On 06/24/2012 8:18 AM, Thomas Bogendoerfer wrote:
> On Sat, Jun 23, 2012 at 03:30:39PM -0400, Joshua Kinard wrote:
>> On 06/05/2012 3:08 PM, David Daney wrote:
>>> On 06/05/2012 12:00 PM, Jim Faulkner wrote:
>>>> Hi all, I haven't been able to boot any kernels after linux 3.2 on my
>>>> SGI O2. I've tried linux-3.3.5 and linux-3.4.0, but neither would boot.
>>>> Unfortunately I don't have further information such as a kernel panic,
>>>> since I don't get any video output after the kernel is loaded. I've
>>>> attached my linux-3.4 .config. Anybody know what patches I might need
>>>> to get the latest kernels booting on this system?
>>> I have had problems as well.
>>> Someone should configure a serial console and early printk to see if they
>>> can see what is happening.
>> Early printk on O2 systems probably has the same issues as on IP22/IP28
>> where it's overwriting PROM memory somewhere. The system will hang very
>> early in the bootmem allocator if you kludge early printk to work on these
> O2 is easy since it's simply a 16550 as console port chip, so no PROM
> is needed and no problems like on IP22/28.
All I know is if I add SYS_HAS_EARLY_PRINTK to config IP32 in
arch/mips/Kconfig, then enable Early Printk in menuconfig and boot the
resultant kernel, it freezes right there. It's not CPU-model dependent,
because this same O2 used to run an R5000 @ 180MHz and had the same problems
>> CPU revision is: 00002733 (RM7000)
> I guess the problem is here, iirc there was another RM7000 error report
> due to some cache code restructering or it's RM7000 O2 problem in general.
> I don't own a RM7000 cpu module for O2 ...
> On my R5k O2 3.4.3 runs without problems like early 3.5 did:
> mips:~# uname -a
> Linux mips.franken.de 3.4.3 #2 Sun Jun 24 14:12:05 CEST 2012 mips64 GNU/Linux
> mips:~# cat /proc/cpuinfo
> system type : SGI O2
> processor : 0
> cpu model : R5000 V2.1 FPU V1.0
> BogoMIPS : 179.45
I got 3.4 working fine. It was me enabling USB support (OHCI, EHCI), and
the SCSI bus probe was hanging for some reason (trying to read a thumb drive
that I plugged in, I believe). When I disabled OHCI and kept only EHCI,
then I get an unhandled kernel unaligned access oops. Not sure the cause of
it, but it's triggered by ffffffff802560e4 ehci_urb_done+0x24/0x100.
I've also noticed that meth is starting to suffer worse. I get Rx
underflows like its going out of style, as well as a large number of stack
traces triggered by the interrupt handler somewhere:
kswapd0: page allocation failure: order:1, mode:0x4020
But these don't crash the kernel, nor cause corruption in network traffic.
Seems to just be noise. I have the box compiling gcc-4.6.3 for N32 now.
"The past tempts us, the present confuses us, the future frightens us. And
our lives slip away, moment by moment, lost in that vast, terrible in-between."
--Emperor Turhan, Centauri Republic