<br><br><div class="gmail_quote">On Mon, May 14, 2012 at 11:27 PM, Christoph Hellwig <span dir="ltr"><<a href="mailto:hch@infradead.org" target="_blank">hch@infradead.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I have looked over the patch a bit more and I think there is an issue<br>
in it as it decouples the epoch update formerly in __sd_leave_done<br>
from the vnode information update the line below - it seems like we<br>
might now hit a window where the vnode has alreayd been incremented<br>
by operations still operate on the old vnode confifuration. But at<br>
least to me it seems like we could update the vnode configuration<br>
earlier, too.<br></blockquote><div>Your are right. </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
another thing I noticed is that I think your patch makes the ->dispatch<br>
cluster driver operation obsolete. Once we never register the event<br>
handler for the cluster driver FD it can be handled netirely inside the<br>
cluster driver, e.g. just call register_event from inside the init<br>
method and don't even tell the core about it.<br></blockquote><div>We have called register_event(cdrv_fd, ...) in create_cluster().</div></div><br><br clear="all"><div><br></div>-- <br>Yunkai Zhang<div>Work at Taobao</div>
<br>