Official thread: "zen-sources" - Part 7

Message

pappy_mcfae · Post by **pappy_mcfae** » Fri Sep 04, 2009 5:10 am

cheater1034 wrote:use that patch over 2.6.30 (not 2.6.30.1-2.6.30.5) - if you get the same results let me know

Firstly, you can drop the (not 2.6.30.1-2.6.30.5) stuff. I have more open kernel source on one machine (over one hundred gigs) than you've probably seen, so I know just a bit about kernels. I just cranked out 316 complete .config files, so you can imagine I know my way around make xconfig, patching, and working with kernel source.

I had a similar problem but couldnt get any information to help get it fixed

the problem i had was intermittent boot, sometimes it booted, sometimes it would stop before starting init

That's not the only bug I have, either. So far, this machine, a Celeron-based laptop, is the only machine in my corral that can successfully run with 2.6.30-BFS. My 64 bit won't boot (will try the patch later tonight), and my x86 P-IV stopped dead in its tracks while compiling glibc (my test, so to speak). Attempting to restart emerge resulted in emerge failing. Rebooting fixed things for about an hour, then it was back to acting up.

I have installed the source on my P-III server, and will hopefully get BFS running on it as well. Since it's headless, that might cut out the possibility of X issues.

So far, not so good. But the fact that it runs on this machine is a good sign. I have yet to see the miracles, but I have only had this kernel running for about a day. That's hardly time to put it through its paces.

Anyway, I will give the new patch a try and report back later this evening or sometime tomorrow.

Blessed be!
Pappy

ponciarello · Post by **ponciarello** » Fri Sep 04, 2009 6:27 am

just for laughs

from http://ck.kolivas.org/patches/bfs/bfs-faq.txt

Are you looking at getting this into mainline?

LOL.

No really, are you?

LOL.

Really really, are you?

No.

for what's worth con, i'll always test your shiz

pappy_mcfae · Post by **pappy_mcfae** » Fri Sep 04, 2009 9:29 am

Results:

a. The amd64 booted twice, failed boot four times.
b. Once booted, it failed an emerge -aev world eighteen packages into the process. This was a CLI session, not X.
c. Files on the desktop started disappearing during an attempt to emerge glibc in X using konsole. Konsole was very snappy until things died.
d. glibc failed emerge because emerge stopped.

On the good side of things, this laptop ran with the bfs kernel all day long. It only became weird when it was in the process of recompiling a kernel while another machine was distributing compilation to it with distcc. Under normal circumstances, I can do all that, plus read the forums, and jam with kaffeine. With bfs, it brought the machine to its knees; made it completely forget what it was doing.

All that said, BFS is fucking cool! When the bugs get killed, I can see how amazingly fast my computers will run. I'm definitely into this project. Keep those patches coming!

Blessed be!
Pappy

jongeek · Post by **jongeek** » Fri Sep 04, 2009 12:44 pm

Hi, folks. I'm not yet on zen-sources, but I wanted to check out BFS since I've been having _serious_ latency issues on my desktop/server at home (running gentoo-sources-2.6.30-r5). I found this thread while searching for info on Gentoo w/BFS, and will be looking into zen-sources for my home box. Last night I installed vanilla-sources-2.6.30.5, and patched it with the BFS 204 patch. After a reboot, my latency issues are gone, baby, gone ! I mean wow - I am freakin' amazed.

Normally my machine (older, yes, specs to follow) is a dog while running kde 4.3 by itself, much less trying to do any serious multi-tasking. I was having several second delays just running alt-tab to switch between open windows. konsoles would take serious time to redraw when restoring from a minimized state, and it could take 30 seconds just to unlock the screensaver.

With BFS, I was able to emerge kde 4.3.1 using emerge --jobs=2, MAKEOPTS="-j2", and PORTAGE_NICENESS="19", watch video with smplayer, read email, surf the web and a few other things. No noticeable latency performing any of these tasks. My system is responsive like it hasn't been since I was running XFCE. Oh, and my desktop doubles as a home CIFS file server, DNS server, and SMTP/IMAP server.

