MacBook Air M1: Firefox crash visiting https://www.shadertoy.com (stack smashing detected)

Hi !

Got consistent crash visiting https://www.shadertoy.com home page on MacBook Air M1 (j313) running 6.5.6-403.asahi.fc39.aarch64+16k.

This is a fresh FC39 install with KDE.

See backtrace below, nothing scary in system logs.

Any idea?

*** stack smashing detected ***: terminated

Thread 29 "CanvasRenderer" received signal SIGABRT, Aborted.
[Switching to Thread 0xffff31c03160 (LWP 6968)]
__pthread_kill_implementation (threadid=281471516422496, signo=signo@entry=6, no_tid=no_tid@entry=0) at pthread_kill.c:44
44            return INTERNAL_SYSCALL_ERROR_P (ret) ? INTERNAL_SYSCALL_ERRNO (ret) : 0;
(gdb) bt
#0  __pthread_kill_implementation (threadid=281471516422496, signo=signo@entry=6, no_tid=no_tid@entry=0) at pthread_kill.c:44
#1  0x0000ffff939922f8 [PAC] in __pthread_kill_internal (signo=6, threadid=<optimized out>) at pthread_kill.c:78
#2  0x0000ffff93945800 in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
#3  0x0000ffff93930288 [PAC] in __GI_abort () at abort.c:79
#4  0x0000ffff93984fbc [PAC] in __libc_message (fmt=fmt@entry=0xffff93a6ec08 "*** %s ***: terminated\n") at ../sysdeps/posix/libc_fatal.c:150
#5  0x0000ffff93a0f520 [PAC] in __GI___fortify_fail (msg=msg@entry=0xffff93a6ec20 "stack smashing detected") at fortify_fail.c:24
#6  0x0000ffff93a10578 [PAC] in __stack_chk_fail () at stack_chk_fail.c:24
#7  0x0000fffed1eea540 [PAC] in agx_emit_parallel_copies () from /usr/lib64/dri/apple_dri.so
#8  0x0000fffed1ee8d9c [PAC] in agx_ra () from /usr/lib64/dri/apple_dri.so
#9  0x0000fffed1ed3e10 [PAC] in agx_compile_shader_nir () from /usr/lib64/dri/apple_dri.so
#10 0x0000fffe000000ea [PAC] in ?? ()
#11 0x0003fffe000000ea in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
1 Like

Just tested with the exact same model, and I’m encountering a crash as well.

This is a known limitation of the driver (excessively complex shaders are not supported) and not a bug. The next driver update will abort with a verbose error message instead of crashing like that. The missing feature (register spilling) will be implemented in the future.

The home page used to load in the past, but some complex shadertoys are known not to. I guess one of those happens to be featured on the front page right now.

Thanks. Saw the « don’t report » message in one of mesa’s MR. Did not realize the crash was related.