When compiling with O2 optimization, the compiler optimizes the code
in a way that causes irq variable to be corrupted. The getipsr()
function reads IPSR into r0, but the subsequent inline assembly that
sets FPSCR also uses r0 without declaring it as clobbered. This causes
the compiler to reuse r0 for the immediate value (0x40000), overwriting
the IRQ number read from IPSR.
The issue manifests as:
- getipsr() correctly reads IPSR (e.g., 0xf for IRQ 15)
- Compiler optimizes and reuses r0 for ARM_FPSCR_LTPSIZE_NONE (0x40000)
- irq variable gets the wrong value 0x40000 instead of actual IRQ number
- This leads to assertion failures in irq_dispatch due to invalid IRQ
Root cause analysis from disassembly:
mrs r0, IPSR ; Read IPSR to r0
mov.w r0, #262144 ; Compiler overwrites r0 with 0x40000!
vmsr fpscr, r0 ; Set FPSCR
str r0, [sp, #4] ; Store corrupted 0x40000 as irq
...
ldr r0, [sp, #4] ; Load corrupted value
bl irq_dispatch ; Call with wrong IRQ number 0x40000
Fix by adding r0 to the clobber list in the inline assembly, which
forces the compiler to save irq value before using r0 for FPSCR setup.
This issue only occurs at O2 optimization level and affects ARMv8-M
architecture with FPU enabled.
Signed-off-by: xuxingliang <xuxingliang@xiaomi.com>
armv8m:use ip save pc
armv7m:use ip save pc, r11 save sp
nsh_main backtrace before:
```
0 nxsem_wait_slow (sem=0x10001e0 <g_uart0port+32>) at semaphore/sem_wait.c:202
1 0x0000882c in nxsem_wait (sem=0x10001e0 <g_uart0port+32>) at semaphore/sem_wait.c:270
2 0x0000e7e2 in uart_read (filep=0x608037d4, buffer=0x60c013eb " h\024\300`", buflen=1)
at serial/serial.c:1203
3 0x0001aad4 in file_readv_compat (filep=0x608037d4, uio=0x60c01344) at vfs/fs_read.c:81
4 0x0001abae in file_readv (filep=0x608037d4, uio=0x60c01344) at vfs/fs_read.c:165
5 0x0001ac2e in nx_readv (fd=0, iov=0x60c01388, iovcnt=1) at vfs/fs_read.c:256
6 0x0001ac84 in readv (fd=0, iov=0x60c01388, iovcnt=1) at vfs/fs_read.c:318
7 0x0001acca in read (fd=0, buf=0x60c013eb, nbytes=1) at vfs/fs_read.c:352
8 0x000100ea in STUB_read (nbr=56, parm1=0, parm2=1623200747, parm3=1) at stubs/STUB_read.c:9
9 0x000014d2 in dispatch_syscall () at armv8-m/arm_svcall.c:82
```
nsh_main backtrace after:
```
0 nxsem_wait_slow (sem=0x10001f0 <g_uart0port+32>) at semaphore/sem_wait.c:202
1 0x000088c0 in nxsem_wait (sem=0x10001f0 <g_uart0port+32>) at semaphore/sem_wait.c:270
2 0x0000e882 in uart_read (filep=0x6080380c, buffer=0x60c013eb " h\024\300`", buflen=1)
at serial/serial.c:1203
3 0x0001ab74 in file_readv_compat (filep=0x6080380c, uio=0x60c01344) at vfs/fs_read.c:81
4 0x0001ac4e in file_readv (filep=0x6080380c, uio=0x60c01344) at vfs/fs_read.c:165
5 0x0001acce in nx_readv (fd=0, iov=0x60c01388, iovcnt=1) at vfs/fs_read.c:256
6 0x0001ad24 in readv (fd=0, iov=0x60c01388, iovcnt=1) at vfs/fs_read.c:318
7 0x0001ad6a in read (fd=0, buf=0x60c013eb, nbytes=1) at vfs/fs_read.c:352
8 0x0001018a in STUB_read (nbr=56, parm1=0, parm2=1623200747, parm3=1) at stubs/STUB_read.c:9
9 0x00001ece in dispatch_syscall () at armv8-m/arm_dispatch_syscall.S:111
10 0x2000a874 in read (parm1=0, parm2=0x60c013eb, parm3=1) at proxies/PROXY_read.c:9
11 0x20001152 in readline_getc (vtbl=0x60c01468) at readline_fd.c:73
12 0x20003994 in readline_common (vtbl=0x60c01468, buf=0x60c017f8 "\n", buflen=80)
at readline_common.c:513
13 0x200012ec in readline_fd (buf=0x60c017f8 "\n", buflen=80, infd=0, outfd=1) at readline_fd.c:236
14 0x200010a8 in nsh_session (pstate=0x60c01540, login=1, argc=1, argv=0x60c00548)
at nsh_session.c:229
15 0x20000eb0 in nsh_consolemain (argc=1, argv=0x60c00548) at nsh_consolemain.c:75
16 0x20000e4c in nsh_main (argc=1, argv=0x60c00548) at nsh_main.c:74
17 0x20000062 in nxtask_startup (entrypt=0x20000e11 <nsh_main>, argc=1, argv=0x60c00548)
at sched/task_startup.c:72
18 0x00009ac4 in nxtask_start () at task/task_start.c:106
19 0x60c00548 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
```
Signed-off-by: anjiahao <anjiahao@xiaomi.com>
* Explicit registration of ttys0 when no console defined
* Fixes overwriting registration of lpuart1 on ttys1 when no console on ttys0
* Change regstration logic to check for console assignment
* See issue 17116
Signed-off-by: dirksavage88 <dirksavage88@gmail.com>
Previously only up to 31 PIO instructions were accepted.
But the hardware allowed 32 instructions.
Now we accept 32 instructions.
See the corresponding commit in the pico-sdk repository:
6f7dc67791
Signed-off-by: Lars Kruse <devel@sumpfralle.de>
The specific Cortex-R52 implementation could be configured with
a Single-Precision-only FPU (SP-only) and no Neon unit.
Executing double-precision instructions (e.g., `vadd.f64`) triggers
an Undefined Instruction exception.
The standard `-mfpu=fp-armv8` implicitly enables double-precision,
which is unsafe for this hardware.
`-mfpu=fpv5-sp-d16` is selected as the closest architectural match.
- It enforces Single Precision code generation (preventing crashes).
- It enables VFPv4/FPv5 features like FMA (Fused Multiply-Add) supported by the CR52 FPU.
- It restricts the register set to d0-d15, matching the hardware constraints.
This ensures the compiler utilizes hardware FPU and FMA acceleration
without emitting illegal double-precision instructions.
Signed-off-by: xiezhanpeng3 <xiezhanpeng3@lixiang.com>
Initial commit for adding rs-485 support to serial driver. More is needed as well as testing.
Style fixes.
Added unconfigure functionality.
Added Unconfigure configs to CONFIG.
More style fixes.
Changed depend spacing from 4 spaces to tab.
Fixed priv types, changed serialin to getreg.
Do not unconfigure tx/rx/dir pins by default.
Signed-off-by: kywwilson11 <kwilson@2g-eng.com>
The parameter count of read had to be a multiple of 4, otherwise the
result was truncated. Now all values are possible.
Signed-off-by: Mathias Duckeck <mathias.duckeck@avat.de>
- Fix EtherCAT signals drive strengh (from errata PORTS_CM.H002), caused bus faults.
- Changed xmc4_ecat.c for compile time pin definition.
- Fixed xmc4_ecat.c register not written (reset value already used).
- Removed EXTCLK for xmc4800 as pin is used for ECAT.
- Clean xmc4 familly board.h and clocks config.
Signed-off-by: adriendesp <adrien.desproges@gmail.com>
This commit adds support for SPI_SETDELAY operation on SAMv7 QSPI
peripheral running in SPI mode. The logic is the same as for standard
SPI peripheral, just different registers. The change allows to set
custom delays between transfers, chip selects and so on.
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
During I2C read, one-too-many byte is read, which can lead to the I2C bus
getting stuck. This is likely due to the STOP condition being set at the
wrong time or being missed completely. The chip offers a shortcut, such
that the STOP condition is set automatically after the last byte is being
written/read.
Signed-off-by: Max Kriegleder <max.kriegleder@gmail.com>
When TICKLESS not enabled, up_alarm_set_lowerhalf will call start, if we
overwrite the compare register will cause TICK mode no longer isr.
Signed-off-by: buxiasen <buxiasen@xiaomi.com>
In multi-core environment where NuttX runs on each core, if one core
sets the RTC time, the RTC value gotten on other cores is incorrect.
This is caused by clock_gettime(CLOCK_MONOTONIC) function used to get
elapsed time, which uses a core-specific global varaiable g_basetime
as the base time.
To fix this, update the g_basetime from the backup SRAM that can be
shared between cores in setting/getting the RTC time.
Signed-off-by: SPRESENSE <41312067+SPRESENSE@users.noreply.github.com>
- Added PWM driver support for the RA4M1 microcontroller using the GPT timer.
- This driver supports Saw-wave mode and one of the two output channels (A or B).
- Added necessary configurations in CMakeLists.txt, Kconfig, and Make.defs.
- Created new header file for GPT.
Signed-off-by: leocafonso <leocafonso@gmail.com>
Use clock_get_sched_tclock_systime_ticks() to access system ticks
in order to fix the build error and improve code safety.
Signed-off-by: Chengdong Wang <wangchengdong@lixiang.com>
The .align N syntax in GHS means the alignment is N Bytes. In GCC
compiler, it means 2**N Bytes. Therefore, .align 32 in GHS and
.align 5 in GCC are equivalent.
Co-authored-by: Chengdong Wang <wangchengdong@lixiang.com>
Signed-off-by: xiezhanpeng3 <xiezhanpeng3@lixiang.com>
The $+0x0 is not accepted by Ghs compiler. Therefore, we change it to a
more common syntax " b ." for endless self loop.
Signed-off-by: xiezhanpeng3 <xiezhanpeng3@lixiang.com>
The GreenHills Compiler uses Unified Assembler Language by default and
it does not recognize .syntax unified instruction. Therefore, remove
them if GHS compiler is used.
Co-authored-by: Chengdong Wang <wangchengdong@lixiang.com>
Signed-off-by: xiezhanpeng3 <xiezhanpeng3@lixiang.com>
The GreenHills Compiler uses different grammer for the .type directive
in asm, this patch adds support for the grammer for arm-v8r.
Similar change for arm-v7r was done in pull request #12883.
Co-authored-by: Chengdong Wang <wangchengdong@lixiang.com>
Signed-off-by: xiezhanpeng3 <xiezhanpeng3@lixiang.com>
1. up_irq_save should not mask fiq if CONFIG_ARCH_HIPRI_INTERRUPT=y
2. up_irq_save should mask fiq if CONFIG_ARCH_TRUSTZONE_SECURE=y
3. up_irq_save should mask irq if CONFIG_ARCH_TRUSTZONE_SECURE=n
4. add up_secure_irq in arm64
5. add ARCH_HAVE_TRUSTZONE support for ARCH_CORTEX_R82
Signed-off-by: hujun5 <hujun5@xiaomi.com>
Fix a transfer issue in the Kinetis DSPI driver when operating
with transmit and receive FIFOs disabled (`MCR[DIS_TXF]=1`,
`MCR[DIS_RXF]=1`). In this mode, the DSPI module behaves as a
simple double-buffered SPI interface without TX staging.
When FIFOs are disabled, `PUSHR` acts as a single 32-bit
command/data register. Partial (16-bit) writes to its upper or
lower halves can result in incomplete or corrupted transfers.
This patch ensures the full 32-bit packet is prepared and
written in a single operation.
* Resolves broken SPI transactions with LAN9252 (EasyCAT).
* Improves reliability in non-FIFO DSPI configurations.
* No impact on DMA or FIFO-enabled modes.
Signed-off-by: trns1997 <trns1997@gmail.com>
Replace the spinlock/sched_lock pair in arm/stm32 with
spin_lock_irqsave_nopreempt() to improve code clarity and consistency.
Signed-off-by: Chengdong Wang wangchengdong@lixiang.com
Replace the spinlock/sched_lock pair in arm/sam34 with
spin_lock_irqsave_nopreempt() to improve code clarity and consistency.
Signed-off-by: Chengdong Wang wangchengdong@lixiang.com
Replace the spinlock/sched_lock pair in arm/s32k3xx with
spin_lock_irqsave_nopreempt() to improve code clarity and consistency.
Signed-off-by: Chengdong Wang wangchengdong@lixiang.com
Replace the spinlock/sched_lock pair in arm/s32k1xx with
spin_lock_irqsave_nopreempt() to improve code clarity and consistency.
Signed-off-by: Chengdong Wang wangchengdong@lixiang.com
