drm/i915: don't track relative-constants-mode
Subsystems (2)
PATCH v2 1/1 drm/i915: don't track relative-constants-mode
'relative_constants_mode' has always been tracked per-device, but this has actually been wrong ever since hardware contexts were introduced, as the INSTPM register is saved (and automatically restored) as part of the render ring context. The software per-device value could therefore get out of sync with the hardware per-context value.
Rather than track this correctly (per-context in LRC modes, per-device in legacy ringbuffer mode), a simpler solution is just to give up trying to track it at all, and always surround each render batch that uses a non-default mode with the instructions to set it before and reset it to default after.
Test case (if anyone wants to write it) would be to create two contexts, submit a batch with a non-default mode in the first context, submit a batch with the default mode in the other context, submit another batch in the first context, but this time in default mode. Without this patch, the driver will fail to insert the instructions to reset INSTPM into the first context's ringbuffer.
v2:
toggle mode on and off for non-default batches, rather than setting it explicitly for every batch,
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92448
--- drivers/gpu/drm/i915/i915_drv.h | 2 - drivers/gpu/drm/i915/i915_gem.c | 3 -- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 70 ++++++++++++++++++------------ 3 files changed, 43 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +0 -2
@@ -1728,8 +1728,6 @@ struct drm_i915_private {const struct intel_device_info info; -int relative_constants_mode; -void __iomem *regs;struct intel_uncore uncore;
drivers/gpu/drm/i915/i915_gem.c +0 -3
@@ -4615,9 +4615,6 @@ int i915_gem_init(struct drm_device *dev) i915_gem_idle_work_handler);init_waitqueue_head (&dev_priv->gpu_error.wait_queue);init_waitqueue_head (&dev_priv->gpu_error.reset_queue); - - dev_priv->relative_constants_mode =I915_EXEC_CONSTANTS_REL_GENERAL ; -init_waitqueue_head (&dev_priv->pending_flip_queue); dev_priv->mm.interruptible = true;
drivers/gpu/drm/i915/i915_gem_execbuffer.c +43 -27
98 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
drivers/gpu/drm/i915/i915_gem_execbuffer.c static void eb_export_fence(struct drm_i915_gem_object *obj, · line 1536
Quoted 10 lines · Dave Gordon
drivers/gpu/drm/i915/i915_gem_execbuffer.c +6
-0@@ -1520,6 +1528,14 @@ static void eb_export_fence(struct drm_i915_gem_object *obj,if (ret)return ret; +if (instp_mode !=I915_EXEC_CONSTANTS_REL_GENERAL ) { +/* Restore default value of INSTPM */ + ret =emit_instpm (params->request, instp_mask, +I915_EXEC_CONSTANTS_REL_GENERAL ); +if (ret) +return ret;
We can't return an error at this point, we are committed to executing the batch. And we do expect to see at least an EINTR here occasionally. -Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 5 lines · Chris Wilson
Quoted 4 lines · Dave Gordon
drivers/gpu/drm/i915/i915_gem_execbuffer.c +0
-0@@ -1520,6 +1528,14 @@ static void eb_export_fence(struct drm_i915_gem_object *obj,if (ret)return ret;
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem_execbuffer.c static void eb_export_fence(struct drm_i915_gem_object *obj, · line 1536
Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_execbuffer.c +6
-0+if (instp_mode !=I915_EXEC_CONSTANTS_REL_GENERAL ) { +/* Restore default value of INSTPM */ + ret =emit_instpm (params->request, instp_mask, +I915_EXEC_CONSTANTS_REL_GENERAL ); +if (ret) +return ret;We can't return an error at this point, we are committed to executing the batch. And we do expect to see at least an EINTR here occasionally. -Chris
Note the 'if (ret) return ret' just above. We already quit early on error, this just adds yet one more case. If you want to throw half-formed command streams at the GPU you'll need to fix all the other early exits, so adding one more doesn't really make it any worse.
In any case, it's the result from the ring_begin() which universally triggers an early exit, but on the other hand should never fail because we reserved enough space up front.
.Dave.
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx