From: Saeed Mahameed <saeed@kernel.org>
To: "David S. Miller" <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
Eric Dumazet <edumazet@google.com>
Cc: Saeed Mahameed <saeedm@nvidia.com>,
netdev@vger.kernel.org, Tariq Toukan <tariqt@nvidia.com>,
Shay Drory <shayd@nvidia.com>, Mark Bloch <mbloch@nvidia.com>
Subject: [net-next V2 05/15] net/mlx5: Replace global mlx5_intf_lock with HCA devcom component lock
Date: Thu, 12 Oct 2023 12:27:40 -0700 [thread overview]
Message-ID: <20231012192750.124945-6-saeed@kernel.org> (raw)
In-Reply-To: <20231012192750.124945-1-saeed@kernel.org>
From: Shay Drory <shayd@nvidia.com>
mlx5_intf_lock is used to sync between LAG changes and its slaves
mlx5 core dev aux devices changes, which means every time mlx5 core
dev add/remove aux devices, mlx5 is taking this global lock, even if
LAG functionality isn't supported over the core dev.
This cause a bottleneck when probing VFs/SFs in parallel.
Hence, replace mlx5_intf_lock with HCA devcom component lock, or no
lock if LAG functionality isn't supported.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
drivers/net/ethernet/mellanox/mlx5/core/dev.c | 37 +++++--------------
.../net/ethernet/mellanox/mlx5/core/lag/lag.c | 35 +++++++++++++++---
.../net/ethernet/mellanox/mlx5/core/lag/lag.h | 1 +
.../ethernet/mellanox/mlx5/core/lag/mpesw.c | 9 ++++-
.../ethernet/mellanox/mlx5/core/lib/devcom.c | 7 ++++
.../ethernet/mellanox/mlx5/core/lib/devcom.h | 1 +
.../ethernet/mellanox/mlx5/core/mlx5_core.h | 8 ++--
7 files changed, 59 insertions(+), 39 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/dev.c b/drivers/net/ethernet/mellanox/mlx5/core/dev.c
index 6e3a8c22881f..cf0477f53dc4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/dev.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/dev.c
@@ -38,8 +38,6 @@
#include "devlink.h"
#include "lag/lag.h"
-/* intf dev list mutex */
-static DEFINE_MUTEX(mlx5_intf_mutex);
static DEFINE_IDA(mlx5_adev_ida);
static bool is_eth_rep_supported(struct mlx5_core_dev *dev)
@@ -337,9 +335,9 @@ static void del_adev(struct auxiliary_device *adev)
void mlx5_dev_set_lightweight(struct mlx5_core_dev *dev)
{
- mutex_lock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
dev->priv.flags |= MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
- mutex_unlock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
}
bool mlx5_dev_is_lightweight(struct mlx5_core_dev *dev)
@@ -355,7 +353,7 @@ int mlx5_attach_device(struct mlx5_core_dev *dev)
int ret = 0, i;
devl_assert_locked(priv_to_devlink(dev));
- mutex_lock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
priv->flags &= ~MLX5_PRIV_FLAGS_DETACH;
for (i = 0; i < ARRAY_SIZE(mlx5_adev_devices); i++) {
if (!priv->adev[i]) {
@@ -400,7 +398,7 @@ int mlx5_attach_device(struct mlx5_core_dev *dev)
break;
}
}
- mutex_unlock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
return ret;
}
@@ -413,7 +411,7 @@ void mlx5_detach_device(struct mlx5_core_dev *dev, bool suspend)
int i;
devl_assert_locked(priv_to_devlink(dev));
- mutex_lock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
for (i = ARRAY_SIZE(mlx5_adev_devices) - 1; i >= 0; i--) {
if (!priv->adev[i])
continue;
@@ -443,7 +441,7 @@ void mlx5_detach_device(struct mlx5_core_dev *dev, bool suspend)
priv->adev[i] = NULL;
}
priv->flags |= MLX5_PRIV_FLAGS_DETACH;
- mutex_unlock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
}
int mlx5_register_device(struct mlx5_core_dev *dev)
