Chris Wilson [Mon, 9 Dec 2013 10:37:24 +0000 (10:37 +0000)]
drm/i915: Repeat eviction search after idling the GPU
With the advent of hw context support, we gained some objects that are
pinned for the duration of their request. That is we can make aperture
space available by idling the GPU and in the process performing a
context switch back to the always-pinned default context. As such, we
should not conclude that there is no space in the aperture for the
current object until we have unpinned any such context objects.
Note that we also have the problem of outstanding pageflips preventing
eviction of their framebuffer objects to resolve.
Testcase: igt/gem_ctx_exec/eviction
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72507 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Tested-by: lu hua <huax.lu@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 5 Dec 2013 14:42:34 +0000 (15:42 +0100)]
drm/i915: Fix use-after-free in do_switch
So apparently under ridiculous amounts of memory pressure we can get
into trouble in do_switch when we try to move the old hw context
backing storage object onto the active lists.
With list debugging enabled that usually results in us chasing a
poisoned pointer - which means we've hit upon a vma that has been
removed from all lrus with list_del (and then deallocated, so it's a
real use-after free).
Ian Lister has done some great callchain chasing and noticed that we
can reenter do_switch:
i915_gem_do_execbuffer()
i915_switch_context()
do_switch()
from = ring->last_context;
i915_gem_object_pin()
i915_gem_object_bind_to_gtt()
ret = drm_mm_insert_node_in_range_generic();
// If the above call fails then it will try i915_gem_evict_something()
// If that fails it will call i915_gem_evict_everything() ...
i915_gem_evict_everything()
i915_gpu_idle()
i915_switch_context(DEFAULT_CONTEXT)
Like with everything else where the shrinker or eviction code can
invalidate pointers we need to reload relevant state.
Note that there's no need to recheck whether a context switch is still
required because:
- Doing a switch to the same context is harmless (besides wasting a
bit of energy).
- This can only happen with the default context. But since that one's
pinned we'll never call down into evict_everything under normal
circumstances. Note that there's a little driver bringup fun
involved namely that we could recourse into do_switch for the
initial switch. Atm we're fine since we assign the context pointer
only after the call to do_switch at driver load or resume time. And
in the gpu reset case we skip the entire setup sequence (which might
be a bug on its own, but definitely not this one here).
Cc'ing stable since apparently ChromeOS guys are seeing this in the
wild (and not just on artificial stress tests), see the reference.
Note that in upstream code doesn't calle evict_everything directly
from evict_something, that's an extension in this product branch. But
we can still hit upon this bug (and apparently we do, see the linked
backtraces). I've noticed this while trying to construct a testcase
for this bug and utterly failed to provoke it. It looks like we need
to driver the system squarly into the lowmem wall and provoke the
shrinker to evict the context object by doing the last-ditch
evict_everything call.
Aside: There's currently no means to get a badly-fragmenting hw
context object away from a bad spot in the upstream code. We should
fix this by at least adding some code to evict_something to handle hw
contexts.
References: https://code.google.com/p/chromium/issues/detail?id=248191 Reported-by: Ian Lister <ian.lister@intel.com> Cc: Ian Lister <ian.lister@intel.com> Cc: stable@vger.kernel.org Cc: Ben Widawsky <benjamin.widawsky@intel.com> Cc: Stéphane Marchesin <marcheu@chromium.org> Cc: Bloomfield, Jon <jon.bloomfield@intel.com> Tested-by: Rafael Barbalho <rafael.barbalho@intel.com> Reviewed-by: Ian Lister <ian.lister@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Fri, 6 Dec 2013 09:17:53 +0000 (10:17 +0100)]
drm/i915: fix pm init ordering
Shovel a bit more of the the code into the setup function, and call
it earlier. Otherwise lockdep is unhappy since we cancel the delayed
resume work before it's initialized.
While at it also shovel the pc8 setup code into the same functions.
I wanted to also ditch the header declaration of the hws pc8 functions,
but for unfathomable reasons that stuff is in intel_display.c instead
of intel_pm.c.
Chris Wilson [Wed, 4 Dec 2013 14:52:06 +0000 (14:52 +0000)]
drm/i915: Hold mutex across i915_gem_release
Inorder to serialise the closing of the file descriptor and its
subsequent release of client requests with i915_gem_free_request(), we
need to hold the struct_mutex in i915_gem_release(). Failing to do so
has the potential to trigger an OOPS, later with a use-after-free.
Testcase: igt/gem_close_race
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70874
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71029 Reported-by: Eric Anholt <eric@anholt.net> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Mon, 2 Dec 2013 17:39:09 +0000 (17:39 +0000)]
drm/i915: Do not clobber config status after a forced restore of hw state
We call intel_modeset_setup_hw_state() along two paths, driver
load/resume and after a lid event notification. During initialisation of
the driver, it is imperative that we reset the config state. This
correctly sets up the initial connector statuses and prepares the
hardware for a thorough probing. However, during a lid event, we only
want to undo the damage caused by the bios by resetting our last known
mode. In this cirumstance, we do not want to clobber our desired state.
In order to try and keep sanity between the config state and our own
tracking, do the drm_mode_config_reset() first along the load/resume
paths before reading out the hw state and apply any definite known
corrections.
v2: "As discussed on irc I don't think we should force the connector
state to anything here: Imo connector->status should reflect what we
believe to be the true output connection state, whereas connector->encoder
reflects whether this connector is wired up to a pipe. And since we no
longer reject modeset on disconnected connectors and never nuked the pipe
if the connector gets disconnected there's no reason for that - such policy
is userspace's job.
drm/i915/crt: explicitly set up HOTPLUG_BITS on resume"
so sayeth Daniel.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: stable@vger.kernel.org (v3.8 and later) Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ville Syrjälä [Mon, 2 Dec 2013 09:08:06 +0000 (11:08 +0200)]
drm/i915: Take modeset locks around intel_modeset_setup_hw_state()
Some lower level things get angry if we don't have modeset locks
during intel_modeset_setup_hw_state(). Actually the resume and
lid_notify codepaths alreday hold the locks, but the init codepath
doesn't, so fix that.
Note: This slipped through since we only disable pipes if the
plane/pipe linking doesn't match. Which is only relevant on older
gen3 mobile machines, if the BIOS fails to set up our preferred
linking.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: stable@vger.kernel.org Tested-and-reported-by: Paul Bolle <pebolle@tiscali.nl>
[danvet: Add note now that I could confirm my theory with the log
files Paul Bolle provided.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Fri, 29 Nov 2013 11:44:59 +0000 (11:44 +0000)]
drm/i915: Pin pages whilst allocating for dma-buf vmap()
During the vmap() routine for the dma-buf, we first grab the pages and
then try to allocate a temporary array to pass to the vmap(). However,
the shrinker can and will reap any object that is unbound if the
allocation for the array first fails. This includes the object which we
are attempting to vmap(). The solution is to mark the object's pages as
pinned whilst we try the allocation to prevent the use-after-free
introduced by the potential shrinkage.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ville Syrjälä [Fri, 29 Nov 2013 12:56:12 +0000 (14:56 +0200)]
drm/i915: MI_PREDICATE_RESULT_2 is HSW only
The MI_PREDICATE_RESULT_2 register exits only on HSW. On other
platforms the same offset is either reserved, or contains some
other register. So write the register only on HSW.
Ville Syrjälä [Fri, 29 Nov 2013 11:13:42 +0000 (13:13 +0200)]
drm/i915: Make the DERRMR SRM target global GTT
The ring scratch pages don't have a PPGTT mapping, so the DERRM SRM
should target the global GTT instead.
v2: Add MI_SRM_LRM_GLOBAL_GTT define for -fixes
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Paulo Zanoni [Thu, 21 Nov 2013 15:47:16 +0000 (13:47 -0200)]
drm/i915: use the correct force_wake function at the PC8 code
When I submitted the first patch adding these force wake functions,
Chris Wilson observed that I was using the wrong functions, so I sent
a second version of the patch to correct this problem. The problem is
that v1 was merged instead of v2.
I was able to notice the problem when running the
debugfs-forcewake-user subtest of pm_pc8 from intel-gpu-tools.
Cc: stable@vger.kernel.org Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ville Syrjälä [Thu, 28 Nov 2013 20:10:38 +0000 (22:10 +0200)]
drm/i915: Fix pipe CSC post offset calculation
We were miscalculating the pipe CSC post offset for the full->limited
range conversion. The resulting post offset was double what it was
supposed to be, which caused blacks to come out grey when using
limited range output on HSW+.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71769 Cc: stable@vger.kernel.org Tested-by: Lauri Mylläri <lauri.myllari@gmail.com> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ville Syrjälä [Fri, 1 Nov 2013 16:22:39 +0000 (18:22 +0200)]
drm/i915: Check VBT for eDP ports on VLV
VLV can have eDP on either port B or C, or even both. Based on the
VBT spec, intel_dpd_is_edp() should work on VLV too, assuming we
check the correct ports.
So instead of hardcoding port D, rename the function to
intel_dp_is_edp() and pass the port as a parameter, and use it
on VLV ports B and C.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71051 Tested-by: Robert Hooker <robert.hooker@canonical.com> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Wrestle the patch to apply and compile properly.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Jesse Barnes [Wed, 27 Nov 2013 19:10:26 +0000 (11:10 -0800)]
drm/i915: use crtc_htotal in watermark calculations to match fastboot v2
This value is more correct, and matches what we read out in the fastboot
code. Without this, the watermark code will panic after the first mode
setting activity after a fastboot.
v2: fix up HSW ->clock usage too (Ville)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Tue, 26 Nov 2013 11:23:15 +0000 (11:23 +0000)]
drm/i915: Pin relocations for the duration of constructing the execbuffer
As the execbuffer dispatch grows ever more complex and involves multiple
stages of moving objects into the aperture, we need to take greater care
that we do not evict our execbuffer objects prior to dispatch. This is
relatively simple as we can just keep the objects pinned for not just
the relocation but until we are finished.
One such example is the possibility of the context switch causing an
eviction or hitting the shrinker in order to fit its object into the
aperture.
Link: http://lists.freedesktop.org/archives/intel-gfx/2013-November/036166.html Reported-by: "Siluvery, Arun" <arun.siluvery@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ben Widawsky <benjamin.widawsky@intel.com> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: stable@vger.kernel.org Acked-by: Ben Widawsky <ben@bwidawsk.net> Tested-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Add the additional explanations from Chris to the commit
message.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
actually seems to prefer the default PTE encoding as age 3. Presumably,
this is in part due to the use of MOCS to override the PTE encodings
when appropriate.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=69870 Tested-by: mengmeng.meng@intel.com Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org> Reviewed-by: Eric Anholt <eric@anholt.net Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drm/i915/dp: set sink to power down mode on dp disable
but for DDI, where we've never done this.
Signed-off-by: Jani Nikula <jani.nikula@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Tue, 19 Nov 2013 02:32:37 +0000 (18:32 -0800)]
drm/i915: Do not enable package C8 on unsupported hardware
If the hardware does not support package C8, then do not even schedule
work to enable it. Thereby we can eliminate a bunch of dangerous work.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Tue, 19 Nov 2013 02:32:36 +0000 (18:32 -0800)]
drm/i915: Hold pc8 lock around toggling pc8.gpu_idle
We need to hold the pc8 lock around toggling the value of gpu_idle.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Mon, 18 Nov 2013 08:00:59 +0000 (09:00 +0100)]
drm/i915: encoder->get_config is no longer optional
We must have one to fill out the adjusted_mode.crtc_clock. And with
the tv encoder fixed up every encoder we have has a ->get_config
callback. So we can drop the checks.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Mon, 18 Nov 2013 08:00:58 +0000 (09:00 +0100)]
drm/i915/tv: add ->get_config callback
We need this to properly fill in adjusted_mode.crtc_clock, otherwise
the state checker gets unhappy. This seems to have been forgotten in
the big clock rework in
drm/i915: Fix port_clock and adjusted_mode.clock readout all over
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
To make this work both for resets and for BIOS takeover just add the
forcewake clearing call back to intel_uncore_early_sanitize.
We need to clear the forcewake in early sanitize so that the forcewak
dance in intel_uncore_init (to figure out whether we have mt or legacy
forcewake on ivb) works. That cleanup fits in nicely with the general
topic of early_sanitize to prepare for the very first mmio ops.
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reported-by: Jörg Otte <jrg.otte@gmail.com> Cc: Jörg Otte <jrg.otte@gmail.com>
References: https://lkml.org/lkml/2013/11/16/40 Cc: stable@vger.kernel.org (for 3.12 only) Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Mon, 4 Nov 2013 07:13:45 +0000 (08:13 +0100)]
drm/i915: flush cursors harder
Apparently they need the same treatment as primary planes. This fixes
modesetting failures because of stuck cursors (!) on Thomas' i830M
machine.
I've figured while at it I'll also roll it out for the ivb 3 pipe
version of this function. I didn't do this for i845/i865 since Bspec
says the update mechanism works differently, and there's some
additional rules about what can be updated in which order.
Tested-by: Thomas Richter <thor@math.tu-berlin.de> Cc: stable@vger.kernel.org Cc: Thomas Richter <thor@math.tu-berlin.de> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Duncan Laurie [Thu, 14 Nov 2013 01:59:43 +0000 (17:59 -0800)]
i915: Use 120MHz LVDS SSC clock for gen5/gen6/gen7
We had been using a DMI table workaround to select the right
frequency for devices, but this is fragile and must be updated
with every new platform.
Instead the default case when VBT is missing is changed to use
120MHz clock for LVDS SSC for these generations.
The docs for 2010-Core, SandyBridge, and IvyBridge all indicate
that the reference frequency for LVDS is 120MHz:
"2010 Core"
http://intellinuxgraphics.org/IHD_OS_Vol3_Part3r2.pdf
page 38
Reference Frequency: 120MHz for CRT and LVDS. 100MHz for the FDI.
"2011 SandyBridge"
http://intellinuxgraphics.org/documentation/SNB/IHD_OS_Vol3_Part3.pdf
page 33
Reference Frequency: 120MHz for CRT, HDMI, LVDS. 100MHz for the FDI.
"2012 IvyBridge"
http://intellinuxgraphics.org/documentation/IVB/IHD_OS_Vol3_Part4.pdf
page 27
Reference Frequency: 120 MHz for CRT, HDMI, LVDS, 100MHz for the FDI.
Signed-off-by: Duncan Laurie <dlaurie@chromium.org>
[olof: Fixup for recent base, switched from if/else to single call] Signed-off-by: Olof Johansson <olof@lixom.net> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Jani Nikula [Tue, 12 Nov 2013 15:10:13 +0000 (17:10 +0200)]
drm/i915/dp: set sink to power down mode on dp disable
We used to put the local sink and any downstream sinks to power down
mode at disable or dpms off using the DPCD SET_POWER register, until
this was broken by
Dave Airlie [Wed, 13 Nov 2013 23:53:15 +0000 (09:53 +1000)]
Merge branch 'drm-next-3.13' of git://people.freedesktop.org/~agd5f/linux into drm-next
One last patch I keep forgetting to include. Fix for EDID quirk
handling. Been on the list and reviewed for several months now,
I just keep forgetting about it.
* 'drm-next-3.13' of git://people.freedesktop.org/~agd5f/linux:
drm/edid: compare actual vrefresh for all modes for quirks
Dave Airlie [Wed, 13 Nov 2013 23:52:44 +0000 (09:52 +1000)]
Merge tag 'drm-intel-fixes-2013-11-12' of git://people.freedesktop.org/~danvet/drm-intel into drm-next
Just one patch to fix compile fail for CONFIG_ACPI=n. Figured I better
send this out quickly to minimize the broken build span. Otherwise no
bugfixes (besides some bdw stuff) anywhere in sight.
* tag 'drm-intel-fixes-2013-11-12' of git://people.freedesktop.org/~danvet/drm-intel:
drm/i915/opregion: fix build error on CONFIG_ACPI=n
Dave Airlie [Wed, 13 Nov 2013 23:51:43 +0000 (09:51 +1000)]
Merge branch 'vmwgfx-next-3.13' of git://people.freedesktop.org/~thomash/linux into drm-next
A resource eviction fix, and a fix for compilation / sparse problems
from the previous pull.
* 'vmwgfx-next-3.13' of git://people.freedesktop.org/~thomash/linux:
drm/vmwgfx: Fix a couple of compile / sparse warnings and errors
drm/vmwgfx: Resource evict fixes
Fix a long-standing TTM issue where we manipulated the vma page_prot
bits while mmap_sem was taken in read mode only. We now make a local
copy of the vma structure which we pass when we set the ptes.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Thomas Hellstrom [Tue, 12 Nov 2013 07:49:26 +0000 (23:49 -0800)]
drm/vmwgfx: Fix a couple of compile / sparse warnings and errors
Fixes
*) an implicit function declaration on mips,
*) a defined but not used label on !CONFIG_INTEL_IOMMU
*) Hopefully a couple of sparse warnings where we implicitly typecast
integer to __le32 and vice versa.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com> Reviewed-by: Brian Paul <brianp@vmware.com>
Alex Deucher [Thu, 15 Aug 2013 15:42:14 +0000 (11:42 -0400)]
drm/edid: compare actual vrefresh for all modes for quirks
The vrefresh field of the mode is 0 for most modes
fetched from the EDID (e.g., established timings).
When dealing with monitors that have a bogus preferred
mode, we may not always select the mode we want because
we compare the target refresh to the mode's vrefresh which
is 0 in a lot of cases.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Dave Airlie [Sun, 10 Nov 2013 21:25:33 +0000 (07:25 +1000)]
Merge branch 'gma500-next' of git://github.com/patjak/drm-gma500 into drm-next
SDVO support for minnowboard
* 'gma500-next' of git://github.com/patjak/drm-gma500:
drm/gma500/mrst: Add SDVO to output init
drm/gma500/mrst: Don't blindly guess a mode for LVDS
drm/gma500/mrst: Setup GMBUS for oaktrail/mrst
drm/gma500/mrst: Replace WMs and chickenbits with values from EMGD
drm/gma500/mrst: Add aux register writes to SDVO
drm/gma500/mrst: Properly route oaktrail hdmi hooks
drm/gma500/mrst: Add aux register writes when programming pipe
drm/gma500/mrst: Add SDVO clock calculation
drm/gma500: Add aux device support for gmbus
drm/gma500: Add support for aux pci vdc device
drm/gma500: Add chip specific sdvo masks
drm/gma500: Add Minnowboard to the IS_MRST() macro
Laurent Pinchart [Mon, 28 Oct 2013 22:49:22 +0000 (23:49 +0100)]
drm: shmob_drm: Convert to clk_prepare/unprepare
Turn clk_enable() and clk_disable() calls into clk_prepare_enable() and
clk_disable_unprepare() to get ready for the migration to the common
clock framework.
Cc: David Airlie <airlied@linux.ie> Cc: dri-devel@lists.freedesktop.org Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
Dave Airlie [Sun, 10 Nov 2013 08:35:33 +0000 (18:35 +1000)]
Merge tag 'bdw-stage1-2013-11-08-v2' of git://people.freedesktop.org/~danvet/drm-intel into drm-next
So here's the Broadwell pull request. From a kernel driver pov there's
two areas with big changes in Broadwell:
- Completely new enumerated interrupt bits. On the plus side it now looks
fairly unform and sane.
- Completely new pagetable layout.
To ensure minimal impact on existing platforms we've refactored both the
irq and low-level gtt handling code a lot in anticipation of the bdw push.
So now bdw enabling in these areas just plugs in a bunch of vfuncs.
Otherwise it's all fairly harmless adjusting of switch cases and
if-ladders to shovel bdw into the right blocks. So minimized impact on
existing platforms. I've also merged the bdw-stage1 branch into our
-nightly integration branch for the past week to make sure we don't break
anything.
Note that there's still quite a flurry or patches floating around, but
I've figured I'll push this out. I plan to keep the bdw fixes separate
from my usual -fixes stream so that you can reject them easily in case it
still looks like too much churn. Also, bdw is for now hidden behind the
preliminary hw enabling module option. So there's no real pressure to get
follow-up patches all into 3.13.
* tag 'bdw-stage1-2013-11-08-v2' of git://people.freedesktop.org/~danvet/drm-intel: (75 commits)
drm/i915: Mask the vblank interrupt on bdw by default
drm/i915: Wire up cpu fifo underrun reporting support for bdw
drm/i915: Optimize gen8_enable|disable_vblank functions
drm/i915: Wire up pipe CRC support for bdw
drm/i915: Wire up PCH interrupts for bdw
drm/i915: Wire up port A aux channel
drm/i915: Fix up the bdw pipe interrupt enable lists
drm/i915: Optimize pipe irq handling on bdw
drm/i915/bdw: Take render error interrupt out of the mask
drm/i915/bdw: Add BDW PCH check first
drm/i915: Use hsw_crt_get_config on BDW
drm/i915/bdw: Change dp aux timeout to 600us on DDIA
drm/i915/bdw: Enable trickle feed on Broadwell
drm/i915/bdw: WaSingleSubspanDispatchOnAALinesAndPoints
drm/i915/bdw: conservative SBE VUE cache mode
drm/i915/bdw: Limit SDE poly depth FIFO to 2
drm/i915/bdw: Sampler power bypass disable
ddrm/i915/bdw: Disable centroid pixel perf optimization
drm/i915/bdw: BWGTLB clock gate disable
drm/i915/bdw: Implement edp PSR workarounds
...
Dave Airlie [Sun, 10 Nov 2013 08:33:17 +0000 (18:33 +1000)]
Merge branch 'drm-next-3.13' of git://people.freedesktop.org/~agd5f/linux into drm-next
A few more patches for 3.13. The big one here is Hawaii support.
I wanted to get that out sooner, but was sick earlier this week. That
said, it's mostly self contained, so it shouldn't impact other asics.
The rest are just bug fixes and a merge fix.
* 'drm-next-3.13' of git://people.freedesktop.org/~agd5f/linux: (23 commits)
Revert "drm/radeon/audio: don't set speaker allocation on DCE4+"
drm/radeon/audio: improve ACR calculation
drm/radeon/audio: correct ACR table
drm/radeon: fix mismerge of drm-next with 3.12
drm/radeon: add pci ids for hawaii
drm/radeon: fill in radeon_asic_init for hawaii
drm/radeon: modesetting updates for hawaii
drm/radeon: atombios.h updates for hawaii
drm/radeon: update cik_get_csb_buffer for hawaii
drm/radeon: add hawaii dpm support
drm/radeon/cik: add hawaii UVD support
drm/radeon: update firmware loading for hawaii
drm/radeon: update rb setup for hawaii
drm/radeon: add golden register settings for hawaii
drm/radeon: update cik_tiling_mode_table_init() for hawaii
drm/radeon: minor updates to cik.c for hawaii
drm/radeon: update cik_gpu_init() for hawaii
drm/radeon: add Hawaii chip family
drm/radeon: fix-up some float to fixed conversion thinkos
drm/radeon: use HDP_MEM_COHERENCY_FLUSH_CNTL for sdma as well
...
Dave Airlie [Sun, 10 Nov 2013 08:27:31 +0000 (18:27 +1000)]
Merge branch 'msm-next' of git://people.freedesktop.org/~robclark/linux into drm-next
prime support, inactive rework, render nodes
* 'msm-next' of git://people.freedesktop.org/~robclark/linux:
drm/msm/mdp4: page_flip cleanups/fixes
drm/msm: EBUSY status handling in msm_gem_fault()
drm/msm: rework inactive-work
drm/msm: add plane support
drm/msm: resync generated headers
drm/msm: support render nodes
drm/msm: prime support
Jani Nikula [Fri, 8 Nov 2013 08:10:31 +0000 (10:10 +0200)]
drm/i915/opregion: fix build error on CONFIG_ACPI=n
Fix CONFIG_ACPI=n build fail
CC drivers/gpu/drm/i915/intel_opregion.o
drivers/gpu/drm/i915/intel_opregion.c: In function ‘intel_opregion_setup’:
drivers/gpu/drm/i915/intel_opregion.c:879:2: error: ‘asle_work’ undeclared (first use in this function)
drivers/gpu/drm/i915/intel_opregion.c:879:2: note: each undeclared identifier is reported only once for each function it appears in
make[4]: *** [drivers/gpu/drm/i915/intel_opregion.o] Error 1
drm/i915: move opregion asle request handling to a work queue
Reported-by: Jim Davis <jim.epost@gmail.com>
Reference: http://lkml.kernel.org/r/CA+r1ZhjcFpr5KKVX0pLCOP8cAyZoiYO=UyqYMJtNSV-Kt_p7xQ@mail.gmail.com Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's try this again for 3.13. It's required for proper
interaction with alsa. Was disabled previously in 3.12
to be on the safe side since it caused problems on older
asics.
Pierre Ossman [Wed, 6 Nov 2013 19:09:08 +0000 (20:09 +0100)]
drm/radeon/audio: improve ACR calculation
In order to have any realistic chance of calculating proper
ACR values, we need to be able to calculate both N and CTS,
not just CTS. We still aim for the ideal N as specified in
the HDMI spec though.
Pierre Ossman [Wed, 6 Nov 2013 19:00:32 +0000 (20:00 +0100)]
drm/radeon/audio: correct ACR table
The values were taken from the HDMI spec, but they assumed
exact x/1.001 clocks. Since we round the clocks, we also need
to calculate different N and CTS values.
Note that the N for 25.2/1.001 MHz at 44.1 kHz audio is out of
spec. Hopefully this mode is rarely used and/or HDMI sinks
tolerate overly large values of N.
Ilija Hadzic [Sun, 3 Nov 2013 03:00:20 +0000 (23:00 -0400)]
drm/radeon/kms: add crtc_disable function for legacy crtc
To plug the VRAM memory leak (see previous patch for
details) we must unpin the frame buffer when disabling the
CRTC. This warrants the addition of disable function for legacy
CRTC, which puts the CRTC in DPMS-OFF state and unpins the
frame buffer if there is one associated with the CRTC.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Ilija Hadzic [Sun, 3 Nov 2013 03:00:19 +0000 (23:00 -0400)]
drm/radeon/kms: unpin fb in atombios crtc disable
When drm_helper_disable_unused_functions calls disable
function of the CRTC, it also sets the crtc->fb pointer
to NULL. This can later (when the mode on that CRTC is setup
again from user space) cause ***_do_set_base functions to
"think" that there is no old buffer and skip the unpinning
code. Consequently, the buffer that has been NULL-ified in
drm_helper_disable_unused_functions will never be unpinned
causing a leak in VRAM.
This patch plugs the leak by unpinning the frame buffer
in crtc_disable function.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Anssi Hannula [Mon, 28 Oct 2013 23:19:16 +0000 (01:19 +0200)]
drm/radeon/audio: fix missing multichannel PCM SAD in some cases
The current code writing SADs to the audio registers seems to assume
that there is at most a single SAD per audio format.
However, that is not the case. Especially for PCM it is somewhat common
for sinks to have two SADs, one for 8-channel and one for 2-channel
audio, which may have different supported sample rates (i.e. the sink
supports stereo audio at higher sample rates than multichannel audio).
Because of this, only the 2-channel SAD may be used if it appears before
the 8-channel SAD. Unless other SADs require otherwise, this may cause
the ALSA HDA driver to allow stereo playback only.
Fix the code to pick the PCM SAD with the highest number of channels,
while merging the rate masks of PCM SADs with lower amount of channels
into the additional stereo rate mask byte.
Technically there are even more cases to handle (multiple non-PCM SADs
of the same type, more than two PCM SADs with varying channel counts,
etc), but those have not actually been encountered in the field and
handling them would be non-trivial.
Daniel Vetter [Thu, 7 Nov 2013 14:31:52 +0000 (15:31 +0100)]
drm/i915: Mask the vblank interrupt on bdw by default
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 7 Nov 2013 10:05:46 +0000 (11:05 +0100)]
drm/i915: Wire up cpu fifo underrun reporting support for bdw
HW engineers have listened and given us again a real interrupt with
masking and status regs. Yay!
For consistency with other platforms call the #define FIFO_UNDERRUN.
Eventually we also might need to have some enable/disable functions
for bdw display interrupts, but for now open-coding seems to be good
enough.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's cache the IMR value like on other platforms. This is needed to
implement the underrun reporting since then we'll have two places that
change the same register at runtime.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 7 Nov 2013 10:05:43 +0000 (11:05 +0100)]
drm/i915: Wire up PCH interrupts for bdw
Gives us hotplug, gmbus, dp aux and south errors (underrun
reporting!).
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Acked-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 7 Nov 2013 13:49:55 +0000 (14:49 +0100)]
drm/i915: Wire up port A aux channel
Useful for dp aux to work better. Also stop enabling the port A
hotplug event - eDP panels are expected to fire that interupt and
we're not really ready to deal with them. This is consistent with how
we handle port A on ilk-hsw.
The more important bit is that we must delay the enabling of hotplug
interrupts until all the encoders are fully set up. But we need irq
support earlier than that, hence hotplug interrupts can only be
enabled in the ->hpd_irq_setup callback.
v2: Drop the _HOTPLUG, it isn't (Ville).
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 7 Nov 2013 13:49:24 +0000 (14:49 +0100)]
drm/i915: Fix up the bdw pipe interrupt enable lists
- Pipe underrun can't just be enabled, we need some support code like
on ilk-hsw to make this happen. So drop it for now.
- CRC error is a special mode of the CRC hardware that we don't use,
so again drop it. Real CRC support for bdw will be added later.
- All the other error bits are about faults, so rename the #define and
adjust the output.
v2: Use pipe_name as pointed out by Ville. Ville's comment was on a
previous patch, but it was easier to squash in here.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Daniel Vetter [Thu, 7 Nov 2013 10:05:40 +0000 (11:05 +0100)]
drm/i915: Optimize pipe irq handling on bdw
We have a per-pipe bit in the master irq control register, so use it.
This allows us to drop the masks for aggregate interrupt bits and be a
bit more explicit in the code. It also removes one indentation level.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
v3: Update with workaround names.
Use for_each_pipe() and CHICKEN_PIPESL_1(pipe) macro (Ville)
Cc: Art Runyan <arthur.j.runyan@intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by: Jesse Barnes <jbarnes@virtuosugeek.org>
[danvet: Kill redundant IS_BDW check and remove the copious amount of
uneeded lines added.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky [Sun, 3 Nov 2013 04:07:52 +0000 (21:07 -0700)]
drm/i915/bdw: Create a separate BDW rps enable
This is mostly what we have for HSW with the exceptions of:
no writes:
GEN6_RC1_WAKE_RATE_LIMIT
GEN6_RC6pp_WAKE_RATE_LIMIT
GEN6_RC1e_THRESHOLD
GEN6_RC6p_THRESHOLD
GEN6_RC6pp_THRESHOLD
GEN6_RP_DOWN_TIMEOUT - use 1s instead of 1.28s
Don't try to overclock, or program ring/IA frequency tables since we
don't quite have sufficient docs yet.
NOTE: These values do not reflect the changes made recently by Chris.
Since we have no evidence yet what the proper way to tweak for this
platform is, I think it is good to go, and can be optimized by Chris, or
whomever, later.
Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Drop spurious hunk and drop TODO - having per-platform rps
register frobbing code is in my opinion preferred, now that all the
infrastructure functions are extracted.] Reviewed-by: Jesse Barnes <jbarnes@virtuosugeek.org> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky [Sun, 3 Nov 2013 04:07:50 +0000 (21:07 -0700)]
drm/i915/bdw: Don't wait for c0 threads on forcewake
It's no longer a required workaround on BDW.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Move compile fix from a later patch to this one.] Reviewed-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky [Sun, 3 Nov 2013 04:07:49 +0000 (21:07 -0700)]
drm/i915/bdw: Use HSW formula for ring freq scaling
The current formula we use for HSW is not what is in current docs.
However, changing to the HSW formula on my HSW does not improve power
usage, and decreases performance by about 5% in limited xonotic testing.
For gen8, until we know otherwise, or run experiments, let's use
the HSW formula - which should be the same used in the Windows driver
(and thus help make an apples-applies comparison) on gen8.
v2: Use >= 8 instead of > 7 to be consistent with all other gen
checks.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v1) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky [Tue, 5 Nov 2013 06:45:05 +0000 (22:45 -0800)]
drm/i915/bdw: Support eDP PSR
Broadwell PSR support is a superset of Haswell. With this simple
register base calculation, everything that worked on HSW for eDP PSR
should work on BDW.
Note that Broadwell provides additional PSR support. This is not
addressed at this time.
v2: Make the HAS_PSR include BDW
v3: Use the correct offset (I had incorrectly used one from my faulty
brain) (Art!)
v4: It helps if you git add
v5: Be explicit about not setting min link entry time for BDW. This
should be no functional change over v4 (Jani)
Reviewed-by: Art Runyan <arthur.j.runyan@intel.com> Reviewed-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ben Widawsky [Sun, 3 Nov 2013 04:07:47 +0000 (21:07 -0700)]
drm/i915/bdw: Use The GT mailbox for IPS enable/disable
v2: Squash in fixup from Ben to synchronize the GT mailbox commands.
CC: Art Runyan <arthur.j.runyan@intel.com> Reviewed-by: Art Runyan <arthur.j.runyan@intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ville Syrjälä [Sun, 3 Nov 2013 04:07:46 +0000 (21:07 -0700)]
drm/i915/bdw: Add Broadwell display FIFO limits
Broadwell has bigger display FIFOs than Haswell. Otherwise the
two are very similar.
v2: Fix FBC WM_LP shift for BDW
v3: Rebase on top of the big Haswell wm rework.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v2) Reviewed-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Paulo Zanoni [Sun, 3 Nov 2013 04:07:45 +0000 (21:07 -0700)]
drm/i915/bdw: check DPD on port D when setting the DDI buffers
Use the eDP values on platforms where port D is eDP. This doesn't
affect Haswell since it uses the same DDI buffer values for eDP and
DP.
Reviewed-by: Art Runyan <arthur.j.runyan@intel.com> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Paulo Zanoni [Sun, 3 Nov 2013 04:07:44 +0000 (21:07 -0700)]
drm/i915/bdw: BDW also has only 2 FDI lanes
So treat it like Haswell.
Reviewed-by: Art Runyan <arthur.j.runyan@intel.com> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>