perf intel-pt: Fix IPC with CYC threshold

The code assumed every CYC-eligible packet has a CYC packet, which is not
the case when CYC thresholds are used. Fix by checking if a CYC packet is
actually present in that case.

Fixes: 5b1dc0fd ("perf intel-pt: Add support for samples to contain IPC ratio")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/r/20210205175350.23817-4-adrian.hunter@intel.comSigned-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/util/intel-pt-decoder/intel-pt-decoder.c

@@ -24,6 +24,13 @@
 #include "intel-pt-decoder.h"
 #include "intel-pt-log.h"
 
+#define BITULL(x) (1ULL << (x))
+
+/* IA32_RTIT_CTL MSR bits */
+#define INTEL_PT_CYC_ENABLE		BITULL(1)
+#define INTEL_PT_CYC_THRESHOLD		(BITULL(22) | BITULL(21) | BITULL(20) | BITULL(19))
+#define INTEL_PT_CYC_THRESHOLD_SHIFT	19
+
 #define INTEL_PT_BLK_SIZE 1024
 
 #define BIT63 (((uint64_t)1 << 63))
@@ -167,6 +174,8 @@ struct intel_pt_decoder {
 	uint64_t sample_tot_cyc_cnt;
 	uint64_t base_cyc_cnt;
 	uint64_t cyc_cnt_timestamp;
+	uint64_t ctl;
+	uint64_t cyc_threshold;
 	double tsc_to_cyc;
 	bool continuous_period;
 	bool overflow;
@@ -204,6 +213,14 @@ static uint64_t intel_pt_lower_power_of_2(uint64_t x)
 	return x << i;
 }
 
+static uint64_t intel_pt_cyc_threshold(uint64_t ctl)
+{
+	if (!(ctl & INTEL_PT_CYC_ENABLE))
+		return 0;
+
+	return (ctl & INTEL_PT_CYC_THRESHOLD) >> INTEL_PT_CYC_THRESHOLD_SHIFT;
+}
+
 static void intel_pt_setup_period(struct intel_pt_decoder *decoder)
 {
 	if (decoder->period_type == INTEL_PT_PERIOD_TICKS) {
@@ -245,12 +262,15 @@ struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
 
 	decoder->flags              = params->flags;
 
+	decoder->ctl                = params->ctl;
 	decoder->period             = params->period;
 	decoder->period_type        = params->period_type;
 
 	decoder->max_non_turbo_ratio    = params->max_non_turbo_ratio;
 	decoder->max_non_turbo_ratio_fp = params->max_non_turbo_ratio;
 
+	decoder->cyc_threshold = intel_pt_cyc_threshold(decoder->ctl);
+
 	intel_pt_setup_period(decoder);
 
 	decoder->mtc_shift = params->mtc_period;
@@ -2017,6 +2037,7 @@ static int intel_pt_hop_trace(struct intel_pt_decoder *decoder, bool *no_tip, in
 
 static int intel_pt_walk_trace(struct intel_pt_decoder *decoder)
 {
+	int last_packet_type = INTEL_PT_PAD;
 	bool no_tip = false;
 	int err;
 
@@ -2025,6 +2046,12 @@ static int intel_pt_walk_trace(struct intel_pt_decoder *decoder)
 		if (err)
 			return err;
 next:
+		if (decoder->cyc_threshold) {
+			if (decoder->sample_cyc && last_packet_type != INTEL_PT_CYC)
+				decoder->sample_cyc = false;
+			last_packet_type = decoder->packet.type;
+		}
+
 		if (decoder->hop) {
 			switch (intel_pt_hop_trace(decoder, &no_tip, &err)) {
 			case HOP_IGNORE:

tools/perf/util/intel-pt-decoder/intel-pt-decoder.h

@@ -244,6 +244,7 @@ struct intel_pt_params {
 	void *data;
 	bool return_compression;
 	bool branch_enable;
+	uint64_t ctl;
 	uint64_t period;
 	enum intel_pt_period_type period_type;
 	unsigned max_non_turbo_ratio;

tools/perf/util/intel-pt.c

@@ -893,6 +893,18 @@ static bool intel_pt_sampling_mode(struct intel_pt *pt)
 	return false;
 }
 
+static u64 intel_pt_ctl(struct intel_pt *pt)
+{
+	struct evsel *evsel;
+	u64 config;
+
+	evlist__for_each_entry(pt->session->evlist, evsel) {
+		if (intel_pt_get_config(pt, &evsel->core.attr, &config))
+			return config;
+	}
+	return 0;
+}
+
 static u64 intel_pt_ns_to_ticks(const struct intel_pt *pt, u64 ns)
 {
 	u64 quot, rem;
@@ -1026,6 +1038,7 @@ static struct intel_pt_queue *intel_pt_alloc_queue(struct intel_pt *pt,
 	params.data = ptq;
 	params.return_compression = intel_pt_return_compression(pt);
 	params.branch_enable = intel_pt_branch_enable(pt);
+	params.ctl = intel_pt_ctl(pt);
 	params.max_non_turbo_ratio = pt->max_non_turbo_ratio;
 	params.mtc_period = intel_pt_mtc_period(pt);
 	params.tsc_ctc_ratio_n = pt->tsc_ctc_ratio_n;