accel/amdxdna: Fix command hang on suspended hardware context

[PATCH V1] accel/amdxdna: Fix command hang on suspended hardware context

Posted by Lizhi Hou 1 month, 2 weeks ago

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")
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
---
 drivers/accel/amdxdna/aie2_ctx.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c
index 89c0ab896f92..622ddbf7fb6f 100644
--- a/drivers/accel/amdxdna/aie2_ctx.c
+++ b/drivers/accel/amdxdna/aie2_ctx.c
@@ -53,6 +53,7 @@ static void aie2_hwctx_stop(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwct
 {
 	drm_sched_stop(&hwctx->priv->sched, bad_job);
 	aie2_destroy_context(xdna->dev_handle, hwctx);
+	drm_sched_start(&hwctx->priv->sched, 0);
 }
 
 static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwctx)
@@ -80,7 +81,6 @@ static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hw
 	}
 
 out:
-	drm_sched_start(&hwctx->priv->sched, 0);
 	XDNA_DBG(xdna, "%s restarted, ret %d", hwctx->name, ret);
 	return ret;
 }
@@ -297,19 +297,23 @@ aie2_sched_job_run(struct drm_sched_job *sched_job)
 	struct dma_fence *fence;
 	int ret;
 
-	if (!hwctx->priv->mbox_chann)
+	ret = amdxdna_pm_resume_get(hwctx->client->xdna);
+	if (ret)
+		return NULL;
+
+	if (!hwctx->priv->mbox_chann) {
+		amdxdna_pm_suspend_put(hwctx->client->xdna);
 		return NULL;
+	}
 
-	if (!mmget_not_zero(job->mm))
+	if (!mmget_not_zero(job->mm)) {
+		amdxdna_pm_suspend_put(hwctx->client->xdna);
 		return ERR_PTR(-ESRCH);
+	}
 
 	kref_get(&job->refcnt);
 	fence = dma_fence_get(job->fence);
 
-	ret = amdxdna_pm_resume_get(hwctx->client->xdna);
-	if (ret)
-		goto out;
-
 	if (job->drv_cmd) {
 		switch (job->drv_cmd->opcode) {
 		case SYNC_DEBUG_BO:
-- 
2.34.1

Re: [PATCH V1] accel/amdxdna: Fix command hang on suspended hardware context

Posted by Mario Limonciello 1 month, 2 weeks ago

On 2/11/26 2:53 PM, Lizhi Hou wrote:
> 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")
> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
> ---
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
>   drivers/accel/amdxdna/aie2_ctx.c | 18 +++++++++++-------
>   1 file changed, 11 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c
> index 89c0ab896f92..622ddbf7fb6f 100644
> --- a/drivers/accel/amdxdna/aie2_ctx.c
> +++ b/drivers/accel/amdxdna/aie2_ctx.c
> @@ -53,6 +53,7 @@ static void aie2_hwctx_stop(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwct
>   {
>   	drm_sched_stop(&hwctx->priv->sched, bad_job);
>   	aie2_destroy_context(xdna->dev_handle, hwctx);
> +	drm_sched_start(&hwctx->priv->sched, 0);
>   }
>   
>   static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hwctx)
> @@ -80,7 +81,6 @@ static int aie2_hwctx_restart(struct amdxdna_dev *xdna, struct amdxdna_hwctx *hw
>   	}
>   
>   out:
> -	drm_sched_start(&hwctx->priv->sched, 0);
>   	XDNA_DBG(xdna, "%s restarted, ret %d", hwctx->name, ret);
>   	return ret;
>   }
> @@ -297,19 +297,23 @@ aie2_sched_job_run(struct drm_sched_job *sched_job)
>   	struct dma_fence *fence;
>   	int ret;
>   
> -	if (!hwctx->priv->mbox_chann)
> +	ret = amdxdna_pm_resume_get(hwctx->client->xdna);
> +	if (ret)
> +		return NULL;
> +
> +	if (!hwctx->priv->mbox_chann) {
> +		amdxdna_pm_suspend_put(hwctx->client->xdna);
>   		return NULL;
> +	}
>   
> -	if (!mmget_not_zero(job->mm))
> +	if (!mmget_not_zero(job->mm)) {
> +		amdxdna_pm_suspend_put(hwctx->client->xdna);
>   		return ERR_PTR(-ESRCH);
> +	}
>   
>   	kref_get(&job->refcnt);
>   	fence = dma_fence_get(job->fence);
>   
> -	ret = amdxdna_pm_resume_get(hwctx->client->xdna);
> -	if (ret)
> -		goto out;
> -
>   	if (job->drv_cmd) {
>   		switch (job->drv_cmd->opcode) {
>   		case SYNC_DEBUG_BO: