add_interrupt_randomness() can cause significant cpu overhead on
interrupt-heavy workloads. We try to limit that overhead by bailing out
quickly once we have sampled a few bits of entropy this second. If
there is enough entropy around it doesn't hurt to waste the excess.
However, we also waste entropy early in boot when we haven't even
initialized the pools yet. With this patch we initialize the pools in
1-2s while it takes 10-20s without this patch. Actual numbers depend on
hardware and fluctuate from boot to boot, but in all cases I have tested
there is a clear improvement.
Signed-off-by: Joern Engel <jo...@logfs.org>
---
drivers/char/random.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index 9cd6968e2f92..514f67a98b88 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -898,7 +898,8 @@ void add_interrupt_randomness(int irq, int irq_flags)
add_interrupt_bench(cycles);
if ((fast_pool->count < 64) &&
- !time_after(now, fast_pool->last + HZ))
+ !time_after(now, fast_pool->last + HZ) &&
+ nonblocking_pool.initialized)
return;
r = nonblocking_pool.initialized ? &input_pool : &nonblocking_pool;
--
2.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
