Re: [patch] get_request starvation fix

["Dieter Nützel" <Dieter.Nuetzel@hamburg.de>]

On Friday, 8. February 2002 20:44, Andrew Morton wrote:
> Dieter Nützel wrote:
> > On Fri, Feb 08 2002, Andrew Morton wrote:
> > > Here's a patch which addresses the get_request starvation problem.
> >
> > [snip]
> >
> > > Also, you noted the other day that a LILO run *never* terminated when
> > > there was a dbench running.  This is in fact not due to request
> > > starvation.  It's due to livelock in invalidate_bdev(), which is called
> > > via ioctl(BLKFLSBUF) by LILO.  invalidate_bdev() will never terminate
> > > as long as another task is generating locked buffers against the
> > > device.
> >
> > [snip]
> >
> > Could this below related?
> > I get system looks through lilo (bzlilo) from time to time with all
> > latest kernels + O(1) + -aa + preempt
>
> Depends what you mean by "system locks".

Hard lock up :-(

Nothing in the log files.
No SYSRQ works. Only reset. --- But ReiserFS works. Some few corrupted 
*.o.flag files and most of the time a broken /usr/src/vmlinux or freshly 
/boot/vmlinuz file.

> The invalidate_bdev() problem won't lock the machine.

I see.

> In other words: if you run dbench, then lilo, lilo will not complete
> until after dbench terminates.

dbench wasn't run before
Only several "sync" commands (by hand) during kernel build helps.

> If you're seeing actual have-to-hit-reset lockups then that'll
> be due to something quite different.

Sadly, yes.

Thanks,
	Dieter