2.6.9-nitro4 "Less is more"

[miseiler]

Nevermind, I can answer that one: No.

Applying the patch and then attempting compilation gives the following error:

Code: Select all

fs/built-in.o(.text+0x771af): In function `mask_ok_common':
: undefined reference to `generic_permission'

A little bit of searching turned up this post.

So I called ld fs/built-in.o, and got about a half a meg's worth of errors I won't reprint here. The patch provided changes fs/reiser4/plugin/security/perm.c, and that's where the mask_ok_common function is, so we can safely assume the problem lies there. Further research follows:

Code: Select all

static int
mask_ok_common(struct inode *inode, int mask)
{
 return generic_permission(inode, mask, NULL);
}

Note how the method used to solve the problem posted in the link above is now the problem itself. Curious now, I go back to the patch and see if the line has been added:

Code: Select all

$ cat reiser4-from-2.6.9-cko3-to-2.6.9-nitro4.patch |grep generic
+       return generic_permission(inode, mask, NULL);

Ah. Yes, the patch has definitely caused the problem there, and indeed the line "return vfs_permission(inode, mask);" was removed, which was of course the problem in linux-2.6.9-rc1-mm5 again as indicated by the link above. Curious, I went back to undo that specific instance in the patched nitro4 and tried to compile it again.

Aaaand it compiles. No debug output besides all the warnings everyone's used to. Unfortunately, I can't test the kernel right now since I have class, but if anyone's curious...just apply the cko3 to nitro4 patch Pepek gave us, then undo the change on L16 of fs/reiser4/plugin/security/perm.c and it should compile.

[lenk]

I can't access the addons directory, getting 403 forbidden. http://sepi.be/nitro/2.6.9-nitro4/addons/

[Enderson]

Aplied the patches against development-sources-2.6.9 OK?

I can't find the reisr4 in menu? Is that supposed to be there? or just selecting reiserfs suppor it will compile support for reiser4 ?

[EDIT]
OK I found!

Unselected the 4k stack at kernel hack.
[EDIT}

[miseiler]

Aplied the patches against development-sources-2.6.9 OK?

No, not okay. This is applied against vanilla 2.6.9 sources. Just use the ebuild provided in the first post. If your patch applied cleanly, I'd be pretty surprised.

I can't find the reisr4 in menu? Is that supposed to be there? or just selecting reiserfs suppor it will compile support for reiser4 ?

I guess they didn't apply cleanly, since reiser4 support IS in the menu for the nitro4 kernel. It is not in the development-sources kernel. My guess is you're just trying to compile development-sources.

[IainCE]

I can't find the reisr4 in menu? Is that supposed to be there? or just selecting reiserfs suppor it will compile support for reiser4 ?

Disable 4k stacks

[Pepek]

Nevermind, I can answer that one: No.

Applying the patch and then attempting compilation gives the following error:

Code: Select all

fs/built-in.o(.text+0x771af): In function `mask_ok_common':
: undefined reference to `generic_permission'

A little bit of searching turned up this post.

So I called ld fs/built-in.o, and got about a half a meg's worth of errors I won't reprint here. The patch provided changes fs/reiser4/plugin/security/perm.c, and that's where the mask_ok_common function is, so we can safely assume the problem lies there. Further research follows:

Code: Select all

static int
mask_ok_common(struct inode *inode, int mask)
{
 return generic_permission(inode, mask, NULL);
}

Note how the method used to solve the problem posted in the link above is now the problem itself. Curious now, I go back to the patch and see if the line has been added:

Code: Select all

$ cat reiser4-from-2.6.9-cko3-to-2.6.9-nitro4.patch |grep generic
+       return generic_permission(inode, mask, NULL);

Ah. Yes, the patch has definitely caused the problem there, and indeed the line "return vfs_permission(inode, mask);" was removed, which was of course the problem in linux-2.6.9-rc1-mm5 again as indicated by the link above. Curious, I went back to undo that specific instance in the patched nitro4 and tried to compile it again.

Aaaand it compiles. No debug output besides all the warnings everyone's used to. Unfortunately, I can't test the kernel right now since I have class, but if anyone's curious...just apply the cko3 to nitro4 patch Pepek gave us, then undo the change on L16 of fs/reiser4/plugin/security/perm.c and it should compile.

Sorry about that stupid mistake. generic_permission should be vfs_permission (only since kernel 2.6.10-rc1 this should be generic_permission). I fix that and now my patch for reiser4 should be good.

Cheers.

