]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/log
mirror_ubuntu-zesty-kernel.git
6 years agodrm/vc4: Verify at boot that CMA doesn't cross a 256MB boundary.
Eric Anholt [Mon, 12 Oct 2015 15:58:08 +0000 (08:58 -0700)]
drm/vc4: Verify at boot that CMA doesn't cross a 256MB boundary.

I've seen lots of users cranking CMA up higher, so throw an error if
they do.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Fix OOPSes from trying to cache a partially constructed BO.
Eric Anholt [Thu, 9 Feb 2017 17:23:34 +0000 (09:23 -0800)]
drm/vc4: Fix OOPSes from trying to cache a partially constructed BO.

If a CMA allocation failed, the partially constructed BO would be
unreferenced through the normal path, and we might choose to put it in
the BO cache.  If we then reused it before it expired from the cache,
the kernel would OOPS.

Signed-off-by: Eric Anholt <eric@anholt.net>
Fixes: c826a6e10644 ("drm/vc4: Add a BO cache.")
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Fulfill user BO creation requests from the kernel BO cache.
Eric Anholt [Wed, 8 Feb 2017 23:00:54 +0000 (15:00 -0800)]
drm/vc4: Fulfill user BO creation requests from the kernel BO cache.

The from_cache flag was actually "the BO is invisible to userspace",
so we can repurpose to just zero out a cached BO and return it to
userspace.

Improves wall time for a loop of 5 glsl-algebraic-add-add-1 by
-1.44989% +/- 0.862891% (n=28, 1 outlier removed from each that
appeared to be other system noise)

Note that there's an intel-gpu-tools test to check for the proper
zeroing behavior here, which we continue to pass.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Fix sending of page flip completion events in FKMS mode.
Eric Anholt [Thu, 2 Feb 2017 17:42:18 +0000 (09:42 -0800)]
drm/vc4: Fix sending of page flip completion events in FKMS mode.

In the rewrite of vc4_crtc.c for fkms, I dropped the part of the
CRTC's atomic flush handler that moved the completion event from the
proposed atomic state change to the CRTC's current state.  That meant
that when full screen pageflipping happened (glxgears -fullscreen in
X, compton, por weston), the app would end up blocked firever waiting
to draw its next frame.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Add DRM_DEBUG_ATOMIC for the insides of fkms.
Eric Anholt [Thu, 2 Feb 2017 01:10:09 +0000 (17:10 -0800)]
drm/vc4: Add DRM_DEBUG_ATOMIC for the insides of fkms.

Trying to debug weston on fkms involved figuring out what calls I was
making to the firmware.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Name the primary and cursor planes in fkms.
Eric Anholt [Thu, 2 Feb 2017 01:09:18 +0000 (17:09 -0800)]
drm/vc4: Name the primary and cursor planes in fkms.

This makes debugging nicer, compared to trying to remember what the
IDs are.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoBCM270X: Disable VEC unless vc4-kms-v3d is present.
Eric Anholt [Mon, 23 Jan 2017 19:41:54 +0000 (11:41 -0800)]
BCM270X: Disable VEC unless vc4-kms-v3d is present.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoARM: dts: bcm283x: Enable the VEC IP on all RaspberryPi boards
Boris Brezillon [Fri, 2 Dec 2016 13:48:13 +0000 (14:48 +0100)]
ARM: dts: bcm283x: Enable the VEC IP on all RaspberryPi boards

Enable the VEC IP on all RaspberryPi boards.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
(cherry picked from commit 5ab1a37c6027c114a87a1ae32cfc5ef303d643c5)
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoARM: dts: bcm283x: Add VEC node in bcm283x.dtsi
Boris Brezillon [Fri, 2 Dec 2016 13:48:12 +0000 (14:48 +0100)]
ARM: dts: bcm283x: Add VEC node in bcm283x.dtsi

Add the VEC (Video EnCoder) node definition in bcm283x.dtsi.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
(cherry picked from commit b899c45208d6f204a6da9a1132577993eeecf0fb)
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Add DSI driver
Eric Anholt [Wed, 10 Feb 2016 19:42:32 +0000 (11:42 -0800)]
drm/vc4: Add DSI driver

The DSI0 and DSI1 blocks on the 2835 are related hardware blocks.
Some registers move around, and the featureset is slightly different,
as DSI1 (the 4-lane DSI) is a later version of the hardware block.
This driver doesn't yet enable DSI0, since we don't have any hardware
to test against, but it does put a lot of the register definitions and
code in place.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Add support for feeding DSI encoders from the pixel valve.
Eric Anholt [Thu, 11 Feb 2016 00:17:29 +0000 (16:17 -0800)]
drm/vc4: Add support for feeding DSI encoders from the pixel valve.

We have to set a different pixel format, which tells the hardware to
use the pix_width field that's fed in sideband from the DSI encoder to
divide the "pixel" clock.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/vc4: Set up SCALER_DISPCTRL at boot.
Eric Anholt [Thu, 15 Sep 2016 14:25:23 +0000 (15:25 +0100)]
drm/vc4: Set up SCALER_DISPCTRL at boot.

We want the HVS on, obviously, and we also want DSP3 (PV1's source) to
be muxed from HVS channel 2 like we expect in vc4_crtc.c.  The
firmware wasn't setting the DSP3 mux up when both the LCD and HDMI
were disabled.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoBCM270X: Enable the DSI panel node in the VC4 overlay.
Eric Anholt [Thu, 2 Jun 2016 22:09:35 +0000 (15:09 -0700)]
BCM270X: Enable the DSI panel node in the VC4 overlay.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoARM: bcm2835: dt: Add the DSI module nodes and clocks.
Eric Anholt [Tue, 13 Dec 2016 23:15:10 +0000 (15:15 -0800)]
ARM: bcm2835: dt: Add the DSI module nodes and clocks.

The modules stay disabled by default, and if you want to enable DSI
you'll need an overlay that connects a panel to it.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoBCM270X: Add the DSI panel to the defconfig.
Eric Anholt [Thu, 2 Jun 2016 19:29:45 +0000 (12:29 -0700)]
BCM270X: Add the DSI panel to the defconfig.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agodrm/panel: Add support for the Raspberry Pi 7" Touchscreen.
Eric Anholt [Tue, 26 Apr 2016 20:46:13 +0000 (13:46 -0700)]
drm/panel: Add support for the Raspberry Pi 7" Touchscreen.

This driver communicates with the Atmel microcontroller for sequencing
the poweron of the TC358762 DSI-DPI bridge and controlling the
backlight PWM.

The following lines are required in config.txt, to keep the firmware
from trying to bash our I2C lines and steal the DSI interrupts:

    disable_touchscreen=1
    ignore_lcd=2
    mask_gpu_interrupt1=0x1000

This means that the firmware won't power on the panel at boot time (no
rainbow) and the touchscreen input won't work.  The native input
driver for the touchscreen still needs to be written.

v2: Set the same default orientation as the closed source firmware
    used, which is the best for viewing angle.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoclk: bcm2835: Add leaf clock measurement support, disabled by default
Eric Anholt [Tue, 17 Jan 2017 20:31:57 +0000 (07:31 +1100)]
clk: bcm2835: Add leaf clock measurement support, disabled by default

This proved incredibly useful during debugging of the DSI driver, to
see if our clocks were running at rate we requested.  Let's leave it
here for the next person interacting with clocks on the platform (and
so that hopefully we can just hook it up to debugfs some day).

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
(cherry picked from commit 3f9195811d8d829556c4cd88d3f9e56a80d5ba60)
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoclk: bcm2835: Register the DSI0/DSI1 pixel clocks.
Eric Anholt [Tue, 17 Jan 2017 20:31:56 +0000 (07:31 +1100)]
clk: bcm2835: Register the DSI0/DSI1 pixel clocks.

The DSI pixel clocks are muxed from clocks generated in the analog phy
by the DSI driver.  In order to set them as parents, we need to do the
same name lookup dance on them as we do for our root oscillator.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
(cherry picked from commit 8a39e9fa578229fd4604266c6ebb1a3a77d7994c)
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoclk: bcm2835: Don't rate change PLLs on behalf of DSI PLL dividers.
Eric Anholt [Tue, 17 Jan 2017 20:31:55 +0000 (07:31 +1100)]
clk: bcm2835: Don't rate change PLLs on behalf of DSI PLL dividers.

Our core PLLs are intended to be configured once and left alone.  With
the SET_RATE_PARENT, asking to set the PLLD_DSI1 clock rate would
change PLLD just to get closer to the requested DSI clock, thus
changing PLLD_PER, the UART and ethernet PHY clock rates downstream of
it, and breaking ethernet.

We *do* want PLLH to change so that PLLH_AUX can be exactly the value
we want, though.  Thus, we need to have a per-divider policy of
whether to pass rate changes up.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
(cherry picked from commit 55486091bd1e1c5ed28c43c0d6b3392468a9adb5)
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agovchiq_arm: Tweak the logging output
Phil Elwell [Mon, 7 Mar 2016 15:05:11 +0000 (15:05 +0000)]
vchiq_arm: Tweak the logging output

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoclk: bcm2835: Mark GPIO clocks enabled at boot as critical.
Eric Anholt [Tue, 10 May 2016 00:28:18 +0000 (17:28 -0700)]
clk: bcm2835: Mark GPIO clocks enabled at boot as critical.

These divide off of PLLD_PER and are used for the ethernet and wifi
PHYs source PLLs.  Neither of them is currently represented by a phy
device that would grab the clock for us.

This keeps other drivers from killing the networking PHYs when they
disable their own clocks and trigger PLLD_PER's refcount going to 0.

v2: Skip marking as critical if they aren't on at boot.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agopinctrl-bcm2835: Fix interrupt handling for GPIOs 28-31 and 46-53
Phil Elwell [Tue, 24 Feb 2015 13:40:50 +0000 (13:40 +0000)]
pinctrl-bcm2835: Fix interrupt handling for GPIOs 28-31 and 46-53

Contrary to the documentation, the BCM2835 GPIO controller actually has
four interrupt lines - one each for the three IRQ groups and one common. Rather
confusingly, the GPIO interrupt groups don't correspond directly with the GPIO
control banks. Instead, GPIOs 0-27 generate IRQ GPIO0, 28-45 GPIO1 and
46-53 GPIO2.

Awkwardly, the GPIOS for IRQ GPIO1 straddle two 32-entry GPIO banks, so it is
cleaner to split out a function to process the interrupts for a single GPIO
bank.

This bug has only just been observed because GPIOs above 27 can only be
accessed on an old Raspberry Pi with the optional P5 header fitted, where
the pins are often used for I2S instead.

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agobuild/arm64: Add rules for .dtbo files for dts overlays
Khem Raj [Sat, 11 Feb 2017 01:57:08 +0000 (17:57 -0800)]
build/arm64: Add rules for .dtbo files for dts overlays

We now create overlays as .dtbo files.

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoASoC: bcm2835_i2s.c: relax the ch2 register setting for 8 channels
Matt Flax [Wed, 8 Mar 2017 10:13:24 +0000 (21:13 +1100)]
ASoC: bcm2835_i2s.c: relax the ch2 register setting for 8 channels

This patch allows ch2 registers to be set for 8 channels of audio.

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoAdd support for the AudioInjector.net Octo sound card
Matt Flax [Wed, 8 Mar 2017 09:04:13 +0000 (20:04 +1100)]
Add support for the AudioInjector.net Octo sound card

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agobcm2835-v4l2: Fix buffer overflow problem
Dave Stevenson [Fri, 10 Mar 2017 14:43:15 +0000 (14:43 +0000)]
bcm2835-v4l2: Fix buffer overflow problem

https://github.com/raspberrypi/linux/issues/1447
port_parameter_get() failed to account for the header
(u32 id and u32 size) in the size before memcpying
the response into the response buffer, so overrunning
the provided buffer by 8 bytes.

Account for those bytes, and also a belt-and-braces
check to ensure we never copy more than *value_size
bytes into value.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoconfig: Add RTL8XXXU wifi module
popcornmix [Sun, 5 Mar 2017 11:46:41 +0000 (11:46 +0000)]
config: Add RTL8XXXU wifi module

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoconfig: Add CONFIG_CRYPTO_LZ4
Phil Elwell [Tue, 7 Mar 2017 19:48:23 +0000 (19:48 +0000)]
config: Add CONFIG_CRYPTO_LZ4

Enabling this options allows LZ4 memory compression.

Fixes: https://github.com/raspberrypi/linux/issues/1875
Signed-off-by: Phil Elwell <phil@raspberrypi.org>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoclk-bcm2835: Read max core clock from firmware
Phil Elwell [Mon, 6 Mar 2017 09:06:18 +0000 (09:06 +0000)]
clk-bcm2835: Read max core clock from firmware

The VPU is responsible for managing the core clock, usually under
direction from the bcm2835-cpufreq driver but not via the clk-bcm2835
driver. Since the core frequency can change without warning, it is
safer to report the maximum clock rate to users of the core clock -
I2C, SPI and the mini UART - to err on the safe side when calculating
clock divisors.

If the DT node for the clock driver includes a reference to the
firmware node, use the firmware API to query the maximum core clock
instead of reading the divider registers.

Prior to this patch, a "100KHz" I2C bus was sometimes clocked at about
160KHz. In particular, switching to the 4.9 kernel was likely to break
SenseHAT usage on a Pi3.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoUBUNTU: Start new release
Tim Gardner [Fri, 17 Mar 2017 15:24:56 +0000 (09:24 -0600)]
UBUNTU: Start new release

Ignore: yes
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoUBUNTU: Ubuntu-raspi2-4.10.0-1000.2
Tim Gardner [Mon, 6 Mar 2017 17:29:21 +0000 (10:29 -0700)]
UBUNTU: Ubuntu-raspi2-4.10.0-1000.2

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoUBUNTU: Start new release
Tim Gardner [Mon, 6 Mar 2017 17:21:14 +0000 (10:21 -0700)]
UBUNTU: Start new release

Ignore: yes
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoUBUNTU: Ubuntu-raspi2-4.10.0-1000.1
Tim Gardner [Fri, 3 Mar 2017 18:26:16 +0000 (11:26 -0700)]
UBUNTU: Ubuntu-raspi2-4.10.0-1000.1

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoUBUNTU: Enabled raspi2 packaging
Tim Gardner [Thu, 2 Mar 2017 15:45:01 +0000 (08:45 -0700)]
UBUNTU: Enabled raspi2 packaging

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
6 years agoBCM270X_DT: Add bcm2708-rpi-0-w.dts
Phil Elwell [Tue, 24 May 2016 15:30:05 +0000 (16:30 +0100)]
BCM270X_DT: Add bcm2708-rpi-0-w.dts

Add DT support for the Pi Zero W. N.B. It will not be loaded
automatically without a corresponding change to the firmware.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agobcm2835-gpio-exp: Driver for GPIO expander via mailbox service
Dave Stevenson [Mon, 20 Feb 2017 17:01:21 +0000 (17:01 +0000)]
bcm2835-gpio-exp: Driver for GPIO expander via mailbox service

Pi3 and Compute Module 3 have a GPIO expander that the
VPU communicates with.
There is a mailbox service that now allows control of this
expander, so add a kernel driver that can make use of it.

Pwr_led node added to device-tree for Pi3.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
6 years agoBCM2835-V4L2: Correctly denote key frames in encoded data
Dave Stevenson [Mon, 13 Feb 2017 13:11:41 +0000 (13:11 +0000)]
BCM2835-V4L2: Correctly denote key frames in encoded data

Forward MMAL key frame flags to the V4L2 buffers.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
6 years agoBCM2835-V4L2: Ensure H264 header bytes get a sensible timestamp
Dave Stevenson [Mon, 13 Feb 2017 11:10:50 +0000 (11:10 +0000)]
BCM2835-V4L2: Ensure H264 header bytes get a sensible timestamp

H264 header come off VC with 0 timestamps, which means they get a
strange timestamp when processed with VC/kernel start times,
particularly if used with the inline header option.
Remember the last frame timestamp and use that if set, or otherwise
use the kernel start time.

https://github.com/raspberrypi/linux/issues/1836

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
6 years agoamba_pl011: Round input clock up
Phil Elwell [Wed, 1 Mar 2017 16:07:39 +0000 (16:07 +0000)]
amba_pl011: Round input clock up

The UART clock is initialised to be as close to the requested
frequency as possible without exceeding it. Now that there is a
clock manager that returns the actual frequencies, an expected
48MHz clock is reported as 47999625. If the requested baudrate
== requested clock/16, there is no headroom and the slight
reduction in actual clock rate results in failure.

Detect cases where it looks like a "round" clock was chosen and
adjust the reported clock to match that "round" value. As the
code comment says:

/*
 * If increasing a clock by less than 0.1% changes it
 * from ..999.. to ..000.., round up.
 */

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoclk-bcm2835: Correct the prediv logic
Phil Elwell [Wed, 1 Mar 2017 16:06:53 +0000 (16:06 +0000)]
clk-bcm2835: Correct the prediv logic

If a clock has the prediv flag set, both the integer and fractional
parts must be scaled when calculating the resulting frequency.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoAdd overlay for ads1115 ADCs (#1864)
Scott Ellis [Wed, 1 Mar 2017 12:22:24 +0000 (07:22 -0500)]
Add overlay for ads1115 ADCs (#1864)

6 years agoAdd support for Fe-Pi audio sound card. (#1867)
Fe-Pi [Wed, 1 Mar 2017 11:42:43 +0000 (04:42 -0700)]
Add support for Fe-Pi audio sound card. (#1867)

Fe-Pi Audio Sound Card is based on NXP SGTL5000 codec.
Mechanical specification of the board is the same the Raspberry Pi Zero.
3.5mm jacks for Headphone/Mic, Line In, and Line Out.

Signed-off-by: Henry Kupis <fe-pi@cox.net>
6 years agosound: Support for Dion Audio LOCO-V2 DAC-AMP HAT
Miquel [Fri, 24 Feb 2017 19:51:06 +0000 (20:51 +0100)]
sound: Support for Dion Audio LOCO-V2 DAC-AMP HAT

Signed-off-by: Miquel Blauw <info@dionaudio.nl>
6 years agoconfig: add slcan kernel module
popcornmix [Fri, 27 Jan 2017 18:49:30 +0000 (18:49 +0000)]
config: add slcan kernel module

See: https://github.com/raspberrypi/linux/issues/1819

6 years agoconfig: Add ads1015 driver to config
Scott Ellis [Thu, 23 Feb 2017 16:56:20 +0000 (11:56 -0500)]
config: Add ads1015 driver to config

6 years agodmaengine: bcm2835: Fix cyclic DMA period splitting
Matthias Reichl [Mon, 20 Feb 2017 19:01:16 +0000 (20:01 +0100)]
dmaengine: bcm2835: Fix cyclic DMA period splitting

The code responsible for splitting periods into chunks that
can be handled by the DMA controller missed to update total_len,
the number of bytes processed in the current period, when there
are more chunks to follow.

Therefore total_len was stuck at 0 and the code didn't work at all.
This resulted in a wrong control block layout and audio issues because
the cyclic DMA callback wasn't executing on period boundaries.

Fix this by adding the missing total_len update.

Signed-off-by: Matthias Reichl <hias@horus.com>
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Tested-by: Clive Messer <clive.messer@digitaldreamtime.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
6 years agoclk-bcm2835: Add claim-clocks property
Phil Elwell [Mon, 13 Feb 2017 17:20:08 +0000 (17:20 +0000)]
clk-bcm2835: Add claim-clocks property

The claim-clocks property can be used to prevent PLLs and dividers
from being marked as critical. It contains a vector of clock IDs,
as defined by dt-bindings/clock/bcm2835.h.

Use this mechanism to claim PLLD_DSI0, PLLD_DSI1, PLLH_AUX and
PLLH_PIX for the vc4_kms_v3d driver.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agobrcmfmac: Mute expected startup 'errors'
Phil Elwell [Fri, 17 Feb 2017 15:26:13 +0000 (15:26 +0000)]
brcmfmac: Mute expected startup 'errors'

The brcmfmac WiFi driver always complains about the '00' country code
and the firmware version is reported as an error. Modify the driver to
ignore '00' silently and display firmware version at INFO level.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoBCM270X_DT: Add SMSC ethernet controller to DT
Phil Elwell [Fri, 17 Feb 2017 09:47:11 +0000 (09:47 +0000)]
BCM270X_DT: Add SMSC ethernet controller to DT

With an ethernet node in the DT, a suitable firmware can populate the
local-mac-address property, removing the need for a downstream patch
to the driver to read its MAC address from a module parameter.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agosound: Suppress error message about deferrals
Phil Elwell [Thu, 9 Feb 2017 14:40:33 +0000 (14:40 +0000)]
sound: Suppress error message about deferrals

Since driver load deferrals are expected and will already
have resulted in a kernel message, suppress an essentially
duplicate error message from the RPi audio board drivers.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agosound: Demote deferral errors to INFO level
Phil Elwell [Thu, 9 Feb 2017 14:36:44 +0000 (14:36 +0000)]
sound: Demote deferral errors to INFO level

At present there is no mechanism to specify driver load order,
which can lead to deferrals and repeated retries until successful.
Since this situation is expected, reduce the dmesg level to
INFO and mention that the operation will be retried.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoirq-bcm2836: Avoid "Invalid trigger warning"
Phil Elwell [Thu, 9 Feb 2017 14:33:30 +0000 (14:33 +0000)]
irq-bcm2836: Avoid "Invalid trigger warning"

Initialise the level for each IRQ to avoid a warning from the
arm arch timer code.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoclk-bcm2835: Mark used PLLs and dividers CRITICAL
Phil Elwell [Mon, 13 Feb 2017 17:20:08 +0000 (17:20 +0000)]
clk-bcm2835: Mark used PLLs and dividers CRITICAL

The VPU configures and relies on several PLLs and dividers. Mark all
enabled dividers and their PLLs as CRITICAL to prevent the kernel from
switching them off.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agodwc_otg: fix summarize urb->actual_length for isochronous transfers
Martin Cerveny [Mon, 13 Feb 2017 16:23:47 +0000 (17:23 +0100)]
dwc_otg: fix summarize urb->actual_length for isochronous transfers

Kernel does not copy input data of ISO transfers to userspace
if actual_length is set only in ISO transfers and not summarized
in urb->actual_length. Fixes raspberrypi/linux#903

6 years agoconfig: enable Cirrus Logic Audio Card
Matthias Reichl [Sun, 22 Jan 2017 11:49:37 +0000 (12:49 +0100)]
config: enable Cirrus Logic Audio Card

Signed-off-by: Matthias Reichl <hias@horus.com>
6 years agoASoC: Add driver for Cirrus Logic Audio Card
Matthias Reichl [Sun, 22 Jan 2017 11:49:37 +0000 (12:49 +0100)]
ASoC: Add driver for Cirrus Logic Audio Card

Note: due to problems with deferred probing of regulators
the following softdep should be added to a modprobe.d file

softdep arizona-spi pre: arizona-ldo1

Signed-off-by: Matthias Reichl <hias@horus.com>
6 years agoBCM270x DT: Consolidate audio card overlays
Matthias Reichl [Sun, 22 Jan 2017 11:49:36 +0000 (12:49 +0100)]
BCM270x DT: Consolidate audio card overlays

Reference 3.3V / 5V system rails instead of instantiating local
regulators.

Add missing power supply properties for codecs where these are
required according to the DT bindings docs.

Signed-off-by: Matthias Reichl <hias@horus.com>
6 years agoBCM270x DT: expose 3.3V and 5V system rails
Matthias Reichl [Sun, 22 Jan 2017 11:49:36 +0000 (12:49 +0100)]
BCM270x DT: expose 3.3V and 5V system rails

Signed-off-by: Matthias Reichl <hias@horus.com>
6 years agoconfig: Enable regulator support
Matthias Reichl [Sun, 22 Jan 2017 11:49:36 +0000 (12:49 +0100)]
config: Enable regulator support

Signed-off-by: Matthias Reichl <hias@horus.com>
6 years agosquash: fix order of sound/soc/bcm makefile
popcornmix [Mon, 13 Feb 2017 15:33:47 +0000 (15:33 +0000)]
squash: fix order of sound/soc/bcm makefile

6 years agoARM64: Force hardware emulation of deprecated instructions.
Michael Zoran [Sat, 11 Feb 2017 09:18:31 +0000 (01:18 -0800)]
ARM64: Force hardware emulation of deprecated instructions.

6 years agoUpdate vfpmodule.c
Claggy3 [Sat, 11 Feb 2017 14:00:30 +0000 (14:00 +0000)]
Update vfpmodule.c

Christopher Alexander Tobias Schulze - May 2, 2015, 11:57 a.m.
This patch fixes a problem with VFP state save and restore related
to exception handling (panic with message "BUG: unsupported FP
instruction in kernel mode") present on VFP11 floating point units
(as used with ARM1176JZF-S CPUs, e.g. on first generation Raspberry
Pi boards). This patch was developed and discussed on

   https://github.com/raspberrypi/linux/issues/859

A precondition to see the crashes is that floating point exception
traps are enabled. In this case, the VFP11 might determine that a FPU
operation needs to trap at a point in time when it is not possible to
signal this to the ARM11 core any more. The VFP11 will then set the
FPEXC.EX bit and store the trapped opcode in FPINST. (In some cases,
a second opcode might have been accepted by the VFP11 before the
exception was detected and could be reported to the ARM11 - in this
case, the VFP11 also sets FPEXC.FP2V and stores the second opcode in
FPINST2.)

If FPEXC.EX is set, the VFP11 will "bounce" the next FPU opcode issued
by the ARM11 CPU, which will be seen by the ARM11 as an undefined opcode
trap. The VFP support code examines the FPEXC.EX and FPEXC.FP2V bits
to decide what actions to take, i.e., whether to emulate the opcodes
found in FPINST and FPINST2, and whether to retry the bounced instruction.

If a user space application has left the VFP11 in this "pending trap"
state, the next FPU opcode issued to the VFP11 might actually be the
VSTMIA operation vfp_save_state() uses to store the FPU registers
to memory (in our test cases, when building the signal stack frame).
In this case, the kernel crashes as described above.

This patch fixes the problem by making sure that vfp_save_state() is
always entered with FPEXC.EX cleared. (The current value of FPEXC has
already been saved, so this does not corrupt the context. Clearing
FPEXC.EX has no effects on FPINST or FPINST2. Also note that many
callers already modify FPEXC by setting FPEXC.EN before invoking
vfp_save_state().)

This patch also addresses a second problem related to FPEXC.EX: After
returning from signal handling, the kernel reloads the VFP context
from the user mode stack. However, the current code explicitly clears
both FPEXC.EX and FPEXC.FP2V during reload. As VFP11 requires these
bits to be preserved, this patch disables clearing them for VFP
implementations belonging to architecture 1. There should be no
negative side effects: the user can set both bits by executing FPU
opcodes anyway, and while user code may now place arbitrary values
into FPINST and FPINST2 (e.g., non-VFP ARM opcodes) the VFP support
code knows which instructions can be emulated, and rejects other
opcodes with "unhandled bounce" messages, so there should be no
security impact from allowing reloading FPEXC.EX and FPEXC.FP2V.

Signed-off-by: Christopher Alexander Tobias Schulze <cat.schulze@alice-dsl.net>
6 years agogpio_mem: Remove unnecessary dev_info output (#1830)
JamesH65 [Mon, 6 Feb 2017 15:24:47 +0000 (15:24 +0000)]
gpio_mem: Remove unnecessary dev_info output (#1830)

The open function was spamming syslog every time
called, so have removed call completely.

6 years agoAdd overlay for mcp3008 adc (#1818)
Scott Ellis [Fri, 27 Jan 2017 11:42:42 +0000 (06:42 -0500)]
Add overlay for mcp3008 adc (#1818)

Some example usage:

SPI0.0
dtparam=spi=on
dtoverlay=mcp3008:spi0-0-present

SPI0.1
dtparam=spi=on
dtoverlay=mcp3008:spi0-1-present

SPI0.0 and SPI0.1
dtparam=spi=on
dtoverlay=mcp3008:spi0-0-present,spi0-1-present

SPI1.0
dtparam=spi=on
dtoverlay=spi1-1cs
dtoverlay=mcp3008:spi1-0-present

SPI1.2
dtparam=spi=on
dtoverlay=spi1-1cs:cs0_pin=16
dtoverlay=mcp3008:spi1-0-present

SPI1.0 and SPI1.1
dtoverlay=spi1-2cs
dtoverlay=mcp3008:spi1-0-present,spi1-1-present

Changing the speed

SPI0.0
dtparam=spi=on
dtoverlay=mcp3008:spi0-0-present,spi0-0-speed=2000000

6 years agoconfig: Add SND_SOC_ADAU7002 codec module
Phil Elwell [Wed, 25 Jan 2017 21:17:23 +0000 (21:17 +0000)]
config: Add SND_SOC_ADAU7002 codec module

As there is now an overlay requiring it, build the codec module.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoASoC: A simple-card overlay for ADAU7002
chris johnson [Sun, 22 Jan 2017 03:27:31 +0000 (03:27 +0000)]
ASoC: A simple-card overlay for ADAU7002

Usage: `dtoverlay=adau7002-simple`

6 years agoconfig: Add CONFIG_TCP_CONG_BBR
popcornmix [Mon, 16 Jan 2017 16:33:54 +0000 (16:33 +0000)]
config: Add CONFIG_TCP_CONG_BBR
See: https://github.com/raspberrypi/linux/issues/1784

6 years agoconfig: Add additional network scheduling modules
popcornmix [Wed, 25 Jan 2017 11:30:38 +0000 (11:30 +0000)]
config: Add additional network scheduling modules

6 years agoBCM270X_DT: Add reference to audio_pins to CM dtb
Phil Elwell [Mon, 23 Jan 2017 17:36:50 +0000 (17:36 +0000)]
BCM270X_DT: Add reference to audio_pins to CM dtb

The CM1 dtb contains an empty audio_pins node, but no reference to it.
Adding the usual pinctrl reference from the audio node enables the
audremap overlay (and others) to easily turn on audio.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoconfig: Add CONFIG_USB_DWC2 and CONFIG_USB_GADGET
Phil Elwell [Thu, 19 Jan 2017 11:18:55 +0000 (11:18 +0000)]
config: Add CONFIG_USB_DWC2 and CONFIG_USB_GADGET

The introduction of CM3 makes gadget mode on 2709 a useful option,
so enable the building of the required modules. Note that these
modules are not loaded by default and must be enabled with a DT
overlay.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoARM64: Use dwc_otg driver by default for USB.
Michael Zoran [Sun, 15 Jan 2017 05:46:04 +0000 (21:46 -0800)]
ARM64: Use dwc_otg driver by default for USB.

If it breaks on anybody, they can use the standard device tree
overlays to switch back to the dwc2 driver.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Enable DWC_OTG Driver In ARM64 Build Config(bcmrpi3_defconfig)
Michael Zoran [Sun, 15 Jan 2017 05:45:03 +0000 (21:45 -0800)]
ARM64: Enable DWC_OTG Driver In ARM64 Build Config(bcmrpi3_defconfig)

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Round-Robin dispatch IRQs between CPUs.
Michael Zoran [Sun, 15 Jan 2017 05:43:57 +0000 (21:43 -0800)]
ARM64: Round-Robin dispatch IRQs between CPUs.

IRQ-CPU mapping is round robined on ARM64 to increase
concurrency and allow multiple interrupts to be serviced
at a time.  This reduces the need for FIQ.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64/DWC_OTG: Port dwc_otg driver to ARM64
Michael Zoran [Sun, 15 Jan 2017 05:33:51 +0000 (21:33 -0800)]
ARM64/DWC_OTG: Port dwc_otg driver to ARM64

In ARM64, the FIQ mechanism used by this driver is not current
implemented.   As a workaround, reqular IRQ is used instead
of FIQ.

In a separate change, the IRQ-CPU mapping is round robined
on ARM64 to increase concurrency and allow multiple interrupts
to be serviced at a time.  This reduces the need for FIQ.

Tests Run:

This mechanism is most likely to break when multiple USB devices
are attached at the same time.  So the system was tested under
stress.

Devices:

1. USB Speakers playing back a FLAC audio through VLC
   at 96KHz.(Higher then typically, but supported on my speakers).

2. sftp transferring large files through the buildin ethernet
   connection which is connected through USB.

