2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* INET An implementation of the TCP/IP protocol suite for the LINUX
|
|
|
|
* operating system. INET is implemented using the BSD Socket
|
|
|
|
* interface as the means of communication with the user level.
|
|
|
|
*
|
|
|
|
* ROUTE - implementation of the IP router.
|
|
|
|
*
|
2005-05-05 17:16:16 -06:00
|
|
|
* Authors: Ross Biro
|
2005-04-16 16:20:36 -06:00
|
|
|
* Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
|
|
|
|
* Alan Cox, <gw4pts@gw4pts.ampr.org>
|
|
|
|
* Linus Torvalds, <Linus.Torvalds@helsinki.fi>
|
|
|
|
* Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
|
|
|
|
*
|
|
|
|
* Fixes:
|
|
|
|
* Alan Cox : Verify area fixes.
|
|
|
|
* Alan Cox : cli() protects routing changes
|
|
|
|
* Rui Oliveira : ICMP routing table updates
|
|
|
|
* (rco@di.uminho.pt) Routing table insertion and update
|
|
|
|
* Linus Torvalds : Rewrote bits to be sensible
|
|
|
|
* Alan Cox : Added BSD route gw semantics
|
2007-02-09 07:24:47 -07:00
|
|
|
* Alan Cox : Super /proc >4K
|
2005-04-16 16:20:36 -06:00
|
|
|
* Alan Cox : MTU in route table
|
|
|
|
* Alan Cox : MSS actually. Also added the window
|
|
|
|
* clamper.
|
|
|
|
* Sam Lantinga : Fixed route matching in rt_del()
|
|
|
|
* Alan Cox : Routing cache support.
|
|
|
|
* Alan Cox : Removed compatibility cruft.
|
|
|
|
* Alan Cox : RTF_REJECT support.
|
|
|
|
* Alan Cox : TCP irtt support.
|
|
|
|
* Jonathan Naylor : Added Metric support.
|
|
|
|
* Miquel van Smoorenburg : BSD API fixes.
|
|
|
|
* Miquel van Smoorenburg : Metrics.
|
|
|
|
* Alan Cox : Use __u32 properly
|
|
|
|
* Alan Cox : Aligned routing errors more closely with BSD
|
|
|
|
* our system is still very different.
|
|
|
|
* Alan Cox : Faster /proc handling
|
|
|
|
* Alexey Kuznetsov : Massive rework to support tree based routing,
|
|
|
|
* routing caches and better behaviour.
|
2007-02-09 07:24:47 -07:00
|
|
|
*
|
2005-04-16 16:20:36 -06:00
|
|
|
* Olaf Erb : irtt wasn't being copied right.
|
|
|
|
* Bjorn Ekwall : Kerneld route support.
|
|
|
|
* Alan Cox : Multicast fixed (I hope)
|
|
|
|
* Pavel Krauz : Limited broadcast fixed
|
|
|
|
* Mike McLagan : Routing by source
|
|
|
|
* Alexey Kuznetsov : End of old history. Split to fib.c and
|
|
|
|
* route.c and rewritten from scratch.
|
|
|
|
* Andi Kleen : Load-limit warning messages.
|
|
|
|
* Vitaly E. Lavrov : Transparent proxy revived after year coma.
|
|
|
|
* Vitaly E. Lavrov : Race condition in ip_route_input_slow.
|
|
|
|
* Tobias Ringstrom : Uninitialized res.type in ip_route_output_slow.
|
|
|
|
* Vladimir V. Ivanov : IP rule info (flowid) is really useful.
|
|
|
|
* Marc Boucher : routing by fwmark
|
|
|
|
* Robert Olsson : Added rt_cache statistics
|
|
|
|
* Arnaldo C. Melo : Convert proc stuff to seq_file
|
2005-07-05 16:00:32 -06:00
|
|
|
* Eric Dumazet : hashed spinlocks and rt_check_expire() fixes.
|
2006-03-25 02:38:55 -07:00
|
|
|
* Ilia Sotnikov : Ignore TOS on PMTUD and Redirect
|
|
|
|
* Ilia Sotnikov : Removed TOS from hash calculations
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <asm/uaccess.h>
|
|
|
|
#include <asm/system.h>
|
|
|
|
#include <linux/bitops.h>
|
|
|
|
#include <linux/types.h>
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/mm.h>
|
2005-07-05 15:58:19 -06:00
|
|
|
#include <linux/bootmem.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/socket.h>
|
|
|
|
#include <linux/sockios.h>
|
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/in.h>
|
|
|
|
#include <linux/inet.h>
|
|
|
|
#include <linux/netdevice.h>
|
|
|
|
#include <linux/proc_fs.h>
|
|
|
|
#include <linux/init.h>
|
2007-09-15 11:55:54 -06:00
|
|
|
#include <linux/workqueue.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <linux/skbuff.h>
|
|
|
|
#include <linux/inetdevice.h>
|
|
|
|
#include <linux/igmp.h>
|
|
|
|
#include <linux/pkt_sched.h>
|
|
|
|
#include <linux/mroute.h>
|
|
|
|
#include <linux/netfilter_ipv4.h>
|
|
|
|
#include <linux/random.h>
|
|
|
|
#include <linux/jhash.h>
|
|
|
|
#include <linux/rcupdate.h>
|
|
|
|
#include <linux/times.h>
|
2007-11-13 22:34:06 -07:00
|
|
|
#include <net/dst.h>
|
2007-09-12 04:01:34 -06:00
|
|
|
#include <net/net_namespace.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <net/protocol.h>
|
|
|
|
#include <net/ip.h>
|
|
|
|
#include <net/route.h>
|
|
|
|
#include <net/inetpeer.h>
|
|
|
|
#include <net/sock.h>
|
|
|
|
#include <net/ip_fib.h>
|
|
|
|
#include <net/arp.h>
|
|
|
|
#include <net/tcp.h>
|
|
|
|
#include <net/icmp.h>
|
|
|
|
#include <net/xfrm.h>
|
2006-07-30 21:43:36 -06:00
|
|
|
#include <net/netevent.h>
|
2007-03-22 12:55:17 -06:00
|
|
|
#include <net/rtnetlink.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#ifdef CONFIG_SYSCTL
|
|
|
|
#include <linux/sysctl.h>
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#define RT_FL_TOS(oldflp) \
|
|
|
|
((u32)(oldflp->fl4_tos & (IPTOS_RT_MASK | RTO_ONLINK)))
|
|
|
|
|
|
|
|
#define IP_MAX_MTU 0xFFF0
|
|
|
|
|
|
|
|
#define RT_GC_TIMEOUT (300*HZ)
|
|
|
|
|
|
|
|
static int ip_rt_max_size;
|
2008-03-22 18:43:59 -06:00
|
|
|
static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT;
|
|
|
|
static int ip_rt_gc_interval __read_mostly = 60 * HZ;
|
|
|
|
static int ip_rt_gc_min_interval __read_mostly = HZ / 2;
|
|
|
|
static int ip_rt_redirect_number __read_mostly = 9;
|
|
|
|
static int ip_rt_redirect_load __read_mostly = HZ / 50;
|
|
|
|
static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1));
|
|
|
|
static int ip_rt_error_cost __read_mostly = HZ;
|
|
|
|
static int ip_rt_error_burst __read_mostly = 5 * HZ;
|
|
|
|
static int ip_rt_gc_elasticity __read_mostly = 8;
|
|
|
|
static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ;
|
|
|
|
static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20;
|
|
|
|
static int ip_rt_min_advmss __read_mostly = 256;
|
|
|
|
static int ip_rt_secret_interval __read_mostly = 10 * 60 * HZ;
|
2008-10-27 13:28:25 -06:00
|
|
|
static int rt_chain_length_max __read_mostly = 20;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2009-06-11 14:10:07 -06:00
|
|
|
static struct delayed_work expires_work;
|
|
|
|
static unsigned long expires_ljiffies;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Interface to generic destination cache.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
|
|
|
|
static void ipv4_dst_destroy(struct dst_entry *dst);
|
|
|
|
static void ipv4_dst_ifdown(struct dst_entry *dst,
|
|
|
|
struct net_device *dev, int how);
|
|
|
|
static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
|
|
|
|
static void ipv4_link_failure(struct sk_buff *skb);
|
|
|
|
static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
|
2008-01-18 04:56:57 -07:00
|
|
|
static int rt_garbage_collect(struct dst_ops *ops);
|
2008-10-27 13:28:25 -06:00
|
|
|
static void rt_emergency_hash_rebuild(struct net *net);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
|
|
|
|
static struct dst_ops ipv4_dst_ops = {
|
|
|
|
.family = AF_INET,
|
2009-02-01 01:45:17 -07:00
|
|
|
.protocol = cpu_to_be16(ETH_P_IP),
|
2005-04-16 16:20:36 -06:00
|
|
|
.gc = rt_garbage_collect,
|
|
|
|
.check = ipv4_dst_check,
|
|
|
|
.destroy = ipv4_dst_destroy,
|
|
|
|
.ifdown = ipv4_dst_ifdown,
|
|
|
|
.negative_advice = ipv4_negative_advice,
|
|
|
|
.link_failure = ipv4_link_failure,
|
|
|
|
.update_pmtu = ip_rt_update_pmtu,
|
2008-05-20 15:32:14 -06:00
|
|
|
.local_out = __ip_local_out,
|
2008-01-30 21:07:45 -07:00
|
|
|
.entries = ATOMIC_INIT(0),
|
2005-04-16 16:20:36 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
#define ECN_OR_COST(class) TC_PRIO_##class
|
|
|
|
|
2007-07-09 16:32:57 -06:00
|
|
|
const __u8 ip_tos2prio[16] = {
|
2005-04-16 16:20:36 -06:00
|
|
|
TC_PRIO_BESTEFFORT,
|
|
|
|
ECN_OR_COST(FILLER),
|
|
|
|
TC_PRIO_BESTEFFORT,
|
|
|
|
ECN_OR_COST(BESTEFFORT),
|
|
|
|
TC_PRIO_BULK,
|
|
|
|
ECN_OR_COST(BULK),
|
|
|
|
TC_PRIO_BULK,
|
|
|
|
ECN_OR_COST(BULK),
|
|
|
|
TC_PRIO_INTERACTIVE,
|
|
|
|
ECN_OR_COST(INTERACTIVE),
|
|
|
|
TC_PRIO_INTERACTIVE,
|
|
|
|
ECN_OR_COST(INTERACTIVE),
|
|
|
|
TC_PRIO_INTERACTIVE_BULK,
|
|
|
|
ECN_OR_COST(INTERACTIVE_BULK),
|
|
|
|
TC_PRIO_INTERACTIVE_BULK,
|
|
|
|
ECN_OR_COST(INTERACTIVE_BULK)
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Route cache.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* The locking scheme is rather straight forward:
|
|
|
|
*
|
|
|
|
* 1) Read-Copy Update protects the buckets of the central route hash.
|
|
|
|
* 2) Only writers remove entries, and they hold the lock
|
|
|
|
* as they look at rtable reference counts.
|
|
|
|
* 3) Only readers acquire references to rtable entries,
|
|
|
|
* they do so with atomic increments and with the
|
|
|
|
* lock held.
|
|
|
|
*/
|
|
|
|
|
|
|
|
struct rt_hash_bucket {
|
|
|
|
struct rtable *chain;
|
2005-07-05 15:55:24 -06:00
|
|
|
};
|
2008-10-27 13:28:25 -06:00
|
|
|
|
2006-07-03 01:24:54 -06:00
|
|
|
#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \
|
|
|
|
defined(CONFIG_PROVE_LOCKING)
|
2005-07-05 15:55:24 -06:00
|
|
|
/*
|
|
|
|
* Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks
|
|
|
|
* The size of this table is a power of two and depends on the number of CPUS.
|
2006-07-03 01:24:59 -06:00
|
|
|
* (on lockdep we have a quite big spinlock_t, so keep the size down there)
|
2005-07-05 15:55:24 -06:00
|
|
|
*/
|
2006-07-03 01:24:59 -06:00
|
|
|
#ifdef CONFIG_LOCKDEP
|
|
|
|
# define RT_HASH_LOCK_SZ 256
|
2005-07-05 15:55:24 -06:00
|
|
|
#else
|
2006-07-03 01:24:59 -06:00
|
|
|
# if NR_CPUS >= 32
|
|
|
|
# define RT_HASH_LOCK_SZ 4096
|
|
|
|
# elif NR_CPUS >= 16
|
|
|
|
# define RT_HASH_LOCK_SZ 2048
|
|
|
|
# elif NR_CPUS >= 8
|
|
|
|
# define RT_HASH_LOCK_SZ 1024
|
|
|
|
# elif NR_CPUS >= 4
|
|
|
|
# define RT_HASH_LOCK_SZ 512
|
|
|
|
# else
|
|
|
|
# define RT_HASH_LOCK_SZ 256
|
|
|
|
# endif
|
2005-07-05 15:55:24 -06:00
|
|
|
#endif
|
|
|
|
|
|
|
|
static spinlock_t *rt_hash_locks;
|
|
|
|
# define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)]
|
2007-12-05 22:15:05 -07:00
|
|
|
|
|
|
|
static __init void rt_hash_lock_init(void)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ,
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!rt_hash_locks)
|
|
|
|
panic("IP: failed to allocate rt_hash_locks\n");
|
|
|
|
|
|
|
|
for (i = 0; i < RT_HASH_LOCK_SZ; i++)
|
|
|
|
spin_lock_init(&rt_hash_locks[i]);
|
|
|
|
}
|
2005-07-05 15:55:24 -06:00
|
|
|
#else
|
|
|
|
# define rt_hash_lock_addr(slot) NULL
|
2007-12-05 22:15:05 -07:00
|
|
|
|
|
|
|
static inline void rt_hash_lock_init(void)
|
|
|
|
{
|
|
|
|
}
|
2005-07-05 15:55:24 -06:00
|
|
|
#endif
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-03-22 18:43:59 -06:00
|
|
|
static struct rt_hash_bucket *rt_hash_table __read_mostly;
|
|
|
|
static unsigned rt_hash_mask __read_mostly;
|
|
|
|
static unsigned int rt_hash_log __read_mostly;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
[IPV4]: rt_cache_stat can be statically defined
Using __get_cpu_var(obj) is slightly faster than per_cpu_ptr(obj,
raw_smp_processor_id()).
1) Smaller code and memory use
For static and small objects, DEFINE_PER_CPU(type, object) is preferred over a
alloc_percpu() : Better and smaller code to access them, and no extra memory
(storing the pointer, and the percpu array of pointers)
x86_64 code before patch
mov 1237577(%rip),%rax # ffffffff803e5990 <rt_cache_stat>
not %rax # part of per_cpu machinery
mov %gs:0x3c,%edx # get cpu number
movslq %edx,%rdx # extend 32 bits cpu number to 64 bits
mov (%rax,%rdx,8),%rax # get the pointer for this cpu
incl 0x38(%rax)
x86_64 code after patch
mov $per_cpu__rt_cache_stat,%rdx
mov %gs:0x48,%rax # get percpu data offset
incl 0x38(%rax,%rdx,1)
2) False sharing avoidance for SMP :
For a small NR_CPUS, the array of per cpu pointers allocated in alloc_percpu()
can be <= 32 bytes. This let slab code gives a part of a cache line. If the
other part of this 64 bytes (or 128 bytes) cache line is used by a mostly
written object, we can have false sharing and expensive per_cpu_ptr() operations.
Size of rt_cache_stat is 64 bytes, so this patch is not a danger of a too big
increase of bss (in UP mode) or static per_cpu data for SMP
(PERCPU_ENOUGH_ROOM is currently 32768 bytes)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-01-17 03:54:36 -07:00
|
|
|
static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
|
2006-01-17 22:58:01 -07:00
|
|
|
#define RT_CACHE_STAT_INC(field) \
|
[PATCH] Define __raw_get_cpu_var and use it
There are several instances of per_cpu(foo, raw_smp_processor_id()), which
is semantically equivalent to __get_cpu_var(foo) but without the warning
that smp_processor_id() can give if CONFIG_DEBUG_PREEMPT is enabled. For
those architectures with optimized per-cpu implementations, namely ia64,
powerpc, s390, sparc64 and x86_64, per_cpu() turns into more and slower
code than __get_cpu_var(), so it would be preferable to use __get_cpu_var
on those platforms.
