On 22-04-2025 22:34, Matthew Brost wrote:
Ensure GPU can make forward progress on an atomic SVM GPU fault by
giving the GPU a timeslice of 5ms
v2:
- Reduce timeslice to 5ms
- Double timeslice on retry
- Split out GPU SVM changes into independent patch
Signed-off-by: Matthew Brost <matthew.br...@intel.com>
---
drivers/gpu/drm/xe/xe_svm.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/xe/xe_svm.c b/drivers/gpu/drm/xe/xe_svm.c
index f749ae367a8f..d5376a76cdd1 100644
--- a/drivers/gpu/drm/xe/xe_svm.c
+++ b/drivers/gpu/drm/xe/xe_svm.c
@@ -783,6 +783,8 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma
*vma,
IS_ENABLED(CONFIG_DRM_XE_DEVMEM_MIRROR) ? SZ_64K : 0,
.devmem_only = atomic && IS_DGFX(vm->xe) &&
IS_ENABLED(CONFIG_DRM_XE_DEVMEM_MIRROR),
+ .timeslice_ms = atomic && IS_DGFX(vm->xe) &&
+ IS_ENABLED(CONFIG_DRM_XE_DEVMEM_MIRROR) ? 5 : 0,
};
struct xe_svm_range *range;
struct drm_gpusvm_range *r;
@@ -819,6 +821,7 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma
*vma,
if (--migrate_try_count >= 0 &&
xe_svm_range_needs_migrate_to_vram(range, vma)) {
err = xe_svm_alloc_vram(vm, tile, range, &ctx);
+ ctx.timeslice_ms <<= 1; /* Double timeslice if we have to
retry */
LGTM
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimi...@intel.com>
if (err) {
if (migrate_try_count || !ctx.devmem_only) {
drm_dbg(&vm->xe->drm,
