } + reloc_offset[i] = total; total += exec[i].relocation_count; } -- cgit v0.10.2 From b514407547890686572606c9dfa4b7f832db9958 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Thu, 17 Jan 2013 10:24:09 +0200 Subject: drm/i915: fix FORCEWAKE posting reads We stopped reading FORCEWAKE for posting reads in commit 8dee3eea3ccd3b6c00a8d3a08dd715d6adf737dd Author: Ben Widawsky Date: Sat Sep 1 22:59:50 2012 -0700 drm/i915: Never read FORCEWAKE and started using something from the same cacheline instead. On the bug reporter's machine this broke entering rc6 states after a suspend/resume cycle. It turns out reading ECOBUS as posting read worked fine, while GTFIFODBG did not, preventing RC6 states after suspend/resume per the bug report referenced below. It's not entirely clear why, but clearly GTFIFODBG was nowhere near the same cacheline or address range as FORCEWAKE. Trying out various registers for posting reads showed that all tested registers for which NEEDS_FORCE_WAKE() (in i915_drv.c) returns true work. Conversely, most (but not quite all) registers for which NEEDS_FORCE_WAKE() returns false do not work. Details in the referenced bug. Based on the above, add posting reads on ECOBUS where GTFIFODBG was previously relied on. In true cargo cult spirit, add posting reads for FORCEWAKE_VLV writes as well, but instead of ECOBUS, use FORCEWAKE_ACK_VLV which is in the same address range as FORCEWAKE_VLV. v2: Add more details to the commit message. No functional changes. Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=52411 Reported-and-tested-by: Alexander Bersenev CC: Ben Widawsky Signed-off-by: Jani Nikula Reviewed-by: Chris Wilson Cc: stable@vger.kernel.org [danvet: add cc: stable and make the commit message a bit clearer that this is a regression fix and what exactly broke.] Signed-off-by: Daniel Vetter diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index e83a117..3280cff 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4250,7 +4250,8 @@ static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv) static void __gen6_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_DISABLE(0xffff)); - POSTING_READ(ECOBUS); /* something from same cacheline, but !FORCEWAKE */ + /* something from same cacheline, but !FORCEWAKE_MT */ + POSTING_READ(ECOBUS); } static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv) @@ -4267,7 +4268,8 @@ static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv) DRM_ERROR("Timed out waiting for forcewake old ack to clear.\n"); I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL)); - POSTING_READ(ECOBUS); /* something from same cacheline, but !FORCEWAKE */ + /* something from same cacheline, but !FORCEWAKE_MT */ + POSTING_READ(ECOBUS); if (wait_for_atomic((I915_READ_NOTRACE(forcewake_ack) & 1), FORCEWAKE_ACK_TIMEOUT_MS)) @@ -4304,14 +4306,16 @@ void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv) static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE, 0); - /* gen6_gt_check_fifodbg doubles as the POSTING_READ */ + /* something from same cacheline, but !FORCEWAKE */ + POSTING_READ(ECOBUS); gen6_gt_check_fifodbg(dev_priv); } static void __gen6_gt_force_wake_mt_put(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL)); - /* gen6_gt_check_fifodbg doubles as the POSTING_READ */ + /* something from same cacheline, but !FORCEWAKE_MT */ + POSTING_READ(ECOBUS); gen6_gt_check_fifodbg(dev_priv); } @@ -4351,6 +4355,8 @@ int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv) static void vlv_force_wake_reset(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE_VLV, _MASKED_BIT_DISABLE(0xffff)); + /* something from same cacheline, but !FORCEWAKE_VLV */ + POSTING_READ(FORCEWAKE_ACK_VLV); } static void vlv_force_wake_get(struct drm_i915_private *dev_priv) @@ -4371,7 +4377,8 @@ static void vlv_force_wake_get(struct drm_i915_private *dev_priv) static void vlv_force_wake_put(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE_VLV, _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL)); - /* The below doubles as a POSTING_READ */ + /* something from same cacheline, but !FORCEWAKE_VLV */ + POSTING_READ(FORCEWAKE_ACK_VLV); gen6_gt_check_fifodbg(dev_priv); } -- cgit v0.10.2 From 20707874fd4fd37e09513f508e642fa8bd06365a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 17 Jan 2013 13:10:50 -0500 Subject: Revert "drm/radeon: do not move bo to different placement at each cs" This reverts commit d025e9e2b890db679f1246037bf65bd4be512627. This