* [pull request][net-next 0/9] Mellanox, mlx5 ethernet updates 2018-09-05
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Saeed Mahameed
Hi Dave,
This pull request provides some updates to mlx5 ethernet driver.
For more information please see tag log below.
Please pull and let me know if there's any problem.
Thanks,
Saeed.
---
The following changes since commit 05dcc71298643256948a2e17db7dbecc748719d2:
net: lan743x_ptp: make function lan743x_ptp_set_sync_ts_insert() static (2018-09-05 08:07:05 -0700)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5e-updates-2018-09-05
for you to fetch changes up to fe1dc069990c1f290ef6b99adb46332c03258f38:
net/mlx5e: don't set CHECKSUM_COMPLETE on SCTP packets (2018-09-05 21:14:57 -0700)
----------------------------------------------------------------
mlx5e-updates-2018-09-05
This series provides updates to mlx5 ethernet driver.
1) Starting with a four patches series to optimize flow counters updates,
>From Vlad Buslov:
==============================================
By default mlx5 driver updates cached counters each second. Update function
consumes noticeable amount of CPU resources. The goal of this patch series
is to optimize update function.
Investigation revealed following bottlenecks in fs counters
implementation:
1) Update code(scheduled each second) iterates over all counters twice.
(first for finding and deleting counters that are marked for deletion,
second iteration is for actually updating the counters)
2) Counters are stored in rb tree. Linear iteration over all rb tree
elements(rb_next in profiling data) consumed ~65% of time spent in
update function.
Following optimizations were implemented:
1) Instead of just marking counters for deletion, store them in
standalone list. This removes first iteration over whole counters tree.
2) Store counters in sorted list to optimize traversing them and remove
calls to rb_next.
First implementation of these changes caused degradation of performance,
instead of improving it. Investigation revealed that there first cache
line of struct mlx5_fc is full and adding anything to it causes amount
of cache misses to double. To mitigate that, following refactorings were
implemented:
- Change 'addlist' list type from double linked to single linked. This
allowes to get free space for one additional pointer that is used to
store deletion list(optimization 1)
- Substitute rb tree with idr. Idr is non-intrusive data structure and
doesn't require adding any new members to struct mlx5_fc. Use free
space that became available for double linked sorted list that is used
for traversing all counters. (optimization 2)
Described changes reduced CPU time spent in mlx5_fc_stats_work from 70%
to 44%. (global perf profile mode)
============================================
The rest of the series are misc updates:
2) From Kamal, Move mlx5e_priv_flags into en_ethtool.c, to avoid a
compilation warning.
3) From Roi Dayan, Move Q counters allocation and drop RQ to init_rx profile
function to avoid allocating Q counters when not required.
4) From Shay Agroskin, Replace PTP clock lock from RW lock to seq lock.
Almost double the packet rate when timestamping is active on multiple TX
queues.
5) From: Natali Shechtman, set ECN for received packets using CQE indication.
6) From: Alaa Hleihel, don't set CHECKSUM_COMPLETE on SCTP packets.
CHECKSUM_COMPLETE is not applicable to SCTP protocol.
----------------------------------------------------------------
Alaa Hleihel (1):
net/mlx5e: don't set CHECKSUM_COMPLETE on SCTP packets
Kamal Heib (1):
net/mlx5e: Move mlx5e_priv_flags into en_ethtool.c
Natali Shechtman (1):
net/mlx5e: Set ECN for received packets using CQE indication
Roi Dayan (1):
net/mlx5e: Move Q counters allocation and drop RQ to init_rx
Shay Agroskin (1):
net/mlx5e: Replace PTP clock lock from RW lock to seq lock
Vlad Buslov (4):
net/mlx5: Change flow counters addlist type to single linked list
net/mlx5: Add new list to store deleted flow counters
net/mlx5: Store flow counters in a list
net/mlx5: Add flow counters idr
drivers/net/ethernet/mellanox/mlx5/core/en.h | 13 +-
.../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 7 +
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 45 +++--
drivers/net/ethernet/mellanox/mlx5/core/en_rep.c | 12 +-
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c | 47 +++++-
drivers/net/ethernet/mellanox/mlx5/core/en_stats.c | 3 +
drivers/net/ethernet/mellanox/mlx5/core/en_stats.h | 2 +
drivers/net/ethernet/mellanox/mlx5/core/fs_core.h | 5 +-
.../net/ethernet/mellanox/mlx5/core/fs_counters.c | 184 +++++++++++----------
.../net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c | 17 +-
.../net/ethernet/mellanox/mlx5/core/lib/clock.c | 34 ++--
.../net/ethernet/mellanox/mlx5/core/lib/clock.h | 8 +-
include/linux/mlx5/driver.h | 11 +-
13 files changed, 235 insertions(+), 153 deletions(-)
^ permalink raw reply
* [net-next 1/9] net/mlx5: Change flow counters addlist type to single linked list
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Vlad Buslov, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Vlad Buslov <vladbu@mellanox.com>
In order to prevent flow counters stats work function from traversing whole
flow counters tree while searching for deleted flow counters, new list to
store deleted flow counters will be added to struct mlx5_fc_stats. However,
the flow counter structure itself has no space left to store any more data
in first cache line. To free space that is needed to store additional list
node, convert current addlist double linked list (two pointers per node) to
atomic single linked list (one pointer per node).
Lockless NULL-terminated single linked list data type doesn't require any
additional external synchronization for operations used by flow counters
module (add single new element, remove all elements from list and traverse
them). Remove addlist_lock that is no longer needed.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../net/ethernet/mellanox/mlx5/core/fs_core.h | 3 +-
.../ethernet/mellanox/mlx5/core/fs_counters.c | 45 +++++++++----------
include/linux/mlx5/driver.h | 4 +-
3 files changed, 23 insertions(+), 29 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
index 32070e5d993d..f68590291e0c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
@@ -36,6 +36,7 @@
#include <linux/refcount.h>
#include <linux/mlx5/fs.h>
#include <linux/rhashtable.h>
+#include <linux/llist.h>
enum fs_node_type {
FS_TYPE_NAMESPACE,
@@ -139,7 +140,7 @@ struct mlx5_fc_cache {
struct mlx5_fc {
struct rb_node node;
- struct list_head list;
+ struct llist_node addlist;
/* last{packets,bytes} members are used when calculating the delta since
* last reading
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index 58af6be13dfa..d996d6cf9e19 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -52,7 +52,9 @@
* access to counter list:
* - create (user context)
* - mlx5_fc_create() only adds to an addlist to be used by
- * mlx5_fc_stats_query_work(). addlist is protected by a spinlock.
+ * mlx5_fc_stats_query_work(). addlist is a lockless single linked list
+ * that doesn't require any additional synchronization when adding single
+ * node.
* - spawn thread to do the actual destroy
*
* - destroy (user context)
@@ -156,28 +158,29 @@ static struct rb_node *mlx5_fc_stats_query(struct mlx5_core_dev *dev,
return node;
}
+static void mlx5_free_fc(struct mlx5_core_dev *dev,
+ struct mlx5_fc *counter)
+{
+ mlx5_cmd_fc_free(dev, counter->id);
+ kfree(counter);
+}
+
static void mlx5_fc_stats_work(struct work_struct *work)
{
struct mlx5_core_dev *dev = container_of(work, struct mlx5_core_dev,
priv.fc_stats.work.work);
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ struct llist_node *tmplist = llist_del_all(&fc_stats->addlist);
unsigned long now = jiffies;
struct mlx5_fc *counter = NULL;
struct mlx5_fc *last = NULL;
struct rb_node *node;
- LIST_HEAD(tmplist);
-
- spin_lock(&fc_stats->addlist_lock);
- list_splice_tail_init(&fc_stats->addlist, &tmplist);
-
- if (!list_empty(&tmplist) || !RB_EMPTY_ROOT(&fc_stats->counters))
+ if (tmplist || !RB_EMPTY_ROOT(&fc_stats->counters))
queue_delayed_work(fc_stats->wq, &fc_stats->work,
fc_stats->sampling_interval);
- spin_unlock(&fc_stats->addlist_lock);
-
- list_for_each_entry(counter, &tmplist, list)
+ llist_for_each_entry(counter, tmplist, addlist)
mlx5_fc_stats_insert(&fc_stats->counters, counter);
node = rb_first(&fc_stats->counters);
@@ -229,9 +232,7 @@ struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
counter->cache.lastuse = jiffies;
counter->aging = true;
- spin_lock(&fc_stats->addlist_lock);
- list_add(&counter->list, &fc_stats->addlist);
- spin_unlock(&fc_stats->addlist_lock);
+ llist_add(&counter->addlist, &fc_stats->addlist);
mod_delayed_work(fc_stats->wq, &fc_stats->work, 0);
}
@@ -268,8 +269,7 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
fc_stats->counters = RB_ROOT;
- INIT_LIST_HEAD(&fc_stats->addlist);
- spin_lock_init(&fc_stats->addlist_lock);
+ init_llist_head(&fc_stats->addlist);
fc_stats->wq = create_singlethread_workqueue("mlx5_fc");
if (!fc_stats->wq)
@@ -284,6 +284,7 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
{
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ struct llist_node *tmplist;
struct mlx5_fc *counter;
struct mlx5_fc *tmp;
struct rb_node *node;
@@ -292,13 +293,9 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
destroy_workqueue(dev->priv.fc_stats.wq);
dev->priv.fc_stats.wq = NULL;
- list_for_each_entry_safe(counter, tmp, &fc_stats->addlist, list) {
- list_del(&counter->list);
-
- mlx5_cmd_fc_free(dev, counter->id);
-
- kfree(counter);
- }
+ tmplist = llist_del_all(&fc_stats->addlist);
+ llist_for_each_entry_safe(counter, tmp, tmplist, addlist)
+ mlx5_free_fc(dev, counter);
node = rb_first(&fc_stats->counters);
while (node) {
@@ -308,9 +305,7 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
rb_erase(&counter->node, &fc_stats->counters);
- mlx5_cmd_fc_free(dev, counter->id);
-
- kfree(counter);
+ mlx5_free_fc(dev, counter);
}
}
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 7a452716de4b..c00549293982 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -584,9 +584,7 @@ struct mlx5_irq_info {
struct mlx5_fc_stats {
struct rb_root counters;
- struct list_head addlist;
- /* protect addlist add/splice operations */
- spinlock_t addlist_lock;
+ struct llist_head addlist;
struct workqueue_struct *wq;
struct delayed_work work;
--
2.17.1
^ permalink raw reply related
* [net-next 4/9] net/mlx5: Add flow counters idr
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Vlad Buslov, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Vlad Buslov <vladbu@mellanox.com>
Previous patch in series changed flow counter storage structure from
rb_tree to linked list in order to improve flow counter traversal
performance. The drawback of such solution is that flow counter lookup by
id becomes linear in complexity.
