Using CONFIGURE_MINIMUM_TASK_STACK_SIZE increases also the interrupt
stack size. This is an issue on some BSPs. Use
CONFIGURE_INIT_TASK_TABLE_SIZE instead.
Update #3433.
Introduce a new internal define _CONFIGURE_MAXIMUM_PROCESSORS and ensure
that it is _CONFIGURE_MAXIMUM_PROCESSORS > 1 only in SMP configurations.
This avoids to allocate data structures for non-existing additional
processors in uniprocessor configuration.
Update #3459.
The previous version worked only on a patched Qemu. Writes to mip are
illegal according to the The RISC-V Instruction Set Manual, Volume II:
Privileged Architecture, Privileged Architecture Version 1.10.
Update #3433.
Remove _CPU_ISR_install_raw_handler() and _CPU_ISR_install_vector()
functions. Applications can install an exception handler via the fatal
error handler to handle synchronous exceptions.
Handle interrupt exceptions via _RISCV_Interrupt_dispatch() which must
be provided by the BSP.
Update #3433.
Optimize _SMP_Inter_processor_interrupt_handler() for the common case in
which the inter-processor interrupt is only used to trigger a thread
dispatch.
On some architectures/simulators it is difficult to provoke an
exception with misaligned or illegal data loads. Use an illegal
instruction instead.
Update #3433.
The context validation support functions _CPU_Context_validate() and
_CPU_Context_volatile_clobber() are used only by one test program
(spcontext01). Move the function declarations to the CPU port
implementation header file.
Add internal fdt_cells() to avoid copy and paste. Test error cases and
default values. Fix typo in fdt_size_cells() documentation comment.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
At present fdt_create() will succeed if there is exactly enough space to
put in the fdt header. However, it sets the off_mem_rsvmap field, a few
bytes past that in order to align the memory reservation block.
Having block pointers pointing past the end of the fdt is pretty ugly, even
if it is just a transient state. Worse, if fdt_resize() is called at
exactly the wrong time, it can end up accessing data past the blob's
allocated space because of this.
So, correct fdt_create() to ensure that there is sufficient space for the
alignment padding as well as the plain header. For paranoia, also add a
check in fdt_resize() to make sure we don't copy data from outside the
blob's bounds.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
At present this function appears to copy only the data before the struct
region and the data in the string region. It does not seem to copy the
struct region itself.
From the arguments of this function it seems that it should support fdt
and buf being different. This patch attempts to fix this problem.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This adds some helpers to load (32 or 64 bit) words from an fdt blob, even
if they're unaligned and we're on a platform that doesn't like plain
unaligned loads and stores. We then use the helpers in a number of places.
There are two purposes for this:
1) This makes libfdt more robust against a blob loaded at an unaligned
address. It's usually good practice to load a blob at a 64-bit
alignment, but it's nice to work even then.
2) Users can use these helpers to load integer values from within property
values. These can often be unaligned, even if the blob as a whole is
aligned, since some property encodings have integers and strings mixed
together without any alignment gaps.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
