kernel: add hardware offload patch for flow tables support

Supports offloading through VLAN, bridge and PPPoE devices as well

Signed-off-by: Felix Fietkau <nbd@nbd.name>
This commit is contained in:
Felix Fietkau 2018-03-12 20:51:55 +01:00
parent f8dbafb234
commit 11a3d27043
7 changed files with 1143 additions and 2 deletions

View file

@ -149,8 +149,9 @@ define KernelPackage/nf-flow
CONFIG_NF_FLOW_TABLE_HW
DEPENDS:=+kmod-nf-conntrack @!LINUX_3_18 @!LINUX_4_4 @!LINUX_4_9
FILES:= \
$(LINUX_DIR)/net/netfilter/nf_flow_table.ko
AUTOLOAD:=$(call AutoProbe,nf_flow_table)
$(LINUX_DIR)/net/netfilter/nf_flow_table.ko \
$(LINUX_DIR)/net/netfilter/nf_flow_table_hw.ko
AUTOLOAD:=$(call AutoProbe,nf_flow_table nf_flow_table_hw)
endef
$(eval $(call KernelPackage,nf-flow))

View file

@ -0,0 +1,580 @@
From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: Thu, 11 Jan 2018 16:32:00 +0100
Subject: [PATCH] netfilter: nf_flow_table: add hardware offload support
This patch adds the infrastructure to offload flows to hardware, in case
the nic/switch comes with built-in flow tables capabilities.
If the hardware comes with no hardware flow tables or they have
limitations in terms of features, the existing infrastructure falls back
to the software flow table implementation.
The software flow table garbage collector skips entries that resides in
the hardware, so the hardware will be responsible for releasing this
flow table entry too via flow_offload_dead().
Hardware configuration, either to add or to delete entries, is done from
the hardware offload workqueue, to ensure this is done from user context
given that we may sleep when grabbing the mdio mutex.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
create mode 100644 net/netfilter/nf_flow_table_hw.c
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -826,6 +826,13 @@ struct xfrmdev_ops {
};
#endif
+struct flow_offload;
+
+enum flow_offload_type {
+ FLOW_OFFLOAD_ADD = 0,
+ FLOW_OFFLOAD_DEL,
+};
+
/*
* This structure defines the management hooks for network devices.
* The following hooks can be defined; unless noted otherwise, they are
@@ -1057,6 +1064,10 @@ struct xfrmdev_ops {
* int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
* u16 flags);
*
+ * int (*ndo_flow_offload)(enum flow_offload_type type,
+ * struct flow_offload *flow);
+ * Adds/deletes flow entry to/from net device flowtable.
+ *
* int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
* Called to change device carrier. Soft-devices (like dummy, team, etc)
* which do not represent real hardware may define this to allow their
@@ -1281,6 +1292,8 @@ struct net_device_ops {
int (*ndo_bridge_dellink)(struct net_device *dev,
struct nlmsghdr *nlh,
u16 flags);
+ int (*ndo_flow_offload)(enum flow_offload_type type,
+ struct flow_offload *flow);
int (*ndo_change_carrier)(struct net_device *dev,
bool new_carrier);
int (*ndo_get_phys_port_id)(struct net_device *dev,
--- a/include/net/netfilter/nf_flow_table.h
+++ b/include/net/netfilter/nf_flow_table.h
@@ -20,11 +20,17 @@ struct nf_flowtable_type {
struct module *owner;
};
+enum nf_flowtable_flags {
+ NF_FLOWTABLE_F_HW = 0x1,
+};
+
struct nf_flowtable {
struct list_head list;
struct rhashtable rhashtable;
const struct nf_flowtable_type *type;
+ u32 flags;
struct delayed_work gc_work;
+ possible_net_t ft_net;
};
enum flow_offload_tuple_dir {
@@ -69,6 +75,7 @@ struct flow_offload_tuple_rhash {
#define FLOW_OFFLOAD_DNAT 0x2
#define FLOW_OFFLOAD_DYING 0x4
#define FLOW_OFFLOAD_TEARDOWN 0x8
+#define FLOW_OFFLOAD_HW 0x10
struct flow_offload {
struct flow_offload_tuple_rhash tuplehash[FLOW_OFFLOAD_DIR_MAX];
@@ -126,6 +133,22 @@ unsigned int nf_flow_offload_ip_hook(voi
unsigned int nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
const struct nf_hook_state *state);
+void nf_flow_offload_hw_add(struct net *net, struct flow_offload *flow,
+ struct nf_conn *ct);
+void nf_flow_offload_hw_del(struct net *net, struct flow_offload *flow);
+
+struct nf_flow_table_hw {
+ struct module *owner;
+ void (*add)(struct net *net, struct flow_offload *flow,
+ struct nf_conn *ct);
+ void (*del)(struct net *net, struct flow_offload *flow);
+};
+
+int nf_flow_table_hw_register(const struct nf_flow_table_hw *offload);
+void nf_flow_table_hw_unregister(const struct nf_flow_table_hw *offload);
+
