[gentoo-amd64] multicore strange issue

[gentoo-amd64] multicore strange issue

[Beso]

[-- Attachment #1: Type: text/plain, Size: 1165 bytes --]

hi,

i've got a strange problem with my new notebook pc. i've bought a pc with an
amd athlon 64 x2 processor with 2 cores, an atheros board and an rs690 ati.
my old pc was a turion 64 with an atheros board and an rs480. now the
components inside are all about the same so i've just copied the old gentoo
system into the new pc and tried to recompile the kernel to have it tuned
for the new processor. so the thing i've done was to activate the symetric
multiprocessing and set to compiled into the kernel of md4, md5, md6 and aes
x86_64 chipers for a luks /home chipering.
the problem is that with the old normal config the kernel detects one
processor and one core and it works, loading the modules, but when enabling
simetric multiprocessing the kernel detects just one processor and one core
and then it fails to load modules because of a different executable format
in the modules.

the strange thing is that opensuse livecd and mandriva live cd recognize the
double processor (/proc/cpuinfo has 2 lines) and is able to step it in the
right way. does anyone has a hint on what to control in the kernel config to
see what's wrong?!

thanks.

-- 
dott. ing. beso

[-- Attachment #2: Type: text/html, Size: 1235 bytes --]

Re: [gentoo-amd64] multicore strange issue

[Drake Donahue]

On Tue, 2008-06-03 at 09:01 +0000, Beso wrote:
> hi,
> 
> i've got a strange problem with my new notebook pc. i've bought a pc
> with an amd athlon 64 x2 processor with 2 cores, an atheros board and
> an rs690 ati. 
> my old pc was a turion 64 with an atheros board and an rs480. now the
> components inside are all about the same so i've just copied the old
> gentoo system into the new pc and tried to recompile the kernel to
> have it tuned for the new processor. so the thing i've done was to
> activate the symetric multiprocessing and set to compiled into the
> kernel of md4, md5, md6 and aes x86_64 chipers for a luks /home
> chipering. 
> the problem is that with the old normal config the kernel detects one
> processor and one core and it works, loading the modules, but when
> enabling simetric multiprocessing the kernel detects just one
> processor and one core and then it fails to load modules because of a
> different executable format in the modules.
> 
> the strange thing is that opensuse livecd and mandriva live cd
> recognize the double processor (/proc/cpuinfo has 2 lines) and is able
> to step it in the right way. does anyone has a hint on what to control
> in the kernel config to see what's wrong?!
> 
> thanks.
> 
> -- 
> dott. ing. beso
If the attempt to boot does not end in a kernel panic; 
use <shift-page up> to get to the top of the console output produced
during boot. If the boot loader has started an SMP kernel, a line
similar to this: "Linux Version 2.6.24-gentoo-r8 (root@livecd) (gcc
version 4.1.1 (gentoo 4.1.1-r3)) #1 SMP Sat May 24 14:06:10 EDT 2008"
should appear.  The key word being SMP. No key word, "SMP", means that
"make menuconfig" (or whatever kernel configuration tool you use needs
to be redone.
As far as I know under, Processor type and features,:
    [*] Symmetric multi-processing support
    Subarchitecture Type (PC-compatible)  ---> 
    Processor family (Opteron/Athlon64/Hammer/K8)  --->
produces an appropriate SMP kernel.
A <shift-page down> should reveal a line like:
CPU has two num_cores
A couple more <shift-page down>:
Initializing CPU#0
About 3 more <shift-page down>:
Brought up 2 CPU's.
About modules, I'm guessing modules compiled for single cpu have
differences from those compiled for multiple cpu. Also guessing
that /etc/modules/autoload.d and/or udev is trying to load single cpu
module(s) that were not written to adjust to a single cpu kernel. 


-- 
gentoo-amd64@lists.gentoo.org mailing list

Re: [gentoo-amd64] multicore strange issue

[Beso]

[-- Attachment #1: Type: text/plain, Size: 2660 bytes --]

2008/6/3 Drake Donahue <donahue95@comcast.net>:

>
> On Tue, 2008-06-03 at 09:01 +0000, Beso wrote:
> > hi,
> >
> > i've got a strange problem with my new notebook pc. i've bought a pc
> > with an amd athlon 64 x2 processor with 2 cores, an atheros board and
> > an rs690 ati.
> > my old pc was a turion 64 with an atheros board and an rs480. now the
> > components inside are all about the same so i've just copied the old
> > gentoo system into the new pc and tried to recompile the kernel to
> > have it tuned for the new processor. so the thing i've done was to
> > activate the symetric multiprocessing and set to compiled into the
> > kernel of md4, md5, md6 and aes x86_64 chipers for a luks /home
> > chipering.
> > the problem is that with the old normal config the kernel detects one
> > processor and one core and it works, loading the modules, but when
> > enabling simetric multiprocessing the kernel detects just one
> > processor and one core and then it fails to load modules because of a
> > different executable format in the modules.
> >
> > the strange thing is that opensuse livecd and mandriva live cd
> > recognize the double processor (/proc/cpuinfo has 2 lines) and is able
> > to step it in the right way. does anyone has a hint on what to control
> > in the kernel config to see what's wrong?!
> >
> > thanks.
> >
> > --
> > dott. ing. beso
> If the attempt to boot does not end in a kernel panic;
> use <shift-page up> to get to the top of the console output produced
> during boot. If the boot loader has started an SMP kernel, a line
> similar to this: "Linux Version 2.6.24-gentoo-r8 (root@livecd) (gcc
> version 4.1.1 (gentoo 4.1.1-r3)) #1 SMP Sat May 24 14:06:10 EDT 2008"
> should appear.  The key word being SMP. No key word, "SMP", means that
> "make menuconfig" (or whatever kernel configuration tool you use needs
> to be redone.
> As far as I know under, Processor type and features,:
>    [*] Symmetric multi-processing support
>    Subarchitecture Type (PC-compatible)  --->
>    Processor family (Opteron/Athlon64/Hammer/K8)  --->
> produces an appropriate SMP kernel.
> A <shift-page down> should reveal a line like:
> CPU has two num_cores
> A couple more <shift-page down>:
> Initializing CPU#0
> About 3 more <shift-page down>:
> Brought up 2 CPU's.
> About modules, I'm guessing modules compiled for single cpu have
> differences from those compiled for multiple cpu. Also guessing
> that /etc/modules/autoload.d and/or udev is trying to load single cpu
> module(s) that were not written to adjust to a single cpu kernel.
>

so grub needs a different line at startup when using multicore?!



-- 
dott. ing. beso

[-- Attachment #2: Type: text/html, Size: 3323 bytes --]

Re: [gentoo-amd64] multicore strange issue

[Konrad Okurowski]

[-- Attachment #1: Type: text/plain, Size: 3048 bytes --]

I'm using turion x2 laptop with smp activated with no problem. Previously I
was using athlon64 3200+ desktop but gentoo on laptop was installed from
scratch.

Kfiaciarka

2008/6/3 Beso <givemesugarr@gmail.com>:

> 2008/6/3 Drake Donahue <donahue95@comcast.net>:
>
>>
>> On Tue, 2008-06-03 at 09:01 +0000, Beso wrote:
>> > hi,
>> >
>> > i've got a strange problem with my new notebook pc. i've bought a pc
>> > with an amd athlon 64 x2 processor with 2 cores, an atheros board and
>> > an rs690 ati.
>> > my old pc was a turion 64 with an atheros board and an rs480. now the
>> > components inside are all about the same so i've just copied the old
>> > gentoo system into the new pc and tried to recompile the kernel to
>> > have it tuned for the new processor. so the thing i've done was to
>> > activate the symetric multiprocessing and set to compiled into the
>> > kernel of md4, md5, md6 and aes x86_64 chipers for a luks /home
>> > chipering.
>> > the problem is that with the old normal config the kernel detects one
>> > processor and one core and it works, loading the modules, but when
>> > enabling simetric multiprocessing the kernel detects just one
>> > processor and one core and then it fails to load modules because of a
>> > different executable format in the modules.
>> >
>> > the strange thing is that opensuse livecd and mandriva live cd
>> > recognize the double processor (/proc/cpuinfo has 2 lines) and is able
>> > to step it in the right way. does anyone has a hint on what to control
>> > in the kernel config to see what's wrong?!
>> >
>> > thanks.
>> >
>> > --
>> > dott. ing. beso
>> If the attempt to boot does not end in a kernel panic;
>> use <shift-page up> to get to the top of the console output produced
>> during boot. If the boot loader has started an SMP kernel, a line
>> similar to this: "Linux Version 2.6.24-gentoo-r8 (root@livecd) (gcc
>> version 4.1.1 (gentoo 4.1.1-r3)) #1 SMP Sat May 24 14:06:10 EDT 2008"
>> should appear.  The key word being SMP. No key word, "SMP", means that
>> "make menuconfig" (or whatever kernel configuration tool you use needs
>> to be redone.
>> As far as I know under, Processor type and features,:
>>    [*] Symmetric multi-processing support
>>    Subarchitecture Type (PC-compatible)  --->
>>    Processor family (Opteron/Athlon64/Hammer/K8)  --->
>> produces an appropriate SMP kernel.
>> A <shift-page down> should reveal a line like:
>> CPU has two num_cores
>> A couple more <shift-page down>:
>> Initializing CPU#0
>> About 3 more <shift-page down>:
>> Brought up 2 CPU's.
>> About modules, I'm guessing modules compiled for single cpu have
>> differences from those compiled for multiple cpu. Also guessing
>> that /etc/modules/autoload.d and/or udev is trying to load single cpu
>> module(s) that were not written to adjust to a single cpu kernel.
>>
>
> so grub needs a different line at startup when using multicore?!
>
>
>
> --
> dott. ing. beso




