netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Saeed Mahameed <saeed@kernel.org>
To: "David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>
Cc: netdev@vger.kernel.org, Maxim Mikityanskiy <maximmi@mellanox.com>,
	Tariq Toukan <tariqt@nvidia.com>,
	Saeed Mahameed <saeedm@nvidia.com>
Subject: [net 09/15] net/mlx5e: kTLS, Use refcounts to free kTLS RX priv context
Date: Thu, 11 Feb 2021 18:56:35 -0800	[thread overview]
Message-ID: <20210212025641.323844-10-saeed@kernel.org> (raw)
In-Reply-To: <20210212025641.323844-1-saeed@kernel.org>

From: Maxim Mikityanskiy <maximmi@mellanox.com>

wait_for_resync is unreliable - if it timeouts, priv_rx will be freed
anyway. However, mlx5e_ktls_handle_get_psv_completion will be called
sooner or later, leading to use-after-free. For example, it can happen
if a CQ error happened, and ICOSQ stopped, but later on the queues are
destroyed, and ICOSQ is flushed with mlx5e_free_icosq_descs.

This patch converts the lifecycle of priv_rx to fully refcount-based, so
that the struct won't be freed before the refcount goes to zero.

Fixes: 0419d8c9d8f8 ("net/mlx5e: kTLS, Add kTLS RX resync support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../mellanox/mlx5/core/en_accel/ktls_rx.c     | 64 +++++++++----------
 1 file changed, 30 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_rx.c
index 0f13b661f7f9..d06532d0baa4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_rx.c
@@ -57,6 +57,20 @@ struct mlx5e_ktls_offload_context_rx {
 	struct mlx5e_ktls_rx_resync_ctx resync;
 };
 
+static bool mlx5e_ktls_priv_rx_put(struct mlx5e_ktls_offload_context_rx *priv_rx)
+{
+	if (!refcount_dec_and_test(&priv_rx->resync.refcnt))
+		return false;
+
+	kfree(priv_rx);
+	return true;
+}
+
+static void mlx5e_ktls_priv_rx_get(struct mlx5e_ktls_offload_context_rx *priv_rx)
+{
+	refcount_inc(&priv_rx->resync.refcnt);
+}
+
 static int mlx5e_ktls_create_tir(struct mlx5_core_dev *mdev, u32 *tirn, u32 rqtn)
 {
 	int err, inlen;
@@ -326,7 +340,7 @@ static void resync_handle_work(struct work_struct *work)
 	priv_rx = container_of(resync, struct mlx5e_ktls_offload_context_rx, resync);
 
 	if (unlikely(test_bit(MLX5E_PRIV_RX_FLAG_DELETING, priv_rx->flags))) {
-		refcount_dec(&resync->refcnt);
+		mlx5e_ktls_priv_rx_put(priv_rx);
 		return;
 	}
 
@@ -334,7 +348,7 @@ static void resync_handle_work(struct work_struct *work)
 	sq = &c->async_icosq;
 
 	if (resync_post_get_progress_params(sq, priv_rx))
-		refcount_dec(&resync->refcnt);
+		mlx5e_ktls_priv_rx_put(priv_rx);
 }
 
 static void resync_init(struct mlx5e_ktls_rx_resync_ctx *resync,
@@ -377,7 +391,11 @@ static int resync_handle_seq_match(struct mlx5e_ktls_offload_context_rx *priv_rx
 	return err;
 }
 
-/* Function is called with elevated refcount, it decreases it. */
+/* Function can be called with the refcount being either elevated or not.
+ * It decreases the refcount and may free the kTLS priv context.
+ * Refcount is not elevated only if tls_dev_del has been called, but GET_PSV was
+ * already in flight.
+ */
 void mlx5e_ktls_handle_get_psv_completion(struct mlx5e_icosq_wqe_info *wi,
 					  struct mlx5e_icosq *sq)
 {
@@ -410,7 +428,7 @@ void mlx5e_ktls_handle_get_psv_completion(struct mlx5e_icosq_wqe_info *wi,
 	tls_offload_rx_resync_async_request_end(priv_rx->sk, cpu_to_be32(hw_seq));
 	priv_rx->stats->tls_resync_req_end++;
 out:
-	refcount_dec(&resync->refcnt);
+	mlx5e_ktls_priv_rx_put(priv_rx);
 	dma_unmap_single(dev, buf->dma_addr, PROGRESS_PARAMS_PADDED_SIZE, DMA_FROM_DEVICE);
 	kfree(buf);
 }
@@ -431,9 +449,9 @@ static bool resync_queue_get_psv(struct sock *sk)
 		return false;
 
 	resync = &priv_rx->resync;
-	refcount_inc(&resync->refcnt);
+	mlx5e_ktls_priv_rx_get(priv_rx);
 	if (unlikely(!queue_work(resync->priv->tls->rx_wq, &resync->work)))
-		refcount_dec(&resync->refcnt);
+		mlx5e_ktls_priv_rx_put(priv_rx);
 
 	return true;
 }
@@ -625,31 +643,6 @@ int mlx5e_ktls_add_rx(struct net_device *netdev, struct sock *sk,
 	return err;
 }
 
-/* Elevated refcount on the resync object means there are
- * outstanding operations (uncompleted GET_PSV WQEs) that
- * will read the resync / priv_rx objects once completed.
- * Wait for them to avoid use-after-free.
- */
-static void wait_for_resync(struct net_device *netdev,
-			    struct mlx5e_ktls_rx_resync_ctx *resync)
-{
-#define MLX5E_KTLS_RX_RESYNC_TIMEOUT 20000 /* msecs */
-	unsigned long exp_time = jiffies + msecs_to_jiffies(MLX5E_KTLS_RX_RESYNC_TIMEOUT);
-	unsigned int refcnt;
-
-	do {
-		refcnt = refcount_read(&resync->refcnt);
-		if (refcnt == 1)
-			return;
-
-		msleep(20);
-	} while (time_before(jiffies, exp_time));
-
-	netdev_warn(netdev,
-		    "Failed waiting for kTLS RX resync refcnt to be released (%u).\n",
-		    refcnt);
-}
-
 void mlx5e_ktls_del_rx(struct net_device *netdev, struct tls_context *tls_ctx)
 {
 	struct mlx5e_ktls_offload_context_rx *priv_rx;
@@ -671,8 +664,7 @@ void mlx5e_ktls_del_rx(struct net_device *netdev, struct tls_context *tls_ctx)
 		wait_for_completion(&priv_rx->add_ctx);
 	resync = &priv_rx->resync;
 	if (cancel_work_sync(&resync->work))
-		refcount_dec(&resync->refcnt);
-	wait_for_resync(netdev, resync);
+		mlx5e_ktls_priv_rx_put(priv_rx);
 
 	priv_rx->stats->tls_del++;
 	if (priv_rx->rule.rule)
@@ -680,5 +672,9 @@ void mlx5e_ktls_del_rx(struct net_device *netdev, struct tls_context *tls_ctx)
 
 	mlx5_core_destroy_tir(mdev, priv_rx->tirn);
 	mlx5_ktls_destroy_key(mdev, priv_rx->key_id);
-	kfree(priv_rx);
+	/* priv_rx should normally be freed here, but if there is an outstanding
+	 * GET_PSV, deallocation will be delayed until the CQE for GET_PSV is
+	 * processed.
+	 */
+	mlx5e_ktls_priv_rx_put(priv_rx);
 }
-- 
2.29.2


  parent reply	other threads:[~2021-02-12  2:59 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-12  2:56 [pull request][net 00/15] mlx5 fixes 2021-02-11 Saeed Mahameed
2021-02-12  2:56 ` [net 01/15] net/mlx5e: E-switch, Fix rate calculation for overflow Saeed Mahameed
2021-02-27 12:14   ` Arnd Bergmann
2021-03-02  0:52     ` Saeed Mahameed
2021-03-02  9:01       ` Arnd Bergmann
2021-02-12  2:56 ` [net 02/15] net/mlx5e: Enable striding RQ for Connect-X IPsec capable devices Saeed Mahameed
2021-02-12  2:56 ` [net 03/15] net/mlx5e: Enable XDP " Saeed Mahameed
2021-02-12  2:56 ` [net 04/15] net/mlx5e: Don't change interrupt moderation params when DIM is enabled Saeed Mahameed
2021-02-12  2:56 ` [net 05/15] net/mlx5e: Change interrupt moderation channel params also when channels are closed Saeed Mahameed
2021-02-12  2:56 ` [net 06/15] net/mlx5: Fix health error state handling Saeed Mahameed
2021-02-12  2:56 ` [net 07/15] net/mlx5e: Replace synchronize_rcu with synchronize_net Saeed Mahameed
2021-02-12  2:56 ` [net 08/15] net/mlx5e: Fix CQ params of ICOSQ and async ICOSQ Saeed Mahameed
2021-02-12  2:56 ` Saeed Mahameed [this message]
2021-02-12  2:56 ` [net 10/15] net/mlx5: Disable devlink reload for multi port slave device Saeed Mahameed
2021-02-12  2:56 ` [net 11/15] net/mlx5: Disallow RoCE on " Saeed Mahameed
2021-02-12  2:56 ` [net 12/15] net/mlx5: Disallow RoCE on lag device Saeed Mahameed
2021-02-12  2:56 ` [net 13/15] net/mlx5: Disable devlink reload for lag devices Saeed Mahameed
2021-02-12  2:56 ` [net 14/15] net/mlx5e: CT: manage the lifetime of the ct entry object Saeed Mahameed
2021-02-12  2:56 ` [net 15/15] net/mlx5e: Check tunnel offload is required before setting SWP Saeed Mahameed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210212025641.323844-10-saeed@kernel.org \
    --to=saeed@kernel.org \
    --cc=davem@davemloft.net \
    --cc=kuba@kernel.org \
    --cc=maximmi@mellanox.com \
    --cc=netdev@vger.kernel.org \
    --cc=saeedm@nvidia.com \
    --cc=tariqt@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).