netdev - Re: [PATCH] net: use READ_ONCE() to read in concurrent environment

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <8f596e31e1d24f418fa1c6b1a2bae5fc00746e33.camel@redhat.com>
Date: Tue, 23 Jan 2024 09:15:18 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: linke li <lilinke99@...com>, netdev@...r.kernel.org
Subject: Re: [PATCH] net: use READ_ONCE() to read in concurrent environment

On Tue, 2024-01-23 at 04:24 +0800, linke li wrote:
> In function sk_stream_wait_memory(), reads of sk->sk_err and sk->sk_shutdown
> is protected using READ_ONCE() in line 145, 146.
> 145: 		ret = sk_wait_event(sk, &current_timeo, READ_ONCE(sk->sk_err) ||
> 146: 				    (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN) ||

The above read happens outside the sk socket lock, see the
sk_wait_event() macro definition...

> 
> But reads in line 133 are not protected. 

... while the above one happens under the sk socket lock. The _ONCE
annotation is not needed.


> This may cause unexpected error
> when other threads change sk->sk_err and sk->sk_shutdown. Function
> sk_stream_wait_connect() has same problem.

Same as above, the access with the _ONCE() annotation is outside the
socket lock and vice versa.


I think this patch is not needed.

Thanks,

Paolo