Kernels, kernels, kernels

iphitus · 2006-10-17 09:27:51

brain0 wrote:

Dusty, it was intended to informational only, but some developer discussions may have sneaked in.

oops. at least the world domination etc plans weren't mentioned -- then we'd be in trouble.

in other news... wooo hooo! i finished school today! now exams

patroclo7 · 2006-10-17 13:20:20

hussam wrote:

phrakture, actually there are uses for suspend on desktops. I suspend my machines when for example there is a thunderstorm so if the electricity goes out, I don't lose my work.
In any case the suspend in the stock kernel works fine but the suspend2 scripts are a bit more powerful.

This comparison is slightly misleading: the suspend2 scripts can be used with any suspension method, included that in the stock kernel: just select the preferred method in /etc/hibernate/hibernate.conf. Thus, the right comparison is between the kernel methods, the scripts are a completely different problem.

brain0 · 2006-10-17 15:55:45

The suspend2 method is still more powerful. If you follow discussions on this topic on lkml, it appears that the reasons for not merging suspend2 are not really technical, but rather political (most of it is NIH syndrome).

Dusty · 2006-10-17 18:36:33

brain0 wrote:

Dusty, it was intended to informational only,

Of course. Just wanted to keep it that way.

I'm typing **#$#ing qwerty right now. Keeps me terse. Don't mean to be handing down orders, its forum "advising".

Dusty

brain0 · 2006-10-18 18:18:21

I updated to linux 2.6.19-rc2-mm1. There are some bugs in this release:

You will get the following messages on boot, but it should work anyway:

BUG: warning at fs/inode.c:1389/i_size_write()
 [<c0185680>] i_size_write+0x70/0x80
 [<c018e5ee>] simple_commit_write+0x7e/0xb0
 [<c018e570>] simple_commit_write+0x0/0xb0
 [<c0179273>] __page_symlink+0x93/0x170
 [<c017936b>] page_symlink+0x1b/0x20
 [<c01b6063>] ramfs_symlink+0x43/0xc0
 [<c0179cb7>] permission+0x137/0x140
 [<c0179f14>] vfs_symlink+0xc4/0x120
 [<c017c692>] sys_symlinkat+0xb2/0xe0
 [<c011d0cc>] do_page_fault+0x32c/0x630
 [<c017c6e0>] sys_symlink+0x20/0x30
 [<c01032c8>] syscall_call+0x7/0xb
 =======================

Also, ide doesn't seem to work for me. But I can use my ide controller with the new libata pata driver without trouble. I hope there will be patches for these issues soon.

iphitus · 2006-10-18 23:28:43

brain0 wrote:

I updated to linux 2.6.19-rc2-mm1. There are some bugs in this release:
You will get the following messages on boot, but it should work anyway:
BUG: warning at fs/inode.c:1389/i_size_write()
 [<c0185680>] i_size_write+0x70/0x80
 [<c018e5ee>] simple_commit_write+0x7e/0xb0
 [<c018e570>] simple_commit_write+0x0/0xb0
 [<c0179273>] __page_symlink+0x93/0x170
 [<c017936b>] page_symlink+0x1b/0x20
 [<c01b6063>] ramfs_symlink+0x43/0xc0
 [<c0179cb7>] permission+0x137/0x140
 [<c0179f14>] vfs_symlink+0xc4/0x120
 [<c017c692>] sys_symlinkat+0xb2/0xe0
 [<c011d0cc>] do_page_fault+0x32c/0x630
 [<c017c6e0>] sys_symlink+0x20/0x30
 [<c01032c8>] syscall_call+0x7/0xb
 =======================
Also, ide doesn't seem to work for me. But I can use my ide controller with the new libata pata driver without trouble. I hope there will be patches for these issues soon.

there's a possible patch for the fs/inode.c issue on -mm in the announce thread on LKML, however the message is harmless in most cases, and is a false positive for another bug.

as for your ide..... in 2.6.19, there is no /dev/hdxx, it's all renamed /dev/sdxx. may as well just use the the libata driver anyway, that's where the development is going, its just a matter of time before the other ones go.

James

brain0 · 2006-10-19 05:54:14

iphitus wrote:

there's a possible patch for the fs/inode.c issue on -mm in the announce thread on LKML, however the message is harmless in most cases, and is a false positive for another bug.

I didn't see it, I'll have another look this evening.

as for your ide..... in 2.6.19, there is no /dev/hdxx, it's all renamed /dev/sdxx. may as well just use the the libata driver anyway, that's where the development is going, its just a matter of time before the other ones go.

