objtool: Fix ORC 'signal' propagation
authorJosh Poimboeuf <jpoimboe@kernel.org>
Thu, 16 Feb 2023 20:34:41 +0000 (12:34 -0800)
committerIngo Molnar <mingo@kernel.org>
Thu, 23 Feb 2023 08:21:50 +0000 (09:21 +0100)
commit00c8f01c4e84637c3db76f368b8687cb61f4dd9d
tree54c3288a6bcc0bbae8b13dc9c132791987f8ddc1
parent1c34496e5856886d565665fb64029ecdeb080ffb
objtool: Fix ORC 'signal' propagation

There have been some recently reported ORC unwinder warnings like:

  WARNING: can't access registers at entry_SYSCALL_64_after_hwframe+0x63/0xcd
  WARNING: stack going in the wrong direction? at __sys_setsockopt+0x2c6/0x5b0 net/socket.c:2271

And a KASAN warning:

  BUG: KASAN: stack-out-of-bounds in unwind_next_frame (arch/x86/include/asm/ptrace.h:136 arch/x86/kernel/unwind_orc.c:455)

It turns out the 'signal' bit isn't getting propagated from the unwind
hints to the ORC entries, making the unwinder confused at times.

Fixes: ffb1b4a41016 ("x86/unwind/orc: Add 'signal' field to ORC metadata")
Reported-by: kernel test robot <oliver.sang@intel.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/97eef9db60cd86d376a9a40d49d77bb67a8f6526.1676579666.git.jpoimboe@kernel.org
tools/objtool/check.c
tools/objtool/include/objtool/cfi.h
tools/objtool/orc_gen.c