accel/amdxdna: Prevent PM resume deadlock in hwctx_sync_debug_bo()

amdxdna_hwctx_sync_debug_bo() invokes the hardware hwctx_sync_debug_bo()
callback while holding xdna->dev_lock.

The callback may call amdxdna_cmd_submit(), which in turn calls
amdxdna_pm_resume_get(). If the device is suspended,
amdxdna_pm_resume_get() may synchronously execute amdxdna_pm_resume(),
which also acquires xdna->dev_lock, resulting in a deadlock.

Avoid the deadlock by calling amdxdna_pm_resume_get() before holding
xdna->dev_lock in both amdxdna_hwctx_sync_debug_bo() and
amdxdna_drm_config_hwctx_ioctl()

Fixes: 7ea046838021 ("accel/amdxdna: Support firmware debug buffer")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260616212429.3620645-1-lizhi.hou@amd.com

diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c
index 54486960cbf5e688bceff79b7919864d9077456a..4fa9abd90cd7dbb0e1cd1b15eecbd13cd8017002 100644 (file)
--- a/drivers/accel/amdxdna/aie2_ctx.c
+++ b/drivers/accel/amdxdna/aie2_ctx.c
@@ -875,7 +875,7 @@ static int aie2_hwctx_cu_config(struct amdxdna_hwctx *hwctx, void *buf, u32 size
        if (!hwctx->cus)
                return -ENOMEM;
 
-       ret = amdxdna_pm_resume_get_locked(xdna);
+       ret = amdxdna_pm_resume_get(xdna);
        if (ret)
                goto free_cus;
 

diff --git a/drivers/accel/amdxdna/amdxdna_ctx.c b/drivers/accel/amdxdna/amdxdna_ctx.c
index 855da8c79a1cdcdf6df00f703d3a98f8e4189b43..67a2abcf173ee544c7eab90ae030a650b1c92c95 100644 (file)
--- a/drivers/accel/amdxdna/amdxdna_ctx.c
+++ b/drivers/accel/amdxdna/amdxdna_ctx.c
@@ -382,16 +382,25 @@ int amdxdna_drm_config_hwctx_ioctl(struct drm_device *dev, void *data, struct dr
                return -EINVAL;
        }
 
-       guard(mutex)(&xdna->dev_lock);
+       ret = amdxdna_pm_resume_get(xdna);
+       if (ret) {
+               XDNA_ERR(xdna, "Resume failed, ret %d", ret);
+               goto free_buf;
+       }
+
+       mutex_lock(&xdna->dev_lock);
        hwctx = xa_load(&client->hwctx_xa, args->handle);
        if (!hwctx) {
                XDNA_DBG(xdna, "PID %d failed to get hwctx %d", client->pid, args->handle);
                ret = -EINVAL;
-               goto free_buf;
+               goto unlock;
        }
 
        ret = xdna->dev_info->ops->hwctx_config(hwctx, args->param_type, val, buf, buf_size);
 
+unlock:
+       mutex_unlock(&xdna->dev_lock);
+       amdxdna_pm_suspend_put(xdna);
 free_buf:
        kfree(buf);
        return ret;
@@ -412,16 +421,25 @@ int amdxdna_hwctx_sync_debug_bo(struct amdxdna_client *client, u32 debug_bo_hdl)
        if (!gobj)
                return -EINVAL;
 
+       ret = amdxdna_pm_resume_get(xdna);
+       if (ret) {
+               XDNA_ERR(xdna, "Resume failed, ret %d", ret);
+               goto put_obj;
+       }
+
        abo = to_xdna_obj(gobj);
-       guard(mutex)(&xdna->dev_lock);
+       mutex_lock(&xdna->dev_lock);
        hwctx = xa_load(&client->hwctx_xa, abo->assigned_hwctx);
        if (!hwctx) {
                ret = -EINVAL;
-               goto put_obj;
+               goto unlock;
        }
 
        ret = xdna->dev_info->ops->hwctx_sync_debug_bo(hwctx, debug_bo_hdl);
 
+unlock:
+       mutex_unlock(&xdna->dev_lock);
+       amdxdna_pm_suspend_put(xdna);
 put_obj:
        drm_gem_object_put(gobj);
        return ret;