[MrApples]

I can't access the addons directory, getting 403 forbidden. http://sepi.be/nitro/2.6.9-nitro4/addons/

its working fine for me

[Pepek]

Sure I can, but first I need to know : with this my patch software suspend with fbsplash works fine or not ?

how is it suposed to work? any settings in hibernate.conf needed? on which vt's have to be splash screens? do you need to boot in silent mode?
when i compile it in the kernel, with or without textmode support, nothing happens.

I don't know exactly how is it supposed to work, because I don't use swsusp, but I think it is supposed to work like bootsplash + swsusp. About settings in hibernate.conf - if in this file are any settings for bootsplash, than that settings are needed for fbsplash too, I think. About on which vt's have to be splash screens - on any you want, because this module check this. About boot in the silent mode - this should work on any mode of fbsplash (both silent and verbose).
You can check is it works, because with this and fbslash enabled (splash is on any vt) swsusp should can suspend machine and then resume it. If hibernate get some errors with suspend/resume about the splash, that means this not work. If hibernate first suspend and than resume w/o any errors and splash is still on that vt's on which one was enabled, that means this work (and than I will be vary happy ).

Cheers.

[lenk]

I can access http://sepi.be/nitro/2.6.9-nitro4/addon ... 2-bk2.diff , but I can't download the files.

[seppe]

Pepek, I tried your gensplash patch for Swsusp, but it doesn't work. Well, I see the gensplash when I boot normally, but I don't see it when I hibernate or resume from a hibernate. But maybe I have to change something in my hibernate.conf?

[cybe]

no LUFS? =(

No, because you can do "emerge lufs" as well to get it

Well, first of all, one needs the lufs kernel module

Code: Select all

(~:$)->mount /mnt/landoFTP
mount failed: No such device
you don't have kernel lufs support (check whether the lufs module is available/loaded)

and I believe the lufs ebuild is still so old that it only has 2.4 kernel support...I've been using the CVS version..

[jdkbx]

I don't know exactly how is it supposed to work, because I don't use swsusp, but I think it is supposed to work like bootsplash + swsusp. About settings in hibernate.conf - if in this file are any settings for bootsplash, than that settings are needed for fbsplash too, I think. About on which vt's have to be splash screens - on any you want, because this module check this. About boot in the silent mode - this should work on any mode of fbsplash (both silent and verbose).
You can check is it works, because with this and fbslash enabled (splash is on any vt) swsusp should can suspend machine and then resume it. If hibernate get some errors with suspend/resume about the splash, that means this not work. If hibernate first suspend and than resume w/o any errors and splash is still on that vt's on which one was enabled, that means this work (and than I will be vary happy ).

Cheers.

right now i'm using fbsplash with the image inside initramfs.cpio.gz, so you can say its compiled in the kernel. i only use verbose mode. when i boot i see the image on vt0. when i shutdown it's still on vt0. as long as i don't start the spalsh service all other terminals are normal. when i suspend, it suspends in text mode. on resume i see the splash image until it the kernel is copied back and it starts to say "reading caches". in that moment the textmode acts like its on a non-fbsplash terminal and puts its output just over the splash image with a black background where it draws. so fbsplash works (like you described it) without your patch. with it, there wasn't any change.

im gonna try with bootsplash support in hibernate.conf now. with and w/o your patch. gonna post results later.

as far as i understand it, bootsplash support means that swsusp2 manipulates the progress bar of a splash image while resuming. if u turn on bootsplash support in hibernate.conf then /usr/local/share/hibernate/scriplets.d/bootsplash will be called. this wants to set things in /proc/splash, that isn't there of course. maybe just adapting it to /proc/fbsplash could help (just guessing).
as far as your patch is concerned, i doesn't do anything on my system. with or without nothing changes. i don't know if bootsplash would work either.

[iBormuth]

Thanks !!!

Nitro is the best if you want reiser4 *and* a stable kernel.

Love it.

[jblechert]

luckily not much trouble, just a minor problem, fbcon runs smooth aslong as X hasn't been started, however, after X was started every time I change the console (not just from X to console but under the several consoles aswell), the screen seems to re-adjust itself and the sound is skipped for about 1/10th second. This also happens when X is closed but has been started before (after a reboot everything works fine again)

[Pepek]

I don't know exactly how is it supposed to work, because I don't use swsusp, but I think it is supposed to work like bootsplash + swsusp. About settings in hibernate.conf - if in this file are any settings for bootsplash, than that settings are needed for fbsplash too, I think. About on which vt's have to be splash screens - on any you want, because this module check this. About boot in the silent mode - this should work on any mode of fbsplash (both silent and verbose).
You can check is it works, because with this and fbslash enabled (splash is on any vt) swsusp should can suspend machine and then resume it. If hibernate get some errors with suspend/resume about the splash, that means this not work. If hibernate first suspend and than resume w/o any errors and splash is still on that vt's on which one was enabled, that means this work (and than I will be vary happy ).

