View previous topic :: View next topic |
Author |
Message |
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Sun Feb 06, 2005 1:14 pm Post subject: |
|
|
You are mixing up two things:
io-scheduler (ie cfq, deadline anticipatory noop)
and cpu scheduler (O(1)->kernel default, staircase->ck, zaphod->nevermind)
____EDIT____ wrote: |
so realtime preemption has nothing to do with io-scheduling, see the line above
|
the realtime preemption has the goal of reducing the latencies of task-swithing to nearly zero and thus provide realtime system responsiveness even under heavy-loads ( ie it replaces system-common locking techniques with semaphores, mutexes )
It is building up on the default O(1) scheduler as the developer of realtime-preempt, ingo molnar also developed the O(1) scheduler.
now, i already managed to get it working with the (almost) whole ck patchset except for staircase cpu-sched, as this replaces the O(1) scheduler and thus makes realtime-preempt patch unusable.
hope this clarifies things,
cheers,
T |
|
Back to top |
|
|
ZiffyWombat n00b
Joined: 06 Sep 2004 Posts: 64 Location: Earth, Europe, Scandinavia, Finland, Nokia
|
|
Back to top |
|
|
zerojay Veteran
Joined: 09 Aug 2003 Posts: 1033
|
Posted: Sun Feb 06, 2005 1:50 pm Post subject: |
|
|
Tiger683 wrote: | You are mixing up two things:
io-scheduler (ie cfq, deadline anticipatory noop)
and cpu scheduler (O(1)->kernel default, staircase->ck, zaphod->nevermind)
____EDIT____ wrote: |
so realtime preemption has nothing to do with io-scheduling, see the line above
|
the realtime preemption has the goal of reducing the latencies of task-swithing to nearly zero and thus provide realtime system responsiveness even under heavy-loads ( ie it replaces system-common locking techniques with semaphores, mutexes )
It is building up on the default O(1) scheduler as the developer of realtime-preempt, ingo molnar also developed the O(1) scheduler.
now, i already managed to get it working with the (almost) whole ck patchset except for staircase cpu-sched, as this replaces the O(1) scheduler and thus makes realtime-preempt patch unusable.
hope this clarifies things,
cheers,
T |
Forgive my lack of knowledge about the subject, but if you could clarify, is real-time something that's really usable at the moment? Are there any drawbacks? Any specific site out there that might better answer my questions? |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Sun Feb 06, 2005 2:11 pm Post subject: |
|
|
Hmm,
the answer is rather no, there is no www.realtime-preempt.com or such...
EDIT:
but have you ever wondered why your mouse might jump around the screen instead of smoothly gliding upon cpu-usage > 40% ?
or why it takes sometimes 2-3 seconds after a click on a folder until the window pops up?
the answer to the questions about it is partly in the rt: in doesn't have to be this way....
PS: i'm tearing my eyes out right now so you can try it out without the drawbacks of the first nitro-rt release (aka USE="want-realtime") |
|
Back to top |
|
|
Legoguy Apprentice
Joined: 22 Dec 2003 Posts: 166 Location: Edmonton, Alberta, Canada
|
Posted: Sun Feb 06, 2005 6:37 pm Post subject: |
|
|
predatorfreak wrote: | Legoguy wrote: | I don't know if it's nitro or not, but after about 2 days on this kernel my mem filled up to nearly 100% - less than 60 MB remaining out of 1gb. top showed no specific program using it all, and rebooting obviously fixed it but we'll have to see if it happens again in a few days. |
Isn't this how the kernel manages ram anyway? The kernel uses ram until full and unallocates whats unneeded and temperaly uses swap for certain things. |
It's not supposed to fill up to 100% and then stay there (and become really slow). It's happening again now, seems like 3 days and it's full. It's at 616MB usage right now after 1d 14:42 uptime, and usage isn't dropping. This only happens with nitro though, non-rt. Something's not right. I have 1GB Lowmem checked, but I've never experienced problems with that before.
This is NOT normal behavior.