@@ -451,10 +449,10 @@ int mlx5_register_device(struct mlx5_core_dev *dev)
int ret;
devl_assert_locked(priv_to_devlink(dev));
- mutex_lock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
dev->priv.flags &= ~MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
ret = mlx5_rescan_drivers_locked(dev);
- mutex_unlock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
if (ret)
mlx5_unregister_device(dev);
@@ -464,10 +462,10 @@ int mlx5_register_device(struct mlx5_core_dev *dev)
void mlx5_unregister_device(struct mlx5_core_dev *dev)
{
devl_assert_locked(priv_to_devlink(dev));
- mutex_lock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
dev->priv.flags = MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
mlx5_rescan_drivers_locked(dev);
- mutex_unlock(&mlx5_intf_mutex);
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
}
static int add_drivers(struct mlx5_core_dev *dev)
@@ -545,7 +543,6 @@ int mlx5_rescan_drivers_locked(struct mlx5_core_dev *dev)
{
struct mlx5_priv *priv = &dev->priv;
- lockdep_assert_held(&mlx5_intf_mutex);
if (priv->flags & MLX5_PRIV_FLAGS_DETACH)
return 0;
@@ -565,17 +562,3 @@ bool mlx5_same_hw_devs(struct mlx5_core_dev *dev, struct mlx5_core_dev *peer_dev
return (fsystem_guid && psystem_guid && fsystem_guid == psystem_guid);
}
-
-void mlx5_dev_list_lock(void)
-{
- mutex_lock(&mlx5_intf_mutex);
-}
-void mlx5_dev_list_unlock(void)
-{
- mutex_unlock(&mlx5_intf_mutex);
-}
-
-int mlx5_dev_list_trylock(void)
-{
- return mutex_trylock(&mlx5_intf_mutex);
-}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.c b/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.c
index f0b57f97739f..d14459e5c04f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.c
@@ -943,6 +943,26 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
}
}
+/* The last mdev to unregister will destroy the workqueue before removing the
+ * devcom component, and as all the mdevs use the same devcom component we are
+ * guaranteed that the devcom is valid while the calling work is running.
+ */
+struct mlx5_devcom_comp_dev *mlx5_lag_get_devcom_comp(struct mlx5_lag *ldev)
+{
+ struct mlx5_devcom_comp_dev *devcom = NULL;
+ int i;
+
+ mutex_lock(&ldev->lock);
+ for (i = 0; i < ldev->ports; i++) {
+ if (ldev->pf[i].dev) {
+ devcom = ldev->pf[i].dev->priv.hca_devcom_comp;
+ break;
+ }
+ }
+ mutex_unlock(&ldev->lock);
+ return devcom;
+}
+
static void mlx5_queue_bond_work(struct mlx5_lag *ldev, unsigned long delay)
{
queue_delayed_work(ldev->wq, &ldev->bond_work, delay);
@@ -953,9 +973,14 @@ static void mlx5_do_bond_work(struct work_struct *work)
struct delayed_work *delayed_work = to_delayed_work(work);
struct mlx5_lag *ldev = container_of(delayed_work, struct mlx5_lag,
bond_work);
+ struct mlx5_devcom_comp_dev *devcom;
int status;
- status = mlx5_dev_list_trylock();
+ devcom = mlx5_lag_get_devcom_comp(ldev);
+ if (!devcom)
+ return;
+
+ status = mlx5_devcom_comp_trylock(devcom);
if (!status) {
mlx5_queue_bond_work(ldev, HZ);
return;
@@ -964,14 +989,14 @@ static void mlx5_do_bond_work(struct work_struct *work)
mutex_lock(&ldev->lock);
if (ldev->mode_changes_in_progress) {
mutex_unlock(&ldev->lock);
- mlx5_dev_list_unlock();
+ mlx5_devcom_comp_unlock(devcom);
mlx5_queue_bond_work(ldev, HZ);
return;
}
mlx5_do_bond(ldev);
mutex_unlock(&ldev->lock);
- mlx5_dev_list_unlock();
+ mlx5_devcom_comp_unlock(devcom);
}
static int mlx5_handle_changeupper_event(struct mlx5_lag *ldev,
@@ -1435,7 +1460,7 @@ void mlx5_lag_disable_change(struct mlx5_core_dev *dev)
