On Mon, 31 Jul 2017 12:21:54 +0200
Peter Zijlstra <[email protected]> wrote:
> So I was looking at text_poke_bp() today and I couldn't make sense of
> the barriers there.
>
> How's for something like so?
>
> ---
> arch/x86/kernel/alternative.c | 22 ++++++++++++++++------
> 1 file changed, 16 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index 32e14d137416..3344d3382e91 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -742,7 +742,16 @@ static void *bp_int3_handler, *bp_int3_addr;
>
> int poke_int3_handler(struct pt_regs *regs)
> {
> - /* bp_patching_in_progress */
> + /*
> + * Having observed our INT3 instruction, we now must observe
> + * bp_patching_in_progress.
> + *
> + * in_progress = TRUE INT3
> + * WMB RMB
> + * write INT3 if (in_progress)
> + *
> + * Idem for bp_int3_handler.
> + */
Looks correct.
> smp_rmb();
>
> if (likely(!bp_patching_in_progress))
> @@ -788,9 +797,8 @@ void *text_poke_bp(void *addr, const void *opcode, size_t
> len, void *handler)
> bp_int3_addr = (u8 *)addr + sizeof(int3);
> bp_patching_in_progress = true;
> /*
> - * Corresponding read barrier in int3 notifier for
> - * making sure the in_progress flags is correctly ordered wrt.
> - * patching
> + * Corresponding read barrier in int3 notifier for making sure the
> + * in_progress and handler are correctly ordered wrt. patching.
> */
This looks correct as well.
> smp_wmb();
>
> @@ -815,9 +823,11 @@ void *text_poke_bp(void *addr, const void *opcode,
> size_t len, void *handler)
> text_poke(addr, opcode, sizeof(int3));
>
> on_each_cpu(do_sync_core, NULL, 1);
> -
> + /*
> + * sync_core() implies an smp_mb() and orders this store against
> + * the writing of the new instruction.
> + */
Yep.
> bp_patching_in_progress = false;
> - smp_wmb();
Heh, I think this was a "lets not leak bp_patching_in_progress" out of
this function. But I don't see any harm if it happens.
As this function was a *very* slow path, that smp_wmb() was a "it's not
really needed, but it wont hurt anything to slap it in there just in
case".
Reviewed-by: Steven Rostedt (VMware) <[email protected]>
-- Steve
>
> return addr;
> }
