Download:
http://www.puppylinux.asia/tpp/SitHeelS ... nd-lib.tgz
user=puppy password=linux
Size: 39,023,664 bytes
md5sum:
8a238e9078f58aa8fa11ea0c399dc0ad 2.6.31.6-rt19-kernel-and-lib.tgz
Warning: The usual disclaimer. Try this kernel only if you know what you're doing and are capable of booting from another Puppy to restore things in case it blows up. It has full printk turned on, so be aware that you will see a lot of logging messages.
Second warning: this kernel breaks the XOrg which is supplied with Karmic Woof. I will post a correctly-compiled XOrg in a day or two.
I tested it on a full install of ttuuxxx's Karmic Woof UPup-450. This is the kernel I describe at
http://www.murga-linux.com/puppy/viewto ... =30#363009
I am supplying it as a .tgz to discourage people from treating it as an ordinary PET package.
Known faults: The Toshiba laptop extras are broken. Not sure if ndiswrapper works or not. Eicon network adapters probably won't work.
I must retract the statement I made in the other thread, about Seamonkey being more stable with this kernel. The Seamonkey which is supplied with Karmic Woof is not more stable.
However, I am finding that the combination of this kernel plus my heavily-patched gcc...is a winner.
If someone will tell me where I can download the complete Debian kernel patchset for 2.6.31.n, preferably 2.6.31.6, or if that's nonexistent then I'll settle for 2.6.30.n, --surely it is somewhere within http://debian.osuosl.org/debian/? (***edited: never mind, found them***) --then I will release this kernel with the Debian patchset, as well as the Ubuntu Karmic Koala 9.10 2.6.31-14.48 patchset (which it has).
2.6.31.6-rt-smp kernel for Karmic Woof UPup-450
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
I tested first in a full install of 4.3.1--yes, that was by accident--i forgot which was where. It booted. Xorg worked. I was using a radeon driver posted by aragon (I think--I 'll check). Network Wizard did not work for my Belkin Wireless USB. An anomaly was that the Wireless Wizard showed eth) and wlan0 as normal + 2 firewire ports that I have (which it does not usually). The wlan0 did not work--it could not bring the network up with ifconfig wlan0 up.
I then installed the 2.6.31.6 on a 4.32 full install. It booted. Xorg Worked. 0n this one I have the ati proprietary fglrx driver 9.10 that I compiled. It now has a problem with scrolling, especially upward, being very slow and jerky--the Ubuntu fglrx driver had this same problem which was fixed when I compiled my own. Also Xorgwizard made some errors in the xorg.conf file--I now have three Section: Device with different drivers, vesa, radeonhd, and fglrx. Come to think of it I am not sure which is running fglrx or radeonhd, but not vesa as the refresh is 85Hz. I had the same problem with the wirelss networking, could not bring network up. I connected the eth0 and that did work--i am using it now to post. GL is not running however so no glxinfo.. I would need the kernel sources and devx to try to (re)compile the ati driver.
SRWare Iron 4.0.227 for linux is running. Desktop looks right. All the basic programs start and run. The scrolling is the only slowdown. After I post I will try to adjust the video driver.
I am back now. I reverted the xorg.conf to the previous, with the radeonhd driver downloaded from Ubuntu package repo. Video is working nicely, no scrolling slowdown. This is a Radeon HD 3850.
Things seem fast, but this computer always seems pretty fast with Puppy (Phenom 9850, 2,8GHz). Certainly no slower than 2.6.30.5.
Anything I can check for you? The only show stopper is the Wireless not working. I think the problem with scrolling might be what happens when one uses a fglrx that is not exactly right for the OS--fixed by compiling or getting exact version. This has the gcc_devx pup you made and kernel sources will come along eventually.
Lately I have been installing the devx routinely (some things seem to run better--crossover in particular), and the latest success with the ati proprietary driver makes me think to compile anything I can (the old gentoo in me
Therefore a Puppy with modern tweaked compiling features seems like a good thing.
GLX should be working--I have added the necessary files and it worked with the other kernel--so xorg is broken there.
Does this only work with a full install? I have tried to install in 4.50, which is a frugal install, but it will not boot. I get a fatal error "cannot load /lib/2,6,31..../dep.mod, no such file ...."
.
