pandora-kernel.git
11 years agodrm/i915: try to train DP even harder
Paulo Zanoni [Fri, 29 Jun 2012 19:03:34 +0000 (16:03 -0300)]
drm/i915: try to train DP even harder

While debugging Haswell link train failures I observed that we never
try the maximum voltage configuration more than once consecutively. We
start the training, the monitor keeps telling us to increase the
voltage, then when we reach the maximum we just go back to the start
(because of the "memset" above "voltage_tries = 0"). When we reach
this point, we keep alternating between the maximum and the minimum
voltages until we give up.

The DP spec suggests that we should try the same voltage 5 times
before giving up. This patch makes us try the maximum voltage at
least 5 times before going back to the minimum voltages.

This patch does not fix any particular bug I'm aware of.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: kill intel_ddc_probe
Daniel Vetter [Wed, 11 Jul 2012 10:31:53 +0000 (12:31 +0200)]
drm/i915: kill intel_ddc_probe

We have way too much lying hardware to rely on a simple "does someone
answer on the ddc i2c address?" check. And now it's unused, so just
kill it.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: check whether we actually received an edid in detect_ddc
Daniel Vetter [Wed, 11 Jul 2012 10:31:52 +0000 (12:31 +0200)]
drm/i915: check whether we actually received an edid in detect_ddc

Somehow detect_ddc manages to fall through all checks when we think
that something responds on the ddc i2c address, but the edid read
failed. Fix this up by explicitly checking for this case.

This fixes a regression on newer chips because since

commit aaa377302b2994fcc2c66741b47da33feb489dca
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sat Jun 16 15:30:32 2012 +0200

    drm/i915/crt: Do not rely upon the HPD presence pin

we use ddc detection also on hotplug capable platforms. And one of
these reads all 0s for any i2c transaction if nothing is connected to
the vga port.

v2: Implement Chris Wilson's review:
- simplify logic, default to "nothing detected"
- kill stale comment
- BUG_ON(!crt->type != ANALOG)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=51900
Tested-by: Yang Guang <guang.a.yang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix up PCH backlight #define mixup
Daniel Vetter [Tue, 10 Jul 2012 22:31:06 +0000 (00:31 +0200)]
drm/i915: fix up PCH backlight #define mixup

I so totally suck.

This can cause a black screen if (for whatever reason) the bios
hasn't set this bit itself.

This regression has been introduced in

commit 7cf4160148136deb31ee5f2802857dd935a38529
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Tue Jun 5 10:07:09 2012 +0200

    drm/i915: clear up backlight #define confusion on gen4+

Tested-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Add comments to explain the BSD tail write workaround
Chris Wilson [Thu, 5 Jul 2012 16:14:01 +0000 (17:14 +0100)]
drm/i915: Add comments to explain the BSD tail write workaround

Having had to dive into the bspec to understand what each stage of the
workaround meant, and how that the ring broadcasting IDLE corresponded
with the GT powering down the ring (i.e. rc6) add comments to aide
the next reader.

And since the register "is used to control all aspects of PSMI and power
saving functions" that makes it quite interesting to inspect with
regards to RC6 hangs, so add it to the error-state.

v2: Rediscover the piece of magic, set the RNCID to 0 before waiting for
the ring to wake up.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Disable the BLT on pre-production SNB hardware
Chris Wilson [Thu, 5 Jul 2012 22:49:40 +0000 (23:49 +0100)]
drm/i915: Disable the BLT on pre-production SNB hardware

It never quite worked despite the numerous workarounds, yet I still see
people trying to use this hardware and filing bug reports. As we no
longer even try to implement the workarounds, since 6a233c78878
(drm/i915/ringbuffer: kill snb blt workaround), simply disable the ring.

v2: Add a message to inform the user about the limited capabilities of
their pre-production hardware.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: initialize power wells in modeset_init_hw
Eugeni Dodonov [Fri, 6 Jul 2012 18:42:36 +0000 (15:42 -0300)]
drm/i915: initialize power wells in modeset_init_hw

This initializes power wells within the modeset_init_hw routine.
Testing has shown that this works for both driver load time and for
suspend-resume code paths.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Only request PM interrupts for the events we handled
Chris Wilson [Thu, 5 Jul 2012 14:02:17 +0000 (15:02 +0100)]
drm/i915: Only request PM interrupts for the events we handled

There is little point waking up every 10ms to service an interrupt which
we then promptly ignore. So only program the the PMIER to enable
interrupts for those events which we do handle, not all of them!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Eugeni Dodonov <eugeni.dodonov@intel.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915/context: Add missing IVB context sizes
Ben Widawsky [Wed, 18 Jul 2012 17:10:10 +0000 (10:10 -0700)]
drm/i915/context: Add missing IVB context sizes

There were some fields missed. Daniel pointed this out in review, and I
know I fixed it, but something happened somehow and some time.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915/context/: s/CTX/CXT
Ben Widawsky [Wed, 18 Jul 2012 17:10:09 +0000 (10:10 -0700)]
drm/i915/context/: s/CTX/CXT

*sigh* the docs had it spelled wrong, corrected it, and then proceeded
to re-do the original error. The original code preserved this history,
and this patch attempts to keep in sync with the current docs.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/sis: fixup sis_mm ioctl structs
Daniel Vetter [Sun, 24 Jun 2012 17:57:24 +0000 (19:57 +0200)]
drm/sis: fixup sis_mm ioctl structs

Userspace uses long in quite a few places more than the kernel. Which
gives me neat proof that I'm the only guy on this side of the galaxy
who ever tried to run glxgears on a 64bit machine with sis graphics on
linux.

Note that the longs in drm_sis_mem_t aren't aligned properly, so this
won't even work with 32bit userspace on 64bit kernel as-is. Hence the
patch can't break that, either.

Nope, I'm not nuts enough to write the 32bit ioctl compat layer for
this and test it with some wine app. Even though hunting the ebay
dungeons for a sis card actually supported by the mesa drivers casts
some doubts on this ...

