[gentoo-user] Problem with new install

[gentoo-user] Problem with new install

[Mike Diehl]

Hi all.

I've never had this much trouble with a server before, but I've been pulling 
my hair out.

The install seemed to go well, but when I rebooted it from it's own hard 
drive, it fails.  fsck claims that it can't open /dev/sda3 or that the 
superblock doesn't describe a valid ext2 filesystem.

However, when I reboot from the live CD, it mounts just fine and fsck says 
it's clean.

Here is the /etc/fstab:
/dev/sda1       /boot           ext2            noauto,noatime  1 2
/dev/sda3       /               ext2            noatime         0 1
/dev/sda2       none            swap            sw              0 0
/dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0
shm                     /dev/shm        tmpfs   nodev,nosuid,noexec     0 0

Here is the /boot/grub/grub.conf file:
default 0
timeout 30
splashimage=(hd0,0)/boot/grub/splash.xpm.gz

title Gentoo Linux 
root (hd0,0)
kernel /bzImage root=/dev/sda3

I've verified that ext2 and ext3 are in the kernel statically.  I've also 
compiled in ALL of the SATA drivers, statically.

What am I missing?

-- 

Take care and have fun,
Mike Diehl.

Re: [gentoo-user] Problem with new install

[Florian Philipp]

[-- Attachment #1: Type: text/plain, Size: 2182 bytes --]

Am 15.10.2010 19:29, schrieb Mike Diehl:
> Hi all.
> 
> I've never had this much trouble with a server before, but I've been pulling 
> my hair out.
> 
> The install seemed to go well, but when I rebooted it from it's own hard 
> drive, it fails.  fsck claims that it can't open /dev/sda3 or that the 
> superblock doesn't describe a valid ext2 filesystem.
> 
> However, when I reboot from the live CD, it mounts just fine and fsck says 
> it's clean.
> 
> Here is the /etc/fstab:
> /dev/sda1       /boot           ext2            noauto,noatime  1 2
> /dev/sda3       /               ext2            noatime         0 1
> /dev/sda2       none            swap            sw              0 0
> /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0
> shm                     /dev/shm        tmpfs   nodev,nosuid,noexec     0 0
> 
> Here is the /boot/grub/grub.conf file:
> default 0
> timeout 30
> splashimage=(hd0,0)/boot/grub/splash.xpm.gz
> 
> title Gentoo Linux 
> root (hd0,0)
> kernel /bzImage root=/dev/sda3
> 
> I've verified that ext2 and ext3 are in the kernel statically.  I've also 
> compiled in ALL of the SATA drivers, statically.
> 
> What am I missing?
> 

*All* of the drivers could be too much. There is a generic driver which
can prevent the "right" driver from taking over. In that case you end up
with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
= CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
CONFIG_IDE_GENERIC.

I think it is the second option that causes that problem. However, you
won't need the first option, either.

Instead of your brute-force "yes to all" approach, newer kernels also
support `make localyesconfig` which takes all modules currently used in
the running kernel and compiles them into the new kernel. It is very
helpful when you already have a good but generic kernel like the one on
your live CD.

If even that doesn't help, it might be possible that the device
numbering has changed and your hard disk is detected as /dev/sdb or so.
Try mounting it by UUID (google for it, please).

Hope this helps,
Florian Philipp


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-user] Problem with new install

[Mike Diehl]

On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:

> 
> *All* of the drivers could be too much. There is a generic driver which
> can prevent the "right" driver from taking over. In that case you end up
> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
> CONFIG_IDE_GENERIC.

I will try this.

> I think it is the second option that causes that problem. However, you
> won't need the first option, either.

Ya, I don't like having EVERYTHING in the kernel, but nothing else was 
working, so I figured I'd give it a try.
 
> Instead of your brute-force "yes to all" approach, newer kernels also
> support `make localyesconfig` which takes all modules currently used in
> the running kernel and compiles them into the new kernel. It is very
> helpful when you already have a good but generic kernel like the one on
> your live CD.

Oh now this is cool.  Thank you.  I'll try this, also.

> If even that doesn't help, it might be possible that the device
> numbering has changed and your hard disk is detected as /dev/sdb or so.
> Try mounting it by UUID (google for it, please).

I've tried changing grub to point to sdb and hda.  Perhaps, I need to change 
etc/fstab...

