Re: help to debug tests failed on riscv64

[Andrius Štikonas]

That is not a kernel crash, that's userspace crash. Also setjmp can't cause 
kernel crash
as it's entirely a userspace thing, it doesn't do any syscalls.

setjmp on mes on riscv64 is known to be broken but it's not really needed right 
now.
When we bootstrap tcc, first tcc binary (tcc-mes) is built with -DHAVE_SETJMP=0.
After that setjmp works (meslibc has a completely different
setjmp implementation for tcc/gcc which does work).

I don't remember what's the issue with 68-truncate-shift.c, Ekaitz might 
remember more.

Andrius

2024 m. sausio 10 d., trečiadienis 07:36:45 GMT Bo YU rašė:
> Hi,
> 
> I am looking at the mes's build issue on Debian riscv64[0].
> 
> There are two test cases failed on here:
> 
> ```
> 
> [0;31mFAIL [m: lib/tests/scaffold/68-truncate-shift.c
> 
>  [0;31mFAIL [m: lib/tests/setjmp/80-setjmp.c
> 
> ```
> And 80-setjmp.c triggers a kernel crash from dmesg log:
> ```
> [1031114.072314] 80-setjmp[3953800]: unhandled signal 11 code 0x1 at
> 0x0000000000000000 in 80-setjmp[10000+3000]
> [1031114.072377] CPU: 1 PID: 3953800 Comm: 80-setjmp Tainted: G
>      L     6.6.8-riscv64 #1  Debian 6.6.8-1
> [1031114.072390] Hardware name: SiFive HiFive Unmatched A00 (DT)
> [1031114.072398] epc : 0000000000010910 ra : 00000000000108b6 sp :
> 0000003ffe89f8c0
> [1031114.072407]  gp : 0000000000014800 tp : 0000003fb11003c0 t0 :
> 0000003ffe89fa48
> [1031114.072416]  t1 : 0000003ffe89fa18 t2 : 0000000000000003 s0 :
> 0000003ffe89f900
> [1031114.072425]  s1 : 0000003ff59320cb a0 : 000000000000000f a1 :
> 000000000aaa3000
> [1031114.072433]  a2 : 0000000000000007 a3 : 0000000000000001 a4 :
> 0000000000011798
> [1031114.072442]  a5 : 0000000000000000 a6 : 0000003fb11151ec a7 :
> 00000000000000d6
> [1031114.072450]  s2 : 0000002adfd384c8 s3 : 0000000000000001 s4 :
> 0000003ff5931a48
> [1031114.072459]  s5 : 0000003ff5931a18 s6 : 0000003fb1126d50 s7 :
> fffffffffffffffe
> [1031114.072468]  s8 : 000000000000006b s9 : 0000000000000101 s10:
> 0000003fb1126d50
> [1031114.072476]  s11: 0000002ada73b148 t3 : 0000000000000000 t4 :
> 0000000000094ed2
> [1031114.072485]  t5 : 0000000000000010 t6 : 0000000003f0b27f
> [1031114.072492] status: 0000000200004020 badaddr: 0000000000000000
> cause: 000000000000000d
> 
> ```
> So I suspected the mes found a riscv kernel issue.
> 
> The question is how to reproduce the test if I get `mes` binary. I can
> disbable to test to get it but was stuck to do next step.
> 
> Please cc me.:)
> 
> Thanks,
> Bo
> 
> [0]: 
> https://buildd.debian.org/status/fetch.php?pkg=mes&arch=riscv64&ver=0.26-1&stamp=1704532404&raw=0
> 
>