Store pointers to flow counters in idr in order to improve lookup
performance to logarithmic again. Idr is non-intrusive data structure and
doesn't require extending flow counter struct with new elements. This means
that idr can be used for lookup, while linked list from previous patch is
used for traversal, and struct mlx5_fc size is <= 2 cache lines.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../ethernet/mellanox/mlx5/core/fs_counters.c | 37 +++++++++++++++++--
include/linux/mlx5/driver.h | 2 +
2 files changed, 35 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index 90ebfee37508..09206c4acd9a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -77,13 +77,18 @@ static struct list_head *mlx5_fc_counters_lookup_next(struct mlx5_core_dev *dev,
u32 id)
{
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ unsigned long next_id = (unsigned long)id + 1;
struct mlx5_fc *counter;
- list_for_each_entry(counter, &fc_stats->counters, list)
- if (counter->id > id)
- return &counter->list;
+ rcu_read_lock();
+ /* skip counters that are in idr, but not yet in counters list */
+ while ((counter = idr_get_next_ul(&fc_stats->counters_idr,
+ &next_id)) != NULL &&
+ list_empty(&counter->list))
+ next_id++;
+ rcu_read_unlock();
- return &fc_stats->counters;
+ return counter ? &counter->list : &fc_stats->counters;
}
static void mlx5_fc_stats_insert(struct mlx5_core_dev *dev,
@@ -214,15 +219,29 @@ struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
counter = kzalloc(sizeof(*counter), GFP_KERNEL);
if (!counter)
return ERR_PTR(-ENOMEM);
+ INIT_LIST_HEAD(&counter->list);
err = mlx5_cmd_fc_alloc(dev, &counter->id);
if (err)
goto err_out;
if (aging) {
+ u32 id = counter->id;
+
counter->cache.lastuse = jiffies;
counter->aging = true;
+ idr_preload(GFP_KERNEL);
+ spin_lock(&fc_stats->counters_idr_lock);
+
+ err = idr_alloc_u32(&fc_stats->counters_idr, counter, &id, id,
+ GFP_NOWAIT);
+
+ spin_unlock(&fc_stats->counters_idr_lock);
+ idr_preload_end();
+ if (err)
+ goto err_out_alloc;
+
llist_add(&counter->addlist, &fc_stats->addlist);
mod_delayed_work(fc_stats->wq, &fc_stats->work, 0);
@@ -230,6 +249,8 @@ struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
return counter;
+err_out_alloc:
+ mlx5_cmd_fc_free(dev, counter->id);
err_out:
kfree(counter);
@@ -245,6 +266,10 @@ void mlx5_fc_destroy(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
return;
if (counter->aging) {
+ spin_lock(&fc_stats->counters_idr_lock);
+ WARN_ON(!idr_remove(&fc_stats->counters_idr, counter->id));
+ spin_unlock(&fc_stats->counters_idr_lock);
+
llist_add(&counter->dellist, &fc_stats->dellist);
mod_delayed_work(fc_stats->wq, &fc_stats->work, 0);
return;
@@ -258,6 +283,8 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
{
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ spin_lock_init(&fc_stats->counters_idr_lock);
+ idr_init(&fc_stats->counters_idr);
INIT_LIST_HEAD(&fc_stats->counters);
init_llist_head(&fc_stats->addlist);
init_llist_head(&fc_stats->dellist);
@@ -283,6 +310,8 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
destroy_workqueue(dev->priv.fc_stats.wq);
dev->priv.fc_stats.wq = NULL;
+ idr_destroy(&fc_stats->counters_idr);
+
tmplist = llist_del_all(&fc_stats->addlist);
llist_for_each_entry_safe(counter, tmp, tmplist, addlist)
mlx5_free_fc(dev, counter);
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 61bed33e6675..2a0c845f6bdb 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -583,6 +583,8 @@ struct mlx5_irq_info {
};
struct mlx5_fc_stats {
+ spinlock_t counters_idr_lock; /* protects counters_idr */
+ struct idr counters_idr;
struct list_head counters;
struct llist_head addlist;
struct llist_head dellist;
--
2.17.1
^ permalink raw reply related
* [net-next 2/9] net/mlx5: Add new list to store deleted flow counters
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Vlad Buslov, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Vlad Buslov <vladbu@mellanox.com>
In order to prevent flow counters stats work function from traversing whole
flow counters tree while searching for deleted flow counters, new list to
store deleted flow counters is added to struct mlx5_fc_stats. Lockless
NULL-terminated single linked list data type is used due to following
reasons:
- This use case only needs to add single element to list and
remove/iterate whole list. Lockless list doesn't require any additional
synchronization for these operations.
- First cache line of flow counter data structure only has space to store
single additional pointer, which precludes usage of double linked list.
Remove flow counter 'deleted' flag that is no longer needed.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../net/ethernet/mellanox/mlx5/core/fs_core.h | 2 +-
.../ethernet/mellanox/mlx5/core/fs_counters.c | 34 +++++++------------
include/linux/mlx5/driver.h | 1 +
3 files changed, 14 insertions(+), 23 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
index f68590291e0c..617d6239c5f3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
@@ -141,6 +141,7 @@ struct mlx5_fc_cache {
struct mlx5_fc {
struct rb_node node;
struct llist_node addlist;
+ struct llist_node dellist;
/* last{packets,bytes} members are used when calculating the delta since
* last reading
@@ -149,7 +150,6 @@ struct mlx5_fc {
u64 lastbytes;
u32 id;
- bool deleted;
bool aging;
struct mlx5_fc_cache cache ____cacheline_aligned_in_smp;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index d996d6cf9e19..f1266f215a31 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -58,7 +58,7 @@
* - spawn thread to do the actual destroy
*
* - destroy (user context)
- * - mark a counter as deleted
+ * - add a counter to lockless dellist
* - spawn thread to do the actual del
*
* - dump (user context)
@@ -171,9 +171,8 @@ static void mlx5_fc_stats_work(struct work_struct *work)
priv.fc_stats.work.work);
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
struct llist_node *tmplist = llist_del_all(&fc_stats->addlist);
+ struct mlx5_fc *counter = NULL, *last = NULL, *tmp;
unsigned long now = jiffies;
- struct mlx5_fc *counter = NULL;
- struct mlx5_fc *last = NULL;
struct rb_node *node;
if (tmplist || !RB_EMPTY_ROOT(&fc_stats->counters))
@@ -183,26 +182,17 @@ static void mlx5_fc_stats_work(struct work_struct *work)
llist_for_each_entry(counter, tmplist, addlist)
mlx5_fc_stats_insert(&fc_stats->counters, counter);
- node = rb_first(&fc_stats->counters);
- while (node) {
- counter = rb_entry(node, struct mlx5_fc, node);
-
- node = rb_next(node);
-
- if (counter->deleted) {
- rb_erase(&counter->node, &fc_stats->counters);
-
- mlx5_cmd_fc_free(dev, counter->id);
-
- kfree(counter);
- continue;
- }
+ tmplist = llist_del_all(&fc_stats->dellist);
+ llist_for_each_entry_safe(counter, tmp, tmplist, dellist) {
+ rb_erase(&counter->node, &fc_stats->counters);
- last = counter;
+ mlx5_free_fc(dev, counter);
}
- if (time_before(now, fc_stats->next_query) || !last)
+ node = rb_last(&fc_stats->counters);
+ if (time_before(now, fc_stats->next_query) || !node)
return;
+ last = rb_entry(node, struct mlx5_fc, node);
node = rb_first(&fc_stats->counters);
while (node) {
@@ -254,13 +244,12 @@ void mlx5_fc_destroy(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
return;
if (counter->aging) {
- counter->deleted = true;
+ llist_add(&counter->dellist, &fc_stats->dellist);
mod_delayed_work(fc_stats->wq, &fc_stats->work, 0);
return;
}
- mlx5_cmd_fc_free(dev, counter->id);
- kfree(counter);
+ mlx5_free_fc(dev, counter);
}
EXPORT_SYMBOL(mlx5_fc_destroy);
@@ -270,6 +259,7 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
fc_stats->counters = RB_ROOT;
init_llist_head(&fc_stats->addlist);
+ init_llist_head(&fc_stats->dellist);
fc_stats->wq = create_singlethread_workqueue("mlx5_fc");
if (!fc_stats->wq)
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index c00549293982..4b53ac64004b 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -585,6 +585,7 @@ struct mlx5_irq_info {
struct mlx5_fc_stats {
struct rb_root counters;
struct llist_head addlist;
+ struct llist_head dellist;
struct workqueue_struct *wq;
struct delayed_work work;
--
2.17.1
^ permalink raw reply related
* [net-next 3/9] net/mlx5: Store flow counters in a list
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Vlad Buslov, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Vlad Buslov <vladbu@mellanox.com>
In order to improve performance of flow counter stats query loop that
traverses all configured flow counters, replace rb_tree with double-linked
list. This change improves performance of traversing flow counters by
removing the tree traversal. (profiling data showed that call to rb_next
was most top CPU consumer)
However, lookup of flow flow counter in list becomes linear, instead of
logarithmic. This problem is fixed by next patch in series, which adds idr
for fast lookup. Idr is to be used because it is not an intrusive data
structure and doesn't require adding any new members to struct mlx5_fc,
which allows its control data part to stay <= 1 cache line in size.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../net/ethernet/mellanox/mlx5/core/fs_core.h | 2 +-
.../ethernet/mellanox/mlx5/core/fs_counters.c | 88 +++++++++----------
include/linux/mlx5/driver.h | 2 +-
3 files changed, 42 insertions(+), 50 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
index 617d6239c5f3..a06f83c0c2b6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
@@ -139,7 +139,7 @@ struct mlx5_fc_cache {
};
struct mlx5_fc {
- struct rb_node node;
+ struct list_head list;
struct llist_node addlist;
struct llist_node dellist;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index f1266f215a31..90ebfee37508 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -73,36 +73,38 @@
* elapsed, the thread will actually query the hardware.
