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
commit ac657f6461 ("drm/i915: Introduce IS_GEN macro") introduced
GEN_FOREVER that was never used.
My first attempt was to rename it to FOREVER since GEN is
already part of the macro. Then I used coccinelle to change all
-INTEL_GEN(e1) >= e2
+INTEL_GEN_RANGE(e1, e2, FOREVER)
-INTEL_GEN(e1) <= e2
+INTEL_GEN_RANGE(e1, 0, e2)
and I liked it.
However I didn't like very much the remaining
INTEL_GEN(dev_priv) < n
and:
INTEL_GEN(e1) < n
INTEL_GEN_RANGE(e1, 0, n - 1)
didn't make much sense either.
So INTEL_GEN use for > or < seems a better unified way for unlimited
bounds. So, no reason to keep GEN_FOREVER here.
Let's kill before someone start using it.
v2: Remove remaining GEN_FOREVER forgotten in a comment. (Daniel)
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20181026195143.20353-2-rodrigo.vivi@intel.com
The way our hardware is designed doesn't seem to let us use the
MI_RECORD_PERF_COUNT command without setting up a circular buffer.
In the case where the user didn't request OA reports to be available
through the i915 perf stream, we can set the OA buffer to the minimum
size to avoid consuming memory which won't be used by the driver.
v2: Simplify oa buffer size exponent selection (Chris)
Reuse vma size field (Lionel)
v3: Restrict size opening parameter to values supported by HW (Chris)
v4: Drop out of date comment (Matt)
Add debug message when buffer size is rejected (Matt)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181023100707.31738-5-lionel.g.landwerlin@intel.com
Add plane alpha blending support with the different blend modes.
This has been tested on a icl to show the correct results,
on earlier platforms small rounding errors cause issues. But this
already happens case with fully transparant or fully opaque RGB8888
fb's.
The recommended HW workaround is to disable alpha blending when the
plane alpha is 0 (transparant, hide plane) or 0xff (opaque, disable blending).
This is easy to implement on any platform, so just do that.
The tests for userspace are also available, and pass on gen11.
Changes since v1:
- Change mistaken < 0xff0 to 0xff00.
- Only set PLANE_KEYMSK_ALPHA_ENABLE when plane alpha < 0xff00, ignore blend mode.
- Rework disabling FBC when per pixel alpha is used.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
[mlankhorst: Change MISSING_CASE default to explicit alpha disable (mattrope)]
Link: https://patchwork.freedesktop.org/patch/msgid/20180815103405.22679-1-maarten.lankhorst@linux.intel.com
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Move max firmware size to the same if ladder with firmware name and
required version. This allows us to detect the missing max size for a
platform without actually loading the firmware, and makes the whole
thing easier to maintain.
We need to move the power get earlier to allow for early return in the
missing platform case. While at it, extend the comment on why we return
with the reference held on errors.
We also need to move the module parameter override later to reuse the
max firmware size, which is independent of the override.
v2: Add comment on why we leak the wakeref on errors (Chris)
v3: Rebase
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180926133414.22073-2-jani.nikula@intel.com
That we use a WB mapping for updating the RING_TAIL register inside the
context image even on !llc machines has been a source of consternation
for every reader. It appears to work on bsw+, but it may just have been
that we have been incredibly bad at detecting the errors.
v2: With extra enthusiasm.
v3: Drop force of map type for pinned default_state as by the time we
pin it, the map type is always WB and doesn't conflict with the earlier
use by ce->state.
v4: Transfer engine->default_state from MAP_WC to MAP_WB on creation so
we do not need the MAP_FORCE littered around the backends
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180914123504.2062-3-chris@chris-wilson.co.uk
drm-misc-next for 4.20:
UAPI Changes:
- Add host endian variants for the most common formats (Gerd)
- Fail ADDFB2 for big-endian drivers that don't advertise BE quirk (Gerd)
- clear smem_start in fbdev for drm drivers to avoid leaking fb addr (Daniel)
Cross-subsystem Changes:
Core Changes:
- fix drm_mode_addfb() on big endian machines (Gerd)
- add timeline point to syncobj find+replace (Chunming)
- more drmP.h removal effort (Daniel)
- split uapi portions of drm_atomic.c into drm_atomic_uapi.c (Daniel)
Driver Changes:
- bochs: Convert open-coded portions to use helpers (Peter)
- vkms: Add cursor support (Haneen)
- udmabuf: Lots of fixups (mostly cosmetic afaict) (Gerd)
- qxl: Convert to use fbdev helper (Peter)
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Chunming Zhou <david1.zhou@amd.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Peter Wu <peter@lekensteyn.nl>
Cc: Haneen Mohammed <hamohammed.sa@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Sean Paul <sean@poorly.run>
Link: https://patchwork.freedesktop.org/patch/msgid/20180913130254.GA156437@art_vandelay
This patch adds support to decode system memory bandwidth and other
parameters for skylake and Gen9+ platforms, which will be used for
arbitrated display memory bandwidth calculation in GEN9 based
platforms and WM latency level-0 Work-around calculation on GEN9+.
Changes Since V1:
- s/memdev_info/dram_info
- create a struct to hold channel info
Changes Since V2:
- rewrite code to adhere i915 coding style
- not valid for GLK
Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180824093225.12598-3-mahesh1.kumar@intel.com
Future gen reduce the number of bits we will have available to
differentiate between contexts, so reduce the lifetime of the ID
assignment from that of the context to its current active cycle (i.e.
only while it is pinned for use by the HW, will it have a constant ID).
This means that instead of a max of 2k allocated contexts (worst case
before fun with bit twiddling), we instead have a limit of 2k in flight
contexts (minus a few that have been pinned by the kernel or by perf).
To reduce the number of contexts id we require, we allocate a context id
on first and mark it as pinned for as long as the GEM context itself is,
that is we keep it pinned it while active on each engine. If we exhaust
our context id space, then we try to reclaim an id from an idle context.
In the extreme case where all context ids are pinned by active contexts,
we force the system to idle in order to recover ids.
We cannot reduce the scope of an HW-ID to an engine (allowing the same
gem_context to have different ids on each engine) as in the future we
will need to preassign an id before we know which engine the
context is being executed on.
v2: Improved commentary (Tvrtko) [I tried at least]
References: https://bugs.freedesktop.org/show_bug.cgi?id=107788
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180904153117.3907-1-chris@chris-wilson.co.uk
Older gen use a physical address for the hardware status page, for which
we use cache-coherent writes. As the writes are into the cpu cache, we use
a normal WB mapped page to read the HWS, used for our seqno tracking.
Anecdotally, I observed lost breadcrumbs writes into the HWS on i965gm,
which so far have not reoccurred with this patch. How reliable that
evidence is remains to be seen.
v2: Explicitly pass the expected physical address to the hw
v3: Also remember the wild writes we once had for HWS above 4G.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20180903152304.31589-2-chris@chris-wilson.co.uk
Instead of defining all registers twice, define just a PCH_GPIO_BASE
that has the same address as PCH_GPIO_A and use that to calculate all
the others. This also brings VLV and !HAS_GMCH_DISPLAY in line, doing
the same thing.
v2: Fix GMBUS registers to be relative to gpio base; create GPIO()
macro to return a particular gpio address and move the enum out of
i915_reg.h (suggested by Jani)
v3: Move base offset inside the GPIO() macro so the GMBUS defines don't
actually need to be changed (suggested by Daniel/Ville)
v4: Move definition of i915_gpio to intel_display.h and remove
GMBUS/GPIO handling from gvt since now they have their own
defines.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180727193647.8639-3-lucas.demarchi@intel.com
The device global init_power_on flag is somewhat arbitrary and makes
debugging power refcounting problems difficult. Instead arrange things
so that all display power domain get has a corresponding put call. After
this change we have the following sequences:
driver loading:
intel_power_domains_init_hw();
<other init steps>
intel_power_domains_enable();
driver unloading:
intel_power_domains_disable();
<other uninit steps>
intel_power_domains_fini_hw();
system suspend:
intel_power_domains_disable();
<other suspend steps>
intel_power_domains_suspend();
system resume:
intel_power_domains_resume();
<other resume steps>
intel_power_domains_enable();
at other times while the driver is loaded:
intel_display_power_get();
...
intel_display_power_put();
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180816123757.3286-2-imre.deak@intel.com
Currently tests modify i915.enable_psr and then do a modeset cycle
to change PSR. We can write a value to i915_edp_psr_debug to force
a certain PSR mode without a modeset.
To retain compatibility with older userspace, we also still allow
the override through the module parameter, and add some tracking
to check whether a debugfs mode is specified.
Changes since v1:
- Rename dev_priv->psr.enabled to .dp, and .hw_configured to .enabled.
- Fix i915_psr_debugfs_mode to match the writes to debugfs.
