[PATCH] perf/riscv: clear snapshot setup state when a CPU goes offline

[Pengpeng Hou]

pmu_sbi_dying_cpu() disables the calling hart's SBI PMU snapshot shared
memory but leaves snapshot_set_done set. When the CPU comes back online,
pmu_sbi_snapshot_setup() returns early without registering the shared
memory again.

Snapshot operations on that hart can then fail because the firmware no
longer has a snapshot area, while the driver still follows its snapshot
paths.

Clear the per-CPU setup flag after snapshot disable succeeds. Retain the
flag and the existing error return if disable fails, so a failed teardown
does not claim that the firmware registration has been removed.

Fixes: a8625217a054 ("drivers/perf: riscv: Implement SBI PMU snapshot function")
Assisted-by: GPT-5
Signed-off-by: Pengpeng Hou <hppiscas@xxxxxxx>
---
The issue was found by our static-analysis tool and manually reviewed.


diff --git a/drivers/perf/riscv_pmu_sbi.c b/drivers/perf/riscv_pmu_sbi.c
index 50220f7b..f38a64c7 100644
--- a/drivers/perf/riscv_pmu_sbi.c
+++ b/drivers/perf/riscv_pmu_sbi.c
@@ -1175,6 +1175,10 @@ static int pmu_sbi_starting_cpu(unsigned int cpu, struct hlist_node *node)
 
 static int pmu_sbi_dying_cpu(unsigned int cpu, struct hlist_node *node)
 {
+	struct riscv_pmu *pmu = hlist_entry_safe(node, struct riscv_pmu, node);
+	struct cpu_hw_events *cpu_hw_evt = per_cpu_ptr(pmu->hw_events, cpu);
+	int ret;
+
 	if (riscv_pmu_use_irq) {
 		disable_percpu_irq(riscv_pmu_irq);
 	}
@@ -1182,8 +1186,12 @@ static int pmu_sbi_dying_cpu(unsigned int cpu, struct hlist_node *node)
 	/* Disable all counters access for user mode now */
 	csr_write(CSR_SCOUNTEREN, 0x0);
 
-	if (sbi_pmu_snapshot_available())
-		return pmu_sbi_snapshot_disable();
+	if (sbi_pmu_snapshot_available()) {
+		ret = pmu_sbi_snapshot_disable();
+		if (ret)
+			return ret;
+		cpu_hw_evt->snapshot_set_done = false;
+	}
 
 	return 0;
 }

base-commit: 4d7d9486c04d917265f64c55bd23b2cc4fe7749c