RDS: flush fmrs before allocating new ones (8576f374) · Commits · 戴 / test

Flushing FMRs is somewhat expensive, and is currently kicked off when the interrupt handler notices that we are getting low. The result of this is that FMR flushing only happens from the interrupt cpus. This spreads the load more effectively by triggering flushes just before we allocate a new FMR. Signed-off-by:

Chris Mason <chris.mason@oracle.com>

net/rds/ib_rdma.c

+5 −2

Original line number	Diff line number	Diff line
		@@ -38,6 +38,8 @@
		#include "ib.h"
		#include "xlist.h"

		struct workqueue_struct *rds_ib_fmr_wq;

		static DEFINE_PER_CPU(unsigned long, clean_list_grace);
		#define CLEAN_LIST_BUSY_BIT 0

		@@ -304,6 +306,9 @@ static struct rds_ib_mr rds_ib_alloc_fmr(struct rds_ib_device rds_ibdev)
		struct rds_ib_mr *ibmr = NULL;
		int err = 0, iter = 0;

		if (atomic_read(&pool->dirty_count) >= pool->max_items / 10)
		queue_delayed_work(rds_ib_fmr_wq, &pool->flush_worker, 10);

		while (1) {
		ibmr = rds_ib_reuse_fmr(pool);
		if (ibmr)
		@@ -691,8 +696,6 @@ out_nolock:
		return ret;
		}

		struct workqueue_struct *rds_ib_fmr_wq;

		int rds_ib_fmr_init(void)
		{
		rds_ib_fmr_wq = create_workqueue("rds_fmr_flushd");

Admin message