drm/i915: check that rpm ref is held when accessing ringbuf in stolen mem
While running some tests on the scheduler patches with rpm enabled I
came across a corruption in the ringbuffer, which was root-caused to
the GPU being suspended while commands were being emitted to the
ringbuffer. The access to memory was failing because the GPU needs to
be awake when accessing stolen memory (where my ringbuffer was located).
Since we have this constraint it looks like a sensible idea to check
that we hold a refcount when we access the rungbuffer.
v2: move the check from ring_begin to ringbuffer iomap time (Chris)
v3: update comment (Chris)
Daniel Vetter [Wed, 27 Jan 2016 13:37:59 +0000 (14:37 +0100)]
drm/i915: Stop depending upon CONFIG_AGP_INTEL
The AGP_INTEL driver provides an interface for very old userspace to
control the GART (though the GART itself was only ever emulated on Intel
systems). The pci bridge discovery code is also used by the i915.ko
driver to set up the GTT on old systems, but it does not require the
old userspace interface. When i915.ko selects the old interface, it
binds another user to the core GTT routines, and in particular creates a
second reference to the scratch pages allocated. This hinders resource
leak debugging for when we unload i915.ko as we want to assert that all
DMA pages have been released, but we appear to leak because of the
secondary interface which persists after i915.ko unloads.
All i915.ko users do not require the old /dev/agpgart interface so stop
selecting it and simplify our debugging by dropping the historical
baggage.
Note that by selecting AGP=n it was already possible to unselect
AGP_INTEL. But since we've dropped support for any of the AGP stuff
long ago there's really no point for this any more.
Also note that we still need INTEL_GTT, which is the underlying,
shared, driver for the graphics GART on gen1-5.
v2: Entirely new commit message (Chris, Ville).
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1453901881-26425-2-git-send-email-daniel.vetter@ffwll.ch
Daniel Vetter [Wed, 27 Jan 2016 13:37:58 +0000 (14:37 +0100)]
agp/intel-gtt: Don't leak the scratch page
Recently discovered by enabling CONFIG_DMA_API_DEBUG in our CI. By the
looks of it broken since forever.
v2: Don't forget to set the scratch page back to wb (Chris). Reuse
intel_gtt_teardown_scratch_page for that (and fix it up to treat
needs_dmar y/n correctly).
Nick Hoath [Fri, 22 Jan 2016 23:10:06 +0000 (23:10 +0000)]
drm/i915: fix context/engine cleanup order
Swap the order of context & engine cleanup, so that contexts are cleaned
up first, and *then* engines. This is a more sensible order anyway, but
in particular has become necessary since the 'intel_ring_initialized()
must be simple and inline' patch, which now uses ring->dev as an
'initialised' flag, so it can now be NULL after engine teardown. This
in turn can cause a problem in the context code, which (used to) check
the ring->dev->struct_mutex -- causing a fault if ring->dev was NULL.
Also rename the cleanup function to reflect what it actually does
(cleanup engines, not a ringbuffer), and fix an annoying whitespace issue.
v2: Also make the fix in i915_load_modeset_init, not just in
i915_driver_unload (Chris Wilson)
v3: Had extra stuff in it.
v4: Reverted extra stuff (so we're back to v2).
Rebased and updated commentary above (Dave Gordon).
Signed-off-by: Nick Hoath <nicholas.hoath@intel.com> Signed-off-by: David Gordon <david.s.gordon@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Dave Gordon <david.s.gordon@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/1453504211-7982-2-git-send-email-david.s.gordon@intel.com
Gabriel Feceoru [Fri, 22 Jan 2016 11:28:45 +0000 (13:28 +0200)]
drm/i915: Handle PipeC fused off on IVB/HSW/BDW
Some Gen7/8 production parts may have the Display Pipe C fused off.
In this case, the display hardware will prevent the enable bit in
PIPE_CONF register (for Pipe C) from being set to 1.
Fixed by adjusting pipe_count to reflect this.
v2: Rename HSW_PIPE_C_DISABLE to IVB_PIPE_C_DISABLE as it already exists
on ivybridge (Ville)
v3: Remove unnecessary MMIO read, correct the description (Damien)
v4: Be more specific in description (Patrick)
Lyude [Thu, 4 Feb 2016 15:43:21 +0000 (10:43 -0500)]
drm/i915/skl: Fix typo in DPLL_CFGCR1 definition
We accidentally point both cfgcr registers for the second shared DPLL to
the same location in i915_reg.h. This results in a lot of hw pipe state
mismatches whenever we try to do a modeset that requires allocating the
DPLL to a CRTC:
[drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in dpll_hw_state.cfgcr1 (expected 0x80000168, found 0x000004a5)
[drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in base.adjusted_mode.crtc_clock (expected 108000, found 49500)
[drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in port_clock (expected 108000, found 49500)
This usually ends up causing blank monitors, since the DPLL never can
get set to the right clock.
Lyude [Tue, 2 Feb 2016 15:49:43 +0000 (10:49 -0500)]
drm/i915/skl: Don't skip mst encoders in skl_ddi_pll_select()
We don't actually check for INTEL_OUTPUT_DP_MST at all in here, as a
result we skip assigning a DPLL to any DP MST ports, which makes link
training fail:
[ 1442.933896] [drm:intel_power_well_enable] enabling DDI D power well
[ 1442.933905] [drm:skl_set_power_well] Enabling DDI D power well
[ 1442.933957] [drm:intel_mst_pre_enable_dp] 0
[ 1442.935474] [drm:intel_dp_set_signal_levels] Using signal levels 00000000
[ 1442.935477] [drm:intel_dp_set_signal_levels] Using vswing level 0
[ 1442.935480] [drm:intel_dp_set_signal_levels] Using pre-emphasis level 0
[ 1442.936190] [drm:intel_dp_set_signal_levels] Using signal levels 05000000
[ 1442.936193] [drm:intel_dp_set_signal_levels] Using vswing level 1
[ 1442.936195] [drm:intel_dp_set_signal_levels] Using pre-emphasis level 1
[ 1442.936858] [drm:intel_dp_set_signal_levels] Using signal levels 08000000
[ 1442.936862] [drm:intel_dp_set_signal_levels] Using vswing level 2
…
[ 1442.998253] [drm:intel_dp_link_training_clock_recovery [i915]] *ERROR* too many full retries, give up
[ 1442.998512] [drm:intel_dp_start_link_train [i915]] *ERROR* failed to train DP, aborting
After which the pipe state goes completely out of sync:
[ 70.075596] [drm:check_crtc_state] [CRTC:25]
[ 70.075696] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in ddi_pll_sel (expected 0x00000000, found 0x00000001)
[ 70.075747] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in shared_dpll (expected -1, found 0)
[ 70.075798] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in dpll_hw_state.ctrl1 (expected 0x00000000, found 0x00000021)
[ 70.075840] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in dpll_hw_state.cfgcr1 (expected 0x00000000, found 0x80400173)
[ 70.075884] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in dpll_hw_state.cfgcr2 (expected 0x00000000, found 0x000003a5)
[ 70.075954] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in base.adjusted_mode.crtc_clock (expected 262750, found 72256)
[ 70.075999] [drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in port_clock (expected 540000, found 148500)
And if you're especially lucky, it keeps going downhill:
drm/i915: Do not disable cxsr when crtc is disabled.
It's safe to assume cxsr is already disabled when the crtc is off.
This prevents an unclaimed register warning when the required power
wells are not enabled.
This is a revert of commit 066cf55b9ce3 "drm/i915: Fix IPS related flicker".
intel_pre_disable_primary already handles this, and now everything
goes through the atomic path there's no need to try to disable ips twice.
drm/i915/bxt: Check BIOS RC6 setup before enabling RC6
RC6 setup is shared between BIOS and Driver. BIOS sets up subset of RC6
setup registers. If those are not setup Driver should not enable RC6.
For implementing this, driver can check RC_CTRL0 and RC_CTRL1 values
to know if BIOS has enabled HW/SW RC6.
This will also enable user to control RC6 using BIOS settings alone.
RC6 related instability can be avoided by disabling via BIOS settings
till driver fixes it.
v2: Had placed logic in gen8 function by mistake. Fixed it.
Ensuring RPM is not enabled in case BIOS disabled RC6.
v3: Need to disable RPM if RC6 is disabled due to BIOS settings. (Daniel)
Runtime PM enabling happens before gen9_enable_rc6.
Moved the updation of enable_rc6 parameter in intel_uncore_sanitize.
v4: Added elaborate check for BIOS RC6 setup. Prepared check_pctx for bxt.
(Imre)
v5: Caching reserved stolen base and size in the driver private data.
Reorganized RC6 setup check. Moved from gen9_enable_rc6 to
intel_uncore_sanitize. (Imre)
v6: Rebasing on the patch submitted by Imre that moves gem_init_stolen
earlier in the load.
v7: Removed PWRCTX_MAXCNT_VCSUNIT1 check as it applies to SKL. (Imre)
v8: Fixed formatting and checkpatch issues. Fixed functional issue where
RC6 ctx size check was missing. (Imre)
However, it seems there are real world devices out there that just
aren't spec compliant, and still work at HBR2 using TPS2. So reduce the
error message to debug logging.
Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com> Cc: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92932 Fixes: 1da7d7131c35 ("drm/i915: ignore link rate in TPS3 selection") Cc: drm-intel-fixes@lists.freedesktop.org Signed-off-by: Jani Nikula <jani.nikula@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1454667370-8001-2-git-send-email-jani.nikula@intel.com
Jani Nikula [Fri, 5 Feb 2016 10:16:09 +0000 (12:16 +0200)]
drm/i915/dp: abstract training pattern selection
Make it cleaner to add more checks in the function. No functional
changes.
Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com> Cc: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> Cc: drm-intel-fixes@lists.freedesktop.org # dependency on the next patch Signed-off-by: Jani Nikula <jani.nikula@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1454667370-8001-1-git-send-email-jani.nikula@intel.com
Paulo Zanoni [Fri, 29 Jan 2016 20:57:39 +0000 (18:57 -0200)]
drm/i915/fbc: unexport the HW level activation functions
The recent introduction of a new caller of dev_priv->fbc.deactivate()
is a good example of why we need unexport those functions. Anything
outside intel_fbc.c should only call the functions exported by
intel_fbc.c, so in order to enforce that, kill the function pointers
stored inside dev_priv->fbc and replace them with functions that can't
be called from outside intel_fbc.c.
This should make it much harder for new code to call these functions
from outside intel_fbc.c.
