This is on Android 12 on a Pixel 6. I am installing a SIGSEGV handler to catch and handle on purpose generated segmentation faults. This works as expected but I am observing a single case where the info->si_addr passed to the handler is not what I expect it to be.
For example let's say that a memory is allocated with mmap at 0x6ecae15000 and size 4194304 bytes. It is protected with PROT_NONE. Then there is a write on address 0x6ecae1e000. A SIGSEGV is triggered and the handler is called but the info->si_addr is 0x277500001a93.
However the ucontext->uc_mcontext.fault_address holds the expected address (0x6ecae1e000).
Any ideas why there's this discrepancy in info->si_addr value? Maybe on arm only ucontext->uc_mcontext.fault_address should be relied upon getting the fault address?
