The ARM manual mentions that:
During execution, PC does not contain the address of the currently executing instruction. The address of the currently executing instruction is typically PC-8 for ARM, or PC-4 for Thumb.
Does this apply for PC
values in crash backtraces as well?
For example, if I have the following backtrace (from Android armv7 device using ARM instructions):
libSomeLib.so!SomeClass::someMethod [someFile.cpp : 638 + 0x4]
r0 = 0x00000001 r1 = 0xffffffff r2 = 0x00000000 r3 = 0xd4fcd71c
r4 = 0xd39b9284 r5 = 0xd39b927c r6 = 0xd39b9278 r7 = 0xc7025520
r8 = 0xc5e1d7b0 r9 = 0xe01136a8 r10 = 0x00000012 r12 = 0xd39b9268
fp = 0xd39b92d4 sp = 0xd39b9268 lr = 0xd4ea1f24 pc = 0xd4ea1f24
...
0xd47cb000 - 0xd5079fff libSomeLib.so ???
Should I look for the crashing instruction inside libSomeLib.so
at 0xd4ea1f24-0xd47cb000=0x006D6F24
or 0xd4ea1f24-0xd47cb000-8=0x006D6F1C
?
The ARM instruction set architecture has three addressing modes: Immediate. The offset is an unsigned integer that is stored as part of the instruction. It can be added to or subtracted from the value in the base register.
According to the ARM IC. In ARM state, the value of the PC is the address of the current instruction plus 8 bytes. In Thumb state: For B, BL, CBNZ, and CBZ instructions, the value of the PC is the address of the current instruction plus 4 bytes.
I made an intentional crash to investigate. Source code:
int* crashPointer = nullptr;
*crashPointer = 7;
Generated assembly:
11feb8: e3005007 movw r5, #7
11febc: e3006000 movw r6, #0 <--- r6 is #0
11fec0: e50b0014 str r0, [fp, #-20] ; 0xffffffec
11fec4: e50b1018 str r1, [fp, #-24] ; 0xffffffe8
11fec8: e51b0014 ldr r0, [fp, #-20] ; 0xffffffec
int* crashPointer = nullptr;
11fecc: e50b601c str r6, [fp, #-28] ; 0xffffffe4 <--- Stores #0 to [fp, #-28] from r6
*crashPointer = 7;
11fed0: e51b101c ldr r1, [fp, #-28] ; 0xffffffe4 <--- Loads #0 from [fp, #-28] to r1
11fed4: e5815000 str r5, [r1] <--- This should crash since it's trying to dereference r1 which is #0
Predicted crash address is 11fed4
Actual crash dump:
signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0x0
r0 ea408800 r1 00000000 r2 00000006 r3 ea6ccfa0
r4 ea6d1b2a r5 00000007 r6 00000000 r7 ea3ff904
r8 00000001 r9 e8b22ec0 sl ea6b5be9 fp ea3ff5c0
ip ea6d1af8 sp ea3ff550 lr ea6d09aa pc ea696ed4 cpsr 600f0010
...
ea577000-ea6d7000 r-xp 00000000 fe:01 1556782 libSomeLib.so
Conclusion:
ea696ed4-ea577000=11fed4
, which is equal to the predicted address.
To summarize: there's no need to add an offset to the PC value in the crash dump (for Android armv7 at least).
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With