microsoft
/
WSL2-Linux-Kernel
зеркало из https://github.com/microsoft/WSL2-Linux-Kernel.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

mlx5-updates-2021-08-19 

This series introduces the support for two new mlx5 features:
 
 1) Sample offload for tunneled traffic
 2) devlink rate objects support
 
 1) From Chris Mi: Sample offload for tunneled traffic
 =====================================================
 
 Background and solution
 -----------------------
 
 Currently the sample offload actions send the encapsulated packet
 to software. This series de-capsulates the packet before performing
 the sampling and set the tunnel properties on the skb metadata
 fields to make the behavior consistent with OVS sFlow.
 
 If de-capsulating first, we can't use the same match like before in
 default table. So instantiate a post action instance to continue
 processing the action list. If HW can preserve reg_c, also use the
 post action instance.
 
 Post action infrastructure
 --------------------------
 
 Some tc actions are modeled in hardware using multiple tables
 causing a tc action list split. For example, CT action is modeled
 by jumping to a ct table which is controlled by nf flow table.
 sFlow jumps in hardware to a sample table, which continues to a
 "default table" where it should continue processing the action list.
 
 Multi table actions are modeled in hardware using a unique fte_id.
 The fte_id is set before jumping to a table. Split actions continue
 to a post-action table where the matched fte_id value continues the
 execution the tc action list.
 
 This series also introduces post action infrastructure. Both ct and
 sample use it.
 
 Sample for tunnel in TC SW
 --------------------------
 
 tc filter add dev vxlan1 protocol ip parent ffff: prio 3		\
 	flower src_mac 24:25:d0:e1:00:00 dst_mac 02:25:d0:13:01:02	\
 	enc_src_ip 192.168.1.14 enc_dst_ip 192.168.1.13			\
 	enc_dst_port 4789 enc_key_id 4					\
 	action sample rate 1 group 6					\
 	action tunnel_key unset						\
 	action mirred egress redirect dev enp4s0f0_1
 
 MLX5 sample HW offload
 ----------------------
 
 For the following typical flow table:
 
 +-------------------------------+
 +       original flow table     +
 +-------------------------------+
 +         original match        +
 +-------------------------------+
 + sample action + other actions +
 +-------------------------------+
 
 We translate the tc filter with sample action to the following HW model:
 
         +---------------------+
         + original flow table +
         +---------------------+
         +   original match    +
         +---------------------+
               | set fte_id (if reg_c preserve cap)
               | do decap
               v
 +------------------------------------------------+
 +                Flow Sampler Object             +
 +------------------------------------------------+
 +                    sample ratio                +
 +------------------------------------------------+
 +    sample table id    |    default table id    +
 +------------------------------------------------+
            |                            |
            v                            v
 +-----------------------------+  +-------------------+
 +        sample table         +  +   default table   +
 +-----------------------------+  +-------------------+
 + forward to management vport +             |
 +-----------------------------+             |
                                     +-------+------+
                                     |              |reg_c preserve cap
                                     |              |or decap action
                                     v              v
                        +-----------------+   +-------------+
                        + per vport table +   + post action +
                        +-----------------+   +-------------+
                        + original match  +
                        +-----------------+
                        + other actions   +
                        +-----------------+
 
 2) From Dmytro Linkin: devlink rate object support for mlx5_core driver
 =======================================================================
 
 HIGH-LEVEL OVERVIEW
 
 Devlink leaf rate objects created per vport (VF/SF, and PF on BlueField)
 in switchdev mode on devlink port registration.
 Implement devlink ops callbacks to create/destroy rate groups, set TX
 rate values of the vport/group, assign vport to the group.
 Driver accepts TX rate values as fraction of 1Mbps.
 
 Refactor existing eswitch QoS infrastructure to be accessible by legacy
 NDO rate API and new devlink rate API. NDO rate API is not
 removed/disabled in switchdev mode to not break existing users. Rate
 values configured with NDO rate API are not visible for devlink
 infrastructure, therefore APIs should not be used simultaneously.
 
 IMPLEMENTATION DETAILS
 
 Driver provide two level rate hierarchy to manage bandwidth - group
 level and vport level. Initially each vport added to internal unlimited
 group created by default. Each rate element (vport or group) receive
 bandwidth relative to its parent element (for groups the parent is a
 physical link itself) in a Round Robin manner, where element get
 bandwidth value according to its weight. Example:
 
 Created four rate groups with tx_share limits:
 
 $ devlink port function rate add \
     pci/0000:06:00.0/group_1 tx_share 30gbit
 $ devlink port function rate add \
     pci/0000:06:00.0/group_2 tx_share 20gbit
 $ devlink port function rate add \
     pci/0000:06:00.0/group_3 tx_share 20gbit
 $ devlink port function rate add \
     pci/0000:06:00.0/group_4 tx_share 10gbit
 
 Weights created in HW for each group are relative to the bigest tx_share
 value, which is 30gbit:
 
 <group_1> 1.0
 <group_2> 0.67
 <group_3> 0.67
 <group_4> 0.33
 
 Assuming link speed is 50 Gbit/sec and each group can sustain such
 amount of traffic, maximum bandwidth is 50 / (1.0 + 0.67 + 0.67 + 0.33)
  = ~18.75 Gbit/sec. Normilized bandwidth values for groups:
 
 <group_1> 18.75 * 1.0  = 18.75 Gbit/sec
 <group_2> 18.75 * 0.67 = 12.5 Gbit/sec
 <group_3> 18.75 * 0.67 = 12.5 Gbit/sec
 <group_4> 18.75 * 0.33 = 6.25 Gbit/sec
 
 If in example above group_1 doesn't produce any traffic, then maximum
 bandwidth becomes 50 / (0.67 + 0.67 + 0.33) = ~30.0 Gbit/sec. Normalized
 values:
 
 <group_2> 30.0 * 0.67 = 20.0 Gbit/sec
 <group_3> 30.0 * 0.67 = 20.0 Gbit/sec
 <group_4> 30.0 * 0.33 = 10.0 Gbit/sec
 
 Same normalization applied to each vport in the group.
 
 Normalized values are internal, therefore driver provides QoS
 tracepoints for next events:
 
 * vport rate element creation/deletion:
 * vport rate element configuration;
 * group rate element creation/deletion;
 * group rate element configuration.
 
 PATCHES OVERVIEW
 
 1 - Moving and isolation of eswitch QoS logic in separate file;
 
 2 - Implement devlink leaf rate object support for vports;
 
 3 - Implement rate groups creation/deletion;
 
 4 - Implement TX rate management for the groups;
 
 5 - Implement parent set for vports;
 
 6 - Eswitch QoS tracepoints.
 -----BEGIN PGP SIGNATURE-----
 
 iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAmEfNKEACgkQSD+KveBX
 +j4DVgf/ZTX3n7xDVrqNgM3hkUOT7QKVCq5zUlDKw1IizE6I+8xB6LO3KmUPyWIn
 +VBXE3c+aqUNZu4XlqdkVn1JskVdfTdAGXIIeBgMQskJ/VFCNqU4E9uieEmpiHnI
 DUUkEI6eBURJSu1KPD0xdMqtdGJE+/KjwmfZFnCsa4uxmRuV7B0BdxzGIA6AMFKn
 +jNS/PFbQM6bUDqP2UUwd97sThtTzDIVH86gu36yK/mdcwdLreqKeuxoHJWePWHC
 qLReBC5OQ9zXD2F1Dv2u3WU7EJT7qyCLNrBUrTwHcR9N0Di+2a6lGvjRL5tjWKKC
 KOrNkkviurmPp+VieJU+rHHYQwjoGQ==
 =XfOY
 -----END PGP SIGNATURE-----

Merge tag 'mlx5-updates-2021-08-19' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2021-08-19

This series introduces the support for two new mlx5 features:

1) Sample offload for tunneled traffic
2) devlink rate objects support

1) From Chris Mi: Sample offload for tunneled traffic
=====================================================

Background and solution
-----------------------

Currently the sample offload actions send the encapsulated packet
to software. This series de-capsulates the packet before performing
the sampling and set the tunnel properties on the skb metadata
fields to make the behavior consistent with OVS sFlow.

If de-capsulating first, we can't use the same match like before in
default table. So instantiate a post action instance to continue
processing the action list. If HW can preserve reg_c, also use the
post action instance.

Post action infrastructure
--------------------------

Some tc actions are modeled in hardware using multiple tables
causing a tc action list split. For example, CT action is modeled
by jumping to a ct table which is controlled by nf flow table.
sFlow jumps in hardware to a sample table, which continues to a
"default table" where it should continue processing the action list.

Multi table actions are modeled in hardware using a unique fte_id.
The fte_id is set before jumping to a table. Split actions continue
to a post-action table where the matched fte_id value continues the
execution the tc action list.

This series also introduces post action infrastructure. Both ct and
sample use it.

Sample for tunnel in TC SW
--------------------------

tc filter add dev vxlan1 protocol ip parent ffff: prio 3		\
	flower src_mac 24:25:d0:e1:00:00 dst_mac 02:25:d0:13:01:02	\
	enc_src_ip 192.168.1.14 enc_dst_ip 192.168.1.13			\
	enc_dst_port 4789 enc_key_id 4					\
	action sample rate 1 group 6					\
	action tunnel_key unset						\
	action mirred egress redirect dev enp4s0f0_1

MLX5 sample HW offload
----------------------

For the following typical flow table:

+-------------------------------+
+       original flow table     +
+-------------------------------+
+         original match        +
+-------------------------------+
+ sample action + other actions +
+-------------------------------+

We translate the tc filter with sample action to the following HW model:

        +---------------------+
        + original flow table +
        +---------------------+
        +   original match    +
        +---------------------+
              | set fte_id (if reg_c preserve cap)
              | do decap
              v
+------------------------------------------------+
+                Flow Sampler Object             +
+------------------------------------------------+
+                    sample ratio                +
+------------------------------------------------+
+    sample table id    |    default table id    +
+------------------------------------------------+
           |                            |
           v                            v
+-----------------------------+  +-------------------+
+        sample table         +  +   default table   +
+-----------------------------+  +-------------------+
+ forward to management vport +             |
+-----------------------------+             |
                                    +-------+------+
                                    |              |reg_c preserve cap
                                    |              |or decap action
                                    v              v
                       +-----------------+   +-------------+
                       + per vport table +   + post action +
                       +-----------------+   +-------------+
                       + original match  +
                       +-----------------+
                       + other actions   +
                       +-----------------+

2) From Dmytro Linkin: devlink rate object support for mlx5_core driver
=======================================================================

HIGH-LEVEL OVERVIEW

Devlink leaf rate objects created per vport (VF/SF, and PF on BlueField)
in switchdev mode on devlink port registration.
Implement devlink ops callbacks to create/destroy rate groups, set TX
rate values of the vport/group, assign vport to the group.
Driver accepts TX rate values as fraction of 1Mbps.

Refactor existing eswitch QoS infrastructure to be accessible by legacy
NDO rate API and new devlink rate API. NDO rate API is not
removed/disabled in switchdev mode to not break existing users. Rate
values configured with NDO rate API are not visible for devlink
infrastructure, therefore APIs should not be used simultaneously.

IMPLEMENTATION DETAILS

Driver provide two level rate hierarchy to manage bandwidth - group
level and vport level. Initially each vport added to internal unlimited
group created by default. Each rate element (vport or group) receive
bandwidth relative to its parent element (for groups the parent is a
physical link itself) in a Round Robin manner, where element get
bandwidth value according to its weight. Example:

Created four rate groups with tx_share limits:

$ devlink port function rate add \
    pci/0000:06:00.0/group_1 tx_share 30gbit
$ devlink port function rate add \
    pci/0000:06:00.0/group_2 tx_share 20gbit
$ devlink port function rate add \
    pci/0000:06:00.0/group_3 tx_share 20gbit
$ devlink port function rate add \
    pci/0000:06:00.0/group_4 tx_share 10gbit

Weights created in HW for each group are relative to the bigest tx_share
value, which is 30gbit:

<group_1> 1.0
<group_2> 0.67
<group_3> 0.67
<group_4> 0.33

Assuming link speed is 50 Gbit/sec and each group can sustain such
amount of traffic, maximum bandwidth is 50 / (1.0 + 0.67 + 0.67 + 0.33)
 = ~18.75 Gbit/sec. Normilized bandwidth values for groups:

<group_1> 18.75 * 1.0  = 18.75 Gbit/sec
<group_2> 18.75 * 0.67 = 12.5 Gbit/sec
<group_3> 18.75 * 0.67 = 12.5 Gbit/sec
<group_4> 18.75 * 0.33 = 6.25 Gbit/sec

If in example above group_1 doesn't produce any traffic, then maximum
bandwidth becomes 50 / (0.67 + 0.67 + 0.33) = ~30.0 Gbit/sec. Normalized
values:

<group_2> 30.0 * 0.67 = 20.0 Gbit/sec
<group_3> 30.0 * 0.67 = 20.0 Gbit/sec
<group_4> 30.0 * 0.33 = 10.0 Gbit/sec

Same normalization applied to each vport in the group.

Normalized values are internal, therefore driver provides QoS
tracepoints for next events:

* vport rate element creation/deletion:
* vport rate element configuration;
* group rate element creation/deletion;
* group rate element configuration.

PATCHES OVERVIEW

1 - Moving and isolation of eswitch QoS logic in separate file;

2 - Implement devlink leaf rate object support for vports;

3 - Implement rate groups creation/deletion;

4 - Implement TX rate management for the groups;

5 - Implement parent set for vports;

6 - Eswitch QoS tracepoints.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
David S. Miller 2021-08-20 12:49:04 +01:00
Родитель e61fbee7be 3202ea65f8
Коммит f96b48c621
24 изменённых файлов: 1813 добавлений и 701 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

44
Documentation/networking/device_drivers/ethernet/mellanox/mlx5.rst
Просмотреть файл

@ -656,3 +656,47 @@ Bridge offloads tracepoints:
$ cat /sys/kernel/debug/tracing/trace
...
ip-5387 [000] ...1 573713: mlx5_esw_bridge_vport_cleanup: vport_num=1
Eswitch QoS tracepoints:
- mlx5_esw_vport_qos_create: trace creation of transmit scheduler arbiter for vport::
$ echo mlx5:mlx5_esw_vport_qos_create >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-23496 [018] .... 73136.838831: mlx5_esw_vport_qos_create: (0000:82:00.0) vport=2 tsar_ix=4 bw_share=0, max_rate=0 group=000000007b576bb3
- mlx5_esw_vport_qos_config: trace configuration of transmit scheduler arbiter for vport::
$ echo mlx5:mlx5_esw_vport_qos_config >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-26548 [023] .... 75754.223823: mlx5_esw_vport_qos_config: (0000:82:00.0) vport=1 tsar_ix=3 bw_share=34, max_rate=10000 group=000000007b576bb3
- mlx5_esw_vport_qos_destroy: trace deletion of transmit scheduler arbiter for vport::
$ echo mlx5:mlx5_esw_vport_qos_destroy >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-27418 [004] .... 76546.680901: mlx5_esw_vport_qos_destroy: (0000:82:00.0) vport=1 tsar_ix=3
- mlx5_esw_group_qos_create: trace creation of transmit scheduler arbiter for rate group::
$ echo mlx5:mlx5_esw_group_qos_create >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-26578 [008] .... 75776.022112: mlx5_esw_group_qos_create: (0000:82:00.0) group=000000008dac63ea tsar_ix=5
- mlx5_esw_group_qos_config: trace configuration of transmit scheduler arbiter for rate group::
$ echo mlx5:mlx5_esw_group_qos_config >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-27303 [020] .... 76461.455356: mlx5_esw_group_qos_config: (0000:82:00.0) group=000000008dac63ea tsar_ix=5 bw_share=100 max_rate=20000
- mlx5_esw_group_qos_destroy: trace deletion of transmit scheduler arbiter for group::
$ echo mlx5:mlx5_esw_group_qos_destroy >> /sys/kernel/debug/tracing/set_event
$ cat /sys/kernel/debug/tracing/trace
...
<...>-27418 [006] .... 76547.187258: mlx5_esw_group_qos_destroy: (0000:82:00.0) group=000000007b576bb3 tsar_ix=1

13
drivers/net/ethernet/mellanox/mlx5/core/Makefile
Просмотреть файл

@ -44,19 +44,22 @@ mlx5_core-$(CONFIG_MLX5_CLS_ACT) += en_tc.o en/rep/tc.o en/rep/neigh.o \
lib/fs_chains.o en/tc_tun.o \
esw/indir_table.o en/tc_tun_encap.o \
en/tc_tun_vxlan.o en/tc_tun_gre.o en/tc_tun_geneve.o \
en/tc_tun_mplsoudp.o diag/en_tc_tracepoint.o
en/tc_tun_mplsoudp.o diag/en_tc_tracepoint.o \
en/tc/post_act.o
mlx5_core-$(CONFIG_MLX5_TC_CT) += en/tc_ct.o
mlx5_core-$(CONFIG_MLX5_TC_SAMPLE) += en/tc/sample.o
#
# Core extra
#
mlx5_core-$(CONFIG_MLX5_ESWITCH) += eswitch.o eswitch_offloads.o eswitch_offloads_termtbl.o \
ecpf.o rdma.o esw/legacy.o
ecpf.o rdma.o esw/legacy.o \
esw/devlink_port.o esw/vporttbl.o esw/qos.o
mlx5_core-$(CONFIG_MLX5_ESWITCH) += esw/acl/helper.o \
esw/acl/egress_lgcy.o esw/acl/egress_ofld.o \
esw/acl/ingress_lgcy.o esw/acl/ingress_ofld.o \
esw/devlink_port.o esw/vporttbl.o
mlx5_core-$(CONFIG_MLX5_TC_SAMPLE) += esw/sample.o
esw/acl/ingress_lgcy.o esw/acl/ingress_ofld.o
mlx5_core-$(CONFIG_MLX5_BRIDGE) += esw/bridge.o en/rep/bridge.o
mlx5_core-$(CONFIG_MLX5_MPFS) += lib/mpfs.o

8
drivers/net/ethernet/mellanox/mlx5/core/devlink.c
Просмотреть файл

@ -7,6 +7,7 @@
#include "fw_reset.h"
#include "fs_core.h"
#include "eswitch.h"
#include "esw/qos.h"
#include "sf/dev/dev.h"
#include "sf/sf.h"
@ -292,6 +293,13 @@ static const struct devlink_ops mlx5_devlink_ops = {
.eswitch_encap_mode_get = mlx5_devlink_eswitch_encap_mode_get,
.port_function_hw_addr_get = mlx5_devlink_port_function_hw_addr_get,
.port_function_hw_addr_set = mlx5_devlink_port_function_hw_addr_set,
.rate_leaf_tx_share_set = mlx5_esw_devlink_rate_leaf_tx_share_set,
.rate_leaf_tx_max_set = mlx5_esw_devlink_rate_leaf_tx_max_set,
.rate_node_tx_share_set = mlx5_esw_devlink_rate_node_tx_share_set,
.rate_node_tx_max_set = mlx5_esw_devlink_rate_node_tx_max_set,
.rate_node_new = mlx5_esw_devlink_rate_node_new,
.rate_node_del = mlx5_esw_devlink_rate_node_del,
.rate_leaf_parent_set = mlx5_esw_devlink_rate_parent_set,
#endif
#ifdef CONFIG_MLX5_SF_MANAGER
.port_new = mlx5_devlink_sf_port_new,

3
drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
Просмотреть файл

@ -7,6 +7,8 @@
#include "mod_hdr.h"
#include "lib/fs_ttc.h"
struct mlx5e_post_act;
enum {
MLX5E_TC_FT_LEVEL = 0,
MLX5E_TC_TTC_FT_LEVEL,
@ -19,6 +21,7 @@ struct mlx5e_tc_table {
struct mutex t_lock;
struct mlx5_flow_table *t;
struct mlx5_fs_chains *chains;
struct mlx5e_post_act *post_act;
struct rhashtable ht;

48
drivers/net/ethernet/mellanox/mlx5/core/en/rep/tc.c
Просмотреть файл

@ -17,7 +17,7 @@
#include "en/mapping.h"
#include "en/tc_tun.h"
#include "lib/port_tun.h"
#include "esw/sample.h"
#include "en/tc/sample.h"
struct mlx5e_rep_indr_block_priv {
struct net_device *netdev;
@ -516,7 +516,6 @@ void mlx5e_rep_tc_netdevice_event_unregister(struct mlx5e_rep_priv *rpriv)
mlx5e_rep_indr_block_unbind);
}
#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
static bool mlx5e_restore_tunnel(struct mlx5e_priv *priv, struct sk_buff *skb,
struct mlx5e_tc_update_priv *tc_priv,
u32 tunnel_id)
@ -609,12 +608,13 @@ static bool mlx5e_restore_tunnel(struct mlx5e_priv *priv, struct sk_buff *skb,
return true;
}
static bool mlx5e_restore_skb(struct sk_buff *skb, u32 chain, u32 reg_c1,
struct mlx5e_tc_update_priv *tc_priv)
static bool mlx5e_restore_skb_chain(struct sk_buff *skb, u32 chain, u32 reg_c1,
struct mlx5e_tc_update_priv *tc_priv)
{
struct mlx5e_priv *priv = netdev_priv(skb->dev);
u32 tunnel_id = (reg_c1 >> ESW_TUN_OFFSET) & TUNNEL_ID_MASK;
#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
if (chain) {
struct mlx5_rep_uplink_priv *uplink_priv;
struct mlx5e_rep_priv *uplink_rpriv;
@ -636,9 +636,25 @@ static bool mlx5e_restore_skb(struct sk_buff *skb, u32 chain, u32 reg_c1,
zone_restore_id))
return false;
}
#endif /* CONFIG_NET_TC_SKB_EXT */
return mlx5e_restore_tunnel(priv, skb, tc_priv, tunnel_id);
}
#endif /* CONFIG_NET_TC_SKB_EXT */
static void mlx5e_restore_skb_sample(struct mlx5e_priv *priv, struct sk_buff *skb,
struct mlx5_mapped_obj *mapped_obj,
struct mlx5e_tc_update_priv *tc_priv)
{
if (!mlx5e_restore_tunnel(priv, skb, tc_priv, mapped_obj->sample.tunnel_id)) {
netdev_dbg(priv->netdev,
"Failed to restore tunnel info for sampled packet\n");
return;
}
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
mlx5e_tc_sample_skb(skb, mapped_obj);
#endif /* CONFIG_MLX5_TC_SAMPLE */
mlx5_rep_tc_post_napi_receive(tc_priv);
}
bool mlx5e_rep_tc_update_skb(struct mlx5_cqe64 *cqe,
struct sk_buff *skb,
@ -647,7 +663,7 @@ bool mlx5e_rep_tc_update_skb(struct mlx5_cqe64 *cqe,
struct mlx5_mapped_obj mapped_obj;
struct mlx5_eswitch *esw;
struct mlx5e_priv *priv;
u32 reg_c0, reg_c1;
u32 reg_c0;
int err;
reg_c0 = (be32_to_cpu(cqe->sop_drop_qpn) & MLX5E_TC_FLOW_ID_MASK);
@ -659,8 +675,6 @@ bool mlx5e_rep_tc_update_skb(struct mlx5_cqe64 *cqe,
*/
skb->mark = 0;
reg_c1 = be32_to_cpu(cqe->ft_metadata);
priv = netdev_priv(skb->dev);
esw = priv->mdev->priv.eswitch;
err = mapping_find(esw->offloads.reg_c0_obj_pool, reg_c0, &mapped_obj);
@ -671,18 +685,14 @@ bool mlx5e_rep_tc_update_skb(struct mlx5_cqe64 *cqe,
return false;
}
#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
if (mapped_obj.type == MLX5_MAPPED_OBJ_CHAIN)
return mlx5e_restore_skb(skb, mapped_obj.chain, reg_c1, tc_priv);
#endif /* CONFIG_NET_TC_SKB_EXT */
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
if (mapped_obj.type == MLX5_MAPPED_OBJ_SAMPLE) {
mlx5_esw_sample_skb(skb, &mapped_obj);
if (mapped_obj.type == MLX5_MAPPED_OBJ_CHAIN) {
u32 reg_c1 = be32_to_cpu(cqe->ft_metadata);
return mlx5e_restore_skb_chain(skb, mapped_obj.chain, reg_c1, tc_priv);
} else if (mapped_obj.type == MLX5_MAPPED_OBJ_SAMPLE) {
mlx5e_restore_skb_sample(priv, skb, &mapped_obj, tc_priv);
return false;
}
#endif /* CONFIG_MLX5_TC_SAMPLE */
if (mapped_obj.type != MLX5_MAPPED_OBJ_SAMPLE &&
mapped_obj.type != MLX5_MAPPED_OBJ_CHAIN) {
} else {
netdev_dbg(priv->netdev, "Invalid mapped object type: %d\n", mapped_obj.type);
return false;
}

164
drivers/net/ethernet/mellanox/mlx5/core/en/tc/post_act.c Normal file
Просмотреть файл

@ -0,0 +1,164 @@
// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
// Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#include "en_tc.h"
#include "post_act.h"
#include "mlx5_core.h"
struct mlx5e_post_act {
enum mlx5_flow_namespace_type ns_type;
struct mlx5_fs_chains *chains;
struct mlx5_flow_table *ft;
struct mlx5e_priv *priv;
struct xarray ids;
};
struct mlx5e_post_act_handle {
enum mlx5_flow_namespace_type ns_type;
struct mlx5_flow_attr *attr;
struct mlx5_flow_handle *rule;
u32 id;
};
#define MLX5_POST_ACTION_BITS (mlx5e_tc_attr_to_reg_mappings[FTEID_TO_REG].mlen)
#define MLX5_POST_ACTION_MAX GENMASK(MLX5_POST_ACTION_BITS - 1, 0)
#define MLX5_POST_ACTION_MASK MLX5_POST_ACTION_MAX
struct mlx5e_post_act *
mlx5e_tc_post_act_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
enum mlx5_flow_namespace_type ns_type)
{
struct mlx5e_post_act *post_act;
int err;
if (ns_type == MLX5_FLOW_NAMESPACE_FDB &&
!MLX5_CAP_ESW_FLOWTABLE_FDB(priv->mdev, ignore_flow_level)) {
mlx5_core_warn(priv->mdev, "firmware level support is missing\n");
err = -EOPNOTSUPP;
goto err_check;
} else if (!MLX5_CAP_FLOWTABLE_NIC_RX(priv->mdev, ignore_flow_level)) {
mlx5_core_warn(priv->mdev, "firmware level support is missing\n");
err = -EOPNOTSUPP;
goto err_check;
}
post_act = kzalloc(sizeof(*post_act), GFP_KERNEL);
if (!post_act) {
err = -ENOMEM;
goto err_check;
}
post_act->ft = mlx5_chains_create_global_table(chains);
if (IS_ERR(post_act->ft)) {
err = PTR_ERR(post_act->ft);
mlx5_core_warn(priv->mdev, "failed to create post action table, err: %d\n", err);
goto err_ft;
}
post_act->chains = chains;
post_act->ns_type = ns_type;
post_act->priv = priv;
xa_init_flags(&post_act->ids, XA_FLAGS_ALLOC1);
return post_act;
err_ft:
kfree(post_act);
err_check:
return ERR_PTR(err);
}
void
mlx5e_tc_post_act_destroy(struct mlx5e_post_act *post_act)
{
if (IS_ERR_OR_NULL(post_act))
return;
xa_destroy(&post_act->ids);
mlx5_chains_destroy_global_table(post_act->chains, post_act->ft);
kfree(post_act);
}
struct mlx5e_post_act_handle *
mlx5e_tc_post_act_add(struct mlx5e_post_act *post_act, struct mlx5_flow_attr *attr)
{
u32 attr_sz = ns_to_attr_sz(post_act->ns_type);
struct mlx5e_post_act_handle *handle = NULL;
struct mlx5_flow_attr *post_attr = NULL;
struct mlx5_flow_spec *spec = NULL;
int err;
handle = kzalloc(sizeof(*handle), GFP_KERNEL);
spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
post_attr = mlx5_alloc_flow_attr(post_act->ns_type);
if (!handle || !spec || !post_attr) {
kfree(post_attr);
kvfree(spec);
kfree(handle);
return ERR_PTR(-ENOMEM);
}
memcpy(post_attr, attr, attr_sz);
post_attr->chain = 0;
post_attr->prio = 0;
post_attr->ft = post_act->ft;
post_attr->inner_match_level = MLX5_MATCH_NONE;
post_attr->outer_match_level = MLX5_MATCH_NONE;
post_attr->action &= ~(MLX5_FLOW_CONTEXT_ACTION_DECAP);
handle->ns_type = post_act->ns_type;
/* Splits were handled before post action */
if (handle->ns_type == MLX5_FLOW_NAMESPACE_FDB)
post_attr->esw_attr->split_count = 0;
err = xa_alloc(&post_act->ids, &handle->id, post_attr,
XA_LIMIT(1, MLX5_POST_ACTION_MAX), GFP_KERNEL);
if (err)
goto err_xarray;
/* Post action rule matches on fte_id and executes original rule's
* tc rule action
*/
mlx5e_tc_match_to_reg_match(spec, FTEID_TO_REG,
handle->id, MLX5_POST_ACTION_MASK);
handle->rule = mlx5_tc_rule_insert(post_act->priv, spec, post_attr);
if (IS_ERR(handle->rule)) {
err = PTR_ERR(handle->rule);
netdev_warn(post_act->priv->netdev, "Failed to add post action rule");
goto err_rule;
}
handle->attr = post_attr;
kvfree(spec);
return handle;
err_rule:
xa_erase(&post_act->ids, handle->id);
err_xarray:
kfree(post_attr);
kvfree(spec);
kfree(handle);
return ERR_PTR(err);
}
void
mlx5e_tc_post_act_del(struct mlx5e_post_act *post_act, struct mlx5e_post_act_handle *handle)
{
mlx5_tc_rule_delete(post_act->priv, handle->rule, handle->attr);
xa_erase(&post_act->ids, handle->id);
kfree(handle->attr);
kfree(handle);
}
struct mlx5_flow_table *
mlx5e_tc_post_act_get_ft(struct mlx5e_post_act *post_act)
{
return post_act->ft;
}
/* Allocate a header modify action to write the post action handle fte id to a register. */
int
mlx5e_tc_post_act_set_handle(struct mlx5_core_dev *dev,
struct mlx5e_post_act_handle *handle,
struct mlx5e_tc_mod_hdr_acts *acts)
{
return mlx5e_tc_match_to_reg_set(dev, acts, handle->ns_type, FTEID_TO_REG, handle->id);
}

35
drivers/net/ethernet/mellanox/mlx5/core/en/tc/post_act.h Normal file
Просмотреть файл

@ -0,0 +1,35 @@
/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/* Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved. */
#ifndef __MLX5_POST_ACTION_H__
#define __MLX5_POST_ACTION_H__
#include "en.h"
#include "lib/fs_chains.h"
struct mlx5_flow_attr;
struct mlx5e_priv;
struct mlx5e_tc_mod_hdr_acts;
struct mlx5e_post_act *
mlx5e_tc_post_act_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
enum mlx5_flow_namespace_type ns_type);
void
mlx5e_tc_post_act_destroy(struct mlx5e_post_act *post_act);
struct mlx5e_post_act_handle *
mlx5e_tc_post_act_add(struct mlx5e_post_act *post_act, struct mlx5_flow_attr *attr);
void
mlx5e_tc_post_act_del(struct mlx5e_post_act *post_act, struct mlx5e_post_act_handle *handle);
struct mlx5_flow_table *
mlx5e_tc_post_act_get_ft(struct mlx5e_post_act *post_act);
int
mlx5e_tc_post_act_set_handle(struct mlx5_core_dev *dev,
struct mlx5e_post_act_handle *handle,
struct mlx5e_tc_mod_hdr_acts *acts);
#endif /* __MLX5_POST_ACTION_H__ */

470
drivers/net/ethernet/mellanox/mlx5/core/esw/sample.c → drivers/net/ethernet/mellanox/mlx5/core/en/tc/sample.c
Просмотреть файл

@ -4,7 +4,8 @@
#include <linux/skbuff.h>
#include <net/psample.h>
#include "en/mapping.h"
#include "esw/sample.h"
#include "en/tc/post_act.h"
#include "sample.h"
#include "eswitch.h"
#include "en_tc.h"
#include "fs_core.h"
@ -17,17 +18,18 @@ static const struct esw_vport_tbl_namespace mlx5_esw_vport_tbl_sample_ns = {
.flags = MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT | MLX5_FLOW_TABLE_TUNNEL_EN_DECAP,
};
struct mlx5_esw_psample {
struct mlx5e_priv *priv;
struct mlx5e_tc_psample {
struct mlx5_eswitch *esw;
struct mlx5_flow_table *termtbl;
struct mlx5_flow_handle *termtbl_rule;
DECLARE_HASHTABLE(hashtbl, 8);
struct mutex ht_lock; /* protect hashtbl */
DECLARE_HASHTABLE(restore_hashtbl, 8);
struct mutex restore_lock; /* protect restore_hashtbl */
struct mlx5e_post_act *post_act;
};
struct mlx5_sampler {
struct mlx5e_sampler {
struct hlist_node hlist;
u32 sampler_id;
u32 sample_ratio;
@ -36,29 +38,32 @@ struct mlx5_sampler {
int count;
};
struct mlx5_sample_flow {
struct mlx5_sampler *sampler;
struct mlx5_sample_restore *restore;
struct mlx5e_sample_flow {
struct mlx5e_sampler *sampler;
struct mlx5e_sample_restore *restore;
struct mlx5_flow_attr *pre_attr;
struct mlx5_flow_handle *pre_rule;
struct mlx5_flow_handle *rule;
struct mlx5_flow_attr *post_attr;
struct mlx5_flow_handle *post_rule;
struct mlx5e_post_act_handle *post_act_handle;
};
struct mlx5_sample_restore {
struct mlx5e_sample_restore {
struct hlist_node hlist;
struct mlx5_modify_hdr *modify_hdr;
struct mlx5_flow_handle *rule;
struct mlx5e_post_act_handle *post_act_handle;
u32 obj_id;
int count;
};
static int
sampler_termtbl_create(struct mlx5_esw_psample *esw_psample)
sampler_termtbl_create(struct mlx5e_tc_psample *tc_psample)
{
struct mlx5_core_dev *dev = esw_psample->priv->mdev;
struct mlx5_eswitch *esw = dev->priv.eswitch;
struct mlx5_eswitch *esw = tc_psample->esw;
struct mlx5_flow_table_attr ft_attr = {};
struct mlx5_flow_destination dest = {};
struct mlx5_core_dev *dev = esw->dev;
struct mlx5_flow_namespace *root_ns;
struct mlx5_flow_act act = {};
int err;
@ -79,20 +84,20 @@ sampler_termtbl_create(struct mlx5_esw_psample *esw_psample)
ft_attr.prio = FDB_SLOW_PATH;
ft_attr.max_fte = 1;
ft_attr.level = 1;
esw_psample->termtbl = mlx5_create_auto_grouped_flow_table(root_ns, &ft_attr);
if (IS_ERR(esw_psample->termtbl)) {
err = PTR_ERR(esw_psample->termtbl);
tc_psample->termtbl = mlx5_create_auto_grouped_flow_table(root_ns, &ft_attr);
if (IS_ERR(tc_psample->termtbl)) {
err = PTR_ERR(tc_psample->termtbl);
mlx5_core_warn(dev, "failed to create termtbl, err: %d\n", err);
return err;
}
act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
dest.vport.num = esw->manager_vport;
esw_psample->termtbl_rule = mlx5_add_flow_rules(esw_psample->termtbl, NULL, &act, &dest, 1);
if (IS_ERR(esw_psample->termtbl_rule)) {
err = PTR_ERR(esw_psample->termtbl_rule);
tc_psample->termtbl_rule = mlx5_add_flow_rules(tc_psample->termtbl, NULL, &act, &dest, 1);
if (IS_ERR(tc_psample->termtbl_rule)) {
err = PTR_ERR(tc_psample->termtbl_rule);
mlx5_core_warn(dev, "failed to create termtbl rule, err: %d\n", err);
mlx5_destroy_flow_table(esw_psample->termtbl);
mlx5_destroy_flow_table(tc_psample->termtbl);
return err;
}
@ -100,14 +105,14 @@ sampler_termtbl_create(struct mlx5_esw_psample *esw_psample)
}
static void
sampler_termtbl_destroy(struct mlx5_esw_psample *esw_psample)
sampler_termtbl_destroy(struct mlx5e_tc_psample *tc_psample)
{
mlx5_del_flow_rules(esw_psample->termtbl_rule);
mlx5_destroy_flow_table(esw_psample->termtbl);
mlx5_del_flow_rules(tc_psample->termtbl_rule);
mlx5_destroy_flow_table(tc_psample->termtbl);
}
static int
sampler_obj_create(struct mlx5_core_dev *mdev, struct mlx5_sampler *sampler)
sampler_obj_create(struct mlx5_core_dev *mdev, struct mlx5e_sampler *sampler)
{
u32 in[MLX5_ST_SZ_DW(create_sampler_obj_in)] = {};
u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
@ -163,16 +168,16 @@ sampler_cmp(u32 sample_ratio1, u32 default_table_id1, u32 sample_ratio2, u32 def
return sample_ratio1 != sample_ratio2 || default_table_id1 != default_table_id2;
}
static struct mlx5_sampler *
sampler_get(struct mlx5_esw_psample *esw_psample, u32 sample_ratio, u32 default_table_id)
static struct mlx5e_sampler *
sampler_get(struct mlx5e_tc_psample *tc_psample, u32 sample_ratio, u32 default_table_id)
{
struct mlx5_sampler *sampler;
struct mlx5e_sampler *sampler;
u32 hash_key;
int err;
mutex_lock(&esw_psample->ht_lock);
mutex_lock(&tc_psample->ht_lock);
hash_key = sampler_hash(sample_ratio, default_table_id);
hash_for_each_possible(esw_psample->hashtbl, sampler, hlist, hash_key)
hash_for_each_possible(tc_psample->hashtbl, sampler, hlist, hash_key)
if (!sampler_cmp(sampler->sample_ratio, sampler->default_table_id,
sample_ratio, default_table_id))
goto add_ref;
@ -183,42 +188,49 @@ sampler_get(struct mlx5_esw_psample *esw_psample, u32 sample_ratio, u32 default_
goto err_alloc;
}
sampler->sample_table_id = esw_psample->termtbl->id;
sampler->sample_table_id = tc_psample->termtbl->id;
sampler->default_table_id = default_table_id;
sampler->sample_ratio = sample_ratio;
err = sampler_obj_create(esw_psample->priv->mdev, sampler);
err = sampler_obj_create(tc_psample->esw->dev, sampler);
if (err)
goto err_create;
hash_add(esw_psample->hashtbl, &sampler->hlist, hash_key);
hash_add(tc_psample->hashtbl, &sampler->hlist, hash_key);
add_ref:
sampler->count++;
mutex_unlock(&esw_psample->ht_lock);
mutex_unlock(&tc_psample->ht_lock);
return sampler;
err_create:
kfree(sampler);
err_alloc:
mutex_unlock(&esw_psample->ht_lock);
mutex_unlock(&tc_psample->ht_lock);
return ERR_PTR(err);
}
static void
sampler_put(struct mlx5_esw_psample *esw_psample, struct mlx5_sampler *sampler)
sampler_put(struct mlx5e_tc_psample *tc_psample, struct mlx5e_sampler *sampler)
{
mutex_lock(&esw_psample->ht_lock);
mutex_lock(&tc_psample->ht_lock);
if (--sampler->count == 0) {
hash_del(&sampler->hlist);
sampler_obj_destroy(esw_psample->priv->mdev, sampler->sampler_id);
sampler_obj_destroy(tc_psample->esw->dev, sampler->sampler_id);
kfree(sampler);
}
mutex_unlock(&esw_psample->ht_lock);
mutex_unlock(&tc_psample->ht_lock);
}
/* obj_id is used to restore the sample parameters.
* Set fte_id in original flow table, then match it in the default table.
* Only set it for NICs can preserve reg_c or decap action. For other cases,
* use the same match in the default table.
* Use one header rewrite for both obj_id and fte_id.
*/
static struct mlx5_modify_hdr *
sample_metadata_rule_get(struct mlx5_core_dev *mdev, u32 obj_id)
sample_modify_hdr_get(struct mlx5_core_dev *mdev, u32 obj_id,
struct mlx5e_post_act_handle *handle)
{
struct mlx5e_tc_mod_hdr_acts mod_acts = {};
struct mlx5_modify_hdr *modify_hdr;
@ -229,6 +241,12 @@ sample_metadata_rule_get(struct mlx5_core_dev *mdev, u32 obj_id)
if (err)
goto err_set_regc0;
if (handle) {
err = mlx5e_tc_post_act_set_handle(mdev, handle, &mod_acts);
if (err)
goto err_post_act;
}
modify_hdr = mlx5_modify_header_alloc(mdev, MLX5_FLOW_NAMESPACE_FDB,
mod_acts.num_actions,
mod_acts.actions);
@ -241,23 +259,40 @@ sample_metadata_rule_get(struct mlx5_core_dev *mdev, u32 obj_id)
return modify_hdr;
err_modify_hdr:
err_post_act:
dealloc_mod_hdr_actions(&mod_acts);
err_set_regc0:
return ERR_PTR(err);
}
static struct mlx5_sample_restore *
sample_restore_get(struct mlx5_esw_psample *esw_psample, u32 obj_id)
static u32
restore_hash(u32 obj_id, struct mlx5e_post_act_handle *post_act_handle)
{
struct mlx5_core_dev *mdev = esw_psample->priv->mdev;
struct mlx5_eswitch *esw = mdev->priv.eswitch;
struct mlx5_sample_restore *restore;
return jhash_2words(obj_id, hash32_ptr(post_act_handle), 0);
}
static bool
restore_equal(struct mlx5e_sample_restore *restore, u32 obj_id,
struct mlx5e_post_act_handle *post_act_handle)
{
return restore->obj_id == obj_id && restore->post_act_handle == post_act_handle;
}
static struct mlx5e_sample_restore *
sample_restore_get(struct mlx5e_tc_psample *tc_psample, u32 obj_id,
struct mlx5e_post_act_handle *post_act_handle)
{
struct mlx5_eswitch *esw = tc_psample->esw;
struct mlx5_core_dev *mdev = esw->dev;
struct mlx5e_sample_restore *restore;
struct mlx5_modify_hdr *modify_hdr;
u32 hash_key;
int err;
mutex_lock(&esw_psample->restore_lock);
hash_for_each_possible(esw_psample->restore_hashtbl, restore, hlist, obj_id)
if (restore->obj_id == obj_id)
mutex_lock(&tc_psample->restore_lock);
hash_key = restore_hash(obj_id, post_act_handle);
hash_for_each_possible(tc_psample->restore_hashtbl, restore, hlist, hash_key)
if (restore_equal(restore, obj_id, post_act_handle))
goto add_ref;
restore = kzalloc(sizeof(*restore), GFP_KERNEL);
@ -266,8 +301,9 @@ sample_restore_get(struct mlx5_esw_psample *esw_psample, u32 obj_id)
goto err_alloc;
}
restore->obj_id = obj_id;
restore->post_act_handle = post_act_handle;
modify_hdr = sample_metadata_rule_get(mdev, obj_id);
modify_hdr = sample_modify_hdr_get(mdev, obj_id, post_act_handle);
if (IS_ERR(modify_hdr)) {
err = PTR_ERR(modify_hdr);
goto err_modify_hdr;
@ -280,10 +316,10 @@ sample_restore_get(struct mlx5_esw_psample *esw_psample, u32 obj_id)
goto err_restore;
}
hash_add(esw_psample->restore_hashtbl, &restore->hlist, obj_id);
hash_add(tc_psample->restore_hashtbl, &restore->hlist, hash_key);
add_ref:
restore->count++;
mutex_unlock(&esw_psample->restore_lock);
mutex_unlock(&tc_psample->restore_lock);
return restore;
err_restore:
@ -291,26 +327,26 @@ err_restore:
err_modify_hdr:
kfree(restore);
err_alloc:
mutex_unlock(&esw_psample->restore_lock);
mutex_unlock(&tc_psample->restore_lock);
return ERR_PTR(err);
}
static void
sample_restore_put(struct mlx5_esw_psample *esw_psample, struct mlx5_sample_restore *restore)
sample_restore_put(struct mlx5e_tc_psample *tc_psample, struct mlx5e_sample_restore *restore)
{
mutex_lock(&esw_psample->restore_lock);
mutex_lock(&tc_psample->restore_lock);
if (--restore->count == 0)
hash_del(&restore->hlist);
mutex_unlock(&esw_psample->restore_lock);
mutex_unlock(&tc_psample->restore_lock);
if (!restore->count) {
mlx5_del_flow_rules(restore->rule);
mlx5_modify_header_dealloc(esw_psample->priv->mdev, restore->modify_hdr);
mlx5_modify_header_dealloc(tc_psample->esw->dev, restore->modify_hdr);
kfree(restore);
}
}
void mlx5_esw_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj)
void mlx5e_tc_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj)
{
u32 trunc_size = mapped_obj->sample.trunc_size;
struct psample_group psample_group = {};
@ -325,6 +361,87 @@ void mlx5_esw_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj
psample_sample_packet(&psample_group, skb, mapped_obj->sample.rate, &md);
}
static int
add_post_rule(struct mlx5_eswitch *esw, struct mlx5e_sample_flow *sample_flow,
struct mlx5_flow_spec *spec, struct mlx5_flow_attr *attr,
u32 *default_tbl_id)
{
struct mlx5_esw_flow_attr *esw_attr = attr->esw_attr;
u32 attr_sz = ns_to_attr_sz(MLX5_FLOW_NAMESPACE_FDB);
struct mlx5_vport_tbl_attr per_vport_tbl_attr;
struct mlx5_flow_table *default_tbl;
struct mlx5_flow_attr *post_attr;
int err;
/* Allocate default table per vport, chain and prio. Otherwise, there is
* only one default table for the same sampler object. Rules with different
* prio and chain may overlap. For CT sample action, per vport default
* table is needed to resotre the metadata.
*/
per_vport_tbl_attr.chain = attr->chain;
per_vport_tbl_attr.prio = attr->prio;
per_vport_tbl_attr.vport = esw_attr->in_rep->vport;
per_vport_tbl_attr.vport_ns = &mlx5_esw_vport_tbl_sample_ns;
default_tbl = mlx5_esw_vporttbl_get(esw, &per_vport_tbl_attr);
if (IS_ERR(default_tbl)) {
err = PTR_ERR(default_tbl);
goto err_default_tbl;
}
*default_tbl_id = default_tbl->id;
post_attr = mlx5_alloc_flow_attr(MLX5_FLOW_NAMESPACE_FDB);
if (!post_attr) {
err = -ENOMEM;
goto err_attr;
}
sample_flow->post_attr = post_attr;
memcpy(post_attr, attr, attr_sz);
/* Perform the original matches on the default table.
* Offload all actions except the sample action.
*/
post_attr->chain = 0;
post_attr->prio = 0;
post_attr->ft = default_tbl;
post_attr->flags = MLX5_ESW_ATTR_FLAG_NO_IN_PORT;
/* When offloading sample and encap action, if there is no valid
* neigh data struct, a slow path rule is offloaded first. Source
* port metadata match is set at that time. A per vport table is
* already allocated. No need to match it again. So clear the source
* port metadata match.
*/
mlx5_eswitch_clear_rule_source_port(esw, spec);
sample_flow->post_rule = mlx5_eswitch_add_offloaded_rule(esw, spec, post_attr);
if (IS_ERR(sample_flow->post_rule)) {
err = PTR_ERR(sample_flow->post_rule);
goto err_rule;
}
return 0;
err_rule:
kfree(post_attr);
err_attr:
mlx5_esw_vporttbl_put(esw, &per_vport_tbl_attr);
err_default_tbl:
return err;
}
static void
del_post_rule(struct mlx5_eswitch *esw, struct mlx5e_sample_flow *sample_flow,
struct mlx5_flow_attr *attr)
{
struct mlx5_esw_flow_attr *esw_attr = attr->esw_attr;
struct mlx5_vport_tbl_attr tbl_attr;
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->post_rule, sample_flow->post_attr);
kfree(sample_flow->post_attr);
tbl_attr.chain = attr->chain;
tbl_attr.prio = attr->prio;
tbl_attr.vport = esw_attr->in_rep->vport;
tbl_attr.vport_ns = &mlx5_esw_vport_tbl_sample_ns;
mlx5_esw_vporttbl_put(esw, &tbl_attr);
}
/* For the following typical flow table:
*
* +-------------------------------+
@ -342,8 +459,9 @@ void mlx5_esw_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj
* +---------------------+
* + original match +
* +---------------------+
* |
* v
* | set fte_id (if reg_c preserve cap)
* | do decap (if required)
* v
* +------------------------------------------------+
* + Flow Sampler Object +
* +------------------------------------------------+
@ -353,80 +471,82 @@ void mlx5_esw_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj
* +------------------------------------------------+
* | |
* v v
* +-----------------------------+ +----------------------------------------+
* + sample table + + default table per <vport, chain, prio> +
* +-----------------------------+ +----------------------------------------+
* + forward to management vport + + original match +
* +-----------------------------+ +----------------------------------------+
* + other actions +
* +----------------------------------------+
* +-----------------------------+ +-------------------+
* + sample table + + default table +
* +-----------------------------+ +-------------------+
* + forward to management vport + |
* +-----------------------------+ |
* +-------+------+
* | |reg_c preserve cap
* | |or decap action
* v v
* +-----------------+ +-------------+
* + per vport table + + post action +
* +-----------------+ +-------------+
* + original match +
* +-----------------+
* + other actions +
* +-----------------+
*/
struct mlx5_flow_handle *
mlx5_esw_sample_offload(struct mlx5_esw_psample *esw_psample,
mlx5e_tc_sample_offload(struct mlx5e_tc_psample *tc_psample,
struct mlx5_flow_spec *spec,
struct mlx5_flow_attr *attr)
struct mlx5_flow_attr *attr,
u32 tunnel_id)
{
struct mlx5e_post_act_handle *post_act_handle = NULL;
struct mlx5_esw_flow_attr *esw_attr = attr->esw_attr;
struct mlx5_vport_tbl_attr per_vport_tbl_attr;
struct mlx5_esw_flow_attr *pre_esw_attr;
struct mlx5_mapped_obj restore_obj = {};
struct mlx5_sample_flow *sample_flow;
struct mlx5_sample_attr *sample_attr;
struct mlx5_flow_table *default_tbl;
struct mlx5e_sample_flow *sample_flow;
struct mlx5e_sample_attr *sample_attr;
struct mlx5_flow_attr *pre_attr;
struct mlx5_eswitch *esw;
u32 default_tbl_id;
u32 obj_id;
int err;
if (IS_ERR_OR_NULL(esw_psample))
if (IS_ERR_OR_NULL(tc_psample))
return ERR_PTR(-EOPNOTSUPP);
/* If slow path flag is set, eg. when the neigh is invalid for encap,
* don't offload sample action.
*/
esw = esw_psample->priv->mdev->priv.eswitch;
esw = tc_psample->esw;
if (attr->flags & MLX5_ESW_ATTR_FLAG_SLOW_PATH)
return mlx5_eswitch_add_offloaded_rule(esw, spec, attr);
sample_flow = kzalloc(sizeof(*sample_flow), GFP_KERNEL);
if (!sample_flow)
return ERR_PTR(-ENOMEM);
esw_attr->sample->sample_flow = sample_flow;
sample_attr = attr->sample_attr;
sample_attr->sample_flow = sample_flow;
/* Allocate default table per vport, chain and prio. Otherwise, there is
* only one default table for the same sampler object. Rules with different
* prio and chain may overlap. For CT sample action, per vport default
* table is needed to resotre the metadata.
/* For NICs with reg_c_preserve support or decap action, use
* post action instead of the per vport, chain and prio table.
* Only match the fte id instead of the same match in the
* original flow table.
*/
per_vport_tbl_attr.chain = attr->chain;
per_vport_tbl_attr.prio = attr->prio;
per_vport_tbl_attr.vport = esw_attr->in_rep->vport;
per_vport_tbl_attr.vport_ns = &mlx5_esw_vport_tbl_sample_ns;
default_tbl = mlx5_esw_vporttbl_get(esw, &per_vport_tbl_attr);
if (IS_ERR(default_tbl)) {
err = PTR_ERR(default_tbl);
goto err_default_tbl;
}
if (MLX5_CAP_GEN(esw->dev, reg_c_preserve) ||
attr->action & MLX5_FLOW_CONTEXT_ACTION_DECAP) {
struct mlx5_flow_table *ft;
/* Perform the original matches on the default table.
* Offload all actions except the sample action.
*/
esw_attr->sample->sample_default_tbl = default_tbl;
/* When offloading sample and encap action, if there is no valid
* neigh data struct, a slow path rule is offloaded first. Source
* port metadata match is set at that time. A per vport table is
* already allocated. No need to match it again. So clear the source
* port metadata match.
*/
mlx5_eswitch_clear_rule_source_port(esw, spec);
sample_flow->rule = mlx5_eswitch_add_offloaded_rule(esw, spec, attr);
if (IS_ERR(sample_flow->rule)) {
err = PTR_ERR(sample_flow->rule);
goto err_offload_rule;
ft = mlx5e_tc_post_act_get_ft(tc_psample->post_act);
default_tbl_id = ft->id;
post_act_handle = mlx5e_tc_post_act_add(tc_psample->post_act, attr);
if (IS_ERR(post_act_handle)) {
err = PTR_ERR(post_act_handle);
goto err_post_act;
}
sample_flow->post_act_handle = post_act_handle;
} else {
err = add_post_rule(esw, sample_flow, spec, attr, &default_tbl_id);
if (err)
goto err_post_rule;
}
/* Create sampler object. */
sample_flow->sampler = sampler_get(esw_psample, esw_attr->sample->rate, default_tbl->id);
sample_flow->sampler = sampler_get(tc_psample, sample_attr->rate, default_tbl_id);
if (IS_ERR(sample_flow->sampler)) {
err = PTR_ERR(sample_flow->sampler);
goto err_sampler;
@ -434,16 +554,17 @@ mlx5_esw_sample_offload(struct mlx5_esw_psample *esw_psample,
/* Create an id mapping reg_c0 value to sample object. */
restore_obj.type = MLX5_MAPPED_OBJ_SAMPLE;
restore_obj.sample.group_id = esw_attr->sample->group_num;
restore_obj.sample.rate = esw_attr->sample->rate;
restore_obj.sample.trunc_size = esw_attr->sample->trunc_size;
restore_obj.sample.group_id = sample_attr->group_num;
restore_obj.sample.rate = sample_attr->rate;
restore_obj.sample.trunc_size = sample_attr->trunc_size;
restore_obj.sample.tunnel_id = tunnel_id;
err = mapping_add(esw->offloads.reg_c0_obj_pool, &restore_obj, &obj_id);
if (err)
goto err_obj_id;
esw_attr->sample->restore_obj_id = obj_id;
sample_attr->restore_obj_id = obj_id;
/* Create sample restore context. */
sample_flow->restore = sample_restore_get(esw_psample, obj_id);
sample_flow->restore = sample_restore_get(tc_psample, obj_id, post_act_handle);
if (IS_ERR(sample_flow->restore)) {
err = PTR_ERR(sample_flow->restore);
goto err_sample_restore;
@ -455,21 +576,23 @@ mlx5_esw_sample_offload(struct mlx5_esw_psample *esw_psample,
pre_attr = mlx5_alloc_flow_attr(MLX5_FLOW_NAMESPACE_FDB);
if (!pre_attr) {
err = -ENOMEM;
goto err_alloc_flow_attr;
goto err_alloc_pre_flow_attr;
}
sample_attr = kzalloc(sizeof(*sample_attr), GFP_KERNEL);
if (!sample_attr) {
err = -ENOMEM;
goto err_alloc_sample_attr;
}
pre_esw_attr = pre_attr->esw_attr;
pre_attr->action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
/* For decap action, do decap in the original flow table instead of the
* default flow table.
*/
if (tunnel_id)
pre_attr->action |= MLX5_FLOW_CONTEXT_ACTION_DECAP;
pre_attr->modify_hdr = sample_flow->restore->modify_hdr;
pre_attr->flags = MLX5_ESW_ATTR_FLAG_SAMPLE;
pre_attr->inner_match_level = attr->inner_match_level;
pre_attr->outer_match_level = attr->outer_match_level;
pre_attr->chain = attr->chain;
pre_attr->prio = attr->prio;
pre_esw_attr->sample = sample_attr;
pre_esw_attr->sample->sampler_id = sample_flow->sampler->sampler_id;
pre_attr->sample_attr = attr->sample_attr;
sample_attr->sampler_id = sample_flow->sampler->sampler_id;
pre_esw_attr = pre_attr->esw_attr;
pre_esw_attr->in_mdev = esw_attr->in_mdev;
pre_esw_attr->in_rep = esw_attr->in_rep;
sample_flow->pre_rule = mlx5_eswitch_add_offloaded_rule(esw, spec, pre_attr);
@ -479,108 +602,113 @@ mlx5_esw_sample_offload(struct mlx5_esw_psample *esw_psample,
}
sample_flow->pre_attr = pre_attr;
return sample_flow->rule;
return sample_flow->post_rule;
err_pre_offload_rule:
kfree(sample_attr);
err_alloc_sample_attr:
kfree(pre_attr);
err_alloc_flow_attr:
sample_restore_put(esw_psample, sample_flow->restore);
err_alloc_pre_flow_attr:
sample_restore_put(tc_psample, sample_flow->restore);
err_sample_restore:
mapping_remove(esw->offloads.reg_c0_obj_pool, obj_id);
err_obj_id:
sampler_put(esw_psample, sample_flow->sampler);
sampler_put(tc_psample, sample_flow->sampler);
err_sampler:
/* For sample offload, rule is added in default_tbl. No need to call
* mlx5_esw_chains_put_table()
*/
attr->prio = 0;
attr->chain = 0;
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->rule, attr);
err_offload_rule:
mlx5_esw_vporttbl_put(esw, &per_vport_tbl_attr);
err_default_tbl:
if (!post_act_handle)
del_post_rule(esw, sample_flow, attr);
err_post_rule:
if (post_act_handle)
mlx5e_tc_post_act_del(tc_psample->post_act, post_act_handle);
err_post_act:
kfree(sample_flow);
return ERR_PTR(err);
}
void
mlx5_esw_sample_unoffload(struct mlx5_esw_psample *esw_psample,
mlx5e_tc_sample_unoffload(struct mlx5e_tc_psample *tc_psample,
struct mlx5_flow_handle *rule,
struct mlx5_flow_attr *attr)
{
struct mlx5_esw_flow_attr *esw_attr = attr->esw_attr;
struct mlx5_sample_flow *sample_flow;
struct mlx5e_sample_flow *sample_flow;
struct mlx5_vport_tbl_attr tbl_attr;
struct mlx5_flow_attr *pre_attr;
struct mlx5_eswitch *esw;
if (IS_ERR_OR_NULL(esw_psample))
if (IS_ERR_OR_NULL(tc_psample))
return;
/* If slow path flag is set, sample action is not offloaded.
* No need to delete sample rule.
*/
esw = esw_psample->priv->mdev->priv.eswitch;
esw = tc_psample->esw;
if (attr->flags & MLX5_ESW_ATTR_FLAG_SLOW_PATH) {
mlx5_eswitch_del_offloaded_rule(esw, rule, attr);
return;
}
sample_flow = esw_attr->sample->sample_flow;
pre_attr = sample_flow->pre_attr;
memset(pre_attr, 0, sizeof(*pre_attr));
esw = esw_psample->priv->mdev->priv.eswitch;
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->pre_rule, pre_attr);
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->rule, attr);
/* The following delete order can't be changed, otherwise,
* will hit fw syndromes.
*/
sample_flow = attr->sample_attr->sample_flow;
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->pre_rule, sample_flow->pre_attr);
if (!sample_flow->post_act_handle)
mlx5_eswitch_del_offloaded_rule(esw, sample_flow->post_rule,
sample_flow->post_attr);
sample_restore_put(esw_psample, sample_flow->restore);
mapping_remove(esw->offloads.reg_c0_obj_pool, esw_attr->sample->restore_obj_id);
sampler_put(esw_psample, sample_flow->sampler);
tbl_attr.chain = attr->chain;
tbl_attr.prio = attr->prio;
tbl_attr.vport = esw_attr->in_rep->vport;
tbl_attr.vport_ns = &mlx5_esw_vport_tbl_sample_ns;
mlx5_esw_vporttbl_put(esw, &tbl_attr);
sample_restore_put(tc_psample, sample_flow->restore);
mapping_remove(esw->offloads.reg_c0_obj_pool, attr->sample_attr->restore_obj_id);
sampler_put(tc_psample, sample_flow->sampler);
if (sample_flow->post_act_handle) {
mlx5e_tc_post_act_del(tc_psample->post_act, sample_flow->post_act_handle);
} else {
tbl_attr.chain = attr->chain;
tbl_attr.prio = attr->prio;
tbl_attr.vport = esw_attr->in_rep->vport;
tbl_attr.vport_ns = &mlx5_esw_vport_tbl_sample_ns;
mlx5_esw_vporttbl_put(esw, &tbl_attr);
kfree(sample_flow->post_attr);
}
kfree(pre_attr->esw_attr->sample);
kfree(pre_attr);
kfree(sample_flow->pre_attr);
kfree(sample_flow);
}
struct mlx5_esw_psample *
mlx5_esw_sample_init(struct mlx5e_priv *priv)
struct mlx5e_tc_psample *
mlx5e_tc_sample_init(struct mlx5_eswitch *esw, struct mlx5e_post_act *post_act)
{
struct mlx5_esw_psample *esw_psample;
struct mlx5e_tc_psample *tc_psample;
int err;
esw_psample = kzalloc(sizeof(*esw_psample), GFP_KERNEL);
if (!esw_psample)
tc_psample = kzalloc(sizeof(*tc_psample), GFP_KERNEL);
if (!tc_psample)
return ERR_PTR(-ENOMEM);
esw_psample->priv = priv;
err = sampler_termtbl_create(esw_psample);
if (IS_ERR_OR_NULL(post_act)) {
err = PTR_ERR(post_act);
goto err_post_act;
}
tc_psample->post_act = post_act;
tc_psample->esw = esw;
err = sampler_termtbl_create(tc_psample);
if (err)
goto err_termtbl;
goto err_post_act;
mutex_init(&esw_psample->ht_lock);
mutex_init(&esw_psample->restore_lock);
mutex_init(&tc_psample->ht_lock);
mutex_init(&tc_psample->restore_lock);
return esw_psample;
return tc_psample;
err_termtbl:
kfree(esw_psample);
err_post_act:
kfree(tc_psample);
return ERR_PTR(err);
}
void
mlx5_esw_sample_cleanup(struct mlx5_esw_psample *esw_psample)
mlx5e_tc_sample_cleanup(struct mlx5e_tc_psample *tc_psample)
{
if (IS_ERR_OR_NULL(esw_psample))
if (IS_ERR_OR_NULL(tc_psample))
return;
mutex_destroy(&esw_psample->restore_lock);
mutex_destroy(&esw_psample->ht_lock);
sampler_termtbl_destroy(esw_psample);
kfree(esw_psample);
mutex_destroy(&tc_psample->restore_lock);
mutex_destroy(&tc_psample->ht_lock);
sampler_termtbl_destroy(tc_psample);
kfree(tc_psample);
}