3. Keyboard and mouse attached and being used.

Although I do occasionally hear some glitches, the music seems to
play quite well.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoconfig: Add CONFIG_MD_M25P80 and CONFIG_MD_SPI_NOR
Phil Elwell [Tue, 17 Jan 2017 14:39:39 +0000 (14:39 +0000)]
config: Add CONFIG_MD_M25P80 and CONFIG_MD_SPI_NOR

See: https://github.com/raspberrypi/linux/issues/1781

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoBCM270X_DT: Enable UART0 on CM3
Phil Elwell [Tue, 17 Jan 2017 11:34:58 +0000 (11:34 +0000)]
BCM270X_DT: Enable UART0 on CM3

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agospi-bcm2835: Disable forced software CS
Phil Elwell [Fri, 1 Jul 2016 21:09:24 +0000 (22:09 +0100)]
spi-bcm2835: Disable forced software CS

Select software CS in bcm2708_common.dtsi, and disable the automatic
conversion in the driver to allow hardware CS to be re-enabled with an
overlay.

See: https://github.com/raspberrypi/linux/issues/1547

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoBCM270X_DT: Add spi0-cs overlay
Phil Elwell [Mon, 16 Jan 2017 14:53:12 +0000 (14:53 +0000)]
BCM270X_DT: Add spi0-cs overlay

The spi0-cs overlay allows the software chip selectts to be modified
using the cs0_pin and cs1_pin parameters.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoARM64: Enable RTL8187/RTL8192CU wifi in build config
Michael Zoran [Sun, 15 Jan 2017 15:31:59 +0000 (07:31 -0800)]
ARM64: Enable RTL8187/RTL8192CU wifi in build config

These drivers build now, so they can be enabled back
in the build configuration just like they are for
32 bit.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Fix build break for RTL8187/RTL8192CU wifi
Michael Zoran [Sun, 15 Jan 2017 15:25:18 +0000 (07:25 -0800)]
ARM64: Fix build break for RTL8187/RTL8192CU wifi

These drivers use an ASM function from the base
system to compute the ipv6 checksum.  These functions
are not available on ARM64, probably because nobody
has bother to write them.  The base system does have
a generic "C" version, so a simple fix is to include
the header to use the generic version on ARM64 only.

A longer term solution would be to submit the necessary
ASM function to the upstream source.

With this change, these drivers now compile without
any errors on ARM64.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Enable Kernel Address Space Randomization (#1792)
Electron752 [Sat, 14 Jan 2017 10:54:26 +0000 (02:54 -0800)]
ARM64: Enable Kernel Address Space Randomization (#1792)

Randomization allows the mapping between virtual addresses and physical
address to be different on each boot.  This makes it more difficult
to exploit security vulnerabilities that require knowledge of fixed
hardware addresses.

The firmware generates a 8 byte random number during bootup and stores
it in the device tree under chosen/kaslr-seed. This number is used
to randomize the address mapping.

This change enables this feature in the build configuration for ARM64.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Run bcmrpi3_defconfig through savedefconfig.
Michael Zoran [Fri, 13 Jan 2017 03:14:03 +0000 (19:14 -0800)]
ARM64: Run bcmrpi3_defconfig through savedefconfig.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Enable HDMI audio and vc04_services in bcmrpi3_defconfig
Michael Zoran [Fri, 13 Jan 2017 03:10:07 +0000 (19:10 -0800)]
ARM64: Enable HDMI audio and vc04_services in bcmrpi3_defconfig

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64/SND_BCM2835: Port it to arm64.
Michael Zoran [Fri, 13 Jan 2017 03:05:46 +0000 (19:05 -0800)]
ARM64/SND_BCM2835: Port it to arm64.

In the messages sent to VCHIQ, SND_BCM2835 passes a callback
and a context into two 32 bit pointers.  Since this
message is interpreted by the firmware, it can't be easily
changed.  Luckily only one of these fields is actual used,
so on ARM64 only use one of the fields to store the upper
half of the 64 bit callback pointer.

The Kconfig is also changed to allow ARM audio to work for ARM64
as well.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Fix bad cast in vc04_services
Michael Zoran [Fri, 13 Jan 2017 02:56:54 +0000 (18:56 -0800)]
ARM64: Fix bad cast in vc04_services

The function vchiq_copy_from_user contains a non-portable
cast to uint32_t.  Convert this to a cast to unsigned long
which is portable.

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoARM64: Make it work again on 4.9 (#1790)
Electron752 [Thu, 12 Jan 2017 15:07:08 +0000 (07:07 -0800)]
ARM64: Make it work again on 4.9 (#1790)

* Invoke the dtc compiler with the same options used in arm mode.
* ARM64 now uses the bcm2835 platform just like ARM32.
* ARM64: Update bcmrpi3_defconfig

Signed-off-by: Michael Zoran <mzoran@crowfest.net>
6 years agoBCM270X_DT: Add pi3-disable-wifi overlay
Phil Elwell [Wed, 11 Jan 2017 13:01:21 +0000 (13:01 +0000)]
BCM270X_DT: Add pi3-disable-wifi overlay

pi3-disable-wifi is a minimal overlay to disable the onboard WiFi.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agoAdd driver_name paramater
Aaron Shaw [Tue, 10 Jan 2017 16:11:04 +0000 (16:11 +0000)]
Add driver_name paramater

Add driver_name parameter for use with 5.1 passthrough audio in LibreElec and other Kodi OSs

6 years agoAdd driver_name property
Aaron Shaw [Tue, 10 Jan 2017 16:05:41 +0000 (16:05 +0000)]
Add driver_name property

Add driver name property for use with 5.1 passthrough audio in LibreElec and other Kodi based OSs

6 years agopisound improvements:
Giedrius Trainavicius [Thu, 5 Jan 2017 00:38:16 +0000 (02:38 +0200)]
pisound improvements:

* Added a writable sysfs object to enable scripts / user space software
to blink MIDI activity LEDs for variable duration.
* Improved hw_param constraints setting.
* Added compatibility with S16_LE sample format.
* Exposed some simple placeholder volume controls, so the card appears
in volumealsa widget.

