drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
Subsystems (2)
PATCH v3 4/5 drm/i915: Add a fault injection point after WOPCM init
In reply to Jakub Bartmiński--- drivers/gpu/drm/i915/i915_gem.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/gpu/drm/i915/i915_gem.c +5 -0
@@ -5475,6 +5475,11 @@ int i915_gem_init(struct drm_i915_private *dev_priv)if (ret) goto err_uc_misc; +if (i915_inject_load_failure ()) { + ret = -E2BIG ; + goto err_uc_misc; + } +/* This is just a security blanket to placate dragons. * On some systems, we very sporadically observe that the first TLBs * used by the CS may be stale, despite us poking the TLB reset. If
Signature · 6 lines
2.17.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH v3 1/5 drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
It would seem that we are using uninitialized WOPCM variables when setting the GuC pin bias. The pin bias has to be set after the WOPCM, but before the call to i915_gem_contexts_init where the first contexts are pinned so the safest place to set it seems to be right after initializing the relevant variables in intel_wopcm_init.
--- drivers/gpu/drm/i915/intel_guc.c | 21 --------------------- drivers/gpu/drm/i915/intel_wopcm.c | 19 +++++++++++++++++++ 2 files changed, 19 insertions(+), 21 deletions(-)
drivers/gpu/drm/i915/intel_guc.c +0 -21
@@ -27,8 +27,6 @@#include "intel_guc_submission.h" #include "i915_drv.h" -static void guc_init_ggtt_pin_bias (struct intel_guc *guc); -static void gen8_guc_raise_irq (struct intel_guc *guc) {struct drm_i915_private *dev_priv =guc_to_i915 (guc);
@@ -142,8 +140,6 @@ int intel_guc_init_misc(struct intel_guc *guc)struct drm_i915_private *i915 =guc_to_i915 (guc);int ret; -guc_init_ggtt_pin_bias (guc); - ret =guc_init_wq (guc);if (ret)return ret;
@@ -611,23 +607,6 @@ int intel_guc_resume(struct intel_guc *guc)* actual GuC WOPCM size. */ -/** -* guc_init_ggtt_pin_bias() - Initialize the GuC ggtt_pin_bias value. -* @guc: intel_guc structure. -* -* This function will calculate and initialize the ggtt_pin_bias value based on -* overall WOPCM size and GuC WOPCM size. -*/ -static void guc_init_ggtt_pin_bias (struct intel_guc *guc) -{ -struct drm_i915_private *i915 =guc_to_i915 (guc); - -GEM_BUG_ON (!i915->wopcm.size); -GEM_BUG_ON (i915->wopcm.size < i915->wopcm.guc.base); - - guc->ggtt_pin_bias = i915->wopcm.size - i915->wopcm.guc.base; -} -/** * intel_guc_allocate_vma() - Allocate a GGTT VMA for GuC usage * @guc: the guc
drivers/gpu/drm/i915/intel_wopcm.c +19 -0
@@ -140,6 +140,23 @@ static inline int check_hw_restriction(struct drm_i915_private *i915,return err; } +/** +* wopcm_init_guc_ggtt_pin_bias() - Initialize the GuC ggtt_pin_bias value. +* @wopcm: pointer to intel_wopcm. +* +* This function will calculate and initialize the GuC ggtt_pin_bias value based +* on overall WOPCM size and GuC WOPCM size. +*/ +static void wopcm_init_guc_ggtt_pin_bias (struct intel_wopcm *wopcm) +{ +struct drm_i915_private *i915 =wopcm_to_i915 (wopcm); + +GEM_BUG_ON (!wopcm->size); +GEM_BUG_ON (wopcm->size < wopcm->guc.base); + + i915->guc.ggtt_pin_bias = wopcm->size - wopcm->guc.base; +} +/** * intel_wopcm_init() - Initialize the WOPCM structure. * @wopcm: pointer to intel_wopcm.
@@ -207,6 +224,8 @@ int intel_wopcm_init(struct intel_wopcm *wopcm) wopcm->guc.base = guc_wopcm_base; wopcm->guc.size = guc_wopcm_size; +wopcm_init_guc_ggtt_pin_bias (wopcm); +return 0 ; }
Signature · 6 lines
2.17.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH v3 2/5 drm/i915/guc: Move the pin bias value from GuC to GGTT
Removing the pin bias from GuC allows us to not check for GuC every time we pin a context, which fixes the assertion error on unresolved GuC platform default in mock contexts selftest.