Paulo Zanoni [Fri, 29 Jan 2016 20:57:38 +0000 (18:57 -0200)]
drm/i915: don't deactivate FBC at skylake_disable_primary_plane
FBC is already deactivated at this point.
Besides, nothing should be calling these lower-level function
pointers. A few months ago, the only caller of
dev_priv->fbc.deactivate was intel_pipe_set_base_atomic(), which was
the kgdboc function. But the following commit added it to the SKL
function:
commit a8d201af68506b375b701d0d8dbe8487034256f2
Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Date: Thu Jan 7 11:54:11 2016 +0100
drm/i915: Use plane state for primary plane updates.
Tim Gore [Thu, 4 Feb 2016 11:49:34 +0000 (11:49 +0000)]
drm/i915: implement WaIncreaseDefaultTLBEntries
WaIncreaseDefaultTLBEntries increases the number of TLB
entries available for GPGPU workloads and gives significant
( > 10% ) performance gain for some OCL benchmarks.
Put this in a new function that can be a place for
workarounds that are GT related but not required per ring.
This function is called on driver load and also after a
reset and on resume, so it is safe for workarounds that get
clobbered in these situations. This function currently has
just this one workaround.
v2: This was originally split into 3 patches but following
review feedback was squashed into 1.
I have not incorporated some style comments from Chris
Wilson as I felt that after defining and intialising a
temporary variable and then adding an additional if block
to only write the register if the temporary variable had
been set, this didn't really give a net gain.
drm/i915: mark GEM object pages dirty when mapped & written by the CPU
introduced a check into i915_gem_object_get_dirty_pages() that returned
a NULL pointer when called with a bad object, one that was not backed by
shmemfs. This WARN was too strict as we can work on all struct page
backed objects, and resulted in a WARN + GPF for existing userspace. In
order to differentiate the various types of objects, add a new flags field
to the i915_gem_object_ops struct to describe their capabilities, with
the first flag being whether the object has struct pages.
v2: Drop silly const before an integer in the structure declaration.
Testcase: igt/gem_userptr_blits/relocations Reported-and-tested-by: Kristian Høgsberg Kristensen <krh@bitplanet.net> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: Kristian Høgsberg Kristensen <krh@bitplanet.net> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Dave Gordon <david.s.gordon@intel.com> Reviewed-by: Kristian Høgsberg Kristensen <krh@bitplanet.net> Tested-by: Michal Winiarski <michal.winiarski@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1453487551-16799-1-git-send-email-chris@chris-wilson.co.uk
Matt Roper [Wed, 3 Feb 2016 06:06:51 +0000 (22:06 -0800)]
drm/i915: Pretend cursor is always on for ILK-style WM calculations (v2)
Due to our lack of two-step watermark programming, our driver has
historically pretended that the cursor plane is always on for the
purpose of watermark calculations; this helps avoid serious flickering
when the cursor turns off/on (e.g., when the user moves the mouse
pointer to a different screen). That workaround was accidentally
dropped as we started working toward atomic watermark updates. Since we
still aren't quite there yet with two-stage updates, we need to
resurrect the workaround and treat the cursor as always active.
v2: Tweak cursor width calculations slightly to more closely match the
logic we used before the atomic overhaul began. (Ville)
Cc: simdev11@outlook.com Cc: manfred.kitzbichler@gmail.com Cc: drm-intel-fixes@lists.freedesktop.org Reported-by: simdev11@outlook.com Reported-by: manfred.kitzbichler@gmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93892 Fixes: 43d59eda1 ("drm/i915: Eliminate usage of plane_wm_parameters from ILK-style WM code (v2)") Signed-off-by: Matt Roper <matthew.d.roper@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1454479611-6804-1-git-send-email-matthew.d.roper@intel.com
Matt Roper [Thu, 28 Jan 2016 23:09:37 +0000 (15:09 -0800)]
drm/i915: Check DDI max lanes after applying BXT workaround
In commit bfb9faab8 we added a workaround for some BXT BIOS that fail to
properly initialize the DDI_A_4_LANES bit of the control register (4
lanes is the only valid configuration on BXT since there is no DDI E to
share with). A recent patch added some additional checks on this
register bit before the workaround gets applied; this breaks eDP on BXT
in some settings. Some minor code shuffling is all we need to restore
the workaround.
Matt Roper [Wed, 6 Jan 2016 17:53:41 +0000 (09:53 -0800)]
drm/i915/bxt: Don't save/restore eDP panel power during suspend (v3)
Our attempts save/restore panel power state in i915_suspend.c are
causing unclaimed register warnings on BXT since the registers for this
platform differ from older platforms.
The big hammer suspend/resume shouldn't be necessary for PP since the
connector/encoder hooks should already handle this. In theory we could
remove this for all platforms, but in practice it's likely that would
cause some regressions since older platforms with LVDS may have
incomplete PP handling. For now we'll leave the PCH save/restore alone
and change the non-PCH branch to only operate on gen <= 4 so that BXT
and future platforms aren't included.
v2: Typo fix: s/||/&&/
v3: Change non-PCH condition to a gen <= 4 test rather than listing
VLV/CHV/BXT as specific platforms to exclude; should be more
future-proof as we add new platforms. (Daniel)
Imre Deak [Fri, 29 Jan 2016 12:52:28 +0000 (14:52 +0200)]
drm/i915: Add debug info for failed MSI enabling
While not being able to enable MSI interrupts may be a normal
circumstance, for debugging it may still be a useful information, so
emit an info about this.
