Re: [PATCH 13/15] perf intel-pt: Add support for recording AUX area samples

[Arnaldo Carvalho de Melo]

Em Fri, Nov 15, 2019 at 02:42:23PM +0200, Adrian Hunter escreveu:
> Set up the default number of mmap pages, default sample size and default
> psb_period for AUX area sampling. Add documentation also.
> 
> Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
> ---
>  tools/perf/Documentation/intel-pt.txt | 59 ++++++++++++++++++-
>  tools/perf/arch/x86/util/auxtrace.c   |  2 +
>  tools/perf/arch/x86/util/intel-pt.c   | 81 ++++++++++++++++++++++++++-
>  3 files changed, 139 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/perf/Documentation/intel-pt.txt b/tools/perf/Documentation/intel-pt.txt
> index e0d9e7dd4f17..2cf2d9e9d0da 100644
> --- a/tools/perf/Documentation/intel-pt.txt
> +++ b/tools/perf/Documentation/intel-pt.txt
> @@ -434,6 +434,56 @@ pwr_evt		Enable power events.  The power events provide information about
>  		"0" otherwise.
>  
>  
> +AUX area sampling option
> +------------------------
> +
> +To select Intel PT "sampling" the AUX area sampling option can be used:
> +
> +	--aux-sample
> +
> +Optionally it can be followed by the sample size in bytes e.g.
> +
> +	--aux-sample=8192
> +
> +In addition, the Intel PT event to sample must be defined e.g.
> +
> +	-e intel_pt//u
> +
> +Samples on other events will be created containing Intel PT data e.g. the
> +following will create Intel PT samples on the branch-misses event, note the
> +events must be grouped using {}:
> +
> +	perf record --aux-sample -e '{intel_pt//u,branch-misses:u}'
> +
> +An alternative to '--aux-sample' is to add the config term 'aux-sample-size' to
> +events.  In this case, the grouping is implied e.g.
> +
> +	perf record -e intel_pt//u -e branch-misses/aux-sample-size=8192/u
> +
> +is the same as:
> +
> +	perf record -e '{intel_pt//u,branch-misses/aux-sample-size=8192/u}'
> +
> +but allows for also using an address filter e.g.:
> +
> +	perf record -e intel_pt//u --filter 'filter * @/bin/ls' -e branch-misses/aux-sample-size=8192/u -- ls


Can we improve this warning?

[root@quaco ~]# perf record -e intel_pt//u --filter 'filter * @/bin/ls' -e branch-misses/aux-sample-size=8192/u -- ls
Error:
The sys_perf_event_open() syscall returned with 7 (Argument list too long) for event (branch-misses/aux-sample-size=8192/u).
/bin/dmesg | grep -i perf may provide additional information.

[root@quaco ~]#
[root@quaco ~]# uname -a
Linux quaco 5.4.0-rc8 #1 SMP Mon Nov 18 06:15:31 -03 2019 x86_64 x86_64 x86_64 GNU/Linux
[root@quaco ~]#

> +
> +It is important to select a sample size that is big enough to contain at least
> +one PSB packet.  If not a warning will be displayed:
> +
> +	Intel PT sample size (%zu) may be too small for PSB period (%zu)
> +
> +The calculation used for that is: if sample_size <= psb_period + 256 display the
> +warning.  When sampling is used, psb_period defaults to 0 (2KiB).
> +
> +The default sample size is 4KiB.
> +
> +The sample size is passed in aux_sample_size in struct perf_event_attr.  The
> +sample size is limited by the maximum event size which is 64KiB.  It is
> +difficult to know how big the event might be without the trace sample attached,
> +but the tool validates that the sample size is not greater than 60KiB.
> +
> +
>  new snapshot option
>  -------------------
>  
> @@ -487,8 +537,8 @@ their mlock limit (which defaults to 64KiB but is not multiplied by the number
>  of cpus).
>  
>  In full-trace mode, powers of two are allowed for buffer size, with a minimum
> -size of 2 pages.  In snapshot mode, it is the same but the minimum size is
> -1 page.
> +size of 2 pages.  In snapshot mode or sampling mode, it is the same but the
> +minimum size is 1 page.
>  
>  The mmap size and auxtrace mmap size are displayed if the -vv option is used e.g.
>  
> @@ -501,12 +551,17 @@ Intel PT modes of operation
>  
>  Intel PT can be used in 2 modes:
>  	full-trace mode
> +	sample mode
>  	snapshot mode
>  
>  Full-trace mode traces continuously e.g.
>  
>  	perf record -e intel_pt//u uname
>  
> +Sample mode attaches a Intel PT sample to other events e.g.
> +
> +	perf record --aux-sample -e intel_pt//u -e branch-misses:u
> +
>  Snapshot mode captures the available data when a signal is sent e.g.
>  
>  	perf record -v -e intel_pt//u -S ./loopy 1000000000 &
> diff --git a/tools/perf/arch/x86/util/auxtrace.c b/tools/perf/arch/x86/util/auxtrace.c
> index 96f4a2c11893..092543cad324 100644
> --- a/tools/perf/arch/x86/util/auxtrace.c
> +++ b/tools/perf/arch/x86/util/auxtrace.c
> @@ -26,6 +26,8 @@ struct auxtrace_record *auxtrace_record__init_intel(struct evlist *evlist,
>  	bool found_bts = false;
>  
>  	intel_pt_pmu = perf_pmu__find(INTEL_PT_PMU_NAME);
> +	if (intel_pt_pmu)
> +		intel_pt_pmu->auxtrace = true;
>  	intel_bts_pmu = perf_pmu__find(INTEL_BTS_PMU_NAME);
>  
>  	evlist__for_each_entry(evlist, evsel) {
> diff --git a/tools/perf/arch/x86/util/intel-pt.c b/tools/perf/arch/x86/util/intel-pt.c
> index d6d26256915f..20df442fdf36 100644
> --- a/tools/perf/arch/x86/util/intel-pt.c
> +++ b/tools/perf/arch/x86/util/intel-pt.c
> @@ -17,6 +17,7 @@
>  #include "../../util/event.h"
>  #include "../../util/evlist.h"
>  #include "../../util/evsel.h"
> +#include "../../util/evsel_config.h"
>  #include "../../util/cpumap.h"
>  #include "../../util/mmap.h"
>  #include <subcmd/parse-options.h>
> @@ -551,6 +552,43 @@ static int intel_pt_validate_config(struct perf_pmu *intel_pt_pmu,
>  					evsel->core.attr.config);
>  }
>  
> +static void intel_pt_config_sample_mode(struct perf_pmu *intel_pt_pmu,
> +					struct evsel *evsel)
> +{
> +	struct perf_evsel_config_term *term;
> +	u64 user_bits = 0, bits;
> +
> +	term = perf_evsel__get_config_term(evsel, CFG_CHG);
> +	if (term)
> +		user_bits = term->val.cfg_chg;
> +
> +	bits = perf_pmu__format_bits(&intel_pt_pmu->format, "psb_period");
> +
> +	/* Did user change psb_period */
> +	if (bits & user_bits)
> +		return;
> +
> +	/* Set psb_period to 0 */
> +	evsel->core.attr.config &= ~bits;
> +}
> +
> +static void intel_pt_min_max_sample_sz(struct evlist *evlist,
> +				       size_t *min_sz, size_t *max_sz)
> +{
> +	struct evsel *evsel;
> +
> +	evlist__for_each_entry(evlist, evsel) {
> +		size_t sz = evsel->core.attr.aux_sample_size;
> +
> +		if (!sz)
> +			continue;
> +		if (min_sz && (sz < *min_sz || !*min_sz))
> +			*min_sz = sz;
> +		if (max_sz && sz > *max_sz)
> +			*max_sz = sz;
> +	}
> +}
> +
>  /*
>   * Currently, there is not enough information to disambiguate different PEBS
>   * events, so only allow one.
> @@ -606,6 +644,11 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
>  		return -EINVAL;
>  	}
>  
> +	if (opts->auxtrace_snapshot_mode && opts->auxtrace_sample_mode) {
> +		pr_err("Snapshot mode (" INTEL_PT_PMU_NAME " PMU) and sample trace cannot be used together\n");
> +		return -EINVAL;
> +	}
> +
>  	if (opts->use_clockid) {
>  		pr_err("Cannot use clockid (-k option) with " INTEL_PT_PMU_NAME "\n");
>  		return -EINVAL;
> @@ -617,6 +660,9 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
>  	if (!opts->full_auxtrace)
>  		return 0;
>  
> +	if (opts->auxtrace_sample_mode)
> +		intel_pt_config_sample_mode(intel_pt_pmu, intel_pt_evsel);
> +
>  	err = intel_pt_validate_config(intel_pt_pmu, intel_pt_evsel);
>  	if (err)
>  		return err;
> @@ -666,6 +712,34 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
>  				    opts->auxtrace_snapshot_size, psb_period);
>  	}
>  
> +	/* Set default sizes for sample mode */
> +	if (opts->auxtrace_sample_mode) {
> +		size_t psb_period = intel_pt_psb_period(intel_pt_pmu, evlist);
> +		size_t min_sz = 0, max_sz = 0;
> +
> +		intel_pt_min_max_sample_sz(evlist, &min_sz, &max_sz);
> +		if (!opts->auxtrace_mmap_pages && !privileged &&
> +		    opts->mmap_pages == UINT_MAX)
> +			opts->mmap_pages = KiB(256) / page_size;
> +		if (!opts->auxtrace_mmap_pages) {
> +			size_t sz = round_up(max_sz, page_size) / page_size;
> +
> +			opts->auxtrace_mmap_pages = roundup_pow_of_two(sz);
> +		}
> +		if (max_sz > opts->auxtrace_mmap_pages * (size_t)page_size) {
> +			pr_err("Sample size %zu must not be greater than AUX area tracing mmap size %zu\n",
> +			       max_sz,
> +			       opts->auxtrace_mmap_pages * (size_t)page_size);
> +			return -EINVAL;
> +		}
> +		pr_debug2("Intel PT min. sample size: %zu max. sample size: %zu\n",
> +			  min_sz, max_sz);
> +		if (psb_period &&
> +		    min_sz <= psb_period + INTEL_PT_PSB_PERIOD_NEAR)
> +			ui__warning("Intel PT sample size (%zu) may be too small for PSB period (%zu)\n",
> +				    min_sz, psb_period);
> +	}
> +
>  	/* Set default sizes for full trace mode */
>  	if (opts->full_auxtrace && !opts->auxtrace_mmap_pages) {
>  		if (privileged) {
> @@ -682,7 +756,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr,
>  		size_t sz = opts->auxtrace_mmap_pages * (size_t)page_size;
>  		size_t min_sz;
>  
> -		if (opts->auxtrace_snapshot_mode)
> +		if (opts->auxtrace_snapshot_mode || opts->auxtrace_sample_mode)
>  			min_sz = KiB(4);
>  		else
>  			min_sz = KiB(8);
> @@ -1136,5 +1210,10 @@ struct auxtrace_record *intel_pt_recording_init(int *err)
>  	ptr->itr.parse_snapshot_options = intel_pt_parse_snapshot_options;
>  	ptr->itr.reference = intel_pt_reference;
>  	ptr->itr.read_finish = intel_pt_read_finish;
> +	/*
> +	 * Decoding starts at a PSB packet. Minimum PSB period is 2K so 4K
> +	 * should give at least 1 PSB per sample.
> +	 */
> +	ptr->itr.default_aux_sample_size = 4096;
>  	return &ptr->itr;
>  }
> -- 
> 2.17.1

-- 

- Arnaldo