netdev - Re: [PATCH] virtio_net: add local_bh_disable() around u64_stats_update

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

Message-Id: <20181018.162308.2295937118791060714.davem@davemloft.net>
Date:   Thu, 18 Oct 2018 16:23:08 -0700 (PDT)
From:   David Miller <davem@...emloft.net>
To:     bigeasy@...utronix.de
Cc:     stephen@...workplumber.org, netdev@...r.kernel.org,
        virtualization@...ts.linux-foundation.org, tglx@...utronix.de,
        makita.toshiaki@....ntt.co.jp, mst@...hat.com, jasowang@...hat.com
Subject: Re: [PATCH] virtio_net: add local_bh_disable() around
 u64_stats_update_begin

From: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Date: Thu, 18 Oct 2018 10:43:13 +0200

> on 32bit, lockdep notices that virtnet_open() and refill_work() invoke
> try_fill_recv() from process context while virtnet_receive() invokes the
> same function from BH context. The problem that the seqcounter within
> u64_stats_update_begin() may deadlock if it is interrupted by BH and
> then acquired again.
> 
> Introduce u64_stats_update_begin_bh() which disables BH on 32bit
> architectures. Since the BH might interrupt the worker, this new
> function should not limited to SMP like the others which are expected
> to be used in softirq.
> 
> With this change we might lose increments but this is okay. The
> important part that the two 32bit parts of the 64bit counter are not
> corrupted.
> 
> Fixes: 461f03dc99cf6 ("virtio_net: Add kick stats").
> Suggested-by: Stephen Hemminger <stephen@...workplumber.org>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>

Trying to get down to the bottom of this:

1) virtnet_receive() runs from softirq but only if NAPI is active and
   enabled.  It is in this context that it invokes try_fill_recv().

2) refill_work() runs from process context, but disables NAPI (and
   thus invocation of virtnet_receive()) before calling
   try_fill_recv().

3) virtnet_open() invokes from process context as well, but before the
   NAPI instances are enabled, it is same as case #2.

4) virtnet_restore_up() is the same situations as #3.

Therefore I agree that this is a false positive, and simply lockdep
cannot see the NAPI synchronization done by case #2.

I think we shouldn't add unnecessary BH disabling here, and instead
find some way to annotate this for lockdep's sake.

Thank you.