I then installed the 2.6.31.6 on a 4.32 full install. It booted. Xorg Worked. 0n this one I have the ati proprietary fglrx driver 9.10 that I compiled. It now has a problem with scrolling, especially upward, being very slow and jerky--the Ubuntu fglrx driver had this same problem which was fixed when I compiled my own. Also Xorgwizard made some errors in the xorg.conf file--I now have three Section: Device with different drivers, vesa, radeonhd, and fglrx. Come to think of it I am not sure which is running fglrx or radeonhd, but not vesa as the refresh is 85Hz. I had the same problem with the wirelss networking, could not bring network up. I connected the eth0 and that did work--i am using it now to post. GL is not running however so no glxinfo.. I would need the kernel sources and devx to try to (re)compile the ati driver.
SRWare Iron 4.0.227 for linux is running. Desktop looks right. All the basic programs start and run. The scrolling is the only slowdown. After I post I will try to adjust the video driver.
I am back now. I reverted the xorg.conf to the previous, with the radeonhd driver downloaded from Ubuntu package repo. Video is working nicely, no scrolling slowdown. This is a Radeon HD 3850.
Things seem fast, but this computer always seems pretty fast with Puppy (Phenom 9850, 2,8GHz). Certainly no slower than 2.6.30.5.
Anything I can check for you? The only show stopper is the Wireless not working. I think the problem with scrolling might be what happens when one uses a fglrx that is not exactly right for the OS--fixed by compiling or getting exact version. This has the gcc_devx pup you made and kernel sources will come along eventually.
Lately I have been installing the devx routinely (some things seem to run better--crossover in particular), and the latest success with the ati proprietary driver makes me think to compile anything I can (the old gentoo in me
Therefore a Puppy with modern tweaked compiling features seems like a good thing.GLX should be working--I have added the necessary files and it worked with the other kernel--so xorg is broken there.
Does this only work with a full install? I have tried to install in 4.50, which is a frugal install, but it will not boot. I get a fatal error "cannot load /lib/2,6,31..../dep.mod, no such file ...."
.
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
Hi playdayz, thanks very much for testing. I'm curious what would happen if you recompile fglrx under this kernel.
I think the most efficient use of my time will be to go ahead and add the Debian 2.6.31 patchset into my kernel source, also I think I'll give it the Gentoo patches as well, and then I'll post the source as well as a compiled kernel.
Also I want to build an XOrg 7.5 (xserver 1.7) patched up to even newer than the Fedora fc12 version. Mesa has a new version just out, 7.6.1, and glut has a fairly recent new version too.
So I think what I will devote my time to, is just finishing-up the Openbox variant which I started building last August, and in it I will incorporate the new kernel and XOrg.
And now to start giving my kernel source the Debian and Gentoo patches, and compile a fully-patched XOrg 7.5. I'll be back up for air in two or three days.
If it runs on the Broadcom 4312 chipset, then that is a known problem; if you run lsmod and see b43, the kernel is correctly identifying the Belkin but the b43 driver doesn't work. There is a proprietary firmware set from Broadcom which does work... (***edited: I supply it, below***)playdayz wrote:...Network Wizard did not work for my Belkin Wireless USB...
If recompiling fglrx under this kernel doesn't work...the scrolling problem you are seeing is, I suspect, the same problem as has been solved recently by the Fedora crew with their updated xorg-x11-server-1.7.1-3.fc12. But you can't just drop it into Puppy, you need to update the whole of XOrg.playdayz wrote:I then installed the 2.6.31.6 on a 4.32 full install. It booted. Xorg Worked. 0n this one I have the ati proprietary fglrx driver 9.10 that I compiled. It now has a problem with scrolling, especially upward, being very slow and jerky...
I think the most efficient use of my time will be to go ahead and add the Debian 2.6.31 patchset into my kernel source, also I think I'll give it the Gentoo patches as well, and then I'll post the source as well as a compiled kernel.
Also I want to build an XOrg 7.5 (xserver 1.7) patched up to even newer than the Fedora fc12 version. Mesa has a new version just out, 7.6.1, and glut has a fairly recent new version too.
So I think what I will devote my time to, is just finishing-up the Openbox variant which I started building last August, and in it I will incorporate the new kernel and XOrg.
When I compiled 2.6.31.5, a full install would not boot because the remount rw command in rc.sysinit was failing to remount. I solved the problem by inserting the same command immediately underneath the remount rw line, but unconditional (i.e. no dependency on state of initrd). I suspect the same problem may be occurring on a frugal install, possibly in the init.playdayz wrote:Does this only work with a full install? I have tried to install in 4.50, which is a frugal install, but it will not boot. I get a fatal error "cannot load /lib/2,6,31..../dep.mod, no such file ...."
