perf/x86/intel: Introduce setup_pebs_sample_data()
authorYan, Zheng <zheng.z.yan@intel.com>
Wed, 6 May 2015 19:33:48 +0000 (15:33 -0400)
committerIngo Molnar <mingo@kernel.org>
Sun, 7 Jun 2015 14:08:40 +0000 (16:08 +0200)
Move code that sets up the PEBS sample data to a separate function.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-3-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/kernel/cpu/perf_event_intel_ds.c

index 4802d5dec2222ea6951afc28a2b5b7bc0a4f5827..a5fe561c4902765f528f23024a8754814e31b968 100644 (file)
@@ -853,8 +853,10 @@ static inline u64 intel_hsw_transaction(struct pebs_record_hsw *pebs)
        return txn;
 }
 
-static void __intel_pmu_pebs_event(struct perf_event *event,
-                                  struct pt_regs *iregs, void *__pebs)
+static void setup_pebs_sample_data(struct perf_event *event,
+                                  struct pt_regs *iregs, void *__pebs,
+                                  struct perf_sample_data *data,
+                                  struct pt_regs *regs)
 {
 #define PERF_X86_EVENT_PEBS_HSW_PREC \
                (PERF_X86_EVENT_PEBS_ST_HSW | \
@@ -866,30 +868,25 @@ static void __intel_pmu_pebs_event(struct perf_event *event,
         */
        struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
        struct pebs_record_hsw *pebs = __pebs;
-       struct perf_sample_data data;
-       struct pt_regs regs;
        u64 sample_type;
        int fll, fst, dsrc;
        int fl = event->hw.flags;
 
-       if (!intel_pmu_save_and_restart(event))
-               return;
-
        sample_type = event->attr.sample_type;
        dsrc = sample_type & PERF_SAMPLE_DATA_SRC;
 
        fll = fl & PERF_X86_EVENT_PEBS_LDLAT;
        fst = fl & (PERF_X86_EVENT_PEBS_ST | PERF_X86_EVENT_PEBS_HSW_PREC);
 
-       perf_sample_data_init(&data, 0, event->hw.last_period);
+       perf_sample_data_init(data, 0, event->hw.last_period);
 
-       data.period = event->hw.last_period;
+       data->period = event->hw.last_period;
 
        /*
         * Use latency for weight (only avail with PEBS-LL)
         */
        if (fll && (sample_type & PERF_SAMPLE_WEIGHT))
-               data.weight = pebs->lat;
+               data->weight = pebs->lat;
 
        /*
         * data.data_src encodes the data source
@@ -902,7 +899,7 @@ static void __intel_pmu_pebs_event(struct perf_event *event,
                        val = precise_datala_hsw(event, pebs->dse);
                else if (fst)
                        val = precise_store_data(pebs->dse);
-               data.data_src.val = val;
+               data->data_src.val = val;
        }
 
        /*
@@ -915,58 +912,70 @@ static void __intel_pmu_pebs_event(struct perf_event *event,
         * PERF_SAMPLE_IP and PERF_SAMPLE_CALLCHAIN to function properly.
         * A possible PERF_SAMPLE_REGS will have to transfer all regs.
         */
-       regs = *iregs;
-       regs.flags = pebs->flags;
-       set_linear_ip(&regs, pebs->ip);
-       regs.bp = pebs->bp;
-       regs.sp = pebs->sp;
+       *regs = *iregs;
+       regs->flags = pebs->flags;
+       set_linear_ip(regs, pebs->ip);
+       regs->bp = pebs->bp;
+       regs->sp = pebs->sp;
 
        if (sample_type & PERF_SAMPLE_REGS_INTR) {
-               regs.ax = pebs->ax;
-               regs.bx = pebs->bx;
-               regs.cx = pebs->cx;
-               regs.dx = pebs->dx;
-               regs.si = pebs->si;
-               regs.di = pebs->di;
-               regs.bp = pebs->bp;
-               regs.sp = pebs->sp;
-
-               regs.flags = pebs->flags;
+               regs->ax = pebs->ax;
+               regs->bx = pebs->bx;
+               regs->cx = pebs->cx;
+               regs->dx = pebs->dx;
+               regs->si = pebs->si;
+               regs->di = pebs->di;
+               regs->bp = pebs->bp;
+               regs->sp = pebs->sp;
+
+               regs->flags = pebs->flags;
 #ifndef CONFIG_X86_32
-               regs.r8 = pebs->r8;
-               regs.r9 = pebs->r9;
-               regs.r10 = pebs->r10;
-               regs.r11 = pebs->r11;
-               regs.r12 = pebs->r12;
-               regs.r13 = pebs->r13;
-               regs.r14 = pebs->r14;
-               regs.r15 = pebs->r15;
+               regs->r8 = pebs->r8;
+               regs->r9 = pebs->r9;
+               regs->r10 = pebs->r10;
+               regs->r11 = pebs->r11;
+               regs->r12 = pebs->r12;
+               regs->r13 = pebs->r13;
+               regs->r14 = pebs->r14;
+               regs->r15 = pebs->r15;
 #endif
        }
 
        if (event->attr.precise_ip > 1 && x86_pmu.intel_cap.pebs_format >= 2) {
-               regs.ip = pebs->real_ip;
-               regs.flags |= PERF_EFLAGS_EXACT;
-       } else if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(&regs))
-               regs.flags |= PERF_EFLAGS_EXACT;
+               regs->ip = pebs->real_ip;
+               regs->flags |= PERF_EFLAGS_EXACT;
+       } else if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(regs))
+               regs->flags |= PERF_EFLAGS_EXACT;
        else
-               regs.flags &= ~PERF_EFLAGS_EXACT;
+               regs->flags &= ~PERF_EFLAGS_EXACT;
 
        if ((sample_type & PERF_SAMPLE_ADDR) &&
            x86_pmu.intel_cap.pebs_format >= 1)
-               data.addr = pebs->dla;
+               data->addr = pebs->dla;
 
        if (x86_pmu.intel_cap.pebs_format >= 2) {
                /* Only set the TSX weight when no memory weight. */
                if ((sample_type & PERF_SAMPLE_WEIGHT) && !fll)
-                       data.weight = intel_hsw_weight(pebs);
+                       data->weight = intel_hsw_weight(pebs);
 
                if (sample_type & PERF_SAMPLE_TRANSACTION)
-                       data.txn = intel_hsw_transaction(pebs);
+                       data->txn = intel_hsw_transaction(pebs);
        }
 
        if (has_branch_stack(event))
-               data.br_stack = &cpuc->lbr_stack;
+               data->br_stack = &cpuc->lbr_stack;
+}
+
+static void __intel_pmu_pebs_event(struct perf_event *event,
+                                  struct pt_regs *iregs, void *__pebs)
+{
+       struct perf_sample_data data;
+       struct pt_regs regs;
+
+       if (!intel_pmu_save_and_restart(event))
+               return;
+
+       setup_pebs_sample_data(event, iregs, __pebs, &data, &regs);
 
        if (perf_event_overflow(event, &data, &regs))
                x86_pmu_stop(event, 0);