| Commit message (Collapse) | Author | Files | Lines |
|---|
|
A speciality of the RTEMS build system was the make preinstall step. It
copied header files from arbitrary locations into the build tree. The
header files were included via the -Bsome/build/tree/path GCC command
line option.
This has at least seven problems:
* The make preinstall step itself needs time and disk space.
* Errors in header files show up in the build tree copy. This makes it
hard for editors to open the right file to fix the error.
* There is no clear relationship between source and build tree header
files. This makes an audit of the build process difficult.
* The visibility of all header files in the build tree makes it
difficult to enforce API barriers. For example it is discouraged to
use BSP-specifics in the cpukit.
* An introduction of a new build system is difficult.
* Include paths specified by the -B option are system headers. This
may suppress warnings.
* The parallel build had sporadic failures on some hosts.
This patch removes the make preinstall step. All installed header
files are moved to dedicated include directories in the source tree.
Let @RTEMS_CPU@ be the target architecture, e.g. arm, powerpc, sparc,
etc. Let @RTEMS_BSP_FAMILIY@ be a BSP family base directory, e.g.
erc32, imx, qoriq, etc.
The new cpukit include directories are:
* cpukit/include
* cpukit/score/cpu/@RTEMS_CPU@/include
* cpukit/libnetworking
The new BSP include directories are:
* bsps/include
* bsps/@RTEMS_CPU@/include
* bsps/@RTEMS_CPU@/@RTEMS_BSP_FAMILIY@/include
There are build tree include directories for generated files.
The include directory order favours the most general header file, e.g.
it is not possible to override general header files via the include path
order.
The "bootstrap -p" option was removed. The new "bootstrap -H" option
should be used to regenerate the "headers.am" files.
Update #3254.
|
|
Rename PER_CPU_WATCHDOG_RELATIVE in PER_CPU_WATCHDOG_MONOTONIC to
highlight the corresponding POSIX CLOCK_MONOTONIC.
Rename PER_CPU_WATCHDOG_ABSOLUTE in PER_CPU_WATCHDOG_REALTIME to
highlight the corresponding POSIX CLOCK_REALTIME.
Update #3117.
Update #3182.
|
|
Rename struct Scheduler_Control to _Scheduler_Control to allow its use
in standard header files, e.g. <pthread.h>.
Update #3112.
|
|
This fixes the CPU ports with relaxed alignment restrictions, e.g. type
alignment is less than the type size.
Close #2822.
Close #2823.
|
|
Close #2820.
|
|
Update #2809.
|
|
Update #2751.
|
|
Update #2797.
|
|
This makes it possible to adjust the scheduler of a processor at
run-time.
Update #2797.
|
|
|
|
|
|
Update #2556.
|
|
Only use CPU_Per_CPU_control if it contains at least one filed. In GNU
C empty structures have a size of zero. In C++ structures have a
non-zero size. In case CPU_PER_CPU_CONTROL_SIZE is defined to zero,
then this structure is not used anymore.
|
|
According to the C11 standard only atomic read-modify-write operations
guarantee that the last value written in modification order is read, see
"7.17.3 Order and consistency". Thus we must use a read-modify-write in
_SMP_Inter_processor_interrupt_handler() to make sure we read an
up-to-date message.
|
|
Update #2555.
|
|
Rename _ISR_Disable() into _ISR_Local_disable(). Rename _ISR_Enable()
into _ISR_Local_enable(). Remove _Debug_Is_owner_of_giant().
This is a preparation to remove the Giant lock.
Update #2555.
|
|
Rename _ISR_Disable_without_giant() into _ISR_Local_disable(). Rename
_ISR_Enable_without_giant() into _ISR_Local_enable().
This is a preparation to remove the Giant lock.
Update #2555.
|
|
|
|
The rtems_extension_create() no longer uses the Giant lock. Ensure that
we call _User_extensions_Add_set() only in the right context.
Update #2555.
|
|
The CPU time used of a thread was previously maintained per-processor
mostly during _Thread_Dispatch(). However, on SMP configurations the
actual processor of a thread is difficult to figure out since thread
dispatching is a highly asynchronous process (e.g. via inter-processor
interrupts). Only the intended processor of a thread is known to the
scheduler easily. Do the CPU usage accounting during thread heir
updates in the context of the scheduler operations. Provide the
function _Thread_Get_CPU_time_used() to get the CPU usage of a thread
using proper locks to get a consistent value.
Close #2627.
|
|
Use a red-black tree instead of delta chains.
Close #2344.
Update #2554.
Update #2555.
Close #2606.
|
|
The use case for this is the Cortex-A9 MPCore which has per-processor
registers (only accessible by a particular processor) for the global
timer used by the clock driver. This might be useful for other drivers
as well.
Update #2554.
|
|
|
|
Rename Per_CPU_Control::started into Per_CPU_Control::online to match
standard nomenclature.
|
|
Update #2408.
|
|
According to the C11 and C++11 memory models only a read-modify-write
operation guarantees that we read the last value written in modification
order. Avoid the sequential consistent thread fence and instead use the
inter-processor interrupt to set the thread dispatch necessary
indicator.
|
|
Add a leading underscore to the structure name to allow forward
declarations in standard header files provided by Newlib and GCC.
|
|
Add a SMP lock statistics pointer to SMP_lock_Stats_context and drop the
SMP lock statistics parameter from _SMP_ticket_lock_Release().
|
|
The problem is that empty structures have a different size in C and C++.
|
|
Avoid Thread_Control typedef in <rtems/score/percpu.h>. This helps to
get rid of the <rtems/score/percpu.h> include in <rtems/score/thread.h>
which exposes a lot of implementation details.
|
|
|
|
Add PER_CPU_OFFSET_HEIR. Move Per_CPU_Control::executing and
Per_CPU_Control::heir for easy offset calculation.
|
|
Replace _Per_CPU_State_wait_for_ready_to_start_multitasking() with
_Per_CPU_State_wait_for_non_initial_state(). Implement this function.
|
|
The current implementation of task migration in RTEMS has some
implications with respect to the interrupt latency. It is crucial to
preserve the system invariant that a task can execute on at most one
processor in the system at a time. This is accomplished with a boolean
indicator in the task context. The processor architecture specific
low-level task context switch code will mark that a task context is no
longer executing and waits that the heir context stopped execution
before it restores the heir context and resumes execution of the heir
task. So there is one point in time in which a processor is without a
task. This is essential to avoid cyclic dependencies in case multiple
tasks migrate at once. Otherwise some supervising entity is necessary to
prevent life-locks. Such a global supervisor would lead to scalability
problems so this approach is not used. Currently the thread dispatch is
performed with interrupts disabled. So in case the heir task is
currently executing on another processor then this prolongs the time of
disabled interrupts since one processor has to wait for another
processor to make progress.
It is difficult to avoid this issue with the interrupt latency since
interrupts normally store the context of the interrupted task on its
stack. In case a task is marked as not executing we must not use its
task stack to store such an interrupt context. We cannot use the heir
stack before it stopped execution on another processor. So if we enable
interrupts during this transition we have to provide an alternative task
independent stack for this time frame. This issue needs further
investigation.
|
|
|
|
|
|
Add optional method _CPU_Get_current_per_CPU_control() to obtain the
per-CPU control of the current processor.
This is optional. Not every CPU port needs this. It is only an
optional optimization variant. In case this macro is undefined, the
default implementation using the current processor index will be used.
|
|
Use "cpu" for an arbitrary Per_CPU_Control variable.
Use "cpu_self" for the Per_CPU_Control of the current processor.
Use "cpu_index" for an arbitrary processor index.
Use "cpu_index_self" for the processor index of the current processor.
Use "cpu_count" for the processor count obtained via
_SMP_Get_processor_count().
Use "cpu_max" for the processor maximum obtained by
rtems_configuration_get_maximum_processors().
|
|
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.
|
|
|
|
Add _Per_CPU_State_wait_for_ready_to_start_multitasking(). Add new
fatal SMP error SMP_FATAL_SHUTDOWN_EARLY.
|
|
|
|
|
|
Add per-CPU profiling stats API. Implement the thread dispatch disable
level profiling. The interrupt profiling must be implemented in CPU
port specific parts (mostly assembler code). Add a support function
_Profiling_Outer_most_interrupt_entry_and_exit() for this purpose.
|
|
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.
|
|
|
|
Rename _SMP_Request_other_cores_to_perform_first_context_switch() into
_SMP_Request_start_multitasking() since this requests now a multitasking
start on all configured and available processors. The name corresponds
_Thread_Start_multitasking() and
_SMP_Start_multitasking_on_secondary_processor() actions issued in
response to this request. Move in source file to right place.
Rename PER_CPU_STATE_READY_TO_BEGIN_MULTITASKING into
PER_CPU_STATE_READY_TO_START_MULTITASKING.
Rename PER_CPU_STATE_BEGIN_MULTITASKING into
PER_CPU_STATE_REQUEST_START_MULTITASKING.
Rename _SMP_Request_other_cores_to_shutdown() into
_SMP_Request_shutdown().
Add a per-CPU state lock to protect all changes. This was necessary to
offer a controlled shutdown of the system (atomic read/writes alone are
not sufficient for this kind of synchronization).
Add documentation for Per_CPU_State.
Delete debug output.
New tests smptests/smpfatal01 and smptests/smpfatal02.
|
|
Collect SMP implementation specific parts in the
<rtems/score/smpimpl.h> header file.
|
|
Use a per-CPU thread dispatch disable level. So instead of one global
thread dispatch disable level we have now one instance per processor.
This is a major performance improvement for SMP. On non-SMP
configurations this may simplifiy the interrupt entry/exit code.
The giant lock is still present, but it is now decoupled from the thread
dispatching in _Thread_Dispatch(), _Thread_Handler(),
_Thread_Restart_self() and the interrupt entry/exit. Access to the
giant lock is now available via _Giant_Acquire() and _Giant_Release().
The giant lock is still implicitly acquired via
_Thread_Dispatch_decrement_disable_level().
The giant lock is only acquired for high-level operations in interrupt
handlers (e.g. release of a semaphore, sending of an event).
As a side-effect this change fixes the lost thread dispatch necessary
indication bug in _Thread_Dispatch().
A per-CPU thread dispatch disable level greatly simplifies the SMP
support for the interrupt entry/exit code since no spin locks have to be
acquired in this area. It is only necessary to get the current
processor index and use this to calculate the address of the own per-CPU
control. This reduces the interrupt latency considerably.
All elements for the interrupt entry/exit code are now part of the
Per_CPU_Control structure: thread dispatch disable level, ISR nest level
and thread dispatch necessary. Nothing else is required (except CPU
port specific stuff like on SPARC).
|
|
Add and use _Per_CPU_Release_all().
The context switch user extensions are invoked in _Thread_Dispatch().
This change is necessary to avoid the giant lock in _Thread_Dispatch().
|