The old ide subsystem is still in place and still supposed to work. Not all drivers have been ported to libata, and they haven't been tested very much. It is even recommended to stay with the ide subsystem for now until libata got more testing and bugfixing.
With 2.6.19-rc1-mm1, I could use both libata and ide, but with -rc2-mm1, I can't use ide any more - the hd* devices just don't show up in /sys/block.

superstoned · 2006-10-21 17:47:41

ehm, any news on a beyond-2.6.18 release? got some fighting now between mkinitcpio and mkinitrd going on... ;-)

tomk · 2006-10-21 21:14:20

iphitus will update beyond next month, after he has completed his exams.

Snarkout · 2006-10-21 21:39:43

Every time I see this thread title it reminds me of that "..badger, badger, badger..." flash from a few years ago.

brazzmonkey · 2006-10-21 21:40:54

(it'd be nice to have evolution-mission kernel as well...)

superstoned · 2006-10-21 21:56:48

tomk wrote:

iphitus will update beyond next month, after he has completed his exams.

great. i hope he'll get nice grades

Xarturkhann · 2006-10-21 22:46:04

brazzmonkey wrote:

(it'd be nice to have evolution-mission kernel as well...)

I agreed...

brain0 · 2006-10-23 14:41:52

Okay, kernel26mm 2.6.19.rc2.mm2 is up.

The only problem I have seen so far is that legacy IDE still doesn't work. Can anyone else confirm that for his IDE controller? I have it working with pata_sis (libata), but not with sis5513 (legacy ide).

byte · 2006-10-24 04:50:01

Installed kernel26mm, confirmed to work with both ide and libata.

mkinitcpio.conf for ide:
MODULES="ide_disk alim15x3 jfs"

mkinitcpio.conf-libata:
MODULES="sd_mod pata_ali jfs"

Just the 'base' hook for both of them.

brain0 · 2006-10-24 10:12:11

brain0 wrote:

* If you boot from sata you NEED mkinitcpio 0.5.8 or later. You also need to use the "ata" hook instead of the "sata" hook for now. Actually, you can use "ata sata", so your mkinitcpio.conf will work for both the stock and the -mm kernel.
* If you boot from ide, you can leave everything as is. If you like, you can replace "ide" with "ata" in your mkinitcpio configuration (this - again - requires mkinitcpio 0.5.8 or later). But if you do this, all your ide devices will show up as sdX instead of hdX. If you do this and have both sata and ide, you may run into trouble with the controllers randomly switching their order, so you have to define it by explicitly loading the modules in the right order. (note for jmicron users: jmicron is only available in libata, not in ide)

mkinitcpio 0.5.9 or later change this behaviour again (it will stay that way now). This ONLY affects -mm users now:

* the "ata" hook is no more, it was only a temporary solution
* sata users have to use the "sata" hook again. It now works with older (<=2.6.18) and newer kernels. When we move to 2.6.19, sata users won't notice any changes.
* ide users now have the choice between the "ide" hook (old ide subsystem) and the "pata" hook (new libata-based ide drivers). When we move to 2.6.19, most ide users will stay with ide, because pata isn't completed yet. Some users (in particular jmicron) will have to move to pata, because thier drivers won't be merged into the legacy ide subsystem.

@byte: The "bug" was a mistake on mkinitcpio, -mm works fine here with ide and pata now. You didn't habe the bug, because you added all modules manually.

brain0 · 2006-10-30 15:05:07

I uploaded 2.6.19-rc3-mm1. The "no processes left in this runlevel" bug is fixed, but it now seems that wireless is broken. Please test and have fun.

brain0 · 2006-10-31 20:01:27

Okay. 2.6.19-rc4-mm1 is up. The ACPI dsdt patch applies again as Andrew has (according to lkml) reverted some patches, among them the acpi patch. It doesn't seem as broken as rc3-mm1., although my wireless (rt2500) is still weird, essid length is off-by-one. But I got it working and am now online with rc4-mm1, works good so far.

Cagnulein · 2006-11-01 08:48:50

# uname -a
Linux Cagnulein 2.6.19-rc4-mm1 #1 SMP PREEMPT Tue Oct 31 13:50:54 CET 2006 i686 AMD Athlon(tm) AuthenticAMD GNU/Linux

Works very well with new nvidia beta driver
Good job

brain0 · 2006-11-01 09:45:56

In case you use ide: Did you also try pata? It needs some testing

Cagnulein · 2006-11-01 10:35:27

yeah, i use 2 ide hard disk and a ide dvd-writer...how could i test pata? could you give some links?
Thanks

hightower · 2006-11-01 11:03:23

* ide users now have the choice between the "ide" hook (old ide subsystem) and the "pata" hook (new libata-based ide drivers). When we move to 2.6.19, most ide users will stay with ide, because pata isn't completed yet. Some users (in particular jmicron) will have to move to pata, because thier drivers won't be merged into the legacy ide subsystem.

