Rémi Bernon (@rbernon) commented about dlls/ntdll/unix/signal_arm64.c:
"mov sp, x10\n\t" /* we're now on the kernel stack, stitch unwind info with previous frame */ __ASM_CFI_CFA_IS_AT2(x22, 0x98, 0x02) /* frame->syscall_cfa */
__ASM_CFI_REG_IS_AT2(sp, x22, 0x98, 0x02)
I'm not sure to understand why you removed the sp info here (vs not in the i386 / x86_64 dispatchers) and why you didn't include it in user_mode_abort_thread?
Is it because it's not been modified before, and so we assume that the default and implicit rule of `.cfi_val_offset %sp,0`, or old sp == cfa (or whatever it is) is still valid?
Feels a bit brittle if that ever changes isn't it? Especially in the ARM dispatchers where we could add intermediate user stack CFI like on x86.