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill i915/i830 ids from drm_pciids.h
Daniel Vetter [Tue, 25 Oct 2011 23:03:05 +0000 (01:03 +0200)]
drm: kill i915/i830 ids from drm_pciids.h

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: unconditionally clean up dma buffers of closing clients
Daniel Vetter [Tue, 25 Oct 2011 22:31:26 +0000 (00:31 +0200)]
drm: unconditionally clean up dma buffers of closing clients

With the last patch to ditch DMA_QUEUE support, we should be able
to call the dma cleanup uncoditionally, even when the master has
disappeared.

Do so because it just makes more sense.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill dma queue support
Daniel Vetter [Tue, 25 Oct 2011 22:54:41 +0000 (00:54 +0200)]
drm: kill dma queue support

Absolutely unused. All the values are only ever initialized and
then used at most in some debug printout functions.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: ditch strange DRIVER_DMA_QUEUE only error bail-out
Daniel Vetter [Tue, 25 Oct 2011 22:53:57 +0000 (00:53 +0200)]
drm: ditch strange DRIVER_DMA_QUEUE only error bail-out

Only one driver (i810) even sets that flag. Now the actual locking
code uncoditionally promotes lock->context to an unsigned int.

Closer inspection of the userspace reveals that the drm lock context
is defined as an unsigned int (at least on linux). I suspect we just
have a strange case of signedness confusion going on.

Tested on my i815, doesn't seem to break anything.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill reclaim_buffers callback
Daniel Vetter [Tue, 25 Oct 2011 22:20:57 +0000 (00:20 +0200)]
drm: kill reclaim_buffers callback

All leftover users either haven't set DRIVER_HAVE_DMA, in which
case this will never be called, or use the drm_core implementation.

Call that directly in the only callsite.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/savage: clean up reclaim_buffers
Daniel Vetter [Tue, 25 Oct 2011 22:14:15 +0000 (00:14 +0200)]
drm/savage: clean up reclaim_buffers

The reclaim_buffers function of the savage driver actually wants to run
with the hw_lock held - at least there are printks in the call-chain
to that effect. But the drm core only calls reclaim_buffers as used
by savage _after_ forcefully dropping the hwlock (in case it's still
hold by the closing fd).

So do the same idlelock dance as for the other dma drivers and hope
that papers over any issues.

v2: Don't let the idlelock linger around.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill reclaim_buffers_locked
Daniel Vetter [Tue, 25 Oct 2011 21:57:28 +0000 (23:57 +0200)]
drm: kill reclaim_buffers_locked

i810 was the last user of this code, with that gone, kill it.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoRevert "Revert "drm/i810: cleanup reclaim_buffers""
Daniel Vetter [Mon, 11 Jun 2012 19:50:23 +0000 (21:50 +0200)]
Revert "Revert "drm/i810: cleanup reclaim_buffers""

This reverts commit 6e877b576ddf7cde5db2e9a6dcb56fef0ea77e64,
reinstating the original commit:

commit 87499ffdcb1c70f66988cd8febc4ead0ba2f9118
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Tue Oct 25 23:51:24 2011 +0200

    drm/i810: cleanup reclaim_buffers

    My dear old i815 always hits the deadlocked on reclaim_buffers
    warning. Switch over to the idlelock duct-tape on hope that
    works better. I've fired up my i815 and now closing glxgears doesn't
    take 5 seconds anymore. \o/

The original problem with that was that I've moved it ahead in the
series so that it could be included despite some patches not being
ready quite yet. The little problem is that this patch required some
of the previous rework to work correctly.

Now that everything is in the right order again, this actually works
on my i810 and does speed up closing gl apps as the original commit
claimed. Without hanging the machine, as the revert says.

Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill reclaim_buffers_idlelocked functions
Daniel Vetter [Tue, 25 Oct 2011 21:42:29 +0000 (23:42 +0200)]
drm: kill reclaim_buffers_idlelocked functions

The only two users are now folded into the drivers preclose functions,
so this is unused.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/sis: clean up reclaim_buffers
Daniel Vetter [Tue, 25 Oct 2011 21:42:59 +0000 (23:42 +0200)]
drm/sis: clean up reclaim_buffers

Like for via.

v2: Actually drop the idlelock again if taken.

v3: Fixup.

v4: Fixup the "has master" vs. "is master" confusion the refactor
introduced.

v5: Drop the idlelock in the early return path.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/via: clean up reclaim_buffers
Daniel Vetter [Tue, 25 Oct 2011 21:37:09 +0000 (23:37 +0200)]
drm/via: clean up reclaim_buffers

A few things
- kill reclaim_buffers, it's never ever called because via does not set
  DRIVER_HAVE_DMA
- inline the idlelock dance into the buffer reclaim logic and make it
  a simple preclose cleanup function
- directly call the the dma_quiescent function and kill the needless
  if check.

v2: Actually drop the idlelock when we take it. Reported by James
Simmons.

v3: Rebased onto latest drm-next.

v4: Fixup the refactor.

v5: More fixup the refactor - I've accidentally changed the check for
any master to checking whether the closing fd is the master.

v6: Don't forget to drop the idlelock in the early return path, too.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/udl: port over blanking code from udlfb.
Dave Airlie [Tue, 26 Jun 2012 13:53:07 +0000 (14:53 +0100)]
drm/udl: port over blanking code from udlfb.

This ports over the dpms code from udlfb, and should mean
a better chance of turning on some udl devices.

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon/kms: auto detect pcie link speed from root port
Dave Airlie [Wed, 27 Jun 2012 07:35:54 +0000 (08:35 +0100)]
drm/radeon/kms: auto detect pcie link speed from root port

This check the root ports supported link speeds and enables
GEN2 mode if the 5.0 GT link speed is available.

The first 3.0 cards are SI so they will probably need more investigation.

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/pci: add support for getting the supported link bw.
Dave Airlie [Wed, 27 Jun 2012 07:35:53 +0000 (08:35 +0100)]
drm/pci: add support for getting the supported link bw.

This should work for PCIE3.0 as well.

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agopci_regs: define LNKSTA2 pcie cap + bits.
Dave Airlie [Wed, 27 Jun 2012 07:35:52 +0000 (08:35 +0100)]
pci_regs: define LNKSTA2 pcie cap + bits.

We need these for detecting the max link speed for drm drivers.

Acked-by: Bjorn Helgaas <bhelgass@google.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: improve GPU lockup debugging info on r6xx/r7xx/r8xx/r9xx
Jerome Glisse [Wed, 27 Jun 2012 16:25:01 +0000 (12:25 -0400)]
drm/radeon: improve GPU lockup debugging info on r6xx/r7xx/r8xx/r9xx

Print various CP register that have valuable informations regarding
GPU lockup.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/mgag200: fix null pointer dereference
Devendra Naga [Sat, 7 Jul 2012 09:22:15 +0000 (09:22 +0000)]
drm/mgag200: fix null pointer dereference

we are referencing the pointer after doing alloc_apertures,
as alloc_apertures kzallocs, the kzalloc may fail and we get a NULL.

so we need to check for NULL before we dereference this pointer

Signed-off-by: Devendra Naga <devendra.aaru@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: Try harder to avoid HW cursor ending on a multiple of 128 columns.
Michel Dänzer [Tue, 17 Jul 2012 17:02:09 +0000 (19:02 +0200)]
drm/radeon: Try harder to avoid HW cursor ending on a multiple of 128 columns.

This could previously fail if either of the enabled displays was using a
horizontal resolution that is a multiple of 128, and only the leftmost column
of the cursor was (supposed to be) visible at the right edge of that display.

The solution is to move the cursor one pixel to the left in that case.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=33183

Cc: stable@vger.kernel.org
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Make the .mode_fixup() operations mode argument a const pointer
Laurent Pinchart [Tue, 17 Jul 2012 15:56:50 +0000 (17:56 +0200)]
drm: Make the .mode_fixup() operations mode argument a const pointer

The passed mode must not be modified by the operation, make it const.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: remove the list_head from drm_mode_set
Daniel Vetter [Wed, 11 Jul 2012 14:28:08 +0000 (16:28 +0200)]
drm: remove the list_head from drm_mode_set

It's unused. At it confused me quite a bit until I've discovered that.

Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/fb helper: don't call drm_crtc_helper_set_config
Daniel Vetter [Wed, 11 Jul 2012 14:28:07 +0000 (16:28 +0200)]
drm/fb helper: don't call drm_crtc_helper_set_config

Go through the interface vtable instead, because not everyone might be
using the crtc helper code.

Cc: dri-devel@lists.freedesktop.org
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/fb-helper: delay hotplug handling when partially bound
Daniel Vetter [Fri, 15 Jun 2012 09:01:22 +0000 (11:01 +0200)]
drm/fb-helper: delay hotplug handling when partially bound

Ok, this requires quite a dance to actually hit:
1) We plug in a 2nd screen, enable it in both X and (by vt-switching)
in the fbcon.
2) We disable that screen again in with xrandr.
3) We vt-switch again, so that fbcon displays on the 2nd screen, but X
on the first screen. This obviously needs a driver that doesn't switch
off unused functions when regaining the VT.
3) When X controls the vt, we unplug that screen.

Now drm_fb_helper_hotplug_event we noticed that that some crtcs are
bound, but because we still have the fbcon on the 2nd screeen we also
have bound set. Which means the fbcon wrongly assumes it's in control
of everything an happily disables the output on the 2nd screen, but
enables its fb on the first screen.

Work around this issue by counting how many crtcs are bound and how
many are bound to fbcon and assuming that when fbcon isn't bound to
all of them, it better not touch the output configuration.

Conceptually this is the same as only restoring the fbcon output
configuration on the driver's ->lastclose, when we're sure that no one
else is using kms. So this should be consistent with existing kms
drivers.

Chris has created a separate patch for the intel ddx, but I think we
should fix this issue here regardless - the fbcon messing with the
output config while it's not fully in control simply isn't a too
polite behaviour.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50772
Tested-by: Maxim A. Nikulin <M.A.Nikulin@gmail.com>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'next' of git://people.freedesktop.org/~deathsimple/linux into drm-next
Dave Airlie [Fri, 20 Jul 2012 01:20:54 +0000 (21:20 -0400)]
Merge branch 'next' of git://people.freedesktop.org/~deathsimple/linux into drm-next

This contains all the radeon documentation rebased on top of the ib fixes.

* 'next' of git://people.freedesktop.org/~deathsimple/linux:
  drm/radeon: fix SS setup for DCPLL
  drm/radeon: fix up pll selection on DCE5/6
  drm/radeon: start to document evergreen.c
  drm/radeon: start to document the functions r100.c
  drm/radeon: document VM functions in radeon_gart.c (v3)
  drm/radeon: document non-VM functions in radeon_gart.c (v2)
  drm/radeon: document radeon_ring.c (v4)
  drm/radeon: document radeon_fence.c (v2)
  drm/radeon: document radeon_asic.c
  drm/radeon: document radeon_irq_kms.c
  drm/radeon: document radeon_kms.c
  drm/radeon: document radeon_device.c (v2)
  drm/radeon: add rptr save support for r1xx-r5xx
  drm/radeon: update rptr saving logic for memory buffers
  drm/radeon: remove radeon_ring_index()
  drm/radeon: update ib_execute for SI (v2)
  drm/radeon: fix const IB handling v2
  drm/radeon: let sa manager block for fences to wait for v2
  drm/radeon: return an error if there is nothing to wait for

11 years agodrm/radeon: fix SS setup for DCPLL
Alex Deucher [Tue, 17 Jul 2012 18:02:44 +0000 (14:02 -0400)]
drm/radeon: fix SS setup for DCPLL

Need to actually set the SS parameters rather than just 0.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: fix up pll selection on DCE5/6
Alex Deucher [Tue, 17 Jul 2012 18:02:43 +0000 (14:02 -0400)]
drm/radeon: fix up pll selection on DCE5/6

Selecting ATOM_PPLL_INVALID should be equivalent as the
DCPLL or PPLL0 are already programmed for the DISPCLK, but
the preferred method is to always specify the PLL selected.
SetPixelClock will check the parameters and skip the
programming if the PLL is already set up.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: start to document evergreen.c
Alex Deucher [Tue, 17 Jul 2012 18:02:42 +0000 (14:02 -0400)]
drm/radeon: start to document evergreen.c

Still a lot to do.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: start to document the functions r100.c
Alex Deucher [Tue, 17 Jul 2012 18:02:41 +0000 (14:02 -0400)]
drm/radeon: start to document the functions r100.c