My machine has an ASUS M2NPV-VM mobo, AMD Athlon 64 X2 3800+ CPU, 2GB RAM, NVIDIA 6150 onboard video, three SATA HDD, one external firewire LaCie HDD, and an IDE DVD-RW.

I'm going to learn more about what other features are in zen-sources beside the BFS patch, and then give zen-sources a shot over the weekend. For now, I wanted to chime in and let everyone know what an amazing difference BFS has made on my box. Maybe it can even serve as a stable platform to compare to for those of you having issues with BFS.

Huzzah !

aTan · Post by **aTan** » Fri Sep 04, 2009 1:17 pm

BFS 204 is buggy for me. when I try to run emerge --sync it freezes in the middle of rsyncing and cpu's wait is about 40-50% (one core? have two), I can't kill it. and it is possible that it damaged my Ext4 with OS. but not sure. don't want to restore whole desktop from backup one more time. i have ~amd64 system. tried gentoo-sources-2.6.30-r4 patched with BFS.

btw 205 is out.

Ant P. · Post by **Ant P.** » Fri Sep 04, 2009 6:09 pm

I must've got lucky this time around, BFS has been rock stable for me.

If anyone's wondering, it isn't a placebo effect either - I can actually max out my 3-core CPU with folding@home (4 processes * 75% each) whereas mainline CFS was getting me maybe 80% per core - and it clearly wasn't maxed out as cpufreq was bouncing up and down endlessly.

That's not really praise of BFS though, it just proves CFS is a huge pile of crap.

kernelOfTruth · Post by **kernelOfTruth** » Fri Sep 04, 2009 9:04 pm

cheater1034 wrote:
@kerneloftruth
Please pull master-2.6.30 to test bfs 203 - if it has the same problem please download the patch from http://ck.kolivas.org/patches/bfs/ and try a plain old 2.6.30 kernel and see if you exhibit the same behavior, then i will aid in you reporting it to con.

---------

I see what I can do, currently I'm moving back home from holidays so there won't be much time - even though I want to have BFS working here soon

it simply leads to a much higher productivity (much less lags, higher response, etc.)

I've read the diff from 202 to 203 and it looked more likely to be of cosmetic nature than real fixing things (regarding this race condition / maxing out issue),

according to recent posts - there doesn't seem to be any improvements in that regard yet

aTan wrote:BFS 204 is buggy for me. when I try to run emerge --sync it freezes in the middle of rsyncing and cpu's wait is about 40-50% (one core? have two), I can't kill it. and it is possible that it damaged my Ext4 with OS. but not sure. don't want to restore whole desktop from backup one more time. i have ~amd64 system. tried gentoo-sources-2.6.30-r4 patched with BFS.

btw 205 is out.

that's exactly the kind of behavior I'm also seeing (not exactly syncing the portage tree but) syncing huge masses of data with rsync (or launching lots of apps -> heavy I/O) leads to that behavior ...

cheater1034 · Post by **cheater1034** » Fri Sep 04, 2009 9:18 pm

From the FAQ, just useful to know:

Currently known problems?

1. Intermittent boot failures on some hardware.
2. Stuck tasks on the same hardware as 1. after extended periods, suggesting
a common problem.
3. Stuck tasks after extensive use of trace functions (ptrace etc.)
4. Failure to suspend on some hardware.
5. More likely to show up bugs in *other* code due to being much more
aggressive at using multiple CPUs so race conditions will show up more
frequently.

THESE ARE OPTIONAL FOR LOWEST LATENCY. YOU DO NOT NEED THESE!
Configure your kernel with 1000Hz, preempt ON and disable dynamic ticks.