With this change the intel_guc_ggtt_offset function has to know the full declaration of the drm_i915_private structure so it can no longer be located in the intel_guc.h header file due to include order.
v2:
This also makes it so that there's no need to set GuC variables from within the WOPCM init function or to move the WOPCM init, while keeping the correct initialization order. Also for mock tests the pin bias is left at 0 and we make sure that the pin bias with GuC will not be smaller than without GuC.
v3:
Avoid unused i915 in intel_guc_ggtt_offset if debug is disabled.
Testcase: igt/drv_selftest/mock_contexts #GuC
--- drivers/gpu/drm/i915/i915_gem_context.c | 10 +------- drivers/gpu/drm/i915/i915_gem_gtt.c | 22 ++++++++++++---- drivers/gpu/drm/i915/i915_gem_gtt.h | 2 ++ drivers/gpu/drm/i915/intel_guc.c | 31 ++++++++++++++++++++--- drivers/gpu/drm/i915/intel_guc.h | 28 +------------------- drivers/gpu/drm/i915/intel_huc.c | 2 +- drivers/gpu/drm/i915/intel_uc_fw.c | 2 +- drivers/gpu/drm/i915/intel_wopcm.c | 19 -------------- drivers/gpu/drm/i915/intel_wopcm.h | 8 ++++++ drivers/gpu/drm/i915/selftests/mock_gtt.c | 2 ++ 10 files changed, 60 insertions(+), 66 deletions(-)
drivers/gpu/drm/i915/i915_gem_context.c +1 -9
@@ -329,15 +329,7 @@ __create_hw_context(struct drm_i915_private *dev_priv, ctx->desc_template =default_desc_template (dev_priv, dev_priv->mm.aliasing_ppgtt); -/* -* GuC requires the ring to be placed in Non-WOPCM memory. If GuC is not -* present or not in use we still need a small bias as ring wraparound -* at offset 0 sometimes hangs. No idea why. -*/ -if (USES_GUC (dev_priv)) - ctx->ggtt_offset_bias = dev_priv->guc.ggtt_pin_bias; -else - ctx->ggtt_offset_bias =I915_GTT_PAGE_SIZE ; + ctx->ggtt_offset_bias = dev_priv->ggtt.pin_bias;return ctx;
drivers/gpu/drm/i915/i915_gem_gtt.c +17 -5
@@ -2901,7 +2901,7 @@ void i915_gem_fini_aliasing_ppgtt(struct drm_i915_private *i915) ggtt->vm.vma_ops.unbind_vma = ggtt_unbind_vma; } -int i915_gem_init_ggtt (struct drm_i915_private *dev_priv) +int i915_gem_init_ggtt (struct drm_i915_private *i915) {/* Let GEM Manage all of the aperture. *
@@ -2912,12 +2912,24 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)* aperture. One page should be enough to keep any prefetching inside * of the aperture. */ -struct i915_ggtt *ggtt = &dev_priv->ggtt; +struct i915_ggtt *ggtt = &i915->ggtt;unsigned long hole_start, hole_end;struct drm_mm_node *entry;int ret; - ret =intel_vgt_balloon (dev_priv); +/* +* We need a small bias as ring wraparound at offset 0 +* sometimes hangs. No idea why. +*/ + ggtt->pin_bias =I915_GTT_PAGE_SIZE ; + +/* Any objects shared with GuC can't overlap with WOPCM. */ +if (USES_GUC (i915)) { + ggtt->pin_bias =max (ggtt->pin_bias, +intel_wopcm_get_pin_bias (&i915->wopcm)); + } + + ret =intel_vgt_balloon (i915);if (ret)return ret;
@@ -2940,8 +2952,8 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)/* And finally clear the reserved guard page */ ggtt->vm.clear_range (&ggtt->vm, ggtt->vm.total -PAGE_SIZE ,PAGE_SIZE ); -if (USES_PPGTT (dev_priv) && !USES_FULL_PPGTT (dev_priv)) { - ret =i915_gem_init_aliasing_ppgtt (dev_priv); +if (USES_PPGTT (i915) && !USES_FULL_PPGTT (i915)) { + ret =i915_gem_init_aliasing_ppgtt (i915);if (ret) goto err; }
