Unfortunately, it seems that the HPD IRQ storm problem from the early
days of Intel GPUs was never entirely solved, only mostly. Within the
last couple of days, I got a bug report from one of our customers who
had been having issues with their machine suddenly booting up very
slowly after having updated. The amount of time it took to boot went
from around 30 seconds, to over 6 minutes consistently.
After some investigation, I discovered that i915 was reporting massive
amounts of short HPD IRQ spam on this system from the DisplayPort port,
despite there not being anything actually connected. The symptoms would
start with one "long" HPD IRQ being detected at boot:
[ 1.891398] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00440000, dig 0x00440000, pins 0x000000a0
[ 1.891436] [drm:intel_hpd_irq_handler [i915]] digital hpd port B - long
[ 1.891472] [drm:intel_hpd_irq_handler [i915]] Received HPD interrupt on PIN 5 - cnt: 0
[ 1.891508] [drm:intel_hpd_irq_handler [i915]] digital hpd port D - long
[ 1.891544] [drm:intel_hpd_irq_handler [i915]] Received HPD interrupt on PIN 7 - cnt: 0
[ 1.891592] [drm:intel_dp_hpd_pulse [i915]] got hpd irq on port B - long
[ 1.891628] [drm:intel_dp_hpd_pulse [i915]] got hpd irq on port D - long
…
followed by constant short IRQs afterwards:
[ 1.895091] [drm:intel_encoder_hotplug [i915]] [CONNECTOR:66:DP-1] status updated from unknown to disconnected
[ 1.895129] [drm:i915_hotplug_work_func [i915]] Connector DP-3 (pin 7) received hotplug event.
[ 1.895165] [drm:intel_dp_detect [i915]] [CONNECTOR:72:DP-3]
[ 1.895275] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00200000, dig 0x00200000, pins 0x00000080
[ 1.895312] [drm:intel_hpd_irq_handler [i915]] digital hpd port D - short
[ 1.895762] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00200000, dig 0x00200000, pins 0x00000080
[ 1.895799] [drm:intel_hpd_irq_handler [i915]] digital hpd port D - short
[ 1.896239] [drm:intel_dp_aux_xfer [i915]] dp_aux_ch timeout status 0x71450085
[ 1.896293] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00200000, dig 0x00200000, pins 0x00000080
[ 1.896330] [drm:intel_hpd_irq_handler [i915]] digital hpd port D - short
[ 1.896781] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00200000, dig 0x00200000, pins 0x00000080
[ 1.896817] [drm:intel_hpd_irq_handler [i915]] digital hpd port D - short
[ 1.897275] [drm:intel_get_hpd_pins [i915]] hotplug event received, stat 0x00200000, dig 0x00200000, pins 0x00000080
The customer's system in question has a GM45 GPU, which is apparently
well known for hotplugging storms.
So, workaround this impressively broken hardware by changing the default
HPD storm threshold from 5 to 50. Then, make long IRQs count for 10, and
short IRQs count for 1. This makes it so that 5 long IRQs will trigger
an HPD storm, and on systems with short HPD storm detection 50 short
IRQs will trigger an HPD storm. 50 short IRQs amounts to 100ms of
constant pulsing, which seems like a good middleground between being too
sensitive and not being sensitive enough (which would cause visible
stutters in userspace every time a storm occurs).
And just to be extra safe: we don't enable this by default on systems
with MST support. There's too high of a chance of MST support triggering
storm detection, and systems that are new enough to support MST are a
lot less likely to have issues with IRQ storms anyway.
As a note: this patch was tested using a ThinkPad T450s and a Chamelium
to simulate the short IRQ storms.
