On Mon, 26 Feb 2007 08:43:29 -0700
Sverre Froyen <sverre@viewmark.com> wrote:

> On Friday 23 February 2007, Hisashi T Fujinaka wrote:
> > it appears to be stuck in ioflush?  I'm going to try to get a dump
> > to generate a backtrace.
> 
> I see hangs whenever heavy file system IO takes place (like the daily
> scripts that run every night or building a NetBSD distribution).
> Often, currently running programs will continue to work, but new ones
> will not start.  Typing return at shell prompt (I'm using bash) will
> not give me another prompt.  ^C and ^Z have no effect.  I can
> (usually) switch from the X-windows display back to console 1 (with
> ctrl-alt-f1) where I can access DDB.  The DDB backtrace always
> procduces
> 
> DDB lost frame for netbsd: Xintr_legacy1+0xb4..
> 
> a ps shows several programs (e.g., qmgr, korganizer, kdeinit (times
> 4)) in state vnlock and ioflush in lfs_dir (/usr and /home are LFS, /
> and /var are FFS, /tmp is tmpfs).
> 
> This is i386 current from 23 Feb (basically GENERIC_LAPTOP).
> 
> I'm doing a cvs update this morning and I will try a new kernel, but
> I do not see any source changes that appear relevant.  Is there any
> (other) information that would be useful and that I should try to
> capture when/if it hangs again?
> 

I suspect it's related to the vnlock hang I posted about a few days
ago.  (I couldn't get a trace; by the time I had the leisure to try,
the offending process had been swapped out.)



		--Steve Bellovin, http://www.cs.columbia.edu/~smb