Imre Deak [Fri, 29 Jan 2016 12:52:26 +0000 (14:52 +0200)]
drm/i915: Sanity check DP AUX message buffer and size
While we are calling intel_dp_aux_transfer() with msg->size=0 whenever
msg->buffer is NULL, passing NULL to memcpy() is undefined according to
the ISO C standard. I haven't found any notes about this in the GNU C's
or the kernel's documentation of the function and can't imagine what it
would do with the NULL ptr. To better document this use of the
parameters it still make sense to add an explicit check for this to the
code.
Imre Deak [Thu, 28 Jan 2016 14:04:12 +0000 (16:04 +0200)]
drm/i915/bxt: update list of PCIIDs
Add PCIIDs for new versions of the SOC, based on BSpec. Also add the
name of the versions as code comment where this is available. The new
versions don't have any changes visible to the kernel driver.
Patrik Jakobsson [Wed, 20 Jan 2016 14:31:20 +0000 (15:31 +0100)]
drm/i915/skl/kbl: Add support for pipe fusing
On SKL and KBL we can have pipe A/B/C disabled by fuse settings. The
pipes must be fused in descending order (e.g. C, B+C, A+B+C). We simply
decrease info->num_pipes if we find a valid fused out config.
v2: Don't store the pipe disabled mask in device info (Damien)
v3: Don't check FUSE_STRAP register for pipe c disabled
Rodrigo Vivi [Mon, 1 Feb 2016 20:02:08 +0000 (12:02 -0800)]
drm/i915: Instrument PSR parameter for debuging with link standby x link off.
Unfortunately we don't know all panels and platforms out there and we
found internal prototypes without VBT proper set but where only
link in standby worked well.
So, before enable PSR by default let's instrument the PSR parameter
in a way that we can identify different panels out there that might
require or work better with link standby mode.
It is also useful to say that for backward compatibility I'm not
changing the meaning of this flag. So "0" still means disabled
and "1" means enabled with full support and maximum power savings.
v2: Use positive value instead of negative for different operation mode
as suggested by Daniel.
v3: As Paulo suggested use 2 to force link standby and 3 to force link
fully on. Also split the link_standby introduction in a separated patch.
v4: Use DRM_ERROR for link off request on platforms that don't support and
Remove the quirk promise.
Rodrigo Vivi [Mon, 1 Feb 2016 20:02:07 +0000 (12:02 -0800)]
drm/i915: Add PSR main link standby support back
Link standby support has been deprecated with 'commit 89251b177
("drm/i915: PSR: deprecate link_standby support for core platforms.")'
The reason for that is that main link in full off offers more power
savings and on HSW and BDW implementations on source side had known
bugs with link standby.
However that same HSD report only mentions BDW and HSW and tells that
a fix was going to new platforms. Since on Skylake link standby
didn't cause the bad blank flickering screens seen on HSW and BDW
let's respect VBT again for this and future platforms.
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Rodrigo Vivi [Mon, 1 Feb 2016 20:02:06 +0000 (12:02 -0800)]
drm/i915: PSR simplify port and link standby checks.
Current code not just block link_standby for non DDI platforms but also
block PSR from work on other ports B/C/D/E.
So, besides change any behaviour let's just fix the mess a bit here and
reuse HSW check to block the other ports and reduce the second if only to
link stadnby request.
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Rodrigo Vivi [Fri, 29 Jan 2016 22:44:59 +0000 (14:44 -0800)]
drm/i915: Sink CRC: tune down error message at stop to debug_kms.
When we stop the sink CRC calculation we wait a while until the counter
is reset to zero and return -ETIMEDOUT. However the sink crc was
calculated already by this point so we just ignore this return at
the main function.
So, let's also ignore the message and put it as a debug message instead
of an error one. The message might still be useful when debuging
test failures so we could be able to know something was not going so
well with sink crc stop.
Paulo Zanoni [Mon, 18 Jan 2016 17:56:58 +0000 (15:56 -0200)]
drm/i915/fbc: refactor some small functions called only once
The FBC fixes we've been doing in the last months required a lot of
refactor, so functions that were once big and called from different
spots are now small and called only once. IMHO now it's better to just
move the contents of these functions to their only callers since this
reduces the number of indirections while reading the code.
While at it, also improve the related comments a little bit.
Paulo Zanoni [Mon, 18 Jan 2016 17:45:56 +0000 (15:45 -0200)]
drm/i915/fbc: don't store/check a pointer to the FB
We already make sure we run intel_fbc_update_update during modesets
and page flips, and this function takes care of deactivating FBC, so
it shouldn't be possible for us to reach the condition we check at
intel_fbc_work_fn. So instead of grabbing framebuffer references and
adding a lot of code to track when we need to free them, just don't
track anything at all since we shouldn't need to.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:56 +0000 (11:35 -0200)]
drm/i915/fbc: call intel_fbc_pre_update earlier during page flips
Make sure we do the pre_update - which also deactivates FBC - before
we actually schedule the page flip, just to make sure we don't
flip to the new FB with FBC still activated for the previous FB.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:55 +0000 (11:35 -0200)]
drm/i915/fbc: don't store the fb_id on reg_params
We don't actually use fb_id anywhere. We already compare all
parameters that matter to the hardware: pixel format, stride,
fence_reg and ggtt_offset. The ID shouldn't make a difference.
