Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-09 Thread Guillaume Cottenceau 

Reinhard Katzmann <[EMAIL PROTECTED]> writes:


[...]

> > Unfortunately there is no trustable way to detect if a filesystem
> > is compiled in a kernel so we don't "support" kernels for which
> > fs'es (other than ext2) are compiled in.
> 
> Ok, but what about an option to read /boot/config- as defined
> with "mkinitrd   ? Of course this has the
> disadvantage that mkinitrd needs to know about all filesystem types
> which can be used to boot (I think at least ext2,3,reiserfs,xfs,jfs)
> and it's option in the config file (which might change and break
> this option at times).

Well it needs a mapping between fs module name and kernel option
name such as "ext3" <=> "CONFIG_EXT3_FS".

I'm reluctant in sending too much worktime into something we
don't even provide (e.g. a kernel with compiled in FS'es).


> > > When I invoke mkinitrd and look at the result, in /lib you find the following
> > > file (initrd image):
> > > lib:
> > > insgesamt 83
> > > drwxr-xr-x2 root root 1024 Jan  7 10:16 ./
> > > drwxr-xr-x   10 root root 1024 Jan  7 10:16 ../
> > > -rw-r--r--1 root root81514 Jan  5 19:18 ext3.o
> > > 
> > > locate ext3.o
> > > /usr/src/linux-2.4.17-1mdk/fs/ext3/.ext3.o.flags
> > > /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> > > 
> > > ll /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> > > -rw-r--r--1 root root81514 Jan  5 19:18 
>/usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> > 
> > Lines 108 and 116 of mkinitrd should prevent this from happen.
> > Can you trace the problem more precisely ?
> 
> Well the problem is simple: There is a link inside every kernel directory
> which points to the directory where the build of this kernel was done.

Uh.. I didn't know that.. sorry.. Yet actually for "stock"
RPM-provided kernels this link is broken so it's not very
harmful.

Yet I don't know how to elegantly solve this problem then.


> I don't know exactly what sh_find does in line 108 but it does not

It's meant to do the same job as "find". (it's here because
mkinitrd is targetted to work with /usr not mounted)

> ignore the links but dereference them. I don't know about find but I
> find it annoying that there is no possibility to tell 'ls -lR' to not
> dereference links (what should be the default option but it still
> dereferences it). Ok I'm getting Offtopic here ;-)
> 
> With best regards,
> 
> Reinhard Katzmann

-- 
Guillaume Cottenceau - http://people.mandrakesoft.com/~gc/

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-09 Thread Reinhard Katzmann 

Hi Guillaume!

On Wed, Jan 09, 2002 at 05:55:46PM +0100, Guillaume Cottenceau wrote:
> Reinhard Katzmann <[EMAIL PROTECTED]> writes:
> 
>  
> > > > - mkinitrd does not check the the kernel config (/boot/config-2.4.17-xyz)
> > > >   correctly, modules are added from the kernel source tree (!)
> > > >   which are inside the kernel (for me this was ext3, ide-mod, ide-probe-mod,
> > > >   ide-disk). If I don't add them to IGNOREMODS insmoding will (of course)
> > > >   fail when booting the kernel.
> > > 
> > > I don't follow you. Modules are taken from /lib/modules/$kernver
> > > so what's the problem?
> > 
> > No they are not. For example I have ext3 in my kernel like this:
> > CONFIG_EXT3_FS=y
> 
> Unfortunately there is no trustable way to detect if a filesystem
> is compiled in a kernel so we don't "support" kernels for which
> fs'es (other than ext2) are compiled in.
> 

Ok, but what about an option to read /boot/config- as defined
with "mkinitrd   ? Of course this has the
disadvantage that mkinitrd needs to know about all filesystem types
which can be used to boot (I think at least ext2,3,reiserfs,xfs,jfs)
and it's option in the config file (which might change and break
this option at times).

> > When I invoke mkinitrd and look at the result, in /lib you find the following
> > file (initrd image):
> > lib:
> > insgesamt 83
> > drwxr-xr-x2 root root 1024 Jan  7 10:16 ./
> > drwxr-xr-x   10 root root 1024 Jan  7 10:16 ../
> > -rw-r--r--1 root root81514 Jan  5 19:18 ext3.o
> > 
> > locate ext3.o
> > /usr/src/linux-2.4.17-1mdk/fs/ext3/.ext3.o.flags
> > /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> > 
> > ll /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> > -rw-r--r--1 root root81514 Jan  5 19:18 
>/usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> 
> Lines 108 and 116 of mkinitrd should prevent this from happen.
> Can you trace the problem more precisely ?

