2.6.2_rc2-love4 aka "Jimmy Has A Cheeto"

Message

ejohnson · Post by **ejohnson** » Sun Feb 01, 2004 10:57 pm

ett_gramse_nap wrote:Works like a charm over here! Had som trouble steeming from disabling systemvpl (or something like that) with X but that isn't a love-issue...

Still curious about what I wrote in my earlier post thou... Can someone please explain or posts some links regarding nick's swappiness stuff!

Hi,
vm swappiness is default in the kernel, and you can find the patches here and a simplifed artical on his scheduler from lwn.

There isn't as much documentation on vm_swappiness, which is a method of keeping certain types of commonly used data in memory, before its swaped out to the hard disk. This makes for a snappy response when apps are called after being idle and for an increase in responsiveness in commonly used apps. For servers and large databases I think it might have a few negative effects, but I've yet to try that

EDIT: I'm not that sure if Steel300 uses the patches from Nick's page on KernelTrap because the one's included in love are special built I believe.

Donpasquale · Post by **Donpasquale** » Sun Feb 01, 2004 11:08 pm

I just tested out the new love sources on my amd64 machine. I got a compilererror at the very beginning of hte process:

In file included from kernel/signal.c:27:
include/linux/compat_siginfo.h:73: error: syntax error before "compat_timer_t"
include/linux/compat_siginfo.h:73: warning: no semicolon at end of struct or union
include/linux/compat_siginfo.h:73: warning: no semicolon at end of struct or union
include/linux/compat_siginfo.h:109: error: syntax error before '}' token
include/linux/compat_siginfo.h:109: warning: type defaults to `int' in declaration of `_sifields'
include/linux/compat_siginfo.h:109: warning: data definition has no type or storage class
include/linux/compat_siginfo.h:110: error: syntax error before '}' token
include/linux/compat_siginfo.h:110: warning: type defaults to `int' in declaration of `compat_siginfo_t'
include/linux/compat_siginfo.h:110: warning: data definition has no type or storage class
include/linux/compat_siginfo.h: In function `compat_copy_siginfo':
include/linux/compat_siginfo.h:161: error: structure has no member named `_sifields'
include/linux/compat_siginfo.h:161: error: structure has no member named `_sifields'
include/linux/compat_siginfo.h: At top level:
include/linux/compat_siginfo.h:166: error: syntax error before '*' token
include/linux/compat_siginfo.h:166: warning: function declaration isn't a prototype
kernel/signal.c: In function `copy_siginfo_to_user':
kernel/signal.c error: syntax error before ')' token
make[1]: *** [kernel/signal.o] Error 1
make: *** [kernel] Error 2

I hope someone could fix it. The love sources is very nice. Only got that prob with the scheduler.

thx

steel300 · Post by **steel300** » Mon Feb 02, 2004 12:38 am

To all who want a broken-out directory:

I will put one up when I get the time. Some call me lazy, but I'm just very busy. 2 fulltime jobs, full time school, Com Sci club board of directors, buying a house, getting married, developing several open source projects, and keeping love-sources current take up more time than most people have. Posting a broken-out directory is a lot harder than it sounds for me. I don't have one area where I keep patches. I have several patches scattered everywhere. If you pm me, and request a patch, I'll be happy to email it to you, but I just don't have time for a proper broken-out directory.

Also, not all patches apply cleanly, and thus, won't revert properly. To avoid certain headaches, it's easier to forego the whole process all together. I'm really a nice guy, but, in all honesty, I trust very few people. In my line of work, trust is not given out, but earned over many years. I have learned to never offer information. A broken-out directory contradicts my entire line of thinking. If you need a patch or two or three, email me and let me know. I will be more than happy to respond and email you back.

ett_gramse_nap · Post by **ett_gramse_nap** » Mon Feb 02, 2004 6:15 am

ejohnson wrote:
ett_gramse_nap wrote:Works like a charm over here! Had som trouble steeming from disabling systemvpl (or something like that) with X but that isn't a love-issue...

Still curious about what I wrote in my earlier post thou... Can someone please explain or posts some links regarding nick's swappiness stuff!
Hi,
vm swappiness is default in the kernel, and you can find the patches here and a simplifed artical on his scheduler from lwn.

There isn't as much documentation on vm_swappiness, which is a method of keeping certain types of commonly used data in memory, before its swaped out to the hard disk. This makes for a snappy response when apps are called after being idle and for an increase in responsiveness in commonly used apps. For servers and large databases I think it might have a few negative effects, but I've yet to try that

EDIT: I'm not that sure if Steel300 uses the patches from Nick's page on KernelTrap because the one's included in love are special built I believe.