Besides, we already update the FBC data at every modeset/flip, so this
can't change behind our backs.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:54 +0000 (11:35 -0200)]
drm/i915/fbc: don't print no_fbc_reason to dmesg
Our dmesg messages started being misleading after we converted to the
enable+activate model: we always print "Disabling FBC", even when
we're just deactivating it. So, for example, when I boot my machine
and do "dmesg | grep -i fbc", I see:
[drm:intel_fbc_enable] Enabling FBC on pipe A
[drm:set_no_fbc_reason] Disabling FBC: framebuffer not tiled or fenced
but then, if I read the debugfs file, I will see:
$ sudo cat i915_fbc_status
FBC enabled
Compressing: yes
so we can conclude that dmesg is misleading, since FBC is actually
enabled. What happened is that we deactivated FBC due to fbcon not
being tiled, but when we silently reactivated it when the display
manager started. We don't print activation messages since there may be
way too many of these operations per second during normal desktop
usage.
One possible solution would be to change set_no_fbc_reason to
correctly differentiate between disable and deactivation, but we
removed support from printing activation/deactivation messages in the
past because they were too frequent. So instead of doing this, let's
just not print anything on dmesg, and leave the debugfs file if the
user needs to investigate something. We already print when we enable
and disable FBC anyway on a given pipe, so this should already help
triaging bugs.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:52 +0000 (11:35 -0200)]
drm/i915/fbc: make FBC work with fastboot
Move intel_fbc_enable to a place where it is called regardless of the
"modeset" variable, and make sure intel_fbc_enable can be called
multiple times without intel_fbc_disable being called.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:51 +0000 (11:35 -0200)]
drm/i915/fbc: move intel_fbc_{enable, disable} call one level up
Instead of duplicating the calls for every platform, let's just put
them in the correct places inside intel_atomic_commit. This will also
make it easier for us to move the enable call in order to support
fasbtoot.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:50 +0000 (11:35 -0200)]
drm/i915/fbc: choose the new FBC CRTC during atomic check
This opens the possibility of implementing nicer schemes to choose the
CRTC, such as checking the amount of stolen memory available, or
choosing the best pipe on platforms that don't die FBC to pipe or
plane A.
This code was written for another refactor that I ended up discarding,
so I don't actually need it, but I figured this patch would be an
improvement on its own so I kept it on the series.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:49 +0000 (11:35 -0200)]
drm/i915: simplify struct drm_device access at intel_atomic_check()
We already have a dev variable, there's no need to access state->dev.
Also, I plan to add another dev_priv user here, so declare one for the
current user.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:48 +0000 (11:35 -0200)]
drm/i915/fbc: rewrite the multiple_pipes_ok() code for locking
Older FBC platforms have this restriction where FBC can't be enabled
if multiple pipes are enabled. In the current code, we disable FBC
before the second pipe becomes visible.
One of the problems with this code is that the current
multiple_pipes_ok() implementation just iterates through all CRTCs
looking at their states, but it doesn't make sure that the state
locks are grabbed. It also can't just grab the locks for every CRTC
since this would kill one of the biggest advantages of atomic
modesetting.
After the recent FBC changes, we now have the appropriate locks for
the given CRTC, so we can just try to maintain the state of each CRTC
and update it once intel_fbc_pre_update is called.
As a last note, I don't have gen 2/3 machines to test this code. My
current plan is to enable FBC on just the newer platforms, so this
patch is just an attempt to get the gen 2/3 code at least looking
sane, so if one day someone decide to fix FBC on these platforms, they
may have less work to do.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:47 +0000 (11:35 -0200)]
drm/i915/fbc: make sure we cancel the work function at fbc_disable
Just to be sure nothing will survive a module unload. We need to do
this after the unlock in order to make sure the function won't get
stuck trying to grab the lock we already own while we wait for it to
finish.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:45 +0000 (11:35 -0200)]
drm/i915/fbc: unexport intel_fbc_deactivate
With the addition and usage of intel_fbc_pre_update,
intel_fbc_deactivate is not used anymore outside intel_fbc.c, so kill
the exported function and rename __intel_fbc_deactivate.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:44 +0000 (11:35 -0200)]
drm/i915/fbc: fix the FBC state checking code
We'll now call intel_fbc_pre_update instead of intel_fbc_deactivate
during atomic commits. This will continue to guarantee that we
deactivate FBC and it will also update the state checking structures
at the correct time. Then, later, at the point where we were calling
intel_fbc_update, we'll only need to call intel_fbc_post_update.
Also add the proper warnings in case we don't have the appropriate
locks. Daniel mentioned the warnings will have to be removed for async
commits, but let's keep them here while we can.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:43 +0000 (11:35 -0200)]
drm/i915/fbc: split intel_fbc_update into pre and post update
So now pre_update will be responsible for unconditionally deactivating
FBC and updating the state cache, while post_update will be
responsible for checking if it can be enabled, then enabling it.
This is one more step into proper locking.
Notice that intel_fbc_flush now calls post_update directly. The FBC
flush can only happen for drawing operations - since we explicitly
ignore the flips -, so the FBC state is not expected to have changed
at this point. With this we can just run post_update, which will make
sure we won't deactivate+reactivate FBC as would be the case now if we
called pre_update + post_update.
Per the new atomic locking rules, we need to cache the CRTC, plane and
FB state structures we use so we can access them later without needing
more locks. So do this.
