Red Hat Bugzilla – Bug 164605
kernel-smp panics on megaraid2 (dell 1850) - x86_64
Last modified: 2007-11-30 17:07:08 EST
From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; fr-FR; rv:1.7.8) Gecko/20050511 Firefox/1.0.4 Description of problem: I got a Dell PowerEdge 1850, with a Perc 4e/Si RAID controller. I installed rhel 3 U5 x86_64 version . With the uniproc kernel, this machine boots fine, but with the smp kernel, I got a kernel panic. Version-Release number of selected component (if applicable): kernel-smp-2.4.21-32.0.1.ELsmp How reproducible: Always Steps to Reproduce: 1. Install RHEL3_U5_x86_64 2. Boot the smp kernel 3. Kernel panics. Additional info:
Created attachment 117263 [details] Console capture of the kernel panic
Can you boot with mem=2000m? How much RAM do you have installed? 4G is the magic number so far as 64bit bugs go. I saw this exact bug on a non-RedHat kernel earlier tonight with 4G EMT64 system. I used the mem= command line parameter to scrape 50 megs off the top and that booted for me. On my system because of the way the memory is laid out I guess, I use mem=4600m and it shows that I have around 4G installed when I type free. FC3 works on this system no problem.
BTW. I used both the megaraid and the megaraid2 drivers and they both had the same error with the pci_map_single. Also can you configure IOMMU in your BIOS? Sometimes you can. I tried poking around with the iommu= boot params but that seemed like a voodoo approach because I wasn't sure what I was doing. For example, iommu=memaper=2 cuts down on the ammount of RAM available and it works but you may as well do the mem= trick because it's cleaner.
The Dell PowerEdge 1850 is EMT64 as well. This bug is similar to bug 160799 and bug 146199 except that for you it is the install kernel that's goofing. Install with mem=2000m and change to the correct kernel afterwards. Let us know how that works out for you.
Use the correct (ia32e) RPM. By the way, in U6 we've changed the panic you'll get when you do this to use a message that is more explicit about what's wrong. That's what bug 160799 was about.