incubator-nuttx/sched/Kconfig

1987 lines
66 KiB
Plaintext

#
# For a description of the syntax of this configuration file,
# see the file kconfig-language.txt in the NuttX tools repository.
#
menuconfig DISABLE_OS_API
bool "Disable NuttX interfaces"
default y
---help---
The following can be used to disable categories of
APIs supported by the OS. If the compiler supports
weak functions, then it should not be necessary to
disable functions unless you want to restrict usage
of those APIs.
if DISABLE_OS_API
config DISABLE_POSIX_TIMERS
bool "Disable POSIX timers"
default DEFAULT_SMALL
---help---
Disable support for the the entire POSIX timer family
including timer_create(), timer_gettime(), timer_settime(),
etc.
NOTE: This option will also disable getitimer() and
setitimer() which are not, strictly speaking, POSIX timers.
config DISABLE_PTHREAD
bool "Disable pthread support"
default DEFAULT_SMALL
config DISABLE_MQUEUE
bool "Disable POSIX message queue support"
default DEFAULT_SMALL
config DISABLE_MQUEUE_SYSV
bool "Disable System V message queue support"
default DISABLE_MQUEUE
---help---
Disable System V message queue support
config DISABLE_ENVIRON
bool "Disable environment variable support"
default DEFAULT_SMALL
endif # DISABLE_OS_API
config DISABLE_IDLE_LOOP
bool "Disable idle loop support"
default n
---help---
This option allows nx_start to return instead of
entering the idle loop.
menu "Clocks and Timers"
config ARCH_HAVE_TICKLESS
bool
config SCHED_TICKLESS
bool "Support tick-less OS"
default n
depends on ARCH_HAVE_TICKLESS
---help---
By default, system time is driven by a periodic timer interrupt. An
alternative configurations is a tick-less configuration in which
there is no periodic timer interrupt. Instead an interval timer is
used to schedule the next OS time event. This option selects that
tick-less OS option. If the tick-less OS is selected, then there are
additional platform specific interfaces that must be provided as
defined in include/nuttx/arch.h
if SCHED_TICKLESS
config SCHED_TICKLESS_TICK_ARGUMENT
bool "Scheduler use tick argument"
default n
---help---
Enables use of tick argument in scheduler. If enabled, then the
board-specific logic must provide the following functions:
int up_timer_gettick(FAR clock_t *ticks);
If SCHED_TICKLESS_ALARM is enabled, then these additional interfaces are
expected:
int up_alarm_tick_cancel(FAR clock_t *ticks);
int up_alarm_tick_start(clock_t ticks);
Otherwise, these additional interfaces are expected:
int up_timer_tick_cancel(FAR clock_t *ticks);
int up_timer_tick_start(FAR clock_t ticks);
config SCHED_TICKLESS_ALARM
bool "Tickless alarm"
default n
---help---
The tickless option can be supported either via a simple interval
timer (plus elapsed time) or via an alarm. The interval timer allows
programming events to occur after an interval. With the alarm,
you can set a time in the future and get an event when that alarm
goes off. This option selects the use of an alarm.
The advantage of an alarm is that it avoids some small timing
errors; the advantage of the use of the interval timer is that
the hardware requirement may be less.
config SCHED_TICKLESS_LIMIT_MAX_SLEEP
bool "Max sleep period (in microseconds)"
default n
---help---
Enables use of the g_oneshot_maxticks variable. This variable is
initialized by platform-specific logic at runtime to the maximum
delay that the timer can wait (in configured clock ticks). The
RTOS tickless logic will then limit all requested delays to this
value.
endif
config USEC_PER_TICK
int "System timer tick period (microseconds)"
default 10000 if !SCHED_TICKLESS
default 100 if SCHED_TICKLESS
---help---
In the "normal" configuration where system time is provided by a
periodic timer interrupt, the default system timer is expected to
run at 100Hz or USEC_PER_TICK=10000. This setting must be defined
to inform of NuttX the interval that the processor hardware is
providing system timer interrupts to the OS.
If SCHED_TICKLESS is selected, then there are no system timer
interrupts. In this case, USEC_PER_TICK does not control any timer
rates. Rather, it only determines the resolution of time reported
by clock_systime_ticks() and the resolution of times that can be set for
certain delays including watchdog timers and delayed work. In this
case there is a trade-off: It is better to have the USEC_PER_TICK as
low as possible for higher timing resolution. However, the time
is currently held in 'unsigned int' on some systems, this may be
16-bits but on most contemporary systems it will be 32-bits. In
either case, smaller values of USEC_PER_TICK will reduce the range
of values that delays that can be represented. So the trade-off is
between range and resolution (you could also modify the code to use
a 64-bit value if you really want both).
The default, 100 microseconds, will provide for a range of delays
up to 120 hours.
This value should never be less than the underlying resolution of
the timer. Error may ensue.
if !SCHED_TICKLESS
config SYSTEMTICK_EXTCLK
bool "Use external clock"
default n
depends on ARCH_HAVE_EXTCLK
---help---
Use external clock for system tick. When enabled, the platform-specific
logic must start its own timer interrupt to make periodic calls to the
nxsched_process_timer() or the functions called within. The purpose is
to move the scheduling off the processor clock to allow entering low
power states that would disable that clock.
config SYSTEMTICK_HOOK
bool "System timer hook"
default n
---help---
Enable a call to a user-provided, board-level function on each timer
tick. This permits custom actions that may be performed on each
timer tick. The form of the user-provided function is:
void board_timerhook(void);
(prototyped in include/nuttx/board.h).
endif # !SCHED_TICKLESS
config SYSTEM_TIME64
bool "64-bit system clock"
default n
---help---
The system timer is incremented at the rate determined by
USEC_PER_TICK, typically at 100Hz. The count at any given time is
then the "uptime" in units of system timer ticks. By default, the
system time is 32-bits wide. Those defaults provide a range of about
497 days which is probably a sufficient range for "uptime".
However, if the system timer rate is significantly higher than 100Hz
and/or if a very long "uptime" is required, then this option can be
selected to support a 64-bit wide timer.
config ARCH_HAVE_ADJTIME
bool
default n
config CLOCK_ADJTIME
bool "Support adjtime function"
default n
depends on ARCH_HAVE_ADJTIME || RTC_ADJTIME
---help---
Enables usage of adjtime() interface used to correct the system time
clock. This requires specific architecture support.
Adjustment can affect system timer period and/or high-resolution RTC.
These are implemented by interfaces up_adjtime() and up_rtc_adjtime().
This is not a POSIX interface but derives from 4.3BSD, System V.
It is also supported for Linux compatibility.
if CLOCK_ADJTIME
config CLOCK_ADJTIME_SLEWLIMIT_PPM
int "Adjtime slew limit"
default 20000
range 1 1000000
---help---
Set limit of adjtime() clock slewing as parts per million.
In real time systems we do not want the time to adjust too quickly.
For example CLOCK_ADJTIME_SLEWLIMIT=1000 will slow or speed the timer
tick period by at most 0.1 percent of the nominal value.
config CLOCK_ADJTIME_PERIOD_MS
int "Adjtime period"
default 970
range 1 3600000
---help---
Define system clock adjustment period in milliseconds.
The adjustment commanded by adjtime() call is applied over this time period.
endif
config ARCH_HAVE_TIMEKEEPING
bool
default n
config CLOCK_TIMEKEEPING
bool "Support timekeeping algorithms"
default n
depends on ARCH_HAVE_TIMEKEEPING
---help---
CLOCK_TIMEKEEPING enables experimental time management algorithms.
config JULIAN_TIME
bool "Enables Julian time conversions"
default n
---help---
Enables Julian time conversions
config START_YEAR
int "Start year"
default 2018
range 1970 2106
---help---
NuttX uses an unsigned 32-bit integer for time_t which provides a
range from 1970 to 2106.
config START_MONTH
int "Start month"
default 1
range 1 12
config START_DAY
int "Start day"
default 1
range 1 31
config PREALLOC_TIMERS
int "Number of pre-allocated POSIX timers"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
depends on !DISABLE_POSIX_TIMERS
---help---
The number of pre-allocated POSIX timer structures. The system manages a
pool of preallocated timer structures to minimize dynamic allocations. Set to
zero for all dynamic allocations.
config PERF_OVERFLOW_CORRECTION
bool "Compensate perf count overflow"
depends on SYSTEM_TIME64 && (ALARM_ARCH || TIMER_ARCH || ARCH_PERF_EVENTS)
default n
---help---
If this option is enabled, then the perf event will be enabled
by default.
