[2.6.39 regression] i915/kms: garbled screen because of 49183b281 (was: Re: Linux 2.6.39-rc7)

[2.6.39 regression] i915/kms: garbled screen because of 49183b281 (was: Re: Linux 2.6.39-rc7)

[Melchior FRANZ]

* Linus Torvalds -- Tuesday 10 May 2011:
> But please do test, just to make sure that 39-final is good.

> Chris Wilson (4):
>       drm/i915: Only enable the plane after setting the fb base (pre-ILK)

This patch introduces a bug on my infamous "Acer Travelmate
5735Z-452G32Mnss": when KMS takes over, the frame buffer contents
get completely garbled up on screen, with colored stripes and
unreadable text (photo on request). Only when X11 is started, the
screen gets restored again. Closing and re-opening the lid partly
cures the mess, too: it makes the font readable, though horizontally
stretched.

Reverting 49183b2818de6899383bb82bc032f9344d6791ff fixes the
bug.

m.





drm.debug=0x02

[    2.604831] [drm] Initialized drm 1.1.0 20060810
[    2.648409] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[    2.648415] i915 0000:00:02.0: setting latency timer to 64
[    2.708949] [drm:intel_opregion_setup], graphic opregion physical addr: 0x7ba8c018
[    2.708987] [drm:intel_opregion_setup], Public ACPI methods supported
[    2.708989] [drm:intel_opregion_setup], SWSCI supported
[    2.708991] [drm:intel_opregion_setup], ASLE supported
[    2.709047] i915 0000:00:02.0: irq 44 for MSI/MSI-X
[    2.709053] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[    2.709054] [drm] Driver supports precise vblank timestamp query.
[    2.735626] [drm:init_status_page], render ring hws offset: 0x00000000
[    2.735747] [drm:init_status_page], bsd ring hws offset: 0x00021000
[    2.735852] [drm:intel_parse_bios], Using VBT from OpRegion: $VBT CANTIGA        d
[    2.779279] [drm:intel_panel_get_backlight], get backlight PWM = 0
[    2.779287] vgaarb: device changed decodes: PCI:0000:00:02.0,olddecodes=io+mem,decodes=io+mem:owns=io+mem
[    2.779616] [drm:intel_panel_set_backlight], set backlight PWM = 0
[    2.779619] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[    2.779626] [drm:intel_opregion_asle_intr], non asle set request??
[    3.022059] [drm:gm45_get_vblank_counter], trying to get vblank count for disabled pipe A
[    3.022065] [drm:gm45_get_vblank_counter], trying to get vblank count for disabled pipe A
[    3.074270] checking generic (80000000 3ff0000) vs hw (80000000 10000000)
[    3.074274] fb: conflicting fb hw usage inteldrmfb vs VESA VGA - removing generic driver
[    3.074293] Console: switching to colour dummy device 80x25
[    3.074806] fbcon: inteldrmfb (fb0) is primary device
[    3.109085] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[    3.109087] [drm:intel_panel_set_backlight], set backlight PWM = 736950
[    3.109090] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[    3.109098] [drm:intel_opregion_asle_intr], non asle set request??
[    3.109111] Console: switching to colour frame buffer device 170x48
[    3.111779] fb0: inteldrmfb frame buffer device
[    3.111780] drm: registered panic notifier
[    3.158015] scsi 4:0:0:0: Direct-Access     Generic- Multi-Card       1.00 PQ: 0 ANSI: 0 CCS
[    3.413081] acpi device:07: registered as cooling_device2
[    3.413431] input: Video Bus as /devices/LNXSYSTM:00/device:00/PNP0A08:00/LNXVIDEO:01/input/input6
[    3.413532] ACPI: Video Device [OVGA] (multi-head: yes  rom: no  post: no)
[    3.413982] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0
[    4.059019] sd 4:0:0:0: [sdb] 15720448 512-byte logical blocks: (8.04 GB/7.49 GiB)
[    4.059746] sd 4:0:0:0: [sdb] Write Protect is off
[    4.059750] sd 4:0:0:0: [sdb] Mode Sense: 03 00 00 00
[    4.059752] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[    4.061879] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[    4.066269]  sdb: sdb1
[    4.067995] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[    4.068056] sd 4:0:0:0: [sdb] Attached SCSI removable disk
[    5.016884] md: linear personality registered for level -1
[   94.800469] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[   94.800473] [drm:intel_panel_set_backlight], set backlight PWM = 72250
[   94.800477] [drm:intel_panel_get_max_backlight], max backlight PWM = 736950
[   94.800482] [drm:intel_opregion_asle_intr], non asle set request??
[   94.800485] [drm:intel_opregion_asle_intr], non asle set request??

[PATCH] Revert "drm/i915: Only enable the plane after setting the fb base (pre-ILK)"

[Keith Packard]

[-- Attachment #1: Type: text/plain, Size: 2045 bytes --]


This reverts commit 49183b2818de6899383bb82bc032f9344d6791ff.

Franz Melchior:

 This patch introduces a bug on my infamous "Acer Travelmate
 5735Z-452G32Mnss": when KMS takes over, the frame buffer contents
 get completely garbled up on screen, with colored stripes and
 unreadable text (photo on request). Only when X11 is started, the
 screen gets restored again. Closing and re-opening the lid partly
 cures the mess, too: it makes the font readable, though horizontally
 stretched.

