Refactored of I/O expander support to simplify addition of new devices.
This had added ISR support for PCA9555.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
GPIO expander ISRs were dropping interrupts as each pin had its own
thread which cleared the INT signal when reading the GPIO register.
Thus pending interrupts for GPIO on threads waiting to be scheduled
were dropped. Now a single thread polls the GPIO register when INT goes
low and sets interrupt flags for each GPIO line. The threads created
by mraa_gpio_isr() now poll the appropriate interrupt flag.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
Synchronization must be done at "context" level so that i2c bus
cannot be unexpectedly changed by a different thread.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
- FT4222 GPIO ISR implementation was specific to C/C++ and ignored
language binding support. This is now fixed via adv_func updates.
- GPIO ISR code has been refactored to reduce i2c traffic when using
I/O expanders
- Added support for built-in FT4222 GPIO interrupts
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Switch to using calloc on all calls to malloc where the memory isn't
initialized. For things like the mraa_board_t, not allocating all to zero
causes issues such as with the sub_platform member, where if that's not zero
mraa_get_platform_type will try to dereference a random memory location for the
sub_platform->platform_name, which can result in segmentation faults and other
issues.
Note that in some places where immediately after the malloc call is a copy
operation, there is no need for calloc, as all the memory gets overwritten
anyways, but in cases where there may or may not be memory written to (such as
in mraa_file_contains, with reading from a file), even though in most cases the
memory is overwritten, it could be the case that the read operation does
nothing, but the memory still has non-zero values, by virtue of the fact it
wasn't overwritten.
Signed-off-by: Ian Johnson <ijohnson@wolfram.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
First four pins (0 to 3) are now built-in FTDI GPIO/I2C pins
If i2c GPIO expander is detected, next 8 pins (4 to 11) are its GPIO
If i2c switch is detected an extra 4 i2c busses are added
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
mraa_adv_func_t->gpio_init_internal_replace() now has mraa_gpio_context
parameter so that override function can access both physical and logical
pin numbers.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
mraa_i2c_init_raw now works again without platfrorm defined.
More API calls now support sub-platform
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
Now platform_type is no longer global it tracks with board instance
so that platform and sub-platform can report different types.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
Existing FTDI D2xx support is just for serial comms library
but FT4222 needs a different library for i2c operation.
Signed-off-by: Henry Bruce <henry.bruce@intel.com>
Signed-off-by: Brendan Le Foll <brendan.le.foll@intel.com>
