On Wed, Nov 14, 2018 at 06:11:39PM +0000, Will Deacon wrote: [...] > On Fri, Oct 19, 2018 at 12:24:04PM +0100, Will Deacon wrote: [...] > > FWIW: I think we should be entertaining a prctl() interface to use a new > > key on a per-thread basis. Obviously, this would need to be used with care > > (e.g. you'd fork(); use the prctl() and then you'd better not return from > > the calling function!). > > > > Assuming we want this (Kees -- I was under the impression that everything in > > Android would end up with the same key otherwise?), then the question is > > do we want: > > > > - prctl() get/set operations for the key, or > > - prctl() set_random_key operation, or > > - both of the above? > > > > Part of the answer to that may lie in the requirements of CRIU, where I > > strongly suspect they need explicit get/set operations, although these > > could be gated on CONFIG_CHECKPOINT_RESTORE=y. > > I managed to speak to the CRIU developers at LPC. The good news is that > their preference is for a ptrace()-based interface for getting and setting > the keys, so the only prctl() operation we need is to set a random key > (separately for A and B). That's good if it works for them, and it seems the cleaner approach. _If_ they run the new thread up to a checkpoint, restoring the memory and doing all the setup that requires in-thread syscalls, then stop it in ptrace to finally inject the regs, then it makes sense to set the keys at that stop -- i.e., you set the keys atomically* with the final setting of the thread's PC. (* with respect to the target thread) So long as you're confident they've understood the implications of ptrauth for CRIU, I guess this can work. (In other news, they will also need to do some work to support SVE, but that's unrelated to ptrauth.) Cheers ---Dave
