Re: deadlock on flt_noram5

[yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)]

hi,

> YAMAMOTO Takashi <yamt%mwd.biglobe.ne.jp@localhost> wrote:
> 
>> the problem is not specific to puffs.  see the CAVEATS section of
>> mount_nfs(8).
> 
> But here I am not loopback-mounting the fs. 

what puffs does is an equivalent of loopback nfs mount wrt the problem;
the file server (either puffs's one or nfsd) freely allocate pages
even when it's acting as a behalf of the pagedaemon.

> 
> I note that HEAD behaves much better than netbsd-5 for the same
> situation. It terminates processes, but survives, where netbsd-5 kernel
> just freeze. There is definitvely something to fix here.

while some changes might improved the chance for systems to survive,
the above problem still exists as far as i know.

YAMAMOTO Takashi

> 
> -- 
> Emmanuel Dreyfus
> http://hcpnet.free.fr/pubz
> manu%netbsd.org@localhost