On Mon, 23 Apr 2018, Sebastian Ott wrote:
> This happend once during boot and I could not reproduce this since, but I
> think the following patch should fix the issue:
I can now recreate the issue reliably. The following patch works for me:
------>8
>From 6bbfab86beae8fe8b6efeffd08c9ad43d1c2ba68 Mon Sep 17 00:00:00 2001
From: Sebastian Ott <seb...@linux.ibm.com>
Date: Mon, 23 Apr 2018 14:59:20 +0200
Subject: [PATCH] random: don't sleep during allocations
Since numa_crng_init() can be called from irq context let's
use __GFP_ATOMIC for allocations.
Fixes: 8ef35c866f88 ("random: set up the NUMA crng instances...")
Signed-off-by: Sebastian Ott <seb...@linux.ibm.com>
---
drivers/char/random.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index 3cd3aae24d6d..98634b8025d5 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -795,10 +795,10 @@ static void numa_crng_init(void)
struct crng_state *crng;
struct crng_state **pool;
- pool = kcalloc(nr_node_ids, sizeof(*pool), GFP_KERNEL|__GFP_NOFAIL);
+ pool = kcalloc(nr_node_ids, sizeof(*pool), __GFP_NOFAIL | GFP_ATOMIC);
for_each_online_node(i) {
crng = kmalloc_node(sizeof(struct crng_state),
- GFP_KERNEL | __GFP_NOFAIL, i);
+ __GFP_NOFAIL | GFP_ATOMIC, i);
spin_lock_init(&crng->lock);
crng_initialize(crng);
pool[i] = crng;
--
2.13.4
