2.6.10-nitro4: The Kernel Formerly Known As Hercules

[Gentree]

low latency patch would be the mingo-ll patch that comes from -ck... meaning you've already got it.

Oh good. At least I won't have to hunt around for it.

I was refering to Ingo Molnar's low-latency patches . I misread something on morph-sources and thought he had added it rather than re-added something he had taken out. It appears this is indeed part of ck , so sorry for getting you all exited for nothing.

The good news is , no extra work required , nitro is all I could ever want !

BTW do you realise you're a kernel junkie , bet you cant go 24 hrs .
Luckily help is available, you can now buy patches ... !!

I'm going to see if I use nitro4 before you make it obselete . Good luck.

[discomfitor]

Tiger683: I suppose I could get rid of it if it's really causing problems for you, but I could have sworn I fixed everything related to remap_page_range (I was aware of the issues and attempted to fix them with apparent success). Could you give me a specific example of something that it breaks?

xiaosuo: I can't foresee there being any other use flags added any time soon. The idea behind want-mm is to give two separate nitros: one that is incredibly stable, and one that is more experimental in nature while possibly being "less than stable". The bk patch is experimental stuff, so it will stay in the want-mm. I may change the flag in the future to something like "build-experimental" or something to reflect the nature of this, but that's the only real change I can think of.

Gentree: I've been a kernel junkie since I started rolling my own just prior to love-sources' conception. If I don't have something new to boot every day, I start to get edgy from the stability. I'll always be running the mm/bk version of nitro . My plan for staying off the kernel patching for a few days involves buying the patches, as you said, and playing massive amounts of World of Warcraft (which I did too much of already) since it runs so amazingly well on linux. See my ultra-successful howto thread over here .

[Tiger683]

This patch ONLY removes remap_page_range from kernel headers. Its just a Kernel-developers perversion about deprecated API features. It breaks fglrx(aka. proprietary ATI drivers). Well, lets say it breaks them too...

EDIT:
Performance patches is one thing, but removing features before third party sw adapts is a little TOO close to future kernels....

[Gentree]

Anyone got dmix/jackd working on recentn nitro (2.6.9/10) ?

I've been banging my head against this particular wall for 2 full days and I need to get a handle on what's bugging it.

TIA.

Darckness : seems like you're a bit a wine conaisseur , have you ever tried to get Dragon Speach rec. working through wine?

It's the only thing I miss from Windows (Not that I ever used it much since Win.MTBF<Dragon.trainingPeriod. But for the time it stayed up it was impressive).

[discomfitor]

Tiger683: Good point. I'll make sure to leave that out in future versions.

Gentree: I have dmix/jackd working. Then again, I also have an Audigy 2 ZS which supports hardware mixing. As for Dragon Speech, I've never even heard of it. Can't say I've tried it, but I can say that it's a real pain to get external functions working in wine (printers, microphones) if they aren't working out of the box. What's Dragon Speech do? I might be able to suggest a replacement app since I do quite a bit of sound work!

Somebody with ATI drivers that is having problems building them and is technically proficient:
Try changing all occurances of remap_page_range in the ati drivers to io_remap_page_range and tell me what happens.

[Tiger683]

will try as soon as i get home today.
till now i succeeded by both replacing page with pfn and adding back the remap page r. to the kernel , but the first might be baaaad, but it works and loads...with no_dri in driver enabled it also loads X.....

sometime later Ill try your ver.......
EDIT:
From what i just found on Inet, this might be the golden shot, Ill report back later...

[nlabadie]

nlabadie: there's two versions of nitro4. One uses mm patches, the other does not. Which are you using?

Nevermind, it was the one without mm-sources. After SSH'ing into the laptop running nitro-sources-2.6.10-r4 (keyboard not working), I discovered the following error after running dmesg:

Code: Select all

ACPI: PS/2 Keyboard Controller [KBC] at I/O 0x60, 0x66, irq 1
ACPI: PS/2 Mouse Controller [PS2M] at irq 12
i8042.c: Can't read CTR while initializing i8042.

