Installing PSP 8.11 on RHAS4U6 crashes machine during install

[Roger Heflin]

I attempted to install PSP 8.11 on a RHAS4U6 machine that had just been upgraded from RHAS4U4 (with an older PSP) to U6, when installing the PSP the machine kernel panics with the error shown at the end of the message

Any ideas on this?

Dec 1 21:15:01 machinename kernel: kernel BUG at arch/i386/mm/pageattr.c:146!
Dec 1 21:15:01 machinename kernel: invalid operand: 0000 [#1]
Dec 1 21:15:01 machinename kernel: SMP
Dec 1 21:15:01 machinename kernel: Modules linked in: mptctl mptbase 8021q sg cpqci(U) ipmi_si(U) ipmi_devintf(U) ipmi_msghandler(U) nfsd exportfs lockd nfs_acl md5 ipv6 autofs4 i2c_dev i2c_core sunrpc dm_mirror dm_multipath joydev button battery ac ehci_hcd uhci_hcd hw_random tg3 bnx2 bonding(U) ext3 jbd dm_mod qla2400 qla2322 qla2xxx scsi_transport_fc cciss sd_mod scsi_mod
Dec 1 21:15:01 machinename kernel: CPU: 1
Dec 1 21:15:01 machinename kernel: EIP: 0060:[<0211bd7c>] Tainted: P VLI
Dec 1 21:15:01 machinename kernel: EFLAGS: 00010002 (2.6.9-67.ELhugemem)
Dec 1 21:15:01 machinename kernel: EIP is at __change_page_attr+0x2a5/0x38e
Dec 1 21:15:01 machinename kernel: eax: 00000080 ebx: 04ebf800 ecx: 00000080 edx: 00000000
Dec 1 21:15:01 machinename kernel: esi: 00000063 edi: 80000000 ebp: 030000e0 esp: ccd9bea4
Dec 1 21:15:01 machinename kernel: ds: 007b es: 007b ss: 0068
Dec 1 21:15:01 machinename kernel: Process rmmod (pid: 20337, threadinfo=ccd9b000 task=d1bab430)
Dec 1 21:15:01 machinename kernel: Stack: f5e000e3 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Dec 1 21:15:01 machinename kernel: 00000000 00000000 03000000 f7fc0000 02007df8 00000063 80000000 04ebf800
Dec 1 21:15:01 machinename kernel: 00000007 00000007 00000001 cc8e0000 04ebf800 00000000 00000000 00000001
Dec 1 21:15:01 machinename kernel: Call Trace:
Dec 1 21:15:01 machinename kernel: [<0211be92>] change_page_attr+0x2d/0x50
Dec 1 21:15:01 machinename kernel: [<0211b8d0>] iounmap+0x71/0x85
Dec 1 21:15:01 machinename kernel: [] CpqCiDrvOsFree+0x3c/0xd3 [cpqci]
Dec 1 21:15:01 machinename kernel: [] CpqCiDrvDeleteAdapter+0x36/0x59 [cpqci]
Dec 1 21:15:01 machinename kernel: [] cleanup_module_helper+0x82/0x98 [cpqci]
Dec 1 21:15:01 machinename kernel: [] cleanup_module+0x18/0x1b [cpqci]
Dec 1 21:15:01 machinename kernel: [<02135a8e>] sys_delete_module+0x13b/0x184
Dec 1 21:15:01 machinename kernel: [<0215004a>] unmap_vma_list+0xe/0x17
Dec 1 21:15:01 machinename kernel: [<021503a7>] do_munmap+0x135/0x143
Dec 1 21:15:01 machinename kernel: [<0211ae35>] do_page_fault+0x0/0x5f7
Dec 1 21:15:01 machinename kernel: Code: 24 30 89 c3 89 f0 09 da 89 54 24 44 09 c8 8b 4c 24 44 89 44 24 40 8b 5c 24 40 8b 07 8b 57 04 f0 0f c7 0f 75 f5 f0 ff 4d 04 eb 08 <0f> 0b 92 00 09 1c 2e 02 a1 0c 14 32 02 a8 08 0f 84 ca 00 00 00
Dec 1 21:15:01 machinename kernel: <0>Fatal exception: panic in 5 seconds

[Ivan Kuznetsov]

Hello

Dec 1 21:15:01 machinename kernel: Process rmmod (pid: 20337, threadinfo=ccd9b000 task=d1bab430)

Your server panics while a module remove. Some modules can't be removed after loaded.

It's preferred to uninstall previus version of PSP, reboot and install a new version

WBR, Ivan Kuznetsov

[Roger Heflin]

I understand some modules cannot be rmmod'ed, WHY is the psp trying to remove modules, it should be putting in new modules to the kernel modules directories, and building a new initrd and then the changes take effect after the next reboot, not trying to rmmod modules that cannot be removed, rmmoding the modules is a dangerous and totally unnecessary step. Is there a way to make it not rmmod things that cannot be removed?

[Ivan Kuznetsov]

What model of server and old version of PSP do you have?

I can't find 'cpqci' module (and hprsm-*.rpm package too) in current PSP 8.15. I do not have PSP 8.11 in my library so I can't see is it there.

I suppose that functionality of cpqci is included in hp-ilo module. Scripts in hp-ilo-*.rpm tries to 'rmmod hp-ilo' while upgrading or uninstalling the RPM. Probably the similar scripts are in hprsm-*.rpm. If rmmod'ing of cpqci triggers kernel panic you can try to uninstall the RPM with '--noscrips' option

[Roger Heflin]

I am not for sure what the old version was, but alot of the HP rpm names have 7.7 in it (the machine was 4u4), so I am guessing that 7.7 was the old version.

8.11 is what we are using on other production nodes, so that choice was to install what was already being used in production.

The machine is a:
ProLiant BL460c G1

I am not sure where cpqci came from it is not located in /lib/modules anywere that I can find. It looks like the hprsm startup script puts it in.

I will see if the not running the scripts on uninstall works.

Is there are suggested order to do a OS and/or PSP upgrade in such a way that works?

Ie:
RHAS4u4 -> upgrade psp to 8.11
then u4 -> u6

What I did was:
RHAS4u4 -> u6
then PSP 7.7 to 8.11.

We potentially have a number of machines that will need this done, so an idea of what would have the highest change of working is useful. And ongoing we will be doing this with other combinations of OS/PSP in the future.

Additional though, do you know if there is a chance that we could put these rpms on a yum server and have them taken care of on the one yum update step? That way we could "package" a given PSP with a given OS update that we believe to work, and not be in the situation of more PSP/OS update combinations potentially causing issues.