<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<p><br>
</p>
<div class="moz-cite-prefix">On 2021-12-14 12:03 p.m., Andrey
Grodzovsky wrote:<br>
</div>
<blockquote type="cite" cite="mid:20f336a4-0bfd-9988-ee3b-a8206f045f7e@amd.com">
<blockquote type="cite" style="color: #007cff;">-
<br>
- if (job != NULL) {
<br>
- /* mark this fence has a parent job */
<br>
- set_bit(AMDGPU_FENCE_FLAG_EMBED_IN_JOB_BIT,
&fence->flags);
<br>
+ if (job)
<br>
+ dma_fence_init(fence, &amdgpu_job_fence_ops,
<br>
+ &ring->fence_drv.lock,
<br>
+ adev->fence_context + ring->idx,
seq);
<br>
+ else
<br>
+ dma_fence_init(fence, &amdgpu_fence_ops,
<br>
+ &ring->fence_drv.lock,
<br>
+ adev->fence_context + ring->idx,
seq);
<br>
}
<br>
</blockquote>
<br>
<br>
It's probably me missing something but why can't we just move
setting of AMDGPU_FENCE_FLAG_EMBED_IN_JOB_BIT
<br>
to before dma_fence_init here or even into amdgpu_job_alloc
instead of all the refactoring ?
<br>
<br>
Andrey
</blockquote>
<p><br>
</p>
<p>My bad, I see now that dma_fence_init just overrides flags to 0
and immediately call the trace .<br>
But why then can't we add dma_fence_init_with_flags wrapper to
dma-fence ?</p>
<p>Andrey<br>
</p>
</body>
</html>