Hi,

thanks for your help,

[email protected] írta:
> Hi,
> 
> Laszlo Attila Toth:
>>> CONFIG_AUFS_ROBR?
>> If AUFS is enabled without any suboption, the private_data is still NULL.
> 
> Unfortunately I cannot fully understand what you wrote due to my poor
> English. Will you describe in other words? Do you mean CONFIG_AUFS_ROBR
> makes the private_data NULL?
> 

Hm, my English is also far from perfect. When I enable AUFS in 'make 
menuconfig', additional options appear, I called them - perhaps 
improperly - suboptions. They weren't enabled.


> 
>> It is related to named pipes:
>       :::
>> So, /dev/xconsole isn't opened by the aufs code but calling fchown()=20
>> uses aufs_setattr().
>>
>> I have no idea how can it be.
> 
> I see.
> The special files (named pipe, char device, etc.) are handled like this.
> It is not a problem.
> 
> I may find a cause of your problem.
> As you read the comment in aufs_setattr(), it expects that ATTR_FILE is
> set by ftruncate(2) only. But your fchown(2) sets ATTR_FILE too. I
> checked ubuntu intrepid source files and found they modified
> fchown. This is unexpected situation for aufs (both aufs1 and aufs2).
> I guess their changes triggered the problem in aufs.
> 

It is an AppArmor-related change, and really unexpected. I assumed that 
the kernel source is consistent, this is why I searched the cause 
elsewhere, although I found some strange bug earlier in the 
Ubuntu-shipped extra modules.

In this case checking for the private_data member and disabling 
ATTR_FILE if it is NULL seems to be a good solution. Since that change 
we didn't found any error.

Laszlo

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july

Reply via email to