netdev - Re: [PATCH net] bnx2x: bail out if unable to acquire stats

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: <CADVnQyn-i1SsdqPR1SN5pqbE6sFFD-5_HLqVa6hXmCu2-f8h1Q@mail.gmail.com>
Date:	Fri, 6 Sep 2013 23:53:32 -0400
From:	Neal Cardwell <ncardwell@...gle.com>
To:	Dmitry Kravkov <dkravkov@...il.com>
Cc:	Dmitry Kravkov <dmitry@...adcom.com>,
	Michal Schmidt <mschmidt@...hat.com>,
	"davem@...emloft.net" <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Ariel Elior <ariele@...adcom.com>,
	Eilon Greenstein <eilong@...adcom.com>,
	Havard Skinnemoen <hskinnemoen@...gle.com>,
	Eric Dumazet <edumazet@...gle.com>
Subject: Re: [PATCH net] bnx2x: bail out if unable to acquire stats_sema

On Fri, Sep 6, 2013 at 2:40 AM, Dmitry Kravkov <dkravkov@...il.com> wrote:
> On Wed, Sep 4, 2013 at 8:17 PM, Neal Cardwell <ncardwell@...gle.com> wrote:
>> On Tue, Sep 3, 2013 at 11:51 AM, Dmitry Kravkov <dmitry@...adcom.com> wrote:
>>>> -----Original Message-----
>>>> From: Michal Schmidt [mailto:mschmidt@...hat.com]
>>>> Sent: Tuesday, September 03, 2013 6:46 PM
>>>> To: davem@...emloft.net
>>>> Cc: netdev@...r.kernel.org; Dmitry Kravkov; Ariel Elior; Eilon Greenstein
>>>> Subject: [PATCH net] bnx2x: bail out if unable to acquire stats_sema
>>>>
>>>> If we fail to acquire stats_sema in the specified time limit, the chip is
>>>> probably dead. It probably does not matter whether we try to continue or
>>>> not, but certainly we should not up() the semaphore afterwards.
>>>>
>>>> Signed-off-by: Michal Schmidt <mschmidt@...hat.com>
>>>> ---
>>>>  drivers/net/ethernet/broadcom/bnx2x/bnx2x_stats.c | 24
>>>> +++++++++++++++++------
>>
>> It seems like this patch has the downside that if the down_timeout()
>> fails, then bnx2x_stats_handle() ends up updating the stats state
>> machine's state without really executing the real body of the
>> action().
>>
>> In fact it seems like there is a more general pre-existing problem of
>> this flavor with the bnx2x stats state machine: the
>> bnx2x_stats_handle() function updates the state machine
>> bp->stats_state while holding the spin lock, but does not execute the
>> action() while holding any sort of synchronization, so AFAICT there is
>> nothing to guarantee that the state machine actions happen in the
>> order the state machine wants them to happen. For example, if stats
>> events fire such that we want to execute actions that disable and then
>> enable stats, we could instead end up executing the actions in the
>> order that would attempt to enable and then disable them, if we get
>> unlucky with respect to when interrupts fire, etc.
>>
>> It seems to me that instead of having all of the callees of
>> bnx2x_stats_handle() try to down/up the semaphore, instead
>> bnx2x_stats_handle() should try to down the stats_sema at the top, and
>> then if successful, it should change the bp->stats_state, call the
>> action, and up the stats_sema. Would that work?
>>
> handle() is called from sleepable context (open/close) and timer
> context, then it's not possible to use semaphore for pretection

But it seems like all of the callees of bnx2x_stats_handle() are
already using the stats_sema for protection, and the only difference
is that bnx2x_stats_update uses down_trylock and the other callees use
down_timeout. What about making this explicit in bnx2x_stats_handle,
with something like:

if (event == STATS_EVENT_UPDATE) {
  if (down_trylock(&bp->stats_sema)) {
       BNX2X_ERR("stats down_trylock failed\n");
       goto out;
  }
else {
  if (down_timeout(&bp->stats_sema, HZ/10)) {
       BNX2X_ERR("stats down_timeout failed\n");
       goto out;
  }
}
bp->stats_state = ...;
action = ...;
action(bp);
up(&bp->stats_sema);

That would allow us to protect the bnx2x state machine so that the
action() events happen in the correct order, and we don't (e.g.)
accidentally end up doing an enable-then-disable when we wanted
disable-then-enable.

Would that work?

neal


>> neal
>> --
>> To unsubscribe from this list: send the line "unsubscribe netdev" in
>> the body of a message to majordomo@...r.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>
>
> --
> Best regards,
> Dmitry
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html