On 7/16/2025 12:57 PM, Gang Ba wrote:
If vm belongs to another process, this is fclose after fork,
wait may enable signaling KFD eviction fence and cause parent process queue
evicted.
The commit message does not target the issue description. amdgpu_flush
got trigger from child process when it makes execve system call because
render node has O_CLOEXEC flag. fork only does not close inherited file
descriptors from child process. The back trace below also shows that.
Regards
Xiaogang
[677852.634569] amdkfd_fence_enable_signaling+0x56/0x70 [amdgpu]
[677852.634814] __dma_fence_enable_signaling+0x3e/0xe0
[677852.634820] dma_fence_wait_timeout+0x3a/0x140
[677852.634825] amddma_resv_wait_timeout+0x7f/0xf0 [amdkcl]
[677852.634831] amdgpu_vm_wait_idle+0x2d/0x60 [amdgpu]
[677852.635026] amdgpu_flush+0x34/0x50 [amdgpu]
[677852.635208] filp_flush+0x38/0x90
[677852.635213] filp_close+0x14/0x30
[677852.635216] do_close_on_exec+0xdd/0x130
[677852.635221] begin_new_exec+0x1da/0x490
[677852.635225] load_elf_binary+0x307/0xea0
[677852.635231] ? srso_alias_return_thunk+0x5/0xfbef5
[677852.635235] ? ima_bprm_check+0xa2/0xd0
[677852.635240] search_binary_handler+0xda/0x260
[677852.635245] exec_binprm+0x58/0x1a0
[677852.635249] bprm_execve.part.0+0x16f/0x210
[677852.635254] bprm_execve+0x45/0x80
[677852.635257] do_execveat_common.isra.0+0x190/0x200
Suggested-by: Christian König<christian.koe...@amd.com>
Signed-off-by: Gang Ba<gang...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index ea9b0f050f79..2f75f967f95f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -2414,13 +2414,13 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev,
uint32_t min_vm_size,
*/
long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
{
- timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv,
- DMA_RESV_USAGE_BOOKKEEP,
- true, timeout);
+ guard(mutex)(&vm->eviction_lock);
+
+ timeout = drm_sched_entity_flush(&vm->immediate, timeout);
if (timeout <= 0)
return timeout;
- return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
+ return drm_sched_entity_flush(&vm->delayed, timeout);
}
static void amdgpu_vm_destroy_task_info(struct kref *kref)
