| 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
| ||
|
Message-ID: <20200715211330.GR183694@krava>
Date: Wed, 15 Jul 2020 23:13:30 +0200
From: Jiri Olsa <jolsa@...hat.com>
To: Ian Rogers <irogers@...gle.com>
Cc: Jiri Olsa <jolsa@...nel.org>,
Arnaldo Carvalho de Melo <acme@...nel.org>,
lkml <linux-kernel@...r.kernel.org>,
Ingo Molnar <mingo@...nel.org>,
Namhyung Kim <namhyung@...nel.org>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Peter Zijlstra <a.p.zijlstra@...llo.nl>,
Michael Petlan <mpetlan@...hat.com>,
Andi Kleen <ak@...ux.intel.com>,
Kajol Jain <kjain@...ux.ibm.com>,
John Garry <john.garry@...wei.com>,
"Paul A. Clarke" <pc@...ibm.com>,
Stephane Eranian <eranian@...gle.com>
Subject: Re: [PATCH 09/18] perf metric: Collect referenced metrics in struct
metric_ref_node
On Tue, Jul 14, 2020 at 05:07:04PM -0700, Ian Rogers wrote:
> On Sun, Jul 12, 2020 at 6:27 AM Jiri Olsa <jolsa@...nel.org> wrote:
> >
> > Collecting referenced metrics in struct metric_ref_node object,
> > so we can process them later on.
> >
> > The change will parse nested metric names out of expression and
> > 'resolve' them.
> >
> > All referenced metrics are dissolved into one context, meaning all
> > nested metrics events and added to the parent context.
> >
> > Signed-off-by: Jiri Olsa <jolsa@...nel.org>
> > ---
> > tools/perf/util/metricgroup.c | 147 ++++++++++++++++++++++++++++++----
> > 1 file changed, 132 insertions(+), 15 deletions(-)
> >
> > diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> > index f0b0a053bfd2..9923eef1e2d4 100644
> > --- a/tools/perf/util/metricgroup.c
> > +++ b/tools/perf/util/metricgroup.c
> > @@ -102,12 +102,20 @@ void metricgroup__rblist_exit(struct rblist *metric_events)
> > rblist__exit(metric_events);
> > }
> >
> > +struct metric_ref_node {
> > + const char *metric_name;
> > + const char *metric_expr;
> > + struct list_head list;
> > +};
>
> Perhaps a comment like this above:
> /* A node in the list of referenced metrics. metric_expr is held as a
> convenience to avoid a search through the metric list. */
sounds good
>
> > +
> > struct egroup {
> > struct list_head nd;
> > struct expr_parse_ctx pctx;
> > const char *metric_name;
> > const char *metric_expr;
> > const char *metric_unit;
> > + struct list_head refs;
> > + int refs_cnt;
>
> A comment would be nice here as refs is a pretty overloaded term. For
> example, is refs_cnt a reference count for memory management or the
> length of the refs list? I'm unpopular and would use a long variable
> name here of something like referenced_metrics for the list.
how about metric_refs ? ;-)
>
> > int runtime;
> > bool has_constraint;
> > };
> > @@ -574,27 +582,66 @@ int __weak arch_get_runtimeparam(void)
> > static int __add_metric(struct list_head *group_list,
> > struct pmu_event *pe,
> > bool metric_no_group,
> > - int runtime)
> > + int runtime,
> > + struct egroup **egp)
>
> Rather than egp perhaps parent_eg or parent_metric?
I'm changing that later on, but perhaps not to parent_metric,
which sounds good, I'll check
>
> > {
> > + struct metric_ref_node *ref;
> > struct egroup *eg;
> >
> > - eg = malloc(sizeof(*eg));
> > - if (!eg)
> > - return -ENOMEM;
> > + if (*egp == NULL) {
> > + /*
> > + * We got in here for the master group,
> > + * allocate it and put it on the list.