41
drivers/net/ethernet/mellanox/mlx5/core/en/tc/sample.h Normal file
Просмотреть файл

@ -0,0 +1,41 @@
/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/* Copyright (c) 2021 Mellanox Technologies. */
#ifndef __MLX5_EN_TC_SAMPLE_H__
#define __MLX5_EN_TC_SAMPLE_H__
#include "eswitch.h"
struct mlx5_flow_attr;
struct mlx5e_tc_psample;
struct mlx5e_post_act;
struct mlx5e_sample_attr {
u32 group_num;
u32 rate;
u32 trunc_size;
u32 restore_obj_id;
u32 sampler_id;
struct mlx5e_sample_flow *sample_flow;
};
void mlx5e_tc_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj);
struct mlx5_flow_handle *
mlx5e_tc_sample_offload(struct mlx5e_tc_psample *sample_priv,
struct mlx5_flow_spec *spec,
struct mlx5_flow_attr *attr,
u32 tunnel_id);
void
mlx5e_tc_sample_unoffload(struct mlx5e_tc_psample *sample_priv,
struct mlx5_flow_handle *rule,
struct mlx5_flow_attr *attr);
struct mlx5e_tc_psample *
mlx5e_tc_sample_init(struct mlx5_eswitch *esw, struct mlx5e_post_act *post_act);
void
mlx5e_tc_sample_cleanup(struct mlx5e_tc_psample *tc_psample);
#endif /* __MLX5_EN_TC_SAMPLE_H__ */

154
drivers/net/ethernet/mellanox/mlx5/core/en/tc_ct.c
Просмотреть файл

@ -19,6 +19,7 @@
#include "en/tc_ct.h"
#include "en/mod_hdr.h"
#include "en/mapping.h"
#include "en/tc/post_act.h"
#include "en.h"
#include "en_tc.h"
#include "en_rep.h"
@ -32,10 +33,6 @@
#define MLX5_CT_STATE_RELATED_BIT BIT(5)
#define MLX5_CT_STATE_INVALID_BIT BIT(6)
#define MLX5_FTE_ID_BITS (mlx5e_tc_attr_to_reg_mappings[FTEID_TO_REG].mlen)
#define MLX5_FTE_ID_MAX GENMASK(MLX5_FTE_ID_BITS - 1, 0)
#define MLX5_FTE_ID_MASK MLX5_FTE_ID_MAX
#define MLX5_CT_LABELS_BITS (mlx5e_tc_attr_to_reg_mappings[LABELS_TO_REG].mlen)
#define MLX5_CT_LABELS_MASK GENMASK(MLX5_CT_LABELS_BITS - 1, 0)
@ -46,14 +43,13 @@ struct mlx5_tc_ct_priv {
struct mlx5_core_dev *dev;
const struct net_device *netdev;
struct mod_hdr_tbl *mod_hdr_tbl;
struct idr fte_ids;
struct xarray tuple_ids;
struct rhashtable zone_ht;
struct rhashtable ct_tuples_ht;
struct rhashtable ct_tuples_nat_ht;
struct mlx5_flow_table *ct;
struct mlx5_flow_table *ct_nat;
struct mlx5_flow_table *post_ct;
struct mlx5e_post_act *post_act;
struct mutex control_lock; /* guards parallel adds/dels */
struct mapping_ctx *zone_mapping;
struct mapping_ctx *labels_mapping;
@ -64,11 +60,9 @@ struct mlx5_tc_ct_priv {
struct mlx5_ct_flow {
struct mlx5_flow_attr *pre_ct_attr;
struct mlx5_flow_attr *post_ct_attr;
struct mlx5_flow_handle *pre_ct_rule;
struct mlx5_flow_handle *post_ct_rule;
struct mlx5e_post_act_handle *post_act_handle;
struct mlx5_ct_ft *ft;
u32 fte_id;
u32 chain_mapping;
};
@ -768,7 +762,7 @@ mlx5_tc_ct_entry_add_rule(struct mlx5_tc_ct_priv *ct_priv,
MLX5_FLOW_CONTEXT_ACTION_FWD_DEST |
MLX5_FLOW_CONTEXT_ACTION_COUNT;
attr->dest_chain = 0;
attr->dest_ft = ct_priv->post_ct;
attr->dest_ft = mlx5e_tc_post_act_get_ft(ct_priv->post_act);
attr->ft = nat ? ct_priv->ct_nat : ct_priv->ct;
attr->outer_match_level = MLX5_MATCH_L4;
attr->counter = entry->counter->counter;
@ -1432,7 +1426,7 @@ static int tc_ct_pre_ct_add_rules(struct mlx5_ct_ft *ct_ft,
ctstate |= MLX5_CT_STATE_NAT_BIT;
mlx5e_tc_match_to_reg_match(spec, CTSTATE_TO_REG, ctstate, ctstate);
dest.ft = ct_priv->post_ct;
dest.ft = mlx5e_tc_post_act_get_ft(ct_priv->post_act);
rule = mlx5_add_flow_rules(ft, spec, &flow_act, &dest, 1);
if (IS_ERR(rule)) {
err = PTR_ERR(rule);
@ -1716,9 +1710,9 @@ mlx5_tc_ct_del_ft_cb(struct mlx5_tc_ct_priv *ct_priv, struct mlx5_ct_ft *ft)
* | do decap
* v
* +---------------------+
* + pre_ct/pre_ct_nat + if matches +---------------------+
* + zone+nat match +---------------->+ post_ct (see below) +
* +---------------------+ set zone +---------------------+
* + pre_ct/pre_ct_nat + if matches +-------------------------+
* + zone+nat match +---------------->+ post_act (see below) +
* +---------------------+ set zone +-------------------------+
* | set zone
* v
* +--------------------+
@ -1732,7 +1726,7 @@ mlx5_tc_ct_del_ft_cb(struct mlx5_tc_ct_priv *ct_priv, struct mlx5_ct_ft *ft)
* | do nat (if needed)
* v
* +--------------+
* + post_ct + original filter actions
* + post_act + original filter actions
* + fte_id match +------------------------>
* +--------------+
*/
@ -1746,19 +1740,15 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
struct mlx5e_priv *priv = netdev_priv(ct_priv->netdev);
struct mlx5e_tc_mod_hdr_acts pre_mod_acts = {};
u32 attr_sz = ns_to_attr_sz(ct_priv->ns_type);
struct mlx5_flow_spec *post_ct_spec = NULL;
struct mlx5e_post_act_handle *handle;
struct mlx5_flow_attr *pre_ct_attr;
struct mlx5_modify_hdr *mod_hdr;
struct mlx5_flow_handle *rule;
struct mlx5_ct_flow *ct_flow;
int chain_mapping = 0, err;
struct mlx5_ct_ft *ft;
u32 fte_id = 1;
post_ct_spec = kvzalloc(sizeof(*post_ct_spec), GFP_KERNEL);
ct_flow = kzalloc(sizeof(*ct_flow), GFP_KERNEL);
if (!post_ct_spec || !ct_flow) {
kvfree(post_ct_spec);
if (!ct_flow) {
kfree(ct_flow);
return ERR_PTR(-ENOMEM);
}
@ -1773,14 +1763,13 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
}
ct_flow->ft = ft;
err = idr_alloc_u32(&ct_priv->fte_ids, ct_flow, &fte_id,
MLX5_FTE_ID_MAX, GFP_KERNEL);
if (err) {
netdev_warn(priv->netdev,
"Failed to allocate fte id, err: %d\n", err);
goto err_idr;
handle = mlx5e_tc_post_act_add(ct_priv->post_act, attr);
if (IS_ERR(handle)) {
err = PTR_ERR(handle);
ct_dbg("Failed to allocate post action handle");
goto err_post_act_handle;
}
ct_flow->fte_id = fte_id;
ct_flow->post_act_handle = handle;
/* Base flow attributes of both rules on original rule attribute */
ct_flow->pre_ct_attr = mlx5_alloc_flow_attr(ct_priv->ns_type);
@ -1789,15 +1778,8 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
goto err_alloc_pre;
}
ct_flow->post_ct_attr = mlx5_alloc_flow_attr(ct_priv->ns_type);
if (!ct_flow->post_ct_attr) {
err = -ENOMEM;
goto err_alloc_post;
}
pre_ct_attr = ct_flow->pre_ct_attr;
memcpy(pre_ct_attr, attr, attr_sz);
memcpy(ct_flow->post_ct_attr, attr, attr_sz);
/* Modify the original rule's action to fwd and modify, leave decap */
pre_ct_attr->action = attr->action & MLX5_FLOW_CONTEXT_ACTION_DECAP;
@ -1823,10 +1805,9 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
goto err_mapping;
}
err = mlx5e_tc_match_to_reg_set(priv->mdev, &pre_mod_acts, ct_priv->ns_type,
FTEID_TO_REG, fte_id);
err = mlx5e_tc_post_act_set_handle(priv->mdev, handle, &pre_mod_acts);
if (err) {
ct_dbg("Failed to set fte_id register mapping");
ct_dbg("Failed to set post action handle");
goto err_mapping;
}
@ -1857,33 +1838,6 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
}
pre_ct_attr->modify_hdr = mod_hdr;
/* Post ct rule matches on fte_id and executes original rule's
* tc rule action
*/
mlx5e_tc_match_to_reg_match(post_ct_spec, FTEID_TO_REG,
fte_id, MLX5_FTE_ID_MASK);
/* Put post_ct rule on post_ct flow table */
ct_flow->post_ct_attr->chain = 0;
ct_flow->post_ct_attr->prio = 0;
ct_flow->post_ct_attr->ft = ct_priv->post_ct;
/* Splits were handled before CT */
if (ct_priv->ns_type == MLX5_FLOW_NAMESPACE_FDB)
ct_flow->post_ct_attr->esw_attr->split_count = 0;
ct_flow->post_ct_attr->inner_match_level = MLX5_MATCH_NONE;
ct_flow->post_ct_attr->outer_match_level = MLX5_MATCH_NONE;
ct_flow->post_ct_attr->action &= ~(MLX5_FLOW_CONTEXT_ACTION_DECAP);
rule = mlx5_tc_rule_insert(priv, post_ct_spec,
ct_flow->post_ct_attr);
ct_flow->post_ct_rule = rule;
if (IS_ERR(ct_flow->post_ct_rule)) {
err = PTR_ERR(ct_flow->post_ct_rule);
ct_dbg("Failed to add post ct rule");
goto err_insert_post_ct;
}
/* Change original rule point to ct table */
pre_ct_attr->dest_chain = 0;
pre_ct_attr->dest_ft = nat ? ft->pre_ct_nat.ft : ft->pre_ct.ft;
@ -1897,28 +1851,21 @@ __mlx5_tc_ct_flow_offload(struct mlx5_tc_ct_priv *ct_priv,
attr->ct_attr.ct_flow = ct_flow;
dealloc_mod_hdr_actions(&pre_mod_acts);
kvfree(post_ct_spec);
return rule;
return ct_flow->pre_ct_rule;
err_insert_orig:
mlx5_tc_rule_delete(priv, ct_flow->post_ct_rule,
ct_flow->post_ct_attr);
err_insert_post_ct:
mlx5_modify_header_dealloc(priv->mdev, pre_ct_attr->modify_hdr);
err_mapping:
dealloc_mod_hdr_actions(&pre_mod_acts);
mlx5_chains_put_chain_mapping(ct_priv->chains, ct_flow->chain_mapping);
err_get_chain:
kfree(ct_flow->post_ct_attr);
err_alloc_post:
kfree(ct_flow->pre_ct_attr);
err_alloc_pre:
idr_remove(&ct_priv->fte_ids, fte_id);
err_idr:
mlx5e_tc_post_act_del(ct_priv->post_act, handle);
err_post_act_handle:
mlx5_tc_ct_del_ft_cb(ct_priv, ft);
err_ft:
kvfree(post_ct_spec);
kfree(ct_flow);
netdev_warn(priv->netdev, "Failed to offload ct flow, err %d\n", err);
return ERR_PTR(err);
@ -2029,16 +1976,13 @@ __mlx5_tc_ct_delete_flow(struct mlx5_tc_ct_priv *ct_priv,
pre_ct_attr);
mlx5_modify_header_dealloc(priv->mdev, pre_ct_attr->modify_hdr);
if (ct_flow->post_ct_rule) {
mlx5_tc_rule_delete(priv, ct_flow->post_ct_rule,
ct_flow->post_ct_attr);
if (ct_flow->post_act_handle) {
mlx5_chains_put_chain_mapping(ct_priv->chains, ct_flow->chain_mapping);
idr_remove(&ct_priv->fte_ids, ct_flow->fte_id);
mlx5e_tc_post_act_del(ct_priv->post_act, ct_flow->post_act_handle);
mlx5_tc_ct_del_ft_cb(ct_priv, ct_flow->ft);
}
kfree(ct_flow->pre_ct_attr);
kfree(ct_flow->post_ct_attr);
kfree(ct_flow);
}
@ -2064,11 +2008,6 @@ static int
mlx5_tc_ct_init_check_esw_support(struct mlx5_eswitch *esw,
const char **err_msg)
{
if (!MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ignore_flow_level)) {
*err_msg = "firmware level support is missing";
return -EOPNOTSUPP;
}
if (!mlx5_eswitch_vlan_actions_supported(esw->dev, 1)) {
/* vlan workaround should be avoided for multi chain rules.
* This is just a sanity check as pop vlan action should
@ -2097,21 +2036,10 @@ mlx5_tc_ct_init_check_esw_support(struct mlx5_eswitch *esw,
return 0;
}
static int
mlx5_tc_ct_init_check_nic_support(struct mlx5e_priv *priv,
const char **err_msg)
{
if (!MLX5_CAP_FLOWTABLE_NIC_RX(priv->mdev, ignore_flow_level)) {
*err_msg = "firmware level support is missing";
return -EOPNOTSUPP;
}
return 0;
}
static int
mlx5_tc_ct_init_check_support(struct mlx5e_priv *priv,
enum mlx5_flow_namespace_type ns_type,
struct mlx5e_post_act *post_act,
const char **err_msg)
{
struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
@ -2122,10 +2050,14 @@ mlx5_tc_ct_init_check_support(struct mlx5e_priv *priv,
*err_msg = "tc skb extension missing";
return -EOPNOTSUPP;
#endif
if (IS_ERR_OR_NULL(post_act)) {
*err_msg = "tc ct offload not supported, post action is missing";
return -EOPNOTSUPP;
}
if (ns_type == MLX5_FLOW_NAMESPACE_FDB)
return mlx5_tc_ct_init_check_esw_support(esw, err_msg);
else
return mlx5_tc_ct_init_check_nic_support(priv, err_msg);
return 0;
}
#define INIT_ERR_PREFIX "tc ct offload init failed"
@ -2133,7 +2065,8 @@ mlx5_tc_ct_init_check_support(struct mlx5e_priv *priv,
struct mlx5_tc_ct_priv *
mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
struct mod_hdr_tbl *mod_hdr,
enum mlx5_flow_namespace_type ns_type)
enum mlx5_flow_namespace_type ns_type,
struct mlx5e_post_act *post_act)
{
struct mlx5_tc_ct_priv *ct_priv;
struct mlx5_core_dev *dev;
@ -2142,11 +2075,9 @@ mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
int err;
dev = priv->mdev;
err = mlx5_tc_ct_init_check_support(priv, ns_type, &msg);
err = mlx5_tc_ct_init_check_support(priv, ns_type, post_act, &msg);
if (err) {
mlx5_core_warn(dev,
"tc ct offload not supported, %s\n",
msg);
mlx5_core_warn(dev, "tc ct offload not supported, %s\n", msg);
goto err_support;
}
@ -2194,16 +2125,7 @@ mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
goto err_ct_nat_tbl;
}
ct_priv->post_ct = mlx5_chains_create_global_table(chains);
if (IS_ERR(ct_priv->post_ct)) {
err = PTR_ERR(ct_priv->post_ct);
mlx5_core_warn(dev,
"%s, failed to create post ct table err: %d\n",
INIT_ERR_PREFIX, err);
goto err_post_ct_tbl;
}
idr_init(&ct_priv->fte_ids);
ct_priv->post_act = post_act;
mutex_init(&ct_priv->control_lock);
rhashtable_init(&ct_priv->zone_ht, &zone_params);
rhashtable_init(&ct_priv->ct_tuples_ht, &tuples_ht_params);
@ -2211,8 +2133,6 @@ mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
return ct_priv;
err_post_ct_tbl:
mlx5_chains_destroy_global_table(chains, ct_priv->ct_nat);
err_ct_nat_tbl:
mlx5_chains_destroy_global_table(chains, ct_priv->ct);
err_ct_tbl:
@ -2237,7 +2157,6 @@ mlx5_tc_ct_clean(struct mlx5_tc_ct_priv *ct_priv)
chains = ct_priv->chains;
mlx5_chains_destroy_global_table(chains, ct_priv->post_ct);
mlx5_chains_destroy_global_table(chains, ct_priv->ct_nat);
mlx5_chains_destroy_global_table(chains, ct_priv->ct);
mapping_destroy(ct_priv->zone_mapping);
@ -2247,7 +2166,6 @@ mlx5_tc_ct_clean(struct mlx5_tc_ct_priv *ct_priv)
rhashtable_destroy(&ct_priv->ct_tuples_nat_ht);
rhashtable_destroy(&ct_priv->zone_ht);
mutex_destroy(&ct_priv->control_lock);
idr_destroy(&ct_priv->fte_ids);
kfree(ct_priv);
}

6
drivers/net/ethernet/mellanox/mlx5/core/en/tc_ct.h
Просмотреть файл

@ -92,7 +92,8 @@ struct mlx5_ct_attr {
struct mlx5_tc_ct_priv *
mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
struct mod_hdr_tbl *mod_hdr,
enum mlx5_flow_namespace_type ns_type);
enum mlx5_flow_namespace_type ns_type,
struct mlx5e_post_act *post_act);
void
mlx5_tc_ct_clean(struct mlx5_tc_ct_priv *ct_priv);
@ -132,7 +133,8 @@ mlx5e_tc_ct_restore_flow(struct mlx5_tc_ct_priv *ct_priv,
static inline struct mlx5_tc_ct_priv *
mlx5_tc_ct_init(struct mlx5e_priv *priv, struct mlx5_fs_chains *chains,
struct mod_hdr_tbl *mod_hdr,
enum mlx5_flow_namespace_type ns_type)
enum mlx5_flow_namespace_type ns_type,
struct mlx5e_post_act *post_act)
{
return NULL;
}

4
drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
Просмотреть файл

@ -60,6 +60,7 @@ struct mlx5e_neigh_update_table {
struct mlx5_tc_ct_priv;
struct mlx5e_rep_bond;
struct mlx5e_tc_tun_encap;
struct mlx5e_post_act;
struct mlx5_rep_uplink_priv {
/* Filters DB - instantiated by the uplink representor and shared by
@ -88,8 +89,9 @@ struct mlx5_rep_uplink_priv {
/* maps tun_enc_opts to a unique id*/
struct mapping_ctx *tunnel_enc_opts_mapping;
struct mlx5e_post_act *post_act;
struct mlx5_tc_ct_priv *ct_priv;
struct mlx5_esw_psample *esw_psample;
struct mlx5e_tc_psample *tc_psample;
/* support eswitch vports bonding */
struct mlx5e_rep_bond *bond;

57
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
Просмотреть файл

@ -47,6 +47,7 @@
#include <net/bareudp.h>
#include <net/bonding.h>
#include "en.h"
#include "en/tc/post_act.h"
#include "en_rep.h"
#include "en/rep/tc.h"
#include "en/rep/neigh.h"
@ -60,7 +61,7 @@
#include "en/mod_hdr.h"
#include "en/tc_priv.h"
#include "en/tc_tun_encap.h"
#include "esw/sample.h"
#include "en/tc/sample.h"
#include "lib/devcom.h"
#include "lib/geneve.h"
#include "lib/fs_chains.h"
@ -246,7 +247,7 @@ get_ct_priv(struct mlx5e_priv *priv)
}
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
static struct mlx5_esw_psample *
static struct mlx5e_tc_psample *
get_sample_priv(struct mlx5e_priv *priv)
{
struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
@ -257,7 +258,7 @@ get_sample_priv(struct mlx5e_priv *priv)
uplink_rpriv = mlx5_eswitch_get_uplink_priv(esw, REP_ETH);
uplink_priv = &uplink_rpriv->uplink_priv;
return uplink_priv->esw_psample;
return uplink_priv->tc_psample;
}
return NULL;
@ -1147,7 +1148,8 @@ mlx5e_tc_offload_fdb_rules(struct mlx5_eswitch *esw,
mod_hdr_acts);
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
} else if (flow_flag_test(flow, SAMPLE)) {
rule = mlx5_esw_sample_offload(get_sample_priv(flow->priv), spec, attr);
rule = mlx5e_tc_sample_offload(get_sample_priv(flow->priv), spec, attr,
mlx5e_tc_get_flow_tun_id(flow));
#endif
} else {
rule = mlx5_eswitch_add_offloaded_rule(esw, spec, attr);
@ -1186,7 +1188,7 @@ void mlx5e_tc_unoffload_fdb_rules(struct mlx5_eswitch *esw,
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
if (flow_flag_test(flow, SAMPLE)) {
mlx5_esw_sample_unoffload(get_sample_priv(flow->priv), flow->rule[0], attr);
mlx5e_tc_sample_unoffload(get_sample_priv(flow->priv), flow->rule[0], attr);
return;
}
#endif
@ -1550,6 +1552,7 @@ static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv,
else
mlx5e_detach_mod_hdr(priv, flow);
}
kfree(attr->sample_attr);
kvfree(attr->parse_attr);
kvfree(attr->esw_attr->rx_tun_attr);
@ -1559,7 +1562,6 @@ static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv,
if (flow_flag_test(flow, L3_TO_L2_DECAP))
mlx5e_detach_decap(priv, flow);
kfree(flow->attr->esw_attr->sample);
kfree(flow->attr);
}
@ -1624,17 +1626,22 @@ static void mlx5e_tc_del_flow(struct mlx5e_priv *priv,
}
}
static int flow_has_tc_fwd_action(struct flow_cls_offload *f)
static bool flow_requires_tunnel_mapping(u32 chain, struct flow_cls_offload *f)
{
struct flow_rule *rule = flow_cls_offload_flow_rule(f);
struct flow_action *flow_action = &rule->action;
const struct flow_action_entry *act;
int i;
if (chain)
return false;
flow_action_for_each(i, act, flow_action) {
switch (act->id) {
case FLOW_ACTION_GOTO:
return true;
case FLOW_ACTION_SAMPLE:
return true;
default:
continue;
}
@ -1875,7 +1882,7 @@ static int parse_tunnel_attr(struct mlx5e_priv *priv,
return -EOPNOTSUPP;
needs_mapping = !!flow->attr->chain;
sets_mapping = !flow->attr->chain && flow_has_tc_fwd_action(f);
sets_mapping = flow_requires_tunnel_mapping(flow->attr->chain, f);
*match_inner = !needs_mapping;
if ((needs_mapping || sets_mapping) &&
@ -3716,13 +3723,13 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
struct mlx5e_tc_flow_parse_attr *parse_attr;
struct mlx5e_rep_priv *rpriv = priv->ppriv;
struct mlx5e_sample_attr sample_attr = {};
const struct ip_tunnel_info *info = NULL;
struct mlx5_flow_attr *attr = flow->attr;
int ifindexes[MLX5_MAX_FLOW_FWD_VPORTS];
bool ft_flow = mlx5e_is_ft_flow(flow);
const struct flow_action_entry *act;
struct mlx5_esw_flow_attr *esw_attr;
struct mlx5_sample_attr sample = {};
bool encap = false, decap = false;
u32 action = attr->action;
int err, i, if_count = 0;
@ -3993,10 +4000,10 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
NL_SET_ERR_MSG_MOD(extack, "Sample action with connection tracking is not supported");
return -EOPNOTSUPP;
}
sample.rate = act->sample.rate;
sample.group_num = act->sample.psample_group->group_num;
sample_attr.rate = act->sample.rate;
sample_attr.group_num = act->sample.psample_group->group_num;
if (act->sample.truncate)
sample.trunc_size = act->sample.trunc_size;
sample_attr.trunc_size = act->sample.trunc_size;
flow_flag_set(flow, SAMPLE);
break;
default:
@ -4081,10 +4088,10 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
* no errors after parsing.
*/
if (flow_flag_test(flow, SAMPLE)) {
esw_attr->sample = kzalloc(sizeof(*esw_attr->sample), GFP_KERNEL);
if (!esw_attr->sample)
attr->sample_attr = kzalloc(sizeof(*attr->sample_attr), GFP_KERNEL);
if (!attr->sample_attr)
return -ENOMEM;
*esw_attr->sample = sample;
*attr->sample_attr = sample_attr;
}
return 0;
@ -4682,7 +4689,7 @@ static int apply_police_params(struct mlx5e_priv *priv, u64 rate,
rate_mbps = max_t(u32, rate, 1);
}
err = mlx5_esw_modify_vport_rate(esw, vport_num, rate_mbps);
err = mlx5_esw_qos_modify_vport_rate(esw, vport_num, rate_mbps);
if (err)
NL_SET_ERR_MSG_MOD(extack, "failed applying action to hardware");
@ -4895,8 +4902,9 @@ int mlx5e_tc_nic_init(struct mlx5e_priv *priv)
goto err_chains;
}
tc->post_act = mlx5e_tc_post_act_init(priv, tc->chains, MLX5_FLOW_NAMESPACE_KERNEL);
tc->ct = mlx5_tc_ct_init(priv, tc->chains, &priv->fs.tc.mod_hdr,
MLX5_FLOW_NAMESPACE_KERNEL);
MLX5_FLOW_NAMESPACE_KERNEL, tc->post_act);
tc->netdevice_nb.notifier_call = mlx5e_tc_netdev_event;
err = register_netdevice_notifier_dev_net(priv->netdev,
@ -4912,6 +4920,7 @@ int mlx5e_tc_nic_init(struct mlx5e_priv *priv)
err_reg:
mlx5_tc_ct_clean(tc->ct);
mlx5e_tc_post_act_destroy(tc->post_act);
mlx5_chains_destroy(tc->chains);
err_chains:
mapping_destroy(chains_mapping);
@ -4950,6 +4959,7 @@ void mlx5e_tc_nic_cleanup(struct mlx5e_priv *priv)
mutex_destroy(&tc->t_lock);
mlx5_tc_ct_clean(tc->ct);
mlx5e_tc_post_act_destroy(tc->post_act);
mapping_destroy(tc->mapping);
mlx5_chains_destroy(tc->chains);
}
@ -4970,13 +4980,16 @@ int mlx5e_tc_esw_init(struct rhashtable *tc_ht)
priv = netdev_priv(rpriv->netdev);
esw = priv->mdev->priv.eswitch;
uplink_priv->post_act = mlx5e_tc_post_act_init(priv, esw_chains(esw),
MLX5_FLOW_NAMESPACE_FDB);
uplink_priv->ct_priv = mlx5_tc_ct_init(netdev_priv(priv->netdev),
esw_chains(esw),
&esw->offloads.mod_hdr,
MLX5_FLOW_NAMESPACE_FDB);
MLX5_FLOW_NAMESPACE_FDB,
uplink_priv->post_act);
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
uplink_priv->esw_psample = mlx5_esw_sample_init(netdev_priv(priv->netdev));
uplink_priv->tc_psample = mlx5e_tc_sample_init(esw, uplink_priv->post_act);
#endif
mapping_id = mlx5_query_nic_system_image_guid(esw->dev);
@ -5022,11 +5035,12 @@ err_enc_opts_mapping:
mapping_destroy(uplink_priv->tunnel_mapping);
err_tun_mapping:
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
mlx5_esw_sample_cleanup(uplink_priv->esw_psample);
mlx5e_tc_sample_cleanup(uplink_priv->tc_psample);
#endif
mlx5_tc_ct_clean(uplink_priv->ct_priv);
netdev_warn(priv->netdev,
"Failed to initialize tc (eswitch), err: %d", err);
mlx5e_tc_post_act_destroy(uplink_priv->post_act);
return err;
}
@ -5043,9 +5057,10 @@ void mlx5e_tc_esw_cleanup(struct rhashtable *tc_ht)
mapping_destroy(uplink_priv->tunnel_mapping);
#if IS_ENABLED(CONFIG_MLX5_TC_SAMPLE)
mlx5_esw_sample_cleanup(uplink_priv->esw_psample);
mlx5e_tc_sample_cleanup(uplink_priv->tc_psample);
#endif
mlx5_tc_ct_clean(uplink_priv->ct_priv);
mlx5e_tc_post_act_destroy(uplink_priv->post_act);
}
int mlx5e_tc_num_filters(struct mlx5e_priv *priv, unsigned long flags)

1
drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
Просмотреть файл

@ -70,6 +70,7 @@ struct mlx5_flow_attr {
struct mlx5_fc *counter;
struct mlx5_modify_hdr *modify_hdr;
struct mlx5_ct_attr ct_attr;
struct mlx5e_sample_attr *sample_attr;
struct mlx5e_tc_flow_parse_attr *parse_attr;
u32 chain;
u16 prio;

26
drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c
Просмотреть файл

@ -91,9 +91,15 @@ int mlx5_esw_offloads_devlink_port_register(struct mlx5_eswitch *esw, u16 vport_
if (err)
goto reg_err;
err = devlink_rate_leaf_create(dl_port, vport);
if (err)
goto rate_err;
vport->dl_port = dl_port;
return 0;
rate_err:
devlink_port_unregister(dl_port);
reg_err:
mlx5_esw_dl_port_free(dl_port);
return err;
@ -109,6 +115,12 @@ void mlx5_esw_offloads_devlink_port_unregister(struct mlx5_eswitch *esw, u16 vpo
vport = mlx5_eswitch_get_vport(esw, vport_num);
if (IS_ERR(vport))
return;
if (vport->dl_port->devlink_rate) {
mlx5_esw_qos_vport_update_group(esw, vport, NULL, NULL);
devlink_rate_leaf_destroy(vport->dl_port);
}
devlink_port_unregister(vport->dl_port);
mlx5_esw_dl_port_free(vport->dl_port);
vport->dl_port = NULL;
@ -148,8 +160,16 @@ int mlx5_esw_devlink_sf_port_register(struct mlx5_eswitch *esw, struct devlink_p
if (err)
return err;
err = devlink_rate_leaf_create(dl_port, vport);
if (err)
goto rate_err;
vport->dl_port = dl_port;
return 0;
rate_err:
devlink_port_unregister(dl_port);
return err;
}
void mlx5_esw_devlink_sf_port_unregister(struct mlx5_eswitch *esw, u16 vport_num)
@ -159,6 +179,12 @@ void mlx5_esw_devlink_sf_port_unregister(struct mlx5_eswitch *esw, u16 vport_num
vport = mlx5_eswitch_get_vport(esw, vport_num);
if (IS_ERR(vport))
return;
if (vport->dl_port->devlink_rate) {
mlx5_esw_qos_vport_update_group(esw, vport, NULL, NULL);
devlink_rate_leaf_destroy(vport->dl_port);
}
devlink_port_unregister(vport->dl_port);
vport->dl_port = NULL;
}

123
drivers/net/ethernet/mellanox/mlx5/core/esw/diag/qos_tracepoint.h Normal file
Просмотреть файл

@ -0,0 +1,123 @@
/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/* Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved. */
#undef TRACE_SYSTEM
#define TRACE_SYSTEM mlx5
#if !defined(_MLX5_ESW_TP_) || defined(TRACE_HEADER_MULTI_READ)
#define _MLX5_ESW_TP_
#include <linux/tracepoint.h>
#include "eswitch.h"
TRACE_EVENT(mlx5_esw_vport_qos_destroy,
TP_PROTO(const struct mlx5_vport *vport),
TP_ARGS(vport),
TP_STRUCT__entry(__string(devname, dev_name(vport->dev->device))
__field(unsigned short, vport_id)
__field(unsigned int, tsar_ix)
),
TP_fast_assign(__assign_str(devname, dev_name(vport->dev->device));
__entry->vport_id = vport->vport;
__entry->tsar_ix = vport->qos.esw_tsar_ix;
),
TP_printk("(%s) vport=%hu tsar_ix=%u\n",
__get_str(devname), __entry->vport_id, __entry->tsar_ix
)
);
DECLARE_EVENT_CLASS(mlx5_esw_vport_qos_template,
TP_PROTO(const struct mlx5_vport *vport, u32 bw_share, u32 max_rate),
TP_ARGS(vport, bw_share, max_rate),
TP_STRUCT__entry(__string(devname, dev_name(vport->dev->device))
__field(unsigned short, vport_id)
__field(unsigned int, tsar_ix)
__field(unsigned int, bw_share)
__field(unsigned int, max_rate)
__field(void *, group)
),
TP_fast_assign(__assign_str(devname, dev_name(vport->dev->device));
__entry->vport_id = vport->vport;
__entry->tsar_ix = vport->qos.esw_tsar_ix;
__entry->bw_share = bw_share;
__entry->max_rate = max_rate;
__entry->group = vport->qos.group;
),
TP_printk("(%s) vport=%hu tsar_ix=%u bw_share=%u, max_rate=%u group=%p\n",
__get_str(devname), __entry->vport_id, __entry->tsar_ix,
__entry->bw_share, __entry->max_rate, __entry->group
)
);
DEFINE_EVENT(mlx5_esw_vport_qos_template, mlx5_esw_vport_qos_create,
TP_PROTO(const struct mlx5_vport *vport, u32 bw_share, u32 max_rate),
TP_ARGS(vport, bw_share, max_rate)
);
DEFINE_EVENT(mlx5_esw_vport_qos_template, mlx5_esw_vport_qos_config,
TP_PROTO(const struct mlx5_vport *vport, u32 bw_share, u32 max_rate),
TP_ARGS(vport, bw_share, max_rate)
);
DECLARE_EVENT_CLASS(mlx5_esw_group_qos_template,
TP_PROTO(const struct mlx5_core_dev *dev,
const struct mlx5_esw_rate_group *group,
unsigned int tsar_ix),
TP_ARGS(dev, group, tsar_ix),
TP_STRUCT__entry(__string(devname, dev_name(dev->device))
__field(const void *, group)
__field(unsigned int, tsar_ix)
),
TP_fast_assign(__assign_str(devname, dev_name(dev->device));
__entry->group = group;
__entry->tsar_ix = tsar_ix;
),
TP_printk("(%s) group=%p tsar_ix=%u\n",
__get_str(devname), __entry->group, __entry->tsar_ix
)
);
DEFINE_EVENT(mlx5_esw_group_qos_template, mlx5_esw_group_qos_create,
TP_PROTO(const struct mlx5_core_dev *dev,
const struct mlx5_esw_rate_group *group,
unsigned int tsar_ix),
TP_ARGS(dev, group, tsar_ix)
);
DEFINE_EVENT(mlx5_esw_group_qos_template, mlx5_esw_group_qos_destroy,
TP_PROTO(const struct mlx5_core_dev *dev,
const struct mlx5_esw_rate_group *group,
unsigned int tsar_ix),
TP_ARGS(dev, group, tsar_ix)
);
TRACE_EVENT(mlx5_esw_group_qos_config,
TP_PROTO(const struct mlx5_core_dev *dev,
const struct mlx5_esw_rate_group *group,
unsigned int tsar_ix, u32 bw_share, u32 max_rate),
TP_ARGS(dev, group, tsar_ix, bw_share, max_rate),
TP_STRUCT__entry(__string(devname, dev_name(dev->device))
__field(const void *, group)
__field(unsigned int, tsar_ix)
__field(unsigned int, bw_share)
__field(unsigned int, max_rate)
),
TP_fast_assign(__assign_str(devname, dev_name(dev->device));
__entry->group = group;
__entry->tsar_ix = tsar_ix;
__entry->bw_share = bw_share;
__entry->max_rate = max_rate;
),
TP_printk("(%s) group=%p tsar_ix=%u bw_share=%u max_rate=%u\n",
__get_str(devname), __entry->group, __entry->tsar_ix,
__entry->bw_share, __entry->max_rate
)
);
#endif /* _MLX5_ESW_TP_ */
/* This part must be outside protection */
#undef TRACE_INCLUDE_PATH
#define TRACE_INCLUDE_PATH esw/diag
#undef TRACE_INCLUDE_FILE
#define TRACE_INCLUDE_FILE qos_tracepoint
#include <trace/define_trace.h>

20
drivers/net/ethernet/mellanox/mlx5/core/esw/legacy.c
Просмотреть файл

@ -11,6 +11,7 @@
#include "mlx5_core.h"
#include "eswitch.h"
#include "fs_core.h"
#include "esw/qos.h"
enum {
LEGACY_VEPA_PRIO = 0,
@ -508,3 +509,22 @@ unlock:
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, u16 vport,
u32 max_rate, u32 min_rate)
{
struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
int err;
if (!mlx5_esw_allowed(esw))
return -EPERM;
if (IS_ERR(evport))
return PTR_ERR(evport);
mutex_lock(&esw->state_lock);
err = mlx5_esw_qos_set_vport_min_rate(esw, evport, min_rate, NULL);
if (!err)
err = mlx5_esw_qos_set_vport_max_rate(esw, evport, max_rate, NULL);
mutex_unlock(&esw->state_lock);
return err;
}

869
drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c Normal file
Просмотреть файл