Code: |
top - 12:49:05 up 1 day, 14:42, 8 users, load average: 0.02, 0.06, 0.02
Tasks: 108 total, 1 running, 107 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.3% us, 0.3% sy, 0.0% ni, 96.3% id, 0.0% wa, 0.2% hi, 2.8% si
Mem: 1032852k total, 1016704k used, 16148k free, 121024k buffers
Swap: 506036k total, 5248k used, 500788k free, 281212k cached
# free -m
total used free shared buffers cached
Mem: 1008 993 15 0 118 274
-/+ buffers/cache: 600 408
Swap: 494 5 489
|
As you can see hardly any of that is disk cache. |
|
Back to top |
|
|
predatorfreak l33t
Joined: 13 Jan 2005 Posts: 708 Location: USA, Michigan.
|
Posted: Sun Feb 06, 2005 11:49 pm Post subject: |
|
|
Tiger683 wrote: | predatorfreak wrote: |
its looking to me like 2.6.11 is going to break quite a few things, heh, I'l likely end up being the last person to move to 2.6.11. |
2.6.11 is not the prob, it is the mm patchset that contains so many stuffbreaking patches, that i'll probably start doing my mm-reloaded and after thorough testing put it into some patchset... As darckness said, we had EXACTLY the same probleam yesterday evening at testing the nitro-e and nitro-rt : complete lockup at boottime ide-probing. I still believe i'll be able to fix it, but i guess darckness already gave up all hopes for regular e version. |
No I ment as in, the fact that currently most of the rc versions of 2.6.11 are breaking a good amount of drivers, such as nvidia and that when 2.6.11 comes out, it will likely take awhile before most drivers can cope with the new changes in 2.6.11. In other words, I'm saying most binary drivers will end up busted with 2.6.11 and a few user space apps depending on these binary drivers. Anyway, I hope I'm wrong and that stuff won't break. _________________ System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up. |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Mon Feb 07, 2005 12:10 am Post subject: |
|
|
predatorfreak wrote: |
I'm saying most binary drivers will end up busted with 2.6.11 and a few user space apps depending on these binary drivers.
|
I think in this question i'll agree with the kernel devs:
Didn't the other vendors have enough time to update?
rc releases of 2.6.11 have been out for quite a long time now,
and i don't think their main purpose was to just make a good
impression.....
cheers
T |
|
Back to top |
|
|
predatorfreak l33t
Joined: 13 Jan 2005 Posts: 708 Location: USA, Michigan.
|
Posted: Mon Feb 07, 2005 12:30 am Post subject: |
|
|
Tiger683 wrote: | predatorfreak wrote: |
I'm saying most binary drivers will end up busted with 2.6.11 and a few user space apps depending on these binary drivers.
|
I think in this question i'll agree with the kernel devs:
Didn't the other vendors have enough time to update?
rc releases of 2.6.11 have been out for quite a long time now,
and i don't think their main purpose was to just make a good
impression.....
cheers
T |
This is true, but take into account that these ARE rc versions and hence not as many users test them (I don't personally run/test rc release) which means less users alert say nVidia to the fact that the current release doesn't work with the 2.6.11-rc* kernel and with nVidia it seems to be that they like to make a release perfect (its generally a few months between offical releases). _________________ System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up. |
|
Back to top |
|
|
annihilation n00b
Joined: 07 Feb 2005 Posts: 8
|
Posted: Mon Feb 07, 2005 11:06 pm Post subject: |
|
|
Hey all, first time using this type of kernel and i came across a few issues, one in mmap.c, i had to change:
current->rlim to current->signal->rlim this seems to have fixed my non smp systems. and now they boot fine and run. Was this a good way to fix?
BUT my smp systems give me an error on the bzimage building.
LD .tmp_vmlinux1
kernel/built-in.o(.text+0xc91): In function `try_to_wake_up':
: undefined reference to `smp_send_reschedule_allbutself'
kernel/built-in.o(.text+0x11cd): In function `finish_task_switch':
: undefined reference to `smp_send_reschedule_allbutself'
make: *** [.tmp_vmlinux1] Error 1
Can any one help please, i have tried a lot of things to fix this but this one i really do not know how to fix. I have tried a dual a64 and a dual e64t, nothing helped, disabling smp fixes this issue.
Any ideas? |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Mon Feb 07, 2005 11:12 pm Post subject: |
|
|
Which nitro version? _________________ Retired gentoo user |
|
Back to top |
|
|
annihilation n00b
Joined: 07 Feb 2005 Posts: 8
|
Posted: Mon Feb 07, 2005 11:19 pm Post subject: |
|
|
This one.
2.6.11-rc2-nitro1: The Bear Necessities with RT enabled. |
|
Back to top |
|
|
discomfitor l33t
Joined: 21 Feb 2003 Posts: 927 Location: None
|
Posted: Mon Feb 07, 2005 11:41 pm Post subject: |
|
|
annihilation: You should probably try the non-rt version then. The rt version has been known to be buggy... _________________ There is no substitute for experience.
Imperfection indicates a lack of effort. |
|
Back to top |
|
|
annihilation n00b
Joined: 07 Feb 2005 Posts: 8
|
Posted: Mon Feb 07, 2005 11:47 pm Post subject: |
|
|
ok, just owndered wot u thought of my hacks if they were ok to do, and also why it was only smp that didnt like it |
|
Back to top |
|
|
ziegs Tux's lil' helper
Joined: 29 Apr 2004 Posts: 119
|
Posted: Sat Apr 02, 2005 4:40 am Post subject: |
|
|
the nitro website appears to be down, and i'd like to switch my kernel over. does anyone know if there's a mirror up somewhere or would anyone be able to email/dcc me the file? |
|
Back to top |
|
|
Lenz Veteran
Joined: 17 Jul 2003 Posts: 1439 Location: Marburg [HE, D, EU]
|
Posted: Sat Apr 02, 2005 4:13 pm Post subject: |
|
|
ziegs wrote: | the nitro website appears to be down, and i'd like to switch my kernel over. does anyone know if there's a mirror up somewhere or would anyone be able to email/dcc me the file? |
You can download the latest patchset and ebuild directly from my server:
nitro-sources-2.6.11-r2.ebuild
patch-2.6.11-nitro2.bz2
Another source of mirrors in the thread of 2.6.11-nitro2-release:
https://forums.gentoo.org/viewtopic-t-311984-start-75.html (you have to scroll down a bit) _________________ .:: Lenz' Signature 1.7b ::.
| Gentoo Linux since v1.4 (08-2003) [Screenshot]
| NetHack Highscore: 5.476.380 Pts., Val-Dwa-Fem-Law, ~ ascended to demigoddess-hood ~
Last edited by Lenz on Sun Apr 03, 2005 4:17 pm; edited 1 time in total |
|
Back to top |
|
|
ziegs Tux's lil' helper
Joined: 29 Apr 2004 Posts: 119
|
Posted: Sat Apr 02, 2005 5:16 pm Post subject: |
|
|
thanks, don't know how i missed that thread |
|
Back to top |
|
|
Thetargos Apprentice
Joined: 28 May 2003 Posts: 155 Location: Mexico City
|
Posted: Sun Apr 17, 2005 6:51 pm Post subject: |
|
|
Thanks for the links, I was rather begining to think the Nitro patches were dead, since I've been having trouble getting into the main page (anyone else noticed this too?) |
|
Back to top |
|
|
b3cks Veteran
Joined: 23 Mar 2004 Posts: 1481 Location: Bremen (GER)
|
|
Back to top |
|
|
Thetargos Apprentice
Joined: 28 May 2003 Posts: 155 Location: Mexico City
|
Posted: Sun Apr 17, 2005 7:15 pm Post subject: |
|
|
Thanks a lot!! I'll give it a whirl on my testing system, and will most likely use the others for my production systems. |
|
Back to top |
|
|
|
|
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
|
|