The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

martadinata · 2012-08-26 10:02:09

there was 3.5.3 released

post-factum · 2012-08-27 13:57:21

nous wrote:

I always use the stock -ARCH config and only modify the options related to your patchset. From what I see, crypt target support (dm-crypt) is there. But I remember I saw a thread on arch-dev-public about systemd and encrypted LVM not getting along and another one about broken encrypted LVM with 3.5.2. Right now, only the initial thread message appears on the mailing list archive, so I'm pasting below:
…

Hmm, I saw that mail, but I've got no systemd on my laptop. Possible real bug in 3.5.x with LVM, not related to systemd?

Last edited by post-factum (2012-08-27 13:57:37)

post-factum · 2012-08-27 13:58:17

martadinata wrote:

there was 3.5.3 released

I should prepare kernel for my new laptop, and then release next version.

post-factum · 2012-08-28 10:09:21

Anyway, Arch 3.5.3 works well for me, so I'll start preparing 3.5.3-pf.

martadinata · 2012-08-28 11:41:48

post-factum wrote:

Anyway, Arch 3.5.3 works well for me, so I'll start preparing 3.5.3-pf.

waiting for the release

nous · 2012-08-28 14:49:37

post-factum wrote:

Anyway, Arch 3.5.3 works well for me, so I'll start preparing 3.5.3-pf.

I take it the LVM bug got fixed in 3.5.3?

kokoko3k · 2012-08-28 15:48:54

Sorry to go slightly off topic, but speaking just about tuxonice, doesn't anybody have problems hibernating when the image size is particulary big?

For me, on two systems with kernel 3.4.x, it always fails when i left (say) firefox open with some tabs, so i keep clearing caches before going to hibernation and everything works, but that way my system is unresponsive at resume.
It fails in "preparing images" stage, and then complains about wacky drivers or so, but the same drivers are loaded when the memory is free, so i guess it is wrong.

The same DOES NOT happen with 3.2 (but is on other hardware too)

jakob · 2012-08-28 16:08:20

kokoko3k wrote:

Sorry to go slightly off topic, but speaking just about tuxonice, doesn't anybody have problems hibernating when the image size is particulary big?

Define particularly big. I only have problems if image size > swap size (~1,9 GiB), which happens only when I leave my browser open so I usually exit it before hibernation…

post-factum · 2012-08-28 18:00:29

nous wrote:

I take it the LVM bug got fixed in 3.5.3?

I've got no idea if it is in 3.5.0-3.5.2 as I've updated from 3.4.9 directly to 3.5.3 .

post-factum · 2012-08-28 18:01:09

kokoko3k wrote:

Sorry to go slightly off topic, but speaking just about tuxonice, doesn't anybody have problems hibernating when the image size is particulary big?
For me, on two systems with kernel 3.4.x, it always fails when i left (say) firefox open with some tabs, so i keep clearing caches before going to hibernation and everything works, but that way my system is unresponsive at resume.
It fails in "preparing images" stage, and then complains about wacky drivers or so, but the same drivers are loaded when the memory is free, so i guess it is wrong.
The same DOES NOT happen with 3.2 (but is on other hardware too)

I guess that is well-known issue with memory allocation, and it hasn't been fixed yet.

nous · 2012-08-29 19:17:21

Yup, I get that too. A work-around is to have hibernate script execute "sync; echo 3 >| /proc/sys/vm/drop_caches" at the beginning of hibernation.

kokoko3k · 2012-08-30 06:44:20

jakob wrote:

kokoko3k wrote:
Sorry to go slightly off topic, but speaking just about tuxonice, doesn't anybody have problems hibernating when the image size is particulary big?
Define particularly big. I only have problems if image size > swap size (~1,9 GiB), which happens only when I leave my browser open so I usually exit it before hibernation…

I'd say >400~500MB

jakob · 2012-08-30 08:21:11

nous wrote:

Yup, I get that too. A work-around is to have hibernate script execute "sync; echo 3 >| /proc/sys/vm/drop_caches" at the beginning of hibernation.

I also had that option in /etc/hibernate/common.conf while I used uswsusp (when TOI didn't work with the first few linux 3.0 releases) but removed it when switching back to TOI and have no lags whatsoever after resume.

However, I've got

DisableWritecacheOne /dev/sda # that's where swap resides, cf. man hibernate.conf

enabled… maybe that makes the difference?

post-factum · 2012-08-30 09:38:22

BTW, LVM trouble is well-known: https://bbs.archlinux.org/viewtopic.php?pid=1153611

leodag · 2012-08-30 23:49:09

Hey everybody, my first post here
I found some weird things on the PKGBUILD, specifically on parts concerning aufs3:

mv include/linux/Kbuild "${pkgdir}/usr/src/linux-${_kernver}/include/"

Here's a citation from aufs3's README:

Never copy $PWD/include/linux/Kbuild.

Well, something seems wrong...

