On Fri 11-10-19 11:49:38, Eric Sandeen wrote:
> One remains: remove_dquot_ref(), because I'm not quite sure how to deal
> with that one w/o taking the i_lock.
Yeah, that will be somewhat tricky. But I think we can modify the standard
iget-iput dance like:
if (need_resched()) {
spin_lock(&inode->i_lock);
if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) {
/*
* Cannot break on this inode, need to do one
* more.
*/
spin_unlock(&inode->i_lock);
continue;
}
__iget(inode);
spin_unlock(&inode->i_lock);
spin_unlock(&sb->s_inode_list_lock);
iput(put_inode);
put_inode = inode;
cond_resched();
spin_lock(&sb->s_inode_list_lock);
}
...
iput(put_inode);
Will you transform this into a proper patch in your series or should I do
it?
Honza
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
