In case percpu_max_size is 1 percpu_ida_free() sets bit in the cpumask and immediately transfers the tag to the pool. As result, the very next call to percpu_ida_alloc() on the same CPU will have to pull a tag from the pool to the local cache and so on. Hence, positive effects of local caching become largely negated.
This update assumes stealing tags is faster than ping-ponging between local caches and the pool and prevents returning tags to the pool in case percpu_max_size is 1. Signed-off-by: Alexander Gordeev <[email protected]> --- lib/percpu_ida.c | 5 ++--- 1 files changed, 2 insertions(+), 3 deletions(-) diff --git a/lib/percpu_ida.c b/lib/percpu_ida.c index 9dd8741..4adc3e5 100644 --- a/lib/percpu_ida.c +++ b/lib/percpu_ida.c @@ -243,9 +243,8 @@ void percpu_ida_free(struct percpu_ida *pool, unsigned tag) */ smp_wmb(); wake_up = true; - } - - if (nr_free == pool->percpu_max_size) { + } else if ((nr_free == pool->percpu_max_size) && + (pool->percpu_max_size > 1)) { spin_lock(&pool->lock); /* -- 1.7.7.6 -- Regards, Alexander Gordeev [email protected] -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
