Commit graph

25 commits

Author SHA1 Message Date
Benjamin Walsh
d211a52fc0 kernel: add defines for delta_ticks_from_prev special values
Use _INACTIVE instead of hardcoding -1.

_EXPIRED is defined as -2 and will be used for an improvement so that
interrupts are not locked for a non-deterministic amount of time while
handling expired timeouts.

_abort_timeout/_abort_thread_timeout return _INACTIVE instead of -1 if
the timeout has already been disabled.

Change-Id: If99226ff316a62c27b2a2e4e874388c3c44a8aeb
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-12-15 15:50:02 -05:00
Benjamin Walsh
88b3691415 kernel/arch: enhance the "ready thread" cache
The way the ready thread cache was implemented caused it to not always
be "hot", i.e. there could be some misses, which happened when the
cached thread was taken out of the ready queue. When that happened, it
was not replaced immediately, since doing so could mean that the
replacement might not run because the flow could be interrupted and
another thread could take its place. This was the more conservative
approach that insured that moving a thread to the cache would never be
wasted.

However, this caused two problems:

1. The cache could not be refilled until another thread context-switched
in, since there was no thread in the cache to compare priorities
against.

2. Interrupt exit code would always have to call into C to find what
thread to run when the current thread was not coop and did not have the
scheduler locked. Furthermore, it was possible for this code path to
encounter a cold cache and then it had to find out what thread to run
the long way.

To fix this, filling the cache is now more aggressive, i.e. the next
thread to put in the cache is found even in the case the current cached
thread is context-switched out. This ensures the interrupt exit code is
much faster on the slow path. In addition, since finding the next thread
to run is now always "get it from the cache", which is a simple fetch
from memory (_kernel.ready_q.cache), there is no need to call the more
complex C code.

On the ARM FRDM K64F board, this improvement is seen:

Before:

1- Measure time to switch from ISR back to interrupted task

   switching time is 215 tcs = 1791 nsec

2- Measure time from ISR to executing a different task (rescheduled)

   switch time is 315 tcs = 2625 nsec

After:

1- Measure time to switch from ISR back to interrupted task

   switching time is 130 tcs = 1083 nsec

2- Measure time from ISR to executing a different task (rescheduled)

   switch time is 225 tcs = 1875 nsec

These are the most dramatic improvements, but most of the numbers
generated by the latency_measure test are improved.

Fixes ZEP-1401.

Change-Id: I2eaac147048b1ec71a93bd0a285e743a39533973
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-12-15 15:50:02 -05:00
Carles Cufi
9849df8c80 kernel: Disable interrupts after tick calculation in k_sleep()
To guarantee that the compiler does not reorder the execution of
irq_lock() with preceding operations, a volatile qualifier is
placed before the declaration of the ticks variable, which then
ensures that irq_lock() is executed after the tick calculation but
before accessing the ready and timeout queues.
Without the volatile keyword interrupts will be disabled during the
calculation of the ticks, which increases interrupt latency
significantly.

Change-Id: I2da82a1282e344f3b8d69e9457b36a4cb1d9ec18
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
2016-12-08 16:37:51 +00:00
Benjamin Walsh
a36e0cf651 kernel: remove K_TIMING thread flag
The fact that a thread is timing out was tracked via two flags: the
K_TIMING thread flag bit, and the thread's timeout's
delta_ticks_from_prev being -1 or not. This duplication could
potentially cause discrepancies if the two flags got out-of-sync, and
there was no benfits to having both.

Since timeouts that are not parts of a thread rely on the value of
delta_ticks_from_prev, standardize on it.

Since the K_TIMING bit is removed from the thread's flags, K_READY would
not reflect the reality anymore. It is removed and replaced by
_is_thread_prevented_froM_running(), which looks at the state flags that
are relevant. A thread that is ready now is not prevented from running
and does not have an active timeout.

Change-Id: I902ef9fb7801b00626df491f5108971817750daa
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-26 14:04:18 +00:00
Benjamin Walsh
358a53cb2f kernel: support for more than 32 total priorities
In addition to more priorities taking more memory to host them, finding
the next thread to run when it is not cached is slower since each extra
set of 32 priorities maps to a loop iteration. That loop is remove
entirely when the number of priorities is less than 32 (31 + the idle
thread).

