Bug 5564
Summary: | Kernel Panic on PCI-Probe - 450NX - HP Netserver LH4r | ||
---|---|---|---|
Product: | Drivers | Reporter: | Frank Overton (frank) |
Component: | PCI | Assignee: | Bartlomiej Zolnierkiewicz (bzolnier) |
Status: | CLOSED PATCH_ALREADY_AVAILABLE | ||
Severity: | high | CC: | acpi-bugzilla, akpm, bzolnier, greg |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.9-22 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: |
debug patch
2.6.14.1 debug screenshot 2.6.9-22 (CentOS 4.2) Max resolution Screenshot |
Description
Frank Overton
2005-11-07 13:05:43 UTC
This is a bug in your bios. Please contact HP to get an update for it. Created attachment 6492 [details]
debug patch
I somehow suspect it's a kernel bug. pci_fixup_i450nx calls pci_scan_bus, which
appears adding new devices too early to me (root bridge scanning will add all
new devices finally). If it works for you, we might also fix pci_fixup_i450gx.
I would mention the patch is against 2.6.14. It can't work for previous kernel version (pci_scan_bus_parented changed recently) Comment on attachment 6492 [details]
debug patch
Relooked at it. Sorry, my guess is wrong.
I am not a coder and my C is quite rusty but might the following be an issue? The following patch appears about the same time as the bug: ===== piix.c 1.20 vs edited ===== --- 1.20/drivers/ide/pci/piix.c Wed Oct 22 09:27:24 2003 +++ edited/piix.c Fri Nov 21 09:28:43 2003 @@ -768,8 +768,8 @@ /* Only on the original revision: IDE DMA can hang */ if(rev == 0x00) no_piix_dma = 1; - /* On all revisions PXB bus lock must be disabled for IDE */ - else if(cfg & (1<<14)) + /* On all revisions below 5 PXB bus lock must be disabled for IDE */ + else if(cfg & (1<<14) && rev < 5) no_piix_dma = 2; } if(no_piix_dma) Could it make a difference that the last line is not... + else if(cfg & (1<<14) && rev < 0x05) Recreated error on kernel 2.6.14.1 with DEBUG defined in arch/i386/pci/pci.h (BTW: Changing to 5 to 0x05 in piix.c makes no difference.) This is the OOPS message: Starting balanced_irq Using IPI Shortcut mode md: Autodetecting RAID arrays md: autorun ... md: ... autorun DONE. RAMDISK: Compressed image found at block 0 logips2pp: Detected unknown logitech mouse model 1 input: PS/2 Logitech Mouse on isa0060/serio1 VFS: Mounted root (ext2 filesystem) RED Hat nash version 3.5.13.6 starting Mounted /proc filesystem Loading scsi_mod.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading megaraid.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading sym53c8xx.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading ext3.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Creating block devices VFS: CAnnot open root device "LABEL=/" or unknown-block(0,0) Please append a correct "root=" boot option Kernel Panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) The output at the URL in the initial report shows the same PCI stack trace with "acpi=off". punting this report to PCI sub-system. It would help if we could get the whole oops trace. Can you get a 50-row screen? I can't remember how to do that with grub, but in lilo you add `vga=extended' to the config file. That'll get us the whole oops. A digital photo of the screen will then save a lot of pain. We've booted RHEL4 (2.6.9-22) on an Intel AD450NX, so this issue is not as simple as "2.6 vs 450NX". It may be specific to the HP Netserver LH4r. Please verify that the system is running an up-to-date BIOS. Odds are RHEL has a patch that we do not know about :( Created attachment 6586 [details]
2.6.14.1 debug screenshot
Screen shot at max resolution of 2.6.14.1 debug
Created attachment 6587 [details]
2.6.9-22 (CentOS 4.2) Max resolution Screenshot
I have verified with HP that I am running the latest BIOS files available. bugme-daemon@bugzilla.kernel.org wrote: > > http://bugzilla.kernel.org/show_bug.cgi?id=5564 > Ah, the dreaded "PCI: device disabled (BIOS)" thing. Is there a fix for that? Is it a simple matter of enabling the thing in BIOS? If so, why? Can't we just ignore what BIOS said and go ahead and do it? Is this new behaviour? Behaviour is new to 2.6 kernels. All 2.4 kernels work fine. Hmmm... The LH4 has two internal RAID controllers. I have disabled one of them in the BIOS and have RAID1 mirror drives configured in the other. I'll can enable both controllers but I don't expect it to make a difference. Enabling both RAID controllers doesn't change a thing. On 11/14/05, Andrew Morton <akpm@osdl.org> wrote: > bugme-daemon@bugzilla.kernel.org wrote: > > > > http://bugzilla.kernel.org/show_bug.cgi?id=5564 > > > > Ah, the dreaded "PCI: device disabled (BIOS)" thing. > > Is there a fix for that? Is it a simple matter of enabling the thing in BIOS? For IDE enabling it in BIOS should be OK. > If so, why? Can't we just ignore what BIOS said and go ahead and do it? In general - yes, the only side-effect would be different ordering of IDE devices. However it doesn't look like IDE problem et all, Andrew please look at this: " This is the OOPS message: Starting balanced_irq Using IPI Shortcut mode md: Autodetecting RAID arrays md: autorun ... md: ... autorun DONE. RAMDISK: Compressed image found at block 0 logips2pp: Detected unknown logitech mouse model 1 input: PS/2 Logitech Mouse on isa0060/serio1 VFS: Mounted root (ext2 filesystem) RED Hat nash version 3.5.13.6 starting Mounted /proc filesystem Loading scsi_mod.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading megaraid.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading sym53c8xx.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading ext3.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Creating block devices VFS: CAnnot open root device "LABEL=/" or unknown-block(0,0) Please append a correct "root=" boot option Kernel Panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) " Frank, I don't see any OOPS here. IMO it is PCI problem in RH 2.6.9-22 kernel which is fixed in 2.6.14.1, plus RHEL4.2 (CentOS4.2) comes with one year old module-init-tools package. You should try upgrading module-init-tools to the latest version, regenerate your initrd and try again with 2.6.14.1. BTW Documentation/Changes in -git tree lists module-init-tools-0.9.10 as a required version which seems a bit unbelievable (they were released on 25-Feb-2003). On 11/14/05, Andrew Morton <akpm@osdl.org> wrote: > bugme-daemon@bugzilla.kernel.org wrote: > > > > http://bugzilla.kernel.org/show_bug.cgi?id=5564 > > > > Ah, the dreaded "PCI: device disabled (BIOS)" thing. > > Is there a fix for that? Is it a simple matter of enabling the thing in BIOS? For IDE enabling it in BIOS should be OK. > If so, why? Can't we just ignore what BIOS said and go ahead and do it? In general - yes, the only side-effect would be different ordering of IDE devices. However it doesn't look like IDE problem et all, Andrew please look at this: " This is the OOPS message: Starting balanced_irq Using IPI Shortcut mode md: Autodetecting RAID arrays md: autorun ... md: ... autorun DONE. RAMDISK: Compressed image found at block 0 logips2pp: Detected unknown logitech mouse model 1 input: PS/2 Logitech Mouse on isa0060/serio1 VFS: Mounted root (ext2 filesystem) RED Hat nash version 3.5.13.6 starting Mounted /proc filesystem Loading scsi_mod.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading megaraid.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading sym53c8xx.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Loading ext3.ko module insmod: QM_MODULES: Error: /bin/insmod exited abnormally! Creating block devices VFS: CAnnot open root device "LABEL=/" or unknown-block(0,0) Please append a correct "root=" boot option Kernel Panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) " Frank, I don't see any OOPS here. IMO it is PCI problem in RH 2.6.9-22 kernel which is fixed in 2.6.14.1, plus RHEL4.2 (CentOS4.2) comes with one year old module-init-tools package. You should try upgrading module-init-tools to the latest version, regenerate your initrd and try again with 2.6.14.1. BTW Documentation/Changes in -git tree lists module-init-tools-0.9.10 as a required version which seems a bit unbelievable (they were released on 25-Feb-2003). Upgrading module-init-tools let 2.6.14.1 boot! Thanks!!! Now being a noob, I'll ask a noob question... Can I run CentOS 4.2 on top of the newer kernel or is too much of the release dependent on a specific kernel version? Thanks again! Generally you should stay with distribution provided kernel, but if your user-space matches newer kernel versions (see Documentation/Changes) you can give it a try. However you are probably better off reporting the problem (+ pointing to this bugzilla entry) to CentOS developers. PS I would like to close this bug as it is fixed in the mainstream kernel. |