RTEMS | noelv/qemu: Misaligned BSP reset code can make QEMU run after application exit (#5366)

[Matteo Concas (@matteo.concas)]

Matteo Concas created an issue: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5366



## Summary

NOEL-V binaries on QEMU might have misaligned `bsp_reset` code causing it to run forever:

```
[ RTEMS shutdown ]
CPU: 0
RTEMS version: 6.1.0.not-released
RTEMS tools: 13.3.0 20240521 (RTEMS 6, RSB no-repo, Newlib 1b3dcfd)
executing thread ID: 0x0a010001
executing thread name: UI1 

*** FATAL ***
fatal source: 9 (RTEMS_FATAL_SOURCE_EXCEPTION)
CPU: 0
mstatus 0x8000000a00007800
mcause  0x0000000000000003
mepc    0x0000000000026000
ra      0x0000000000024bc8
sp      0x00000000000f1ab0
gp      0x000000000007e018
tp      0x00000000000f1c98
t0      0x0000000000000057
t1      0x0000000000000000
t2      0x0000000000000078
s0      0x0000000000000005
s1      0x0000000000000000
a0      0x0000000000000020
a1      0x00000000000f1ab0
a2      0x0000000000000038
a3      0x0000000000000000
a4      0x0000000000024f38
a5      0x00000000000f1ab0
a6      0x0000000000000018
a7      0xfffffffffffffffc
s2      0x00000000000dc628
s3      0x0000000000037ae0
s4      0x0000000000037ae0
s5      0x0000000000000000
s6      0x0000000000000000
s7      0x0000000000000000
s8      0x0000000000000120
s9      0x00000000000379c0
s10     0x0000000000000000
s11     0x0000000000000000
t3      0x0000000000000000
t4      0x0000000000000000
t5      0x0000000000000000
t6      0x0000000000000000
fcsr    0x00000000
RTEMS version: 6.1.0.not-released
RTEMS tools: 13.3.0 20240521 (RTEMS 6, RSB no-repo, Newlib 1b3dcfd)
executing thread ID: 0x0a010001
executing thread name: UI1 

*** FATAL ***
fatal source: 9 (RTEMS_FATAL_SOURCE_EXCEPTION)
CPU: 0
mstatus 0x8000000a00007800
mcause  0x0000000000000003
mepc    0x0000000000026000
ra      0x0000000000024bc8
sp      0x00000000000f1860
gp      0x000000000007e018
tp      0x00000000000f1c98
t0      0x0000000000000057
t1      0x0000000000000000
t2      0x0000000000000078
s0      0x0000000000000009
s1      0x00000000000f1960
a0      0x0000000000000020
a1      0x00000000000f1860
a2      0x0000000000000038
a3      0x0000000000000000
a4      0x0000000000024f38
a5      0x00000000000f1860
a6      0x0000000000000018
a7      0xfffffffffffffffc
s2      0x00000000000dc628
s3      0x0000000000037ae0
s4      0x0000000000037ae0
s5      0x0000000000000000
s6      0x0000000000000000
s7      0x0000000000000000
s8      0x0000000000000120
s9      0x00000000000379c0
s10     0x0000000000000000
s11     0x0000000000000000
t3      0x0000000000000000
t4      0x0000000000000000
t5      0x0000000000000000
t6      0x0000000000000000
fcsr    0x00000000
RTEMS version: 6.1.0.not-released
RTEMS tools: 13.3.0 20240521 (RTEMS 6, RSB no-repo, Newlib 1b3dcfd)
executing thread ID: 0x0a010001
executing thread name: UI1 
```

This keeps going forever.

This does not affect hardware.

## Steps to reproduce

Hard to reproduce as it seems most of the time the code will be correctly aligned.

## Solution

A fix is to align the instruction in `bsps/riscv/noel/start/bsp_fatal_halt.c:bsp_reset()` :

```c
void bsp_reset(rtems_fatal_source source, rtems_fatal_code code)
{
        (void)source;

        uint64_t args[2] = {ADP_Stopped_ApplicationExit, code};
        __asm__ volatile("li a0, %0" ::"i"(TARGET_SYS_EXIT_EXTENDED));
        __asm__ volatile("mv a1, %0" ::"r"(&args));
        __asm__ volatile(".align 4; slli zero, zero, 0x1f");
        __asm__ volatile("ebreak");
        __asm__ volatile("srai zero, zero, 0x7");
        RTEMS_UNREACHABLE();
}
```

### Pre-set options

-- 
View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5366
You're receiving this email because of your account on gitlab.rtems.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/bugs/attachments/20251002/025900e9/attachment-0001.htm>