When enabled, it will always return an increasing count value to
avoid overflow on 32-bit platforms.
endmenu # Clocks and Timers
menu "Tasks and Scheduling"
config SPINLOCK
bool "Support Spinlocks"
default n
---help---
Enables support for spinlocks. Spinlocks are used primarily for
synchronization in SMP configurations but are available for general
synchronization between CPUs. Use in a single CPU configuration would
most likely be fatal. Note, however, that this does not depend on
CONFIG_ARCH_HAVE_MULTICPU. This permits the use of spinlocks in
other novel architectures.
if SPINLOCK
config TICKET_SPINLOCK
bool "Use ticket Spinlocks"
default n
---help---
Use ticket spinlock algorithm.
config RW_SPINLOCK
bool "Support read-write Spinlocks"
default y
---help---
Spinlocks are spilit into read and write lock.
Reader can take read lock simultaneously and only one writer
can take write lock.
endif # SPINLOCK
config IRQCHAIN
bool "Enable multi handler sharing a IRQ"
default n
---help---
Enable support for IRQCHAIN.
if IRQCHAIN
config PREALLOC_IRQCHAIN
int "Number of pre-allocated irq chains"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
---help---
The number of pre-allocated irq chain structures. The system manages
a pool of preallocated irq chain structures to minimize dynamic
allocations. You will, however, get better performance and memory
usage if this value is tuned to minimize such allocations.
endif # IRQCHAIN
config IRQCOUNT
bool
default n
config SMP
bool "Symmetric Multi-Processing (SMP)"
default n
depends on ARCH_HAVE_MULTICPU
depends on ARCH_HAVE_TESTSET
depends on ARCH_INTERRUPTSTACK != 0
select SPINLOCK
select IRQCOUNT
---help---
Enables support for Symmetric Multi-Processing (SMP) on a multi-CPU
platform.
N.B. SMP mode requires the use of ARCH_INTERRUPTSTACK:
CPU0 thread0 -> IRQ enter -> add thread0 to block_list -> IRQ leave(crash)
||
/\
/ \
CPU1 thread1 -> block_task -> take thread0 from block_list -> run thread0
CPU0 IRQ handler use thread0's stack, but thread0 may switch to CPU1, that
will caused IRQ handler stack corruption.
if SMP
config SMP_NCPUS
int "Number of CPUs"
default 4
range 1 32
---help---
This value identifies the number of CPUs supported by the processor
that will be used for SMP.
If CONFIG_DEBUG_FEATURES is enabled, then the value one is permitted
for CONFIG_SMP_NCPUS. This is not normally a valid setting for an
SMP configuration. However, running the SMP logic in a single CPU
configuration is useful during certain testing.
config SMP_DEFAULT_CPUSET
hex "Default CPU bit set"
default 0xffffffff
---help---
Set the Default CPU bits. The way to use the unset CPU is to call the
sched_setaffinity function to bind a task to the CPU. bit0 means CPU0.
config SMP_CALL
bool "Support SMP function call"
default n
---help---
Enable to support SMP function call.
endif # SMP
choice
prompt "Initialization Task"
default INIT_ENTRY if !BUILD_KERNEL
default INIT_FILE if !BINFMT_DISABLE
default INIT_NONE if BINFMT_DISABLE
config INIT_NONE
bool "None"
config INIT_ENTRY
bool "Via application entry"
depends on !BUILD_KERNEL
config INIT_FILE
bool "Via executable file"
depends on !BINFMT_DISABLE
endchoice # Initialization task
config INIT_ARGS
string "Application argument list"
depends on !INIT_NONE
---help---
The argument list for user applications. e.g.:
"\"arg1\",\"arg2\",\"arg3\""
config INIT_STACKSIZE
int "Main thread stack size"
default DEFAULT_TASK_STACKSIZE
---help---
The size of the stack to allocate for the user initialization thread
that is started as soon as the OS completes its initialization.
config INIT_PRIORITY
int "init thread priority"
default 100
---help---
The priority of the user initialization thread.
if INIT_ENTRY
config INIT_ENTRYPOINT
string "Application entry point"
default "main"
---help---
The name of the entry point for user applications. For the example
applications this is of the form 'app_main' where 'app' is the application
name. If not defined, INIT_ENTRYPOINT defaults to "main".
Note that main must take "argc" and "argv" arguments:
int main(int argc, FAR char *argv[])
Otherwise, if using a signature such as "int main(void)" a compilation
error will result:
> $ make
> CC: CustomHello.c <command-line>: error: conflicting types for
> 'custom_hello_main'
> CustomHello.c:3:5: note: in expansion of macro 'main'
> 3 | int main(void)
> | ^~~~
config INIT_ENTRYNAME
string "Application entry name"
default INIT_ENTRYPOINT
endif # INIT_ENTRY
if INIT_FILE
config INIT_FILEPATH
string "Application initialization path"
default "/bin/init"
---help---
The name of the entry point for user applications. For the example
applications this is of the form 'app_main' where 'app' is the application
name. If not defined, INIT_ENTRYPOINT defaults to "main".
config INIT_SYMTAB
string "Symbol table"
default "NULL" if !EXECFUNCS_HAVE_SYMTAB
default EXECFUNCS_SYMTAB_ARRAY if EXECFUNCS_HAVE_SYMTAB
depends on BUILD_FLAT
---help---
The name of other global array that holds the exported symbol table.
The special string "NULL" may be provided if there is no symbol
table. Quotation marks will be stripped when config.h is generated.
NOTE: This setting cannot be used in protected or kernel builds.
Any kernel mode symbols tables would not be usable for resolving
symbols in user mode executables.
config INIT_NEXPORTS
string "Symbol table size"
default "0" if !EXECFUNCS_HAVE_SYMTAB
default EXECFUNCS_NSYMBOLS_VAR if EXECFUNCS_HAVE_SYMTAB
depends on BUILD_FLAT
---help---
The size of the symbol table. NOTE that is is logically a numeric
value but is represent by a string. That allows you to put
sizeof(something) or a macro or a global variable name for the
symbol table size. Quotation marks will be stripped when config.h
is generated.
NOTE: This setting cannot be used in protected or kernel builds.
Any kernel mode symbols tables would not be usable for resolving
symbols in user mode executables.
menuconfig INIT_MOUNT
bool "Auto-mount init file system"
default n
depends on !DISABLE_MOUNTPOINT
---help---
In order to use the the initial startup program when CONFIG_INIT_FILEPATH
is provided, it is necessary to mount the initial file system that
provides init program. Normally this mount is done in the board-specific
initialization logic. However, if the mount is very simple, it can be
performed by the OS bring-up logic itself by selecting this option.
if INIT_MOUNT
config INIT_MOUNT_SOURCE
string "The block device to mount"
default "/dev/ram0"
config INIT_MOUNT_TARGET
string "Path to the mounted file system"
default "/bin"
config INIT_MOUNT_FSTYPE
string "The file system type to mount"
default "romfs"
config INIT_MOUNT_FLAGS
hex "Flags passed to mount"
default 0
config INIT_MOUNT_DATA
string "Additional data passed to mount"
default ""
endif # INIT_MOUNT
endif # INIT_FILE
menuconfig ETC_ROMFS
bool "Auto-mount etc baked-in ROMFS image"
default n
depends on !DISABLE_MOUNTPOINT && FS_ROMFS
---help---
Mount a ROMFS filesystem at /etc and provide a system init
script at /etc/init.d/rc.sysinit and a startup script
at /etc/init.d/rcS. The default system init script will mount
a FAT FS RAMDISK at /tmp but the logic is easily extensible.