*/
-static void mlx5_fc_stats_insert(struct rb_root *root, struct mlx5_fc *counter)
+static struct list_head *mlx5_fc_counters_lookup_next(struct mlx5_core_dev *dev,
+ u32 id)
{
- struct rb_node **new = &root->rb_node;
- struct rb_node *parent = NULL;
-
- while (*new) {
- struct mlx5_fc *this = rb_entry(*new, struct mlx5_fc, node);
- int result = counter->id - this->id;
-
- parent = *new;
- if (result < 0)
- new = &((*new)->rb_left);
- else
- new = &((*new)->rb_right);
- }
+ struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ struct mlx5_fc *counter;
+
+ list_for_each_entry(counter, &fc_stats->counters, list)
+ if (counter->id > id)
+ return &counter->list;
+
+ return &fc_stats->counters;
+}
+
+static void mlx5_fc_stats_insert(struct mlx5_core_dev *dev,
+ struct mlx5_fc *counter)
+{
+ struct list_head *next = mlx5_fc_counters_lookup_next(dev, counter->id);
- /* Add new node and rebalance tree. */
- rb_link_node(&counter->node, parent, new);
- rb_insert_color(&counter->node, root);
+ list_add_tail(&counter->list, next);
}
-/* The function returns the last node that was queried so the caller
+/* The function returns the last counter that was queried so the caller
* function can continue calling it till all counters are queried.
*/
-static struct rb_node *mlx5_fc_stats_query(struct mlx5_core_dev *dev,
+static struct mlx5_fc *mlx5_fc_stats_query(struct mlx5_core_dev *dev,
struct mlx5_fc *first,
u32 last_id)
{
+ struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+ struct mlx5_fc *counter = NULL;
struct mlx5_cmd_fc_bulk *b;
- struct rb_node *node = NULL;
+ bool more = false;
u32 afirst_id;
int num;
int err;
@@ -132,14 +134,16 @@ static struct rb_node *mlx5_fc_stats_query(struct mlx5_core_dev *dev,
goto out;
}
- for (node = &first->node; node; node = rb_next(node)) {
- struct mlx5_fc *counter = rb_entry(node, struct mlx5_fc, node);
+ counter = first;
+ list_for_each_entry_from(counter, &fc_stats->counters, list) {
struct mlx5_fc_cache *c = &counter->cache;
u64 packets;
u64 bytes;
- if (counter->id > last_id)
+ if (counter->id > last_id) {
+ more = true;
break;
+ }
mlx5_cmd_fc_bulk_get(dev, b,
counter->id, &packets, &bytes);
@@ -155,7 +159,7 @@ static struct rb_node *mlx5_fc_stats_query(struct mlx5_core_dev *dev,
out:
mlx5_cmd_fc_bulk_free(b);
- return node;
+ return more ? counter : NULL;
}
static void mlx5_free_fc(struct mlx5_core_dev *dev,
@@ -173,33 +177,30 @@ static void mlx5_fc_stats_work(struct work_struct *work)
struct llist_node *tmplist = llist_del_all(&fc_stats->addlist);
struct mlx5_fc *counter = NULL, *last = NULL, *tmp;
unsigned long now = jiffies;
- struct rb_node *node;
- if (tmplist || !RB_EMPTY_ROOT(&fc_stats->counters))
+ if (tmplist || !list_empty(&fc_stats->counters))
queue_delayed_work(fc_stats->wq, &fc_stats->work,
fc_stats->sampling_interval);
llist_for_each_entry(counter, tmplist, addlist)
- mlx5_fc_stats_insert(&fc_stats->counters, counter);
+ mlx5_fc_stats_insert(dev, counter);
tmplist = llist_del_all(&fc_stats->dellist);
llist_for_each_entry_safe(counter, tmp, tmplist, dellist) {
- rb_erase(&counter->node, &fc_stats->counters);
+ list_del(&counter->list);
mlx5_free_fc(dev, counter);
}
- node = rb_last(&fc_stats->counters);
- if (time_before(now, fc_stats->next_query) || !node)
+ if (time_before(now, fc_stats->next_query) ||
+ list_empty(&fc_stats->counters))
return;
- last = rb_entry(node, struct mlx5_fc, node);
-
- node = rb_first(&fc_stats->counters);
- while (node) {
- counter = rb_entry(node, struct mlx5_fc, node);
+ last = list_last_entry(&fc_stats->counters, struct mlx5_fc, list);
- node = mlx5_fc_stats_query(dev, counter, last->id);
- }
+ counter = list_first_entry(&fc_stats->counters, struct mlx5_fc,
+ list);
+ while (counter)
+ counter = mlx5_fc_stats_query(dev, counter, last->id);
fc_stats->next_query = now + fc_stats->sampling_interval;
}
@@ -257,7 +258,7 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
{
struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
- fc_stats->counters = RB_ROOT;
+ INIT_LIST_HEAD(&fc_stats->counters);
init_llist_head(&fc_stats->addlist);
init_llist_head(&fc_stats->dellist);
@@ -277,7 +278,6 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
struct llist_node *tmplist;
struct mlx5_fc *counter;
struct mlx5_fc *tmp;
- struct rb_node *node;
cancel_delayed_work_sync(&dev->priv.fc_stats.work);
destroy_workqueue(dev->priv.fc_stats.wq);
@@ -287,16 +287,8 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
llist_for_each_entry_safe(counter, tmp, tmplist, addlist)
mlx5_free_fc(dev, counter);
- node = rb_first(&fc_stats->counters);
- while (node) {
- counter = rb_entry(node, struct mlx5_fc, node);
-
- node = rb_next(node);
-
- rb_erase(&counter->node, &fc_stats->counters);
-
+ list_for_each_entry_safe(counter, tmp, &fc_stats->counters, list)
mlx5_free_fc(dev, counter);
- }
}
int mlx5_fc_query(struct mlx5_core_dev *dev, struct mlx5_fc *counter,
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 4b53ac64004b..61bed33e6675 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -583,7 +583,7 @@ struct mlx5_irq_info {
};
struct mlx5_fc_stats {
- struct rb_root counters;
+ struct list_head counters;
struct llist_head addlist;
struct llist_head dellist;
--
2.17.1
^ permalink raw reply related
* [net-next 5/9] net/mlx5e: Move mlx5e_priv_flags into en_ethtool.c
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Kamal Heib, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Kamal Heib <kamalheib1@gmail.com>
Move the definition of mlx5e_priv_flags into en_ethtool.c because it's
only used there.
Fixes: 4e59e2888139 ("net/mlx5e: Introduce net device priv flags infrastructure")
Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 7 -------
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c | 7 +++++++
2 files changed, 7 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index db2cfcd21d43..de0f7702c86a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -204,13 +204,6 @@ struct mlx5e_umr_wqe {
extern const char mlx5e_self_tests[][ETH_GSTRING_LEN];
-static const char mlx5e_priv_flags[][ETH_GSTRING_LEN] = {
- "rx_cqe_moder",
- "tx_cqe_moder",
- "rx_cqe_compress",
- "rx_striding_rq",
-};
-
enum mlx5e_priv_flag {
MLX5E_PFLAG_RX_CQE_BASED_MODER = (1 << 0),
MLX5E_PFLAG_TX_CQE_BASED_MODER = (1 << 1),
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
index 98dd3e0ada72..8cd338ceb237 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
@@ -135,6 +135,13 @@ void mlx5e_build_ptys2ethtool_map(void)
ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT);
}
+static const char mlx5e_priv_flags[][ETH_GSTRING_LEN] = {
+ "rx_cqe_moder",
+ "tx_cqe_moder",
+ "rx_cqe_compress",
+ "rx_striding_rq",
+};
+
int mlx5e_ethtool_get_sset_count(struct mlx5e_priv *priv, int sset)
{
int i, num_stats = 0;
--
2.17.1
^ permalink raw reply related
* [net-next 6/9] net/mlx5e: Move Q counters allocation and drop RQ to init_rx
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Roi Dayan, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Roi Dayan <roid@mellanox.com>
Not all profiles query the HW Q counters in update_stats() callback.
HW Q couners are limited per device and in case of representors all
their Q counters are allocated on the parent PF device.
Avoid reundant allocation of HW Q counters by moving the allocation
to init_rx profile callback.
