Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
Gentoo Installation on SGI O2+ (R12k 400MHz)
View unanswered posts
View posts from last 24 hours

 
Reply to topic    Gentoo Forums Forum Index Gentoo on Alternative Architectures
View previous topic :: View next topic  
Author Message
cantrop
n00b
n00b


Joined: 11 Sep 2005
Posts: 29

PostPosted: Wed Nov 01, 2006 6:42 pm    Post subject: Gentoo Installation on SGI O2+ (R12k 400MHz) Reply with quote

Hello

I try to install Gentoo on an 400 MHz R12k O2+. I'm in the chroot after netbooting at the moment and try to compile a recent kernel. Before I upgraded to gcc-4.1 according to the upgrade-guide. Does anybody have a useful config for mips-sources-2.6.17.10 ? Are there any other hints to follow beside the handbook for this special machine?
Here are my CFLAGS="-O2 -march=mips4 -pipe -mabi=32 -fomit-frame-pointer -ftracer -fforce-addr"
and USE="unicode nls ip32r10k bzip2"

Thanks, Cantrop

For those who write the Handbook:
I had some issues with the Gentoo MIPS Handbook during the installation so far:

First I tried the Live CD but it did'nt boot. Ok the O2+ seems not to be supported so I went on with netbooting.

2.d. Netbooting on SGI Workstations

I used the netboot image from http://dev.gentoo.org/~kumba/mips/netboot/testing/ip32/r10k/

Although it tried to get the ip adress from dhcp after "resetenv" I had to issue these commands to get the linux system to load:
setenv OSLoadOptions "ip=dhcp"
setenv OSLoadPartition /dev/nfs

4.d. Creating Filesystems
On my system /dev/sda3 did not exist. I had to create it using
"mknod sda3 b 8 3"

4.e. Mounting
the root partition is mounted on /mnt/gentoo but then the handbook goes on and uses /gentoo in the following part (part 5)...

... and then uses /mnt/gentoo again in part 6.

6.a. Chrooting
cp -L /etc/resolv.conf /mnt/gentoo/etc/resolv.conf

The "-L" option is not implemented in the busybox cp
(cp: illegal option -- L
BusyBox v1.00 (2006.01.13-07:36+0000) multi-call binary)
Back to top
View user's profile Send private message
Redhatter
Retired Dev
Retired Dev


Joined: 20 Sep 2003
Posts: 548
Location: Brisbane, QLD, Australia

PostPosted: Thu Nov 02, 2006 3:47 am    Post subject: Reply with quote

Did you read the hardware requirements page?

o Gentoo/MIPS Linux Hardware Requirements
o Gentoo/MIPS Hardware Compatability Database

Those sites quite clearly state that R1x000 O2s are not supported. If you know how to fix the cache coherancy issues faced with this system, you're welcome to have a go, that's what the netboot image is there for, but I'll guarantee that the machine will fall over practically every time. Last time Kumba tried it, an emerge --sync brought the machine to its knees.
_________________
Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...

Gentoo/MIPS Cobalt developer, Mozilla herd member.
Back to top
View user's profile Send private message
cantrop
n00b
n00b


Joined: 11 Sep 2005
Posts: 29

PostPosted: Thu Nov 02, 2006 5:29 am    Post subject: Reply with quote

Thats a pity although my "emerge --sync" worked. But I'm getting this all the time:
"CRIME CPU error at 0x010062bf0 status 0x00000004"

Maybe I should try some of my OCTANE(2) systems? I have an old Indigo2 lying around too. According to the Compatibility Database they should be better supported. But if you don't mind I'm also trying to go a bit further with the O2+. I'll guess that I have to find out for myself whether I should create a 64bit or 32bit vmlinux...

Thanks, Cantrop
Back to top
View user's profile Send private message
hanc
n00b
n00b


Joined: 31 Oct 2006
Posts: 11
Location: Marianske Lazne, Czech Republic

PostPosted: Thu Nov 02, 2006 6:23 am    Post subject: Reply with quote

Code:
zcat /proc/config > /usr/linux/.config
is a good starting poing for configuring a new kernel.
BTW, mips-sources doesn't work on my Octane. I'm using manualy downloaded vanilla MIPS sources with the speedracer patches applied.

hanc
Back to top
View user's profile Send private message
Redhatter
Retired Dev
Retired Dev


Joined: 20 Sep 2003
Posts: 548
Location: Brisbane, QLD, Australia

PostPosted: Thu Nov 02, 2006 7:18 am    Post subject: Reply with quote

cantrop: Octane2 should definately work. They're considered stable running Gentoo.

hanc: Ahh okay... this is interesting. I've got an Octane sitting in my garage now, waiting for me to finish my exams. So in a bit over a fortnight, I should have an opportunity to give Gentoo a try on Octane. :-)

(BTW: I think you meant zcat /proc/config.gz > /usr/src/linux/.config ;-))
_________________
Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...

