Commit 19efbd93 authored by Kirill Tkhai's avatar Kirill Tkhai Committed by David S. Miller

net: Kill net_mutex

We take net_mutex, when there are !async pernet_operations
registered, and read locking of net_sem is not enough. But
we may get rid of taking the mutex, and just change the logic
to write lock net_sem in such cases. This obviously reduces
the number of lock operations, we do.
Signed-off-by: default avatarKirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent ffc385b9
...@@ -35,7 +35,6 @@ extern int rtnl_trylock(void); ...@@ -35,7 +35,6 @@ extern int rtnl_trylock(void);
extern int rtnl_is_locked(void); extern int rtnl_is_locked(void);
extern wait_queue_head_t netdev_unregistering_wq; extern wait_queue_head_t netdev_unregistering_wq;
extern struct mutex net_mutex;
extern struct rw_semaphore net_sem; extern struct rw_semaphore net_sem;
#ifdef CONFIG_PROVE_LOCKING #ifdef CONFIG_PROVE_LOCKING
......
...@@ -60,8 +60,11 @@ struct net { ...@@ -60,8 +60,11 @@ struct net {
struct list_head list; /* list of network namespaces */ struct list_head list; /* list of network namespaces */
struct list_head cleanup_list; /* namespaces on death row */ struct list_head cleanup_list; /* namespaces on death row */
struct list_head exit_list; /* Use only net_mutex */ struct list_head exit_list; /* To linked to call pernet exit
* methods on dead net (net_sem
* read locked), or to unregister
* pernet ops (net_sem wr locked).
*/
struct user_namespace *user_ns; /* Owning user namespace */ struct user_namespace *user_ns; /* Owning user namespace */
struct ucounts *ucounts; struct ucounts *ucounts;
spinlock_t nsid_lock; spinlock_t nsid_lock;
...@@ -89,7 +92,7 @@ struct net { ...@@ -89,7 +92,7 @@ struct net {
/* core fib_rules */ /* core fib_rules */
struct list_head rules_ops; struct list_head rules_ops;
struct list_head fib_notifier_ops; /* protected by net_mutex */ struct list_head fib_notifier_ops; /* protected by net_sem */
struct net_device *loopback_dev; /* The loopback */ struct net_device *loopback_dev; /* The loopback */
struct netns_core core; struct netns_core core;
...@@ -316,7 +319,7 @@ struct pernet_operations { ...@@ -316,7 +319,7 @@ struct pernet_operations {
/* /*
* Indicates above methods are allowed to be executed in parallel * Indicates above methods are allowed to be executed in parallel
* with methods of any other pernet_operations, i.e. they are not * with methods of any other pernet_operations, i.e. they are not
* need synchronization via net_mutex. * need write locked net_sem.
*/ */
bool async; bool async;
}; };
......
...@@ -29,8 +29,6 @@ ...@@ -29,8 +29,6 @@
static LIST_HEAD(pernet_list); static LIST_HEAD(pernet_list);
static struct list_head *first_device = &pernet_list; static struct list_head *first_device = &pernet_list;
/* Used only if there are !async pernet_operations registered */
DEFINE_MUTEX(net_mutex);
LIST_HEAD(net_namespace_list); LIST_HEAD(net_namespace_list);
EXPORT_SYMBOL_GPL(net_namespace_list); EXPORT_SYMBOL_GPL(net_namespace_list);
...@@ -407,6 +405,7 @@ struct net *copy_net_ns(unsigned long flags, ...@@ -407,6 +405,7 @@ struct net *copy_net_ns(unsigned long flags,
{ {
struct ucounts *ucounts; struct ucounts *ucounts;
struct net *net; struct net *net;
unsigned write;
int rv; int rv;
if (!(flags & CLONE_NEWNET)) if (!(flags & CLONE_NEWNET))
...@@ -424,20 +423,26 @@ struct net *copy_net_ns(unsigned long flags, ...@@ -424,20 +423,26 @@ struct net *copy_net_ns(unsigned long flags,
refcount_set(&net->passive, 1); refcount_set(&net->passive, 1);
net->ucounts = ucounts; net->ucounts = ucounts;
get_user_ns(user_ns); get_user_ns(user_ns);
again:
rv = down_read_killable(&net_sem); write = READ_ONCE(nr_sync_pernet_ops);
if (write)
rv = down_write_killable(&net_sem);
else
rv = down_read_killable(&net_sem);
if (rv < 0) if (rv < 0)
goto put_userns; goto put_userns;
if (nr_sync_pernet_ops) {
rv = mutex_lock_killable(&net_mutex); if (!write && unlikely(READ_ONCE(nr_sync_pernet_ops))) {
if (rv < 0) up_read(&net_sem);
goto up_read; goto again;
} }
rv = setup_net(net, user_ns); rv = setup_net(net, user_ns);
if (nr_sync_pernet_ops)
mutex_unlock(&net_mutex); if (write)
up_read: up_write(&net_sem);
up_read(&net_sem); else
up_read(&net_sem);
if (rv < 0) { if (rv < 0) {
put_userns: put_userns:
put_user_ns(user_ns); put_user_ns(user_ns);
...@@ -485,15 +490,23 @@ static void cleanup_net(struct work_struct *work) ...@@ -485,15 +490,23 @@ static void cleanup_net(struct work_struct *work)
struct net *net, *tmp, *last; struct net *net, *tmp, *last;
struct list_head net_kill_list; struct list_head net_kill_list;
LIST_HEAD(net_exit_list); LIST_HEAD(net_exit_list);
unsigned write;
/* Atomically snapshot the list of namespaces to cleanup */ /* Atomically snapshot the list of namespaces to cleanup */
spin_lock_irq(&cleanup_list_lock); spin_lock_irq(&cleanup_list_lock);
list_replace_init(&cleanup_list, &net_kill_list); list_replace_init(&cleanup_list, &net_kill_list);
spin_unlock_irq(&cleanup_list_lock); spin_unlock_irq(&cleanup_list_lock);
again:
write = READ_ONCE(nr_sync_pernet_ops);
if (write)
down_write(&net_sem);
else
down_read(&net_sem);
down_read(&net_sem); if (!write && unlikely(READ_ONCE(nr_sync_pernet_ops))) {
if (nr_sync_pernet_ops) up_read(&net_sem);
mutex_lock(&net_mutex); goto again;
}
/* Don't let anyone else find us. */ /* Don't let anyone else find us. */
rtnl_lock(); rtnl_lock();
...@@ -528,14 +541,14 @@ static void cleanup_net(struct work_struct *work) ...@@ -528,14 +541,14 @@ static void cleanup_net(struct work_struct *work)
list_for_each_entry_reverse(ops, &pernet_list, list) list_for_each_entry_reverse(ops, &pernet_list, list)
ops_exit_list(ops, &net_exit_list); ops_exit_list(ops, &net_exit_list);
if (nr_sync_pernet_ops)
mutex_unlock(&net_mutex);
/* Free the net generic variables */ /* Free the net generic variables */
list_for_each_entry_reverse(ops, &pernet_list, list) list_for_each_entry_reverse(ops, &pernet_list, list)
ops_free_list(ops, &net_exit_list); ops_free_list(ops, &net_exit_list);
up_read(&net_sem); if (write)
up_write(&net_sem);
else
up_read(&net_sem);
/* Ensure there are no outstanding rcu callbacks using this /* Ensure there are no outstanding rcu callbacks using this
* network namespace. * network namespace.
...@@ -563,8 +576,6 @@ static void cleanup_net(struct work_struct *work) ...@@ -563,8 +576,6 @@ static void cleanup_net(struct work_struct *work)
void net_ns_barrier(void) void net_ns_barrier(void)
{ {
down_write(&net_sem); down_write(&net_sem);
mutex_lock(&net_mutex);
mutex_unlock(&net_mutex);
up_write(&net_sem); up_write(&net_sem);
} }
EXPORT_SYMBOL(net_ns_barrier); EXPORT_SYMBOL(net_ns_barrier);
......
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