Changes since v1:
- Don't use two separate thresholds, just make long IRQs count for 10
each and short IRQs count for 1. This simplifies the code a bit
- Ville Syrjälä
Changes since v2:
- Document @long_hpd in intel_hpd_irq_storm_detect, no functional
changes
Changes since v4:
- Remove !! in long_hpd assignment - Ville Syrjälä
- queue_hp = true - Ville Syrjälä
Signed-off-by: Lyude Paul <lyude@redhat.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181106213017.14563-6-lyude@redhat.com
This is rather confusing to look at as-is:
dev_priv->display.hpd_irq_setup(dev_priv); in intel_hpd_irq_handler()
handles disabling the actual HPD IRQ, while
intel_hpd_irq_storm_disable() handles moving the HPD pin state over from
MARK_DISABLED to DISABLED along with enabling polling for it.
Changes since v3:
- Rename i915_hpd_irq_storm_disable() to
i915_hpd_irq_storm_switch_to_polling() - Rodrigo Vivi
Signed-off-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181106213017.14563-5-lyude@redhat.com
Currently in intel_hpd_irq_storm_detect() when we detect that the last
recorded hotplug wasn't within the period defined by
HPD_STORM_DETECT_DELAY, we make the mistake of resetting the HPD count
to 0 without incrementing it. This results in us only enabling storm
detection when we go +2 above the threshold, e.g. an HPD threshold of 5
would not trigger a storm until we reach a total of 7 hotplugs.
So: rework the code a bit so we reset the HPD count when
HPD_STORM_DETECT_DELAY has passed, then increment the count afterwards.
Also, clean things up a bit to make it easier to undertand.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181106213017.14563-4-lyude@redhat.com
Exercising the gpu reloc path strenuously revealed an issue where the
updated relocations (from MI_STORE_DWORD_IMM) were not being observed
upon execution. After some experiments with adding pipecontrols (a lot
of pipecontrols (32) as gen4/5 do not have a bit to wait on earlier pipe
controls or even the current on), it was discovered that we merely
needed to delay the EMIT_INVALIDATE by several flushes. It is important
to note that it is the EMIT_INVALIDATE as opposed to the EMIT_FLUSH that
needs the delay as opposed to what one might first expect -- that the
delay is required for the TLB invalidation to take effect (one presumes
to purge any CS buffers) as opposed to a delay after flushing to ensure
the writes have landed before triggering invalidation.
Testcase: igt/gem_tiled_fence_blits
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181105094305.5767-1-chris@chris-wilson.co.uk
We're no longer programming any watermarks when we're disabling
a pipe. That means ilk_wm_merge() & co. will keep considering
the any pipe that is getting disabled as still enabled. Thus we
either get no LP1+ watermakrs (ilk-ivb), or we get suboptimal
ones (hsw-bdw).
This seems to have been broken by commit b6b178a772 ("drm/i915:
Calculate ironlake intermediate watermarks correctly, v2."). Before
that we apparently had some difference between the intermediate
and optimal watermarks and so we would program the optiomal ones.
Now intermediate and optimal are identical for disabled pipes
and so we don't program either.
Fix this by programming the intermediate watermarks even for
disabled pipes. We were already doing that for skl+. We'll
leave out gmch platforms for now since those do the merging
in a different manner and should work as is. We'll want to
unify this eventually, but play it safe for now and just put
in a FIXME.
Cc: stable@vger.kernel.org
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: b6b178a772 ("drm/i915: Calculate ironlake intermediate watermarks correctly, v2.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181025130536.29024-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #irc
(cherry picked from commit a748faea3b)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Plane input CSC needs to be enabled to convert frambuffers from
YUV to RGB. This is needed for bottom 3 planes on ICL, rest of
the planes have hardcoded conversion and taken care by the legacy
code.
This patch defines the co-efficient values for YUV to RGB conversion
in BT709 and BT601 formats. It programs the coefficients and enables
the plane input csc unit in hardware.
This has been verified and tested by Maarten and the change is working
as expecpted.
v2: Addressed Maarten's and Ville's review comments and added the
coefficients in a 2D array instead of independent Macros.
v3: Added individual coefficient matrix (9 values) instead of 6
register values as per Maarten's comment. Also addresed a shift
issue with B channel coefficient.
v4: Added support for Limited Range Color Handling
v5: Fixed Matt and Maarten's review comments.
v6: Added human readable matrix values for YUV to RGB Conversion along
with just the bspec register values, as per Matt's suggestion.
v7: Refactored the code, move csc coefficient programming function to
intel_sprite.c and made it static as per Ville's review comment.
v8: Addressed Ville's review comment. Called the coefficient programming
from within the skl_program_plane and used I915_WRITE_FW instead of
I915_WRITE.
v9: Fixed Ville's review comments.
