Many PCI and other drivers performs snd_pcm_period_elapsed() simply in
its interrupt handler, so the sync_stop operation is just to call
synchronize_irq(). Instead of putting this call multiple times,
introduce the common card->sync_irq field. When this field is set,
PCM core performs synchronize_irq() for sync-stop operation. Each
driver just needs to copy its local IRQ number to card->sync_irq, and
that's all we need.
Link: https://lore.kernel.org/r/20191117085308.23915-8-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The standard programming model of a PCM sound driver is to process
snd_pcm_period_elapsed() from an interrupt handler. When a running
stream is stopped, PCM core calls the trigger-STOP PCM ops, sets the
stream state to SETUP, and moves on to the next step. This is
performed in an atomic manner -- this could be called from the interrupt
context, after all.
The problem is that, if the stream goes further and reaches to the
CLOSE state immediately, the stream might be still being processed in
snd_pcm_period_elapsed() in the interrupt context, and hits a NULL
dereference. Such a crash happens because of the atomic operation,
and we can't wait until the stream-stop finishes.
For addressing such a problem, this commit adds a new PCM ops,
sync_stop. This gets called at the appropriate places that need a
sync with the stream-stop, i.e. at hw_params, prepare and hw_free.
Some drivers already have a similar mechanism implemented locally, and
we'll refactor the code later.
Link: https://lore.kernel.org/r/20191117085308.23915-7-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Currently PCM ioctl ops is a mandatory field but almost all drivers
simply pass snd_pcm_lib_ioctl. For simplicity, allow to set NULL in
the field and call snd_pcm_lib_ioctl() as default.
Link: https://lore.kernel.org/r/20191117085308.23915-4-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Update the documentation for the newly introduced managed buffer
allocation mode accordingly. The old preallocation is no longer
recommended.
Link: https://lore.kernel.org/r/20191117085308.23915-3-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch adds the support for the feature to automatically allocate
and free PCM buffers, so called "managed buffer allocation" mode.
It's set up via new PCM helpers, snd_pcm_set_managed_buffer() and
snd_pcm_set_managed_buffer_all(), both of which correspond to the
existing preallocator helpers, snd_pcm_lib_preallocate_pages() and
snd_pcm_lib_preallocate_pages_for_all(). When the new helper is used,
it not only performs the pre-allocation of buffers, but also it
manages to call snd_pcm_lib_malloc_pages() before the PCM hw_params
ops and snd_lib_pcm_free() after the PCM hw_free ops inside PCM core,
respectively. This allows drivers to drop the explicit calls of the
memory allocation / release functions, and it will be a good amount of
code reduction in the end of this patch series.
When the PCM substream is set to the managed buffer allocation mode,
the managed_buffer_alloc flag is set in the substream object. Since
some drivers want to know when a buffer is newly allocated or
re-allocated at hw_params callback (e.g. want to set up the additional
stuff for the given buffer only at allocation time), now PCM core
turns on buffer_changed flag when the buffer has changed.
The standard conversions to use the new API will be straightforward:
- Replace snd_pcm_lib_preallocate*() calls with the corresponding
snd_pcm_set_managed_buffer*(); the arguments should be unchanged
- Drop superfluous snd_pcm_lib_malloc() and snd_pcm_lib_free() calls;
the check of snd_pcm_lib_malloc() returns should be replaced with
the check of runtime->buffer_changed flag.
- If hw_params or hw_free becomes empty, drop them from PCM ops
Link: https://lore.kernel.org/r/20191117085308.23915-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Show and change sound card timer source with read-write info
file in proc filesystem. Initial string can still be set as
module parameter.
The timer source string value can be changed at any time,
but it is latched by PCM substream open callback (the first one
for a particular cable). At this point it is actually used, that
is the string is parsed, and the timer is looked up and opened.
The timer source is set for a loopback card (the same as initial
setting by module parameter), but every cable uses the value,
current at the moment of open.
Setting the value to empty string switches the timer to jiffies.
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-8-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
to do synchronous audio forwarding between hardware sound card and aloop
devices. Such an audio route could look like the following:
Sound card -> Loopback application -> ALSA loop device -> arecord
In this case the loopback device should use the sound timer of the sound
card. Without this patch the loopback application has to implement an
adaptive sample rate converter to align the different clocks of the
different ALSA devices.
The used timer can be selected by referring to a sound card, its device
and subdevice, when loading the module:
$ modprobe snd_aloop enable=1 timer_source=[<card>[.<dev>[.<subdev>]]]
<card> is the name (id) of the sound card or a card number.
<dev> and <subdev> are device and subdevice numbers (defaults are 0).
Empty string as a value of timer_source= parameter enables previous
functionality (using jiffies timer).
Signed-off-by: Timo Wischer <twischer@de.adit-jv.com>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-7-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This commit does not change the behaviour. It only separates the jiffies
timer specific implementation from the generic part.
Signed-off-by: Timo Wischer <twischer@de.adit-jv.com>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-5-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This commit only refactors the implementation. It does not change the
behaviour.
It is required to support other timers (e.g sound timer).
Signed-off-by: Timo Wischer <twischer@de.adit-jv.com>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-4-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This is required for additional timer implementations which could detect
errors and want to throw them.
Signed-off-by: Timo Wischer <twischer@de.adit-jv.com>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-3-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Describe the unit of the variables used to calculate the hw pointer
depending on jiffies ticks.
Signed-off-by: Timo Wischer <twischer@de.adit-jv.com>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20191120174955.6410-2-andrew_gabbasov@mentor.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
geneve RFC (draft-ietf-nvo3-geneve-14) allows a geneve packet to carry
multiple geneve opts, so it's necessary for lwtunnel to support adding
multiple geneve opts in one lwtunnel route. But vxlan and erspan opts
are still only allowed to add one option.
With this patch, iproute2 could make it like:
# ip r a 1.1.1.0/24 encap ip id 1 geneve_opts 0:0:12121212,1:2:12121212 \
dst 10.1.0.2 dev geneve1
# ip r a 1.1.1.0/24 encap ip id 1 vxlan_opts 456 \
dst 10.1.0.2 dev erspan1
# ip r a 1.1.1.0/24 encap ip id 1 erspan_opts 1:123:0:0 \
dst 10.1.0.2 dev erspan1
Which are pretty much like cls_flower and act_tunnel_key.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* Fix a merge damage that causes issues with high-throuput on AX200+;
* Support TX/RX antennas reporting;
* Small fix in DVM's BT link-quality code;
* Bump supported FW API version to 52;
* Yet another scan FW API update;
* Some clean-ups;
-----BEGIN PGP SIGNATURE-----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=0a16
-----END PGP SIGNATURE-----
Merge tag 'iwlwifi-next-for-kalle-2019-11-20' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-next
Patches intended for v5.5
* Fix a merge damage that causes issues with high-throuput on AX200+;
* Support TX/RX antennas reporting;
* Small fix in DVM's BT link-quality code;
* Bump supported FW API version to 52;
* Yet another scan FW API update;
* Some clean-ups;
These changes just clean up few typos, and there is one non-critical
correction of missing put_device() after calling of_platform_populate()
for display controller.
Naturally none of this is urgent and can be merged when suitable.
-----BEGIN PGP SIGNATURE-----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=2Ql6
-----END PGP SIGNATURE-----
Merge tag 'omap-for-v5.5/soc-late-signed' into omap-for-v5.5/ti-sysc-late
Few late SoC changes for v5.5 merge window
These changes just clean up few typos, and there is one non-critical
correction of missing put_device() after calling of_platform_populate()
for display controller.
Naturally none of this is urgent and can be merged when suitable.
The DRA7 CPSW MDIO functional clock (gmac_clkctrl DRA7_GMAC_GMAC_CLKCTRL 0)
is specified incorrectly, which is caused incorrect MDIO bus clock
configuration MDCLK. The correct CPSW MDIO functional clock is
gmac_main_clk (125MHz), which is the same as CPSW fck. Hence fix it.
Fixes: 1faa415c9c ("ARM: dts: Add fck for cpsw mdio for omap variants")
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
am57xx-beagle-x15 revb1 and revc have 3.3V connected to the eMMC I/O
lines. Update the pinmux name to reflect this.
Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
The MMC card detection GPIO polarity is active low on TAO3530, like in many
other similar boards. Now the card is not detected and it is unable to
mount rootfs from an SD card.
Fix this by using the correct polarity.
This incorrect polarity was defined already in the commit 30d95c6d70
("ARM: dts: omap3: Add Technexion TAO3530 SOM omap3-tao3530.dtsi") in v3.18
kernel and later changed to use defined GPIO constants in v4.4 kernel by
the commit 3a637e008e ("ARM: dts: Use defined GPIO constants in flags
cell for OMAP2+ boards").
While the latter commit did not introduce the issue I'm marking it with
Fixes tag due the v4.4 kernels still being maintained.
Fixes: 3a637e008e ("ARM: dts: Use defined GPIO constants in flags cell for OMAP2+ boards")
Cc: linux-stable <stable@vger.kernel.org> # 4.4+
Signed-off-by: Jarkko Nikula <jarkko.nikula@bitmer.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Adjust indentation from spaces to tab (+optional two spaces) as in
coding style with command like:
$ sed -e 's/^ /\t/' -i */Kconfig
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20191120133252.6365-1-krzk@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
This reverts commit 957ce0c6b8 (ASoC: soc-pcm: check symmetry after
hw_params).
That commit cause soc_pcm_params_symmetry can't take effect.
cpu_dai->rate, cpu_dai->channels and cpu_dai->sample_bits
are updated in the middle of soc_pcm_hw_params, so move
soc_pcm_params_symmetry to the end of soc_pcm_hw_params is
not a good solution, for judgement of symmetry in the function
is always true.
FIXME:
According to the comments of that commit, I think the case
described in the commit should disable symmetric_rates
in Back-End, rather than changing the position of
soc_pcm_params_symmetry.
Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Link: https://lore.kernel.org/r/1573555602-5403-1-git-send-email-shengjiu.wang@nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
kmemleak is used internally for a long time and as there isn't
any issue with it we can finally enable it in upstream.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Now have pixel clock PLL driver and we can change pixel clock rate
so we don't need to enforce one exact video mode. Moreover enforcing
video mode is harmful in case of we enforce mode which isn't
supported by the monitor we are using.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Use PLL driver instead of fixed-clock for PGU pixel clock.
That allows us to support wider range of graphic modes.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Adjust indentation from spaces to tab (+optional two spaces) as in
coding style with command like:
$ sed -e 's/^ /\t/' -i */Kconfig
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Link: https://lore.kernel.org/r/20191120133916.13595-1-krzk@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Adjust indentation from spaces to tab (+optional two spaces) as in
coding style with command like:
$ sed -e 's/^ /\t/' -i */Kconfig
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Link: https://lore.kernel.org/r/20191120133949.13996-1-krzk@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Tell the regulator framework to retrieve regulator init
data from the 'regulator' subnode, or from the parent mfd
device's platform data.
Example:
i2c0 {
tps61052@33 {
compatible = "ti,tps61052";
reg = <0x33>;
regulator {
regulator-min-microvolt = <5000000>;
regulator-max-microvolt = <5000000>;
regulator-always-on;
};
};
};
Tree: next-20191118
Signed-off-by: Sven Van Asbroeck <TheSven73@gmail.com>
Link: https://lore.kernel.org/r/20191119154611.29625-3-TheSven73@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This driver currently requires platform data to specify the
operational mode and regulator init data (in case of regulator
mode).
Optionally specify the operational mode by looking at the name
of the devicetree child node.
Example: put chip in regulator mode:
i2c0 {
tps61052@33 {
compatible = "ti,tps61052";
reg = <0x33>;
regulator {
regulator-min-microvolt = <5000000>;
regulator-max-microvolt = <5000000>;
regulator-always-on;
};
};
};
Tree: linux-next
Signed-off-by: Sven Van Asbroeck <TheSven73@gmail.com>
Link: https://lore.kernel.org/r/20191119154611.29625-2-TheSven73@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RST (reset-gpios) is low active so the driver must handle it
accordingly.
Add comments to explain clearly how the line is used.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Link: https://lore.kernel.org/r/20191120131753.6831-3-peter.ujfalusi@ti.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Use the standard name for the gpion in DT: reset-gpios
Document that the RST line is low active and update the example
accordingly.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Link: https://lore.kernel.org/r/20191120131753.6831-2-peter.ujfalusi@ti.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This ASCII string can carry additional information about
soundcard components or configuration. Add the possibility
to set this string via the ASoC card.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Cc: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20191119174933.25526-1-perex@perex.cz
Signed-off-by: Mark Brown <broonie@kernel.org>
The 0V9 and 1V8 supplies power the PCIe block in the SoC itself, and
are thus fundamental to PCIe being usable at all. As such, it makes
sense to treat them as non-optional and rely on dummy regulators if
not explicitly described.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Now that all the PCI host drivers are using pci_parse_request_of_pci_ranges(),
make devm_of_pci_get_host_bridge_resources() static.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Now that the helpers provide the inbound resources in the host bridge
'dma_ranges' resource list, convert Renesas R-Car PCIe host bridge to
use the resource list to setup the inbound addresses.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Simon Horman <horms@verge.net.au>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Now that the helpers provide the inbound resources in the host bridge
'dma_ranges' resource list, convert Broadcom iProc host bridge to use
the resource list to setup the inbound addresses.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Ray Jui <rjui@broadcom.com>
Cc: Scott Branden <sbranden@broadcom.com>
Cc: bcm-kernel-feedback-list@broadcom.com
Now that the helpers provide the inbound resources in the host bridge
'dma_ranges' resource list, convert the Xgene host bridge to use the
resource list to setup the inbound addresses.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Cc: Toan Le <toan@os.amperecomputing.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Now that the helpers provide the inbound resources in the host bridge
'dma_ranges' resource list, convert the v3-semi host bridge to use
the resource list to setup the inbound addresses.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Now that the helpers provide the inbound resources in the host bridge
'dma_ranges' resource list, convert Faraday ftpci100 host bridge to use
the resource list to setup the inbound addresses.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Extend devm_of_pci_get_host_bridge_resources() and
pci_parse_request_of_pci_ranges() helpers to also parse the inbound
addresses from DT 'dma-ranges' and populate a resource list with the
translated addresses. This will help ensure 'dma-ranges' is always
parsed in a consistent way.
