x86/entry: 32-bit C exit conversion and code deletion

[Resend, this time to lkml, too. Sigh, I'm bad at this email thing.]

As promised, here's the 32-bit code. Lightly tested. Brian, thanks for cleaning up the vm86 mess.

I can imagine some objections to patch 1. I wonder how many users actually care about 32-bit audit performance. It ought to be relatively straightforward to come up with a clean opportunistic sysexit implementation for 32-bit kernels, which would fix the performance regression (mostly, at least) and would speed up lots of other workloads.

Are these okay as is, or is opportunistic sysexit a prerequisite?

Applies to tip/x86/asm.

P.S. Denys, I think we should do opportunistic sysretl on 64-bit kernels. That would be much nicer on top of your sysexit cleanup series.

Andy Lutomirski (3):
x86/entry/32: Remove 32-bit syscall audit optimizations x86/entry/32: Migrate to C exit path
x86/entry: Remove do_notify_resume, syscall_trace_leave, and their TIF
masks

 arch/x86/entry/common.c            |  57 --------------------
 arch/x86/entry/entry_32.S          | 108 +++++--------------------------------
 arch/x86/include/asm/ptrace.h      |   1 -
 arch/x86/include/asm/signal.h      |   1 -
 arch/x86/include/asm/thread_info.h |  16 ------
 5 files changed, 12 insertions(+), 171 deletions(-)

2.4.3