Bug 86478

Environment:
--------------------------
Platform: Haswell
Kernel: (drm-intel-next-queued) latest
77c1aa84de0096792de673aa1c64c36b38553cf5

dmesg log is same as bug Bug 86233:
https://bugs.freedesktop.org/attachment.cgi?id=109444

Bug detailed description:
-----------------------------
When boot up ,and then plug in VGA ,error messages show as follows:

*ERROR* uncleared pch fifo underrun on pch transcoder A
*ERROR* PCH transcoder A FIFO underrun
*ERROR* uncleared fifo underrun on pipe B
*ERROR* CPU pipe B FIFO underrun

Steps:
---------------------------
Boot up
plug in VGA

Bisect Result
---------------------------
The _culprit_ commit introduced the following two error lines.
*ERROR* uncleared pch fifo underrun on pch transcoder A
*ERROR* uncleared fifo underrun on pipe B

29c6b0c5ee33d54ada586040eca2a258f0ce0650 is the first bad commit
commit 29c6b0c5ee33d54ada586040eca2a258f0ce0650
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date:   Fri May 16 19:40:24 2014 +0300

    drm/i915: Simplify the uncleared FIFO underrun detection

    Checking whether the error interrupt was enabled or not isn't really
    necessary when we check for uncleared FIFO underruns. If it was enabled
    we'll race with the interrupt handler a bit, but that seems OK as we
    still claim the interrupt.

    Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
    Reviewed-by: Thomas Wood <thomas.wood@intel.com>
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

The additional log output is just a bit more paranoid reporting if you have multiple underruns. Which is usually the case. In other words the patch you've bisected to works as planned (and is actually a bugfix for this debug code, which the commit message fails to state).

So there's just the issue with fifo underruns when you plug in VGA.

Please test this patch:

http://patchwork.freedesktop.org/patch/37391/

hi,Daniel.I tried with this patch,but it seems doesn't work.Error message still exits.

(In reply to Li Xu from comment #4)
> hi,Daniel.I tried with this patch,but it seems doesn't work.Error message
> still exits.

Sorry forgot to mention that this will only address "uncleared pch fifo underrun on pch transcoder A", not the non-pch underrun. Can you please recheck whether that message is gone?

The non-pch underrun is likely a different issue.

The error "uncleared pch fifo underrun on pch transcoder A" still exits.

commit b68362278af94e1171f5be9d4e44988601fb0439
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Mon Nov 24 17:02:45 2014 +0100

    drm/i915: More cautious with pch fifo underruns

pushed to drm-intel-fixes, please retest.

The two error messages still exits.

Attach VGA, following cases also have this error:
igt/kms_setmode/clone-exclusive-crtc
igt/kms_3d

(In reply to lu hua from comment #9)
> Attach VGA, following cases also have this error:
> igt/kms_setmode/clone-exclusive-crtc
> igt/kms_3d

Those 2 cases also have this error on SNB.

*** Bug 84951 has been marked as a duplicate of this bug. ***

*** Bug 87990 has been marked as a duplicate of this bug. ***

Reported twice on HSW but I also see on BDW when alternating to fbcon and back to X. I believe there is an error in the enable/disable enable/old logic somewhere but I couldn't find. If this is the case it should affect many other platforms not just BDW and HSW.

*** Bug 85621 has been marked as a duplicate of this bug. ***

Sure about #85621 being a dupe of this? AFAIK that is about missing a bandwidth check for modes exceeding cdclk (not a regression), and this does not look related.

Created attachment 112692 [details]
20150122_fifo.log

I've noticed the error message "*ERROR* uncleared fifo underrun on pipe A" since kernel 3.17, not sure what the best way to reproduce, sometimes it will appear when changing to tty2 and others when the system is stressed (e.g. high GPU usage by Chromium).

See attached log, noticed that I only needed to change to tty2 for the error message to appear.

Worth mentioning that I also see regression of xf86-video-intel 2.99.917, running X on vt1, changing to tty2 and back will make the desktop (Gnome) unresponsive. Chris Wilson tested this in Ubuntu and didn't saw any problem so this might be Arch packaging specific, I'm waiting for someone else to confirm it at https://bugs.archlinux.org/task/43534 .

system: Acer C720, Haswell GT1, 2995U
kernel: 3.19rc5
xf86-video-intel: 2.99.916
mesa: 10.4.2
xorg-server: 1.16.3
distro: Arch Linux x86_64

Maybe this will help: when compiling xf86-video-intel-2.99.916 with flag --enable-debug=full and attached the journald output and Xorg.0.log.
Notice that adding the full debug flag seems to make a change that cause X to crash when changing to vt2 (but fifo underrun error still appears).

Created attachment 112711 [details]
20150123-01-xorg_fast_crash-journal.log

Created attachment 112712 [details]
20150123-01-xorg_fast_crash-Xorg.0.log.log.tar.xz

BDW doesn't have vga or a pch transcoder ...

Ok there's too many bugs in here all at once.

Note to everyone: fifo underruns are not all made equal. This one here was about the _PCH_ fifo underrun on _HSW_ (which is only possible with VGA).

There's a lot more bugs around here which are different, so I'll close this bug as too confusing.

Please retest with latest drm-intel-nightly, and if you still have underruns open a _new_ bug report.

Thanks, Daniel

On HSW with the latest drm-intel-nightly kernel. plug and unplug VGA and test kms_3d, I don't meet this error. Close it.

I'm having this with an Intel NUC DC3217IYE on ArchLinux, kernel 3.19.3

journalctl output:
Process 282 (sddm-greeter) of user 996 dumped core
[drm:intel_set_pch_fifo_underrun_reporting [i915]] *ERROR* intel uncleared pch fifo underrun on pch [drm:intel_pch_fifo_underrun_irq_handler [i915]] *ERROR* PCH transcoder a FIFO underrun

Bráulio, as stated in comment #21, please open a new bug report.

Closing verified+fixed.