[lkml]   [2005]   [Apr]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
Subject[PATCH] mtime attribute is not being updated on client
Hi Trond,

The acregmin (default=3) and acregmax (default=60) NFS attributes that
control the min and max attribute cache lifetimes don't appear to be
working after the first few timeouts. Using a test program that loops
on the following sequence:
- write to a file on an NFS3 mounted filesystem from the client
- sleep for one second
- stat the file to get the mtime
you see the mtime change once after ~56 seconds, but no further mtime
changes are detected by the test.

nfs_refresh_inode() currently bypasses the inode vs fattr mtime comparison
if data_unstable is true. At the end of nfs_refresh_inode(), it resets the
attribute timer. Since nfs_refresh_inode() is being called after every
write to the server (which occurs more often than the attribute timer is
set to expire), the attribute timer never expires again for this file past
the ~56 sec point.

In nfs_refresh_inode() I believe the mtime comparison should be moved outside
the check for data_unstable. The server might already have a newer value for
mtime than the value on the client. With this change, the test sees the mtime
change after each write completes on the server.


The following patch is for 2.6.12-rc2:

diff -Nura base/fs/nfs/inode.c new/fs/nfs/inode.c
--- base/fs/nfs/inode.c 2005-04-07 16:04:40.933611205 -0400
+++ new/fs/nfs/inode.c 2005-04-07 16:12:34.484299540 -0400
@@ -1176,9 +1176,11 @@

/* Verify a few of the more important attributes */
+ if (!timespec_equal(&inode->i_mtime, &fattr->mtime))
+ nfsi->flags |= NFS_INO_INVALID_ATTR;
if (!data_unstable) {
- if (!timespec_equal(&inode->i_mtime, &fattr->mtime)
- || cur_size != new_isize)
+ if (cur_size != new_isize)
nfsi->flags |= NFS_INO_INVALID_ATTR;
} else if (S_ISREG(inode->i_mode) && new_isize > cur_size)
nfsi->flags |= NFS_INO_INVALID_ATTR;

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: 2005-04-08 03:04    [W:0.063 / U:0.520 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site