[lkml]   [2005]   [Sep]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: truncate(2) sometimes updates ctime and sometimes ctime and mtime!
On Fri, 2005-09-30 at 08:36 -0700, Linus Torvalds wrote:
> On Fri, 30 Sep 2005, Anton Altaparmakov wrote:
> > There is an inconsistency in the way truncate works which was introduced
> > (relatively) recently.
> >
> > fs/open.c::sys_truncate
> > -> do_sys_truncate
> > -> do_truncate does:
> >
> > newattrs.ia_size = length;
> > newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
> >
> > down(&dentry->d_inode->i_sem);
> > err = notify_change(dentry, &newattrs);
> > up(&dentry->d_inode->i_sem);
> >
> > This changes the ctime only.
> Hmm.. That looks wrong, partly because I don't think it should even set
> ATTR_CTIME _either_. However, I don't see any recent changes to that code,
> so it must have been logn for a long time. That line in do_truncate() has
> been like that since at _least_ 2002.

Indeed. What _has_ changed recently is that IIRC do_truncate used to
set ATTR_SIZE | ATTR_CTIME | ATTR_MTIME at which point the two were
consistent. But (relatively) recently someone removed the ATTR_MTIME
from do_truncate and forgot to remove it from inode_setattr...

> So what changed to make you notice?

I have finished coding the initial cut of ntfs_truncate() (regular files
only, i.e. uncompressed, unencrypted, and
not-attribute-list-attribute-containing). (-:

Now I am at the testing it stage and I was doing:

$ stat ntfs/file
$ truncate ntfs/file <someval>
$ stat ntfs/file

And noticed that the ctime was chaning but not the mtime so I read the
man page which suggested mtime ought to change so I had a quick look at
relevant kernel code and the inconsistency stared me in the face. (-;

> The _code_ actually expects the low-level filesystem to just do it when it
> does the actual truncate itself. That's certainly what ext3 does, for
> example.

Ah, I thought the VFS took care of _all_ {a,c,m} time changes and we
filesystem developers only ever had to write the changed times to

Obviously I was wrong...

I will have a look at ext2/3.

> A comment or two to that effect might be useful, though.

Indeed, it would be useful for unsuspecting fs developers like me. (-;

> In other words: some attributes are "implicit". For example, mtime is
> supposed to be set automatically by the filesystem whenever it sees a
> write. The VFS layer will _not_ do a "inode state notify" event for that.

Right. I seem to remember this being different some time ago. IIRC the
kernel did update {a,m,c}time in many places...

> The same is true of truncation.
> But I agree that it's inconsistent. Anybody have any deep opinions?

I certainly don't. It just should be consistent. (-:

Best regards,

Anton Altaparmakov <aia21 at> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on
WWW: &

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-09-30 18:02    [W:0.044 / U:0.960 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site