Well the problem is simple: There is a link inside every kernel directory
which points to the directory where the build of this kernel was done.
I don't know exactly what sh_find does in line 108 but it does not
ignore the links but dereference them. I don't know about find but I
find it annoying that there is no possibility to tell 'ls -lR' to not
dereference links (what should be the default option but it still
dereferences it). Ok I'm getting Offtopic here ;-)

With best regards,

Reinhard Katzmann
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50570/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-09 Thread Guillaume Cottenceau 

Reinhard Katzmann <[EMAIL PROTECTED]> writes:


[...]

> It works, but I don't think it's correct:
> 
> mkdir -p $MNTIMAGE/sbin
> mkdir -p $MNTIMAGE/etc
> mkdir -p $MNTIMAGE/dev
> mkdir -p $MNTIMAGE/loopfs
> mkdir -p $MNTIMAGE/proc
> mkdir -p $MNTIMAGE/sysroot
> ln -s bin $MNTIMAGE/sbin
> 
> The result is: (ls -l bin sbin)
> bin:
> insgesamt 36
> -rwxr-xr-x1 root root24220 Jan  7 09:55 insmod*
> -rwxr-xr-x1 root root11128 Jan  7 09:55 nash*
> 
> sbin:
> insgesamt 0
> lrwxrwxrwx1 root root3 Jan  7 10:11 bin -> bin
> lrwxrwxrwx1 root root9 Jan  7 09:55 modprobe -> /bin/nash
> 
> Either the sbin mkdir should be removed or the link command.

Right.

 
> > > - mkinitrd does not check the the kernel config (/boot/config-2.4.17-xyz)
> > >   correctly, modules are added from the kernel source tree (!)
> > >   which are inside the kernel (for me this was ext3, ide-mod, ide-probe-mod,
> > >   ide-disk). If I don't add them to IGNOREMODS insmoding will (of course)
> > >   fail when booting the kernel.
> > 
> > I don't follow you. Modules are taken from /lib/modules/$kernver
> > so what's the problem?
> 
> No they are not. For example I have ext3 in my kernel like this:
> CONFIG_EXT3_FS=y

Unfortunately there is no trustable way to detect if a filesystem
is compiled in a kernel so we don't "support" kernels for which
fs'es (other than ext2) are compiled in.


> When I invoke mkinitrd and look at the result, in /lib you find the following
> file (initrd image):
> lib:
> insgesamt 83
> drwxr-xr-x2 root root 1024 Jan  7 10:16 ./
> drwxr-xr-x   10 root root 1024 Jan  7 10:16 ../
> -rw-r--r--1 root root81514 Jan  5 19:18 ext3.o
> 
> locate ext3.o
> /usr/src/linux-2.4.17-1mdk/fs/ext3/.ext3.o.flags
> /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> 
> ll /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
> -rw-r--r--1 root root81514 Jan  5 19:18 
>/usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o

Lines 108 and 116 of mkinitrd should prevent this from happen.
Can you trace the problem more precisely ?


-- 
Guillaume Cottenceau - http://people.mandrakesoft.com/~gc/

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-09 Thread Guillaume Cottenceau 

Reinhard Katzmann <[EMAIL PROTECTED]> writes:


[...]

> Hmm, there is a case in mkinitrd, where initrd is not built:
> 
> if [ "$rootfs" != "ext3" -a -z "$rootfsopts" -a -n "$ifneeded" -a -z "$MODULES" 
> ]; then
> if [ -n "$verbose" ]; then
> echo "Rootfs is not ext3, there is no rootfs special options, and"
> echo "no modules are needed -- not building initrd image."
> fi
> exit 0
> fi
> 
> If I did not use ext3 but ext2 for me this script would not build any
> initrd image.

Yes.
 

> One more thing about mkinitrd:
> I found that rootdev variable is defined twice:
> 
> 1) rootdev=$(awk '{ if ($2 == "/") { print $1; }}' $fstab)
> 
> 2) (more below)
> 
> if [ -n "$loopDev" ]; then
> 
> ...
> 
> rootdev=/safedev/loop7
> else
> echo "echo Creating root device" >> $RCFILE
> echo "mkrootdev /dev/root" >> $RCFILE
> rootdev=/dev/root
> fi
> 
> The first rootdev definition is nowhere used, and if I use it,
> booting will fail (because mkrootdev either is missing or
> creates the wrong device). No problem here, but I think the
> 1) line could be removed from the script, couldn't it ?

Seems right (but not thrilling ;p).