+extern struct work_struct nf_flow_offload_hw_work;
+
#define MODULE_ALIAS_NF_FLOWTABLE(family) \
MODULE_ALIAS("nf-flowtable-" __stringify(family))
--- a/include/uapi/linux/netfilter/nf_tables.h
+++ b/include/uapi/linux/netfilter/nf_tables.h
@@ -1341,6 +1341,7 @@ enum nft_object_attributes {
* @NFTA_FLOWTABLE_HOOK: netfilter hook configuration(NLA_U32)
* @NFTA_FLOWTABLE_USE: number of references to this flow table (NLA_U32)
* @NFTA_FLOWTABLE_HANDLE: object handle (NLA_U64)
+ * @NFTA_FLOWTABLE_FLAGS: flags (NLA_U32)
*/
enum nft_flowtable_attributes {
NFTA_FLOWTABLE_UNSPEC,
@@ -1350,6 +1351,7 @@ enum nft_flowtable_attributes {
NFTA_FLOWTABLE_USE,
NFTA_FLOWTABLE_HANDLE,
NFTA_FLOWTABLE_PAD,
+ NFTA_FLOWTABLE_FLAGS,
__NFTA_FLOWTABLE_MAX
};
#define NFTA_FLOWTABLE_MAX (__NFTA_FLOWTABLE_MAX - 1)
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -686,6 +686,15 @@ config NF_FLOW_TABLE
To compile it as a module, choose M here.
+config NF_FLOW_TABLE_HW
+ tristate "Netfilter flow table hardware offload module"
+ depends on NF_FLOW_TABLE
+ help
+ This option adds hardware offload support for the flow table core
+ infrastructure.
+
+ To compile it as a module, choose M here.
+
config NETFILTER_XTABLES
tristate "Netfilter Xtables support (required for ip_tables)"
default m if NETFILTER_ADVANCED=n
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -116,6 +116,7 @@ obj-$(CONFIG_NF_FLOW_TABLE) += nf_flow_t
nf_flow_table-objs := nf_flow_table_core.o nf_flow_table_ip.o
obj-$(CONFIG_NF_FLOW_TABLE_INET) += nf_flow_table_inet.o
+obj-$(CONFIG_NF_FLOW_TABLE_HW) += nf_flow_table_hw.o
# generic X tables
obj-$(CONFIG_NETFILTER_XTABLES) += x_tables.o xt_tcpudp.o
--- a/net/netfilter/nf_flow_table_core.c
+++ b/net/netfilter/nf_flow_table_core.c
@@ -199,10 +199,16 @@ int flow_offload_add(struct nf_flowtable
}
EXPORT_SYMBOL_GPL(flow_offload_add);
+static inline bool nf_flow_in_hw(const struct flow_offload *flow)
+{
+ return flow->flags & FLOW_OFFLOAD_HW;
+}
+
static void flow_offload_del(struct nf_flowtable *flow_table,
struct flow_offload *flow)
{
struct flow_offload_entry *e;
+ struct net *net = read_pnet(&flow_table->ft_net);
rhashtable_remove_fast(&flow_table->rhashtable,
&flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].node,
@@ -214,6 +220,9 @@ static void flow_offload_del(struct nf_f
e = container_of(flow, struct flow_offload_entry, flow);
clear_bit(IPS_OFFLOAD_BIT, &e->ct->status);
+ if (nf_flow_in_hw(flow))
+ nf_flow_offload_hw_del(net, flow);
+
flow_offload_free(flow);
}
@@ -307,6 +316,7 @@ static int nf_flow_offload_gc_step(struc
rhashtable_walk_start(&hti);
while ((tuplehash = rhashtable_walk_next(&hti))) {
+ bool teardown;
if (IS_ERR(tuplehash)) {
err = PTR_ERR(tuplehash);
if (err != -EAGAIN)
@@ -319,9 +329,13 @@ static int nf_flow_offload_gc_step(struc
flow = container_of(tuplehash, struct flow_offload, tuplehash[0]);
- if (nf_flow_has_expired(flow) ||
- (flow->flags & (FLOW_OFFLOAD_DYING |
- FLOW_OFFLOAD_TEARDOWN)))
+ teardown = flow->flags & (FLOW_OFFLOAD_DYING |
+ FLOW_OFFLOAD_TEARDOWN);
+
+ if (nf_flow_in_hw(flow) && !teardown)
+ continue;
+
+ if (nf_flow_has_expired(flow) || teardown)
flow_offload_del(flow_table, flow);
}
out:
@@ -456,10 +470,43 @@ int nf_flow_dnat_port(const struct flow_
}
EXPORT_SYMBOL_GPL(nf_flow_dnat_port);
+static const struct nf_flow_table_hw __rcu *nf_flow_table_hw_hook __read_mostly;
+
+static int nf_flow_offload_hw_init(struct nf_flowtable *flow_table)
+{
+ const struct nf_flow_table_hw *offload;
+
+ if (!rcu_access_pointer(nf_flow_table_hw_hook))
+ request_module("nf-flow-table-hw");
+
+ rcu_read_lock();
+ offload = rcu_dereference(nf_flow_table_hw_hook);
+ if (!offload)
+ goto err_no_hw_offload;
+
+ if (!try_module_get(offload->owner))
+ goto err_no_hw_offload;
+
+ rcu_read_unlock();
+
+ return 0;
+
+err_no_hw_offload:
+ rcu_read_unlock();
+
+ return -EOPNOTSUPP;
+}
+
int nf_flow_table_init(struct nf_flowtable *flowtable)
{
int err;
+ if (flowtable->flags & NF_FLOWTABLE_F_HW) {
+ err = nf_flow_offload_hw_init(flowtable);
+ if (err)
+ return err;
+ }
+
INIT_DEFERRABLE_WORK(&flowtable->gc_work, nf_flow_offload_work_gc);
err = rhashtable_init(&flowtable->rhashtable,
@@ -497,6 +544,8 @@ static void nf_flow_table_iterate_cleanu
{
nf_flow_table_iterate(flowtable, nf_flow_table_do_cleanup, dev);
flush_delayed_work(&flowtable->gc_work);