if ETC_ROMFS
config ETC_CROMFS
bool "Support CROMFS (compressed) start-up script"
default n
depends on FS_CROMFS
---help---
Mount a CROMFS filesystem at /etc and provide a compressed system
init script at /etc/init.d/rc.sysinit and a startup script
at /etc/init.d/rcS.
config ETC_ROMFSMOUNTPT
string "Mountpoint of the etc romfs image"
default "/etc"
config ETC_ROMFSDEVNO
int "ROMFS block device minor number"
default 0
---help---
This is the minor number of the ROMFS block device. The default is
'0' corresponding to /dev/ram0.
config ETC_ROMFSSECTSIZE
int "ROMFS sector size"
default 64
---help---
This is the sector size to use with the ROMFS volume. Since the
default volume is very small, this defaults to 64 but should be
increased if the ROMFS volume were to be become large. Any value
selected must be a power of 2.
config ETC_FATDEVNO
int "FAT block device minor number"
default 1
depends on FS_FAT
---help---
When the default rcS file used when ETC_ROMFS is selected, it
will mount a FAT FS under /tmp. This is the minor number of the FAT
FS block device. The default is '1' corresponding to /dev/ram1.
config ETC_FATSECTSIZE
int "FAT sector size"
default 512
depends on FS_FAT
---help---
When the default rcS file used when ETC_ROMFS is selected, it
will mount a FAT FS under /tmp. This is the sector size use with the
FAT FS. Default is 512.
config ETC_FATNSECTORS
int "FAT number of sectors"
default 1024
depends on FS_FAT
---help---
When the default rcS file used when ETC_ROMFS is selected, it
will mount a FAT FS under /tmp. This is the number of sectors to use
with the FAT FS. Default is 1024. The amount of memory used by the
FAT FS will be ETC_FATSECTSIZE * ETC_FATNSECTORS bytes.
config ETC_FATMOUNTPT
string "FAT mount point"
default "/tmp"
depends on FS_FAT
---help---
When the default rcS file used when ETC_ROMFS is selected, it
will mount a FAT FS under /tmp. This is the location where the FAT
FS will be mounted. Default is "/tmp".
endif # ETC_ROMFS
config RR_INTERVAL
int "Round robin timeslice (MSEC)"
default 0
---help---
The round robin timeslice will be set this number of milliseconds;
Round robin scheduling (SCHED_RR) is enabled by setting this
interval to a positive, non-zero value.
config SCHED_SPORADIC
bool "Support sporadic scheduling"
default n
select SCHED_SUSPENDSCHEDULER
select SCHED_RESUMESCHEDULER
---help---
Build in additional logic to support sporadic scheduling
(SCHED_SPORADIC).
if SCHED_SPORADIC
config SCHED_SPORADIC_MAXREPL
int "Maximum number of replenishments"
default 3
range 1 255
---help---
Controls the size of allocated replenishment structures and, hence,
also limits the maximum number of replenishments.
config SPORADIC_INSTRUMENTATION
bool "Sporadic scheduler monitor hooks"
default n
---help---
Enables instrumentation in the sporadic scheduler to monitor
scheduler behavior. If enabled, then the board-specific logic must
provide the following functions:
void arch_sporadic_start(FAR struct tcb_s *tcb);
void arch_sporadic_lowpriority(FAR struct tcb_s *tcb);
void arch_sporadic_suspend(FAR struct tcb_s *tcb);
void arch_sporadic_resume(FAR struct tcb_s *tcb);
endif # SCHED_SPORADIC
config TASK_NAME_SIZE
int "Maximum task name size"
default 31
---help---
Specifies the maximum size of a task name to save in the TCB.
Useful if scheduler instrumentation is selected. Set to zero to
disable. Excludes the NUL terminator; the actual allocated size
will be TASK_NAME_SIZE + 1. The default of 31 then results in
a align-able 32-byte allocation.
config SCHED_HAVE_PARENT
bool "Support parent/child task relationships"
default n
---help---
Remember the ID of the parent task when a new child task is
created. This support enables some additional features (such as
SIGCHLD) and modifies the behavior of other interfaces. For
example, it makes waitpid() more standards complete by restricting
the waited-for tasks to the children of the caller. Default:
disabled.
config SCHED_CHILD_STATUS
bool "Retain child exit status"
default n
depends on SCHED_HAVE_PARENT
---help---
If this option is selected, then the exit status of the child task
will be retained after the child task exits. This option should be
selected if you require knowledge of a child process's exit status.
Without this setting, wait(), waitpid() or waitid() may fail. For
example, if you do:
1) Start child task
2) Wait for exit status (using wait(), waitpid(), or waitid()).
This can fail because the child task may run to completion before
the wait begins. There is a non-standard work-around in this case:
The above sequence will work if you disable pre-emption using
sched_lock() prior to starting the child task, then re-enable pre-
emption with sched_unlock() after the wait completes. This works
because the child task is not permitted to run until the wait is in
place.
The standard solution would be to enable SCHED_CHILD_STATUS. In
this case the exit status of the child task is retained after the
child exits and the wait will successful obtain the child task's
exit status whether it is called before the child task exits or not.
Warning: If you enable this feature, then your application must
either (1) take responsibility for reaping the child status with wait(),
waitpid(), or waitid(), or (2) suppress retention of child status.
If you do not reap the child status, then you have a memory leak and
your system will eventually fail.
Retention of child status can be suppressed on the parent using logic like:
struct sigaction sa;
sa.sa_handler = SIG_IGN;
sa.sa_flags = SA_NOCLDWAIT;
int ret = sigaction(SIGCHLD, &sa, NULL);
if SCHED_CHILD_STATUS
config PREALLOC_CHILDSTATUS
int "Number of pre-allocated child status"
default 0
---help---
To prevent runaway child status allocations and to improve
allocation performance, child task exit status structures are pre-
allocated when the system boots. This setting determines the number
of child status structures that will be pre-allocated.
However, the number of child status structures may need to be
significantly larger because this number includes the maximum number
of tasks that are running PLUS the number of tasks that have exit'ed
without having their exit status reaped (via wait(), waitid(), or
waitpid()).
Obviously, if tasks spawn children indefinitely and never have the
exit status reaped, then you may have a memory leak! If you enable
the SCHED_CHILD_STATUS feature, then your application must take
responsibility for either (1) reaping the child status with wait(),
waitpid(), or waitid() or it must (2) suppress retention of child
status. Otherwise, your system will eventually fail.
Retention of child status can be suppressed on the parent using logic like:
struct sigaction sa;
sa.sa_handler = SIG_IGN;
sa.sa_flags = SA_NOCLDWAIT;
int ret = sigaction(SIGCHLD, &sa, NULL);
config DEBUG_CHILDSTATUS
bool "Enable Child Status Debug Output"
default n
depends on SCHED_CHILD_STATUS && DEBUG_FEATURES
---help---
Very detailed... I am sure that you do not want this.
endif # SCHED_CHILD_STATUS
config SCHED_WAITPID
bool "Enable waitpid() API"
default n
depends on SCHED_HAVE_PARENT || !BUILD_KERNEL
---help---
Enables the waitpid() interface in a default, non-standard mode
(non-standard in the sense that the waited for PID need not be child
of the caller). If SCHED_HAVE_PARENT is also defined, then this
setting will modify the behavior or waitpid() (making more spec
compliant) and will enable the waitid() and wait() interfaces as
well. Note that SCHED_HAVE_PARENT must be defined in BUILD_KERNEL if
SCHED_WAITPID is needed.
config SCHED_DUMP_LEAK
bool "Enable catch task memory leak"
default n
---help---
When this option is enabled, the task's outstanding memory allocations
are printed using syslog. This helps catch any memory allocated by the
task that remains unreleased when the task exits.
config SCHED_USER_IDENTITY
bool "Support per-task User Identity"
default n
---help---
This selection enables functionality of getuid(), setuid(), getgid(),
setgid(). If this option is not selected, then stub, root-only
versions of these interfaces are available. When selected, these
interfaces will associate a UID and/or GID with each task group.
Those can then be managed using the interfaces. Child tasks will
inherit the UID and GID of its parent.
config SCHED_THREAD_LOCAL
bool "Support __thread/thread_local keyword"
default n
depends on ARCH_HAVE_THREAD_LOCAL
---help---
This option enables architecture-specific TLS support (__thread/thread_local keyword)
Note: Toolchain must be compiled with '--enable-tls' enabled
endmenu # Tasks and Scheduling
menu "Pthread Options"
depends on !DISABLE_PTHREAD
config PTHREAD_MUTEX_TYPES
bool "Enable mutex types"
default n
---help---
Set to enable support for recursive and errorcheck mutexes. Enables
pthread_mutexattr_settype().
choice
prompt "pthread mutex robustness"
default PTHREAD_MUTEX_ROBUST if !DEFAULT_SMALL
default PTHREAD_MUTEX_UNSAFE if DEFAULT_SMALL
config PTHREAD_MUTEX_ROBUST
bool "Robust mutexes"
---help---
Support only the robust form of the NORMAL mutex.
config PTHREAD_MUTEX_UNSAFE
bool "Traditional unsafe mutexes"
---help---
Support only the traditional non-robust form of the NORMAL mutex.
You should select this option only for backward compatibility with
software you may be porting or, perhaps, if you are trying to minimize
footprint.
config PTHREAD_MUTEX_BOTH
bool "Both robust and unsafe mutexes"
---help---
Support both forms of NORMAL mutexes.
endchoice # pthread mutex robustness
choice
prompt "Default NORMAL mutex robustness"
default PTHREAD_MUTEX_DEFAULT_ROBUST
depends on PTHREAD_MUTEX_BOTH
config PTHREAD_MUTEX_DEFAULT_ROBUST
bool "Robust default"
---help---
The default is robust NORMAL mutexes (non-standard)
config PTHREAD_MUTEX_DEFAULT_UNSAFE
bool "Unsafe default"
---help---
The default is traditional unsafe NORMAL mutexes (standard)
endchoice # Default NORMAL mutex robustness
choice
prompt "Default pthread mutex protocol"
default PTHREAD_MUTEX_DEFAULT_PRIO_NONE
config PTHREAD_MUTEX_DEFAULT_PRIO_NONE
bool "PTHREAD_PRIO_NONE default"
---help---
By default, pthread mutexes utilize PTHREAD_PRIO_NONE protocol (standard).
config PTHREAD_MUTEX_DEFAULT_PRIO_INHERIT
bool "PTHREAD_PRIO_INHERIT default"
depends on PRIORITY_INHERITANCE
---help---
By default, pthread mutexes utilize PTHREAD_PRIO_INHERIT protocol
(POSIX non-standard but a reasonable choice for most real-time systems).
endchoice # Default pthread mutex protocol
config PTHREAD_CLEANUP_STACKSIZE
int "pthread cleanup stack size"
default 0
range 0 255
---help---
The maximum number of cleanup actions that may be pushed by
pthread_cleanup_push().
if 0 disable the interfaces pthread_cleanup_push() and pthread_cleanup_pop().
This setting will increase the size of EVERY
pthread task control block by about n * CONFIG_PTHREAD_CLEANUP_STACKSIZE
where n is the size of a pointer, 2 * sizeof(uintptr_t), this would be
8 for a CPU with 32-bit addressing and 4 for a CPU with 16-bit
addressing.
config CANCELLATION_POINTS
bool "Cancellation points"
default n
---help---
Enable POSIX cancellation points for pthread_cancel(). If selected,
cancellation points will also used with the task_delete() API even if
pthreads are not enabled.
endmenu # Pthread Options
menu "Performance Monitoring"
config SCHED_SUSPENDSCHEDULER
bool
default n
config SCHED_RESUMESCHEDULER
bool
default n
config SCHED_IRQMONITOR
bool "Enable IRQ monitoring"
default n
depends on FS_PROCFS
---help---
Enabling counting of interrupts from all interrupt sources. These
counts will be available in the mounted procfs file systems at the
top-level file, "irqs".
config SCHED_CRITMONITOR
bool "Enable Critical Section monitoring"
default n
depends on FS_PROCFS
select SCHED_SUSPENDSCHEDULER
select SCHED_RESUMESCHEDULER
select IRQCOUNT
---help---
Enables logic that monitors the duration of time that a thread keeps
interrupts or pre-emption disabled. These global locks can have
negative consequences to real time performance: Disabling interrupts
adds jitter in the time when an interrupt request is asserted until
the hardware can respond with the interrupt. Disabling pre-emption
adds jitter in the time from when the event is posted in the
interrupt handler until the task that responds to the event can run.
if SCHED_CRITMONITOR
config SCHED_CRITMONITOR_MAXTIME_THREAD
int "THREAD max execution time"
default 0
---help---
Thread execution time should be smaller than
SCHED_CRITMONITOR_MAXTIME_THREAD, or system will give a warning.
For debugging system latency, 0 means disabled.
config SCHED_CRITMONITOR_MAXTIME_WQUEUE
int "WORK queue max execution time"
default SCHED_CRITMONITOR_MAXTIME_THREAD
---help---
Worker execution time should be smaller than
SCHED_CRITMONITOR_MAXTIME_WQUEUE, or system will give a warning.
For debugging system latency, 0 means disabled.
config SCHED_CRITMONITOR_MAXTIME_PREEMPTION
int "Pre-emption (sched_lock) max holding time"
default SCHED_CRITMONITOR_MAXTIME_WQUEUE
---help---
Pre-emption holding time should be smaller than
SCHED_CRITMONITOR_MAXTIME_PREEMPTION, or system will give a warning.
For debugging system latency, 0 means disabled.
config SCHED_CRITMONITOR_MAXTIME_CSECTION
int "Csection (enter_critical_section) max holding time"
default SCHED_CRITMONITOR_MAXTIME_PREEMPTION
---help---
Csection holding time should be smaller than
SCHED_CRITMONITOR_MAXTIME_CSECTION, or system will give a warning.
For debugging system latency, 0 means disabled.
config SCHED_CRITMONITOR_MAXTIME_IRQ
int "IRQ max execution time"
default SCHED_CRITMONITOR_MAXTIME_CSECTION
---help---
IRQ handler execution time should be smaller than
SCHED_CRITMONITOR_MAXTIME_IRQ, or system will give a warning.
For debugging system latency, 0 means disabled.
config SCHED_CRITMONITOR_MAXTIME_WDOG
int "WDOG callback max execution time"
default SCHED_CRITMONITOR_MAXTIME_IRQ
---help---
Wdog callback execution time should be smaller than
SCHED_CRITMONITOR_MAXTIME_WDOG, or system will give a warning.
For debugging system latency, 0 means disabled.
endif # SCHED_CRITMONITOR
config SCHED_CRITMONITOR_MAXTIME_PANIC
bool "Monitor timeout panic"
depends on \
SCHED_CRITMONITOR_MAXTIME_THREAD > 0 || \
SCHED_CRITMONITOR_MAXTIME_WDOG > 0 || \
SCHED_CRITMONITOR_MAXTIME_WQUEUE > 0 || \
SCHED_CRITMONITOR_MAXTIME_PREEMPTION > 0 || \
SCHED_CRITMONITOR_MAXTIME_CSECTION > 0 || \
SCHED_CRITMONITOR_MAXTIME_IRQ > 0
default n
---help---
If this option is enabled, a panic will be triggered when
IRQ/WQUEUE/PREEMPTION execution time exceeds SCHED_CRITMONITOR_MAXTIME_xxx
choice
prompt "Select CPU load clock source"
default SCHED_CPULOAD_NONE
---help---
If this option is selected, the timer interrupt handler will monitor
if the system is IDLE or busy at the time of that the timer interrupt
occurs. This is a very coarse measurement, but over a period of time,
it can very accurately determine the percentage of the time that the
CPU is IDLE.
The statistics collected in this could be used, for example, in the
PROCFS file system to provide CPU load measurements when read.
