On 2022-04-28 09:51, Felix Kuehling
wrote:
Am 2022-04-27 um 19:07 schrieb Philip Yang:yes, thanks for catching this, I will send another patch to fix this as I have merged this patch.
Otherwise userspace resubmit the BOs again will trigger kernel WARNING
and fail the command submission.
Signed-off-by: Philip Yang <Philip.Yang@xxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 18 +++++++++++++++---
1 file changed, 15 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index ea28942b0ede..44dcbef80229 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -550,7 +550,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
There is another error case a few lines up, if kvmalloc_array fails. I think that needs to be handled with a goto out_free_user_pages as well.
if (r) {
kvfree(e->user_pages);
e->user_pages = NULL;
The above two lines can be removed. The loop under out_free_user_pages should handle this.
No, out_free_user_pages will call
amdgpu_ttm_tt_get_user_pages_done with gtt->range is NULL for
this BO, and trigger kernel WARNING.
Thanks,
Philip
Regards,
Felix
- return r;
+ goto out_free_user_pages;
}
for (i = 0; i < bo->tbo.ttm->num_pages; i++) {
@@ -567,7 +567,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
if (unlikely(r != 0)) {
if (r != -ERESTARTSYS)
DRM_ERROR("ttm_eu_reserve_buffers failed.\n");
- goto out;
+ goto out_free_user_pages;
}
amdgpu_bo_list_for_each_entry(e, p->bo_list) {
@@ -649,7 +649,19 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
}
ttm_eu_backoff_reservation(&p->ticket, &p->validated);
}
-out:
+
+out_free_user_pages:
+ if (r) {
+ amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
+ struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+
+ if (!e->user_pages)
+ continue;
+ amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
+ kvfree(e->user_pages);
+ e->user_pages = NULL;
+ }
+ }
return r;
}
