<br><br>
<div class="gmail_quote">On Fri, May 25, 2012 at 9:46 PM, Christoph Hellwig <span dir="ltr"><<a href="mailto:hch@infradead.org" target="_blank">hch@infradead.org</a>></span> wrote:<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px 0px 0.8ex;PADDING-LEFT:1ex" class="gmail_quote">I like this idea, but there's a few more things that should be<br>investigated in this area:<br><br>1) shouldn't the recovery code use struct vnode_info through the code?<br>
2) given that we know the old epoch configuration before starting<br> recovery shouldn't we just pass the old and new one in through<br> start_recovery?<br></blockquote>
<div> </div>
<div> </div>
<div>Getting new and old nodes/vnodes need read data from disk. It should occur at recovery thread. IMHO, we should not pass them to start_recovery.</div>
<div> </div>
<div>There are other files calling vnode_info besides group.c now. If the calling need not read fields of vnode_info, there is no error. Perhaps we can use vnode_info in recovery structure directly and not use sd_vnodes any more. Then the vnode_info definition can remain at group.c. </div>
<div> </div>
<div>Thanks</div>
<div>Haiti </div></div>