Fix error from the last pull request.  Making sure we shut the panel off
is more correct and saves power.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Ported over from the old UMS list.  Unfortunately they're still
necessary especially on older laptop platforms.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=22126.
Tested-by: NXavier <shiningxc@gmail.com>
Tested-by: NDiego Escalante Urrelo <diegoe@gnome.org>
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NEric Anholt <eric@anholt.net>

The crtc mode fixup is run after the encoders adjust the mode to fit on
their output, so don't reset the mode!

Fixes:

  Bug 29057 - display corruption under 800x600 on netbook
              (1024x600) with 'Full Aspect' scaling
  https://bugs.freedesktop.org/show_bug.cgi?id=29057Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: NXun Fang <xunx.fang@intel.com>
Signed-off-by: NEric Anholt <eric@anholt.net>

At module unload time we'll tear down the fbdev state.  We do so under
the struct mutex, so we shouldn't try to use the unlocked variant of
the GEM object unreference function or we may deadlock.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NEric Anholt <eric@anholt.net>

We should only free the compressed llb if we allocated it in the first
place otherwise we'll panic at unload time.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NEric Anholt <eric@anholt.net>

We're really supposed to do this to avoid trouble with underflows when
multiple planes are active.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=26987.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Nfangxun <xunx.fang@intel.com>
Signed-off-by: NEric Anholt <eric@anholt.net>

This fixes a potential modesetting error during boot with plymouth on
Broadwater and Crestline introduced with 9df47c. The framebuffer was
hard-coding an alignment of 64K, but the modesetting code required the
documented alignment of 128K. The result was that we would attempt to
unbind the pinned fbcon buffer, triggering an ERROR and ultimately
failing the mode change.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NEric Anholt <eric@anholt.net>

When enabling the eDP port, we need to make sure the panel is turned on
after training the link.  If we don't, it likely won't come back after
suspend or may not come up at all.

For unknown reasons, unlocking the panel regs before initiating a power
on sequence is necessary.  There are known bugs in the PCH panel
sequencing logic, apparently this is one possible workaround.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=28739.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: N"Paulo J. S. Silva" <pjssilva@gmail.com>
Signed-off-by: NEric Anholt <eric@anholt.net>

In some cases, unlocking the panel regs is safe and can help us avoid a
flickery, full mode set sequence.  So define the unlock key and use it.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NEric Anholt <eric@anholt.net>

Fixes an Ironlake laptop with a 68.940MHz 1280x800 panel and 120MHz SSC
reference clock.

More generally, the 0.488% tolerance used before is just too tight to
reliably find a PLL setting.  I extracted the search algorithm and
modified it to find the dot clocks with maximum error over the valid
range for the given output type:

http://people.freedesktop.org/~ajax/intel_g4x_find_best_pll.c

This gave:

Worst dotclock for Ironlake DAC refclk is 350000kHz (error 0.00571)
Worst dotclock for Ironlake SL-LVDS refclk is 102321kHz (error 0.00524)
Worst dotclock for Ironlake DL-LVDS refclk is 219642kHz (error 0.00488)
Worst dotclock for Ironlake SL-LVDS SSC refclk is 84374kHz (error 0.00529)
Worst dotclock for Ironlake DL-LVDS SSC refclk is 183035kHz (error 0.00488)
Worst dotclock for G4X SDVO refclk is 267600kHz (error 0.00448)
Worst dotclock for G4X HDMI refclk is 334400kHz (error 0.00478)
Worst dotclock for G4X SL-LVDS refclk is 95571kHz (error 0.00449)
Worst dotclock for G4X DL-LVDS refclk is 224000kHz (error 0.00510)
Signed-off-by: NAdam Jackson <ajax@redhat.com>
Cc: stable@kernel.org
Signed-off-by: NEric Anholt <eric@anholt.net>

We cannot the initial configuration set by the BIOS not to have a dither
mode enabled which conflicts with our enabling the Spatial Temporal 1
dither mode for PCH. In particular, the BIOS may either enable temporal
dithering or the Spatial Temporal 2 with the result that we enable pure
temporal dithering. Temporal dithering looks bad and is perceived as a
flicker.

