linux-kernel - Re: mmotm 2009-01-14-20-31 uploaded (gfs2)

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: <4974BE8A.6000808@oracle.com>
Date:	Mon, 19 Jan 2009 09:55:22 -0800
From:	Randy Dunlap <randy.dunlap@...cle.com>
To:	Steven Whitehouse <swhiteho@...hat.com>
CC:	Andrew Morton <akpm@...ux-foundation.org>,
	linux-kernel@...r.kernel.org, cluster-devel@...hat.com
Subject: Re: mmotm 2009-01-14-20-31 uploaded (gfs2)

Steven Whitehouse wrote:
> Hi,
> 
> On Mon, 2009-01-19 at 09:05 -0800, Randy Dunlap wrote:
>> Steven Whitehouse wrote:
>>> Hi,
>>>
>>> On Fri, 2009-01-16 at 09:35 -0800, Andrew Morton wrote:
>>>> On Fri, 16 Jan 2009 09:06:23 -0800 Randy Dunlap <randy.dunlap@...cle.com> wrote:
>>>>
>>>>>>>> which is not ideal, but I don't see any easy way to avoid the #ifdef,
>>>>>>>>
>>>>>>> Take a look in fs.h:
>>>>>>>
>>>>>>> #define generic_setlease(a, b, c) ({ -EINVAL; })
>>>>>>>
>>>>>>> If that wasn't a stupid macro, your code would have compiled and ran
>>>>>>> just as intended.
>>>>>>>
>>>>>> There doesn't seem to be an easy answer though. If I #define it to NULL,
>>>>>> that upsets other parts of the code that rely on that macro, and if I
>>>>>> turn it into a inline function which returns -EINVAL, then presumably I
>>>>>> can't take its address for my file_operations.
>>>>> No, gcc will allow &inline_func and out-of-line it if it is needed (AFAIK;
>>>>> I've seen a few cases of that).
>>>>>
>>>> yup.  It measn that we'll get a separate private copy of the
>>>> generic_setlease() code in each compilation unit which takes its
>>>> address, but I don't think that would kill us.
>>>>
>>>> The prevention is of course to put the stub function in a core kernel
>>>> .c file and export it to modules.
>>>>
>>> Having looked into this in a bit more detail now, it seems that this
>>> particular function (generic_setlease) is one of a number appearing in
>>> fs.h which are replaced by macros in the case that CONFIG_FILE_LOCKING
>>> is not set.
>>>
>>> So rather than just do the one function, it seemed to make sense to me
>>> to make them all the same. So this uses inline functions as originally
>>> proposed. If you'd prefer that we don't inline them and instead have a
>>> fs/no-locks.c or something like that with stub functions in it, then I"m
>>> happy to revise the patch accordingly.
>> Acked-by/Tested-by: Randy Dunlap <randy.dunlap@...cle.com>
>>
> Ok, Thanks. I'll send a proper patch to a suitable tree. Presumably the
> VFS tree would be best for this?

That sounds correct, but no guarantees.

>>> This patch passes my compile tests, modulo a small change that I need to
>>> make in GFS2 to suppress a warning (not attached). That seems to be
>>> related to yet another set of macros which appear only with
>>> CONFIG_FILE_LOCKING not set. Maybe I should update those to be inline
>>> functions as well....
>> You mean these?  Probably should update them as well.
>>
>> #else /* !CONFIG_FILE_LOCKING */
>> #define locks_mandatory_locked(a) ({ 0; })
>> #define locks_mandatory_area(a, b, c, d, e) ({ 0; })
>> #define __mandatory_lock(a) ({ 0; })
>> #define mandatory_lock(a) ({ 0; })
>> #define locks_verify_locked(a) ({ 0; })
>> #define locks_verify_truncate(a, b, c) ({ 0; })
>> #define break_lease(a, b) ({ 0; })
>> #endif /* CONFIG_FILE_LOCKING */
>>
>>
> Yes, I'll do a patch for those as well then,

Thanks.

-- 
~Randy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/