Still a lot more to do.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document VM functions in radeon_gart.c (v3)
Alex Deucher [Tue, 17 Jul 2012 18:02:40 +0000 (14:02 -0400)]
drm/radeon: document VM functions in radeon_gart.c (v3)

Document the VM functions in radeon_gart.c

v2: adjust per Christian's suggestions
v3: adjust to Christians's latest changes

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document non-VM functions in radeon_gart.c (v2)
Alex Deucher [Tue, 17 Jul 2012 18:02:39 +0000 (14:02 -0400)]
drm/radeon: document non-VM functions in radeon_gart.c (v2)

Document the non-VM functions in radeon_gart.c

v2: adjust per Christian's suggestions

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_ring.c (v4)
Alex Deucher [Tue, 17 Jul 2012 18:02:38 +0000 (14:02 -0400)]
drm/radeon: document radeon_ring.c (v4)

Adds documentation to most of the functions in
radeon_ring.c

v2: adjust per Christian's suggestions
v3: adjust per Christian's latest patches
v4: adjust per my latest changes

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_fence.c (v2)
Alex Deucher [Tue, 17 Jul 2012 18:02:37 +0000 (14:02 -0400)]
drm/radeon: document radeon_fence.c (v2)

Adds documentation to most of the functions in
radeon_fence.c

v2: address Christian's comments:
- split common concept description into it's own comment
- fix description of intr parameter
- Improve description of -EDEADLK error

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_asic.c
Alex Deucher [Tue, 17 Jul 2012 18:02:36 +0000 (14:02 -0400)]
drm/radeon: document radeon_asic.c

Adds documentation to most of the functions in
radeon_asic.c

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_irq_kms.c
Alex Deucher [Tue, 17 Jul 2012 18:02:35 +0000 (14:02 -0400)]
drm/radeon: document radeon_irq_kms.c

Adds documentation to most of the functions in
radeon_irq_kms.c

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_kms.c
Alex Deucher [Tue, 17 Jul 2012 18:02:34 +0000 (14:02 -0400)]
drm/radeon: document radeon_kms.c

Adds documentation to most of the functions in
radeon_kms.c

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: document radeon_device.c (v2)
Alex Deucher [Tue, 17 Jul 2012 18:02:33 +0000 (14:02 -0400)]
drm/radeon: document radeon_device.c (v2)

Adds documentation to most of the functions in
radeon_device.c

v2: split out general descriptions as per Christian's
comments.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: add rptr save support for r1xx-r5xx
Alex Deucher [Tue, 17 Jul 2012 18:02:32 +0000 (14:02 -0400)]
drm/radeon: add rptr save support for r1xx-r5xx

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: update rptr saving logic for memory buffers
Alex Deucher [Tue, 17 Jul 2012 18:02:31 +0000 (14:02 -0400)]
drm/radeon: update rptr saving logic for memory buffers

Add support for using memory buffers rather than
scratch registers.  Some rings may not be able to
write to scratch registers.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: remove radeon_ring_index()
Alex Deucher [Tue, 17 Jul 2012 18:02:30 +0000 (14:02 -0400)]
drm/radeon: remove radeon_ring_index()

Just store the index in the ring structure.
Idea taken from one of Jerome's wip rptr patches.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: update ib_execute for SI (v2)
Alex Deucher [Tue, 17 Jul 2012 18:02:29 +0000 (14:02 -0400)]
drm/radeon: update ib_execute for SI (v2)

When submitting a CONST_IB, emit a SWITCH_BUFFER
packet before the CONST_IB.  This isn't strictly necessary
(the driver will work fine without it), but is good practice
and allows for more flexible DE/CE sychronization options
in the future.  Current userspace drivers do not take
advantage of the CE yet.

v2: - clean up code flow a bit
    - no need to flush caches for CONST IB

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: fix const IB handling v2
Christian König [Fri, 13 Jul 2012 11:06:00 +0000 (13:06 +0200)]
drm/radeon: fix const IB handling v2

Const IBs are executed on the CE not the CP, so we can't
fence them in the normal way.

So submit them directly before the IB instead, just as
the documentation says.

v2: keep the extra documentation

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: let sa manager block for fences to wait for v2
Christian König [Wed, 11 Jul 2012 19:07:57 +0000 (21:07 +0200)]
drm/radeon: let sa manager block for fences to wait for v2

Otherwise we can encounter out of memory situations under extreme load.

v2: add documentation for the new function

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: return an error if there is nothing to wait for
Christian König [Wed, 11 Jul 2012 15:12:11 +0000 (17:12 +0200)]
drm/radeon: return an error if there is nothing to wait for

Otherwise the sa managers out of memory
handling doesn't work.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm: Disallow DRM_IOCTL_MODESET_CTL for KMS drivers
Laurent Pinchart [Tue, 29 May 2012 22:58:09 +0000 (00:58 +0200)]
drm: Disallow DRM_IOCTL_MODESET_CTL for KMS drivers

DRM_IOCTL_MODESET_CTL must only be used for UMS drivers. Make it a no-op
for KMS drivers.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Michel Dänzer <michel@daenzer.net>
Signed-off-by: Dave Airlie <airlied@gmail.com>
11 years agoMerge branch 'next' of git://people.freedesktop.org/~deathsimple/linux into drm-next
Dave Airlie [Tue, 17 Jul 2012 09:46:49 +0000 (19:46 +1000)]
Merge branch 'next' of git://people.freedesktop.org/~deathsimple/linux into drm-next

This merges Christian work that has been hanging around on the list.