Signed-off-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 6 +++
.../net/ethernet/mellanox/mlx5/core/en_main.c | 45 +++++++++----------
.../net/ethernet/mellanox/mlx5/core/en_rep.c | 12 ++++-
.../ethernet/mellanox/mlx5/core/ipoib/ipoib.c | 17 ++++++-
4 files changed, 55 insertions(+), 25 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index de0f7702c86a..01a967e717e7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -898,6 +898,12 @@ void mlx5e_destroy_mdev_resources(struct mlx5_core_dev *mdev);
int mlx5e_refresh_tirs(struct mlx5e_priv *priv, bool enable_uc_lb);
/* common netdev helpers */
+void mlx5e_create_q_counters(struct mlx5e_priv *priv);
+void mlx5e_destroy_q_counters(struct mlx5e_priv *priv);
+int mlx5e_open_drop_rq(struct mlx5e_priv *priv,
+ struct mlx5e_rq *drop_rq);
+void mlx5e_close_drop_rq(struct mlx5e_rq *drop_rq);
+
int mlx5e_create_indirect_rqt(struct mlx5e_priv *priv);
int mlx5e_create_indirect_tirs(struct mlx5e_priv *priv);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 5a7939e70190..d14c4051edd8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -3049,8 +3049,8 @@ static int mlx5e_alloc_drop_cq(struct mlx5_core_dev *mdev,
return mlx5e_alloc_cq_common(mdev, param, cq);
}
-static int mlx5e_open_drop_rq(struct mlx5e_priv *priv,
- struct mlx5e_rq *drop_rq)
+int mlx5e_open_drop_rq(struct mlx5e_priv *priv,
+ struct mlx5e_rq *drop_rq)
{
struct mlx5_core_dev *mdev = priv->mdev;
struct mlx5e_cq_param cq_param = {};
@@ -3094,7 +3094,7 @@ static int mlx5e_open_drop_rq(struct mlx5e_priv *priv,
return err;
}
-static void mlx5e_close_drop_rq(struct mlx5e_rq *drop_rq)
+void mlx5e_close_drop_rq(struct mlx5e_rq *drop_rq)
{
mlx5e_destroy_rq(drop_rq);
mlx5e_free_rq(drop_rq);
@@ -4726,7 +4726,7 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev)
mlx5e_tls_build_netdev(priv);
}
-static void mlx5e_create_q_counters(struct mlx5e_priv *priv)
+void mlx5e_create_q_counters(struct mlx5e_priv *priv)
{
struct mlx5_core_dev *mdev = priv->mdev;
int err;
@@ -4744,7 +4744,7 @@ static void mlx5e_create_q_counters(struct mlx5e_priv *priv)
}
}
-static void mlx5e_destroy_q_counters(struct mlx5e_priv *priv)
+void mlx5e_destroy_q_counters(struct mlx5e_priv *priv)
{
if (priv->q_counter)
mlx5_core_dealloc_q_counter(priv->mdev, priv->q_counter);
@@ -4783,9 +4783,17 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
struct mlx5_core_dev *mdev = priv->mdev;
int err;
+ mlx5e_create_q_counters(priv);
+
+ err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
+ if (err) {
+ mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
+ goto err_destroy_q_counters;
+ }
+
err = mlx5e_create_indirect_rqt(priv);
if (err)
- return err;
+ goto err_close_drop_rq;
err = mlx5e_create_direct_rqts(priv);
if (err)
@@ -4821,6 +4829,10 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
mlx5e_destroy_direct_rqts(priv);
err_destroy_indirect_rqts:
mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+err_close_drop_rq:
+ mlx5e_close_drop_rq(&priv->drop_rq);
+err_destroy_q_counters:
+ mlx5e_destroy_q_counters(priv);
return err;
}
@@ -4832,6 +4844,8 @@ static void mlx5e_cleanup_nic_rx(struct mlx5e_priv *priv)
mlx5e_destroy_indirect_tirs(priv);
mlx5e_destroy_direct_rqts(priv);
mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+ mlx5e_close_drop_rq(&priv->drop_rq);
+ mlx5e_destroy_q_counters(priv);
}
static int mlx5e_init_nic_tx(struct mlx5e_priv *priv)
@@ -4975,7 +4989,6 @@ struct net_device *mlx5e_create_netdev(struct mlx5_core_dev *mdev,
int mlx5e_attach_netdev(struct mlx5e_priv *priv)
{
- struct mlx5_core_dev *mdev = priv->mdev;
const struct mlx5e_profile *profile;
int err;
@@ -4986,28 +4999,16 @@ int mlx5e_attach_netdev(struct mlx5e_priv *priv)
if (err)
goto out;
- mlx5e_create_q_counters(priv);
-
- err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
- if (err) {
- mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
- goto err_destroy_q_counters;
- }
-
err = profile->init_rx(priv);
if (err)
- goto err_close_drop_rq;
+ goto err_cleanup_tx;
if (profile->enable)
profile->enable(priv);
return 0;
-err_close_drop_rq:
- mlx5e_close_drop_rq(&priv->drop_rq);
-
-err_destroy_q_counters:
- mlx5e_destroy_q_counters(priv);
+err_cleanup_tx:
profile->cleanup_tx(priv);
out:
@@ -5025,8 +5026,6 @@ void mlx5e_detach_netdev(struct mlx5e_priv *priv)
flush_workqueue(priv->wq);
profile->cleanup_rx(priv);
- mlx5e_close_drop_rq(&priv->drop_rq);
- mlx5e_destroy_q_counters(priv);
profile->cleanup_tx(priv);
cancel_delayed_work_sync(&priv->update_stats_work);
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index c9cc9747d21d..f6eead24931f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -999,14 +999,21 @@ static int mlx5e_init_rep_rx(struct mlx5e_priv *priv)
struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
struct mlx5e_rep_priv *rpriv = priv->ppriv;
struct mlx5_eswitch_rep *rep = rpriv->rep;
+ struct mlx5_core_dev *mdev = priv->mdev;
struct mlx5_flow_handle *flow_rule;
int err;
mlx5e_init_l2_addr(priv);
+ err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
+ if (err) {
+ mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
+ return err;
+ }
+
err = mlx5e_create_direct_rqts(priv);
if (err)
- return err;
+ goto err_close_drop_rq;
err = mlx5e_create_direct_tirs(priv);
if (err)
@@ -1027,6 +1034,8 @@ static int mlx5e_init_rep_rx(struct mlx5e_priv *priv)
mlx5e_destroy_direct_tirs(priv);
err_destroy_direct_rqts:
mlx5e_destroy_direct_rqts(priv);
+err_close_drop_rq:
+ mlx5e_close_drop_rq(&priv->drop_rq);
return err;
}
@@ -1037,6 +1046,7 @@ static void mlx5e_cleanup_rep_rx(struct mlx5e_priv *priv)
mlx5_del_flow_rules(rpriv->vport_rx_rule);
mlx5e_destroy_direct_tirs(priv);
mlx5e_destroy_direct_rqts(priv);
+ mlx5e_close_drop_rq(&priv->drop_rq);
}
static int mlx5e_init_rep_tx(struct mlx5e_priv *priv)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c b/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c
index 3dd9f885ed8b..a825ed093efd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c
@@ -349,11 +349,20 @@ static void mlx5i_destroy_flow_steering(struct mlx5e_priv *priv)
static int mlx5i_init_rx(struct mlx5e_priv *priv)
{
+ struct mlx5_core_dev *mdev = priv->mdev;
int err;
+ mlx5e_create_q_counters(priv);
+
+ err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
+ if (err) {
+ mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
+ goto err_destroy_q_counters;
+ }
+
err = mlx5e_create_indirect_rqt(priv);
if (err)
- return err;
+ goto err_close_drop_rq;
err = mlx5e_create_direct_rqts(priv);
if (err)
@@ -381,6 +390,10 @@ static int mlx5i_init_rx(struct mlx5e_priv *priv)
mlx5e_destroy_direct_rqts(priv);
err_destroy_indirect_rqts:
mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+err_close_drop_rq:
+ mlx5e_close_drop_rq(&priv->drop_rq);
+err_destroy_q_counters:
+ mlx5e_destroy_q_counters(priv);
return err;
}
@@ -391,6 +404,8 @@ static void mlx5i_cleanup_rx(struct mlx5e_priv *priv)
mlx5e_destroy_indirect_tirs(priv);
mlx5e_destroy_direct_rqts(priv);
mlx5e_destroy_rqt(priv, &priv->indir_rqt);
+ mlx5e_close_drop_rq(&priv->drop_rq);
+ mlx5e_destroy_q_counters(priv);
}
static const struct mlx5e_profile mlx5i_nic_profile = {
--
2.17.1
^ permalink raw reply related
* [net-next 7/9] net/mlx5e: Replace PTP clock lock from RW lock to seq lock
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Shay Agroskin, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Shay Agroskin <shayag@mellanox.com>
Changed "priv.clock.lock" lock from 'rw_lock' to 'seq_lock'
in order to improve packet rate performance.
Tested on Intel(R) Xeon(R) CPU E5-2660 v2 @ 2.20GHz.
Sent 64b packets between two peers connected by ConnectX-5,
and measured packet rate for the receiver in three modes:
no time-stamping (base rate)
time-stamping using rw_lock (old lock) for critical region
time-stamping using seq_lock (new lock) for critical region
Only the receiver time stamped its packets.
The measured packet rate improvements are:
Single flow (multiple TX rings to single RX ring):
without timestamping: 4.26 (M packets)/sec
with rw-lock (old lock): 4.1 (M packets)/sec
with seq-lock (new lock): 4.16 (M packets)/sec
1.46% improvement
Multiple flows (multiple TX rings to six RX rings):
without timestamping: 22 (M packets)/sec
with rw-lock (old lock): 11.7 (M packets)/sec
with seq-lock (new lock): 21.3 (M packets)/sec
82.05% improvement
The packet rate improvement is due to the lack of atomic operations
for the 'readers' by the seq-lock.
Since there are much more 'readers' than 'writers' contention
on this lock, almost all atomic operations are saved.
this results in a dramatic decrease in overall
cache misses.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../ethernet/mellanox/mlx5/core/lib/clock.c | 34 +++++++++----------
.../ethernet/mellanox/mlx5/core/lib/clock.h | 8 +++--
include/linux/mlx5/driver.h | 2 +-
3 files changed, 23 insertions(+), 21 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 3f767cde4c1d..0d90b1b4a3d3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -111,10 +111,10 @@ static void mlx5_pps_out(struct work_struct *work)
for (i = 0; i < clock->ptp_info.n_pins; i++) {
u64 tstart;
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
tstart = clock->pps_info.start[i];
clock->pps_info.start[i] = 0;
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
if (!tstart)
continue;
@@ -132,10 +132,10 @@ static void mlx5_timestamp_overflow(struct work_struct *work)
overflow_work);
unsigned long flags;
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
timecounter_read(&clock->tc);
mlx5_update_clock_info_page(clock->mdev);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
schedule_delayed_work(&clock->overflow_work, clock->overflow_period);
}
@@ -147,10 +147,10 @@ static int mlx5_ptp_settime(struct ptp_clock_info *ptp,
u64 ns = timespec64_to_ns(ts);
unsigned long flags;
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
timecounter_init(&clock->tc, &clock->cycles, ns);
mlx5_update_clock_info_page(clock->mdev);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
return 0;
}
@@ -162,9 +162,9 @@ static int mlx5_ptp_gettime(struct ptp_clock_info *ptp, struct timespec64 *ts)
u64 ns;
unsigned long flags;
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
ns = timecounter_read(&clock->tc);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
*ts = ns_to_timespec64(ns);
@@ -177,10 +177,10 @@ static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
ptp_info);
unsigned long flags;
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
timecounter_adjtime(&clock->tc, delta);
mlx5_update_clock_info_page(clock->mdev);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
return 0;
}
@@ -203,12 +203,12 @@ static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta)
adj *= delta;
diff = div_u64(adj, 1000000000ULL);
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
timecounter_read(&clock->tc);
clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff :
clock->nominal_c_mult + diff;
mlx5_update_clock_info_page(clock->mdev);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
return 0;
}
@@ -307,12 +307,12 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
ts.tv_nsec = rq->perout.start.nsec;
ns = timespec64_to_ns(&ts);
cycles_now = mlx5_read_internal_timer(mdev);
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
nsec_now = timecounter_cyc2time(&clock->tc, cycles_now);
nsec_delta = ns - nsec_now;
cycles_delta = div64_u64(nsec_delta << clock->cycles.shift,
clock->cycles.mult);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
time_stamp = cycles_now + cycles_delta;
field_select = MLX5_MTPPS_FS_PIN_MODE |
MLX5_MTPPS_FS_PATTERN |
@@ -471,14 +471,14 @@ void mlx5_pps_event(struct mlx5_core_dev *mdev,
ts.tv_sec += 1;
ts.tv_nsec = 0;
ns = timespec64_to_ns(&ts);
- write_lock_irqsave(&clock->lock, flags);
+ write_seqlock_irqsave(&clock->lock, flags);
nsec_now = timecounter_cyc2time(&clock->tc, cycles_now);
nsec_delta = ns - nsec_now;
cycles_delta = div64_u64(nsec_delta << clock->cycles.shift,
clock->cycles.mult);
clock->pps_info.start[pin] = cycles_now + cycles_delta;
schedule_work(&clock->pps_info.out_work);
- write_unlock_irqrestore(&clock->lock, flags);
+ write_sequnlock_irqrestore(&clock->lock, flags);
break;
default:
mlx5_core_err(mdev, " Unhandled event\n");
@@ -498,7 +498,7 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
mlx5_core_warn(mdev, "invalid device_frequency_khz, aborting HW clock init\n");
return;
}
- rwlock_init(&clock->lock);
+ seqlock_init(&clock->lock);
clock->cycles.read = read_internal_timer;
clock->cycles.shift = MLX5_CYCLES_SHIFT;
clock->cycles.mult = clocksource_khz2mult(dev_freq,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
index 02e2e4575e4f..263cb6e2aeee 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.h
@@ -46,11 +46,13 @@ static inline int mlx5_clock_get_ptp_index(struct mlx5_core_dev *mdev)
static inline ktime_t mlx5_timecounter_cyc2time(struct mlx5_clock *clock,
u64 timestamp)
{
+ unsigned int seq;
u64 nsec;
- read_lock(&clock->lock);
- nsec = timecounter_cyc2time(&clock->tc, timestamp);
- read_unlock(&clock->lock);
+ do {
+ seq = read_seqbegin(&clock->lock);
+ nsec = timecounter_cyc2time(&clock->tc, timestamp);
+ } while (read_seqretry(&clock->lock, seq));
return ns_to_ktime(nsec);
}
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 2a0c845f6bdb..b7fce2c9443d 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -805,7 +805,7 @@ struct mlx5_pps {
};
struct mlx5_clock {
- rwlock_t lock;
+ seqlock_t lock;
struct cyclecounter cycles;
struct timecounter tc;
struct hwtstamp_config hwtstamp_config;
--
2.17.1
^ permalink raw reply related
* [net-next 9/9] net/mlx5e: don't set CHECKSUM_COMPLETE on SCTP packets
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Alaa Hleihel, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Alaa Hleihel <alaa@mellanox.com>
CHECKSUM_COMPLETE is not applicable to SCTP protocol.
Setting it for SCTP packets leads to CRC32c validation failure.
Fixes: bbceefce9adf ("net/mlx5e: Support RX CHECKSUM_COMPLETE")
Signed-off-by: Alaa Hleihel <alaa@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index 2175d6972dc3..424bc89184c6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -755,6 +755,14 @@ static __be32 mlx5e_get_fcs(struct sk_buff *skb)
return fcs_bytes;
}
+static u8 get_ip_proto(struct sk_buff *skb, __be16 proto)
+{
+ void *ip_p = skb->data + sizeof(struct ethhdr);
+
+ return (proto == htons(ETH_P_IP)) ? ((struct iphdr *)ip_p)->protocol :
+ ((struct ipv6hdr *)ip_p)->nexthdr;
+}
+
static inline void mlx5e_handle_csum(struct net_device *netdev,
struct mlx5_cqe64 *cqe,
struct mlx5e_rq *rq,
@@ -775,6 +783,9 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
}
if (likely(is_last_ethertype_ip(skb, &network_depth, &proto))) {
+ if (unlikely(get_ip_proto(skb, proto) == IPPROTO_SCTP))
+ goto csum_unnecessary;
+
skb->ip_summed = CHECKSUM_COMPLETE;
skb->csum = csum_unfold((__force __sum16)cqe->check_sum);
if (network_depth > ETH_HLEN)
@@ -792,6 +803,7 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
return;
}
+csum_unnecessary:
if (likely((cqe->hds_ip_ext & CQE_L3_OK) &&
(cqe->hds_ip_ext & CQE_L4_OK))) {
skb->ip_summed = CHECKSUM_UNNECESSARY;
--
2.17.1
^ permalink raw reply related
* [net-next 8/9] net/mlx5e: Set ECN for received packets using CQE indication
From: Saeed Mahameed @ 2018-09-06 4:33 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Natali Shechtman, Saeed Mahameed
In-Reply-To: <20180906043331.31958-1-saeedm@mellanox.com>
From: Natali Shechtman <natali@mellanox.com>
In multi-host (MH) NIC scheme, a single HW port serves multiple hosts
or sockets on the same host.
The HW uses a mechanism in the PCIe buffer which monitors
the amount of consumed PCIe buffers per host.
On a certain configuration, under congestion,
the HW emulates a switch doing ECN marking on packets using ECN
indication on the completion descriptor (CQE).
The driver needs to set the ECN bits on the packet SKB,
such that the network stack can react on that, this commit does that.
Signed-off-by: Natali Shechtman <natali@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
.../net/ethernet/mellanox/mlx5/core/en_rx.c | 35 ++++++++++++++++---
.../ethernet/mellanox/mlx5/core/en_stats.c | 3 ++
.../ethernet/mellanox/mlx5/core/en_stats.h | 2 ++
3 files changed, 35 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index 6a959e8b1f9d..2175d6972dc3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -37,6 +37,7 @@
#include <net/busy_poll.h>
#include <net/ip6_checksum.h>
#include <net/page_pool.h>
+#include <net/inet_ecn.h>
#include "en.h"
#include "en_tc.h"
#include "eswitch.h"
@@ -690,12 +691,29 @@ static inline void mlx5e_skb_set_hash(struct mlx5_cqe64 *cqe,
skb_set_hash(skb, be32_to_cpu(cqe->rss_hash_result), ht);
}
-static inline bool is_last_ethertype_ip(struct sk_buff *skb, int *network_depth)
+static inline bool is_last_ethertype_ip(struct sk_buff *skb, int *network_depth,
+ __be16 *proto)
{
- __be16 ethertype = ((struct ethhdr *)skb->data)->h_proto;
+ *proto = ((struct ethhdr *)skb->data)->h_proto;
+ *proto = __vlan_get_protocol(skb, *proto, network_depth);
+ return (*proto == htons(ETH_P_IP) || *proto == htons(ETH_P_IPV6));
+}
+
+static inline void mlx5e_enable_ecn(struct mlx5e_rq *rq, struct sk_buff *skb)
+{
+ int network_depth = 0;
+ __be16 proto;
+ void *ip;
+ int rc;
- ethertype = __vlan_get_protocol(skb, ethertype, network_depth);
- return (ethertype == htons(ETH_P_IP) || ethertype == htons(ETH_P_IPV6));
+ if (unlikely(!is_last_ethertype_ip(skb, &network_depth, &proto)))
+ return;
+
+ ip = skb->data + network_depth;
+ rc = ((proto == htons(ETH_P_IP)) ? IP_ECN_set_ce((struct iphdr *)ip) :
+ IP6_ECN_set_ce(skb, (struct ipv6hdr *)ip));
+
+ rq->stats->ecn_mark += !!rc;
}
static __be32 mlx5e_get_fcs(struct sk_buff *skb)
@@ -745,6 +763,7 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
{
struct mlx5e_rq_stats *stats = rq->stats;
int network_depth = 0;
+ __be16 proto;
if (unlikely(!(netdev->features & NETIF_F_RXCSUM)))
goto csum_none;
@@ -755,7 +774,7 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
return;
}
- if (likely(is_last_ethertype_ip(skb, &network_depth))) {
+ if (likely(is_last_ethertype_ip(skb, &network_depth, &proto))) {
skb->ip_summed = CHECKSUM_COMPLETE;
skb->csum = csum_unfold((__force __sum16)cqe->check_sum);
if (network_depth > ETH_HLEN)
@@ -790,6 +809,8 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
stats->csum_none++;
}
+#define MLX5E_CE_BIT_MASK 0x80
+
static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,
u32 cqe_bcnt,
struct mlx5e_rq *rq,
@@ -834,6 +855,10 @@ static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,
skb->mark = be32_to_cpu(cqe->sop_drop_qpn) & MLX5E_TC_FLOW_ID_MASK;
mlx5e_handle_csum(netdev, cqe, rq, skb, !!lro_num_seg);
+ /* checking CE bit in cqe - MSB in ml_path field */
+ if (unlikely(cqe->ml_path & MLX5E_CE_BIT_MASK))
+ mlx5e_enable_ecn(rq, skb);
+
skb->protocol = eth_type_trans(skb, netdev);
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
index 6839481f7697..90c7607b1f44 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
@@ -53,6 +53,7 @@ static const struct counter_desc sw_stats_desc[] = {
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_lro_packets) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_lro_bytes) },
+ { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_ecn_mark) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_removed_vlan_packets) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_csum_unnecessary) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_csum_none) },
@@ -144,6 +145,7 @@ void mlx5e_grp_sw_update_stats(struct mlx5e_priv *priv)
s->rx_bytes += rq_stats->bytes;
s->rx_lro_packets += rq_stats->lro_packets;
s->rx_lro_bytes += rq_stats->lro_bytes;
+ s->rx_ecn_mark += rq_stats->ecn_mark;
s->rx_removed_vlan_packets += rq_stats->removed_vlan_packets;
s->rx_csum_none += rq_stats->csum_none;
s->rx_csum_complete += rq_stats->csum_complete;
@@ -1144,6 +1146,7 @@ static const struct counter_desc rq_stats_desc[] = {
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, xdp_redirect) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, lro_packets) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, lro_bytes) },
+ { MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, ecn_mark) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, removed_vlan_packets) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, wqe_err) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, mpwqe_filler_cqes) },
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
index a4c035aedd46..a5fb3dc27f50 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
@@ -66,6 +66,7 @@ struct mlx5e_sw_stats {
u64 tx_nop;
u64 rx_lro_packets;
u64 rx_lro_bytes;
+ u64 rx_ecn_mark;
u64 rx_removed_vlan_packets;
u64 rx_csum_unnecessary;
u64 rx_csum_none;
@@ -184,6 +185,7 @@ struct mlx5e_rq_stats {
u64 csum_none;
u64 lro_packets;
u64 lro_bytes;
+ u64 ecn_mark;
u64 removed_vlan_packets;
u64 xdp_drop;
u64 xdp_redirect;
--
2.17.1
^ permalink raw reply related
* Re: [PATCH] ath9k: add reset for airtime station debugfs
From: Toke Høiland-Jørgensen @ 2018-09-06 9:27 UTC (permalink / raw)
To: Louie Lu, kvalo-sgV2jX0FEOL9JmXXK+q4OQ
Cc: ath9k-devel-A+ZNKFmMK5xy9aJCnZT0Uw, davem-fT/PcQaiUtIeIZ0/mPfg9Q,
linux-wireless-u79uwXL29TY76Z2rM5mHXA,
netdev-u79uwXL29TY76Z2rM5mHXA
In-Reply-To: <20180906035323.GA11337@nems_nctu_lu>
Louie Lu <git-sgRGTR89XAc@public.gmane.org> writes:
> Let user can reset station airtime status by debugfs, it will
> reset all airtime deficit to ATH_AIRTIME_QUANTUM and reset rx/tx
> airtime accumulate to 0.
No objections to the patch, but I'm curious which issues you were
debugging that led you to needing it? :)
-Toke
^ permalink raw reply
* Re: [PATCH v2 net-next] packet: add sockopt to ignore outgoing packets
From: David Miller @ 2018-09-06 5:10 UTC (permalink / raw)
To: vincent.whitchurch; +Cc: netdev, willemb, rabinv
In-Reply-To: <20180903142336.32122-1-vincent.whitchurch@axis.com>
From: Vincent Whitchurch <vincent.whitchurch@axis.com>
Date: Mon, 3 Sep 2018 16:23:36 +0200
> Currently, the only way to ignore outgoing packets on a packet socket is
> via the BPF filter. With MSG_ZEROCOPY, packets that are looped into
> AF_PACKET are copied in dev_queue_xmit_nit(), and this copy happens even
> if the filter run from packet_rcv() would reject them. So the presence
> of a packet socket on the interface takes away the benefits of
> MSG_ZEROCOPY, even if the packet socket is not interested in outgoing
> packets. (Even when MSG_ZEROCOPY is not used, the skb is unnecessarily
> cloned, but the cost for that is much lower.)
>
> Add a socket option to allow AF_PACKET sockets to ignore outgoing
> packets to solve this. Note that the *BSDs already have something
> similar: BIOCSSEESENT/BIOCSDIRECTION and BIOCSDIRFILT.
>
> The first intended user is lldpd.
>
> Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
> ---
> v2: Stricter value validation.
> Moved ignore check out of skb_loop_sk().
Applied, thank you.
^ permalink raw reply
* Re: [PATCH] mac80211: fix TX status reporting for ieee80211s
From: Yuan-Chi Pang @ 2018-09-06 9:45 UTC (permalink / raw)
To: Johannes Berg; +Cc: David S. Miller, linux-wireless, netdev, linux-kernel
In-Reply-To: <1536224269-69350-1-git-send-email-fu3mo6goo@gmail.com>
On Thu, 2018-09-06 at 16:57 +0800, Yuan-Chi Pang wrote:
> TX status reporting to ieee80211s is through
> ieee80211s_update_metric.
> There are two problems about ieee80211s_update_metric:
>
> 1. The purpose is to estimate the fail probability
> to a specific link. No need to restrict to data frame.
>
> 2. Current implementation does not work if wireless driver does not
> pass tx_status with skb.
>
> Fix this by removing ieee80211_is_data condition, passing
> ieee80211_tx_status directly to ieee80211s_update_metric, and
> putting it in both __ieee80211_tx_status and ieee80211_tx_status_ext.
>
> Signed-off-by: Yuan-Chi Pang <fu3mo6goo@gmail.com>
> ---
> net/mac80211/mesh.h | 2 +-
> net/mac80211/mesh_hwmp.c | 8 ++------
> net/mac80211/status.c | 4 +++-
> 3 files changed, 6 insertions(+), 8 deletions(-)
>
> diff --git a/net/mac80211/mesh.h b/net/mac80211/mesh.h
> index ee56f18..7ffe56a 100644
> --- a/net/mac80211/mesh.h
> +++ b/net/mac80211/mesh.h
> @@ -217,7 +217,7 @@ void mesh_rmc_free(struct ieee80211_sub_if_data
> *sdata);
> int mesh_rmc_init(struct ieee80211_sub_if_data *sdata);
> void ieee80211s_init(void);
> void ieee80211s_update_metric(struct ieee80211_local *local,
> - struct sta_info *sta, struct sk_buff
> *skb);
> + struct sta_info *sta, struct
> ieee80211_tx_status *st);
> void ieee80211_mesh_init_sdata(struct ieee80211_sub_if_data *sdata);
> void ieee80211_mesh_teardown_sdata(struct ieee80211_sub_if_data
> *sdata);
> int ieee80211_start_mesh(struct ieee80211_sub_if_data *sdata);
I received the comment of my previous submission that I should fix
indentation. But if I use the same indentation as before, I got "line
over 80 characters" warning by checkpatch. What should I do?
> diff --git a/net/mac80211/mesh_hwmp.c b/net/mac80211/mesh_hwmp.c
> index daf9db3..4286673 100644
> --- a/net/mac80211/mesh_hwmp.c
> +++ b/net/mac80211/mesh_hwmp.c
> @@ -295,15 +295,11 @@ int mesh_path_error_tx(struct
> ieee80211_sub_if_data *sdata,
> }
>
> void ieee80211s_update_metric(struct ieee80211_local *local,
> - struct sta_info *sta, struct sk_buff *skb)
> + struct sta_info *sta, struct ieee80211_tx_status
> *st)
> {
> - struct ieee80211_tx_info *txinfo = IEEE80211_SKB_CB(skb);
> - struct ieee80211_hdr *hdr = (struct ieee80211_hdr *) skb-
> >data;
> + struct ieee80211_tx_info *txinfo = st->info;
> int failed;
>
> - if (!ieee80211_is_data(hdr->frame_control))
> - return;
> -
> failed = !(txinfo->flags & IEEE80211_TX_STAT_ACK);
>
> /* moving average, scaled to 100.
