Re: 2.6.27-rc1: strange fstab issue (Re: 2.6.27-rc1 + selinux new options = no httpd)

[Gene Heskett <gene.heskett@gmail.com>]

On Friday 01 August 2008, Eric Paris wrote:
>On Fri, 2008-08-01 at 09:39 -0400, Gene Heskett wrote:
>> On Thursday 31 July 2008, Rafael J. Wysocki wrote:
>> Update by Gene below.
>>
>> >On Thursday, 31 of July 2008, James Morris wrote:
>> >> On Thu, 31 Jul 2008, Gene Heskett wrote:
>> >> > >Which new options?
>> >> >
>> >> > Make xconfig-->security options:
>> >> >
>> >> > XFRM Networking security hooks
>> >> >
>> >> >  and several others just below it.  Unforch, I can't copy/paste the
>> >> > screen.
>> >>
>> >> I can't really imagine what that is (although if you enable the secmark
>> >> controls under the main SELinux menu, which are disabled by default,
>> >> there could be problems).
>> >
>> >On a possibly related note, I've been observing a strange issue on one of
>> >my test boxes with OpenSUSE 10.3 recently.   Namely, the fsck complains
>> >that there's no passno value in the fstab, although it obviously is
>> > present.
>> >
>> >Strangely enough, if the kernel is compiled with CONFIG_SECURITY_SELINUX
>> > unset, the fsck doesn't complain about the missing passno field any
>> > more.
>> >
>> >Thanks,
>> >Rafael
>>
>> I just did a 2.6.27-rc1 rebuild on a pure, all defaults 'make oldconfig'
>> from my 2.6.26 final .config moved to that src tree.
>>
>> httpd is still being denied access to its log files and dies during the
>> bootup.
>>
>> This is a showstopper for me.
>
>Stephen Smalley just sent me a private note.  Apparently he is having
>e-mail trouble but he did point out the most likely problem.  Can you
>add the patch from
>
>http://marc.info/?l=linux-kernel&m=121726661110266&w=2

Bingo!
The first version there was off about 10 line numbers so I just added the "|
MAY_APPEND", as the second version shows and that was it.  Thanks.

>And give it a whirl?  Sorry, but we think the problem is that the VFS
>stopped passing all of the relevant information down to the security
>system.  https is only allowed to append to its log files, not actually
>'write.'  Since the VFS is longer differentiating those two operations
>you are getting then denial for write.
>
>I'll try to get this pushed into linus's tree quickly.

Looks like its a good to go fix from this angle.  Thanks Eric.
You could even put a tested by: Gene Heskett in it I suppose. :)

>-Eric



-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Fashion is a form of ugliness so intolerable that we have to alter it
every six months.
		-- Oscar Wilde