The dmesg from nitro-sources-2.6.10-r2 (keyboard working):

Code: Select all

i8042.c: Warning: Keylock active.
serio: i8042 AUX port at 0x60,0x64 irq 12
serio: i8042 KBD port at 0x60,0x64 irq 1

Hrmm. Searched for the error "Can't read CTR while initializing i8042" and found multiple references (mostly involving mm-sources) about passing "i8042.noacpi" at boot. Added "i8042.noacpi" to grub.conf, rebooted, and everything is working again.

[cybe]

this patchset has the shfs support???

i dont find it

I patch my sources with the LUFS patch that has dissapeared lately from the nitro sources... it needs just a slight little modification by hand and it will work

[cybe]

2.6.10-nitro4 isnt as good on my setup as the two previous ones I used .. it makes my music skip very easily while nitro2 is goood

[Gentree]

What's Dragon Speech do? I might be able to suggest a replacement app since I do quite a bit of sound work!

Dragon NaturallySpeaking (Prefered) voice recognition software. now owned by Lerhard & Hauspie is just about the best and longest standing PC voice rec. software. Sadly win only.

Like the now abandonded IBM viaVoice and Phillips FreeSpeach (as in beer) it allows you to dicatate text into the mike that it converts to a character stream inserted into the keyboard buffer. ie it truely replaces typing and is not application dependant. (Not just word recognition like the old soundBlaster Voice tools.)

I had this pretty well set up under win98 at one stage and was able to dictate a page of A4 in one hit , speaking very quickly.

Incredibly it only needed about 10 corrections on the page!

I was gob-smacked. Sadly my windows installation screwed up shortly afterwards and the hours I'd spent configuring and training the software to my voice patterns went with it.

The software remains one of the most impressive things I've seen a PC do. Shame it depends on a flakey OS. Hence my interest in seeing it Linux.

I bought v5.5 Fr and it was about as effective with native French dictation which is even more impressive in view of the number of homophones there are in the french language.

Years back they were giving away earlier versions on magazine covers so some free versions should be findable on web or p2p.

Version 4 was good. version 5 and later excellent. Must be at 8 or 9 now but I think the engine changes little after v5, its more front-end niceties and IE integration gimicks.

If you have a dual boot its worth taking a look. Very effective even with minimal training/setup.

Sorry that was abit long but you did ask

[discomfitor]

cybe: are you using the same config as with nitro2? I've had no issues with music skipping/other irritations of that sort. What player do you use?

Gentree: that sounds really cool! I think you've sparked my interest in voice dictation. Probably a bad thing since I'll be working on something related to this for a few days now, but I'll let you know if I find anything.

[LordArthas]

It doesn't work here on my AMD64 (I upgraded from, and then fell-back to, 2.6.9-nitro2). With the same configuration it hangs at boot time after 20 or 30 lines of reports. The last 3 lines are:

Code: Select all

ata3: SATA max UDMA/133 cmd 0x9F0 ctl 0xBF2 bmdma 0xDC00 irq 11
ata4: SATA max UDMA/133 cmd 0x970 ctl 0xB72 bmdma 0xDC08 irq 11
ata3: dev 0 ATA, max UDMA/133, 240119615 sectors:

After this, it just hangs there. I guess there's some problem with SATA. With 2.6.9-nitro2 I get this:

Code: Select all

ata3: SATA max UDMA/133 cmd 0x9F0 ctl 0xBF2 bmdma 0xDC00 irq 11
ata4: SATA max UDMA/133 cmd 0x970 ctl 0xB72 bmdma 0xDC08 irq 11
ata3: dev 0 cfg 49:2f00 82:7c6b 83:7b09 84:4003 85:7c68 86:3a01 87:4003 88:407f
ata3: dev 0 ATA, max UDMA/133, 240119615 sectors:

and then of course booting continues. Output is identical except for the third line, which is missing when booting with 2.6.10-nitro4.

Thanks, M.

[Trevoke]

Is the SMP problem fixed?

