Jann Horn <ja...@google.com> writes: > I'm actually trying to get rid of the ->mm access in > __ptrace_may_access() entirely by moving the dumpability and the > user_ns into the signal_struct, but I don't have patches for that > ready (yet).
Do you have a plan for dealing with old linux-threads style threads where you have two processes that share the same mm, but have different signal_structs. I don't think it is required to share any other structures except mm_struct when you share mm_struct. Maybe sighand_struct. Not to derail your idea. Only needing to look at signal_struct sounds very nice. I just know we have some other somewhat bizarre cases the kernel still supports. Eric
