| Age | Commit message (Collapse) | Author |
|---|
|
Framebuffer suspend event comes before framebuffer
blank event when system entering suspend mode.
So, if another driver calls FB_EVENT_BLANK
notifier chain, i.e, Android IPUv3 fb driver,
the suspend and blank event handlers registered
in TVE driver will be called. We used to disable
TVE clock in suspend handler, so it causes system
hang when we try to access TVE register when go
to blank handler. This patch fixes this issue.
Signed-off-by: Liu Ying <Ying.Liu@freescale.com>
|
|
This patch add pm_cap for mx_sdhci driver, so the client mmc driver can use
sdio_set_host_pm_flags() to set the MMC_PM_KEEP_POWER.
Also if the driver found MMC_PM_WAKE_SDIO_IRQ in pm_flags,
it should not unregister irq of this host, since the client
maybe use this irq to support wake up.
Signed-off-by: Zhang Jiejing <jiejing.zhang@freescale.com>
|
|
When using SNAPSHOT update scheme, submitting FULL mode updates can easily
lead to an endlessly looping sequence of collisions if any updates are active
when the new FULL mode update is submitted. Thus, we must first flush any
updates out before submitting a new FULL mode update.
Signed-off-by: Danny Nold <dannynold@freescale.com>
|
|
The vpu code in driver for mx32/mx37 is out of date already, and
we won't support the platforms, so clean the code for removing it.
Signed-off-by: Sammy He <r62914@freescale.com>
|
|
improve mpr121 capacitive key sensitivity by change threshold of
release and touch and add a quick charge bit in init which is not
documented in MPR121's data sheet.
move macro define from header to source file.
Signed-off-by: Zhang Jiejing <jiejing.zhang@freescale.com>
|
|
- This patch changes the way that buffers are allocated and used for the
case where PxP alignment limitations are not met (a copy from the
update source buffer into a temporary buffer before initiating the
PxP processing task is required). Previously, a buffer was allocated
on a per-update basis, resulting in a large number of copy buffers. Now,
we allocate just one copy buffer that can be reused any time the PxP
alignment workaround is required. This is okay because it is assured
that only one PxP transaction can be prepared at a time.
Signed-off-by: Danny Nold <dannynold@freescale.com>
|
|
Remove build warning message from ipu, mc13892 and gps drivers
Signed-off-by: Lily Zhang <r58066@freescale.com>
|
|
kernel compilation with a warning about fec.
drivers/net/fec.c:929: warning: initialization from incompatible pointer type.
This is caused by wrongly define get_link callback return type to int.
Signed-off-by: Zeng Zhaoming <b32542@freescale.com>
|
|
The problem is in drivers/mmc/host/mx_sdhci.c.
DIGCTL_BASE_ADDR is undefined.
Define it for mx25/31/35 for pass building.
Signed-off-by: Dong Aisheng <b29396@freescale.com>
|
|
|
|
|
|
On setting the alarm and if the wakeup event has been enabled, configure
the PMIC IRQ as wakeup event.
|
|
This prints a message when if the alarm is to trigger immediately.
|
|
|
|
Unmask the RTC alarm IRQ, and mask other unwanted IRQs by default, as
the battery charge events.
Also, configure the battery not to charge by default, let the battery
driver do this if needed.
|
|
These are the same changes implemented in the MX51 that allows to use
the power/wakeup GPIO interface.
|
|
Only set the PMIC wake up source if configured to use the power on button.
|
|
They were continuously polluting the console at our default loglevel.
|
|
|
|
This commit adds support for the LQ121K1LG11 LVDS display
on the CERTMX53 board. To enable this support, select that
display in the video variable:
video=displayfb:LVDS@LQ121K1LG11
and configure the lvds bridge with following kernel cmdline
entry:
std_bootarg=ldb=single,di=0,ch0_map=JEIDA
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
blanking period"
This reverts commit 192d207beb81ce5ec17968a7359da8ba42e5578b.
This code was causing a problem on some diplays by moving
to the right in 1 pixel the framebuffer.
This was happening on the LQ70 display and also on some
screens with the HDMI interface.
Fixes #40184.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This is the other half of 7b97ff0a7e268d03396314c077cd04f26d3869bf.
On a half duplex SPI transfer, a NULL RX or TX buffer can be passed to
indicate it. This caused a NULL pointer exception.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
When setting the mode to Measurement (0x01 to Mode Control Register
at address 0x16), the DRDY pin was by default output to INT1/DRDY pin
causing an interrupt. It looks like the interrupt is not correctly
handled and hangs the system.
To avoid this, when setting any mode set also DRPD flag to
not output Data Ready status to DRDY/INT1 pin.
The interrupt logic must be reviewed because it hangs the system.
Currently, this driver does not correctly support interrupts.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
|
|
|
|
Any I2C bus should work with the driver. In the CCWMX51 it has been
tested with both I2C_MXC and I2C_IMX (the default).
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The mma7455 driver is not working with the IMX_I2C driver, but it is
with the MXC_I2C one.
This commit modifies the I2C send and receive driver functions to
use i2c_master_send() and i2c_master_rcve() instead of directly
i2c_transfer. This change makes the driver work with both I2C bus drivers.
This fixes Vantive 40023.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This option was moved to platform code in ENGR00140893-1 (9672393), in
order to have a different value between MX51 and MX53 on freescale's
multiplatform kernel.
Freescale's recommendation is to disable it in the MX51 and enable it on
the MX53.
However, on our testing we have found it working better on some conditions
with the IRAM enabled on the MX51.
This commit reintroduces the option as a configuration element, and defaults
it to enabled as that was the DEL 5.6 default.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The power button will now suspend and resume the device.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
When using fbdevsink, the application needs more time to provide buffers.
|
|
Some framebuffer elements like fbdevsink will not use this output driver and
hence the buffer_cnt will be zero. In that case, we should't try to continue
with the mxc_v4l2_streamon.
|
|
When the mfw_v4lsink gstreamer element is used for camera capture, the
mxc_streamon needs to be scheduled before mxc_v4lout_streamon, otherwise
the application won't supply any more buffers to the output ready_q and
the capture will not resume.
|
|
When calling mxc_v4l2out_streamoff from suspend, we don't want it to
empty the ready queue as the application on resume will expect it to
be as it was before suspend.
|
|
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This fixes the video suspend and resume issues for videos that need to be
resized using the IPU post processor. The PP was not being reinitialized
correctly on suspend/resume and the video was frozen on the screen.
This fixes #40009.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
A patch by Freescale was causing a kernel warning when removing the
USB cable on cc9m2443js platforms.
Wrap this code under CONFIG_ARCH_MX5 define.
This fixes vantive 39792.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit 373ad26fb851d6b3fdb2b8db89f396d9b916d7f2)
|
|
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit 240ab1b12fdfe3978c175b44c6f0709290c8b581)
|
|
The CSI registers lose their configuration on suspend. Make sure they are
reinitialized on resume.
|
|
We need to call mxc_streamoff on suspend, otherwise the CSI_MEM0 DMA
channel (15) will not be disabled and the IPU clock will be left enabled
on suspend. That will make the target not react to wake up events.
However, we don't want mxc_streamoff to reinitialize the buffer queues as
the application will desynchronize.
This is part of a set of fixes for #40007.
|
|
The application needs to queue a buffer before we can re initialize the
stream.
|
|
This condition has no effect on the functionality, don't scare the user with
an error message.
|
|
With the camera capture on, the MEM_PP_MEM channel is initialized, but on
suspend the channel is not uninitialized. As a consequence the suspend
happens with the IPU clock enabled, and the MX CPU will not respond to
the programmed wake up events.
On this commit we implement the un-initialization and re-initialization
of that IPU channel on the new suspend and resume functions.
Also, we need to call mxc_streamoff from mxc_v4l2_suspend to disable
the CSI_MEM0 IPU channel. Unless all active IPU channels are disabled
the IPU clock will be active and the wake up event will hang the target.
|
|
|
|
This reverts commit 27092fa602c11d0abb03ea05601fdb4a8dd15cff.
This fixes #37737.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Support for the two cameras was incomplete and caused the
camera2 be not detected if the camera1 was not connected
(see vantive 40526). Also, two devices (video0 and video1)
were created no matter if only one camera was enabled in the
kernel config.
The commit, wraps the different arrays for the two camera sensors
into the kernel config defines for each camera. The function that
returns the CSI of the camera is substituted to return the index
of the camera in the arrays, instead.
The CSI of the camera is retrieved from field 'id' of the
mxc_v4l2_devices array.
The cameras (if enabled in kernel config) are registered, no matter
if the sensor is detected, otherwise camera2 would not be detected
if camera1 is not connected.
This fixes vantive 40526.
NOTE: a race appeared between function init_camera_struct() in
mxc_v4l2_capture.c and function mt9v111_probe() in mt9v111.c.
The first function registers the device which calls the probe.
Apparently the race happened when accessing the i2c_client pointer.
Moving this instruction below in the probe() made the race disappear
but it might show up again in the future.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit 870484c800401ab91945af30713756fd292a313f)
|
|
- When card detect pin's polling was enabled, the driver always started with
card unplugged state. It caused double initialization.
- SD response timeout was too tight. Sometimes the FIM didn't have enough
time to answer, if there was an SD communication error.
Signed-off-by: Robert Hodaszi <robert.hodaszi@digi.com>
|
|
Add a configuration option to force all resets to be cold resets.
This is a workaround for v40123.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
Conflicts:
drivers/watchdog/mxc_wdt.c
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
When the DMA descriptor ring was full, the socket buffer was not
freed, which could lead to a memory leak as the rx irq routine
keeps allocating socket buffers.
This happened under the following scenario: netpipe test when target
is forced to advertise 10baseT modes only.
This fixes vantive #40455.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit 64269671ee2aa18c4490fe319004a8aeae13698c)
|
|
cause invalid driver state
Firmware restart function was not worked before -> fixed.
If the user removed the SD-card under communication or without unmount, the
driver got into an invalid state. It was because e.g. in case of a transmit
error, the FIM firmware interrupted the communication after the problematic
block, but the driver sent more blocks in a DMA buffer, so the FIM interpreted
the remaining data as new commands.
It is fixed in the FIM firmware also, see in the FIM repository:
bf7188b061315ce9c8cd56c8cc37414851fe3556
In the fim_sd_send_command() function, the command response validater routine
moved before the block writing. It causes a shorter command timeout time.
Signed-off-by: Robert Hodaszi <robert.hodaszi@digi.com>
|
|
For some reason, after resuming with PHY_INT_EN by connecting
the cable the interrupt line is hold low and the interrupt status
is active. Writing a 1 to INT_STS doesn't seem to clear the interrupt
either, preventing the driver from doing future wakeups.
This commit resets PHY_INT_EN and restores its previous status.
This clears the interrupt.
Fixes vantive 39967.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit 23762275c384e14940f11e8f3979f588e746195f)
|
