From: Eugene Shatokhin Subject: Re: [PATCH] usbnet: Fix two races between usbnet_stop() and the BH Date: Wed, 19 Aug 2015 10:57:09 +0300 Message-ID: <55D436D5.6010105@rosalab.ru> References: <55CE1D7E.2070400@rosalab.ru> <1439571516-11862-1-git-send-email-eugene.shatokhin@rosalab.ru> <20150818.185407.1667358232705414236.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: oneukum-IBi9RG/b67k@public.gmane.org, netdev-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-usb-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: David Miller Return-path: In-Reply-To: <20150818.185407.1667358232705414236.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org> Sender: linux-usb-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: netdev.vger.kernel.org 19.08.2015 04:54, David Miller =D0=BF=D0=B8=D1=88=D0=B5=D1=82: > From: Eugene Shatokhin > Date: Fri, 14 Aug 2015 19:58:36 +0300 > >> 2. The second race is on dev->flags. >> >> dev->flags is set to 0 here: >> *0 usbnet_stop (usbnet.c:816) >> /* deferred work (task, timer, softirq) must also stop. >> * can't flush_scheduled_work() until we drop rtnl (later), >> * else workers could deadlock; so make workers a NOP. >> */ >> dev->flags =3D 0; >> del_timer_sync (&dev->delay); >> tasklet_kill (&dev->bh); >> >> And here, the code clears EVENT_RX_KILL bit in dev->flags, which may >> execute concurrently with the above operation: >> *0 clear_bit (bitops.h:113, inlined) >> *1 usbnet_bh (usbnet.c:1475) >> /* restart RX again after disabling due to high error rate */ >> clear_bit(EVENT_RX_KILL, &dev->flags); >> >> It seems, setting dev->flags to 0 is not necessarily atomic w.r.t. >> clear_bit() and other bit operations with dev->flags. It is safer to >> make it atomic and this way, make the race harmless. >> >> While at it, the checking of EVENT_NO_RUNTIME_PM bit of dev->flags i= n >> usbnet_stop() was fixed too: the bit should be checked before dev->f= lags >> is cleared. > > The fix for this is excessive. > > Instead of all of this madness, looping over expensive clear_bit() > atomics, just do whatever it takes to make sure that usbnet_bh() is > quiesced and cannot execute any more. Then you can safely clear > dev->flags normally. > If I understand it correctly, it is to make sure usbnet_bh() is not=20 scheduled again that dev->flags should be set to 0 first, one way or=20 another. That is what this madness is for. tasklet_kill() will wait then for the already running instance of=20 usbnet_bh() (if one is running). After that, it is guaranteed BH is not= =20 running and will not be re-scheduled. As for the performance concerns, I doubt that usbnet_stop() is anywhere= =20 on the critical path. I have been testing this patch for some time and=20 haven't seen any new performance issues with it yet. If needed, it is possible to measure and compare the time needed for=20 usbnet_stop() before and after this patch and try to estimate the impac= t=20 of this on the overall performance. Regards, Eugene -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html