Re: [PATCH 02/19] perf/hisilicon: Fix group validation

On Wed, Aug 13, 2025 at 06:00:54PM +0100, Robin Murphy wrote:
> The group validation logic shared by the HiSilicon HNS3/PCIe drivers is
> a bit off, in that given a software group leader, it will consider that
> event *in place of* the actual new event being opened. At worst this
> could theoretically allow an unschedulable group if the software event
> config happens to look like one of the hardware siblings.
> 
> The uncore framework avoids that particular issue,

What is "the uncore framework"? I'm not sure exactly what you're
referring to, nor how that composes with the problem described above.

> but all 3 also share the common issue of not preventing racy access to
> the sibling list,

Can you please elaborate on this racy access to the silbing list? I'm
not sure exactly what you're referring to.

> and some redundant checks which can be cleaned up.
> 
> Signed-off-by: Robin Murphy <[email protected]>
> ---
>  drivers/perf/hisilicon/hisi_pcie_pmu.c   | 17 ++++++-----------
>  drivers/perf/hisilicon/hisi_uncore_pmu.c | 23 +++++++----------------
>  drivers/perf/hisilicon/hns3_pmu.c        | 17 ++++++-----------
>  3 files changed, 19 insertions(+), 38 deletions(-)
> 
> diff --git a/drivers/perf/hisilicon/hisi_pcie_pmu.c b/drivers/perf/hisilicon/hisi_pcie_pmu.c
> index c5394d007b61..3b0b2f7197d0 100644
> --- a/drivers/perf/hisilicon/hisi_pcie_pmu.c
> +++ b/drivers/perf/hisilicon/hisi_pcie_pmu.c
> @@ -338,21 +338,16 @@ static bool hisi_pcie_pmu_validate_event_group(struct perf_event *event)
>  	int counters = 1;
>  	int num;
>  
> -	event_group[0] = leader;
> -	if (!is_software_event(leader)) {
> -		if (leader->pmu != event->pmu)
> -			return false;
> +	if (leader == event)
> +		return true;
>  
> -		if (leader != event && !hisi_pcie_pmu_cmp_event(leader, event))
> -			event_group[counters++] = event;
> -	}
> +	event_group[0] = event;
> +	if (leader->pmu == event->pmu && !hisi_pcie_pmu_cmp_event(leader, event))
> +		event_group[counters++] = leader;

Looking at this, the existing logic to share counters (which
hisi_pcie_pmu_cmp_event() is trying to permit) looks to be bogus, given
that the start/stop callbacks will reprogram the HW counters (and hence
can fight with one another).

I suspect that can be removed *entirely*, and this can be simplified
down to allocating N counters, without a quadratic event comparison.  We
don't try to share counters in other PMU drivers, and there was no
rationale for trying to do this when this wa introduced in commit:

  8404b0fbc7fbd42e ("drivers/perf: hisi: Add driver for HiSilicon PCIe PMU")

The 'link' tag in that comment goes to v13, which doesn't link to prior
postings, so I'm not going to dig further.

Mark.

>  
>  	for_each_sibling_event(sibling, event->group_leader) {
> -		if (is_software_event(sibling))
> -			continue;
> -
>  		if (sibling->pmu != event->pmu)
> -			return false;
> +			continue;
>  
>  		for (num = 0; num < counters; num++) {
>  			/*
> diff --git a/drivers/perf/hisilicon/hisi_uncore_pmu.c b/drivers/perf/hisilicon/hisi_uncore_pmu.c
> index a449651f79c9..3c531b36cf25 100644
> --- a/drivers/perf/hisilicon/hisi_uncore_pmu.c
> +++ b/drivers/perf/hisilicon/hisi_uncore_pmu.c
> @@ -101,26 +101,17 @@ static bool hisi_validate_event_group(struct perf_event *event)
>  	/* Include count for the event */
>  	int counters = 1;
>  
> -	if (!is_software_event(leader)) {
> -		/*
> -		 * We must NOT create groups containing mixed PMUs, although
> -		 * software events are acceptable
> -		 */
> -		if (leader->pmu != event->pmu)
> -			return false;
> +	if (leader == event)
> +		return true;
>  
> -		/* Increment counter for the leader */
> -		if (leader != event)
> -			counters++;
> -	}
> +	/* Increment counter for the leader */
> +	if (leader->pmu == event->pmu)
> +		counters++;
>  
>  	for_each_sibling_event(sibling, event->group_leader) {
> -		if (is_software_event(sibling))
> -			continue;
> -		if (sibling->pmu != event->pmu)
> -			return false;
>  		/* Increment counter for each sibling */
> -		counters++;
> +		if (sibling->pmu == event->pmu)
> +			counters++;
>  	}
>  
>  	/* The group can not count events more than the counters in the HW */
> diff --git a/drivers/perf/hisilicon/hns3_pmu.c b/drivers/perf/hisilicon/hns3_pmu.c
> index c157f3572cae..382e469257f9 100644
> --- a/drivers/perf/hisilicon/hns3_pmu.c
> +++ b/drivers/perf/hisilicon/hns3_pmu.c
> @@ -1058,21 +1058,16 @@ static bool hns3_pmu_validate_event_group(struct perf_event *event)
>  	int counters = 1;
>  	int num;
>  
> -	event_group[0] = leader;
> -	if (!is_software_event(leader)) {
> -		if (leader->pmu != event->pmu)
> -			return false;
> +	if (leader == event)
> +		return true;
>  
> -		if (leader != event && !hns3_pmu_cmp_event(leader, event))
> -			event_group[counters++] = event;
> -	}
> +	event_group[0] = event;
> +	if (leader->pmu == event->pmu && !hns3_pmu_cmp_event(leader, event))
> +		event_group[counters++] = leader;
>  
>  	for_each_sibling_event(sibling, event->group_leader) {
> -		if (is_software_event(sibling))
> -			continue;
> -
>  		if (sibling->pmu != event->pmu)
> -			return false;
> +			continue;
>  
>  		for (num = 0; num < counters; num++) {
>  			/*
> -- 
> 2.39.2.101.g768bb238c484.dirty
>