lkml.org 
[lkml]   [2020]   [Jul]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patches in this message
/
Date
From
SubjectRe: [PATCH 4/4] x86/cpu: Use SERIALIZE in sync_core() when available
On Mon, Jul 27, 2020 at 05:49:28AM -0700, hpa@zytor.com wrote:
> On July 27, 2020 1:36:19 AM PDT, peterz@infradead.org wrote:
> >On Sun, Jul 26, 2020 at 10:55:15PM -0700, hpa@zytor.com wrote:
> >> For a really overenginered solution, but which might perform
> >> unnecessary poorly on existing hardware:
> >>
> >> asm volatile("1: .byte 0xf, 0x1, 0xe8; 2:"
> >> _ASM_EXTABLE(1b,2b));
> >
> >Ha! cute, you take an #UD ?
> >
> >We could optimize the #UD exception handler for this I suppose, but
> >that makes it an even worse hack. The simple alternative() seems like
> >a much simpler approach.
>
> If this is in any way performance critical, then no :)

Yeah, I'm not sure.. the 'funny' thing is that typically call
sync_core() from an IPI anyway. And the synchronous broadcast IPI is by
far the most expensive part of that.

Something like this...

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 20e07feb4064..528e049ee1d9 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -989,12 +989,13 @@ void *text_poke_kgdb(void *addr, const void *opcode, size_t len)

static void do_sync_core(void *info)
{
- sync_core();
+ /* IRET implies sync_core() */
}

void text_poke_sync(void)
{
on_each_cpu(do_sync_core, NULL, 1);
+ sync_core();
}

struct text_poke_loc {

> Taking the #UD
> has the cute property that we end up IRET on the way back, so we don't
> even need a fix-up path.

I got that, what I had in mind was making sure #UD avoids the overhead
of doing exception entry/exit by adding an early exit.

Something like so:

diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index 8493f55e1167..a3f41d645944 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -96,6 +96,16 @@ __always_inline int is_valid_bugaddr(unsigned long addr)
return *(unsigned short *)addr == INSN_UD2;
}

+__always_inline int handle_serialize(struct pt_regs *regs)
+{
+ const char serialize[3] = { 0x0f, 0xe8, 0x02 };
+
+ if (regs->ip < TASK_SIZE_MAX)
+ return 0;
+
+ return !memcmp((const void *)regs->ip, serialize, 3);
+}
+
static nokprobe_inline int
do_trap_no_signal(struct task_struct *tsk, int trapnr, const char *str,
struct pt_regs *regs, long error_code)
@@ -252,8 +262,13 @@ DEFINE_IDTENTRY_RAW(exc_invalid_op)
* handle it before exception entry to avoid recursive WARN
* in case exception entry is the one triggering WARNs.
*/
- if (!user_mode(regs) && handle_bug(regs))
- return;
+ if (!user_mode(regs)) {
+ if (handle_bug(regs))
+ return;
+
+ if (handle_serialize(regs))
+ return;
+ }

state = idtentry_enter(regs);
instrumentation_begin();
\
 
 \ /
  Last update: 2020-07-27 15:06    [W:0.095 / U:0.108 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site