mlx5: use RCU lock in mlx5_eq_cq_get()

Browse source

mlx5_eq_cq_get() is called in IRQ handler, the spinlock inside
gets a lot of contentions when we test some heavy workload
with 60 RX queues and 80 CPU's, and it is clearly shown in the
flame graph.

In fact, radix_tree_lookup() is perfectly fine with RCU read lock,
we don't have to take a spinlock on this hot path. This is pretty
much similar to commit 291c566a28
("net/mlx4_core: Fix racy CQ (Completion Queue) free"). Slow paths
are still serialized with the spinlock, and with synchronize_irq()
it should be safe to just move the fast path to RCU read lock.

This patch itself reduces the latency by about 50% for our memcached
workload on a 4.14 kernel we test. In upstream, as pointed out by Saeed,
this spinlock gets some rework in commit 02d92f7903
("net/mlx5: CQ Database per EQ"), so the difference could be smaller.

Cc: Saeed Mahameed <saeedm@mellanox.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

...

This commit is contained in:

Cong Wang 2019-02-06 15:00:19 -08:00 • committed by Saeed Mahameed

parent 46861e3e88

commit 1fbf1252df

1 changed files with 6 additions and 6 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download patch file Download diff file Expand all files Collapse all files

12

drivers/net/ethernet/mellanox/mlx5/core/eq.c

View file