On Mon 24-11-14 23:33:35, Ted Tso wrote:

On Tue, Nov 25, 2014 at 12:52:39PM +1100, Dave Chinner wrote:

quoted

Eviction is too late for this. I'm pretty sure that it won't get
this far as iput_final() should catch the I_DIRTY_TIME in the !drop
case via write_inode_now().

Actually, the tracepoint for fs_lazytime_evict() does get triggered
from time to time; but only when the inode is evicted due to memory
pressure, i.e., via the evict_inodes() path.

I thought about possibly doing this in iput_final(), but that would
mean that whenever we closed the last fd on the file, we would push
the inode out to disk.  For files that we are writing, that's not so
bad; but if we enable strictatime with lazytime, then we would be
updating the atime for inodes that had been only been read on every
close --- which in the case of say, all of the files in the kernel
tree, would be a little unfortunate.

  Actually, I'd also prefer to do the writing from iput_final(). My main
reason is that shrinker starts behaving very differently when you put
inodes with I_DIRTY_TIME to the LRU. See inode_lru_isolate() and in
particular:
        /*
         * Referenced or dirty inodes are still in use. Give them another
         * pass
         * through the LRU as we canot reclaim them now.
         */
        if (atomic_read(&inode->i_count) ||
            (inode->i_state & ~I_REFERENCED)) {
                list_del_init(&inode->i_lru);
                spin_unlock(&inode->i_lock);
                this_cpu_dec(nr_unused);
                return LRU_REMOVED;
        }

Regarding your concern that we'd write the inode when file is closed -
that's not true. We'll write the inode only after corresponding dentry is
evicted and thus drops inode reference. That doesn't seem too bad to me.

								Honza
-- 
Jan Kara [off-list ref]
SUSE Labs, CR

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs