View previous topic :: View next topic |
Author |
Message |
GenKreton l33t
Joined: 20 Sep 2003 Posts: 828 Location: Cambridge, MA
|
Posted: Wed Oct 08, 2003 2:57 am Post subject: |
|
|
Yeah, what exactly are these love-sources for? |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 9:08 am Post subject: |
|
|
Ian wrote: | Can someone either give me a very dumbed down description of what the scheduler does/how it works, or point me to a dumbed down version?
I think I get it, but I'm not sure, and I'm just confusing myself now :p. |
Nick Piggin explained this on lkml several times - but if there's a demand I could ask him to answer a few questions on the design.
But basically it uses process priority (not niceness as other schedulers) to determine timeslice lenght. AFAIK. |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 9:16 am Post subject: |
|
|
GenKreton wrote: | Yeah, what exactly are these love-sources for? |
dumb question... world domination of course.
Well, started out as a desktop kernel, I suppose that's still true... mainly it's a fun little task I undertook to test new ideas, atm. these come mainly from Nick Piggin. But something like Reiser4 has also got a place in -love (though excluded right now for technical reasons). As well as all those patches that get posted on lkml to fix small bugs, add features I like (but nobody else likes, like better CPU selection).
I aim for a stable kernel tweaked for the desktop - but world domination seems like a lofty goal. |
|
Back to top |
|
|
neenee Veteran
Joined: 20 Jul 2003 Posts: 1786
|
Posted: Wed Oct 08, 2003 9:45 am Post subject: |
|
|
hmm.. i recall something being said about athlons
locking up and mem=nopentium; that mem=no-
pentium would not be needed with this patch? has
that been confirmed? if so; i can remove that line
from my grub config. also, i have no-hlt there. |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 9:55 am Post subject: |
|
|
neenee wrote: | hmm.. i recall something being said about athlons
locking up and mem=nopentium; that mem=no-
pentium would not be needed with this patch? has
that been confirmed? if so; i can remove that line
from my grub config. also, i have no-hlt there. |
-love includes a new workaround to the athon prefetch flaw - you can safely remove mem=nopentium I think, but I've gotten no feedback on that patch yet so I have no way of saying that for sure.
but AMD has tested it on their machines for months and they sayth "no crashy" so I think we can say it's safe... Andi Kleen is the AMD fix0r god |
|
Back to top |
|
|
asimon l33t
Joined: 27 Jun 2002 Posts: 979 Location: Germany, Old Europe
|
Posted: Wed Oct 08, 2003 10:40 am Post subject: |
|
|
2.6.0-test6-love1's initrd driver is not working. Neither the "RAMDISK driver initialized" message is shown nor are any initrds detected. |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 10:57 am Post subject: |
|
|
asimon wrote: | 2.6.0-test6-love1's initrd driver is not working. Neither the "RAMDISK driver initialized" message is shown nor are any initrds detected. |
Known issue - Andrew broke it...
This is the patch that breaks stuff - you could try patch -p1 -R -i /path/to/patch/name.patch it.
http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test6/2.6.0-test6-mm4/broken-out/RD0-initrd-B6.patch
But it's a fairly large patch and it affects a nasty lot of the initrd code, so odds are that it will break something else.
Last edited by Lovechild on Wed Oct 08, 2003 11:24 am; edited 1 time in total |
|
Back to top |
|
|
neenee Veteran
Joined: 20 Jul 2003 Posts: 1786
|
Posted: Wed Oct 08, 2003 11:19 am Post subject: |
|
|
i will remove the no-hlt + mem=nopentium
in a bit and reboot and hope for the best |
|
Back to top |
|
|
asimon l33t
Joined: 27 Jun 2002 Posts: 979 Location: Germany, Old Europe
|
Posted: Wed Oct 08, 2003 11:34 am Post subject: |
|
|
Lovechild wrote: |
Known issue - Andrew broke it... |
Yes, Evil Andrew! It's broken since test6-mm2.
Thanks for the pointer. I already tried reverting this patch (actually I tried it with mm4 not with love1 but I think the outcome will be the same). Without this patch the initrd driver gets initialized and even detects the initrd but then follows a kernel oops and panic. So it's as you guessed: just reverting it will break the kernel. |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 11:53 am Post subject: |
|
|
asimon wrote: | Lovechild wrote: |
Known issue - Andrew broke it... |
Yes, Evil Andrew! It's broken since test6-mm2.