Note that in tickless mode of operation (SCHED_TICKLESS) there is
no system timer interrupt and CPU load measurements will not be
possible unless you provide an alternative clock to drive the
sampling and select SCHED_CPULOAD_EXTCLK.
config SCHED_CPULOAD_NONE
bool "None CPU load clock source"
---help---
If this option is enabled, the system will not support CPU load
measurement.
config SCHED_CPULOAD_SYSCLK
bool "Use system clock"
depends on !SCHED_TICKLESS
---help---
If this option is enabled, the system clock is used for cpu load
measurement by default.
There is a serious issue for the accuracy of measurements if the
system clock is used, however. NuttX threads are often started at
the time of the system timer expiration. Others may be stopped at
the time of the system timer expiration (if round-robin time-slicing
is enabled). Such thread behavior occurs synchronously with the
system timer and, hence, is not randomly sampled. As a consequence,
the CPU load attributed to these threads that run synchronously with
they system timer may be grossly in error.
The CPU load measurements are determined by sampling the active
tasks periodically at the occurrence to a timer expiration.
If tickless is enabled, SYSCLK should not be used. Its error will be
very large, and using it for analysis will lead to wrong conclusions.
config SCHED_CPULOAD_EXTCLK
bool "Use external clock"
---help---
There is a serious issue for the accuracy of measurements if the
system clock is used, however. NuttX threads are often started at
the time of the system timer expiration. Others may be stopped at
the time of the system timer expiration (if round-robin time-slicing
is enabled). Such thread behavior occurs synchronously with the
system timer and, hence, is not randomly sampled. As a consequence,
the CPU load attributed to these threads that run synchronously with
they system timer may be grossly in error.
The solution is to use some other clock that runs at a different
rate and has timer expirations that are asynchronous with the
system timer. Then truly accurate load measurements can be
achieved. This option enables use of such an "external" clock. The
implementation of the clock must be provided by platform-specific
logic; that platform-specific logic must call the system function
nxsched_process_cpuload_ticks() at each timer expiration with interrupts
disabled.
config SCHED_CPULOAD_CRITMONITOR
bool "Use critical monitor"
depends on SCHED_CRITMONITOR
---help---
Use the perfcounter in the core of the chip as a counter, no need to
use an external timer. Need to depend on SCHED_CRITMONITOR.
When the task is suspended, call nxsched_critmon_cpuload_ticks to count
the recent running time of the task
endchoice
if SCHED_CPULOAD_EXTCLK
config SCHED_CPULOAD_TICKSPERSEC
int "External clock rate"
default 100
---help---
If an external clock is used to drive the sampling for the CPU load
calculations, then this value must be provided. This value provides
the rate of the external clock interrupts in units of ticks per
second. The default value of 100 corresponds to a 100Hz clock. NOTE:
that 100Hz is the default frequency of the system time and, hence,
the worst possible choice in most cases.
choice
prompt "Select CPU load timer"
default CPULOAD_ONESHOT
config CPULOAD_ONESHOT
bool "Use Oneshot timer"
---help---
Use an MCU-specific oneshot timer as the external clock. The
oneshot timer must be configured by board specific logic which must
then call:
void nxsched_oneshot_extclk(FAR struct oneshot_lowerhalf_s *lower);
To start the CPU load measurement. See include/nuttx/clock.h
NOTE that in this configuration, CONFIG_SCHED_CPULOAD_TICKSPERSEC is
the sample rate that will be accomplished by programming the oneshot
time repeatedly. If CPULOAD_ONESHOT_ENTROPY is also selected, then
the underly frequency driving the oneshot timer must be
significantly faster than CONFIG_SCHED_CPULOAD_TICKSPERSE to permit
precise modulation the sample periods.
config CPULOAD_PERIOD
bool "Use Period timer"
---help---
Use an MCU-specific period timer as the external clock. The
period timer must be configured by board specific logic which must
then call:
void nxsched_period_extclk(FAR struct timer_lowerhalf_s *lower);
To start the CPU load measurement. See include/nuttx/clock.h
NOTE that in this configuration, CONFIG_SCHED_CPULOAD_TICKSPERSEC is
the sample rate that will be accomplished by programming the period
time.
endchoice
config CPULOAD_ENTROPY
int "Bits of entropy"
default 6
range 0 30
---help---
This is the number of bits of entropy that will be applied. The
oneshot will be set to this interval:
CPULOAD_ONESHOT_NOMINAL - (CPULOAD_ONESHOT_ENTROPY / 2) +
error + nrand(CPULOAD_ONESHOT_ENTROPY)
Where
CPULOAD_ONESHOT_NOMINAL is the nominal sample interval implied
by CONFIG_SCHED_CPULOAD_TICKSPERSEC in units of microseconds.
CPULOAD_ONESHOT_ENTROPY is (1 << CONFIG_CPULOAD_ENTROPY),
and 'error' is an error value that is retained from interval to
interval so that although individual intervals are randomized,
the average will still be CONFIG_SCHED_CPULOAD_TICKSPERSEC.
This special value of zero disables entropy.
endif # SCHED_CPULOAD_EXTCLK
config SCHED_CPULOAD_TIMECONSTANT
int "CPU load time constant"
depends on !SCHED_CPULOAD_NONE
default 2
---help---
The accumulated CPU count is divided by two when the accumulated
tick count exceeds this time constant. This time constant is in
units of seconds.
menuconfig SCHED_INSTRUMENTATION
bool "System performance monitor hooks"
default n
select SCHED_SUSPENDSCHEDULER
select SCHED_RESUMESCHEDULER
---help---
Enables instrumentation in scheduler to monitor system performance.
If enabled, then the board-specific logic must provide the following
functions (see include/sched.h):
void sched_note_start(FAR struct tcb_s *tcb);
void sched_note_stop(FAR struct tcb_s *tcb);
If CONFIG_SMP is enabled, then these additional interfaces are
expected:
void sched_note_cpu_start(FAR struct tcb_s *tcb, int cpu);
void sched_note_cpu_started(FAR struct tcb_s *tcb);
if SCHED_INSTRUMENTATION
config SCHED_INSTRUMENTATION_CPUSET
hex "CPU bit set"
default 0xffff
depends on SMP && SCHED_INSTRUMENTATION_FILTER
---help---
Monitor only CPUs in the bitset. Bit 0=CPU0, Bit1=CPU1, etc.
config SCHED_INSTRUMENTATION_FILTER
bool "Instrumentation filter"
default n
---help---
Enables the filter logic for the instrumentation. If this option
is enabled, the instrumentation data passed to sched_note_add()
can be filtered by syscall and IRQ number.
The filter logic can be configured by sched_note_filter APIs defined in
include/nuttx/sched_note.h.
config SCHED_INSTRUMENTATION_FILTER_DEFAULT_MODE
hex "Default instrumentation filter mode"
depends on SCHED_INSTRUMENTATION_FILTER
default 0x3f
---help---
Default mode of the instrumentation filter logic.
Bit 0 = Enable instrumentation
Bit 1 = Enable switch instrumentation
Bit 2 = Enable syscall instrumentation
Bit 3 = Enable IRQ instrumentation
Bit 4 = Enable dump instrumentation
Bit 5 = Enable collecting syscall arguments
config SCHED_INSTRUMENTATION_SWITCH
bool "Use note switch for instrumentation"
default n
---help---
Use note switch for instrumentation.
void sched_note_suspend(FAR struct tcb_s *tcb);
void sched_note_resume(FAR struct tcb_s *tcb);
If CONFIG_SMP is enabled, then these additional interfaces are
expected:
void sched_note_cpu_pause(FAR struct tcb_s *tcb, int cpu);
void sched_note_cpu_paused(FAR struct tcb_s *tcb);
void sched_note_cpu_resume(FAR struct tcb_s *tcb, int cpu);
void sched_note_cpu_resumed(FAR struct tcb_s *tcb);
NOTE: These are internal OS interfaces and are called at very
critical locations in the OS. There is very little that can be
done in these interfaces. For example, normal devices may not be
used; syslog output cannot be performed.
config SCHED_INSTRUMENTATION_PREEMPTION
bool "Preemption monitor hooks"
default n
---help---
Enables additional hooks for changes to pre-emption state. Board-
specific logic must provide this additional logic.
void sched_note_premption(FAR struct tcb_s *tcb, bool state);
config SCHED_INSTRUMENTATION_CSECTION
bool "Critical section monitor hooks"
default n
select IRQCOUNT
---help---
Enables additional hooks for entry and exit from critical sections.
