add an equivalent to glibc's PTR_MANGLE for jmp_buf (setjmp/longjmp)

[thestinger]

This wouldn't be very useful for the handler registration functions (atexit, at_quick_exit, pthread_atfork) because the pages they're stored in are read-only when a new handler isn't being registered. The potential for a race is tiny because these functions are called a very limited number of times during initialization phases (start-up, loading a library, calling a library for the first time).

It's the only thing that can be done to protect jmp_buf though, and sadly setjmp / longjmp are widely used...

[thestinger]

This is being implemented upstream. I nudged them in the right direction:

https://android-review.googlesource.com/#/c/170157/

[thestinger]

This is being implemented upstream. I nudged them in the right direction:

https://android-review.googlesource.com/#/c/170157/

[thestinger]

This is now implemented upstream and could just be backported. However, there were quite a few changes to the assembly code after Lollipop so it's going to be a bit painful.

[thestinger]

This is now implemented upstream and could just be backported. However, there were quite a few changes to the assembly code after Lollipop so it's going to be a bit painful.

[thestinger]

This is now backported.

[thestinger]

This is now backported.