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

kostett · 2012-06-13 18:30:49

yahoo, i downloaded an update from pfkernel repo (3.4.1-3) and all works now! hybernate in my hands, very thank you, dear nous

nous · 2012-06-13 21:15:37

kostett wrote:

yahoo, i downloaded an update from pfkernel repo (3.4.1-3) and all works now! hybernate in my hands, very thank you, dear nous

Glad it worked (for me too). Now all I want is zram and zcache build again with TOI.

gee · 2012-06-14 02:46:39

Hey nous,

it seems the os version including "-pf" somehow breaks strace:

strace xbmc
> strace: Bad OS release string: '3.4-pf'

Any idea there?

Thanks

nous · 2012-06-14 05:10:39

gee wrote:

Hey nous,
it seems the os version including "-pf" somehow breaks strace:
strace xbmc
> strace: Bad OS release string: '3.4-pf'
Any idea there?
Thanks

I've seen that too and I was puzzled. Strange thing is, it works with '3.4.1-ARCH'...

[EDIT]
Just took a look at strace.c:

/* u.release has this form: "3.2.9[-some-garbage]" */
1424         rel = 0;
1425         p = u.release;
1426         for (;;) {
1427                 if (!(*p >= '0' && *p <= '9'))
1428                         error_msg_and_die("Bad OS release string: '%s'", u.release);
1429                 /* Note: this open-codes KERNEL_VERSION(): */
1430                 rel = (rel << 8) | atoi(p);
1431                 if (rel >= KERNEL_VERSION(1,0,0))
1432                         break;
1433                 while (*p >= '0' && *p <= '9')
1434                         p++;
1435                 if (*p != '.')
1436                         error_msg_and_die("Bad OS release string: '%s'", u.release);
1437                 p++;
1438         }
1439         return rel;

I also saw a patch to allow for 2-digit kernel releases at strace-devel.

Last edited by nous (2012-06-14 05:27:59)

kostett · 2012-06-14 16:06:38

hmm, at now, when i start hibernation my computer enving (hanging, locked, not responded to my actions, i don't know how to say that in english) on the "Doing atomic copy/restore" stage..
also, now i can't to launch any DE - after installing new broadcom-wl pkgbuild package i have an lbkmod error on launch gdm and warning when rc.d was loading modules
i'll try to re-install kernel, because removing of broadcom-wl package didn't help
added: yes, it helped, i'm from linux-pf now.. now i'll try to hibernate
added2: O_o i did remake links to kmod in all sbin folders and it works now.. i think that's not for long time and it appear to be crashed soon..
added3: oh shit, that's not works
i found a precedent and solution!!! nous, see here - http://lists.tuxonice.net/pipermail/tux … 06823.html
for Ъ: man has problem same as my and he solve it made recompiling the kernel without TRANSPARENT_HUGEPAGE, it makes TOI works without any problem

Last edited by kostett (2012-06-14 16:44:31)

big_bum · 2012-06-14 19:29:30

Hey I'm just wondering if in the generic kernel IA32 emulation is enabled. In the -pf-core2 it's not.
Is this the single diff from a generic build and a processor specific one?

nous · 2012-06-14 21:34:11

kostett wrote:

i found a precedent and solution!!! nous, see here - http://lists.tuxonice.net/pipermail/tux … 06823.html
for Ъ: man has problem same as my and he solve it made recompiling the kernel without TRANSPARENT_HUGEPAGE, it makes TOI works without any problem

Could you try compiling a kernel, having changed the "Transparent Hugepage Support sysfs defaults" to "madvise" instead of the default "always" and see if you can hibernate? It's in the "Processor type and features" menu. TRANSPARENT_HUGEPAGE is an important option and I wouldn't want to disable it so hastily.

big_bum wrote:

Hey I'm just wondering if in the generic kernel IA32 emulation is enabled. In the -pf-core2 it's not.
Is this the single diff from a generic build and a processor specific one?

