From: Stephen Hemminger <shemminger@vyatta•com>
To: "David S. Miller" <davem@davemloft•net>,
Hideaki YOSHIFUJI <yoshfuji@linux-ipv6•org>
Cc: netdev@vger•kernel.org
Subject: [PATCH 06/12] IPv6: convert addrconf hash list to RCU
Date: Tue, 02 Mar 2010 15:32:49 -0800 [thread overview]
Message-ID: <20100302234003.262703041@vyatta.com> (raw)
In-Reply-To: 20100302233243.259794027@vyatta.com
[-- Attachment #1: ipv6-addrconf-rcu.patch --]
[-- Type: text/plain, Size: 9668 bytes --]
Convert from reader/writer lock to RCU and spinlock for addrconf
hash list.
Adds an additional helper macro for hlist_for_each_entry_continue_rcu
to handle the continue case.
Signed-off-by: Stephen Hemminger <shemminger@vyatta•com>
---
include/linux/rculist.h | 13 +++++++
include/net/if_inet6.h | 1
net/ipv6/addrconf.c | 80 +++++++++++++++++++++++++-----------------------
3 files changed, 57 insertions(+), 37 deletions(-)
--- a/net/ipv6/addrconf.c 2010-03-02 14:29:25.444951749 -0800
+++ b/net/ipv6/addrconf.c 2010-03-02 14:33:59.033450885 -0800
@@ -127,7 +127,7 @@ static int ipv6_count_addresses(struct i
* Configured unicast address hash table
*/
static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
-static DEFINE_RWLOCK(addrconf_hash_lock);
+static DEFINE_SPINLOCK(addrconf_hash_lock);
static void addrconf_verify(unsigned long);
@@ -523,8 +523,13 @@ static int addrconf_fixup_forwarding(str
}
#endif
-/* Nobody refers to this ifaddr, destroy it */
+static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
+{
+ struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
+ kfree(ifp);
+}
+/* Nobody refers to this ifaddr, destroy it */
void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
{
WARN_ON(ifp->if_next != NULL);
@@ -545,7 +550,7 @@ void inet6_ifa_finish_destroy(struct ine
}
dst_release(&ifp->rt->u.dst);
- kfree(ifp);
+ call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
}
static void
@@ -616,7 +621,7 @@ ipv6_add_addr(struct inet6_dev *idev, co
goto out2;
}
- write_lock(&addrconf_hash_lock);
+ spin_lock(&addrconf_hash_lock);
/* Ignore adding duplicate addresses on an interface */
if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
@@ -670,9 +675,9 @@ ipv6_add_addr(struct inet6_dev *idev, co
/* Add to big hash table */
hash = ipv6_addr_hash(addr);
- hlist_add_head(&ifa->addr_lst, &inet6_addr_lst[hash]);
+ hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
in6_ifa_hold(ifa);
- write_unlock(&addrconf_hash_lock);
+ spin_unlock(&addrconf_hash_lock);
write_lock(&idev->lock);
/* Add to inet6_dev unicast addr list. */
@@ -699,7 +704,7 @@ out2:
return ifa;
out:
- write_unlock(&addrconf_hash_lock);
+ spin_unlock(&addrconf_hash_lock);
goto out2;
}
@@ -717,10 +722,10 @@ static void ipv6_del_addr(struct inet6_i
ifp->dead = 1;
- write_lock_bh(&addrconf_hash_lock);
- hlist_del_init(&ifp->addr_lst);
+ spin_lock_bh(&addrconf_hash_lock);
+ hlist_del_init_rcu(&ifp->addr_lst);
__in6_ifa_put(ifp);
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
write_lock_bh(&idev->lock);
#ifdef CONFIG_IPV6_PRIVACY
@@ -1274,8 +1279,8 @@ int ipv6_chk_addr(struct net *net, struc
struct hlist_node *node;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr) &&
@@ -1285,7 +1290,8 @@ int ipv6_chk_addr(struct net *net, struc
break;
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
+
return ifp != NULL;
}
EXPORT_SYMBOL(ipv6_chk_addr);
@@ -1341,8 +1347,8 @@ struct inet6_ifaddr *ipv6_get_ifaddr(str
struct hlist_node *node;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr)) {
@@ -1353,7 +1359,7 @@ struct inet6_ifaddr *ipv6_get_ifaddr(str
}
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
return ifp;
}
