current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[Troy Benjegerdes]

Okay, I pulled 2_4_devel this morning and built and SMP kernel my MTX.. it
booted up, but some weird things started happening..

First, exportfs died on bootup..

Starting internet superserver: inetd.
Exporting directories for NFS kernel
daemon.../etc/rc2.d/S20nfs-kernel-server: line 80:   214 Illegal
instruction     $PREFIX/sbin/exportfs -r
done.
Starting NFS kernel daemon: nfsd mountd.


Then, when I tried to log in over serial console, the login process
appeared to die after I entered 'root'.  If I kept trying, I would
occasional actually get a password prompt.


I think there's something going wrong either in the cache flush avoidance
stuff, or in the generic smp VM.

(I'm going to finally update linuxppc_2_4 for MTX SMP later today, so I'll
can hopefully find out if it's a generic VM problem or not)

--
Troy Benjegerdes | master of mispeeling | 'da hozer' |  hozer@drgw.net
-----"If this message isn't misspelled, I didn't write it" -- Me -----
"Why do musicians compose symphonies and poets write poems? They do it
because life wouldn't have any meaning for them if they didn't. That's
why I draw cartoons. It's my life." -- Charles Shulz

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[Troy Benjegerdes]

On Tue, Oct 30, 2001 at 12:12:21PM -0600, Troy Benjegerdes wrote:
>
> Okay, I pulled 2_4_devel this morning and built and SMP kernel my MTX.. it
> booted up, but some weird things started happening..

I am 'reliably' able to reproduce the following:

kalmia:/var/log# /etc/init.d/squid restart
Restarting proxy server:
/etc/init.d/squid: line 3:   450 Segmentation fault      start-stop-daemon
--quiet --start --pidfile $PIDFILE --exec $DAEMON -- $SQUID_ARGS
</dev/null
squid.



--
Troy Benjegerdes | master of mispeeling | 'da hozer' |  hozer@drgw.net
-----"If this message isn't misspelled, I didn't write it" -- Me -----
"Why do musicians compose symphonies and poets write poems? They do it
because life wouldn't have any meaning for them if they didn't. That's
why I draw cartoons. It's my life." -- Charles Shulz

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[David Boutcher]

> I am 'reliably' able to reproduce the following:
>
> kalmia:/var/log# /etc/init.d/squid restart
> Restarting proxy server:
> /etc/init.d/squid: line 3:   450 Segmentation fault
start-stop-daemon

Me too.  Same symptoms on iSeries.   There were some memory related changes
that went in today...I suspect the problems are related to that.

Dave Boutcher
Senior Technical Staff Member
iSeries Linux Development and Architecture


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[Benjamin Herrenschmidt]

>
>
>Then, when I tried to log in over serial console, the login process
>appeared to die after I entered 'root'.  If I kept trying, I would
>occasional actually get a password prompt.
>
>
>I think there's something going wrong either in the cache flush avoidance
>stuff, or in the generic smp VM.
>
>(I'm going to finally update linuxppc_2_4 for MTX SMP later today, so I'll
>can hopefully find out if it's a generic VM problem or not)

It's pre5 ? I think Tom's fix is wrong. It should probably be
copy_user_page or copy_cow_page (I use the last one). Without that,
the cache avoidance will probably win a direct ticket to hell ;)

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[David Boutcher]

> It's pre5 ? I think Tom's fix is wrong. It should probably be
> copy_user_page or copy_cow_page (I use the last one). Without that,
> the cache avoidance will probably win a direct ticket to hell ;)

Changing the fix to copy_cow_page fixed the problem...so I pushed that
change into the devel tree.  Hopefully Tom doesn't mind.

Dave Boutcher
Senior Technical Staff Member
iSeries Linux Development and Architecture


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[Tom Rini]

On Tue, Oct 30, 2001 at 02:28:40PM -0600, David Boutcher wrote:
> > It's pre5 ? I think Tom's fix is wrong. It should probably be
> > copy_user_page or copy_cow_page (I use the last one). Without that,
> > the cache avoidance will probably win a direct ticket to hell ;)
>
> Changing the fix to copy_cow_page fixed the problem...so I pushed that
> change into the devel tree.  Hopefully Tom doesn't mind.

Not at all.  I just heard Paul say copy_highpage was gone in _devel and
something about copy_mem_page. :)  I need to go and get my breakable
PPCs up and running again....

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: current linuxppc_2_4_devel SMP strangeness.. (processes dieing)

[Troy Benjegerdes]

On Tue, Oct 30, 2001 at 01:32:38PM -0700, Tom Rini wrote:
>
> On Tue, Oct 30, 2001 at 02:28:40PM -0600, David Boutcher wrote:
> > > It's pre5 ? I think Tom's fix is wrong. It should probably be
> > > copy_user_page or copy_cow_page (I use the last one). Without that,
> > > the cache avoidance will probably win a direct ticket to hell ;)
> >
> > Changing the fix to copy_cow_page fixed the problem...so I pushed that
> > change into the devel tree.  Hopefully Tom doesn't mind.
>
> Not at all.  I just heard Paul say copy_highpage was gone in _devel and
> something about copy_mem_page. :)  I need to go and get my breakable
> PPCs up and running again....

Let's see.. I think that's three changesets that made a change to memory.c
today ;)

Benh pushed one with 'copy_user_page' instead, but it looks like David's
version won the merge.. So is the copy_cow_page optimization usefull in
this case, or not?

--
Troy Benjegerdes | master of mispeeling | 'da hozer' |  hozer@drgw.net
-----"If this message isn't misspelled, I didn't write it" -- Me -----
"Why do musicians compose symphonies and poets write poems? They do it
because life wouldn't have any meaning for them if they didn't. That's
why I draw cartoons. It's my life." -- Charles Shulz

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/