Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
accel/amdxdna: Fix command hang on suspended hardware context
When a hardware context is suspended, the job scheduler is stopped. If a
command is submitted while the context is suspended, the job is queued in
the scheduler but aie2_sched_job_run() is never invoked to restart the
hardware context. As a result, the command hangs.
Fix this by modifying the hardware context suspend routine to keep the job
scheduler running so that queued jobs can trigger context restart properly.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260211205341.722982-1-lizhi.hou@amd.com
+11
-7
+11
-7
drivers/accel/amdxdna/aie2_ctx.c
+11
-7
drivers/accel/amdxdna/aie2_ctx.c
···
53
53
{
54
54
drm_sched_stop(&hwctx->priv->sched, bad_job);
55
55
aie2_destroy_context(xdna->dev_handle, hwctx);
56
+
drm_sched_start(&hwctx->priv->sched, 0);
56
57
}
57
58
58
59
static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwctx)
···
81
80
}
82
81
83
82
out:
84
-
drm_sched_start(&hwctx->priv->sched, 0);
85
83
XDNA_DBG(xdna, "%s restarted, ret %d", hwctx->name, ret);
86
84
return ret;
87
85
}
···
297
297
struct dma_fence *fence;
298
298
int ret;
299
299
300
-
if (!hwctx->priv->mbox_chann)
300
+
ret = amdxdna_pm_resume_get(hwctx->client->xdna);
301
+
if (ret)
301
302
return NULL;
302
303
303
-
if (!mmget_not_zero(job->mm))
304
+
if (!hwctx->priv->mbox_chann) {
305
+
amdxdna_pm_suspend_put(hwctx->client->xdna);
306
+
return NULL;
307
+
}
308
+
309
+
if (!mmget_not_zero(job->mm)) {
310
+
amdxdna_pm_suspend_put(hwctx->client->xdna);
304
311
return ERR_PTR(-ESRCH);
312
+
}
305
313
306
314
kref_get(&job->refcnt);
307
315
fence = dma_fence_get(job->fence);
308
-
309
-
ret = amdxdna_pm_resume_get(hwctx->client->xdna);
310
-
if (ret)
311
-
goto out;
312
316
313
317
if (job->drv_cmd) {
314
318
switch (job->drv_cmd->opcode) {