linux-kernel - Re: [PATCH 1/2] perf, x86: Implement event scheduler helper functions

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 14 Sep 2011 16:43:28 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Robert Richter <robert.richter@....com>
Cc:	Ingo Molnar <mingo@...e.hu>, Stephane Eranian <eranian@...gle.com>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/2] perf, x86: Implement event scheduler helper
 functions

On Sat, 2011-09-10 at 16:49 +0200, Robert Richter wrote:
> This patch introduces x86 perf scheduler code helper functions. We
> need this to later add more complex functionality to support
> overlapping counter constraints (next patch).
> 
> The algorithm is modified so that the range of weight values is now
> generated from the constraints. There shouldn't be other functional
> changes.
> 
> With the helper functions the scheduler is controlled. There are
> functions to initialize, traverse the event list, find unused counters
> etc. The scheduler keeps its own state.
> 
> Cc: Stephane Eranian <eranian@...gle.com>
> Signed-off-by: Robert Richter <robert.richter@....com>
> ---
>  arch/x86/kernel/cpu/perf_event.c |  158 +++++++++++++++++++++++++-------------
>  1 files changed, 105 insertions(+), 53 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
> index 594d425..44ec767 100644
> --- a/arch/x86/kernel/cpu/perf_event.c
> +++ b/arch/x86/kernel/cpu/perf_event.c
> @@ -790,18 +790,118 @@ static inline int is_x86_event(struct perf_event *event)
>  	return event->pmu == &pmu;
>  }
>  
> +struct sched_state {
> +	int	weight;
> +	int	event;
> +	int	counter;
> +	int	unassigned;
> +	unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
> +};

Maybe add a few comments here? Took me a while to figure out unassigned
is the number of unassigned events.

> +static struct sched_state *perf_sched_find_counter(struct perf_sched *sched)
> +{
> +	struct event_constraint *c;
> +	int idx;
> +
> +	if (!sched->state.unassigned)
> +		return NULL;

So bail when we're done and there's nothing left to assign.

> +	c = sched->constraints[sched->state.event];
> +
> +	idx = sched->state.counter;

Which is typically 0, but this could be a restart, at which point we
continue looking where we left off.

> +	/* for each bit in idxmsk starting from idx */
> +	while (idx < X86_PMC_IDX_MAX) {
> +		idx = find_next_bit(c->idxmsk, X86_PMC_IDX_MAX, idx);
> +		if (idx == X86_PMC_IDX_MAX)
> +			break;
> +		if (!__test_and_set_bit(idx, sched->state.used))
> +			break;
> +		idx++;
> +	}

#define for_each_set_bit_continue(bit, addr, size) 	\
	for( ; (bit) < (size);				\
	       (bit) = find_next_bit((addr), (size), (bit) + 1))

	for_each_set_bit_continue(idx, c->idxmask, X86_PMC_IDX_MAX) {
		if (!__test_and_set_bit(idx, sched->state.used))
			break;
	}

> +	sched->state.counter = idx;
> +
> +	if (idx >= X86_PMC_IDX_MAX)
> +		return NULL;

OK, so its important to assign idx to counter even if we're too big,
because of the restart, right? That wants a comment.

> +
> +	return &sched->state;
> +}
> +
> +static int perf_sched_next_event(struct perf_sched *sched)
> +{
> +	struct event_constraint *c;
> +
> +	if (!sched->state.unassigned || !--sched->state.unassigned)
> +		return 0;

Shouldn't we avoid getting here if there's nothing to do? I get
the !--unassigned case, but am a bit puzzled by the !unassigned case.

> +	do {
> +		/* next event */
> +		sched->state.event++;
> +		if (sched->state.event >= sched->max_events) {
> +			/* next weight */
> +			sched->state.event = 0;
> +			sched->state.weight++;
> +			if (sched->state.weight > sched->max_weight)
> +				return 0;
> +		}
> +		c = sched->constraints[sched->state.event];
> +	} while (c->weight != sched->state.weight);
> +
> +	sched->state.counter = 0;	/* start with first counter */
> +
> +	return 1;
> +}

fair enough..

Looks ok otherwise, just a tad hard to grok in one go.. a few comments
could go a long way. I'm sure I'll have forgotten how it works in a few
weeks.

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