Blob Blame Raw
From e36a5a8613df42339773ebf48e07d063ad7484e8 Mon Sep 17 00:00:00 2001
From: Fam Zheng <famz@redhat.com>
Date: Mon, 5 Sep 2016 01:18:15 +0200
Subject: [PATCH] nbd-server: Set O_NONBLOCK on client fd

RH-Author: Fam Zheng <famz@redhat.com>
Message-id: <1473038295-7193-1-git-send-email-famz@redhat.com>
Patchwork-id: 72141
O-Subject: [RHEL-7.3 qemu-kvm PATCH] nbd-server: Set O_NONBLOCK on client fd
Bugzilla: 1285453
RH-Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
RH-Acked-by: Paolo Bonzini <pbonzini@redhat.com>
RH-Acked-by: Max Reitz <mreitz@redhat.com>

Upstream: upstream uses IO channels that is not present in downstream.
Backporting that deserves a separate and deliberate justification BZ,
for 7.4.

Even with 2a68d80 (nbd-server: Coroutine based negotiation), QEMU still
hangs when client hangs, because recvmsg the socket fd is blocking. Set
the O_NONBLOCK to fix this.

Analyzed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
---
 nbd.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/nbd.c b/nbd.c
index c20e57e..8a32e18 100644
--- a/nbd.c
+++ b/nbd.c
@@ -1281,6 +1281,7 @@ static coroutine_fn void nbd_co_client_start(void *opaque)
     if (exp) {
         nbd_export_get(exp);
     }
+    qemu_set_nonblock(client->sock);
     if (nbd_negotiate(data)) {
         nbd_client_close(client);
         goto out;
-- 
1.8.3.1