if (!ldev)
return;
- mlx5_dev_list_lock();
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
mutex_lock(&ldev->lock);
ldev->mode_changes_in_progress++;
@@ -1443,7 +1468,7 @@ void mlx5_lag_disable_change(struct mlx5_core_dev *dev)
mlx5_disable_lag(ldev);
mutex_unlock(&ldev->lock);
- mlx5_dev_list_unlock();
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
}
void mlx5_lag_enable_change(struct mlx5_core_dev *dev)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.h b/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.h
index 481e92f39fe6..50fcb1eee574 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag/lag.h
@@ -112,6 +112,7 @@ void mlx5_disable_lag(struct mlx5_lag *ldev);
void mlx5_lag_remove_devices(struct mlx5_lag *ldev);
int mlx5_deactivate_lag(struct mlx5_lag *ldev);
void mlx5_lag_add_devices(struct mlx5_lag *ldev);
+struct mlx5_devcom_comp_dev *mlx5_lag_get_devcom_comp(struct mlx5_lag *ldev);
static inline bool mlx5_lag_is_supported(struct mlx5_core_dev *dev)
{
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag/mpesw.c b/drivers/net/ethernet/mellanox/mlx5/core/lag/mpesw.c
index 0857eebf4f07..82889f30506e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag/mpesw.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag/mpesw.c
@@ -129,9 +129,14 @@ static void disable_mpesw(struct mlx5_lag *ldev)
static void mlx5_mpesw_work(struct work_struct *work)
{
struct mlx5_mpesw_work_st *mpesww = container_of(work, struct mlx5_mpesw_work_st, work);
+ struct mlx5_devcom_comp_dev *devcom;
struct mlx5_lag *ldev = mpesww->lag;
- mlx5_dev_list_lock();
+ devcom = mlx5_lag_get_devcom_comp(ldev);
+ if (!devcom)
+ return;
+
+ mlx5_devcom_comp_lock(devcom);
mutex_lock(&ldev->lock);
if (ldev->mode_changes_in_progress) {
mpesww->result = -EAGAIN;
@@ -144,7 +149,7 @@ static void mlx5_mpesw_work(struct work_struct *work)
disable_mpesw(ldev);
unlock:
mutex_unlock(&ldev->lock);
- mlx5_dev_list_unlock();
+ mlx5_devcom_comp_unlock(devcom);
complete(&mpesww->comp);
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c
index f4d5c300ddd6..e8e50563e956 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c
@@ -401,3 +401,10 @@ void mlx5_devcom_comp_unlock(struct mlx5_devcom_comp_dev *devcom)
return;
up_write(&devcom->comp->sem);
}
+
+int mlx5_devcom_comp_trylock(struct mlx5_devcom_comp_dev *devcom)
+{
+ if (IS_ERR_OR_NULL(devcom))
+ return 0;
+ return down_write_trylock(&devcom->comp->sem);
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h
index ed249bafd07c..1549edd19aa8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h
@@ -54,5 +54,6 @@ void *mlx5_devcom_get_next_peer_data_rcu(struct mlx5_devcom_comp_dev *devcom,
void mlx5_devcom_comp_lock(struct mlx5_devcom_comp_dev *devcom);
void mlx5_devcom_comp_unlock(struct mlx5_devcom_comp_dev *devcom);
+int mlx5_devcom_comp_trylock(struct mlx5_devcom_comp_dev *devcom);
#endif /* __LIB_MLX5_DEVCOM_H__ */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
index bbc96aa6cbcc..0b60b02518c1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
@@ -41,6 +41,7 @@
#include <linux/mlx5/cq.h>
#include <linux/mlx5/fs.h>
#include <linux/mlx5/driver.h>
+#include "lib/devcom.h"
extern uint mlx5_core_debug_mask;
@@ -250,9 +251,6 @@ int mlx5_register_device(struct mlx5_core_dev *dev);
void mlx5_unregister_device(struct mlx5_core_dev *dev);
void mlx5_dev_set_lightweight(struct mlx5_core_dev *dev);
bool mlx5_dev_is_lightweight(struct mlx5_core_dev *dev);
-void mlx5_dev_list_lock(void);
-void mlx5_dev_list_unlock(void);
-int mlx5_dev_list_trylock(void);
void mlx5_fw_reporters_create(struct mlx5_core_dev *dev);
int mlx5_query_mtpps(struct mlx5_core_dev *dev, u32 *mtpps, u32 mtpps_size);
@@ -291,9 +289,9 @@ static inline int mlx5_rescan_drivers(struct mlx5_core_dev *dev)
{
int ret;
- mlx5_dev_list_lock();
+ mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
ret = mlx5_rescan_drivers_locked(dev);
- mlx5_dev_list_unlock();
+ mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
return ret;
}
--
2.41.0
next prev parent reply other threads:[~2023-10-12 19:27 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-12 19:27 [pull request][net-next V2 00/15] mlx5 updates 2023-10-10 Saeed Mahameed
2023-10-12 19:27 ` [net-next V2 01/15] net/mlx5: Parallelize vhca event handling Saeed Mahameed
2023-10-12 21:13 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 02/15] net/mlx5: Redesign SF active work to remove table_lock Saeed Mahameed
2023-10-12 21:18 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 03/15] net/mlx5: Avoid false positive lockdep warning by adding lock_class_key Saeed Mahameed
2023-10-12 21:23 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 04/15] net/mlx5: Refactor LAG peer device lookout bus logic to mlx5 devcom Saeed Mahameed
2023-10-12 21:26 ` Jacob Keller
2023-10-12 21:46 ` Saeed Mahameed
2023-10-12 19:27 ` Saeed Mahameed [this message]
2023-10-12 21:28 ` [net-next V2 05/15] net/mlx5: Replace global mlx5_intf_lock with HCA devcom component lock Jacob Keller
2023-10-12 19:27 ` [net-next V2 06/15] net/mlx5: Remove unused declaration Saeed Mahameed
2023-10-12 21:29 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 07/15] net/mlx5: fix config name in Kconfig parameter documentation Saeed Mahameed
2023-10-12 19:27 ` [net-next V2 08/15] net/mlx5: Use PTR_ERR_OR_ZERO() to simplify code Saeed Mahameed
2023-10-12 19:27 ` [net-next V2 09/15] net/mlx5e: " Saeed Mahameed
2023-10-12 21:30 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 10/15] net/mlx5e: Refactor rx_res_init() and rx_res_free() APIs Saeed Mahameed
2023-10-12 21:31 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 11/15] net/mlx5e: Refactor mlx5e_rss_set_rxfh() and mlx5e_rss_get_rxfh() Saeed Mahameed
2023-10-12 21:32 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 12/15] net/mlx5e: Refactor mlx5e_rss_init() and mlx5e_rss_free() API's Saeed Mahameed
2023-10-12 21:33 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 13/15] net/mlx5e: Preparations for supporting larger number of channels Saeed Mahameed
2023-10-12 21:35 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 14/15] net/mlx5e: Increase max supported channels number to 256 Saeed Mahameed
2023-10-12 21:35 ` Jacob Keller
2023-10-12 19:27 ` [net-next V2 15/15] net/mlx5e: Allow IPsec soft/hard limits in bytes Saeed Mahameed
2023-10-12 21:37 ` Jacob Keller
2023-10-14 1:10 ` [pull request][net-next V2 00/15] mlx5 updates 2023-10-10 Jakub Kicinski
2023-10-14 17:19 ` 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=20231012192750.124945-6-saeed@kernel.org \
--to=saeed@kernel.org \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=kuba@kernel.org \
--cc=mbloch@nvidia.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=saeedm@nvidia.com \
--cc=shayd@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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.