Signed-off-by: Giedrius Trainavicius <giedrius@blokas.io>
6 years agoDocument the si7020 option
Phil Elwell [Tue, 3 Jan 2017 21:27:46 +0000 (21:27 +0000)]
Document the si7020 option

6 years agoAdd support for Silicon Labs Si7013/20/21 humidity/temperature sensor.
Alex Tucker [Tue, 13 Dec 2016 19:50:18 +0000 (19:50 +0000)]
Add support for Silicon Labs Si7013/20/21 humidity/temperature sensor.

6 years agoraspberrypi-firmware: Export the general transaction function.
Eric Anholt [Wed, 14 Sep 2016 08:16:19 +0000 (09:16 +0100)]
raspberrypi-firmware: Export the general transaction function.

The vc4-firmware-kms module is going to be doing the MBOX FB call.

Signed-off-by: Eric Anholt <eric@anholt.net>
6 years agoi2c: bcm2835: Add debug support
Noralf Trønnes [Tue, 1 Nov 2016 14:15:41 +0000 (15:15 +0100)]
i2c: bcm2835: Add debug support

This adds a debug module parameter to aid in debugging transfer issues
by printing info to the kernel log. When enabled, status values are
collected in the interrupt routine and msg info in
bcm2835_i2c_start_transfer(). This is done in a way that tries to avoid
affecting timing. Having printk in the isr can mask issues.

debug values (additive):
1: Print info on error
2: Print info on all transfers
3: Print messages before transfer is started

The value can be changed at runtime:
/sys/module/i2c_bcm2835/parameters/debug

Example output, debug=3:
[  747.114448] bcm2835_i2c_xfer: msg(1/2) write addr=0x54, len=2 flags= [i2c1]
[  747.114463] bcm2835_i2c_xfer: msg(2/2) read addr=0x54, len=32 flags= [i2c1]
[  747.117809] start_transfer: msg(1/2) write addr=0x54, len=2 flags= [i2c1]
[  747.117825] isr: remain=2, status=0x30000055 : TA TXW TXD TXE  [i2c1]
[  747.117839] start_transfer: msg(2/2) read addr=0x54, len=32 flags= [i2c1]
[  747.117849] isr: remain=32, status=0xd0000039 : TA RXR TXD RXD  [i2c1]
[  747.117861] isr: remain=20, status=0xd0000039 : TA RXR TXD RXD  [i2c1]
[  747.117870] isr: remain=8, status=0x32 : DONE TXD RXD  [i2c1]

Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
6 years agodrm/vc4: Add a mode for using the closed firmware for display.
Eric Anholt [Wed, 14 Sep 2016 07:39:33 +0000 (08:39 +0100)]
drm/vc4: Add a mode for using the closed firmware for display.

Signed-off-by: Eric Anholt <eric@anholt.net>
6 years agoraspberrypi-firmware: Define the MBOX channel in the header.
Eric Anholt [Wed, 14 Sep 2016 08:18:09 +0000 (09:18 +0100)]
raspberrypi-firmware: Define the MBOX channel in the header.

Signed-off-by: Eric Anholt <eric@anholt.net>
6 years agovchiq_arm: Avoid use of mutex in add_completion
Phil Elwell [Mon, 20 Jun 2016 12:51:44 +0000 (13:51 +0100)]
vchiq_arm: Avoid use of mutex in add_completion

Claiming the completion_mutex within add_completion did prevent some
messages appearing twice, but provokes a deadlock caused by vcsm using
vchiq within a page fault handler.

Revert the use of completion_mutex, and instead fix the original
problem using more memory barriers.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agovchiq_arm: Add completion records under the mutex
Phil Elwell [Thu, 21 Apr 2016 12:49:32 +0000 (13:49 +0100)]
vchiq_arm: Add completion records under the mutex

An issue was observed when flushing openmax components
which generate a large number of messages returning
buffers to host.

We occasionally found a duplicate message from 16
messages prior, resulting in a buffer returned twice.

While only one thread adds completions, without the
mutex you don't get the protection of the automatic
memory barrier you get with synchronisation objects.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agovchiq_arm: Service callbacks must not fail
Phil Elwell [Wed, 23 Mar 2016 20:53:47 +0000 (20:53 +0000)]
vchiq_arm: Service callbacks must not fail

Service callbacks are not allowed to return an error. The internal callback
that delivers events and messages to user tasks does not enqueue them if
the service is closing, but this is not an error and should not be
reported as such.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
6 years agovchiq_arm: Access the dequeue_pending flag locked
Phil Elwell [Wed, 23 Mar 2016 14:16:25 +0000 (14:16 +0000)]
vchiq_arm: Access the dequeue_pending flag locked

Reading through this code looking for another problem (now found in userland)
the use of dequeue_pending outside a lock didn't seem safe.

Signed-off-by: Phil Elwell <phil@raspberrypi.org>