From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail-it0-f65.google.com ([209.85.214.65]:35744 "EHLO mail-it0-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750848AbdGGOQA (ORCPT ); Fri, 7 Jul 2017 10:16:00 -0400 MIME-Version: 1.0 In-Reply-To: References: <87bmowd0mh.fsf@kamboji.qca.qualcomm.com> From: Geert Uytterhoeven Date: Fri, 7 Jul 2017 16:15:58 +0200 Message-ID: (sfid-20170707_161631_133941_403D543B) Subject: Re: ath10k: ret used but uninitialized To: Arnd Bergmann Cc: Kalle Valo , Erik Stromdahl , "ath10k@lists.infradead.org" , linux-wireless , "netdev@vger.kernel.org" , Linux Kernel Mailing List Content-Type: text/plain; charset="UTF-8" Sender: linux-wireless-owner@vger.kernel.org List-ID: Hi Arnd, On Fri, Jul 7, 2017 at 4:14 PM, Arnd Bergmann wrote: > On Fri, Jul 7, 2017 at 12:04 PM, Kalle Valo wrot= e: >> Erik Stromdahl writes: >>>> With gcc 4.1.2: >>>> >>>> drivers/net/wireless/ath/ath10k/sdio.c: In function >>>> =E2=80=98ath10k_sdio_mbox_rxmsg_pending_handler=E2=80=99: >>>> drivers/net/wireless/ath/ath10k/sdio.c:676: warning: =E2=80=98ret=E2= =80=99 may be used >>>> uninitialized in this function >>>> >>>>> + >>>>> + *done =3D true; >>>>> + >>>>> + /* Copy the lookahead obtained from the HTC register table in= to our >>>>> + * temp array as a start value. >>>>> + */ >>>>> + lookaheads[0] =3D msg_lookahead; >>>>> + >>>>> + timeout =3D jiffies + SDIO_MBOX_PROCESSING_TIMEOUT_HZ; >>>> >>>> Although very unlikely due to the long timeout, if the code is preempt= ed here, >>>> and the loop below never entered, ret will indeed be uninitialized. >>>> >>>> It's unclear to me what the proper initialization would be, though, so >>>> that's why I didn't send a patch. >>>> >>> I think it would be best to use 0 as initial value of ret in this case. >>> This will make all other interrupts be processed in a normal way. >>> >>> Kalle: Should I create a new patch (initializing ret with zero)? >> >> Yes, please send a new patch fixing this. >> >> But I don't like that much with the style of initialising ret to zero, >> it tends to hide things. Instead my preference is something like below >> where the error handling is more explicit and easier to find where it's >> exactly failing. But that's just an example how I would try to solve it, >> it still lacks the handling of -ECANCEL etc. > > I think I would simply replace the "while() {}" loop with "do{} while()", > as that would guarantee it to be run at least once in a way that the > compiler can see. Right, that's probably the simplest and cleanest solution. Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k= .org In personal conversations with technical people, I call myself a hacker. Bu= t when I'm talking to journalists I just say "programmer" or something like t= hat. -- Linus Torvalds