Re: apparent regressions from TLB range flushing page set

From: Jan Beulich
Date: Wed Aug 22 2012 - 03:44:38 EST


>>> Alex Shi <alex.shi@xxxxxxxxx> 08/22/12 5:27 AM >>>
>> Second, the UV code doesn't flush the full range at all, it simply
>> ignores its 'end' parameter (and hence also the "all" indicator).
>
>Sure. the following rfc patch try to fix it. untested since no hardware.

Sure - this needs to be looked at by a person knowing UV (and I would
have thought a change like the one we're discussing here would also
have required an ack from such a person), but ...

>--- a/arch/x86/platform/uv/tlb_uv.c
>+++ b/arch/x86/platform/uv/tlb_uv.c
>@@ -280,12 +280,12 @@ static void bau_process_message(struct msg_desc *mdp, struct bau_control *bcp,
> /*
> * This must be a normal message, or retry of a normal message
> */
>- if (msg->address == TLB_FLUSH_ALL) {
>+ if (msg->end == 0) {

How would "end" end up being 0 here? Don't you rather mean "start and
end on the same page"? And even if you do, aren't you then losing the
intended optimization?

>+ __flush_tlb_one(msg->start);
>+ stat->d_onetlb++;
>+ } else {
> local_flush_tlb();
> stat->d_alltlb++;
>- } else {
>- __flush_tlb_one(msg->address);
>- stat->d_onetlb++;
> }
> stat->d_requestee++;
>
>@@ -1113,7 +1114,8 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
>
> record_send_statistics(stat, locals, hubs, remotes, bau_desc);
>
>- bau_desc->payload.address = start;
>+ bau_desc->payload.start = start;
>+ bau_desc->payload.end = end;
> bau_desc->payload.sending_cpu = cpu;
> /*
> * uv_flush_send_and_wait returns 0 if all cpu's were messaged,

Jan

--
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/