11 years agodrm/radeon: implement ring saving on reset v4
Christian König [Mon, 9 Jul 2012 09:52:44 +0000 (11:52 +0200)]
drm/radeon: implement ring saving on reset v4

Try to save whatever is on the rings when
we encounter an lockup.

v2: Fix spelling error. Free saved ring data if reset fails.
    Add documentation for the new functions.
v3: Some more spelling fixes
v4: It doesn't make sense to save anything if all fences
    are signaled

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: record what is next valid wptr for each ring v4
Christian König [Fri, 6 Jul 2012 14:22:55 +0000 (16:22 +0200)]
drm/radeon: record what is next valid wptr for each ring v4

Before emitting any indirect buffer, emit the offset of the next
valid ring content if any. This allow code that want to resume
ring to resume ring right after ib that caused GPU lockup.

v2: use scratch registers instead of storing it into memory
v3: skip over the surface sync for ni and si as well
v4: use SET_CONFIG_REG instead of PACKET0

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: move radeon_ib_ring_tests out of chipset code
Christian König [Sat, 7 Jul 2012 10:47:58 +0000 (12:47 +0200)]
drm/radeon: move radeon_ib_ring_tests out of chipset code

Making it easier to control when it is executed.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove vm_manager start/suspend
Christian König [Thu, 5 Jul 2012 12:32:00 +0000 (14:32 +0200)]
drm/radeon: remove vm_manager start/suspend

Just restore the page table instead. Addressing three
problem with this change:

1. Calling vm_manager_suspend in the suspend path is
   problematic cause it wants to wait for the VM use
   to end, which in case of a lockup never happens.

2. In case of a locked up memory controller
   unbinding the VM seems to make it even more
   unstable, creating an unrecoverable lockup
   in the end.

3. If we want to backup/restore the leftover ring
   content we must not unbind VMs in between.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove r600_blit_suspend
Christian König [Thu, 5 Jul 2012 14:05:28 +0000 (16:05 +0200)]
drm/radeon: remove r600_blit_suspend

Just reinitialize the shader content on resume instead.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove ip_pool start/suspend
Christian König [Thu, 5 Jul 2012 09:55:34 +0000 (11:55 +0200)]
drm/radeon: remove ip_pool start/suspend

The IB pool is in gart memory, so it is completely
superfluous to unpin / repin it on suspend / resume.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: make cp init on cayman more robust
Christian König [Wed, 4 Jul 2012 19:36:53 +0000 (21:36 +0200)]
drm/radeon: make cp init on cayman more robust

It's not critical, but the current code isn't
100% correct.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove FIXME comment from chipset suspend
Christian König [Thu, 5 Jul 2012 11:33:41 +0000 (13:33 +0200)]
drm/radeon: remove FIXME comment from chipset suspend

For a normal suspend/resume we allready wait for
the rings to be empty, and for a suspend/reasume
in case of a lockup we REALLY don't want to wait
for anything.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix fence init after resume
Christian König [Sat, 7 Jul 2012 11:10:39 +0000 (13:10 +0200)]
drm/radeon: fix fence init after resume

Start with last signaled fence number instead
of last emitted one.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix fence value access
Christian König [Mon, 9 Jul 2012 08:52:39 +0000 (10:52 +0200)]
drm/radeon: fix fence value access

It is possible that radeon_fence_process is called
after writeback is disabled for suspend, leading
to an invalid read of register 0x0.

This fixes a problem for me where the fence value
is temporary incremented by 0x100000000 on
suspend/resume.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix ring commit padding
Christian König [Sat, 7 Jul 2012 10:11:32 +0000 (12:11 +0200)]
drm/radeon: fix ring commit padding

We don't need to pad anything if the number of dwords
written to the ring already matches the requirements.

Fixes some "writting more dword to ring than expected"
warnings.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add an exclusive lock for GPU reset v2
Jerome Glisse [Mon, 2 Jul 2012 16:45:19 +0000 (12:45 -0400)]
drm/radeon: add an exclusive lock for GPU reset v2

GPU reset need to be exclusive, one happening at a time. For this
add a rw semaphore so that any path that trigger GPU activities
have to take the semaphore as a reader thus allowing concurency.

The GPU reset path take the semaphore as a writer ensuring that
no concurrent reset take place.

v2: init rw semaphore

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix fence related segfault in CS
Christian König [Tue, 3 Jul 2012 12:05:41 +0000 (14:05 +0200)]
drm/radeon: fix fence related segfault in CS

Don't return success if scheduling the IB fails, otherwise
we end up with an oops in ttm_eu_fence_buffer_objects.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: add error handling to radeon_vm_unbind_locked
Christian König [Mon, 25 Jun 2012 13:13:50 +0000 (15:13 +0200)]
drm/radeon: add error handling to radeon_vm_unbind_locked

Waiting for a fence can fail for different reasons,
the most common is a deadlock.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add error handling to fence_wait_empty_locked
Christian König [Fri, 29 Jun 2012 09:33:12 +0000 (11:33 +0200)]
drm/radeon: add error handling to fence_wait_empty_locked

Instead of returning the error handle it directly
and while at it fix the comments about the ring lock.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm: Add colouring to the range allocator
Chris Wilson [Tue, 10 Jul 2012 10:15:23 +0000 (11:15 +0100)]
drm: Add colouring to the range allocator

In order to support snoopable memory on non-LLC architectures (so that
we can bind vgem objects into the i915 GATT for example), we have to
avoid the prefetcher on the GPU from crossing memory domains and so
prevent allocation of a snoopable PTE immediately following an uncached
PTE. To do that, we need to extend the range allocator with support for
tracking and segregating different node colours.

This will be used by i915 to segregate memory domains within the GTT.

v2: Now with more drm_mm helpers and less driver interference.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Airlie <airlied@redhat.com
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@gmail.com>
11 years agodrm: fail gracefully when proc isn't setup.
Dave Airlie [Fri, 6 Jul 2012 17:06:42 +0000 (18:06 +0100)]
drm: fail gracefully when proc isn't setup.

If drm can't find proc it should fail more gracefully, than just
oopsing, this tests drm_class is NULL, and sets it to NULL in the
fail paths.

