2.4.18-ben0 on G4 SMP: do_IRQ waiting for irq lock

2.4.18-ben0 on G4 SMP: do_IRQ waiting for irq lock

[Harald Welte]

Hi!

I've tried to run 2.4.18-ben0 on my G4 SMP (dual 450 MHz) box.

I've tried both, a kernel with all IRQ's routed to CPU0 - and a kernel
where both CPU's can handle IRQ's.

Unfortunately in both kernel images, I don't even get past all init scripts,
at some (random) point during bootup:

do_IRQ waiting for irq lock (holder=1)

occurs.

Any idea what's causing the problem?  I've already tried to boot in
single user mode - and it's working fine.  Then I've started every single
init script in /etc/rc2.d (it's debian) by hand - and the problem didn't
occur.

Am I right assuming that there is a deadlock where one CPU is holding
the IRQ lock while the other wants to grab it?

And how could this happen on a kernel where all IRQ's are routed to one
CPU ??

Thanks.

--
Live long and prosper
- Harald Welte / laforge@gnumonks.org               http://www.gnumonks.org/
============================================================================
GCS/E/IT d- s-: a-- C+++ UL++++$ P+++ L++++$ E--- W- N++ o? K- w--- O- M+
V-- PS++ PE-- Y++ PGP++ t+ 5-- !X !R tv-- b+++ !DI !D G+ e* h--- r++ y+(*)

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

Re: 2.4.18-ben0 on G4 SMP: do_IRQ waiting for irq lock

[benh]

>I've tried to run 2.4.18-ben0 on my G4 SMP (dual 450 MHz) box.
>
>I've tried both, a kernel with all IRQ's routed to CPU0 - and a kernel
>where both CPU's can handle IRQ's.

For some yet unknown reason, letting interrupts be distributed on all
CPUs will cause occasional lockups on dual G4s. However, with interrupts
routed to CPU0, I have the machine running rock solid here.

>Unfortunately in both kernel images, I don't even get past all init scripts,
>at some (random) point during bootup:
>
>do_IRQ waiting for irq lock (holder=1)
>
>occurs.
>
>Any idea what's causing the problem?  I've already tried to boot in
>single user mode - and it's working fine.  Then I've started every single
>init script in /etc/rc2.d (it's debian) by hand - and the problem didn't
>occur.
>
>Am I right assuming that there is a deadlock where one CPU is holding
>the IRQ lock while the other wants to grab it?
>
>And how could this happen on a kernel where all IRQ's are routed to one
>CPU ??

Could be a driver bug showing up, difficult to say. The problem definitely
doesn't happen here (dual G4 450Mhz, 2x7400 CPU with highmem). The fact that
the IRQ lock can be held by the other CPU is normal though, both CPUs don't
take external IRQs but decrementer ones for example happen on both.

That said, I don't know what is causing your problem.

Ben.


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

Re: 2.4.18-ben0 on G4 SMP: do_IRQ waiting for irq lock

[Harald Welte]

On Fri, Mar 08, 2002 at 07:34:09PM +0100, benh@kernel.crashing.org wrote:
> Could be a driver bug showing up, difficult to say. The problem definitely
> doesn't happen here (dual G4 450Mhz, 2x7400 CPU with highmem). The fact that
> the IRQ lock can be held by the other CPU is normal though, both CPUs don't
> take external IRQs but decrementer ones for example happen on both.
>
> That said, I don't know what is causing your problem.

I've now managed to get the kernel running - so far without any crashes for
one day.

The problem seemed connected with the cups daemon.  If I don't start (or use)
any part of the cups (common unix printing system), it runs fine.

I don't know what cups is trying to do, since there are no local printers
configured.  Everything it is supposed to do is to submit remote printjobs
to my printer server.

As of drivers, the only hardware driver (apart from standard pmu, aty128fb,
sungem) I'm using is tulip for a 21x4x network board.

But the problems are not related with this network interface, since I'm
heavily using it for NFS, etc.  I've also tried to pingflood the machine
on both network interfaces - no problem.

> Ben.

--
Live long and prosper
- Harald Welte / laforge@gnumonks.org               http://www.gnumonks.org/
============================================================================
GCS/E/IT d- s-: a-- C+++ UL++++$ P+++ L++++$ E--- W- N++ o? K- w--- O- M+
V-- PS++ PE-- Y++ PGP++ t+ 5-- !X !R tv-- b+++ !DI !D G+ e* h--- r++ y+(*)

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

Re: 2.4.18-ben0 on G4 SMP: do_IRQ waiting for irq lock

[roger blofeld]

Harald,
I've seen the same problem with cups on my G4 dual
450.

In addition, my machine hangs whenever the macserial
driver is closed (e.g., open /dev/ttyS0 with minicom,
the exit).

Does your machine do this as well?

-roger

On Sat, 9 Mar 2002 10:43:28 +0100,
laforge@gnumonks.org wrote:
>On Fri, Mar 08, 2002 at 07:34:09PM +0100,
benh@kernel.crashing.org wrote:
>> Could be a driver bug showing up, difficult to say.
The problem definitely
>> doesn't happen here (dual G4 450Mhz, 2x7400 CPU
with highmem). The fact that
>> the IRQ lock can be held by the other CPU is normal
though, both CPUs don't
>> take external IRQs but decrementer ones for example
happen on both.
>>
>> That said, I don't know what is causing your
problem.
>
>I've now managed to get the kernel running - so far
without any crashes for
>one day.
>
>The problem seemed connected with the cups daemon. If
I don't start (or use)
>any part of the cups (common unix printing system),
it runs fine.
>
>I don't know what cups is trying to do, since there
are no local printers
>configured. Everything it is supposed to do is to
submit remote printjobs
>to my printer server.
>
>As of drivers, the only hardware driver (apart from
standard pmu, aty128fb,
>sungem) I'm using is tulip for a 21x4x network board.
>
>But the problems are not related with this network
interface, since I'm
>heavily using it for NFS, etc. I've also tried to
pingflood the machine
>on both network interfaces - no problem.
>
>> Ben.


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