From e041c683412d5bf44dc2b109053e3b837b71742d Mon Sep 17 00:00:00 2001 From: Alan Stern Date: Mon, 27 Mar 2006 01:16:30 -0800 Subject: [PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern Signed-off-by: Chandra Seetharaman Signed-off-by: Jes Sorensen Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- net/bluetooth/hci_core.c | 8 +++---- net/core/dev.c | 42 +++++++++++++++++++--------------- net/decnet/dn_dev.c | 10 ++++---- net/ipv4/devinet.c | 16 +++++++------ net/ipv4/netfilter/ip_conntrack_core.c | 6 ++--- net/ipv6/addrconf.c | 10 ++++---- net/netfilter/nf_conntrack_core.c | 6 ++--- net/netlink/af_netlink.c | 9 ++++---- 8 files changed, 57 insertions(+), 50 deletions(-) (limited to 'net') diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c index 9106354c781..a49a6975092 100644 --- a/net/bluetooth/hci_core.c +++ b/net/bluetooth/hci_core.c @@ -73,23 +73,23 @@ DEFINE_RWLOCK(hci_cb_list_lock); struct hci_proto *hci_proto[HCI_MAX_PROTO]; /* HCI notifiers list */ -static struct notifier_block *hci_notifier; +static ATOMIC_NOTIFIER_HEAD(hci_notifier); /* ---- HCI notifications ---- */ int hci_register_notifier(struct notifier_block *nb) { - return notifier_chain_register(&hci_notifier, nb); + return atomic_notifier_chain_register(&hci_notifier, nb); } int hci_unregister_notifier(struct notifier_block *nb) { - return notifier_chain_unregister(&hci_notifier, nb); + return atomic_notifier_chain_unregister(&hci_notifier, nb); } static void hci_notify(struct hci_dev *hdev, int event) { - notifier_call_chain(&hci_notifier, event, hdev); + atomic_notifier_call_chain(&hci_notifier, event, hdev); } /* ---- HCI requests ---- */ diff --git a/net/core/dev.c b/net/core/dev.c index 8e1dc305122..a3ab11f3415 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -193,7 +193,7 @@ static inline struct hlist_head *dev_index_hash(int ifindex) * Our notifier list */ -static struct notifier_block *netdev_chain; +static BLOCKING_NOTIFIER_HEAD(netdev_chain); /* * Device drivers call our routines to queue packets here. We empty the @@ -736,7 +736,8 @@ int dev_change_name(struct net_device *dev, char *newname) if (!err) { hlist_del(&dev->name_hlist); hlist_add_head(&dev->name_hlist, dev_name_hash(dev->name)); - notifier_call_chain(&netdev_chain, NETDEV_CHANGENAME, dev); + blocking_notifier_call_chain(&netdev_chain, + NETDEV_CHANGENAME, dev); } return err; @@ -750,7 +751,7 @@ int dev_change_name(struct net_device *dev, char *newname) */ void netdev_features_change(struct net_device *dev) { - notifier_call_chain(&netdev_chain, NETDEV_FEAT_CHANGE, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_FEAT_CHANGE, dev); } EXPORT_SYMBOL(netdev_features_change); @@ -765,7 +766,8 @@ EXPORT_SYMBOL(netdev_features_change); void netdev_state_change(struct net_device *dev) { if (dev->flags & IFF_UP) { - notifier_call_chain(&netdev_chain, NETDEV_CHANGE, dev); + blocking_notifier_call_chain(&netdev_chain, + NETDEV_CHANGE, dev); rtmsg_ifinfo(RTM_NEWLINK, dev, 0); } } @@ -862,7 +864,7 @@ int dev_open(struct net_device *dev) /* * ... and announce new interface. */ - notifier_call_chain(&netdev_chain, NETDEV_UP, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_UP, dev); } return ret; } @@ -885,7 +887,7 @@ int dev_close(struct net_device *dev) * Tell people we are going down, so that they can * prepare to death, when device is still operating. */ - notifier_call_chain(&netdev_chain, NETDEV_GOING_DOWN, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_GOING_DOWN, dev); dev_deactivate(dev); @@ -922,7 +924,7 @@ int dev_close(struct net_device *dev) /* * Tell people we are down */ - notifier_call_chain(&netdev_chain, NETDEV_DOWN, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_DOWN, dev); return 0; } @@ -953,7 +955,7 @@ int register_netdevice_notifier(struct notifier_block *nb) int err; rtnl_lock(); - err = notifier_chain_register(&netdev_chain, nb); + err = blocking_notifier_chain_register(&netdev_chain, nb); if (!err) { for (dev = dev_base; dev; dev = dev->next) { nb->notifier_call(nb, NETDEV_REGISTER, dev); @@ -981,7 +983,7 @@ int unregister_netdevice_notifier(struct notifier_block *nb) int err; rtnl_lock(); - err = notifier_chain_unregister(&netdev_chain, nb); + err = blocking_notifier_chain_unregister(&netdev_chain, nb); rtnl_unlock(); return err; } @@ -992,12 +994,12 @@ int unregister_netdevice_notifier(struct notifier_block *nb) * @v: pointer passed unmodified to notifier function * * Call all network notifier blocks. Parameters and return value - * are as for notifier_call_chain(). + * are as for blocking_notifier_call_chain(). */ int call_netdevice_notifiers(unsigned long val, void *v) { - return notifier_call_chain(&netdev_chain, val, v); + return blocking_notifier_call_chain(&netdev_chain, val, v); } /* When > 0 there are consumers of rx skb time stamps */ @@ -2242,7 +2244,8 @@ int dev_change_flags(struct net_device *dev, unsigned flags) if (dev->flags & IFF_UP && ((old_flags ^ dev->flags) &~ (IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE))) - notifier_call_chain(&netdev_chain, NETDEV_CHANGE, dev); + blocking_notifier_call_chain(&netdev_chain, + NETDEV_CHANGE, dev); if ((flags ^ dev->gflags) & IFF_PROMISC) { int inc = (flags & IFF_PROMISC) ? +1 : -1; @@ -2286,8 +2289,8 @@ int dev_set_mtu(struct net_device *dev, int new_mtu) else dev->mtu = new_mtu; if (!err && dev->flags & IFF_UP) - notifier_call_chain(&netdev_chain, - NETDEV_CHANGEMTU, dev); + blocking_notifier_call_chain(&netdev_chain, + NETDEV_CHANGEMTU, dev); return err; } @@ -2303,7 +2306,8 @@ int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) return -ENODEV; err = dev->set_mac_address(dev, sa); if (!err) - notifier_call_chain(&netdev_chain, NETDEV_CHANGEADDR, dev); + blocking_notifier_call_chain(&netdev_chain, + NETDEV_CHANGEADDR, dev); return err; } @@ -2359,7 +2363,7 @@ static int dev_ifsioc(struct ifreq *ifr, unsigned int cmd) return -EINVAL; memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); - notifier_call_chain(&netdev_chain, + blocking_notifier_call_chain(&netdev_chain, NETDEV_CHANGEADDR, dev); return 0; @@ -2813,7 +2817,7 @@ int register_netdevice(struct net_device *dev) write_unlock_bh(&dev_base_lock); /* Notify protocols, that a new device appeared. */ - notifier_call_chain(&netdev_chain, NETDEV_REGISTER, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_REGISTER, dev); /* Finish registration after unlock */ net_set_todo(dev); @@ -2892,7 +2896,7 @@ static void netdev_wait_allrefs(struct net_device *dev) rtnl_lock(); /* Rebroadcast unregister notification */ - notifier_call_chain(&netdev_chain, + blocking_notifier_call_chain(&netdev_chain, NETDEV_UNREGISTER, dev); if (test_bit(__LINK_STATE_LINKWATCH_PENDING, @@ -3148,7 +3152,7 @@ int unregister_netdevice(struct net_device *dev) /* Notify protocols, that we are about to destroy this device. They should clean all the things. */ - notifier_call_chain(&netdev_chain, NETDEV_UNREGISTER, dev); + blocking_notifier_call_chain(&netdev_chain, NETDEV_UNREGISTER, dev); /* * Flush the multicast chain diff --git a/net/decnet/dn_dev.c b/net/decnet/dn_dev.c index cc7b9d9255e..d2ae9893ca1 100644 --- a/net/decnet/dn_dev.c +++ b/net/decnet/dn_dev.c @@ -68,7 +68,7 @@ __le16 decnet_address = 0; static DEFINE_RWLOCK(dndev_lock); static struct net_device *decnet_default_device; -static struct notifier_block *dnaddr_chain; +static BLOCKING_NOTIFIER_HEAD(dnaddr_chain); static struct dn_dev *dn_dev_create(struct net_device *dev, int *err); static void dn_dev_delete(struct net_device *dev); @@ -446,7 +446,7 @@ static void dn_dev_del_ifa(struct dn_dev *dn_db, struct dn_ifaddr **ifap, int de } rtmsg_ifa(RTM_DELADDR, ifa1); - notifier_call_chain(&dnaddr_chain, NETDEV_DOWN, ifa1); + blocking_notifier_call_chain(&dnaddr_chain, NETDEV_DOWN, ifa1); if (destroy) { dn_dev_free_ifa(ifa1); @@ -481,7 +481,7 @@ static int dn_dev_insert_ifa(struct dn_dev *dn_db, struct dn_ifaddr *ifa) dn_db->ifa_list = ifa; rtmsg_ifa(RTM_NEWADDR, ifa); - notifier_call_chain(&dnaddr_chain, NETDEV_UP, ifa); + blocking_notifier_call_chain(&dnaddr_chain, NETDEV_UP, ifa); return 0; } @@ -1285,12 +1285,12 @@ void dn_dev_devices_on(void) int register_dnaddr_notifier(struct notifier_block *nb) { - return notifier_chain_register(&dnaddr_chain, nb); + return blocking_notifier_chain_register(&dnaddr_chain, nb); } int unregister_dnaddr_notifier(struct notifier_block *nb) { - return notifier_chain_unregister(&dnaddr_chain, nb); + return blocking_notifier_chain_unregister(&dnaddr_chain, nb); } #ifdef CONFIG_PROC_FS diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c index 44fdf1413e2..81c2f788529 100644 --- a/net/ipv4/devinet.c +++ b/net/ipv4/devinet.c @@ -81,7 +81,7 @@ static struct ipv4_devconf ipv4_devconf_dflt = { static void rtmsg_ifa(int event, struct in_ifaddr *); -static struct notifier_block *inetaddr_chain; +static BLOCKING_NOTIFIER_HEAD(inetaddr_chain); static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, int destroy); #ifdef CONFIG_SYSCTL @@ -267,7 +267,8 @@ static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, *ifap1 = ifa->ifa_next; rtmsg_ifa(RTM_DELADDR, ifa); - notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa); + blocking_notifier_call_chain(&inetaddr_chain, + NETDEV_DOWN, ifa); inet_free_ifa(ifa); } else { promote = ifa; @@ -291,7 +292,7 @@ static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, So that, this order is correct. */ rtmsg_ifa(RTM_DELADDR, ifa1); - notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1); + blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1); if (promote) { @@ -303,7 +304,8 @@ static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, promote->ifa_flags &= ~IFA_F_SECONDARY; rtmsg_ifa(RTM_NEWADDR, promote); - notifier_call_chain(&inetaddr_chain, NETDEV_UP, promote); + blocking_notifier_call_chain(&inetaddr_chain, + NETDEV_UP, promote); for (ifa = promote->ifa_next; ifa; ifa = ifa->ifa_next) { if (ifa1->ifa_mask != ifa->ifa_mask || !inet_ifa_match(ifa1->ifa_address, ifa)) @@ -366,7 +368,7 @@ static int inet_insert_ifa(struct in_ifaddr *ifa) Notifier will trigger FIB update, so that listeners of netlink will know about new ifaddr */ rtmsg_ifa(RTM_NEWADDR, ifa); - notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa); + blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa); return 0; } @@ -938,12 +940,12 @@ u32 inet_confirm_addr(const struct net_device *dev, u32 dst, u32 local, int scop int register_inetaddr_notifier(struct notifier_block *nb) { - return notifier_chain_register(&inetaddr_chain, nb); + return blocking_notifier_chain_register(&inetaddr_chain, nb); } int unregister_inetaddr_notifier(struct notifier_block *nb) { - return notifier_chain_unregister(&inetaddr_chain, nb); + return blocking_notifier_chain_unregister(&inetaddr_chain, nb); } /* Rename ifa_labels for a device name change. Make some effort to preserve existing diff --git a/net/ipv4/netfilter/ip_conntrack_core.c b/net/ipv4/netfilter/ip_conntrack_core.c index 9e34034729a..ceaabc18202 100644 --- a/net/ipv4/netfilter/ip_conntrack_core.c +++ b/net/ipv4/netfilter/ip_conntrack_core.c @@ -80,8 +80,8 @@ static int ip_conntrack_vmalloc; static unsigned int ip_conntrack_next_id; static unsigned int ip_conntrack_expect_next_id; #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS -struct notifier_block *ip_conntrack_chain; -struct notifier_block *ip_conntrack_expect_chain; +ATOMIC_NOTIFIER_HEAD(ip_conntrack_chain); +ATOMIC_NOTIFIER_HEAD(ip_conntrack_expect_chain); DEFINE_PER_CPU(struct ip_conntrack_ecache, ip_conntrack_ecache); @@ -92,7 +92,7 @@ __ip_ct_deliver_cached_events(struct ip_conntrack_ecache *ecache) { DEBUGP("ecache: delivering events for %p\n", ecache->ct); if (is_confirmed(ecache->ct) && !is_dying(ecache->ct) && ecache->events) - notifier_call_chain(&ip_conntrack_chain, ecache->events, + atomic_notifier_call_chain(&ip_conntrack_chain, ecache->events, ecache->ct); ecache->events = 0; ip_conntrack_put(ecache->ct); diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c index 01c62a0d374..445006ee452 100644 --- a/net/ipv6/addrconf.c +++ b/net/ipv6/addrconf.c @@ -143,7 +143,7 @@ static void inet6_prefix_notify(int event, struct inet6_dev *idev, struct prefix_info *pinfo); static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev); -static struct notifier_block *inet6addr_chain; +static ATOMIC_NOTIFIER_HEAD(inet6addr_chain); struct ipv6_devconf ipv6_devconf = { .forwarding = 0, @@ -593,7 +593,7 @@ out2: read_unlock_bh(&addrconf_lock); if (likely(err == 0)) - notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa); + atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa); else { kfree(ifa); ifa = ERR_PTR(err); @@ -688,7 +688,7 @@ static void ipv6_del_addr(struct inet6_ifaddr *ifp) ipv6_ifa_notify(RTM_DELADDR, ifp); - notifier_call_chain(&inet6addr_chain,NETDEV_DOWN,ifp); + atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp); addrconf_del_timer(ifp); @@ -3767,12 +3767,12 @@ static void addrconf_sysctl_unregister(struct ipv6_devconf *p) int register_inet6addr_notifier(struct notifier_block *nb) { - return notifier_chain_register(&inet6addr_chain, nb); + return atomic_notifier_chain_register(&inet6addr_chain, nb); } int unregister_inet6addr_notifier(struct notifier_block *nb) { - return notifier_chain_unregister(&inet6addr_chain,nb); + return atomic_notifier_chain_unregister(&inet6addr_chain,nb); } /* diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c index 0ae281d9bfc..56389c83557 100644 --- a/net/netfilter/nf_conntrack_core.c +++ b/net/netfilter/nf_conntrack_core.c @@ -90,8 +90,8 @@ static int nf_conntrack_vmalloc; static unsigned int nf_conntrack_next_id; static unsigned int nf_conntrack_expect_next_id; #ifdef CONFIG_NF_CONNTRACK_EVENTS -struct notifier_block *nf_conntrack_chain; -struct notifier_block *nf_conntrack_expect_chain; +ATOMIC_NOTIFIER_HEAD(nf_conntrack_chain); +ATOMIC_NOTIFIER_HEAD(nf_conntrack_expect_chain); DEFINE_PER_CPU(struct nf_conntrack_ecache, nf_conntrack_ecache); @@ -103,7 +103,7 @@ __nf_ct_deliver_cached_events(struct nf_conntrack_ecache *ecache) DEBUGP("ecache: delivering events for %p\n", ecache->ct); if (nf_ct_is_confirmed(ecache->ct) && !nf_ct_is_dying(ecache->ct) && ecache->events) - notifier_call_chain(&nf_conntrack_chain, ecache->events, + atomic_notifier_call_chain(&nf_conntrack_chain, ecache->events, ecache->ct); ecache->events = 0; diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c index d00a9034cb5..2a233ffcf61 100644 --- a/net/netlink/af_netlink.c +++ b/net/netlink/af_netlink.c @@ -123,7 +123,7 @@ static void netlink_destroy_callback(struct netlink_callback *cb); static DEFINE_RWLOCK(nl_table_lock); static atomic_t nl_table_users = ATOMIC_INIT(0); -static struct notifier_block *netlink_chain; +static ATOMIC_NOTIFIER_HEAD(netlink_chain); static u32 netlink_group_mask(u32 group) { @@ -469,7 +469,8 @@ static int netlink_release(struct socket *sock) .protocol = sk->sk_protocol, .pid = nlk->pid, }; - notifier_call_chain(&netlink_chain, NETLINK_URELEASE, &n); + atomic_notifier_call_chain(&netlink_chain, + NETLINK_URELEASE, &n); } if (nlk->module) @@ -1695,12 +1696,12 @@ static struct file_operations netlink_seq_fops = { int netlink_register_notifier(struct notifier_block *nb) { - return notifier_chain_register(&netlink_chain, nb); + return atomic_notifier_chain_register(&netlink_chain, nb); } int netlink_unregister_notifier(struct notifier_block *nb) { - return notifier_chain_unregister(&netlink_chain, nb); + return atomic_notifier_chain_unregister(&netlink_chain, nb); } static const struct proto_ops netlink_ops = { -- cgit v1.2.3-18-g5258