kernel freeze on HP Pavilion dv9000 Turion64x2

[Marcus]

Binary package hint: linux-image-2.6.20-15-generic

Hi,

Ubuntu 7.04 freezes on my new HP Pavilion. I think this is kernel related. I use linux-image-2.6.20-15-generic for i386 (I can't use the 64bit version, but that's another problem I will report later).

The freeze only happens when I am in console or splash screen mode. When I'm using X I have no problems and can use the computer without problems. But during start up or shutdown or when I am running programs in the console(not xterm but /dev/tty1) the computer freezes. I also see that the computer freezes more easily when I run program that produces a lot of output to stdout, such as "dmesg". After the freeze the cpu fan starts to go a little bit faster, but not full speed, so the kernel is probably busy with something.

I tried to do some research. So, I striped down my initrd to only contain modules needed to find the hard drive then I started ubuntu with the kernel arg "init=/bin/bash". This should bring down the computer to simplest mode. Now I can run dmesg a lot of times without any crash. Next thing I did was to load the ehci_hcd module which detects my usb system. I tried to type some letters on the usb keyboard. usbhid is not loaded yet so the letters did not show up on the console, but when I run dmesg again (from the built in keyboard) the kernel freezes!!. So I was thinking this is usb related. I added all usb modules to the blacklist and rebooted my computer normally. This didn't help, the computer still freezes when I switch to consolemode and run dmesg.

[Marcus (marcus-liljedahl)]

The problem seems to disappear if the kernel option noapic is added. Now an other problem shows up instead. After a while the kernel says the following:

[ 2358.262653] irq 7: nobody cared (try booting with the "irqpoll" option)
[ 2358.262659]
[ 2358.262660] Call Trace:
[ 2358.262663] <IRQ> [<ffffffff802bd285>] __report_bad_irq+0x35/0x90
[ 2358.262698] [<ffffffff802bd500>] note_interrupt+0x220/0x280
[ 2358.262720] [<ffffffff8809e465>] :usbcore:usb_hcd_irq+0x25/0x60
[ 2358.262730] [<ffffffff802be243>] handle_level_irq+0xe3/0x140
[ 2358.262734] [<ffffffff8026223c>] call_softirq+0x1c/0x28
[ 2358.262742] [<ffffffff80270189>] do_IRQ+0x89/0x100
[ 2358.262746] [<ffffffff8026e860>] default_idle+0x0/0x50
[ 2358.262751] [<ffffffff80261631>] ret_from_intr+0x0/0xa
[ 2358.262754] <EOI> [<ffffffff802316b0>] unix_poll+0x0/0xa0
[ 2358.262769] [<ffffffff8026e889>] default_idle+0x29/0x50
[ 2358.262775] [<ffffffff8024b14b>] cpu_idle+0x9b/0xd0
[ 2358.262782] [<ffffffff8027a435>] start_secondary+0x4d5/0x4f0
[ 2358.262811]
[ 2358.262812] handlers:
[ 2358.262815] [<ffffffff8809e440>] (usb_hcd_irq+0x0/0x60 [usbcore])
[ 2358.262827] Disabling IRQ #7

The usb mouse and keyboard work really bad after this kernel message.

The mouse and keyboard start to work normally again if the ehci_hcd module is removed. I guess that the ochi_hcd module takes over all usb functionality in that case. I can modprobe ehci_hcd module again and the system works normal.

[Marcus (marcus-liljedahl)]

I continue to describe this bug...

I said in the first message that the problems showed up as quick as I load the ehci_hcd and ochi_hcd modules, but other modules seems to freeze the computer too. I think I will try to focus on ehci_hcd in this bug/thread because many other problems are related to that module.

I forgot to mention that with the noapic option the 64bit version av ubuntu works as good as the 32bit. I also forgot to mention that when the ehci_hcd module is removed and loaded again, as I described in the last message the problem can happen any time again, unless I remove the module and never load it again.

Next thing I did was to add the irqpoll option to the kernel. And yes!, I get rid of the error message above, but the ehci_hcd module has another problem. It looks like it takes 100% from one of the cpu cores. I didn't realize this in the beginning because the system works really fine even if one of the cores are busy. But when I remove the module the fan goes down to zero speed and the system responds much faster. The ehci_hcd module does only take 100% cpu if the noapic option is added, whitout that option the cpu isn't loaded, but on the other hand the system freezes quite quick.

So, I can get rid of the error message either by either adding the irqpoll option, or removing the ehci_hcd module, and the last case is the best because the module takes 100% of one cpu cores.

[lemonman (jd-schroeder)]

Confirmed on my dv9317cl as well.

I also tried booting straight into bash, and managed to freeze the system. You have to get multiple processes going, I did something like this:

while true; do (dmesg &); done

which really flooded the console. I suspect this is something with SMP deadlocking, since if you have a single process going it doesn't seem to ever freeze.

[Marcus (marcus-liljedahl)]

Im not sure it is related to SMP, at least not on my computer. Because running "while true; do (dmesg &); done >outfile" does not result in freeze. But running "cat outfile" (same file) results in freeze. I have also trie maxcpus=1 as kernel options, but this does not solve the problem.

[lemonman (jd-schroeder)]

I've switched to a framebuffer console, and my system is now stable.

[Marcus (marcus-liljedahl)]

Frame buffer console doesn't help for me. Which frame buffer driver did you use?

[lemonman (jd-schroeder)]

It's either vesa or vga, I compiled them both in. I use mode 0x300 (640x480).

Also make sure you're using the latest kernel (I find 2.6.22 or later works best); earlier ones had other IRQ problems that would freeze up regardless of the console.

[Marcus (marcus-liljedahl)]

Are you using 32bit or 64bit ubuntu?

[Keyper7 (keyper7)]

I can confirm this problem on Feisty and also that using a framebuffer console solved the problem.

[Keyper7 (keyper7)]

Sorry, I forgot: I had the same problem but on a dv6258se

[Launchpad Janitor (janitor)]

Beginning with the Hardy Heron 8.04 development cycle, all open Ubuntu kernel bugs need to be reported against the "linux" kernel package. We are automatically migrating this bug to the new "linux" package. However, development has already began for the upcoming Intrepid Ibex 8.10 release. It would be helpful if you could test the upcoming release and verify if this is still an issue - http://www.ubuntu.com/testing . If the issue still exists, please update this report by changing the Status of the "linux" task from "Incomplete" to "New". We appreciate your patience and understanding as we make this transition. Thanks!