causes corruption for a number of users and needs further investigation in the next cycle. https://bugzilla.kernel.org/show_bug.cgi?id=52491 https://bugs.freedesktop.org/show_bug.cgi?id=58659 http://lists.freedesktop.org/archives/dri-devel/2013-January/032961.html Signed-off-by: Alex Deucher diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 768e8fe..a08f657 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -324,7 +324,6 @@ struct radeon_bo { struct list_head list; /* Protected by tbo.reserved */ u32 placements[3]; - u32 busy_placements[3]; struct ttm_placement placement; struct ttm_buffer_object tbo; struct ttm_bo_kmap_obj kmap; diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 883c95d..d3aface 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -84,6 +84,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) rbo->placement.fpfn = 0; rbo->placement.lpfn = 0; rbo->placement.placement = rbo->placements; + rbo->placement.busy_placement = rbo->placements; if (domain & RADEON_GEM_DOMAIN_VRAM) rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM; @@ -104,14 +105,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) if (!c) rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; rbo->placement.num_placement = c; - - c = 0; - rbo->placement.busy_placement = rbo->busy_placements; - if (rbo->rdev->flags & RADEON_IS_AGP) { - rbo->busy_placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT; - } else { - rbo->busy_placements[c++] = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; - } rbo->placement.num_busy_placement = c; } @@ -357,6 +350,7 @@ int radeon_bo_list_validate(struct list_head *head) { struct radeon_bo_list *lobj; struct radeon_bo *bo; + u32 domain; int r; r = ttm_eu_reserve_buffers(head); @@ -366,9 +360,17 @@ int radeon_bo_list_validate(struct list_head *head) list_for_each_entry(lobj, head, tv.head) { bo = lobj->bo; if (!bo->pin_count) { + domain = lobj->wdomain ? lobj->wdomain : lobj->rdomain; + + retry: + radeon_ttm_placement_from_domain(bo, domain); r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); if (unlikely(r)) { + if (r != -ERESTARTSYS && domain == RADEON_GEM_DOMAIN_VRAM) { + domain |= RADEON_GEM_DOMAIN_GTT; + goto retry; + } return r; } } -- cgit v0.10.2 From 630541863b29f88c7ab34e647758344e4cd1eafd Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 16 Jan 2013 14:25:44 +1000 Subject: ttm: don't destroy old mm_node on memcpy failure When we are using memcpy to move objects around, and we fail to memcpy due to lack of memory to populate or failure to finish the copy, we don't want to destroy the mm_node that has been copied into old_copy. While working on a new kms driver that uses memcpy, if I overallocated bo's up to the memory limits, and eviction failed, then machine would oops soon after due to having an active bo with an already freed drm_mm embedded in it, freeing it a second time didn't end well. Reviewed-by: Jerome Glisse Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index d73d6e3..44420fc 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -344,8 +344,12 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo, if (ttm->state == tt_unpopulated) { ret = ttm->bdev->driver->ttm_tt_populate(ttm); - if (ret) + if (ret) { + /* if we fail here don't nuke the mm node + * as the bo still owns it */ + old_copy.mm_node = NULL; goto out1; + } } add = 0; @@ -371,8 +375,11 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo, prot); } else ret = ttm_copy_io_page(new_iomap, old_iomap, page); - if (ret) + if (ret) { + /* failing here, means keep old copy as-is */ + old_copy.mm_node = NULL; goto out1; + } } mb(); out2: -- cgit v0.10.2 From 014b34409fb2015f63663b6cafdf557fdf289628 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 16 Jan 2013 15:58:34 +1000 Subject: ttm: on move memory failure don't leave a node dangling if we have a move notify callback, when moving fails, we call move notify the opposite way around, however this ends up with *mem containing the mm_node from the bo, which means we double free it. This is a follow on to the previous fix. Reviewed-by: Jerome Glisse Signed-off-by: Dave Airlie diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index 33d20be..52b20b1 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -434,6 +434,7 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, bo->mem = tmp_mem; bdev->driver->move_notify(bo, mem); bo->mem = *mem; + *mem = tmp_mem; } goto out_err; -- cgit v0.10.2