For all the complexity of handling affinity for CPU hotplug, what we've apparently managed to overlook is that arm_cmn_init_irqs() has in fact always been setting the *initial* affinity of all IRQs to CPU 0, not the CPU we subsequently choose for event scheduling. Oh dear.
Cc: stable@vger.kernel.org Fixes: 0ba64770a2f2 ("perf: Add Arm CMN-600 PMU driver") Signed-off-by: Robin Murphy robin.murphy@arm.com --- drivers/perf/arm-cmn.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c index aa2908313558..e385f187a084 100644 --- a/drivers/perf/arm-cmn.c +++ b/drivers/perf/arm-cmn.c @@ -2558,6 +2558,7 @@ static int arm_cmn_probe(struct platform_device *pdev)
cmn->dev = &pdev->dev; cmn->part = (unsigned long)device_get_match_data(cmn->dev); + cmn->cpu = cpumask_local_spread(0, dev_to_node(cmn->dev)); platform_set_drvdata(pdev, cmn);
if (cmn->part == PART_CMN600 && has_acpi_companion(cmn->dev)) { @@ -2585,7 +2586,6 @@ static int arm_cmn_probe(struct platform_device *pdev) if (err) return err;
- cmn->cpu = cpumask_local_spread(0, dev_to_node(cmn->dev)); cmn->pmu = (struct pmu) { .module = THIS_MODULE, .parent = cmn->dev,
On Mon, 12 May 2025, Robin Murphy wrote:
For all the complexity of handling affinity for CPU hotplug, what we've apparently managed to overlook is that arm_cmn_init_irqs() has in fact always been setting the *initial* affinity of all IRQs to CPU 0, not the CPU we subsequently choose for event scheduling. Oh dear.
Cc: stable@vger.kernel.org Fixes: 0ba64770a2f2 ("perf: Add Arm CMN-600 PMU driver") Signed-off-by: Robin Murphy robin.murphy@arm.com
Good catch.
Reviewed-by: Ilkka Koskinen ilkka@os.amperecomputing.com
drivers/perf/arm-cmn.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c index aa2908313558..e385f187a084 100644 --- a/drivers/perf/arm-cmn.c +++ b/drivers/perf/arm-cmn.c @@ -2558,6 +2558,7 @@ static int arm_cmn_probe(struct platform_device *pdev)
cmn->dev = &pdev->dev; cmn->part = (unsigned long)device_get_match_data(cmn->dev);
cmn->cpu = cpumask_local_spread(0, dev_to_node(cmn->dev)); platform_set_drvdata(pdev, cmn);
if (cmn->part == PART_CMN600 && has_acpi_companion(cmn->dev)) {
@@ -2585,7 +2586,6 @@ static int arm_cmn_probe(struct platform_device *pdev) if (err) return err;
- cmn->cpu = cpumask_local_spread(0, dev_to_node(cmn->dev)); cmn->pmu = (struct pmu) { .module = THIS_MODULE, .parent = cmn->dev,
-- 2.39.2.101.g768bb238c484.dirty
On Mon, 12 May 2025 18:11:54 +0100, Robin Murphy wrote:
For all the complexity of handling affinity for CPU hotplug, what we've apparently managed to overlook is that arm_cmn_init_irqs() has in fact always been setting the *initial* affinity of all IRQs to CPU 0, not the CPU we subsequently choose for event scheduling. Oh dear.
Applied to will (for-next/perf), thanks!
[1/1] perf/arm-cmn: Initialise cmn->cpu earlier https://git.kernel.org/will/c/597704e20106
Cheers,
linux-stable-mirror@lists.linaro.org
-
Ilkka Koskinen -
Robin Murphy -
Will Deacon