Revises the names of 2 types to make them better reflect the
data structures they represent, namely:
- block set: a collection of memory pool blocks of a given size
- quad-block: four contiguous memory pool blocks of a given size
Change-Id: I3e424586e97157eea185fba6836e2e89d10d9cd6
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Eliminates the memory pool field indicating the size of the
pool's memory buffer, since it isn't used anywhere.
(Anyway, it could be computed by multiplying the maximum
block size by the number of maximum-sized blocks.)
Change-Id: Ia11554bdc2b246a1ba0ea33f05c5e6ce6a32ca13
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Alters the use of the field so that a 1 bit indicates that the
associated block is available (i.e. can be allocated from the pool),
and a 0 bit indicates the block is unavailable (i.e. it is already
allocated or does not exist).
The revised definition of this field is more intuitive, since it follows
the conventional "1 = thing is present, 0 = this is not present" model.
Change-Id: Id133d1940aca0dd1c3c1672f989d0c0bb083ebc5
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
No longer initializes the quad-block descriptors for block sets,
except for the one containing the largest size blocks. The descriptors
for the other block sets don't need initialization since sysgen already
ensures the block pointer field is NULL, and the block status field can
be anything (since it is ignored if the pointer is NULL).
Change-Id: I72cdf772329ef3e6d1babd9da11706d11611e61a
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Adds missing check to prevent defragmentation algorithm from
going past the end of a block set's array of quad-blocks.
Optimizes quad-block deletion algorithm so that it simply moves
the final array entry into the slot for the deleted entry,
rather than shifting all of the entries in between them.
Also enhances comments to make it easier to understand
what is going on.
Change-Id: Ic281c6f3c6dd5df9ec532a302b4103f3d929665b
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Now use a standard routine for identifying which block set to use,
rather than duplicating the identification code in multiple places.
Change-Id: I2f6577879c23183f3f91e4418d1ea0b2f6eec184
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Revises algorithm to call an existing routine to free an allocated
block, rather than duplicating the logic to do the freeing.
Also enhances comments to make it easier to understand what is
going on.
Change-Id: Ie7a83a8a2e978e8a685c27e07acb5b70aea37afa
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Revises algorithm for deallocating an existing block to make it
more compact (and efficient). Address calculation is simplified
and unnecessary error checking is removed (& replaced by asserts).
Also enhances routine name and comments to make it easier to
understand what is going on.
Change-Id: If1803ee09c8f4f73693c0cff9e433a7938f14398
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Enhances comments to make it easier to understand what is going on.
(Doesn't change actual defragmentation algorithm.)
Change-Id: I294f55b8f233d88c01ce30ba9ccff88000dc7936
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Enhances routine name and comments to make it easier to understand
what is going on. (Doesn't change actual allocation algorithm.)
Change-Id: I7d982ed8eeda3a2edd4602c5a10003aa8532457d
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Revises algorithm for allocating an existing free block to make it
more compact (and efficient) by using the same logic to handle the
allocation of any block within a quad-block.
Also enhances routine name and comments to make it easier to
understand what is going on.
Change-Id: I55ca513e6f85df2b548502262e4dbe6bb272596b
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Converts check for a request for an impossibly large block
from a run-time check to an assertion. (The run-time check logic
was faulty anyway, since an excessive request would result
in a negative value for "offset", which would then get used
as an array index!)
Also enhances comments to make it easier to understand what is
going on.
Change-Id: I3faa86be177dcbc76912e23fabc2d24724fcba18
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Makes a couple of minor optimizations to initialization algorithm:
- Now initializes block status array for largest size blocks once,
rather than twice.
- Doesn't initialize "count" field of block size descriptors to zero,
since sysgen ensures that this happens automatically.
Also enhances comments for initialization routine to make it easier
to understand what is going on.
Change-Id: I00d907c0f2a86f5b6ea8a63475b40074fe89357c
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Eliminates memory pool field associated with a capability that was
never implemented. (This field was initialized, but never subsequently
referenced.)
Change-Id: I58cf8c4bb846a66b4c8754654ffc3ff55abcff7f
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Used by ARC, ARM, Nios II. x86 has alternate code done in assembly.
Linker scripts had some alarming comments about data/BSS overlap,
but the beginning of BSS is aligned so this can't happen even if
the end of data isn't.
The common code doesn't use fake pointer values for the number of
words in these sections, don't compute or export them.
