System events are similar to normal events. They offer a second set of
events. These events are intended for internal RTEMS use and should not
be used by applications (with the exception of the transient system
event).
The work areas (RTEMS work space and C program heap) will be initialized
now in a separate step and are no longer part of
rtems_initialize_data_structures(). Initialization is performed with
tables of Heap_Area entries. This allows usage of scattered memory
areas present on various small scale micro-controllers.
The sbrk() support API changes also. The bsp_sbrk_init() must now deal
with a minimum size for the first memory chunk to take the configured
work space size into account.
The _Heap_Extend() has now the same signature as _Heap_Initialize().
The 4th parameter is ignored (page size in _Heap_Initialize()).
Add Heap_Area and Heap_Initialization_or_extend_handler.
Add and test _Heap_No_extend().
This helps to do a table based heap initialization and extension.
Create a table of Heap_Area elements and iterate through it. Set the
handler to _Heap_Initialize() in the first iteration and then to
_Heap_Extend().
Use _Workspace_Allocate_or_fatal_error() consistently in case auto
extend is turned off. This helps to avoid undefined behaviour in
_API_Mutex_Allocate() in case _API_Mutex_Information() fails.
The changes in _Thread_Dispatch() of commits
dad36c52b8 and
d4dc7c8196 introduced a severe bug which
destroys the real-time properties of RTEMS completely.
Consider the following scenario. We have three tasks L (lowest
priority), M (middle priority), and H (highest priority). Now let a
thread dispatch from M to L happen. An interrupt occurs in
_Thread_Dispatch() here:
void _Thread_Dispatch( void )
{
[...]
post_switch:
_ISR_Enable( level );
<-- INTERRUPT
<-- AFTER INTERRUPT
_Thread_Unnest_dispatch();
_API_extensions_Run_postswitch();
}
The interrupt event makes task H ready. The interrupt code will see
_Thread_Dispatch_disable_level > 0 and thus doesn't perform a
_Thread_Dispatch(). Now we return to position "AFTER INTERRUPT". This
means task L executes now although task H is ready! Task H will execute
once someone calls _Thread_Dispatch().
This code was reviewed as part of coverage analysis improvements. The
uncovered range had unclear documentation and the code itself was also
cleaned up to be easier to understand.
Author: Krzysztof Mięsowicz <krzysztof.miesowicz@gmail.com>
Due to the fact that CBS extends the EDF scheduler, most parts of EDF are
reused. One part is the allocation of the per-thread data. However, the CBS
per-thread structure defines a further element (while including the rest of
Scheduler_EDF_Per_thread)
The 64-bit timestamps were introduced to simplify the timestamp
calculations. This works well since nearly all operations are
additions. The previous _TOD_Tickle_ticks() implementation had a
serious performance regression in case of 64-bit timestamps due to the
usage of two 64-bit divisions which are quite expensive on some
architectures.
A new field seconds_trigger in TOD_Control is introduced to trigger the
_Watchdog_Tickle_seconds() in _TOD_Tickle_ticks(). This avoids the
64-bit divisions completely and only 32-bit additions are used.
Port
+ v850 does not have appear to have any optimized bit scan instructions
+ v850 does have single instructions for wap u16 and u32
+ Code path optimization preferences set
+ Add BSP variants for each GCC CPU model flag and a README
- v850e1 variant does not work (fails during BSP initialization)
BSP for GDB v850 Simulator
+ linkcmds matches defaults in GDB simulator with RTEMS mods
+ crt1.c added from v850 newlib port for __main()
+ BSP exits cleanly
+ printk and console I/O work
+ uses clock tick from IDLE task
+ Tests not requiring real clock ISR work
Documentation
+ CPU Supplment chapter for v850 added
PowerPC cores with the SPE (Signal Processing Extension) have 64-bit
general-purpose registers. The SPE context switch code has been merged
with the standard context switch code. The context switch may use cache
operations to increase the performance. It will be ensured that the
context is 32-byte aligned (PPC_DEFAULT_CACHE_LINE_SIZE). This
increases the overall memory size of the context area in the thread
control block slightly. The general-purpose registers GPR2 and GPR13
are no longer part of the context. The BSP must initialize these
registers during startup (usually initialized by the __eabi() function).
The new BSP option BSP_USE_DATA_CACHE_BLOCK_TOUCH can be used to enable
the dcbt instruction in the context switch.
The new BSP option BSP_USE_SYNC_IN_CONTEXT_SWITCH can be used to enable
sync and isync instructions in the context switch. This should be not
necessary in most cases.
These modifications were required by hand after running the script.
In some cases, the file names did not match patterns. In others,
the format of the file did not match any common patterns.
Script does what is expected and tries to do it as
smartly as possible.
+ remove occurrences of two blank comment lines
next to each other after Id string line removed.
+ remove entire comment blocks which only exited to
contain CVS Ids
+ If the processing left a blank line at the top of
a file, it was removed.
In particular CPU_INTERRUPT_NUMBER_OF_VECTORS and
CPU_INTERRUPT_MAXIMUM_VECTOR_NUMBER are only used on
Simple Vectored Architectures, so do not depend on
them being defined. This disables as much as possible
that is specific to the Simple Vectored Model and
not expected to be used on architectures which use
the Programmable Interrupt Controller model for
interrupt handler vectoring.
The implementation of RBTree_Next was using an awkward construction to detect
and avoid accessing the false root of the red-black tree. To deal with the
false root, RBTree_Next was comparing node parents with the control node.
Instead the false root can be detected by checking if the grandparent of a
node exists; the grandparent of the tree's true root is NULL by definition
so the root of the tree is found while walking up the tree by checking for
the non-existence of a grandparent.
This change propagates into the predecessor/successor and iterate functions.
When inserting to a red-black tree with duplicates the min and max pointers are
not updated properly. We need to check the key of the min/max node against the
insert node since the insert point could be the child of a node with an
identical key to the min/max node.
During node extraction from a red-black tree the min and max values are updated
incorrectly. We need to use the successor/predecessor functions to find the
next/previous node when we remove the min/max from the tree.
Make default for rtems_rbtree functions be unprotected (preemption enabled)
unless an unprotected variant e.g. rtems_rbtree_xxx_unprotected is available.
The increase of the stack size by CPU_STACK_ALIGNMENT in
_Thread_Stack_Allocate() is disadvantageous. This may lead to a huge
over allocation for specialized stack allocators. The
CPU_STACK_ALIGNMENT is at most 16 on all current RTEMS CPU ports. The
mimimum stack size ensured by _Stack_Ensure_minimum() must be
considerable larger than this value, otherwise stack overflows will
likely occur. Thus the _Stack_Adjust_size() is also superfluous.
