On 2022/11/20 0:13, Chuck Lever III wrote:

quoted

On Nov 18, 2022, at 11:34 PM, Zhang Xiaoxu [off-list ref] wrote:

If rdma receive buffer allocate failed, should call rpcrdma_regbuf_free()
to free the send buffer, otherwise, the buffer data will be leaked.

Fixes: 8cec3dba76a4 ("xprtrdma: rpcrdma_regbuf alignment")

Actually Fixes: bb93a1ae2bf4 ("xprtrdma: Allocate req's regbufs
at xprt create time") might be better. bb93a1ae2bf4 is the commit
that incorrectly added the kfree() to rpcrdma_req_create().
Even though 8cec3dba76a4 is what split the regbufs into two
allocations, your fix isn't applicable to the rpcrdma_req_create()
that's in 8cec3dba76a4.
Thanks, I will change the fix tag and add your reviewed-by in v2.

quoted

Signed-off-by: Zhang Xiaoxu <redacted>

I'm guessing the error path is getting exercised more once
f20f18c95630 ("xprtrdma: Prevent memory allocations from driving
a reclaim") has been applied. Good catch!

Reviewed-by: Chuck Lever <redacted>

quoted

---
net/sunrpc/xprtrdma/verbs.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 44b87e4274b4..b098fde373ab 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c

@@ -831,7 +831,7 @@ struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt,

	return req;

out3:
-	kfree(req->rl_sendbuf);
+	rpcrdma_regbuf_free(req->rl_sendbuf);
out2:
	kfree(req);
out1:
-- 
2.31.1

--
Chuck Lever