Change-Id: I4291c2a6d0222d0a3e95c140deae7539ebab3cc3
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
We now allow use of -mgpopt=global and -mgpopt=data. The 'global'
option is now the default instead of compiler-default local, expanding
global pointer usage to all small data in the system.
For systems where all RAM is less than 64K, the 'data' option may be
appropriate.
Some fixes had to be made to the system in order to get around some
issues:
* prep_c.c no longer uses fake linker variables to figure out the size
of data or BSS, as these gave the linker fits as it tried to compute
relative addresses to them.
* _k_task_ptr_idle is create by sysgen and placed in a special section.
Any small data in a special section needs to be declared extern
with __attribute__((section)) else the compiler will assume it's in
.sdata.
* same situation with extern references to k_pipe_t (fixed pipe_priv
test)
For legacy applications being ported to Nios II which do things that
freak out global pointer calculation, it can be disabled entirely.
Change-Id: I5eb86ee8aefb8e2fac49c5cdd104ee19cea23f6f
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Introduce the family of nano_fifo_put_list and nano_fifo_put_slist APIs,
which allow queuing a list of elements on a nanokernel fifo in one
shot. When called from an ISR or a fiber, the behaviour is not really
different than calling nano_fifo_put for each element to enqueue.
However, when called from a task, it allows the task to enqueue the full
list without yielding to fibers that were waiting on the fifo.
All fibers currently waiting on the fifo will be awakened and given an
element from the list in their order of priority. When some elements are
not matched with a receiver, they are queued normally.
There are two ways of passing a list: with either an ad-hoc queue, by
passing the head and the tail elements, or with a sys_slist_t object.
For the latter, the object must be reinitialized afterwards.
Change-Id: I6ac077f556dc39995191e9149c4a047a3433826f
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
It is now safe to introduce the callback since nano_timer_init now
calls _nano_timeout_init which does takes care of initializing all
the fields properly.
Change-Id: I5735eeebef233a0a541ec8b2a354b65da98082fc
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
With the introduction of _nano_timeout_init it prefered to call it
to initialize the _nano_timeout fields properly.
Change-Id: I83e9c63f9bb2903c508264d1199d2c687c330ec8
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This will allow initializing a nano timeout that is not associated with
a thread.
Change-Id: Ic71175b0059396b19a0e3616f4fab570071c3d48
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This adds a delayed version of nano_work API which is useful when
handling timeouts since the same stack/workqueue can be shared.
Change-Id: Iac43796fe96deb0a9c8976c91a65104b57779b00
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Convert leading whitespace into tabs in Kconfig files. Also replaced
double spaces between config and <prompt>.
Change-Id: I341c718ecf4143529b477c239bbde88e18f37062
Signed-off-by: Kumar Gala <kumar.gala@linaro.org>
By default, kernel event logger is using the system timer. But on
some platforms where the timer driver maintains the system timer
cycle accumulator in software, such as ones using the LOAPIC timer,
the system timer behavior leads to timestamp errors. For example,
the timer interrupt is logged with a wrong timestamp since the HW
timer value has been reset (periodic mode) but accumulated value not
updated yet (done later in the ISR).
This patch is adding the possibility to register a timer callback
function that will be used by the kernel event logger. For example,
on Quark SE, this allows using RTC or AON counter which accuracy is
sufficient and behavior more straight forward compared to system
timer.
Change-Id: I754c7557350ef29fc10701e62a35a5425e035f11
Signed-off-by: Fabrice Olivero <fabrice.olivero@intel.com>
This adds a callback to struct _nano_timeout which is called in ISR
context allowing more flexible handling of timeouts.
Change-Id: If837b0b51b24dfffebac6f99f4d66fdf01c164f0
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Overflowing the k_server command stack will now trigger an __ASSERT()
when CONFIG_ASSERT=y.
Change-Id: Icf6f4242ab8a9897918769415a0f7485602d7630
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
Add a generic API for drivers to start workqueues and submit work
items. This is needed by drivers which need to schedule code that might
sleep from an ISR to run in fiber context.
Also add the option to start a system-wide workqueue.
Both additions are optional. They can be deactivated for systems that
do not need them.
Change-Id: Ia843568fde5daf6d4279ef7bf241c26c1e3dcfb7
Signed-off-by: Vlad Dogaru <vlad.dogaru@intel.com>
Moved comments from code to header.
Jira: ZEP-160
Change-Id: Ifd0f3c930289256e682b5941d77433aca3d3f941
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Added CONFIG_KERNEL_EVENT_PROFILER_DYNAMIC flag for enabling that
capability. When set, nothing will be logged by default
Change-Id: I03552483e5a6bfd9e2505eda56908f0d0ae98618
Signed-off-by: Fabrice Olivero <fabrice.olivero@intel.com>
OS tick period (usually 10 ms) is not sufficiently precise for task
execution analysis (like processing CPU load per context)
With that modification, the timestamp used by the kernel event logger
is 32-bit LSB of platform HW timer (for example Lakemont APIC timer
for Quark SE).
This timer period is very small and leads to timestamp wraparound
happening quite often (e.g. every 134s for Quark SE).
This wraparound must be considered when analyzing kernel event logger
data and care must be taken when tickless idle is enabled and sleep
duration can exceed maximum HW timer value.
Change-Id: Idc545da8f828a7357a69d83ff25c9afd09dab3c4
Signed-off-by: Fabrice Olivero <fabrice.olivero@intel.com>
Add a way of finding if a microkernel task is the idle task instead of
of duplicating:
task_priority_get() == (CONFIG_NUM_TASK_PRIORITIES - 1)
which could be subject to change.
Only available for microkernel, since there is no such concept in the
nanokernel.
Change-Id: Ie8930981f1a2ac5ff16f905f4eb4e333c8b59c5d
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
In order to build test/kernel/test_stackprot/microkernel for the ARC,
the _MOVE_INSTR needs to define what the move instruction is for this
target.
Change-Id: I087cc5baa4c41297ce52323556e94aab424aa891
Signed-off-by: Chuck Jordan <cjordan@synopsys.com>
Locks interrupts in the microkernel routines _k_state_bit_reset() and
_k_state_bit_set(). This is a necessary pre-requisite for allowing
microkernel objects to pend on nanokernel objects since that feature
will require the manipulation of the microkernel queues in the context
of an ISR as well as the kernel server fiber.
Change-Id: I2d263707e0d3aed75bba971df878daa3d7ae1d11
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds support that allows fibers and ISRs to invoke the microkernel
no-op kernel service request. This is useful for cases when the
nanokernel needs to invoke the microkernel task scheduler.
Change-Id: I1f4b2a39ac6b5e44bb1b6c6b3cd6034262bbada8
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Fixes the timeout recalculation in the following routines:
nano_task_fifo_get()
nano_task_lifo_get()
nano_task_sem_take()
_nano_task_sleep()
Without this fix, a task that called one of the previously listed
routines could in theory sleep/wait up to almost twice the requested
timeout.
Change-Id: I53196be84e65874e94a62d5b0be1b7aaaaeda91f
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds the nanokernel infrastructure to permit microkernel tasks
to block/unblock on nanokernel objects. Multiple tasks may wait
on a nanokernel object's dedicated task wait queue.
It is important to note that when data is posted to the object
all the tasks on that object's dedicated task wait queue may be
woken up but the data is not immediately given to any of the tasks.
This is done to maintain consistent behavior with the nanokernel
as in a nanokernel system, fibers are given preference in both
waiting on and getting data from a nanokernel object.
Change-Id: Ia5c7f21ae59a367d9fec23dafc3a918d9e767db5
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds the microkernel infrastructure to permit a microkernel task
to [un]block on a nanokernel object. Unlike tasks that [un]block
on microkernel objects, the work for [un]blocking tasks on nanokernel
objects will not always be done in the kernel service fiber. One of
the repercussions of this is that in many cases the microkernel task
scheduler must be explicitly invoked (by issuing a no-op kernel
service call).
Origin: Original
Change-Id: I2b145668cef142a7a4034e191116fcb344a9b8b3
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Fibers initialize this back pointer to NULL as they are (by definition)
not microkernel tasks. Microkernel tasks initialize it to their
corresponding 'ktask_t'.
However for nanokernel systems, the back pointer is always NULL. This
is because there is only one task in a nanokernel system (the background
task) and it can not pend on a nanokernel object--it must poll.
Change-Id: I9840fecc44224bef63d09d587d703720cf33ad57
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds the TF_NANO wait flag reason to indicate that a task is pending
on a nanokernel object.
