Re: ULE nice bugs are fixed.

On Fri, 11 Apr 2003, Steve Kargl wrote:

> On Fri, Apr 11, 2003 at 05:55:42PM -0400, Jeff Roberson wrote:
> >
> > > On Fri, Apr 11, 2003 at 02:49:35PM -0400, Jeff Roberson wrote:
> > > > On Fri, 11 Apr 2003, Steve Kargl wrote:
> > >
> > > I started to recompile the kernel and while sitting here decided
> > > to load linux-mozilla.  The system rebooted before linux-mozilla
> > > displayed a window.  I'm not sure this ULE related.
> > >
> > > Whoops.  The system just panic after coming back up from
> > > a linux-mozilla induce reboot.  I've left the machine
> > > at the db> prompt.
> > >
> >
> > Is this with the change in my patch or without?  Odd that I'm not seeing
> > this.  I cant remember if I asked for the output of 'call kseq_print(0)'
> > but could you get that as well?
> >
>
> It happens with a kernel without the change at line 180
> and it happens with a kernel with the change :-(
>
> db> call kseq_print(0)
>    load:            0
>    load ITHD:       0
>    load REALTIME:   0
>    load TIMESHARE:  0
>    load IDLE:       0
>    nicemin:         -4
>    nice counts:
>           0 = -1
>           4 = 1
> 0x8
>
>
> I am suspicious that background fsck is exasperating
> the situation.  I'm trying to build a 4BSD kernel to
> see if it is ULE specific without much luck.
>

Ah, unusual.  nicemin is -4 but there is no process with a -4 nice on the
run queue.  This is very suspicious.  Notice the -1 count for 0.  This
makes me think that the nice of a process was set to 0 without going
through the scheduler.  I bet it will be quick to track down.

Let me see if I can find it.  I dont run background fsck here which is why
I'm not seeing this.

Oh, here it is.

Line 303 and 647 of ffs_snapshot.  Change the "td->td_ksegrp->kg_nice = x"
to "sched_nice(td->td_ksegrp, x);"  let me know if that fixes it (it
will).

Cheers,
Jeff