[OpenAFS] Strange access problems on one client

Derrick Brashear shadow@gmail.com
Thu, 11 Oct 2007 23:20:35 -0400


------=_Part_10273_15211296.1192159235287
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On 10/11/07, Jeffrey Altman <jaltman@columbia.edu> wrote:
>
> Marc Dionne wrote:
> > Now I'm curious as to why the specific value returned by DirHash even
> > matters.  Even with the gcc bug it would be consistent for a given
> > string on the same client.  Does this value have to match what's
> > computed for the same string on the file server side?
>
> The hash is used to look up the name in the AFS3 directory hash table
> that is received by the client from the file server.
>
> > (looks like
> > DirHash is also used in fileserver) If so, it looks like the original
> > code could produce different values with different size ints on the
> > server and client.
>
> Good thing we are lucky that 'int' is 32-bits on all of the platforms we
> care about.   I agree this should be afs_int32.


the entire dir package is bad about specifying int, long or afs_int32;
making stylistic changes now is probably not the best idea.

------=_Part_10273_15211296.1192159235287
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

<br><br><div><span class="gmail_quote">On 10/11/07, <b class="gmail_sendername">Jeffrey Altman</b> &lt;<a href="mailto:jaltman@columbia.edu">jaltman@columbia.edu</a>&gt; wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Marc Dionne wrote:<br>&gt; Now I&#39;m curious as to why the specific value returned by DirHash even<br>&gt; matters.&nbsp;&nbsp;Even with the gcc bug it would be consistent for a given<br>&gt; string on the same client.&nbsp;&nbsp;Does this value have to match what&#39;s
<br>&gt; computed for the same string on the file server side?<br><br>The hash is used to look up the name in the AFS3 directory hash table<br>that is received by the client from the file server.<br><br>&gt; (looks like<br>
&gt; DirHash is also used in fileserver) If so, it looks like the original<br>&gt; code could produce different values with different size ints on the<br>&gt; server and client.<br><br>Good thing we are lucky that &#39;int&#39; is 32-bits on all of the platforms we
<br>care about.&nbsp;&nbsp; I agree this should be afs_int32.</blockquote><div><br>the entire dir package is bad about specifying int, long or afs_int32; making stylistic changes now is probably not the best idea. <br></div><br></div>
<br>

------=_Part_10273_15211296.1192159235287--