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge tag 'drm-intel-next-2012-07-06' of git://people.freedesktop.org/~danvet/drm...
Dave Airlie [Sat, 14 Jul 2012 08:15:21 +0000 (18:15 +1000)]
Merge tag 'drm-intel-next-2012-07-06' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

Daniel writes:
New pull for -next. Highlights:
- rc6/turbo support for hsw (Eugeni)
- improve corner-case of the reset handling code - gpu reset handling
  should be rock-solid now
- support for fb offset > 4096 pixels on gen4+ (yeah, you need some fairly
  big screens to hit that)
- the "Flush Me Harder" patch to fix the gen6+ fallout from disabling the
  flushing_list
- no more /dev/agpgart on gen6+!
- HAS_PCH_xxx improvements from Paulo
- a few minor bits&pieces all over, most of it in thew hsw code

* tag 'drm-intel-next-2012-07-06' of git://people.freedesktop.org/~danvet/drm-intel: (40 commits)
  drm/i915: program FDI_RX TP and FDI delays
  drm/i915: introduce for_each_encoder_on_crtc
  drm/i915: adjust framebuffer base address on gen4+
  drm/i915: introduce crtc->dspaddr_offset
  drm/i915: Reject page flips with changed format/offset/pitch
  drm/i915: Zero initialize mode_cmd
  drm/i915: don't return a spurious -EIO from intel_ring_begin
  drm/i915: properly SIGBUS on I/O errors
  drm/i915: don't hang userspace when the gpu reset is stuck
  drm/i915: non-interruptible sleeps can't handle -EAGAIN
  drm/i915: don't trylock in the gpu reset code
  drm/i915: fix PIPE_DDI_PORT_MASK
  drm/i915: prevent bogus intel_update_fbc notifications
  drm/i915: re-initialize DDI buffer translations after resume
  drm/i915: don't ironlake_init_pch_refclk() on LPT
  drm/i915: get rid of dev_priv->info->has_pch_split
  drm/i915: add PCH_NONE to enum intel_pch
  drm/i915: prefer wide & slow to fast & narrow in DP configs
  drm/i915: fix up ilk rc6 disabling confusion
  drm/i915: move force wake support into intel_pm
  ...

11 years agodrm/i915: program FDI_RX TP and FDI delays
Eugeni Dodonov [Wed, 4 Jul 2012 23:15:16 +0000 (20:15 -0300)]
drm/i915: program FDI_RX TP and FDI delays

This is required for a stable FDI connection.

v2: fix and simplify the FDI_RX_MISC bits as noticed by Paulo Zanoni.

CC: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: introduce for_each_encoder_on_crtc
Daniel Vetter [Thu, 5 Jul 2012 07:50:24 +0000 (09:50 +0200)]
drm/i915: introduce for_each_encoder_on_crtc

We already have this pattern at quite a few places, and moving part of
the modeset helper stuff into the driver will add more.

v2: Don't clobber the crtc struct name with the macro parameter ...

v3: Convert two more places noticed by Paulo Zanoni.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: adjust framebuffer base address on gen4+
Daniel Vetter [Thu, 5 Jul 2012 10:17:30 +0000 (12:17 +0200)]
drm/i915: adjust framebuffer base address on gen4+

The tileoffset register only supports a limited offset in x/y of 4096,
so for giant screen configuration with a shared fb we wrap around.

Fix this by computing a linear offset in tiles (pages) and only use
the tileoffset register to offset within the tile.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: introduce crtc->dspaddr_offset
Daniel Vetter [Thu, 5 Jul 2012 10:17:29 +0000 (12:17 +0200)]
drm/i915: introduce crtc->dspaddr_offset

To avoid recomputing the display framebuffer offset on gen2/3
pageflips. This is also prep work to do similar trickery on gen4+

Also:
- kill "Start", such upper-case remnants from the ddx must surely die.
- rename "Offset" to linear_offset, to make it clearer that on gen4+
  this is only used by the hw for linear buffers, for tiled buffers it
  uses the TILEOFF register.
- call DSAPADDR DSPLINOFF on gen4+ for the same reason (and because
  the documentation really renamed the register).

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Reject page flips with changed format/offset/pitch
Ville Syrjälä [Thu, 24 May 2012 18:08:59 +0000 (21:08 +0300)]
drm/i915: Reject page flips with changed format/offset/pitch

MI display flips can't handle some changes in the framebuffer
format or layout. Return an error in such cases.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Zero initialize mode_cmd
Ville Syrjälä [Thu, 24 May 2012 18:08:56 +0000 (21:08 +0300)]
drm/i915: Zero initialize mode_cmd

Zero initialize the mode_cmd structure when creating the kernel
framebuffer. Avoids having uninitialized data in offsets[0] for
instance.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't return a spurious -EIO from intel_ring_begin
Daniel Vetter [Wed, 4 Jul 2012 20:52:50 +0000 (22:52 +0200)]
drm/i915: don't return a spurious -EIO from intel_ring_begin

The issue with this check is that it results in userspace receiving an
-EIO while the gpu reset hasn't completed, resulting in fallback to sw
rendering or worse.

Now there's also a stern comment in intel_ring_wait_seqno saying that
intel_ring_begin should not return -EAGAIN, ever, because some callers
can't handle that. But after an audit of the callsites I don't see any
issues. I guess the last problematic spot disappeared with the removal
of the pipelined fencing code.

So do the right thing and call check_wedge, which should properly
decide whether an -EAGAIN or -EIO is appropriate if wedged is set.

Note that the early check for a wedged gpu before touching the ring is
rather important (and it took me quite some time of acting like the
densest doofus to figure that out): If we don't do that and the gpu
died for good, not having been resurrect by the reset code, userspace
can merrily fill up the entire ring until it notices that something is
amiss.

Allowing userspace to emit more render, despite that we know that it
will fail can't lead to anything good (and by experience can lead to
all sorts of havoc, including angering the OOM gods and hard-hanging
the hw for good).

v2: Fix EAGAIN mispell, noticed by Chris Wilson.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: properly SIGBUS on I/O errors
Daniel Vetter [Wed, 4 Jul 2012 20:18:42 +0000 (22:18 +0200)]
drm/i915: properly SIGBUS on I/O errors

... instead of looping endless with no hope of ever serving that
page-fault. We only need to break out of this loop when the gpu died,
to run the reset work (and hopefully resurrect it).

To clarify questions Chris raised on irc: This is about handling I/O
errors not from our own code, but e.g. when the disk died when trying
to swap in a gem bo. So this patch remidies the issue that the current
handling only handles gpu-death-induced cases of -EIO. Admittedly,
dying disks are much rarer than hanging gpus ...To clarify questions
Chris raised on irc: This is about handling I/O errors not from our
own code, but e.g. when the disk died when trying to swap in a gem bo.
So this patch remidies the issue that the current handling only
handles gpu-death-induced cases of -EIO. Admittedly, dying disks are
much rarer than hanging gpus ...

