intel_put_shared_regs_event_constraints(cpuc, event);
 }
 
-static int intel_pmu_hw_config(struct perf_event *event)
+static void intel_pebs_aliases_core2(struct perf_event *event)
 {
-       int ret = x86_pmu_hw_config(event);
-
-       if (ret)
-               return ret;
-
-       if (event->attr.precise_ip &&
-           (event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) {
+       if ((event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) {
                /*
                 * Use an alternative encoding for CPU_CLK_UNHALTED.THREAD_P
                 * (0x003c) so that we can use it with PEBS.
                 */
                u64 alt_config = X86_CONFIG(.event=0xc0, .inv=1, .cmask=16);
 
+               alt_config |= (event->hw.config & ~X86_RAW_EVENT_MASK);
+               event->hw.config = alt_config;
+       }
+}
+
+static void intel_pebs_aliases_snb(struct perf_event *event)
+{
+       if ((event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) {
+               /*
+                * Use an alternative encoding for CPU_CLK_UNHALTED.THREAD_P
+                * (0x003c) so that we can use it with PEBS.
+                *
+                * The regular CPU_CLK_UNHALTED.THREAD_P event (0x003c) isn't
+                * PEBS capable. However we can use UOPS_RETIRED.ALL
+                * (0x01c2), which is a PEBS capable event, to get the same
+                * count.
+                *
+                * UOPS_RETIRED.ALL counts the number of cycles that retires
+                * CNTMASK micro-ops. By setting CNTMASK to a value (16)
+                * larger than the maximum number of micro-ops that can be
+                * retired per cycle (4) and then inverting the condition, we
+                * count all cycles that retire 16 or less micro-ops, which
+                * is every cycle.
+                *
+                * Thereby we gain a PEBS capable cycle counter.
+                */
+               u64 alt_config = X86_CONFIG(.event=0xc2, .umask=0x01, .inv=1, .cmask=16);
 
                alt_config |= (event->hw.config & ~X86_RAW_EVENT_MASK);
                event->hw.config = alt_config;
        }
+}
+
+static int intel_pmu_hw_config(struct perf_event *event)
+{
+       int ret = x86_pmu_hw_config(event);
+
+       if (ret)
+               return ret;
+
+       if (event->attr.precise_ip && x86_pmu.pebs_aliases)
+               x86_pmu.pebs_aliases(event);
 
        if (intel_pmu_needs_lbr_smpl(event)) {
                ret = intel_pmu_setup_lbr_filter(event);
        .max_period             = (1ULL << 31) - 1,
        .get_event_constraints  = intel_get_event_constraints,
        .put_event_constraints  = intel_put_event_constraints,
+       .pebs_aliases           = intel_pebs_aliases_core2,
 
        .format_attrs           = intel_arch3_formats_attr,
 
 
                x86_pmu.event_constraints = intel_snb_event_constraints;
                x86_pmu.pebs_constraints = intel_snb_pebs_event_constraints;
+               x86_pmu.pebs_aliases = intel_pebs_aliases_snb;
                x86_pmu.extra_regs = intel_snb_extra_regs;
                /* all extra regs are per-cpu when HT is on */
                x86_pmu.er_flags |= ERF_HAS_RSP_1;