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1541099420-12419-3-git-send-email-uma.shankar@intel.com
Defined the plane input csc coefficient registers and macros.
6 registers are used to program a total of 9 coefficients,
added macros to define each of them for all the planes
supporting the feature on pipes. On ICL, bottom 3 planes
have this capability.
v2: Segregated the register macro definition as separate patch
as per Maarten's suggestion.
v3: Removed a redundant 3rd Pipe register definition and
simplified the equally spaced register definition by adding an
offset as per Matt's comment.
v4: No Change
v5: Renamed the register Macro as per Matt's suggestion.
v6: No Change
v7: No Change
v8: No Change
v9: No Change
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1541099420-12419-2-git-send-email-uma.shankar@intel.com
We're no longer programming any watermarks when we're disabling
a pipe. That means ilk_wm_merge() & co. will keep considering
the any pipe that is getting disabled as still enabled. Thus we
either get no LP1+ watermakrs (ilk-ivb), or we get suboptimal
ones (hsw-bdw).
This seems to have been broken by commit b6b178a772 ("drm/i915:
Calculate ironlake intermediate watermarks correctly, v2."). Before
that we apparently had some difference between the intermediate
and optimal watermarks and so we would program the optiomal ones.
Now intermediate and optimal are identical for disabled pipes
and so we don't program either.
Fix this by programming the intermediate watermarks even for
disabled pipes. We were already doing that for skl+. We'll
leave out gmch platforms for now since those do the merging
in a different manner and should work as is. We'll want to
unify this eventually, but play it safe for now and just put
in a FIXME.
Cc: stable@vger.kernel.org
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: b6b178a772 ("drm/i915: Calculate ironlake intermediate watermarks correctly, v2.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181025130536.29024-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #irc
Pull drm fixes from Dave Airlie:
"Pretty much a normal fixes pull pre-rc1, mostly amdgpu fixes, one i915
link training regression fix, and a couple of minor panel/bridge fixes
and a panel quirk"
* tag 'drm-next-2018-11-02' of git://anongit.freedesktop.org/drm/drm: (37 commits)
drm/amdgpu: revert "enable gfxoff in non-sriov and stutter mode by default"
drm/amd/pp: Print warning if od_sclk/mclk out of range
drm/amd/pp: Fix pp_sclk/mclk_od not work on Vega10
drm/amd/pp: Fix pp_sclk/mclk_od not work on smu7
drm/amd/powerplay: no MGPU fan boost enablement on DPM disabled
drm/amdgpu: Fix skipping hangged job reset during gpu recover.
drm/amd/powerplay: revise Vega20 pptable version check
drm/amd/display: set backlight level limit to 1
drm/panel: simple: Innolux TV123WAM is actually P120ZDG-BF1
dt-bindings: drm/panel: simple: Innolux TV123WAM is actually P120ZDG-BF1
drm/bridge: ti-sn65dsi86: Remove the mystery delay
drm/panel: simple: Add "no-hpd" delay for Innolux TV123WAM
drm/panel: simple: Support panels with HPD where HPD isn't connected
dt-bindings: drm/panel: simple: Add no-hpd property
drm/edid: Add 6 bpc quirk for BOE panel.
drm/amdgpu: fix reporting of failed msg sent to SMU (v2)
drm/amdgpu: Fix compute ring 1.0.0 failure after reset
drm/amdgpu: fix VM leaf walking
drm/amdgpu: fix amdgpu_vm_fini
drm/amd/powerplay: commonize the API for retrieving current clocks
...
Although there's nothing crucial missing, it's been a long time since
the last backmerge. Catch up with drm-next.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
