linux/net/netlink/af_netlink.c

2740 lines
63 KiB
C
Raw Normal View History

/*
* NETLINK Kernel-user communication protocol.
*
* Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
* Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
* Patrick McHardy <kaber@trash.net>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*
* Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
* added netlink_proto_exit
* Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
* use nlk_sk, as sk->protinfo is on a diet 8)
* Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
* - inc module use count of module that owns
* the kernel socket in case userspace opens
* socket of same protocol
* - remove all module support, since netlink is
* mandatory if CONFIG_NET=y these days
*/
#include <linux/module.h>
#include <linux/capability.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/stat.h>
#include <linux/socket.h>
#include <linux/un.h>
#include <linux/fcntl.h>
#include <linux/termios.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/uaccess.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/rtnetlink.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/notifier.h>
#include <linux/security.h>
#include <linux/jhash.h>
#include <linux/jiffies.h>
#include <linux/random.h>
#include <linux/bitops.h>
#include <linux/mm.h>
#include <linux/types.h>
#include <linux/audit.h>
#include <linux/mutex.h>
#include <linux/vmalloc.h>
#include <linux/if_arp.h>
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
#include <linux/rhashtable.h>
#include <asm/cacheflush.h>
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
#include <linux/hash.h>
genetlink: synchronize socket closing and family removal In addition to the problem Jeff Layton reported, I looked at the code and reproduced the same warning by subscribing and removing the genl family with a socket still open. This is a fairly tricky race which originates in the fact that generic netlink allows the family to go away while sockets are still open - unlike regular netlink which has a module refcount for every open socket so in general this cannot be triggered. Trying to resolve this issue by the obvious locking isn't possible as it will result in deadlocks between unregistration and group unbind notification (which incidentally lockdep doesn't find due to the home grown locking in the netlink table.) To really resolve this, introduce a "closing socket" reference counter (for generic netlink only, as it's the only affected family) in the core netlink code and use that in generic netlink to wait for all the sockets that are being closed at the same time as a generic netlink family is removed. This fixes the race that when a socket is closed, it will should call the unbind, but if the family is removed at the same time the unbind will not find it, leading to the warning. The real problem though is that in this case the unbind could actually find a new family that is registered to have a multicast group with the same ID, and call its mcast_unbind() leading to confusing. Also remove the warning since it would still trigger, but is now no longer a problem. This also moves the code in af_netlink.c to before unreferencing the module to avoid having the same problem in the normal non-genl case. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 10:37:14 +00:00
#include <linux/genetlink.h>
#include <linux/net_namespace.h>
#include <net/net_namespace.h>
#include <net/sock.h>
#include <net/scm.h>
#include <net/netlink.h>
#include "af_netlink.h"
struct listeners {
struct rcu_head rcu;
unsigned long masks[0];
};
/* state bits */
#define NETLINK_S_CONGESTED 0x0
static inline int netlink_is_kernel(struct sock *sk)
{
return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET;
}
struct netlink_table *nl_table __read_mostly;
EXPORT_SYMBOL_GPL(nl_table);
static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS];
static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = {
"nlk_cb_mutex-ROUTE",
"nlk_cb_mutex-1",
"nlk_cb_mutex-USERSOCK",
"nlk_cb_mutex-FIREWALL",
"nlk_cb_mutex-SOCK_DIAG",
"nlk_cb_mutex-NFLOG",
"nlk_cb_mutex-XFRM",
"nlk_cb_mutex-SELINUX",
"nlk_cb_mutex-ISCSI",
"nlk_cb_mutex-AUDIT",
"nlk_cb_mutex-FIB_LOOKUP",
"nlk_cb_mutex-CONNECTOR",
"nlk_cb_mutex-NETFILTER",
"nlk_cb_mutex-IP6_FW",
"nlk_cb_mutex-DNRTMSG",
"nlk_cb_mutex-KOBJECT_UEVENT",
"nlk_cb_mutex-GENERIC",
"nlk_cb_mutex-17",
"nlk_cb_mutex-SCSITRANSPORT",
"nlk_cb_mutex-ECRYPTFS",
"nlk_cb_mutex-RDMA",
"nlk_cb_mutex-CRYPTO",
"nlk_cb_mutex-SMC",
"nlk_cb_mutex-23",
"nlk_cb_mutex-24",
"nlk_cb_mutex-25",
"nlk_cb_mutex-26",
"nlk_cb_mutex-27",
"nlk_cb_mutex-28",
"nlk_cb_mutex-29",
"nlk_cb_mutex-30",
"nlk_cb_mutex-31",
"nlk_cb_mutex-MAX_LINKS"
};
static int netlink_dump(struct sock *sk);
static void netlink_skb_destructor(struct sk_buff *skb);
/* nl_table locking explained:
* Lookup and traversal are protected with an RCU read-side lock. Insertion
* and removal are protected with per bucket lock while using RCU list
* modification primitives and may run in parallel to RCU protected lookups.
* Destruction of the Netlink socket may only occur *after* nl_table_lock has
* been acquired * either during or after the socket has been removed from
* the list and after an RCU grace period.
*/
DEFINE_RWLOCK(nl_table_lock);
EXPORT_SYMBOL_GPL(nl_table_lock);
static atomic_t nl_table_users = ATOMIC_INIT(0);
#define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
static BLOCKING_NOTIFIER_HEAD(netlink_chain);
static DEFINE_SPINLOCK(netlink_tap_lock);
static struct list_head netlink_tap_all __read_mostly;
static const struct rhashtable_params netlink_rhashtable_params;
static inline u32 netlink_group_mask(u32 group)
{
return group ? 1 << (group - 1) : 0;
}
static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb,
gfp_t gfp_mask)
{
unsigned int len = skb_end_offset(skb);
struct sk_buff *new;
new = alloc_skb(len, gfp_mask);
if (new == NULL)
return NULL;
NETLINK_CB(new).portid = NETLINK_CB(skb).portid;
NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group;
NETLINK_CB(new).creds = NETLINK_CB(skb).creds;
memcpy(skb_put(new, len), skb->data, len);
return new;
}
int netlink_add_tap(struct netlink_tap *nt)
{
if (unlikely(nt->dev->type != ARPHRD_NETLINK))
return -EINVAL;
spin_lock(&netlink_tap_lock);
list_add_rcu(&nt->list, &netlink_tap_all);
spin_unlock(&netlink_tap_lock);
__module_get(nt->module);
return 0;
}
EXPORT_SYMBOL_GPL(netlink_add_tap);
static int __netlink_remove_tap(struct netlink_tap *nt)
{
bool found = false;
struct netlink_tap *tmp;
spin_lock(&netlink_tap_lock);
list_for_each_entry(tmp, &netlink_tap_all, list) {
if (nt == tmp) {
list_del_rcu(&nt->list);
found = true;
goto out;
}
}
pr_warn("__netlink_remove_tap: %p not found\n", nt);
out:
spin_unlock(&netlink_tap_lock);
if (found)
module_put(nt->module);
return found ? 0 : -ENODEV;
}
int netlink_remove_tap(struct netlink_tap *nt)
{
int ret;
ret = __netlink_remove_tap(nt);
synchronize_net();
return ret;
}
EXPORT_SYMBOL_GPL(netlink_remove_tap);
static bool netlink_filter_tap(const struct sk_buff *skb)
{
struct sock *sk = skb->sk;
/* We take the more conservative approach and
* whitelist socket protocols that may pass.
*/
switch (sk->sk_protocol) {
case NETLINK_ROUTE:
case NETLINK_USERSOCK:
case NETLINK_SOCK_DIAG:
case NETLINK_NFLOG:
case NETLINK_XFRM:
case NETLINK_FIB_LOOKUP:
case NETLINK_NETFILTER:
case NETLINK_GENERIC:
return true;
}
return false;
}
static int __netlink_deliver_tap_skb(struct sk_buff *skb,
struct net_device *dev)
{
struct sk_buff *nskb;
struct sock *sk = skb->sk;
int ret = -ENOMEM;
dev_hold(dev);
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
if (is_vmalloc_addr(skb->head))
nskb = netlink_to_full_skb(skb, GFP_ATOMIC);
else
nskb = skb_clone(skb, GFP_ATOMIC);
if (nskb) {
nskb->dev = dev;
nskb->protocol = htons((u16) sk->sk_protocol);
netlink: specify netlink packet direction for nlmon In order to facilitate development for netlink protocol dissector, fill the unused field skb->pkt_type of the cloned skb with a hint of the address space of the new owner (receiver) socket in the notion of "to kernel" resp. "to user". At the time we invoke __netlink_deliver_tap_skb(), we already have set the new skb owner via netlink_skb_set_owner_r(), so we can use that for netlink_is_kernel() probing. In normal PF_PACKET network traffic, this field denotes if the packet is destined for us (PACKET_HOST), if it's broadcast (PACKET_BROADCAST), etc. As we only have 3 bit reserved, we can use the value (= 6) of PACKET_FASTROUTE as it's _not used_ anywhere in the whole kernel and not supported anywhere, and packets of such type were never exposed to user space, so there are no overlapping users of such kind. Thus, as wished, that seems the only way to make both PACKET_* values non-overlapping and therefore device agnostic. By using those two flags for netlink skbs on nlmon devices, they can be made available and picked up via sll_pkttype (previously unused in netlink context) in struct sockaddr_ll. We now have these two directions: - PACKET_USER (= 6) -> to user space - PACKET_KERNEL (= 7) -> to kernel space Partial `ip a` example strace for sa_family=AF_NETLINK with detected nl msg direction: syscall: direction: sendto(3, ...) = 40 /* to kernel */ recvmsg(3, ...) = 3404 /* to user */ recvmsg(3, ...) = 1120 /* to user */ recvmsg(3, ...) = 20 /* to user */ sendto(3, ...) = 40 /* to kernel */ recvmsg(3, ...) = 168 /* to user */ recvmsg(3, ...) = 144 /* to user */ recvmsg(3, ...) = 20 /* to user */ Signed-off-by: Daniel Borkmann <dborkman@redhat.com> Signed-off-by: Jakub Zawadzki <darkjames-ws@darkjames.pl> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-12-23 13:35:56 +00:00
nskb->pkt_type = netlink_is_kernel(sk) ?
PACKET_KERNEL : PACKET_USER;
skb_reset_network_header(nskb);
ret = dev_queue_xmit(nskb);
if (unlikely(ret > 0))
ret = net_xmit_errno(ret);
}
dev_put(dev);
return ret;
}
static void __netlink_deliver_tap(struct sk_buff *skb)
{
int ret;
struct netlink_tap *tmp;
if (!netlink_filter_tap(skb))
return;
list_for_each_entry_rcu(tmp, &netlink_tap_all, list) {
ret = __netlink_deliver_tap_skb(skb, tmp->dev);
if (unlikely(ret))
break;
}
}
static void netlink_deliver_tap(struct sk_buff *skb)
{
rcu_read_lock();
if (unlikely(!list_empty(&netlink_tap_all)))
__netlink_deliver_tap(skb);
rcu_read_unlock();
}
static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src,
struct sk_buff *skb)
{
if (!(netlink_is_kernel(dst) && netlink_is_kernel(src)))
netlink_deliver_tap(skb);
}
static void netlink_overrun(struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) {
if (!test_and_set_bit(NETLINK_S_CONGESTED,
&nlk_sk(sk)->state)) {
sk->sk_err = ENOBUFS;
sk->sk_error_report(sk);
}
}
atomic_inc(&sk->sk_drops);
}
static void netlink_rcv_wake(struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
if (skb_queue_empty(&sk->sk_receive_queue))
clear_bit(NETLINK_S_CONGESTED, &nlk->state);
if (!test_bit(NETLINK_S_CONGESTED, &nlk->state))
wake_up_interruptible(&nlk->wait);
}
static void netlink_skb_destructor(struct sk_buff *skb)
{
if (is_vmalloc_addr(skb->head)) {
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
if (!skb->cloned ||
!atomic_dec_return(&(skb_shinfo(skb)->dataref)))
vfree(skb->head);
skb->head = NULL;
}
if (skb->sk != NULL)
sock_rfree(skb);
}
static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
{
WARN_ON(skb->sk != NULL);
skb->sk = sk;
skb->destructor = netlink_skb_destructor;
atomic_add(skb->truesize, &sk->sk_rmem_alloc);
sk_mem_charge(sk, skb->truesize);
}
static void netlink_sock_destruct(struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
if (nlk->cb_running) {
if (nlk->cb.done)
nlk->cb.done(&nlk->cb);
module_put(nlk->cb.module);
kfree_skb(nlk->cb.skb);
}
skb_queue_purge(&sk->sk_receive_queue);
if (!sock_flag(sk, SOCK_DEAD)) {
printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
return;
}
WARN_ON(atomic_read(&sk->sk_rmem_alloc));
WARN_ON(atomic_read(&sk->sk_wmem_alloc));
WARN_ON(nlk_sk(sk)->groups);
}
static void netlink_sock_destruct_work(struct work_struct *work)
{
struct netlink_sock *nlk = container_of(work, struct netlink_sock,
work);
sk_free(&nlk->sk);
}
/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
* SMP. Look, when several writers sleep and reader wakes them up, all but one
* immediately hit write lock and grab all the cpus. Exclusive sleep solves
* this, _but_ remember, it adds useless work on UP machines.
*/
void netlink_table_grab(void)
__acquires(nl_table_lock)
{
might_sleep();
write_lock_irq(&nl_table_lock);
if (atomic_read(&nl_table_users)) {
DECLARE_WAITQUEUE(wait, current);
add_wait_queue_exclusive(&nl_table_wait, &wait);
for (;;) {
set_current_state(TASK_UNINTERRUPTIBLE);
if (atomic_read(&nl_table_users) == 0)
break;
write_unlock_irq(&nl_table_lock);
schedule();
write_lock_irq(&nl_table_lock);
}
__set_current_state(TASK_RUNNING);
remove_wait_queue(&nl_table_wait, &wait);
}
}
void netlink_table_ungrab(void)
__releases(nl_table_lock)
{
write_unlock_irq(&nl_table_lock);
wake_up(&nl_table_wait);
}
static inline void
netlink_lock_table(void)
{
/* read_lock() synchronizes us to netlink_table_grab */
read_lock(&nl_table_lock);
atomic_inc(&nl_table_users);
read_unlock(&nl_table_lock);
}
static inline void
netlink_unlock_table(void)
{
if (atomic_dec_and_test(&nl_table_users))
wake_up(&nl_table_wait);
}
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
struct netlink_compare_arg
{
possible_net_t pnet;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
u32 portid;
};
/* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
#define netlink_compare_arg_len \
(offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
static inline int netlink_compare(struct rhashtable_compare_arg *arg,
const void *ptr)
{
const struct netlink_compare_arg *x = arg->key;
const struct netlink_sock *nlk = ptr;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
return nlk->portid != x->portid ||
!net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet));
}
static void netlink_compare_arg_init(struct netlink_compare_arg *arg,
struct net *net, u32 portid)
{
memset(arg, 0, sizeof(*arg));
write_pnet(&arg->pnet, net);
arg->portid = portid;
}
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
struct net *net)
{
struct netlink_compare_arg arg;
netlink_compare_arg_init(&arg, net, portid);
return rhashtable_lookup_fast(&table->hash, &arg,
netlink_rhashtable_params);
}
static int __netlink_insert(struct netlink_table *table, struct sock *sk)
{
struct netlink_compare_arg arg;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid);
return rhashtable_lookup_insert_key(&table->hash, &arg,
&nlk_sk(sk)->node,
netlink_rhashtable_params);
}
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
{
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
struct netlink_table *table = &nl_table[protocol];
struct sock *sk;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
rcu_read_lock();
sk = __netlink_lookup(table, portid, net);
if (sk)
sock_hold(sk);
rcu_read_unlock();
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
return sk;
}
static const struct proto_ops netlink_ops;
static void
netlink_update_listeners(struct sock *sk)
{
struct netlink_table *tbl = &nl_table[sk->sk_protocol];
unsigned long mask;
unsigned int i;
struct listeners *listeners;
listeners = nl_deref_protected(tbl->listeners);
if (!listeners)
return;
for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
mask = 0;
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
sk_for_each_bound(sk, &tbl->mc_list) {
if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
mask |= nlk_sk(sk)->groups[i];
}
listeners->masks[i] = mask;
}
/* this function is only called with the netlink table "grabbed", which
* makes sure updates are visible before bind or setsockopt return. */
}
static int netlink_insert(struct sock *sk, u32 portid)
{
struct netlink_table *table = &nl_table[sk->sk_protocol];
int err;
lock_sock(sk);
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY;
if (nlk_sk(sk)->bound)
goto err;
err = -ENOMEM;
rhashtable: Per bucket locks & deferred expansion/shrinking Introduces an array of spinlocks to protect bucket mutations. The number of spinlocks per CPU is configurable and selected based on the hash of the bucket. This allows for parallel insertions and removals of entries which do not share a lock. The patch also defers expansion and shrinking to a worker queue which allows insertion and removal from atomic context. Insertions and deletions may occur in parallel to it and are only held up briefly while the particular bucket is linked or unzipped. Mutations of the bucket table pointer is protected by a new mutex, read access is RCU protected. In the event of an expansion or shrinking, the new bucket table allocated is exposed as a so called future table as soon as the resize process starts. Lookups, deletions, and insertions will briefly use both tables. The future table becomes the main table after an RCU grace period and initial linking of the old to the new table was performed. Optimization of the chains to make use of the new number of buckets follows only the new table is in use. The side effect of this is that during that RCU grace period, a bucket traversal using any rht_for_each() variant on the main table will not see any insertions performed during the RCU grace period which would at that point land in the future table. The lookup will see them as it searches both tables if needed. Having multiple insertions and removals occur in parallel requires nelems to become an atomic counter. Signed-off-by: Thomas Graf <tgraf@suug.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-02 22:00:20 +00:00
if (BITS_PER_LONG > 32 &&
unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX))
goto err;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
nlk_sk(sk)->portid = portid;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
sock_hold(sk);
err = __netlink_insert(table, sk);
if (err) {
netlink: make sure -EBUSY won't escape from netlink_insert Linus reports the following deadlock on rtnl_mutex; triggered only once so far (extract): [12236.694209] NetworkManager D 0000000000013b80 0 1047 1 0x00000000 [12236.694218] ffff88003f902640 0000000000000000 ffffffff815d15a9 0000000000000018 [12236.694224] ffff880119538000 ffff88003f902640 ffffffff81a8ff84 00000000ffffffff [12236.694230] ffffffff81a8ff88 ffff880119c47f00 ffffffff815d133a ffffffff81a8ff80 [12236.694235] Call Trace: [12236.694250] [<ffffffff815d15a9>] ? schedule_preempt_disabled+0x9/0x10 [12236.694257] [<ffffffff815d133a>] ? schedule+0x2a/0x70 [12236.694263] [<ffffffff815d15a9>] ? schedule_preempt_disabled+0x9/0x10 [12236.694271] [<ffffffff815d2c3f>] ? __mutex_lock_slowpath+0x7f/0xf0 [12236.694280] [<ffffffff815d2cc6>] ? mutex_lock+0x16/0x30 [12236.694291] [<ffffffff814f1f90>] ? rtnetlink_rcv+0x10/0x30 [12236.694299] [<ffffffff8150ce3b>] ? netlink_unicast+0xfb/0x180 [12236.694309] [<ffffffff814f5ad3>] ? rtnl_getlink+0x113/0x190 [12236.694319] [<ffffffff814f202a>] ? rtnetlink_rcv_msg+0x7a/0x210 [12236.694331] [<ffffffff8124565c>] ? sock_has_perm+0x5c/0x70 [12236.694339] [<ffffffff814f1fb0>] ? rtnetlink_rcv+0x30/0x30 [12236.694346] [<ffffffff8150d62c>] ? netlink_rcv_skb+0x9c/0xc0 [12236.694354] [<ffffffff814f1f9f>] ? rtnetlink_rcv+0x1f/0x30 [12236.694360] [<ffffffff8150ce3b>] ? netlink_unicast+0xfb/0x180 [12236.694367] [<ffffffff8150d344>] ? netlink_sendmsg+0x484/0x5d0 [12236.694376] [<ffffffff810a236f>] ? __wake_up+0x2f/0x50 [12236.694387] [<ffffffff814cad23>] ? sock_sendmsg+0x33/0x40 [12236.694396] [<ffffffff814cb05e>] ? ___sys_sendmsg+0x22e/0x240 [12236.694405] [<ffffffff814cab75>] ? ___sys_recvmsg+0x135/0x1a0 [12236.694415] [<ffffffff811a9d12>] ? eventfd_write+0x82/0x210 [12236.694423] [<ffffffff811a0f9e>] ? fsnotify+0x32e/0x4c0 [12236.694429] [<ffffffff8108cb70>] ? wake_up_q+0x60/0x60 [12236.694434] [<ffffffff814cba09>] ? __sys_sendmsg+0x39/0x70 [12236.694440] [<ffffffff815d4797>] ? entry_SYSCALL_64_fastpath+0x12/0x6a It seems so far plausible that the recursive call into rtnetlink_rcv() looks suspicious. One way, where this could trigger is that the senders NETLINK_CB(skb).portid was wrongly 0 (which is rtnetlink socket), so the rtnl_getlink() request's answer would be sent to the kernel instead to the actual user process, thus grabbing rtnl_mutex() twice. One theory would be that netlink_autobind() triggered via netlink_sendmsg() internally overwrites the -EBUSY error to 0, but where it is wrongly originating from __netlink_insert() instead. That would reset the socket's portid to 0, which is then filled into NETLINK_CB(skb).portid later on. As commit d470e3b483dc ("[NETLINK]: Fix two socket hashing bugs.") also puts it, -EBUSY should not be propagated from netlink_insert(). It looks like it's very unlikely to reproduce. We need to trigger the rhashtable_insert_rehash() handler under a situation where rehashing currently occurs (one /rare/ way would be to hit ht->elasticity limits while not filled enough to expand the hashtable, but that would rather require a specifically crafted bind() sequence with knowledge about destination slots, seems unlikely). It probably makes sense to guard __netlink_insert() in any case and remap that error. It was suggested that EOVERFLOW might be better than an already overloaded ENOMEM. Reference: http://thread.gmane.org/gmane.linux.network/372676 Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Herbert Xu <herbert@gondor.apana.org.au> Acked-by: Thomas Graf <tgraf@suug.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-06 22:26:41 +00:00
/* In case the hashtable backend returns with -EBUSY
* from here, it must not escape to the caller.
*/
if (unlikely(err == -EBUSY))
err = -EOVERFLOW;
if (err == -EEXIST)
err = -EADDRINUSE;
sock_put(sk);
goto err;
}
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
/* We need to ensure that the socket is hashed and visible. */
smp_wmb();
nlk_sk(sk)->bound = portid;
err:
release_sock(sk);
return err;
}
static void netlink_remove(struct sock *sk)
{
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
struct netlink_table *table;
table = &nl_table[sk->sk_protocol];
if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node,
netlink_rhashtable_params)) {
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
__sock_put(sk);
}
netlink_table_grab();
if (nlk_sk(sk)->subscriptions) {
__sk_del_bind_node(sk);
netlink_update_listeners(sk);
}
genetlink: synchronize socket closing and family removal In addition to the problem Jeff Layton reported, I looked at the code and reproduced the same warning by subscribing and removing the genl family with a socket still open. This is a fairly tricky race which originates in the fact that generic netlink allows the family to go away while sockets are still open - unlike regular netlink which has a module refcount for every open socket so in general this cannot be triggered. Trying to resolve this issue by the obvious locking isn't possible as it will result in deadlocks between unregistration and group unbind notification (which incidentally lockdep doesn't find due to the home grown locking in the netlink table.) To really resolve this, introduce a "closing socket" reference counter (for generic netlink only, as it's the only affected family) in the core netlink code and use that in generic netlink to wait for all the sockets that are being closed at the same time as a generic netlink family is removed. This fixes the race that when a socket is closed, it will should call the unbind, but if the family is removed at the same time the unbind will not find it, leading to the warning. The real problem though is that in this case the unbind could actually find a new family that is registered to have a multicast group with the same ID, and call its mcast_unbind() leading to confusing. Also remove the warning since it would still trigger, but is now no longer a problem. This also moves the code in af_netlink.c to before unreferencing the module to avoid having the same problem in the normal non-genl case. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 10:37:14 +00:00
if (sk->sk_protocol == NETLINK_GENERIC)
atomic_inc(&genl_sk_destructing_cnt);
netlink_table_ungrab();
}
static struct proto netlink_proto = {
.name = "NETLINK",
.owner = THIS_MODULE,
.obj_size = sizeof(struct netlink_sock),
};
static int __netlink_create(struct net *net, struct socket *sock,
struct mutex *cb_mutex, int protocol,
int kern)
{
struct sock *sk;
struct netlink_sock *nlk;
sock->ops = &netlink_ops;
sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern);
if (!sk)
return -ENOMEM;
sock_init_data(sock, sk);
nlk = nlk_sk(sk);
if (cb_mutex) {
nlk->cb_mutex = cb_mutex;
} else {
nlk->cb_mutex = &nlk->cb_def_mutex;
mutex_init(nlk->cb_mutex);
lockdep_set_class_and_name(nlk->cb_mutex,
nlk_cb_mutex_keys + protocol,
nlk_cb_mutex_key_strings[protocol]);
}
init_waitqueue_head(&nlk->wait);
sk->sk_destruct = netlink_sock_destruct;
sk->sk_protocol = protocol;
return 0;
}
static int netlink_create(struct net *net, struct socket *sock, int protocol,
int kern)
{
struct module *module = NULL;
struct mutex *cb_mutex;
struct netlink_sock *nlk;
int (*bind)(struct net *net, int group);
void (*unbind)(struct net *net, int group);
int err = 0;
sock->state = SS_UNCONNECTED;
if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
return -ESOCKTNOSUPPORT;
if (protocol < 0 || protocol >= MAX_LINKS)
return -EPROTONOSUPPORT;
netlink_lock_table();
#ifdef CONFIG_MODULES
if (!nl_table[protocol].registered) {
netlink_unlock_table();
request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
netlink_lock_table();
}
#endif
if (nl_table[protocol].registered &&
try_module_get(nl_table[protocol].module))
module = nl_table[protocol].module;
netlink: fix for too early rmmod Netlink code does module autoload if protocol userspace is asking for is not ready. However, module can dissapear right after it was autoloaded. Example: modprobe/rmmod stress-testing and xfrm_user.ko providing NETLINK_XFRM. netlink_create() in such situation _will_ create userspace socket and _will_not_ pin module. Now if module was removed and we're going to call ->netlink_rcv into nothing: BUG: unable to handle kernel paging request at ffffffffa02f842a ^^^^^^^^^^^^^^^^ modules are loaded near these addresses here IP: [<ffffffffa02f842a>] 0xffffffffa02f842a PGD 161f067 PUD 1623063 PMD baa12067 PTE 0 Oops: 0010 [#1] PREEMPT SMP DEBUG_PAGEALLOC last sysfs file: /sys/devices/pci0000:00/0000:00:1f.2/host0/target0:0:0/0:0:0:0/block/sda/uevent CPU 1 Pid: 11515, comm: ip Not tainted 2.6.33-rc5-netns-00594-gaaa5728-dirty #6 P5E/P5E RIP: 0010:[<ffffffffa02f842a>] [<ffffffffa02f842a>] 0xffffffffa02f842a RSP: 0018:ffff8800baa3db48 EFLAGS: 00010292 RAX: ffff8800baa3dfd8 RBX: ffff8800be353640 RCX: 0000000000000000 RDX: ffffffff81959380 RSI: ffff8800bab7f130 RDI: 0000000000000001 RBP: ffff8800baa3db58 R08: 0000000000000001 R09: 0000000000000000 R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000011 R13: ffff8800be353640 R14: ffff8800bcdec240 R15: ffff8800bd488010 FS: 00007f93749656f0(0000) GS:ffff880002300000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: ffffffffa02f842a CR3: 00000000ba82b000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process ip (pid: 11515, threadinfo ffff8800baa3c000, task ffff8800bab7eb30) Stack: ffffffff813637c0 ffff8800bd488000 ffff8800baa3dba8 ffffffff8136397d <0> 0000000000000000 ffffffff81344adc 7fffffffffffffff 0000000000000000 <0> ffff8800baa3ded8 ffff8800be353640 ffff8800bcdec240 0000000000000000 Call Trace: [<ffffffff813637c0>] ? netlink_unicast+0x100/0x2d0 [<ffffffff8136397d>] netlink_unicast+0x2bd/0x2d0 netlink_unicast_kernel: nlk->netlink_rcv(skb); [<ffffffff81344adc>] ? memcpy_fromiovec+0x6c/0x90 [<ffffffff81364263>] netlink_sendmsg+0x1d3/0x2d0 [<ffffffff8133975b>] sock_sendmsg+0xbb/0xf0 [<ffffffff8106cdeb>] ? __lock_acquire+0x27b/0xa60 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff8106db22>] ? __lock_release+0x82/0x170 [<ffffffff810a190e>] ? might_fault+0xbe/0xd0 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff81344c77>] ? verify_iovec+0x47/0xd0 [<ffffffff8133a509>] sys_sendmsg+0x1a9/0x360 [<ffffffff813c2be5>] ? _raw_spin_unlock_irqrestore+0x65/0x70 [<ffffffff8106aced>] ? trace_hardirqs_on+0xd/0x10 [<ffffffff813c2bc2>] ? _raw_spin_unlock_irqrestore+0x42/0x70 [<ffffffff81197004>] ? __up_read+0x84/0xb0 [<ffffffff8106ac95>] ? trace_hardirqs_on_caller+0x145/0x190 [<ffffffff813c207f>] ? trace_hardirqs_on_thunk+0x3a/0x3f [<ffffffff8100262b>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<ffffffffa02f842a>] 0xffffffffa02f842a RSP <ffff8800baa3db48> CR2: ffffffffa02f842a If module was quickly removed after autoloading, return -E. Return -EPROTONOSUPPORT if module was quickly removed after autoloading. Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-01-30 10:05:05 +00:00
else
err = -EPROTONOSUPPORT;
cb_mutex = nl_table[protocol].cb_mutex;
bind = nl_table[protocol].bind;
unbind = nl_table[protocol].unbind;
netlink_unlock_table();
netlink: fix for too early rmmod Netlink code does module autoload if protocol userspace is asking for is not ready. However, module can dissapear right after it was autoloaded. Example: modprobe/rmmod stress-testing and xfrm_user.ko providing NETLINK_XFRM. netlink_create() in such situation _will_ create userspace socket and _will_not_ pin module. Now if module was removed and we're going to call ->netlink_rcv into nothing: BUG: unable to handle kernel paging request at ffffffffa02f842a ^^^^^^^^^^^^^^^^ modules are loaded near these addresses here IP: [<ffffffffa02f842a>] 0xffffffffa02f842a PGD 161f067 PUD 1623063 PMD baa12067 PTE 0 Oops: 0010 [#1] PREEMPT SMP DEBUG_PAGEALLOC last sysfs file: /sys/devices/pci0000:00/0000:00:1f.2/host0/target0:0:0/0:0:0:0/block/sda/uevent CPU 1 Pid: 11515, comm: ip Not tainted 2.6.33-rc5-netns-00594-gaaa5728-dirty #6 P5E/P5E RIP: 0010:[<ffffffffa02f842a>] [<ffffffffa02f842a>] 0xffffffffa02f842a RSP: 0018:ffff8800baa3db48 EFLAGS: 00010292 RAX: ffff8800baa3dfd8 RBX: ffff8800be353640 RCX: 0000000000000000 RDX: ffffffff81959380 RSI: ffff8800bab7f130 RDI: 0000000000000001 RBP: ffff8800baa3db58 R08: 0000000000000001 R09: 0000000000000000 R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000011 R13: ffff8800be353640 R14: ffff8800bcdec240 R15: ffff8800bd488010 FS: 00007f93749656f0(0000) GS:ffff880002300000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: ffffffffa02f842a CR3: 00000000ba82b000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process ip (pid: 11515, threadinfo ffff8800baa3c000, task ffff8800bab7eb30) Stack: ffffffff813637c0 ffff8800bd488000 ffff8800baa3dba8 ffffffff8136397d <0> 0000000000000000 ffffffff81344adc 7fffffffffffffff 0000000000000000 <0> ffff8800baa3ded8 ffff8800be353640 ffff8800bcdec240 0000000000000000 Call Trace: [<ffffffff813637c0>] ? netlink_unicast+0x100/0x2d0 [<ffffffff8136397d>] netlink_unicast+0x2bd/0x2d0 netlink_unicast_kernel: nlk->netlink_rcv(skb); [<ffffffff81344adc>] ? memcpy_fromiovec+0x6c/0x90 [<ffffffff81364263>] netlink_sendmsg+0x1d3/0x2d0 [<ffffffff8133975b>] sock_sendmsg+0xbb/0xf0 [<ffffffff8106cdeb>] ? __lock_acquire+0x27b/0xa60 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff8106db22>] ? __lock_release+0x82/0x170 [<ffffffff810a190e>] ? might_fault+0xbe/0xd0 [<ffffffff810a18c3>] ? might_fault+0x73/0xd0 [<ffffffff81344c77>] ? verify_iovec+0x47/0xd0 [<ffffffff8133a509>] sys_sendmsg+0x1a9/0x360 [<ffffffff813c2be5>] ? _raw_spin_unlock_irqrestore+0x65/0x70 [<ffffffff8106aced>] ? trace_hardirqs_on+0xd/0x10 [<ffffffff813c2bc2>] ? _raw_spin_unlock_irqrestore+0x42/0x70 [<ffffffff81197004>] ? __up_read+0x84/0xb0 [<ffffffff8106ac95>] ? trace_hardirqs_on_caller+0x145/0x190 [<ffffffff813c207f>] ? trace_hardirqs_on_thunk+0x3a/0x3f [<ffffffff8100262b>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<ffffffffa02f842a>] 0xffffffffa02f842a RSP <ffff8800baa3db48> CR2: ffffffffa02f842a If module was quickly removed after autoloading, return -E. Return -EPROTONOSUPPORT if module was quickly removed after autoloading. Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-01-30 10:05:05 +00:00
if (err < 0)
goto out;
err = __netlink_create(net, sock, cb_mutex, protocol, kern);
if (err < 0)
goto out_module;
local_bh_disable();
sock_prot_inuse_add(net, &netlink_proto, 1);
local_bh_enable();
nlk = nlk_sk(sock->sk);
nlk->module = module;
nlk->netlink_bind = bind;
nlk->netlink_unbind = unbind;
out:
return err;
out_module:
module_put(module);
goto out;
}
static void deferred_put_nlk_sk(struct rcu_head *head)
{
struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu);
struct sock *sk = &nlk->sk;
if (!atomic_dec_and_test(&sk->sk_refcnt))
return;
if (nlk->cb_running && nlk->cb.done) {
INIT_WORK(&nlk->work, netlink_sock_destruct_work);
schedule_work(&nlk->work);
return;
}
sk_free(sk);
}
static int netlink_release(struct socket *sock)
{
struct sock *sk = sock->sk;
struct netlink_sock *nlk;
if (!sk)
return 0;
netlink_remove(sk);
sock_orphan(sk);
nlk = nlk_sk(sk);
/*
* OK. Socket is unlinked, any packets that arrive now
* will be purged.
*/
genetlink: synchronize socket closing and family removal In addition to the problem Jeff Layton reported, I looked at the code and reproduced the same warning by subscribing and removing the genl family with a socket still open. This is a fairly tricky race which originates in the fact that generic netlink allows the family to go away while sockets are still open - unlike regular netlink which has a module refcount for every open socket so in general this cannot be triggered. Trying to resolve this issue by the obvious locking isn't possible as it will result in deadlocks between unregistration and group unbind notification (which incidentally lockdep doesn't find due to the home grown locking in the netlink table.) To really resolve this, introduce a "closing socket" reference counter (for generic netlink only, as it's the only affected family) in the core netlink code and use that in generic netlink to wait for all the sockets that are being closed at the same time as a generic netlink family is removed. This fixes the race that when a socket is closed, it will should call the unbind, but if the family is removed at the same time the unbind will not find it, leading to the warning. The real problem though is that in this case the unbind could actually find a new family that is registered to have a multicast group with the same ID, and call its mcast_unbind() leading to confusing. Also remove the warning since it would still trigger, but is now no longer a problem. This also moves the code in af_netlink.c to before unreferencing the module to avoid having the same problem in the normal non-genl case. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 10:37:14 +00:00
/* must not acquire netlink_table_lock in any way again before unbind
* and notifying genetlink is done as otherwise it might deadlock
*/
if (nlk->netlink_unbind) {
int i;
for (i = 0; i < nlk->ngroups; i++)
if (test_bit(i, nlk->groups))
nlk->netlink_unbind(sock_net(sk), i + 1);
}
if (sk->sk_protocol == NETLINK_GENERIC &&
atomic_dec_return(&genl_sk_destructing_cnt) == 0)
wake_up(&genl_sk_destructing_waitq);
sock->sk = NULL;
wake_up_interruptible_all(&nlk->wait);
skb_queue_purge(&sk->sk_write_queue);
if (nlk->portid && nlk->bound) {
struct netlink_notify n = {
.net = sock_net(sk),
.protocol = sk->sk_protocol,
.portid = nlk->portid,
};
blocking_notifier_call_chain(&netlink_chain,
[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 <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 09:16:30 +00:00
NETLINK_URELEASE, &n);
}
module_put(nlk->module);
if (netlink_is_kernel(sk)) {
netlink_table_grab();
BUG_ON(nl_table[sk->sk_protocol].registered == 0);
if (--nl_table[sk->sk_protocol].registered == 0) {
struct listeners *old;
old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
kfree_rcu(old, rcu);
nl_table[sk->sk_protocol].module = NULL;
nl_table[sk->sk_protocol].bind = NULL;
nl_table[sk->sk_protocol].unbind = NULL;
nl_table[sk->sk_protocol].flags = 0;
nl_table[sk->sk_protocol].registered = 0;
}
netlink_table_ungrab();
}
kfree(nlk->groups);
nlk->groups = NULL;
local_bh_disable();
sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
local_bh_enable();
call_rcu(&nlk->rcu, deferred_put_nlk_sk);
return 0;
}
static int netlink_autobind(struct socket *sock)
{
struct sock *sk = sock->sk;
struct net *net = sock_net(sk);
struct netlink_table *table = &nl_table[sk->sk_protocol];
s32 portid = task_tgid_vnr(current);
int err;
s32 rover = -4096;
bool ok;
retry:
cond_resched();
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
rcu_read_lock();
ok = !__netlink_lookup(table, portid, net);
rcu_read_unlock();
if (!ok) {
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
/* Bind collision, search negative portid values. */
if (rover == -4096)
/* rover will be in range [S32_MIN, -4097] */
rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN);
else if (rover >= -4096)
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
rover = -4097;
portid = rover--;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
goto retry;
}
err = netlink_insert(sk, portid);
if (err == -EADDRINUSE)
goto retry;
[NETLINK]: Fix two socket hashing bugs. 1) netlink_release() should only decrement the hash entry count if the socket was actually hashed. This was causing hash->entries to underflow, which resulting in all kinds of troubles. On 64-bit systems, this would cause the following conditional to erroneously trigger: err = -ENOMEM; if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) goto err; 2) netlink_autobind() needs to propagate the error return from netlink_insert(). Otherwise, callers will not see the error as they should and thus try to operate on a socket with a zero pid, which is very bad. However, it should not propagate -EBUSY. If two threads race to autobind the socket, that is fine. This is consistent with the autobind behavior in other protocols. So bug #1 above, combined with this one, resulted in hangs on netlink_sendmsg() calls to the rtnetlink socket. We'd try to do the user sendmsg() with the socket's pid set to zero, later we do a socket lookup using that pid (via the value we stashed away in NETLINK_CB(skb).pid), but that won't give us the user socket, it will give us the rtnetlink socket. So when we try to wake up the receive queue, we dive back into rtnetlink_rcv() which tries to recursively take the rtnetlink semaphore. Thanks to Jakub Jelink for providing backtraces. Also, thanks to Herbert Xu for supplying debugging patches to help track this down, and also finding a mistake in an earlier version of this fix. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-26 22:31:51 +00:00
/* If 2 threads race to autobind, that is fine. */
if (err == -EBUSY)
err = 0;
return err;
}
/**
* __netlink_ns_capable - General netlink message capability test
* @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
* @user_ns: The user namespace of the capability to use
* @cap: The capability to use
*
* Test to see if the opener of the socket we received the message
* from had when the netlink socket was created and the sender of the
* message has has the capability @cap in the user namespace @user_ns.
*/
bool __netlink_ns_capable(const struct netlink_skb_parms *nsp,
struct user_namespace *user_ns, int cap)
{
netlink: Only check file credentials for implicit destinations It was possible to get a setuid root or setcap executable to write to it's stdout or stderr (which has been set made a netlink socket) and inadvertently reconfigure the networking stack. To prevent this we check that both the creator of the socket and the currentl applications has permission to reconfigure the network stack. Unfortunately this breaks Zebra which always uses sendto/sendmsg and creates it's socket without any privileges. To keep Zebra working don't bother checking if the creator of the socket has privilege when a destination address is specified. Instead rely exclusively on the privileges of the sender of the socket. Note from Andy: This is exactly Eric's code except for some comment clarifications and formatting fixes. Neither I nor, I think, anyone else is thrilled with this approach, but I'm hesitant to wait on a better fix since 3.15 is almost here. Note to stable maintainers: This is a mess. An earlier series of patches in 3.15 fix a rather serious security issue (CVE-2014-0181), but they did so in a way that breaks Zebra. The offending series includes: commit aa4cf9452f469f16cea8c96283b641b4576d4a7b Author: Eric W. Biederman <ebiederm@xmission.com> Date: Wed Apr 23 14:28:03 2014 -0700 net: Add variants of capable for use on netlink messages If a given kernel version is missing that series of fixes, it's probably worth backporting it and this patch. if that series is present, then this fix is critical if you care about Zebra. Cc: stable@vger.kernel.org Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-05-30 18:04:00 +00:00
return ((nsp->flags & NETLINK_SKB_DST) ||
file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) &&
ns_capable(user_ns, cap);
}
EXPORT_SYMBOL(__netlink_ns_capable);
/**
* netlink_ns_capable - General netlink message capability test
* @skb: socket buffer holding a netlink command from userspace
* @user_ns: The user namespace of the capability to use
* @cap: The capability to use
*
* Test to see if the opener of the socket we received the message
* from had when the netlink socket was created and the sender of the
* message has has the capability @cap in the user namespace @user_ns.
*/
bool netlink_ns_capable(const struct sk_buff *skb,
struct user_namespace *user_ns, int cap)
{
return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap);
}
EXPORT_SYMBOL(netlink_ns_capable);
/**
* netlink_capable - Netlink global message capability test
* @skb: socket buffer holding a netlink command from userspace
* @cap: The capability to use
*
* Test to see if the opener of the socket we received the message
* from had when the netlink socket was created and the sender of the
* message has has the capability @cap in all user namespaces.
*/
bool netlink_capable(const struct sk_buff *skb, int cap)
{
return netlink_ns_capable(skb, &init_user_ns, cap);
}
EXPORT_SYMBOL(netlink_capable);
/**
* netlink_net_capable - Netlink network namespace message capability test
* @skb: socket buffer holding a netlink command from userspace
* @cap: The capability to use
*
* Test to see if the opener of the socket we received the message
* from had when the netlink socket was created and the sender of the
* message has has the capability @cap over the network namespace of
* the socket we received the message from.
*/
bool netlink_net_capable(const struct sk_buff *skb, int cap)
{
return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap);
}
EXPORT_SYMBOL(netlink_net_capable);
static inline int netlink_allowed(const struct socket *sock, unsigned int flag)
{
return (nl_table[sock->sk->sk_protocol].flags & flag) ||
ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
}
static void
netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
{
struct netlink_sock *nlk = nlk_sk(sk);
if (nlk->subscriptions && !subscriptions)
__sk_del_bind_node(sk);
else if (!nlk->subscriptions && subscriptions)
sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
nlk->subscriptions = subscriptions;
}
static int netlink_realloc_groups(struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
unsigned int groups;
unsigned long *new_groups;
int err = 0;
netlink_table_grab();
groups = nl_table[sk->sk_protocol].groups;
if (!nl_table[sk->sk_protocol].registered) {
err = -ENOENT;
goto out_unlock;
}
if (nlk->ngroups >= groups)
goto out_unlock;
new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
if (new_groups == NULL) {
err = -ENOMEM;
goto out_unlock;
}
memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
nlk->groups = new_groups;
nlk->ngroups = groups;
out_unlock:
netlink_table_ungrab();
return err;
}
static void netlink_undo_bind(int group, long unsigned int groups,
struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
int undo;
if (!nlk->netlink_unbind)
return;
for (undo = 0; undo < group; undo++)
if (test_bit(undo, &groups))
nlk->netlink_unbind(sock_net(sk), undo + 1);
}
static int netlink_bind(struct socket *sock, struct sockaddr *addr,
int addr_len)
{
struct sock *sk = sock->sk;
struct net *net = sock_net(sk);
struct netlink_sock *nlk = nlk_sk(sk);
struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
int err;
long unsigned int groups = nladdr->nl_groups;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
bool bound;
if (addr_len < sizeof(struct sockaddr_nl))
return -EINVAL;
if (nladdr->nl_family != AF_NETLINK)
return -EINVAL;
/* Only superuser is allowed to listen multicasts */
if (groups) {
if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
return -EPERM;
err = netlink_realloc_groups(sk);
if (err)
return err;
}
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
bound = nlk->bound;
if (bound) {
/* Ensure nlk->portid is up-to-date. */
smp_rmb();
if (nladdr->nl_pid != nlk->portid)
return -EINVAL;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
}
if (nlk->netlink_bind && groups) {
int group;
for (group = 0; group < nlk->ngroups; group++) {
if (!test_bit(group, &groups))
continue;
err = nlk->netlink_bind(net, group + 1);
if (!err)
continue;
netlink_undo_bind(group, groups, sk);
return err;
}
}
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
/* No need for barriers here as we return to user-space without
* using any of the bound attributes.
*/
if (!bound) {
err = nladdr->nl_pid ?
netlink_insert(sk, nladdr->nl_pid) :
netlink_autobind(sock);
if (err) {
netlink_undo_bind(nlk->ngroups, groups, sk);
return err;
}
}
if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
return 0;
netlink_table_grab();
netlink_update_subscriptions(sk, nlk->subscriptions +
hweight32(groups) -
hweight32(nlk->groups[0]));
nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups;
netlink_update_listeners(sk);
netlink_table_ungrab();
return 0;
}
static int netlink_connect(struct socket *sock, struct sockaddr *addr,
int alen, int flags)
{
int err = 0;
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
if (alen < sizeof(addr->sa_family))
return -EINVAL;
if (addr->sa_family == AF_UNSPEC) {
sk->sk_state = NETLINK_UNCONNECTED;
nlk->dst_portid = 0;
nlk->dst_group = 0;
return 0;
}
if (addr->sa_family != AF_NETLINK)
return -EINVAL;
if ((nladdr->nl_groups || nladdr->nl_pid) &&
!netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
return -EPERM;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
/* No need for barriers here as we return to user-space without
* using any of the bound attributes.
*/
if (!nlk->bound)
err = netlink_autobind(sock);
if (err == 0) {
sk->sk_state = NETLINK_CONNECTED;
nlk->dst_portid = nladdr->nl_pid;
nlk->dst_group = ffs(nladdr->nl_groups);
}
return err;
}
static int netlink_getname(struct socket *sock, struct sockaddr *addr,
int *addr_len, int peer)
{
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
nladdr->nl_family = AF_NETLINK;
nladdr->nl_pad = 0;
*addr_len = sizeof(*nladdr);
if (peer) {
nladdr->nl_pid = nlk->dst_portid;
nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
} else {
nladdr->nl_pid = nlk->portid;
nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
}
return 0;
}
static int netlink_ioctl(struct socket *sock, unsigned int cmd,
unsigned long arg)
{
/* try to hand this ioctl down to the NIC drivers.
*/
return -ENOIOCTLCMD;
}
static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
{
struct sock *sock;
struct netlink_sock *nlk;
sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
if (!sock)
return ERR_PTR(-ECONNREFUSED);
/* Don't bother queuing skb if kernel socket has no input function */
nlk = nlk_sk(sock);
if (sock->sk_state == NETLINK_CONNECTED &&
nlk->dst_portid != nlk_sk(ssk)->portid) {
sock_put(sock);
return ERR_PTR(-ECONNREFUSED);
}
return sock;
}
struct sock *netlink_getsockbyfilp(struct file *filp)
{
struct inode *inode = file_inode(filp);
struct sock *sock;
if (!S_ISSOCK(inode->i_mode))
return ERR_PTR(-ENOTSOCK);
sock = SOCKET_I(inode)->sk;
if (sock->sk_family != AF_NETLINK)
return ERR_PTR(-EINVAL);
sock_hold(sock);
return sock;
}
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
static struct sk_buff *netlink_alloc_large_skb(unsigned int size,
int broadcast)
{
struct sk_buff *skb;
void *data;
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
if (size <= NLMSG_GOODSIZE || broadcast)
return alloc_skb(size, GFP_KERNEL);
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
size = SKB_DATA_ALIGN(size) +
SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
data = vmalloc(size);
if (data == NULL)
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
return NULL;
net: fix crash in build_skb() When I added pfmemalloc support in build_skb(), I forgot netlink was using build_skb() with a vmalloc() area. In this patch I introduce __build_skb() for netlink use, and build_skb() is a wrapper handling both skb->head_frag and skb->pfmemalloc This means netlink no longer has to hack skb->head_frag [ 1567.700067] kernel BUG at arch/x86/mm/physaddr.c:26! [ 1567.700067] invalid opcode: 0000 [#1] PREEMPT SMP KASAN [ 1567.700067] Dumping ftrace buffer: [ 1567.700067] (ftrace buffer empty) [ 1567.700067] Modules linked in: [ 1567.700067] CPU: 9 PID: 16186 Comm: trinity-c182 Not tainted 4.0.0-next-20150424-sasha-00037-g4796e21 #2167 [ 1567.700067] task: ffff880127efb000 ti: ffff880246770000 task.ti: ffff880246770000 [ 1567.700067] RIP: __phys_addr (arch/x86/mm/physaddr.c:26 (discriminator 3)) [ 1567.700067] RSP: 0018:ffff8802467779d8 EFLAGS: 00010202 [ 1567.700067] RAX: 000041000ed8e000 RBX: ffffc9008ed8e000 RCX: 000000000000002c [ 1567.700067] RDX: 0000000000000004 RSI: 0000000000000000 RDI: ffffffffb3fd6049 [ 1567.700067] RBP: ffff8802467779f8 R08: 0000000000000019 R09: ffff8801d0168000 [ 1567.700067] R10: ffff8801d01680c7 R11: ffffed003a02d019 R12: ffffc9000ed8e000 [ 1567.700067] R13: 0000000000000f40 R14: 0000000000001180 R15: ffffc9000ed8e000 [ 1567.700067] FS: 00007f2a7da3f700(0000) GS:ffff8801d1000000(0000) knlGS:0000000000000000 [ 1567.700067] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 1567.700067] CR2: 0000000000738308 CR3: 000000022e329000 CR4: 00000000000007e0 [ 1567.700067] Stack: [ 1567.700067] ffffc9000ed8e000 ffff8801d0168000 ffffc9000ed8e000 ffff8801d0168000 [ 1567.700067] ffff880246777a28 ffffffffad7c0a21 0000000000001080 ffff880246777c08 [ 1567.700067] ffff88060d302e68 ffff880246777b58 ffff880246777b88 ffffffffad9a6821 [ 1567.700067] Call Trace: [ 1567.700067] build_skb (include/linux/mm.h:508 net/core/skbuff.c:316) [ 1567.700067] netlink_sendmsg (net/netlink/af_netlink.c:1633 net/netlink/af_netlink.c:2329) [ 1567.774369] ? sched_clock_cpu (kernel/sched/clock.c:311) [ 1567.774369] ? netlink_unicast (net/netlink/af_netlink.c:2273) [ 1567.774369] ? netlink_unicast (net/netlink/af_netlink.c:2273) [ 1567.774369] sock_sendmsg (net/socket.c:614 net/socket.c:623) [ 1567.774369] sock_write_iter (net/socket.c:823) [ 1567.774369] ? sock_sendmsg (net/socket.c:806) [ 1567.774369] __vfs_write (fs/read_write.c:479 fs/read_write.c:491) [ 1567.774369] ? get_lock_stats (kernel/locking/lockdep.c:249) [ 1567.774369] ? default_llseek (fs/read_write.c:487) [ 1567.774369] ? vtime_account_user (kernel/sched/cputime.c:701) [ 1567.774369] ? rw_verify_area (fs/read_write.c:406 (discriminator 4)) [ 1567.774369] vfs_write (fs/read_write.c:539) [ 1567.774369] SyS_write (fs/read_write.c:586 fs/read_write.c:577) [ 1567.774369] ? SyS_read (fs/read_write.c:577) [ 1567.774369] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 1567.774369] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 1567.774369] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 1567.774369] system_call_fastpath (arch/x86/kernel/entry_64.S:261) Fixes: 79930f5892e ("net: do not deplete pfmemalloc reserve") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Sasha Levin <sasha.levin@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-04-24 23:05:01 +00:00
skb = __build_skb(data, size);
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
if (skb == NULL)
vfree(data);
net: fix crash in build_skb() When I added pfmemalloc support in build_skb(), I forgot netlink was using build_skb() with a vmalloc() area. In this patch I introduce __build_skb() for netlink use, and build_skb() is a wrapper handling both skb->head_frag and skb->pfmemalloc This means netlink no longer has to hack skb->head_frag [ 1567.700067] kernel BUG at arch/x86/mm/physaddr.c:26! [ 1567.700067] invalid opcode: 0000 [#1] PREEMPT SMP KASAN [ 1567.700067] Dumping ftrace buffer: [ 1567.700067] (ftrace buffer empty) [ 1567.700067] Modules linked in: [ 1567.700067] CPU: 9 PID: 16186 Comm: trinity-c182 Not tainted 4.0.0-next-20150424-sasha-00037-g4796e21 #2167 [ 1567.700067] task: ffff880127efb000 ti: ffff880246770000 task.ti: ffff880246770000 [ 1567.700067] RIP: __phys_addr (arch/x86/mm/physaddr.c:26 (discriminator 3)) [ 1567.700067] RSP: 0018:ffff8802467779d8 EFLAGS: 00010202 [ 1567.700067] RAX: 000041000ed8e000 RBX: ffffc9008ed8e000 RCX: 000000000000002c [ 1567.700067] RDX: 0000000000000004 RSI: 0000000000000000 RDI: ffffffffb3fd6049 [ 1567.700067] RBP: ffff8802467779f8 R08: 0000000000000019 R09: ffff8801d0168000 [ 1567.700067] R10: ffff8801d01680c7 R11: ffffed003a02d019 R12: ffffc9000ed8e000 [ 1567.700067] R13: 0000000000000f40 R14: 0000000000001180 R15: ffffc9000ed8e000 [ 1567.700067] FS: 00007f2a7da3f700(0000) GS:ffff8801d1000000(0000) knlGS:0000000000000000 [ 1567.700067] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 1567.700067] CR2: 0000000000738308 CR3: 000000022e329000 CR4: 00000000000007e0 [ 1567.700067] Stack: [ 1567.700067] ffffc9000ed8e000 ffff8801d0168000 ffffc9000ed8e000 ffff8801d0168000 [ 1567.700067] ffff880246777a28 ffffffffad7c0a21 0000000000001080 ffff880246777c08 [ 1567.700067] ffff88060d302e68 ffff880246777b58 ffff880246777b88 ffffffffad9a6821 [ 1567.700067] Call Trace: [ 1567.700067] build_skb (include/linux/mm.h:508 net/core/skbuff.c:316) [ 1567.700067] netlink_sendmsg (net/netlink/af_netlink.c:1633 net/netlink/af_netlink.c:2329) [ 1567.774369] ? sched_clock_cpu (kernel/sched/clock.c:311) [ 1567.774369] ? netlink_unicast (net/netlink/af_netlink.c:2273) [ 1567.774369] ? netlink_unicast (net/netlink/af_netlink.c:2273) [ 1567.774369] sock_sendmsg (net/socket.c:614 net/socket.c:623) [ 1567.774369] sock_write_iter (net/socket.c:823) [ 1567.774369] ? sock_sendmsg (net/socket.c:806) [ 1567.774369] __vfs_write (fs/read_write.c:479 fs/read_write.c:491) [ 1567.774369] ? get_lock_stats (kernel/locking/lockdep.c:249) [ 1567.774369] ? default_llseek (fs/read_write.c:487) [ 1567.774369] ? vtime_account_user (kernel/sched/cputime.c:701) [ 1567.774369] ? rw_verify_area (fs/read_write.c:406 (discriminator 4)) [ 1567.774369] vfs_write (fs/read_write.c:539) [ 1567.774369] SyS_write (fs/read_write.c:586 fs/read_write.c:577) [ 1567.774369] ? SyS_read (fs/read_write.c:577) [ 1567.774369] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63) [ 1567.774369] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2594 kernel/locking/lockdep.c:2636) [ 1567.774369] ? trace_hardirqs_on_thunk (arch/x86/lib/thunk_64.S:42) [ 1567.774369] system_call_fastpath (arch/x86/kernel/entry_64.S:261) Fixes: 79930f5892e ("net: do not deplete pfmemalloc reserve") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Sasha Levin <sasha.levin@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-04-24 23:05:01 +00:00
else
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
skb->destructor = netlink_skb_destructor;
return skb;
}
/*
* Attach a skb to a netlink socket.
* The caller must hold a reference to the destination socket. On error, the
* reference is dropped. The skb is not send to the destination, just all
* all error checks are performed and memory in the queue is reserved.
* Return values:
* < 0: error. skb freed, reference to sock dropped.
* 0: continue
* 1: repeat lookup - reference dropped while waiting for socket memory.
*/
int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
long *timeo, struct sock *ssk)
{
struct netlink_sock *nlk;
nlk = nlk_sk(sk);
if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
test_bit(NETLINK_S_CONGESTED, &nlk->state))) {
DECLARE_WAITQUEUE(wait, current);
if (!*timeo) {
if (!ssk || netlink_is_kernel(ssk))
netlink_overrun(sk);
sock_put(sk);
kfree_skb(skb);
return -EAGAIN;
}
__set_current_state(TASK_INTERRUPTIBLE);
add_wait_queue(&nlk->wait, &wait);
if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
test_bit(NETLINK_S_CONGESTED, &nlk->state)) &&
!sock_flag(sk, SOCK_DEAD))
*timeo = schedule_timeout(*timeo);
__set_current_state(TASK_RUNNING);
remove_wait_queue(&nlk->wait, &wait);
sock_put(sk);
if (signal_pending(current)) {
kfree_skb(skb);
return sock_intr_errno(*timeo);
}
return 1;
}
netlink_skb_set_owner_r(skb, sk);
return 0;
}
static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
{
int len = skb->len;
netlink_deliver_tap(skb);
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
skb_queue_tail(&sk->sk_receive_queue, skb);
sk->sk_data_ready(sk);
return len;
}
int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
{
int len = __netlink_sendskb(sk, skb);
sock_put(sk);
return len;
}
void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
{
kfree_skb(skb);
sock_put(sk);
}
static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
{
int delta;
WARN_ON(skb->sk != NULL);
delta = skb->end - skb->tail;
if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize)
return skb;
if (skb_shared(skb)) {
struct sk_buff *nskb = skb_clone(skb, allocation);
if (!nskb)
return skb;
consume_skb(skb);
skb = nskb;
}
pskb_expand_head(skb, 0, -delta,
(allocation & ~__GFP_DIRECT_RECLAIM) |
__GFP_NOWARN | __GFP_NORETRY);
return skb;
}
static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
struct sock *ssk)
{
int ret;
struct netlink_sock *nlk = nlk_sk(sk);
ret = -ECONNREFUSED;
if (nlk->netlink_rcv != NULL) {
ret = skb->len;
netlink_skb_set_owner_r(skb, sk);
NETLINK_CB(skb).sk = ssk;
netlink_deliver_tap_kernel(sk, ssk, skb);
nlk->netlink_rcv(skb);
consume_skb(skb);
} else {
kfree_skb(skb);
}
sock_put(sk);
return ret;
}
int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
u32 portid, int nonblock)
{
struct sock *sk;
int err;
long timeo;
skb = netlink_trim(skb, gfp_any());
timeo = sock_sndtimeo(ssk, nonblock);
retry:
sk = netlink_getsockbyportid(ssk, portid);
if (IS_ERR(sk)) {
kfree_skb(skb);
return PTR_ERR(sk);
}
if (netlink_is_kernel(sk))
return netlink_unicast_kernel(sk, skb, ssk);
if (sk_filter(sk, skb)) {
err = skb->len;
kfree_skb(skb);
sock_put(sk);
return err;
}
err = netlink_attachskb(sk, skb, &timeo, ssk);
if (err == 1)
goto retry;
if (err)
return err;
return netlink_sendskb(sk, skb);
}
EXPORT_SYMBOL(netlink_unicast);
int netlink_has_listeners(struct sock *sk, unsigned int group)
{
int res = 0;
struct listeners *listeners;
BUG_ON(!netlink_is_kernel(sk));
rcu_read_lock();
listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
res = test_bit(group - 1, listeners->masks);
rcu_read_unlock();
return res;
}
EXPORT_SYMBOL_GPL(netlink_has_listeners);
static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
{
struct netlink_sock *nlk = nlk_sk(sk);
if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
!test_bit(NETLINK_S_CONGESTED, &nlk->state)) {
netlink_skb_set_owner_r(skb, sk);
__netlink_sendskb(sk, skb);
return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
}
return -1;
}
struct netlink_broadcast_data {
struct sock *exclude_sk;
struct net *net;
u32 portid;
u32 group;
int failure;
netlink: change return-value logic of netlink_broadcast() Currently, netlink_broadcast() reports errors to the caller if no messages at all were delivered: 1) If, at least, one message has been delivered correctly, returns 0. 2) Otherwise, if no messages at all were delivered due to skb_clone() failure, return -ENOBUFS. 3) Otherwise, if there are no listeners, return -ESRCH. With this patch, the caller knows if the delivery of any of the messages to the listeners have failed: 1) If it fails to deliver any message (for whatever reason), return -ENOBUFS. 2) Otherwise, if all messages were delivered OK, returns 0. 3) Otherwise, if no listeners, return -ESRCH. In the current ctnetlink code and in Netfilter in general, we can add reliable logging and connection tracking event delivery by dropping the packets whose events were not successfully delivered over Netlink. Of course, this option would be settable via /proc as this approach reduces performance (in terms of filtered connections per seconds by a stateful firewall) but providing reliable logging and event delivery (for conntrackd) in return. This patch also changes some clients of netlink_broadcast() that may report ENOBUFS errors via printk. This error handling is not of any help. Instead, the userspace daemons that are listening to those netlink messages should resync themselves with the kernel-side if they hit ENOBUFS. BTW, netlink_broadcast() clients include those that call cn_netlink_send(), nlmsg_multicast() and genlmsg_multicast() since they internally call netlink_broadcast() and return its error value. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 07:56:36 +00:00
int delivery_failure;
int congested;
int delivered;
gfp_t allocation;
struct sk_buff *skb, *skb2;
int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
void *tx_data;
};
static void do_one_broadcast(struct sock *sk,
struct netlink_broadcast_data *p)
{
struct netlink_sock *nlk = nlk_sk(sk);
int val;
if (p->exclude_sk == sk)
return;
if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
!test_bit(p->group - 1, nlk->groups))
return;
if (!net_eq(sock_net(sk), p->net)) {
if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID))
return;
if (!peernet_has_id(sock_net(sk), p->net))
return;
if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns,
CAP_NET_BROADCAST))
return;
}
if (p->failure) {
netlink_overrun(sk);
return;
}
sock_hold(sk);
if (p->skb2 == NULL) {
if (skb_shared(p->skb)) {
p->skb2 = skb_clone(p->skb, p->allocation);
} else {
p->skb2 = skb_get(p->skb);
/*
* skb ownership may have been set when
* delivered to a previous socket.
