[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210810035923.345745-6-saeed@kernel.org>
Date: Mon, 9 Aug 2021 20:59:16 -0700
From: Saeed Mahameed <saeed@...nel.org>
To: "David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org, Leon Romanovsky <leonro@...dia.com>,
Tariq Toukan <tariqt@...dia.com>,
Maxim Mikityanskiy <maximmi@...dia.com>,
Saeed Mahameed <saeedm@...dia.com>
Subject: [net 05/12] net/mlx5e: Destroy page pool after XDP SQ to fix use-after-free
From: Maxim Mikityanskiy <maximmi@...dia.com>
mlx5e_close_xdpsq does the cleanup: it calls mlx5e_free_xdpsq_descs to
free the outstanding descriptors, which relies on
mlx5e_page_release_dynamic and page_pool_release_page. However,
page_pool_destroy is already called by this point, because
mlx5e_close_rq runs before mlx5e_close_xdpsq.
This commit fixes the use-after-free by swapping mlx5e_close_xdpsq and
mlx5e_close_rq.
The commit cited below started calling page_pool_destroy directly from
the driver. Previously, the page pool was destroyed under a call_rcu
from xdp_rxq_info_unreg_mem_model, which would defer the deallocation
until after the XDPSQ is cleaned up.
Fixes: 1da4bbeffe41 ("net: core: page_pool: add user refcnt and reintroduce page_pool_destroy")
Signed-off-by: Maxim Mikityanskiy <maximmi@...dia.com>
Reviewed-by: Tariq Toukan <tariqt@...dia.com>
Signed-off-by: Saeed Mahameed <saeedm@...dia.com>
---
.../net/ethernet/mellanox/mlx5/core/en_main.c | 20 +++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 37c440837945..fd250f7bcd88 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -1891,30 +1891,30 @@ static int mlx5e_open_queues(struct mlx5e_channel *c,
if (err)
goto err_close_icosq;
+ err = mlx5e_open_rxq_rq(c, params, &cparam->rq);
+ if (err)
+ goto err_close_sqs;
+
if (c->xdp) {
err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq, NULL,
&c->rq_xdpsq, false);
if (err)
- goto err_close_sqs;
+ goto err_close_rq;
}
- err = mlx5e_open_rxq_rq(c, params, &cparam->rq);
- if (err)
- goto err_close_xdp_sq;
-
err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq, NULL, &c->xdpsq, true);
if (err)
- goto err_close_rq;
+ goto err_close_xdp_sq;
return 0;
-err_close_rq:
- mlx5e_close_rq(&c->rq);
-
err_close_xdp_sq:
if (c->xdp)
mlx5e_close_xdpsq(&c->rq_xdpsq);
+err_close_rq:
+ mlx5e_close_rq(&c->rq);
+
err_close_sqs:
mlx5e_close_sqs(c);
@@ -1949,9 +1949,9 @@ static int mlx5e_open_queues(struct mlx5e_channel *c,
static void mlx5e_close_queues(struct mlx5e_channel *c)
{
mlx5e_close_xdpsq(&c->xdpsq);
- mlx5e_close_rq(&c->rq);
if (c->xdp)
mlx5e_close_xdpsq(&c->rq_xdpsq);
+ mlx5e_close_rq(&c->rq);
mlx5e_close_sqs(c);
mlx5e_close_icosq(&c->icosq);
mlx5e_close_icosq(&c->async_icosq);
--
2.31.1
Powered by blists - more mailing lists