Thanks! It's always nice to learn something new!

Regor · Post by **Regor** » Mon Feb 02, 2004 9:15 am

steel300 wrote:I'm just very busy. 2 fulltime jobs, full time school, Com Sci club board of directors, buying a house, getting married, developing several open source projects, and keeping love-sources current take up more time than most people have.

Wow!

You have all that going on and still agreed to take on maintaining love-sources!? You, sir, are aptly named - you are the man of steel!

I'm loving love-sources, so thanks for all the effort!

grzewho · Post by **grzewho** » Mon Feb 02, 2004 9:15 am

grzewho wrote:the vesafb driver (or bootsplash, dunno) gives lots of garbage output to syslog

any fixes ?

*bump

even when i try to see dmesg output the only thing i get is bootsplash (or vesafb) errors

help me please, i know that there was a fix for some older love-sources with this issue

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 2:04 pm

grzewho wrote:
grzewho wrote:the vesafb driver (or bootsplash, dunno) gives lots of garbage output to syslog

any fixes ?
*bump

even when i try to see dmesg output the only thing i get is bootsplash (or vesafb) errors

help me please, i know that there was a fix for some older love-sources with this issue

I only know of the newer NVidia drivers doing this. There could be issues with fb and the bootsplash, but I don't go for splashes and havn't heard of anything like that yet. hehe.

If you don't use nVidia, then there is a problem. Any examples would help.

PrakashP · Post by **PrakashP** » Mon Feb 02, 2004 4:04 pm

YES! The new Nvidia driver seems to fix the lock-up issue I was having. So currently I am using rc1-love1. Unfortunately newer kernels have acpi issues, as already some other people found out. I mailed at lkml and hope to get some fixes.

Jake · Post by **Jake** » Mon Feb 02, 2004 6:11 pm

People requesting a broken-out can find runtime-selectable IO scheduler and rieser4 patches fixed for 2.6.2-rc3 at http://people.msoe.edu/~maciejej/patches/

steel300 helped me with the runtime-selectable IO scheduler fixes.

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 6:22 pm

PrakashKC wrote:YES! The new Nvidia driver seems to fix the lock-up issue I was having. So currently I am using rc1-love1. Unfortunately newer kernels have acpi issues, as already some other people found out. I mailed at lkml and hope to get some fixes.

PrakashKC,

That's good about the nVidia drivers. Are these the same nForce2 problems of a few months ago? I've read your posts on lkml about the subject

I'm using an a7n8x board with rc2-love4 and have had it running for a few days with out any cpu disconnect problems. I of course always deselect that feature in the kernel power management section anyway.

theonlymcc · Post by **theonlymcc** » Mon Feb 02, 2004 6:29 pm

I just tried these newest love-sources. My machine locks up. I have tried changing the PCI Access mode. Any other suggestions? I was using 2.6.2-rc1-mm3 and no probs.

steel300 · Post by **steel300** » Mon Feb 02, 2004 6:45 pm

Jake wrote:People requesting a broken-out can find runtime-selectable IO scheduler and rieser4 patches fixed for 2.6.2-rc3 at http://people.msoe.edu/~maciejej/patches/

steel300 helped me with the runtime-selectable IO scheduler fixes.

Thanks for the link. I'm glad to see that there are people with the time to pick up some aspects where I'm lacking.

buckoven · Post by **buckoven** » Mon Feb 02, 2004 7:01 pm

hi folks!
with love4 my cd-burner won't work anymore (the device doesn't exist any longer), with love1 it works like a charm.
anyone experienced similar problems here?
btw, i'm not using ide-scsi anymore...
thanks for your help!

greets,
buckoven

PrakashP · Post by **PrakashP** » Mon Feb 02, 2004 7:26 pm

@ejohnson

Nope, what you were referring is the APIC/CPU Disconnect issue. Currently I don't use APIC, as my CPU runs cooler now.

The newer lock-ups I was experiencing began with 2.6.1-mm1 and it was due to a problem with kernel and nvidia binary drivers. The new drivers seem to fix that issue.

But as I said above a new issue with ACPI (not APIC, it is still not resolved) has emerged. I have posted on lkml and hope to get some fixes. But interesting that new love kernel works for you. Do you have siimage controller and sata drive? If yes, try using libata und ACPI. I locks up for me (haven't tried love4 though) with 2.6.2-rc2 and rc3 and 2.6.2-mm2 but not mm1. Using ide siimage driver makes the system boot but let serious errors about "irq, nobody cared!" (no lock-up though) occur after a while.

