When running the in-kernel Dhrystone benchmark with CONFIG_DEBUG_PREEMPT=y:
BUG: using smp_processor_id() in preemptible [00000000] code: bash/938
Fix this by not using smp_processor_id() directly, but instead wrapping the whole benchmark inside a get_cpu()/put_cpu() pair. This makes sure the whole benchmark is run on the same CPU core, and the reported values are consistent.
Fixes: d5528cc16893f1f6 ("lib: add Dhrystone benchmark test") Reported-by: Tobias Klausmann klausman@schwarzvogel.de Link: https://bugzilla.kernel.org/show_bug.cgi?id=217179 Signed-off-by: Geert Uytterhoeven geert+renesas@glider.be --- lib/dhry_run.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/lib/dhry_run.c b/lib/dhry_run.c index f9d33efa6d090604..f15ac666e9d38bd2 100644 --- a/lib/dhry_run.c +++ b/lib/dhry_run.c @@ -31,6 +31,7 @@ MODULE_PARM_DESC(iterations,
static void dhry_benchmark(void) { + unsigned int cpu = get_cpu(); int i, n;
if (iterations > 0) { @@ -45,9 +46,10 @@ static void dhry_benchmark(void) }
report: + put_cpu(); if (n >= 0) - pr_info("CPU%u: Dhrystones per Second: %d (%d DMIPS)\n", - smp_processor_id(), n, n / DHRY_VAX); + pr_info("CPU%u: Dhrystones per Second: %d (%d DMIPS)\n", cpu, + n, n / DHRY_VAX); else if (n == -EAGAIN) pr_err("Please increase the number of iterations\n"); else