morph-sources-2.6.12-morph7 (realtime-preempt, updates)

Message

yardbird · Post by **yardbird** » Wed Jul 13, 2005 12:41 am

nightfrost wrote:runing cherries now, seemingly w/o problems. swsusp2 feels a tad faster. Time for old nightfrost to go to sleep. Good night, all.

Thanks and goodnight all from me too, I need sleep

Ateo · Post by **Ateo** » Wed Jul 13, 2005 4:15 am

yardbird wrote:Ok, thanks for the report. I can reproduce the problem, and it's most likely a suspend2 issue. Enabling suspend2 in kernel config or disabling the "suspend2" USE flag and re-emerging (so that the suspend2 patch is not applied when emerging the kernel) should solve the problem. Sorry about that, but I guess we'll have to wait for another suspend2 version from upstream.

Actually, my fault. I have become accustomed to kernel ebuilds without flags (other than symlink and doc) that it didn't even occur to me to disable the suspend2 flag. And doing so, worked (obviously)...

This must be a newly introduced bug with suspend2 because with morph5, I was able to compile the kernel without suspend2 enabled and *with* the suspend2 patch applied (via USE="suspend2")... Regardless, at least your ebuild gives the choice of applying the patch or not. So that's a good thing...

... Off topic

I thought I pass this by you. You may or may not be able to help but it's worth taking a shot and asking...

My laptop has a 4 in 1 flash card reader.

lspci wrote:0000:06:09.3 Unknown mass storage controller: Texas Instruments Texas Instruments PCIxx21 Integrated FlashMedia Controller

I've configured the kernel as I would for a usb hard drive (all the scsi stuff) and assume it's going to show up as a scsi drive (/dev/sgX or something of the like) but I am unable to access it.

Do you know if there is support for this? I've tried all the relative options in the kernel but the device never seems to exist. I'm probably out of luck since this is a relatively new motherboard. Probably under 4.. 5 months I think... I've googled it but nothing useful comes up...

If this was too much too ask, don't bother answering.. hehe. I think I already know what the answer is gonna be...

Thanks for Morph sources. It's doing my laptop justice.

landon · Post by **landon** » Wed Jul 13, 2005 7:11 am

yardbird wrote: @landon: thanks for reporting this. I've found a patch here:

http://seclists.org/lists/linux-kernel/ ... /0614.html

According to that thread the patch solves the bug. I've committed it to the upcoming morph-sources, but would you please test it to see if it does solve the issue?

Sorry for the rather late reply. Yes, it patches cleanly, and does the trick.

I decided to stick with your morph6 work, though

yardbird · Post by **yardbird** » Wed Jul 13, 2005 8:45 am

@ateo: yep, the new suspend2 patch probably introduced the bug. I have little experience with card readers. I set up one on an AMD64 laptop: it was seen as a SCSI disk IIRC and it was accessible through /dev/sdX. But things are not always so simple, and there are card readers that do not use this standard interface. Maybe if you re-emerge pciutils you could get more info about it (when you emerge pciutils it tries to download the latest PCI devices definitions from the net).

@landon: thanks for the report, hope you enjoy morph-sources

Ateo · Post by **Ateo** » Wed Jul 13, 2005 4:40 pm

yardbird wrote:@ateo: yep, the new suspend2 patch probably introduced the bug. I have little experience with card readers. I set up one on an AMD64 laptop: it was seen as a SCSI disk IIRC and it was accessible through /dev/sdX. But things are not always so simple, and there are card readers that do not use this standard interface. Maybe if you re-emerge pciutils you could get more info about it (when you emerge pciutils it tries to download the latest PCI devices definitions from the net).

Ok. Thanks for the tip. I'll keep trying to get it mounted...

Also, one more question. When using suspend2, it is my understanding that I must have 2 commands to pass onto the kernel (from bootloader). One with the resume2=swap:/dev/hdaX (for a fresh boot) and noresume2 to return to the state before hibernation. Am I understanding this correctly?

In other words, I need to option for 2 kernels (on with resume2 and the other with noresume2)...

Thanks for Morph..

nightfrost · Post by **nightfrost** » Wed Jul 13, 2005 5:13 pm

Ateo wrote: Also, one more question. When using suspend2, it is my understanding that I must have 2 commands to pass onto the kernel (from bootloader). One with the resume2=swap:/dev/hdaX (for a fresh boot) and noresume2 to return to the state before hibernation. Am I understanding this correctly?

In other words, I need to option for 2 kernels (on with resume2 and the other with noresume2)...

Thanks for Morph..

Nope, you just need resume2=swap:/dev/hdaX (if you're using the swapwriter of course). If there's no image saved in the swap space (i.e. you haven't hibernated), the kernel will just boot normally.

Ateo · Post by **Ateo** » Wed Jul 13, 2005 5:19 pm

nightfrost wrote:Nope, you just need resume2=swap:/dev/hdaX (if you're using the swapwriter of course). If there's no image saved in the swap space (i.e. you haven't hibernated), the kernel will just boot normally.

Ok. Thanks. I misunderstood the docs then....

nightfrost · Post by **nightfrost** » Thu Jul 14, 2005 7:51 am

Mysteriously, I have suddenly started to recieve these errors. I don't know where they come from, but suspend2 has been working with this very kernel before. Weird

Code: Select all

hibernate: Less than 3 seconds since last resumed. Not suspending.
hibernate: Aborting suspend due to errors in CheckLastResume (use --force to override).

Any ideas?

EDIT: PS: The problem is of course solved with the --force flag, but I don't see why I would need it suddenly...

nightfrost · Post by **nightfrost** » Thu Jul 14, 2005 11:37 am

Hmmm... I might have solved the problem. Here's an attempt to make a long story short:

I started having problems with acpid not registering events, claiming /proc/acpi/events to be busy. All other things acpi worked though, leaving me confused. I finally found a solution to this problem (the acpid daemon must apparently be started before dbus/hal are started, probably since hal nowadays reads acpi events as well). Anyway, after fixing those issues I haven't been getting the swsusp2 error, and the --force flag is no longer needed. But I've only tried it a couple of times, so I'm not sure if this actually is a permanent solution. Anyway, just thought I'd let this out in case anyone bumps into similar issues...

mrfree · Post by **mrfree** » Mon Jul 18, 2005 9:19 am

If you use morph-sources with suspend2, you probably like this ebuild http://bugs.gentoo.org/show_bug.cgi?id=99410 for the new suspend2 userspaceUI (for suspend2-v2.1.9.8 or later)

kohno · Post by **kohno** » Fri Jul 22, 2005 1:10 am

Is there win4lin patch in this source?

yardbird · Post by **yardbird** » Sat Jul 23, 2005 12:53 am

kohno wrote:Is there win4lin patch in this source?

Sorry, not yet in the 2.6.12 series (it is there in the 2.6.11 series though), but it will be back soon.

---------------------

I've released 2.6.12-morph7. It adds (optionally of course) the realtime-preempt patch from Ingo Molnar. It is supposed to reduce latency on desktop system. Read on the website for more details. Apart from this there are the usual updates. Cheers

ahorn · Post by **ahorn** » Sat Jul 23, 2005 6:13 am

http://dev.gentoo.org/~dsd/genpatches/p ... .12-10.htm these upstream patches are all the way great. please think about them.

do you plan to release a 2.6.13 pre-release?

and thanks for including the ext3-stuff right into the mainpatchset.

_sys/sid · Post by **_sys/sid** » Sat Jul 23, 2005 4:45 pm

Code: Select all

drivers/built-in.o(.text+0xdf3d0): In function `pi_init':
: multiple definition of `pi_init'
kernel/built-in.o(.bss+0x98f4): first defined here
ld: Warning: size of symbol `pi_init' changed from 4 in kernel/built-in.o to 897 in drivers/built-in.o
ld: Warning: type of symbol `pi_init' changed from 1 to 2 in drivers/built-in.o
make: *** [.tmp_vmlinux1] Error 1

FouiniX · Post by **FouiniX** » Sun Jul 24, 2005 1:09 pm

Same problem on Arch Linux

yardbird · Post by **yardbird** » Sun Jul 24, 2005 2:55 pm

_sys/sid wrote:

Code: Select all

drivers/built-in.o(.text+0xdf3d0): In function `pi_init':
: multiple definition of `pi_init'
kernel/built-in.o(.bss+0x98f4): first defined here
ld: Warning: size of symbol `pi_init' changed from 4 in kernel/built-in.o to 897 in drivers/built-in.o
ld: Warning: type of symbol `pi_init' changed from 1 to 2 in drivers/built-in.o
make: *** [.tmp_vmlinux1] Error 1

Could you post a bit more of the error output? Did you use "make oldconfig" by chance?

Tanisete · Post by **Tanisete** » Mon Jul 25, 2005 12:51 pm

Hi to all!!!

Some stupid questions:

- I have and amd64 system... Can i use the 1glowmem patch (I think it's only for x86)?
- How can i use the full memory of my system (highmem?)?
- If I use highmem... does it make my system slower??

Thanks a lot, for the answers, and, specially, for morph-sources

.

broch · Post by **broch** » Mon Jul 25, 2005 1:06 pm

1. you have 64-bit kernel so..
2 and 3. memory limitations concerns only 32-bit. systems. 64-bit systems will utilise memory better than 32-bit
64-bit does not have any (reasonable) memory limit. And 64-bit kernel does not have HIHGMEM option at all.

Your question suggest that you have built 32-bit system (taht is why you see HIGHMEM option), so use HIGHMEM option and yes on 32-bit system, you can use 32-bit patches. CPU (AMD64) does not matter here.

Tanisete · Post by **Tanisete** » Mon Jul 25, 2005 2:07 pm

No, I'm using an x86_64 kernel, because i don't have that options. I was only asking, because my system is only using 512MB RAM, and I don't know why. So, if it's not related to a patch... How can I make my system to use the whole memory? The modules are detected, of course

Thanks again!!

broch · Post by **broch** » Mon Jul 25, 2005 5:48 pm

it must be something else then: I have 2GB of RAM recognized at the installation time by 32-bit or 64-bit kernel.

Tanisete · Post by **Tanisete** » Mon Jul 25, 2005 9:01 pm

Yeah, i get the modules recognized... and yeah, it was another thing

.

Now i'm trying to compile new morph-sources (x86_64), and i got some errors in the beginning:

Code: Select all

In file included from arch/x86_64/kernel/mce.c:17:
include/linux/fs.h: In function `lock_super':
include/linux/fs.h:851: aviso: declaración implícita de la función `down'
include/linux/fs.h: In function `unlock_super':
include/linux/fs.h:857: aviso: declaración implícita de la función `up'
arch/x86_64/kernel/mce.c: In function `mce_read':
arch/x86_64/kernel/mce.c:383: aviso: el tipo de dato por defecto es `int' en la declaración
 de `DECLARE_MUTEX'
arch/x86_64/kernel/mce.c:383: aviso: nombres de parámetros (sin tipos) en la declaración de
 la función
arch/x86_64/kernel/mce.c:392: error: `mce_read_sem' no se declaró aquí (primer uso en esta                                                   función)
arch/x86_64/kernel/mce.c:392: error: (Cada identificador no declarado solamente se reporta                                                   una vez
arch/x86_64/kernel/mce.c:392: error: para cada funcion en la que aparece.)
make[1]: *** [arch/x86_64/kernel/mce.o] Error 1
make: *** [arch/x86_64/kernel] Error 2

I compiled it with "fuse" and "rtp" flags, and i activated rtp with voluntary preemtp (no more options activated). I haven't selected suspend, nor supermount. Can anyone help me??

Thanks a lot!!

yardbird · Post by **yardbird** » Thu Jul 28, 2005 1:02 pm

Tanisete wrote:Yeah, i get the modules recognized... and yeah, it was another thing

.

Now i'm trying to compile new morph-sources (x86_64), and i got some errors in the beginning:

Code: Select all

In file included from arch/x86_64/kernel/mce.c:17:
include/linux/fs.h: In function `lock_super':
include/linux/fs.h:851: aviso: declaración implícita de la función `down'
include/linux/fs.h: In function `unlock_super':
include/linux/fs.h:857: aviso: declaración implícita de la función `up'
arch/x86_64/kernel/mce.c: In function `mce_read':
arch/x86_64/kernel/mce.c:383: aviso: el tipo de dato por defecto es `int' en la declaración
 de `DECLARE_MUTEX'
arch/x86_64/kernel/mce.c:383: aviso: nombres de parámetros (sin tipos) en la declaración de
 la función
arch/x86_64/kernel/mce.c:392: error: `mce_read_sem' no se declaró aquí (primer uso en esta                                                   función)
arch/x86_64/kernel/mce.c:392: error: (Cada identificador no declarado solamente se reporta                                                   una vez
arch/x86_64/kernel/mce.c:392: error: para cada funcion en la que aparece.)
make[1]: *** [arch/x86_64/kernel/mce.o] Error 1
make: *** [arch/x86_64/kernel] Error 2

I compiled it with "fuse" and "rtp" flags, and i activated rtp with voluntary preemtp (no more options activated). I haven't selected suspend, nor supermount. Can anyone help me??

Thanks a lot!!

Ugh, sorry for the late reply. I've been pretty much busy lately, damn social^H^H^H^H^H^H real life

I'll try to investigate this error as soon as possible. As a workaround, have you tried to disable MCE-related options (under the "Processor type and features" menu, IIRC)? Does this happen on vanilla too? Are you using suspend2?

Tanisete · Post by **Tanisete** » Thu Jul 28, 2005 2:27 pm

Hi Yardbird!!

Don't worry about that!
I'm using an amd64 without suspend or supermount, but i can't deactivate this option:

Code: Select all

CONFIG_X86_MCE=y

In fact, every time i make menuconfig, that option does not appear, and if I erase it, i appears again... really weird...

Thanks a lot!!

Tanisete · Post by **Tanisete** » Thu Jul 28, 2005 9:09 pm

I've compiled morph-sources without a problem without rtp patch... so it should be something related to it... anyone experienced similar problems??

Thanks a lot!!

yardbird · Post by **yardbird** » Sat Jul 30, 2005 12:35 pm

Tanisete wrote:I'm using an amd64 without suspend or supermount, but i can't deactivate this option:
Code: Select all
CONFIG_X86_MCE=y
In fact, every time i make menuconfig, that option does not appear, and if I erase it, i appears again... really weird...

Then it is either automatically selected by some other option or it is mandatory on AMD64.

I've compiled morph-sources without a problem without rtp patch... so it should be something related to it... anyone experienced similar problems??

Seems like you've found the culprit

The next release will have an updated rtp patch, so there's a chance the problem will be solved. In the meantime I'm afraid you will have to drop the rtp patch...