linux-kernel - Re: [PATCH] tools: Remove conflicting BITS_PER

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]

Message-ID: <b9a95224-ece8-a23b-e759-48d3f89b3067@nokia.com>
Date:   Mon, 24 Sep 2018 10:33:20 +0200
From:   Alexander Sverdlin <alexander.sverdlin@...ia.com>
To:     Arnaldo Carvalho de Melo <acme@...hat.com>
CC:     Arnaldo Carvalho de Melo <acme@...nel.org>,
        <linux-kernel@...r.kernel.org>,
        Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH] tools: Remove conflicting BITS_PER_LONG define

Hello Arnaldo,

>>>> Em Wed, Sep 12, 2018 at 07:02:32PM +0200, Alexander Sverdlin escreveu:
>>>>>   CC       .../tools/objtool/builtin-check.o
>>>>>   ...
>>>>> In file included from .../tools/arch/x86/include/uapi/asm/bitsperlong.h:11:0,
>>>>>                  from .../tools/include/asm-generic/bitops/__ffs.h:6,
>>>>>                  from .../tools/include/asm-generic/bitops.h:16,
>>>>>                  from .../tools/include/linux/bitops.h:35,
>>>>>                  from .../tools/include/linux/hashtable.h:13,
>>>>>                  from elf.h:24,
>>>>>                  from check.h:22,
>>>>>                  from builtin-check.c:30:
>>>>> .../tools/include/asm-generic/bitsperlong.h:8:0: error: "BITS_PER_LONG" redefined [-Werror]
>>>>>  #define BITS_PER_LONG (__CHAR_BIT__ * __SIZEOF_LONG__)

[...]

finally I more or less know what happens here. In the actual Linux we have two files defining the
same define:

>> # 1 ".../tools/include/linux/bitops.h" 1

#ifndef BITS_PER_LONG
# define BITS_PER_LONG __WORDSIZE
#endif

>> # 6 ".../tools/include/asm-generic/bitsperlong.h" 2

#ifdef __SIZEOF_LONG__
#define BITS_PER_LONG (__CHAR_BIT__ * __SIZEOF_LONG__)
#else
#define BITS_PER_LONG __WORDSIZE
#endif

So the two defines only work together if bitsperlong.h is included first.
In objtool both files are included and for most people bitsperlong.h is indeed
included first.

> I'll try and get one for building a x86_64 tools/perf,
> tools/lib/{api,bpf,traceevent} to see if I manage to reproduce the
> problem you're reporting.

One way to reproduce the reverted include order is to take a HOST compiler
built against old Linux headers, namely 2.6.30 and older.
This is because of the changes in asm/types.h.

It would be possible to put a guard into bitsperlong.h (#ifndef BITS_PER_LONG),
but it just doesn't look correct to me that we have two files defining the same
thing once quite simple, in the second case even more tricky.

-- 
Best regards,
Alexander Sverdlin.