Notice that there are some pieces of the FBC code that look at things
that are only computed during the modeset, so we can't just can't
precompute whether FBC can be activated during the update_state_cache
stage. We may be able to do this later.
Paulo Zanoni [Thu, 21 Jan 2016 20:07:17 +0000 (18:07 -0200)]
drm/i915/fbc: unconditionally update FBC during atomic commits
We unconditionally disable/update FBC even during the page flip
IOCTLs, and an unconditional disable/update at every atomic commit
touching the primary plane shouldn't impact PC state residency
noticeably. Besides, the code that checks for rotation is a good hint
that we may be forgetting something else, so let's leave all the
decisions to intel_fbc.c, making the code much safer.
Once we have the code to properly make FBC enable/update decisions
based on atomic states, with proper locking, then we'll be able to
evaluate whether it will be worth trying to optimize the cases where a
disable isn't needed.
v2: Upstream moved and now our patch needs to remove dev_priv.
Paulo Zanoni [Tue, 19 Jan 2016 13:35:39 +0000 (11:35 -0200)]
drm/i915/fbc: don't use the frontbuffer tracking subsystem for flips
Before this patch, page flips would call intel_frontbuffer_flip() and
intel_frontbuffer_flip_complete(), which would call intel_fbc_flush(),
which would call intel_fbc_update(). The problem is that drawing
operations also trigger intel_fbc_flush() calls, so it's not
guaranteed that we have the CRTC and FB locks grabbed when
intel_fbc_flush() happens, since the call trace may come from the
rendering path.
We're trying to make the FBC code grab the appropriate CRTC/FB locks,
so split the drawing and the flipping logic in order to achieve that
in later patches. So now the frontbuffer tracking code is just going
to be used for frontbuffer drawing, and intel_fbc_update() is going to
be used directly for actual page flips.
As a note, we don't need to call intel_fbc_flip() during the two
places where we call intel_frontbuffer_flip() since in one of them we
already have an intel_fbc_update() call, and in the other we have the
planes disabled.
Paulo Zanoni [Mon, 11 Jan 2016 19:44:36 +0000 (17:44 -0200)]
drm/i915/fbc: replace frequent dev_priv->fbc.x with fbc->x
We say "dev_priv->fbc.something" way too many times in our code while
we could be saying just "fbc->something" with a previous declaration
of fbc. This has been bothering me for a while but I didn't want to
patch it since I wanted to fix the real problems first. But as I add
more code I keep thinking about it, especially since it makes the code
easier to read and it can make us fit 80 columns easier, so let's just
do the change now.
While at it, also rename from i915_fbc to intel_fbc because the whole
FBC code uses intel_fbc.
The early return inside __intel_fbc_update does not completely check
all the parameters that affect the FBC register values. For example,
we currently lack looking at crtc->adjusted_y (for the fence Y offset)
and all the parameters that affect the CFB size (for i8xx).
Instead of just adding the missing parameters to the check and hoping
that any changes to the fbc_activate functions also come with a
matching change to the __intel_fbc_update check, introduce a new
structure where we store these parameters and use the structure at the
fbc_activate function. Of course, it's still possible to access
everything from dev_priv in those functions, but IMHO the new code
will be harder to break.
Paulo Zanoni [Thu, 21 Jan 2016 20:03:05 +0000 (18:03 -0200)]
drm/i915/fbc: wait for a vblank instead of 50ms when enabling
Instead of waiting for 50ms, just wait until the next vblank, since
it's the minimum requirement. The whole infrastructure of FBC is based
on vblanks, so waiting for X vblanks instead of X milliseconds sounds
like the correct way to go. Besides, 50ms may be less than a vblank on
super slow modes that may or may not exist.
There are some small improvements in PC state residency (due to the
fact that we're now using 16ms for the common modes instead of 50ms),
but the biggest advantage is still the correctness of being
vblank-based instead of time-based.
v2:
- Rebase after changing the patch order.
- Update the commit message.
v3:
- Fix bogus vblank_get() instead of vblank_count() (Ville).
- Don't forget to call drm_crtc_vblank_{get,put} (Chris, Ville)
- Adjust the performance details on the commit message.
v4:
- Don't grab the FBC mutex just to grab the vblank (Maarten)
Gerd Hoffmann [Mon, 25 Jan 2016 11:02:28 +0000 (12:02 +0100)]
drm/i915: refine qemu south bridge detection
The test for the qemu q35 south bridge added by commit
"39bfcd52 drm/i915: more virtual south bridge detection"
also matches on real hardware. Having the check for
virtual systems last in the list is not enough to avoid
that ...
Refine the check by additionally verifying the pci
subsystem id to see whenever it *really* is qemu.
Ville Syrjälä [Wed, 20 Jan 2016 19:05:28 +0000 (21:05 +0200)]
drm/i915: Fix intel_tile_width() parameters
The fb_modifiers and cpp arguments passed to intel_tile_width() in
intel_fill_fb_ggtt_view() got accidentally swapped around. I'm pretty
sure I fixed this already, but could be I lost the fix accidentally
during some rebases or something. Anyway, fix it up for real.
Ville Syrjälä [Wed, 20 Jan 2016 19:05:26 +0000 (21:05 +0200)]
drm/i915: Standardize on 'cpp' for bytes per pixel
We more or less randomly call the "bytes per pixel" value
'cpp', 'bytes_per_pixel', 'pixel_size', or even 'bpp'. Let's just pick
one and stick to it. I've chosen 'cpp'.