Change-Id: Ic20ef79398da7d9118bdf775e22b8f8a31501f7f
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
We really should have more faith in the compiler, it generates
code to implement this exactly like the arch-specific assembly
versions, and on ARM is actually 4 bytes shorter.
FUNC_NO_FP used to disable the usual C preamble to update the
frame/stack pointers, which is how the sizes are still the same
or less. It's debatable how useful the occasional use of
FUNC_NO_FP is in practice since it hinders debugging and in a
production build frame pointers should be globally disabled, but
we can address that later.
Change-Id: I6c4b64ab3e3a9b6f91d52fa8c92e6e79a986fc77
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Of the 3 related functions;
_thread_essential_set()
_thread_essential_clear()
_is_thread_essential()
The first two are parameter-less and always operate on
"_nanokernel.current". The last one takes a 'thread' parameter but will
operate on _nanokernel.current if the parameter is NULL. All calls to
_is_thread_essential() pass NULL!
This change makes the 3 functions consistent by removing the parameter
to the 3rd function. This should also be marginally more efficient,
though consistency was the motivation. This change corrects the doc
preamble to all 3 functions.
(These functions would probably be better as inlines. Also, the choice
of when to use wrappers seems a bit arbitrary. E.g. there's nothing
for setting/testing the "FIBER" flag.)
Change-Id: Ie3589f8a28b227c6d7a3a31b664d3b3e6e9c6d17
Signed-off-by: Geoff Thorpe <geoff.thorpe@nxp.com>
This wrapper exists so that the ESSENTIAL bit doesn't have to be set
explicitly in _nanokernel.current. (And a note is added to nano_init to
avoid a nasty gotcha there.)
Change-Id: I4026ac0cfeace60e36abdbccf15554759f2dbf7a
Signed-off-by: Geoff Thorpe <geoff.thorpe@nxp.com>
These C variants of atomic operations can work on any arch,
have platforms select them if they don't have ASM equivalents.
Change-Id: I38eb03bb58beff865681ee56ef7bc0fcded1e906
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
The file device.c uses errno codes so it should include errno.h.
Change-Id: I93e806a9b20b2c9bcb245ac5e86fba70486c9591
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Certain Low power SOC states (e.g. deep sleep) will result in device
IP blocks losing state. In such a scenario it can be useful to have
a mechanism for devices (driver code) to signal the power manager /
policy that they are in the middle of a transaction.
We expect the device driver code to make a call to
device_busy_set(device *) before initiating a transaction and
device_busy_clear(device *) on completion. It is expected that device
driver developers will add this as necessary in their drivers.
Further an API is provided for power manager application / policy to
check this. Based on this the power manager / policy can decide
whether or not to go into a particular power state.
Change-Id: I0fedd90b98e182cd41b53c7f9e08655532822faa
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
This changes the behavior of device_get_binding() so that
it returns NULL if driver_api is not set. This provides
a way for driver to state that it has not been initialized
properly, and prevents app from using it since no reference
to the device struct will be returned.
This implements the idea specified in [1]. The idea is to
reuse an existing resource by piggy-backing onto driver_api,
thus avoiding an extra "device state" variable in the device
struct. This differs from the code specified in the mailing
list by checking driver_api for NULL first. This avoids
the unnecessary strcmp() if driver_api is NULL.
[1] https://lists.zephyrproject.org/archives/list/devel@lists.zephyrproject.org/message/MZB5PYBSRHV3NIEHJYXYQVLTPFIIHPB3/
Change-Id: I978b1a6683cd56c8a72532d6368c47e67515c82d
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Fixes a problem where the nanokernel FIFO state information could get
out of sync due to a timeout.
The nanokernel FIFO structure nano_fifo now maintains two separate
queues: one for waiting fibers and the other for posted data. This
permits the safe and independent querying of the queues as needed
when getting and/or putting data from/on the nanokernel FIFO.
Change-Id: Ifbcb5004558b06fc55cad2a955f5be20e716b392
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Device drivers may use task_sleep() routine during the initialization.
As device driver initialization is carried by idle task, it can not
be sheduled out as any other task. Idle task goes into a wait loop
instead.
In order to invoke task_sleep() device drivers must enable
CONFIG_NANO_TIMEOUTS option.
Change-Id: Ib73a2ad1f3c0bda44c24f2417e102bfaa3a13a15
Signed-off-by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