@ -0,0 +1,869 @@
// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
/* Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved. */
#include "eswitch.h"
#include "esw/qos.h"
#include "en/port.h"
#define CREATE_TRACE_POINTS
#include "diag/qos_tracepoint.h"
/* Minimum supported BW share value by the HW is 1 Mbit/sec */
#define MLX5_MIN_BW_SHARE 1
#define MLX5_RATE_TO_BW_SHARE(rate, divider, limit) \
min_t(u32, max_t(u32, DIV_ROUND_UP(rate, divider), MLX5_MIN_BW_SHARE), limit)
struct mlx5_esw_rate_group {
u32 tsar_ix;
u32 max_rate;
u32 min_rate;
u32 bw_share;
struct list_head list;
};
static int esw_qos_tsar_config(struct mlx5_core_dev *dev, u32 *sched_ctx,
u32 parent_ix, u32 tsar_ix,
u32 max_rate, u32 bw_share)
{
u32 bitmask = 0;
if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
return -EOPNOTSUPP;
MLX5_SET(scheduling_context, sched_ctx, parent_element_id, parent_ix);
MLX5_SET(scheduling_context, sched_ctx, max_average_bw, max_rate);
MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
return mlx5_modify_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
sched_ctx,
tsar_ix,
bitmask);
}
static int esw_qos_group_config(struct mlx5_eswitch *esw, struct mlx5_esw_rate_group *group,
u32 max_rate, u32 bw_share, struct netlink_ext_ack *extack)
{
u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_core_dev *dev = esw->dev;
int err;
err = esw_qos_tsar_config(dev, sched_ctx,
esw->qos.root_tsar_ix, group->tsar_ix,
max_rate, bw_share);
if (err)
NL_SET_ERR_MSG_MOD(extack, "E-Switch modify group TSAR element failed");
trace_mlx5_esw_group_qos_config(dev, group, group->tsar_ix, bw_share, max_rate);
return err;
}
static int esw_qos_vport_config(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
u32 max_rate, u32 bw_share,
struct netlink_ext_ack *extack)
{
u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_esw_rate_group *group = vport->qos.group;
struct mlx5_core_dev *dev = esw->dev;
u32 parent_tsar_ix;
void *vport_elem;
int err;
if (!vport->qos.enabled)
return -EIO;
parent_tsar_ix = group ? group->tsar_ix : esw->qos.root_tsar_ix;
MLX5_SET(scheduling_context, sched_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
element_attributes);
MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
err = esw_qos_tsar_config(dev, sched_ctx, parent_tsar_ix, vport->qos.esw_tsar_ix,
max_rate, bw_share);
if (err) {
esw_warn(esw->dev,
"E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
NL_SET_ERR_MSG_MOD(extack, "E-Switch modify TSAR vport element failed");
return err;
}
trace_mlx5_esw_vport_qos_config(vport, bw_share, max_rate);
return 0;
}
static u32 esw_qos_calculate_min_rate_divider(struct mlx5_eswitch *esw,
struct mlx5_esw_rate_group *group,
bool group_level)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
struct mlx5_vport *evport;
u32 max_guarantee = 0;
unsigned long i;
if (group_level) {
struct mlx5_esw_rate_group *group;
list_for_each_entry(group, &esw->qos.groups, list) {
if (group->min_rate < max_guarantee)
continue;
max_guarantee = group->min_rate;
}
} else {
mlx5_esw_for_each_vport(esw, i, evport) {
if (!evport->enabled || !evport->qos.enabled ||
evport->qos.group != group || evport->qos.min_rate < max_guarantee)
continue;
max_guarantee = evport->qos.min_rate;
}
}
if (max_guarantee)
return max_t(u32, max_guarantee / fw_max_bw_share, 1);
/* If vports min rate divider is 0 but their group has bw_share configured, then
* need to set bw_share for vports to minimal value.
*/
if (!group_level && !max_guarantee && group->bw_share)
return 1;
return 0;
}
static u32 esw_qos_calc_bw_share(u32 min_rate, u32 divider, u32 fw_max)
{
if (divider)
return MLX5_RATE_TO_BW_SHARE(min_rate, divider, fw_max);
return 0;
}
static int esw_qos_normalize_vports_min_rate(struct mlx5_eswitch *esw,
struct mlx5_esw_rate_group *group,
struct netlink_ext_ack *extack)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
u32 divider = esw_qos_calculate_min_rate_divider(esw, group, false);
struct mlx5_vport *evport;
unsigned long i;
u32 bw_share;
int err;
mlx5_esw_for_each_vport(esw, i, evport) {
if (!evport->enabled || !evport->qos.enabled || evport->qos.group != group)
continue;
bw_share = esw_qos_calc_bw_share(evport->qos.min_rate, divider, fw_max_bw_share);
if (bw_share == evport->qos.bw_share)
continue;
err = esw_qos_vport_config(esw, evport, evport->qos.max_rate, bw_share, extack);
if (err)
return err;
evport->qos.bw_share = bw_share;
}
return 0;
}
static int esw_qos_normalize_groups_min_rate(struct mlx5_eswitch *esw, u32 divider,
struct netlink_ext_ack *extack)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
struct mlx5_esw_rate_group *group;
u32 bw_share;
int err;
list_for_each_entry(group, &esw->qos.groups, list) {
bw_share = esw_qos_calc_bw_share(group->min_rate, divider, fw_max_bw_share);
if (bw_share == group->bw_share)
continue;
err = esw_qos_group_config(esw, group, group->max_rate, bw_share, extack);
if (err)
return err;
group->bw_share = bw_share;
/* All the group's vports need to be set with default bw_share
* to enable them with QOS
*/
err = esw_qos_normalize_vports_min_rate(esw, group, extack);
if (err)
return err;
}
return 0;
}
int mlx5_esw_qos_set_vport_min_rate(struct mlx5_eswitch *esw,
struct mlx5_vport *evport,
u32 min_rate,
struct netlink_ext_ack *extack)
{
u32 fw_max_bw_share, previous_min_rate;
bool min_rate_supported;
int err;
lockdep_assert_held(&esw->state_lock);
fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
fw_max_bw_share >= MLX5_MIN_BW_SHARE;
if (min_rate && !min_rate_supported)
return -EOPNOTSUPP;
if (min_rate == evport->qos.min_rate)
return 0;
previous_min_rate = evport->qos.min_rate;
evport->qos.min_rate = min_rate;
err = esw_qos_normalize_vports_min_rate(esw, evport->qos.group, extack);
if (err)
evport->qos.min_rate = previous_min_rate;
return err;
}
int mlx5_esw_qos_set_vport_max_rate(struct mlx5_eswitch *esw,
struct mlx5_vport *evport,
u32 max_rate,
struct netlink_ext_ack *extack)
{
u32 act_max_rate = max_rate;
bool max_rate_supported;
int err;
lockdep_assert_held(&esw->state_lock);
max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
if (max_rate && !max_rate_supported)
return -EOPNOTSUPP;
if (max_rate == evport->qos.max_rate)
return 0;
/* If parent group has rate limit need to set to group
* value when new max rate is 0.
*/
if (evport->qos.group && !max_rate)
act_max_rate = evport->qos.group->max_rate;
err = esw_qos_vport_config(esw, evport, act_max_rate, evport->qos.bw_share, extack);
if (!err)
evport->qos.max_rate = max_rate;
return err;
}
static int esw_qos_set_group_min_rate(struct mlx5_eswitch *esw, struct mlx5_esw_rate_group *group,
u32 min_rate, struct netlink_ext_ack *extack)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
struct mlx5_core_dev *dev = esw->dev;
u32 previous_min_rate, divider;
int err;
if (!(MLX5_CAP_QOS(dev, esw_bw_share) && fw_max_bw_share >= MLX5_MIN_BW_SHARE))
return -EOPNOTSUPP;
if (min_rate == group->min_rate)
return 0;
previous_min_rate = group->min_rate;
group->min_rate = min_rate;
divider = esw_qos_calculate_min_rate_divider(esw, group, true);
err = esw_qos_normalize_groups_min_rate(esw, divider, extack);
if (err) {
group->min_rate = previous_min_rate;
NL_SET_ERR_MSG_MOD(extack, "E-Switch group min rate setting failed");
/* Attempt restoring previous configuration */
divider = esw_qos_calculate_min_rate_divider(esw, group, true);
if (esw_qos_normalize_groups_min_rate(esw, divider, extack))
NL_SET_ERR_MSG_MOD(extack, "E-Switch BW share restore failed");
}
return err;
}
static int esw_qos_set_group_max_rate(struct mlx5_eswitch *esw,
struct mlx5_esw_rate_group *group,
u32 max_rate, struct netlink_ext_ack *extack)
{
struct mlx5_vport *vport;
unsigned long i;
int err;
if (group->max_rate == max_rate)
return 0;
err = esw_qos_group_config(esw, group, max_rate, group->bw_share, extack);
if (err)
return err;
group->max_rate = max_rate;
/* Any unlimited vports in the group should be set
* with the value of the group.
*/
mlx5_esw_for_each_vport(esw, i, vport) {
if (!vport->enabled || !vport->qos.enabled ||
vport->qos.group != group || vport->qos.max_rate)
continue;
err = esw_qos_vport_config(esw, vport, max_rate, vport->qos.bw_share, extack);
if (err)
NL_SET_ERR_MSG_MOD(extack,
"E-Switch vport implicit rate limit setting failed");
}
return err;
}
static int esw_qos_vport_create_sched_element(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
u32 max_rate, u32 bw_share)
{
u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_esw_rate_group *group = vport->qos.group;
struct mlx5_core_dev *dev = esw->dev;
u32 parent_tsar_ix;
void *vport_elem;
int err;
parent_tsar_ix = group ? group->tsar_ix : esw->qos.root_tsar_ix;
MLX5_SET(scheduling_context, sched_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx, element_attributes);
MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
MLX5_SET(scheduling_context, sched_ctx, parent_element_id, parent_tsar_ix);
MLX5_SET(scheduling_context, sched_ctx, max_average_bw, max_rate);
MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
err = mlx5_create_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
sched_ctx,
&vport->qos.esw_tsar_ix);
if (err) {
esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
return err;
}
return 0;
}
static int esw_qos_update_group_scheduling_element(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
struct mlx5_esw_rate_group *curr_group,
struct mlx5_esw_rate_group *new_group,
struct netlink_ext_ack *extack)
{
u32 max_rate;
int err;
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
vport->qos.esw_tsar_ix);
if (err) {
NL_SET_ERR_MSG_MOD(extack, "E-Switch destroy TSAR vport element failed");
return err;
}
vport->qos.group = new_group;
max_rate = vport->qos.max_rate ? vport->qos.max_rate : new_group->max_rate;
/* If vport is unlimited, we set the group's value.
* Therefore, if the group is limited it will apply to
* the vport as well and if not, vport will remain unlimited.
*/
err = esw_qos_vport_create_sched_element(esw, vport, max_rate, vport->qos.bw_share);
if (err) {
NL_SET_ERR_MSG_MOD(extack, "E-Switch vport group set failed.");
goto err_sched;
}
return 0;
err_sched:
vport->qos.group = curr_group;
max_rate = vport->qos.max_rate ? vport->qos.max_rate : curr_group->max_rate;
if (esw_qos_vport_create_sched_element(esw, vport, max_rate, vport->qos.bw_share))
esw_warn(esw->dev, "E-Switch vport group restore failed (vport=%d)\n",
vport->vport);
return err;
}
static int esw_qos_vport_update_group(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
struct mlx5_esw_rate_group *group,
struct netlink_ext_ack *extack)
{
struct mlx5_esw_rate_group *new_group, *curr_group;
int err;
if (!vport->enabled)
return -EINVAL;
curr_group = vport->qos.group;
new_group = group ?: esw->qos.group0;
if (curr_group == new_group)
return 0;
err = esw_qos_update_group_scheduling_element(esw, vport, curr_group, new_group, extack);
if (err)
return err;
/* Recalculate bw share weights of old and new groups */
if (vport->qos.bw_share) {
esw_qos_normalize_vports_min_rate(esw, curr_group, extack);
esw_qos_normalize_vports_min_rate(esw, new_group, extack);
}
return 0;
}
static struct mlx5_esw_rate_group *
esw_qos_create_rate_group(struct mlx5_eswitch *esw, struct netlink_ext_ack *extack)
{
u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_esw_rate_group *group;
u32 divider;
int err;
if (!MLX5_CAP_QOS(esw->dev, log_esw_max_sched_depth))
return ERR_PTR(-EOPNOTSUPP);
group = kzalloc(sizeof(*group), GFP_KERNEL);
if (!group)
return ERR_PTR(-ENOMEM);
MLX5_SET(scheduling_context, tsar_ctx, parent_element_id,
esw->qos.root_tsar_ix);
err = mlx5_create_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
tsar_ctx,
&group->tsar_ix);
if (err) {
NL_SET_ERR_MSG_MOD(extack, "E-Switch create TSAR for group failed");
goto err_sched_elem;
}
list_add_tail(&group->list, &esw->qos.groups);
divider = esw_qos_calculate_min_rate_divider(esw, group, true);
if (divider) {
err = esw_qos_normalize_groups_min_rate(esw, divider, extack);
if (err) {
NL_SET_ERR_MSG_MOD(extack, "E-Switch groups normalization failed");
goto err_min_rate;
}
}
trace_mlx5_esw_group_qos_create(esw->dev, group, group->tsar_ix);
return group;
err_min_rate:
list_del(&group->list);
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
group->tsar_ix);
if (err)
NL_SET_ERR_MSG_MOD(extack, "E-Switch destroy TSAR for group failed");
err_sched_elem:
kfree(group);
return ERR_PTR(err);
}
static int esw_qos_destroy_rate_group(struct mlx5_eswitch *esw,
struct mlx5_esw_rate_group *group,
struct netlink_ext_ack *extack)
{
u32 divider;
int err;
list_del(&group->list);
divider = esw_qos_calculate_min_rate_divider(esw, NULL, true);
err = esw_qos_normalize_groups_min_rate(esw, divider, extack);
if (err)
NL_SET_ERR_MSG_MOD(extack, "E-Switch groups' normalization failed");
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
group->tsar_ix);
if (err)
NL_SET_ERR_MSG_MOD(extack, "E-Switch destroy TSAR_ID failed");
trace_mlx5_esw_group_qos_destroy(esw->dev, group, group->tsar_ix);
kfree(group);
return err;
}
static bool esw_qos_element_type_supported(struct mlx5_core_dev *dev, int type)
{
switch (type) {
case SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_TASR;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_VPORT;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT_TC:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_VPORT_TC;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_PARA_VPORT_TC:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_PARA_VPORT_TC;
}
return false;
}
void mlx5_esw_qos_create(struct mlx5_eswitch *esw)
{
u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_core_dev *dev = esw->dev;
__be32 *attr;
int err;
if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
return;
if (!esw_qos_element_type_supported(dev, SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR))
return;
mutex_lock(&esw->state_lock);
if (esw->qos.enabled)
goto unlock;
MLX5_SET(scheduling_context, tsar_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR);
attr = MLX5_ADDR_OF(scheduling_context, tsar_ctx, element_attributes);
*attr = cpu_to_be32(TSAR_ELEMENT_TSAR_TYPE_DWRR << 16);
err = mlx5_create_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
tsar_ctx,
&esw->qos.root_tsar_ix);
if (err) {
esw_warn(dev, "E-Switch create root TSAR failed (%d)\n", err);
goto unlock;
}
INIT_LIST_HEAD(&esw->qos.groups);
if (MLX5_CAP_QOS(dev, log_esw_max_sched_depth)) {
esw->qos.group0 = esw_qos_create_rate_group(esw, NULL);
if (IS_ERR(esw->qos.group0)) {
esw_warn(dev, "E-Switch create rate group 0 failed (%ld)\n",
PTR_ERR(esw->qos.group0));
goto err_group0;
}
}
esw->qos.enabled = true;
unlock:
mutex_unlock(&esw->state_lock);
return;
err_group0:
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
esw->qos.root_tsar_ix);
if (err)
esw_warn(esw->dev, "E-Switch destroy root TSAR failed (%d)\n", err);
mutex_unlock(&esw->state_lock);
}
void mlx5_esw_qos_destroy(struct mlx5_eswitch *esw)
{
struct devlink *devlink = priv_to_devlink(esw->dev);
int err;
devlink_rate_nodes_destroy(devlink);
mutex_lock(&esw->state_lock);
if (!esw->qos.enabled)
goto unlock;
if (esw->qos.group0)
esw_qos_destroy_rate_group(esw, esw->qos.group0, NULL);
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
esw->qos.root_tsar_ix);
if (err)
esw_warn(esw->dev, "E-Switch destroy root TSAR failed (%d)\n", err);
esw->qos.enabled = false;
unlock:
mutex_unlock(&esw->state_lock);
}
int mlx5_esw_qos_vport_enable(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
u32 max_rate, u32 bw_share)
{
int err;
lockdep_assert_held(&esw->state_lock);
if (!esw->qos.enabled)
return 0;
if (vport->qos.enabled)
return -EEXIST;
vport->qos.group = esw->qos.group0;
err = esw_qos_vport_create_sched_element(esw, vport, max_rate, bw_share);
if (!err) {
vport->qos.enabled = true;
trace_mlx5_esw_vport_qos_create(vport, bw_share, max_rate);
}
return err;
}
void mlx5_esw_qos_vport_disable(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
{
int err;
lockdep_assert_held(&esw->state_lock);
if (!esw->qos.enabled || !vport->qos.enabled)
return;
WARN(vport->qos.group && vport->qos.group != esw->qos.group0,
"Disabling QoS on port before detaching it from group");
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
vport->qos.esw_tsar_ix);
if (err)
esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
vport->qos.enabled = false;
trace_mlx5_esw_vport_qos_destroy(vport);
}
int mlx5_esw_qos_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num, u32 rate_mbps)
{
u32 ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_vport *vport;
u32 bitmask;
vport = mlx5_eswitch_get_vport(esw, vport_num);
if (IS_ERR(vport))
return PTR_ERR(vport);
if (!vport->qos.enabled)
return -EOPNOTSUPP;
MLX5_SET(scheduling_context, ctx, max_average_bw, rate_mbps);
bitmask = MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
return mlx5_modify_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
ctx,
vport->qos.esw_tsar_ix,
bitmask);
}
#define MLX5_LINKSPEED_UNIT 125000 /* 1Mbps in Bps */
/* Converts bytes per second value passed in a pointer into megabits per
* second, rewriting last. If converted rate exceed link speed or is not a
* fraction of Mbps - returns error.
*/
static int esw_qos_devlink_rate_to_mbps(struct mlx5_core_dev *mdev, const char *name,
u64 *rate, struct netlink_ext_ack *extack)
{
u32 link_speed_max, reminder;
u64 value;
int err;
err = mlx5e_port_max_linkspeed(mdev, &link_speed_max);
if (err) {
NL_SET_ERR_MSG_MOD(extack, "Failed to get link maximum speed");
return err;
}
value = div_u64_rem(*rate, MLX5_LINKSPEED_UNIT, &reminder);
if (reminder) {
pr_err("%s rate value %lluBps not in link speed units of 1Mbps.\n",
name, *rate);
NL_SET_ERR_MSG_MOD(extack, "TX rate value not in link speed units of 1Mbps");
return -EINVAL;
}
if (value > link_speed_max) {
pr_err("%s rate value %lluMbps exceed link maximum speed %u.\n",
name, value, link_speed_max);
NL_SET_ERR_MSG_MOD(extack, "TX rate value exceed link maximum speed");
return -EINVAL;
}
*rate = value;
return 0;
}
/* Eswitch devlink rate API */
int mlx5_esw_devlink_rate_leaf_tx_share_set(struct devlink_rate *rate_leaf, void *priv,
u64 tx_share, struct netlink_ext_ack *extack)
{
struct mlx5_vport *vport = priv;
struct mlx5_eswitch *esw;
int err;
esw = vport->dev->priv.eswitch;
if (!mlx5_esw_allowed(esw))
return -EPERM;
err = esw_qos_devlink_rate_to_mbps(vport->dev, "tx_share", &tx_share, extack);
if (err)
return err;
mutex_lock(&esw->state_lock);
err = mlx5_esw_qos_set_vport_min_rate(esw, vport, tx_share, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_leaf_tx_max_set(struct devlink_rate *rate_leaf, void *priv,
u64 tx_max, struct netlink_ext_ack *extack)
{
struct mlx5_vport *vport = priv;
struct mlx5_eswitch *esw;
int err;
esw = vport->dev->priv.eswitch;
if (!mlx5_esw_allowed(esw))
return -EPERM;
err = esw_qos_devlink_rate_to_mbps(vport->dev, "tx_max", &tx_max, extack);
if (err)
return err;
mutex_lock(&esw->state_lock);
err = mlx5_esw_qos_set_vport_max_rate(esw, vport, tx_max, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_node_tx_share_set(struct devlink_rate *rate_node, void *priv,
u64 tx_share, struct netlink_ext_ack *extack)
{
struct mlx5_core_dev *dev = devlink_priv(rate_node->devlink);
struct mlx5_eswitch *esw = dev->priv.eswitch;
struct mlx5_esw_rate_group *group = priv;
int err;
err = esw_qos_devlink_rate_to_mbps(dev, "tx_share", &tx_share, extack);
if (err)
return err;
mutex_lock(&esw->state_lock);
err = esw_qos_set_group_min_rate(esw, group, tx_share, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_node_tx_max_set(struct devlink_rate *rate_node, void *priv,
u64 tx_max, struct netlink_ext_ack *extack)
{
struct mlx5_core_dev *dev = devlink_priv(rate_node->devlink);
struct mlx5_eswitch *esw = dev->priv.eswitch;
struct mlx5_esw_rate_group *group = priv;
int err;
err = esw_qos_devlink_rate_to_mbps(dev, "tx_max", &tx_max, extack);
if (err)
return err;
mutex_lock(&esw->state_lock);
err = esw_qos_set_group_max_rate(esw, group, tx_max, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_node_new(struct devlink_rate *rate_node, void **priv,
struct netlink_ext_ack *extack)
{
struct mlx5_esw_rate_group *group;
struct mlx5_eswitch *esw;
int err = 0;
esw = mlx5_devlink_eswitch_get(rate_node->devlink);
if (IS_ERR(esw))
return PTR_ERR(esw);
mutex_lock(&esw->state_lock);
if (esw->mode != MLX5_ESWITCH_OFFLOADS) {
NL_SET_ERR_MSG_MOD(extack,
"Rate node creation supported only in switchdev mode");
err = -EOPNOTSUPP;
goto unlock;
}
group = esw_qos_create_rate_group(esw, extack);
if (IS_ERR(group)) {
err = PTR_ERR(group);
goto unlock;
}
*priv = group;
unlock:
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_node_del(struct devlink_rate *rate_node, void *priv,
struct netlink_ext_ack *extack)
{
struct mlx5_esw_rate_group *group = priv;
struct mlx5_eswitch *esw;
int err;
esw = mlx5_devlink_eswitch_get(rate_node->devlink);
if (IS_ERR(esw))
return PTR_ERR(esw);
mutex_lock(&esw->state_lock);
err = esw_qos_destroy_rate_group(esw, group, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_qos_vport_update_group(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
struct mlx5_esw_rate_group *group,
struct netlink_ext_ack *extack)
{
int err;
mutex_lock(&esw->state_lock);
err = esw_qos_vport_update_group(esw, vport, group, extack);
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_esw_devlink_rate_parent_set(struct devlink_rate *devlink_rate,
struct devlink_rate *parent,
void *priv, void *parent_priv,
struct netlink_ext_ack *extack)
{
struct mlx5_esw_rate_group *group;
struct mlx5_vport *vport = priv;
if (!parent)
return mlx5_esw_qos_vport_update_group(vport->dev->priv.eswitch,
vport, NULL, extack);
group = parent_priv;
return mlx5_esw_qos_vport_update_group(vport->dev->priv.eswitch, vport, group, extack);
}

41
drivers/net/ethernet/mellanox/mlx5/core/esw/qos.h Normal file
Просмотреть файл

@ -0,0 +1,41 @@
/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/* Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved. */
#ifndef __MLX5_ESW_QOS_H__
#define __MLX5_ESW_QOS_H__
#ifdef CONFIG_MLX5_ESWITCH
int mlx5_esw_qos_set_vport_min_rate(struct mlx5_eswitch *esw,
struct mlx5_vport *evport,
u32 min_rate,
struct netlink_ext_ack *extack);
int mlx5_esw_qos_set_vport_max_rate(struct mlx5_eswitch *esw,
struct mlx5_vport *evport,
u32 max_rate,
struct netlink_ext_ack *extack);
void mlx5_esw_qos_create(struct mlx5_eswitch *esw);
void mlx5_esw_qos_destroy(struct mlx5_eswitch *esw);
int mlx5_esw_qos_vport_enable(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
u32 max_rate, u32 bw_share);
void mlx5_esw_qos_vport_disable(struct mlx5_eswitch *esw, struct mlx5_vport *vport);
int mlx5_esw_devlink_rate_leaf_tx_share_set(struct devlink_rate *rate_leaf, void *priv,
u64 tx_share, struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_leaf_tx_max_set(struct devlink_rate *rate_leaf, void *priv,
u64 tx_max, struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_node_tx_share_set(struct devlink_rate *rate_node, void *priv,
u64 tx_share, struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_node_tx_max_set(struct devlink_rate *rate_node, void *priv,
u64 tx_max, struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_node_new(struct devlink_rate *rate_node, void **priv,
struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_node_del(struct devlink_rate *rate_node, void *priv,
struct netlink_ext_ack *extack);
int mlx5_esw_devlink_rate_parent_set(struct devlink_rate *devlink_rate,
struct devlink_rate *parent,
void *priv, void *parent_priv,
struct netlink_ext_ack *extack);
#endif
#endif

42
drivers/net/ethernet/mellanox/mlx5/core/esw/sample.h
Просмотреть файл

@ -1,42 +0,0 @@
/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/* Copyright (c) 2021 Mellanox Technologies. */
#ifndef __MLX5_EN_TC_SAMPLE_H__
#define __MLX5_EN_TC_SAMPLE_H__
#include "en.h"
#include "eswitch.h"
struct mlx5e_priv;
struct mlx5_flow_attr;
struct mlx5_esw_psample;
struct mlx5_sample_attr {
u32 group_num;
u32 rate;
u32 trunc_size;
u32 restore_obj_id;
u32 sampler_id;
struct mlx5_flow_table *sample_default_tbl;
struct mlx5_sample_flow *sample_flow;
};
void mlx5_esw_sample_skb(struct sk_buff *skb, struct mlx5_mapped_obj *mapped_obj);
struct mlx5_flow_handle *
mlx5_esw_sample_offload(struct mlx5_esw_psample *sample_priv,
struct mlx5_flow_spec *spec,
struct mlx5_flow_attr *attr);
void
mlx5_esw_sample_unoffload(struct mlx5_esw_psample *sample_priv,
struct mlx5_flow_handle *rule,
struct mlx5_flow_attr *attr);
struct mlx5_esw_psample *
mlx5_esw_sample_init(struct mlx5e_priv *priv);
void
mlx5_esw_sample_cleanup(struct mlx5_esw_psample *esw_psample);
#endif /* __MLX5_EN_TC_SAMPLE_H__ */

310
drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
Просмотреть файл

@ -38,6 +38,7 @@
#include <linux/mlx5/mpfs.h>
#include "esw/acl/lgcy.h"
#include "esw/legacy.h"
#include "esw/qos.h"
#include "mlx5_core.h"
#include "lib/eq.h"
#include "eswitch.h"
@ -740,201 +741,6 @@ static void esw_vport_change_handler(struct work_struct *work)
mutex_unlock(&esw->state_lock);
}
static bool element_type_supported(struct mlx5_eswitch *esw, int type)
{
const struct mlx5_core_dev *dev = esw->dev;
switch (type) {
case SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_TASR;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_VPORT;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT_TC:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_VPORT_TC;
case SCHEDULING_CONTEXT_ELEMENT_TYPE_PARA_VPORT_TC:
return MLX5_CAP_QOS(dev, esw_element_type) &
ELEMENT_TYPE_CAP_MASK_PARA_VPORT_TC;
}
return false;
}
/* Vport QoS management */
static void esw_create_tsar(struct mlx5_eswitch *esw)
{
u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
struct mlx5_core_dev *dev = esw->dev;
__be32 *attr;
int err;
if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
return;
if (!element_type_supported(esw, SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR))
return;
if (esw->qos.enabled)
return;
MLX5_SET(scheduling_context, tsar_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR);
attr = MLX5_ADDR_OF(scheduling_context, tsar_ctx, element_attributes);
*attr = cpu_to_be32(TSAR_ELEMENT_TSAR_TYPE_DWRR << 16);
err = mlx5_create_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
tsar_ctx,
&esw->qos.root_tsar_id);
if (err) {
esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
return;
}
esw->qos.enabled = true;
}
static void esw_destroy_tsar(struct mlx5_eswitch *esw)
{
int err;
if (!esw->qos.enabled)
return;
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
esw->qos.root_tsar_id);
if (err)
esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
esw->qos.enabled = false;
}
static int esw_vport_enable_qos(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
u32 initial_max_rate, u32 initial_bw_share)
{
u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
struct mlx5_core_dev *dev = esw->dev;
void *vport_elem;
int err = 0;
if (!esw->qos.enabled)
return 0;
if (vport->qos.enabled)
return -EEXIST;
MLX5_SET(scheduling_context, sched_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
element_attributes);
MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
esw->qos.root_tsar_id);
MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
initial_max_rate);
MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
err = mlx5_create_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
sched_ctx,
&vport->qos.esw_tsar_ix);
if (err) {
esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
return err;
}
vport->qos.enabled = true;
return 0;
}
static void esw_vport_disable_qos(struct mlx5_eswitch *esw,
struct mlx5_vport *vport)
{
int err;
if (!vport->qos.enabled)
return;
err = mlx5_destroy_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
vport->qos.esw_tsar_ix);
if (err)
esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
vport->qos.enabled = false;
}
static int esw_vport_qos_config(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
u32 max_rate, u32 bw_share)
{
u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
struct mlx5_core_dev *dev = esw->dev;
void *vport_elem;
u32 bitmask = 0;
int err = 0;
if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
return -EOPNOTSUPP;
if (!vport->qos.enabled)
return -EIO;
MLX5_SET(scheduling_context, sched_ctx, element_type,
SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
element_attributes);
MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
esw->qos.root_tsar_id);
MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
max_rate);
MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
err = mlx5_modify_scheduling_element_cmd(dev,
SCHEDULING_HIERARCHY_E_SWITCH,
sched_ctx,
vport->qos.esw_tsar_ix,
bitmask);
if (err) {
esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
vport->vport, err);
return err;
}
return 0;
}
int mlx5_esw_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num,
u32 rate_mbps)
{
u32 ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
struct mlx5_vport *vport;
vport = mlx5_eswitch_get_vport(esw, vport_num);
if (IS_ERR(vport))
return PTR_ERR(vport);
if (!vport->qos.enabled)
return -EOPNOTSUPP;
MLX5_SET(scheduling_context, ctx, max_average_bw, rate_mbps);
return mlx5_modify_scheduling_element_cmd(esw->dev,
SCHEDULING_HIERARCHY_E_SWITCH,
ctx,
vport->qos.esw_tsar_ix,
MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW);
}
static void node_guid_gen_from_mac(u64 *node_guid, const u8 *mac)
{
((u8 *)node_guid)[7] = mac[0];
@ -976,7 +782,7 @@ static int esw_vport_setup(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
return err;
/* Attach vport to the eswitch rate limiter */
esw_vport_enable_qos(esw, vport, vport->qos.max_rate, vport->qos.bw_share);
mlx5_esw_qos_vport_enable(esw, vport, vport->qos.max_rate, vport->qos.bw_share);
if (mlx5_esw_is_manager_vport(esw, vport_num))
return 0;
@ -1013,7 +819,7 @@ static void esw_vport_cleanup(struct mlx5_eswitch *esw, struct mlx5_vport *vport
vport_num, 1,
MLX5_VPORT_ADMIN_STATE_DOWN);
esw_vport_disable_qos(esw, vport);
mlx5_esw_qos_vport_disable(esw, vport);
esw_vport_cleanup_acl(esw, vport);
}
@ -1454,7 +1260,7 @@ int mlx5_eswitch_enable_locked(struct mlx5_eswitch *esw, int mode, int num_vfs)
mlx5_eswitch_update_num_of_vfs(esw, num_vfs);
esw_create_tsar(esw);
mlx5_esw_qos_create(esw);
esw->mode = mode;
@ -1484,7 +1290,7 @@ abort:
if (mode == MLX5_ESWITCH_OFFLOADS)
mlx5_rescan_drivers(esw->dev);
esw_destroy_tsar(esw);
mlx5_esw_qos_destroy(esw);
mlx5_esw_acls_ns_cleanup(esw);
return err;
}
@ -1553,7 +1359,7 @@ void mlx5_eswitch_disable_locked(struct mlx5_eswitch *esw, bool clear_vf)
if (old_mode == MLX5_ESWITCH_OFFLOADS)
mlx5_rescan_drivers(esw->dev);
esw_destroy_tsar(esw);
mlx5_esw_qos_destroy(esw);
mlx5_esw_acls_ns_cleanup(esw);
if (clear_vf)
@ -2050,110 +1856,6 @@ int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
return err;
}
static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
struct mlx5_vport *evport;
u32 max_guarantee = 0;
unsigned long i;
mlx5_esw_for_each_vport(esw, i, evport) {
if (!evport->enabled || evport->qos.min_rate < max_guarantee)
continue;
max_guarantee = evport->qos.min_rate;
}
if (max_guarantee)
return max_t(u32, max_guarantee / fw_max_bw_share, 1);
return 0;
}
static int normalize_vports_min_rate(struct mlx5_eswitch *esw)
{
u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
u32 divider = calculate_vports_min_rate_divider(esw);
struct mlx5_vport *evport;
u32 vport_max_rate;
u32 vport_min_rate;
unsigned long i;
u32 bw_share;
int err;
mlx5_esw_for_each_vport(esw, i, evport) {
if (!evport->enabled)
continue;
vport_min_rate = evport->qos.min_rate;
vport_max_rate = evport->qos.max_rate;
bw_share = 0;
if (divider)
bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
divider,
fw_max_bw_share);
if (bw_share == evport->qos.bw_share)
continue;
err = esw_vport_qos_config(esw, evport, vport_max_rate,
bw_share);
if (!err)
evport->qos.bw_share = bw_share;
else
return err;
}
return 0;
}
int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, u16 vport,
u32 max_rate, u32 min_rate)
{
struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
u32 fw_max_bw_share;
u32 previous_min_rate;
bool min_rate_supported;
bool max_rate_supported;
int err = 0;
if (!mlx5_esw_allowed(esw))
return -EPERM;
if (IS_ERR(evport))
return PTR_ERR(evport);
fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
fw_max_bw_share >= MLX5_MIN_BW_SHARE;
max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
return -EOPNOTSUPP;
mutex_lock(&esw->state_lock);
if (min_rate == evport->qos.min_rate)
goto set_max_rate;
previous_min_rate = evport->qos.min_rate;
evport->qos.min_rate = min_rate;
err = normalize_vports_min_rate(esw);
if (err) {
evport->qos.min_rate = previous_min_rate;
goto unlock;
}
set_max_rate:
if (max_rate == evport->qos.max_rate)
goto unlock;
err = esw_vport_qos_config(esw, evport, max_rate, evport->qos.bw_share);
if (!err)
evport->qos.max_rate = max_rate;
unlock:
mutex_unlock(&esw->state_lock);
return err;
}
int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
u16 vport_num,
struct ifla_vf_stats *vf_stats)

21
drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
Просмотреть файл

@ -46,7 +46,7 @@
#include "lib/fs_chains.h"
#include "sf/sf.h"
#include "en/tc_ct.h"
#include "esw/sample.h"
#include "en/tc/sample.h"
enum mlx5_mapped_obj_type {
MLX5_MAPPED_OBJ_CHAIN,
@ -61,6 +61,7 @@ struct mlx5_mapped_obj {
u32 group_id;
u32 rate;
u32 trunc_size;
u32 tunnel_id;
} sample;
};
};
@ -75,11 +76,6 @@ struct mlx5_mapped_obj {
#define MLX5_MAX_MC_PER_VPORT(dev) \
(1 << MLX5_CAP_GEN(dev, log_max_current_mc_list))
#define MLX5_MIN_BW_SHARE 1
#define MLX5_RATE_TO_BW_SHARE(rate, divider, limit) \
min_t(u32, max_t(u32, (rate) / (divider), MLX5_MIN_BW_SHARE), limit)
#define mlx5_esw_has_fwd_fdb(dev) \
MLX5_CAP_ESW_FLOWTABLE(dev, fdb_multi_path_to_table)
@ -181,6 +177,7 @@ struct mlx5_vport {
u32 bw_share;
u32 min_rate;
u32 max_rate;
struct mlx5_esw_rate_group *group;
} qos;
u16 vport;
@ -309,7 +306,9 @@ struct mlx5_eswitch {
struct {
bool enabled;
u32 root_tsar_id;
u32 root_tsar_ix;
struct mlx5_esw_rate_group *group0;
struct list_head groups; /* Protected by esw->state_lock */
} qos;
struct mlx5_esw_bridge_offloads *br_offloads;
@ -335,8 +334,7 @@ int mlx5_esw_offloads_vport_metadata_set(struct mlx5_eswitch *esw, bool enable);
u32 mlx5_esw_match_metadata_alloc(struct mlx5_eswitch *esw);
void mlx5_esw_match_metadata_free(struct mlx5_eswitch *esw, u32 metadata);
int mlx5_esw_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num,
u32 rate_mbps);
int mlx5_esw_qos_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num, u32 rate_mbps);
/* E-Switch API */
int mlx5_eswitch_init(struct mlx5_core_dev *dev);
@ -359,6 +357,10 @@ int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
u16 vport_num, bool setting);
int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, u16 vport,
u32 max_rate, u32 min_rate);
int mlx5_esw_qos_vport_update_group(struct mlx5_eswitch *esw,
struct mlx5_vport *vport,
struct mlx5_esw_rate_group *group,
struct netlink_ext_ack *extack);
int mlx5_eswitch_set_vepa(struct mlx5_eswitch *esw, u8 setting);
int mlx5_eswitch_get_vepa(struct mlx5_eswitch *esw, u8 *setting);
int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
@ -469,7 +471,6 @@ struct mlx5_esw_flow_attr {
} dests[MLX5_MAX_FLOW_FWD_VPORTS];
struct mlx5_rx_tun_attr *rx_tun_attr;
struct mlx5_pkt_reformat *decap_pkt_reformat;
struct mlx5_sample_attr *sample;
};
int mlx5_devlink_eswitch_mode_set(struct devlink *devlink, u16 mode,

11
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
Просмотреть файл

@ -187,12 +187,12 @@ esw_cleanup_decap_indir(struct mlx5_eswitch *esw,
static int
esw_setup_sampler_dest(struct mlx5_flow_destination *dest,
struct mlx5_flow_act *flow_act,
struct mlx5_esw_flow_attr *esw_attr,
struct mlx5_flow_attr *attr,
int i)
{
flow_act->flags |= FLOW_ACT_IGNORE_FLOW_LEVEL;
dest[i].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_SAMPLER;
dest[i].sampler_id = esw_attr->sample->sampler_id;
dest[i].sampler_id = attr->sample_attr->sampler_id;
return 0;
}
@ -435,7 +435,7 @@ esw_setup_dests(struct mlx5_flow_destination *dest,
attr->flags |= MLX5_ESW_ATTR_FLAG_SRC_REWRITE;
if (attr->flags & MLX5_ESW_ATTR_FLAG_SAMPLE) {
esw_setup_sampler_dest(dest, flow_act, esw_attr, *i);
esw_setup_sampler_dest(dest, flow_act, attr, *i);
(*i)++;
} else if (attr->dest_ft) {
esw_setup_ft_dest(dest, flow_act, esw, attr, spec, *i);
@ -540,10 +540,7 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
flow_act.modify_hdr = attr->modify_hdr;
/* esw_attr->sample is allocated only when there is a sample action */
if (esw_attr->sample && esw_attr->sample->sample_default_tbl) {
fdb = esw_attr->sample->sample_default_tbl;
} else if (split) {
if (split) {
fwd_attr.chain = attr->chain;
fwd_attr.prio = attr->prio;
fwd_attr.vport = esw_attr->in_rep->vport;

3
include/linux/mlx5/mlx5_ifc.h
Просмотреть файл

@ -865,7 +865,8 @@ struct mlx5_ifc_qos_cap_bits {
u8 nic_bw_share[0x1];
u8 nic_rate_limit[0x1];
u8 packet_pacing_uid[0x1];
u8 reserved_at_c[0x14];
u8 log_esw_max_sched_depth[0x4];
u8 reserved_at_10[0x10];
u8 reserved_at_20[0xb];
u8 log_max_qos_nic_queue_group[0x5];