drivers/gpu/drm/i915/i915_gem_gtt.h +2 -0
@@ -396,6 +396,8 @@ struct i915_ggtt {int mtrr; +u32 pin_bias; +struct drm_mm_node error_capture; };
drivers/gpu/drm/i915/intel_guc.c +27 -4
@@ -607,6 +607,29 @@ int intel_guc_resume(struct intel_guc *guc)* actual GuC WOPCM size. */ +/** +* intel_guc_ggtt_offset() - Get and validate the GGTT offset of @vma +* @guc: intel_guc structure. +* @vma: i915 graphics virtual memory area. +* +* GuC does not allow any gfx GGTT address that falls into range +* [0, ggtt.pin_bias), which is reserved for Boot ROM, SRAM and WOPCM. +* Currently, in order to exclude [0, ggtt.pin_bias) address space from +* GGTT, all gfx objects used by GuC are allocated with intel_guc_allocate_vma() +* and pinned with PIN_OFFSET_BIAS along with the value of ggtt.pin_bias. +* +* Return: GGTT offset of the @vma. +*/ +u32 intel_guc_ggtt_offset (struct intel_guc *guc,struct i915_vma *vma) +{ +u32 offset =i915_ggtt_offset (vma); + +GEM_BUG_ON (offset <i915_vm_to_ggtt (vma->vm)->pin_bias); +GEM_BUG_ON (range_overflows_t (u64, offset, vma->size,GUC_GGTT_TOP )); + +return offset; +} +/** * intel_guc_allocate_vma() - Allocate a GGTT VMA for GuC usage * @guc: the guc
@@ -622,21 +645,21 @@ int intel_guc_resume(struct intel_guc *guc)*/ struct i915_vma *intel_guc_allocate_vma (struct intel_guc *guc,u32 size) { -struct drm_i915_private *dev_priv =guc_to_i915 (guc); +struct drm_i915_private *i915 =guc_to_i915 (guc);struct drm_i915_gem_object *obj;struct i915_vma *vma;int ret; - obj =i915_gem_object_create (dev_priv, size); + obj =i915_gem_object_create (i915, size);if (IS_ERR (obj))return ERR_CAST (obj); - vma =i915_vma_instance (obj, &dev_priv->ggtt.vm,NULL ); + vma =i915_vma_instance (obj, &i915->ggtt.vm,NULL );if (IS_ERR (vma)) goto err; ret =i915_vma_pin (vma,0 ,PAGE_SIZE , -PIN_GLOBAL |PIN_OFFSET_BIAS | guc->ggtt_pin_bias); +PIN_GLOBAL |PIN_OFFSET_BIAS | i915->ggtt.pin_bias);if (ret) { vma =ERR_PTR (ret); goto err;
drivers/gpu/drm/i915/intel_guc.h +1 -27
@@ -49,9 +49,6 @@ struct intel_guc {struct intel_guc_log log;struct intel_guc_ct ct; -/* Offset where Non-WOPCM memory starts. */ -u32 ggtt_pin_bias; -/* Log snapshot if GuC errors during load */ struct drm_i915_gem_object *load_err_log;
@@ -123,30 +120,7 @@ static inline void intel_guc_to_host_event_handler(struct intel_guc *guc)/* GuC addresses above GUC_GGTT_TOP also don't map through the GTT */ #define GUC_GGTT_TOP 0xFEE00000 - -/** -* intel_guc_ggtt_offset() - Get and validate the GGTT offset of @vma -* @guc: intel_guc structure. -* @vma: i915 graphics virtual memory area. -* -* GuC does not allow any gfx GGTT address that falls into range -* [0, GuC ggtt_pin_bias), which is reserved for Boot ROM, SRAM and WOPCM. -* Currently, in order to exclude [0, GuC ggtt_pin_bias) address space from -* GGTT, all gfx objects used by GuC are allocated with intel_guc_allocate_vma() -* and pinned with PIN_OFFSET_BIAS along with the value of GuC ggtt_pin_bias. -* -* Return: GGTT offset of the @vma. -*/ -static inline u32 intel_guc_ggtt_offset (struct intel_guc *guc, -struct i915_vma *vma) -{ -u32 offset =i915_ggtt_offset (vma); - -GEM_BUG_ON (offset < guc->ggtt_pin_bias); -GEM_BUG_ON (range_overflows_t (u64, offset, vma->size,GUC_GGTT_TOP )); - -return offset; -} +u32 intel_guc_ggtt_offset (struct intel_guc *guc,struct i915_vma *vma);void intel_guc_init_early (struct intel_guc *guc);void intel_guc_init_send_regs (struct intel_guc *guc);
drivers/gpu/drm/i915/intel_huc.c +1 -1
@@ -63,7 +63,7 @@ int intel_huc_auth(struct intel_huc *huc)return -ENOEXEC ; vma =i915_gem_object_ggtt_pin (huc->fw.obj,NULL ,0 ,0 , -PIN_OFFSET_BIAS | guc->ggtt_pin_bias); +PIN_OFFSET_BIAS | i915->ggtt.pin_bias);if (IS_ERR (vma)) { ret =PTR_ERR (vma);DRM_ERROR ("HuC: Failed to pin huc fw object %d\n" , ret);
drivers/gpu/drm/i915/intel_uc_fw.c +1 -1
@@ -222,7 +222,7 @@ int intel_uc_fw_upload(struct intel_uc_fw *uc_fw, goto fail; } - ggtt_pin_bias =to_i915 (uc_fw->obj->base.dev)->guc.ggtt_pin_bias; + ggtt_pin_bias =to_i915 (uc_fw->obj->base.dev)->ggtt.pin_bias; vma =i915_gem_object_ggtt_pin (uc_fw->obj,NULL ,0 ,0 ,PIN_OFFSET_BIAS | ggtt_pin_bias);if (IS_ERR (vma)) {
drivers/gpu/drm/i915/intel_wopcm.c +0 -19
@@ -140,23 +140,6 @@ static inline int check_hw_restriction(struct drm_i915_private *i915,return err; } -/** -* wopcm_init_guc_ggtt_pin_bias() - Initialize the GuC ggtt_pin_bias value. -* @wopcm: pointer to intel_wopcm. -* -* This function will calculate and initialize the GuC ggtt_pin_bias value based -* on overall WOPCM size and GuC WOPCM size. -*/ -static void wopcm_init_guc_ggtt_pin_bias (struct intel_wopcm *wopcm) -{ -struct drm_i915_private *i915 =wopcm_to_i915 (wopcm); - -GEM_BUG_ON (!wopcm->size); -GEM_BUG_ON (wopcm->size < wopcm->guc.base); - - i915->guc.ggtt_pin_bias = wopcm->size - wopcm->guc.base; -} -/** * intel_wopcm_init() - Initialize the WOPCM structure. * @wopcm: pointer to intel_wopcm.
@@ -224,8 +207,6 @@ int intel_wopcm_init(struct intel_wopcm *wopcm) wopcm->guc.base = guc_wopcm_base; wopcm->guc.size = guc_wopcm_size; -wopcm_init_guc_ggtt_pin_bias (wopcm); -return 0 ; }
drivers/gpu/drm/i915/intel_wopcm.h +8 -0
@@ -7,6 +7,7 @@#ifndef _INTEL_WOPCM_H_#define _INTEL_WOPCM_H_ +#include "i915_gem.h" #include <linux/types.h> /**
@@ -24,6 +25,13 @@ struct intel_wopcm { } guc; }; +static inline u32 intel_wopcm_get_pin_bias (struct intel_wopcm *wopcm) +{ +GEM_BUG_ON (!wopcm->size); +GEM_BUG_ON (wopcm->size < wopcm->guc.base); +return wopcm->size - wopcm->guc.base; +} +void intel_wopcm_init_early (struct intel_wopcm *wopcm);int intel_wopcm_init (struct intel_wopcm *wopcm);int intel_wopcm_init_hw (struct intel_wopcm *wopcm);
drivers/gpu/drm/i915/selftests/mock_gtt.c +2 -0
@@ -117,6 +117,8 @@ void mock_init_ggtt(struct drm_i915_private *i915) ggtt->vm.vma_ops.set_pages = ggtt_set_pages; ggtt->vm.vma_ops.clear_pages = clear_pages; + ggtt->pin_bias =0 ; +i915_address_space_init (&ggtt->vm, i915); }
Signature · 6 lines
2.17.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH v3 3/5 drm/i915: Remove unnecessary ggtt_offset_bias from i915_gem_context
In reply to Jakub BartmińskiSince ggtt_offset_bias is now stored in ggtt.pin_bias, it is duplicated inside i915_gem_context, and can instead be accessed directly from ggtt.
v3:
Added a helper function to retrieve the ggtt.pin_bias from the vma.
--- drivers/gpu/drm/i915/i915_gem_context.c | 2 -- drivers/gpu/drm/i915/i915_gem_context.h | 3 --- drivers/gpu/drm/i915/intel_lrc.c | 14 +++++++++----- 3 files changed, 9 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/i915_gem_context.c +0 -2
@@ -329,8 +329,6 @@ __create_hw_context(struct drm_i915_private *dev_priv, ctx->desc_template =default_desc_template (dev_priv, dev_priv->mm.aliasing_ppgtt); - ctx->ggtt_offset_bias = dev_priv->ggtt.pin_bias; -return ctx; err_pid:
drivers/gpu/drm/i915/i915_gem_context.h +0 -3
@@ -147,9 +147,6 @@ struct i915_gem_context {struct i915_sched_attr sched; -/** ggtt_offset_bias: placement restriction for context objects */ -u32 ggtt_offset_bias; -/** engine: per-engine logical HW state */ struct intel_context {struct i915_gem_context *gem_context;
drivers/gpu/drm/i915/intel_lrc.c +9 -5
@@ -1346,6 +1346,11 @@ static void execlists_context_unpin(struct intel_context *ce)i915_gem_context_put (ce->gem_context); } +static inline u32 i915_ggtt_pin_bias (struct i915_vma *vma) +{ +return i915_vm_to_ggtt (vma->vm)->pin_bias; +} +static int __context_pin (struct i915_gem_context *ctx,struct i915_vma *vma) {unsigned int flags;
@@ -1362,10 +1367,8 @@ static int __context_pin(struct i915_gem_context *ctx, struct i915_vma *vma)return err; } - flags =PIN_GLOBAL |PIN_HIGH ; -if (ctx->ggtt_offset_bias) - flags |=PIN_OFFSET_BIAS | ctx->ggtt_offset_bias; - + flags =PIN_GLOBAL |PIN_HIGH | +PIN_OFFSET_BIAS |i915_ggtt_pin_bias (vma);return i915_vma_pin (vma,0 ,GEN8_LR_CONTEXT_ALIGN , flags); }
@@ -1392,7 +1395,8 @@ __execlists_context_pin(struct intel_engine_cs *engine, goto unpin_vma; } - ret =intel_ring_pin (ce->ring, ctx->i915, ctx->ggtt_offset_bias); + ret =intel_ring_pin (ce->ring, ctx->i915, +i915_ggtt_pin_bias (ce->ring->vma));if (ret) goto unpin_map;
Signature · 6 lines
2.17.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH v3 5/5 HAX enable GuC for CI
In reply to Jakub BartmińskiFrom: Michal Wajdeczko <[email protected]>
--- drivers/gpu/drm/i915/i915_params.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_params.h +1 -1
@@ -47,7 +47,7 @@ struct drm_printer; param(int, disable_power_well, -1) \ param(int, enable_ips, 1) \ param(int, invert_brightness, 0) \ - param(int, enable_guc, 0) \ + param(int, enable_guc, -1) \ param(int, guc_log_level, -1) \ param(char *, guc_firmware_path, NULL) \param (char *, huc_firmware_path, NULL) \
Signature · 6 lines
2.17.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH v3 3/5] drm/i915: Remove unnecessary ggtt_offset_bias from i915_gem_context
In reply to Jakub BartmińskiQuoting Jakub Bartmiński (2018-07-19 09:35:40)
drivers/gpu/drm/i915/intel_lrc.c static void execlists_context_unpin(struct intel_context *ce) · line 1351
Quoted 3 lines
drivers/gpu/drm/i915/intel_lrc.c +3
-0+static inline u32 i915_ggtt_pin_bias (struct i915_vma *vma) +{ +return i915_vm_to_ggtt (vma->vm)->pin_bias;
The hint was to do this earlier in the series to avoid pulling it randomly from i915->ggtt. If we allocate the vma, then we should have the vm/i915_ggtt already to hand. If not, pulling it from vma->vm looks more consistent than delving into a seemingly unrelated address space.