Fixes:

  Bug 29248 - [Arrandale] Annoying flicker on internal panel, goes away
              after suspend to RAM
  https://bugs.freedesktop.org/show_bug.cgi?id=29248Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NEric Anholt <eric@anholt.net>

If a framebuffer is shared across CRTCs, the x,y position of one of them
is likely to be something other than the origin (e.g. for extended
desktop configs).  So calculate the offset at flip time so such
configurations can work.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=28518.
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: NThomas M. <tmezzadra@gmail.com>
Tested-by: Nfangxun <xunx.fang@intel.com>
Cc: stable@kernel.org
Signed-off-by: NEric Anholt <eric@anholt.net>

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=29062Reported-by: NAndres Cimmarusti <acimmarusti@gmail.com>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Return the crtc_id, not the counter value.  They are not
necessarily the same.

Cc: Jerome Glisse <glisse@freedesktop.org>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The commit 1e8655f8
    drm/ttm: Fix build on architectures without AGP

looks at TTM_HAS_AGP before it has been set in ttm_bo_driver.h

Move the conditional inclusion of <asm/agp.h> *after* we have included
ttm_bo_driver.h
Signed-of-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Add delay after turning off the LVDS encoder.

Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=16389Tested-by: NJan Kreuzer <kontrollator@gmx.de>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

We only add/remove crtcs at driver load, you cannot remove when
the GPU is running a CS packet since the fd is open, when
GPU hotplugging on radeons actually is needed all this locking
needs a review and I've started re-working kms core locking to deal
with this better. But for now avoid long delays in CS processing when
hotplug detect is happening in a different thread.

this fixes a regression introduced with hotplug detection.
Signed-off-by: NDave Airlie <airlied@redhat.com>

Add the shrinkers missed in the first conversion of the API in
commit 7f8275d0 ("mm: add context argument to
shrinker callback").
Signed-off-by: NDave Chinner <dchinner@redhat.com>

Reported-by: NAlexander Y. Fomichev <git.user@gmail.com>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Connector is actually DVI rather than HDMI.
Reported-by: NtrapDoor <trapdoor6@gmail.com>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

this fixes a regression since the fbcon rework.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

On nv50 it became impossible to attempt a PCI ROM shadow of the VBIOS,
which will break some setups.

This patch also removes the different ordering of shadow methods for
pre-nv50 chipsets.  The reason for the different ordering was paranoia,
but it should hopefully be OK to try shadowing PRAMIN first.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

This fixes a regression caused by b2ea4aa6
due to the way shared ddc with multiple digital connectors was handled.

You generally have two cases where DDC lines are shared:
- HDMI + VGA
- HDMI + DVI-D

HDMI + VGA is easy to deal with because you can check the EDID for the
to see if the attached monitor is digital.  A shared DDC line with two
digital connectors is more complex.  You can't use the hdmi bits in the
EDID since they may not be there with DVI<->HDMI adapters.  In this case
all we can do is check the HPD pins to see which is connected as we have
no way of knowing using the EDID.

Reported-by: trapdoor6@gmail.com
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

A lot of 945GMs have had stability issues for a long time, this manifested as X hangs, blitter engine hangs, and lots of crashes.

one such report is at:
https://bugs.freedesktop.org/show_bug.cgi?id=20560

along with numerous distro bugzillas.

This only took a week of digging and hair ripping to figure out.

Tracked down and tested on a 945GM Lenovo T60,
previously running
x11perf -copypixwin500
or
x11perf -copywinpix500
repeatedly would cause the GPU to wedge within 4 or 5 tries, with random busy bits set.

After this patch no hangs were observed.

cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

The i915 memory arbiter has a register full of configuration
bits which are currently not defined in the driver header file.
Signed-off-by: NKeith Packard <keithp@keithp.com>
cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

copy_to_user() returns the number of bytes remaining to be copied, but
we want to return a negative error code.  This gets copied to user
space.
Signed-off-by: NDan Carpenter <error27@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

If kzalloc() fails exit with -ENOMEM.
Signed-off-by: NKulikov Vasiliy <segooon@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Make inclusion of <asm/agp.h> conditional on TTM_HAS_AGP.  The use
of the functions declared in it is already conditional.
Reported-by: NGeert Stappers <stappers@stappers.nl>
Signed-off-by: NBen Hutchings <ben@decadent.org.uk>
Tested-by: NGeert Stappers <stappers@stappers.nl>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The current shrinker implementation requires the registered callback
to have global state to work from. This makes it difficult to shrink
caches that are not global (e.g. per-filesystem caches). Pass the shrinker
structure to the callback so that users can embed the shrinker structure
in the context the shrinker needs to operate on and get back to it in the
callback via container_of().
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

The hibernate issues that got fixed in commit 985b823b ("drm/i915:
fix hibernation since i915 self-reclaim fixes") turn out to have been
incomplete.  Vefa Bicakci tested lots of hibernate cycles, and without
the __GFP_RECLAIMABLE flag the system eventually fails to resume.

With the flag added, Vefa can apparently hibernate forever (or until he
gets bored running his automated scripts, whichever comes first).

The reclaimable flag was there originally, and was one of the flags that
were dropped (unintentionally) by commit 4bdadb97 ("drm/i915:
Selectively enable self-reclaim") that introduced all these problems,
but I didn't want to just blindly add back all the flags in commit
985b823b, and it looked like __GFP_RECLAIM wasn't necessary.  It
clearly was.

I still suspect that there is some subtle reason we're missing that
causes the problems, but __GFP_RECLAIMABLE is certainly not wrong to use
in this context, and is what the code historically used.  And we have no
idea what the causes the corruption without it.
Reported-and-tested-by: NM. Vefa Bicakci <bicave@superonline.com>
Cc: Dave Airlie <airlied@gmail.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: stable@kernel.org
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The asics in question have the following requirements with regard to
their gart setups:

1. The GART aperture size has to be in the form of 2^X bytes, where X is from 25 to 31
2. The GART aperture MC base has to be aligned to a boundary equal to the size of the
aperture.
3. The GART page table has to be aligned to the boundary equal to the size of the table.
4. The GART page table size is: table_entry_size * (aperture_size / page_size)
5. The GART page table has to be allocated in non-paged, non-cached, contiguous system
memory.

This patch takes care 2.  The rest should already be handled properly.

This fixes a regression noticed by: Torsten Kaiser <just.for.lkml@googlemail.com>
Tested-by: NTorsten Kaiser <just.for.lkml@googlemail.com>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Check ulBootUpMemoryClock on AMD IGPs.

Fix regression noticed by Torsten Kaiser <just.for.lkml@googlemail.com>
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

fixes fdo bug 26915
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

We don't currently update the DPMS status of the connector (both in the
connector itself and the connector's DPMS property) in the fb helper
code.  This means that if the kernel FB core has blanked the screen,
sysfs will still show a DPMS status of "on".  It also means that when X
starts, it will try to light up the connectors, but the drm_crtc_helper
code will ignore the DPMS change since according to the connector, the
DPMS status is already on.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=28436 (the annoying
"my screen was blanked when I started X and now it won't light up" bug).
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Connectors with a shared ddc line can be connected to different
encoders.

Reported by Pasi Kärkkäinen <pasik@iki.fi> on dri-devel
Signed-off-by: NAlex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

Repeated ttm_page_alloc_init/fini fails noisily because the pool
manager kobj isn't zeroed out between uses (we could do just that but
statically allocated kobjects are generally considered a bad thing).
Move it to kzalloc'ed memory.

Note that this patch drops the refcounting behavior of the pool
allocator init/fini functions: it would have led to a race condition
in its current form, and anyway it was never exploited.

This fixes a regression with reloading kms modules at runtime, since
page allocator was introduced.
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Oops, when merging the extra details following an OOM, I missed that
driver_private is now NULL and the correct way to convert from the
drm_gem_object into the drm_i915_gem_object is to use to_intel_bo().

BUG: unable to handle kernel NULL pointer dereference at 00000069
IP: [<c11a4a02>] i915_gem_do_execbuffer+0x71f/0xbb6
*pde = 00000000
Oops: 0000 [#1] SMP
last sysfs file: /sys/devices/virtual/vc/vcsa3/uevent

Pid: 10993, comm: X Not tainted 2.6.35-rc2+ #67 /
EIP: 0060:[<c11a4a02>] EFLAGS: 00213202 CPU: 0
EIP is at i915_gem_do_execbuffer+0x71f/0xbb6
EAX: f647e8a8 EBX: 00000000 ECX: 00000003 EDX: 00000000
ESI: 00424000 EDI: 00000000 EBP: f6508e48 ESP: f6508dd4
 DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
Process X (pid: 10993, ti=f6508000 task=f6432880 task.ti=f6508000)
Stack:
 f6508de0 f7130000 00000001 00000000 00000000 f647e8a8 00000000 f64f8480
<0> f7974414 00000000 00000006 00000000 00000000 f6578000 00000008 00000006
<0> f6797880 00400000 00000000 ffffffe4 f7974400 000000d0 000000d0 000001c0
Call Trace:
 [<c11a4f3a>] ? i915_gem_execbuffer2+0xa1/0xe7
 [<c118ab96>] ? drm_ioctl+0x22c/0x2fa
 [<c11a4e99>] ? i915_gem_execbuffer2+0x0/0xe7
 [<c107e88c>] ? do_sync_read+0x8f/0xca
 [<c1088cbd>] ? vfs_ioctl+0x2c/0x96
 [<c118a96a>] ? drm_ioctl+0x0/0x2fa
 [<c10891f4>] ? do_vfs_ioctl+0x429/0x45a
 [<c107e5c9>] ? fsnotify_access+0x54/0x5f
 [<c107ee1c>] ? vfs_read+0x9a/0xae
 [<c1089258>] ? sys_ioctl+0x33/0x4d
 [<c1002610>] ? sysenter_do_call+0x12/0x26
Code: d0 89 4d c4 31 c9 89 45 d8 eb 44 8b 45 cc 8b 14 88 8b 42 50 89 45
bc 8b 45 a0 8b 52 38 89 55 d0 31 d2 f6 40 20 01 74 0d 8b 55 bc <f6> 42
69 30 0f 95 c2 0f b6 d2 8b 45 d0 c7 45 d4 00 00 00 00 89
EIP: [<c11a4a02>] i915_gem_do_execbuffer+0x71f/0xbb6 SS:ESP 0068:f6508dd4
CR2: 0000000000000069
---[ end trace 3f1d514b34d39381 ]---
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NEric Anholt <eric@anholt.net>

On some machines (currently only the Toshiba Tecra A11 is known), the GPU
locks up when modeset is forced on LID open. This patch adds a new DMI
blacklist and omits modesetting for all matches.

Fixes https://bugzilla.kernel.org/show_bug.cgi?id=15550Signed-off-by: NThomas Bächler <thomas@archlinux.org>
Signed-off-by: NEric Anholt <eric@anholt.net>

The refreshed patch had a copy & paste bug.
Reported-by: NSimon Farnsworth <simon.farnsworth@onelan.com>
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Since commit 4bdadb97 ("drm/i915:
Selectively enable self-reclaim"), we've been passing GFP_MOVABLE to the
i915 page allocator where we weren't before due to some over-eager
removal of the page mapping gfp_flags games the code used to play.

This caused hibernate on Intel hardware to result in a lot of memory
corruptions on resume.  See for example

  http://bugzilla.kernel.org/show_bug.cgi?id=13811

Reported-by: Evengi Golov (in bugzilla)
Signed-off-by: NDave Airlie <airlied@redhat.com>
Tested-by: NM. Vefa Bicakci <bicave@superonline.com>
Cc: stable@kernel.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>