Thanks for the pointer. I already tried reverting this patch (actually I tried it with mm4 not with love1 but I think the outcome will be the same). Without this patch the initrd driver gets initialized and even detects the initrd but then follows a kernel oops and panic. So it's as you guessed: just reverting it will break the kernel. |
did you inform him that he killed the damn thing, or do you just like complianing to us better? |
|
Back to top |
|
|
asimon l33t
Joined: 27 Jun 2002 Posts: 979 Location: Germany, Old Europe
|
Posted: Wed Oct 08, 2003 12:13 pm Post subject: |
|
|
Lovechild wrote: |
did you inform him that he killed the damn thing |
Of course, but the last two mm versions didn't fixed it and the mail from the guy who posted the very same issue on lkml wasn't so far answered either. Seems to be a complicated issue. I guess Andrew doesn't use initrd himself. Maybe test7 or mm5 will fix it...
EDIT: BTW, my comment was not meant to "complain", but just to inform people who need initrd and who wanted to test love1 so that they don't have to spend time in vain. It will not work for such systems. |
|
Back to top |
|
|
didl Retired Dev
Joined: 09 Sep 2003 Posts: 1106 Location: Pittsburgh, PA
|
Posted: Wed Oct 08, 2003 3:47 pm Post subject: Re: Love-sources - The "Wailing like Bob Dylan" re |
|
|
Hi,
Lovechild wrote: |
echo cfq > /sys/block/hda/queue/io_scheduler
to /etc/conf.d/local.start
|
Could somebody give me their recommendations on what scheduler to use
for a desktop machine( i.e. cfq, as ....). I don't know any details about either implementation
and their differences.
Otherwise - the love sources really rock on my box - thanks!
|
|
Back to top |
|
|
GoinBald n00b
Joined: 19 Apr 2003 Posts: 39 Location: Canada
|
Posted: Wed Oct 08, 2003 3:56 pm Post subject: Yet Another Happy Customer |
|
|
I'd really like to thank you for this excellent patchset.
Great job!!!!
j |
|
Back to top |
|
|
bssteph l33t
Joined: 26 Feb 2003 Posts: 652 Location: Wisconsin
|
Posted: Wed Oct 08, 2003 7:32 pm Post subject: |
|
|
Okay, so I'm busy as all hell Tuesday and get some free time today. Check my posts on the forums. Don't check anything new. Think Lovechild hasn't gotten test6-love1 out. Patch -mm4 with v15a myself. Party. Remember that Lovechild makes separate posts for each version. See love1 is out. Grumble at myself.
Downloading the love set now :) |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 8:08 pm Post subject: |
|
|
bssteph wrote: | Okay, so I'm busy as all hell Tuesday and get some free time today. Check my posts on the forums. Don't check anything new. Think Lovechild hasn't gotten test6-love1 out. Patch -mm4 with v15a myself. Party. Remember that Lovechild makes separate posts for each version. See love1 is out. Grumble at myself.
Downloading the love set now |
not really, I try to make a new thread for each major revision like when we went for -wli/-mm based, and now with each linus milestone release update to -love (or if I fix a major bug which might be critical to update).
Quote: |
Hi,
Lovechild wrote:
echo cfq > /sys/block/hda/queue/io_scheduler
to /etc/conf.d/local.start
Could somebody give me their recommendations on what scheduler to use
for a desktop machine( i.e. cfq, as ....). I don't know any details about either implementation
and their differences.
Otherwise - the love sources really rock on my box - thanks!
|
For a pure desktop I would always go for CFQ, for a server AS or deadline (only deadline if AS has serious bugs). In the future CFQ might get some AS elements added, but that's probably 2.7 work.
You will want CFQ because it give equal rights to the IO bandwidth to all processes so you are sure nothing it hoging your system. CFQ also tends too boost startup times, especially under load. |
|
Back to top |
|
|
bssteph l33t
Joined: 26 Feb 2003 Posts: 652 Location: Wisconsin
|
Posted: Wed Oct 08, 2003 8:27 pm Post subject: |
|
|
Lovechild wrote: |
not really, I try to make a new thread for each major revision like when we went for -wli/-mm based, and now with each linus milestone release update to -love (or if I fix a major bug which might be critical to update).
|
That's what I meant to say... I think...
