[OpenAFS-devel] still seeing iput errors on 2.2.19...

[Neulinger, Nathan]

I don't think so, but am not sure. The machines that I have typically seen
this on have been samba servers, or a server that I run htdig on. I can
understand the samba server possibly having large numbers of refs, but
65535+ seems a bit rediculous to me.

Would anyone happen to have a snippet that I could put into this routine to
tell me something useful about the vnode in question - i.e. so I could track
down if these are specific files?

Unfortunately/fortunately, this particular bug does not happen very often
any more. I used to see it alot more frequently.

-- Nathan

> -----Original Message-----
> From: Chaskiel M Grundman [mailto:cg2v+@andrew.cmu.edu]
> Sent: Monday, April 09, 2001 1:53 PM
> To: openafs-devel@openafs.org
> Subject: Re: [OpenAFS-devel] still seeing iput errors on 2.2.19...
> 
> 
> Excerpts from internet.computing.openafs.devel: 9-Apr-101
> [OpenAFS-devel] still seein.. by "Neulinger, Nathan"@umr. 
> > What is the 0xffff0000? I don't see i_count being and'ed 
> with anything like
> > that anywhere else in the kernel sources. Does afs have a 
> 2^16-1 limit on
> > i_count or something like that?
> In some versions of afs, there is/was a piece of code that
> *intentionally* got extra refs on the AFS root vnode (e.g. the one
> corresponding to /afs). I'm guessing that this code is a (too-) clever
> attempt to catch a case where that code was risking an overflow of the
> i_count. 
> 
> Are you doing something that actually has that many refs on an AFS
> vnode? It's possible that you are seeing this bug in action. 
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo.cgi/openafs-devel
>