On Tue, Mar 07, 2023 at 12:06:47AM +0800, Vernon Yang wrote:
After commit 596ff4a09b89 ("cpumask: re-introduce constant-sized cpumask
optimizations"), when NR_CPUS <= BITS_PER_LONG, small_cpumask_bits used
a macro instead of variable-sized for efficient.
If no further cpus set, the cpumask_next() returns small_cpumask_bits,
it must greater than or equal to nr_cpumask_bits, so fix it to correctly.
Signed-off-by: Vernon Yang <redacted>
Hi Vernon,
In all that cases, nr_cpu_ids must be used. The difference is that
nr_cpumask_bits is an upper limit for possible CPUs, and it's derived
from compile-time NR_CPUS, unless CPUMASK_OFFSTACK is enabled.
nr_cpu_ids is an actual number of CPUS as counted on boot.
So, nr_cpu_ids is always equal or less than nr_cpumask_bits, and we'd
compare with the smaller number.
Nor sure, but maybe it's worth to introduce a macro like:
#define valid_cpuid(cpu) (cpu) < nr_cpu_ids
Thanks,
Yury
quoted hunk ↗ jump to hunk
---
drivers/char/random.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index ce3ccd172cc8..d76f12a5f74f 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -1311,7 +1311,7 @@ static void __cold try_to_generate_entropy(void)
/* Basic CPU round-robin, which avoids the current CPU. */
do {
cpu = cpumask_next(cpu, &timer_cpus);
- if (cpu == nr_cpumask_bits)
+ if (cpu >= nr_cpumask_bits)
cpu = cpumask_first(&timer_cpus);
} while (cpu == smp_processor_id() && num_cpus > 1);
--
2.34.1