Cheers.

right now i'm using fbsplash with the image inside initramfs.cpio.gz, so you can say its compiled in the kernel. i only use verbose mode. when i boot i see the image on vt0. when i shutdown it's still on vt0. as long as i don't start the spalsh service all other terminals are normal. when i suspend, it suspends in text mode. on resume i see the splash image until it the kernel is copied back and it starts to say "reading caches". in that moment the textmode acts like its on a non-fbsplash terminal and puts its output just over the splash image with a black background where it draws. so fbsplash works (like you described it) without your patch. with it, there wasn't any change.

im gonna try with bootsplash support in hibernate.conf now. with and w/o your patch. gonna post results later.

as far as i understand it, bootsplash support means that swsusp2 manipulates the progress bar of a splash image while resuming. if u turn on bootsplash support in hibernate.conf then /usr/local/share/hibernate/scriplets.d/bootsplash will be called. this wants to set things in /proc/splash, that isn't there of course. maybe just adapting it to /proc/fbsplash could help (just guessing).
as far as your patch is concerned, i doesn't do anything on my system. with or without nothing changes. i don't know if bootsplash would work either.

I was adapted this in my module from /proc/splash to the /proc/fbsplash, but if the hibernate wants to set things in the /proc/splash, than I have to look to the code of hibernate, but in this week I will be very busy, so I don't know when I did it.

Cheers.

[Pepek]

luckily not much trouble, just a minor problem, fbcon runs smooth aslong as X hasn't been started, however, after X was started every time I change the console (not just from X to console but under the several consoles aswell), the screen seems to re-adjust itself and the sound is skipped for about 1/10th second. This also happens when X is closed but has been started before (after a reboot everything works fine again)

If you use vesa-tng try w/o vesa-tng, but with vesa.

Cheers.

[HecHacker1]

I am having a major problem with this kernel... Once i reach a certain number of open windows with Firefox, Thunderbird, and Gxine my computer begins to freeze up. When I look at the RAM and Swap disk monitor that XFCE4 has, I see my RAM shoot up from the normal 100MB usage all the way to fill up my ram! Thats 1GB's worth. Also, after the RAM fills it begins to fill the SWAP disk which is also 1GB. Once the swap disk fills, then both of them fall to normal levels for a split second, only to shoot right back up and start the cycle again...

Once I do a ctrl+alt+backspace to force quit X then everything goes back to normal.

The only new things since the last Nitro kernel is a new build of firefox and the new Folding@Home 5.02 client. I think that something is wrong with the kernel once I reach a certain amount of MB's.

Help.

[infirit]

For the people asking for the lufs kernel module patch!
Download this and apply with

Code: Select all

cd /usr/src/linux-2.6.9-nitro4
bzcat lufs-for-2.6.9-nitro4.patch.bz2 | patch -p1 --dry-run

If you do not see any rejects remove the "--dry-run" part.

[edit]DO NOT TRY THIS PATCH WITH ANY 2.6.10 KERNEL, it is broken because of a change in 2.6.10[/edit]

[jblechert]

If you use vesa-tng try w/o vesa-tng, but with vesa.

Cheers.

I use splash so I don't think this is an option, anyways, I wasn't asking for help, I can live with this bug, I just wanted to point out that it is there, but thanks anyway

[Pepek]

If you use vesa-tng try w/o vesa-tng, but with vesa.

Cheers.

I use splash so I don't think this is an option, anyways, I wasn't asking for help, I can live with this bug, I just wanted to point out that it is there, but thanks anyway

Fbsplash works fine with vesafb and with vesafb-tng, so you can try this w/o vesafb-tng.

Cheers.

[redshift]

I'd like to see the BadRam patch back in, though it seemingly hasn't been updated at the official site since 2.6.5. If no one has created a new patch, I'll post a patch for the next version. I hacked it into nitro4 today and it works fine.

[Francis85]

I was wondering if it could be possible to add the device-mapper patches

http://sources.redhat.com/dm/patches.html

ftp://sources.redhat.com/pub/dm/patches ... m1.tar.bz2

This patch is needed to use dmraid properly (see http://people.redhat.com/~heinzm/sw/dmraid/readme )

[Nate_S]

This new nitro is feelin great so far; first time I've gotten vesafb-tng to work.

