proc: try to remove use of FOLL_FORCE entirely
authorLinus Torvalds <torvalds@linux-foundation.org>
Tue, 9 May 2017 15:45:16 +0000 (08:45 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 9 May 2017 15:45:16 +0000 (08:45 -0700)
commit8ee74a91ac304ad2e9e794eb96ed76f0634dec40
tree923bffc9af9f033a87e7e1ed52e3b2226ddd27cc
parent2868b2513aa732a99ea4a0a6bf10dc93c1f3dac2
proc: try to remove use of FOLL_FORCE entirely

We fixed the bugs in it, but it's still an ugly interface, so let's see
if anybody actually depends on it.  It's entirely possible that nothing
actually requires the whole "punch through read-only mappings"
semantics.

For example, gdb definitely uses the /proc/<pid>/mem interface, but it
looks like it mainly does it for regular reads of the target (that don't
need FOLL_FORCE), and looking at the gdb source code seems to fall back
on the traditional ptrace(PTRACE_POKEDATA) interface if it needs to.

If this breaks something, I do have a (more complex) version that only
enables FOLL_FORCE when somebody has PTRACE_ATTACH'ed to the target,
like the comment here used to say ("Maybe we should limit FOLL_FORCE to
actual ptrace users?").

Cc: Kees Cook <keescook@chromium.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Eric Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/base.c