Fixes ZEP-1303.

Change-Id: I3205df90d379a0f4456ff1d7f1aaa67ad2cddf15
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-18 23:45:34 +00:00
Benjamin Walsh
a4e033fdba kernel: fix obsolete access to fields in K_DEBUG() calls
When moving arch-specific thread structure to arch-agnostic, some field
accesses were missed when used in K_DEBUG statements, which are turned
off by default.

Change-Id: Ife0f49b8185a0db468deab73555f7034f20ca3e8
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-18 23:08:47 +00:00
Benjamin Walsh
f6ca7de09c kernel/arch: consolidate tTCS and TNANO definitions
There was a lot of duplication between architectures for the definition
of threads and the "nanokernel" guts. These have been consolidated.

Now, a common file kernel/unified/include/kernel_structs.h holds the
common definitions. Architectures provide two files to complement it:
kernel_arch_data.h and kernel_arch_func.h. The first one contains at
least the struct _thread_arch and struct _kernel_arch data structures,
as well as the struct _callee_saved and struct _caller_saved register
layouts. The second file contains anything that needs what is provided
by the common stuff in kernel_structs.h. Those two files are only meant
to be included in kernel_structs.h in very specific locations.

The thread data structure has been separated into three major parts:
common struct _thread_base and struct k_thread, and arch-specific struct
_thread_arch. The first and third ones are included in the second.

The struct s_NANO data structure has been split into two: common struct
_kernel and arch-specific struct _kernel_arch. The latter is included in
the former.

Offsets files have also changed: nano_offsets.h has been renamed
kernel_offsets.h and is still included by the arch-specific offsets.c.
Also, since the thread and kernel data structures are now made of
sub-structures, offsets have to be added to make up the full offset.
Some of these additions have been consolidated in shorter symbols,
available from kernel/unified/include/offsets_short.h, which includes an
arch-specific offsets_arch_short.h. Most of the code include
offsets_short.h now instead of offsets.h.

Change-Id: I084645cb7e6db8db69aeaaf162963fe157045d5a
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-12 07:04:52 -05:00
Benjamin Walsh
445830dcec kernel: add k_is_preempt_thread()
Useful for finding out if the current thread is protected against
preemption when using non-preemption to protect data structures.

Change-Id: Ib545a3609af3646ba49eeeb5a2c50dc51af010d4
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-10 23:18:56 +00:00
Benjamin Walsh
d7ad176be6 kernel: export k_sched_lock and k_sched_unlock.
Oversight. These functions are used extensively in the kernel guts, but
are also supposed to be an API.

k_sched_lock used to be implemented as a static inline. However, until
the header files are cleaned-up, and everything, including applications
get access to the kernel internal data structures, it must be
implemented as a function. To reduce the cost to the internals of the
kernel, the new internal _sched_lock() contains the same implemetation,
but is inlined.

Change-Id: If2f61d7714f87d81ddbeed69fedd111b8ce01376
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-10 23:18:55 +00:00
Benjamin Walsh
3cc2ba9f9c kernel: add __ASSERT() for thread priorities
Verify the thread priorities are within the bounds when starting a new
thread and when changing the priority of a thread.

Change-Id: I007b3b249e4b80235b6439cbee44cad2f31973bb
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-11-08 20:27:31 -05:00
Peter Mitsis
348eb4c4fb unified: Update kernel.h doxygen comments
Jira: ZEP-981
Change-Id: I3797f5e8824b4a01153f3c1fe0e070d3b107f596
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
2016-10-31 23:56:04 +00:00
Allan Stephens
6c98c4d378 unified: Ensure delays do not time out prematurely
Ensures that all APIs which accept a timeout value wait for at least
the specified amount of time, and do not time out prematurely.

* The kernel now waits for the next system clock tick to occur before
  the timeout interval is considered to have started. (That is, the only
  way to ensure a delay of N tick intervals is to wait for N+1 ticks
  to occur.)

* Gets rid of ticks -> milliseconds -> ticks conversion in task_sleep()
  and fiber_sleep() legacy APIs, since this introduces rounding that
  -- coupled with the previous change -- can alter the number of ticks
  being requested during the sleep operation.

* Corrects work queue API that was incorrectly shown to use a delay
  measured in ticks, rather than milliseconds.

Change-Id: I8b04467237b24fb0364c8f344d872457418c18da
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
2016-10-21 15:33:29 +00:00
Benjamin Walsh
b4b108de4d unified: rename sched.h to ksched.h
Build breaks when enabling CONFIG_NEWLIB_LIBC because it has its own
sched.h file.

This is a bad symptom of a greater issue: the build system passes many
'-I<path>' options to the compiler, and that allows including header
files by simply specifying their names (when located somewhere else than
<zephyr>/include/) and can cause clashes when several files in different
locations have the same name, like in this case.

Fixes ZEP-1062.

Change-Id: I81d1d69ee6669a609cd0c420b1b8f870d17dcb67
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-14 20:49:36 +00:00
Benjamin Walsh
3751123fa9 unified: align prototype and definition of k_thread_priority_set
Change-Id: I7299ded2fc66acb5806499ea26e1e5daedc7cc9c
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-14 20:49:36 +00:00
Allan Stephens
399d0ad55a unified: Rationalize thread priority APIs
* Gets rid of k_current_priority_get(). Users can just call
  k_thread_priority_get(k_current_get()) instead.

* Declares k_thread_priority_get() in kernel.h, where it
  really belongs.

* Removes duplicate declaration of k_thread_priority_set().

Change-Id: I616ae6f2e06c95ecba3b92324186b3fa29162fd1
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
2016-10-13 13:53:59 +00:00
Benjamin Walsh
ce9f782a7f unified: use _is_thread_pending() rather than poking in _timeout
That's what the interfaces are for.

Change-Id: Ida0b4d6561c7848a63bcb06537e0dbaafd8dbbfd
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:43 +00:00
Benjamin Walsh
b7ef0cba5f unified: remove last instances of struct tcs
Change-Id: I956bf0e96266e68ac1743f02a82ffafe77ebb0e8
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:38 +00:00
Benjamin Walsh
688973e5ab unified: streamline "timeout add" internal interfaces.
Rename _do_timeout_add to _add_timeout, rename _TIMEOUT_ADD to
_add_thread_timeout to better reflect their functionalities. Have the
latter call the former, remove _do_timeout_add and
_nano_timer_timeout_add.

Change-Id: Ica86bea10d99d72bf78379598a942d277e7002d0
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:36 +00:00
Benjamin Walsh
7caef4581d unified: streamline "timeout abort" internal interface
Rename _do_timeout_abort to _abort_timeout, rename _timeout_abort to
_abort_thread_timeout to better reflect their functionalities. Have the
latter call the former, remove _do_timeout_abort and
_nano_timer_timeout_abort.

Change-Id: I0fea9474b19a2eb47a37489eb06c0d1d56886c9c
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:34 +00:00
Benjamin Walsh
35497d6c5e unified: cache the next thread to run
When adding a thread to the ready queue, it is often known at that time
if the thread added will be the next one to run or not. So, instead of
simply updating the ready queues and the bitmask, also cache what that
thread is, so that when the scheduler is invoked, it can simply fetch it
from there. This is only done if there is a thread in the cache, since
the way the cache is updated is by comparing the priorities of the
thread being added and the cached thread.

When a thread is removed from the ready queue, if it is currently the
cached thread, it is also removed from the cache. The cache is not
updated at this time, since this would be a preemptive fetching that
could be overriden before the newly cached thread would even be
scheduled in.

Finally, when a thread is scheduled in, it now becomes the cached thread
since the fact that it is running means that by definition it was the
next one to run.

Doing this can speed up considerably some context switch times,
especially when a thread is preempted by an interrupt and the same
thread is scheduled when the interrupt exits.

Change-Id: I6dc8391cfca566699bb9b217eafe6bc6a063c8bb
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:31 +00:00
Benjamin Walsh
7bc86c0344 unified: use sys_dlist_peek_head_not_empty()
When fetching the next thread to run, we know at least one thread is
available.