It's not. Somehow I overwrote config.x86_64 with the i686 config, which obviously lacks 32bit emulation. Fixing and uploading 3.4.1-4.

kostett · 2012-06-15 05:55:43

nous wrote:

Could you try compiling a kernel, having changed the "Transparent Hugepage Support sysfs defaults" to "madvise" instead of the default "always" and see if you can hibernate? It's in the "Processor type and features" menu. TRANSPARENT_HUGEPAGE is an important option and I wouldn't want to disable it so hastily.

i can but my internet speed is 16 kbit/s now and it will be a further seven days, so, building process will be looong..
i'll be thank if you make for me that kernel.. i'll test it

Last edited by kostett (2012-06-15 07:37:47)

big_bum · 2012-06-15 19:49:35

nous wrote:

It's not. Somehow I overwrote config.x86_64 with the i686 config, which obviously lacks 32bit emulation. Fixing and uploading 3.4.1-4.

Updated the kernel and everything runs fine!

Thank you sir!

Oh, and some other thing: Why isn't the kernel built with the 3.4.x number? uname shows 3.4-pf but pacman -Qi shows the 3.4.1-4. Is this the way you wanted to be, using just the basename of the kernel?

Last edited by big_bum (2012-06-15 19:53:18)

nous · 2012-06-15 21:41:21

big_bum wrote:

Oh, and some other thing: Why isn't the kernel built with the 3.4.x number? uname shows 3.4-pf but pacman -Qi shows the 3.4.1-4. Is this the way you wanted to be, using just the basename of the kernel?

Stripping the patchlevel number is intentional. This ensures binary modules like nvidia continue to work without rebuilding on every bump.

gee · 2012-06-16 06:37:21

nous wrote:

gee wrote:
Hey nous,
it seems the os version including "-pf" somehow breaks strace:
strace xbmc
> strace: Bad OS release string: '3.4-pf'
Any idea there?
Thanks
I've seen that too and I was puzzled. Strange thing is, it works with '3.4.1-ARCH'...
[EDIT]
Just took a look at strace.c:
/* u.release has this form: "3.2.9[-some-garbage]" */
1424         rel = 0;
1425         p = u.release;
1426         for (;;) {
1427                 if (!(*p >= '0' && *p <= '9'))
1428                         error_msg_and_die("Bad OS release string: '%s'", u.release);
1429                 /* Note: this open-codes KERNEL_VERSION(): */
1430                 rel = (rel << 8) | atoi(p);
1431                 if (rel >= KERNEL_VERSION(1,0,0))
1432                         break;
1433                 while (*p >= '0' && *p <= '9')
1434                         p++;
1435                 if (*p != '.')
1436                         error_msg_and_die("Bad OS release string: '%s'", u.release);
1437                 p++;
1438         }
1439         return rel;
I also saw a patch to allow for 2-digit kernel releases at strace-devel.

Weird.
Maybe it will be fixed in the next version of strace then.

thank you.

Lockheed · 2012-06-17 13:26:28

I would suggest adding this patch to the kernel as right now it poses issues when mounting some SD cards:
http://talk.maemo.org/showthread.php?t=72789

nous · 2012-06-18 07:13:26

Lockheed wrote:

I would suggest adding this patch to the kernel as right now it poses issues when mounting some SD cards:
http://talk.maemo.org/showthread.php?t=72789

Do you still have issues with 3.4.2?

Lockheed · 2012-06-18 08:46:16

nous wrote:

Lockheed wrote:
I would suggest adding this patch to the kernel as right now it poses issues when mounting some SD cards:
http://talk.maemo.org/showthread.php?t=72789
Do you still have issues with 3.4.2?

Didn't upgrade yet. Was there any change with this in mind done by you, because if not, I am sure it is not fixed.

By the way, I am getting error during module insallation after compilation of kernel since at least 2 last versions:
Error: module LZO not found.

EDIT:
Yup, as I suspected - still doesn't work on the newest kernel.

[    6.252037] mmc0: SDHCI controller on PCI [0000:15:00.2] using DMA
[  112.789058] mmc0: error -110 whilst initialising SD card

It mounts only every one out of 7 times on average.