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 7:48 pm

PrakashKC wrote:irq, nobody cared!"

haha! I've always liked the warnings that Linux apps provide. Kinda reminds me of a good friend trying to ease in some bad news.

I havn't tried libata. I've read about it and assumed it was default in love, but it looks like I'll need to apply the patch. I do have two SATA drives and I use ACPI with out warnings, so I'll look into the use of both at the same time.

Do you have a link to the latest libata version? Or the version you use?

PrakashP · Post by **PrakashP** » Mon Feb 02, 2004 7:53 pm

I just use the version which is provided. libata is in linux and mm kernel and thus it is also in love kernel.

Birnenpfluecker · Post by **Birnenpfluecker** » Mon Feb 02, 2004 7:54 pm

Sorry if I ask something stupid. But where is supermount? Have I overread something? Has it moved to smewhere? It is not under file systems --> pseudo filesystems...
Installed love sources first time, all other things work.

steel300 · Post by **steel300** » Mon Feb 02, 2004 7:56 pm

Birnenpfluecker wrote:Sorry if I ask something stupid. But where is supermount? Have I overread something? Has it moved to smewhere? It is not under file systems --> pseudo filesystems...
Installed love sources first time, all other things work.

Supermount isn't included in love-source for a number reasons which I won't get into here. You haven't overlooked anything, it's just not included.

Birnenpfluecker · Post by **Birnenpfluecker** » Mon Feb 02, 2004 8:02 pm

thx for the real quick answer. The I will try something else.

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 8:03 pm

PrakashKC wrote:I just use the version which is provided. libata is in linux and mm kernel and thus it is also in love kernel.

So, that's why I havn't seen any record of it in Steel300's patchsets. Ok cool. Is it related to large transfers of info or should the warnings appear at boot?

Also Prakash, which version of the nVidia kernel are you using?

PrakashP · Post by **PrakashP** » Mon Feb 02, 2004 8:10 pm

As I said, it it IN the kernel, so of course it won't appear on steel300's list.

Aaarghh, this is driving me nuts. I just wanted to reproduce the oops at bott, but now it doesn't seem to happen... I think it is the nvidia driver which makes the difference. Now I am using lates: 53.36-r1 AFAIK, so the old driver seems to put the card in a state where some bug in ACPI code is triggered. Now it does not. So I will try rc3 again and see whether I get an oops. Otherwise I'll switch to love4 ASAP.

sklettke · Post by **sklettke** » Mon Feb 02, 2004 8:21 pm

buckoven wrote:hi folks!
with love4 my cd-burner won't work anymore (the device doesn't exist any longer), with love1 it works like a charm.
anyone experienced similar problems here?
btw, i'm not using ide-scsi anymore...
thanks for your help!

buckoven, try doing this:

Code: Select all

ln -sf /dev/ide/host0/bus1/target1/lun0/cd /dev/sg1

I'm using love4 (and no ide-scsi) and it worked for me.
BTW, it came from this topic: http://forums.gentoo.org/viewtopic.php? ... c&start=25

Scott

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 8:24 pm

@Prakash

hah. I'm still in the stone age of 5328's, with their nasy attitude towards the kernel. But I just noticed that the 5336 series you're using is not for 64bit like I had thought, so I'll give them a shot as well.

AFAIK those bugs should be just warnings that take up your syslog buffer and really shouldn't effect your stability ... well maybe your mental stablitiy when trying to check dmesg. Goodluck with love4

PrakashP · Post by **PrakashP** » Mon Feb 02, 2004 8:41 pm

So I booted up love4.

Nevertheless 2.6.2-rc3 still gives me oopses. "Good." So we can debug. Unfortunately IU don't have serial console and dmesg to floppy didn't work. *sigh* This is gonna be tough.

ejohnson · Post by **ejohnson** » Mon Feb 02, 2004 8:50 pm

PrakashKC wrote:So I booted up love4. Nevertheless 2.6.2-rc3 still gives me oopses. "Good." So we can debug. Unfortunately IU don't have serial console and dmesg to floppy didn't work. *sigh* This is gonna be tough.

Congrats on love4

The long way I've found to debug dmesg material is to roll back the vid drivers to the 44.xx or even 28.xx series which don't produce all of the excess warnings. Then I can actually debug the non-nVidia related problems.

If I had a problem between the boot process and nVidia I wouldn't know what I'd do, well maybe add a script at boot to dump the dmesg.

2.6.2_rc2-love4 aka "Jimmy Has A Cheeto"

cd burninng gone with love4

Re: cd burninng gone with love4