On Tue, Nov 25, 2025 at 03:48:42PM +0530, Nareshkumar Gollakoti wrote: > The variable "fence" should be initialized to NULL, > and any usage of fence should be guarded > by a check to ensure it is not NULL > > Signed-off-by: Nareshkumar Gollakoti <[email protected]> > --- > drivers/gpu/drm/xe/xe_pagefault.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/xe/xe_pagefault.c > b/drivers/gpu/drm/xe/xe_pagefault.c > index afb06598b6e1..401f1835939b 100644 > --- a/drivers/gpu/drm/xe/xe_pagefault.c > +++ b/drivers/gpu/drm/xe/xe_pagefault.c > @@ -70,7 +70,7 @@ static int xe_pagefault_handle_vma(struct xe_gt *gt, struct > xe_vma *vma, > struct xe_tile *tile = gt_to_tile(gt); > struct xe_validation_ctx ctx; > struct drm_exec exec; > - struct dma_fence *fence; > + struct dma_fence *fence = NULL; > int err, needs_vram; > > lockdep_assert_held_write(&vm->lock); > @@ -122,8 +122,10 @@ static int xe_pagefault_handle_vma(struct xe_gt *gt, > struct xe_vma *vma, > } > } > > - dma_fence_wait(fence, false); > - dma_fence_put(fence); > + if (fence) { > + dma_fence_wait(fence, false); > + dma_fence_put(fence); > + }
Maybe I'm blind, but I don't see the problem in the existing code. How did this get reported as an issue which you are trying to fix here? Also I think Rodrigo mentioned this fixup! are not convention in DRM, rather 'Fixes' tags. Matt > > unlock_dma_resv: > xe_validation_ctx_fini(&ctx); > -- > 2.43.0 >
