[PATCH 2/8] perf/amd/ibs: Remove pointless sample period check

Ravi Bangoria posted 8 patches 1 month, 3 weeks ago
[PATCH 2/8] perf/amd/ibs: Remove pointless sample period check
Posted by Ravi Bangoria 1 month, 3 weeks ago
Valid perf event sample period value for IBS PMUs (Fetch and Op both)
is limited to multiple of 0x10. perf_ibs_init() has this check:

  if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
          return -EINVAL;

But it's broken since hwc->sample_period will always be 0 when
event->attr.sample_freq is 0 (irrespective of event->attr.freq value.)

One option to fix this is to change the condition:

  - if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
  + if (!event->attr.freq && hwc->sample_period & 0x0f)

However, that will break all userspace tools which have been using IBS
event with sample_period not multiple of 0x10.

Another option is to remove the condition altogether and mask lower
nibble _silently_, same as what current code is inadvertently doing.
I'm preferring this approach as it keeps the existing behavior.

Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
---
 arch/x86/events/amd/ibs.c | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/arch/x86/events/amd/ibs.c b/arch/x86/events/amd/ibs.c
index 347353b9eb70..6b55a8520166 100644
--- a/arch/x86/events/amd/ibs.c
+++ b/arch/x86/events/amd/ibs.c
@@ -294,13 +294,8 @@ static int perf_ibs_init(struct perf_event *event)
 		if (config & perf_ibs->cnt_mask)
 			/* raw max_cnt may not be set */
 			return -EINVAL;
-		if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
-			/*
-			 * lower 4 bits can not be set in ibs max cnt,
-			 * but allowing it in case we adjust the
-			 * sample period to set a frequency.
-			 */
-			return -EINVAL;
+
+		/* Silently mask off lower nibble. IBS hw mandates it. */
 		hwc->sample_period &= ~0x0FULL;
 		if (!hwc->sample_period)
 			hwc->sample_period = 0x10;
-- 
2.46.2
Re: [PATCH 2/8] perf/amd/ibs: Remove pointless sample period check
Posted by Namhyung Kim 1 month, 3 weeks ago
Hello,

On Mon, Oct 07, 2024 at 03:48:04AM +0000, Ravi Bangoria wrote:
> Valid perf event sample period value for IBS PMUs (Fetch and Op both)
> is limited to multiple of 0x10. perf_ibs_init() has this check:
> 
>   if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
>           return -EINVAL;
> 
> But it's broken since hwc->sample_period will always be 0 when
> event->attr.sample_freq is 0 (irrespective of event->attr.freq value.)

Right, hwc->sample_period is set to event->attr.sample_period and it's
the same as sample_freq since there are in a union.

struct perf_event_attr {
	__u32			type;
	__u32			size;
	__u64			config;
	union {
		__u64		sample_period;
		__u64		sample_freq;
	};
	...

perf_event_alloc() sets the hwc->sample_period and changes it only when
both attr->freq and attr->sample_freq are not zero.

> 
> One option to fix this is to change the condition:
> 
>   - if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
>   + if (!event->attr.freq && hwc->sample_period & 0x0f)

Right, I believe this is the intention.

> 
> However, that will break all userspace tools which have been using IBS
> event with sample_period not multiple of 0x10.

Correct.

> 
> Another option is to remove the condition altogether and mask lower
> nibble _silently_, same as what current code is inadvertently doing.
> I'm preferring this approach as it keeps the existing behavior.

Agreed.  The condition never worked and should be safe to remove.

Thanks,
Namhyung

> 
> Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
> ---
>  arch/x86/events/amd/ibs.c | 9 ++-------
>  1 file changed, 2 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/events/amd/ibs.c b/arch/x86/events/amd/ibs.c
> index 347353b9eb70..6b55a8520166 100644
> --- a/arch/x86/events/amd/ibs.c
> +++ b/arch/x86/events/amd/ibs.c
> @@ -294,13 +294,8 @@ static int perf_ibs_init(struct perf_event *event)
>  		if (config & perf_ibs->cnt_mask)
>  			/* raw max_cnt may not be set */
>  			return -EINVAL;
> -		if (!event->attr.sample_freq && hwc->sample_period & 0x0f)
> -			/*
> -			 * lower 4 bits can not be set in ibs max cnt,
> -			 * but allowing it in case we adjust the
> -			 * sample period to set a frequency.
> -			 */
> -			return -EINVAL;
> +
> +		/* Silently mask off lower nibble. IBS hw mandates it. */
>  		hwc->sample_period &= ~0x0FULL;
>  		if (!hwc->sample_period)
>  			hwc->sample_period = 0x10;
> -- 
> 2.46.2
>