-- 
pozdrawiam
Konrad Okurowski <dm.konrad@gmail.com>
Key Fingerprint = D993 3823 F20D 3B8D 2719 3012 2FE8

[-- Attachment #2: Type: text/html, Size: 4015 bytes --]

Re: [gentoo-amd64] multicore strange issue

[Drake Donahue]

On Tue, 2008-06-03 at 15:15 +0000, Beso wrote:
> 2008/6/3 Drake Donahue <donahue95@comcast.net>:
>         
>         
>         On Tue, 2008-06-03 at 09:01 +0000, Beso wrote:
>         > hi,
>         >
>         > i've got a strange problem with my new notebook pc. i've
>         bought a pc
>         > with an amd athlon 64 x2 processor with 2 cores, an atheros
>         board and
>         > an rs690 ati.
>         > my old pc was a turion 64 with an atheros board and an
>         rs480. now the
>         > components inside are all about the same so i've just copied
>         the old
>         > gentoo system into the new pc and tried to recompile the
>         kernel to
>         > have it tuned for the new processor. so the thing i've done
>         was to
>         > activate the symetric multiprocessing and set to compiled
>         into the
>         > kernel of md4, md5, md6 and aes x86_64 chipers for a
>         luks /home
>         > chipering.
>         > the problem is that with the old normal config the kernel
>         detects one
>         > processor and one core and it works, loading the modules,
>         but when
>         > enabling simetric multiprocessing the kernel detects just
>         one
>         > processor and one core and then it fails to load modules
>         because of a
>         > different executable format in the modules.
>         >
>         > the strange thing is that opensuse livecd and mandriva live
>         cd
>         > recognize the double processor (/proc/cpuinfo has 2 lines)
>         and is able
>         > to step it in the right way. does anyone has a hint on what
>         to control
>         > in the kernel config to see what's wrong?!
>         >
>         > thanks.
>         >
>         > --
>         > dott. ing. beso
>         
>         If the attempt to boot does not end in a kernel panic;
>         use <shift-page up> to get to the top of the console output
>         produced
>         during boot. If the boot loader has started an SMP kernel, a
>         line
>         similar to this: "Linux Version 2.6.24-gentoo-r8 (root@livecd)
>         (gcc
>         version 4.1.1 (gentoo 4.1.1-r3)) #1 SMP Sat May 24 14:06:10
>         EDT 2008"
>         should appear.  The key word being SMP. No key word, "SMP",
>         means that
>         "make menuconfig" (or whatever kernel configuration tool you
>         use needs
>         to be redone.
>         As far as I know under, Processor type and features,:
>            [*] Symmetric multi-processing support
>            Subarchitecture Type (PC-compatible)  --->
>            Processor family (Opteron/Athlon64/Hammer/K8)  --->
>         produces an appropriate SMP kernel.
>         A <shift-page down> should reveal a line like:
>         CPU has two num_cores
>         A couple more <shift-page down>:
>         Initializing CPU#0
>         About 3 more <shift-page down>:
>         Brought up 2 CPU's.
>         About modules, I'm guessing modules compiled for single cpu
>         have
>         differences from those compiled for multiple cpu. Also
>         guessing
>         that /etc/modules/autoload.d and/or udev is trying to load
>         single cpu
>         module(s) that were not written to adjust to a single cpu
>         kernel.
> 
> so grub needs a different line at startup when using multicor

I had assumed you had established a multiboot grub with one entry
pointed to the old kernel and another entry pointed to a newly compiled
SMP kernel.

>  
> 
> 
> 
> 
> -- 
> dott. ing. beso

-- 
gentoo-amd64@lists.gentoo.org mailing list

[gentoo-amd64] Re: multicore strange issue

[Duncan]

Beso <givemesugarr@gmail.com> posted
d257c3560806030201y23df89fdy1cf12ae0bb3b772a@mail.gmail.com, excerpted
below, on  Tue, 03 Jun 2008 09:01:04 +0000:

> the problem is that with the old normal config the kernel detects one
> processor and one core and it works, loading the modules, but when
> enabling simetric multiprocessing the kernel detects just one processor
> and one core and then it fails to load modules because of a different
> executable format in the modules.

This indicates that the modules and the kernel aren't matched.  IOW, you 
apparently compiled the new kernel but failed to compile the matching 
modules, so it's trying to use the old ones, which don't match the new 
kernel, so it won't load them.

Even if it's the exact same kernel you compiled, if you change options, 
particularly something as big as SMP, you'll need to recompile the kernel 
modules as well.

Also note that if you use an initrd/initramfs, you'll need to redo it, 
loading the new modules as appropriate into it as well as placing them in 
the normal /lib/modules/<kern-ver>/ dir.  I don't run an initramfs/
initrd, so have never learned the technical intricacies thereof, but 
wherever it's trying to load them from, you don't have the right ones 
there for it to load.

One way or the other, with or without the initrd/initramfs, the kernel is 
trying to load modules that don't match the kernel, and failing.  Correct 
that, and you will have fixed at least that problem, altho it's possible 
you'll then have others to fix as well.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-amd64@lists.gentoo.org mailing list

Re: [gentoo-amd64] Re: multicore strange issue

[Konrad Okurowski]

[-- Attachment #1: Type: text/plain, Size: 2326 bytes --]

Hi,
IMO when performing "make" after make clean and make menuconfig (assuming
you want to change something in kernel) builds kernel image and modules as
well. After that I run rm -rf /lib/modules/<kernel dir> and make && make
modules_install. Then there shouldn't be any problem. (also [emerge -av
$(qdepends -CNQ virtual/linux-sources)] would be nice).

Kfiaciarka

2008/6/3 Duncan <1i5t5.duncan@cox.net>:

> Beso <givemesugarr@gmail.com> posted
> d257c3560806030201y23df89fdy1cf12ae0bb3b772a@mail.gmail.com, excerpted
> below, on  Tue, 03 Jun 2008 09:01:04 +0000:
>
> > the problem is that with the old normal config the kernel detects one
> > processor and one core and it works, loading the modules, but when
> > enabling simetric multiprocessing the kernel detects just one processor
> > and one core and then it fails to load modules because of a different
> > executable format in the modules.
>
> This indicates that the modules and the kernel aren't matched.  IOW, you
> apparently compiled the new kernel but failed to compile the matching
> modules, so it's trying to use the old ones, which don't match the new
> kernel, so it won't load them.
>
> Even if it's the exact same kernel you compiled, if you change options,
> particularly something as big as SMP, you'll need to recompile the kernel
> modules as well.
>
> Also note that if you use an initrd/initramfs, you'll need to redo it,
> loading the new modules as appropriate into it as well as placing them in
> the normal /lib/modules/<kern-ver>/ dir.  I don't run an initramfs/
> initrd, so have never learned the technical intricacies thereof, but
> wherever it's trying to load them from, you don't have the right ones
> there for it to load.
>
> One way or the other, with or without the initrd/initramfs, the kernel is
> trying to load modules that don't match the kernel, and failing.  Correct
> that, and you will have fixed at least that problem, altho it's possible
> you'll then have others to fix as well.
>
> --
> Duncan - List replies preferred.   No HTML msgs.
> "Every nonfree program has a lord, a master --
> and if you use the program, he is your master."  Richard Stallman
>
> --
> gentoo-amd64@lists.gentoo.org mailing list
>
>


-- 
pozdrawiam
Konrad Okurowski <dm.konrad@gmail.com>
Key Fingerprint = D993 3823 F20D 3B8D 2719 3012 2FE8

[-- Attachment #2: Type: text/html, Size: 3155 bytes --]

Re: [gentoo-amd64] Re: multicore strange issue

[Beso]

[-- Attachment #1: Type: text/plain, Size: 3296 bytes --]

2008/6/3 Konrad Okurowski <dm.konrad@gmail.com>:

> Hi,
> IMO when performing "make" after make clean and make menuconfig (assuming
> you want to change something in kernel) builds kernel image and modules as
> well. After that I run rm -rf /lib/modules/<kernel dir> and make && make
> modules_install. Then there shouldn't be any problem. (also [emerge -av
> $(qdepends -CNQ virtual/linux-sources)] would be nice).
>
> Kfiaciarka
>
> 2008/6/3 Duncan <1i5t5.duncan@cox.net>:
>
> Beso <givemesugarr@gmail.com> posted
>> d257c3560806030201y23df89fdy1cf12ae0bb3b772a@mail.gmail.com, excerpted
>> below, on  Tue, 03 Jun 2008 09:01:04 +0000:
>>
>> > the problem is that with the old normal config the kernel detects one
>> > processor and one core and it works, loading the modules, but when
>> > enabling simetric multiprocessing the kernel detects just one processor
>> > and one core and then it fails to load modules because of a different
>> > executable format in the modules.
>>
>> This indicates that the modules and the kernel aren't matched.  IOW, you
>> apparently compiled the new kernel but failed to compile the matching
>> modules, so it's trying to use the old ones, which don't match the new
>> kernel, so it won't load them.
>>
>> Even if it's the exact same kernel you compiled, if you change options,
>> particularly something as big as SMP, you'll need to recompile the kernel
>> modules as well.
>>
>> Also note that if you use an initrd/initramfs, you'll need to redo it,
>> loading the new modules as appropriate into it as well as placing them in
>> the normal /lib/modules/<kern-ver>/ dir.  I don't run an initramfs/
>> initrd, so have never learned the technical intricacies thereof, but
>> wherever it's trying to load them from, you don't have the right ones
>> there for it to load.
>>
>> One way or the other, with or without the initrd/initramfs, the kernel is
>> trying to load modules that don't match the kernel, and failing.  Correct
>> that, and you will have fixed at least that problem, altho it's possible
>> you'll then have others to fix as well.
>>
>> --
>> Duncan - List replies preferred.   No HTML msgs.
>> "Every nonfree program has a lord, a master --
>> and if you use the program, he is your master."  Richard Stallman
>>
>> --
>> gentoo-amd64@lists.gentoo.org mailing list
>>
>>
>
>
> --
> pozdrawiam
> Konrad Okurowski <dm.konrad@gmail.com>
> Key Fingerprint = D993 3823 F20D 3B8D 2719 3012 2FE8
>

i've found out the problem: i'm using a /boot not on the /partition and on
the new pc i've set the partition table in a different way, so boot didn't
mount at startup. the make modules_install && make install would put the
right stuff in the right place, but when restarting the kernel image read
was the old one (the one without smp support) and so the different
executable file format errror, using the new smp modules and the old
single-core kernel image. i'm really feeling really stupid for not having
noticed this earlier, but maybe since the issue got out at about 23.00 after
a full day of work maybe it's plausible.
thanks to all of you who have answered to my question.

ps. duncan, when you've got some spare time could you please post the
scripts for the kernel automation build if it's not bothersome. thanks
again.

-- 
dott. ing. beso

[-- Attachment #2: Type: text/html, Size: 4508 bytes --]

[gentoo-amd64] Re: multicore strange issue

[Duncan]

Beso <givemesugarr@gmail.com> posted
d257c3560806040034n1b0e28bes90044be3b316ee02@mail.gmail.com, excerpted
below, on  Wed, 04 Jun 2008 07:34:09 +0000:

> the make modules_install && make install would put the right stuff in
> the right place, but when restarting the kernel image read was the old
> one (the one without smp support) and so the different executable file
> format errror, using the new smp modules and the old single-core kernel
> image. i'm really feeling really stupid for not having noticed this
> earlier, but maybe since the issue got out at about 23.00 after a full
> day of work maybe it's plausible. thanks to all of you who have answered
> to my question.

Don't worry about it.  I'm sure most of us eventually hit that or a 
similar modules/kernel mismatch error.  I know I've seen it when I boot 
my rootbak partition (on which I keep a backup system image, from a time 
it was known to be generally working) and the kernel in /boot is newer 
than the modules on rootbak, one of the reasons I recognized it.

> ps. duncan, when you've got some spare time could you please post the
> scripts for the kernel automation build if it's not bothersome. thanks
> again.

I have it on my list for this weekend if not earlier.  If I haven't 
posted anything by Sunday, poke me. =8^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-amd64@lists.gentoo.org mailing list