drivers/gpu/drm/i915/intel_lrc.c static int __context_pin(struct i915_gem_context *ctx, struct i915_vma *vma) · line 1371
Quoted 13 lines
static int
__context_pin(struct i915_gem_context *ctx, struct i915_vma *vma){ unsigned int flags;drivers/gpu/drm/i915/intel_lrc.c +2
-4@@ -1362,10 +1367,8 @@ static int __context_pin(struct i915_gem_context *ctx, struct i915_vma *vma)return err; } - flags =PIN_GLOBAL |PIN_HIGH ; -if (ctx->ggtt_offset_bias) - flags |=PIN_OFFSET_BIAS | ctx->ggtt_offset_bias; - + flags =PIN_GLOBAL |PIN_HIGH | +PIN_OFFSET_BIAS |i915_ggtt_pin_bias (vma);
Looks messy :(
drivers/gpu/drm/i915/intel_lrc.c __execlists_context_pin(struct intel_engine_cs *engine, · line 1399
Quoted 10 lines
drivers/gpu/drm/i915/intel_lrc.c +2
-1return i915_vma_pin (vma,0 ,GEN8_LR_CONTEXT_ALIGN , flags); }@@ -1392,7 +1395,8 @@ __execlists_context_pin(struct intel_engine_cs *engine, goto unpin_vma; } - ret =intel_ring_pin (ce->ring, ctx->i915, ctx->ggtt_offset_bias); + ret =intel_ring_pin (ce->ring, ctx->i915, +i915_ggtt_pin_bias (ce->ring->vma));
This indicates that intel_ring_pin() now knows the pin_bias (as it stored on the i915_ggtt and not on the context) and so no longer needs it pass it from the context.
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH v3 2/5] drm/i915/guc: Move the pin bias value from GuC to GGTT
In reply to Jakub BartmińskiQuoting Jakub Bartmiński (2018-07-19 09:35:39)
Quoted 82 lines
Removing the pin bias from GuC allows us to not check for GuC every time we pin a context, which fixes the assertion error on unresolved GuC platform default in mock contexts selftest.
With this change the
intel_guc_ggtt_offsetfunction has to know the full declaration of thedrm_i915_privatestructure so it can no longer be located in theintel_guc.hheader file due to include order.v2:
This also makes it so that there's no need to set GuC variables from within the WOPCM init function or to move the WOPCM init, while keeping the correct initialization order. Also for mock tests the pin bias is left at 0 and we make sure that the pin bias with GuC will not be smaller than without GuC.v3:
Avoid unused i915 inintel_guc_ggtt_offsetif debug is disabled.Testcase: igt/drv_selftest/mock_contexts #GuC
Signed-off-by Jakub Bartmiński <[email protected]>Cc Chris Wilson <[email protected]>Cc Michał Winiarski <[email protected]>Cc Michal Wajdeczko <[email protected]>--- drivers/gpu/drm/i915/i915_gem_context.c | 10 +------- drivers/gpu/drm/i915/i915_gem_gtt.c | 22 ++++++++++++---- drivers/gpu/drm/i915/i915_gem_gtt.h | 2 ++ drivers/gpu/drm/i915/intel_guc.c | 31 ++++++++++++++++++++--- drivers/gpu/drm/i915/intel_guc.h | 28 +------------------- drivers/gpu/drm/i915/intel_huc.c | 2 +- drivers/gpu/drm/i915/intel_uc_fw.c | 2 +- drivers/gpu/drm/i915/intel_wopcm.c | 19 -------------- drivers/gpu/drm/i915/intel_wopcm.h | 8 ++++++ drivers/gpu/drm/i915/selftests/mock_gtt.c | 2 ++ 10 files changed, 60 insertions(+), 66 deletions(-)drivers/gpu/drm/i915/i915_gem_context.c +1
-9@@ -329,15 +329,7 @@ __create_hw_context(struct drm_i915_private *dev_priv, ctx->desc_template =default_desc_template (dev_priv, dev_priv->mm.aliasing_ppgtt); -/* -* GuC requires the ring to be placed in Non-WOPCM memory. If GuC is not -* present or not in use we still need a small bias as ring wraparound -* at offset 0 sometimes hangs. No idea why. -*/ -if (USES_GUC (dev_priv)) - ctx->ggtt_offset_bias = dev_priv->guc.ggtt_pin_bias; -else - ctx->ggtt_offset_bias =I915_GTT_PAGE_SIZE ; + ctx->ggtt_offset_bias = dev_priv->ggtt.pin_bias;return ctx;drivers/gpu/drm/i915/i915_gem_gtt.c +3