-- 
Guillaume Cottenceau - http://people.mandrakesoft.com/~gc/

RE: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-08 Thread Borsenkow Andrej 

> 
> I'd like to know if the devfs_noroot patch is available in CVS
> or anywhere or if you can send it to me, so I don't need to
> download ~30 GB kernel-source over modem line :-)
> 

And have you tried simply get a look in CVS?

http://www.linux-mandrake.com/cgi-bin/cvsweb.cgi/SPECS/kernel/linux-2.4.
17-devfs_no_root.patch


-andrej

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-08 Thread Reinhard Katzmann 

Hi Juan!

I'd like to know if the devfs_noroot patch is available in CVS
or anywhere or if you can send it to me, so I don't need to
download ~30 GB kernel-source over modem line :-)

> Get mandrake 2.4.17.1mdk kernel, and search for a patch called
> devfs_noroot, problem is that devfs changed behaviour in one stable
> release :(  and you are not able anymore to do an:
> unlink("/dev/root");
> 
> that change happened in 2.4.17-preX.  The workaround in Mandrake
> kernel is that we don't create that file anymore by devfs.
> 
> Now you now what it takes so long to release an official 2.4.17 kernel
> in mandrake, not all the reasons was the holidays :)
> 
> Notice that the patch(workaround is trivial), but finding that two
> damned lines was the difficult part :(


> reinhard> I also tested without initrd image. This fails as well:
> 
> as and added bonus, now it is possible (again) to boot with
> devfs=nomount and it works :)

Good news :) When I tested I of course tried both with and without
devfs. Good to here that mkinitrd is not the cause of the trouble :):

Regards,

Reinhard Katzmann
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50471/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-08 Thread Juan Quintela 

> "reinhard" == Reinhard Katzmann <[EMAIL PROTECTED]> writes:

reinhard> Hi Juan!

Hi

could you check if Mandrake kernel 2.4.17 (either linus or Mandrake)
boots for you?  it took a while to found that damned bug here, but now
I am able to boot all my machines with/whithout devfs (with plain
2.4.17 they just hang during boot as yours).

Later, Juan.

>> >> VFS: Mounted root (ext2 filesystem).
>> >> Mounted devfs on /dev.
>> >> Red Hat nash version 3.1.6-mdk starting.
>> >> Mounting /proc filesystem.
>> >> Creating root device.
>> >> Mounting root filesystem with flags noatime
>> >> kjournald starting.  Commit interval 5 seconds
>> >> EXT3-fs: mounted filesystem with ordered data mode.
>> >> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
>> >> Freeing unused kernel memory:708k freed.
>> >> Kernel panic: No init found. Try passing init= option to kernel.
>> 
guillaume> Booh :-(.
>> 
>> No idea what is happening here, It take a lot of time to get 2.4.17
>> out due to the bug that it didn't boot :(  Could you told us what is
>> your hardware?

reinhard> Yes, I also include a part of my .config (the CPU part). I have a
reinhard> Sony Vaio PCG-FX-301 Laptop / a AMD K62-based desktop. I concentrate
reinhard> on the laptop though it does not work with both computers. The
reinhard> Laptop has an 800 Mhz Mobile Duron Prozessor (powertweak segfaults with
reinhard> this processor, but that's another story...) with 256 MB RAM and
reinhard> 10 Gig IDE 2.5" hard disk and 8xDVD-Drive. More info from 2.4.16
reinhard> Kernel output (dmesg, incomplete):

reinhard> Local APIC disabled by BIOS -- reenabling.
reinhard> Found and enabled local APIC!
reinhard> Kernel command line: root=/dev/hda1  vga=791
reinhard> Initializing CPU#0
reinhard> Detected 800.048 MHz processor.
reinhard> Memory: 255284k/262080k available (971k kernel code, 6408k reserved, 352k 
data, 
reinhard> 704k init, 0k highmem)
reinhard> PCI: PCI BIOS revision 2.10
reinhard> Trying to stomp on Athlon bug...
reinhard> Unknown bridge resource 2: assuming transparent
reinhard> VP_IDE: IDE controller on PCI bus 00 dev 39
reinhard> VP_IDE: chipset revision 16
reinhard> hda: FUJITSU MHM2100AT, ATA DISK drive
reinhard> hdc: TOSHIBA DVD-ROM SD-C2502, ATAPI CD/DVD-ROM drive
reinhard> eth0: RealTek RTL8139 Fast Ethernet at 0xd1125800, 08:00:46:1e:51:6a, IRQ 10

reinhard> It has a VIA VT82C686 chipset and also has USB, I-Link (IEEE1394, Unknown
reinhard> Device 8020), Sound  AC97/Audio onboard, ATI Rage Mobility P/M AGP graphic
reinhard> onboard. Host bridge: Unknown device 0305, PCI Bridge: Unknown device b115.
reinhard> Part of my kernel config:

