Commit 5cd667b0 authored by Alex Wang's avatar Alex Wang Committed by Pravin B Shelar

openvswitch: Allow each vport to have an array of 'port_id's.

In order to allow handlers directly read upcalls from datapath,
we need to support per-handler netlink socket for each vport in
datapath.  This commit makes this happen.  Also, it is guaranteed
to be backward compatible with previous branch.
Signed-off-by: default avatarAlex Wang <alexw@nicira.com>
Acked-by: default avatarThomas Graf <tgraf@redhat.com>
Signed-off-by: default avatarPravin B Shelar <pshelar@nicira.com>
parent f6e67532
......@@ -118,6 +118,9 @@ struct ovs_vport_stats {
/* Allow last Netlink attribute to be unaligned */
#define OVS_DP_F_UNALIGNED (1 << 0)
/* Allow datapath to associate multiple Netlink PIDs to each vport */
#define OVS_DP_F_VPORT_PIDS (1 << 1)
/* Fixed logical ports. */
#define OVSP_LOCAL ((__u32)0)
......@@ -203,9 +206,10 @@ enum ovs_vport_type {
* this is the name of the network device. Maximum length %IFNAMSIZ-1 bytes
* plus a null terminator.
* @OVS_VPORT_ATTR_OPTIONS: Vport-specific configuration information.
* @OVS_VPORT_ATTR_UPCALL_PID: The Netlink socket in userspace that
* OVS_PACKET_CMD_MISS upcalls will be directed to for packets received on
* this port. A value of zero indicates that upcalls should not be sent.
* @OVS_VPORT_ATTR_UPCALL_PID: The array of Netlink socket pids in userspace
* among which OVS_PACKET_CMD_MISS upcalls will be distributed for packets
* received on this port. If this is a single-element array of value 0,
* upcalls should not be sent.
* @OVS_VPORT_ATTR_STATS: A &struct ovs_vport_stats giving statistics for
* packets sent or received through the vport.
*
......@@ -228,7 +232,8 @@ enum ovs_vport_attr {
OVS_VPORT_ATTR_TYPE, /* u32 OVS_VPORT_TYPE_* constant. */
OVS_VPORT_ATTR_NAME, /* string name, up to IFNAMSIZ bytes long */
OVS_VPORT_ATTR_OPTIONS, /* nested attributes, varies by vport type */
OVS_VPORT_ATTR_UPCALL_PID, /* u32 Netlink PID to receive upcalls */
OVS_VPORT_ATTR_UPCALL_PID, /* array of u32 Netlink socket PIDs for */
/* receiving upcalls */
OVS_VPORT_ATTR_STATS, /* struct ovs_vport_stats */
__OVS_VPORT_ATTR_MAX
};
......
......@@ -266,7 +266,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
upcall.cmd = OVS_PACKET_CMD_MISS;
upcall.key = &key;
upcall.userdata = NULL;
upcall.portid = p->upcall_portid;
upcall.portid = ovs_vport_find_upcall_portid(p, skb);
ovs_dp_upcall(dp, skb, &upcall);
consume_skb(skb);
stats_counter = &stats->n_missed;
......@@ -1373,7 +1373,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
parms.options = NULL;
parms.dp = dp;
parms.port_no = OVSP_LOCAL;
parms.upcall_portid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
parms.upcall_portids = a[OVS_DP_ATTR_UPCALL_PID];
ovs_dp_change(dp, a);
......@@ -1632,8 +1632,8 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
nla_put_string(skb, OVS_VPORT_ATTR_NAME, vport->ops->get_name(vport)) ||
nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_portid))
nla_put_string(skb, OVS_VPORT_ATTR_NAME,
vport->ops->get_name(vport)))
goto nla_put_failure;
ovs_vport_get_stats(vport, &vport_stats);
......@@ -1641,6 +1641,9 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
&vport_stats))
goto nla_put_failure;
if (ovs_vport_get_upcall_portids(vport, skb))
goto nla_put_failure;
err = ovs_vport_get_options(vport, skb);
if (err == -EMSGSIZE)
goto error;
......@@ -1762,7 +1765,7 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
parms.options = a[OVS_VPORT_ATTR_OPTIONS];
parms.dp = dp;
parms.port_no = port_no;
parms.upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
parms.upcall_portids = a[OVS_VPORT_ATTR_UPCALL_PID];
vport = new_vport(&parms);
err = PTR_ERR(vport);
......@@ -1812,8 +1815,14 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
goto exit_unlock_free;
}
if (a[OVS_VPORT_ATTR_UPCALL_PID])
vport->upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
if (a[OVS_VPORT_ATTR_UPCALL_PID]) {
struct nlattr *ids = a[OVS_VPORT_ATTR_UPCALL_PID];
err = ovs_vport_set_upcall_portids(vport, ids);
if (err)
goto exit_unlock_free;
}
err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
info->snd_seq, 0, OVS_VPORT_CMD_NEW);
......