Compiles and boots fine, but I cannot install the ivtv drivers on it...

Code: Select all

/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c:427:59: macro "create_workqueue" requires 2 arguments, but only 1 given
/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c: In function `ivtv_init_struct':
/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c:427: error: `create_workqueue' undeclared (first use in this function)
/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c:427: error: (Each undeclared identifier is reported only once
/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c:427: error: for each function it appears in.)
/var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver/ivtv-driver.c:432:57: macro "create_workqueue" requires 2 arguments, but only 1 given

Go to the /var/tmp/portage/ivtv-0.2.0_rc2/work/ivtv-0.2.0-rc2f/driver directory, than open that file (ivtv-driver.c) and change line number 427 from this :

Code: Select all

	itv->vbi_work_queues = create_workqueue("ivtv_vbi");

to this :

Code: Select all

	itv->vbi_work_queues = create_workqueue("ivtv_vbi", PF_SYNCTHREAD);

, than change line 432 too from this :

Code: Select all

	itv->fw_work_queues = create_workqueue("ivtv_fw");

to this :

Code: Select all

	itv->fw_work_queues = create_workqueue("ivtv_fw", PF_SYNCTHREAD);

and than try to compile (now you can't do emerge ..., but you have to use `ebuild /path/to/that/ebuild merge`).

Cheers.

Thanks a lot. I was having the same problem and this fixed it such that it would compile.

however, if you don't mind, I'm having one more problem. When compiling, it gives this:

Code: Select all

  CC [M] /usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.o
/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.c: In function `load_fw_direct':
/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.c:106: warning: implicit declaration of function `open'
/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.c:112: warning: implicit declaration of function `lseek'
/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.c:123: warning: implicit declaration of function `read'
/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-firmware.c:129: warning: implicit declaration of function `close'


and later this...


  LD [M]  /usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv-fb.o
  Building modules, stage 2.
  MODPOST
*** Warning: "close" [/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv.ko] undefined!
*** Warning: "open" [/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv.ko] undefined!
*** Warning: "read" [/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv.ko] undefined!
*** Warning: "lseek" [/usr/local/src/ivtv-0.1.10-pre2-ck100z/driver/ivtv.ko] undefined!

when I try to modprobe it, it fails and gives this in dmesg:

Code: Select all

ivtv: Unknown symbol lseek
ivtv: Unknown symbol read
ivtv: Unknown symbol open
ivtv: Unknown symbol close

Thanks in advance for any help

-Nate

[Phlogiston]

For LIRC problems look here:

http://forums.gentoo.org/viewtopic.php? ... highlight=

Greets

[kos]

I'm having two problems with 2.6.9-nitroX kernels (tried 2.6.9-nitro1 and now -nitro4):

1. it traps on startkde, somewhere in artsd. Here's what screen of death says: (sorry, had to hand-write it)

Code: Select all

unable to handle paging request at virtual address ...
pde = 0000000
Oops: 0002 [#1]
Modules linked in: .. (my usual modules)
..
EIP is at enqueue_task+0x2b/0x40
..
process artsd (...)
Stack ...
Call Trace:
    scheduker_tick+0x2cb/0x320
    do_timer+0x29/0xd0
    timer_interrupt+0x53/0x140
    handle_IRQ_event+0x34/0x60
    do_IRQ+0x90/0x140
    common_interrupt+0x18/0x20
Code: ....
<0> Kernel panic - not syncing! Fatal exception in interrupt

I don't have this problem on 2.6.8.1-nitro6. I also tried to play with pci=noacpi, noapic, disableapic -- no difference.

2. I have problems with mounts. I have two binds in my fstab:

Code: Select all

/wd/ftp  /home/ftp      reiser4         bind                    0 0
/home/ftp  /home/kos/depot/ftproot      reiser4         rbind                   0 0

They work perfectly in 2.6.8.1-ntro6, but 2.6.9-nitro(1,4) forgets to mount them on boot. When I say "mount -a", only first one (/home/ftp) gets mounted. To mount the second one I have to explicitly say "mount /home/kos/depot/ftproot".

PS I see you are discussing fbsplash+swsusp2, I've changed (hacked, better to say) bootsplash scriplet to use fbsplash: http://www.livotov.org/pub/bootsplash
hibernate.conf settings will look like this:

Code: Select all

Bootsplash on
BootsplashConfig ThinkLinux

Where ThinkLinux is my fbsplash theme.
You can use it as a starting point to implement a clean fbsplash scriplet.