问题
I was examining some ARM crashes on Android and realised some memory addresses positions are not 2-byte aligned (THUMB/ARM) but some are. Does any one know why is that so? Here is an example tombstone:
Stack frame #00 pc 00072c9a /system/lib/libdvm.so (dvmAbort+170)
Stack frame #01 pc 00050f6b /system/lib/libdvm.so (ScopedCheck::checkInstanceFieldID(_jobject*, _jfieldID*)+379)
Stack frame #02 pc 0005db56 /system/lib/libdvm.so (Check_SetBooleanField(_JNIEnv*, _jobject*, _jfieldID*, unsigned char)+150)
Stack frame #03 pc 00003a93 /system/lib/libmytestt.so (Java_com_test_jni_MyLibWrapper_getMonitorSample+259)
Stack frame #04 pc 0002d850 /system/lib/libdvm.so (dvmPlatformInvoke+80)
Stack frame #05 pc 00085b48 /system/lib/libdvm.so (dvmCallJNIMethod(unsigned int const*, JValue*, Method const*, Thread*)+648)
Stack frame #06 pc 000505db /system/lib/libdvm.so (dvmCheckCallJNIMethod(unsigned int const*, JValue*, Method const*, Thread*)+59)
Stack frame #07 pc 0008b96d /system/lib/libdvm.so (dvmResolveNativeMethod(unsigned int const*, JValue*, Method const*, Thread*)+317)
Stack frame #08 pc 00170ff1 /system/lib/libdvm.so
One can see that #00, #02 are 2 byte aligned while #01 and #03 are not. Any explanation of why is it so? Thanks.
回答1:
I would guess that the odd addresses indicate that those functions are in THUMB mode, while the even addresses are in ARM mode. When returning from a function with bx lr, the processor will switch from ARM mode to THUMB or vice versa depending on the lowest bit of the address, so I'd guess this is what you end up seeing in the backtrace as well.
来源：https://stackoverflow.com/questions/29381889/core-dump-addresses-not-byte-aligned