ext3 and like, mount point in module

ext3 and like, mount point in module

[Tomas Bortoli]

Hi! I'm looking for a way to get the mount point of a file system from the EXT3 linux kernel module.
The folder of the module is (in linux 4.0.5) /fs/ext3
Is it possible? Which is the variable that contains it?

Thanks all
 		 	   		  

Re: ext3 and like, mount point in module

[Richard Weinberger]

On Sat, Aug 1, 2015 at 2:27 PM, Tomas Bortoli <tomasbortoli@hotmail.it> wrote:
> Hi! I'm looking for a way to get the mount point of a file system from the EXT3 linux kernel module.
> The folder of the module is (in linux 4.0.5) /fs/ext3
> Is it possible? Which is the variable that contains it?

The VFS manages mount points, not filesystem drivers.
What problem are you drying to solve?

-- 
Thanks,
//richard

RE: ext3 and like, mount point in module

[Tomas Bortoli]

Thanks for the clarification!
I'm trying to make a patch to slightly improve security in file system.
It consists in removing the ".." dir entry in the "/" dir of the file system mounted on the root mount point.
This could prevent an attacker from using a long series of "../../../" etc in a  transversal directory attack 
with unknown initial relative path to reach the root dir for sure and then move from there.
The dangerousness depends from which is the flaw
Do you think it's worth it?

----------------------------------------
> Date: Sat, 1 Aug 2015 14:46:33 +0200
> Subject: Re: ext3 and like, mount point in module
> From: richard.weinberger@gmail.com
> To: tomasbortoli@hotmail.it
> CC: linux-fsdevel@vger.kernel.org
>
> On Sat, Aug 1, 2015 at 2:27 PM, Tomas Bortoli <tomasbortoli@hotmail.it> wrote:
>> Hi! I'm looking for a way to get the mount point of a file system from the EXT3 linux kernel module.
>> The folder of the module is (in linux 4.0.5) /fs/ext3
>> Is it possible? Which is the variable that contains it?
>
> The VFS manages mount points, not filesystem drivers.
> What problem are you drying to solve?
>
> --
> Thanks,
> //richard
> --
> 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
 		 	   		  --
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

Re: ext3 and like, mount point in module

[Richard Weinberger]

Am 01.08.2015 um 15:10 schrieb Tomas Bortoli:
> Thanks for the clarification!
> I'm trying to make a patch to slightly improve security in file system.
> It consists in removing the ".." dir entry in the "/" dir of the file system mounted on the root mount point.

What about chroot/namespaces/etc? :)

> This could prevent an attacker from using a long series of "../../../" etc in a  transversal directory attack 
> with unknown initial relative path to reach the root dir for sure and then move from there.
> The dangerousness depends from which is the flaw
> Do you think it's worth it?

I'm not sure if it is worth the hassle, I bet some applications depend on that behavior.
But you can give it a try, I'd insert a negative dentry for ".." if ".." is child of the current
root.

Thanks,
//richard

Re: ext3 and like, mount point in module

[Al Viro]

On Sat, Aug 01, 2015 at 03:27:17PM +0200, Richard Weinberger wrote:

> I'm not sure if it is worth the hassle, I bet some applications depend on that behavior.
> But you can give it a try, I'd insert a negative dentry for ".." if ".." is child of the current
> root.

Leaving aside the fact that .. handling does not depend on the corresponding
directory entry, why would such logics belong in the filesystem itself
anyway?  And not in the caller of ->lookup(), which *does* have all the
information needed.

RE: ext3 and like, mount point in module

[Tomas Bortoli]

> What about chroot/namespaces/etc? :)

I think chroot would work even so. Why would a program call chroot("..") on the root dir of the root fs? 
It's unnecessary and tricky. A local program has all the information to know relative paths and move consequently.
Anyway it is possible.


> I'm not sure if it is worth the hassle, I bet some applications depend on that behavior.
> But you can give it a try, I'd insert a negative dentry for ".." if ".." is child of the current
> root.

If with negative dentry you mean not putting ".." I've understood. 
But, ".." is needed for file system mounted not on the root point, otherwise they would be disconnected from the "parent" fs.
So, the "mount_point" variable is needed and it is not present in the file system module itself, as you said. 
Where can I find the vfs module? 
And, If I've understood what Al Viro said, it's correct to work on vfs rather than on the specific file system module.

Thanks all!
 		 	   		  

Re: ext3 and like, mount point in module

[Eric W. Biederman]

Tomas Bortoli <tomasbortoli@hotmail.it> writes:

>> What about chroot/namespaces/etc? :)
>
> I think chroot would work even so. Why would a program call chroot("..") on the root dir of the root fs? 
> It's unnecessary and tricky. A local program has all the information to know relative paths and move consequently.
> Anyway it is possible.
>
>
>> I'm not sure if it is worth the hassle, I bet some applications depend on that behavior.
>> But you can give it a try, I'd insert a negative dentry for ".." if ".." is child of the current
>> root.
>
> If with negative dentry you mean not putting ".." I've understood. 
> But, ".." is needed for file system mounted not on the root point, otherwise they would be disconnected from the "parent" fs.
> So, the "mount_point" variable is needed and it is not present in the file system module itself, as you said. 
> Where can I find the vfs module? 
> And, If I've understood what Al Viro said, it's correct to work on vfs rather than on the specific file system module.

fs/namei.c  follow_dotdot and follow_dotdot_rcu

You might also take a gander at the patches I have recently posted that
address a similiar but different issue with .. and bind mounts.  That
should at least show you where the code is.

Eric