Ville Syrjälä [Wed, 20 Jan 2016 19:05:23 +0000 (21:05 +0200)]
drm/i915: Pass stride to rotate_pages()
Pass stride in addition to width and height to rotate_pages(). For now
width and stride are the same, but once framebuffer offsets enter the
scene that may no longer be the case.
Dave Gordon [Thu, 28 Jan 2016 10:48:09 +0000 (10:48 +0000)]
Fix pointer tests in error-handling paths
In the error-handling paths of i915_gem_do_execbuffer() and
intel_crtc_page_flip(), the local pointer-to-request variables
were expected to be either valid pointers or NULL. Since
2682708 drm/i915: simplify allocation of driver-internal requests
they could also be ERR_PTR() values, so the tests need to be
updated to accommodate this case.
Tvrtko Ursulin [Thu, 28 Jan 2016 10:29:57 +0000 (10:29 +0000)]
drm/i915: Fix premature LRC unpin in GuC mode
In GuC mode LRC pinning lifetime depends exclusively on the
request liftime. Since that is terminated by the seqno update
that opens up a race condition between GPU finishing writing
out the context image and the driver unpinning the LRC.
To extend the LRC lifetime we will employ a similar approach
to what legacy ringbuffer submission does.
We will start tracking the last submitted context per engine
and keep it pinned until it is replaced by another one.
Note that the driver unload path is a bit fragile and could
benefit greatly from efforts to unify the legacy and exec
list submission code paths.
At the moment i915_gem_context_fini has special casing for the
two which are potentialy not needed, and also depends on
i915_gem_cleanup_ringbuffer running before itself.
v2:
* Move pinning into engine->emit_request and actually fix
the reference/unreference logic. (Chris Wilson)
* ring->dev can be NULL on driver unload so use a different
route towards it.
v3:
* Rebase.
* Handle the reset path. (Chris Wilson)
* Exclude default context from the pinning - it is impossible
to get it right before default context special casing in
general is eliminated.
v4:
* Rebased & moved context tracking to
intel_logical_ring_advance_and_submit.
Tvrtko Ursulin [Thu, 28 Jan 2016 10:29:56 +0000 (10:29 +0000)]
drm/i915: Extract context unpinning to its own function
Will enable cleaner implementation of a following fix and
easier code unification in the future.
Idea and code by Chris Wilson.
v2: Do not return before last_contexts on engines are unpinned.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Chris Wilson <chris@chris-wilson.co.uk>
Tvrtko Ursulin [Thu, 28 Jan 2016 10:29:55 +0000 (10:29 +0000)]
drm/i915: Make LRC pinning own a reference to the context
Will simplify the following fix and sounds logical.
v2: Add some whitespace to separate logic better. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Nick Hoath <nicholas.hoath@intel.com>
Tvrtko Ursulin [Thu, 28 Jan 2016 10:29:54 +0000 (10:29 +0000)]
drm/i915: Make LRC (un)pinning work on context and engine
Previously intel_lr_context_(un)pin were operating on requests
which is in conflict with their names.
If we make them take a context and an engine, it makes the names
make more sense and it also makes future fixes possible.
v2: Rebase for default_context/kernel_context change.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Nick Hoath <nicholas.hoath@intel.com>
Imre Deak [Tue, 19 Jan 2016 13:26:32 +0000 (15:26 +0200)]
drm/i915: Move stolen memory initialization earlier during loading
The only device specific dependency of the stolen memory setup is the
MMIO mapping and the stolen memory size. Both are already available in
i915_gtt_init(), so move the stolen initialization to there. The
clean-up code for i915_gtt_init() is in i915_global_gtt_cleanup(), so
move the stolen memory clean-up code there too.
This will be needed by an upcoming patch that needs the details of the
memory we reserve, but the change is also part of our generic goal to
move the initialization of resources with no or little dependencies on
other device specific resources towards the beginning of the init
sequence.
Imre Deak [Tue, 19 Jan 2016 13:26:31 +0000 (15:26 +0200)]
drm/i915: Move MCHBAR setup earlier during init
Move the MCHBAR setup right after the MMIO setup, since the two things
are logically related and the MCHBAR setup code doesn't depend on any
other device specific resource. We'll also need MCHBAR to be ready
earlier in an upcoming patch, so this is also a preparation for that.
Factor out the init/clean-up code to separate functions to make things
clearer in the i915_driver_load()/unload() functions.
Imre Deak [Tue, 19 Jan 2016 13:26:30 +0000 (15:26 +0200)]
drm/i915: Move allocation of various workqueues earlier during init
Workqueue initalization doesn't depend on any other device specific
resource, so move it close to the beginning, so we don't need to
consider them when thinking about dependencies for other resources.
Also factor out things to separate init/cleanup functions to make
i915_driver_load()/unload() clearer, atm it's somewhat difficult to
follow there in what order resources are inited/cleaned-up.
Imre Deak [Tue, 19 Jan 2016 13:26:29 +0000 (15:26 +0200)]
drm/i915: Sanitize i915_gem_load() init and clean-up
Factor out common clean-up code for the GEM load time init function.
Also rename i915_gem_load() to i915_gem_load_init() to have a better
match with its new clean-up function.
