Age | Commit message (Collapse) | Author |
|
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Otherwise, even when CONFIG_USB is disabled, the wake up devices are
registered and this could cause problems.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This commit will configure the USB reset line and perform a USB host1
reset on host1 initialization.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This commit configures the SD pins connected to the wireless
module as gpios inputs when the driver is removed. This is
needed to avoid some kind of current feedback that happens
through the SD pins that causes that the reset level of the
redpine module doesnt go low enough causing module reload
problems (at least in WinCE).
Configuring the pins as inputs seems to solve the problem.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This commit fixes some initialization/configuration problems
that were happening with some screens due to the generation
of multiple hot plug events. This commit disables the interrupts
during a configurable period to avoid this reconfigurations that
some screens doesnt tolerate well.
Fixes #42030 for most screens. The NANNSPREE still doesnt work
very well but that monitor presents a strange behavior (also in
WinCE).
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
When reading a 10bits number, the 2s complement MSB is the sign. This bit
cannot be just shifted to bit 16, the number has to be sign extended.
0abcdefghij => 00000abcdefghij
1abcdefghij => 11111abcdefghij
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
When reading 10bit registers, the value of the MSB is latched after the
LSB is read. The MSB read is expected right after the LSB access otherwise
read errors occur.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The masking/unmasking of events in suspend/resume is not specific to
any platform.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The ADC suspend happens early in the suspend process and it is the best
place to suspend not wanted events that may abort the suspend.
This fixes Vantive 42007.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Add functions to get the currently enabled events and to ascertain whether
certain event is masked on a provided event mask.
These functions will be used to mask/unmask events in suspend/resume.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Otherwise some symbols may clash in other MX platforms builds.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Some PHYs do not reset properly without it. We have seen problems on
CCXMX53 and CC9M2443.
Fixes Vantive #41269.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This commit fixes #41819. The PAD of the hw flow control
lines was incorrectly set. Note that its necessary to
change 2 resistors in the base board to test the hw flow
control on the UART1.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This code was not executed for mx51 but it was executed on mx53,
preventing this platform from being discovered by ADDP.
This fixes vantive 42101.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
When using our GPIO module to write to a gpio dev node (even
if it is a single node) if we use a value bigger than 1
(e.g. write(fd, 0xff)) the value is propagated to the driver's
gpio_set function, which eventually sets more bits
than the GPIO it was supposed to write.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
(cherry picked from commit c62fe2a5a9e249466af14e13b498aff07ec58f7f)
|
|
Test of idigibuttonled on ccwmx53js revealed that the status
of the user LEDs (gpios as outputs) always read as 0,
independently of their real status.
Investigation led to the SION (Software Input On Field) bit on
the MUX_CTL_PAD register. From the manual:
Software Input On Field. Force the selected mux mode Input path
no matter of MUX_MODE functionality.
1 Force input path
0 Input Path is determined by functionality of the selected
mux mode (regular).
Default value of SION is 0, which makes the Linux driver always
reads 0 on the PSR (pad status register) of the gpio pins configured
as outputs, no matter the real value of the pin.
For the UserLEDs we want SION to be set, so that we can read the
real value of the pins.
This fixes vantive 42103.
Signed-off-by: Hector Palacios <hector.palacios@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Javier Viguera <javier.viguera@digi.com>
|
|
Even though it should not be needed as the write_many on the same function
also "parks" the PMIC, meaning it does a last access to a safe register
to avoid a PMIC I2C bug, some modules need this last access to work.
Also, change the parking access to the STATUSA safe register, as the one
we had does not seem to be safe in all cases. That is the same register
used in WCE.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This reverts commit 9521e4b256644b5bf430ea9c1df3f1dd71618685.
Which was commited again by mistake
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This commit adds the platform functions called before
suspending to ram and after resuming from suspend.
In this functions, we mask and unmask respectively
irqs to avoid being wakeup by undesired events, like
the touch interface.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This commit increases the DMA_ZONE_SIZE from 96 to 184 MB
to allow playing fullhd videos in a fullhd screen.
With some videos, we where getting dma allocation failures.
See 41803 for additional details.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This commit adds the code to:
-Select a custom logo that can be used as splash screen.
-Center that logo in the screen.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This commit sets again the singal strenght of the video lines
to high, when using high res. Without this, there were some
wrong pixels in fullhd displays. This was different than the
results that have been observed in linux.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
(cherry picked from commit f1663e328951db554e709256d611e52d9795e0d7)
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
This fixes issues with the first buffer captured by the sensor when
connected to the MX5X CSI.
Fixes Vantive 41833.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This could happen for something like a signal, not necessarily an error.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The MMA7455 seems to have a hardware bug that makes it keep the data line
low if an I2C read transfer is interrupted between writing the register
address and reading the data.
For the CCWMX53 this means leaving the bus where the PMIC is inaccessible,
making it impossible to suspend or reboot, and needing a cold boot with
the coin cell disabled to recover.
Without a definitive fix for the hardware problem I have implemented a
patch to make the situation highly unlikely.
1) The register read is changed from two i2c transfers (one for writing
the register address and one for reading) to a single I2C transfer.
2) I have also added a 10ms delay between sysfs reads which helps
minimize the window of opportunity for the signal to interrupt the
I2C transfer.
These makes the problem very rare both for the MX51 and MX53.
I have not been able to repro it manually after the changes, but probably
an automated test could.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This partially reverts commit '314c1b4' because after commit '5b13ce7'
only one video interface is registered.
Vantive: #41837
Signed-off-by: Javier Viguera <javier.viguera@digi.com>
|
|
The mc13892 driver from the BSP assumed that the battery was always
connected. Otherwise the battery driver would not be loaded.
This fix allows for the driver to work even when booting without a
battery. It modifies the battery status function to use the charger
current measured via the ADC to detect the presence of a battery and
its charging status instead of the sense bit. Also, because now the
driver can tell whether a battery is present or not, it fixes some of
the places where the driver assumed the battery would always be connected.
This fixes 42033.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
(cherry picked from commit 093ee454469036f25f104c2d544ba144d9a1ed37)
|
|
This commit enables the support for the gpios of the da9053.
This allows to manupulate that gpios (for instance in the
wireless drivers, where the wifi power enable comes from a
pmic gpio).
Not doing a power cycling on the redpine module was causing
some detection issues on some ccwmx53 modules.
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
|
|
Although DEL is working fine with the previous I2C park fixes, we have
seen spurious events coming from the PMIC running Android.
This was traced down to using the read_many function. If the events
registers were read using four different reads the contents of the
registers are read OK. But if we use the read_many, some times the
register contents are read incorrectly.
Using the read_many function from Dialog's latest patch seems to solve
this, although both implementations should work OK.
Given that Dialog's PMIC has not well documented I2C access issues,
we may not get a complete understanding on the problem.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This commit limits to 1 the number of framebuffer devices that
can be registered (+ the overlay). This avoids problems where
2 big displays where connected (fullhd + vga@1280x1024, for
instance).
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
(cherry picked from commit a697c49edfd6ad9e629834bbc87cfae3a9c9781f)
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
(cherry picked from commit aa2de2fc2a5af3e8382340df89db69d70d77483e)
|
|
Signed-off-by: Pedro Perez de Heredia <pedro.perez@digi.com>
(cherry picked from commit f843b14a08a5f26ecd47ebcfde9110db1769723a)
|
|
The kernel API can update the alarm time while the alarm is disabled. Do
not set the alarm when updating the time unless explicitely told so.
This helps Vantive 41951.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Add a basic voltage based capacity calculation to the driver. It uses
static charge and discharge lookup tables for 25 celsius, characterized
for a hopefully generic Li-Ion battery.
This simple implementation disregards temperature and age, and it also
does not include hysteresis algorithms so the capacity may vary upon
charger connection and disconnection.
User space can still implement better algorithms using the information
provided in the sysfs.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
(cherry picked from commit 35fea3d126c91a4f25d9269ad58dc8a1c52fcd1f)
|
|
Dialog is about to release samples of a new PMIC version which contains a
hardware fix for this issue.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
The patch to the PMIC I2C accesses to make them safe does not completely
guarantee that a reboot can not happen and that interrupts cannot make
unsafe accesses to the PMIC.
This commit assures a last safe PMIC access on reboot.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
There is a Dialog bug on the PMIC which can leave the I2C bus inaccessible
leaving an unbootable system.
The workaround suggested by Dialog is to flush all accesses with a last
write to a safe register.
They provided a patch which when applied does not work well, so this is a
rework simplified version of that patch.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
It's safer if the touch thread stops accessing the PMIC early on reboot.
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
Signed-off-by: Alex Gonzalez <alex.gonzalez@digi.com>
|
|
This reverts commit 87ac4650973e1e2d9ff372ff41acf59c289b60a2.
|