reinhard> CONFIG_X86=y
reinhard> CONFIG_ISA=y
reinhard> # CONFIG_SBUS is not set
reinhard> CONFIG_UID16=y
reinhard> #
reinhard> # Processor type and features
reinhard> #
reinhard> CONFIG_MK7=y
reinhard> CONFIG_X86_WP_WORKS_OK=y
reinhard> CONFIG_X86_INVLPG=y
reinhard> CONFIG_X86_CMPXCHG=y
reinhard> CONFIG_X86_XADD=y
reinhard> CONFIG_X86_BSWAP=y
reinhard> CONFIG_X86_POPAD_OK=y
reinhard> CONFIG_RWSEM_XCHGADD_ALGORITHM=y
reinhard> CONFIG_X86_L1_CACHE_SHIFT=6
reinhard> CONFIG_X86_TSC=y
reinhard> CONFIG_X86_GOOD_APIC=y
reinhard> CONFIG_X86_USE_3DNOW=y
reinhard> CONFIG_X86_PGE=y
reinhard> CONFIG_X86_USE_PPRO_CHECKSUM=y
reinhard> CONFIG_MICROCODE=m
reinhard> CONFIG_X86_MSR=m
reinhard> CONFIG_X86_CPUID=m
reinhard> CONFIG_E820_PROC=y
reinhard> CONFIG_NOHIGHMEM=y
reinhard> CONFIG_MTRR=y
reinhard> CONFIG_X86_UP_APIC=y
reinhard> CONFIG_X86_UP_IOAPIC=y
reinhard> CONFIG_X86_LOCAL_APIC=y
reinhard> CONFIG_X86_IO_APIC=y
reinhard> #
reinhard> # General setup
reinhard> #
reinhard> CONFIG_NET=y
reinhard> CONFIG_PCI=y
reinhard> # CONFIG_PCI_GOBIOS is not set
reinhard> # CONFIG_PCI_GODIRECT is not set
reinhard> CONFIG_PCI_GOANY=y
reinhard> CONFIG_PCI_BIOS=y
reinhard> CONFIG_PCI_DIRECT=y
reinhard> CONFIG_PCI_NAMES=y
reinhard> CONFIG_EISA=y
reinhard> # CONFIG_MCA is not set
reinhard> CONFIG_HOTPLUG=y

reinhard> The only thing I can update is the pivot_root command from util-linux,
reinhard> is this necessary ?

reinhard> With best regards,

reinhard> Reinhard Katzmann
reinhard> -- 
reinhard> Software-Engineer, Developer for Embedded Devices
reinhard> Project: HyperPen Tablet USB Driver for Linux 
reinhard> GnuPG Public Key available on request

-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-08 Thread Juan Quintela 

> "reinhard" == Reinhard Katzmann <[EMAIL PROTECTED]> writes:

reinhard> Hi!
reinhard> I'm unable to boot my self-made 2.4.17 kernel. When I try it
reinhard> I get the following messages:

Get mandrake 2.4.17.1mdk kernel, and search for a patch called
devfs_noroot, problem is that devfs changed behaviour in one stable
release :(  and you are not able anymore to do an:
unlink("/dev/root");

that change happened in 2.4.17-preX.  The workaround in Mandrake
kernel is that we don't create that file anymore by devfs.

Now you now what it takes so long to release an official 2.4.17 kernel
in mandrake, not all the reasons was the holidays :)

Notice that the patch(workaround is trivial), but finding that two
damned lines was the difficult part :(

reinhard> VFS: Mounted root (ext2 filesystem).
reinhard> Mounted devfs on /dev.
reinhard> Red Hat nash version 3.1.6-mdk starting.
reinhard> Mounting /proc filesystem.
reinhard> Creating root device.
reinhard> Mounting root filesystem with flags noatime
reinhard> kjournald starting.  Commit interval 5 seconds
reinhard> EXT3-fs: mounted filesystem with ordered data mode.
reinhard> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
reinhard> Freeing unused kernel memory:708k freed.
reinhard> Kernel panic: No init found. Try passing init= option to kernel.

reinhard> Note: pivotroot returns error 22. What does this mean ?

