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: <20180629132724.ztcfelcnb2dt2u35@lakrids.cambridge.arm.com>
Date:   Fri, 29 Jun 2018 14:27:25 +0100
From:   Mark Rutland <mark.rutland@....com>
To:     Suzuki K Poulose <suzuki.poulose@....com>
Cc:     linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        will.deacon@....com, robin.murphy@....com, julien.thierry@....com
Subject: Re: [PATCH v3 4/7] arm_pmu: Tidy up clear_event_idx call backs

On Tue, Jun 19, 2018 at 11:15:39AM +0100, Suzuki K Poulose wrote:
> The armpmu uses get_event_idx callback to allocate an event
> counter for a given event, which marks the selected counter
> as "used". Now, when we delete the counter, the arm_pmu goes
> ahead and clears the "used" bit and then invokes the "clear_event_idx"
> call back, which kind of splits the job between the core code
> and the backend. Tidy this up by relying on the clear_event_idx
> to do the book keeping, if available. Otherwise, let the core
> driver do the default "clear" bit operation. This will be useful
> for adding the chained event support, where we leave the event
> idx maintenance to the backend.
> 
> Also, when an event is removed from the PMU, reset the hw.idx
> to indicate that a counter is not allocated for this event,
> to help the backends do better checks. This will be also used
> for the chain counter support.
> 
> Cc: Mark Rutland <mark.rutland@....com>
> Cc: Will Deacon <will.deacon@....com>
> Reviewed-by: Julien Thierry <julien.thierry@....com>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@....com>

TBH, I think it might be better to make clear_event_idx() mandatory,
such that each backend provides both clear / set, next to each other in
the code.

It's a bit more code, but it would restore orthogonality to the API.

Mark.

> ---
> Changes since v2:
>  - Reset the event counter after an event is removed.
> ---
>  arch/arm/kernel/perf_event_v7.c |  2 ++
>  drivers/perf/arm_pmu.c          | 17 +++++++++++++----
>  2 files changed, 15 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c
> index fd7ce01..765d265 100644
> --- a/arch/arm/kernel/perf_event_v7.c
> +++ b/arch/arm/kernel/perf_event_v7.c
> @@ -1637,6 +1637,7 @@ static void krait_pmu_clear_event_idx(struct pmu_hw_events *cpuc,
>  	bool venum_event = EVENT_VENUM(hwc->config_base);
>  	bool krait_event = EVENT_CPU(hwc->config_base);
>  
> +	clear_bit(hwc->idx, cpuc->used_mask);
>  	if (venum_event || krait_event) {
>  		bit = krait_event_to_bit(event, region, group);
>  		clear_bit(bit, cpuc->used_mask);
> @@ -1966,6 +1967,7 @@ static void scorpion_pmu_clear_event_idx(struct pmu_hw_events *cpuc,
>  	bool venum_event = EVENT_VENUM(hwc->config_base);
>  	bool scorpion_event = EVENT_CPU(hwc->config_base);
>  
> +	clear_bit(hwc->idx, cpuc->used_mask);
>  	if (venum_event || scorpion_event) {
>  		bit = scorpion_event_to_bit(event, region, group);
>  		clear_bit(bit, cpuc->used_mask);
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index e3766a8..6e10e8c 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -227,6 +227,16 @@ static void armpmu_start(struct perf_event *event, int flags)
>  	armpmu->enable(event);
>  }
>  
> +static void armpmu_clear_event_idx(struct arm_pmu *armpmu,
> +				   struct pmu_hw_events *hw_events,
> +				   struct perf_event *event)
> +{
> +	if (armpmu->clear_event_idx)
> +		armpmu->clear_event_idx(hw_events, event);
> +	else
> +		clear_bit(event->hw.idx, hw_events->used_mask);
> +}
> +
>  static void
>  armpmu_del(struct perf_event *event, int flags)
>  {
> @@ -237,11 +247,10 @@ armpmu_del(struct perf_event *event, int flags)
>  
>  	armpmu_stop(event, PERF_EF_UPDATE);
>  	hw_events->events[idx] = NULL;
> -	clear_bit(idx, hw_events->used_mask);
> -	if (armpmu->clear_event_idx)
> -		armpmu->clear_event_idx(hw_events, event);
> -
> +	armpmu_clear_event_idx(armpmu, hw_events, event);
>  	perf_event_update_userpage(event);
> +	/* Clear the allocated counter */
> +	hwc->idx = -1;
>  }
>  
>  static int
> -- 
> 2.7.4
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