linux-kernel - Re: Bug Report: 2cc7b6 exposes violations of type limits

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <YJ079LaADbsxWnbu@yury-ThinkPad>
Date:   Thu, 13 May 2021 07:47:16 -0700
From:   Yury Norov <yury.norov@...il.com>
To:     Max Görner <max@...ilie-goerner.eu>
Cc:     akpm@...ux-foundation.org, linux-kernel@...r.kernel.org
Subject: Re: Bug Report: 2cc7b6 exposes violations of type limits

[CC: LKML]

On Thu, May 13, 2021 at 12:05:46PM +0200, Max Görner wrote:
> Dear Mr Norov, dear Mr Morton,
> 
> with this e-mail I want to report a bug in the kernel to you. I am approaching
> both of you because you are the author and the maintainer, respectively.
> 
> I am the maintainer of a community fork of a WIFI driver (rtl88x2bu,
> https://github.com/cilynx/rtl88x2bu/tree/5.8.7.1_35809.20191129_COEX20191120-7777).
> Usually new kernel versions introduce new incompatibilities which break the
> compilation. I then add a new `#if LINUX_VERSION_CODE` branch in which I fix
> things.
> 
> I am proud to say that the driver compiles with `-O3 -Wall -Werror` with only a
> modest number of `-Wno-...` directives. I usually compile with
> `-O2 -Wall -Werror`, though.
> 
> In the merge window for Linux Kernel v5.13 Mr Norov added commit
> 2cc7b6a44ac21d31b398b03f4845c53152070416 which looks fairly innocent to me.
> However, it exposes some information to GCC, allowing GCC to deduce that some
> comparisons do not make sense. Now GCC complains that there are tests whether
> an `unsigned long` is `< 0`.
> 
> However, these problems appear only when I try to compile the WIFI driver, not
> when compiling the kernel itself. I suspect that the problems would appear if
> the kernel would be compiled with `-Wall -Werror`, but I do not know how to do
> that. I attached the error messages in a separate file.
> 
> I think having type limit violations in the code is bad in itself. It also
> breaks some external code (my driver). With this bug report I want to kindly
> ask you if you could fix these type limit violations.
> 
> 
> Thank you very much.
> 
> 
> Best regards
> 
> Max Görner

> make ARCH=x86_64 CROSS_COMPILE= -C /home/mgoerner/Werkbank/Hobbys/Linux-Kernel/linux M=/home/mgoerner/Werkbank/Hobbys/rtl88x2bu  modules
> make[1]: Verzeichnis „/home/mgoerner/Werkbank/Hobbys/Linux-Kernel/linux“ wird betreten
>   CC [M]  /home/mgoerner/Werkbank/Hobbys/rtl88x2bu/core/rtw_cmd.o
> In file included from ./include/linux/bits.h:22,
>                  from ./include/linux/bitops.h:6,
>                  from ./include/linux/kernel.h:12,
>                  from ./include/linux/list.h:9,
>                  from ./include/linux/module.h:12,
>                  from /home/mgoerner/Werkbank/Hobbys/rtl88x2bu/include/basic_types.h:76,
>                  from /home/mgoerner/Werkbank/Hobbys/rtl88x2bu/include/drv_types.h:26,
>                  from /home/mgoerner/Werkbank/Hobbys/rtl88x2bu/core/rtw_cmd.c:17:
> ./include/asm-generic/bitops/find.h: In function ‘find_first_bit’:
> ./include/linux/bits.h:25:28: error: comparison of unsigned expression in ‘< 0’ is always false [-Werror=type-limits]
>    25 |   __builtin_constant_p((l) > (h)), (l) > (h), 0)))
>       |                            ^

[...]

Kernel disables -Werror=type-limits in scripts/Makefile.extrawarn. It
looks missed -Wno-type-limits in your scripts.