The NMI can asynchronously interrupt the load/store exception handler. This
does occur frequently as the NMI handler code does invoke load/store
exceptions, and the load/store exception handler is heavly used. This was
corrupting the load/store exception handler saved state and thus randomly
corrupting registers a0 to a6 of the interruptee.
This patch fixes the problem, while receiving R1b type of response from the card.
In case of R1b response the continuous stream that the card sends must be checked whether the signal indicates busy status and wait until a non zero response.
The issue was found by using f_mkdir("directory_name").
The function always returned FR_DISK_ERR due to an unhadled case in sdio_write_sectors().
According to this documentation: http://www.convict.lu/pdf/ProdManualSDCardv1.9.pdf#page=92 stop transmission command should be sent after writing multiple blocks.
This patch fixes this bug also adds a test case for it.
If `sdk_ets_timer_arm_ms_us` is called right after `vPortExitCritical` inside `sdk_ets_timer_disarm` on line 247 (by an ISR for example, which was my case), execution will loop forever on line 196, since `timer->next!=ETS_TIMER_NOT_ARMED`, which was supposed to be changed in line 248.
By delaying the IRS or context switch until `sdk_ets_timer_disarm` the problem is fixed.
* Introduce sbBYTES_TO_STORE_MESSAGE_LENGTH
* Fix bug in ucStreamBufferGetStreamBufferType() - which is only used by the
Percepio trace tool.
* Update the line within vTaskStartScheduler() that was setting xTickCount to 0
to instead set it to configINITIAL_TICK_COUNT.
* Introduce xMessageBufferNextLengthBytes() and tests for the same.
* Add call to traceTASK_SWITCHED_IN() in vTaskStartScheduler() so trace tools
can see the first task to run.
* Correct definition of StaticTask_t in the case that portUSE_MPU_WRAPPERS is
set to 1.
* prvTaskCheckFreeStackSpace() now returns configSTACK_DEPTH_TYPE to allow
return values greater than max uint16_t value if required.
* xStreamBufferSend() and xStreamBufferReceive() no longer clear task
notification bits - clearing was unnecessary as only the task notification
state is used.
* Correct out of date comment in tasks.c.
* Fix typo in comment in queue.h.
A number of the ROM functions were pointing to wrappers that saved and
restored the $a0 register, but the functions they called either did
not use $a0 or saved and restored $a0 as needed anyway.
This change means that there is only one set of global stdin, stdout, and stderr
FILE streams shared by all the threads. This reduces memory usage and avoids
having to close these streams before threads exit. These streams still have a
lock to synchronise access.
Look-up tables were used for determining the delay loop counts before.
Based on these hand-tuned values, the loop overhead was estimated for
each option -- 80 and 160 MHz, fast and slow GPIO access. Instead of the
great number of tunable parameters one now only has to tune the overhead
values if the code is changed.
Functions were added to the API which allow setting an arbitrary
frequency. API backward compatibility is retained.
i2c: fix potential overflow situation
