Re: [ABOMINATION] x86: Fast interrupt return to userspace

[Linus Torvalds]

On Tue, May 6, 2014 at 2:00 PM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> I'll do profiles and test the kernel compile too, but the raw timings
> are certainly promising. The "sysret" hack is pretty disgusting, and
> it's broken too. sysret doesn't do some things iret does (like TF flag
> etc), so it's not complete, but it's clearly good enough to run tests
> on. It will definitely break ptrace() and friends.

It clearly breaks other things too, and there seems to be bugs in
there. I got this, for example:

WARNING: CPU: 2 PID: 1174 at kernel/smp.c:230
smp_call_function_single+0x81/0xa0()
CPU: 2 PID: 1174 Comm: gdbus Tainted: G        W
3.15.0-rc4-00260-g38583f095c5a-dirty #2
Call Trace:
  dump_stack+0x45/0x56
  warn_slowpath_common+0x73/0x90
  warn_slowpath_null+0x15/0x20
  smp_call_function_single+0x81/0xa0
  smp_call_function_many+0x21c/0x260
  flush_tlb_page+0x6d/0xb0
  ptep_clear_flush+0x2c/0x40
  do_wp_page+0x208/0x6e0
  handle_mm_fault+0x79c/0x1060
  __do_page_fault+0x15e/0x510
  do_page_fault+0xc/0x10
  page_fault+0x22/0x30
  retint_swapgs+0x6/0x10

which is because interrupts are disabled in that
install_sysret_trampoline() path that touches user space. It's
possibly sufficient to just move the "cli" to below the call to it. I
eventually ended up with a hung machine, possibly related to this,
possibly something else.

Whatever. I got enough profile data to say that it seems to have cut
'iret' overhead by at least two thirds. So it may not *work*, but from
a "hey look, some random numbers" standpoint it is worth playing with.

             Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/