+ if (flowtable->flags & NF_FLOWTABLE_F_HW)
+ flush_work(&nf_flow_offload_hw_work);
}
void nf_flow_table_cleanup(struct net *net, struct net_device *dev)
@@ -510,6 +559,26 @@ void nf_flow_table_cleanup(struct net *n
}
EXPORT_SYMBOL_GPL(nf_flow_table_cleanup);
+struct work_struct nf_flow_offload_hw_work;
+EXPORT_SYMBOL_GPL(nf_flow_offload_hw_work);
+
+/* Give the hardware workqueue the chance to remove entries from hardware.*/
+static void nf_flow_offload_hw_free(struct nf_flowtable *flowtable)
+{
+ const struct nf_flow_table_hw *offload;
+
+ flush_work(&nf_flow_offload_hw_work);
+
+ rcu_read_lock();
+ offload = rcu_dereference(nf_flow_table_hw_hook);
+ if (!offload) {
+ rcu_read_unlock();
+ return;
+ }
+ module_put(offload->owner);
+ rcu_read_unlock();
+}
+
void nf_flow_table_free(struct nf_flowtable *flow_table)
{
mutex_lock(&flowtable_lock);
@@ -519,9 +588,58 @@ void nf_flow_table_free(struct nf_flowta
nf_flow_table_iterate(flow_table, nf_flow_table_do_cleanup, NULL);
WARN_ON(!nf_flow_offload_gc_step(flow_table));
rhashtable_destroy(&flow_table->rhashtable);
+ if (flow_table->flags & NF_FLOWTABLE_F_HW)
+ nf_flow_offload_hw_free(flow_table);
}
EXPORT_SYMBOL_GPL(nf_flow_table_free);
+/* Must be called from user context. */
+void nf_flow_offload_hw_add(struct net *net, struct flow_offload *flow,
+ struct nf_conn *ct)
+{
+ const struct nf_flow_table_hw *offload;
+
+ rcu_read_lock();
+ offload = rcu_dereference(nf_flow_table_hw_hook);
+ if (offload)
+ offload->add(net, flow, ct);
+ rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(nf_flow_offload_hw_add);
+
+/* Must be called from user context. */
+void nf_flow_offload_hw_del(struct net *net, struct flow_offload *flow)
+{
+ const struct nf_flow_table_hw *offload;
+
+ rcu_read_lock();
+ offload = rcu_dereference(nf_flow_table_hw_hook);
+ if (offload)
+ offload->del(net, flow);
+ rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(nf_flow_offload_hw_del);
+
+int nf_flow_table_hw_register(const struct nf_flow_table_hw *offload)
+{
+ if (rcu_access_pointer(nf_flow_table_hw_hook))
+ return -EBUSY;
+
+ rcu_assign_pointer(nf_flow_table_hw_hook, offload);
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(nf_flow_table_hw_register);
+
+void nf_flow_table_hw_unregister(const struct nf_flow_table_hw *offload)
+{
+ WARN_ON(rcu_access_pointer(nf_flow_table_hw_hook) != offload);
+ rcu_assign_pointer(nf_flow_table_hw_hook, NULL);
+
+ synchronize_rcu();
+}
+EXPORT_SYMBOL_GPL(nf_flow_table_hw_unregister);
+
static int nf_flow_table_netdev_event(struct notifier_block *this,
unsigned long event, void *ptr)
{
--- /dev/null
+++ b/net/netfilter/nf_flow_table_hw.c
@@ -0,0 +1,169 @@
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/netfilter.h>
+#include <linux/rhashtable.h>
+#include <linux/netdevice.h>
+#include <net/netfilter/nf_flow_table.h>
+#include <net/netfilter/nf_conntrack.h>
+#include <net/netfilter/nf_conntrack_core.h>
+#include <net/netfilter/nf_conntrack_tuple.h>
+
+static DEFINE_SPINLOCK(flow_offload_hw_pending_list_lock);
+static LIST_HEAD(flow_offload_hw_pending_list);
+
+static DEFINE_MUTEX(nf_flow_offload_hw_mutex);
+
+struct flow_offload_hw {
+ struct list_head list;
+ enum flow_offload_type type;
+ struct flow_offload *flow;
+ struct nf_conn *ct;
+ possible_net_t flow_hw_net;
+};
+
+static int do_flow_offload_hw(struct net *net, struct flow_offload *flow,
+ int type)
+{
+ struct net_device *indev;
+ int ret, ifindex;
+
+ ifindex = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.iifidx;
+ indev = dev_get_by_index(net, ifindex);
+ if (WARN_ON(!indev))
+ return 0;
+
+ mutex_lock(&nf_flow_offload_hw_mutex);
+ ret = indev->netdev_ops->ndo_flow_offload(type, flow);
+ mutex_unlock(&nf_flow_offload_hw_mutex);
+
+ dev_put(indev);
+
+ return ret;
+}
+
+static void flow_offload_hw_work_add(struct flow_offload_hw *offload)
+{
+ struct net *net;
+ int ret;
+
+ if (nf_ct_is_dying(offload->ct))
+ return;
+
+ net = read_pnet(&offload->flow_hw_net);
+ ret = do_flow_offload_hw(net, offload->flow, FLOW_OFFLOAD_ADD);
+ if (ret >= 0)
+ offload->flow->flags |= FLOW_OFFLOAD_HW;
+}
+
+static void flow_offload_hw_work_del(struct flow_offload_hw *offload)
+{
+ struct net *net = read_pnet(&offload->flow_hw_net);
+
+ do_flow_offload_hw(net, offload->flow, FLOW_OFFLOAD_DEL);
+}
+
+static void flow_offload_hw_work(struct work_struct *work)
+{
+ struct flow_offload_hw *offload, *next;
+ LIST_HEAD(hw_offload_pending);
+
+ spin_lock_bh(&flow_offload_hw_pending_list_lock);
+ list_replace_init(&flow_offload_hw_pending_list, &hw_offload_pending);
+ spin_unlock_bh(&flow_offload_hw_pending_list_lock);
+
+ list_for_each_entry_safe(offload, next, &hw_offload_pending, list) {
+ switch (offload->type) {
+ case FLOW_OFFLOAD_ADD:
+ flow_offload_hw_work_add(offload);
+ break;
+ case FLOW_OFFLOAD_DEL:
+ flow_offload_hw_work_del(offload);
+ break;
+ }
+ if (offload->ct)
+ nf_conntrack_put(&offload->ct->ct_general);
+ list_del(&offload->list);
+ kfree(offload);
+ }
+}
+
+static void flow_offload_queue_work(struct flow_offload_hw *offload)
+{
+ spin_lock_bh(&flow_offload_hw_pending_list_lock);
+ list_add_tail(&offload->list, &flow_offload_hw_pending_list);
+ spin_unlock_bh(&flow_offload_hw_pending_list_lock);
+
+ schedule_work(&nf_flow_offload_hw_work);
+}
+
+static void flow_offload_hw_add(struct net *net, struct flow_offload *flow,
+ struct nf_conn *ct)
+{
+ struct flow_offload_hw *offload;
+
+ offload = kmalloc(sizeof(struct flow_offload_hw), GFP_ATOMIC);
+ if (!offload)
+ return;
+
+ nf_conntrack_get(&ct->ct_general);
+ offload->type = FLOW_OFFLOAD_ADD;
+ offload->ct = ct;
+ offload->flow = flow;
+ write_pnet(&offload->flow_hw_net, net);
+
+ flow_offload_queue_work(offload);
+}
+
+static void flow_offload_hw_del(struct net *net, struct flow_offload *flow)
+{
+ struct flow_offload_hw *offload;
+
+ offload = kmalloc(sizeof(struct flow_offload_hw), GFP_ATOMIC);
+ if (!offload)
+ return;
+
+ offload->type = FLOW_OFFLOAD_DEL;
+ offload->ct = NULL;
+ offload->flow = flow;
+ write_pnet(&offload->flow_hw_net, net);
+
+ flow_offload_queue_work(offload);
+}
+
+static const struct nf_flow_table_hw flow_offload_hw = {
+ .add = flow_offload_hw_add,
+ .del = flow_offload_hw_del,
+ .owner = THIS_MODULE,
+};
+
+static int __init nf_flow_table_hw_module_init(void)
+{
+ INIT_WORK(&nf_flow_offload_hw_work, flow_offload_hw_work);
+ nf_flow_table_hw_register(&flow_offload_hw);
+
+ return 0;
+}
+
+static void __exit nf_flow_table_hw_module_exit(void)
+{
+ struct flow_offload_hw *offload, *next;
+ LIST_HEAD(hw_offload_pending);
+
+ nf_flow_table_hw_unregister(&flow_offload_hw);
+ cancel_work_sync(&nf_flow_offload_hw_work);
+
+ list_for_each_entry_safe(offload, next, &hw_offload_pending, list) {
+ if (offload->ct)
+ nf_conntrack_put(&offload->ct->ct_general);
+ list_del(&offload->list);
+ kfree(offload);
+ }
+}
+
+module_init(nf_flow_table_hw_module_init);
+module_exit(nf_flow_table_hw_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
+MODULE_ALIAS("nf-flow-table-hw");
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -4866,6 +4866,14 @@ static int nf_tables_flowtable_parse_hoo
if (err < 0)
goto err1;
+ for (i = 0; i < n; i++) {
+ if (flowtable->data.flags & NF_FLOWTABLE_F_HW &&
+ !dev_array[i]->netdev_ops->ndo_flow_offload) {
+ err = -EOPNOTSUPP;
+ goto err1;
+ }
+ }
+
ops = kzalloc(sizeof(struct nf_hook_ops) * n, GFP_KERNEL);
if (!ops) {
err = -ENOMEM;
@@ -4996,10 +5004,19 @@ static int nf_tables_newflowtable(struct
}
flowtable->data.type = type;
+ write_pnet(&flowtable->data.ft_net, net);
+
err = type->init(&flowtable->data);
if (err < 0)
goto err3;
+ if (nla[NFTA_FLOWTABLE_FLAGS]) {
+ flowtable->data.flags =
+ ntohl(nla_get_be32(nla[NFTA_FLOWTABLE_FLAGS]));
+ if (flowtable->data.flags & ~NF_FLOWTABLE_F_HW)
+ goto err4;
+ }
+
err = nf_tables_flowtable_parse_hook(&ctx, nla[NFTA_FLOWTABLE_HOOK],
flowtable);
if (err < 0)
@@ -5097,7 +5114,8 @@ static int nf_tables_fill_flowtable_info
nla_put_string(skb, NFTA_FLOWTABLE_NAME, flowtable->name) ||
nla_put_be32(skb, NFTA_FLOWTABLE_USE, htonl(flowtable->use)) ||
nla_put_be64(skb, NFTA_FLOWTABLE_HANDLE, cpu_to_be64(flowtable->handle),
- NFTA_FLOWTABLE_PAD))
+ NFTA_FLOWTABLE_PAD) ||
+ nla_put_be32(skb, NFTA_FLOWTABLE_FLAGS, htonl(flowtable->data.flags)))
goto nla_put_failure;
nest = nla_nest_start(skb, NFTA_FLOWTABLE_HOOK);
--- a/net/netfilter/nft_flow_offload.c
+++ b/net/netfilter/nft_flow_offload.c
@@ -121,6 +121,9 @@ static void nft_flow_offload_eval(const
if (ret < 0)
goto err_flow_add;
+ if (flowtable->flags & NF_FLOWTABLE_F_HW)
+ nf_flow_offload_hw_add(nft_net(pkt), flow, ct);
+
return;
err_flow_add:

View file

@ -0,0 +1,302 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Thu, 15 Mar 2018 20:46:31 +0100
Subject: [PATCH] netfilter: nf_flow_table: support hw offload through
virtual interfaces
There are hardware offload devices that support offloading VLANs and
PPPoE devices. Additionally, it is useful to be able to offload packets
routed through bridge interfaces as well.
Add support for finding the path to the offload device through these
virtual interfaces, while collecting useful parameters for the offload
device, like VLAN ID/protocol, PPPoE session and Ethernet MAC address.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -827,6 +827,7 @@ struct xfrmdev_ops {
#endif
struct flow_offload;
+struct flow_offload_hw_path;
enum flow_offload_type {
FLOW_OFFLOAD_ADD = 0,
@@ -1064,8 +1065,15 @@ enum flow_offload_type {
* int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
* u16 flags);
*
+ * int (*ndo_flow_offload_check)(struct flow_offload_hw_path *path);
+ * For virtual devices like bridges, vlan, and pppoe, fill in the
+ * underlying network device that can be used for offloading connections.
+ * Return an error if offloading is not supported.
+ *
* int (*ndo_flow_offload)(enum flow_offload_type type,
- * struct flow_offload *flow);
+ * struct flow_offload *flow,
+ * struct flow_offload_hw_path *src,
+ * struct flow_offload_hw_path *dest);
* Adds/deletes flow entry to/from net device flowtable.
*
* int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
@@ -1292,8 +1300,11 @@ struct net_device_ops {
int (*ndo_bridge_dellink)(struct net_device *dev,
struct nlmsghdr *nlh,
u16 flags);
+ int (*ndo_flow_offload_check)(struct flow_offload_hw_path *path);
int (*ndo_flow_offload)(enum flow_offload_type type,
- struct flow_offload *flow);
+ struct flow_offload *flow,
+ struct flow_offload_hw_path *src,
+ struct flow_offload_hw_path *dest);
int (*ndo_change_carrier)(struct net_device *dev,
bool new_carrier);
int (*ndo_get_phys_port_id)(struct net_device *dev,
--- a/include/net/netfilter/nf_flow_table.h
+++ b/include/net/netfilter/nf_flow_table.h
@@ -86,6 +86,21 @@ struct flow_offload {
};
};
+#define FLOW_OFFLOAD_PATH_ETHERNET BIT(0)
+#define FLOW_OFFLOAD_PATH_VLAN BIT(1)
+#define FLOW_OFFLOAD_PATH_PPPOE BIT(2)
+
+struct flow_offload_hw_path {
+ struct net_device *dev;
+ u32 flags;
+
+ u8 eth_src[ETH_ALEN];
+ u8 eth_dest[ETH_ALEN];
+ u16 vlan_proto;
+ u16 vlan_id;
+ u16 pppoe_sid;
+};
+
#define NF_FLOW_TIMEOUT (30 * HZ)
struct nf_flow_route {
--- a/net/netfilter/nf_flow_table_hw.c
+++ b/net/netfilter/nf_flow_table_hw.c
@@ -19,48 +19,75 @@ struct flow_offload_hw {
enum flow_offload_type type;
struct flow_offload *flow;
struct nf_conn *ct;
- possible_net_t flow_hw_net;
+
+ struct flow_offload_hw_path src;
+ struct flow_offload_hw_path dest;
};
-static int do_flow_offload_hw(struct net *net, struct flow_offload *flow,
- int type)
+static void flow_offload_check_ethernet(struct flow_offload_tuple *tuple,
+ struct flow_offload_hw_path *path)
{
- struct net_device *indev;
- int ret, ifindex;
+ struct net_device *dev = path->dev;
+ struct neighbour *n;
- ifindex = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.iifidx;
- indev = dev_get_by_index(net, ifindex);
- if (WARN_ON(!indev))
- return 0;
+ if (dev->type != ARPHRD_ETHER)
+ return;
- mutex_lock(&nf_flow_offload_hw_mutex);
- ret = indev->netdev_ops->ndo_flow_offload(type, flow);
- mutex_unlock(&nf_flow_offload_hw_mutex);
+ memcpy(path->eth_src, path->dev->dev_addr, ETH_ALEN);
+ n = dst_neigh_lookup(tuple->dst_cache, &tuple->src_v4);
+ if (!n)
+ return;
- dev_put(indev);
+ memcpy(path->eth_dest, n->ha, ETH_ALEN);
+ path->flags |= FLOW_OFFLOAD_PATH_ETHERNET;
+ neigh_release(n);
+}
- return ret;
+static int flow_offload_check_path(struct net *net,
+ struct flow_offload_tuple *tuple,
+ struct flow_offload_hw_path *path)
+{
+ struct net_device *dev;
+
+ dev = dev_get_by_index_rcu(net, tuple->iifidx);
+ if (!dev)
+ return -ENOENT;
+
+ path->dev = dev;
+ flow_offload_check_ethernet(tuple, path);
+
+ if (dev->netdev_ops->ndo_flow_offload_check)
+ return dev->netdev_ops->ndo_flow_offload_check(path);
+
+ return 0;
}
-static void flow_offload_hw_work_add(struct flow_offload_hw *offload)