> diff --git a/net/mac80211/status.c b/net/mac80211/status.c
> index 9a6d720..dfcd419 100644
> --- a/net/mac80211/status.c
> +++ b/net/mac80211/status.c
> @@ -811,7 +811,7 @@ static void __ieee80211_tx_status(struct
> ieee80211_hw *hw,
>
> rate_control_tx_status(local, sband, status);
> if (ieee80211_vif_is_mesh(&sta->sdata->vif))
> - ieee80211s_update_metric(local, sta, skb);
> + ieee80211s_update_metric(local, sta,
> status);
>
> if (!(info->flags & IEEE80211_TX_CTL_INJECTED) &&
> acked)
> ieee80211_frame_acked(sta, skb);
> @@ -972,6 +972,8 @@ void ieee80211_tx_status_ext(struct ieee80211_hw
> *hw,
> }
>
> rate_control_tx_status(local, sband, status);
> + if (ieee80211_vif_is_mesh(&sta->sdata->vif))
> + ieee80211s_update_metric(local, sta,
> status);
> }
>
> if (acked || noack_success) {
^ permalink raw reply
* Re: [PATCH] mac80211: fix TX status reporting for ieee80211s
From: Johannes Berg @ 2018-09-06 9:47 UTC (permalink / raw)
To: Yuan-Chi Pang; +Cc: David S. Miller, linux-wireless, netdev, linux-kernel
In-Reply-To: <1536227104.68915.5.camel@gmail.com>
On Thu, 2018-09-06 at 17:45 +0800, Yuan-Chi Pang wrote:
>
> I received the comment of my previous submission that I should fix
> indentation. But if I use the same indentation as before, I got "line
> over 80 characters" warning by checkpatch. What should I do?
>
> > diff --git a/net/mac80211/mesh_hwmp.c b/net/mac80211/mesh_hwmp.c
> > index daf9db3..4286673 100644
> > --- a/net/mac80211/mesh_hwmp.c
> > +++ b/net/mac80211/mesh_hwmp.c
> > @@ -295,15 +295,11 @@ int mesh_path_error_tx(struct
> > ieee80211_sub_if_data *sdata,
> > }
> >
> > void ieee80211s_update_metric(struct ieee80211_local *local,
> > - struct sta_info *sta, struct sk_buff *skb)
> > + struct sta_info *sta, struct ieee80211_tx_status
> > *st)
Here? Seems easy - add a linebreak between sta and status?
johannes
^ permalink raw reply
* Re: [Patch net] tipc: orphan sock in tipc_release()
From: David Miller @ 2018-09-06 5:15 UTC (permalink / raw)
To: xiyou.wangcong; +Cc: netdev, tipc-discussion, jon.maloy, ying.xue
In-Reply-To: <20180904021241.11426-2-xiyou.wangcong@gmail.com>
From: Cong Wang <xiyou.wangcong@gmail.com>
Date: Mon, 3 Sep 2018 19:12:41 -0700
> Before we unlock the sock in tipc_release(), we have to
> detach sk->sk_socket from sk, otherwise a parallel
> tipc_sk_fill_sock_diag() could stil read it after we
> free this socket.
>
> Fixes: c30b70deb5f4 ("tipc: implement socket diagnostics for AF_TIPC")
> Reported-and-tested-by: syzbot+48804b87c16588ad491d@syzkaller.appspotmail.com
> Cc: Jon Maloy <jon.maloy@ericsson.com>
> Cc: Ying Xue <ying.xue@windriver.com>
> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Applied and queued up for -stable, thanks.
^ permalink raw reply
* Re: [PATCH v2 net-next] failover: Add missing check to validate 'slave_dev' in net_failover_slave_unregister
From: David Miller @ 2018-09-06 5:16 UTC (permalink / raw)
To: yuehaibing
Cc: sridhar.samudrala, stephen, dan.carpenter, alexander.h.duyck,
jeffrey.t.kirsher, liran.alon, joao.m.martins, netdev,
kernel-janitors
In-Reply-To: <1536029786-21710-1-git-send-email-yuehaibing@huawei.com>
From: YueHaibing <yuehaibing@huawei.com>
Date: Tue, 4 Sep 2018 02:56:26 +0000
> Fixes gcc '-Wunused-but-set-variable' warning:
>
> drivers/net/net_failover.c: In function 'net_failover_slave_unregister':
> drivers/net/net_failover.c:598:35: warning:
> variable 'primary_dev' set but not used [-Wunused-but-set-variable]
>
> There should check the validity of 'slave_dev'.
>
> Fixes: cfc80d9a1163 ("net: Introduce net_failover driver")
>
> Signed-off-by: YueHaibing <yuehaibing@huawei.com>
> ---
> v2: use WARN_ON_ONCE as Liran Alon suggested
Applied.
^ permalink raw reply
* Re: [PATCH net-next 0/3] nfp: improve the new rtsym helpers
From: David Miller @ 2018-09-06 5:18 UTC (permalink / raw)
To: jakub.kicinski; +Cc: netdev, oss-drivers
In-Reply-To: <20180904143733.16362-1-jakub.kicinski@netronome.com>
From: Jakub Kicinski <jakub.kicinski@netronome.com>
Date: Tue, 4 Sep 2018 07:37:30 -0700
> This set fixes a bug in ABS rtsym handling I added in net-next,
> it expands the error checking and reporting on the rtsym accesses.
Series applied.
^ permalink raw reply
* Re: [PATCH net-next] nfp: separate VXLAN and GRE feature handling
From: David Miller @ 2018-09-06 5:18 UTC (permalink / raw)
To: jakub.kicinski; +Cc: netdev, oss-drivers
In-Reply-To: <20180904152833.20351-1-jakub.kicinski@netronome.com>
From: Jakub Kicinski <jakub.kicinski@netronome.com>
Date: Tue, 4 Sep 2018 08:28:33 -0700
> VXLAN and GRE FW features have to currently be both advertised
> for the driver to enable them. Separate the handling.
>
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Applied.
^ permalink raw reply
* Re: [PATCH net] net/sched: fix memory leak in act_tunnel_key_init()
From: David Miller @ 2018-09-06 5:19 UTC (permalink / raw)
To: dcaratti; +Cc: xiyou.wangcong, simon.horman, amir, jhs, netdev
In-Reply-To: <d0a72d1371e790505a8141592c6af904c9b24031.1536079973.git.dcaratti@redhat.com>
From: Davide Caratti <dcaratti@redhat.com>
Date: Tue, 4 Sep 2018 19:00:19 +0200
> If users try to install act_tunnel_key 'set' rules with duplicate values
> of 'index', the tunnel metadata are allocated, but never released. Then,
> kmemleak complains as follows:
...
> This problem theoretically happens also in case users attempt to setup a
> geneve rule having wrong configuration data, or when the kernel fails to
> allocate 'params_new'. Ensure that tunnel_key_init() releases the tunnel
> metadata also in the above conditions.
>
> Addresses-Coverity-ID: 1373974 ("Resource leak")
> Fixes: d0f6dd8a914f4 ("net/sched: Introduce act_tunnel_key")
> Fixes: 0ed5269f9e41f ("net/sched: add tunnel option support to act_tunnel_key")
> Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Applied and queued up for -stable.
^ permalink raw reply
* Re: [PATCH RFC] net/mlx5_en: switch to Toeplitz RSS hash by default
From: Saeed Mahameed @ 2018-09-06 5:24 UTC (permalink / raw)
To: Konstantin Khlebnikov
Cc: Tariq Toukan, Linux Netdev List, Saeed Mahameed, Gal Pressman,
Or Gerlitz, David S. Miller
In-Reply-To: <a081d829-35cf-5dd7-2772-db27e818e717@yandex-team.ru>
On Sun, Sep 2, 2018 at 2:55 AM, Konstantin Khlebnikov
<khlebnikov@yandex-team.ru> wrote:
> On 02.09.2018 12:29, Tariq Toukan wrote:
>>
>>
>>
>> On 31/08/2018 2:29 PM, Konstantin Khlebnikov wrote:
>>>
>>> XOR (MLX5_RX_HASH_FN_INVERTED_XOR8) gives only 8 bits.
>>> It seems not enough for RFS. All other drivers use toeplitz.
>>>
>>> Driver mlx4_en uses Toeplitz by default and warns if hash XOR is used
>>> together with NETIF_F_RXHASH (enabled by default too): "Enabling both
>>> XOR Hash function and RX Hashing can limit RPS functionality".
>>>
>>> XOR is default in mlx5_en since commit 2be6967cdbc9
>>> ("net/mlx5e: Support ETH_RSS_HASH_XOR").
>>>
>>> Hash function could be set via ethtool. But it would be nice to have
>>> single standard for drivers or proper description why this one is
>>> special.
>>>
>>> Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
>>> ---
>>
>>
>> Hi Konstantin,
>>
>> Thanks for the patch.
>>
>> I understand the motivation.
>>
>> This change affects the default out-of-the-box behavior and requires a
>> full performance cycle. We'll run performance regression tomorrow, results
>> should be ready by EOW.
>> > I'll update.
>
>
> Ok, thank you.
>
> The only mention I've found in your documentation
> http://www.mellanox.com/related-docs/prod_software/Mellanox_EN_for_Linux_User_Manual_v4_4.pdf
>
> is
> ---
> 1.1.10 RSS Support
> 1.1.10.1 RSS Hash Function
> The device has the ability to use XOR as the RSS distribution function,
> instead of the default
> Toplitz function.
> The XOR function can be better distributed among driver's receive queues in
> small number of
> streams, where it distributes each TCP/UDP stream to a different queue.
> ---
>
> So Toeplitz is supposed to be default hash function for all versions of
> drivers and hardware.
>
> Also XOR8 seems vulnerable for ddos - hash is predictable, no random\secret
> vector, only 8 bits.
> So, it's easy to route all flows into one point. As we got it by accident.
>
> Moreover, in kernel 4.4.y hash switch via ethtool is broken and does not
> work =)
>
is it broken in mlx5 only or for the whole kernel ?
If it is mlx5 then this might be the reason:
commit 2d75b2bc8a8c0ce5567a6ecef52e194d117efe3f
net/mlx5e: Add ethtool RSS configuration options
was submitted to kernel 4.3
and an important fix for hash function change was submitted to 4.5:
commit bdfc028de1b3cd59490d5413a5c87b0fa50040c2
Author: Tariq Toukan <tariqt@mellanox.com>
Date: Mon Feb 29 21:17:12 2016 +0200
net/mlx5e: Fix ethtool RX hash func configuration change
We should modify TIRs explicitly to apply the new RSS configuration.
The light ndo close/open calls do not "refresh" them.
Fixes: 2d75b2bc8a8c ('net/mlx5e: Add ethtool RSS configuration options')
>>
>> Regards,
>> Tariq
^ permalink raw reply
* [PATCH] cfg80211: remove redundant check of !scan_plan
From: Colin King @ 2018-09-06 9:58 UTC (permalink / raw)
To: Johannes Berg, David S . Miller, linux-wireless, netdev
Cc: kernel-janitors, linux-kernel
From: Colin Ian King <colin.king@canonical.com>
The check for !scan_plan is redunant as this has been checked
in the proceeding statement and the code returns -ENOBUFS if
it is true. Remove the redundant check.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
net/wireless/nl80211.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c
index d5f9b5235cdd..b4908bcb0d77 100644
--- a/net/wireless/nl80211.c
+++ b/net/wireless/nl80211.c
@@ -10690,8 +10690,7 @@ static int nl80211_send_wowlan_nd(struct sk_buff *msg,
if (!scan_plan)
return -ENOBUFS;
- if (!scan_plan ||
- nla_put_u32(msg, NL80211_SCHED_SCAN_PLAN_INTERVAL,
+ if (nla_put_u32(msg, NL80211_SCHED_SCAN_PLAN_INTERVAL,
req->scan_plans[i].interval) ||
(req->scan_plans[i].iterations &&
nla_put_u32(msg, NL80211_SCHED_SCAN_PLAN_ITERATIONS,
--
2.17.1
^ permalink raw reply related
* Re: [PATCH bpf-next] bpf/verifier: properly clear union members after a ctx read
From: Alexei Starovoitov @ 2018-09-06 5:32 UTC (permalink / raw)
To: Edward Cree; +Cc: daniel, ast, netdev
In-Reply-To: <719de66a-0cd4-262a-d2d9-578df3a3fdca@solarflare.com>
On Wed, Sep 05, 2018 at 02:47:22PM +0100, Edward Cree wrote:
> On 05/09/18 03:23, Alexei Starovoitov wrote:
> > So would you agree it's fair to add
> > Fixes: f1174f77b50c ("bpf/verifier: rework value tracking")
> > ?
> Sure. Though I don't think it needs backporting, as it's a conservative
> bug (i.e. it merely prevents pruning, but that's safe security-wise).
agree. No backport necessary.
> > How about patch like the following:
> > ------------
> > From 422fd975ed78645ab67d2eb50ff6e1ff6fb3de32 Mon Sep 17 00:00:00 2001
> > From: Alexei Starovoitov <ast@kernel.org>
> > Date: Tue, 4 Sep 2018 19:13:44 -0700
> > Subject: [PATCH] bpf/verifier: fix verifier instability
> >
> > Fixes: f1174f77b50c ("bpf/verifier: rework value tracking")
> > Debugged-by: Edward Cree <ecree@solarflare.com>
> > Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> Acked-by: Edward Cree <ecree@solarflare.com>
Thanks. I copy-pasted your commit log and pushed to bpf-next.
Much appreciate the time and effort spent on root causing these issues.
^ permalink raw reply
* Re: [PATCH net 0/3] net/iucv: fixes 2018-09-05
From: David Miller @ 2018-09-06 5:32 UTC (permalink / raw)
To: jwi; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180905145512.30165-1-jwi@linux.ibm.com>
From: Julian Wiedmann <jwi@linux.ibm.com>
Date: Wed, 5 Sep 2018 16:55:09 +0200
> please apply three straight-forward fixes for iucv. One that prevents
> leaking the skb on malformed inbound packets, one to fix the error
> handling on transmit error, and one to get rid of a compile warning.
Series applied, thank you.
^ permalink raw reply
* Re: [PATCH bpf-next 0/4] i40e AF_XDP zero-copy buffer leak fixes
From: Alexei Starovoitov @ 2018-09-06 5:55 UTC (permalink / raw)
To: Björn Töpel
Cc: Jakub Kicinski, ast, Daniel Borkmann, Netdev, Jeff Kirsher,
intel-wired-lan, Björn Töpel, Karlsson, Magnus,
Magnus Karlsson
In-Reply-To: <CAJ+HfNjQCP7c7gCVJi=LRxK2LvQrNvn6cPfRACW6TPzM8339SA@mail.gmail.com>
On Wed, Sep 05, 2018 at 09:15:14PM +0200, Björn Töpel wrote:
> Den ons 5 sep. 2018 kl 19:14 skrev Jakub Kicinski
> <jakub.kicinski@netronome.com>:
> >
> > On Tue, 4 Sep 2018 20:11:01 +0200, Björn Töpel wrote:
> > > From: Björn Töpel <bjorn.topel@intel.com>
> > >
> > > This series addresses an AF_XDP zero-copy issue that buffers passed
> > > from userspace to the kernel was leaked when the hardware descriptor
> > > ring was torn down.
> > >
> > > The patches fixes the i40e AF_XDP zero-copy implementation.
> > >
> > > Thanks to Jakub Kicinski for pointing this out!
> > >
> > > Some background for folks that don't know the details: A zero-copy
> > > capable driver picks buffers off the fill ring and places them on the
> > > hardware Rx ring to be completed at a later point when DMA is
> > > complete. Similar on the Tx side; The driver picks buffers off the Tx
> > > ring and places them on the Tx hardware ring.
> > >
> > > In the typical flow, the Rx buffer will be placed onto an Rx ring
> > > (completed to the user), and the Tx buffer will be placed on the
> > > completion ring to notify the user that the transfer is done.
> > >
> > > However, if the driver needs to tear down the hardware rings for some
> > > reason (interface goes down, reconfiguration and such), the userspace
> > > buffers cannot be leaked. They have to be reused or completed back to
> > > userspace.
> > >
> > > The implementation does the following:
> > >
> > > * Outstanding Tx descriptors will be passed to the completion
> > > ring. The Tx code has back-pressure mechanism in place, so that
> > > enough empty space in the completion ring is guaranteed.
> > >
> > > * Outstanding Rx descriptors are temporarily stored on a stash/reuse
> > > queue. The reuse queue is based on Jakub's RFC. When/if the HW rings
> > > comes up again, entries from the stash are used to re-populate the
> > > ring.
> > >
> > > * When AF_XDP ZC is enabled, disallow changing the number of hardware
> > > descriptors via ethtool. Otherwise, the size of the stash/reuse
> > > queue can grow unbounded.
> > >
> > > Going forward, introducing a "zero-copy allocator" analogous to Jesper
> > > Brouer's page pool would be a more robust and reuseable solution.
> > >
> > > Jakub: I've made a minor checkpatch-fix to your RFC, prior adding it
> > > into this series.
> >
> > Thanks for the fix! :)
> >
> > Out of curiosity, did checking the reuse queue have a noticeable impact
> > in your test (i.e. always using the _rq() helpers)? You seem to be
> > adding an indirect call, would that not be way worse on a retpoline
> > kernel?
>
> Do you mean the indirection in __i40e_alloc_rx_buffers_zc (patch #3)?
> The indirect call is elided by the __always_inline -- without that
> retpoline took 2.5Mpps worth of Rx. :-(
>
> I'm only using the _rq helpers in the configuration/slow path, so the
> fast-path is unchanged.
Applied to bpf-next. Thanks.
^ permalink raw reply
* Re: [PATCH] ath9k: add reset for airtime station debugfs
From: Toke Høiland-Jørgensen @ 2018-09-06 10:40 UTC (permalink / raw)
To: Louie Lu
Cc: kvalo-sgV2jX0FEOL9JmXXK+q4OQ, ath9k-devel-A+ZNKFmMK5xy9aJCnZT0Uw,
davem-fT/PcQaiUtIeIZ0/mPfg9Q,
linux-wireless-u79uwXL29TY76Z2rM5mHXA,
netdev-u79uwXL29TY76Z2rM5mHXA
In-Reply-To: <CADMDV=ws-iM6+L+CLyrGK=srV56jbYgBbOVWg2AxEX0gR6t8WQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
Louie Lu <git-sgRGTR89XAc@public.gmane.org> writes:
> Toke Høiland-Jørgensen <toke-LJ9M9ZcSy1A@public.gmane.org> 於 2018年9月6日 週四 下午5:27寫道：
>
>> Louie Lu <git-sgRGTR89XAc@public.gmane.org> writes:
>>
>> > Let user can reset station airtime status by debugfs, it will
>> > reset all airtime deficit to ATH_AIRTIME_QUANTUM and reset rx/tx
>> > airtime accumulate to 0.
>>
>> No objections to the patch, but I'm curious which issues you were
>> debugging that led you to needing it? :)
>>
> I'm testing to get the packet queue time + airtime in
> ath_tx_process_buffer,
Right; I've been thinking that it would be useful to make the CoDel
enqueue time available to drivers. And minstrel, for that matter
(lowering the number of retries for packets that has queued for a long
time, for instance). Good to hear that others are looking into something
similar :)
> it would be useful if I can reset the station airtime accumulated
> value, so I can observe in each test round (e.g. 5 ping) airtime
> accumulated
>
> Also to reset the deficit to make sure it run like fresh one.
Yup, makes sense.
-Toke
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox