@@ -326,8 +326,7 @@ static int riscv_pmu_event_init(struct perf_event *event)
u64 event_config = 0;
uint64_t cmask;
- /* driver does not support branch stack sampling */
- if (has_branch_stack(event))
+ if (has_branch_stack(event) && !riscv_pmu_ctr_supported(rvpmu))
return -EOPNOTSUPP;
hwc->flags = 0;
@@ -675,7 +675,7 @@ static void pmu_sched_task(struct perf_event_pmu_context *pmu_ctx,
{
struct riscv_pmu *pmu = to_riscv_pmu(pmu_ctx->pmu);
- /* Call CTR specific Sched hook. */
+ riscv_pmu_ctr_sched_task(pmu_ctx, sched_in);
}
static int rvpmu_sbi_find_num_ctrs(void)
@@ -935,17 +935,25 @@ static irqreturn_t rvpmu_ovf_handler(int irq, void *dev)
hw_evt = &event->hw;
riscv_pmu_event_update(event);
perf_sample_data_init(&data, 0, hw_evt->last_period);
- if (riscv_pmu_event_set_period(event)) {
- /*
- * Unlike other ISAs, RISC-V don't have to disable interrupts
- * to avoid throttling here. As per the specification, the
- * interrupt remains disabled until the OF bit is set.
- * Interrupts are enabled again only during the start.
- * TODO: We will need to stop the guest counters once
- * virtualization support is added.
- */
- perf_event_overflow(event, &data, regs);
+ if (!riscv_pmu_event_set_period(event))
+ continue;
+
+ if (needs_branch_stack(event)) {
+ riscv_pmu_ctr_consume(cpu_hw_evt, event);
+ perf_sample_save_brstack(
+ &data, event,
+ &cpu_hw_evt->branches->branch_stack, NULL);
}
+
+ /*
+ * Unlike other ISAs, RISC-V don't have to disable interrupts
+ * to avoid throttling here. As per the specification, the
+ * interrupt remains disabled until the OF bit is set.
+ * Interrupts are enabled again only during the start.
+ * TODO: We will need to stop the guest counters once
+ * virtualization support is added.
+ */
+ perf_event_overflow(event, &data, regs);
}
rvpmu_start_overflow_mask(pmu, overflowed_ctrs);
@@ -1103,10 +1111,12 @@ static void rvpmu_ctr_start(struct perf_event *event, u64 ival)
else
rvpmu_sbi_ctr_start(event, ival);
-
if ((hwc->flags & PERF_EVENT_FLAG_USER_ACCESS) &&
(hwc->flags & PERF_EVENT_FLAG_USER_READ_CNT))
rvpmu_set_scounteren((void *)event);
+
+ if (needs_branch_stack(event))
+ riscv_pmu_ctr_enable(event);
}
static void rvpmu_ctr_stop(struct perf_event *event, unsigned long flag)
@@ -1128,6 +1138,9 @@ static void rvpmu_ctr_stop(struct perf_event *event, unsigned long flag)
} else {
rvpmu_sbi_ctr_stop(event, flag);
}
+
+ if (needs_branch_stack(event))
+ riscv_pmu_ctr_disable(event);
}
static int rvpmu_find_ctrs(void)
@@ -1161,6 +1174,9 @@ static int rvpmu_find_ctrs(void)
static int rvpmu_event_map(struct perf_event *event, u64 *econfig)
{
+ if (needs_branch_stack(event) && !riscv_pmu_ctr_valid(event))
+ return -EOPNOTSUPP;
+
if (static_branch_likely(&riscv_pmu_cdeleg_available) && !pmu_sbi_is_fw_event(event))
return rvpmu_deleg_event_map(event, econfig);
else
@@ -1207,6 +1223,8 @@ static int rvpmu_starting_cpu(unsigned int cpu, struct hlist_node *node)
enable_percpu_irq(riscv_pmu_irq, IRQ_TYPE_NONE);
}
+ riscv_pmu_ctr_starting_cpu();
+
return 0;
}
@@ -1218,6 +1236,7 @@ static int rvpmu_dying_cpu(unsigned int cpu, struct hlist_node *node)
/* Disable all counters access for user mode now */
csr_write(CSR_SCOUNTEREN, 0x0);
+ riscv_pmu_ctr_dying_cpu();
return 0;
}
@@ -1331,6 +1350,29 @@ static void riscv_pmu_destroy(struct riscv_pmu *pmu)
cpuhp_state_remove_instance(CPUHP_AP_PERF_RISCV_STARTING, &pmu->node);
}
+static int branch_records_alloc(struct riscv_pmu *pmu)
+{
+ struct branch_records __percpu *tmp_alloc_ptr;
+ struct branch_records *records;
+ struct cpu_hw_events *events;
+ int cpu;
+
+ if (!riscv_pmu_ctr_supported(pmu))
+ return 0;
+
+ tmp_alloc_ptr = alloc_percpu_gfp(struct branch_records, GFP_KERNEL);
+ if (!tmp_alloc_ptr)
+ return -ENOMEM;
+
+ for_each_possible_cpu(cpu) {
+ events = per_cpu_ptr(pmu->hw_events, cpu);
+ records = per_cpu_ptr(tmp_alloc_ptr, cpu);
+ events->branches = records;
+ }
+
+ return 0;
+}
+
static void rvpmu_event_init(struct perf_event *event)
{
/*
@@ -1490,6 +1532,12 @@ static int rvpmu_device_probe(struct platform_device *pdev)
pmu->pmu.attr_groups = riscv_cdeleg_pmu_attr_groups;
else
pmu->pmu.attr_groups = riscv_sbi_pmu_attr_groups;
+
+ riscv_pmu_ctr_init(pmu);
+ ret = branch_records_alloc(pmu);
+ if (ret)
+ goto out_ctr_finish;
+
pmu->cmask = cmask;
pmu->ctr_start = rvpmu_ctr_start;
pmu->ctr_stop = rvpmu_ctr_stop;
@@ -1506,7 +1554,7 @@ static int rvpmu_device_probe(struct platform_device *pdev)
ret = cpuhp_state_add_instance(CPUHP_AP_PERF_RISCV_STARTING, &pmu->node);
if (ret)
- return ret;
+ goto out_ctr_finish;
ret = riscv_pm_pmu_register(pmu);
if (ret)
@@ -1523,6 +1571,9 @@ static int rvpmu_device_probe(struct platform_device *pdev)
out_unregister:
riscv_pmu_destroy(pmu);
+out_ctr_finish:
+ riscv_pmu_ctr_finish(pmu);
+
out_free:
kfree(pmu);
return ret;
This integrates recently added CTR ext support in riscv_pmu_dev driver to enable branch stack sampling using PMU events. This mainly adds CTR enable/disable callbacks in rvpmu_ctr_stop() and rvpmu_ctr_start() function to start/stop branch recording along with the event. PMU overflow handler rvpmu_ovf_handler() is also updated to sample CTR entries in case of the overflow for the particular event programmed to records branches. The recorded entries are fed to core perf for further processing. Signed-off-by: Rajnesh Kanwal <rkanwal@rivosinc.com> --- drivers/perf/riscv_pmu_common.c | 3 +- drivers/perf/riscv_pmu_dev.c | 77 +++++++++++++++++++++++++++------ 2 files changed, 65 insertions(+), 15 deletions(-)