Interrupts are disabled while within a critical section. Board-
specific logic must provide this additional logic.
void sched_note_csection(FAR struct tcb_s *tcb, bool state);
config SCHED_INSTRUMENTATION_SPINLOCKS
bool "Spinlock monitor hooks"
default n
---help---
Enables additional hooks for spinlock state. Board-specific logic
must provide this additional logic.
void sched_note_spinlock(FAR struct tcb_s *tcb, FAR volatile spinlock_t *spinlock, int type)
config SCHED_INSTRUMENTATION_SYSCALL
bool "System call monitor hooks"
default n
depends on ARCH_HAVE_SYSCALL_HOOKS
---help---
Enables additional hooks for entry and exit from system call.
Board-specific logic must provide this additional logic.
void sched_note_syscall_enter(int nr, int argc, ...);
void sched_note_syscall_leave(int nr, uintptr_t result);
config SCHED_INSTRUMENTATION_IRQHANDLER
bool "Interrupt handler monitor hooks"
default n
---help---
Enables additional hooks for interrupt handler. Board-specific logic
must provide this additional logic.
void sched_note_irqhandler(int irq, FAR void *handler, bool enter);
config SCHED_INSTRUMENTATION_DUMP
bool "Use note dump for instrumentation"
default n
---help---
Use note dump for instrumentation.
void sched_note_string(FAR const char *buf);
void sched_note_dump(uint32_t module, uint8_t event, FAR const void *buf, size_t len);
void sched_note_vprintf(FAR const char *fmt, va_list va);
void sched_note_vbprintf(uint32_t module, uint8_t event, FAR const char *fmt, va_list va);
void sched_note_printf(FAR const char *fmt, ...) printf_like(1, 2);
void sched_note_bprintf(uint32_t module, uint8_t event, FAR const char *fmt, ...);
config SCHED_INSTRUMENTATION_FUNCTION
bool "Enable function auto-tracing"
default n
---help---
After enabling this option, you can automatically trace the function instrumentation without adding tracepoint manually.
This is similar to the Function Trace effect of the linux kernel
Add CFLAGS += -finstrument-functions to the makefile to track the required modules.
The following compilation option can exclude files that do not want to be tracked in this module
CFLAGS += -finstrument-functions-exclude-file-list=xxx
The following compilation option can exclude functions that do not want to be tracked in this module
CFLAGS += -finstrument-functions-exclude-function-list=xxx
For a more detailed description of compilation options,
refer to the "Program Instrumentation Options" chapter in the gcc documentation
endif # SCHED_INSTRUMENTATION
endmenu # Performance Monitoring
menu "System Auto Instrumentation"
config SCHED_STACK_RECORD
int "Maximum stack backtrace to record"
default 0
---help---
Specifies the maximum number of stack backtrace to record in the
TCB. Useful if scheduler instrumentation is selected. Set to zero
to disable.Through instrumentation, record the backtrace at
the deepest point in the stack.
endmenu
menu "Files and I/O"
config DEV_CONSOLE
bool "Enable /dev/console"
default y
---help---
Set if architecture-specific logic provides /dev/console at boot-up
time. Enables stdout, stderr, stdin in the start-up application.
You need this setting if your console device is ready at boot time.
For example, if you are using a serial console, then /dev/console
(aka, /dev/ttyS0) will be available when the application first starts.
You must not select DEV_CONSOLE if you console device comes up later
and is not ready until after the application starts. At this time,
the only console device that behaves this way is a USB serial console.
When the application first starts, the USB is (probably) not yet
connected and /dev/console will not be created until later when the
host connects to the USB console.
config FDCLONE_DISABLE
bool "Disable cloning of file descriptors"
default n
---help---
Disable cloning of all file descriptors by task_create() when a new
ask is started. If set, all files/drivers will appear to be closed
in the new task.
config FDCLONE_STDIO
bool "Disable clone file descriptors without stdio"
default n
---help---
Disable cloning of all but the first three file descriptors (stdin,
stdout, stderr) by task_create() when a new task is started. If set,
all files/drivers will appear to be closed in the new task except
for stdin, stdout, and stderr.
config NFILE_DESCRIPTORS_PER_BLOCK
int "The number of file descriptors per block"
default 8
range 3 99999
---help---
The number of file descriptors per block(one for each open)
config FILE_STREAM
bool "Enable FILE stream"
default !DEFAULT_SMALL
---help---
Enable the standard buffered input/output support
config NAME_MAX
int "Maximum size of a file name"
default 32
---help---
The maximum size of a file name.
config PATH_MAX
int "Maximum size of path name"
default 256
---help---
The maximum size of path name.
endmenu # Files and I/O
menuconfig PRIORITY_INHERITANCE
bool "Enable priority inheritance"
default n
---help---
Set to enable support for priority inheritance on mutexes and semaphores.
When this option is enabled, the initial configuration of all seamphores
and mutexes will be with priority inheritance enabled. That configuration
may not be appropriate in all cases (such as when the semaphore or mutex
is used for signaling). In such cases, priority inheritance can be
disabled for individual semaphores by calling:
int ret = sem_setprotocol(&sem, SEM_PRIO_NONE);
From applications, the functionally equivalent OS internal interface,
nxsem_set_protocol(), should be used within the OS
And for individual pthread mutexes by setting the protocol attribute
before initializing the mutex:
int ret = pthread_mutexattr_setprotocol(&attr, PTHREAD_PRIO_NONE);
if PRIORITY_INHERITANCE
config SEM_PREALLOCHOLDERS
int "Number of pre-allocated holders"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
---help---
This setting is only used if priority inheritance is enabled.
It defines the maximum number of different threads (minus one) that
can take counts on a semaphore with priority inheritance support.
This may be set to zero if priority inheritance is disabled OR if you
are only using semaphores as mutexes (only one holder) OR if no more
than two threads participate using a counting semaphore.
endif # PRIORITY_INHERITANCE
menu "RTOS hooks"
config BOARD_EARLY_INITIALIZE
bool "Custom board early initialization"
default n
---help---
There are three points in time where you can insert custom,
board-specific initialization logic:
1) <arch>_board_initialize(): This function is used only for
initialization of very low-level things like configuration of
GPIO pins, power setting. The OS has not been initialized
at this point, so you cannot allocate memory or initialize
device drivers at this phase.
2) The next level of initialization is performed by a call to
up_initialize() (in arch/<arch>/src/common/up_initialize.c).
The OS has been initialized at this point and it is okay to
initialize drivers in this phase.
At this same point in time, the OS will also call a board-
specific initialization function named board_early_initialize()
if CONFIG_BOARD_EARLY_INITIALIZE is selected. The context in
which board_early_initialize() executes is suitable for early
initialization of most, simple device drivers and is a logical,
board-specific extension of up_initialize().
board_early_initialize() runs on the startup, initialization thread.
Some initialization operations cannot be performed on the start-up,
initialization thread. That is because the initialization thread
cannot wait for event. Waiting may be required, for example, to
mount a file system or or initialize a device such as an SD card.
For this reason, such driver initialize must be deferred to
board_late_initialize().
3) And, finally, just before the user application code starts.
If CONFIG_BOARD_LATE_INITIALIZE is selected, then an additional
initialization call will be performed in the boot-up sequence to a
function called board_late_initialize(). board_late_initialize()
will be called after up_initialize() is called and just before the
main application is started. This additional initialization
phase may be used, for example, to initialize more complex,
board-specific device drivers.
