iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

openvswitch: Reduce locking requirements.

Browse Source 
Reduce and clarify locking requirements for ovs_flow_cmd_alloc_info(),
ovs_flow_cmd_fill_info() and ovs_flow_cmd_build_info().

A datapath pointer is available only when holding a lock.  Change
ovs_flow_cmd_fill_info() and ovs_flow_cmd_build_info() to take a
dp_ifindex directly, rather than a datapath pointer that is then
(only) used to get the dp_ifindex.  This is useful, since the
dp_ifindex is available even when the datapath pointer is not, both
before and after taking a lock, which makes further critical section
reduction possible.

Make ovs_flow_cmd_alloc_info() take an 'acts' argument instead a
'flow' pointer.  This allows some future patches to do the allocation
before acquiring the flow pointer.

The locking requirements after this patch are:

ovs_flow_cmd_alloc_info(): May be called without locking, must not be
called while holding the RCU read lock (due to memory allocation).
If 'acts' belong to a flow in the flow table, however, then the
caller must hold ovs_mutex.

ovs_flow_cmd_fill_info(): Either ovs_mutex or RCU read lock must be held.

ovs_flow_cmd_build_info(): This calls both of the above, so the caller
must hold ovs_mutex.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
This commit is contained in:
Jarno Rajahalme 2014-05-05 14:28:07 -07:00 committed by Pravin B Shelar
parent 86ec8dbae2
commit 0e9796b4af
1 changed files with 29 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
net/openvswitch/datapath.c
View File

@ -663,7 +663,7 @@ static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
} }
/* Called with ovs_mutex or RCU read lock. */ /* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
struct sk_buff *skb, u32 portid, struct sk_buff *skb, u32 portid,
u32 seq, u32 flags, u8 cmd) u32 seq, u32 flags, u8 cmd)
{ {
@ -680,7 +680,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
if (!ovs_header) if (!ovs_header)
return -EMSGSIZE; return -EMSGSIZE;
ovs_header->dp_ifindex = get_dpifindex(dp); ovs_header->dp_ifindex = dp_ifindex;
/* Fill flow key. */ /* Fill flow key. */
nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY); nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
@ -703,6 +703,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
nla_nest_end(skb, nla); nla_nest_end(skb, nla);
ovs_flow_stats_get(flow, &stats, &used, &tcp_flags); ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
if (used && if (used &&
nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used))) nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
goto nla_put_failure; goto nla_put_failure;
@ -730,9 +731,9 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
const struct sw_flow_actions *sf_acts; const struct sw_flow_actions *sf_acts;
sf_acts = rcu_dereference_ovsl(flow->sf_acts); sf_acts = rcu_dereference_ovsl(flow->sf_acts);
err = ovs_nla_put_actions(sf_acts->actions, err = ovs_nla_put_actions(sf_acts->actions,
sf_acts->actions_len, skb); sf_acts->actions_len, skb);
if (!err) if (!err)
nla_nest_end(skb, start); nla_nest_end(skb, start);
else { else {
@ -753,41 +754,40 @@ error:
return err; return err;
} }
/* Must be called with ovs_mutex. */ /* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow, static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
struct genl_info *info, struct genl_info *info,
bool always) bool always)
{ {
struct sk_buff *skb; struct sk_buff *skb;
size_t len;
if (!always && !ovs_must_notify(info, &ovs_dp_flow_multicast_group)) if (!always && !ovs_must_notify(info, &ovs_dp_flow_multicast_group))
return NULL; return NULL;
len = ovs_flow_cmd_msg_size(ovsl_dereference(flow->sf_acts)); skb = genlmsg_new_unicast(ovs_flow_cmd_msg_size(acts), info, GFP_KERNEL);
skb = genlmsg_new_unicast(len, info, GFP_KERNEL);
if (!skb) if (!skb)
return ERR_PTR(-ENOMEM); return ERR_PTR(-ENOMEM);
return skb; return skb;
} }
/* Must be called with ovs_mutex. */ /* Called with ovs_mutex. */
static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow, static struct sk_buff *ovs_flow_cmd_build_info(const struct sw_flow *flow,
struct datapath *dp, int dp_ifindex,
struct genl_info *info, struct genl_info *info, u8 cmd,
u8 cmd, bool always) bool always)
{ {
struct sk_buff *skb; struct sk_buff *skb;
int retval; int retval;
skb = ovs_flow_cmd_alloc_info(flow, info, always); skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
always);
if (!skb || IS_ERR(skb)) if (!skb || IS_ERR(skb))
return skb; return skb;
retval = ovs_flow_cmd_fill_info(flow, dp, skb, info->snd_portid, retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
info->snd_seq, 0, cmd); info->snd_portid, info->snd_seq, 0,
cmd);
BUG_ON(retval < 0); BUG_ON(retval < 0);
return skb; return skb;
} }
@ -868,8 +868,8 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
goto err_flow_free; goto err_flow_free;
} }
reply = ovs_flow_cmd_build_info(flow, dp, info, reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
OVS_FLOW_CMD_NEW, false); info, OVS_FLOW_CMD_NEW, false);
} else { } else {
/* We found a matching flow. */ /* We found a matching flow. */
/* Bail out if we're not allowed to modify an existing flow. /* Bail out if we're not allowed to modify an existing flow.
@ -895,8 +895,9 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
rcu_assign_pointer(flow->sf_acts, acts); rcu_assign_pointer(flow->sf_acts, acts);
ovs_nla_free_flow_actions(old_acts); ovs_nla_free_flow_actions(old_acts);
} }
reply = ovs_flow_cmd_build_info(flow, dp, info,
OVS_FLOW_CMD_NEW, false); reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
info, OVS_FLOW_CMD_NEW, false);
/* Clear stats. */ /* Clear stats. */
if (a[OVS_FLOW_ATTR_CLEAR]) if (a[OVS_FLOW_ATTR_CLEAR])
@ -958,7 +959,8 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
goto unlock; goto unlock;
} }
reply = ovs_flow_cmd_build_info(flow, dp, info, OVS_FLOW_CMD_NEW, true); reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
OVS_FLOW_CMD_NEW, true);
if (IS_ERR(reply)) { if (IS_ERR(reply)) {
err = PTR_ERR(reply); err = PTR_ERR(reply);
goto unlock; goto unlock;
@ -1005,7 +1007,8 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
goto unlock; goto unlock;
} }
reply = ovs_flow_cmd_alloc_info(flow, info, false); reply = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
false);
if (IS_ERR(reply)) { if (IS_ERR(reply)) {
err = PTR_ERR(reply); err = PTR_ERR(reply);
goto unlock; goto unlock;
@ -1014,7 +1017,8 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
ovs_flow_tbl_remove(&dp->table, flow); ovs_flow_tbl_remove(&dp->table, flow);
if (reply) { if (reply) {
err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid, err = ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex,
reply, info->snd_portid,
info->snd_seq, 0, info->snd_seq, 0,
OVS_FLOW_CMD_DEL); OVS_FLOW_CMD_DEL);
BUG_ON(err < 0); BUG_ON(err < 0);
@ -1054,7 +1058,7 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
if (!flow) if (!flow)
break; break;
if (ovs_flow_cmd_fill_info(flow, dp, skb, if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
NETLINK_CB(cb->skb).portid, NETLINK_CB(cb->skb).portid,
cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh->nlmsg_seq, NLM_F_MULTI,
OVS_FLOW_CMD_NEW) < 0) OVS_FLOW_CMD_NEW) < 0)