Re: [PATCH v3 1/1] vfs: iversion truncate bug fix

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, 2012-01-05 at 14:49 -0500, Mimi Zohar wrote:
> On Thu, 2012-01-05 at 13:14 -0500, Christoph Hellwig wrote:
> > On Thu, Jan 05, 2012 at 03:53:54PM +1100, Dave Chinner wrote:
> > > That's seems like a rather unreliable way of detecting that a file
> > > has changed to me.  I mean, only ext4 uses inode_inc_version() when
> > > it internally dirties an inode, and only ext4 sets the MS_I_VERSION
> > > so that inode_inc_version is only called for ext4 inodes when
> > > timestamps change.
> > 
> > And even ext4 only does it when using the non-default "i_version"
> > mount option.
> > 
> > > Hence just adding an increment to the truncate case doesn't seem to
> > > be sufficient to me. e.g. what about the equivalent case of having a
> > > hole punched in the file via fallocate? The file has definitely
> > > changed, but i_version won't change....
> > > 
> > > Perhaps bumping i_version in __mark_inode_dirty() might be the best
> > > way to capture all changes (other than timestamp updates) to any
> > > inode regardless of the filesystem type?
> > 
> > It has the same problem as the timestamp updates doing that right now -
> > the fs can't do locking around it, and it can't return errors.  That's
> > something affecting at least btrfs, xfs and IIRC ubifs, and probably
> > the cluster filesystems as well.  The right answer is to replace the
> > timestmap updates which are the only places doing that with a method
> > as Josef had planned to do, and then we can include the i_version
> > updates in there.
> > 
> > That assumes we figure out a coherent way to do it - except for the
> > conditional abuse in file_updates_times it's currently entirely under
> > fs control.  So the best way to fix it would be to:
> > 
> >  - move the fs-private use into those filesystems actually using it.
> >    Note that a lot less actually check for it rather than just updating
> >    it based on some cargo cult, and most only do so for directories.
> >  - figure a why what exact change count semantics NFS (and IMA) want
> >    and find a way to implement them so that the fs can tell the callers
> >    that they don't exist.
> > 
> > Btw, does IMA care about these beeing persistent?
> 
> By 'persistent' I assume you mean across boots.  IMA (and IMA-appraisal)
> measure and appraise files the first time they're accessed/executed.  So
> no, it does not need to be persistent. IMA/IMA-appraisal just need some
> way to detect file change in order to know whether the file needs to be
> re-measured/appraised on subsequent access.

One other IMA requirement would be the ability to detect if a file
mmapped executable, after the point that it has been locked from
modification, has changed since the last time measured/appraised.  The
concern here, in particular, is the ability to detect file change while
holding the mmap_sem.

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]
  Powered by Linux