[PATCH] random: wait for system_wq before pushing entropy into output pools

[Konstantin Khlebnikov]

First prototype of tool which injects random delays in random places
(implemented on top of kernel address sanitizer by Andrey Ryabinin)
have caught race right at the first try.

credit_entropy_bits() schedules work for pushing entropy from one pool
into another too early, when system_wq isn't yet created. It's called from
add_interrupt_randomness, so this path starts working right after enabling
interrupts in start_kernel, but system workqueues are initialized much
later in early_initcall(init_workqueues). It's impossible to create them
earlier because threads have to be forked from kthreadd.

This patch skips pushing if system_wq isn't here yet. Delaying this till
SYSTEM_RUNNING state might lead to significant lost of precious entropy.

Signed-off-by: Konstantin Khlebnikov <k.khlebnikov@xxxxxxxxxxx>
Reported-and-tested-by: Andrey Ryabinin <a.ryabinin@xxxxxxxxxxx>
---
 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 04645c0..d4a684f 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -683,7 +683,7 @@ retry:
 		 * full.
 		 */
 		if (entropy_bits > random_write_wakeup_bits &&
-		    r->initialized &&
+		    system_wq && r->initialized &&
 		    r->entropy_total >= 2*random_read_wakeup_bits) {
 			static struct entropy_store *last = &blocking_pool;
 			struct entropy_store *other = &blocking_pool;
@@ -695,7 +695,7 @@ retry:
 				last = other;
 			if (last->entropy_count <=
 			    3 * last->poolinfo->poolfracbits / 4) {
-				schedule_work(&last->push_work);
+				queue_work(system_wq, &last->push_work);
 				r->entropy_total = 0;
 			}
 		}

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/