When running with the kernel unmapped whilst at EL0, the virtually-addressed SPE buffer is also unmapped, which can lead to buffer faults if userspace profiling is enabled.
This patch prohibits SPE profiling of userspace when arm_kernel_unmapped_at_el0(). Signed-off-by: Will Deacon <[email protected]> --- drivers/perf/arm_spe_pmu.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c index 8ce262fc2561..c028db8973a4 100644 --- a/drivers/perf/arm_spe_pmu.c +++ b/drivers/perf/arm_spe_pmu.c @@ -675,6 +675,13 @@ static int arm_spe_pmu_event_init(struct perf_event *event) return -EOPNOTSUPP; /* + * If kernelspace is unmapped when running at EL0, then the SPE + * buffer will fault and prematurely terminate the AUX session. + */ + if (arm64_kernel_unmapped_at_el0() && !attr->exclude_user) + dev_warn_once(&spe_pmu->pdev->dev, "unable to write to profiling buffer from EL0. Try passing \"kaiser=off\" on the kernel command line"); + + /* * Feedback-directed frequency throttling doesn't work when we * have a buffer of samples. We'd need to manually count the * samples in the buffer when it fills up and adjust the event -- 2.1.4
