Re: 319554f284dd ("inet: don't use sk_v6_rcv_saddr directly") causes bind port regression

[Josef Bacik <jbacik@fb.com>]

[-- Attachment #1: Type: text/plain, Size: 731 bytes --]

Alright thanks, this should fix it.

Josef

On 9/13/17, 12:14 PM, "Cole Robinson" <crobinso@redhat.com> wrote:

On 09/13/2017 01:40 PM, Cole Robinson wrote:
> On 09/13/2017 01:28 PM, Josef Bacik wrote:
>> Sorry I thought I had made this other fix, can you apply this on top of the other one and try that?  I have more things to try if this doesn’t work, sorry you are playing go between, but I want to make sure I know _which_ fix actually fixes the problem, and then clean up in followup patches.  Thanks,
>>
> 
> I'm the bug reporter. I'll combine the two patches and report back
> 

Nope, issue is still present with both patches applied. Tried my own build and
a package Laura provided

Thanks,
Cole




[-- Attachment #2: 0001-net-don-t-fast-patch-mismatched-sockets-in-STRICT-mo.patch --]
[-- Type: application/octet-stream, Size: 1182 bytes --]

From 8d3da6d30a7c63034d4ff7d4fc33ea9e3f23cf41 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@toxicpanda.com>
Date: Wed, 13 Sep 2017 12:42:13 -0700
Subject: [PATCH] net: don't fast patch mismatched sockets in STRICT mode

With FASTREUSE_STRICT we may have sockets on the tb that don't match our
fast_sk information, so if our new socket don't strictly match the
fast_sk info we need to not go down the fast path and do the
inet_csk_bind_conflict so we can look at the other sockets on the tb.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 net/ipv4/inet_connection_sock.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index ff8b15a99e42..fe9cf4862de2 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -263,6 +263,8 @@ static inline int sk_reuseport_match(struct inet_bind_bucket *tb,
 	 */
 	if (tb->fastreuseport == FASTREUSEPORT_ANY)
 		return 1;
+	if (tb->fast_ipv6_only && tb->fast_sk_family != sk->sk_family)
+		return 0;
 #if IS_ENABLED(CONFIG_IPV6)
 	if (tb->fast_sk_family == AF_INET6)
 		return ipv6_rcv_saddr_equal(&tb->fast_v6_rcv_saddr,
-- 
2.13.5