Waiting for events, use of I2C, SPI, etc are permissible in the
context of board_late_initialize(). That is because
board_late_initialize() will run on a temporary, internal kernel
thread.
config BOARD_LATE_INITIALIZE
bool "Custom board late initialization"
default n
---help---
There are three points in time where you can insert custom,
board-specific initialization logic:
1) <arch>_board_initialize(): This function is used only for
initialization of very low-level things like configuration of
GPIO pins, power setting. The OS has not been initialized
at this point, so you cannot allocate memory or initialize
device drivers at this phase.
2) The next level of initialization is performed by a call to
up_initialize() (in arch/<arch>/src/common/up_initialize.c).
The OS has been initialized at this point and it is okay to
initialize drivers in this phase.
At this same point in time, the OS will also call a board-
specific initialization function named board_early_initialize()
if CONFIG_BOARD_EARLY_INITIALIZE is selected. The context in
which board_early_initialize() executes is suitable for early
initialization of most, simple device drivers and is a logical,
board-specific extension of up_initialize().
board_early_initialize() runs on the startup, initialization thread.
Some initialization operations cannot be performed on the start-up,
initialization thread. That is because the initialization thread
cannot wait for event. Waiting may be required, for example, to
mount a file system or or initialize a device such as an SD card.
For this reason, such driver initialize must be deferred to
board_late_initialize().
3) And, finally, just before the user application code starts.
If CONFIG_BOARD_LATE_INITIALIZE is selected, then an additional
initialization call will be performed in the boot-up sequence to a
function called board_late_initialize(). board_late_initialize()
will be called after up_initialize() is called and just before the
main application is started. This additional initialization
phase may be used, for example, to initialize more complex,
board-specific device drivers.
Waiting for events, use of I2C, SPI, etc are permissible in the
context of board_late_initialize(). That is because
board_late_initialize() will run on a temporary, internal kernel
thread.
if BOARD_LATE_INITIALIZE
config BOARD_INITTHREAD_STACKSIZE
int "Board initialization thread stack size"
default DEFAULT_TASK_STACKSIZE
---help---
The size of the stack to allocate when starting the board
initialization thread.
config BOARD_INITTHREAD_PRIORITY
int "Board initialization thread priority"
default 240
---help---
The priority of the board initialization thread. This priority is
not a critical setting. No other application threads will be
started until the board initialization is completed. Hence, there
is very little competition for the CPU.
endif # BOARD_LATE_INITIALIZE
config SCHED_STARTHOOK
bool "Enable startup hook"
default n
---help---
Enable a non-standard, internal OS API call nxtask_starthook().
nxtask_starthook() registers a function that will be called on task
startup before that actual task entry point is called. The
starthook is useful, for example, for setting up automatic
configuration of C++ constructors.
endmenu # RTOS hooks
menu "Signal Configuration"
config SIG_PREALLOC_ACTIONS
int "Number of pre-allocated sigactions"
default 4
---help---
The number of pre-allocated sigaction structures.
config SIG_ALLOC_ACTIONS
int "Num of sigactions to allocate per time"
default 1
---help---
The number of sigactions to allocate per time. Note that
if this number is larger than 1, the allocation won't be
returned to the heap but kept in a free list for reuse.
config SIG_PREALLOC_IRQ_ACTIONS
int "Number of pre-allocated irq actions"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
---help---
The number of pre-allocated irq action structures.
config SIG_EVTHREAD
bool "Support SIGEV_THREAD"
default n
depends on !BUILD_KERNEL && SCHED_WORKQUEUE
select LIBC_USRWORK if BUILD_PROTECTED
---help---
Built in support for the SIGEV_THREAD signal deliver method.
NOTE: The current implementation uses a work queue to notify the
client. This, however, would only work in the FLAT build. A
different mechanism would need to be development to support this
feature on the PROTECTED or KERNEL build.
config SIG_EVTHREAD_HPWORK
bool "SIGEV_EVTHREAD use HPWORK"
default n
depends on SIG_EVTHREAD && SCHED_HPWORK
---help---
if selected, SIGEV_THREAD will use the high priority work queue.
If not, it will use the low priority work queue (if available).
REVISIT: This solution is non-optimal. Some notifications should
be high priority and others should be lower priority. Ideally, you
should be able to determine which work queue is used on a
notification-by-notification basis.
menuconfig SIG_DEFAULT
bool "Default signal actions"
default n
---help---
Enable to support default signal actions.
if SIG_DEFAULT
comment "Per-signal Default Actions"
config SIG_SIGKILL_ACTION
bool "Enable all SIGKILL signals"
default y
---help---
Enable the default action for SIGHUP SIGILL SIGTRAP SIGABRT SIGBUS
SIGFPE SIGINT SIGKILL SIGSEGV SIGQUIT SIGTERM SIGXCPU SIGXFSZ and
SIGSYS (terminate the task).
config SIG_SIGUSR1_ACTION
bool "SIGUSR1"
default n
---help---
Enable the default action for SIGUSR1 (terminate the task)
Make sure that your applications are expecting this POSIX behavior.
Backward compatible behavior would require that the application use
sigaction() to ignore SIGUSR1.
config SIG_SIGUSR2_ACTION
bool "SIGUSR2"
default n
---help---
Enable the default action for SIGUSR2 (terminate the task)
Make sure that your applications are expecting this POSIX behavior.
Backward compatible behavior would require that the application use
sigaction() to ignore SIGUSR2.
config SIG_SIGPIPE_ACTION
bool "SIGPIPE"
default n
---help---
Enable the default action for SIGPIPE (terminate the task).
config SIG_SIGALRM_ACTION
bool "SIGALRM SIGVTALRM"
default n
---help---
Enable the default action for SIGALRM AND SIGVTALRM(terminate the task)
Make sure that your applications are expecting this POSIX behavior.
Backward compatible behavior would require that the application use
sigaction() to ignore SIGALRM.
config SIG_SIGSTOP_ACTION
bool "SIGSTOP SIGTSTP SIGCONT SIGTTIN SIGTTOU"
default y
---help---
Enable the default action for SIGSTOP SIGTSTP SIGCONT SIGTTIN SIGTTOU
(suspend the task) and SIGCONT (resume the task).
config SIG_SIGPROF_ACTION
bool "SIGPROF"
default n
---help---
Enable the default action for SIGPROF (nxsig_abnormal_termination)
Make sure that your applications are expecting this POSIX behavior.
Backward compatible behavior would require that the application use
sigaction() to ignore SIGPROF.
config SIG_SIGPOLL_ACTION
bool "SIGPOLL"
default n
depends on FS_AIO
---help---
Enable the default action for SIGPOLL (terminate the task)
Make sure that your applications are expecting this POSIX behavior.
Backward compatible behavior would require that the application use
sigaction() to ignore SIGPOLL.
endif # SIG_DEFAULT
endmenu # Signal Configuration
menu "Message Queue Options"
depends on !DISABLE_MQUEUE || !DISABLE_MQUEUE_SYSV
config PREALLOC_MQ_MSGS
int "Number of pre-allocated messages"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
---help---
The number of pre-allocated message structures. The system manages
a pool of preallocated message structures to minimize dynamic allocations
config PREALLOC_MQ_IRQ_MSGS
int "Number of pre-allocated irq messages"
default 4 if DEFAULT_SMALL
default 8 if !DEFAULT_SMALL
---help---
The number of pre-allocated irq message structures.
config MQ_MAXMSGSIZE
int "Maximum message size"
default 32
---help---
Message structures are allocated with a fixed payload size given by this
setting (does not include other message structure overhead.
config DISABLE_MQUEUE_NOTIFICATION
bool "Disable POSIX message queue notification"
default DEFAULT_SMALL
---help---
Disable POSIX message queue notification
endmenu # POSIX Message Queue Options
config MODULE
bool "Enable loadable OS modules"
default n
select LIBC_MODLIB
select ARCH_USE_TEXT_HEAP if ARCH_HAVE_TEXT_HEAP
---help---
Enable support for loadable OS modules. Default: n
menu "Work queue support"
config SCHED_WORKQUEUE
# bool "Enable worker thread"
bool
default n
---help---
Create dedicated "worker" threads to handle delayed or asynchronous
processing.
