diff options
author | Eric Dumazet <eric.dumazet@gmail.com> | 2009-10-30 03:11:27 -0400 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2009-11-02 02:55:08 -0500 |
commit | 72c9528bab94cc052d00ce241b8e85f5d71e45f0 (patch) | |
tree | fcec7a40e0242e659474a4d9f501d9213225aa55 /net/core/dev.c | |
parent | 68d8287ce1e1da3c99881385a93e74f68c454fc2 (diff) |
net: Introduce dev_get_by_name_rcu()
Some workloads hit dev_base_lock rwlock pretty hard.
We can use RCU lookups to avoid touching this rwlock
(and avoid touching netdevice refcount)
netdevices are already freed after a RCU grace period, so this patch
adds no penalty at device dismantle time.
However, it adds a synchronize_rcu() call in dev_change_name()
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/core/dev.c')
-rw-r--r-- | net/core/dev.c | 49 |
1 files changed, 40 insertions, 9 deletions
diff --git a/net/core/dev.c b/net/core/dev.c index 94f42a15fff1..f54d8b8a434b 100644 --- a/net/core/dev.c +++ b/net/core/dev.c | |||
@@ -213,7 +213,7 @@ static int list_netdevice(struct net_device *dev) | |||
213 | 213 | ||
214 | write_lock_bh(&dev_base_lock); | 214 | write_lock_bh(&dev_base_lock); |
215 | list_add_tail(&dev->dev_list, &net->dev_base_head); | 215 | list_add_tail(&dev->dev_list, &net->dev_base_head); |
216 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); | 216 | hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); |
217 | hlist_add_head_rcu(&dev->index_hlist, | 217 | hlist_add_head_rcu(&dev->index_hlist, |
218 | dev_index_hash(net, dev->ifindex)); | 218 | dev_index_hash(net, dev->ifindex)); |
219 | write_unlock_bh(&dev_base_lock); | 219 | write_unlock_bh(&dev_base_lock); |
@@ -230,7 +230,7 @@ static void unlist_netdevice(struct net_device *dev) | |||
230 | /* Unlink dev from the device chain */ | 230 | /* Unlink dev from the device chain */ |
231 | write_lock_bh(&dev_base_lock); | 231 | write_lock_bh(&dev_base_lock); |
232 | list_del(&dev->dev_list); | 232 | list_del(&dev->dev_list); |
233 | hlist_del(&dev->name_hlist); | 233 | hlist_del_rcu(&dev->name_hlist); |
234 | hlist_del_rcu(&dev->index_hlist); | 234 | hlist_del_rcu(&dev->index_hlist); |
235 | write_unlock_bh(&dev_base_lock); | 235 | write_unlock_bh(&dev_base_lock); |
236 | } | 236 | } |
@@ -599,6 +599,32 @@ struct net_device *__dev_get_by_name(struct net *net, const char *name) | |||
599 | EXPORT_SYMBOL(__dev_get_by_name); | 599 | EXPORT_SYMBOL(__dev_get_by_name); |
600 | 600 | ||
601 | /** | 601 | /** |
602 | * dev_get_by_name_rcu - find a device by its name | ||
603 | * @net: the applicable net namespace | ||
604 | * @name: name to find | ||
605 | * | ||
606 | * Find an interface by name. | ||
607 | * If the name is found a pointer to the device is returned. | ||
608 | * If the name is not found then %NULL is returned. | ||
609 | * The reference counters are not incremented so the caller must be | ||
610 | * careful with locks. The caller must hold RCU lock. | ||
611 | */ | ||
612 | |||
613 | struct net_device *dev_get_by_name_rcu(struct net *net, const char *name) | ||
614 | { | ||
615 | struct hlist_node *p; | ||
616 | struct net_device *dev; | ||
617 | struct hlist_head *head = dev_name_hash(net, name); | ||
618 | |||
619 | hlist_for_each_entry_rcu(dev, p, head, name_hlist) | ||
620 | if (!strncmp(dev->name, name, IFNAMSIZ)) | ||
621 | return dev; | ||
622 | |||
623 | return NULL; | ||
624 | } | ||
625 | EXPORT_SYMBOL(dev_get_by_name_rcu); | ||
626 | |||
627 | /** | ||
602 | * dev_get_by_name - find a device by its name | 628 | * dev_get_by_name - find a device by its name |
603 | * @net: the applicable net namespace | 629 | * @net: the applicable net namespace |
604 | * @name: name to find | 630 | * @name: name to find |
@@ -614,11 +640,11 @@ struct net_device *dev_get_by_name(struct net *net, const char *name) | |||
614 | { | 640 | { |
615 | struct net_device *dev; | 641 | struct net_device *dev; |
616 | 642 | ||
617 | read_lock(&dev_base_lock); | 643 | rcu_read_lock(); |
618 | dev = __dev_get_by_name(net, name); | 644 | dev = dev_get_by_name_rcu(net, name); |
619 | if (dev) | 645 | if (dev) |
620 | dev_hold(dev); | 646 | dev_hold(dev); |
621 | read_unlock(&dev_base_lock); | 647 | rcu_read_unlock(); |
622 | return dev; | 648 | return dev; |
623 | } | 649 | } |
624 | EXPORT_SYMBOL(dev_get_by_name); | 650 | EXPORT_SYMBOL(dev_get_by_name); |
@@ -960,7 +986,12 @@ rollback: | |||
960 | 986 | ||
961 | write_lock_bh(&dev_base_lock); | 987 | write_lock_bh(&dev_base_lock); |
962 | hlist_del(&dev->name_hlist); | 988 | hlist_del(&dev->name_hlist); |
963 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); | 989 | write_unlock_bh(&dev_base_lock); |
990 | |||
991 | synchronize_rcu(); | ||
992 | |||
993 | write_lock_bh(&dev_base_lock); | ||
994 | hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); | ||
964 | write_unlock_bh(&dev_base_lock); | 995 | write_unlock_bh(&dev_base_lock); |
965 | 996 | ||
966 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); | 997 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); |
@@ -1062,9 +1093,9 @@ void dev_load(struct net *net, const char *name) | |||
1062 | { | 1093 | { |
1063 | struct net_device *dev; | 1094 | struct net_device *dev; |
1064 | 1095 | ||
1065 | read_lock(&dev_base_lock); | 1096 | rcu_read_lock(); |
1066 | dev = __dev_get_by_name(net, name); | 1097 | dev = dev_get_by_name_rcu(net, name); |
1067 | read_unlock(&dev_base_lock); | 1098 | rcu_read_unlock(); |
1068 | 1099 | ||
1069 | if (!dev && capable(CAP_NET_ADMIN)) | 1100 | if (!dev && capable(CAP_NET_ADMIN)) |
1070 | request_module("%s", name); | 1101 | request_module("%s", name); |