The load/compare and RMW to wait_count need protection. Using atomic
operations should resolve both issues.
NOTE:
The assumption that the user will call pthread_cond_signal /
pthread_cond_broadcast with the mutex given to pthread_cond_wait held is
simply not true. It MAY hold it, but it is not forced. Thus, using the
user space lock for protecting the wait counter as well is not valid!
The pthread_cond_signal() or pthread_cond_broadcast() functions may be called by a thread whether or not it currently owns the mutex that threads calling pthread_cond_wait() or pthread_cond_timedwait() have associated with the condition variable during their waits; however, if predictable scheduling behaviour is required, then that mutex is locked by the thread calling pthread_cond_signal() or pthread_cond_broadcast().
[1] https://pubs.opengroup.org/onlinepubs/7908799/xsh/pthread_cond_signal.html
just read the code, still have the deadlock after below change:
| commit 3b6f463ce2
| Author: chenrun1 <chenrun1@xiaomi.com>
| Date: Tue Dec 31 14:31:46 2024 +0800
|
| syslog_inbuffer:Fixed the lock reentry issue in single core + spinlock scenario
|
| Summary:
| When spinlock.h is enabled in a single-core scenario, spinlock reentry and exceptions may occur when spin_lock_irqsave_wo_note is used in syslog_add_intbuffer
|
| Signed-off-by: chenrun1 <chenrun1@xiaomi.com>
Signed-off-by: chao an <anchao.archer@bytedance.com>
Summary:
- Fixed a typo in the Rust guide ("Pleae" -> "Please")
- Improved list formatting for better readability and consistency
Impact:
- Improves documentation quality and readability
- Ensures users have complete and accurate configuration information
- Maintains consistency with the actual requirements for Rust development in NuttX
Signed-off-by: Huang Qi <huangqi3@xiaomi.com>
elf_emit_tcb_note: nitems(status.pr_regs) is 18, g_tcbinfo.regs_num is 17, then g_tcbinfo.reg_off.p[17] has been out of bounds
Signed-off-by: wanggang26 <wanggang26@xiaomi.com>
common/espressif/esp_pcnt.c: In function 'esp_pcnt_isr_default':
common/espressif/esp_pcnt.c:396:41: warning: passing argument 1 of 'spin_lock_irqsave' makes pointer from integer without a cast [-Wint-conversion]
396 | flags = spin_lock_irqsave(unit->lock);
| ~~~~^~~~~~
| |
| spinlock_t {aka unsigned char}
In file included from common/espressif/esp_pcnt.c:44:
/home/hujun5/下载/vela_sim/nuttx/include/nuttx/spinlock.h:507:55: note: expected 'volatile spinlock_t *' {aka 'volatile unsigned char *'} but argument is of type 'spinlock_t' {aka 'unsigned char'}
507 | irqstate_t spin_lock_irqsave(FAR volatile spinlock_t *lock)
| ~~~~~~~~~~~~~~~~~~~~~^~~~
Signed-off-by: hujun5 <hujun5@xiaomi.com>
This matches how `PICO_XOSC_STARTUP_DELAY_MULTIPLIER` is set in
pico-sdk.
This should fix issues on startup for some units of these boards. It
seems not all are affected. See
https://github.com/raspberrypi/pico-sdk/pull/457 and the issue linked
there for more information.
Signed-off-by: Ian Douglas Scott <ian@iandouglasscott.com>
This follows the formula in the "Startup Delay" section of the
datasheets, where `BOARD_XOSC_STARTUPDELAY` is the time in milliseconds,
which is then multiplied by the clock fequency in kHz, divided by 256
cycles, and rounded up.
This differs from `pico-sdk`, which just multiplied the value of the
expression it previously used by a `PICO_XOSC_STARTUP_DELAY_MULTIPLIER`.
This seems less correct, though it should be essentially the same except
rounding. (`BOARD_XOSC_STARTUPDELAY_MULTIPLIER` also wouldn't fit nicely
the way the `#define`s are aligned.)
This sets `BOARD_XOSC_STARTUPDELAY` to `1` for all boards, so this will
produce the same output and not change behavior.
Signed-off-by: Ian Douglas Scott <ian@iandouglasscott.com>
Previously this used a hard-coded value of 12 MHz, but this presumably
should use `BOARD_XOSC_FREQ`. Similarly to how it is done in pico-sdk.
It looks like all boards are currently defined to use 12 MHz XOSC
clocks, so this won't change behavior with any of them.
Signed-off-by: Ian Douglas Scott <ian@iandouglasscott.com>
align them with the usage habits of makefile
cmake:
cmake -B build -DBOARD_CONFIG=sim/nsh -GNinja -DEXTRAFLAGS="-DCONFIG_AAA=1 -DCONFIG_BBB=1 -Werror"
makefile:
make -j12 EXTRAFLAGS="-DCONFIG_AAA=1 -DCONFIG_BBB=1 -Werror"
Signed-off-by: chao an <anchao.archer@bytedance.com>
No reason to keep asking for these over and over. Also fix bumping
the CPU booted counter when SBI is enabled.
Must also read g_hart_use_sbi before bumping the CPU booted counter.
add support of set custom command options to config.h
Use the -D parameter to pass the config to header file.
NOTE which must begin with the CONFIG_ prefix.
eg:
cmake -DCONFIG_AAA=1 -DCONFIG_BBB=1 -B build -DBOARD_CONFIG=sim/nsh -GNinja
Signed-off-by: chao an <anchao.archer@bytedance.com>
This adds support for qemu-armv7r board with flat mode sample
config for using with QEMU emulated Cortex-R5 processor.
Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com>
This adds support for QEMU Cortex-R5 virtual processor on existing QEMU
Cortex-A7 code base with profile support in `armv7-r/` and `armv7-a/`.
Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com>
This revise armv7-r/ header files needed to support QEMU cortex-r5
virtual process for armv7-r family.
Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com>
As armv7-a addrenv layout is not in line with other archs, we need
align it with others by defining ARCH_DATA_RESERVE_SIZE as same as
MM_PGSIZE for now to keep gnu-elf.ld.in unified.
Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com>
This fixes compilation error:
```
/tmp/nuttx/arch/arm/src/armv7-r/gic.h:699:17: error:
'GIC_ICDSGIR_NSATT_GRP1' undeclared; did you mean 'GIC_ICDSGIR_NSATT'?
```
Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com>
