Comment 19 for bug 904569

I tested the same kernel and newer ones and had the same problem on one of my systems. Did this bisect:

git bisect start
# good: [ac6766564c0305ca020fe747dfd7dbdf0881369d] Linux 3.0.12
git bisect good ac6766564c0305ca020fe747dfd7dbdf0881369d
# bad: [d986a8dbfd7358bfbda116650c4caf8a3b90d865] Linux 3.0.13
git bisect bad d986a8dbfd7358bfbda116650c4caf8a3b90d865
# good: [08d618b2080d8b3afac6db1a361c54d827b8d044] drm/radeon/kms: add some new pci ids
git bisect good 08d618b2080d8b3afac6db1a361c54d827b8d044
# good: [c060a3d5e9bba4271331b69b9e2c53105999b97f] x86: Fix "Acer Aspire 1" reboot hang
git bisect good c060a3d5e9bba4271331b69b9e2c53105999b97f
# good: [edb9a31845c5ba0ff325daa58f17f881d60d1559] xfs: force buffer writeback before blocking on the ilock in inode reclaim
git bisect good edb9a31845c5ba0ff325daa58f17f881d60d1559
# good: [d80dee54533aa4bfe29def921edb31715fdba214] tick-broadcast: Stop active broadcast device when replacing it
git bisect good d80dee54533aa4bfe29def921edb31715fdba214
# good: [0bbf5c70251286fbc3b7aac5e7961b4568115bfd] oprofile: Fix crash when unloading module (hr timer mode)
git bisect good 0bbf5c70251286fbc3b7aac5e7961b4568115bfd
# bad: [b01b383bbd04e9dcf7d9fe6ca3751b77ccdc533c] clockevents: Set noop handler in clockevents_exchange_device()
git bisect bad b01b383bbd04e9dcf7d9fe6ca3751b77ccdc533c
# good: [4078977c46f627f553ed2d8ea047b9bf25dee48d] clocksource: Fix bug with max_deferment margin calculation
git bisect good 4078977c46f627f553ed2d8ea047b9bf25dee48d

Result:

b01b383bbd04e9dcf7d9fe6ca3751b77ccdc533c is the first bad commit
commit b01b383bbd04e9dcf7d9fe6ca3751b77ccdc533c
Author: Thomas Gleixner <email address hidden>
Date: Fri Dec 2 16:02:45 2011 +0100

    clockevents: Set noop handler in clockevents_exchange_device()

    commit de28f25e8244c7353abed8de0c7792f5f883588c upstream.

    If a device is shutdown, then there might be a pending interrupt,
    which will be processed after we reenable interrupts, which causes the
    original handler to be run. If the old handler is the (broadcast)
    periodic handler the shutdown state might hang the kernel completely.

    Signed-off-by: Thomas Gleixner <email address hidden>
    Signed-off-by: Greg Kroah-Hartman <email address hidden>

:040000 040000 b5dba6238e4accc62febbb9bd67c89c27eeb077b 67fa1fc6a94b92a1fa3cc14018e447a0d387451a M kernel

A revert fixed the issue, also tested with 3.0.14.