HIGHMEM + serial console + SMP + 2GB freeze on pmac

HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Troy Benjegerdes]

I just compiled the latest 2_4_devel with CONFIG_SMP,
CONFIG_SERIAL_CONSOLE, CONFIG_XMON, and CONFIG_HIGHMEM, and I'm running
into some problems.

This machine is an ealier Sawtooth G4 (7400) with 4 memory slots and 2GB
of ram (I know, Macos does not support this).

First, It booted up normally, but when I logged into KDE it got about
halfway through the login and froze.

Then I booted with 'console=ttyS0,38400'.. it froze after

smp_core99_probe: found 1 cpus
Waiting on wait_init_idle (map = 0x0)
All processors have done init_idle
PCI: Probing PCI hardware

Then I tried 'console=ttyS0,38400, mem=1500M'.. still died on probing PCI
hardware.

I'm going to try several other combinations of things, including taking
one memory stick out since this could actually be that the hardware
doesn't like that much memory.

--
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 Schulz

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

Re: HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Troy Benjegerdes]

> I'm going to try several other combinations of things, including taking
> one memory stick out since this could actually be that the hardware
> doesn't like that much memory.

There is at least one problem with serial console on pmac.. I removed the
memory (except for 1 512mb dimm)


If I turn serial console on, I get a machine check in
serial_console_write during the PCI probing.

This only shows up if Xmon is NOT compiled in.. since xmon tries to use
the port (and presumably dies also)

--
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 Schulz

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

Re: HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Troy Benjegerdes]

On Thu, Dec 06, 2001 at 06:25:24PM -0600, Troy Benjegerdes wrote:
>
> > I'm going to try several other combinations of things, including taking
> > one memory stick out since this could actually be that the hardware
> > doesn't like that much memory.
>
> There is at least one problem with serial console on pmac.. I removed the
> memory (except for 1 512mb dimm)
>
>
> If I turn serial console on, I get a machine check in
> serial_console_write during the PCI probing.
>
> This only shows up if Xmon is NOT compiled in.. since xmon tries to use
> the port (and presumably dies also)

Another note: The serial works fine after the machine is booted up.

--
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 Schulz

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

Re: HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Timothy A. Seufert]

At 6:07 PM -0600 12/6/01, Troy Benjegerdes wrote:

>This machine is an ealier Sawtooth G4 (7400) with 4 memory slots and 2GB
>of ram (I know, Macos does not support this).

[...]

>I'm going to try several other combinations of things, including taking
>one memory stick out since this could actually be that the hardware
>doesn't like that much memory.

MacOS X supports 2GB on Uni-N machines with four sockets, so the
hardware supports it.  I think the 1.5GB thing was just a consequence
of Classic MacOS having a fairly inflexible address space layout.
--
Tim Seufert

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

Re: HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Michel Lanners]

On   6 Dec, this message from Troy Benjegerdes echoed through cyberspace:
>
>> I'm going to try several other combinations of things, including taking
>> one memory stick out since this could actually be that the hardware
>> doesn't like that much memory.
>
> There is at least one problem with serial console on pmac.. I removed the
> memory (except for 1 512mb dimm)
>
>
> If I turn serial console on, I get a machine check in
> serial_console_write during the PCI probing.
>
> This only shows up if Xmon is NOT compiled in.. since xmon tries to use
> the port (and presumably dies also)

At one point there was a problem with Mac serial support doing a
kmalloc() (IIRC...), which would evidently fail at the very early
stages, such as when opening a serial console (one of the first things
done after the kernel runs) or maybe when using xmon early on, since the
VM subsystem is not yet initialized at that time.

BenH allegedly found a fix and commited it; at least his 2.4.12 rsync
and 2.4.16 kernel.org do run fine wth serial console here...

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


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

Re: HIGHMEM + serial console + SMP + 2GB freeze on pmac

[Troy Benjegerdes]

On Fri, Dec 07, 2001 at 02:25:14PM -0800, Timothy A. Seufert wrote:
> At 6:07 PM -0600 12/6/01, Troy Benjegerdes wrote:
>
> >This machine is an ealier Sawtooth G4 (7400) with 4 memory slots and 2GB
> >of ram (I know, Macos does not support this).
>
> [...]
>
> >I'm going to try several other combinations of things, including taking
> >one memory stick out since this could actually be that the hardware
> >doesn't like that much memory.
>
> MacOS X supports 2GB on Uni-N machines with four sockets, so the
> hardware supports it.  I think the 1.5GB thing was just a consequence
> of Classic MacOS having a fairly inflexible address space layout.

Yeah, it's not the RAM that's the problem. I've been running various
highmem configs with 2GB for a couple days now.

The following change to arch/ppc/mm/init.c *seems* to fix the hang problem
for me. (And for Olaf Herring on some power3 machines)

===== init.c 1.62 vs edited =====
52c52
< #define MAX_LOW_MEM   (0xF0000000UL - KERNELBASE)
---
> #define MAX_LOW_MEM   (0xE0000000UL - KERNELBASE)

Benh said on IRC he thinks we might have a problem running out of
ioremap/vmmalloc space. I'm inclined to agree with him, although for now
I'd rather just change MAX_LOW_MEM for now until someone figures out a way
to detect vmalloc/ioremap collionsions with lowmem RAM mappings.

--
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 Schulz

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