Thank you.

> Hope this helps,
> Florian Philipp

-- 

Take care and have fun,
Mike Diehl.

Re: [gentoo-user] Problem with new install

[Dan Cowsill]

[-- Attachment #1: Type: text/plain, Size: 702 bytes --]

On 15/10/2010 10:47 AM, Mike Diehl wrote:
> On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
>> Instead of your brute-force "yes to all" approach, newer kernels also
>> support `make localyesconfig` which takes all modules currently used in
>> the running kernel and compiles them into the new kernel. It is very
>> helpful when you already have a good but generic kernel like the one on
>> your live CD.
> 
> Oh now this is cool.  Thank you.  I'll try this, also.
> 

I agree, that is pretty cool!

My two cents:  Don't discount the possibility of hardware failure.  Is
the device new, or old?  Have you kept an eye on the SMART attributes of
the device?

Cheers,
D



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 260 bytes --]

Re: [gentoo-user] Problem with new install

[Per-Erik Westerberg]

fre 2010-10-15 klockan 11:29 -0600 skrev Mike Diehl:
> Hi all.
> 
> I've never had this much trouble with a server before, but I've been pulling 
> my hair out.
> 
> The install seemed to go well, but when I rebooted it from it's own hard 
> drive, it fails.  fsck claims that it can't open /dev/sda3 or that the 
> superblock doesn't describe a valid ext2 filesystem.
> 
> However, when I reboot from the live CD, it mounts just fine and fsck says 
> it's clean.
> 
> Here is the /etc/fstab:
> /dev/sda1       /boot           ext2            noauto,noatime  1 2
> /dev/sda3       /               ext2            noatime         0 1
> /dev/sda2       none            swap            sw              0 0
> /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0
> shm                     /dev/shm        tmpfs   nodev,nosuid,noexec     0 0
> 
> Here is the /boot/grub/grub.conf file:
> default 0
> timeout 30
> splashimage=(hd0,0)/boot/grub/splash.xpm.gz
> 
> title Gentoo Linux 
> root (hd0,0)
> kernel /bzImage root=/dev/sda3
> 
> I've verified that ext2 and ext3 are in the kernel statically.  I've also 
> compiled in ALL of the SATA drivers, statically.
> 
> What am I missing?
> 

Hi,

I had exactly the same problem when I did a reinstall, I used the kernel
".config" I got from the live CD (zcat /proc/config.gz > .config) as a
base when configuring the kernel. I solved it by not using that and
instead start from scratch without an initial ".config". I guess that
there were some options that was conflicting.

  BR / P-E

Re: [gentoo-user] Problem with new install

[Mike Diehl]

On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
> Am 15.10.2010 19:29, schrieb Mike Diehl:
> > Hi all.
> > 
> > I've never had this much trouble with a server before, but I've been
> > pulling my hair out.
> > 
> > The install seemed to go well, but when I rebooted it from it's own hard
> > drive, it fails.  fsck claims that it can't open /dev/sda3 or that the
> > superblock doesn't describe a valid ext2 filesystem.

> *All* of the drivers could be too much. There is a generic driver which
> can prevent the "right" driver from taking over. In that case you end up
> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
> CONFIG_IDE_GENERIC.
> I think it is the second option that causes that problem. However, you
> won't need the first option, either.

I tried this, first without success.  I then ran through all combinations of 
sda3, sdb3, hda3, hdb3 in /etc/fstab.  This didn't work.

> Instead of your brute-force "yes to all" approach, newer kernels also
> support `make localyesconfig` which takes all modules currently used in
> the running kernel and compiles them into the new kernel. It is very
> helpful when you already have a good but generic kernel like the one on
> your live CD.

I tried this, next.  At least now, I believe I have a viable kernel.  But it 
still didn't work.

> If even that doesn't help, it might be possible that the device
> numbering has changed and your hard disk is detected as /dev/sdb or so.
> Try mounting it by UUID (google for it, please).

I tried this.  Only now, fsck.ext2 tells me that it can't resolve the UUID.

Here is the new fstab:
/dev/sda1       /boot           ext2            noauto,noatime  1 2                           
                                                                                              
UUID=ba7511dd-a5f9-48d8-8102-cf71c08a0c7b     /       ext2    noatime    0 1   
                                                                                              
/dev/sda2       none            swap            sw              0 0                           
/dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0  

At this point, I'm going to move the drive to a different port on the SATA 
chain; shouldn't change anything, but I'm running out of ideas.  I'll also 
check the BIOS for anything stupid-obvious.