+static int do_flow_offload_hw(struct flow_offload_hw *offload)
{
- struct net *net;
+ struct net_device *src_dev = offload->src.dev;
+ struct net_device *dest_dev = offload->dest.dev;
int ret;
- if (nf_ct_is_dying(offload->ct))
- return;
+ ret = src_dev->netdev_ops->ndo_flow_offload(offload->type,
+ offload->flow,
+ &offload->src,
+ &offload->dest);
- net = read_pnet(&offload->flow_hw_net);
- ret = do_flow_offload_hw(net, offload->flow, FLOW_OFFLOAD_ADD);
- if (ret >= 0)
- offload->flow->flags |= FLOW_OFFLOAD_HW;
+ /* restore devices in case the driver mangled them */
+ offload->src.dev = src_dev;
+ offload->dest.dev = dest_dev;
+
+ return ret;
}
-static void flow_offload_hw_work_del(struct flow_offload_hw *offload)
+static void flow_offload_hw_free(struct flow_offload_hw *offload)
{
- struct net *net = read_pnet(&offload->flow_hw_net);
-
- do_flow_offload_hw(net, offload->flow, FLOW_OFFLOAD_DEL);
+ dev_put(offload->src.dev);
+ dev_put(offload->dest.dev);
+ if (offload->ct)
+ nf_conntrack_put(&offload->ct->ct_general);
+ list_del(&offload->list);
+ kfree(offload);
}
static void flow_offload_hw_work(struct work_struct *work)
@@ -73,18 +100,22 @@ static void flow_offload_hw_work(struct
spin_unlock_bh(&flow_offload_hw_pending_list_lock);
list_for_each_entry_safe(offload, next, &hw_offload_pending, list) {
+ mutex_lock(&nf_flow_offload_hw_mutex);
switch (offload->type) {
case FLOW_OFFLOAD_ADD:
- flow_offload_hw_work_add(offload);
+ if (nf_ct_is_dying(offload->ct))
+ break;
+
+ if (do_flow_offload_hw(offload) >= 0)
+ offload->flow->flags |= FLOW_OFFLOAD_HW;
break;
case FLOW_OFFLOAD_DEL:
- flow_offload_hw_work_del(offload);
+ do_flow_offload_hw(offload);
break;
}
- if (offload->ct)
- nf_conntrack_put(&offload->ct->ct_general);
- list_del(&offload->list);
- kfree(offload);
+ mutex_unlock(&nf_flow_offload_hw_mutex);
+
+ flow_offload_hw_free(offload);
}
}
@@ -97,20 +128,55 @@ static void flow_offload_queue_work(stru
schedule_work(&nf_flow_offload_hw_work);
}
+static struct flow_offload_hw *
+flow_offload_hw_prepare(struct net *net, struct flow_offload *flow)
+{
+ struct flow_offload_hw_path src = {};
+ struct flow_offload_hw_path dest = {};
+ struct flow_offload_tuple *tuple;
+ struct flow_offload_hw *offload = NULL;
+
+ rcu_read_lock_bh();
+
+ tuple = &flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple;
+ if (flow_offload_check_path(net, tuple, &src))
+ goto out;
+
+ tuple = &flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple;
+ if (flow_offload_check_path(net, tuple, &dest))
+ goto out;
+
+ if (!src.dev->netdev_ops->ndo_flow_offload)
+ goto out;
+
+ offload = kzalloc(sizeof(struct flow_offload_hw), GFP_ATOMIC);
+ if (!offload)
+ goto out;
+
+ dev_hold(src.dev);
+ dev_hold(dest.dev);
+ offload->src = src;
+ offload->dest = dest;
+ offload->flow = flow;
+
+out:
+ rcu_read_unlock_bh();
+
+ return offload;
+}
+
static void flow_offload_hw_add(struct net *net, struct flow_offload *flow,
struct nf_conn *ct)
{
struct flow_offload_hw *offload;
- offload = kmalloc(sizeof(struct flow_offload_hw), GFP_ATOMIC);
+ offload = flow_offload_hw_prepare(net, flow);
if (!offload)
return;
nf_conntrack_get(&ct->ct_general);
offload->type = FLOW_OFFLOAD_ADD;
offload->ct = ct;
- offload->flow = flow;
- write_pnet(&offload->flow_hw_net, net);
flow_offload_queue_work(offload);
}
@@ -119,14 +185,11 @@ static void flow_offload_hw_del(struct n
{
struct flow_offload_hw *offload;
- offload = kmalloc(sizeof(struct flow_offload_hw), GFP_ATOMIC);
+ offload = flow_offload_hw_prepare(net, flow);
if (!offload)
return;
offload->type = FLOW_OFFLOAD_DEL;
- offload->ct = NULL;
- offload->flow = flow;
- write_pnet(&offload->flow_hw_net, net);
flow_offload_queue_work(offload);
}
@@ -153,12 +216,8 @@ static void __exit nf_flow_table_hw_modu
nf_flow_table_hw_unregister(&flow_offload_hw);
cancel_work_sync(&nf_flow_offload_hw_work);
- list_for_each_entry_safe(offload, next, &hw_offload_pending, list) {
- if (offload->ct)
- nf_conntrack_put(&offload->ct->ct_general);
- list_del(&offload->list);
- kfree(offload);
- }
+ list_for_each_entry_safe(offload, next, &hw_offload_pending, list)
+ flow_offload_hw_free(offload);
}
module_init(nf_flow_table_hw_module_init);

