Hi Trond,
This patch keeps the positive return values of nfs_commit_inode() from
leaking out to fsync(). Without this "cvs update" to an nfs dir
breaks.
-Jim
--- linux-2.6/fs/nfs/write.c.orig 2004-04-16 16:22:19.000000000 -0700
+++ linux-2.6/fs/nfs/write.c 2004-04-23 14:04:26.000000000 -0700
@@ -357,8 +357,10 @@
goto out;
}
err = nfs_commit_inode(inode, 0, 0, wb_priority(wbc));
- if (err > 0)
+ if (err > 0) {
wbc->nr_to_write -= err;
+ err = 0;
+ }
out:
clear_bit(BDI_write_congested, &bdi->state);
wake_up_all(&nfs_write_congestion);
Jim,
Maybe we could add some explicit comment here ?
Regards,
Fabian