And now to start giving my kernel source the Debian and Gentoo patches, and compile a fully-patched XOrg 7.5. I'll be back up for air in two or three days.
Last edited by Sit Heel Speak on Fri 20 Nov 2009, 00:26, edited 4 times in total.
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
Heh. Bear in mind that most of the debug and trace capabilities are turned on. These slow it down. I spent a lot of time in researching the kernel compile-time switches which affect performance, and I think I've got the whole enchilada right. And Ingo Molnar and his friends have done a lot of tuning since 2.6.30.5.playdayz wrote:Things seem fast, but this computer always seems pretty fast with Puppy (Phenom 9850, 2,8GHz). Certainly no slower than 2.6.30.5....
I did one more experiment, a full install of 4.50 to see if there were any significant differences between it and 4.32 vis a vis your kernel. Basically not. It booted. Xorg worked the first time, with vesa. Network Wizard still did not pick up my wireless (4.50 with the 2.6.30.5 kernel does install the wireless automatically); eth0 worked. And oh yes, the browser crashed immediately until I deleted flashplayer 10, just like with 4.50 2.6.30.5. I was able to install the ubuntu xorg_server_radeonhd and manually edit xorg.conf and the radeonhd driver worked fine.
To compile the ati proprietary driver I would need the kernel sources--so that will wait until you have the patches down.
This was great fun. Keep it up!
To compile the ati proprietary driver I would need the kernel sources--so that will wait until you have the patches down.
This was great fun. Keep it up!
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
(deleted by poster, superfluous)
Last edited by Sit Heel Speak on Sat 21 Nov 2009, 07:35, edited 1 time in total.
hey SHS, don't sweat the belkin if you have some productive work to do on the kernel. I will try the suggestions tomorrow
Anyway, I was just reporting some slight info on speed. I use the Peacekeeper browser benchmark a lot, to compare different tweaks, different configurations, different cpu speeds, and so forth. It is a common factor. So, I ran Peacekeeper in Iron in 4.50 with the only difference being the two kernels. With the 2.6.30.5 I got a 4,000--this is my typical setup and typical score. With the 2.6.31.6 I got 3,800, which seems good considering the debugging settings as you say (higher score is better).
I also installed the 4 glx files from ubuntu: mesa-dri, mesa-glx, mesa-utiltities, etc. And the radeonhd driver also from ubuntu, and glx is working with the 2.6.31.6 in 4.50.
Anyway, I was just reporting some slight info on speed. I use the Peacekeeper browser benchmark a lot, to compare different tweaks, different configurations, different cpu speeds, and so forth. It is a common factor. So, I ran Peacekeeper in Iron in 4.50 with the only difference being the two kernels. With the 2.6.30.5 I got a 4,000--this is my typical setup and typical score. With the 2.6.31.6 I got 3,800, which seems good considering the debugging settings as you say (higher score is better).
I also installed the 4 glx files from ubuntu: mesa-dri, mesa-glx, mesa-utiltities, etc. And the radeonhd driver also from ubuntu, and glx is working with the 2.6.31.6 in 4.50.
Last edited by playdayz on Fri 20 Nov 2009, 19:12, edited 1 time in total.
My belkin uses the ZD1211rw module. I did find that the module that comes with both Puppy 4.31 and also Ubuntu 9.10 is a bit flaky anyway.
I compiled my own wireless drivers from the latest sources and it worked a lot better.
http://sourceforge.net/projects/zd1211/develop
This would have to be compiled against whatever kernel you want to use.
In any case, using a RT kernel with wireless is a bit silly, since the less drivers and services the better, I would have thought
I compiled my own wireless drivers from the latest sources and it worked a lot better.
http://sourceforge.net/projects/zd1211/develop
This would have to be compiled against whatever kernel you want to use.
In any case, using a RT kernel with wireless is a bit silly, since the less drivers and services the better, I would have thought
ASUS A1000, 800Mhz PIII Coppermine!, 192Mb RAM, 10Gb IBM Travelstar HDD, Build date August 2001.
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
You may have noticed that my 2.6.31.6 kernel is 9.2MB in size, whereas Barry's kernels are around 2MB.playdayz wrote:Whoa, here is the lsmod from 4.50 with 2.6.31.6...