Change-Id: I568c33a61b6a0a6d6a7f79c337caecffd5ef70b6
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-10-10 21:27:30 +00:00
Peter Mitsis
653ff0b949 unified: Remove check in _reschedule_threads()
The routine _reschedule_threads() does not need to check that the
scheduler is locked as this is done as part of _must_switch_threads().

Change-Id: I701b811512836eef591b2adcd708991ec9324b3d
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
2016-09-23 18:25:49 +00:00
Peter Mitsis
68d1f4b562 unified: Add timeslice support
Change-Id: I5b6c1ef5c015d1ddaea21b1c5447336b1b04db39
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
2016-09-20 15:28:54 +00:00
Peter Mitsis
bd69658571 unified: Add _is_next_thread_current()
The routine _is_next_thread_current() checks if the current thread
is still the highest priority thread that is ready to run.

It is useful for determining if a thread must be swapped out when
timeslicing is in effect.

Change-Id: Ide7b89742a64f6082ca4c679a4b2fbd60792e30f
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
2016-09-20 15:28:54 +00:00
Benjamin Walsh
456c6daa9f unified: initial unified kernel implementation
Summary of what this includes:

    initialization:

    Copy from nano_init.c, with the following changes:

    - the main thread is the continuation of the init thread, but an idle
      thread is created as well

    - _main() initializes threads in groups and starts the EXE group

    - the ready queues are initialized

    - the main thread is marked as non-essential once the system init is
      done

    - a weak main() symbol is provided if the application does not provide a
      main() function

    scheduler:

    Not an exhaustive list, but basically provide primitives for:

    - adding/removing a thread to/from a wait queue
    - adding/removing a thread to/from the ready queue
    - marking thread as ready
    - locking/unlocking the scheduler
      - instead of locking interrupts
    - getting/setting thread priority
      - checking what state (coop/preempt) a thread is currenlty running in
    - rescheduling threads
    - finding what thread is the next to run
    - yielding/sleeping/aborting sleep
    - finding the current thread

    threads:

    - Add operationns on threads, such as creating and starting them.

    standardized handling of kernel object return codes:

    - Kernel objects now cause _Swap() to return the following values:
         0      => operation successful
        -EAGAIN => operation timed out
        -Exxxxx => operation failed for another reason

    - The thread's swap_data field can be used to return any additional
    information required to complete the operation, such as the actual
    result of a successful operation.

    timeouts:

    - same as nano timeouts, renamed to simply 'timeouts'
    - the kernel is still tick-based, but objects take timeout values in
      ms for forward compatibility with a tickless kernel.

    semaphores:

      - Port of the nanokernel semaphores, which have the same basic behaviour
      as the microkernel ones. Semaphore groups are not yet implemented.

      - These semaphores are enhanced in that they accept an initial count and a
      count limit. This allows configuring them as binary semaphores, and also
      provisioning them without having to "give" the semaphore multiple times
      before using them.

    mutexes:

    - Straight port of the microkernel mutexes. An init function is added to
    allow defining them at runtime.

    pipes:

    - straight port

    timers:

    - amalgamation of nano and micro timers, with all functionalities
      intact.

    events:

    - re-implementation, using semaphores and workqueues.

    mailboxes:

    - straight port

    message queues:

    - straight port of  microkernel FIFOs

    memory maps:

    - straight port

    workqueues:

    - Basically, have all APIs follow the k_ naming rule, and use the _timeout
    subsystem from the unified kernel directory, and not the _nano_timeout
    one.

    stacks:

    - Port of the nanokernel stacks. They can now have multiple threads
    pending on them and threads can wait with a timeout.

    LIFOs:

    - Straight port of the nanokernel LIFOs.

    FIFOs:

    - Straight port of the nanokernel FIFOs.

Work by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
         Peter Mitsis <peter.mitsis@windriver.com>
         Allan Stephens <allan.stephens@windriver.com>
         Benjamin Walsh <benjamin.walsh@windriver.com>

Change-Id: Id3cadb3694484ab2ca467889cfb029be3cd3a7d6
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
2016-09-13 17:12:55 -04:00