Possible race condition when trap in spawn thread (using WASI threads)

[eloparco]

To reproduce the problem, follow the description here #1869 (comment).

[loganek]

As this is difficult to reproduce issue, I'd suggest compiling the runtime with sanitizers and see if we can get some info from that.

[hritikgupta]

I wasn't able to reproduce the issue, but tried compiling runtime with 2 sanitiser modes i.e. address and thread as suggested,

with address, it gives the below error (this is apparently pointing to a line in some other file in summary):

==47563==ERROR: AddressSanitizer: SEGV on unknown address 0x68a20000689f (pc 0x000104142cf0 bp 0x00016bccabf0 sp 0x00016bccabb0 T0)
==47563==The signal is caused by a UNKNOWN memory access.
    #0 0x104142cf0 in bh_list_elem_next bh_list.c:93
    #1 0x1041a5cec in wasm_cluster_wait_for_all_except_self thread_manager.c:965
    #2 0x104156580 in wasm_application_execute_main wasm_application.c:218
    #3 0x10413a4d4 in app_instance_main main.c:87
    #4 0x1041394f4 in main main.c:660
    #5 0x104555088 in start+0x204 (dyld:arm64e+0x5088)

==47563==Register values:
 x[0] = 0x000068a20000689f   x[1] = 0x0000000000000000   x[2] = 0x0000000000008000   x[3] = 0x0000000000000000  
 x[4] = 0x00000001041a90d4   x[5] = 0x000000016bcca900   x[6] = 0x000000016bcca900   x[7] = 0x0000000000000000  
 x[8] = 0x000068a20000689f   x[9] = 0x0000000000000002  x[10] = 0x0000000000000000  x[11] = 0x0000000000000002  
x[12] = 0x0000000000000002  x[13] = 0x0000000000000000  x[14] = 0x0000000000000000  x[15] = 0x0000000000000000  
x[16] = 0x0000000180dc2d4c  x[17] = 0x00000001049c0740  x[18] = 0x0000000000000000  x[19] = 0x000000016bccb100  
x[20] = 0x0000000104137a84  x[21] = 0x00000001045b0070  x[22] = 0x0000000000000000  x[23] = 0x0000000000000000  
x[24] = 0x0000000000000000  x[25] = 0x0000000000000000  x[26] = 0x0000000000000000  x[27] = 0x0000000000000000  
x[28] = 0x0000000000000000     fp = 0x000000016bccabf0     lr = 0x00000001041a4c54     sp = 0x000000016bccabb0  
AddressSanitizer can not provide additional info.
SUMMARY: AddressSanitizer: SEGV bh_list.c:93 in bh_list_elem_next
==47563==ABORTING
zsh: abort      ./iwasm -v=5 wasm-apps/thread_termination.wasm

the above error was also intermittent.

whereas with thread, the sanitiser didn't give any warning/error.

[loganek]

I wasn't able to reproduce the issue reported by @eloparco , but I'm regularly able to reproduce a different one:

iwasm: ../nptl/pthread_mutex_lock.c:81: __pthread_mutex_lock: Assertion `mutex->__data.__owner == 0' failed.

It might be related as it crashes in atomic.notify. Attaching coredump and executable for analysis.
assertion.zip

[eloparco]

Fixed by multiple commits for thread safety in main branch.