[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200810213909.GJ3982@worktop.programming.kicks-ass.net>
Date: Mon, 10 Aug 2020 23:39:09 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Kan Liang <kan.liang@...ux.intel.com>
Cc: acme@...nel.org, mingo@...hat.com, linux-kernel@...r.kernel.org,
mark.rutland@....com, alexander.shishkin@...ux.intel.com,
jolsa@...hat.com, eranian@...gle.com, ak@...ux.intel.com,
dave.hansen@...el.com, kirill.shutemov@...ux.intel.com
Subject: Re: [PATCH V6 01/16] perf/core: Add PERF_SAMPLE_DATA_PAGE_SIZE
On Mon, Aug 10, 2020 at 02:24:21PM -0700, Kan Liang wrote:
> Current perf can report both virtual addresses and physical addresses,
> but not the page size. Without the page size information of the utilized
> page, users cannot decide whether to promote/demote large pages to
> optimize memory usage.
>
> Add a new sample type for the data page size.
>
> Current perf already has a facility to collect data virtual addresses.
> A page walker is required to walk the pages tables and calculate the
> page size from a given virtual address.
>
> On some platforms, e.g., X86, the page walker is invoked in an NMI
> handler. So the page walker must be IRQ-safe and low overhead. Besides,
> the page walker should work for both user and kernel virtual address.
> The existing generic page walker, e.g., walk_page_range_novma(), is a
> little bit complex and doesn't guarantee the IRQ-safe. The follow_page()
> is only for user-virtual address.
>
> Add a new function perf_get_page_size() to walk the page tables and
> calculate the page size. In the function:
> - Interrupts have to be disabled to prevent any teardown of the page
> tables.
> - The size of a normal page is from the pre-defined page size macros.
> - The size of a compound page is retrieved from the helper function,
> page_size().
>
> Suggested-by: Peter Zijlstra <peterz@...radead.org>
> Signed-off-by: Kan Liang <kan.liang@...ux.intel.com>
> /* default value for data source */
> diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
> index 52ca2093831c..32484accc7a3 100644
> --- a/include/uapi/linux/perf_event.h
> +++ b/include/uapi/linux/perf_event.h
> @@ -143,8 +143,9 @@ enum perf_event_sample_format {
> PERF_SAMPLE_PHYS_ADDR = 1U << 19,
> PERF_SAMPLE_AUX = 1U << 20,
> PERF_SAMPLE_CGROUP = 1U << 21,
> + PERF_SAMPLE_DATA_PAGE_SIZE = 1U << 22,
>
> - PERF_SAMPLE_MAX = 1U << 22, /* non-ABI */
> + PERF_SAMPLE_MAX = 1U << 23, /* non-ABI */
>
> __PERF_SAMPLE_CALLCHAIN_EARLY = 1ULL << 63, /* non-ABI; internal use */
> };
> @@ -7151,6 +7269,9 @@ void perf_prepare_sample(struct perf_event_header *header,
> }
> #endif
>
> + if (sample_type & PERF_SAMPLE_DATA_PAGE_SIZE)
> + data->data_page_size = perf_get_page_size(data->addr);
> +
We could just require SAMPLE_DATA_PAGE requires SAMPLE_ADDR.
Powered by blists - more mailing lists