Tested-by: Srinath Mannam <srinath.mannam@broadcom.com>
Tested-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com> # for AArdvark
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Srinath Mannam <srinath.mannam@broadcom.com>
Reviewed-by: Andrew Murray <andrew.murray@arm.com>
Acked-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
Cc: Jingoo Han <jingoohan1@gmail.com>
Cc: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Will Deacon <will@kernel.org>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Toan Le <toan@os.amperecomputing.com>
Cc: Ley Foon Tan <lftan@altera.com>
Cc: Tom Joseph <tjoseph@cadence.com>
Cc: Ray Jui <rjui@broadcom.com>
Cc: Scott Branden <sbranden@broadcom.com>
Cc: bcm-kernel-feedback-list@broadcom.com
Cc: Ryder Lee <ryder.lee@mediatek.com>
Cc: Karthikeyan Mitran <m.karthikeyan@mobiveil.co.in>
Cc: Hou Zhiqiang <Zhiqiang.Hou@nxp.com>
Cc: Simon Horman <horms@verge.net.au>
Cc: Shawn Lin <shawn.lin@rock-chips.com>
Cc: Heiko Stuebner <heiko@sntech.de>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: rfi@lists.rocketboards.org
Cc: linux-mediatek@lists.infradead.org
Cc: linux-renesas-soc@vger.kernel.org
Cc: linux-rockchip@lists.infradead.org
In commit a79ca656b6 ("drm/i915: Push the wakeref->count deferral to
the backend"), I erroneously concluded that we last modify the engine
inside __i915_request_commit() meaning that we could enable concurrent
submission for userspace as we enqueued this request. However, this
falls into a trap with other users of the engine->kernel_context waking
up and submitting their request before the idle-switch is queued, with
the result that the kernel_context is executed out-of-sequence most
likely upsetting the GPU and certainly ourselves when we try to retire
the out-of-sequence requests.
As such we need to hold onto the effective engine->kernel_context mutex
lock (via the engine pm mutex proxy) until we have finish queuing the
request to the engine.
v2: Serialise against concurrent intel_gt_retire_requests()
v3: Describe the hairy locking scheme with intel_gt_retire_requests()
for future reference.
v4: Combine timeline->lock and engine pm release; it's hairy.
Fixes: a79ca656b6 ("drm/i915: Push the wakeref->count deferral to the backend")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191120165514.3955081-2-chris@chris-wilson.co.uk
The general concept was that intel_timeline.active_count was locked by
the intel_timeline.mutex. The exception was for power management, where
the engine->kernel_context->timeline could be manipulated under the
global wakeref.mutex.
This was quite solid, as we always manipulated the timeline only while
we held an engine wakeref.
And then we started retiring requests outside of struct_mutex, only
using the timelines.active_list and the timeline->mutex. There we
started manipulating intel_timeline.active_count outside of an engine
wakeref, and so introduced a race between __engine_park() and
intel_gt_retire_requests(), a race that could result in the
engine->kernel_context not being added to the active timelines and so
losing requests, which caused us to keep the system permanently powered
up [and unloadable].
The race would be easy to close if we could take the engine wakeref for
the timeline before we retire -- except timelines are not bound to any
engine and so we would need to keep all active engines awake. The
alternative is to guard intel_timeline_enter/intel_timeline_exit for use
outside of the timeline->mutex.
Fixes: e5dadff4b0 ("drm/i915: Protect request retirement with timeline->mutex")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191120165514.3955081-1-chris@chris-wilson.co.uk