It should also be mentioned that elevator=cfq as a boot argument still works. Sets everything to said elevator, so if you're lazy or want most devices to be cfq and the others something else, you can do that and save a couple lines in local.start |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 8:29 pm Post subject: |
|
|
bssteph wrote: | Lovechild wrote: |
not really, I try to make a new thread for each major revision like when we went for -wli/-mm based, and now with each linus milestone release update to -love (or if I fix a major bug which might be critical to update).
|
That's what I meant to say... I think...
It should also be mentioned that elevator=cfq as a boot argument still works. Sets everything to said elevator, so if you're lazy or want most devices to be cfq and the others something else, you can do that and save a couple lines in local.start |
Where's the fun in that? |
|
Back to top |
|
|
bssteph l33t
Joined: 26 Feb 2003 Posts: 652 Location: Wisconsin
|
Posted: Wed Oct 08, 2003 8:32 pm Post subject: |
|
|
Lovechild wrote: | bssteph wrote: | Lovechild wrote: |
not really, I try to make a new thread for each major revision like when we went for -wli/-mm based, and now with each linus milestone release update to -love (or if I fix a major bug which might be critical to update).
|
That's what I meant to say... I think...
It should also be mentioned that elevator=cfq as a boot argument still works. Sets everything to said elevator, so if you're lazy or want most devices to be cfq and the others something else, you can do that and save a couple lines in local.start |
Where's the fun in that? |
Invisioning an elevator car sending the data from disk up through the kernel to your app on the top floor? |
|
Back to top |
|
|
didl Retired Dev
Joined: 09 Sep 2003 Posts: 1106 Location: Pittsburgh, PA
|
Posted: Wed Oct 08, 2003 9:33 pm Post subject: |
|
|
Lovechild wrote: |
For a pure desktop I would always go for CFQ, for a server AS or deadline (only deadline if AS has serious bugs). In the future CFQ might get some AS elements added, but that's probably 2.7 work.
You will want CFQ because it give equal rights to the IO bandwidth to all processes so you are sure nothing it hoging your system. CFQ also tends too boost startup times, especially under load. |
Thanks for the heads up, Lovechild |
|
Back to top |
|
|
nepenthe Guru
Joined: 14 Jul 2002 Posts: 317 Location: Sanford, ME
|
Posted: Wed Oct 08, 2003 10:44 pm Post subject: |
|
|
sweet 2.6.0-test7 is out =p |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Wed Oct 08, 2003 11:32 pm Post subject: |
|
|
nepenthe wrote: | sweet 2.6.0-test7 is out =p |
now all I need to fuck up my day tomorrow is Andrew putting out test7-mm1 |
|
Back to top |
|
|
nepenthe Guru
Joined: 14 Jul 2002 Posts: 317 Location: Sanford, ME
|
Posted: Wed Oct 08, 2003 11:49 pm Post subject: |
|
|
lol was thinking it must be pain... I feel for ya man
But love sources rock... keep on keepin on =p |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Thu Oct 09, 2003 12:00 am Post subject: |
|
|
nepenthe wrote: | lol was thinking it must be pain... I feel for ya man
But love sources rock... keep on keepin on =p |
I only just got the test6 update down the way I liked it... the good news is that -test7 is the last real cleanup release for a while so from now on stability will increase and we will have more solid base on which to put the bootsplah port - so I'm actually happy that test7 was released today. I've been waiting for Linus to declare a hard freeze for sometime now - seeing as we would want 2.6.0 out by late December or early January we can't keep changing the system over and over again. |
|
Back to top |
|
|
nepenthe Guru
Joined: 14 Jul 2002 Posts: 317 Location: Sanford, ME
|
Posted: Thu Oct 09, 2003 12:17 am Post subject: |
|
|
bootsplash port?? Haven't heard about that, will have to look into =)
hehe so you're saying it won't be to long until next love sources |
|
Back to top |
|
|
Lovechild Advocate
Joined: 17 May 2002 Posts: 2858 Location: Århus, Denmark
|
Posted: Thu Oct 09, 2003 12:32 am Post subject: |
|
|
nepenthe wrote: | bootsplash port?? Haven't heard about that, will have to look into =)
hehe so you're saying it won't be to long until next love sources |
Well some of us decided it would be good fun to try to port bootsplash to kernel 2.6 - we haven't started though but I intend to dedicate time tomorrow to reading the code so I can get an idea of the work required. |
|
Back to top |
|
|
|