So, I guess I'm still stuck!

> Hope this helps,
> Florian Philipp

-- 

Take care and have fun,
Mike Diehl.

Re: [gentoo-user] Problem with new install

[Mike Diehl]

One more interesting tidbit:

When it boots, I can see that VFS was able to mount ROOT on device 8:3.

So, it doesn't seem to be hardware or driver related.  Looks like something 
with the filesystem?  fsck.ext2 is the only thing complaining.  However, when 
I boot from CD, I can mount it just fine.

Anyone seen this before?

Mike.

On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
> Am 15.10.2010 19:29, schrieb Mike Diehl:
> > Hi all.
> > 
> > I've never had this much trouble with a server before, but I've been
> > pulling my hair out.
> > 
> > The install seemed to go well, but when I rebooted it from it's own hard
> > drive, it fails.  fsck claims that it can't open /dev/sda3 or that the
> > superblock doesn't describe a valid ext2 filesystem.
> > 
> > However, when I reboot from the live CD, it mounts just fine and fsck
> > says it's clean.
> > 
> > Here is the /etc/fstab:
> > /dev/sda1       /boot           ext2            noauto,noatime  1 2
> > /dev/sda3       /               ext2            noatime         0 1
> > /dev/sda2       none            swap            sw              0 0
> > /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0
> > 0 shm                     /dev/shm        tmpfs   nodev,nosuid,noexec   
> >  0 0
> > 
> > Here is the /boot/grub/grub.conf file:
> > default 0
> > timeout 30
> > splashimage=(hd0,0)/boot/grub/splash.xpm.gz
> > 
> > title Gentoo Linux
> > root (hd0,0)
> > kernel /bzImage root=/dev/sda3
> > 
> > I've verified that ext2 and ext3 are in the kernel statically.  I've also
> > compiled in ALL of the SATA drivers, statically.
> > 
> > What am I missing?
> 
> *All* of the drivers could be too much. There is a generic driver which
> can prevent the "right" driver from taking over. In that case you end up
> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
> CONFIG_IDE_GENERIC.
> 
> I think it is the second option that causes that problem. However, you
> won't need the first option, either.
> 
> Instead of your brute-force "yes to all" approach, newer kernels also
> support `make localyesconfig` which takes all modules currently used in
> the running kernel and compiles them into the new kernel. It is very
> helpful when you already have a good but generic kernel like the one on
> your live CD.
> 
> If even that doesn't help, it might be possible that the device
> numbering has changed and your hard disk is detected as /dev/sdb or so.
> Try mounting it by UUID (google for it, please).
> 
> Hope this helps,
> Florian Philipp

-- 

Take care and have fun,
Mike Diehl.

Re: [gentoo-user] Problem with new install

[Alan McKinnon]

Apparently, though unproven, at 19:47 on Friday 15 October 2010, Mike Diehl 
did opine thusly:

> > Instead of your brute-force "yes to all" approach, newer kernels also
> > support `make localyesconfig` which takes all modules currently used in
> > the running kernel and compiles them into the new kernel. It is very
> > helpful when you already have a good but generic kernel like the one on
> > your live CD.
> 
> Oh now this is cool.  Thank you.  I'll try this, also.

You could also copy the working kernel and it's modules off the LiveCD onto 
the hard disk. It won't be perfect but at least you'd have something that 
works and can boot to assist troubleshooting. Booting and running off LiveCDs 
is both convenient and painful.


-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] Problem with new install

[Florian Philipp]

[-- Attachment #1: Type: text/plain, Size: 3429 bytes --]