Most of the drivers that Barry makes as loadable kernel modules (.ko files under /lib/modules), I'm compiling directly into the kernel. Most of the size increase comes from the fact that mine has all the scsi drivers in the universe compiled directly in. Consequently, you can boot straight from an scsi disk with this kernel. Also you should be able to boot straight from a sata drive, a usb-connected hard disk, an external drive on a firewire port, maybe even an LS-120 or an SD compact flash.
Drivers which are compiled directly into the kernel do not show up in lsmod.
Besides compiling the support for every conceivable boot device into the kernel, in general, I compiled into this kernel all the drivers which appear, at least with a superficial look through Google, to have reached a sufficient level of mature reliability that I don't have much doubt of them.
In short, while I'm debugging it, I want to see in lsmod only the unknown or questionable ones. Well, I don't consider usbhid to be questionable, but for some unknown reason the make does not succeed if usbhid is compiled into the kernel.
I will look into the zd1211rw situation tomorrow (~ 18 hours). I have ZD1211 debugging turned on, so it should leave traces of its attempt to start, and its failure, in /var/log/messages. Also. if you run iwevent in another rxvt window while you attempt to bring the Belkin up, perhaps you will see clues.
You can delete your lsmod listing, above, and save John some diskspace.
just for reference, wireless,
my belkin wireless g usb needs zd1211rw, which *is* being loaded by 2.6.31.6.
Then it appears that zd1211rw needs mac80211 and cfg80211, which are not present in lsmod.in 2.6.31.6
Here how it looks in lsmod from 2.6.30.5 in 4.50.
zd1211rw 44816 0
mac80211 166056 1 zd1211rw
cfg80211 64972 2 zd1211rw,mac80211
my belkin wireless g usb needs zd1211rw, which *is* being loaded by 2.6.31.6.
Then it appears that zd1211rw needs mac80211 and cfg80211, which are not present in lsmod.in 2.6.31.6
Here how it looks in lsmod from 2.6.30.5 in 4.50.
zd1211rw 44816 0
mac80211 166056 1 zd1211rw
cfg80211 64972 2 zd1211rw,mac80211
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
Both are compiled directly into the kernel. Perhaps I should instead compile all wireless LAN adapter drivers as modules. I see where pnethood is broken by my kernel, because its bash script expects cifs to be loaded as a module. Perhaps there is a similar need in the setup script for zd1211rw. So, when I compile the Debian-Ubuntu combination kernel, I will make the wireless drivers (and cifs) be all modules. And we will see what happens.playdayz wrote:...Then it appears that zd1211rw needs mac80211 and cfg80211, which are not present in lsmod.in 2.6.31.6...
-
plaguedogs
- Posts: 60
- Joined: Sat 18 Oct 2008, 22:04
-
plaguedogs
- Posts: 60
- Joined: Sat 18 Oct 2008, 22:04
i have this kernel running on real hardware, it all works good. xorg, wireless, i even have firewire drivers (which ive never had before). my sound doesnt work, but its been broken since 2.6.30.5.
when i booted my virtual box, i was able to watch the very verbose messages. a split second flash to the puppy boot splash with weird colours, yellow pink and green. then it would go back to verbose messaging.
on my real hardware, i get about 20seconds of boot messages then my screen blacked out untill the little white "x" shows up, you know....the mouse cursor when xwin starts.
i tried to hack it into woof, after a few tries i had an iso.... kernel panic.
btw, im using spup-451 full hd, i built it yesterday.
when i booted my virtual box, i was able to watch the very verbose messages. a split second flash to the puppy boot splash with weird colours, yellow pink and green. then it would go back to verbose messaging.
on my real hardware, i get about 20seconds of boot messages then my screen blacked out untill the little white "x" shows up, you know....the mouse cursor when xwin starts.
i tried to hack it into woof, after a few tries i had an iso.... kernel panic.
btw, im using spup-451 full hd, i built it yesterday.
-
Sit Heel Speak
- Posts: 2595
- Joined: Fri 31 Mar 2006, 03:22
- Location: downwind
What is the sound system?
I'm still refining the DOTconfig. The original, breaks PNetHood, but I have solved that problem. Now the networking switches are fairly close to what Barry uses, and I've restored most of the drivers to modules which people are used to seeing in lsmod.
Still applying the Debian patchset...
I'm still refining the DOTconfig. The original, breaks PNetHood, but I have solved that problem. Now the networking switches are fairly close to what Barry uses, and I've restored most of the drivers to modules which people are used to seeing in lsmod.