This seems to have been lost in:

commit d9bc7e9f32716901c617e1f0fb6ce0f74f172686
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Feb 7 13:09:31 2011 +0000

    drm/i915: Fix infinite loop regression from 21dd3734

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't hang userspace when the gpu reset is stuck
Daniel Vetter [Wed, 4 Jul 2012 20:18:41 +0000 (22:18 +0200)]
drm/i915: don't hang userspace when the gpu reset is stuck

With the gpu reset no longer using a trylock we've increased the
chances of userspace getting stuck quite a bit. To make that
(hopefully) rare case more paletable time out when waiting for the gpu
reset code to complete and signal this little issue to the caller by
returning -EIO.

This should help userspace to somewhat gracefully fall back and
hopefully allow the user to grab some logs and reboot the machine
(instead of staring at a frozen X screen in agony).

Suggested by Chris Wilson because I've been stubborn about allowing
the gpu reset code no to fail, ever (by removing the trylock).

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: non-interruptible sleeps can't handle -EAGAIN
Daniel Vetter [Wed, 4 Jul 2012 20:54:13 +0000 (22:54 +0200)]
drm/i915: non-interruptible sleeps can't handle -EAGAIN

So don't return -EAGAIN, even in the case of a gpu hang. Remap it to
-EIO instead. Note that this isn't really an issue with
interruptability, but more that we have quite a few codepaths (mostly
around kms stuff) that simply can't handle any errors and hence not
even -EAGAIN. Instead of adding proper failure paths so that we could
restart these ioctls we've opted for the cheap way out of sleeping
non-interruptibly.  Which works everywhere but when the gpu dies,
which this patch fixes.

So essentially interruptible == false means 'wait for the gpu or die
trying'.'

This patch is a bit ugly because intel_ring_begin is all non-interruptible
and hence only returns -EIO. But as the comment in there says,
auditing all the callsites would be a pain.

To avoid duplicating code, reuse i915_gem_check_wedge in __wait_seqno
and intel_wait_ring_buffer. Also use the opportunity to clarify the
different cases in i915_gem_check_wedge a bit with comments.

v2: Don't access dev_priv->mm.interruptible from check_wedge - we
might not hold dev->struct_mutex, making this racy. Instead pass
interruptible in as a parameter. I've noticed this because I've hit a
BUG_ON(!mutex_is_locked) at the top of check_wedge. This has been
added in

commit b4aca0106c466b5a0329318203f65bac2d91b682
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Wed Apr 25 20:50:12 2012 -0700

    drm/i915: extract some common olr+wedge code

although that commit is missing any justification for this. I guess
it's just copy&paste, because the same commit add the same BUG_ON
check to check_olr, where it indeed makes sense.

But in check_wedge everything we access is protected by other means,
so this is superflous. And because it now gets in the way (we add a
new caller in __wait_seqno, which can be called without
dev->struct_mutext) let's just remove it.

v3: Group all the i915_gem_check_wedge refactoring into this patch, so
that this patch here is all about not returning -EAGAIN to callsites
that can't handle syscall restarting.

v4: Add clarification what interuptible == fales means in our code,
requested by Ben Widawsky.

v5: Fix EAGAIN mispell noticed by Chris Wilson.

Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't trylock in the gpu reset code
Daniel Vetter [Wed, 4 Jul 2012 20:18:39 +0000 (22:18 +0200)]
drm/i915: don't trylock in the gpu reset code

Simply failing to reset the gpu because someone else might still hold
the mutex isn't a great idea - I see reliable silent reset failures.
And gpu reset simply needs to be reliable and Just Work.

"But ... the deadlocks!"

We already kick all processes waiting for the gpu before launching the
reset work item. New waiters need to check the wedging state anyway
and then bail out. If we have places that can deadlock, we simply need
to fix them.

"But ... testing!"

We have the gpu hangman, and if the current gpu load gem_exec_nop
isn't good enough to hit a specific case, we can add a new one.

"But ...  don't we return -EAGAIN for non-interruptible calls to
wait_seqno now?"

Yep, but this problem already exists in the current code. A follow up
patch will remedy this by returning -EIO for non-interruptible sleeps
if the gpu died and the low-level wait bails out with -EAGAIN.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix PIPE_DDI_PORT_MASK
Paulo Zanoni [Fri, 29 Jun 2012 18:39:32 +0000 (15:39 -0300)]
drm/i915: fix PIPE_DDI_PORT_MASK

Only bits 30:28, bit 31 is PIPE_DDI_FUNC_ENABLE.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: prevent bogus intel_update_fbc notifications
Eugeni Dodonov [Thu, 28 Jun 2012 20:11:43 +0000 (17:11 -0300)]
drm/i915: prevent bogus intel_update_fbc notifications

This pollutes dmesg output even if we do not have FBC for the device, so
move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: re-initialize DDI buffer translations after resume
Eugeni Dodonov [Thu, 28 Jun 2012 18:55:35 +0000 (15:55 -0300)]
drm/i915: re-initialize DDI buffer translations after resume

This is necessary for the modesetting to work correctly after a
suspend-resume cycle. Without this, the pipes and clocks got the correct
configuration, but the underlying DDI buffers configuration was lost.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't ironlake_init_pch_refclk() on LPT
Paulo Zanoni [Tue, 3 Jul 2012 18:57:33 +0000 (15:57 -0300)]
drm/i915: don't ironlake_init_pch_refclk() on LPT

This function is used to set the PCH_DREF_CONTROL register, which does
not exist on LPT anymore.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: get rid of dev_priv->info->has_pch_split
Paulo Zanoni [Tue, 3 Jul 2012 18:57:32 +0000 (15:57 -0300)]
drm/i915: get rid of dev_priv->info->has_pch_split

Previously we had has_pch_split to tell us whether we had a PCH or not
and we also had dev_priv->pch_type to tell us which kind of PCH it
was, but it could only be used if we were 100% sure we did have a PCH.
Now that PCH_NONE was added to dev_priv->pch_type we don't need
has_pch_split anymore: we can just check for pch_type != PCH_NONE.

The HAS_PCH_{IBX,CPT,LPT} macros use dev_priv->pch_type, so they can
only be called after intel_detect_pch. The HAS_PCH_SPLIT macro looks
at dev_priv->info->has_pch_split, which is available earlier.

Since the goal is to implement HAS_PCH_SPLIT using dev_priv->pch_type
instead of dev_priv->info->has_pch_split, we need to make sure that
intel_detect_pch is called before any calls to HAS_PCH_SPLIT are made.
So we moved the intel_detect_pch call to an earlier stage.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: add PCH_NONE to enum intel_pch
Paulo Zanoni [Tue, 3 Jul 2012 21:48:16 +0000 (18:48 -0300)]
drm/i915: add PCH_NONE to enum intel_pch

And rely on the fact that it's 0 to assume that machines without a PCH
will have PCH_NONE as dev_priv->pch_type.

Just today I finally realized that HAS_PCH_IBX is true for machines
without a PCH. IMHO this is totally counter-intuitive and I don't
think it's a good idea to assume that we're going to check for
HAS_PCH_IBX only after we check for HAS_PCH_SPLIT.

I believe that in the future we'll have more PCH types and checks
like:

    if (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev))

will become more and more common. There's a good chance that we may
break non-PCH machines by adding these checks in code that runs on all
machines. I also believe that the HAS_PCH_SPLIT check will become less
common as we add more and more different PCH types. We'll probably
start replacing checks like:

    if (HAS_PCH_SPLIT(dev))
        foo();
    else
        bar();

with:

    if (HAS_PCH_NEW(dev))
        baz();
    else if (HAS_PCH_OLD(dev) || HAS_PCH_IBX(dev))
        foo();
    else
        bar();

and this may break gen 2/3/4.

As far as we have investigated, this patch will affect the behavior of
intel_hdmi_dpms and intel_dp_link_down on gen 4. In both functions the
code inside the HAS_PCH_IBX check is for IBX-specific workarounds, so
we should be safe. If we start bisecting gen 2/3/4 bugs to this commit
we should consider replacing the HAS_PCH_IBX checks with something
else.

V2: Improve commit message, list possible side effects and solution.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: prefer wide & slow to fast & narrow in DP configs
Jesse Barnes [Thu, 21 Jun 2012 22:13:50 +0000 (15:13 -0700)]
drm/i915: prefer wide & slow to fast & narrow in DP configs

High frequency link configurations have the potential to cause trouble
with long and/or cheap cables, so prefer slow and wide configurations
instead.  This patch has the potential to cause trouble for eDP
configurations that lie about available lanes, so if we run into that we
can make it conditional on eDP.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45801
Tested-by: peter@colberg.org
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix up ilk rc6 disabling confusion
Daniel Vetter [Fri, 29 Jun 2012 21:32:16 +0000 (23:32 +0200)]
drm/i915: fix up ilk rc6 disabling confusion

While creating the new enable/disable_gt_powersave functions in

commit 8090c6b9daa04dda649ac0a2209601042abfb0a4
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sun Jun 24 16:42:32 2012 +0200

    drm/i915: wrap up gt powersave enabling functions

I've botched up the handling of ironlake_disable_rc6. Fix this up by
calling it at the right place. Note though that ironlake_disable_rc6
does a bit more than just disabling rc6 - it also tears down all the
allocated context objects.

Hence we need to move intel_teardown_rc6 out and directly call it from
intel_modeset_cleanup.

Also properly mark ironlake_enable_rc6 as static and kill the un-used
declaration in i915_drv.h.

Note: In review a question popped out why disable_rc6 also tears down
the backing object and why we should move that out - it's simply for
consistency with gen6+ rps code, which does it that way.

Cc: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: move force wake support into intel_pm
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:11 +0000 (11:51 -0300)]
drm/i915: move force wake support into intel_pm

This commit moves force wake support routines into intel_pm modules, and
exports the gen6_gt_check_fifodbg routine (used in I915_READ).

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: enable RC6 workaround on Haswell
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:10 +0000 (11:51 -0300)]
drm/i915: enable RC6 workaround on Haswell

For Haswell, on some of the early hardware revisions, it is possible to
run into issues when RC6 state is enabled and when pipes change state.

v2: add comment saying that this is for early revisions only.

v3: beautify as suggested by Daniel Vetter.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: introduce haswell_init_clock_gating
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:09 +0000 (11:51 -0300)]
drm/i915: introduce haswell_init_clock_gating

This is based on Ivy Bridge clock gating for now, but is subject to
changes in the future.

Note: Compared to the ivb clock gating this drops the the IDICOS
medium uncore sharing tuned in

commit 208482232de3590cee4757dfabe5d8cee8c6e626
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Fri May 4 18:58:59 2012 -0700

    drm/i915: set IDICOS to medium uncore resources

Eugeni wants to benchmark the effect of this first.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
[danvet: added note]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: disable RC6 when disabling rps
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:08 +0000 (11:51 -0300)]
drm/i915: disable RC6 when disabling rps

We weren't disabling RC6 bits when bringing down RPS.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: enable RC6 by default on Haswell
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:07 +0000 (11:51 -0300)]
drm/i915: enable RC6 by default on Haswell

It should be working so let's turn it on by default and catch any possible
issues faster.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: slightly improve gt enable/disable routines
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:06 +0000 (11:51 -0300)]
drm/i915: slightly improve gt enable/disable routines

Just a cosmetic change to simplify the if statement.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: add RPS configuration for Haswell
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:05 +0000 (11:51 -0300)]
drm/i915: add RPS configuration for Haswell

Most of the RPS and RC6 enabling functionality is similar to what we had
on Gen6/Gen7, so we preserve most of the registers.

Note that Haswell only has RC6, so account for that as well. As suggested
by Daniel Vetter, to reduce the amount of changes in the patch, we still
write the RC6p/RC6pp thresholds, but those are ignored on Haswell.

Note: Some discussion about the nature of the new tuning constants
popped up in review - the answer is that we don't know why they've
changed, but the guide from VPG with the magic numbers simply has
different values now.

v2: Squash fix for ?: vs | operation precende bug into this patch.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Added note to commit message. Squashed fix.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: support Haswell force waking
Eugeni Dodonov [Mon, 2 Jul 2012 14:51:04 +0000 (11:51 -0300)]
drm/i915: support Haswell force waking

There is a different ACK register for force wake on Haswell, so account
for that.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>