Last edited by Lockheed (2012-06-18 10:57:50)

kostett · 2012-06-18 13:06:47

nous wrote:

kostett wrote:
i found a precedent and solution!!! nous, see here - http://lists.tuxonice.net/pipermail/tux … 06823.html
for Ъ: man has problem same as my and he solve it made recompiling the kernel without TRANSPARENT_HUGEPAGE, it makes TOI works without any problem
Could you try compiling a kernel, having changed the "Transparent Hugepage Support sysfs defaults" to "madvise" instead of the default "always" and see if you can hibernate? It's in the "Processor type and features" menu. TRANSPARENT_HUGEPAGE is an important option and I wouldn't want to disable it so hastily.

nous, when I to make a kernel without an TRANSPARENT_HUGEPAGE, my kernel hasn't any tuxonice modules, but them was be enabled in config.. could you make for me an exclusive kernel packages which hasn't TRANSPARENT_HUGEPAGE support? i'll make testing it fully and i'll report to you about its results
so, it may be more comfortable if we will chat via e-mail or jabber.. my are e-mail - kostett@siemenz.ru and xmpp - kostett_tx@jabber.ru
i'll be glad to see you there

Last edited by kostett (2012-06-18 13:11:17)

nous · 2012-06-18 17:41:19

Lockheed wrote:

nous wrote:
Lockheed wrote:
I would suggest adding this patch to the kernel as right now it poses issues when mounting some SD cards:
http://talk.maemo.org/showthread.php?t=72789
Do you still have issues with 3.4.2?
Didn't upgrade yet. Was there any change with this in mind done by you, because if not, I am sure it is not fixed.
By the way, I am getting error during module insallation after compilation of kernel since at least 2 last versions:
Error: module LZO not found.
EDIT:
Yup, as I suspected - still doesn't work on the newest kernel.
[    6.252037] mmc0: SDHCI controller on PCI [0000:15:00.2] using DMA
[  112.789058] mmc0: error -110 whilst initialising SD card
It mounts only every one out of 7 times on average.

Ok, I'll implement this patch on the next occasion. The fixing patch is class10sd_dto14_fix.diff.txt, right?

LZO is built-in, remove it from mkinitcpio.conf.

nous · 2012-06-18 17:52:31

With 3.4.2, zcache and zram compile and work again.

You might have to append the keywords 'zcache' and 'zram.num_devices=N' to the kernel boot parameters, where N your CPU cores, in order to activate zcache and allocate all memory to the CPUs for zram. Also, install zramswap from the AUR and add zramswap to /etc/rc.conf DAEMONS array.

The results are impressive, most impressive.

Last edited by nous (2012-06-18 18:19:31)

Lockheed · 2012-06-18 18:58:24

nous wrote:

Lockheed wrote:
nous wrote:
Do you still have issues with 3.4.2?
Didn't upgrade yet. Was there any change with this in mind done by you, because if not, I am sure it is not fixed.
By the way, I am getting error during module insallation after compilation of kernel since at least 2 last versions:
Error: module LZO not found.
EDIT:
Yup, as I suspected - still doesn't work on the newest kernel.
[    6.252037] mmc0: SDHCI controller on PCI [0000:15:00.2] using DMA
[  112.789058] mmc0: error -110 whilst initialising SD card
It mounts only every one out of 7 times on average.
Ok, I'll implement this patch on the next occasion. The fixing patch is class10sd_dto14_fix.diff.txt, right?

I thought so but I tried compiling the kernel this way, and it did not take effect. I think this file is only for one specific card readers and does not apply to all hardware. So, I looked for other similar content in other drivers in src/mmc/host/ files, but came up short.
I am not an expert in linux drivers.

Maybe what I was trying to do was wrong, because your link handles it differently. I put 'dto = 14;' just before the line with 'reg &= ~DTO_MASK;', and in your file it is higher up.

By the way, looks like the newest version of the kernel has a bug:
I use xscreensaver and it kicks in asking for password when I resume from suspend. But since the latest upgrade, upon resume I can only see mouse cursor, and not the xscreensaver box for entering the password. That means I have to hard-reboot the laptop.