*/
skb_orphan(p->skb2);
}
}
if (p->skb2 == NULL) {
netlink_overrun(sk);
/* Clone failed. Notify ALL listeners. */
p->failure = 1;
if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR)
p->delivery_failure = 1;
goto out;
}
if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
kfree_skb(p->skb2);
p->skb2 = NULL;
goto out;
}
if (sk_filter(sk, p->skb2)) {
kfree_skb(p->skb2);
p->skb2 = NULL;
goto out;
}
NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net);
if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED)
NETLINK_CB(p->skb2).nsid_is_set = true;
val = netlink_broadcast_deliver(sk, p->skb2);
if (val < 0) {
netlink_overrun(sk);
if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR)
p->delivery_failure = 1;
} else {
p->congested |= val;
p->delivered = 1;
p->skb2 = NULL;
}
out:
sock_put(sk);
}
int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
u32 group, gfp_t allocation,
int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
void *filter_data)
{
struct net *net = sock_net(ssk);
struct netlink_broadcast_data info;
struct sock *sk;
skb = netlink_trim(skb, allocation);
info.exclude_sk = ssk;
info.net = net;
info.portid = portid;
info.group = group;
info.failure = 0;
netlink: change return-value logic of netlink_broadcast() Currently, netlink_broadcast() reports errors to the caller if no messages at all were delivered: 1) If, at least, one message has been delivered correctly, returns 0. 2) Otherwise, if no messages at all were delivered due to skb_clone() failure, return -ENOBUFS. 3) Otherwise, if there are no listeners, return -ESRCH. With this patch, the caller knows if the delivery of any of the messages to the listeners have failed: 1) If it fails to deliver any message (for whatever reason), return -ENOBUFS. 2) Otherwise, if all messages were delivered OK, returns 0. 3) Otherwise, if no listeners, return -ESRCH. In the current ctnetlink code and in Netfilter in general, we can add reliable logging and connection tracking event delivery by dropping the packets whose events were not successfully delivered over Netlink. Of course, this option would be settable via /proc as this approach reduces performance (in terms of filtered connections per seconds by a stateful firewall) but providing reliable logging and event delivery (for conntrackd) in return. This patch also changes some clients of netlink_broadcast() that may report ENOBUFS errors via printk. This error handling is not of any help. Instead, the userspace daemons that are listening to those netlink messages should resync themselves with the kernel-side if they hit ENOBUFS. BTW, netlink_broadcast() clients include those that call cn_netlink_send(), nlmsg_multicast() and genlmsg_multicast() since they internally call netlink_broadcast() and return its error value. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 07:56:36 +00:00
info.delivery_failure = 0;
info.congested = 0;
info.delivered = 0;
info.allocation = allocation;
info.skb = skb;
info.skb2 = NULL;
info.tx_filter = filter;
info.tx_data = filter_data;
/* While we sleep in clone, do not allow to change socket list */
netlink_lock_table();
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
do_one_broadcast(sk, &info);
consume_skb(skb);
netlink_unlock_table();
if (info.delivery_failure) {
kfree_skb(info.skb2);
netlink: change return-value logic of netlink_broadcast() Currently, netlink_broadcast() reports errors to the caller if no messages at all were delivered: 1) If, at least, one message has been delivered correctly, returns 0. 2) Otherwise, if no messages at all were delivered due to skb_clone() failure, return -ENOBUFS. 3) Otherwise, if there are no listeners, return -ESRCH. With this patch, the caller knows if the delivery of any of the messages to the listeners have failed: 1) If it fails to deliver any message (for whatever reason), return -ENOBUFS. 2) Otherwise, if all messages were delivered OK, returns 0. 3) Otherwise, if no listeners, return -ESRCH. In the current ctnetlink code and in Netfilter in general, we can add reliable logging and connection tracking event delivery by dropping the packets whose events were not successfully delivered over Netlink. Of course, this option would be settable via /proc as this approach reduces performance (in terms of filtered connections per seconds by a stateful firewall) but providing reliable logging and event delivery (for conntrackd) in return. This patch also changes some clients of netlink_broadcast() that may report ENOBUFS errors via printk. This error handling is not of any help. Instead, the userspace daemons that are listening to those netlink messages should resync themselves with the kernel-side if they hit ENOBUFS. BTW, netlink_broadcast() clients include those that call cn_netlink_send(), nlmsg_multicast() and genlmsg_multicast() since they internally call netlink_broadcast() and return its error value. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 07:56:36 +00:00
return -ENOBUFS;
}
consume_skb(info.skb2);
netlink: change return-value logic of netlink_broadcast() Currently, netlink_broadcast() reports errors to the caller if no messages at all were delivered: 1) If, at least, one message has been delivered correctly, returns 0. 2) Otherwise, if no messages at all were delivered due to skb_clone() failure, return -ENOBUFS. 3) Otherwise, if there are no listeners, return -ESRCH. With this patch, the caller knows if the delivery of any of the messages to the listeners have failed: 1) If it fails to deliver any message (for whatever reason), return -ENOBUFS. 2) Otherwise, if all messages were delivered OK, returns 0. 3) Otherwise, if no listeners, return -ESRCH. In the current ctnetlink code and in Netfilter in general, we can add reliable logging and connection tracking event delivery by dropping the packets whose events were not successfully delivered over Netlink. Of course, this option would be settable via /proc as this approach reduces performance (in terms of filtered connections per seconds by a stateful firewall) but providing reliable logging and event delivery (for conntrackd) in return. This patch also changes some clients of netlink_broadcast() that may report ENOBUFS errors via printk. This error handling is not of any help. Instead, the userspace daemons that are listening to those netlink messages should resync themselves with the kernel-side if they hit ENOBUFS. BTW, netlink_broadcast() clients include those that call cn_netlink_send(), nlmsg_multicast() and genlmsg_multicast() since they internally call netlink_broadcast() and return its error value. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 07:56:36 +00:00
if (info.delivered) {
mm, page_alloc: distinguish between being unable to sleep, unwilling to sleep and avoiding waking kswapd __GFP_WAIT has been used to identify atomic context in callers that hold spinlocks or are in interrupts. They are expected to be high priority and have access one of two watermarks lower than "min" which can be referred to as the "atomic reserve". __GFP_HIGH users get access to the first lower watermark and can be called the "high priority reserve". Over time, callers had a requirement to not block when fallback options were available. Some have abused __GFP_WAIT leading to a situation where an optimisitic allocation with a fallback option can access atomic reserves. This patch uses __GFP_ATOMIC to identify callers that are truely atomic, cannot sleep and have no alternative. High priority users continue to use __GFP_HIGH. __GFP_DIRECT_RECLAIM identifies callers that can sleep and are willing to enter direct reclaim. __GFP_KSWAPD_RECLAIM to identify callers that want to wake kswapd for background reclaim. __GFP_WAIT is redefined as a caller that is willing to enter direct reclaim and wake kswapd for background reclaim. This patch then converts a number of sites o __GFP_ATOMIC is used by callers that are high priority and have memory pools for those requests. GFP_ATOMIC uses this flag. o Callers that have a limited mempool to guarantee forward progress clear __GFP_DIRECT_RECLAIM but keep __GFP_KSWAPD_RECLAIM. bio allocations fall into this category where kswapd will still be woken but atomic reserves are not used as there is a one-entry mempool to guarantee progress. o Callers that are checking if they are non-blocking should use the helper gfpflags_allow_blocking() where possible. This is because checking for __GFP_WAIT as was done historically now can trigger false positives. Some exceptions like dm-crypt.c exist where the code intent is clearer if __GFP_DIRECT_RECLAIM is used instead of the helper due to flag manipulations. o Callers that built their own GFP flags instead of starting with GFP_KERNEL and friends now also need to specify __GFP_KSWAPD_RECLAIM. The first key hazard to watch out for is callers that removed __GFP_WAIT and was depending on access to atomic reserves for inconspicuous reasons. In some cases it may be appropriate for them to use __GFP_HIGH. The second key hazard is callers that assembled their own combination of GFP flags instead of starting with something like GFP_KERNEL. They may now wish to specify __GFP_KSWAPD_RECLAIM. It's almost certainly harmless if it's missed in most cases as other activity will wake kswapd. Signed-off-by: Mel Gorman <mgorman@techsingularity.net> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Michal Hocko <mhocko@suse.com> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Christoph Lameter <cl@linux.com> Cc: David Rientjes <rientjes@google.com> Cc: Vitaly Wool <vitalywool@gmail.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-07 00:28:21 +00:00
if (info.congested && gfpflags_allow_blocking(allocation))
yield();
return 0;
}
return -ESRCH;
}
EXPORT_SYMBOL(netlink_broadcast_filtered);
int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
u32 group, gfp_t allocation)
{
return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
NULL, NULL);
}
EXPORT_SYMBOL(netlink_broadcast);
struct netlink_set_err_data {
struct sock *exclude_sk;
u32 portid;
u32 group;
int code;
};
static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
{
struct netlink_sock *nlk = nlk_sk(sk);
int ret = 0;
if (sk == p->exclude_sk)
goto out;
if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
goto out;
if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
!test_bit(p->group - 1, nlk->groups))
goto out;
if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) {
ret = 1;
goto out;
}
sk->sk_err = p->code;
sk->sk_error_report(sk);
out:
return ret;
}
/**
* netlink_set_err - report error to broadcast listeners
* @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
* @portid: the PORTID of a process that we want to skip (if any)
* @group: the broadcast group that will notice the error
* @code: error code, must be negative (as usual in kernelspace)
*
* This function returns the number of broadcast listeners that have set the
* NETLINK_NO_ENOBUFS socket option.