Am 15.10.2010 21:23, schrieb Mike Diehl:
> On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
>> Am 15.10.2010 19:29, schrieb Mike Diehl:
>>> Hi all.
>>>
>>> I've never had this much trouble with a server before, but I've been
>>> pulling my hair out.
>>>
>>> The install seemed to go well, but when I rebooted it from it's own hard
>>> drive, it fails.  fsck claims that it can't open /dev/sda3 or that the
>>> superblock doesn't describe a valid ext2 filesystem.
> 
>> *All* of the drivers could be too much. There is a generic driver which
>> can prevent the "right" driver from taking over. In that case you end up
>> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
>> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
>> CONFIG_IDE_GENERIC.
>> I think it is the second option that causes that problem. However, you
>> won't need the first option, either.
> 
> I tried this, first without success.  I then ran through all combinations of 
> sda3, sdb3, hda3, hdb3 in /etc/fstab.  This didn't work.
> 
>> Instead of your brute-force "yes to all" approach, newer kernels also
>> support `make localyesconfig` which takes all modules currently used in
>> the running kernel and compiles them into the new kernel. It is very
>> helpful when you already have a good but generic kernel like the one on
>> your live CD.
> 
> I tried this, next.  At least now, I believe I have a viable kernel.  But it 
> still didn't work.
> 
>> If even that doesn't help, it might be possible that the device
>> numbering has changed and your hard disk is detected as /dev/sdb or so.
>> Try mounting it by UUID (google for it, please).
> 
> I tried this.  Only now, fsck.ext2 tells me that it can't resolve the UUID.
> 
> Here is the new fstab:
> /dev/sda1       /boot           ext2            noauto,noatime  1 2                           
>                                                                                               
> UUID=ba7511dd-a5f9-48d8-8102-cf71c08a0c7b     /       ext2    noatime    0 1   
>                                                                                               
> /dev/sda2       none            swap            sw              0 0                           
> /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0  
> 
> At this point, I'm going to move the drive to a different port on the SATA 
> chain; shouldn't change anything, but I'm running out of ideas.  I'll also 
> check the BIOS for anything stupid-obvious.
> 
> So, I guess I'm still stuck!
> 

Hmm, sounds like a serious problem. I suggest you try to get into an
early stage during boot and try to move forward from there. Try to add
'1' to the parameters in order to get into single-user mode. You can
also try 'init=/bin/bash'.

There are lots of other options you can try. For a long time, 'noapic'
(not 'noapci') was my first candidate for odd boot issues. Take a look
at /usr/sr/linux/Documentation/kernel-parameters.txt for more options.

Also, which kernel sources are you using and which live CD (with which
kernel version)?  Is there a specific reason why you use ext2 for root?
What kind of system do you run, anyway? And, just by chance, you are not
using an extremely large (>1TB) drive which might happen to have 4kB
blocks instead of 512 B?

Regards,
Florian Philipp


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-user] Problem with new install

[Bill Longman]

On 10/15/2010 12:23 PM, Mike Diehl wrote:
> On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
>> Am 15.10.2010 19:29, schrieb Mike Diehl:
>>> Hi all.
>>>
>>> I've never had this much trouble with a server before, but I've been
>>> pulling my hair out.
>>>
>>> The install seemed to go well, but when I rebooted it from it's own hard
>>> drive, it fails.  fsck claims that it can't open /dev/sda3 or that the
>>> superblock doesn't describe a valid ext2 filesystem.
> 
>> *All* of the drivers could be too much. There is a generic driver which
>> can prevent the "right" driver from taking over. In that case you end up
>> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
>> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
>> CONFIG_IDE_GENERIC.
>> I think it is the second option that causes that problem. However, you
>> won't need the first option, either.
> 
> I tried this, first without success.  I then ran through all combinations of 
> sda3, sdb3, hda3, hdb3 in /etc/fstab.  This didn't work.
> 
>> Instead of your brute-force "yes to all" approach, newer kernels also
>> support `make localyesconfig` which takes all modules currently used in
>> the running kernel and compiles them into the new kernel. It is very
>> helpful when you already have a good but generic kernel like the one on
>> your live CD.
> 
> I tried this, next.  At least now, I believe I have a viable kernel.  But it 
> still didn't work.
> 
>> If even that doesn't help, it might be possible that the device
>> numbering has changed and your hard disk is detected as /dev/sdb or so.
>> Try mounting it by UUID (google for it, please).
> 
> I tried this.  Only now, fsck.ext2 tells me that it can't resolve the UUID.
> 
> Here is the new fstab:
> /dev/sda1       /boot           ext2            noauto,noatime  1 2                           
>                                                                                               
> UUID=ba7511dd-a5f9-48d8-8102-cf71c08a0c7b     /       ext2    noatime    0 1   
>                                                                                               
> /dev/sda2       none            swap            sw              0 0                           
> /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0 0  
> 
> At this point, I'm going to move the drive to a different port on the SATA 
> chain; shouldn't change anything, but I'm running out of ideas.  I'll also 
> check the BIOS for anything stupid-obvious.

You might also want to jump into grub's shell and look around in /dev
for devices. If they are there, you know the kernel is providing the
modules correctly. They might not be sda but hda or vice-versa. Also,
you can use lsmod and make sure.

Re: [gentoo-user] Problem with new install

[Mike Diehl]

Well, I got it to boot.  Now that it's up, I can see what the problem was/is.  
There are no disk device files in /dev.

I created the appropriate nodes and things went much better.

The problem was that CONFIG_SYSFS_DEPRECATED_V2 was on

I just closed up the chassis and it will ship in the morning.

Thank you for your help!

Mike.

On Friday 15 October 2010 3:02:27 pm Florian Philipp wrote:
> Am 15.10.2010 21:23, schrieb Mike Diehl:
> > On Friday 15 October 2010 11:40:34 am Florian Philipp wrote:
> >> Am 15.10.2010 19:29, schrieb Mike Diehl:
> >>> Hi all.
> >>> 
> >>> I've never had this much trouble with a server before, but I've been
> >>> pulling my hair out.
> >>> 
> >>> The install seemed to go well, but when I rebooted it from it's own
> >>> hard drive, it fails.  fsck claims that it can't open /dev/sda3 or
> >>> that the superblock doesn't describe a valid ext2 filesystem.
> >> 
> >> *All* of the drivers could be too much. There is a generic driver which
> >> can prevent the "right" driver from taking over. In that case you end up
> >> with a /dev/hda node and no DMA. Try to deactivate "Generic ATA support"
> >> = CONFIG_ATA_GENERIC and "generic/default IDE chipset support" =
> >> CONFIG_IDE_GENERIC.
> >> I think it is the second option that causes that problem. However, you
> >> won't need the first option, either.
> > 
> > I tried this, first without success.  I then ran through all combinations
> > of sda3, sdb3, hda3, hdb3 in /etc/fstab.  This didn't work.
> > 
> >> Instead of your brute-force "yes to all" approach, newer kernels also
> >> support `make localyesconfig` which takes all modules currently used in
> >> the running kernel and compiles them into the new kernel. It is very
> >> helpful when you already have a good but generic kernel like the one on
> >> your live CD.
> > 
> > I tried this, next.  At least now, I believe I have a viable kernel.  But
> > it still didn't work.
> > 
> >> If even that doesn't help, it might be possible that the device
> >> numbering has changed and your hard disk is detected as /dev/sdb or so.
> >> Try mounting it by UUID (google for it, please).
> > 
> > I tried this.  Only now, fsck.ext2 tells me that it can't resolve the
> > UUID.
> > 
> > Here is the new fstab:
> > /dev/sda1       /boot           ext2            noauto,noatime  1 2
> > 
> > UUID=ba7511dd-a5f9-48d8-8102-cf71c08a0c7b     /       ext2    noatime   
> > 0 1
> > 
> > /dev/sda2       none            swap            sw              0 0
> > /dev/cdrom              /mnt/cdrom      auto            noauto,ro       0
> > 0
> > 
> > At this point, I'm going to move the drive to a different port on the
> > SATA chain; shouldn't change anything, but I'm running out of ideas. 
> > I'll also check the BIOS for anything stupid-obvious.
> > 
> > So, I guess I'm still stuck!
> 
> Hmm, sounds like a serious problem. I suggest you try to get into an
> early stage during boot and try to move forward from there. Try to add
> '1' to the parameters in order to get into single-user mode. You can
> also try 'init=/bin/bash'.
> 
> There are lots of other options you can try. For a long time, 'noapic'
> (not 'noapci') was my first candidate for odd boot issues. Take a look
> at /usr/sr/linux/Documentation/kernel-parameters.txt for more options.
> 
> Also, which kernel sources are you using and which live CD (with which
> kernel version)?  Is there a specific reason why you use ext2 for root?
> What kind of system do you run, anyway? And, just by chance, you are not
> using an extremely large (>1TB) drive which might happen to have 4kB
> blocks instead of 512 B?
> 
> Regards,
> Florian Philipp

-- 

Take care and have fun,
Mike Diehl.