View file

@ -0,0 +1,56 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Thu, 15 Mar 2018 20:49:58 +0100
Subject: [PATCH] net: 8021q: support hardware flow table offload
Add the VLAN ID and protocol information
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/8021q/vlan_dev.c
+++ b/net/8021q/vlan_dev.c
@@ -29,8 +29,10 @@
#include <linux/net_tstamp.h>
#include <linux/etherdevice.h>
#include <linux/ethtool.h>
+#include <linux/netfilter.h>
#include <net/arp.h>
#include <net/switchdev.h>
+#include <net/netfilter/nf_flow_table.h>
#include "vlan.h"
#include "vlanproc.h"
@@ -762,6 +764,25 @@ static int vlan_dev_get_iflink(const str
return real_dev->ifindex;
}
+static int vlan_dev_flow_offload_check(struct flow_offload_hw_path *path)
+{
+ struct net_device *dev = path->dev;
+ struct vlan_dev_priv *vlan = vlan_dev_priv(dev);
+
+ if (path->flags & FLOW_OFFLOAD_PATH_VLAN)
+ return -EEXIST;
+
+ path->flags |= FLOW_OFFLOAD_PATH_VLAN;
+ path->vlan_proto = vlan->vlan_proto;
+ path->vlan_id = vlan->vlan_id;
+ path->dev = vlan->real_dev;
+
+ if (vlan->real_dev->netdev_ops->ndo_flow_offload_check)
+ return vlan->real_dev->netdev_ops->ndo_flow_offload_check(path);
+
+ return 0;
+}
+
static const struct ethtool_ops vlan_ethtool_ops = {
.get_link_ksettings = vlan_ethtool_get_link_ksettings,
.get_drvinfo = vlan_ethtool_get_drvinfo,
@@ -799,6 +820,7 @@ static const struct net_device_ops vlan_
.ndo_fix_features = vlan_dev_fix_features,
.ndo_get_lock_subclass = vlan_dev_get_lock_subclass,
.ndo_get_iflink = vlan_dev_get_iflink,
+ .ndo_flow_offload_check = vlan_dev_flow_offload_check,
};
static void vlan_dev_free(struct net_device *dev)

View file

@ -0,0 +1,55 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Thu, 15 Mar 2018 20:50:37 +0100
Subject: [PATCH] net: bridge: support hardware flow table offload
Look up the real device and pass it on
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/bridge/br_device.c
+++ b/net/bridge/br_device.c
@@ -18,6 +18,8 @@
#include <linux/ethtool.h>
#include <linux/list.h>
#include <linux/netfilter_bridge.h>
+#include <linux/netfilter.h>
+#include <net/netfilter/nf_flow_table.h>
#include <linux/uaccess.h>
#include "br_private.h"
@@ -340,6 +342,26 @@ static const struct ethtool_ops br_ethto
.get_link = ethtool_op_get_link,
};
+static int br_flow_offload_check(struct flow_offload_hw_path *path)
+{
+ struct net_device *dev = path->dev;
+ struct net_bridge *br = netdev_priv(dev);
+ struct net_bridge_fdb_entry *dst;
+
+ if (!(path->flags & FLOW_OFFLOAD_PATH_ETHERNET))
+ return -EINVAL;
+
+ dst = br_fdb_find_rcu(br, path->eth_dest, path->vlan_id);
+ if (!dst || !dst->dst)
+ return -ENOENT;
+
+ path->dev = dst->dst->dev;
+ if (path->dev->netdev_ops->ndo_flow_offload_check)
+ return path->dev->netdev_ops->ndo_flow_offload_check(path);
+
+ return 0;
+}
+
static const struct net_device_ops br_netdev_ops = {
.ndo_open = br_dev_open,
.ndo_stop = br_dev_stop,
@@ -367,6 +389,7 @@ static const struct net_device_ops br_ne
.ndo_bridge_setlink = br_setlink,
.ndo_bridge_dellink = br_dellink,
.ndo_features_check = passthru_features_check,
+ .ndo_flow_offload_check = br_flow_offload_check,
};
static struct device_type br_type = {

