Skip to contentSkip to conversation

hw/vfio-user: wait for proxy close correctly

PATCH Superseded John Levon 3 participants 5 messages Jul 10, 2025 – Jul 15, 2025

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

John Levon

author
#0 lore raw

PATCH 1/1 hw/vfio-user: wait for proxy close correctly

Coverity reported:

CID 1611806: Concurrent data access violations (BAD_CHECK_OF_WAIT_COND)

A wait is performed without a loop. If there is a spurious wakeup, the condition may not be satisfied.

Fix this by checking ->state for VFIO_PROXY_CLOSED in a loop.

---
 hw/vfio-user/proxy.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
hw/vfio-user/proxy.c +6 -4
index c418954440..2275d3fe39 100644
@@ -32,7 +32,6 @@ static void vfio_user_recycle(VFIOUserProxy *proxy, VFIOUserMsg *msg);
 
 static void vfio_user_recv(void *opaque);
 static void vfio_user_send(void *opaque);
-static void vfio_user_cb(void *opaque);
 
 static void vfio_user_request(void *opaque);
 
@@ -492,7 +491,7 @@ static void vfio_user_send(void *opaque)
     }
 }
 
-static void vfio_user_cb(void *opaque)
+static void vfio_user_close_cb(void *opaque)
 {
     VFIOUserProxy *proxy = opaque;
 
@@ -984,8 +983,11 @@ void vfio_user_disconnect(VFIOUserProxy *proxy)
      * handler to run after the proxy fd handlers were
      * deleted above.
      */
-    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_cb, proxy);
-    qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_close_cb, proxy);
+
+    while (proxy->state != VFIO_PROXY_CLOSED) {
+        qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    }
 
     /* we now hold the only ref to proxy */
     qemu_mutex_unlock(&proxy->lock);
Signature · 1 line
2.43.0
Reviewed-by on v1 1/1
#1 lore raw

On 10/07/2025 16:47, John Levon wrote:

(added Cedric)

Quoted 39 lines

Coverity reported:

CID 1611806: Concurrent data access violations (BAD_CHECK_OF_WAIT_COND)

A wait is performed without a loop. If there is a spurious wakeup, the condition may not be satisfied.

Fix this by checking ->state for VFIO_PROXY_CLOSED in a loop.

---
  hw/vfio-user/proxy.c | 10 ++++++----
  1 file changed, 6 insertions(+), 4 deletions(-)
hw/vfio-user/proxy.c +1 -2
index c418954440..2275d3fe39 100644
@@ -32,7 +32,6 @@ static void vfio_user_recycle(VFIOUserProxy *proxy, VFIOUserMsg *msg);
 
  static void vfio_user_recv(void *opaque);
  static void vfio_user_send(void *opaque);
-static void vfio_user_cb(void *opaque);
 
  static void vfio_user_request(void *opaque);
 
@@ -492,7 +491,7 @@ static void vfio_user_send(void *opaque)
      }
  }
 
-static void vfio_user_cb(void *opaque)
+static void vfio_user_close_cb(void *opaque)
  {
      VFIOUserProxy *proxy = opaque;
 
@@ -984,8 +983,11 @@ void vfio_user_disconnect(VFIOUserProxy *proxy)
       * handler to run after the proxy fd handlers were
       * deleted above.
       */

hw/vfio-user/proxy.c void vfio_user_disconnect(VFIOUserProxy *proxy) · line 993

Quoted 10 lines
hw/vfio-user/proxy.c +5 -2
-    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_cb, proxy);
-    qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_close_cb, proxy);
+
+    while (proxy->state != VFIO_PROXY_CLOSED) {
+        qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    }
 
      /* we now hold the only ref to proxy */
      qemu_mutex_unlock(&proxy->lock);

It think it is worth mentioning the function rename in the commit message, otherwise looks good to me:

ATB,

Mark.

Reviewed-by
#2 lore raw
Quoted 3 lines · Mark Cave-Ayland

On 10/07/2025 16:47, John Levon wrote:

(added Cedric)

ah. Thanks Mark.

This reminds me that we should have maintainers/reviewers that can send PRs for the vfio-user component.

John,

Could you please send a patch adding me and Mark may be ?

Quoted 4 lines
Quoted 4 lines

Coverity reported:

CID 1611806: Concurrent data access violations (BAD_CHECK_OF_WAIT_COND)

Please prefer :

Resolves: Coverity CID 1611805

Quoted 43 lines
Quoted 30 lines

A wait is performed without a loop. If there is a spurious wakeup, the condition may not be satisfied.

Fix this by checking ->state for VFIO_PROXY_CLOSED in a loop.

---
  hw/vfio-user/proxy.c | 10 ++++++----

  1 file changed, 6 insertions(+), 4 deletions(-)
hw/vfio-user/proxy.c +1 -2
index c418954440..2275d3fe39 100644
@@ -32,7 +32,6 @@ static void vfio_user_recycle(VFIOUserProxy *proxy, VFIOUserMsg *msg);
   static void vfio_user_recv(void *opaque);
   static void vfio_user_send(void *opaque);
-static void vfio_user_cb(void *opaque);
   static void vfio_user_request(void *opaque);
@@ -492,7 +491,7 @@ static void vfio_user_send(void *opaque)
       }
   }
-static void vfio_user_cb(void *opaque)
+static void vfio_user_close_cb(void *opaque)
   {
       VFIOUserProxy *proxy = opaque;
@@ -984,8 +983,11 @@ void vfio_user_disconnect(VFIOUserProxy *proxy)
        * handler to run after the proxy fd handlers were
        * deleted above.

hw/vfio-user/proxy.c void vfio_user_disconnect(VFIOUserProxy *proxy) · line 992

Quoted 10 lines
hw/vfio-user/proxy.c +5 -2
        */
-    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_cb, proxy);
-    qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    aio_bh_schedule_oneshot(proxy->ctx, vfio_user_close_cb, proxy);
+
+    while (proxy->state != VFIO_PROXY_CLOSED) {
+        qemu_cond_wait(&proxy->close_cv, &proxy->lock);
+    }
       /* we now hold the only ref to proxy */
       qemu_mutex_unlock(&proxy->lock);

It think it is worth mentioning the function rename in the commit message, otherwise looks good to me:

Quoted 1 line

Thanks,

C.

#3 lore raw
Quoted 5 lines · Cédric Le Goater
Quoted 3 lines · Mark Cave-Ayland

On 10/07/2025 16:47, John Levon wrote:

(added Cedric)

Quoted 8 lines

ah. Thanks Mark.

This reminds me that we should have maintainers/reviewers that can send PRs for the vfio-user component.

John,

Could you please send a patch adding me and Mark may be ?

I'm still coming up to speed with the internals of QEMU's vfio implementation, so this might be a bit premature. But I'd like to think I will get there soon :)

BTW I have a series here that fixes up some of the QOM issues with vfio, but I'm guessing it's a little too close to freeze for this?

ATB,

Mark.

#4 lore raw
Quoted 1 line

BTW I have a series here that fixes up some of the QOM issues with vfio, but I'm guessing it's a little too close to freeze for this?

Please send and we'll see.

Thanks,

C.