Read /usr/include/asm/errno.h to know what errors mean, in this case
it is EINVAL.

reinhard> This is with mkinitrd-3.1.6-16mdk, I also -12mdk (my previous version)
reinhard> without success. This version worked fine with the 2.4.16 kernel. I
reinhard> also updated e2fsprogs to 1.25-2mdk and use the recent modutils 
reinhard> 2.4.12-1mdk. (the modutils were also used with 2.4.16 kernel, while
reinhard> e2fsprogs/mkinitrd was updated when the 2.4.17 kernel did not boot).
reinhard> I used make oldconfig with my old 2.4.16 kernel config and left all
reinhard> new 2.4.17 options out.

reinhard> I also tested without initrd image. This fails as well:

as and added bonus, now it is possible (again) to boot with
devfs=nomount and it works :)

Later, Juan.


-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-07 Thread Reinhard Katzmann 

Hi!

On Mon, Jan 07, 2002 at 12:18:54PM +0100, Juan Quintela wrote:
> > "guillaume" == Guillaume Cottenceau <[EMAIL PROTECTED]> writes:
> 
> >> VFS: Mounted root (ext2 filesystem).
> >> Mounted devfs on /dev.
> >> Freeing unused kernel memory:708k freed.
> >> Kernel panic: Attempted to kill init!
> 
> guillaume> Ooh this is also strange but looks like a kernel problem? Or does
> guillaume> boot without pivot_rooting no more supported, dear kernel team?
> 
> I think it is not supported anymore, but for the mkinitrd :(

Hmm, there is a case in mkinitrd, where initrd is not built:

if [ "$rootfs" != "ext3" -a -z "$rootfsopts" -a -n "$ifneeded" -a -z "$MODULES" 
]; then
if [ -n "$verbose" ]; then
echo "Rootfs is not ext3, there is no rootfs special options, and"
echo "no modules are needed -- not building initrd image."
fi
exit 0
fi

If I did not use ext3 but ext2 for me this script would not build any
initrd image.

> 
> having to take a look at that later.

One more thing about mkinitrd:
I found that rootdev variable is defined twice:

1) rootdev=$(awk '{ if ($2 == "/") { print $1; }}' $fstab)

2) (more below)

if [ -n "$loopDev" ]; then

...

rootdev=/safedev/loop7
else
echo "echo Creating root device" >> $RCFILE
echo "mkrootdev /dev/root" >> $RCFILE
rootdev=/dev/root
fi

The first rootdev definition is nowhere used, and if I use it,
booting will fail (because mkrootdev either is missing or
creates the wrong device). No problem here, but I think the
1) line could be removed from the script, couldn't it ?

If I use the normal way with /dev/root as root device mounting
seems to work fine (according to the output). I did not find a
way how to test if the mounting really worked, any idea how I
could do that ? nash has no ls or something similar and external
ls ("/sysroot/bin/ls" or "exec /sysroot/bin/ls") always fails.

With best regards,

Reinhard Katzmann 
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50407/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-07 Thread Reinhard Katzmann 

Hi Juan!

> >> VFS: Mounted root (ext2 filesystem).
> >> Mounted devfs on /dev.
> >> Red Hat nash version 3.1.6-mdk starting.
> >> Mounting /proc filesystem.
> >> Creating root device.
> >> Mounting root filesystem with flags noatime
> >> kjournald starting.  Commit interval 5 seconds
> >> EXT3-fs: mounted filesystem with ordered data mode.
> >> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
> >> Freeing unused kernel memory:708k freed.
> >> Kernel panic: No init found. Try passing init= option to kernel.
> 
> guillaume> Booh :-(.
> 
> No idea what is happening here, It take a lot of time to get 2.4.17
> out due to the bug that it didn't boot :(  Could you told us what is
> your hardware?

Yes, I also include a part of my .config (the CPU part). I have a
Sony Vaio PCG-FX-301 Laptop / a AMD K62-based desktop. I concentrate
on the laptop though it does not work with both computers. The
Laptop has an 800 Mhz Mobile Duron Prozessor (powertweak segfaults with
this processor, but that's another story...) with 256 MB RAM and
10 Gig IDE 2.5" hard disk and 8xDVD-Drive. More info from 2.4.16
Kernel output (dmesg, incomplete):

Local APIC disabled by BIOS -- reenabling.
Found and enabled local APIC!
Kernel command line: root=/dev/hda1  vga=791
Initializing CPU#0
Detected 800.048 MHz processor.
Memory: 255284k/262080k available (971k kernel code, 6408k reserved, 352k data, 
704k init, 0k highmem)
PCI: PCI BIOS revision 2.10
Trying to stomp on Athlon bug...
Unknown bridge resource 2: assuming transparent
VP_IDE: IDE controller on PCI bus 00 dev 39
VP_IDE: chipset revision 16
hda: FUJITSU MHM2100AT, ATA DISK drive
hdc: TOSHIBA DVD-ROM SD-C2502, ATAPI CD/DVD-ROM drive
eth0: RealTek RTL8139 Fast Ethernet at 0xd1125800, 08:00:46:1e:51:6a, IRQ 10

