Re: [BUG] mtime&ctime updated when it should not

[Andrew Morton]

Jan Kara <jack@xxxxxxx> wrote:
>
>   Hello,
> 
>   one user pointed my attention to the fact that when the write fails
> (for example when the user quota is exceeded) the modification time is
> still updated (the problem appears both in 2.4 and 2.6). According to
> SUSv3 that should not happen because the specification says that mtime
> and ctime should be marked for update upon a successful completition
> of a write (not that it would forbid updating the times in other cases
> but I find it at least a bit nonintuitive).

hrm.  Doesn't sound super-important.  But..

>   The easiest fix would be probably to "backup" the times at the
> beginning of the write and restore the original values when the write
> fails (simply not updating the times would require more surgery because
> for example vmtruncate() is called when the write fails and it also
> updates the times).
>   So should I write the patch or is the current behaviour considered
> correct?

Isn't this sufficient?


diff -puN mm/filemap.c~a mm/filemap.c
--- 25/mm/filemap.c~a	2003-09-01 12:16:13.000000000 -0700
+++ 25-akpm/mm/filemap.c	2003-09-01 12:17:04.000000000 -0700
@@ -1696,7 +1696,6 @@ generic_file_aio_write_nolock(struct kio
 		goto out;
 
 	remove_suid(file->f_dentry);
-	inode_update_time(inode, 1);
 
 	/* coalesce the iovecs and go direct-to-BIO for O_DIRECT */
 	if (unlikely(file->f_flags & O_DIRECT)) {
@@ -1811,7 +1810,12 @@ generic_file_aio_write_nolock(struct kio
 	}
 	
 out_status:	
-	err = written ? written : status;
+	if (written) {
+		err = written;
+		inode_update_time(inode, 1);
+	} else {
+		err = status;
+	}
 out:
 	pagevec_lru_add(&lru_pvec);
 	current->backing_dev_info = 0;

_

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