From 8e62438a1ee74ceeac77bb4c680ceaaf3f860488 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Sat, 20 Feb 2021 12:33:03 +0200 Subject: [PATCH 1/2] drm/i915: Workaround async flip + VT-d corruption on HSW/BDW MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit On HSW/BDW with VT-d active the first tile row scanned out after the first async flip of the frame often ends up corrupted. Whether the corruption happens or not depends on the scanline on which the async flip happens, but the behaviour seems very consistent. Ie. the same set of scanlines (which are most scanlines) always show the corruption. And another set of scanlines (far less of them) never shows the corruption. I discovered that disabling the fetch-stride stretching feature cures the corruption. This is some kind of TLB related prefetch thing AFAIK. We already disable it on SNB primary planes due to a documented workaround. The hardware folks indicated that disabling this should be fine, so let's go with that. And while we're here, let's document the relevant bits on all pre-skl platforms. Fixes: 2a636e240c77 ("drm/i915: Implement async flip for ivb/hsw") Fixes: cda195f13abd ("drm/i915: Implement async flips for bdw") Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20210220103303.3448-1-ville.syrjala@linux.intel.com Reviewed-by: Karthik B S (cherry picked from commit b7a7053ab2ec558b8ae4e55f62ea8f1f58e14f5c) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/i915_reg.h | 23 ++++++++++++++++++++++- drivers/gpu/drm/i915/intel_pm.c | 16 +++++++++++++++- 2 files changed, 37 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 7146cd0f3256..aaf1f0045b16 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3316,7 +3316,18 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define ILK_DISPLAY_CHICKEN1 _MMIO(0x42000) #define ILK_FBCQ_DIS (1 << 22) -#define ILK_PABSTRETCH_DIS (1 << 21) +#define ILK_PABSTRETCH_DIS REG_BIT(21) +#define ILK_SABSTRETCH_DIS REG_BIT(20) +#define IVB_PRI_STRETCH_MAX_MASK REG_GENMASK(21, 20) +#define IVB_PRI_STRETCH_MAX_X8 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 0) +#define IVB_PRI_STRETCH_MAX_X4 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 1) +#define IVB_PRI_STRETCH_MAX_X2 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 2) +#define IVB_PRI_STRETCH_MAX_X1 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 3) +#define IVB_SPR_STRETCH_MAX_MASK REG_GENMASK(19, 18) +#define IVB_SPR_STRETCH_MAX_X8 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 0) +#define IVB_SPR_STRETCH_MAX_X4 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 1) +#define IVB_SPR_STRETCH_MAX_X2 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 2) +#define IVB_SPR_STRETCH_MAX_X1 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 3) /* @@ -8039,6 +8050,16 @@ enum { #define _CHICKEN_PIPESL_1_A 0x420b0 #define _CHICKEN_PIPESL_1_B 0x420b4 +#define HSW_PRI_STRETCH_MAX_MASK REG_GENMASK(28, 27) +#define HSW_PRI_STRETCH_MAX_X8 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 0) +#define HSW_PRI_STRETCH_MAX_X4 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 1) +#define HSW_PRI_STRETCH_MAX_X2 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 2) +#define HSW_PRI_STRETCH_MAX_X1 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 3) +#define HSW_SPR_STRETCH_MAX_MASK REG_GENMASK(26, 25) +#define HSW_SPR_STRETCH_MAX_X8 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 0) +#define HSW_SPR_STRETCH_MAX_X4 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 1) +#define HSW_SPR_STRETCH_MAX_X2 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 2) +#define HSW_SPR_STRETCH_MAX_X1 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 3) #define HSW_FBCQ_DIS (1 << 22) #define BDW_DPRS_MASK_VBLANK_SRD (1 << 0) #define CHICKEN_PIPESL_1(pipe) _MMIO_PIPE(pipe, _CHICKEN_PIPESL_1_A, _CHICKEN_PIPESL_1_B) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 0c3e63f27c29..97b57acc02e2 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7245,11 +7245,16 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv) intel_uncore_write(&dev_priv->uncore, CHICKEN_PAR1_1, intel_uncore_read(&dev_priv->uncore, CHICKEN_PAR1_1) | DPA_MASK_VBLANK_SRD); - /* WaPsrDPRSUnmaskVBlankInSRD:bdw */ for_each_pipe(dev_priv, pipe) { + /* WaPsrDPRSUnmaskVBlankInSRD:bdw */ intel_uncore_write(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe), intel_uncore_read(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe)) | BDW_DPRS_MASK_VBLANK_SRD); + + /* Undocumented but fixes async flip + VT-d corruption */ + if (intel_vtd_active()) + intel_uncore_rmw(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe), + HSW_PRI_STRETCH_MAX_MASK, HSW_PRI_STRETCH_MAX_X1); } /* WaVSRefCountFullforceMissDisable:bdw */ @@ -7285,11 +7290,20 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv) static void hsw_init_clock_gating(struct drm_i915_private *dev_priv) { + enum pipe pipe; + /* WaFbcAsynchFlipDisableFbcQueue:hsw,bdw */ intel_uncore_write(&dev_priv->uncore, CHICKEN_PIPESL_1(PIPE_A), intel_uncore_read(&dev_priv->uncore, CHICKEN_PIPESL_1(PIPE_A)) | HSW_FBCQ_DIS); + for_each_pipe(dev_priv, pipe) { + /* Undocumented but fixes async flip + VT-d corruption */ + if (intel_vtd_active()) + intel_uncore_rmw(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe), + HSW_PRI_STRETCH_MAX_MASK, HSW_PRI_STRETCH_MAX_X1); + } + /* This is required by WaCatErrorRejectionIssue:hsw */ intel_uncore_write(&dev_priv->uncore, GEN7_SQ_CHICKEN_MBCUNIT_CONFIG, intel_uncore_read(&dev_priv->uncore, GEN7_SQ_CHICKEN_MBCUNIT_CONFIG) | From 6a77c6bb7260bd5000f95df454d9f8cdb1af7132 Mon Sep 17 00:00:00 2001 From: Umesh Nerlige Ramappa Date: Fri, 5 Mar 2021 13:09:47 -0800 Subject: [PATCH 2/2] i915/perf: Start hrtimer only if sampling the OA buffer SAMPLE_OA parameter enables sampling of OA buffer and results in a call to init the OA buffer which initializes the OA unit head/tail pointers. The OA_EXPONENT parameter controls the periodicity of the OA reports in the OA buffer and results in starting a hrtimer. Before gen12, all use cases required the use of the OA buffer and i915 enforced this setting when vetting out the parameters passed. In these platforms the hrtimer was enabled if OA_EXPONENT was passed. This worked fine since it was implied that SAMPLE_OA is always passed. With gen12, this changed. Users can use perf without enabling the OA buffer as in OAR use cases. While an OAR use case should ideally not start the hrtimer, we see that passing an OA_EXPONENT parameter will start the hrtimer even though SAMPLE_OA is not specified. This results in an uninitialized OA buffer, so the head/tail pointers used to track the buffer are zero. This itself does not fail, but if we ran a use-case that SAMPLED the OA buffer previously, then the OA_TAIL register is still pointing to an old value. When the timer callback runs, it ends up calculating a wrong/large number of available reports. Since we do a spinlock_irq_save and start processing a large number of reports, NMI watchdog fires and causes a crash. Start the timer only if SAMPLE_OA is specified. v2: - Drop SAMPLE OA check when appending samples (Ashutosh) - Prevent read if OA buffer is not being sampled Fixes: 00a7f0d7155c ("drm/i915/tgl: Add perf support on TGL") Signed-off-by: Umesh Nerlige Ramappa Reviewed-by: Ashutosh Dixit Signed-off-by: Lionel Landwerlin Link: https://patchwork.freedesktop.org/patch/msgid/20210305210947.58751-1-umesh.nerlige.ramappa@intel.com (cherry picked from commit be0bdd67fda9468156c733976688f6487d0c42f7) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/i915_perf.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 112ba5f2ce90..e62ad69606f6 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -603,7 +603,6 @@ static int append_oa_sample(struct i915_perf_stream *stream, { int report_size = stream->oa_buffer.format_size; struct drm_i915_perf_record_header header; - u32 sample_flags = stream->sample_flags; header.type = DRM_I915_PERF_RECORD_SAMPLE; header.pad = 0; @@ -617,10 +616,8 @@ static int append_oa_sample(struct i915_perf_stream *stream, return -EFAULT; buf += sizeof(header); - if (sample_flags & SAMPLE_OA_REPORT) { - if (copy_to_user(buf, report, report_size)) - return -EFAULT; - } + if (copy_to_user(buf, report, report_size)) + return -EFAULT; (*offset) += header.size; @@ -2682,7 +2679,7 @@ static void i915_oa_stream_enable(struct i915_perf_stream *stream) stream->perf->ops.oa_enable(stream); - if (stream->periodic) + if (stream->sample_flags & SAMPLE_OA_REPORT) hrtimer_start(&stream->poll_check_timer, ns_to_ktime(stream->poll_oa_period), HRTIMER_MODE_REL_PINNED); @@ -2745,7 +2742,7 @@ static void i915_oa_stream_disable(struct i915_perf_stream *stream) { stream->perf->ops.oa_disable(stream); - if (stream->periodic) + if (stream->sample_flags & SAMPLE_OA_REPORT) hrtimer_cancel(&stream->poll_check_timer); } @@ -3028,7 +3025,7 @@ static ssize_t i915_perf_read(struct file *file, * disabled stream as an error. In particular it might otherwise lead * to a deadlock for blocking file descriptors... */ - if (!stream->enabled) + if (!stream->enabled || !(stream->sample_flags & SAMPLE_OA_REPORT)) return -EIO; if (!(file->f_flags & O_NONBLOCK)) {