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] [day] [month] [year] [list]
Date:   Fri, 16 Sep 2022 10:52:46 +0200
From:   "Arnd Bergmann" <arnd@...db.de>
To:     "Tomi Valkeinen" <tomi.valkeinen@...asonboard.com>,
        "kernel test robot" <lkp@...el.com>
Cc:     kbuild-all@...ts.01.org, linux-kernel@...r.kernel.org,
        "Linus Walleij" <linus.walleij@...aro.org>,
        Linux-OMAP <linux-omap@...r.kernel.org>,
        "Tomi Valkeinen" <tomi.valkeinen@...com>,
        dri-devel@...ts.freedesktop.org
Subject: Re: drivers/gpu/drm/omapdrm/dss/dsi.c:1126:1: warning: the frame size of 1060
 bytes is larger than 1024 bytes

On Fri, Sep 16, 2022, at 10:24 AM, Tomi Valkeinen wrote:
> On 15/09/2022 21:49, Arnd Bergmann wrote:
>> On Thu, Sep 15, 2022, at 8:17 PM, kernel test robot wrote:
>> I think hte problem is that struct dsi_irq_stats is just too
>> large, at 776 bytes. The interrupts are disabled during a copy
>> from 'dsi->irq_stats' into 'stats'. A trivial workaround would
>> avoid the local copy and keep interrupts disabled through
>> the entire function so it can operate directly on the source
>> data, but that would introduce a longer time with irqs disabled,
>> which might be bad as well.
>> 
>> Since this is only called from a debugfs file, and reading that
>> file is probably not performance critical itself, maybe
>> using kmalloc on the large structure would be best.
>
> I think that makes sense. I have sent a patch using kmalloc.
>
> Oddly enough, I was not able to reproduce the warning with my normal 
> toolchain, gcc-arm-11.2-2022.02-x86_64-arm-none-linux-gnueabihf. I even 
> reduced the frame size limit to 700, and saw warnings from other places, 
> but not from omapdrm.

I had another look and found that this only happens with
CONFIG_INIT_STACK_ALL_PATTERN=y or CONFIG_INIT_STACK_ALL_ZERO=y,
which are only available with gcc-12.x or clang.

It looks like without that, gcc can reduce the size of the
on-stack variable by only copying the members that it actually
needs, see https://pastebin.com/8dDRE1bX for the gcc-11
output.

      Arnd

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