Last edited by Lockheed (2012-06-18 19:00:50)

kostett · 2012-06-18 19:00:32

nous wrote:

With 3.4.2, zcache and zram compile and work again.
You might have to append the keywords 'zcache' and 'zram.num_devices=N' to the kernel boot parameters, where N your CPU cores, in order to activate zcache and allocate all memory to the CPUs for zram. Also, install zramswap from the AUR and add zramswap to /etc/rc.conf DAEMONS array.
The results are impressive, most impressive.

sorry, you say it to me?
OK, I'll try..

el mariachi · 2012-06-18 19:17:53

isn't zcache a virtualmachine thing? also would zram make my system faster if I never need to use swap? (I have 3gb of ram and never use more than 1.5)

jakob · 2012-06-18 20:55:20

nous wrote:

With 3.4.2, zcache and zram compile and work again.
You might have to append the keywords 'zcache' and 'zram.num_devices=N' to the kernel boot parameters, where N your CPU cores, in order to activate zcache and allocate all memory to the CPUs for zram. Also, install zramswap from the AUR and add zramswap to /etc/rc.conf DAEMONS array.
The results are impressive, most impressive.

Hi nous,

how do your box data (RAM) look like? I dont know whether I set up everything right (but it looks good), but I can't feel the great (to be honest any) difference to the speed in compariston to before using zram. But maybe this will only take effect when images get bigger. With a 800 MiB image (4 GB RAM) for 6 GB of swap, there might be no effect observeable… Did anyone else try this?

Best wishes and thanks again for keeping up the fantastic work!

Jakob

nous · 2012-06-18 21:08:42

kostett: I was addressing all users.

el mariachi: zcache transparently (de)compresses swapped pages (using lzo); the overhead is negligible and the disk writes are almost 50% faster. This of course if the system hits swap. Now, with zram deployed, almost twice as much data fit in the physical RAM, as it's made into high priority swapspace(s) (again with negligible overhead), which effectively eliminates the need to swap. In your case, since you don't exhaust your RAM, you won't notice any difference.

Last edited by nous (2012-06-18 21:24:31)

nous · 2012-06-18 21:21:32

jakob wrote:

how do your box data (RAM) look like? I dont know whether I set up everything right (but it looks good), but I can't feel the great (to be honest any) difference to the speed in compariston to before using zram. But maybe this will only take effect when images get bigger. With a 800 MiB image (4 GB RAM) for 6 GB of swap, there might be no effect observeable… Did anyone else try this?

At home, my laptop and desktop have 4GB of RAM and I very rarely use all of it, so there's not much to observe. At work though, I have several core2 workstations with 2GB of RAM running Arch/KDE/LibreOffice, which previously started lagging when someone fired up opera or amarok. In one of them I also run virtualbox with winxp guest OS and it always swapped like crazy. With zram/zcache it's like I've doubled the installed RAM. Read more user feedback on this page.

Lockheed · 2012-06-30 09:22:46

Since the last two kernel updates, I can no longer resume from suspend-to-ram. Teh screen stays blank with a cursor which I can move, but that's it. Hard reset is the only way.

gee · 2012-07-04 07:55:25

Hi nous,

latest kmod from testing branch breaks linux-pf.

I am guessing that is because it is looking for modules in /usr/lib/modules instead of /lib/modules.
Could you change your config for that?
I am guessing it would still work with people using core repo only but not sure.

Thanks

Arch Linux

#476 2012-06-13 18:30:49

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

#477 2012-06-13 21:15:37

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

#478 2012-06-14 02:46:39

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

#479 2012-06-14 05:10:39

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

#480 2012-06-14 16:06:38

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

#481 2012-06-14 19:29:30

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

#482 2012-06-14 21:34:11

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

#483 2012-06-15 05:55:43

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

#484 2012-06-15 19:49:35

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

#485 2012-06-15 21:41:21

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

#486 2012-06-16 06:37:21

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

#487 2012-06-17 13:26:28

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

#488 2012-06-18 07:13:26

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

#489 2012-06-18 08:46:16

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

#490 2012-06-18 13:06:47

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

#491 2012-06-18 17:41:19

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

#492 2012-06-18 17:52:31

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

#493 2012-06-18 18:58:24

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

#494 2012-06-18 19:00:32

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

#495 2012-06-18 19:17:53

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

#496 2012-06-18 20:55:20

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

#497 2012-06-18 21:08:42

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

#498 2012-06-18 21:21:32

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

#499 2012-06-30 09:22:46

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

#500 2012-07-04 07:55:25

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

Board footer