------------

pappy_mcfae wrote:
cheater1034 wrote:use that patch over 2.6.30 (not 2.6.30.1-2.6.30.5) - if you get the same results let me know
Firstly, you can drop the (not 2.6.30.1-2.6.30.5) stuff. I have more open kernel source on one machine (over one hundred gigs) than you've probably seen, so I know just a bit about kernels. I just cranked out 316 complete .config files, so you can imagine I know my way around make xconfig, patching, and working with kernel source.

Sorry for offending you. i just don't like catching crap from con when every time i look in #ck someone is saying something about a problem with bfs while they are using zen-sources. (although this certainly isn't your doing, and when i keep repeating test on 2.6.30 it is meant for everyone to read)

i don't seem to understand why you became so upset over not much of anything though

------------

Report bugs with bfs to con on irc.oftc.net, #ck
If it's a known bug as listed in the FAQ/above then there is probably no need to report it, but if you catch anything slightly different than what is listed go ahead and report, just don't do it while using a zen kernel (test a vanilla kernel with bfs)

I just don't think reporting the bugs with bfs in the zen-sources thread on the gentoo forums is very effective, i can't forward the problem to con since it isn't actually happening on my box - only you can effectively explain to him your hardware and what went wrong.

cheater1034 · Post by **cheater1034** » Sat Sep 05, 2009 2:08 am

Added a bunch of logos for the fun of it!

They all look really good, view them in drivers/video/logo with your image viewer.
Added bsd devil(awesome drawn-looking logo), new bsd one (ugly 3d horns), debian, sidux, exherbo, 2 fedora logos, slackware, and tits logo (no nudity/cartoon)

i know fbsd isnt linux, but who doesn't love the devil logo no matter what os you are on?

pappy_mcfae · Post by **pappy_mcfae** » Sat Sep 05, 2009 4:13 am

I was not upset. I was making a request. I understand the issues with others, but as you said, I am not others. Once I found a patch that would actually work, I was able to figure things out. Getting that first patch was a PITA. Besides, since you are here, I'm probably not going to have contact with Con.

Blessed be!
Pappy

ponciarello · Post by **ponciarello** » Sat Sep 05, 2009 7:06 am

tnx for logos (especially for slack)

writing from latest master-2.6.30 (bfs 205) to which i applied these two patches, all is fine

aTan · Post by **aTan** » Sat Sep 05, 2009 3:10 pm

kernelOfTruth wrote:that's exactly the kind of behavior I'm also seeing (not exactly syncing the portage tree but) syncing huge masses of data with rsync (or launching lots of apps -> heavy I/O) leads to that behavior ...

Have you tried 208? How does it behave?

cheater1034 · Post by **cheater1034** » Sat Sep 05, 2009 3:32 pm

aTan wrote:
kernelOfTruth wrote:that's exactly the kind of behavior I'm also seeing (not exactly syncing the portage tree but) syncing huge masses of data with rsync (or launching lots of apps -> heavy I/O) leads to that behavior ...
Have you tried 208? How does it behave?

i believe this is because of the known ptrace problem, dont believe it's fixed yet

rahulthewall · Post by **rahulthewall** » Sat Sep 05, 2009 5:15 pm

After the latest changes to wireless, the kernel fails to compile.

Code: Select all

CC [M]  net/wireless/scan.o
net/wireless/scan.c: In function ‘cfg80211_inform_bss’:
net/wireless/scan.c:499: error: implicit declaration of function ‘kmemleak_ignore’
make[2]: *** [net/wireless/scan.o] Error 1
make[1]: *** [net/wireless] Error 2
make: *** [net] Error 2

aTan · Post by **aTan** » Sat Sep 05, 2009 6:36 pm

cheater1034 wrote:i believe this is because of the known ptrace problem, dont believe it's fixed yet

right, i've tried it myself and it is the same. it also damaged my ext4 system partition again. but just a little this time.