> > + */
> > + eg = malloc(sizeof(*eg));
> > + if (!eg)
> > + return -ENOMEM;
> > +
> > + expr__ctx_init(&eg->pctx);
> > + eg->metric_name = pe->metric_name;
> > + eg->metric_expr = pe->metric_expr;
> > + eg->metric_unit = pe->unit;
> > + eg->runtime = runtime;
> > + eg->has_constraint = metric_no_group || metricgroup__has_constraint(pe);
> > + INIT_LIST_HEAD(&eg->refs);
> > + eg->refs_cnt = 0;
> > + *egp = eg;
> > + } else {
> > + /*
> > + * We got here for the referenced metric, via the
> > + * recursive metricgroup__add_metric call, add
> > + * it to the master group.
>
> Probably want to avoid the term master here and below:
> https://www.kernel.org/doc/html/latest/process/coding-style.html?highlight=language%20master#naming
> Perhaps parent or referencing metric?
right.. will change
>
> > + */
> > + eg = *egp;
> > +
> > + ref = malloc(sizeof(*ref));
> > + if (!ref)
> > + return -ENOMEM;
> >
> > - expr__ctx_init(&eg->pctx);
> > - eg->metric_name = pe->metric_name;
> > - eg->metric_expr = pe->metric_expr;
> > - eg->metric_unit = pe->unit;
> > - eg->runtime = runtime;
> > - eg->has_constraint = metric_no_group || metricgroup__has_constraint(pe);
> > + ref->metric_name = pe->metric_name;
> > + ref->metric_expr = pe->metric_expr;
> > + list_add(&ref->list, &eg->refs);
> > + eg->refs_cnt++;
> > + eg->has_constraint |= metricgroup__has_constraint(pe);
>
> Why not metric_no_group here? Perhaps use a function to avoid
> duplication with the code above.
oops, I think that has_constraint update should not be there at al,
just for the 'parent' metric, thanks!
>
> > + }
> >
> > + /*
> > + * For both the master and referenced metrics, we parse
> > + * all the metric's IDs and add it to the parent context.
> > + */
> > if (expr__find_other(pe->metric_expr, NULL, &eg->pctx, runtime) < 0) {
> > expr__ctx_clear(&eg->pctx);
> > free(eg);
> > return -EINVAL;
> > }
> >
> > + /*
> > + * We add new group only in the 'master' call,
> > + * so bail out for referenced metric case.
> > + */
> > + if (eg->refs_cnt)
> > + return 0;
> > +
> > if (list_empty(group_list))
> > list_add(&eg->nd, group_list);
> > else {
> > @@ -636,14 +683,63 @@ static struct pmu_event *find_metric(const char *metric, struct pmu_events_map *
> >
> > static int add_metric(struct list_head *group_list,
> > struct pmu_event *pe,
> > - bool metric_no_group)
> > + bool metric_no_group,
> > + struct egroup **egp);
> > +
> > +static int resolve_metric(struct egroup *eg,
> > + bool metric_no_group,
> > + struct list_head *group_list,
> > + struct pmu_events_map *map)
> > +{
> > + struct hashmap_entry *cur;
> > + size_t bkt;
> > + bool all;
> > + int ret;
> > +
> > + /*
> > + * Iterate all the parsed IDs and if there's metric,
> > + * add it to the context.
> > + */
>
> Does this mean that the ID doesn't need to begin "metric:" to
> reference a different metric as per Andi Kleen's request?
correct, that's why I separated find_metric function earlier,
now I see I did not put this to the changelog :-\ sry
>
> > + do {
> > + all = true;
> > + hashmap__for_each_entry((&eg->pctx.ids), cur, bkt) {
> > + struct pmu_event *pe;
> > +
> > + pe = find_metric(cur->key, map);
> > + if (!pe)
> > + continue;
> > +
> > + all = false;
> > + /* The metric key itself needs to go out.. */
> > + expr__del_id(&eg->pctx, cur->key);
> > +
> > + /* ... and it gets resolved to the parent context. */
> > + ret = add_metric(group_list, pe, metric_no_group, &eg);
> > + if (ret)
> > + return ret;
> > +
> > + /*
> > + * We added new metric to hashmap, so we need
> > + * to break the iteration and start over.
> > + */
> > + break;
> > + }
> > + } while (!all);
> > +
> > + return 0;
> > +}
SNIP
Powered by blists - more mailing lists