-3@@ -2901,7 +2901,7 @@ void i915_gem_fini_aliasing_ppgtt(struct drm_i915_private *i915) ggtt->vm.vma_ops.unbind_vma = ggtt_unbind_vma; } -int i915_gem_init_ggtt (struct drm_i915_private *dev_priv) +int i915_gem_init_ggtt (struct drm_i915_private *i915) {/* Let GEM Manage all of the aperture. * @@ -2912,12 +2912,24 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)* aperture. One page should be enough to keep any prefetching inside * of the aperture. */ -struct i915_ggtt *ggtt = &dev_priv->ggtt; +struct i915_ggtt *ggtt = &i915->ggtt;unsigned long hole_start, hole_end;struct drm_mm_node *entry;int ret; - ret =intel_vgt_balloon (dev_priv); +/*
drivers/gpu/drm/i915/i915_gem_gtt.c int i915_gem_init_ggtt(struct drm_i915_private *dev_priv) · line 2930
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +10
-0+* We need a small bias as ring wraparound at offset 0 +* sometimes hangs. No idea why. +*/ + ggtt->pin_bias =I915_GTT_PAGE_SIZE ; + +/* Any objects shared with GuC can't overlap with WOPCM. */ +if (USES_GUC (i915)) { + ggtt->pin_bias =max (ggtt->pin_bias, +intel_wopcm_get_pin_bias (&i915->wopcm)); + }
That is defeating the purpose of the series. We don't want to make these decisions here, we just want to ask if intel_wopcm requires it. i.e., at this point we only know that wopcm may reserve some of the Global GTT for itself, we don't need to complicate that any further here.
USES_GUC() == modparam, ergo it should never be called in live code. -Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH v3 4/5] drm/i915: Add a fault injection point after WOPCM init
In reply to Jakub BartmińskiQuoting Jakub Bartmiński (2018-07-19 09:35:41)
Quoted 13 lines
Signed-off-by Jakub Bartmiński <[email protected]>Cc Chris Wilson <[email protected]>Cc Michał Winiarski <[email protected]>Cc Michal Wajdeczko <[email protected]>--- drivers/gpu/drm/i915/i915_gem.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/i915/i915_gem.c +0
-0
drivers/gpu/drm/i915/i915_gem.c int i915_gem_init(struct drm_i915_private *dev_priv) · line 5481
Quoted 7 lines
drivers/gpu/drm/i915/i915_gem.c +4
-0@@ -5475,6 +5475,11 @@ int i915_gem_init(struct drm_i915_private *dev_priv)if (ret) goto err_uc_misc; +if (i915_inject_load_failure ()) { + ret = -E2BIG ; + goto err_uc_misc; + }
Push it down to the callee. I want us to exercise what happens when that fails. Otherwise we just have static fault injection and never need to worry about the order here.
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✗ Fi.CI.SPARSE: warning for series starting with [v3,1/5] drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
In reply to Jakub Bartmiński== Series Details ==
== Series Details ==
Series: series starting with [v3,1/5] drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
URL : https://patchwork.freedesktop.org/series/46843/
State : warning
== Summary ==
$ dim sparse origin/drm-tip Commit: drm/i915/guc: Fix GuC pin bias and WOPCM initialization order Okay!
Commit: drm/i915/guc: Move the pin bias value from GuC to GGTT +drivers/gpu/drm/i915/i915_gem_gtt.c:2948:34: warning: expression using sizeof(void) +drivers/gpu/drm/i915/i915_gem_gtt.c:2948:34: warning: expression using sizeof(void)
Commit: drm/i915: Remove unnecessary ggtt_offset_bias from i915_gem_context Okay!
Commit: drm/i915: Add a fault injection point after WOPCM init Okay!
Commit: HAX enable GuC for CI
Okay!
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✗ Fi.CI.BAT: failure for series starting with [v3,1/5] drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
In reply to Jakub Bartmiński== Series Details ==
== Series Details ==
Series: series starting with [v3,1/5] drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
URL : https://patchwork.freedesktop.org/series/46843/
State : failure
== Summary ==
= CI Bug Log - changes from CI_DRM_4510 -> Patchwork_9717 =
== Summary - FAILURE ==
Serious unknown changes coming with Patchwork_9717 absolutely need to be verified manually.
