linux-kernel - Re: [PATCH v1] perf parse-events: Make X modifier more respectful of groups

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]
Message-ID: <6ffbfa494002d3ed85a160c2f2a6e1881dfdd304.camel@intel.com>
Date: Fri, 17 Oct 2025 17:54:58 +0000
From: "Falcon, Thomas" <thomas.falcon@...el.com>
To: "alexander.shishkin@...ux.intel.com" <alexander.shishkin@...ux.intel.com>,
	"linux-perf-users@...r.kernel.org" <linux-perf-users@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"peterz@...radead.org" <peterz@...radead.org>, "acme@...nel.org"
	<acme@...nel.org>, "mingo@...hat.com" <mingo@...hat.com>, "Hunter, Adrian"
	<adrian.hunter@...el.com>, "namhyung@...nel.org" <namhyung@...nel.org>,
	"jolsa@...nel.org" <jolsa@...nel.org>, "dapeng1.mi@...ux.intel.com"
	<dapeng1.mi@...ux.intel.com>, "irogers@...gle.com" <irogers@...gle.com>
Subject: Re: [PATCH v1] perf parse-events: Make X modifier more respectful of
 groups

On Fri, 2025-10-17 at 09:57 -0700, Ian Rogers wrote:
> Events with an X modifier were reordered within a group, for example
> slots was made the leader in:
> ```
> $ perf record -e '{cpu/mem-stores/ppu,cpu/slots/uX}' -- sleep 1
> ```
> 
> Fix by making `dont_regroup` evsels always use their index for
> sorting. Make the cur_leader, when fixing the groups, be that of
> `dont_regroup` evsel so that the `dont_regroup` evsel doesn't become
> a
> leader.
> 
> On a tigerlake this patch corrects this and meets expectations in:
> ```
> $ perf stat -e '{cpu/mem-stores/,cpu/slots/uX}' -a -- sleep 0.1
> 
>  Performance counter stats for 'system wide':
> 
>         83,458,652      cpu/mem-stores/
>      2,720,854,880      cpu/slots/uX
> 
>        0.103780587 seconds time elapsed
> 
> $ perf stat -e 'slots,slots:X' -a -- sleep 0.1
> 
>  Performance counter stats for 'system wide':
> 
>        732,042,247      slots                (48.96%)
>        643,288,155      slots:X              (51.04%)
> 
>        0.102731018 seconds time elapsed
> ```
> 

Hi Ian, just confirming the results on an Alder Lake:

% sudo ./perf stat -e '{cpu_core/mem-stores/,cpu_core/slots/uX}' -a --
sleep 0.1

 Performance counter stats for 'system wide':

        23,097,346      cpu_core/mem-stores/
       666,910,974      cpu_core/slots/uX                 

       0.101369982 seconds time elapsed

 % sudo ./perf stat -e 'slots,slots:X' -a -- sleep 0.1  

 Performance counter stats for 'system wide':

     1,775,555,180      cpu_core/slots/                               
(50.08%)
     1,910,073,872      cpu_core/slots/X                              
(49.92%)

       0.101368283 seconds time elapsed

Tested-by: Thomas Falcon <thomas.falcon@...el.com>

Thanks,
Tom

> Closes:
> https://lore.kernel.org/lkml/18f20d38-070c-4e17-bc90-cf7102e1e53d@linux.intel.com/
> Fixes: 035c17893082 ("perf parse-events: Add 'X' modifier to exclude
> an event from being regrouped")
> Signed-off-by: Ian Rogers <irogers@...gle.com>
> ---
>  tools/perf/util/parse-events.c | 16 ++++++++++------
>  1 file changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-
> events.c
> index 3aec86aebdc6..1a5da93f4094 100644
> --- a/tools/perf/util/parse-events.c
> +++ b/tools/perf/util/parse-events.c
> @@ -1973,14 +1973,18 @@ static int evlist__cmp(void *_fg_idx, const
> struct list_head *l, const struct li
>  	 * event's index is used. An index may be forced for events
> that
>  	 * must be in the same group, namely Intel topdown events.
>  	 */
> -	if (*force_grouped_idx != -1 &&
> arch_evsel__must_be_in_group(lhs)) {
> +	if (lhs->dont_regroup) {
> +		lhs_sort_idx = lhs_core->idx;
> +	} else if (*force_grouped_idx != -1 &&
> arch_evsel__must_be_in_group(lhs)) {
>  		lhs_sort_idx = *force_grouped_idx;
>  	} else {
>  		bool lhs_has_group = lhs_core->leader != lhs_core ||
> lhs_core->nr_members > 1;
>  
>  		lhs_sort_idx = lhs_has_group ? lhs_core->leader->idx
> : lhs_core->idx;
>  	}
> -	if (*force_grouped_idx != -1 &&
> arch_evsel__must_be_in_group(rhs)) {
> +	if (rhs->dont_regroup) {
> +		rhs_sort_idx = rhs_core->idx;
> +	} else if (*force_grouped_idx != -1 &&
> arch_evsel__must_be_in_group(rhs)) {
>  		rhs_sort_idx = *force_grouped_idx;
>  	} else {
>  		bool rhs_has_group = rhs_core->leader != rhs_core ||
> rhs_core->nr_members > 1;
> @@ -2078,10 +2082,10 @@ static int
> parse_events__sort_events_and_fix_groups(struct list_head *list)
>  	 */
>  	idx = 0;
>  	list_for_each_entry(pos, list, core.node) {
> -		const struct evsel *pos_leader = evsel__leader(pos);
> +		struct evsel *pos_leader = evsel__leader(pos);
>  		const char *pos_pmu_name = pos->group_pmu_name;
>  		const char *cur_leader_pmu_name;
> -		bool pos_force_grouped = force_grouped_idx != -1 &&
> +		bool pos_force_grouped = force_grouped_idx != -1 &&
> !pos->dont_regroup &&
>  			arch_evsel__must_be_in_group(pos);
>  
>  		/* Reset index and nr_members. */
> @@ -2095,8 +2099,8 @@ static int
> parse_events__sort_events_and_fix_groups(struct list_head *list)
>  		 * groups can't span PMUs.
>  		 */
>  		if (!cur_leader || pos->dont_regroup) {
> -			cur_leader = pos;
> -			cur_leaders_grp = &pos->core;
> +			cur_leader = pos_leader;
> +			cur_leaders_grp = &cur_leader->core;
>  			if (pos_force_grouped)
>  				force_grouped_leader = pos;
>  		}