netdev - Re: [PATCH net-next 7/7] r8169: remove work from irq handler.

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: <20120128111121.GA32163@electric-eye.fr.zoreil.com>
Date:	Sat, 28 Jan 2012 12:11:21 +0100
From:	Francois Romieu <romieu@...zoreil.com>
To:	Michał Mirosław <mirqus@...il.com>
Cc:	davem@...emloft.net, netdev@...r.kernel.org,
	Hayes Wang <hayeswang@...ltek.com>
Subject: Re: [PATCH net-next 7/7] r8169: remove work from irq handler.

Michał Mirosław <mirqus@...il.com> :
> 2012/1/27 Francois Romieu <romieu@...zoreil.com>:
[...]
> > +static void rtl_schedule_task(struct rtl8169_private *tp, enum rtl_flag flag)
> > +{
> > +       spin_lock(&tp->lock);
> > +       if (!test_and_set_bit(flag, tp->wk.flags))
> > +               schedule_work(&tp->wk.work);
> > +       spin_unlock(&tp->lock);
> > +}
> > +
> > +static void rtl_schedule_task_bh(struct rtl8169_private *tp, enum rtl_flag flag)
> > +{
> > +       local_bh_disable();
> > +       rtl_schedule_task(tp, flag);
> > +       local_bh_enable();
> > +}
> > +
> 
> It might be enough to do:
> 
> rtl_schedule_task():
> 
> set_bit(flag, tp->wk.flags);
> schedule_work(&tp->wk.work);
> 
> This will guarantee that the work is done at least once (twice in
> unlikely case that it was being executed just before schedule_work()).

If I understand correctly this remark and the following ones, you are
suggesting to completely avoid the lock to access the bitfield ? 

[...]
> > +/*
> > + * Workqueue context.
> > + */
> > +static void rtl_slow_event_work(struct rtl8169_private *tp)
> > +{
[...]
> > +       if (unlikely(status & RxFIFOOver)) {
> > +               switch (tp->mac_version) {
> > +               /* Work around for rx fifo overflow */
> > +               case RTL_GIGA_MAC_VER_11:
> > +                       netif_stop_queue(dev);
> > +                       rtl_schedule_task_bh(tp, RTL_FLAG_TASK_RESET_PENDING);
> 
> Since this is running from the same task it schedules, it should be
> enough to set_bit() and ensure that rtl_slow_event_work() is first on
> the list in rtl_task().

I saw "first in list" as mildly maintenable. Ok, I buy it if nobody complains.

-- 
Ueimor
--
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