Comment 19 for bug 1454160

Hi crs,
I found the tags in git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git . Which will help you to bisect.
I had a look at the commits between 3.18.3 and 4. The two very last commits before 3.18.4 are my primary suspects:

commit 6b49734139bad7703ccdf0bded005c605f28ca61
Author: Greg Kroah-Hartman <email address hidden>
Date: Mon Mar 18 13:16:44 2013 -0700

    Revert "drm/i915: reorder setup sequence to have irqs for output setup"

    Revert commit 2a9810441fcc26cf3f006f015f8a62094fe57a90 which is
    commit 52d7ecedac3f96fb562cb482c139015372728638 upstream.

    This caused problems in 3.8-stable, but all is fine in 3.9-rc.

    Signed-off-by: Greg Kroah-Hartman <email address hidden>
    Cc: Imre Deak <email address hidden>
    Cc: Daniel Vetter <email address hidden>

commit 2909d2604809e056e5fe3ee102c42cc7899b1739
Author: Greg Kroah-Hartman <email address hidden>
Date: Mon Mar 18 13:13:39 2013 -0700

    Revert "drm/i915: enable irqs earlier when resuming"

    This reverts commit 31f14f4219d2a74b7a6d86c7798f49141b5eccbe which was
    commit 15239099d7a7a9ecdc1ccb5b187ae4cda5488ff9 upstream.

    It caused problems in the 3.8-stable series, but 3.9-rc is just fine.

    Signed-off-by: Greg Kroah-Hartman <email address hidden>
    Cc: Chris Wilson <email address hidden>
    Cc: Mika Kuoppala <email address hidden>
    Cc: Ilya Tumaykin <email address hidden>
    Cc: Daniel Vetter <email address hidden>

Not only are these the only commits mentioning i915, in addition, the diffs seem to remove some workarounds that set a enable_hotplug_processing; to false...

So, bisecting could probably be limited to trying the commit before 2909d2604809e056e5fe3ee102c42cc7899b1739, that commit, and the following commit.

I think it should be rather obvious what is going wrong for someone who is at the slightest familiar with this code.