utimes() with vfat is problematic

utimes() with vfat is problematic

[Jan Engelhardt]

Hi,



vfat does not know about ownership, hence the files are always owned by the
vfat mounter (or whatever the uid= option specified). Which brings 
a problem to userspace programs trying to utime() but which do not 
run as the same user as the vfat mounter, because:


fs/attr.c:53
	ret = -EPERM;
	[...]

	/* Check for setting the inode time. */
	if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET)) {
		if (current->fsuid != inode->i_uid && !capable(CAP_FOWNER))
                        goto error;
        }


To trigger the problem:
	# mount /somevfat -o umask=0,uid=root
	$ touch -d "2005-05-05" /somevfat/myfile

I am not sure how this could be dealt with besides passing -o quiet to 
mount.vfat. Any ideas?



	Jan
-- 

Re: utimes() with vfat is problematic

[Pawel Dziepak]

2007/7/10, Jan Engelhardt <jengelh@computergmbh.de>:
> I am not sure how this could be dealt with besides passing -o quiet to
> mount.vfat. Any ideas?

The problem is in the function utimes. I have mounted vfat partition
as a root, then changed permissions to 777 for all files. However, I
wasn't be able to touch any file (changing date to the  past) as a
normal user. utimes() should check write permission instead of uid/
I didn't read 'touch' source code and it is probably caused by it's
procedures that I was able to change modification date to the current
date.
Single UNIX Specification says crealy that to do utimes on a file user
have to had write permissions or be a file owner.

Pawel

Re: utimes() with vfat is problematic

[Jan Engelhardt]

On Jul 10 2007 12:18, Pawel Dziepak wrote:
>
> Single UNIX Specification says crealy that to do utimes on a file user
> have to had write permissions or be a file owner.

Linux does check for write permission, but _only_ for time=NULL.
Hence it would be helpful if someone knows the exact SUS text,
or whether this is not explicitly specified in SUS, leaving room
for interpretation.



	Jan
-- 

Re: utimes() with vfat is problematic

[Pawel Dziepak]

On 7/10/07, Jan Engelhardt <jengelh@computergmbh.de> wrote:
> Linux does check for write permission, but _only_ for time=NULL.
> Hence it would be helpful if someone knows the exact SUS text,
> or whether this is not explicitly specified in SUS, leaving room
> for interpretation.
I have already checked it strictly. This "problem" occurs not only on
vfat partitions. You can *not* do uptime() for time!=NULL if you are
not a file owner. So, everything in Linux is correct.
It is not possible to solve your problem without changing SUS, it is
not even a problem...
I made mistake in my first post, because I was too implied by piece of
code that you've quoted.

Pawel

Re: utimes() with vfat is problematic

[Bodo Eggert]

Jan Engelhardt <jengelh@computergmbh.de> wrote:

> vfat does not know about ownership, hence the files are always owned by the
> vfat mounter (or whatever the uid= option specified). Which brings
> a problem to userspace programs trying to utime() but which do not
> run as the same user as the vfat mounter, because:
> 
> 
> fs/attr.c:53
> ret = -EPERM;
> [...]
> 
> /* Check for setting the inode time. */
> if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET)) {
> if (current->fsuid != inode->i_uid && !capable(CAP_FOWNER))
>                         goto error;
>         }
> 
> 
> To trigger the problem:
> # mount /somevfat -o umask=0,uid=root
> $ touch -d "2005-05-05" /somevfat/myfile
> 
> I am not sure how this could be dealt with besides passing -o quiet to
> mount.vfat. Any ideas?

Would it be possible to allow any user to modify the fs by adding
"&& current->fsuid != -1"? I think it's commonly the desired behaviour.
Off cause the default behaviour should stay the same.
-- 
Those who hesitate under fire usually do not end up KIA or WIA. 

Friß, Spammer: pof@7o.7eggert.dyndns.org jhojzbk@I.7eggert.dyndns.org
 ysoMi@qRIeghz.7eggert.dyndns.org a-r@s.7eggert.dyndns.org
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html