......@@ -134,10 +134,12 @@ struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
vport->dp = parms->dp;
vport->port_no = parms->port_no;
vport->upcall_portid = parms->upcall_portid;
vport->ops = ops;
INIT_HLIST_NODE(&vport->dp_hash_node);
if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids))
return ERR_PTR(-EINVAL);
vport->percpu_stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
if (!vport->percpu_stats) {
kfree(vport);
......@@ -161,6 +163,10 @@ struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
*/
void ovs_vport_free(struct vport *vport)
{
/* vport is freed from RCU callback or error path, Therefore
* it is safe to use raw dereference.
*/
kfree(rcu_dereference_raw(vport->upcall_portids));
free_percpu(vport->percpu_stats);
kfree(vport);
}
......@@ -326,6 +332,99 @@ int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
return 0;
}
/**
* ovs_vport_set_upcall_portids - set upcall portids of @vport.
*
* @vport: vport to modify.
* @ids: new configuration, an array of port ids.
*
* Sets the vport's upcall_portids to @ids.
*
* Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
* as an array of U32.
*
* Must be called with ovs_mutex.
*/
int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
{
struct vport_portids *old, *vport_portids;
if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
return -EINVAL;
old = ovsl_dereference(vport->upcall_portids);
vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
GFP_KERNEL);
if (!vport_portids)
return -ENOMEM;
vport_portids->n_ids = nla_len(ids) / sizeof(u32);
vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
nla_memcpy(vport_portids->ids, ids, nla_len(ids));
rcu_assign_pointer(vport->upcall_portids, vport_portids);
if (old)
kfree_rcu(old, rcu);
return 0;
}
/**
* ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
*
* @vport: vport from which to retrieve the portids.
* @skb: sk_buff where portids should be appended.
*
* Retrieves the configuration of the given vport, appending the
* %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
* portids to @skb.
*
* Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
* If an error occurs, @skb is left unmodified. Must be called with
* ovs_mutex or rcu_read_lock.
*/
int ovs_vport_get_upcall_portids(const struct vport *vport,
struct sk_buff *skb)
{
struct vport_portids *ids;
ids = rcu_dereference_ovsl(vport->upcall_portids);
if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
ids->n_ids * sizeof(u32), (void *)ids->ids);
else
return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
}
/**
* ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
*
* @vport: vport from which the missed packet is received.
* @skb: skb that the missed packet was received.
*
* Uses the skb_get_hash() to select the upcall portid to send the
* upcall.
*
* Returns the portid of the target socket. Must be called with rcu_read_lock.
*/
u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
{
struct vport_portids *ids;
u32 ids_index;
u32 hash;
ids = rcu_dereference(p->upcall_portids);
if (ids->n_ids == 1 && ids->ids[0] == 0)
return 0;
hash = skb_get_hash(skb);
ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
return ids->ids[ids_index];
}
/**
* ovs_vport_receive - pass up received packet to the datapath for processing
*
......
......@@ -23,6 +23,7 @@
#include <linux/list.h>
#include <linux/netlink.h>
#include <linux/openvswitch.h>
#include <linux/reciprocal_div.h>
#include <linux/skbuff.h>
#include <linux/spinlock.h>
#include <linux/u64_stats_sync.h>
......@@ -52,6 +53,10 @@ void ovs_vport_get_stats(struct vport *, struct ovs_vport_stats *);
int ovs_vport_set_options(struct vport *, struct nlattr *options);
int ovs_vport_get_options(const struct vport *, struct sk_buff *);
int ovs_vport_set_upcall_portids(struct vport *, struct nlattr *pids);
int ovs_vport_get_upcall_portids(const struct vport *, struct sk_buff *);
u32 ovs_vport_find_upcall_portid(const struct vport *, struct sk_buff *);
int ovs_vport_send(struct vport *, struct sk_buff *);
/* The following definitions are for implementers of vport devices: */
......@@ -62,13 +67,27 @@ struct vport_err_stats {
u64 tx_dropped;
u64 tx_errors;
};
/**
* struct vport_portids - array of netlink portids of a vport.
* must be protected by rcu.
* @rn_ids: The reciprocal value of @n_ids.
* @rcu: RCU callback head for deferred destruction.
* @n_ids: Size of @ids array.
* @ids: Array storing the Netlink socket pids to be used for packets received
* on this port that miss the flow table.
*/
struct vport_portids {
struct reciprocal_value rn_ids;
struct rcu_head rcu;
u32 n_ids;
u32 ids[];
};
/**
* struct vport - one port within a datapath
* @rcu: RCU callback head for deferred destruction.
* @dp: Datapath to which this port belongs.
* @upcall_portid: The Netlink port to use for packets received on this port that
* miss the flow table.
* @upcall_portids: RCU protected 'struct vport_portids'.
* @port_no: Index into @dp's @ports array.
* @hash_node: Element in @dev_table hash table in vport.c.
* @dp_hash_node: Element in @datapath->ports hash table in datapath.c.
......@@ -80,7 +99,7 @@ struct vport_err_stats {
struct vport {
struct rcu_head rcu;
struct datapath *dp;
u32 upcall_portid;
struct vport_portids __rcu *upcall_portids;
u16 port_no;
struct hlist_node hash_node;
......@@ -111,7 +130,7 @@ struct vport_parms {
/* For ovs_vport_alloc(). */
struct datapath *dp;
u16 port_no;
u32 upcall_portid;
struct nlattr *upcall_portids;
};
/**
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment