GitHub/LineageOS/G12/android_kernel_amlogic_linux-4.9.git
11 years agodrm/i915: Break up the large vsnprintf() in print_error_buffers()
Chris Wilson [Sat, 29 Jun 2013 22:26:50 +0000 (23:26 +0100)]
drm/i915: Break up the large vsnprintf() in print_error_buffers()

So it appears that I have encountered some bogosity when trying to call
i915_error_printf() with many arguments from print_error_buffers(). The
symptom is that the vsnprintf parser tries to interpret an integer arg
as a character string, the resulting OOPS indicating stack corruption.
Replacing the single call with its 13 format specifiers and arguments
with multiple calls to i915_error_printf() worked fine. This patch goes
one step further and introduced i915_error_puts() to pass the strings
simply.

It may not fix the root cause, but it does prevent my box from dying and
I think helps make print_error_buffers() more friendly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66077
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Refactor the wait_rendering completion into a common routine
Chris Wilson [Sat, 29 Jun 2013 21:05:26 +0000 (22:05 +0100)]
drm/i915: Refactor the wait_rendering completion into a common routine

Harmonise the completion logic between the non-blocking and normal
wait_rendering paths, and move that logic into a common function.

In the process, we note that the last_write_seqno is by definition the
earlier of the two read/write seqnos and so all successful waits will
have passed the last_write_seqno. Therefore we can unconditionally clear
the write seqno and its domains in the completion logic.

v2: Add the missing ring parameter, because sometimes it is good to have
things compile.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Only clear write-domains after a successful wait-seqno
Chris Wilson [Fri, 28 Jun 2013 15:54:08 +0000 (16:54 +0100)]
drm/i915: Only clear write-domains after a successful wait-seqno

In the introduction of the non-blocking wait, I cut'n'pasted the wait
completion code from normal locked path. Unfortunately, this neglected
that the normal path returned early if the wait returned early. The
result is that read-only waits may return whilst the GPU is still
writing to the bo.

Fixes regression from
commit 3236f57a0162391f84b93f39fc1882c49a8998c7 [v3.7]
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Aug 24 09:35:09 2012 +0100

    drm/i915: Use a non-blocking wait for set-to-domain ioctl

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66163
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: correct intel_dp_get_config() function for DevCPT
Xiong Zhang [Fri, 28 Jun 2013 04:59:06 +0000 (12:59 +0800)]
drm/i915: correct intel_dp_get_config() function for DevCPT

On DevCPT, the control register for Transcoder DP Sync Polarity is
TRANS_DP_CTL, not DP_CTL.
Without this patch, Many call trace occur on CPT machine with DP monitor.
The call trace is like: *ERROR* mismatch in adjusted_mode.flags(expected X,found X)

v2: use intel-crtc to simple patch, suggested by Daniel.

Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
[danvet: Extend the encoder->get_config comment to specify that we now
also depend upon intel_encoder->base.crtc being correct. Also bikeshed
s/intel_crtc/crtc/.]
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=65287
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix hpd interrupt register locking
Daniel Vetter [Thu, 27 Jun 2013 15:52:15 +0000 (17:52 +0200)]
drm/i915: fix hpd interrupt register locking

Our interrupt handler (in hardirq context) could race with the timer
(in softirq context), hence we need to hold the spinlock around the
call to ->hdp_irq_setup in intel_hpd_irq_handler, too.

But as an optimization (and more so to clarify things) we don't need
to do the irqsave/restore dance in the hardirq context.

Note also that on ilk+ the race isn't just against the hotplug
reenable timer, but also against the fifo underrun reporting. That one
also modifies the SDEIMR register (again protected by the same
dev_priv->irq_lock).

To lock things down again sprinkle a assert_spin_locked. But exclude
the functions touching SDEIMR for now, I want to extract them all into
a new helper function (like we do already for pipestate, display
interrupts and all the various gt interrupts).

v2: Add the missing 't' Egbert spotted in a comment.

v3: Actually fix the right misspelled comment (Paulo).

Cc: Egbert Eich <eich@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fold the no-irq check into intel_hpd_irq_handler
Daniel Vetter [Thu, 27 Jun 2013 15:52:14 +0000 (17:52 +0200)]
drm/i915: fold the no-irq check into intel_hpd_irq_handler

The usual pattern for our sub-function irq_handlers is that they check
for the no-irq case themselves. This results in more streamlined code
in the upper irq handlers.

v2: Rebase on top of the i965g/gm sdvo hpd fix.

Cc: Egbert Eich <eich@suse.de>
Reviewed-by: Egbert Eich <eich@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fold the queue_work into intel_hpd_irq_handler
Daniel Vetter [Thu, 27 Jun 2013 15:52:13 +0000 (17:52 +0200)]
drm/i915: fold the queue_work into intel_hpd_irq_handler

Everywhere the same.

Note that this patch leaves unnecessary braces behind, but the next
patch will kill those all anyway (including the if itself) so I've
figured I can keep the diff a bit smaller.

v2: Rebase on top of the i965g/gm sdvo hpd fix.

Cc: Egbert Eich <eich@suse.de>
Reviewed-by: Egbert Eich <eich@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fold the hpd_irq_setup call into intel_hpd_irq_handler
Daniel Vetter [Thu, 27 Jun 2013 15:52:12 +0000 (17:52 +0200)]
drm/i915: fold the hpd_irq_setup call into intel_hpd_irq_handler

We already have a vfunc for this (and other parts of the hpd storm
handling code already use it).

v2: Rebase on top of the i965g/gm sdvo hpd fix.

Cc: Egbert Eich <eich@suse.de>
Reviewed-by: Egbert Eich <eich@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: s/hotplug_irq_storm_detect/intel_hpd_irq_handler/
Daniel Vetter [Thu, 27 Jun 2013 15:52:11 +0000 (17:52 +0200)]
drm/i915: s/hotplug_irq_storm_detect/intel_hpd_irq_handler/

The combination of Paulo's fifo underrun detection code and Egbert's
hpd storm handling code unfortunately made the hpd storm handling code
racy.

To avoid duplicating tricky interrupt locking code over all platforms
start with a bit of refactoring. This patch is the very first step
since in the end the irq storm handling code will handle all hotplug
logic (and so also encapsulate the locking nicely).

v2: Rebase on top of the i965g/gm sdvo hpd fix.

Cc: Egbert Eich <eich@suse.de>
Reviewed-by: Egbert Eich <eich@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: close tiny race in the ilk pcu even interrupt setup
Daniel Vetter [Thu, 27 Jun 2013 11:44:59 +0000 (13:44 +0200)]
drm/i915: close tiny race in the ilk pcu even interrupt setup

By the time we write DEIER in the postinstall hook the interrupt
handler could run any time. And it does modify DEIER to handle
interrupts.

Hence the DEIER read-modify-write cycle for enabling the PCU event
source is racy. Close this races the same way we handle vblank
interrupts: Unconditionally enable the interrupt in the IER register,
but conditionally mask it in IMR. The later poses no such race since
the interrupt handler does not touch DEIMR.

Also update the comment, the clearing has already happened
unconditionally above.

v2: Actually shove the updated comment into the right train^W commit,
as spotted by Paulo.

Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix locking around ironlake_enable|disable_display_irq
Daniel Vetter [Thu, 27 Jun 2013 11:44:58 +0000 (13:44 +0200)]
drm/i915: fix locking around ironlake_enable|disable_display_irq

The haswell unclaimed register handling code forgot to take the
spinlock. Since this is in the context of the non-rentrant interupt
handler and we only have one interrupt handler it is sufficient to
just grab the spinlock - we do not need to exclude any other
interrupts from running on the same cpu.

To prevent such gaffles in the future sprinkle assert_spin_locked over
these functions. Unfornately this requires us to hold the spinlock in
the ironlake postinstall hook where it is not strictly required:
Currently that is run in single-threaded context and with userspace
exlcuded from running concurrent ioctls. Add a comment explaining
this.

v2: ivb_can_enable_err_int also needs to be protected by the spinlock.
To ensure this won't happen in the future again also sprinkle a
spinlock assert in there.

v3: Kill the 2nd call to ivb_can_enable_err_int I've accidentally left
behind, spotted by Paulo.

Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Paulo Zanoni <przanoni@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Fix context sizes on HSW
Ben Widawsky [Wed, 26 Jun 2013 04:53:40 +0000 (21:53 -0700)]
drm/i915: Fix context sizes on HSW

With updates to the spec, we can actually see the context layout, and
how many dwords are allocated. That table suggests we need 70720 bytes
per HW context. Rounded up, this is 18 pages. Looking at what lives
after the current 4 pages we use, I can't see too much important (mostly
it's d3d related), but there are a couple of things which look scary. I
am hopeful this can explain some of our odd HSW failures.

v2: Make the context only 17 pages. The power context space isn't used
ever, and execlists aren't used in our driver, making the actual total
66944 bytes.

v3: Add a comment to the code. (Jesse & Paulo)

Reported-by: "Azad, Vinit" <vinit.azad@intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Fix VLV sprite register offsets
Ville Syrjälä [Tue, 25 Jun 2013 11:16:35 +0000 (14:16 +0300)]
drm/i915: Fix VLV sprite register offsets

We forgot to add VLV_DISPLAY_BASE to the VLV sprite registers, which
caused the sprites to not work at all.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agoRevert "drm/i915: Don't use the HDMI port color range bit on Valleyview"
Ville Syrjälä [Tue, 25 Jun 2013 11:16:34 +0000 (14:16 +0300)]
Revert "drm/i915: Don't use the HDMI port color range bit on Valleyview"

The PIPECONF color range bit doesn't appear to be effective, on HDMI
outputs at least. The color range bit in the port register works though,
so let's use it.

I have not yet verified whether the PIPECONF bit works on DP outputs.

This reverts commit 83a2af88f80ebf8104c9e083b786668b00f5b9ce.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: s/LFP/LPF in DPIO PLL register names
Ville Syrjälä [Fri, 14 Jun 2013 11:02:53 +0000 (14:02 +0300)]
drm/i915: s/LFP/LPF in DPIO PLL register names

LPF is short for "low pass filter".

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Fix VLV PLL LPF coefficients for DAC
Ville Syrjälä [Fri, 14 Jun 2013 11:02:52 +0000 (14:02 +0300)]
drm/i915: Fix VLV PLL LPF coefficients for DAC

The current PLL settings produce a rather unstable picture when
I hook up a VLV to my HP ZR24w display via a VGA cable.

According to VLV2A0_DP_eDP_HDMI_DPIO_driver_vbios_notes_9, we should
use the the same LPF coefficients for DAC as we do for HDMI and RBR DP.
And indeed that seems to cure the shivers.

v2: Add the name of the relevant document to the commit message

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Jump to at least RPe on VLV when increasing the GPU frequency
Ville Syrjälä [Tue, 25 Jun 2013 18:38:11 +0000 (21:38 +0300)]
drm/i915: Jump to at least RPe on VLV when increasing the GPU frequency

If the current GPU frquency is below RPe, and we're asked to increase
it, just go directly to RPe. This should provide better performance
faster than letting the frequency trickle up in response to the up
threshold interrupts.

For now just do it for VLV, since that matches quite closely how VLV
used to operate when the rps delayed timer kept things at RPe always.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Don't increase the GPU frequency from the delayed VLV rps timer
Ville Syrjälä [Tue, 25 Jun 2013 18:38:10 +0000 (21:38 +0300)]
drm/i915: Don't increase the GPU frequency from the delayed VLV rps timer

There's little point in increasing the GPU frequency from the delayed
rps work on VLV. Now when the GPU is idle, the GPU frequency actually
keeps dropping gradually until it hits the minimum, whereas previously
it just ping-ponged constantly between RPe and RPe-1.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: GEN6_RP_INTERRUPT_LIMITS doesn't seem to exist on VLV
Ville Syrjälä [Tue, 25 Jun 2013 16:21:06 +0000 (19:21 +0300)]
drm/i915: GEN6_RP_INTERRUPT_LIMITS doesn't seem to exist on VLV

I can't find GEN6_RP_INTERRUPT_LIMITS (0xA014) anywhere in VLV docs.
Reading it always returns zero from what I can tell, and eliminating
it doesn't seem to make any difference to the behaviour of the system.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Make the rps new_delay comparison more readable
Ville Syrjälä [Tue, 25 Jun 2013 16:21:05 +0000 (19:21 +0300)]
drm/i915: Make the rps new_delay comparison more readable

Eliminate the weird inverted logic from the rps new_delay comparison.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Don't wait for Punit after each freq change on VLV
Ville Syrjälä [Tue, 25 Jun 2013 16:21:02 +0000 (19:21 +0300)]
drm/i915: Don't wait for Punit after each freq change on VLV

It seems that even though Punit reports the frequency change to have
been completed, it still reports the old frequency in the status
register for some time.

So rather than polling for Punit to complete the frequency change after
each request, poll before. This gets rid of the spurious "Punit overrode
GPU freq" messages.

This also lets us continue working while Punit is performing the actual
frequency change. As a result, openarena demo088-test1 timedemo average
fps is increased by ~5 fps, and the slowest frame duration is reduced
by ~25%.

The sysfs cur_freq file always reads the current frequency from Punit
anyway, so having rps.cur_delay be slightly off at times doesn't matter.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Clean up VLV rps code a bit
Ville Syrjälä [Tue, 25 Jun 2013 16:21:01 +0000 (19:21 +0300)]
drm/i915: Clean up VLV rps code a bit

Always print both the MHz value and raw register value for rps stuff.

Also kill a somewhat pointless local 'rpe' variable and just use
dev_priv->rps.rpe_delay.

While at it clean up the caps in "GPU" and "Punit" debug messages.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Detect invalid scanout pitches
Chris Wilson [Tue, 25 Jun 2013 16:26:45 +0000 (17:26 +0100)]
drm/i915: Detect invalid scanout pitches

Report back the user error of attempting to setup a CRTC with an invalid
framebuffer pitch. This is trickier than it should be as on gen4, there
is a restriction that tiled surfaces must have a stride less than 16k -
which is less than the largest supported CRTC size.

v2: Fix the limits for gen3
v3: Move check into intel_framebuffer_init() and fix VLV limits. (vsyrjala)
v4: Use idiomatic '>=' for generation checks

References: https://bugs.freedesktop.org/show_bug.cgi?id=65099
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Remove duplicated WaForceL3Serialization:vlv
Ville Syrjälä [Tue, 25 Jun 2013 13:38:21 +0000 (16:38 +0300)]
drm/i915: Remove duplicated WaForceL3Serialization:vlv

No need to apply WaForceL3Serialization:vlv twice.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't scream into dmesg when a modeset fails
Daniel Vetter [Tue, 25 Jun 2013 09:06:52 +0000 (11:06 +0200)]
drm/i915: don't scream into dmesg when a modeset fails

There are legit cases, e.g. when userspace asks for something
impossible. So tune it down to debug output like we do with all other
userspace-triggerable warnings.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66111#c5
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Rebased.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Fix up sdvo hpd pins for i965g/gm
Daniel Vetter [Mon, 24 Jun 2013 19:33:28 +0000 (21:33 +0200)]
drm/i915: Fix up sdvo hpd pins for i965g/gm

Bspec seems to be full of lies, at least it disagress with reality:
Two systems corrobated that SDVO hpd bits are the same as on gen3.

v2: Update comment a bit.

Cc: Arthur Ranyan <arthur.j.runyan@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Reported-and-tested-by: Alex Fiestas <afiestas@kde.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=58405
Cc: stable@vger.kernel.org
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Introduce an HAS_IPS() macro
Damien Lespiau [Mon, 24 Jun 2013 17:29:34 +0000 (18:29 +0100)]
drm/i915: Introduce an HAS_IPS() macro

Follow the trend and don't code conditions with platforms but with
features.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix build warning on format specifier mismatch
Jani Nikula [Fri, 7 Jun 2013 13:03:50 +0000 (16:03 +0300)]
drm/i915: fix build warning on format specifier mismatch

drivers/gpu/drm/i915/i915_gem.c: In function ‘i915_gem_object_bind_to_gtt’:
drivers/gpu/drm/i915/i915_gem.c:3002:3: warning: format ‘%ld’ expects
argument of type ‘long int’, but argument 5 has type ‘size_t’ [-Wformat]

v2: Use %zu instead of %d. Two char patch, and 100% wrong. (Ville)

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: tune down DIDL warning about too many outputs
Daniel Vetter [Mon, 24 Jun 2013 16:32:36 +0000 (18:32 +0200)]
drm/i915: tune down DIDL warning about too many outputs

Nothing the user (nor we) really can do about this, but upsets a nice
quiet boot.

Note that this happens mostly on SDVs where OEMs obviously haven't had
a chance yet to appropriately trim the output list.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=65988
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Amend commit message a bit to clarify a question from Paulo.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: make compact dma scatter lists creation work with SWIOTLB backend.
Konrad Rzeszutek Wilk [Mon, 24 Jun 2013 15:47:48 +0000 (11:47 -0400)]
drm/i915: make compact dma scatter lists creation work with SWIOTLB backend.

Git commit 90797e6d1ec0dfde6ba62a48b9ee3803887d6ed4
("drm/i915: create compact dma scatter lists for gem objects") makes
certain assumptions about the under laying DMA API that are not always
correct.

On a ThinkPad X230 with an Intel HD 4000 with Xen during the bootup
I see:

[drm:intel_pipe_set_base] *ERROR* pin & fence failed
[drm:intel_crtc_set_config] *ERROR* failed to set mode on [CRTC:3], err = -28

Bit of debugging traced it down to dma_map_sg failing (in
i915_gem_gtt_prepare_object) as some of the SG entries were huge (3MB).

That unfortunately are sizes that the SWIOTLB is incapable of handling -
the maximum it can handle is a an entry of 512KB of virtual contiguous
memory for its bounce buffer. (See IO_TLB_SEGSIZE).

Previous to the above mention git commit the SG entries were of 4KB, and
the code introduced by above git commit squashed the CPU contiguous PFNs
in one big virtual address provided to DMA API.

This patch is a simple semi-revert - were we emulate the old behavior
if we detect that SWIOTLB is online. If it is not online then we continue
on with the new compact scatter gather mechanism.

An alternative solution would be for the the '.get_pages' and the
i915_gem_gtt_prepare_object to retry with smaller max gap of the
amount of PFNs that can be combined together - but with this issue
discovered during rc7 that might be too risky.

Reported-and-Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
CC: Chris Wilson <chris@chris-wilson.co.uk>
CC: Imre Deak <imre.deak@intel.com>
CC: Daniel Vetter <daniel.vetter@ffwll.ch>
CC: David Airlie <airlied@linux.ie>
CC: <dri-devel@lists.freedesktop.org>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Fix PCH detect with multiple ISA bridges in VM
Rui Guo [Wed, 19 Jun 2013 13:10:23 +0000 (21:10 +0800)]
drm/i915: Fix PCH detect with multiple ISA bridges in VM

In some virtualized environments (e.g. XEN), there is irrelevant ISA bridge in
the system. To work reliably, we should scan trhough all the ISA bridge
devices and check for the first match, instead of only checking the first one.

Signed-off-by: Rui Guo <firemeteor@users.sourceforge.net>
[danvet: Fixup conflict with the num_pch_pll removal. And add
subsystem header to the commit message headline.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: rename intel_dp_destroy to intel_dp_connector_destroy
Paulo Zanoni [Wed, 12 Jun 2013 20:27:30 +0000 (17:27 -0300)]
drm/i915: rename intel_dp_destroy to intel_dp_connector_destroy

Because it's the function that destroys the connector, not the
encoder. And we already have intel_dp_encoder_destroy.

This has annoyed me for a long time.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: check the return value of intel_dp_i2c_init
Paulo Zanoni [Wed, 12 Jun 2013 20:27:28 +0000 (17:27 -0300)]
drm/i915: check the return value of intel_dp_i2c_init

We've been ignoring this return value, so print a nice backtrace in
case it's not what we expected.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix the "ghost eDP" encoder unwind path
Paulo Zanoni [Wed, 12 Jun 2013 20:27:27 +0000 (17:27 -0300)]
drm/i915: fix the "ghost eDP" encoder unwind path

Because calling intel_dp_encoder_destroy inside
intel_edp_init_connector is just wrong. This is the initialization
path, so we should properly unwind all the initialization through the
whole caller stack.

On the intel_dp_encoder_destroy function we do the following:
1 - Call i2c_del_adapter
2 - Call drm_encoder_cleanup
3 - If edp:
3.1 - Cancel panel_vdd_work
3.2 - Call ironlake_panel_vdd_of_sync
4 - Free the encoder

And here is how we unwind each specific step:
1 - We have intel_dp_init_connector -> intel_dp_i2c_init ->
    i2c_dp_aux_add_bus -> i2c_add_adapter, so we call
    i2c_del_dapter at intel_dp_init_connector
2 - Call it in the same function that called drm_encoder_init
3 - Call it in the same function that called INIT_DELAYED_WORK
4 - Free it in the same function that allocated it

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: fix the "ghost eDP" connector unwind path
Paulo Zanoni [Wed, 12 Jun 2013 20:27:26 +0000 (17:27 -0300)]
drm/i915: fix the "ghost eDP" connector unwind path

Because calling intel_dp_destroy inside intel_edp_init_connector is
just wrong. This is the initialization path, so we should properly
unwind all the initialization through the whole caller stack.

On the intel_dp_destroy function we do the following:
1 - Free edid if it exists
2 - Call intel_panel_fini in case it's eDP
3 - Call drm_sysfs_connector_remove
4 - Call drm_connector_cleanup
5 - Free the connector

And here is how we unwind each specific step:
1 - No need as we still didn't assign anything
2 - No need as we still didn't call intel_panel_init
3 - Call it in the same function that called drm_sysfs_connector_add
4 - Call it in the same function that called drm_connector_init
5 - Free it in the same function that allocated it

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: propagate errors from intel_dp_init_connector
Paulo Zanoni [Wed, 12 Jun 2013 20:27:25 +0000 (17:27 -0300)]
drm/i915: propagate errors from intel_dp_init_connector

In case we detect a "ghost eDP", intel_edp_init_connector frees both
the connector and encoder and then returns. On Haswell, intel_ddi_init
then tries to use the freed encoder on the HDMI initialization path
since the following commit:

commit 21a8e6a4853b2ed39fa4c5188a710f2cf1b92026
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Apr 10 23:28:35 2013 +0200
    drm/i915: don't setup hdmi for port D edp in ddi_init

So now on intel_ddi_init we check for the "ghost eDP" case and return
without trying to initialize HDMI. This way we won't try to read the
freed "intel_encoder" struct in the next "if" statement.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: extract intel_edp_init_connector
Paulo Zanoni [Wed, 12 Jun 2013 20:27:24 +0000 (17:27 -0300)]
drm/i915: extract intel_edp_init_connector

Because intel_dp_init_connector is too big for my poor little brain.
No functional changes.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: don't check encoder at DP connector destroy()
Paulo Zanoni [Wed, 12 Jun 2013 20:27:23 +0000 (17:27 -0300)]
drm/i915: don't check encoder at DP connector destroy()

By the time we call intel_dp_destroy (which destroys the connector)
the encoder may have been destroyed already, so if we use it we may be
reading some free memory. That happens in drm_mode_config_cleanup()
and also inside intel_dp_init_connector() when we detect a ghost eDP.

I also hope this may solve some random memory bugs.

Reported by kmemcheck.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Zoltan Nyul <zoltan.nyul@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agoRevert "drm: kms_helper: don't lose hotplug event"
Dave Airlie [Fri, 28 Jun 2013 10:31:34 +0000 (20:31 +1000)]
Revert "drm: kms_helper: don't lose hotplug event"

This reverts commit 160954b7bca43da7cd3cfbce310e6df919a8216e.

This was rearming the workqueue with a 0 timeout, causing
a WARN_ON, and possible loop.

Daniel writes:
"I've looked a bit into this and I think we need to have a separate
work struct for recovering these lost hotplug events since the
continuous self-rearming case is a real risk (e.g. if a connector
flip-flops all the time). At least I don't see a sane way to block out
re-arming with the current code in a simple way. So reverting the
offender seems like the right thing and I'll go back to the drawing
board for 3.12."

Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/doc: Document the KMS property API
Laurent Pinchart [Sat, 22 Jun 2013 14:10:30 +0000 (16:10 +0200)]
drm/doc: Document the KMS property API

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/doc: Remove outdated note about i915 driver not behaving properly
Laurent Pinchart [Sat, 22 Jun 2013 14:10:29 +0000 (16:10 +0200)]
drm/doc: Remove outdated note about i915 driver not behaving properly

The i915 driver has been fixed not to modify the mode argument of the
encoder mode_fixup operation. Remove the related comment from the
documentation.

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: Improve manual IRQ installation documentation
Laurent Pinchart [Sat, 22 Jun 2013 12:10:59 +0000 (14:10 +0200)]
drm: Improve manual IRQ installation documentation

Define the rules for using irqs from drm drivers.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: add return check for dma_buf_fd
YoungJun Cho [Wed, 26 Jun 2013 01:21:42 +0000 (10:21 +0900)]
drm/prime: add return check for dma_buf_fd

The dma_buf_fd() can return error when it fails to prepare fd,
so the dma_buf needs to be put.

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: reorder drm_prime_add_buf_handle and remove prototype
Seung-Woo Kim [Wed, 26 Jun 2013 01:21:41 +0000 (10:21 +0900)]
drm/prime: reorder drm_prime_add_buf_handle and remove prototype

Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: fix to put an exported dma_buf for adding handle failure
YoungJun Cho [Wed, 26 Jun 2013 01:21:40 +0000 (10:21 +0900)]
drm/prime: fix to put an exported dma_buf for adding handle failure

When drm_prime_add_buf_handle() returns failure for an exported
dma_buf, the dma_buf was already allocated and its refcount was
increased, so it needs to be put.

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: support to cache mapping
Joonyoung Shim [Wed, 19 Jun 2013 06:03:05 +0000 (15:03 +0900)]
drm/prime: support to cache mapping

The drm prime also can support it like GEM CMA supports to cache
mapping. It doesn't allow multiple mappings for one attachment.

[airlied: rebased on top of other prime changes]
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: replace NULL with error value in drm_prime_pages_to_sg
YoungJun Cho [Mon, 24 Jun 2013 07:40:53 +0000 (16:40 +0900)]
drm/prime: replace NULL with error value in drm_prime_pages_to_sg

Instead of NULL, error value is casted with ERR_PTR() for
drm_prime_pages_to_sg() and IS_ERR_OR_NULL() macro is replaced
with IS_ERR() macro for drm_gem_map_dma_buf().

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: fix to check return of dma_map_sg in prime helper
YoungJun Cho [Mon, 24 Jun 2013 06:34:21 +0000 (15:34 +0900)]
drm/prime: fix to check return of dma_map_sg in prime helper

The dma_map_sg(), in map_dma_buf callback operation of prime helper,
can return 0 when it fails to map, so it needs to release related
resources.

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: fix not to assign error value to gem name
YoungJun Cho [Wed, 26 Jun 2013 23:58:33 +0000 (08:58 +0900)]
drm/gem: fix not to assign error value to gem name

If idr_alloc() is failed, obj->name can be error value. Also
it cleans up duplicated flink processing code.

This regression has been introduced in

commit 2e928815c1886fe628ed54623aa98d0889cf5509
Author: Tejun Heo <tj@kernel.org>
Date:   Wed Feb 27 17:04:08 2013 -0800

    drm: convert to idr_alloc()

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: stable@vger.kernel.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: add mutex lock when using drm_gem_mmap_obj
YoungJun Cho [Wed, 26 Jun 2013 23:39:58 +0000 (08:39 +0900)]
drm/gem: add mutex lock when using drm_gem_mmap_obj

The drm_gem_mmap_obj() has to be protected with dev->struct_mutex,
but some caller functions do not. So it adds mutex lock to missing
callers and adds assertion to check whether drm_gem_mmap_obj() is
called with mutex lock or not.

Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/omap: drop the !FB_OMAP2 dep
Dave Airlie [Fri, 28 Jun 2013 02:08:10 +0000 (12:08 +1000)]
drm/omap: drop the !FB_OMAP2 dep

This ends up causing circularity and really let people shoot themselves
in the foot.

Acked-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: get rid of ttm_bo_is_reserved
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:28 +0000 (13:48 +0200)]
drm/ttm: get rid of ttm_bo_is_reserved

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/vmwgfx: get rid of ttm_bo_is_reserved usage
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:27 +0000 (13:48 +0200)]
drm/vmwgfx: get rid of ttm_bo_is_reserved usage

Use lockdep_assert_held instead.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: get rid of ttm_bo_is_reserved usage
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:26 +0000 (13:48 +0200)]
drm/radeon: get rid of ttm_bo_is_reserved usage

Try to use lockdep_assert_held or other alternatives where possible.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: get rid of ttm_bo_is_reserved usage
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:25 +0000 (13:48 +0200)]
drm/ttm: get rid of ttm_bo_is_reserved usage

Use lockdep_assert_held instead.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: inline ttm_bo_reserve and related calls
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:24 +0000 (13:48 +0200)]
drm/ttm: inline ttm_bo_reserve and related calls

Makes lockdep a lot more useful.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: inline reservations
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:23 +0000 (13:48 +0200)]
drm/radeon: inline reservations

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/mgag200: inline reservations
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:22 +0000 (13:48 +0200)]
drm/mgag200: inline reservations

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/cirrus: inline reservations
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:21 +0000 (13:48 +0200)]
drm/cirrus: inline reservations

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ast: inline reservations
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:20 +0000 (13:48 +0200)]
drm/ast: inline reservations

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: convert to the reservation api
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:19 +0000 (13:48 +0200)]
drm/ttm: convert to the reservation api

Now that the code is compatible in semantics, flip the switch.
Use ww_mutex instead of the homegrown implementation.

ww_mutex uses -EDEADLK to signal that the caller has to back off,
and -EALREADY to indicate this buffer is already held by the caller.

ttm used -EAGAIN and -EDEADLK for those, respectively. So some changes
were needed to handle this correctly.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: make flipping lockdep safe
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:18 +0000 (13:48 +0200)]
drm/nouveau: make flipping lockdep safe

cli->mutex was inverted with reservations, and multiple reservations were
used without a ticket, fix both. This commit had to be done after the previous
commit, because otherwise ttm_eu_* calls would use a different seqno counter..

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: make ttm reservation calls behave like reservation calls
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:17 +0000 (13:48 +0200)]
drm/ttm: make ttm reservation calls behave like reservation calls

This commit converts the source of the val_seq counter to
the ww_mutex api. The reservation objects are converted later,
because there is still a lockdep splat in nouveau that has to
resolved first.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoreservation: cross-device reservation support, v4
Maarten Lankhorst [Thu, 27 Jun 2013 11:48:16 +0000 (13:48 +0200)]
reservation: cross-device reservation support, v4

This adds support for a generic reservations framework that can be
hooked up to ttm and dma-buf and allows easy sharing of reservations
across devices.

The idea is that a dma-buf and ttm object both will get a pointer
to a struct reservation_object, which has to be reserved before
anything is done with the contents of the dma-buf.

Changes since v1:
 - Fix locking issue in ticket_reserve, which could cause mutex_unlock
   to be called too many times.
Changes since v2:
 - All fence related calls and members have been taken out for now,
   what's left is the bare minimum to be useful for ttm locking conversion.
Changes since v3:
 - Removed helper functions too. The documentation has an example
   implementation for locking. With the move to ww_mutex there is no
   need to have much logic any more.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ast: do not attempt to acquire a reservation while in an interrupt handler
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:26 +0000 (13:38 +0200)]
drm/ast: do not attempt to acquire a reservation while in an interrupt handler

Mutexes should not be acquired in interrupt context. While the trylock
fastpath is arguably safe on all implementations, the slowpath
unlock path definitely isn't.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/mgag200: do not attempt to acquire a reservation while in an interrupt handler
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:25 +0000 (13:38 +0200)]
drm/mgag200: do not attempt to acquire a reservation while in an interrupt handler

Mutexes should not be acquired in interrupt context. While the trylock
fastpath is arguably safe on all implementations, the slowpath
unlock path definitely isn't.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/cirrus: do not attempt to acquire a reservation while in an interrupt handler
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:24 +0000 (13:38 +0200)]
drm/cirrus: do not attempt to acquire a reservation while in an interrupt handler

Mutexes should not be acquired in interrupt context. While the trylock
fastpath is arguably safe on all implementations, the slowpath
unlock path definitely isn't. This fixes the following lockdep splat:

[   13.044313] ------------[ cut here ]------------
[   13.044367] WARNING: at /c/kernel-tests/src/tip/kernel/mutex.c:858 mutex_trylock+0x87/0x220()
[   13.044378] DEBUG_LOCKS_WARN_ON(in_interrupt())
[   13.044378] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.10.0-rc4-00296-ga2963dd #20
[   13.044379] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
[   13.044390]  0000000000000009 ffff88000de039f8 ffffffff81fc86d5 ffff88000de03a38
[   13.044395]  ffffffff810d511b ffff880000000018 ffff88000f33c690 0000000000000001
[   13.044398]  00000000000003f0 ffff88000f4677c8 0000000000000000 ffff88000de03a98
[   13.044400] Call Trace:
[   13.044412]  <IRQ>  [<ffffffff81fc86d5>] dump_stack+0x19/0x1b
[   13.044441]  [<ffffffff810d511b>] warn_slowpath_common+0x6b/0x90
[   13.044445]  [<ffffffff810d51a6>] warn_slowpath_fmt+0x46/0x50
[   13.044448]  [<ffffffff81fd34d7>] mutex_trylock+0x87/0x220
[   13.044482]  [<ffffffff8186484d>] cirrus_dirty_update+0x1cd/0x330
[   13.044486]  [<ffffffff818649e8>] cirrus_imageblit+0x38/0x50
[   13.044506]  [<ffffffff8165782e>] soft_cursor+0x22e/0x240
[   13.044510]  [<ffffffff81656c31>] bit_cursor+0x581/0x5b0
[   13.044525]  [<ffffffff815de9f4>] ? vsnprintf+0x124/0x670
[   13.044529]  [<ffffffff81651333>] ? get_color.isra.16+0x43/0x130
[   13.044532]  [<ffffffff81653fca>] fbcon_cursor+0x18a/0x1d0
[   13.044535]  [<ffffffff816566b0>] ? update_attr.isra.2+0xa0/0xa0
[   13.044556]  [<ffffffff81754b82>] hide_cursor+0x32/0xa0
[   13.044565]  [<ffffffff81755bd3>] vt_console_print+0x103/0x3b0
[   13.044569]  [<ffffffff810d58ac>] ? print_time+0x9c/0xb0
[   13.044576]  [<ffffffff810d5960>] ? print_prefix+0xa0/0xc0
[   13.044580]  [<ffffffff810d63f6>] call_console_drivers.constprop.6+0x146/0x1f0
[   13.044593]  [<ffffffff815f9b38>] ? do_raw_spin_unlock+0xc8/0x100
[   13.044597]  [<ffffffff810d6f27>] console_unlock+0x2f7/0x460
[   13.044600]  [<ffffffff810d787a>] vprintk_emit+0x59a/0x5e0
[   13.044615]  [<ffffffff81fb676c>] printk+0x4d/0x4f
[   13.044650]  [<ffffffff82ba5511>] print_local_APIC+0x28/0x41c
[   13.044672]  [<ffffffff8114db55>] generic_smp_call_function_single_interrupt+0x145/0x2b0
[   13.044688]  [<ffffffff8106f9e7>] smp_call_function_single_interrupt+0x27/0x40
[   13.044697]  [<ffffffff81fd8f72>] call_function_single_interrupt+0x72/0x80
[   13.044707]  <EOI>  [<ffffffff81078166>] ? native_safe_halt+0x6/0x10
[   13.044717]  [<ffffffff811425cd>] ? trace_hardirqs_on+0xd/0x10
[   13.044738]  [<ffffffff8104f669>] default_idle+0x59/0x120
[   13.044742]  [<ffffffff810501e8>] arch_cpu_idle+0x18/0x40
[   13.044754]  [<ffffffff811320c5>] cpu_startup_entry+0x235/0x410
[   13.044763]  [<ffffffff81f9e781>] rest_init+0xd1/0xe0
[   13.044766]  [<ffffffff81f9e6b5>] ? rest_init+0x5/0xe0
[   13.044778]  [<ffffffff82b93ec2>] start_kernel+0x425/0x493
[   13.044781]  [<ffffffff82b93810>] ? repair_env_string+0x5e/0x5e
[   13.044786]  [<ffffffff82b93595>] x86_64_start_reservations+0x2a/0x2c
[   13.044789]  [<ffffffff82b93688>] x86_64_start_kernel+0xf1/0x100
[   13.044799] ---[ end trace 113ad28772af4058 ]---

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: always select ACPI_VIDEO if ACPI is enabled.
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:23 +0000 (13:38 +0200)]
drm/nouveau: always select ACPI_VIDEO if ACPI is enabled.

Having nouveau builtin would still allow ACPI_VIDEO to be used as external module
if some of the deps for acpi_video have not been met, which would result in a linking
failure. Solve this by selecting all dependencies as well.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: complain loudly if buffer is pinned during destruction
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:22 +0000 (13:38 +0200)]
drm/nouveau: complain loudly if buffer is pinned during destruction

Shouldn't happen, and we invert the struct_mutex with reservation here,
potentially leading to deadlocks. Once reservations become lockdep annotated,
lockdep will go splat on this.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: fixup fbcon failure paths
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:21 +0000 (13:38 +0200)]
drm/nouveau: fixup fbcon failure paths

Add missing calls, and fix a leak from forgetting to call the unpin function.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: unpin notify object in chan_fini
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:20 +0000 (13:38 +0200)]
drm/nouveau: unpin notify object in chan_fini

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/nouveau: implement prime helper unpin function
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:19 +0000 (13:38 +0200)]
drm/nouveau: implement prime helper unpin function

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: implement unpin function, v2
Maarten Lankhorst [Thu, 27 Jun 2013 11:38:18 +0000 (13:38 +0200)]
drm/radeon: implement unpin function, v2

Changes since v1:
- Fixup compiler warning in unpin function.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'drm-next-3.11' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 28 Jun 2013 00:53:23 +0000 (10:53 +1000)]
Merge branch 'drm-next-3.11' of git://people.freedesktop.org/~agd5f/linux into drm-next

This is the pull request for radeon for 3.11.  Highlights include:

- Support for CIK (Sea Islands) asics: 3D, compute, UVD
- DPM (Dynamic Power Management) support for 6xx-SI
- ASPM support for 6xx-SI
- Assorted bug fixes

* 'drm-next-3.11' of git://people.freedesktop.org/~agd5f/linux: (168 commits)
  drm/radeon/SI: fix TDP adjustment in set_power_state
  drm/radeon/NI: fix TDP adjustment in set_power_state
  drm/radeon: fix endian issues in atombios dpm code
  drm/radeon/dpm: fix UVD clock setting on SI
  drm/radeon/dpm: fix UVD clock setting on cayman
  drm/radeon/dpm: add support for setting UVD clock on rv6xx
  drm/radeon/dpm: add support for setting UVD clock on rs780
  drm/radeon: fix typo in ni_print_power_state
  drm/radeon: fix typo in cik_select_se_sh()
  drm/radeon/si: fix typo in function name
  drm/radeon/dpm: fix typo in setting uvd clock
  drm/radeon/dpm: add dpm_set_power_state failure output (si)
  add dpm_set_power_state failure output (7xx-ni)
  drm/radeon/dpm: add dpm_set_power_state failure output (7xx-ni)
  drm/radeon/dpm: add dpm_enable failure output (si)
  drm/radeon/dpm: add dpm_enable failure output (7xx-ni)
  drm/radeon/kms: add dpm support for SI (v7)
  drm/radeon: switch SI to use radeon_ucode.h
  drm/radeon: add SI to r600_is_internal_thermal_sensor()
  drm/radeon/dpm/rs780: properly catch errors in dpm setup
  ...

11 years agoMerge tag 'drm-intel-next-2013-06-18' of git://people.freedesktop.org/~danvet/drm...
Dave Airlie [Thu, 27 Jun 2013 23:50:34 +0000 (09:50 +1000)]
Merge tag 'drm-intel-next-2013-06-18' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

Last 3.11 feature pull. I have a few odds bits and pieces and fixes in my
queue, I'll sort them out later on to see what's for 3.11-fixes and what's
for 3.12. But nothing to hold this here up imo.

Highlights:
- more hangcheck work from Mika and Chris to prepare for arb robustness
- trickle feed fixes from Ville
- first parts of the shared pch pll rework, with some basic hw state
  readout and cross-checking (this shuts up the confused pch pll refcount
  WARN that Linus just recently forwarded)
- Haswell audio power well support from Wang Xingchao (alsa bits acked by
  Takashi)
- some cleanups and asserts sprinkling around the plane/gamma enabling
  sequence from Ville
- more gtt refactoring from Ben
- clear up the adjusted->mode vs. pixel clock vs. port clock confusion
- 30bpp support, this time for real hopefully

* tag 'drm-intel-next-2013-06-18' of git://people.freedesktop.org/~danvet/drm-intel: (97 commits)
  drm/i915: remove a superflous semi-colon
  drm/i915: Kill useless "Enable panel fitter" comments
  drm/i915: Remove extra "ring" from error message
  drm/i915: simplify the reduced clock handling for pch plls
  drm/i915: stop killing pfit on i9xx
  drm/i915: explicitly set up PIPECONF (and gamma table) on haswell
  drm/i915: set up PIPECONF explicitly for i9xx/vlv platforms
  drm/i915: set up PIPECONF explicitly on ilk-ivb
  drm/i915: find guilty batch buffer on ring resets
  drm/i915: store ring hangcheck action
  drm/i915: add batch bo to i915_add_request()
  drm/i915: change i915_add_request to macro
  drm/i915: add i915_gem_context_get_hang_stats()
  drm/i915: add struct i915_ctx_hang_stats
  drm/i915: Try harder to disable trickle feed on VLV
  drm/i915: fix up pch pll enabling for pixel multipliers
  drm/i915: hw state readout and cross-checking for shared dplls
  drm/i915: WARN on lack of shared dpll
  drm/i915: split up intel_modeset_check_state
  drm/i915: extract readout_hw_state from setup_hw_state
  ...

Conflicts:
drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_fb.c
drivers/gpu/drm/i915/intel_sdvo.c

11 years agodrm/radeon/SI: fix TDP adjustment in set_power_state
Alex Deucher [Thu, 27 Jun 2013 23:37:12 +0000 (19:37 -0400)]
drm/radeon/SI: fix TDP adjustment in set_power_state

Fixes hangs with DPM in some cases.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/NI: fix TDP adjustment in set_power_state
Alex Deucher [Thu, 27 Jun 2013 23:08:23 +0000 (19:08 -0400)]
drm/radeon/NI: fix TDP adjustment in set_power_state

Fixes hangs with DPM in some cases.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix endian issues in atombios dpm code
Alex Deucher [Mon, 24 Jun 2013 14:54:16 +0000 (10:54 -0400)]
drm/radeon: fix endian issues in atombios dpm code

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: fix UVD clock setting on SI
Alex Deucher [Tue, 14 May 2013 22:24:34 +0000 (18:24 -0400)]
drm/radeon/dpm: fix UVD clock setting on SI

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: fix UVD clock setting on cayman
Alex Deucher [Tue, 14 May 2013 22:21:17 +0000 (18:21 -0400)]
drm/radeon/dpm: fix UVD clock setting on cayman

The rv770 version was using the wrong power state type.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add support for setting UVD clock on rv6xx
Alex Deucher [Tue, 14 May 2013 22:12:13 +0000 (18:12 -0400)]
drm/radeon/dpm: add support for setting UVD clock on rv6xx

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add support for setting UVD clock on rs780
Alex Deucher [Tue, 14 May 2013 21:55:03 +0000 (17:55 -0400)]
drm/radeon/dpm: add support for setting UVD clock on rs780

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix typo in ni_print_power_state
Alex Deucher [Tue, 14 May 2013 14:35:05 +0000 (10:35 -0400)]
drm/radeon: fix typo in ni_print_power_state

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix typo in cik_select_se_sh()
Alex Deucher [Thu, 18 Apr 2013 20:25:47 +0000 (16:25 -0400)]
drm/radeon: fix typo in cik_select_se_sh()

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/si: fix typo in function name
Alex Deucher [Wed, 17 Apr 2013 20:27:40 +0000 (16:27 -0400)]
drm/radeon/si: fix typo in function name

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: fix typo in setting uvd clock
Alex Deucher [Wed, 3 Apr 2013 19:03:17 +0000 (15:03 -0400)]
drm/radeon/dpm: fix typo in setting uvd clock

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add dpm_set_power_state failure output (si)
Alex Deucher [Thu, 27 Jun 2013 23:33:58 +0000 (19:33 -0400)]
drm/radeon/dpm: add dpm_set_power_state failure output (si)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoadd dpm_set_power_state failure output (7xx-ni)
Alex Deucher [Thu, 27 Jun 2013 23:04:16 +0000 (19:04 -0400)]
add dpm_set_power_state failure output (7xx-ni)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add dpm_set_power_state failure output (7xx-ni)
Alex Deucher [Thu, 28 Mar 2013 14:46:29 +0000 (10:46 -0400)]
drm/radeon/dpm: add dpm_set_power_state failure output (7xx-ni)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add dpm_enable failure output (si)
Alex Deucher [Thu, 28 Mar 2013 14:45:50 +0000 (10:45 -0400)]
drm/radeon/dpm: add dpm_enable failure output (si)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm: add dpm_enable failure output (7xx-ni)
Alex Deucher [Thu, 28 Mar 2013 14:44:28 +0000 (10:44 -0400)]
drm/radeon/dpm: add dpm_enable failure output (7xx-ni)

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/kms: add dpm support for SI (v7)
Alex Deucher [Tue, 25 Jun 2013 21:56:16 +0000 (17:56 -0400)]
drm/radeon/kms: add dpm support for SI (v7)

This adds dpm support for SI asics.  This includes:
- dynamic engine clock scaling
- dynamic memory clock scaling
- dynamic voltage scaling
- dynamic pcie gen1/gen2/gen3 switching
- power containment
- shader power scaling

Set radeon.dpm=1 to enable.

v2: enable hainan support, rebase
v3: guard acpi stuff
v4: fix 64 bit math
v5: fix 64 bit div harder
v6: fix thermal interrupt check noticed by Jerome
v7: attempt fix state enable

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoMerge branch 'drm/for-next' of git://anongit.freedesktop.org/tegra/linux into drm...
Dave Airlie [Thu, 27 Jun 2013 23:29:04 +0000 (09:29 +1000)]
Merge branch 'drm/for-next' of git://anongit.freedesktop.org/tegra/linux into drm-next

These changes are mostly minor fixes to things introduced in 3.10. The
biggest chunk is updates to the host1x firewall which checks job
submissions from userspace and wasn't working properly. All other
patches are mostly one-liners. Nothing new or too exciting this time
around.

* 'drm/for-next' of git://anongit.freedesktop.org/tegra/linux:
  gpu: host1x: Rework CPU syncpoint increment
  gpu: host1x: Fix client_managed type
  gpu: host1x: Fix memory access in syncpt request
  gpu: host1x: Copy gathers before verification
  gpu: host1x: Don't reset firewall between gathers
  gpu: host1x: Check reloc table before usage
  gpu: host1x: Check INCR opcode correctly
  drm/tegra: Remove DRIVER_BUS_PLATFORM from driver_features
  drm/tegra: Fix return value
  drm/tegra: Include header drm/drm.h
  MAINTAINERS: Update Tegra DRM entry
  drm/tegra: fix error return code in gr2d_submit()
  drm/tegra: fix missing unlock on error
  drm/tegra: Honor pixel-format changes
  drm/tegra: Explicitly set irq_enabled
  drm/tegra: Don't disable unused planes

11 years agodrm/radeon: switch SI to use radeon_ucode.h
Alex Deucher [Wed, 27 Mar 2013 19:18:04 +0000 (15:18 -0400)]
drm/radeon: switch SI to use radeon_ucode.h

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add SI to r600_is_internal_thermal_sensor()
Alex Deucher [Tue, 26 Mar 2013 23:25:06 +0000 (19:25 -0400)]
drm/radeon: add SI to r600_is_internal_thermal_sensor()

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm/rs780: properly catch errors in dpm setup
Alex Deucher [Tue, 26 Mar 2013 23:23:19 +0000 (19:23 -0400)]
drm/radeon/dpm/rs780: properly catch errors in dpm setup

We weren't properly catching errors in dpm_enable()
and dpm_set_power_state().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm/r6xx: properly catch errors in dpm setup
Alex Deucher [Tue, 26 Mar 2013 23:18:46 +0000 (19:18 -0400)]
drm/radeon/dpm/r6xx: properly catch errors in dpm setup

We weren't properly catching errors in dpm_enable()
and dpm_set_power_state().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm/r7xx: properly catch errors in dpm setup
Alex Deucher [Tue, 26 Mar 2013 23:09:18 +0000 (19:09 -0400)]
drm/radeon/dpm/r7xx: properly catch errors in dpm setup

We weren't properly catching errors in dpm_enable()
and dpm_set_power_state().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm/trinity: properly catch errors in dpm setup
Alex Deucher [Tue, 26 Mar 2013 23:01:05 +0000 (19:01 -0400)]
drm/radeon/dpm/trinity: properly catch errors in dpm setup

We weren't properly catching errors in dpm_enable()
and dpm_set_power_state().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/dpm/sumo: properly catch errors in dpm setup
Alex Deucher [Tue, 26 Mar 2013 22:55:59 +0000 (18:55 -0400)]
drm/radeon/dpm/sumo: properly catch errors in dpm setup

We weren't properly catching errors in dpm_enable()
and dpm_set_power_state().

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>