View file

@ -0,0 +1,110 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Thu, 15 Mar 2018 21:15:00 +0100
Subject: [PATCH] net: pppoe: support hardware flow table offload
Pass on the PPPoE session ID and the remote MAC address
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/drivers/net/ppp/ppp_generic.c
+++ b/drivers/net/ppp/ppp_generic.c
@@ -56,6 +56,9 @@
#include <net/net_namespace.h>
#include <net/netns/generic.h>
+#include <linux/netfilter.h>
+#include <net/netfilter/nf_flow_table.h>
+
#define PPP_VERSION "2.4.2"
/*
@@ -1383,12 +1386,33 @@ static void ppp_dev_priv_destructor(stru
ppp_destroy_interface(ppp);
}
+static int ppp_flow_offload_check(struct flow_offload_hw_path *path)
+{
+ struct ppp *ppp = netdev_priv(path->dev);
+ struct ppp_channel *chan;
+ struct channel *pch;
+
+ if (ppp->flags & SC_MULTILINK)
+ return -EOPNOTSUPP;
+
+ if (list_empty(&ppp->channels))
+ return -ENODEV;
+
+ pch = list_first_entry(&ppp->channels, struct channel, clist);
+ chan = pch->chan;
+ if (!chan->ops->flow_offload_check)
+ return -EOPNOTSUPP;
+
+ return chan->ops->flow_offload_check(chan, path);
+}
+
static const struct net_device_ops ppp_netdev_ops = {
.ndo_init = ppp_dev_init,
.ndo_uninit = ppp_dev_uninit,
.ndo_start_xmit = ppp_start_xmit,
.ndo_do_ioctl = ppp_net_ioctl,
.ndo_get_stats64 = ppp_get_stats64,
+ .ndo_flow_offload_check = ppp_flow_offload_check,
};
static struct device_type ppp_type = {
--- a/drivers/net/ppp/pppoe.c
+++ b/drivers/net/ppp/pppoe.c
@@ -77,6 +77,8 @@
#include <linux/file.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
+#include <linux/netfilter.h>
+#include <net/netfilter/nf_flow_table.h>
#include <linux/nsproxy.h>
#include <net/net_namespace.h>
@@ -970,8 +972,32 @@ static int pppoe_xmit(struct ppp_channel
return __pppoe_xmit(sk, skb);
}
+static int pppoe_flow_offload_check(struct ppp_channel *chan,
+ struct flow_offload_hw_path *path)
+{
+ struct sock *sk = (struct sock *)chan->private;
+ struct pppox_sock *po = pppox_sk(sk);
+ struct net_device *dev = po->pppoe_dev;
+
+ if (sock_flag(sk, SOCK_DEAD) ||
+ !(sk->sk_state & PPPOX_CONNECTED) || !dev)
+ return -ENODEV;
+
+ path->dev = po->pppoe_dev;
+ path->flags |= FLOW_OFFLOAD_PATH_PPPOE;
+ memcpy(path->eth_src, po->pppoe_dev->dev_addr, ETH_ALEN);
+ memcpy(path->eth_dest, po->pppoe_pa.remote, ETH_ALEN);
+ path->pppoe_sid = be16_to_cpu(po->num);
+
+ if (path->dev->netdev_ops->ndo_flow_offload_check)
+ return path->dev->netdev_ops->ndo_flow_offload_check(path);
+
+ return 0;
+}
+
static const struct ppp_channel_ops pppoe_chan_ops = {
.start_xmit = pppoe_xmit,
+ .flow_offload_check = pppoe_flow_offload_check,
};
static int pppoe_recvmsg(struct socket *sock, struct msghdr *m,
--- a/include/linux/ppp_channel.h
+++ b/include/linux/ppp_channel.h
@@ -32,6 +32,8 @@ struct ppp_channel_ops {
int (*start_xmit)(struct ppp_channel *, struct sk_buff *);
/* Handle an ioctl call that has come in via /dev/ppp. */
int (*ioctl)(struct ppp_channel *, unsigned int, unsigned long);
+
+ int (*flow_offload_check)(struct ppp_channel *, struct flow_offload_hw_path *);
};
struct ppp_channel {

View file

@ -0,0 +1,37 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Sun, 25 Mar 2018 21:10:55 +0200
Subject: [PATCH] netfilter: nf_flow_table: rework hardware offload timeout
handling
Some offload implementations send keepalive packets + explicit
notifications of TCP FIN/RST packets. In this case it is more convenient
to simply let the driver update flow->timeout handling and use the
regular flow offload gc step.
For drivers that manage their own lifetime, a separate flag can be set
to avoid gc timeouts.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/include/net/netfilter/nf_flow_table.h
+++ b/include/net/netfilter/nf_flow_table.h
@@ -76,6 +76,7 @@ struct flow_offload_tuple_rhash {
#define FLOW_OFFLOAD_DYING 0x4
#define FLOW_OFFLOAD_TEARDOWN 0x8
#define FLOW_OFFLOAD_HW 0x10
+#define FLOW_OFFLOAD_KEEP 0x20
struct flow_offload {
struct flow_offload_tuple_rhash tuplehash[FLOW_OFFLOAD_DIR_MAX];
--- a/net/netfilter/nf_flow_table_core.c
+++ b/net/netfilter/nf_flow_table_core.c
@@ -332,7 +332,7 @@ static int nf_flow_offload_gc_step(struc
teardown = flow->flags & (FLOW_OFFLOAD_DYING |
FLOW_OFFLOAD_TEARDOWN);
- if (nf_flow_in_hw(flow) && !teardown)
+ if ((flow->flags & FLOW_OFFLOAD_KEEP) && !teardown)
continue;
if (nf_flow_has_expired(flow) || teardown)