Imre Deak [Tue, 19 Jan 2016 13:26:28 +0000 (15:26 +0200)]
drm/i915: Sanitize GEM shrinker init and clean-up
Factor out the common GEM shrinker clean-up code and call the shrinker
init function from the same function from where the corresponding
shrinker clean-up function is called. Also add sanity checking to the
shrinker and OOM registration calls.
Clarify the name of the label on the error path, making it clear what's
being cleaned up. The kmem_cache_destroy() calls are NOPs on the
corresponding error path.
"Which would mean the offender is in intel_logical_ring_cleanup is most
likely:
...
if (ring->status_page.obj) {
kunmap(sg_page(ring->status_page.obj->pages->sgl));
ring->status_page.obj = NULL;
}
...
"I think that the i915_gem_context_fini will do a final unref on
dev_priv->kernel_context and then the ring buff has a copy which is
left dangling because:
"Where default_ctx_obj == dev_priv->kernel_context->engine[ring->id].state
So indeed looks like the unload ordering is the trigger. In fact it
is almost the same fragility wrt/ kernel_context hidden dependency I
expressed my worry about in an e-mail yesterday or so. It only shows
if CONFIG_DEBUG_SG is set, otherwise it accesses freed memory and
probably just survives."
This causes serious trouble in our CI system since it took out all
gen8+ machines. Not yet clear why this wasn't caught in pre-merge
testing.
Cc: Nick Hoath <nicholas.hoath@intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Tested-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Nick Hoath [Thu, 21 Jan 2016 19:37:45 +0000 (19:37 +0000)]
drm/i915: Fix context/engine cleanup order
Swap the order of context & engine cleanup, so that contexts are cleaned
up first, and *then* engines. This is a more sensible order anyway, but
in particular has become necessary since the 'intel_ring_initialized()
must be simple and inline' patch, which now uses ring->dev as an
'initialised' flag, so it can now be NULL after engine teardown. This
in turn can cause a problem in the context code, which (used to) check
the ring->dev->struct_mutex -- causing a fault if ring->dev was NULL.
Also rename the cleanup function to reflect what it actually does
(cleanup engines, not a ringbuffer), and fix an annoying whitespace issue.
v2: Also make the fix in i915_load_modeset_init, not just in
i915_driver_unload (Chris Wilson)
v3: Had extra stuff in it.
v4: Reverted extra stuff (so we're back to v2).
Rebased and updated commentary above (Dave Gordon).
Signed-off-by: Nick Hoath <nicholas.hoath@intel.com> Signed-off-by: Dave Gordon <david.s.gordon@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2) Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1453405067-32890-3-git-send-email-david.s.gordon@intel.com Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Chris Wilson [Thu, 21 Jan 2016 17:32:43 +0000 (17:32 +0000)]
drm/i915: Improve handling of overlapping objects
The generic interval tree we use to speed up range invalidation is an
augmented rbtree that can report all overlapping intervals for a given
range. Therefore we do not need to degrade to a linear list if we find
overlapping objects. Oops.
Arun Siluvery [Thu, 21 Jan 2016 21:43:53 +0000 (21:43 +0000)]
drm/i915/skl: Enable Per context Preemption granularity control
Per context preemption granularity control is only available from SKL:E0+
Actual WA is to disable percontext preemption granularity control until D0
which is the default case so this is equivalent to the inverse of
WaDisablePerCtxtPreemptionGranularityControl:skl
Arun Siluvery [Thu, 21 Jan 2016 21:43:52 +0000 (21:43 +0000)]
drm/i915/skl: Add GEN8_L3SQCREG4 to HW whitelist
Required for WaDisableLSQCROPERFforOCL:skl
This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.
Arun Siluvery [Thu, 21 Jan 2016 21:43:51 +0000 (21:43 +0000)]
drm/i915/bxt: Add GEN8_L3SQCREG4 to HW whitelist
Required for WaDisableLSQCROPERFforOCL:bxt
According to WA database these are only applicable for BXT:A0 but since
A0 and A1 shares the same GT these are extended for A1 as well.
This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.
Arun Siluvery [Thu, 21 Jan 2016 21:43:50 +0000 (21:43 +0000)]
drm/i915/bxt: Add GEN9_CS_DEBUG_MODE1 to HW whitelist
Required for,
WaDisableObjectLevelPreemptionForTrifanOrPolygon:bxt
WaDisableObjectLevelPreemptionForInstancedDraw:bxt
WaDisableObjectLevelPreemtionForInstanceId:bxt
According to WA database these are only applicable for BXT:A0 but since
A0 and A1 shares the same GT these are extended for A1 as well.
These are also required for SKL until B0 but not adding them because they
are pre-production steppings.
This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.
v2: use lower case in register defines (Nick)
v3: explain purpose of changes (Chris)
Arun Siluvery [Thu, 21 Jan 2016 21:43:49 +0000 (21:43 +0000)]
drm/i915/gen9: Add HDC_CHICKEN1 to HW whitelist
Required for WaAllowUMDToModifyHDCChicken1:skl,bxt
This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.
Arun Siluvery [Thu, 21 Jan 2016 21:43:48 +0000 (21:43 +0000)]
drm/i915/gen9: Add GEN8_CS_CHICKEN1 to HW whitelist
Required for WaEnablePreemptionGranularityControlByUMD:skl,bxt
This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.
projects / linux-beck.git / log