@@ -2691,10 +2697,10 @@ static int addrconf_ifdown(struct net_de
write_unlock_bh(&idev->lock);
/* clear hash table */
- write_lock_bh(&addrconf_hash_lock);
- hlist_del_init(&ifa->addr_lst);
+ spin_lock_bh(&addrconf_hash_lock);
+ hlist_del_init_rcu(&ifa->addr_lst);
__in6_ifa_put(ifa);
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
__ipv6_ifa_notify(RTM_DELADDR, ifa);
atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
@@ -2936,11 +2942,10 @@ static struct inet6_ifaddr *if6_get_firs
for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
struct hlist_node *n;
- hlist_for_each_entry(ifa, n,
- &inet6_addr_lst[state->bucket], addr_lst) {
+ hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
+ addr_lst)
if (net_eq(dev_net(ifa->idev->dev), net))
return ifa;
- }
}
return NULL;
}
@@ -2952,10 +2957,9 @@ static struct inet6_ifaddr *if6_get_next
struct net *net = seq_file_net(seq);
struct hlist_node *n = &ifa->addr_lst;
- hlist_for_each_entry_continue(ifa, n, addr_lst) {
+ hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
if (net_eq(dev_net(ifa->idev->dev), net))
return ifa;
- }
while (++state->bucket < IN6_ADDR_HSIZE) {
hlist_for_each_entry(ifa, n,
@@ -2979,9 +2983,9 @@ static struct inet6_ifaddr *if6_get_idx(
}
static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
- __acquires(addrconf_hash_lock)
+ __acquires(rcu)
{
- read_lock_bh(&addrconf_hash_lock);
+ rcu_read_lock_bh();
return if6_get_idx(seq, *pos);
}
@@ -2995,9 +2999,9 @@ static void *if6_seq_next(struct seq_fil
}
static void if6_seq_stop(struct seq_file *seq, void *v)
- __releases(addrconf_hash_lock)
+ __releases(rcu)
{
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
}
static int if6_seq_show(struct seq_file *seq, void *v)
@@ -3071,8 +3075,8 @@ int ipv6_chk_home_addr(struct net *net,
struct hlist_node *n;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, n, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr) &&
@@ -3081,7 +3085,7 @@ int ipv6_chk_home_addr(struct net *net,
break;
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
return ret;
}
#endif
@@ -3097,7 +3101,8 @@ static void addrconf_verify(unsigned lon
unsigned long now, next;
int i;
- spin_lock_bh(&addrconf_verify_lock);
+ rcu_read_lock_bh();
+ spin_lock(&addrconf_verify_lock);
now = jiffies;
next = now + ADDR_CHECK_FREQUENCY;
@@ -3106,8 +3111,8 @@ static void addrconf_verify(unsigned lon
for (i=0; i < IN6_ADDR_HSIZE; i++) {
restart:
- read_lock(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[i], addr_lst) {
+ hlist_for_each_entry_rcu(ifp, node,
+ &inet6_addr_lst[i], addr_lst) {
unsigned long age;
#ifdef CONFIG_IPV6_PRIVACY
unsigned long regen_advance;
@@ -3129,7 +3134,6 @@ restart:
age >= ifp->valid_lft) {
spin_unlock(&ifp->lock);
in6_ifa_hold(ifp);
- read_unlock(&addrconf_hash_lock);
ipv6_del_addr(ifp);
goto restart;
} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
@@ -3151,7 +3155,6 @@ restart:
if (deprecate) {
in6_ifa_hold(ifp);
- read_unlock(&addrconf_hash_lock);
ipv6_ifa_notify(0, ifp);
in6_ifa_put(ifp);
@@ -3169,7 +3172,7 @@ restart:
in6_ifa_hold(ifp);
in6_ifa_hold(ifpub);
spin_unlock(&ifp->lock);
- read_unlock(&addrconf_hash_lock);
+
spin_lock(&ifpub->lock);
ifpub->regen_count = 0;
spin_unlock(&ifpub->lock);
@@ -3189,12 +3192,12 @@ restart:
spin_unlock(&ifp->lock);
}
}
- read_unlock(&addrconf_hash_lock);
}
addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
add_timer(&addr_chk_timer);
- spin_unlock_bh(&addrconf_verify_lock);
+ spin_unlock(&addrconf_verify_lock);
+ rcu_read_unlock_bh();
}
static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
@@ -4611,10 +4614,10 @@ void addrconf_cleanup(void)
/*
* Check hash table.
*/
- write_lock_bh(&addrconf_hash_lock);
+ spin_lock_bh(&addrconf_hash_lock);
for (i = 0; i < IN6_ADDR_HSIZE; i++)
WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
del_timer(&addr_chk_timer);
rtnl_unlock();
--- a/include/linux/rculist.h 2010-03-01 08:22:23.456662311 -0800
+++ b/include/linux/rculist.h 2010-03-02 14:29:56.641076077 -0800
@@ -428,5 +428,18 @@ static inline void hlist_add_after_rcu(s
({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \
pos = rcu_dereference_raw(pos->next))
+/**
+ * hlist_for_each_entry_continue_rcu - iterate over a hlist continuing after current point
+ * @tpos: the type * to use as a loop cursor.
+ * @pos: the &struct hlist_node to use as a loop cursor.
+ * @member: the name of the hlist_node within the struct.
+ */
+#define hlist_for_each_entry_continue_rcu(tpos, pos, member) \
+ for (pos = rcu_dereference((pos)->next); \
+ pos && ({ prefetch(pos->next); 1; }) && \
+ ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \
+ pos = rcu_dereference(pos->next))
+
+
#endif /* __KERNEL__ */
#endif
--- a/include/net/if_inet6.h 2010-03-02 14:29:25.465077512 -0800
+++ b/include/net/if_inet6.h 2010-03-02 14:31:13.137385294 -0800
@@ -64,6 +64,7 @@ struct inet6_ifaddr {
#endif
int dead;
+ struct rcu_head rcu;
};
struct ip6_sf_socklist {
--
next prev parent reply other threads:[~2010-03-02 23:47 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-03-02 23:32 [PATCH 00/12] IPv6 addrconf changes Stephen Hemminger
2010-03-02 23:32 ` [PATCH 01/12] IPv6: addrconf dad timer unnecessary bh_disable Stephen Hemminger
2010-03-04 8:39 ` David Miller
2010-03-02 23:32 ` [PATCH 02/12] IPv6: addrconf timer race Stephen Hemminger
2010-03-04 8:40 ` David Miller
2010-03-02 23:32 ` [PATCH 03/12] IPv6: addrconf notify when address is unavailable Stephen Hemminger
2010-03-04 8:40 ` David Miller
2010-03-02 23:32 ` [PATCH 04/12] ipv6: convert temporary address list to list macros Stephen Hemminger
2010-03-02 23:32 ` [PATCH 05/12] ipv6: convert addrconf list to hlist Stephen Hemminger
2010-03-02 23:32 ` Stephen Hemminger [this message]
2010-03-02 23:32 ` [PATCH 07/12] ipv6: user better hash for addrconf Stephen Hemminger
2010-03-02 23:32 ` [PATCH 08/12] ipv6: convert idev_list to list macros Stephen Hemminger
2010-03-02 23:32 ` [PATCH 09/12] IPv6: addrconf cleanups Stephen Hemminger
2010-03-02 23:32 ` [PATCH 10/12] IPv6: addrconf checkpatch fixes Stephen Hemminger
2010-03-02 23:32 ` [PATCH 11/12] ipv6: addrconf timer changes Stephen Hemminger
2010-03-02 23:32 ` [PATCH 12/12] IPv6: addrconf cleanup addrconf_verify Stephen Hemminger
2010-03-03 9:16 ` [PATCH 00/12] IPv6 addrconf changes David Miller
2010-03-03 18:14 ` Stephen Hemminger
2010-03-03 18:19 ` [PATCH] IPv6: fix race between cleanup and add/delete address Stephen Hemminger
2010-03-04 8:40 ` David Miller
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20100302234003.262703041@vyatta.com \
--to=shemminger@vyatta$(echo .)com \
--cc=davem@davemloft$(echo .)net \
--cc=netdev@vger$(echo .)kernel.org \
--cc=yoshfuji@linux-ipv6$(echo .)org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox