zynq7000-zedboard: exception abort loop during system startup

[damianloew]

• Happens occasionally

• Issue firstly discovered on 0a9dfbc (to be investigated whether it can be reproduced using earlier commits)

• For now reproduced only on u1 zedboard test unit on CI: https://github.com/phoenix-rtos/phoenix-rtos-project/actions/runs/9249488896/job/25441474802

• Output:

dummyfs: initialized
version 2.2. (NAND) (SUMMARY)  © 2001-2006 Red Hat, Inc.


Exception: 4 #Abort
 r0=00000000  r1=00000000  r2=000000b0  r3=00027ab8
 r4=00026730  r5=00000000  r6=00000003  r7=7fffffa8
 r8=00000000  r9=c0149ab4 r10=c0149ab8  fp=c0161f8c
 ip=fffffffc  sp=c0161ef4  lr=00019b55  pc=00015702
psr=20000030 dfs=00000807 dfa=00000000 ifs=00000000
ifa=00000000

Exception: 4 #Abort

Exception: 4 r0=00000000  r1=00000000  r2=000000b0  r3=00027ab8
 r4=00026730  r5=00000000  r6=00000003  r7=7fffffa8
 r8=00000000  r9=c0149ab4 r10=c0149ab8  fp=c0161f8c
 ip=fffffffc  sp=c0161ef4  lr=00019b55  pc=00015702
psr=20000030 dfs=00000807 dfa=00000000 ifs=00000000
ifa=00000000

 #Abort
 r0=00000000  r1=00000000  r2=000in thread 18, process "/bin/bind" (PID: 8)
000b0  r3=00027ab8
 r4=00026730  r5=00000000  r6=00000003  r7=7fffffa8
 r8=00000000  r9=c0149ab4 r10=c0149ab8  fp=c0161f8c
 ip=fffffffc  sp=c0161ef4  lr=00019b55  pc=00015702
psr=20000030 dfs=00000807 dfa=00000000 ifs=00000000
ifa=00000000

Exception: 4 #Abort

...

[anglov]

arm-phoenix-addr2line 0x00019b55 -e _build/armv7a9-zynq7000-zedboard/prog/psh
/src/libphoenix/stdlib/atexit.c:51
arm-phoenix-addr2line 0x00015702 -e _build/armv7a9-zynq7000-zedboard/prog/psh
/src/libphoenix/arch/arm/memset.S:131

[agkaminski]

arm-phoenix-addr2line 0x00019b55 -e _build/armv7a9-zynq7000-zedboard/prog/psh /src/libphoenix/stdlib/atexit.c:51 arm-phoenix-addr2line 0x00015702 -e _build/armv7a9-zynq7000-zedboard/prog/psh /src/libphoenix/arch/arm/memset.S:131

I've came to same solution by compiling the artifacts myself, but it seems to me to be inconsistent with the character of the exception (instruction that pc is pointing to is not accessing memory, but the abort was caused by NULL dereference). I suspect there was some difference between binary on CI and the one I compiled myself.

[anglov]

arm-phoenix-addr2line 0x00019b55 -e _build/armv7a9-zynq7000-zedboard/prog/psh /src/libphoenix/stdlib/atexit.c:51 arm-phoenix-addr2line 0x00015702 -e _build/armv7a9-zynq7000-zedboard/prog/psh /src/libphoenix/arch/arm/memset.S:131

I've came to same solution by compiling the artifacts myself, but it seems to me to be inconsistent with the character of the exception (instruction that pc is pointing to is not accessing memory, but the abort was caused by NULL dereference). I suspect there was some difference between binary on CI and the one I compiled myself.

I compare md5sum of stripped binary and it's same