Save/Restore FPU registers in C environment is dangerous practive,
which cannot guarantee the compiler won't generate the assembly code
with float point registers, especially in interrupt handling
Signed-off-by: chao.an <anchao@xiaomi.com>
The SPSR is used to store the current value of the CPSR when an exception
is taken so that it can be restored after handling the exception.
Each exception handling mode can access its own SPSR.
User mode and System mode do not have an SPSR because they are not
exception handling modes.
Signed-off-by: chao.an <anchao@xiaomi.com>
The current context save implementation saves registers of each task
to xcp context, which is unnecessary because most of the arm registers are
already saved in the task stack, this commit replace the xcp context with
stack context to improve context switching performance and reduce the tcb
space occupation of tcb instance.
Signed-off-by: chao.an <anchao@xiaomi.com>
Summary:
- I noticed that pthread always crashes when started
if CONFIG_BUILD_KERNEL=y
- This commit fixes this issue
Impact:
- None
Testing:
- Tested with sabre-6quad:netknsh (not merged yet)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- I noticed that make export does not work with swama5d4-ek:knsh
- This commit fixes this issue.
- NOTE: apps/Makefile also needs to be updated.
Impact:
- CONFIG_BUILD_KERNEL=y only
Testing:
- Build (make and make export) with sama5d4-ek:knsh
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Some toolchains may be built without libm support, but using
such toochain should not generate any errors in case if math
functions are not used in the program
Signed-off-by: Petro Karashchenko <petro.karashchenko@gmail.com>
For example, task is blocked by nxsem_wait(sem1), use nxsem_wait(sem2)
in signal handler, and take sem2 successfully, after exit from signal
handler to task, nxsem_wait(sem1) returns OK, but the correct result
should be -EINTR.
Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
When enable DEBUG_TCBINFO config, a global struct will
provide, then debuggers can aware nuttx task infomation.
Signed-off-by: zhuyanlin <zhuyanlin1@xiaomi.com>
If we repeatedly call up_cpu_pause and up_cpu_resume, there would be
cases where the next call to up_cpu_pause happens while the other CPU is
still responding to the previous resume request. In this case the
DEBUGASSERT will trigger. We should allow the first CPU to wait until the
other CPU has finished responding to the resume request.
Signed-off-by: Abdelatif Guettouche <abdelatif.guettouche@espressif.com>
This reverts commit 45672c269d.
Because:
* It's very confusing to have cc as LD.
* I don't see what "-nostartfiles -nodefaultlibs" in LDFLAGS are
supposed to do when we use LD directly. It would be simpler to
remove them from our LDFLAGS.
This feature depends on frame pointer, "-fno-omit-frame-pointer" is mandatory
This feature can not be used in THUMB2 mode if you are using GCC toolchain,
More details please refer:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92172
Signed-off-by: chao.an <anchao@xiaomi.com>
It seems to be caused by the corrupted or wrong CPSR restored on return
from exception. NuttX restores the context using code like this:
msr spsr, r1
GCC translates this to:
msr spsr_fc, r1
As a result, not all SPSR fields are updated on exception return. This
should be:
msr spsr_fsxc, r1
This bug has been fixed by Heesub Shin in:
343243c7c0
Change-Id: Ibc64db7bceecd0fb6ef39284fb5bc467f5603e2e
use the same condition check in declaration and reference
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Change-Id: I7b05316e914708fceeddac394d784ee3720a3c1b
All modern desgin support stack pointer and it's also an
important information, so let's standardize this interface.
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Drop to user-space in kernel/protected build with up_pthread_exit,
now all pthread_cleanup functions executed in user mode.
* A new syscall SYS_pthread_exit added
* A new tcb flag TCB_FLAG_CANCEL_DOING added
* up_pthread_exit implemented for riscv/arm arch
Signed-off-by: Huang Qi <huangqi3@xiaomi.com>
Summary:
- SP_SECTION was introduced to allocate spinlock in non-cachable
region mainly for Cortex-A to stabilize the NuttX SMP kernel
- However, all spinlocks are now allocated in cachable area and
works without any problems
- So SP_SECTION should be removed to simplify the kernel code
Impact:
- None
Testing:
- Build test only
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- The CONFIG_SMP_IDLETHREAD_STACKSIZE was introduced to optimize
the idle stack size for other than CPU0
- However, there are no big differences between the idle stacks.
- This commit removes the config to simplify the kernel code
Impact:
- All SMP configurations
Testing:
- Tested with ostest with the following configs
- spresense:smp, spresense:rndis_smp
- esp32-devkitc:smp (QEMU), maix-bit:smp (QEMU)
- sabre-6quad:smp (QEMU), sabre-6quad:netnsh_smp (QEMU)
- raspberrypi-pico:smp, sim:smp (x86_64)
Signed-off-by: Masayuki Ishikawa <asayuki.Ishikawa@jp.sony.com>
arch: Allocate the space from the beginning in up_stack_frame
and modify the affected portion:
1.Correct the stack dump and check
2.Allocate tls_info_s by up_stack_frame too
3.Move the stack fork allocation from arch to sched
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
All supported arch uses a push-down stack:
The stack grows toward lower addresses in memory. The stack pointer
register points to the lowest, valid working address (the "top" of
the stack). Items on the stack are referenced as positive(include zero)
word offsets from sp.
Which means that for stack in the [begin, begin + size):
1.The initial SP point to begin + size
2.push equals sub and then store
3.pop equals load and then add
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Bug description:
CONFIG_SMP=y
Suppose we have 2 cores in SMP, here is the ps return:
PID GROUP CPU PRI POLICY TYPE NPX STATE STACK USED FILLED COMMAND
0 0 0 0 FIFO Kthread N-- Assigned 004076 000748 18.3% CPU0 IDLE
1 0 1 0 FIFO Kthread N-- Running 004096 000540 13.1% CPU1 IDLE
nsh> kill -4 0
or:
nsh> kill -4 1
system blocked.
Reason:
In func xx_sigdeliver() restore stage, when saved_irqcount == 0, that means
rtcb NOT in critical_section before switch to xx_sigdeliver(), then we need
reset the critical_section state before swith back.
Fix:
Add condition to cover saved_irqcount == 0.
Change-Id: I4af7f95e47f6d78a4094c3757d39b01ac9d533b3
Signed-off-by: ligd <liguiding1@xiaomi.com>
Summary:
- Because this_task() returns the current task of the current CPU
Impact:
- SMP only
Testing:
- Tested with sabre-6quad:smp (QEMU)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- This commit fixes global IRQ control logic
- In previous implementation, g_cpu_irqset for a remote CPU was
set in sched_add_readytorun(), sched_remove_readytorun() and
up_schedule_sigaction()
- In this implementation, they are removed.
- Instead, in the pause handler, call enter_critical_setion()
which will call up_cpu_paused() then acquire g_cpu_irqlock
- So if a new task with irqcount > 1 restarts on the remote CPU,
the CPU will only hold a critical section. Thus, the issue such as
'POSSIBLE FOR TWO CPUs TO HOLD A CRITICAL SECTION' could be resolved.
- Fix nxsched_resume_scheduler() so that it does not call spin_clrbit()
if a CPU does not hold a g_cpu_irqset
- Fix nxtask_exit() so that it acquires g_cpu_irqlock
- Update TODO
Impact:
- All SMP implementations
Testing:
- Tested with smp, ostest with the following configurations
- Tested with spresense:wifi_smp (NCPUS=2,4)
- Tested with sabre-6quad:smp (QEMU, dev board)
- Tested with maix-bit:smp (QEMU)
- Tested with esp32-core:smp (QEMU)
- Tested with lc823450-xgevk:rndis
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- Apply the same logic added to cxd56_cpupause.c
Impact:
- SMP only
Testing:
- Tested with sabre-6quad:smp (QEMU and dev board)
- Run smp and ostest
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- Remove unnecessary d-cache operation to make boot fast
Impact:
- armv7-a SMP only
Testing:
- Tested with sabre-6quad:smp (QEMU and dev board)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- This commit fixes armv7-a deadlocks with D-cache in SMP mode.
- In SMP mode, MMU for SDRAM area must be set to shareable
Impact:
- SMP only
Testing:
- Tested with sabre-6quad:smp (QEMU and dev board)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- This commit fixes kernel stack dump information
Impact:
- Affects armv7-a with kernel build
Testing:
- Built with sama5d4-ek:knsh
- Not tested
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- Remove -4/-8 offset coding in imx_irq.c and arm_vectors.S
- Instead, add SP adjustment after calling setirqstack/setfiqstack
- Fix off-by-one irq/fiq stack allocation in 8-byte aligned arch
- Fix comments on the user stack pointer in arm_vectors.S
- Also, fix up_dumpstate() to extract the user stack pointer
- NOTE: stack pointer alignment is 8-byte
Impact:
- Affects armv7-a with interrupt stack enabled
Testing:
- Tested with sabre-6quad:smp with QEMU
- Tested with sabre-6quad:nsh with QEMU
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Summary:
- This commit fixes CPUx IDLE stack top for SMP
- Also removes SMP_STACK_TOP from smp.h
Impact:
- Affects armv7-a SMP only
Testing:
- Tested with sabre-6quad:smp (QEMU)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
1.Get the stack pointer from sp instead of .Lstkinit's field
2.Make g_idle_topstack point to the end of the idle stack
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
and remove the special handling in the stack dump
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Change-Id: Ia1ef9a427bd4c7f6cee9838d0445f29cfaca3998
PR #1450 broke the Cygwin build. Refer to Issue #1672.
The use of of logic like:
EXTRA_LIBPATHS += -L "${dir ${shell $(CC) $(ARCHCPUFLAGS) --print-file-name=libgcc.a}}"
fails when the Toolchain $(CC) is a native Windows toolchain. That is because the returned path is a Windows-style patch which cannot be handled by the make 'dir' command. Commit 4910d43ab0 reorganized a lot of definitions and replaced the correct code with the use of the limit make 'dir' command. The original code used the Bash dirname command which does not suffer from this limitation; it can handle both POSIX and Windows paths.
This was verified using the stm32f4discover:nsh toolchain with the Windows native ARM Embedded toolchain. That toolchain returns:
arm-none-eabi-gcc --print-file-name=libgcc.a
c:/program files (x86)/gnu tools arm embedded/9 2019-q4-major/bin/../lib/gcc/arm-none-eabi/9.2.1/libgcc.a
1.It make sense to let Toolchain.defs give the default value
2.The board can still change if the default isn't suitable
3.Avoid the same definition spread more than 200 Make.defs
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Change-Id: Ic2649f1c7689bcf59c105ca8db61cad45b6e0e64
since exit will be only callable from userspace and change
the 1st argument from "const uint8_t *" to "const char *"
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Change-Id: I86487d57210ab63109148232da71dbc4d60a563b
Remove support for the Codesourcery, Atollic, DevKitArm, Raisonance, and CodeRed toolchains. Not only are these tools old and no longer used but they are all equivalent to standard ARM EABI toolchains. Retaining specific support has no effect (they are still supported, but now just as generic EABI toolchains).
All complaints fixed except for those that were not possible to fix:
- Used of Mixed case identifier in ESP32 files. These are references to Expressif ROM functions which are outside of the scope of NuttX.
- Remove per-thread errno from the TCB structure (pterrno)
- Remove get_errno() and set_errno() as functions. The macros are still available as stubs and will be needed in the future if we need to access the errno from a different address environment (KERNEL mode).
- Add errno value to the tls_info_s structure definitions
- Move sched/errno to libs/libc/errno. Replace old TCB access to the errno with TLS access to the errno.
The naming standard at https://cwiki.apache.org/confluence/display/NUTTX/Naming+FAQ requires that all architecture-private functions begin with the name of the arch, not up_.
This PR addresses only these name changes for the ARM-private functions up_instack_base() and up_instack_top() which should be called arm_instack_base() and arm_instack_top().
There should be no impact of this change (other that one step toward more consistent naming).
Normal PR checks are sufficient
Summary
The naming standard at https://cwiki.apache.org/confluence/display/NUTTX/Naming+FAQ requires that all MCU-private functions begin with the name of the architecture, not up_.
This PR addresses function-like macro naming that was missing in previous PRs: up_savestate() and up_restorestate() which must be named arm_savestate() and arm_restorestate().
Impact
There should be no impact of this change (other that one step toward more consistent naming).
Testing
stm32f103-minimum:nsh