ponciarello · Post by **ponciarello** » Sat Sep 05, 2009 7:30 pm

it happened also to me now, at the 8th machine i installed it in

but I think because I was trying 2.6.31-zen instead of master-2.6.30 (as usual)

rahulthewall · Post by **rahulthewall** » Sat Sep 05, 2009 7:36 pm

ponciarello wrote:it happened also to me now, at the 8th machine i installed it in

but I think because I was trying 2.6.31-zen instead of master-2.6.30

It's already happened to me - nuked my gentoo partition - am in the process of reinstalling it.

aTan · Post by **aTan** » Sat Sep 05, 2009 10:42 pm

cheater1034 wrote:i believe this is because of the known ptrace problem, dont believe it's fixed yet

looks like it's been fixed. who will try it first?

http://ck.kolivas.org/patches/bfs/clone ... read.patch (it's an email message actually, patch needs to be extracted)

kernelOfTruth · Post by **kernelOfTruth** » Sat Sep 05, 2009 10:47 pm

aTan wrote:
cheater1034 wrote:i believe this is because of the known ptrace problem, dont believe it's fixed yet
looks like it's been fixed. who will try it first?

http://ck.kolivas.org/patches/bfs/clone ... read.patch (it's an email message actually, patch needs to be extracted)

Oleg Nesterov wrote: ... copy_process() uses signal->count as a reference counter, but it is not.
This test case ...

... quickly creates an unkillable task.

oh joy !

this means BFS finally can be used without those unkillable tasks ?

one problem less

now lets see if this also fixes the triggering of the race-condition when using rsync or launching lots of i/o heavy tasks

bollucks · Post by **bollucks** » Sun Sep 06, 2009 12:25 am

aTan wrote:
cheater1034 wrote:i believe this is because of the known ptrace problem, dont believe it's fixed yet
looks like it's been fixed. who will try it first?

http://ck.kolivas.org/patches/bfs/clone ... read.patch (it's an email message actually, patch needs to be extracted)

That's not a fix. It's a patch queued for mainline that was thought to be related and doesn't fix the problem, but it does improve it. These could well be bugs in mainline that only show up in bfs as per known problem #5 in the faq:
5. More likely to show up bugs in *other* code due to being much more
aggressive at using multiple CPUs so race conditions will show up more
frequently.

Oh and aTan, you can apply an email as a patch as is, the top part is just treated as patch comments.

Waninkoko · Post by **Waninkoko** » Sun Sep 06, 2009 12:31 am

That patch already comes with .31-rc9 so just pull last zen tree.

pappy_mcfae · Post by **pappy_mcfae** » Sun Sep 06, 2009 7:55 am

My x86_64 is now running like the proverbial raped ape. This is with the 208 patch and the race condition patch as well.

It's in the midst of emerge glibc. If it survives this, then I'll start an emerge -aev system --keep-going to see how well it gets through that. If this is what it's supposed to do, I like it so far. Everything's still on the desktop, too.

Blessed be!
Pappy

pappy_mcfae · Post by **pappy_mcfae** » Sun Sep 06, 2009 8:34 am

It died just as glibc was about done emerging. The openGL screensaver popped on, and stayed on long enough to confuse emerge. The laptop is also running with the latest patches. It seems rock solid as ever. I'm typing from it with the BFS kernel up and running.

Blessed be!
Pappy

aTan · Post by **aTan** » Sun Sep 06, 2009 8:37 am

pappy_mcfae wrote:It died just as glibc was about done emerging. The openGL screensaver popped on, and stayed on long enough to confuse emerge. The laptop is also running with the latest patches. It seems rock solid as ever. I'm typing from it with the BFS kernel up and running.

Blessed be!
Pappy

Try emerge --sync.

ps 209 is out

pappy_mcfae · Post by **pappy_mcfae** » Sun Sep 06, 2009 4:52 pm

Well, at least I know it works on this machine. I just finished an overnight emerge -aev system, and I woke to a finished job, and a computer still running with BFS. It seems to like single cores...a lot.

Blessed be!
Pappy