View previous topic :: View next topic |
Author |
Message |
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Mon Nov 22, 2004 6:55 pm Post subject: kernel 2.6.x on sparc |
|
|
what is happening with kernel development?
ever since forcing hardware specific patches from development-sources into gentoo-dev-sources, the kernel development for sparc platform has stopped dead in its tracks - the latest available kernel is 2.6.7-r14 which is quite old by now.
second question - whats my chance of using development-sources >=2.6.9 sucessfully on sparc platform ? anyone out there ? |
|
Back to top |
|
|
Mark Clegg Apprentice
Joined: 05 Jan 2004 Posts: 270 Location: ZZ9 Plural Z Alpha
|
Posted: Mon Nov 22, 2004 7:53 pm Post subject: |
|
|
I've had development-sources-2.6.9 on my Ultra 5 briefly. It didn't cause any problems I'm aware of, but I went back to a more supported version because it didn't fix the specific problem I was after. (Mouse baudrate autodetect doesn't)
There is an entry in the Changelog for 2.6.10 that looks promising for that fixs, so I'll be trying that soon.... |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Mon Nov 22, 2004 8:32 pm Post subject: |
|
|
thnx for the fast feedback....
btw, anyone knows what exactly is in sparc hardware-specific patch kit for kernel 2.6? i know about som stuff for broken hme on u1e (for example), but is there anything really significant? |
|
Back to top |
|
|
Weeve Retired Dev
Joined: 30 Oct 2002 Posts: 641
|
Posted: Thu Nov 25, 2004 6:04 am Post subject: |
|
|
Basically what's happening stems from 2 things;
1) Our previous kernel maintainer is no longer able to maintain, so it's taking us a little to switch around
2) development-sources no longer works the way it did up to and including 2.6.7, so we need to switch over to gentoo-dev-sources. One of the problems here is anything greater than 2.6.7 tends to be unstable during periods of high disk I/O, especially on SMP boxes. |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Thu Nov 25, 2004 9:57 pm Post subject: |
|
|
when u say unstable on smp during high i/o - thats exactly the reason why i wanted to go to a newer kernel.
in my experience, kernel 2.6.7 frequently locks up during boot while performing loop delay calibration on smp boxes.
what about alternative schedulers to solve the unstability with high disk i/o? and which filesystems are affected? i'm thinking of converting a raid0 array with high disk i/o to reiserfs4 - whats the score on that one on sparc64 platform? |
|
Back to top |
|
|
Weeve Retired Dev
Joined: 30 Oct 2002 Posts: 641
|
Posted: Fri Nov 26, 2004 2:30 pm Post subject: |
|
|
Basically the only stable filesystem for sparc64 at the moment is ext3, anything else has historically proven to lock up your machine and/or trash your partition table on the drive.
I myself haven't tried different schedulers in my testing but I'm not sure if any of the other developers have either. |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Fri Nov 26, 2004 2:38 pm Post subject: |
|
|
i've used reiserfs3 on number of sparc64 boxes running 2.6.5 and 2.6.7 kernels and never had any problems.
(when i say "using", i don't mean just install and thats it. i have used it on normal disks, on lvm2, on raid0 and never any problems).
my only problems are before mentioned problems which appear on smp platforms. |
|
Back to top |
|
|
Weeve Retired Dev
Joined: 30 Oct 2002 Posts: 641
|
Posted: Fri Nov 26, 2004 2:52 pm Post subject: |
|
|
Reiserfs3 has been hit and miss depending on the person. Some people have good luck w/it and some don't. I don't know that anyone has done any extensive testing with Reiserfs4 on SPARC, so you may be in for a bumpy ride there. |
|
Back to top |
|
|
Mark Clegg Apprentice
Joined: 05 Jan 2004 Posts: 270 Location: ZZ9 Plural Z Alpha
|
Posted: Sun Nov 28, 2004 3:14 pm Post subject: |
|
|
Well I've tried 2.6.10 on my U5 with a make oldconfig from my working 2.6.7.
BUT....
The framebuffer now gives me a blank screen. I can ssh in but the mouse autobaud problem doesn't seem to have gone away, so I'm sticking with 2.6.7 for now. |
|
Back to top |
|
|
hasues n00b
Joined: 04 Mar 2004 Posts: 54
|
Posted: Mon Nov 29, 2004 11:32 pm Post subject: SPARC Woes |
|
|
I'd just be happy if I could get my Sparc 100 running Gentoo, period. I always get a lock after it mounts the root filesystem. It just hangs <shrug>. |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Tue Nov 30, 2004 11:44 am Post subject: |
|
|
hasues, u better open a new thread for that and explain in more detail... |
|
Back to top |
|
|
ba l33t
Joined: 25 May 2003 Posts: 804
|
Posted: Tue Nov 30, 2004 11:50 am Post subject: |
|
|
I just recently installed gentoo-dev-sources-2.6.7(because i need more than 32 supplementary groups max) and noticed that system load(~dayly average) is twice higher than under 2.4. Maybe it's just a coincidence... does anybody noticed same thing? |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Tue Nov 30, 2004 11:52 am Post subject: |
|
|
i just found a serious bug on development-sources 2.6.10-rc2...
it seems that this bug was first discovered on alpha platform in version 2.6.1 (fixed on 2.6.2), but looks like it managed to sneak right back into dev tree...
anyhow, take a look at the last line in:
/arch/sparc64/kernel/vmlinux.lds.S
Code: | /DISCARD/ : { *(.exit.text) *(.exit.data) *(.exitcall.exit) }
|
this has been changed in i386 tree...
i've removed this line and replaced it with
Code: | .exit.text : { *(.exit.text) }
.exit.data : { *(.exit.data) } |
without that, my kernel build was failing with
Quote: | `_exits' referenced in section `.init.text' of drivers/built-in.o: defined in discarded section `.exit.data' of drivers/built-in.o |
but someone more experienced should take a look at this... |
|
Back to top |
|
|
cyan051 n00b
Joined: 21 Aug 2004 Posts: 64
|
Posted: Tue Nov 30, 2004 1:40 pm Post subject: |
|
|
to: ba
no matter what people say on the net, i always found kernel 2.6.7 to present a bigger load on a system on average.
when it comes to specific tasks (single, intensive task) - its definitly faster than 2.4.x, but in idle mode it definitly consumes muuch more resources (both cpu and memory) than 2.4.x
but that doesn't bother me that much if i know that i'll have a faster system when its really needed. |
|
Back to top |
|
|
ashu69 n00b
Joined: 09 Nov 2004 Posts: 45
|
Posted: Wed Dec 01, 2004 9:07 am Post subject: |
|
|
I can echo what cyan said about load vs responsiveness. I have the exact same experience
with 2.6.8.
On the topic of filesystems, i have played with Reiserfs a bit, it gave me a bad feeling of
living on the edge. After having to reinstall the whole system twice, i came back to
good old ext3. I have been running ext3 for much longer now, the only time it gave me
trouble was when i enabled DMA on my WD IDE hard-drive, and i got massive filesystem
corruption in return, so that was my bad.
So yeah, i'd say stick with ext3 for anything more serious then mere curiosity. BTW, i don't
have a SMP box, so YMMV.
HTH,
-a |
|
Back to top |
|
|
nl0pvm n00b
Joined: 08 Nov 2004 Posts: 46 Location: Netherlands
|
Posted: Wed Dec 01, 2004 11:02 am Post subject: ss20 supersparc smp on 2.6.x |
|
|
Hi,
I'm not going to switch to 2.6 until SMP is available for the sparcstation 20. As it is currently not and the committed gentoo developers are not reachable i think that my machine is running 2.4 until it dies |
|
Back to top |
|
|
hasues n00b
Joined: 04 Mar 2004 Posts: 54
|
Posted: Sat Feb 05, 2005 8:42 pm Post subject: 2.6.10 |
|
|
I have 2.6.10 running on my SunBlade 100 now, however, it has only been up for a few days. I'll wait to see if it will lock up. I am having no luck running ALSA with it whatsoever at this point, but I have to research it a bit further. But hey, at least I have a 64 bit kernel Linux box (okay okay, so userland is 32 bit).
Has _________________ Lettuce, Pickles, hold the mayo! |
|
Back to top |
|
|
|