---

On Thu, 12 May 2011 18:49:08 +0200, Melchior FRANZ <melchior.franz@gmail.com> wrote:

> * Linus Torvalds -- Tuesday 10 May 2011:
> > But please do test, just to make sure that 39-final is good.
> 
> > Chris Wilson (4):
> >       drm/i915: Only enable the plane after setting the fb base (pre-ILK)
> 
> This patch introduces a bug on my infamous "Acer Travelmate
> 5735Z-452G32Mnss": when KMS takes over, the frame buffer contents
> get completely garbled up on screen, with colored stripes and
> unreadable text (photo on request). Only when X11 is started, the
> screen gets restored again. Closing and re-opening the lid partly
> cures the mess, too: it makes the font readable, though horizontally
> stretched.
> 
> Reverting 49183b2818de6899383bb82bc032f9344d6791ff fixes the
> bug.

It's Whack-a-Mole time! Fix one laptop, break another. We'll pick 'no
regressions' over 'fixes existing bug' today.

 drivers/gpu/drm/i915/intel_display.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 373c2a0..2166ee0 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -5154,6 +5154,8 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 
 	I915_WRITE(DSPCNTR(plane), dspcntr);
 	POSTING_READ(DSPCNTR(plane));
+	if (!HAS_PCH_SPLIT(dev))
+		intel_enable_plane(dev_priv, plane, pipe);
 
 	ret = intel_pipe_set_base(crtc, x, y, old_fb);
 
-- 
1.7.5.1

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [2.6.39 regression] i915/kms: garbled screen because of 49183b281 (was: Re: Linux 2.6.39-rc7)

[Chris Wilson]

On Thu, 12 May 2011 18:49:08 +0200, Melchior FRANZ <melchior.franz@gmail.com> wrote:
> * Linus Torvalds -- Tuesday 10 May 2011:
> > But please do test, just to make sure that 39-final is good.
> 
> > Chris Wilson (4):
> >       drm/i915: Only enable the plane after setting the fb base (pre-ILK)
> 
> This patch introduces a bug on my infamous "Acer Travelmate
> 5735Z-452G32Mnss": when KMS takes over, the frame buffer contents
> get completely garbled up on screen, with colored stripes and
> unreadable text (photo on request). Only when X11 is started, the
> screen gets restored again. Closing and re-opening the lid partly
> cures the mess, too: it makes the font readable, though horizontally
> stretched.

So we think that enabling the plane at this point is masking a bug in our
modeset, or that some side-effect of writing those registers or waiting
for that vblank has a vital latching or delay that we have not accounted
for.

Can you please grab intel_reg_dumper from 

  http://cgit.freedesktop.org/xorg/app/intel-gpu-tools

and attach its output after passing nomodeset on your kernel boot
parameters (without starting X). That should capture the registers as they
were left by the BIOS. Alternatively, if you load i915.ko as a module, you
can run intel_reg_dumper before loading the module. Then can you attach
the output from intel_reg_dumper from the VT before starting X (whilst the
display is skewiff) and then after (when the display has recovered). That
may help, unless we are right in that it is in the timing of the register
writes.

Thanks,
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: [2.6.39 regression] i915/kms: garbled screen because of 49183b281 (was: Re: Linux 2.6.39-rc7)

[Melchior FRANZ]

[-- Attachment #1: Type: Text/Plain, Size: 1115 bytes --]

* Chris Wilson -- Thursday 12 May 2011:
> So we think that enabling the plane at this point is masking a bug in our
> modeset, or that some side-effect of writing those registers or waiting
> for that vblank has a vital latching or delay that we have not accounted
> for.

Attached are the requested register dumps:

  intel_reg_dump__nomodeset.gz
          \___no kms -- no corruption

  intel_reg_dump__bad.gz
          \___with kms -- corrupted screen contents with staircase
              shaped and horizontally stretched scan lines. Unreadable.

  intel_reg_dump__lid.gz
          \___after closing and reopening lid. The text is now readable,
              but horizontally stretched. It looks like there's a dark
              pixel between any pair of legitimate font pixels.

  intel_reg_dump__x11.gz
           \___after starting x11. Now both the x11 screen and virtual
               terminals look right.


(BTW: there's also still the problem that the backlight remains off
until I press the "screen darker" key. Maybe this problem is related?
https://bugzilla.kernel.org/show_bug.cgi?id=31522).

m.

[-- Attachment #2: intel_reg_dump__nomodeset.gz --]
[-- Type: application/x-gzip, Size: 2315 bytes --]

[-- Attachment #3: intel_reg_dump__bad.gz --]
[-- Type: application/x-gzip, Size: 2344 bytes --]

[-- Attachment #4: intel_reg_dump__lid.gz --]
[-- Type: application/x-gzip, Size: 2342 bytes --]

[-- Attachment #5: intel_reg_dump__x11.gz --]
[-- Type: application/x-gzip, Size: 2568 bytes --]

[3.0-rc1 regression] (was: Re: [2.6.39 regression] i915/kms: garbled screen because of 49183b281)

[Melchior FRANZ]

* Melchior FRANZ -- Thursday 12 May 2011:
> > Chris Wilson (4):
> >       drm/i915: Only enable the plane after setting the fb base (pre-ILK)
> 
> This patch introduces a bug on my infamous "Acer Travelmate
> 5735Z-452G32Mnss": when KMS takes over, the frame buffer contents
> get completely garbled up on screen, with colored stripes and
> unreadable text (photo on request). Only when X11 is started, the
> screen gets restored again. Closing and re-opening the lid partly
> cures the mess, too: it makes the font readable, though horizontally
> stretched.
> 
> Reverting 49183b2818de6899383bb82bc032f9344d6791ff fixes the
> bug.

This got fixed after my bug report, but now it's back with exactly
the same description. Bisection result to come ...

m.