Hi, I have tried to built a linux kernel for Dell T3500 with Intell ICH10 chip and have been failing so far. The error seems to be related to AHCI. The last messages are that loop module has been loaded and next about controller cannot do SNFT, terminiting off CAP_SNFT and next controller reset failling and many line of ahcp: [...] failed to stop engines (-5). Any ideas what I have omitted in kernel? Ubuntu seems to work fine on the box. Wawrzek -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
2010/10/29 Wawrzyniec Niewodniczański <wawrzek@gmail.com>:
Hi, [...] Any ideas what I have omitted in kernel? Ubuntu seems to work fine on the box.
I might be too prompt with this ubuntu. 10.4 seems to work fine, but not 10.10 and it might be difference between configuration with 1 or 2 disks. Wawrzek -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
2010/10/29 Emmanuel Benisty <benisty.e@gmail.com>:
2010/10/29 Wawrzyniec Niewodniczański <wawrzek@gmail.com>:
Any ideas what I have omitted in kernel? Ubuntu seems to work fine on the box.
can you please attach: * your kernel .config * lspci -n * cat /etc/fstab * cat /proc/cpuinfo
Please find in attachment. I have RAID1 with two disk md1 (boot) md2 (LVM2) on LVM root and something else. Wawrzek -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
On Fri, 29 Oct 2010 16:32:15 +0100 Wawrzyniec Niewodniczański wrote:
Hi,
I have tried to built a linux kernel for Dell T3500 with Intell ICH10 chip and have been failing so far. The error seems to be related to AHCI. The last messages are that loop module has been loaded and next about controller cannot do SNFT, terminiting off CAP_SNFT and next controller reset failling and many line of ahcp: [...] failed to stop engines (-5).
Any ideas what I have omitted in kernel? Ubuntu seems to work fine on the box.
Are you referring to this? https://bugzilla.kernel.org/show_bug.cgi?id=16228 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/653238 http://marc.info/?l=linux-kernel&m=128476278029918&w=2 Add 'pci=nocrs' as kernel option, that shold work According to this[1], Ubuntu patches[2] are bit too invasive, so they will probably not be merged until 2.6.37, unless Linus has a Dell laptop too :). So, you can either try to apply Ubuntu patches to vanilla kernel and build that, or use pci=nocrs until new kernel release/Dell BIOS update comes, because this seems to be BIOS related.
Wawrzek
[1]https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/647043/comments/14 [2]http://people.canonical.com/~ogasawara/lp647043/ Good luck. Pedja
Hi, Thanks Pedja for you great email. In my case it's not a laptop but workstation, but anyway I have found that pci='nocrs' workaround works. (BTW, Another one we found is turning off AHCI in BIOS and leaving only ATA). So now I have pure kernel issue - Kernel doesn't see root partition. It gives me the list of available partition with sdaX and sdbX ones, but not trace of RAID/LVM (and I have root on RAID/LVM). I'm pretty sure I have compiled all necessary option to allow boot from /dev/main/root which is on LVM on /dev/md2. I guess that RAID is fine, cause kernel is booting from /dev/md1. Any ideas? Wawrzek PS. I'm attaching my lilo.conf. -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
2010/11/1 Wawrzyniec Niewodniczański <wawrzek@gmail.com>: [...]
/dev/main/root which is on LVM on /dev/md2. I guess that RAID is fine, cause kernel is booting from /dev/md1.
After all it might be wrong assumption. I've just checked some kernel message from about Kernel Panic and there is something that md found and added 0 devices... Wawrzek
2010/11/1 Wawrzyniec Niewodniczański <wawrzek@gmail.com>:
2010/11/1 Wawrzyniec Niewodniczański <wawrzek@gmail.com>:
[...]
/dev/main/root which is on LVM on /dev/md2. I guess that RAID is fine, cause kernel is booting from /dev/md1.
After all it might be wrong assumption. I've just checked some kernel message from about Kernel Panic and there is something that md found and added 0 devices...
I resolve my issue. The problem was partition type. When I changed it to Raid Auto Detect the system started to boot properly. Wawrzek -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
Hi, After another day of investigation I found that both /boot and /root have to be on RAID array with metadata 0.90 (at least for LILO). So now I have: md1 - (sd{a,b}1 - boot) md2 - (sd{a,b}2 - root) md3 - (sd{a,b}3 - LVM) Another thing I have found is that system doesn't add md3 automatic, so LVM is not initialize nor partition on it. I had to add /sbin/mdadm --assemble - scan to /etc/rc (line 16 - before device-mapper nodes creation) Before I'll update Wiki I would like to ensure that may finding are correct. Cheers, Wawrzek -- Wawrzyniec Niewodniczański vel Wawrzek Larry or LarryN Linux User #177124 E-MAIL: wawrzek@gmail.com PhD in Quantum Chemistry WWW: http://wawrzek.name MSc in Molecular Engineering JID: wawrzek@jabber.wroc.pl
Excerpts from Wawrzyniec Niewodniczański's message of Wed Nov 03 13:03:41 +0100 2010:
After another day of investigation I found that both /boot and /root have to be on RAID array with metadata 0.90 (at least for LILO). You could also try extlinux, I think it's superior at least in RAID environments (while being a lot smaller and more modular than GRUB)
Best regards, Moritz
Hi, Wawrzek! Am 03.11.2010 13:03, schrieb Wawrzyniec Niewodniczański:
After another day of investigation I found that both /boot and /root have to be on RAID array with metadata 0.90 (at least for LILO).
So now I have:
md1 - (sd{a,b}1 - boot) md2 - (sd{a,b}2 - root) md3 - (sd{a,b}3 - LVM)
Another thing I have found is that system doesn't add md3 automatic, so LVM is not initialize nor partition on it. I had to add /sbin/mdadm --assemble - scan to /etc/rc (line 16 - before device-mapper nodes creation)
Before I'll update Wiki I would like to ensure that may finding are correct.
Feel free to update my article (mark the LVM specific changes) or create a new one. I cannot really assist you further since I never used LVM and I have no idea how LVM deals with RAIDs. Did you test the RAID by hot-plugging a harddisk (SATA is usually no problem) and rebooting the computer? LILO has to boot from each disk, no matter which one failed. A new spare disk also needs LILO installed in the MBR... Regards, Clemens
participants (5)
-
Clemens Koller -
Emmanuel Benisty -
Moritz Wilhelmy -
Predrag Ivanovic -
Wawrzyniec Niewodniczański