librdmacm: Fix verbs leak due to reentrancy issue

Sean,

I was out of work so I tested the patch "librdmacm: lazy initialization for ib devices" today. The patch works fine for my test case. The only concern I have is that I think the patch might leak verbs context. I think that any call to ucma_init_device must be done with mutex locked. Please correct me if I am wrong.

Thanks, Shamir

---
 src/cma.c |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

index 0dc229e..42769cf 100644
@@ -416,10 +416,13 @@ static int ucma_get_device(struct cma_id_private *id_priv, uint64_t guid)
 
 	return ERR(ENODEV);
 match:
-	if ((ret = ucma_init_device(cma_dev)))
+	pthread_mutex_lock(&mut);
+
+	if ((ret = ucma_init_device(cma_dev))) {
+		pthread_mutex_unlock(&mut);
 		return ret;
+	}
 
-	pthread_mutex_lock(&mut);
 	if (!cma_dev->refcnt++) {
 		cma_dev->pd = ibv_alloc_pd(cma_dev->verbs);
 		if (!cma_dev->pd) {

1.7.9.5

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

I think you're correct; the lock needs to move up. I'll apply your fix. Thanks! --
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

librdmacm git (git://git.openfabrics.org/~shefty/librdmacm.git)
currently only has the lazy initialization patch (23ffef0 librdmacm:
Support lazy initialization) but this patch is missing. can you please add this fix to the git if you approve it? thanks.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

I pushed the fix upstream. Thanks!
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html