*/
int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
{
struct netlink_set_err_data info;
struct sock *sk;
int ret = 0;
info.exclude_sk = ssk;
info.portid = portid;
info.group = group;
/* sk->sk_err wants a positive error value */
info.code = -code;
read_lock(&nl_table_lock);
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
ret += do_one_set_err(sk, &info);
read_unlock(&nl_table_lock);
return ret;
}
EXPORT_SYMBOL(netlink_set_err);
/* must be called with netlink table grabbed */
static void netlink_update_socket_mc(struct netlink_sock *nlk,
unsigned int group,
int is_new)
{
int old, new = !!is_new, subscriptions;
old = test_bit(group - 1, nlk->groups);
subscriptions = nlk->subscriptions - old + new;
if (new)
__set_bit(group - 1, nlk->groups);
else
__clear_bit(group - 1, nlk->groups);
netlink_update_subscriptions(&nlk->sk, subscriptions);
netlink_update_listeners(&nlk->sk);
}
static int netlink_setsockopt(struct socket *sock, int level, int optname,
char __user *optval, unsigned int optlen)
{
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
unsigned int val = 0;
int err;
if (level != SOL_NETLINK)
return -ENOPROTOOPT;
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
if (optlen >= sizeof(int) &&
get_user(val, (unsigned int __user *)optval))
return -EFAULT;
switch (optname) {
case NETLINK_PKTINFO:
if (val)
nlk->flags |= NETLINK_F_RECV_PKTINFO;
else
nlk->flags &= ~NETLINK_F_RECV_PKTINFO;
err = 0;
break;
case NETLINK_ADD_MEMBERSHIP:
case NETLINK_DROP_MEMBERSHIP: {
if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
return -EPERM;
err = netlink_realloc_groups(sk);
if (err)
return err;
if (!val || val - 1 >= nlk->ngroups)
return -EINVAL;
if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) {
err = nlk->netlink_bind(sock_net(sk), val);
if (err)
return err;
}
netlink_table_grab();
netlink_update_socket_mc(nlk, val,
optname == NETLINK_ADD_MEMBERSHIP);
netlink_table_ungrab();
if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind)
nlk->netlink_unbind(sock_net(sk), val);
err = 0;
break;
}
case NETLINK_BROADCAST_ERROR:
if (val)
nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR;
else
nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR;
err = 0;
break;
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
case NETLINK_NO_ENOBUFS:
if (val) {
nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS;
clear_bit(NETLINK_S_CONGESTED, &nlk->state);
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
wake_up_interruptible(&nlk->wait);
} else {
nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS;
}
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
err = 0;
break;
case NETLINK_LISTEN_ALL_NSID:
if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST))
return -EPERM;
if (val)
nlk->flags |= NETLINK_F_LISTEN_ALL_NSID;
else
nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID;
err = 0;
break;
case NETLINK_CAP_ACK:
if (val)
nlk->flags |= NETLINK_F_CAP_ACK;
else
nlk->flags &= ~NETLINK_F_CAP_ACK;
err = 0;
break;
case NETLINK_EXT_ACK:
if (val)
nlk->flags |= NETLINK_F_EXT_ACK;
else
nlk->flags &= ~NETLINK_F_EXT_ACK;
err = 0;
break;
default:
err = -ENOPROTOOPT;
}
return err;
}
static int netlink_getsockopt(struct socket *sock, int level, int optname,
char __user *optval, int __user *optlen)
{
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
int len, val, err;
if (level != SOL_NETLINK)
return -ENOPROTOOPT;
if (get_user(len, optlen))
return -EFAULT;
if (len < 0)
return -EINVAL;
switch (optname) {
case NETLINK_PKTINFO:
if (len < sizeof(int))
return -EINVAL;
len = sizeof(int);
val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0;
if (put_user(len, optlen) ||
put_user(val, optval))
return -EFAULT;
err = 0;
break;
case NETLINK_BROADCAST_ERROR:
if (len < sizeof(int))
return -EINVAL;
len = sizeof(int);
val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0;
if (put_user(len, optlen) ||
put_user(val, optval))
return -EFAULT;
err = 0;
break;
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
case NETLINK_NO_ENOBUFS:
if (len < sizeof(int))
return -EINVAL;
len = sizeof(int);
val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0;
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
if (put_user(len, optlen) ||
put_user(val, optval))
return -EFAULT;
err = 0;
break;
case NETLINK_LIST_MEMBERSHIPS: {
int pos, idx, shift;
err = 0;
netlink_lock_table();
for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) {
if (len - pos < sizeof(u32))
break;
idx = pos / sizeof(unsigned long);
shift = (pos % sizeof(unsigned long)) * 8;
if (put_user((u32)(nlk->groups[idx] >> shift),
(u32 __user *)(optval + pos))) {
err = -EFAULT;
break;
}
}
if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen))
err = -EFAULT;
netlink_unlock_table();
break;
}
case NETLINK_CAP_ACK:
if (len < sizeof(int))
return -EINVAL;
len = sizeof(int);
val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0;
if (put_user(len, optlen) ||
put_user(val, optval))
return -EFAULT;
err = 0;
break;
case NETLINK_EXT_ACK:
if (len < sizeof(int))
return -EINVAL;
len = sizeof(int);
val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0;
if (put_user(len, optlen) || put_user(val, optval))
return -EFAULT;
err = 0;
break;
default:
err = -ENOPROTOOPT;
}
return err;
}
static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
{
struct nl_pktinfo info;
info.group = NETLINK_CB(skb).dst_group;
put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
}
static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg,
struct sk_buff *skb)
{
if (!NETLINK_CB(skb).nsid_is_set)
return;
put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int),
&NETLINK_CB(skb).nsid);
}
static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
{
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
u32 dst_portid;
u32 dst_group;
struct sk_buff *skb;
int err;
struct scm_cookie scm;
netlink: Only check file credentials for implicit destinations It was possible to get a setuid root or setcap executable to write to it's stdout or stderr (which has been set made a netlink socket) and inadvertently reconfigure the networking stack. To prevent this we check that both the creator of the socket and the currentl applications has permission to reconfigure the network stack. Unfortunately this breaks Zebra which always uses sendto/sendmsg and creates it's socket without any privileges. To keep Zebra working don't bother checking if the creator of the socket has privilege when a destination address is specified. Instead rely exclusively on the privileges of the sender of the socket. Note from Andy: This is exactly Eric's code except for some comment clarifications and formatting fixes. Neither I nor, I think, anyone else is thrilled with this approach, but I'm hesitant to wait on a better fix since 3.15 is almost here. Note to stable maintainers: This is a mess. An earlier series of patches in 3.15 fix a rather serious security issue (CVE-2014-0181), but they did so in a way that breaks Zebra. The offending series includes: commit aa4cf9452f469f16cea8c96283b641b4576d4a7b Author: Eric W. Biederman <ebiederm@xmission.com> Date: Wed Apr 23 14:28:03 2014 -0700 net: Add variants of capable for use on netlink messages If a given kernel version is missing that series of fixes, it's probably worth backporting it and this patch. if that series is present, then this fix is critical if you care about Zebra. Cc: stable@vger.kernel.org Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-05-30 18:04:00 +00:00
u32 netlink_skb_flags = 0;
if (msg->msg_flags&MSG_OOB)
return -EOPNOTSUPP;
err = scm_send(sock, msg, &scm, true);
if (err < 0)
return err;
if (msg->msg_namelen) {
err = -EINVAL;
if (addr->nl_family != AF_NETLINK)
goto out;
dst_portid = addr->nl_pid;
dst_group = ffs(addr->nl_groups);
err = -EPERM;
if ((dst_group || dst_portid) &&
!netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
goto out;
netlink: Only check file credentials for implicit destinations It was possible to get a setuid root or setcap executable to write to it's stdout or stderr (which has been set made a netlink socket) and inadvertently reconfigure the networking stack. To prevent this we check that both the creator of the socket and the currentl applications has permission to reconfigure the network stack. Unfortunately this breaks Zebra which always uses sendto/sendmsg and creates it's socket without any privileges. To keep Zebra working don't bother checking if the creator of the socket has privilege when a destination address is specified. Instead rely exclusively on the privileges of the sender of the socket. Note from Andy: This is exactly Eric's code except for some comment clarifications and formatting fixes. Neither I nor, I think, anyone else is thrilled with this approach, but I'm hesitant to wait on a better fix since 3.15 is almost here. Note to stable maintainers: This is a mess. An earlier series of patches in 3.15 fix a rather serious security issue (CVE-2014-0181), but they did so in a way that breaks Zebra. The offending series includes: commit aa4cf9452f469f16cea8c96283b641b4576d4a7b Author: Eric W. Biederman <ebiederm@xmission.com> Date: Wed Apr 23 14:28:03 2014 -0700 net: Add variants of capable for use on netlink messages If a given kernel version is missing that series of fixes, it's probably worth backporting it and this patch. if that series is present, then this fix is critical if you care about Zebra. Cc: stable@vger.kernel.org Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-05-30 18:04:00 +00:00
netlink_skb_flags |= NETLINK_SKB_DST;
} else {
dst_portid = nlk->dst_portid;
dst_group = nlk->dst_group;
}
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
if (!nlk->bound) {
err = netlink_autobind(sock);
if (err)
goto out;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
} else {
/* Ensure nlk is hashed and visible. */
smp_rmb();
}
err = -EMSGSIZE;
if (len > sk->sk_sndbuf - 32)
goto out;
err = -ENOBUFS;
netlink: fix splat in skb_clone with large messages Since (c05cdb1 netlink: allow large data transfers from user-space), netlink splats if it invokes skb_clone on large netlink skbs since: * skb_shared_info was not correctly initialized. * skb->destructor is not set in the cloned skb. This was spotted by trinity: [ 894.990671] BUG: unable to handle kernel paging request at ffffc9000047b001 [ 894.991034] IP: [<ffffffff81a212c4>] skb_clone+0x24/0xc0 [...] [ 894.991034] Call Trace: [ 894.991034] [<ffffffff81ad299a>] nl_fib_input+0x6a/0x240 [ 894.991034] [<ffffffff81c3b7e6>] ? _raw_read_unlock+0x26/0x40 [ 894.991034] [<ffffffff81a5f189>] netlink_unicast+0x169/0x1e0 [ 894.991034] [<ffffffff81a601e1>] netlink_sendmsg+0x251/0x3d0 Fix it by: 1) introducing a new netlink_skb_clone function that is used in nl_fib_input, that sets our special skb->destructor in the cloned skb. Moreover, handle the release of the large cloned skb head area in the destructor path. 2) not allowing large skbuffs in the netlink broadcast path. I cannot find any reasonable use of the large data transfer using netlink in that path, moreover this helps to skip extra skb_clone handling. I found two more netlink clients that are cloning the skbs, but they are not in the sendmsg path. Therefore, the sole client cloning that I found seems to be the fib frontend. Thanks to Eric Dumazet for helping to address this issue. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-28 01:04:23 +00:00
skb = netlink_alloc_large_skb(len, dst_group);
if (skb == NULL)
goto out;
NETLINK_CB(skb).portid = nlk->portid;
NETLINK_CB(skb).dst_group = dst_group;
NETLINK_CB(skb).creds = scm.creds;
netlink: Only check file credentials for implicit destinations It was possible to get a setuid root or setcap executable to write to it's stdout or stderr (which has been set made a netlink socket) and inadvertently reconfigure the networking stack. To prevent this we check that both the creator of the socket and the currentl applications has permission to reconfigure the network stack. Unfortunately this breaks Zebra which always uses sendto/sendmsg and creates it's socket without any privileges. To keep Zebra working don't bother checking if the creator of the socket has privilege when a destination address is specified. Instead rely exclusively on the privileges of the sender of the socket. Note from Andy: This is exactly Eric's code except for some comment clarifications and formatting fixes. Neither I nor, I think, anyone else is thrilled with this approach, but I'm hesitant to wait on a better fix since 3.15 is almost here. Note to stable maintainers: This is a mess. An earlier series of patches in 3.15 fix a rather serious security issue (CVE-2014-0181), but they did so in a way that breaks Zebra. The offending series includes: commit aa4cf9452f469f16cea8c96283b641b4576d4a7b Author: Eric W. Biederman <ebiederm@xmission.com> Date: Wed Apr 23 14:28:03 2014 -0700 net: Add variants of capable for use on netlink messages If a given kernel version is missing that series of fixes, it's probably worth backporting it and this patch. if that series is present, then this fix is critical if you care about Zebra. Cc: stable@vger.kernel.org Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-05-30 18:04:00 +00:00
NETLINK_CB(skb).flags = netlink_skb_flags;
err = -EFAULT;
if (memcpy_from_msg(skb_put(skb, len), msg, len)) {
kfree_skb(skb);
goto out;
}
err = security_netlink_send(sk, skb);
if (err) {
kfree_skb(skb);
goto out;
}
if (dst_group) {
atomic_inc(&skb->users);
netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
}
err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT);
out:
scm_destroy(&scm);
return err;
}
static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
int flags)
{
struct scm_cookie scm;
struct sock *sk = sock->sk;
struct netlink_sock *nlk = nlk_sk(sk);
int noblock = flags&MSG_DONTWAIT;
size_t copied;
struct sk_buff *skb, *data_skb;
int err, ret;
if (flags&MSG_OOB)
return -EOPNOTSUPP;
copied = 0;
skb = skb_recv_datagram(sk, flags, noblock, &err);
if (skb == NULL)
goto out;
data_skb = skb;
net/compat/wext: send different messages to compat tasks Wireless extensions have the unfortunate problem that events are multicast netlink messages, and are not independent of pointer size. Thus, currently 32-bit tasks on 64-bit platforms cannot properly receive events and fail with all kinds of strange problems, for instance wpa_supplicant never notices disassociations, due to the way the 64-bit event looks (to a 32-bit process), the fact that the address is all zeroes is lost, it thinks instead it is 00:00:00:00:01:00. The same problem existed with the ioctls, until David Miller fixed those some time ago in an heroic effort. A different problem caused by this is that we cannot send the ASSOCREQIE/ASSOCRESPIE events because sending them causes a 32-bit wpa_supplicant on a 64-bit system to overwrite its internal information, which is worse than it not getting the information at all -- so we currently resort to sending a custom string event that it then parses. This, however, has a severe size limitation we are frequently hitting with modern access points; this limitation would can be lifted after this patch by sending the correct binary, not custom, event. A similar problem apparently happens for some other netlink users on x86_64 with 32-bit tasks due to the alignment for 64-bit quantities. In order to fix these problems, I have implemented a way to send compat messages to tasks. When sending an event, we send the non-compat event data together with a compat event data in skb_shinfo(main_skb)->frag_list. Then, when the event is read from the socket, the netlink code makes sure to pass out only the skb that is compatible with the task. This approach was suggested by David Miller, my original approach required always sending two skbs but that had various small problems. To determine whether compat is needed or not, I have used the MSG_CMSG_COMPAT flag, and adjusted the call path for recv and recvfrom to include it, even if those calls do not have a cmsg parameter. I have not solved one small part of the problem, and I don't think it is necessary to: if a 32-bit application uses read() rather than any form of recvmsg() it will still get the wrong (64-bit) event. However, neither do applications actually do this, nor would it be a regression. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-01 11:26:02 +00:00
#ifdef CONFIG_COMPAT_NETLINK_MESSAGES
if (unlikely(skb_shinfo(skb)->frag_list)) {
/*
* If this skb has a frag_list, then here that means that we
* will have to use the frag_list skb's data for compat tasks
* and the regular skb's data for normal (non-compat) tasks.
net/compat/wext: send different messages to compat tasks Wireless extensions have the unfortunate problem that events are multicast netlink messages, and are not independent of pointer size. Thus, currently 32-bit tasks on 64-bit platforms cannot properly receive events and fail with all kinds of strange problems, for instance wpa_supplicant never notices disassociations, due to the way the 64-bit event looks (to a 32-bit process), the fact that the address is all zeroes is lost, it thinks instead it is 00:00:00:00:01:00. The same problem existed with the ioctls, until David Miller fixed those some time ago in an heroic effort. A different problem caused by this is that we cannot send the ASSOCREQIE/ASSOCRESPIE events because sending them causes a 32-bit wpa_supplicant on a 64-bit system to overwrite its internal information, which is worse than it not getting the information at all -- so we currently resort to sending a custom string event that it then parses. This, however, has a severe size limitation we are frequently hitting with modern access points; this limitation would can be lifted after this patch by sending the correct binary, not custom, event. A similar problem apparently happens for some other netlink users on x86_64 with 32-bit tasks due to the alignment for 64-bit quantities. In order to fix these problems, I have implemented a way to send compat messages to tasks. When sending an event, we send the non-compat event data together with a compat event data in skb_shinfo(main_skb)->frag_list. Then, when the event is read from the socket, the netlink code makes sure to pass out only the skb that is compatible with the task. This approach was suggested by David Miller, my original approach required always sending two skbs but that had various small problems. To determine whether compat is needed or not, I have used the MSG_CMSG_COMPAT flag, and adjusted the call path for recv and recvfrom to include it, even if those calls do not have a cmsg parameter. I have not solved one small part of the problem, and I don't think it is necessary to: if a 32-bit application uses read() rather than any form of recvmsg() it will still get the wrong (64-bit) event. However, neither do applications actually do this, nor would it be a regression. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-01 11:26:02 +00:00
*
* If we need to send the compat skb, assign it to the
* 'data_skb' variable so that it will be used below for data
* copying. We keep 'skb' for everything else, including
* freeing both later.
net/compat/wext: send different messages to compat tasks Wireless extensions have the unfortunate problem that events are multicast netlink messages, and are not independent of pointer size. Thus, currently 32-bit tasks on 64-bit platforms cannot properly receive events and fail with all kinds of strange problems, for instance wpa_supplicant never notices disassociations, due to the way the 64-bit event looks (to a 32-bit process), the fact that the address is all zeroes is lost, it thinks instead it is 00:00:00:00:01:00. The same problem existed with the ioctls, until David Miller fixed those some time ago in an heroic effort. A different problem caused by this is that we cannot send the ASSOCREQIE/ASSOCRESPIE events because sending them causes a 32-bit wpa_supplicant on a 64-bit system to overwrite its internal information, which is worse than it not getting the information at all -- so we currently resort to sending a custom string event that it then parses. This, however, has a severe size limitation we are frequently hitting with modern access points; this limitation would can be lifted after this patch by sending the correct binary, not custom, event. A similar problem apparently happens for some other netlink users on x86_64 with 32-bit tasks due to the alignment for 64-bit quantities. In order to fix these problems, I have implemented a way to send compat messages to tasks. When sending an event, we send the non-compat event data together with a compat event data in skb_shinfo(main_skb)->frag_list. Then, when the event is read from the socket, the netlink code makes sure to pass out only the skb that is compatible with the task. This approach was suggested by David Miller, my original approach required always sending two skbs but that had various small problems. To determine whether compat is needed or not, I have used the MSG_CMSG_COMPAT flag, and adjusted the call path for recv and recvfrom to include it, even if those calls do not have a cmsg parameter. I have not solved one small part of the problem, and I don't think it is necessary to: if a 32-bit application uses read() rather than any form of recvmsg() it will still get the wrong (64-bit) event. However, neither do applications actually do this, nor would it be a regression. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-01 11:26:02 +00:00
*/
if (flags & MSG_CMSG_COMPAT)
data_skb = skb_shinfo(skb)->frag_list;
net/compat/wext: send different messages to compat tasks Wireless extensions have the unfortunate problem that events are multicast netlink messages, and are not independent of pointer size. Thus, currently 32-bit tasks on 64-bit platforms cannot properly receive events and fail with all kinds of strange problems, for instance wpa_supplicant never notices disassociations, due to the way the 64-bit event looks (to a 32-bit process), the fact that the address is all zeroes is lost, it thinks instead it is 00:00:00:00:01:00. The same problem existed with the ioctls, until David Miller fixed those some time ago in an heroic effort. A different problem caused by this is that we cannot send the ASSOCREQIE/ASSOCRESPIE events because sending them causes a 32-bit wpa_supplicant on a 64-bit system to overwrite its internal information, which is worse than it not getting the information at all -- so we currently resort to sending a custom string event that it then parses. This, however, has a severe size limitation we are frequently hitting with modern access points; this limitation would can be lifted after this patch by sending the correct binary, not custom, event. A similar problem apparently happens for some other netlink users on x86_64 with 32-bit tasks due to the alignment for 64-bit quantities. In order to fix these problems, I have implemented a way to send compat messages to tasks. When sending an event, we send the non-compat event data together with a compat event data in skb_shinfo(main_skb)->frag_list. Then, when the event is read from the socket, the netlink code makes sure to pass out only the skb that is compatible with the task. This approach was suggested by David Miller, my original approach required always sending two skbs but that had various small problems. To determine whether compat is needed or not, I have used the MSG_CMSG_COMPAT flag, and adjusted the call path for recv and recvfrom to include it, even if those calls do not have a cmsg parameter. I have not solved one small part of the problem, and I don't think it is necessary to: if a 32-bit application uses read() rather than any form of recvmsg() it will still get the wrong (64-bit) event. However, neither do applications actually do this, nor would it be a regression. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-01 11:26:02 +00:00
}
#endif
/* Record the max length of recvmsg() calls for future allocations */
nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len);
nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len,
SKB_WITH_OVERHEAD(32768));
copied = data_skb->len;
if (len < copied) {
msg->msg_flags |= MSG_TRUNC;
copied = len;
}
skb_reset_transport_header(data_skb);
err = skb_copy_datagram_msg(data_skb, 0, msg, copied);
if (msg->msg_name) {
DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
addr->nl_family = AF_NETLINK;
addr->nl_pad = 0;
addr->nl_pid = NETLINK_CB(skb).portid;
addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
msg->msg_namelen = sizeof(*addr);
}
if (nlk->flags & NETLINK_F_RECV_PKTINFO)
netlink_cmsg_recv_pktinfo(msg, skb);
if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID)
netlink_cmsg_listen_all_nsid(sk, msg, skb);
memset(&scm, 0, sizeof(scm));
scm.creds = *NETLINK_CREDS(skb);
if (flags & MSG_TRUNC)
copied = data_skb->len;
skb_free_datagram(sk, skb);
if (nlk->cb_running &&
atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
ret = netlink_dump(sk);
if (ret) {
sk->sk_err = -ret;
sk->sk_error_report(sk);
}
}
scm_recv(sock, msg, &scm, flags);
out:
netlink_rcv_wake(sk);
return err ? : copied;
}
static void netlink_data_ready(struct sock *sk)
{
BUG();
}
/*
* We export these functions to other modules. They provide a
* complete set of kernel non-blocking support for message
* queueing.
*/
struct sock *
__netlink_kernel_create(struct net *net, int unit, struct module *module,
struct netlink_kernel_cfg *cfg)
{
struct socket *sock;
struct sock *sk;
struct netlink_sock *nlk;
struct listeners *listeners = NULL;
struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
unsigned int groups;
BUG_ON(!nl_table);
if (unit < 0 || unit >= MAX_LINKS)
return NULL;
if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
return NULL;
if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0)
[NETNS]: Fix race between put_net() and netlink_kernel_create(). The comment about "race free view of the set of network namespaces" was a bit hasty. Look (there even can be only one CPU, as discovered by Alexey Dobriyan and Denis Lunev): put_net() if (atomic_dec_and_test(&net->refcnt)) /* true */ __put_net(net); queue_work(...); /* * note: the net now has refcnt 0, but still in * the global list of net namespaces */ == re-schedule == register_pernet_subsys(&some_ops); register_pernet_operations(&some_ops); (*some_ops)->init(net); /* * we call netlink_kernel_create() here * in some places */ netlink_kernel_create(); sk_alloc(); get_net(net); /* refcnt = 1 */ /* * now we drop the net refcount not to * block the net namespace exit in the * future (or this can be done on the * error path) */ put_net(sk->sk_net); if (atomic_dec_and_test(&...)) /* * true. BOOOM! The net is * scheduled for release twice */ When thinking on this problem, I decided, that getting and putting the net in init callback is wrong. If some init callback needs to have a refcount-less reference on the struct net, _it_ has to be careful himself, rather than relying on the infrastructure to handle this correctly. In case of netlink_kernel_create(), the problem is that the sk_alloc() gets the given namespace, but passing the info that we don't want to get it inside this call is too heavy. Instead, I propose to crate the socket inside an init_net namespace and then re-attach it to the desired one right after the socket is created. After doing this, we also have to be careful on error paths not to drop the reference on the namespace, we didn't get the one on. Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Acked-by: Denis Lunev <den@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-01-31 03:31:06 +00:00
goto out_sock_release_nosk;
sk = sock->sk;
if (!cfg || cfg->groups < 32)
groups = 32;
else
groups = cfg->groups;
listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
if (!listeners)
goto out_sock_release;
sk->sk_data_ready = netlink_data_ready;
if (cfg && cfg->input)
nlk_sk(sk)->netlink_rcv = cfg->input;
if (netlink_insert(sk, 0))
goto out_sock_release;
nlk = nlk_sk(sk);
nlk->flags |= NETLINK_F_KERNEL_SOCKET;
netlink_table_grab();
if (!nl_table[unit].registered) {
nl_table[unit].groups = groups;
rcu_assign_pointer(nl_table[unit].listeners, listeners);
nl_table[unit].cb_mutex = cb_mutex;
nl_table[unit].module = module;
if (cfg) {
nl_table[unit].bind = cfg->bind;
nl_table[unit].unbind = cfg->unbind;
nl_table[unit].flags = cfg->flags;
if (cfg->compare)
nl_table[unit].compare = cfg->compare;
}
nl_table[unit].registered = 1;
} else {
kfree(listeners);
nl_table[unit].registered++;
}
netlink_table_ungrab();
return sk;
out_sock_release:
kfree(listeners);
netlink_kernel_release(sk);
[NETNS]: Fix race between put_net() and netlink_kernel_create(). The comment about "race free view of the set of network namespaces" was a bit hasty. Look (there even can be only one CPU, as discovered by Alexey Dobriyan and Denis Lunev): put_net() if (atomic_dec_and_test(&net->refcnt)) /* true */ __put_net(net); queue_work(...); /* * note: the net now has refcnt 0, but still in * the global list of net namespaces */ == re-schedule == register_pernet_subsys(&some_ops); register_pernet_operations(&some_ops); (*some_ops)->init(net); /* * we call netlink_kernel_create() here * in some places */ netlink_kernel_create(); sk_alloc(); get_net(net); /* refcnt = 1 */ /* * now we drop the net refcount not to * block the net namespace exit in the * future (or this can be done on the * error path) */ put_net(sk->sk_net); if (atomic_dec_and_test(&...)) /* * true. BOOOM! The net is * scheduled for release twice */ When thinking on this problem, I decided, that getting and putting the net in init callback is wrong. If some init callback needs to have a refcount-less reference on the struct net, _it_ has to be careful himself, rather than relying on the infrastructure to handle this correctly. In case of netlink_kernel_create(), the problem is that the sk_alloc() gets the given namespace, but passing the info that we don't want to get it inside this call is too heavy. Instead, I propose to crate the socket inside an init_net namespace and then re-attach it to the desired one right after the socket is created. After doing this, we also have to be careful on error paths not to drop the reference on the namespace, we didn't get the one on. Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Acked-by: Denis Lunev <den@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-01-31 03:31:06 +00:00
return NULL;
out_sock_release_nosk:
sock_release(sock);
return NULL;
}
EXPORT_SYMBOL(__netlink_kernel_create);
void
netlink_kernel_release(struct sock *sk)
{
if (sk == NULL || sk->sk_socket == NULL)
return;
sock_release(sk->sk_socket);
}
EXPORT_SYMBOL(netlink_kernel_release);
int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
{
struct listeners *new, *old;
struct netlink_table *tbl = &nl_table[sk->sk_protocol];
if (groups < 32)
groups = 32;
if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
if (!new)
return -ENOMEM;
old = nl_deref_protected(tbl->listeners);
memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
rcu_assign_pointer(tbl->listeners, new);
kfree_rcu(old, rcu);
}
tbl->groups = groups;
return 0;
}
/**
* netlink_change_ngroups - change number of multicast groups
*
* This changes the number of multicast groups that are available
* on a certain netlink family. Note that it is not possible to
* change the number of groups to below 32. Also note that it does
* not implicitly call netlink_clear_multicast_users() when the
* number of groups is reduced.
*
* @sk: The kernel netlink socket, as returned by netlink_kernel_create().
* @groups: The new number of groups.
*/
int netlink_change_ngroups(struct sock *sk, unsigned int groups)
{
int err;
netlink_table_grab();
err = __netlink_change_ngroups(sk, groups);
netlink_table_ungrab();
return err;
}
void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
{
struct sock *sk;
struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
sk_for_each_bound(sk, &tbl->mc_list)
netlink_update_socket_mc(nlk_sk(sk), group, 0);
}
struct nlmsghdr *
__nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
{
struct nlmsghdr *nlh;
int size = nlmsg_msg_size(len);
nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size));
nlh->nlmsg_type = type;
nlh->nlmsg_len = size;
nlh->nlmsg_flags = flags;
nlh->nlmsg_pid = portid;
nlh->nlmsg_seq = seq;
if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
return nlh;
}
EXPORT_SYMBOL(__nlmsg_put);
/*
* It looks a bit ugly.
* It would be better to create kernel thread.
*/
static int netlink_dump(struct sock *sk)
{
struct netlink_sock *nlk = nlk_sk(sk);
struct netlink_callback *cb;
struct sk_buff *skb = NULL;
struct nlmsghdr *nlh;
struct module *module;
int len, err = -ENOBUFS;
int alloc_min_size;
int alloc_size;
mutex_lock(nlk->cb_mutex);
if (!nlk->cb_running) {
err = -EINVAL;
goto errout_skb;
}
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
goto errout_skb;
/* NLMSG_GOODSIZE is small to avoid high order allocations being
* required, but it makes sense to _attempt_ a 16K bytes allocation
* to reduce number of system calls on dump operations, if user
* ever provided a big enough buffer.
*/
cb = &nlk->cb;
alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
if (alloc_min_size < nlk->max_recvmsg_len) {
alloc_size = nlk->max_recvmsg_len;
skb = alloc_skb(alloc_size,
(GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) |
__GFP_NOWARN | __GFP_NORETRY);
}
if (!skb) {
alloc_size = alloc_min_size;
skb = alloc_skb(alloc_size, GFP_KERNEL);
}
if (!skb)
goto errout_skb;
/* Trim skb to allocated size. User is expected to provide buffer as
* large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at
* netlink_recvmsg())). dump will pack as many smaller messages as
* could fit within the allocated skb. skb is typically allocated
* with larger space than required (could be as much as near 2x the
* requested size with align to next power of 2 approach). Allowing
* dump to use the excess space makes it difficult for a user to have a
* reasonable static buffer based on the expected largest dump of a
* single netdev. The outcome is MSG_TRUNC error.
*/
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
skb_reserve(skb, skb_tailroom(skb) - alloc_size);
netlink_skb_set_owner_r(skb, sk);
len = cb->dump(skb, cb);
if (len > 0) {
mutex_unlock(nlk->cb_mutex);
if (sk_filter(sk, skb))
kfree_skb(skb);
else
__netlink_sendskb(sk, skb);
return 0;
}
nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
if (!nlh)
goto errout_skb;
nl_dump_check_consistent(cb, nlh);
memcpy(nlmsg_data(nlh), &len, sizeof(len));
if (sk_filter(sk, skb))
kfree_skb(skb);
else
__netlink_sendskb(sk, skb);
if (cb->done)
cb->done(cb);
nlk->cb_running = false;
module = cb->module;
skb = cb->skb;
mutex_unlock(nlk->cb_mutex);
module_put(module);
consume_skb(skb);
return 0;
errout_skb:
mutex_unlock(nlk->cb_mutex);
kfree_skb(skb);
return err;
}
int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
const struct nlmsghdr *nlh,
struct netlink_dump_control *control)
{
struct netlink_callback *cb;
struct sock *sk;
struct netlink_sock *nlk;
int ret;
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
atomic_inc(&skb->users);
sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
if (sk == NULL) {
ret = -ECONNREFUSED;
goto error_free;
}
nlk = nlk_sk(sk);
mutex_lock(nlk->cb_mutex);
/* A dump is in progress... */
if (nlk->cb_running) {
ret = -EBUSY;
goto error_unlock;
}
/* add reference of module which cb->dump belongs to */
if (!try_module_get(control->module)) {
ret = -EPROTONOSUPPORT;
goto error_unlock;
}
cb = &nlk->cb;
memset(cb, 0, sizeof(*cb));
cb->start = control->start;
cb->dump = control->dump;
cb->done = control->done;
cb->nlh = nlh;
cb->data = control->data;
cb->module = control->module;
cb->min_dump_alloc = control->min_dump_alloc;
cb->skb = skb;
nlk->cb_running = true;
mutex_unlock(nlk->cb_mutex);
if (cb->start)
cb->start(cb);
ret = netlink_dump(sk);
sock_put(sk);
if (ret)
return ret;
/* We successfully started a dump, by returning -EINTR we
* signal not to send ACK even if it was requested.
*/
return -EINTR;
error_unlock:
sock_put(sk);
mutex_unlock(nlk->cb_mutex);
error_free:
kfree_skb(skb);
return ret;
}
EXPORT_SYMBOL(__netlink_dump_start);
void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err,
const struct netlink_ext_ack *extack)
{
struct sk_buff *skb;
struct nlmsghdr *rep;
struct nlmsgerr *errmsg;
size_t payload = sizeof(*errmsg);
size_t tlvlen = 0;
struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk);
unsigned int flags = 0;
/* Error messages get the original request appened, unless the user
* requests to cap the error message, and get extra error data if
* requested.
*/
if (err) {
if (!(nlk->flags & NETLINK_F_CAP_ACK))
payload += nlmsg_len(nlh);
else
flags |= NLM_F_CAPPED;
if (nlk->flags & NETLINK_F_EXT_ACK && extack) {
if (extack->_msg)
tlvlen += nla_total_size(strlen(extack->_msg) + 1);
if (extack->bad_attr)
tlvlen += nla_total_size(sizeof(u32));
}
} else {
flags |= NLM_F_CAPPED;
if (nlk->flags & NETLINK_F_EXT_ACK &&
extack && extack->cookie_len)
tlvlen += nla_total_size(extack->cookie_len);
}
if (tlvlen)
flags |= NLM_F_ACK_TLVS;
skb = nlmsg_new(payload + tlvlen, GFP_KERNEL);
if (!skb) {
struct sock *sk;
sk = netlink_lookup(sock_net(in_skb->sk),
in_skb->sk->sk_protocol,
NETLINK_CB(in_skb).portid);
if (sk) {
sk->sk_err = ENOBUFS;
sk->sk_error_report(sk);
sock_put(sk);
}
return;
}
rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
NLMSG_ERROR, payload, flags);
errmsg = nlmsg_data(rep);
errmsg->error = err;
memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh));
if (nlk->flags & NETLINK_F_EXT_ACK && extack) {
if (err) {
if (extack->_msg)
WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG,
extack->_msg));
if (extack->bad_attr &&
!WARN_ON((u8 *)extack->bad_attr < in_skb->data ||
(u8 *)extack->bad_attr >= in_skb->data +
in_skb->len))
WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS,
(u8 *)extack->bad_attr -
in_skb->data));
} else {
if (extack->cookie_len)
WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE,
extack->cookie_len,
extack->cookie));
}
}
nlmsg_end(skb, rep);
netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
}
EXPORT_SYMBOL(netlink_ack);
int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
struct nlmsghdr *,
struct netlink_ext_ack *))
{
struct netlink_ext_ack extack = {};
struct nlmsghdr *nlh;
int err;
while (skb->len >= nlmsg_total_size(0)) {
int msglen;
nlh = nlmsg_hdr(skb);
err = 0;
if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
return 0;
/* Only requests are handled by the kernel */
if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
goto ack;
/* Skip control messages */
if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
goto ack;
err = cb(skb, nlh, &extack);
if (err == -EINTR)
goto skip;
ack:
if (nlh->nlmsg_flags & NLM_F_ACK || err)
netlink_ack(skb, nlh, err, &extack);
skip:
msglen = NLMSG_ALIGN(nlh->nlmsg_len);
if (msglen > skb->len)
msglen = skb->len;
skb_pull(skb, msglen);
}
return 0;
}
EXPORT_SYMBOL(netlink_rcv_skb);
/**
* nlmsg_notify - send a notification netlink message
* @sk: netlink socket to use
* @skb: notification message
* @portid: destination netlink portid for reports or 0
* @group: destination multicast group or 0
* @report: 1 to report back, 0 to disable
* @flags: allocation flags
*/
int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
unsigned int group, int report, gfp_t flags)
{
int err = 0;
if (group) {
int exclude_portid = 0;
if (report) {
atomic_inc(&skb->users);
exclude_portid = portid;
}
2009-02-25 07:18:28 +00:00
/* errors reported via destination sk->sk_err, but propagate
* delivery errors if NETLINK_BROADCAST_ERROR flag is set */
err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
}
2009-02-25 07:18:28 +00:00
if (report) {
int err2;
err2 = nlmsg_unicast(sk, skb, portid);
2009-02-25 07:18:28 +00:00
if (!err || err == -ESRCH)
err = err2;
}
return err;
}
EXPORT_SYMBOL(nlmsg_notify);
#ifdef CONFIG_PROC_FS
struct nl_seq_iter {
struct seq_net_private p;
struct rhashtable_iter hti;
int link;
};
static int netlink_walk_start(struct nl_seq_iter *iter)
{
int err;
err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti,
GFP_KERNEL);
if (err) {
iter->link = MAX_LINKS;
return err;
}
err = rhashtable_walk_start(&iter->hti);
return err == -EAGAIN ? 0 : err;
}
static void netlink_walk_stop(struct nl_seq_iter *iter)
{
rhashtable_walk_stop(&iter->hti);
rhashtable_walk_exit(&iter->hti);
}
static void *__netlink_seq_next(struct seq_file *seq)
{
struct nl_seq_iter *iter = seq->private;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
struct netlink_sock *nlk;
do {
for (;;) {
int err;
nlk = rhashtable_walk_next(&iter->hti);
if (IS_ERR(nlk)) {
if (PTR_ERR(nlk) == -EAGAIN)
continue;
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
return nlk;
}
if (nlk)
break;
netlink_walk_stop(iter);
if (++iter->link >= MAX_LINKS)
return NULL;
err = netlink_walk_start(iter);
if (err)
return ERR_PTR(err);
}
} while (sock_net(&nlk->sk) != seq_file_net(seq));
return nlk;
}
static void *netlink_seq_start(struct seq_file *seq, loff_t *posp)
{
struct nl_seq_iter *iter = seq->private;
void *obj = SEQ_START_TOKEN;
loff_t pos;
int err;
iter->link = 0;
err = netlink_walk_start(iter);
if (err)
return ERR_PTR(err);
for (pos = *posp; pos && obj && !IS_ERR(obj); pos--)
obj = __netlink_seq_next(seq);
return obj;
}
static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
++*pos;
return __netlink_seq_next(seq);
}
static void netlink_seq_stop(struct seq_file *seq, void *v)
{
struct nl_seq_iter *iter = seq->private;
if (iter->link >= MAX_LINKS)
return;
netlink_walk_stop(iter);
}
static int netlink_seq_show(struct seq_file *seq, void *v)
{
if (v == SEQ_START_TOKEN) {
seq_puts(seq,
"sk Eth Pid Groups "
"Rmem Wmem Dump Locks Drops Inode\n");
} else {
struct sock *s = v;
struct netlink_sock *nlk = nlk_sk(s);
seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n",
s,
s->sk_protocol,
nlk->portid,
nlk->groups ? (u32)nlk->groups[0] : 0,
sk_rmem_alloc_get(s),
sk_wmem_alloc_get(s),
nlk->cb_running,
netlink: add NETLINK_NO_ENOBUFS socket flag This patch adds the NETLINK_NO_ENOBUFS socket flag. This flag can be used by unicast and broadcast listeners to avoid receiving ENOBUFS errors. Generally speaking, ENOBUFS errors are useful to notify two things to the listener: a) You may increase the receiver buffer size via setsockopt(). b) You have lost messages, you may be out of sync. In some cases, ignoring ENOBUFS errors can be useful. For example: a) nfnetlink_queue: this subsystem does not have any sort of resync method and you can decide to ignore ENOBUFS once you have set a given buffer size. b) ctnetlink: you can use this together with the socket flag NETLINK_BROADCAST_SEND_ERROR to stop getting ENOBUFS errors as you do not need to resync (packets whose event are not delivered are drop to provide reliable logging and state-synchronization). Moreover, the use of NETLINK_NO_ENOBUFS also reduces a "go up, go down" effect in terms of performance which is due to the netlink congestion control when the listener cannot back off. The effect is the following: 1) throughput rate goes up and netlink messages are inserted in the receiver buffer. 2) Then, netlink buffer fills and overruns (set on nlk->state bit 0). 3) While the listener empties the receiver buffer, netlink keeps dropping messages. Thus, throughput goes dramatically down. 4) Then, once the listener has emptied the buffer (nlk->state bit 0 is set off), goto step 1. This effect is easy to trigger with netlink broadcast under heavy load, and it is more noticeable when using a big receiver buffer. You can find some results in [1] that show this problem. [1] http://1984.lsi.us.es/linux/netlink/ This patch also includes the use of sk_drop to account the number of netlink messages drop due to overrun. This value is shown in /proc/net/netlink. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-03-24 23:37:55 +00:00
atomic_read(&s->sk_refcnt),
atomic_read(&s->sk_drops),
sock_i_ino(s)
);
}
return 0;
}
static const struct seq_operations netlink_seq_ops = {
.start = netlink_seq_start,
.next = netlink_seq_next,
.stop = netlink_seq_stop,
.show = netlink_seq_show,
};
static int netlink_seq_open(struct inode *inode, struct file *file)
{
return seq_open_net(inode, file, &netlink_seq_ops,
sizeof(struct nl_seq_iter));
}
static const struct file_operations netlink_seq_fops = {
.owner = THIS_MODULE,
.open = netlink_seq_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release_net,
};
#endif
int netlink_register_notifier(struct notifier_block *nb)
{
return blocking_notifier_chain_register(&netlink_chain, nb);
}
EXPORT_SYMBOL(netlink_register_notifier);
int netlink_unregister_notifier(struct notifier_block *nb)
{
return blocking_notifier_chain_unregister(&netlink_chain, nb);
}
EXPORT_SYMBOL(netlink_unregister_notifier);
static const struct proto_ops netlink_ops = {
.family = PF_NETLINK,
.owner = THIS_MODULE,
.release = netlink_release,
.bind = netlink_bind,
.connect = netlink_connect,
.socketpair = sock_no_socketpair,
.accept = sock_no_accept,
.getname = netlink_getname,
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
.poll = datagram_poll,
.ioctl = netlink_ioctl,
.listen = sock_no_listen,
.shutdown = sock_no_shutdown,
.setsockopt = netlink_setsockopt,
.getsockopt = netlink_getsockopt,
.sendmsg = netlink_sendmsg,
.recvmsg = netlink_recvmsg,
netlink: remove mmapped netlink support mmapped netlink has a number of unresolved issues: - TX zerocopy support had to be disabled more than a year ago via commit 4682a0358639b29cf ("netlink: Always copy on mmap TX.") because the content of the mmapped area can change after netlink attribute validation but before message processing. - RX support was implemented mainly to speed up nfqueue dumping packet payload to userspace. However, since commit ae08ce0021087a5d812d2 ("netfilter: nfnetlink_queue: zero copy support") we avoid one copy with the socket-based interface too (via the skb_zerocopy helper). The other problem is that skbs attached to mmaped netlink socket behave different from normal skbs: - they don't have a shinfo area, so all functions that use skb_shinfo() (e.g. skb_clone) cannot be used. - reserving headroom prevents userspace from seeing the content as it expects message to start at skb->head. See for instance commit aa3a022094fa ("netlink: not trim skb for mmaped socket when dump"). - skbs handed e.g. to netlink_ack must have non-NULL skb->sk, else we crash because it needs the sk to check if a tx ring is attached. Also not obvious, leads to non-intuitive bug fixes such as 7c7bdf359 ("netfilter: nfnetlink: use original skbuff when acking batches"). mmaped netlink also didn't play nicely with the skb_zerocopy helper used by nfqueue and openvswitch. Daniel Borkmann fixed this via commit 6bb0fef489f6 ("netlink, mmap: fix edge-case leakages in nf queue zero-copy")' but at the cost of also needing to provide remaining length to the allocation function. nfqueue also has problems when used with mmaped rx netlink: - mmaped netlink doesn't allow use of nfqueue batch verdict messages. Problem is that in the mmap case, the allocation time also determines the ordering in which the frame will be seen by userspace (A allocating before B means that A is located in earlier ring slot, but this also means that B might get a lower sequence number then A since seqno is decided later. To fix this we would need to extend the spinlocked region to also cover the allocation and message setup which isn't desirable. - nfqueue can now be configured to queue large (GSO) skbs to userspace. Queing GSO packets is faster than having to force a software segmentation in the kernel, so this is a desirable option. However, with a mmap based ring one has to use 64kb per ring slot element, else mmap has to fall back to the socket path (NL_MMAP_STATUS_COPY) for all large packets. To use the mmap interface, userspace not only has to probe for mmap netlink support, it also has to implement a recv/socket receive path in order to handle messages that exceed the size of an rx ring element. Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ken-ichirou MATSUZAWA <chamaken@gmail.com> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Patrick McHardy <kaber@trash.net> Cc: Thomas Graf <tgraf@suug.ch> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-18 14:03:24 +00:00
.mmap = sock_no_mmap,
.sendpage = sock_no_sendpage,
};
static const struct net_proto_family netlink_family_ops = {
.family = PF_NETLINK,
.create = netlink_create,
.owner = THIS_MODULE, /* for consistency 8) */
};
static int __net_init netlink_net_init(struct net *net)
{
#ifdef CONFIG_PROC_FS
if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops))
return -ENOMEM;
#endif
return 0;
}
static void __net_exit netlink_net_exit(struct net *net)
{
#ifdef CONFIG_PROC_FS
remove_proc_entry("netlink", net->proc_net);
#endif
}
static void __init netlink_add_usersock_entry(void)
{
struct listeners *listeners;
int groups = 32;
listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
if (!listeners)
panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
netlink_table_grab();
nl_table[NETLINK_USERSOCK].groups = groups;
rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
nl_table[NETLINK_USERSOCK].registered = 1;
nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
netlink_table_ungrab();
}
static struct pernet_operations __net_initdata netlink_net_ops = {
.init = netlink_net_init,
.exit = netlink_net_exit,
};
static inline u32 netlink_hash(const void *data, u32 len, u32 seed)
{
const struct netlink_sock *nlk = data;
struct netlink_compare_arg arg;
netlink: Replace rhash_portid with bound On Mon, Sep 21, 2015 at 02:20:22PM -0400, Tejun Heo wrote: > > store_release and load_acquire are different from the usual memory > barriers and can't be paired this way. You have to pair store_release > and load_acquire. Besides, it isn't a particularly good idea to OK I've decided to drop the acquire/release helpers as they don't help us at all and simply pessimises the code by using full memory barriers (on some architectures) where only a write or read barrier is needed. > depend on memory barriers embedded in other data structures like the > above. Here, especially, rhashtable_insert() would have write barrier > *before* the entry is hashed not necessarily *after*, which means that > in the above case, a socket which appears to have set bound to a > reader might not visible when the reader tries to look up the socket > on the hashtable. But you are right we do need an explicit write barrier here to ensure that the hashing is visible. > There's no reason to be overly smart here. This isn't a crazy hot > path, write barriers tend to be very cheap, store_release more so. > Please just do smp_store_release() and note what it's paired with. It's not about being overly smart. It's about actually understanding what's going on with the code. I've seen too many instances of people simply sprinkling synchronisation primitives around without any knowledge of what is happening underneath, which is just a recipe for creating hard-to-debug races. > > @@ -1539,7 +1546,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr, > > } > > } > > > > - if (!nlk->portid) { > > + if (!nlk->bound) { > > I don't think you can skip load_acquire here just because this is the > second deref of the variable. That doesn't change anything. Race > condition could still happen between the first and second tests and > skipping the second would lead to the same kind of bug. The reason this one is OK is because we do not use nlk->portid or try to get nlk from the hash table before we return to user-space. However, there is a real bug here that none of these acquire/release helpers discovered. The two bound tests here used to be a single one. Now that they are separate it is entirely possible for another thread to come in the middle and bind the socket. So we need to repeat the portid check in order to maintain consistency. > > @@ -1587,7 +1594,7 @@ static int netlink_connect(struct socket *sock, struct sockaddr *addr, > > !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) > > return -EPERM; > > > > - if (!nlk->portid) > > + if (!nlk->bound) > > Don't we need load_acquire here too? Is this path holding a lock > which makes that unnecessary? Ditto. ---8<--- The commit 1f770c0a09da855a2b51af6d19de97fb955eca85 ("netlink: Fix autobind race condition that leads to zero port ID") created some new races that can occur due to inconcsistencies between the two port IDs. Tejun is right that a barrier is unavoidable. Therefore I am reverting to the original patch that used a boolean to indicate that a user netlink socket has been bound. Barriers have been added where necessary to ensure that a valid portid and the hashed socket is visible. I have also changed netlink_insert to only return EBUSY if the socket is bound to a portid different to the requested one. This combined with only reading nlk->bound once in netlink_bind fixes a race where two threads that bind the socket at the same time with different port IDs may both succeed. Fixes: 1f770c0a09da ("netlink: Fix autobind race condition that leads to zero port ID") Reported-by: Tejun Heo <tj@kernel.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Nacked-by: Tejun Heo <tj@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-22 03:38:56 +00:00
netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid);
return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed);
}
static const struct rhashtable_params netlink_rhashtable_params = {
.head_offset = offsetof(struct netlink_sock, node),
.key_len = netlink_compare_arg_len,
.obj_hashfn = netlink_hash,
.obj_cmpfn = netlink_compare,
.automatic_shrinking = true,
};
static int __init netlink_proto_init(void)
{
int i;
int err = proto_register(&netlink_proto, 0);
if (err != 0)
goto out;
BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
if (!nl_table)
goto panic;
for (i = 0; i < MAX_LINKS; i++) {
if (rhashtable_init(&nl_table[i].hash,
&netlink_rhashtable_params) < 0) {
netlink: Convert netlink_lookup() to use RCU protected hash table Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by: Thomas Graf <tgraf@suug.ch> Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-02 09:47:45 +00:00
while (--i > 0)
rhashtable_destroy(&nl_table[i].hash);
kfree(nl_table);
goto panic;
}
}
INIT_LIST_HEAD(&netlink_tap_all);
netlink_add_usersock_entry();
sock_register(&netlink_family_ops);
register_pernet_subsys(&netlink_net_ops);
/* The netlink device handler may be needed early. */
rtnetlink_init();
out:
return err;
panic:
panic("netlink_init: Cannot allocate nl_table\n");
}
core_initcall(netlink_proto_init);