Still applying the Debian patchset...
-
plaguedogs
- Posts: 60
- Joined: Sat 18 Oct 2008, 22:04
im running a dell inspirion 1525 notebook. intel duo core, onboard video and sound.
the module is snd_hda_intel. it is loaded at boot and shows up in lsmod. just like with k 2.6.30.5, the system really thinks sound is working, i just cant hear it.
heres the first few lines of /proc/asound/card0/codec#2
i have never tried very hard to make it work, i just starting using 2.6.29.6, (by trying hard i mean recompiling the kernel). i have modprobed the module in and out, hoping to kickstart it, but shes still deaf.
ive made other posts about this when the 2.6.30.5 kernel showed up in the 4.3 betas, i think its just this mobo, i have yet to find someone else who has this exact problem.
btw, i was pretty lazy when i tried to fit your package into woof, so it might work if i change the package a little to conform to what woof thinks a kernel.pet should look like. ill let you know if i try again.
the module is snd_hda_intel. it is loaded at boot and shows up in lsmod. just like with k 2.6.30.5, the system really thinks sound is working, i just cant hear it.
heres the first few lines of /proc/asound/card0/codec#2
Code: Select all
Codec: SigmaTel STAC9228
Address: 2
Vendor Id: 0x83847616
Subsystem Id: 0x10280242
Revision Id: 0x100402
No Modem Function Group found
Default PCM:
rates [0x7e0]: 44100 48000 88200 96000 176400 192000
bits [0xe]: 16 20 24
formats [0x1]: PCM
ive made other posts about this when the 2.6.30.5 kernel showed up in the 4.3 betas, i think its just this mobo, i have yet to find someone else who has this exact problem.
btw, i was pretty lazy when i tried to fit your package into woof, so it might work if i change the package a little to conform to what woof thinks a kernel.pet should look like. ill let you know if i try again.
I am going to volunteer what little bit I know here, in the hope that if will help you out, plaguedogs. I only ask forgiveness if this post is of no use to anyone.
On official puppy releases, snd_hda_intel audio usually works when using Alsa 1.0.20, but you might have to go into the Alsamixer and mute the headphone to get anything out of your speakers. (I know, this thread is about a different kernel, but....)
Rerwin previously posted this in the Audio section of this forum:
Some versions of Puppy require the headphone muted on different hardware (AC97 Gateway laptop sound, for example), some don't. I've been meaning to research their Alsa versions and default settings, but have not done so.
Previous versions of Alsa required me to add the 'options snd-hda-intel model=dell' line rerwin speaks of to the end of my modprobe.conf file for sound on my Dell netbook.
-Roy
On official puppy releases, snd_hda_intel audio usually works when using Alsa 1.0.20, but you might have to go into the Alsamixer and mute the headphone to get anything out of your speakers. (I know, this thread is about a different kernel, but....)
Rerwin previously posted this in the Audio section of this forum:
EDIT: More info on modem conflicts can be found in the Puppy 4.3 Final bug thread http://www.murga-linux.com/puppy/viewto ... &start=225As instigator of the Conexant modem support, I feel a responsibility to join in. Your information gives me a lot to work with.
I would like to take this up in the 4.3 Bugs thread, since the problems may be aggravated by the 4.3 addition of those drivers, as well as the patching of the ALSA driver to accommodate Conexant HDA modems.
I am concerned that there is an apparent conflict in Puppy's choice of modem drivers. But I see only the ID of the sound card here and need to know the ID of the modem. That ID triggers the driver selection. It could be that the same ID is accociated with both the Intel and HSF drivers. Could someone post (in 4.3 Bugs) what PupScan (in 4.3, with the zp430305.sfs file included) shows for the PCI interfaces for sound and modems? Also please note which sound and modem drivers are actually loaded (lsmod) in the various cases.
I am also interested in attempting to automate handling of situations requiring special "model =" updates to modprobe.conf. But that appears to be of lower priority, if I understand the conclusions you all have come to. Thank you for any further help you can provide to work this out for everyone.
Some versions of Puppy require the headphone muted on different hardware (AC97 Gateway laptop sound, for example), some don't. I've been meaning to research their Alsa versions and default settings, but have not done so.
Previous versions of Alsa required me to add the 'options snd-hda-intel model=dell' line rerwin speaks of to the end of my modprobe.conf file for sound on my Dell netbook.
-Roy