We have built our IBM Netfinity 4000R servers with RH7.0 and the stock kernel, and used mirroring on our two internal SCSI disks. We had built a 2.2.18 kernel and with no problems. We then built a 2.4.0 kernel (similar config to 2.2.18) and booted. After the two drives are identified, the kernel panics with "unable to mount root fs". Anybody have any ideas how that could happen? We obviously did select RAID in the kernel config. Help!!!
Bit more info. We had actually tried 2.4.0-test9 previously without any problems! More correctly the message we get is ext2: unable to read superblock EXT2-fs: unable to mount root fs. Why would this be??? I can't image it's a kernel bug, but I don't know what we did wrong.
You used stock 2.4.0 tarball? Did you 'make config' based on previous .config (use make oldconfig first) or fresh? Did you remember to use 'mkinitrd' ?
>You used stock 2.4.0 tarball? Did you 'make config' based on previous .config >(use make oldconfig first) > or fresh? > > Did you remember to use 'mkinitrd' ? Yes to all! I've tried with 2.4.0, 2.4.0-ac4, 2.4.0-pre1, 2.4.0-prerelease, test12, test11, test10. No luck. I'm just going to try test9 again.
Solution is described in linux/Documentation/md.txt. You don't need any initrd for this, just a kernel command line in lilo.conf. My system boots off /dev/md0 using RAID 1: image=/boot/vmlinuz-2.4.1-pre10-2 label=linux append = "md=0,/dev/hdc1,/dev/hdb3" read-only root=/dev/md0
nicku, when RAID is built as a module (as it is in our kernels) you *do* need an initrd for RAID root.
Alrighty then! We gave up on RAID for the short term. Now RedHat 7.1Beta is out and we though - cool! maybe this will solve this problem because it installs a 2.4 kernel from scratch. We kickstarted the 7.1B install with RAID 1. Install works great, machine boots (although not the SMP kernel, but that's a different problem). Using RH7.1B as the build machine, we built a 2.4.1 stock kernel (NOT RH's source). Much the same as last time. Install and boot - messages look good, SCSI and RAID modules load fine, then pow: EXT2-fs: unable to read superblock iosfs_read_super: bread failed, dev=09:02, iso_blknum=16, block=32 Kernel panic: VFS: Unable to mount root fs on 09:02 We have tried recompiling the kernel every which way from Sunday to no avail. If I didn't know better I'd say that RH has somehow changed the RAID or filesystem code from the stock kernel. Is that possible? What could we possibly be doing wrong???! HELP!
If you are using an adaptec driver, please try upgrading to wolverine, our latest beta release. If not, please describe your hardware.
Created attachment 11440 [details] Kickstart file for IBM eServer w RAID1
Installed with Wolverine. This time it didn't boot at all. Exactly the same error. Somewhat reassuring actually. Hardware is an IBM eServer 330, dual PIII/800. Adaptec AIC 7892 SCSI. Dual hot-swap Seagate 9GB drives (sorry, don't know model off hand - can look if helpful). Dual eepro100 ethernet. 2GB of RAM. I've attached my kickstart file.
This really looks like a configuration error. Can you please set up a serial console and give us the output from your boot attempt with wolverine installed? 1 connect a null modem cable to another machine 2 run minicom on the other machine, set to 115200n8 no flow control 3 turn on capturing in minicom 4 boot the affected machine with args boot: linux console=ttyS0,115200 console=tty0 5 attach the capture file
Created attachment 11944 [details] Wolverine RAID boot output
Oops, guess I've got to actually update this bug before you see that I've put an attachment on. Here's my log! I hope it's all-revealing!
This is starting to look like it might be an anaconda bug.
Don't have a lot of info here, but my colleague has installed this server with RedHat 7.0 (2.2.18 kernel) with RAID, then built a 2.4.2-ac16 kernel - with RAID and SCSI (aic7xxx) built-in, i.e. not modules. The machine booted! I generally make these things modules, and use an initrd. I don't know what is used in the out-of-the-box Wolverine kernel, but we were wondering if it's maybe a module loading order problem. Just grasping at straws. Also noticed that 2.4.2-pre3 and newer appear to have a new aic7xxx driver. Maybe that has something to do with it...
Woohoo! Built a server with RH7.0 with all updates applied. Built a 2.4.3-pre8 kernel with RAID and SCSI built in (not modules). It works!!! Now I'm not sure whether the problem is fixed by the RH7.0 updates, or by some change in the 2.4 kernel. Sigh. I'll try to find time to test an older 2.4 kernel.
Confirmed that 2.2.18 with SCSI and RAID 1 built in on top of RH7.0 with all updates does not work. Built 2.4.3 (final) with aic7xxx and RAID 1 built in and it works fine! Still getting the md3 overlaps with md2 messages, but it appears to be some sort of kernel problem that is now fixed???
Do you still see the problem with Red Hat Linux 7.1?
Closing due to inactivity. Please reopen if you have any more info.