Another one, from inside build():

mkdir -p "${pkgdir}/usr/src/linux-${_kernver}/include"

The problem is: _kernver is only set on package_linux-pmod(), so you end creating a linux-/ folder in ${pkgdir}/usr/include/

I'll post a patch here later if I have time.

nous · 2012-08-31 07:46:09

leodag wrote:

Hey everybody, my first post here
I found some weird things on the PKGBUILD, specifically on parts concerning aufs3:
mv include/linux/Kbuild "${pkgdir}/usr/src/linux-${_kernver}/include/"
Here's a citation from aufs3's README:
Never copy $PWD/include/linux/Kbuild.
Well, something seems wrong...

I think aufs3-util won't compile without that, but it's been a long time since I put it to test. Any testing is welcome.

leodag wrote:

Another one, from inside build():
mkdir -p "${pkgdir}/usr/src/linux-${_kernver}/include"
The problem is: _kernver is only set on package_linux-pmod(), so you end creating a linux-/ folder in ${pkgdir}/usr/include/
I'll post a patch here later if I have time.

No, $_kernver is exported to all functions below the parent. You can check the directory tree of linux-pf-headers.

Thanks for your feedback, it's really appreciated.

leodag · 2012-08-31 22:30:53

I did some testing here, and this is what I see:

After

mv include/linux/Kbuild "${pkgdir}/usr/src/linux-${_kernver}/include/"

is executed, it would be later overwritten by

for i in acpi asm-generic config crypto drm generated linux math-emu \
  media mtd net pcmcia scsi sound trace video xen; do
  cp -a include/${i} "${pkgdir}/usr/src/linux-${_kernver}/include/"
done

that replaces it by the kernel's file (which BTW was patched by aufs3-kbuild.patch, adding the lines from aufs' Kbuild) so copying it is disnecessary. Considering it's contents are already added by aufs3-kbuild.patch, it doesn't make any difference but it would be less confusing to do 'rm include/linux/Kbuild' (do not just remove the 'mv include/...' part, because if you do that it would overwrite the kernel's include/linux/Kbuild).

And also about $_kernver: I did a test, and put a 'echo ${_kernver}' right before the command that uses it in build(), and it had no output. If I run a ls ${pkgdir}/usr/src/ at the moment it asks me about the kernel's config, it shows a 'linux-' directory, so I don't know why the package doesn't have it.

Last edited by leodag (2012-08-31 22:32:02)

nous · 2012-09-01 20:53:04

leodag wrote:

I did some testing here, and this is what I see:
After
mv include/linux/Kbuild "${pkgdir}/usr/src/linux-${_kernver}/include/"
is executed, it would be later overwritten by
for i in acpi asm-generic config crypto drm generated linux math-emu \
  media mtd net pcmcia scsi sound trace video xen; do
  cp -a include/${i} "${pkgdir}/usr/src/linux-${_kernver}/include/"
done
that replaces it by the kernel's file (which BTW was patched by aufs3-kbuild.patch, adding the lines from aufs' Kbuild) so copying it is disnecessary. Considering it's contents are already added by aufs3-kbuild.patch, it doesn't make any difference but it would be less confusing to do 'rm include/linux/Kbuild' (do not just remove the 'mv include/...' part, because if you do that it would overwrite the kernel's include/linux/Kbuild).
And also about $_kernver: I did a test, and put a 'echo ${_kernver}' right before the command that uses it in build(), and it had no output. If I run a ls ${pkgdir}/usr/src/ at the moment it asks me about the kernel's config, it shows a 'linux-' directory, so I don't know why the package doesn't have it.

Excellent debugging work, I'll fix it in 3.5.3-pf. Thanks.

[EDIT]
3.5.3-pf is up for grabs.

Last edited by nous (2012-09-01 22:03:15)

leodag · 2012-09-02 11:49:02

I was checking the PKGBUILD because I couldn't build aufs3-utils. It still fails, but now I found why:
When make install_headers is executed, it puts the modified headers in usr/include folder in the kernel's source tree, not in ${pkgdir}/. It isn't copied to ${pkgdir}/ later nor anything, so the aufs_type.h file that ends in ${pkgdir}/usr/src/linux.../include/linux/ is the one copied by

for i in acpi asm-generic config crypto drm generated linux math-emu \
  media mtd net pcmcia scsi sound trace video xen; do
  cp -a include/${i} "${pkgdir}/usr/src/linux-${_kernver}/include/"
done

To solve that, I removed the make headers_install from before the c/p from linux-ARCH, and put it near the end of package_linux-pf-headers(), so that it isn't overwritten by the previous code, with the argument INSTALL_HDR_PATH=${pkgdir}/usr/src/linux-${_kernver}.

A patch: http://pastebin.com/Fx27CZX6

I already tested it, and with that I am able to build aufs3-util (with some modifications to it's PKGBUILD, I still have to add 'export CPPFLAGS="${CPPFLAGS}-I/lib/modules/$(uname -r)/build/include"')

Last edited by leodag (2012-09-02 11:53:38)

Pankaj · 2012-09-04 15:10:14

hello everyone, posting for the first time.
after installing linux pf 3.5.3 from the unofficial repos, i am unable to regenerate initramfs. The process fails while building fglrx. Apparently some header files are missing from
/usr/src/3.5-pf/include/linux folder. those headers are absent in linux-pf-headers too. catalyst-install.log complains the following

make: Entering directory `/usr/src/linux-3.5-pf'
CC [M] /tmp/catalyst.ZcLH2r/firegl_public.o
cc1: fatal error: /usr/src/linux-3.5-pf/include/linux/kconfig.h: No such file or directory
compilation terminated.
make[1]: *** [/tmp/catalyst.ZcLH2r/firegl_public.o] Error 1
make: *** [_module_/tmp/catalyst.ZcLH2r] Error 2
make: Leaving directory `/usr/src/linux-3.5-pf

please suggest any solution for this
thanks

nous · 2012-09-04 21:03:28

Pankaj wrote:

hello everyone, posting for the first time.
after installing linux pf 3.5.3 from the unofficial repos, i am unable to regenerate initramfs. The process fails while building fglrx. Apparently some header files are missing from
/usr/src/3.5-pf/include/linux folder. those headers are absent in linux-pf-headers too. catalyst-install.log complains the following
make: Entering directory `/usr/src/linux-3.5-pf'
CC [M] /tmp/catalyst.ZcLH2r/firegl_public.o
cc1: fatal error: /usr/src/linux-3.5-pf/include/linux/kconfig.h: No such file or directory
compilation terminated.
make[1]: *** [/tmp/catalyst.ZcLH2r/firegl_public.o] Error 1
make: *** [_module_/tmp/catalyst.ZcLH2r] Error 2
make: Leaving directory `/usr/src/linux-3.5-pf
please suggest any solution for this
thanks

I just checked my /usr/src/linux-pf/include/linux directory against linux-ARCH's one and indeed there a couple of thousand header files missing. This shouldn't happen as I copy/paste the packaging sections from linux-ARCH into linux-pf, but I'll investigate.

martadinata · 2012-09-05 04:51:24

also failed on recompile aufs3-utils, using my local language, but also can't find the kconfig.h

DKMS make.log for vboxhost-4.1.20 for kernel 3.5-pf (x86_64)
Rab Sep  5 11:46:44 WIT 2012
make: Memasuki direktori `/usr/src/linux-3.5-pf'
  LD      /var/lib/dkms/vboxhost/4.1.20/build/built-in.o
  LD      /var/lib/dkms/vboxhost/4.1.20/build/vboxdrv/built-in.o
  CC [M]  /var/lib/dkms/vboxhost/4.1.20/build/vboxdrv/linux/SUPDrv-linux.o
cc1: fatal error: /usr/src/linux-3.5-pf/include/linux/kconfig.h: Tidak ada berkas atau direktori seperti itu
kompilasi berakhir.
make[2]: *** [/var/lib/dkms/vboxhost/4.1.20/build/vboxdrv/linux/SUPDrv-linux.o] Error 1
make[1]: *** [/var/lib/dkms/vboxhost/4.1.20/build/vboxdrv] Error 2
make: *** [_module_/var/lib/dkms/vboxhost/4.1.20/build] Error 2
make: Meninggalkan direktori `/usr/src/linux-3.5-pf'

Last edited by martadinata (2012-09-05 04:52:28)

nous · 2012-09-05 07:47:15

I got it. It seems leodag's suggestion to relocate the 'make headers_install' directive, left a lot of header files out. I'm working on an updated PKGBUILD right now.

Pankaj · 2012-09-05 08:43:25

nous wrote:

I got it. It seems leodag's suggestion to relocate the 'make headers_install' directive, left a lot of header files out. I'm working on an updated PKGBUILD right now.

Yeah the problem is solved now. Thank you nous.

martadinata · 2012-09-07 03:06:00

kernel panic on hibernate? or just me?

Arch Linux

#576 2012-08-26 10:02:09

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#577 2012-08-27 13:57:21

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#578 2012-08-27 13:58:17

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#579 2012-08-28 10:09:21

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#580 2012-08-28 11:41:48

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#581 2012-08-28 14:49:37

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#582 2012-08-28 15:48:54

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#583 2012-08-28 16:08:20

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#584 2012-08-28 18:00:29

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#585 2012-08-28 18:01:09

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#586 2012-08-29 19:17:21

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#587 2012-08-30 06:44:20

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#588 2012-08-30 08:21:11

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#589 2012-08-30 09:38:22

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#590 2012-08-30 23:49:09

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#591 2012-08-31 07:46:09

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#592 2012-08-31 22:30:53

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#593 2012-09-01 20:53:04

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#594 2012-09-02 11:49:02

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#595 2012-09-04 15:10:14

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#596 2012-09-04 21:03:28

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#597 2012-09-05 04:51:24

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#598 2012-09-05 07:47:15

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#599 2012-09-05 08:43:25

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#600 2012-09-07 03:06:00

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

Board footer