[Tiger683]

Hey Darcknes, Take a look at this one:

http://www.linuxant.com/driverloader/wl ... acks.patch

Worth a try? Looks like a nice thing against stack-overflows.... :]

PS. Does anyone know if REGPARM or 4KSTACKS in some configs or
CFLAGS cause permanently segfaults at bootup?( now I only got -O3 -march=athlon-xp and -pipe as custom CFLAGS...)
Can't test the ati option of io_remap_page range before I get this done.


Tiger.

[Evil Dark Archon]

It doesn't work here on my AMD64 (I upgraded from, and then fell-back to, 2.6.9-nitro2). With the same configuration it hangs at boot time after 20 or 30 lines of reports. The last 3 lines are:

Code: Select all

ata3: SATA max UDMA/133 cmd 0x9F0 ctl 0xBF2 bmdma 0xDC00 irq 11
ata4: SATA max UDMA/133 cmd 0x970 ctl 0xB72 bmdma 0xDC08 irq 11
ata3: dev 0 ATA, max UDMA/133, 240119615 sectors:

After this, it just hangs there. I guess there's some problem with SATA. With 2.6.9-nitro2 I get this:

Code: Select all

ata3: SATA max UDMA/133 cmd 0x9F0 ctl 0xBF2 bmdma 0xDC00 irq 11
ata4: SATA max UDMA/133 cmd 0x970 ctl 0xB72 bmdma 0xDC08 irq 11
ata3: dev 0 cfg 49:2f00 82:7c6b 83:7b09 84:4003 85:7c68 86:3a01 87:4003 88:407f
ata3: dev 0 ATA, max UDMA/133, 240119615 sectors:

and then of course booting continues. Output is identical except for the third line, which is missing when booting with 2.6.10-nitro4.

Thanks, M.

unfortunately i've had that problem with a number of custom kernel patchsets, the only one i don't have the problem on is gentoo-dev-sources with staircase patched on top by myself. I haven't checked -mm though.

[Gentree]

Code: Select all

# last one for >2gig ftp.
CFLAGS="-march=athlon-xp -O3 -pipe -mtune=athlon-xp  -fomit-frame-pointer -falign-functions=64
 -D_FILE_OFFSET_BITS=64"

### readded 22.10.2004 + new ldflags
CXXFLAGS="${CFLAGS}  -fvisibility-inlines-hidden"
LDFLAGS="$LDFLAGS -Wl,-O1"

seems perfect here, no sound glitches mplaying mp3 from reiserfs or R4 partitions.

I regard both REGPARM or 4KSTACKS as def no-nos.

[anir]

Hello, I'd like to know how stable this patchset works in a SMP and SMT environment.

Thanks for every answer!

[Tiger683]

Darckness, do you condider include a -mm3 version?

[discomfitor]

Tiger683: I'll put that 16k stacks patch onto the next one if it doesn't cause problems. As for mm versions, I'll put whatever is the newest one into the next release when I do that. Won't be for a few days though.

As for SMT/SMP stuff I run SMT, and I can say that it runs amazingly on this end. Couldn't tell you about SMP though.

[Tiger683]

It seems to me that the system got rreeeaaalllyyy more responsive tooo...
reasonable, due to the fact that each thread gets bigger stack attached...

[samonli]

I have a mobile radeon 7500,enable DRI in the kernel.when it complied,it said:
Warning:/lib/modules/2.6.10-nitro4/x11-drm/radeon.ko needs unknown symbol remap_page_range.

and i can't enable dri when i start X.

Please help me .thanks very much!

[Hobbit_HK]

Does this Nitro has the Reiser4\HT bug?

[KoolKam]

Patch worked great on my system , upgraded from Nitro 2 , for sure feels a bit more responsive using it through vnc for the desktop mostly so far no problems but have nothing fancy compiled into it either like display drivers and such. I would be interested in trying out the 16k stack patch mentioned above if it was at all stable.

Thanks for another great release !!

[Tiger683]

It is VERY stable , breaks nothing so far and applies almost cleanly against notro4

[KoolKam]

It is VERY stable , breaks nothing so far and applies almost cleanly against notro4

Great What does almost cleanly against nitro 4 mean in how it applies , how do I go about trying this out ,applying this patch?
I tried
#patch -p1 ../patchname.patch

but it didnt seem to patch how do I make it work?

Thanks