It has a VIA VT82C686 chipset and also has USB, I-Link (IEEE1394, Unknown
Device 8020), Sound  AC97/Audio onboard, ATI Rage Mobility P/M AGP graphic
onboard. Host bridge: Unknown device 0305, PCI Bridge: Unknown device b115.
Part of my kernel config:

CONFIG_X86=y
CONFIG_ISA=y
# CONFIG_SBUS is not set
CONFIG_UID16=y
#
# Processor type and features
#
CONFIG_MK7=y
CONFIG_X86_WP_WORKS_OK=y
CONFIG_X86_INVLPG=y
CONFIG_X86_CMPXCHG=y
CONFIG_X86_XADD=y
CONFIG_X86_BSWAP=y
CONFIG_X86_POPAD_OK=y
CONFIG_RWSEM_XCHGADD_ALGORITHM=y
CONFIG_X86_L1_CACHE_SHIFT=6
CONFIG_X86_TSC=y
CONFIG_X86_GOOD_APIC=y
CONFIG_X86_USE_3DNOW=y
CONFIG_X86_PGE=y
CONFIG_X86_USE_PPRO_CHECKSUM=y
CONFIG_MICROCODE=m
CONFIG_X86_MSR=m
CONFIG_X86_CPUID=m
CONFIG_E820_PROC=y
CONFIG_NOHIGHMEM=y
CONFIG_MTRR=y
CONFIG_X86_UP_APIC=y
CONFIG_X86_UP_IOAPIC=y
CONFIG_X86_LOCAL_APIC=y
CONFIG_X86_IO_APIC=y
#
# General setup
#
CONFIG_NET=y
CONFIG_PCI=y
# CONFIG_PCI_GOBIOS is not set
# CONFIG_PCI_GODIRECT is not set
CONFIG_PCI_GOANY=y
CONFIG_PCI_BIOS=y
CONFIG_PCI_DIRECT=y
CONFIG_PCI_NAMES=y
CONFIG_EISA=y
# CONFIG_MCA is not set
CONFIG_HOTPLUG=y

The only thing I can update is the pivot_root command from util-linux,
is this necessary ?

With best regards,

Reinhard Katzmann
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50400/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-07 Thread Juan Quintela 

> "guillaume" == Guillaume Cottenceau <[EMAIL PROTECTED]> writes:

guillaume> Reinhard Katzmann <[EMAIL PROTECTED]> writes:
>> Hi!
>> 
>> I'm unable to boot my self-made 2.4.17 kernel. When I try it
>> I get the following messages:
>> 
>> VFS: Mounted root (ext2 filesystem).
>> Mounted devfs on /dev.
>> Red Hat nash version 3.1.6-mdk starting.
>> Mounting /proc filesystem.
>> Creating root device.
>> Mounting root filesystem with flags noatime
>> kjournald starting.  Commit interval 5 seconds
>> EXT3-fs: mounted filesystem with ordered data mode.
>> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
>> Freeing unused kernel memory:708k freed.
>> Kernel panic: No init found. Try passing init= option to kernel.

guillaume> Booh :-(.

No idea what is happening here, It take a lot of time to get 2.4.17
out due to the bug that it didn't boot :(  Could you told us what is
your hardware?

>> I also tested without initrd image. This fails as well:
>> 
>> VFS: Mounted root (ext2 filesystem).
>> Mounted devfs on /dev.
>> Freeing unused kernel memory:708k freed.
>> Kernel panic: Attempted to kill init!

guillaume> Ooh this is also strange but looks like a kernel problem? Or does
guillaume> boot without pivot_rooting no more supported, dear kernel team?

I think it is not supported anymore, but for the mkinitrd :(
 
>> Any other dependencies which needs to be updated for mkinitrd ? I 
>> only checked "rpm -q --requires mkinitrd" and except the mentioned
>> packages all the other are from original Mandrake 8.1. I did not
>> make changes to my original Mandrake 8.1 except the mentioned kernel
>> packages and initscripts (I'm mainly interested in kernel/driver
>> development USB/IEEE1394 and DVB/DVD).
>> 
>> Also I checked the initrd image and found two strange things:
>> - /sbin is created with mkdir and there a link from bin to sbin
>> is created. (Same with the old version). I commented the mkdir
>> command for testing.

guillaume> This should be correct though.

>> - mkinitrd does not check the the kernel config (/boot/config-2.4.17-xyz)
>> correctly, modules are added from the kernel source tree (!)
>> which are inside the kernel (for me this was ext3, ide-mod, ide-probe-mod,
>> ide-disk). If I don't add them to IGNOREMODS insmoding will (of course)
>> fail when booting the kernel.

guillaume> I don't follow you. Modules are taken from /lib/modules/$kernver
guillaume> so what's the problem?

having to take a look at that later.

Later ,Juan.


-- 
In theory, practice and theory are the same, but in practice they 
are different -- Larry McVoy

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-07 Thread Reinhard Katzmann 

On Mon, Jan 07, 2002 at 06:30:36AM +0100, Guillaume Cottenceau wrote:
> Reinhard Katzmann <[EMAIL PROTECTED]> writes:
> 
> > Note: pivotroot returns error 22. What does this mean ?
> 
> Check /usr/include/asm/errno.h :-).
> 
> (actually it's EINVAL).

Yes, but I don't see any errors with the pivot_root line in the
mkinitrd script (actually this script works fine with 2.4.16), so
it seems really to be a kernel problem...

> > I also tested without initrd image. This fails as well:
> > 
> > VFS: Mounted root (ext2 filesystem).
> > Mounted devfs on /dev.
> > Freeing unused kernel memory:708k freed.
> > Kernel panic: Attempted to kill init!
> 
> Ooh this is also strange but looks like a kernel problem? Or does
> boot without pivot_rooting no more supported, dear kernel team?

That would not be very nice.. I'd prefer to have the possibility to
boot without initrd image and still can use one of the many filesystems :)
(after all initrd makes the kernel bigger, and without effort it's not
easy to keep the (compressed) kernel size below 1 MB).

> > Also I checked the initrd image and found two strange things:
> > - /sbin is created with mkdir and there a link from bin to sbin
> >   is created. (Same with the old version). I commented the mkdir
> >   command for testing.
> 
> This should be correct though.

It works, but I don't think it's correct:

mkdir -p $MNTIMAGE/sbin
mkdir -p $MNTIMAGE/etc
mkdir -p $MNTIMAGE/dev
mkdir -p $MNTIMAGE/loopfs
mkdir -p $MNTIMAGE/proc
mkdir -p $MNTIMAGE/sysroot
ln -s bin $MNTIMAGE/sbin

The result is: (ls -l bin sbin)
bin:
insgesamt 36
-rwxr-xr-x1 root root24220 Jan  7 09:55 insmod*
-rwxr-xr-x1 root root11128 Jan  7 09:55 nash*

sbin:
insgesamt 0
lrwxrwxrwx1 root root3 Jan  7 10:11 bin -> bin
lrwxrwxrwx1 root root9 Jan  7 09:55 modprobe -> /bin/nash

Either the sbin mkdir should be removed or the link command.

> > - mkinitrd does not check the the kernel config (/boot/config-2.4.17-xyz)
> >   correctly, modules are added from the kernel source tree (!)
> >   which are inside the kernel (for me this was ext3, ide-mod, ide-probe-mod,
> >   ide-disk). If I don't add them to IGNOREMODS insmoding will (of course)
> >   fail when booting the kernel.
> 
> I don't follow you. Modules are taken from /lib/modules/$kernver
> so what's the problem?

No they are not. For example I have ext3 in my kernel like this:
CONFIG_EXT3_FS=y

When I invoke mkinitrd and look at the result, in /lib you find the following
file (initrd image):
lib:
insgesamt 83
drwxr-xr-x2 root root 1024 Jan  7 10:16 ./
drwxr-xr-x   10 root root 1024 Jan  7 10:16 ../
-rw-r--r--1 root root81514 Jan  5 19:18 ext3.o

locate ext3.o
/usr/src/linux-2.4.17-1mdk/fs/ext3/.ext3.o.flags
/usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o

ll /usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o
-rw-r--r--1 root root81514 Jan  5 19:18 
/usr/src/linux-2.4.17-1mdk/fs/ext3/ext3.o

With best regards,

Reinhard Katzmann
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50387/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-06 Thread Guillaume Cottenceau 

Reinhard Katzmann <[EMAIL PROTECTED]> writes:

> Hi!
> 
> I'm unable to boot my self-made 2.4.17 kernel. When I try it
> I get the following messages:
> 
> VFS: Mounted root (ext2 filesystem).
> Mounted devfs on /dev.
> Red Hat nash version 3.1.6-mdk starting.
> Mounting /proc filesystem.
> Creating root device.
> Mounting root filesystem with flags noatime
> kjournald starting.  Commit interval 5 seconds
> EXT3-fs: mounted filesystem with ordered data mode.
> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
> Freeing unused kernel memory:708k freed.
> Kernel panic: No init found. Try passing init= option to kernel.

Booh :-(.

I'm currently in vacations so I can't try to fix it until jan-9
(at least). Does everyone else having this problem? Juan, does it
sound familiar/normal to you?


> Note: pivotroot returns error 22. What does this mean ?

Check /usr/include/asm/errno.h :-).

(actually it's EINVAL).

 
> This is with mkinitrd-3.1.6-16mdk, I also -12mdk (my previous version)
> without success. This version worked fine with the 2.4.16 kernel. I
> also updated e2fsprogs to 1.25-2mdk and use the recent modutils 
> 2.4.12-1mdk. (the modutils were also used with 2.4.16 kernel, while
> e2fsprogs/mkinitrd was updated when the 2.4.17 kernel did not boot).
> I used make oldconfig with my old 2.4.16 kernel config and left all
> new 2.4.17 options out.
> 
> I also tested without initrd image. This fails as well:
> 
> VFS: Mounted root (ext2 filesystem).
> Mounted devfs on /dev.
> Freeing unused kernel memory:708k freed.
> Kernel panic: Attempted to kill init!

Ooh this is also strange but looks like a kernel problem? Or does
boot without pivot_rooting no more supported, dear kernel team?
 
> Any other dependencies which needs to be updated for mkinitrd ? I 
> only checked "rpm -q --requires mkinitrd" and except the mentioned
> packages all the other are from original Mandrake 8.1. I did not
> make changes to my original Mandrake 8.1 except the mentioned kernel
> packages and initscripts (I'm mainly interested in kernel/driver
> development USB/IEEE1394 and DVB/DVD).
> 
> Also I checked the initrd image and found two strange things:
> - /sbin is created with mkdir and there a link from bin to sbin
>   is created. (Same with the old version). I commented the mkdir
>   command for testing.

This should be correct though.

> - mkinitrd does not check the the kernel config (/boot/config-2.4.17-xyz)
>   correctly, modules are added from the kernel source tree (!)
>   which are inside the kernel (for me this was ext3, ide-mod, ide-probe-mod,
>   ide-disk). If I don't add them to IGNOREMODS insmoding will (of course)
>   fail when booting the kernel.

I don't follow you. Modules are taken from /lib/modules/$kernver
so what's the problem?


-- 
Guillaume Cottenceau - http://people.mandrakesoft.com/~gc/

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-06 Thread Reinhard Katzmann 

Hi Andrew,

> On ???, 2002-01-06 at 18:05, Reinhard Katzmann wrote:
> > pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
> 
> Have you removed /initrd directory (on real root)?

No, I haven't (and according to the README I should not do it).
(Else I think 2.4.16 would not start, would it ?).

Reinhard Katzmann
-- 
Software-Engineer, Developer for Embedded Devices
Project: HyperPen Tablet USB Driver for Linux 
GnuPG Public Key available on request



msg50372/pgp0.pgp
Description: PGP signature

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-06 Thread Borsenkow Andrej 

On ÷ÓË, 2002-01-06 at 18:05, Reinhard Katzmann wrote:
> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22

Have you removed /initrd directory (on real root)?


-andrej

Re: [Cooker] Unable to boot Kernel 2.4.17.1mdk, mkinitrd-3.1.6 bugs

2002-01-06 Thread Michael Leone 

On Sun, 2002-01-06 at 10:05, Reinhard Katzmann wrote:
> Hi!
> 
> I'm unable to boot my self-made 2.4.17 kernel. When I try it
> I get the following messages:
> 
> VFS: Mounted root (ext2 filesystem).
> Mounted devfs on /dev.
> Red Hat nash version 3.1.6-mdk starting.
> Mounting /proc filesystem.
> Creating root device.
> Mounting root filesystem with flags noatime
> kjournald starting.  Commit interval 5 seconds
> EXT3-fs: mounted filesystem with ordered data mode.
> pivotroot:pivot_root(/sysroot,/sysroot/initrd) failed:22
> Freeing unused kernel memory:708k freed.
> Kernel panic: No init found. Try passing init= option to kernel.
> 
> Note: pivotroot returns error 22. What does this mean ?

I got the same thing, when using generic kernel sources. 

Don't load devfs in lilo.conf, and you'll be able to boot.

-- 

--
Michael J. Leone  Registered Linux user #201348 
ICQ: 50453890 AIM: MikeLeone

PGP Fingerprint: 0AA8 DC47 CB63 AE3F C739 6BF9 9AB4 1EF6 5AA5 BCDF
PGP public key:


Foreman, roving paving crew, Dept. of Roads, Hades.



signature.asc
Description: This is a digitally signed message part