b78c3fcc30
Finish backing out previous change
2016-11-30 08:13:43 -06:00
e6eb2e8bfa
Back out the last change. I just noticed some complexities that need to be verified before this is released onto the world.
2016-11-30 07:41:48 -06:00
8453343426
scheduler instrumentation: Add a little more protection for the SMP case
2016-11-30 07:24:15 -06:00
9ee3f3b933
sched_note: Permit spinlock and critical section notes in in-memory buffer iff sched_not_get() interfaces is disabled.
2016-11-28 18:36:26 -06:00
00215fbc98
sched_note: Add spinlock instrumentation; In SMP configurations, select to log only notes from certain CPUs
2016-11-28 10:33:46 -06:00
d65be718c2
sched_note: Extend OS instrumentation to include some SMP events.
2016-11-27 17:14:57 -06:00
e3fe320e08
SMP: Add support for linking spinlocks into a special, non-cached memory region.
2016-11-26 08:47:03 -06:00
1d06e786e1
SMP: Clean-up and simplication of logic that I implemented late last night.
2016-11-26 07:05:27 -06:00
a0e1af2614
SMP: Fix yet another potential deadlock
2016-11-25 23:04:27 -06:00
eb9f8074c0
Update comments
2016-11-24 09:56:43 -06:00
7bec4ffeec
Update some comments
2016-11-23 17:40:01 -06:00
d4037a30aa
Update some comments
2016-11-23 13:20:18 -06:00
f90525a5d1
SMP: Update some comments; trivial improvement by inlining static function.
2016-11-22 16:48:57 -06:00
bac7153609
SMP: Add logic to avoid a deadlock condition when CPU1 is hung waiting for g_cpu_irqlock and CPU0 is waitin for g_cpu_paused
2016-11-22 11:34:16 -06:00
130bfa3f6b
Remove a assertion condition that appears to rarely cause false-alarm assertions. Teported by Petteri Aimonen
2016-11-21 14:43:56 -06:00
558784d06f
Spinlocks: Added capability to provide architecture-specific memory barriers. This was for i.MX6 but does not help with the SMP problems. It is still a good feature.
2016-11-21 11:55:59 -06:00
2d1765a014
Remove more of those annoying empty comment blocks.
2016-11-21 07:34:27 -06:00
0029c04dad
task_restart: Make sure new task starts with pre-emption disabled and not in a critical section.
2016-11-21 07:33:23 -06:00
cee8d59b58
Update TODO list
2016-11-20 12:26:08 -06:00
e24f281401
This commit adds a new internal interfaces and fixes a problem with three APIs in the SMP configuration. The new internal interface is sched_cpu_pause(tcb). This function will pause a CPU if the task associated with 'tcb' is running on that CPU. This allows a different CPU to modify that OS data stuctures associated with the CPU. When the other CPU is resumed, those modifications can safely take place.
...
The three fixes are to handle cases in the SMP configuration where one CPU does need to make modifications to TCB and data structures on a task that could be running running on another CPU. Those three cases are task_delete(), task_restart(), and execution of signal handles. In all three cases the solutions is basically the same: (1) Call sched_cpu_pause(tcb) to pause the CPU on which the task is running, (2) perform the necessary operations, then (3) call up_cpu_resume() to restart the paused CPU.
2016-11-20 07:57:18 -06:00
f40fbaa4b9
sched/task: task_restart() test not supported on SMP systems. This is not fully implemented.
2016-11-19 11:41:05 -06:00
0d8e191d69
sched/clock: Correct calculation for the case of Tickless mode with a 32-bit timer. In that case, the calculation was returning millisecond accuracy. That is not good when the timer accuracy is < 1 msec.
2016-11-19 09:55:21 -06:00
5ca9128c38
Update comments
2016-11-19 08:33:55 -06:00
ceacacbc63
vfork(): Fix a race condition in the SMP case. Existing logic depended on the fact that the child would not run until waitpid was called because the child had the same priority as the parent. BUT in the SMP case that is not true... the child may run immediately on a different CPU.
2016-11-19 07:30:01 -06:00
69e9f8638d
Most interrupt handling logic interacts with tasks via standard mechanism such as sem_post, sigqueue, mq_send, etc. This all call enter_critical_section and are assumed to be safe in the SMP case.
...
But certain logic interacts with tasks in different ways. The only one that comes to mind are wdogs. There is a tasking interface that to manipulate wdogs, and a different interface in the timer interrupt handling logic to manage wdog expirations.
In the normal case, this is fine. Since the tasking level code calls enter_critical_section, interrupts are disabled an no conflicts can occur. But that may not be the case in the SMP case. Most architectures do not permit disabling interrupts on other CPUs so enter_critical_section must work differently: Locks are required to protect code.
So this change adds locking (via enter_critical section) to wdog expiration logic for the the case if the SMP configuration.
2016-11-18 13:57:30 -06:00
78fd358023
Update some comments.
2016-11-18 08:20:52 -06:00
8602e8a8a9
SMP: irq_csection() has a bad assumption. It assumed that the state of certain variables. That was true on entry into the interrupt handler, but might change to the execution of logic within the interrupt handler.
2016-11-18 07:38:16 -06:00
8e029f019b
Update comments
2016-11-17 17:30:12 -06:00
47b52a2633
Fix an error in manually bringing in Sebastien's changes; Update some comments.
2016-11-17 16:15:06 -06:00
9acd57c819
Fix up some naming and update some comments.
2016-11-17 15:50:39 -06:00
558f4049d4
sched/irq/irq_csection: Fix a bad DEBUG assertion when SMP is enabled.
2016-11-17 15:36:27 -06:00
2c314464a8
Fix a typo in a comment
2016-11-17 08:03:45 -06:00
ccdc11d8ff
Fix typo in variable naming. g_cpu_irqset should be g_cpu_irqlock.
2016-11-17 06:37:24 -06:00
1c1564f0c1
Add logic to handled nested calls to enter_critical_section() from interrupts handlers (with SMP).
2016-11-16 22:45:08 -06:00
0c3207e86e
Fix backward logic in an assertion.
2016-11-16 21:48:27 -06:00
16f0b8fa96
Fix typos in DEBUASSERT statements
2016-11-16 19:58:51 -06:00
f1e4951a8d
Fix some comments
2016-11-16 19:11:31 -06:00
4ca653a1cc
Add some experimental changes to enter/leave_critical_section to deal with the case where interrupts are disabled only on the local CPU
2016-11-16 18:28:35 -06:00
5ce3b399d5
Correct some typos in comments
2016-11-15 16:48:40 -06:00
40d7216a22
Back out 65ab12 and parts of 21f92b
2016-11-15 15:44:43 -06:00
6683f01444
Trivial change to comment
2016-11-15 13:38:43 -06:00
21f92ba601
Review some SMP logic; update comments; refresh configuration.
2016-11-15 09:24:00 -06:00
65ab12fbb9
If SMP is enabled, if any interrupt handler calls enter_critical_section(), it should take the spinlock.
2016-11-15 08:37:58 -06:00
b53866c872
sched/irq: Change order for SMP case in enter_critical_section: Disble local interrupts BEFORE taking spinlock.
2016-11-15 07:54:18 -06:00
484a1b6104
sem_wait() and sem_trywait() no longer modify the errno value UNLESS an error occurs. This allows these functions to be used internallly without clobbering the errno value.
2016-11-09 07:01:49 -06:00
120d29b480
Update some Kconfig comments
2016-11-06 09:06:37 -06:00
8e94d8e7cc
Signal sent from work_signal() may interrupt the low priority worker thread that is already running. For example, the worker thread that is waiting for a semaphore could be woken up by the signal and break any synchronization assumption as a result. It also does not make any sense to send signal if it is already running and busy. This commit fixes it.
2016-11-06 08:00:12 -06:00
796969f6b6
Update TODO. Provide do-nothing stubs for mutex attribute interfaces if features not enabled. pthread_cond includes a signaling semaphore and should call sem_setprotocol.
2016-11-05 11:06:52 -06:00
6f1c5e7b43
Add some comments.
2016-11-05 09:44:29 -06:00
4b0372e542
Missing # on endif
2016-11-03 21:27:52 -06:00
Gregory Nutt