You have to change the hooks in /etc/mkinitcpio.conf

hightower

brain0 · 2006-11-01 11:10:35

Create a copy of your mkinitcpio.conf:

# cp /etc/mkinitcpio.conf /etc/mkinitcpio-pata.conf

Then edit mkinitcpio-pata.conf, remove "ide" and add "pata" instead. Now edit the file /etc/mkinitcpio.d/kernel26mm.preset. It should look like this:

# mkinitcpio preset file for kernel26mm

########################################
# DO NEVER EDIT THIS LINE:
source /etc/mkinitcpio.d/kernel26mm.kver
########################################

PRESETS=('default' 'fallback')

default_config="/etc/mkinitcpio.conf"
default_image="/boot/kernel26mm.img"

fallback_config="/etc/mkinitcpio.d/mkinitcpio-kernel26mm-fallback.conf"
fallback_image="/boot/kernel26mm-fallback.img"

Change it to look like this:

# mkinitcpio preset file for kernel26mm

########################################
# DO NEVER EDIT THIS LINE:
source /etc/mkinitcpio.d/kernel26mm.kver
########################################

PRESETS=('default' 'fallback' 'pata')

default_config="/etc/mkinitcpio.conf"
default_image="/boot/kernel26mm.img"

fallback_config="/etc/mkinitcpio.d/mkinitcpio-kernel26mm-fallback.conf"
fallback_image="/boot/kernel26mm-fallback.img"

pata_config="/etc/mkinitcpio-pata.conf"
pata_image="/boot/kernel26mm-pata.img"

After that, run the command

# mkinitcpio -p kernel26mm

It should now generate three images, the default image, the fallback image and a new pata image. To test pata, you have to try to boot using kernel26mm-pata.img instead of kernel26mm.img.

One problem though: Your hard drives will now be sdXY instead of hdXY. But if you switch back to ide, they will he hdXY again, so changing your bootloader and fstab will always break either of them. That is why you should consider using Persistent block device naming (this also works if you don't experiment with -mm, people have it running with stock arch kernel as well).

If all this sounds awfully complicated to you, then rather let it be before you break your system. I am just curious how well pata works for other people.

Cagnulein · 2006-11-01 11:14:23

mmm...it looks intresting but now i have only one computer so i'd rather have a "usable" one

Maybe next week 8)

brain0 · 2006-11-01 11:16:12

Cagnulein wrote:

mmm...it looks intresting but now i have only one computer so i'd rather have a "usable" one

The point of using persistent naming is that you can switch back and forth between ide and pata. This way you can always use ide again if pata fails. I have it running this way (forgot to boot into -mm this morning, damn *g).

Arch Linux

#26 2006-10-17 09:27:51

Re: Kernels, kernels, kernels

#27 2006-10-17 13:20:20

Re: Kernels, kernels, kernels

#28 2006-10-17 15:55:45

Re: Kernels, kernels, kernels

#29 2006-10-17 18:36:33

Re: Kernels, kernels, kernels

#30 2006-10-18 18:18:21

Re: Kernels, kernels, kernels

#31 2006-10-18 23:28:43

Re: Kernels, kernels, kernels

#32 2006-10-19 05:54:14

Re: Kernels, kernels, kernels

#33 2006-10-21 17:47:41

Re: Kernels, kernels, kernels

#34 2006-10-21 21:14:20

Re: Kernels, kernels, kernels

#35 2006-10-21 21:39:43

Re: Kernels, kernels, kernels

#36 2006-10-21 21:40:54

Re: Kernels, kernels, kernels

#37 2006-10-21 21:56:48

Re: Kernels, kernels, kernels

#38 2006-10-21 22:46:04

Re: Kernels, kernels, kernels

#39 2006-10-23 14:41:52

Re: Kernels, kernels, kernels

#40 2006-10-24 04:50:01

Re: Kernels, kernels, kernels

#41 2006-10-24 10:12:11

Re: Kernels, kernels, kernels

#42 2006-10-30 15:05:07

Re: Kernels, kernels, kernels

#43 2006-10-31 20:01:27

Re: Kernels, kernels, kernels

#44 2006-11-01 08:48:50

Re: Kernels, kernels, kernels

#45 2006-11-01 09:45:56

Re: Kernels, kernels, kernels

#46 2006-11-01 10:35:27

Re: Kernels, kernels, kernels

#47 2006-11-01 11:03:23

Re: Kernels, kernels, kernels

#48 2006-11-01 11:10:35

Re: Kernels, kernels, kernels

#49 2006-11-01 11:14:23

Re: Kernels, kernels, kernels

#50 2006-11-01 11:16:12

Re: Kernels, kernels, kernels

Board footer