Hardware RAID... kernel panic

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

I'm new to installing Gentoo on hardware RAID... I'm using the newest version to date, amd64, on a hardened profile, with hardened-sources. /dev/sda* recognizes all of the partitions on the device. I applied new filesystems to them and started building, assuming the hardware RAID was invisible to the OS. Finish installing, reboot, KP. Great.

Boot back off the livecd, mount, chroot, add support for the RAID device (Symbios Logic SAS1068 PCI-X Fusion-MPT SAS), reboot, still KP. Repeat a few times, and we're at the present.

I did some googling about hw raid, it said to check /dev/mapper from the livecd. I have control only, which I believe indicates that the RAID volume wasn't identified by the livecd. modprobe shows me that it's using mptsas kernel module. The drive is recognized and functions normally as /dev/sda, but the system can't seem to identify /dev/sda at boot.

/dev/mapper contents, lspci -k, and .config below:

DaggyStyle · Watchman Joined: 22 Mar 2006 Posts: 5941

what exact kp are you getting?
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein

krinn · Watchman Joined: 02 May 2003 Posts: 7470

Yes, the disks for the kernel will appears as sda and /dev/mapper will remain empty as it's not for hardward raid but software volume
Depending on your driver it should be more informative in the kernel dmesg
Look for a sample : https://forums.gentoo.org/viewtopic-p-6350807.html#6350807

And if your sign is correct, you have more than just 1 controller, and more than 1 disk attach to them, so you are just assuming like many users that linux should call your disk /dev/sda because linux call it /dev/sda when you use the livecd : it's a mistake.
Look here for explain & way to fix that easy : https://forums.gentoo.org/viewtopic-p-6824250.html#6824250

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

Thanks for the reply.

No this isn't my personal desktop, it's a Dell PowerEdge SC440 server. According to Dell the part number for the RAID controller is UN939, which is listed as "PERC 5/ir SAS PCI-e"

lspci -k tells me:

krinn · Watchman Joined: 02 May 2003 Posts: 7470

No you're incorrect assuming /dev/mapper should have any values : this is logical volume manager, it's not for hardware raid as i already told you, so except if you build software raid over an hardware raid, this will remain empty.

All the "raid6" and "md" messages you see in your dmesg are just because you have enable software raid, unrelated to hardware itself, but no harm to have it.

Because of your short output, canno't really see sda disk status, at least it show your raid array is handle and assign as sg3

Post your actual kernel error when you boot like DaggyStyle told you (in a too short version in my taste), let's expand his quote because kp isn't killer potatoes :

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

krinn · Watchman Joined: 02 May 2003 Posts: 7470

the driver sr is load only and your cdrom appears alone

You have something missing :
- lacking support for scsi disk in kernel (sd)
- lacking support for you disk controller (mptsas)
- support for disk controller is build as module and not include in kernel (mptsas should be Y and not M)

Eheh, and if you think your kernel have the 3, then maybe you've build it, but like many users do, fail to actually install it (there's many ways to fail doing such a simple task : including not mounting boot and kernel is now copy to /boot in your root partition instead of your boot partition, bad target name when copying...)

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

SCSI:

krinn · Watchman Joined: 02 May 2003 Posts: 7470

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

Whoops.

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

I recompiled, verified that /boot was mounted, copied the kernel over, updated grub.conf, and rebooted, same error.

Okay, I just checked the output... for some reason it built an x86 kernel. This would explain why it's not updating as it should as i've been copying the 64-bit kernel. I'm chrooting back in now as we speak... why would it build a 32-bit kernel? my CHOST is the default 64-bit one.

EDIT: /usr/src/linux/arch/x86_64/boot/bzImage is a symlink to /usr/src/linux/arch/x86/boot/bzImage. Is that normal???
_________________
Sysadmin by trade, geek by choice

DESKTOP:
i7-3770K|8800GTS-512|32GB|1x256GB SSD|4x1TB|HD/BDROM|1920x1200+1680x1050
Win8 Pro w/ Hyper-V|Server 2012 VM|Ubuntu VM|Gentoo VM

TABLET: Samsung Ativ 700t
i5-3317U|4GB|128GB SSD|1920x1080 Touch
Win8

krinn · Watchman Joined: 02 May 2003 Posts: 7470

for the symlink, i don't really know, i suppose it's normal, i don't expect kernel to build itself something weird
you can check you your kernel with "file pathtokernel" and lookout it's a 64bits or not

for your kernel boot problem, you still don't see sd answer ? no partitions are shown? and also check the line just after "kernel panic" you should see your kernel version, and it should tell you :
swapper not tainted kernelversion #number

#number is total number of compilation you've made to build that kernel, as your previous pic show #3 at least if you build and install a newer one, that number should be higher.

Feel free to repost a pic if you prefer

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

According to the IRC guys that's normal. x86_64 and i386 were merged into x86.

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found

FcukThisGame · l33t Joined: 20 Apr 2005 Posts: 776 Location: /lost+found