- Rename __i915_edp_psr_write to intel_psr_set_debugfs_mode, simplify
it and move it to intel_psr.c. This keeps all internals in intel_psr.c
- Perform an interruptible wait for hw completion outside of the psr
lock, instead of being forced to trywait and return -EBUSY.
Changes since v2:
- Rebase on top of intel_psr changes.
Changes since v3:
- Assign psr.dp during init. (dhnkrn)
- Add prepared bool, which should be used instead of relying on psr.dp. (dhnkrn)
- Fix -EDEADLK handling in debugfs. (dhnkrn)
- Clean up waiting for idle in intel_psr_set_debugfs_mode.
- Print PSR mode when trying to enable PSR. (dhnkrn)
- Move changing psr debug setting to i915_edp_psr_debug_set. (dhnkrn)
Changes since v4:
- Return error in _set() function.
- Change flag values to make them easier to remember. (dhnkrn)
- Only assign psr.dp once. (dhnkrn)
- Only set crtc_state->has_psr on the crtc with psr.dp.
- Fix typo. (dhnkrn)
Changes since v5:
- Only wait for PSR idle on the PSR connector correctly. (dhnkrn)
- Reinstate WARN_ON(drrs.dp) in intel_psr_enable. (dhnkrn)
- Remove stray comment. (dhnkrn)
- Be silent in intel_psr_compute_config on wrong connector. (dhnkrn)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180809142101.26155-1-maarten.lankhorst@linux.intel.com
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Similarly to the previous patch use a separate request/status HW flag
index defined right after the corresponding control registers instead of
depending for this on the power well IDs. Since the set of
control/status registers varies among the different power wells (on a
single platform), also add a new i915_power_well_registers struct that
we populate and assign to each DDI power well as needed.
Also clarify a bit the code comment describing the function and layout
of the control registers.
This also fixes a problem on ICL, where we incorrectly read the KVMR
control register in hsw_power_well_requesters() even for DDI and AUX
power wells.
v2:
- Clarify platform range tags in code comments. (Paulo)
- Fix line over 80 chars checkpatch warning.
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180806095843.13294-7-imre.deak@intel.com
Atm, we determine the control/status flag offsets within the PUNIT
control/status registers based on the power well's ID. Since the power
well ID enum is global across all platforms, the associated macros to
get the flag offsets involves some magic. This makes checking the
register/bit definitions against the specification more difficult than
necessary. Also the values in the power well ID enum must stay fixed,
making code maintenance of the enum cumbersome.
To solve the above define the control/status flag indices right after
the corresponding registers and use these to derive the control/status
flag values by storing the indices in the i915_power_well_desc struct.
Initializing anonymous union fields require the preceding field in the
struct to be explicitly initialized - even when using named
initializers - and the initialization to be done right before the union
initialization, hence the reordering of the .id fields.
v2:
- Clarify commit log message about anonymous union initializers. (Paulo)
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180806095843.13294-6-imre.deak@intel.com
It makes sense to keep unchanging data const. Extract such fields from
the i915_power_well struct into a new i915_power_well_desc struct that
we initialize during compile time. For the rest of the dynamic
fields allocate an array of i915_power_well objects in i915 dev_priv,
and link to each of these objects their corresponding
i915_power_well_desc object.
v2:
- Fix checkpatch warnings about missing param name in fn declaration and
lines over 80 chars. (Paulo)
- Move check for unique IDs to __set_power_wells().
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
[Fixed checkpatch warn in __set_power_wells()]
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180806095843.13294-5-imre.deak@intel.com
RPS provides a feedback loop where we use the load during the previous
evaluation interval to decide whether to up or down clock the GPU
frequency. Our responsiveness is split into 3 regimes, a high and low
plateau with the intent to keep the gpu clocked high to cover occasional
stalls under high load, and low despite occasional glitches under steady
low load, and inbetween. However, we run into situations like kodi where
we want to stay at low power (video decoding is done efficiently
inside the fixed function HW and doesn't need high clocks even for high
bitrate streams), but just occasionally the pipeline is more complex
than a video decode and we need a smidgen of extra GPU power to present
on time. In the high power regime, we sample at sub frame intervals with
a bias to upclocking, and conversely at low power we sample over a few
frames worth to provide what we consider to be the right levels of
responsiveness respectively. At low power, we more or less expect to be
kicked out to high power at the start of a busy sequence by waitboosting.
Prior to commit e9af4ea2b9 ("drm/i915: Avoid waitboosting on the active
request") whenever we missed the frame or stalled, we would immediate go
full throttle and upclock the GPU to max. But in commit e9af4ea2b9, we
relaxed the waitboosting to only apply if the pipeline was deep to avoid
over-committing resources for a near miss. Sadly though, a near miss is
still a miss, and perceptible as jitter in the frame delivery.
To try and prevent the near miss before having to resort to boosting
after the fact, we use the pageflip queue as an indication that we are
in an "interactive" regime and so should sample the load more frequently
to provide power before the frame misses it vblank. This will make us
more favorable to providing a small power increase (one or two bins) as
required rather than going all the way to maximum and then having to
work back down again. (We still keep the waitboosting mechanism around
just in case a dramatic change in system load requires urgent uplocking,
faster than we can provide in a few evaluation intervals.)
v2: Reduce rps_set_interactive to a boolean parameter to avoid the
confusion of what if they wanted a new power mode after pinning to a
different mode (which to choose?)
v3: Only reprogram RPS while the GT is awake, it will be set when we
wake the GT, and while off warns about being used outside of rpm.
v4: Fix deferred application of interactive mode
v5: s/state/interactive/
v6: Group the mutex with its principle in a substruct
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107111
Fixes: e9af4ea2b9 ("drm/i915: Avoid waitboosting on the active request")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180731132629.3381-1-chris@chris-wilson.co.uk
(cherry picked from commit 60548c554b)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
After disabling resource streamer on ICL (due to it actually not
existing there), I got feedback that there have been some experimental
patches for mesa to use RS years ago, but nothing ever landed or shipped
because there was no performance improvement.
This removes it from kernel keeping the uapi defines around for
compatibility.
v2: - re-add the inadvertent removal of CTX_CTRL_INHIBIT_SYN_CTX_SWITCH
- don't bother trying to document removed params on uapi header:
applications should know that from the query.
(from Chris)
v3: - disable CTX_CTRL_RS_CTX_ENABLE istead of removing it
- reword commit message after Daniele confirmed no performance
regression on his machine
- reword commit message to make clear RS is being removed due to
never been used
v4: - move I915_EXEC_RESOURCE_STREAMER to __I915_EXEC_ILLEGAL_FLAGS so
the check on ioctl() is made much earlier by
i915_gem_check_execbuffer() (suggested by Tvrtko)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180803232443.17193-1-lucas.demarchi@intel.com
RPS provides a feedback loop where we use the load during the previous
evaluation interval to decide whether to up or down clock the GPU
frequency. Our responsiveness is split into 3 regimes, a high and low
plateau with the intent to keep the gpu clocked high to cover occasional
stalls under high load, and low despite occasional glitches under steady
low load, and inbetween. However, we run into situations like kodi where
we want to stay at low power (video decoding is done efficiently
inside the fixed function HW and doesn't need high clocks even for high
bitrate streams), but just occasionally the pipeline is more complex
than a video decode and we need a smidgen of extra GPU power to present
on time. In the high power regime, we sample at sub frame intervals with
a bias to upclocking, and conversely at low power we sample over a few
frames worth to provide what we consider to be the right levels of
responsiveness respectively. At low power, we more or less expect to be
kicked out to high power at the start of a busy sequence by waitboosting.
Prior to commit e9af4ea2b9 ("drm/i915: Avoid waitboosting on the active
request") whenever we missed the frame or stalled, we would immediate go
full throttle and upclock the GPU to max. But in commit e9af4ea2b9, we
relaxed the waitboosting to only apply if the pipeline was deep to avoid
over-committing resources for a near miss. Sadly though, a near miss is
still a miss, and perceptible as jitter in the frame delivery.
To try and prevent the near miss before having to resort to boosting
after the fact, we use the pageflip queue as an indication that we are
in an "interactive" regime and so should sample the load more frequently
to provide power before the frame misses it vblank. This will make us
more favorable to providing a small power increase (one or two bins) as
required rather than going all the way to maximum and then having to
work back down again. (We still keep the waitboosting mechanism around
just in case a dramatic change in system load requires urgent uplocking,
faster than we can provide in a few evaluation intervals.)
v2: Reduce rps_set_interactive to a boolean parameter to avoid the
confusion of what if they wanted a new power mode after pinning to a
different mode (which to choose?)
v3: Only reprogram RPS while the GT is awake, it will be set when we
wake the GT, and while off warns about being used outside of rpm.
v4: Fix deferred application of interactive mode
v5: s/state/interactive/
v6: Group the mutex with its principle in a substruct
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107111
Fixes: e9af4ea2b9 ("drm/i915: Avoid waitboosting on the active request")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180731132629.3381-1-chris@chris-wilson.co.uk