Gentoo/MIPS Cobalt developer, Mozilla herd member.
Back to top
View user's profile Send private message
Kumba
Developer
Developer


Joined: 16 Jul 2002
Posts: 393
Location: Sigma 957

PostPosted: Fri Nov 03, 2006 9:34 pm    Post subject: Reply with quote

cantrop wrote:
Thats a pity although my "emerge --sync" worked. But I'm getting this all the time:
"CRIME CPU error at 0x010062bf0 status 0x00000004"

Maybe I should try some of my OCTANE(2) systems? I have an old Indigo2 lying around too. According to the Compatibility Database they should be better supported. But if you don't mind I'm also trying to go a bit further with the O2+. I'll guess that I have to find out for myself whether I should create a 64bit or 32bit vmlinux...

Thanks, Cantrop



The error you're seeing is basically the CPU trying to deal with the funkiness these systems generate. SGI, for reasons that only a dark god like Cthulu might be able to fathom, made two systems running the R10000-family processor non-cache-coherent. These two systems are the Indog2 and the O2. The R10000 has a feature called "speculative execution", in which the processor tries to guess ahead in the execution order (I think, I may be mixing this up w/ Out-Of-Order execution). The non-coherentness is a fatal blow to this feature, though, and usually results in the machine tripping over its own two feet and smashing its head on a sharp rock.

For the Indigo2 R10000 (IP28), we use a mix of a special gcc patch and kernel modifications to output a "cache barrier" which attempts to mitigate the fatalness before it can happen. For the IP28 kernel, this involved modifying a number of things, including the bus system, networking, and disk drivers. For the O2, I've copied most of the system-agnostic changes into the O2 kernel and built it using the specialized gcc switch to get the kernel you have recently used. The problem is, the networking and disk drivers of the O2 have not been properly modified, so they are not capable of coping very well with the speculative execution side effects that happen. The networking problems, afaik, are semi-transparent, but the disk driver makes it know that it is unhappy by printing the message you got to the screen. Sometimes, it hiccups and dies.

There are two potential solution that I am aware of. #1 is finishing up the work in the style of the IP28 -- making sure cache barriers are put out at all the right spots and that the networking and disk drivers get tweaked to handle the spec exec side effects (this may involve the use of special buffers, and whatnot). The other option, #2, is to utilize a feature of the O2 chipset (CRIME) that's called JUICE. I don't know too much about this option, just that it requires modifying the TLB stuff in the kernel, and may force us to use a 32bit kernel (as I am told that JUICE may be incompatible with XKPHYS, but this isn't known for sure).

To date, no one's stepped up to help us in this regard. The author of the IP28 patch has tried to get his gcc patch accepted upstream a few times, and has met with some success, but I believe he needs to make some more changes that were suggested before upstream will accept it. Then we may see the IP28 patch folded into the main linux-mips stuff.


That's the long story, really. The short form: don't use R10K/R12K-based O2's w/ Linux unless you're willing to help hack the kernel :)



Now I also see you mention an O2+. Is this the Purple O2 system with an RM7000 CPU that's about as rare on eBay as a girl among slashdot readers?


--Kumba
_________________
"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
Back to top
View user's profile Send private message
Redhatter
Retired Dev
Retired Dev


Joined: 20 Sep 2003
Posts: 548
Location: Brisbane, QLD, Australia

PostPosted: Sat Nov 04, 2006 12:50 am    Post subject: Reply with quote

Kumba wrote:
Now I also see you mention an O2+. Is this the Purple O2 system with an RM7000 CPU that's about as rare on eBay as a girl among slashdot readers?


I think SGI produced O2+s with R10000/R12000 CPUs too. That's the impression I get from the SGI website.
_________________
Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...

Gentoo/MIPS Cobalt developer, Mozilla herd member.
Back to top
View user's profile Send private message
cantrop
n00b
n00b


Joined: 11 Sep 2005
Posts: 29

PostPosted: Sat Nov 04, 2006 2:54 pm    Post subject: Reply with quote

Quote:
That's the long story, really. The short form: don't use R10K/R12K-based O2's w/ Linux unless you're willing to help hack the kernel :)


Thanks, for this long answer. Now I understand, that it does not make sense to go further with this machine. I should rather try to get the original IRIX back on it (As I'm not a kernel hacker).

Quote:
Now I also see you mention an O2+. Is this the Purple O2 system with an RM7000 CPU that's about as rare on eBay as a girl among slashdot readers?


It is the purple one you can see at http://www.sgi.com/products/legacy/mips.html on the right side. But it has an MIPS R12000 Processor and runs at 400MHz. I think it was one of the last O2+ they sold.

In the mean time I was able to boot the Live CD on an Indigo2 Impact R10k which is also unsupported by the gentoo mips team :wink: But til now it all went ok (Maybe because the IP28 patches are nearly complete as you mentioned).

Cantrop
Back to top
View user's profile Send private message
Redhatter
Retired Dev
Retired Dev


Joined: 20 Sep 2003
Posts: 548
Location: Brisbane, QLD, Australia

PostPosted: Sun Nov 05, 2006 6:00 am    Post subject: Reply with quote

IP28 is actually supported for the most part, it's just considered unstable. If you don't slaughter the goat right when running make vmlinux, things h0rk up rather badly on reboot.

We don't recommend them to people who are new to MIPS, but these days, they're pretty solid provided you set things up just right.
_________________
Stuart Longland (a.k.a Redhatter, VK4MSL)
I haven't lost my mind - it's backed up on a tape somewhere...

Gentoo/MIPS Cobalt developer, Mozilla herd member.
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Gentoo on Alternative Architectures All times are GMT
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum