[PATCH] random: use offstack cpumask when necessary

Arnd Bergmann posted 1 patch 4 months ago
drivers/char/random.c | 19 ++++++++++++-------
1 file changed, 12 insertions(+), 7 deletions(-)
[PATCH] random: use offstack cpumask when necessary
Posted by Arnd Bergmann 4 months ago
From: Arnd Bergmann <arnd@arndb.de>

The entropy generation function keeps a local cpu mask on the stack, which
can trigger warnings in configurations with a large number of CPUs:

drivers/char/random.c:1292:20: error: stack frame size (1288) exceeds limit (1280) in 'try_to_generate_entropy' [-Werror,-Wframe-larger-than]

Use the cpumask interface to dynamically allocate it in those configurations.

Fixes: 1c21fe00eda7 ("random: spread out jitter callback to different CPUs")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/char/random.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/drivers/char/random.c b/drivers/char/random.c
index b8b24b6ed3fe..04800b75cf73 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -1296,6 +1296,7 @@ static void __cold try_to_generate_entropy(void)
 	struct entropy_timer_state *stack = PTR_ALIGN((void *)stack_bytes, SMP_CACHE_BYTES);
 	unsigned int i, num_different = 0;
 	unsigned long last = random_get_entropy();
+	cpumask_var_t timer_cpus;
 	int cpu = -1;
 
 	for (i = 0; i < NUM_TRIAL_SAMPLES - 1; ++i) {
@@ -1310,13 +1311,15 @@ static void __cold try_to_generate_entropy(void)
 
 	atomic_set(&stack->samples, 0);
 	timer_setup_on_stack(&stack->timer, entropy_timer, 0);
+	if (!alloc_cpumask_var(&timer_cpus, GFP_KERNEL))
+		goto out;;
+
 	while (!crng_ready() && !signal_pending(current)) {
 		/*
 		 * Check !timer_pending() and then ensure that any previous callback has finished
 		 * executing by checking timer_delete_sync_try(), before queueing the next one.
 		 */
 		if (!timer_pending(&stack->timer) && timer_delete_sync_try(&stack->timer) >= 0) {
-			struct cpumask timer_cpus;
 			unsigned int num_cpus;
 
 			/*
@@ -1326,19 +1329,19 @@ static void __cold try_to_generate_entropy(void)
 			preempt_disable();
 
 			/* Only schedule callbacks on timer CPUs that are online. */
-			cpumask_and(&timer_cpus, housekeeping_cpumask(HK_TYPE_TIMER), cpu_online_mask);
-			num_cpus = cpumask_weight(&timer_cpus);
+			cpumask_and(timer_cpus, housekeeping_cpumask(HK_TYPE_TIMER), cpu_online_mask);
+			num_cpus = cpumask_weight(timer_cpus);
 			/* In very bizarre case of misconfiguration, fallback to all online. */
 			if (unlikely(num_cpus == 0)) {
-				timer_cpus = *cpu_online_mask;
-				num_cpus = cpumask_weight(&timer_cpus);
+				*timer_cpus = *cpu_online_mask;
+				num_cpus = cpumask_weight(timer_cpus);
 			}
 
 			/* Basic CPU round-robin, which avoids the current CPU. */
 			do {
-				cpu = cpumask_next(cpu, &timer_cpus);
+				cpu = cpumask_next(cpu, timer_cpus);
 				if (cpu >= nr_cpu_ids)
-					cpu = cpumask_first(&timer_cpus);
+					cpu = cpumask_first(timer_cpus);
 			} while (cpu == smp_processor_id() && num_cpus > 1);
 
 			/* Expiring the timer at `jiffies` means it's the next tick. */
@@ -1354,6 +1357,8 @@ static void __cold try_to_generate_entropy(void)
 	}
 	mix_pool_bytes(&stack->entropy, sizeof(stack->entropy));
 
+	free_cpumask_var(timer_cpus);
+out:
 	timer_delete_sync(&stack->timer);
 	timer_destroy_on_stack(&stack->timer);
 }
-- 
2.39.5
Re: [PATCH] random: use offstack cpumask when necessary
Posted by Jason A. Donenfeld 4 months ago
Hi Arnd,

On Tue, Jun 10, 2025 at 11:27:08AM +0200, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@arndb.de>
> 
> The entropy generation function keeps a local cpu mask on the stack, which
> can trigger warnings in configurations with a large number of CPUs:
> 
> drivers/char/random.c:1292:20: error: stack frame size (1288) exceeds limit (1280) in 'try_to_generate_entropy' [-Werror,-Wframe-larger-than]
> 
> Use the cpumask interface to dynamically allocate it in those configurations.

Thanks. I hadn't seen this interface before. Applied with one nit fixed:

> +		goto out;;

Double semi-colon changed to single semi-colon.

Jason