config WQUEUE_NOTIFIER
bool "Generic work notifier"
default n
depends on SCHED_WORKQUEUE
---help---
Enable building of work queue notifier logic that will execute a
worker function an event occurs. This is is a general purpose
notifier, but was developed specifically to support poll() logic
where the poll must wait for an resources to become available.
config SCHED_HPWORK
bool "High priority (kernel) worker thread"
default n
select SCHED_WORKQUEUE
---help---
Create a dedicated high-priority "worker" thread to handle delayed
processing from interrupt handlers. This feature is required for
some drivers but, if there are no complaints, can be safely
disabled. The high priority worker thread also performs garbage
collection -- completing any delayed memory deallocations from
interrupt handlers. If the high-priority worker thread is disabled,
then that clean up will be performed either by (1) the low-priority
worker thread, if enabled, and if not (2) the IDLE thread instead
(which runs at the lowest of priority and may not be appropriate if
memory reclamation is of high priority)
For other, less-critical asynchronous or delayed process, the
low-priority worker thread is recommended.
if SCHED_HPWORK
config SCHED_HPNTHREADS
int "Number of high-priority worker threads"
default 1
---help---
This options selects multiple, high-priority threads. This is
essentially a "thread pool" that provides multi-threaded servicing
of the high-priority work queue. This breaks the serialization
of the "queue" (hence, it is no longer a queue at all).
CAUTION: Some drivers may use the work queue to serialize
operations. They may also use the high-priority work queue if it is
available. If there are multiple high-priority worker threads, then
this can result in the loss of that serialization. There may be
concurrent driver operations running on different HP threads and
this could lead to a failure. You may need to visit the use of the
HP work queue on your configuration is you select
CONFIG_SCHED_HPNTHREADS > 1
config SCHED_HPWORKPRIORITY
int "High priority worker thread priority"
default 224
---help---
The execution priority of the higher priority worker thread.
The higher priority worker thread is intended to serve as the
"bottom" half for device drivers. As a consequence it must run at
a very high, fixed priority. Typically, it should be the highest
priority thread in your system. Default: 224
For lower priority, application oriented worker thread support,
please consider enabling the lower priority work queue. The lower
priority work queue runs at a lower priority, of course, but has
the added advantage that it supports "priority inheritance" (if
PRIORITY_INHERITANCE is also selected): The priority of the lower
priority worker thread can then be adjusted to match the highest
priority client.
config SCHED_HPWORKSTACKSIZE
int "High priority worker thread stack size"
default DEFAULT_TASK_STACKSIZE
---help---
The stack size allocated for the worker thread. Default: 2K.
endif # SCHED_HPWORK
config SCHED_LPWORK
bool "Low priority (kernel) worker thread"
default n
select SCHED_WORKQUEUE
---help---
If SCHED_LPWORK is defined then a lower-priority work queue will
be created. This lower priority work queue is better suited for
more extended, application oriented processing (such as file system
clean-up operations or asynchronous I/O)
if SCHED_LPWORK
config SCHED_LPNTHREADS
int "Number of low-priority worker threads"
default 1 if !FS_AIO
default 4 if FS_AIO
---help---
This options selects multiple, low-priority threads. This is
essentially a "thread pool" that provides multi-threaded servicing
of the low-priority work queue. This breaks the serialization
of the "queue" (hence, it is no longer a queue at all).
This options is required to support, for example, I/O operations
that stall waiting for input. If there is only a single thread,
then the entire low-priority queue processing stalls in such cases.
Such behavior is necessary to support asynchronous I/O, AIO (for
example).
CAUTION: Some drivers may use the work queue to serialize
operations. They may also use the low-priority work queue if it is
available. If there are multiple low-priority worker threads, then
this can result in the loss of that serialization. There may be
concurrent driver operations running on different LP threads and
this could lead to a failure. You may need to visit the use of the
LP work queue on your configuration is you select
CONFIG_SCHED_LPNTHREADS > 1
config SCHED_LPWORKPRIORITY
int "Low priority worker thread priority"
default 100
---help---
The minimum execution priority of the lower priority worker thread.
The lower priority worker thread is intended support application-
oriented functions. The lower priority work queue runs at a lower
priority, of course, but has the added advantage that it supports
"priority inheritance" (if PRIORITY_INHERITANCE is also selected):
The priority of the lower priority worker thread can then be
adjusted to match the highest priority client. Default: 100
NOTE: This priority inheritance feature is not automatic. The
lower priority worker thread will always a fixed priority unless
you implement logic that calls lpwork_boostpriority() to raise the
priority of the lower priority worker thread (typically called
before scheduling the work) and then call the matching
lpwork_restorepriority() when the work is completed (typically
called within the work handler at the completion of the work).
Currently, only the NuttX asynchronous I/O logic uses this dynamic
prioritization feature.
The higher priority worker thread, on the other hand, is intended
to serve as the "bottom" half for device drivers. As a consequence
it must run at a very high, fixed priority. Typically, it should
be the highest priority thread in your system.
config SCHED_LPWORKPRIOMAX
int "Low priority worker thread maximum priority"
default 176
depends on PRIORITY_INHERITANCE
---help---
The maximum execution priority of the lower priority worker thread.
The lower priority worker thread is intended support application-
oriented functions. The lower priority work queue runs at a lower
priority, of course, but has the added advantage that it supports
"priority inheritance" (if PRIORITY_INHERITANCE is also selected):
The priority of the lower priority worker thread can then be
adjusted to match the highest priority client.
The higher priority worker thread, on the other hand, is intended
to serve as the "bottom" half for device drivers. As a consequence
it must run at a very high, fixed priority. Typically, it should
be the highest priority thread in your system.
This value provides an upper limit on the priority of the lower
priority worker thread. This would be necessary, for example, if
the higher priority worker thread were to defer work to the lower
priority thread. Clearly, in such a case, you would want to limit
the maximum priority of the lower priority work thread. Default:
176
config SCHED_LPWORKSTACKSIZE
int "Low priority worker thread stack size"
default DEFAULT_TASK_STACKSIZE
---help---
The stack size allocated for the lower priority worker thread. Default: 2K.
endif # SCHED_LPWORK
endmenu # Work Queue Support
menu "Stack and heap information"
config DEFAULT_TASK_STACKSIZE
int "The default stack size for tasks"
default 2048
---help---
The default stack size for tasks.
config IDLETHREAD_STACKSIZE
int "Idle thread stack size"
default 1024
---help---
The size of the initial stack used by the IDLE thread. The IDLE thread
is the thread that (1) performs the initial boot of the system up to the
point where start-up application is spawned, and (2) there after is the
IDLE thread that executes only when there is no other thread ready to run.
config PTHREAD_STACK_MIN
int "Minimum pthread stack size"
default 256
---help---
Minimum pthread stack size
config PTHREAD_STACK_DEFAULT
int "Default pthread stack size"
default DEFAULT_TASK_STACKSIZE
---help---
Default pthread stack size
endmenu # Stack and heap information
config SCHED_BACKTRACE
bool "Stack BackTrace"
default "n"
---help---
This option enables stack backtrace support in the NuttX
using the information automatically generated by the
compiler or architecture specific approach when ARCH_HAVE_BACKTRACE
is selected
config GROUP_KILL_CHILDREN_TIMEOUT_MS
int "Group kill children timeout"
default -1
depends on !DISABLE_PTHREAD && SIG_SIGKILL_ACTION
---help---
Kill children a SIGQUIT signal before cancel them,
< 0 means wait until all the child thread exit
> 0 means wait timeout
= 0 means don't do kill signal
config PID_INITIAL_COUNT
int "Initial length of pid table"
default 8 if DEFAULT_SMALL
default 16 if !DEFAULT_SMALL
---help---
This is the initial length of pid table, which the system
can still expand when needed. It is rounded up to power of
two by current implementation. If the number of threads in
your system is known at design time, setting this to it.
config SCHED_EVENTS
bool "Schedule Event objects"
default n
help
This option enables event objects. Threads may wait on event
objects for specific events, but both threads and ISRs may deliver
events to event objects.