On gen8 and onwards, we can mark GPU accesses through the ppGTT as being
read-only, that is cause any GPU write onto that page to be discarded
(not triggering a fault). This is all that we need to finally support
the read-only flag for userptr!
Testcase: igt/gem_userptr_blits/readonly*
Signed-off-by: Chris Wilson <ch...@chris-wilson.co.uk>
Cc: Jon Bloomfield <jon.bloomfi...@intel.com>
Cc: Joonas Lahtinen <joonas.lahti...@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem_userptr.c | 13 +++++++++----
1 file changed, 9 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c
b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 854bd51b9478..d4ee8fa4c379 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -789,10 +789,12 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
return -EFAULT;
if (args->flags & I915_USERPTR_READ_ONLY) {
- /* On almost all of the current hw, we cannot tell the GPU that
a
- * page is readonly, so this is just a placeholder in the uAPI.
+ /*
+ * On almost all of the older hw, we cannot tell the GPU that
+ * a page is readonly.
*/
- return -ENODEV;
+ if (INTEL_GEN(dev_priv) < 8 || !USES_PPGTT(dev_priv))
+ return -ENODEV;
}
obj = i915_gem_object_alloc(dev_priv);
@@ -806,7 +808,10 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC);
obj->userptr.ptr = args->user_ptr;
- obj->userptr.read_only = !!(args->flags & I915_USERPTR_READ_ONLY);
+ if (args->flags & I915_USERPTR_READ_ONLY) {
+ obj->userptr.read_only = true;
+ obj->gt_ro = true;
+ }
/* And keep a pointer to the current->mm for resolving the user pages
* at binding. This means that we need to hook into the mmu_notifier
--
2.17.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
