DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time
@ 2020-01-16  2:07 Zhike Wang
  2020-02-04 18:16 ` Maxime Coquelin
  2020-02-05  9:46 ` Maxime Coquelin
  0 siblings, 2 replies; 3+ messages in thread
From: Zhike Wang @ 2020-01-16  2:07 UTC (permalink / raw)
  To: dev; +Cc: maxime.coquelin, tiwei.bie, zhihong.wang, stable, Zhike Wang

The vhost_user_read_cb() and rte_vhost_driver_unregister() can be
called at the same time by 2 threads. Eg thread1 calls vhost_user_read_cb()
and removes the vsocket from conn_list, then thread2 calls
rte_vhost_driver_unregister() and frees the vsocket since it is NOT in the
conn_list. So thread1 will access invalid memory when trying to reconnect.

The fix is to move the "removing of vsocket from conn_list" to end of the
vhost_user_read_cb(), then avoid the race condition.

The core trace is
Program terminated with signal 11, Segmentation fault.

Signed-off-by: Zhike Wang <wangzhike@jd.com>
---
 lib/librte_vhost/socket.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/lib/librte_vhost/socket.c b/lib/librte_vhost/socket.c
index ebb2ff6..01a9dec 100644
--- a/lib/librte_vhost/socket.c
+++ b/lib/librte_vhost/socket.c
@@ -318,16 +318,16 @@ struct vhost_user {
 
 		vhost_destroy_device(conn->vid);
 
+		if (vsocket->reconnect) {
+			create_unix_socket(vsocket);
+			vhost_user_start_client(vsocket);
+		}
+
 		pthread_mutex_lock(&vsocket->conn_mutex);
 		TAILQ_REMOVE(&vsocket->conn_list, conn, next);
 		pthread_mutex_unlock(&vsocket->conn_mutex);
 
 		free(conn);
-
-		if (vsocket->reconnect) {
-			create_unix_socket(vsocket);
-			vhost_user_start_client(vsocket);
-		}
 	}
 }
 
-- 
1.8.3.1



^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time
  2020-01-16  2:07 [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time Zhike Wang
@ 2020-02-04 18:16 ` Maxime Coquelin
  2020-02-05  9:46 ` Maxime Coquelin
  1 sibling, 0 replies; 3+ messages in thread
From: Maxime Coquelin @ 2020-02-04 18:16 UTC (permalink / raw)
  To: Zhike Wang, dev; +Cc: tiwei.bie, zhihong.wang, stable



On 1/16/20 3:07 AM, Zhike Wang wrote:
> The vhost_user_read_cb() and rte_vhost_driver_unregister() can be
> called at the same time by 2 threads. Eg thread1 calls vhost_user_read_cb()
> and removes the vsocket from conn_list, then thread2 calls
> rte_vhost_driver_unregister() and frees the vsocket since it is NOT in the
> conn_list. So thread1 will access invalid memory when trying to reconnect.
> 
> The fix is to move the "removing of vsocket from conn_list" to end of the
> vhost_user_read_cb(), then avoid the race condition.
> 
> The core trace is
> Program terminated with signal 11, Segmentation fault.

Need to add the Fixes tag;

Fixes: af1475918124 ("vhost: introduce API to start a specific driver")
Cc: stable@dpdk.org

> Signed-off-by: Zhike Wang <wangzhike@jd.com>
> ---
>  lib/librte_vhost/socket.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/lib/librte_vhost/socket.c b/lib/librte_vhost/socket.c
> index ebb2ff6..01a9dec 100644
> --- a/lib/librte_vhost/socket.c
> +++ b/lib/librte_vhost/socket.c
> @@ -318,16 +318,16 @@ struct vhost_user {
>  
>  		vhost_destroy_device(conn->vid);
>  
> +		if (vsocket->reconnect) {
> +			create_unix_socket(vsocket);
> +			vhost_user_start_client(vsocket);
> +		}
> +
>  		pthread_mutex_lock(&vsocket->conn_mutex);
>  		TAILQ_REMOVE(&vsocket->conn_list, conn, next);
>  		pthread_mutex_unlock(&vsocket->conn_mutex);
>  
>  		free(conn);
> -
> -		if (vsocket->reconnect) {
> -			create_unix_socket(vsocket);
> -			vhost_user_start_client(vsocket);
> -		}
>  	}
>  }
>  
> 

No need to send v2, I'll add it when applying to my tree.

Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Thanks,
Maxime


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time
  2020-01-16  2:07 [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time Zhike Wang
  2020-02-04 18:16 ` Maxime Coquelin
@ 2020-02-05  9:46 ` Maxime Coquelin
  1 sibling, 0 replies; 3+ messages in thread
From: Maxime Coquelin @ 2020-02-05  9:46 UTC (permalink / raw)
  To: Zhike Wang, dev; +Cc: tiwei.bie, zhihong.wang, stable



On 1/16/20 3:07 AM, Zhike Wang wrote:
> The vhost_user_read_cb() and rte_vhost_driver_unregister() can be
> called at the same time by 2 threads. Eg thread1 calls vhost_user_read_cb()
> and removes the vsocket from conn_list, then thread2 calls
> rte_vhost_driver_unregister() and frees the vsocket since it is NOT in the
> conn_list. So thread1 will access invalid memory when trying to reconnect.
> 
> The fix is to move the "removing of vsocket from conn_list" to end of the
> vhost_user_read_cb(), then avoid the race condition.
> 
> The core trace is
> Program terminated with signal 11, Segmentation fault.
> 
> Signed-off-by: Zhike Wang <wangzhike@jd.com>
> ---
>  lib/librte_vhost/socket.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)

Applied to dpdk-next-virtio tree.

Thanks,
Maxime


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2020-02-05  9:46 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-01-16  2:07 [dpdk-dev] [PATCH] vhost:fix crash on port deletion and VM shutdown at same time Zhike Wang
2020-02-04 18:16 ` Maxime Coquelin
2020-02-05  9:46 ` Maxime Coquelin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).