perf/x86/intel: Add support for PEBSv3 profiling
authorAndi Kleen <ak@linux.intel.com>
Sun, 10 May 2015 19:22:40 +0000 (12:22 -0700)
committerIngo Molnar <mingo@kernel.org>
Tue, 4 Aug 2015 08:16:56 +0000 (10:16 +0200)
PEBSv3 is the same as the existing PEBSv2 used on Haswell,
but it adds a new TSC field. Add support to the generic
PEBS handler to handle the new format, and overwrite
the perf time stamp using the new native_sched_clock_from_tsc().

Right now the time stamp is just slightly more accurate,
as it is nearer the actual event trigger point. With
the PEBS threshold > 1 patchkit it will be much more accurate,
avoid the problems with MMAP mismatches earlier.
The accurate time stamping is only implemented for
the default trace clock for now.

v2: Use _skl prefix. Check for default clock_id.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1431285767-27027-3-git-send-email-andi@firstfloor.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/kernel/cpu/perf_event_intel_ds.c

index 71fc402..410270a 100644 (file)
@@ -224,6 +224,19 @@ union hsw_tsx_tuning {
 
 #define PEBS_HSW_TSX_FLAGS     0xff00000000ULL
 
+/* Same as HSW, plus TSC */
+
+struct pebs_record_skl {
+       u64 flags, ip;
+       u64 ax, bx, cx, dx;
+       u64 si, di, bp, sp;
+       u64 r8,  r9,  r10, r11;
+       u64 r12, r13, r14, r15;
+       u64 status, dla, dse, lat;
+       u64 real_ip, tsx_tuning;
+       u64 tsc;
+};
+
 void init_debug_store_on_cpu(int cpu)
 {
        struct debug_store *ds = per_cpu(cpu_hw_events, cpu).ds;
@@ -885,7 +898,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
        return 0;
 }
 
-static inline u64 intel_hsw_weight(struct pebs_record_hsw *pebs)
+static inline u64 intel_hsw_weight(struct pebs_record_skl *pebs)
 {
        if (pebs->tsx_tuning) {
                union hsw_tsx_tuning tsx = { .value = pebs->tsx_tuning };
@@ -894,7 +907,7 @@ static inline u64 intel_hsw_weight(struct pebs_record_hsw *pebs)
        return 0;
 }
 
-static inline u64 intel_hsw_transaction(struct pebs_record_hsw *pebs)
+static inline u64 intel_hsw_transaction(struct pebs_record_skl *pebs)
 {
        u64 txn = (pebs->tsx_tuning & PEBS_HSW_TSX_FLAGS) >> 32;
 
@@ -918,7 +931,7 @@ static void setup_pebs_sample_data(struct perf_event *event,
         * unconditionally access the 'extra' entries.
         */
        struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
-       struct pebs_record_hsw *pebs = __pebs;
+       struct pebs_record_skl *pebs = __pebs;
        u64 sample_type;
        int fll, fst, dsrc;
        int fl = event->hw.flags;
@@ -1016,6 +1029,16 @@ static void setup_pebs_sample_data(struct perf_event *event,
                        data->txn = intel_hsw_transaction(pebs);
        }
 
+       /*
+        * v3 supplies an accurate time stamp, so we use that
+        * for the time stamp.
+        *
+        * We can only do this for the default trace clock.
+        */
+       if (x86_pmu.intel_cap.pebs_format >= 3 &&
+               event->attr.use_clockid == 0)
+               data->time = native_sched_clock_from_tsc(pebs->tsc);
+
        if (has_branch_stack(event))
                data->br_stack = &cpuc->lbr_stack;
 }
@@ -1245,6 +1268,13 @@ void __init intel_ds_init(void)
                        x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
                        break;
 
+               case 3:
+                       pr_cont("PEBS fmt3%c, ", pebs_type);
+                       x86_pmu.pebs_record_size =
+                                               sizeof(struct pebs_record_skl);
+                       x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
+                       break;
+
                default:
                        printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type);
                        x86_pmu.pebs = 0;