[1/2] tcp: fix races in tcp_abort()

Commit Message

Youngmin Nam March 14, 2025, 9:24 a.m. UTC

From: Eric Dumazet <[e-mail]>

tcp_abort() has the same issue than the one fixed in the prior patch
in tcp_write_err().

commit 5ce4645c23cf5f048eb8e9ce49e514bababdee85 upstream.

To apply commit bac76cf89816bff06c4ec2f3df97dc34e150a1c4,
this patch must be applied first.

In order to get consistent results from tcp_poll(), we must call
sk_error_report() after tcp_done().

We can use tcp_done_with_error() to centralize this logic.

Fixes: c1e64e298b8c ("net: diag: Support destroying TCP sockets.")
Signed-off-by: Eric Dumazet <[e-mail]>
Acked-by: Neal Cardwell <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>
Cc: <[e-mail]> # v5.10+
[youngmin: Resolved minor conflict in net/ipv4/tcp.c]
Signed-off-by: Youngmin Nam <[e-mail]>
---
 net/ipv4/tcp.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

Comments

Greg KH March 14, 2025, 12:24 p.m. UTC | #1

On Fri, Mar 14, 2025 at 06:24:45PM +0900, Youngmin Nam wrote:
> From: Eric Dumazet <[e-mail]>
> 
> tcp_abort() has the same issue than the one fixed in the prior patch
> in tcp_write_err().
> 
> commit 5ce4645c23cf5f048eb8e9ce49e514bababdee85 upstream.
> 
> To apply commit bac76cf89816bff06c4ec2f3df97dc34e150a1c4,
> this patch must be applied first.
> 
> In order to get consistent results from tcp_poll(), we must call
> sk_error_report() after tcp_done().
> 
> We can use tcp_done_with_error() to centralize this logic.
> 
> Fixes: c1e64e298b8c ("net: diag: Support destroying TCP sockets.")
> Signed-off-by: Eric Dumazet <[e-mail]>
> Acked-by: Neal Cardwell <[e-mail]>
> Link: https://lore.kernel.org/r/[e-mail]
> Signed-off-by: Jakub Kicinski <[e-mail]>
> Cc: <[e-mail]> # v5.10+

Did not apply to 5.10.y, what did you want this added to?

thanks,

greg k-h

Youngmin Nam March 17, 2025, 4:36 a.m. UTC | #2

On Fri, Mar 14, 2025 at 01:24:09PM +0100, Greg KH wrote:
> On Fri, Mar 14, 2025 at 06:24:45PM +0900, Youngmin Nam wrote:
> > From: Eric Dumazet <[e-mail]>
> > 
> > tcp_abort() has the same issue than the one fixed in the prior patch
> > in tcp_write_err().
> > 
> > commit 5ce4645c23cf5f048eb8e9ce49e514bababdee85 upstream.
> > 
> > To apply commit bac76cf89816bff06c4ec2f3df97dc34e150a1c4,
> > this patch must be applied first.
> > 
> > In order to get consistent results from tcp_poll(), we must call
> > sk_error_report() after tcp_done().
> > 
> > We can use tcp_done_with_error() to centralize this logic.
> > 
> > Fixes: c1e64e298b8c ("net: diag: Support destroying TCP sockets.")
> > Signed-off-by: Eric Dumazet <[e-mail]>
> > Acked-by: Neal Cardwell <[e-mail]>
> > Link: https://lore.kernel.org/r/[e-mail]
> > Signed-off-by: Jakub Kicinski <[e-mail]>
> > Cc: <[e-mail]> # v5.10+
> 
> Did not apply to 5.10.y, what did you want this added to?
> 
> thanks,
> 
> greg k-h
> 

Hi Greg,

Sorry about that.

As for 5.10, it seems to have more dependencies for the backport.
I think the maintainer should handle it to ensure a safe backport.

Patch

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7ad82be40f34..9fe164aa185c 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -4630,13 +4630,9 @@  int tcp_abort(struct sock *sk, int err)
 	bh_lock_sock(sk);
 
 	if (!sock_flag(sk, SOCK_DEAD)) {
-		WRITE_ONCE(sk->sk_err, err);
-		/* This barrier is coupled with smp_rmb() in tcp_poll() */
-		smp_wmb();
-		sk_error_report(sk);
 		if (tcp_need_reset(sk->sk_state))
 			tcp_send_active_reset(sk, GFP_ATOMIC);
-		tcp_done(sk);
+		tcp_done_with_error(sk, err);
 	}
 
 	bh_unlock_sock(sk);