Kernel oops at ahci_deinit_port

[Christophe Dumez]

I upgraded My DELL T3500 desktop from Lucid to Maverick today. Unfortunately, I get a kernel oops on startup with the new Kernel:

[ 18.357380] Stack:
[ 18.357436] 00000001 fbadad12 f6f4dd78 f8427d6d f844e000 f6f4dd78 f6df630c 0000001e
[ 18.357768] <0> f844e000 f6f4ddc0 f842856b f842a2ef f8444d0c f75c2920 f842a205 fffffffb
[ 18.358189] <0> fffffffb 0000001e f842a205 f75c2920 f844e000 f75f4060 f844e008 00000000
[ 18.358663] Call Trace:
[ 18.358722] [<f8427d6d>] ? ahci_deinit_port+0x1d/0x90 [libahci]
[ 18.358787] [<f842856b>] ? ahci_init_controller+0x5b/0x110 [libahci]
[ 18.358854] [<f844111e>] ? ahci_pci_init_controller+0x3e/0x50 [ahci]
[ 18.358921] [<f8441cb5>] ? ahci_init_one+0x3f5/0x7d0 [ahci]
[ 18.358988] [<c05ee769>] ? mutex_unlock+0x19/0x20
[ 18.359052] [<c027ac77>] ? sysfs_new_dirent+0x67/0x100
[ 18.359115] [<c027b07a>] ? sysfs_addrm_finish+0x1a/0xb0
[ 18.359181] [<c0375fe3>] ? local_pci_probe+0x13/0x20
[ 18.359244] [<c0376f98>] ? pci_device_probe+0x68/0x90
[ 18.359308] [<c0414d00>] ? really_probe+0x50/0x150
[ 18.359372] [<c041c387>] ? pm_runtime_barrier+0x57/0xb0
[ 18.359435] [<c0414e3c>] ? driver_probe_device+0x3c/0x60
[ 18.359498] [<c0414ee1>] ? __driver_attach+0x81/0x90
[ 18.359562] [<c0414303>] ? bus_for_each_dev+0x53/0x80
[ 18.359625] [<c0414bce>] ? driver_attach+0x1e/0x20
[ 18.359687] [<c0414e60>] ? __driver_attach+0x0/0x90
[ 18.359750] [<c0414595>] ? bus_add_driver+0xd5/0x280
[ 18.359814] [<c0376ed0>] ? pci_device_remove+0x0/0x40
[ 18.359877] [<c04151da>] ? driver_register+0x6a/0x130
[ 18.359940] [<c03771d5>] ? __pci_register_driver+0x45/0xb0
[ 18.360005] [<f8448017>] ? ahci_init+0x17/0x19 [ahci]
[ 18.360068] [<c0103042>] ? do_one_initcall+0x32/0x1a0
[ 18.360132] [<f8448000>] ? ahci_init+0x0/0x19 [ahci]
[ 18.360197] [<c0188f1b>] ? sys_init_module+0x9b/0x1e0
[ 18.360261] [<c0222ec2>] ? sys_write+0x42/0x70
[ 18.360325] [<c010939f>] ? sysenter_do_call+0x12/0x28

Kernel version is: linux-image-2.6.35-22-generic-pae

Also note that I get the following messages just before the crash:
[ 3.267124] ahci 0000:00:1f.2: controller reset failed (0xffffffff)
[ 3.770018] ahci 0000:00:1f.2: failed to stop engine (-5)
[ 4.273087] ahci 0000:00:1f.2: failed to stop engine (-5)
[ 4.776214] ahci 0000:00:1f.2: failed to stop engine (-5)
...

[Christophe Dumez (hydr0g3n)]

I have tried the mainline kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.36-rc7-maverick/
and I get the exact same issue.

[lowey71 (milo-loweys)]

Had exactly the same problem after upgrading with similar Dell T3500 system. 2.6.32-25-generic working fine though.

[Christophe Dumez (hydr0g3n)]

I confirm that v2.6.32-25 is working.
I tested v2.6.36 final (vanilla) and I get the same crash.

[Christophe Dumez (hydr0g3n)]

After more testing, I can say:
- 2.6.34.7 vanilla: NOT WORKING
- 2.6.35-23 from maverick-proposed: NOT WORKING
- 2.6.36 (final) vanilla: NOT WORKING

[trbs (trbs)]

We have the same problem on our developer workstations (Dell T-3500) running Ubuntu Maverick AMD64.

Tried the following kernels:

- 2.6.32-25: WORKING
- 2.6.33-6: WORKING
- 2.6.33-7: WORKING
- 2.6.34-rc1: NOT WORKING
- 2.6.34-1: NOT WORKING

So the problem seems to be introduced with the 2.6.34 kernel.

Our machines report exactly the same error message as given above.

[at (arjan-tijms)]

I just upgraded from Ubuntu 64 bits 10.04 to 10.10 and I too have this problem. I have a Dell T3500 with a W3520, 6GB memory, 2 Quadro FX 570 cards and a X25-M SSD.

When booting Ubuntu 10.10 I get the following message after a while:

ALERT! /dev/disk/by-uuid/[some UUID] does not exist. Dropping to a shell!

When I enable verbose messages in grub.cfg I also see the "failed to stop engine (-5)" messages. Switching back to the 2.6.32-25 kernel lets me boot up again, but needless to say I would prefer the 2.6.35 kernel.

[he536 (henk53602)]

This seems to be a duplicate of:

 - https://bugs.launchpad.net/ubuntu/+bug/653238
 - https://bugs.launchpad.net/ubuntu/+bug/658560
 - https://bugs.launchpad.net/ubuntu/+bug/659149

There was already a bugzilla entry on kernel.org:

 - https://bugzilla.kernel.org/show_bug.cgi?id=16228

Other distros had been affected as well, most of them in June already:

 - https://bugzilla.redhat.com/show_bug.cgi?id=620313
 - http://fedoraproject.org/wiki/KernelCommonProblems (at the bottom of the page)
 - http://forums.gentoo.org/viewtopic-p-6455079.html
 - https://bbs.archlinux.org/viewtopic.php?pid=779136

Inserting pci=nocrs in the kernel line in grub seems to solve it. Since this bug is so well known, I wonder why Ubuntu didn't include it by default for the Dell T3500.

[Bjorn Helgaas (bjorn-helgaas)]

This is a duplicate of bug #653238, not of bug #647043.

Bug #653238 was marked a duplicate of bug #647043, but that is incorrect.