If you think the reported changes have nothing to do with the changes introduced in Patchwork_9717, please notify your bug team to allow them to document this new failure mode, which will reduce false positives in CI.
External URL: https://patchwork.freedesktop.org/api/1.0/series/46843/revisions/1/mbox/
== Possible new issues ==
Here are the unknown changes that may have been introduced in Patchwork_9717:
=== IGT changes ===
==== Possible regressions ====
igt@drv_selftest@live_guc:
fi-kbl-7567u: PASS -> DMESG-WARN
fi-skl-gvtdvm: PASS -> DMESG-WARN
fi-bxt-dsi: NOTRUN -> DMESG-WARN
fi-whl-u: PASS -> DMESG-WARN
fi-kbl-7560u: PASS -> DMESG-WARN
fi-kbl-r: PASS -> DMESG-WARN
fi-kbl-x1275: PASS -> DMESG-WARN
fi-bxt-j4205: PASS -> DMESG-WARN
fi-cfl-s3: PASS -> DMESG-WARN
{fi-cfl-8109u}: PASS -> DMESG-WARN
fi-kbl-7500u: PASS -> DMESG-WARN
fi-cfl-8700k: PASS -> DMESG-WARN
igt@drv_selftest@live_hangcheck:
fi-skl-gvtdvm: PASS -> DMESG-FAIL
fi-cfl-s3: PASS -> DMESG-FAIL
fi-bxt-dsi: NOTRUN -> DMESG-FAIL
fi-bxt-j4205: PASS -> DMESG-FAIL
{fi-skl-iommu}: PASS -> DMESG-FAIL +1
==== Warnings ====
igt@drv_selftest@live_guc:
{fi-skl-iommu}: PASS -> SKIP +1== Known issues ==
Here are the changes found in Patchwork_9717 that come from known issues:
=== IGT changes ===
==== Issues hit ====
igt@drv_selftest@live_guc:
fi-skl-6600u: PASS -> DMESG-WARN (fdo#107175)
fi-skl-6260u: PASS -> DMESG-WARN (fdo#107175)
fi-skl-6700k2: PASS -> DMESG-WARN (fdo#107175)
fi-skl-6770hq: PASS -> DMESG-WARN (fdo#107175)
fi-skl-6700hq: PASS -> DMESG-WARN (fdo#107175)
igt@drv_selftest@live_hangcheck:
fi-skl-6770hq: PASS -> DMESG-FAIL (fdo#107174)
==== Possible fixes ====
igt@debugfs_test@read_all_entries:
fi-snb-2520m: INCOMPLETE (fdo#103713) -> PASS
igt@drv_selftest@live_workarounds:
fi-bdw-5557u: DMESG-FAIL -> PASS
fi-skl-6700hq: DMESG-FAIL -> PASS
igt@kms_pipe_crc_basic@suspend-read-crc-pipe-c:
fi-bxt-dsi: INCOMPLETE (fdo#103927) -> PASS {name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
fdo#103713 https://bugs.freedesktop.org/show_bug.cgi?id=103713
fdo#103927 https://bugs.freedesktop.org/show_bug.cgi?id=103927
fdo#107174 https://bugs.freedesktop.org/show_bug.cgi?id=107174
fdo#107175 https://bugs.freedesktop.org/show_bug.cgi?id=107175
== Participating hosts (47 -> 42) ==
Missing (5): fi-ctg-p8600 fi-ilk-m540 fi-byt-squawks fi-bsw-cyan fi-hsw-4200u
== Build changes ==
* Linux: CI_DRM_4510 -> Patchwork_9717
CI_DRM_4510: 31b9b285a571eda0db7a8fd5206d48fcdd0190b3 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4568: 86f7b724ef18986bc58d35558d22e1ed3f8df4f9 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_9717: 2306b20a6dea3a4f224d21c00f312f89d216e426 @ git://anongit.freedesktop.org/gfx-ci/linux
== Linux commits ==
2306b20a6dea HAX enable GuC for CI0fa25edf61e9 drm/i915: Add a fault injection point after WOPCM init eb5bfd46972b drm/i915: Remove unnecessary ggtt_offset_bias from i915_gem_context e9dda8797e50 drm/i915/guc: Move the pin bias value from GuC to GGTT 5b24a0eaa5b5 drm/i915/guc: Fix GuC pin bias and WOPCM initialization order
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_9717/issues.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx