Broadwater and the rest of gen4 do support being able to saving and reloading context specific registers between contexts, providing isolation of the basic GPU state (as programmable by userspace). This allows userspace to assume that the GPU retains their state from one batch to the next, minimising the amount of state it needs to reload and manually save across batches. v2: CONSTANT_BUFFER woes Running through piglit turned up an interesting issue, a GPU hang inside the context load. The context image includes the CONSTANT_BUFFER command that loads an address into a on-gpu buffer, and the context load was executing that immediately. However, since it was reading from the GTT there is no guarantee that the GTT retains the same configuration as when the context was saved, resulting in stray reads and a GPU hang. Having tried issuing a CONSTANT_BUFFER (to disable the command) from the ring before saving the context to no avail, we resort to patching out the instruction inside the context image before loading. This does impose that gen4 always reissues CONSTANT_BUFFER commands on each batch, but due to the use of a shared GTT that was and will remain a requirement. Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> Cc: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> Cc: Kenneth Graunke <kenneth@xxxxxxxxxxxxx> Reviewed-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> #v1 --- drivers/gpu/drm/i915/intel_engine_cs.c | 2 +- drivers/gpu/drm/i915/intel_gpu_commands.h | 3 +++ drivers/gpu/drm/i915/intel_ringbuffer.c | 17 +++++++++++++++++ 3 files changed, 21 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 148c3e06a2eb..32bd850eec30 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -220,6 +220,7 @@ __intel_engine_context_size(struct drm_i915_private *dev_priv, u8 class) return round_up(GEN6_CXT_TOTAL_SIZE(cxt_size) * 64, PAGE_SIZE); case 5: + case 4: /* * There is a discrepancy here between the size reported * by the register and the size of the context layout @@ -236,7 +237,6 @@ __intel_engine_context_size(struct drm_i915_private *dev_priv, u8 class) cxt_size * 64, cxt_size - 1); return round_up(cxt_size * 64, PAGE_SIZE); - case 4: case 3: case 2: /* For the special day when i810 gets merged. */ diff --git a/drivers/gpu/drm/i915/intel_gpu_commands.h b/drivers/gpu/drm/i915/intel_gpu_commands.h index b96a31bc1080..a95bfd922c41 100644 --- a/drivers/gpu/drm/i915/intel_gpu_commands.h +++ b/drivers/gpu/drm/i915/intel_gpu_commands.h @@ -265,6 +265,9 @@ #define GFX_OP_3DSTATE_BINDING_TABLE_EDIT_PS \ ((0x3<<29)|(0x3<<27)|(0x0<<24)|(0x47<<16)) +#define GFX_OP_CONSTANT_BUFFER \ + (0x3 << 29 | 0x0 << 27 | 0x0 << 24 | 0x2 << 16) + #define MFX_WAIT ((0x3<<29)|(0x1<<27)|(0x0<<16)) #define COLOR_BLT ((0x2<<29)|(0x40<<22)) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 19def67bf1c5..c03d156d59d9 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1734,6 +1734,8 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) len += 2 + (num_rings ? 4*num_rings + 6 : 0); else if (IS_GEN(i915, 5)) len += 2; + else if (IS_GEN(i915, 4)) + len += 4; if (flags & MI_FORCE_RESTORE) { GEM_BUG_ON(flags & MI_RESTORE_INHIBIT); flags &= ~MI_FORCE_RESTORE; @@ -1770,6 +1772,21 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) * this should never take effect and so be a no-op! */ *cs++ = MI_SUSPEND_FLUSH | MI_SUSPEND_FLUSH_EN; + } else if (IS_GEN(i915, 4)) { + /* + * Disable CONSTANT_BUFFER before it is loaded from the context + * image. For as it is loaded, it is executed and the stored + * address may no longer be valid, leading to a GPU hang. + * + * This imposes the requirement that userspace reload their + * CONSTANT_BUFFER on every batch, fortunately a requirement + * they are already accustomed to from before contexts were + * enabled. + */ + *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; + *cs++ = 0; + *cs++ = i915_ggtt_offset(rq->hw_context->state) + 0x1d4; + *cs++ = GFX_OP_CONSTANT_BUFFER; /* inactive */ } if (force_restore) { -- 2.20.1 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx