CodeSonar flagged this as an empty if body. Upon analysis, it turned
out to be an error that we think should never occur but if it did,
there is nothing we could do about it. It would likely just indicate
the thread was deleted before we got here. Adding the _Assert() at least
will flag this if it ever occurs during a debug build and we can discuss
what happened.
Ensure that the global construction is performed in the context of the
first initialization thread. On SMP this was not guaranteed in the
previous implementation.
This lays the proper structure for doing future work on
time adjustment algorithms. Any TOD adjustments should be
requested at the API level and performed at the SCORE level.
Additionally updated a test.
Remove compare function and is unique indicator from the control
structure. Rename RBTree_Compare_function to RBTree_Compare. Rename
rtems_rbtree_compare_function to rtems_rbtree_compare. Provide C++
compatible initializers. Add compare function and is unique indicator
to _RBTree_Find(), _RBTree_Insert(), rtems_rbtree_find() and
rtems_rbtree_insert(). Remove _RBTree_Is_unique() and
rtems_rbtree_is_unique(). Remove compare function and is unique
indicator from _RBTree_Initialize_empty() and
rtems_rbtree_initialize_empty().
Suppose we have two tasks A and B and two processors. Task A is about
to delete task B. Now task B calls rtems_task_wake_after(1) on the
other processor. Task B will block on the Giant lock. Task A
progresses with the task B deletion until it has to wait for
termination. Now task B obtains the Giant lock, sets its state to
STATES_DELAYING, initializes its watchdog timer and waits. Eventually
_Thread_Delay_ended() is called, but now _Thread_Get() returned NULL
since the thread is already marked as deleted. Thus task B remained
forever in the STATES_DELAYING state.
Instead of passing the thread identifier use the thread control block
directly via the watchdog user argument. This makes
_Thread_Delay_ended() also a bit more efficient.
The _Scheduler_Yield() was called by the executing thread with thread
dispatching disabled and interrupts enabled. The rtems_task_suspend()
is explicitly allowed in ISRs:
http://rtems.org/onlinedocs/doc-current/share/rtems/html/c_user/Interrupt-Manager-Directives-Allowed-from-an-ISR.html#Interrupt-Manager-Directives-Allowed-from-an-ISR
Unlike the other scheduler operations the locking was performed inside
the operation. This lead to the following race condition. Suppose a
ISR suspends the executing thread right before the yield scheduler
operation. Now the executing thread is not longer in the set of ready
threads. The typical scheduler operations did not check the thread
state and will now extract the thread again and enqueue it. This
corrupted data structures.
Add _Thread_Yield() and do the scheduler yield operation with interrupts
disabled. This has a negligible effect on the interrupt latency.
The function to change a thread priority was too complex. Simplify it
with a new scheduler operation. This increases the average case
performance due to the simplified logic. The interrupt disabled
critical section is a bit prolonged since now the extract, update and
enqueue steps are executed atomically. This should however not impact
the worst-case interrupt latency since at least for the Deterministic
Priority Scheduler this sequence can be carried out with a wee bit of
instructions and no loops.
Add _Scheduler_Change_priority() to replace the sequence of
- _Thread_Set_transient(),
- _Scheduler_Extract(),
- _Scheduler_Enqueue(), and
- _Scheduler_Enqueue_first().
Delete STATES_TRANSIENT, _States_Is_transient() and
_Thread_Set_transient() since this state is now superfluous.
With this change it is possible to get rid of the
SCHEDULER_SMP_NODE_IN_THE_AIR state. This considerably simplifies the
implementation of the new SMP locking protocols.
Clustered/partitioned scheduling helps to control the worst-case
latencies in the system. The goal is to reduce the amount of shared
state in the system and thus prevention of lock contention. Modern
multi-processor systems tend to have several layers of data and
instruction caches. With clustered/partitioned scheduling it is
possible to honour the cache topology of a system and thus avoid
expensive cache synchronization traffic.
We have clustered scheduling in case the set of processors of a system
is partitioned into non-empty pairwise-disjoint subsets. These subsets
are called clusters. Clusters with a cardinality of one are partitions.
Each cluster is owned by exactly one scheduler instance.
The thread control block contains fields that point to application
configuration dependent memory areas, like the scheduler information,
the API control blocks, the user extension context table, the RTEMS
notepads and the Newlib re-entrancy support. Account for these areas in
the configuration and avoid extra workspace allocations for these areas.
This helps also to avoid heap fragementation and reduces the per thread
memory due to a reduced heap allocation overhead.
The holder field is enough to determine if a mutex is locked or not.
This leads also to better error status codes in case a
rtems_semaphore_release() is done for a mutex without having the
ownership.
The thread deletion is now supported on SMP.
This change fixes the following PRs:
PR1814: SMP race condition between stack free and dispatch
PR2035: psxcancel reveals NULL pointer access in _Thread_queue_Extract()
The POSIX cleanup handler are now called in the right context (should be
called in the context of the terminating thread).
http://pubs.opengroup.org/onlinepubs/009695399/functions/xsh_chap02_09.html
Add a user extension the reflects a thread termination event. This is
used to reclaim the Newlib reentrancy structure (may use file
operations), the POSIX cleanup handlers and the POSIX key destructors.
The man-page for pthread_setspecific does not define the EAGAIN return value.
Further without this patch it was not possible to set keys that have been
already set a new value.
Add test for setting a new value to a already set key.
Add a local context structure to the SMP lock API for acquire and
release pairs. This context can be used to store the ISR level and
profiling information. It may be later used to enable more
sophisticated lock algorithms, e.g. MCS locks.
There is only one lock that cannot be used with a local context. This
is the per-CPU lock since here we would have to transfer the local
context through a context switch which is very complicated.