This defines a __raw_get_cpu_var(x) macro which turns into per_cpu(x,
raw_smp_processor_id()) on architectures that use the generic per-cpu
implementation, and turns into __get_cpu_var(x) on the architectures that
have an optimized per-cpu implementation.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 06:47:14 -06:00
|
|
|
(__raw_get_cpu_var(rt_cache_stat).field++)
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-07-05 20:04:09 -06:00
|
|
|
static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx,
|
|
|
|
int genid)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-04-10 02:54:01 -06:00
|
|
|
return jhash_3words((__force u32)(__be32)(daddr),
|
|
|
|
(__force u32)(__be32)(saddr),
|
2008-07-05 20:04:09 -06:00
|
|
|
idx, genid)
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
& rt_hash_mask;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-07-05 20:04:32 -06:00
|
|
|
static inline int rt_genid(struct net *net)
|
|
|
|
{
|
|
|
|
return atomic_read(&net->ipv4.rt_genid);
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
struct rt_cache_iter_state {
|
2008-02-28 21:50:55 -07:00
|
|
|
struct seq_net_private p;
|
2005-04-16 16:20:36 -06:00
|
|
|
int bucket;
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
int genid;
|
2005-04-16 16:20:36 -06:00
|
|
|
};
|
|
|
|
|
2008-03-25 11:36:06 -06:00
|
|
|
static struct rtable *rt_cache_get_first(struct seq_file *seq)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-03-25 11:36:06 -06:00
|
|
|
struct rt_cache_iter_state *st = seq->private;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct rtable *r = NULL;
|
|
|
|
|
|
|
|
for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) {
|
2008-08-28 02:11:25 -06:00
|
|
|
if (!rt_hash_table[st->bucket].chain)
|
|
|
|
continue;
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_lock_bh();
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
r = rcu_dereference(rt_hash_table[st->bucket].chain);
|
|
|
|
while (r) {
|
2008-03-25 11:36:06 -06:00
|
|
|
if (dev_net(r->u.dst.dev) == seq_file_net(seq) &&
|
2008-02-28 21:50:55 -07:00
|
|
|
r->rt_genid == st->genid)
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
return r;
|
|
|
|
r = rcu_dereference(r->u.dst.rt_next);
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_unlock_bh();
|
|
|
|
}
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
return r;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-03-25 11:36:06 -06:00
|
|
|
static struct rtable *__rt_cache_get_next(struct seq_file *seq,
|
2008-02-28 21:50:33 -07:00
|
|
|
struct rtable *r)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-03-25 11:36:06 -06:00
|
|
|
struct rt_cache_iter_state *st = seq->private;
|
2008-08-28 02:11:25 -06:00
|
|
|
|
2007-02-09 17:19:26 -07:00
|
|
|
r = r->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
while (!r) {
|
|
|
|
rcu_read_unlock_bh();
|
2008-08-28 02:11:25 -06:00
|
|
|
do {
|
|
|
|
if (--st->bucket < 0)
|
|
|
|
return NULL;
|
|
|
|
} while (!rt_hash_table[st->bucket].chain);
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_lock_bh();
|
|
|
|
r = rt_hash_table[st->bucket].chain;
|
|
|
|
}
|
2008-01-10 04:55:57 -07:00
|
|
|
return rcu_dereference(r);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-03-25 11:36:06 -06:00
|
|
|
static struct rtable *rt_cache_get_next(struct seq_file *seq,
|
2008-02-28 21:50:33 -07:00
|
|
|
struct rtable *r)
|
|
|
|
{
|
2008-03-25 11:36:06 -06:00
|
|
|
struct rt_cache_iter_state *st = seq->private;
|
|
|
|
while ((r = __rt_cache_get_next(seq, r)) != NULL) {
|
|
|
|
if (dev_net(r->u.dst.dev) != seq_file_net(seq))
|
2008-02-28 21:50:55 -07:00
|
|
|
continue;
|
2008-02-28 21:50:33 -07:00
|
|
|
if (r->rt_genid == st->genid)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
|
2008-03-25 11:36:06 -06:00
|
|
|
static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-03-25 11:36:06 -06:00
|
|
|
struct rtable *r = rt_cache_get_first(seq);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (r)
|
2008-03-25 11:36:06 -06:00
|
|
|
while (pos && (r = rt_cache_get_next(seq, r)))
|
2005-04-16 16:20:36 -06:00
|
|
|
--pos;
|
|
|
|
return pos ? NULL : r;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
|
|
|
|
{
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
struct rt_cache_iter_state *st = seq->private;
|
|
|
|
if (*pos)
|
2008-03-25 11:36:06 -06:00
|
|
|
return rt_cache_get_idx(seq, *pos - 1);
|
2008-07-05 20:04:32 -06:00
|
|
|
st->genid = rt_genid(seq_file_net(seq));
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
return SEQ_START_TOKEN;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
|
|
|
|
{
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
struct rtable *r;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (v == SEQ_START_TOKEN)
|
2008-03-25 11:36:06 -06:00
|
|
|
r = rt_cache_get_first(seq);
|
2005-04-16 16:20:36 -06:00
|
|
|
else
|
2008-03-25 11:36:06 -06:00
|
|
|
r = rt_cache_get_next(seq, v);
|
2005-04-16 16:20:36 -06:00
|
|
|
++*pos;
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rt_cache_seq_stop(struct seq_file *seq, void *v)
|
|
|
|
{
|
|
|
|
if (v && v != SEQ_START_TOKEN)
|
|
|
|
rcu_read_unlock_bh();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int rt_cache_seq_show(struct seq_file *seq, void *v)
|
|
|
|
{
|
|
|
|
if (v == SEQ_START_TOKEN)
|
|
|
|
seq_printf(seq, "%-127s\n",
|
|
|
|
"Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
|
|
|
|
"Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
|
|
|
|
"HHUptod\tSpecDst");
|
|
|
|
else {
|
|
|
|
struct rtable *r = v;
|
2008-04-24 02:02:16 -06:00
|
|
|
int len;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-04-24 02:02:16 -06:00
|
|
|
seq_printf(seq, "%s\t%08lX\t%08lX\t%8X\t%d\t%u\t%d\t"
|
|
|
|
"%08lX\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n",
|
2005-04-16 16:20:36 -06:00
|
|
|
r->u.dst.dev ? r->u.dst.dev->name : "*",
|
|
|
|
(unsigned long)r->rt_dst, (unsigned long)r->rt_gateway,
|
|
|
|
r->rt_flags, atomic_read(&r->u.dst.__refcnt),
|
|
|
|
r->u.dst.__use, 0, (unsigned long)r->rt_src,
|
|
|
|
(dst_metric(&r->u.dst, RTAX_ADVMSS) ?
|
|
|
|
(int)dst_metric(&r->u.dst, RTAX_ADVMSS) + 40 : 0),
|
|
|
|
dst_metric(&r->u.dst, RTAX_WINDOW),
|
|
|
|
(int)((dst_metric(&r->u.dst, RTAX_RTT) >> 3) +
|
|
|
|
dst_metric(&r->u.dst, RTAX_RTTVAR)),
|
|
|
|
r->fl.fl4_tos,
|
|
|
|
r->u.dst.hh ? atomic_read(&r->u.dst.hh->hh_refcnt) : -1,
|
|
|
|
r->u.dst.hh ? (r->u.dst.hh->hh_output ==
|
|
|
|
dev_queue_xmit) : 0,
|
2008-04-24 02:02:16 -06:00
|
|
|
r->rt_spec_dst, &len);
|
|
|
|
|
|
|
|
seq_printf(seq, "%*s\n", 127 - len, "");
|
2007-02-09 07:24:47 -07:00
|
|
|
}
|
|
|
|
return 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-03-12 15:34:29 -06:00
|
|
|
static const struct seq_operations rt_cache_seq_ops = {
|
2005-04-16 16:20:36 -06:00
|
|
|
.start = rt_cache_seq_start,
|
|
|
|
.next = rt_cache_seq_next,
|
|
|
|
.stop = rt_cache_seq_stop,
|
|
|
|
.show = rt_cache_seq_show,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int rt_cache_seq_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
2008-02-28 21:50:55 -07:00
|
|
|
return seq_open_net(inode, file, &rt_cache_seq_ops,
|
2007-10-10 03:29:29 -06:00
|
|
|
sizeof(struct rt_cache_iter_state));
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-02-12 01:55:35 -07:00
|
|
|
static const struct file_operations rt_cache_seq_fops = {
|
2005-04-16 16:20:36 -06:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.open = rt_cache_seq_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
2008-02-28 21:50:55 -07:00
|
|
|
.release = seq_release_net,
|
2005-04-16 16:20:36 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
if (*pos == 0)
|
|
|
|
return SEQ_START_TOKEN;
|
|
|
|
|
2008-12-29 05:23:42 -07:00
|
|
|
for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!cpu_possible(cpu))
|
|
|
|
continue;
|
|
|
|
*pos = cpu+1;
|
[IPV4]: rt_cache_stat can be statically defined
Using __get_cpu_var(obj) is slightly faster than per_cpu_ptr(obj,
raw_smp_processor_id()).
1) Smaller code and memory use
For static and small objects, DEFINE_PER_CPU(type, object) is preferred over a
alloc_percpu() : Better and smaller code to access them, and no extra memory
(storing the pointer, and the percpu array of pointers)
x86_64 code before patch
mov 1237577(%rip),%rax # ffffffff803e5990 <rt_cache_stat>
not %rax # part of per_cpu machinery
mov %gs:0x3c,%edx # get cpu number
movslq %edx,%rdx # extend 32 bits cpu number to 64 bits
mov (%rax,%rdx,8),%rax # get the pointer for this cpu
incl 0x38(%rax)
x86_64 code after patch
mov $per_cpu__rt_cache_stat,%rdx
mov %gs:0x48,%rax # get percpu data offset
incl 0x38(%rax,%rdx,1)
2) False sharing avoidance for SMP :
For a small NR_CPUS, the array of per cpu pointers allocated in alloc_percpu()
can be <= 32 bytes. This let slab code gives a part of a cache line. If the
other part of this 64 bytes (or 128 bytes) cache line is used by a mostly
written object, we can have false sharing and expensive per_cpu_ptr() operations.
Size of rt_cache_stat is 64 bytes, so this patch is not a danger of a too big
increase of bss (in UP mode) or static per_cpu data for SMP
(PERCPU_ENOUGH_ROOM is currently 32768 bytes)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-01-17 03:54:36 -07:00
|
|
|
return &per_cpu(rt_cache_stat, cpu);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
2008-12-29 05:23:42 -07:00
|
|
|
for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!cpu_possible(cpu))
|
|
|
|
continue;
|
|
|
|
*pos = cpu+1;
|
[IPV4]: rt_cache_stat can be statically defined
Using __get_cpu_var(obj) is slightly faster than per_cpu_ptr(obj,
raw_smp_processor_id()).
1) Smaller code and memory use
For static and small objects, DEFINE_PER_CPU(type, object) is preferred over a
alloc_percpu() : Better and smaller code to access them, and no extra memory
(storing the pointer, and the percpu array of pointers)
x86_64 code before patch
mov 1237577(%rip),%rax # ffffffff803e5990 <rt_cache_stat>
not %rax # part of per_cpu machinery
mov %gs:0x3c,%edx # get cpu number
movslq %edx,%rdx # extend 32 bits cpu number to 64 bits
mov (%rax,%rdx,8),%rax # get the pointer for this cpu
incl 0x38(%rax)
x86_64 code after patch
mov $per_cpu__rt_cache_stat,%rdx
mov %gs:0x48,%rax # get percpu data offset
incl 0x38(%rax,%rdx,1)
2) False sharing avoidance for SMP :
For a small NR_CPUS, the array of per cpu pointers allocated in alloc_percpu()
can be <= 32 bytes. This let slab code gives a part of a cache line. If the
other part of this 64 bytes (or 128 bytes) cache line is used by a mostly
written object, we can have false sharing and expensive per_cpu_ptr() operations.
Size of rt_cache_stat is 64 bytes, so this patch is not a danger of a too big
increase of bss (in UP mode) or static per_cpu data for SMP
(PERCPU_ENOUGH_ROOM is currently 32768 bytes)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-01-17 03:54:36 -07:00
|
|
|
return &per_cpu(rt_cache_stat, cpu);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
return NULL;
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
|
|
|
|
{
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static int rt_cpu_seq_show(struct seq_file *seq, void *v)
|
|
|
|
{
|
|
|
|
struct rt_cache_stat *st = v;
|
|
|
|
|
|
|
|
if (v == SEQ_START_TOKEN) {
|
2005-04-28 13:16:08 -06:00
|
|
|
seq_printf(seq, "entries in_hit in_slow_tot in_slow_mc in_no_route in_brd in_martian_dst in_martian_src out_hit out_slow_tot out_slow_mc gc_total gc_ignored gc_goal_miss gc_dst_overflow in_hlist_search out_hlist_search\n");
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
|
|
|
}
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x "
|
|
|
|
" %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
|
|
|
|
atomic_read(&ipv4_dst_ops.entries),
|
|
|
|
st->in_hit,
|
|
|
|
st->in_slow_tot,
|
|
|
|
st->in_slow_mc,
|
|
|
|
st->in_no_route,
|
|
|
|
st->in_brd,
|
|
|
|
st->in_martian_dst,
|
|
|
|
st->in_martian_src,
|
|
|
|
|
|
|
|
st->out_hit,
|
|
|
|
st->out_slow_tot,
|
2007-02-09 07:24:47 -07:00
|
|
|
st->out_slow_mc,
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
st->gc_total,
|
|
|
|
st->gc_ignored,
|
|
|
|
st->gc_goal_miss,
|
|
|
|
st->gc_dst_overflow,
|
|
|
|
st->in_hlist_search,
|
|
|
|
st->out_hlist_search
|
|
|
|
);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2007-03-12 15:34:29 -06:00
|
|
|
static const struct seq_operations rt_cpu_seq_ops = {
|
2005-04-16 16:20:36 -06:00
|
|
|
.start = rt_cpu_seq_start,
|
|
|
|
.next = rt_cpu_seq_next,
|
|
|
|
.stop = rt_cpu_seq_stop,
|
|
|
|
.show = rt_cpu_seq_show,
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
static int rt_cpu_seq_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
return seq_open(file, &rt_cpu_seq_ops);
|
|
|
|
}
|
|
|
|
|
2007-02-12 01:55:35 -07:00
|
|
|
static const struct file_operations rt_cpu_seq_fops = {
|
2005-04-16 16:20:36 -06:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.open = rt_cpu_seq_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
|
|
|
.release = seq_release,
|
|
|
|
};
|
|
|
|
|
2007-12-05 22:13:48 -07:00
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
2009-11-25 16:40:35 -07:00
|
|
|
static int rt_acct_proc_show(struct seq_file *m, void *v)
|
2007-12-05 22:13:48 -07:00
|
|
|
{
|
2009-11-25 16:40:35 -07:00
|
|
|
struct ip_rt_acct *dst, *src;
|
|
|
|
unsigned int i, j;
|
|
|
|
|
|
|
|
dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
|
|
|
|
if (!dst)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for_each_possible_cpu(i) {
|
|
|
|
src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i);
|
|
|
|
for (j = 0; j < 256; j++) {
|
|
|
|
dst[j].o_bytes += src[j].o_bytes;
|
|
|
|
dst[j].o_packets += src[j].o_packets;
|
|
|
|
dst[j].i_bytes += src[j].i_bytes;
|
|
|
|
dst[j].i_packets += src[j].i_packets;
|
|
|
|
}
|
2007-12-05 22:13:48 -07:00
|
|
|
}
|
|
|
|
|
2009-11-25 16:40:35 -07:00
|
|
|
seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
|
|
|
|
kfree(dst);
|
|
|
|
return 0;
|
|
|
|
}
|
2007-12-05 22:13:48 -07:00
|
|
|
|
2009-11-25 16:40:35 -07:00
|
|
|
static int rt_acct_proc_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
return single_open(file, rt_acct_proc_show, NULL);
|
2007-12-05 22:13:48 -07:00
|
|
|
}
|
2009-11-25 16:40:35 -07:00
|
|
|
|
|
|
|
static const struct file_operations rt_acct_proc_fops = {
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.open = rt_acct_proc_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
|
|
|
.release = single_release,
|
|
|
|
};
|
2007-12-05 22:13:48 -07:00
|
|
|
#endif
|
2007-12-05 22:14:28 -07:00
|
|
|
|
2008-02-28 21:51:18 -07:00
|
|
|
static int __net_init ip_rt_do_proc_init(struct net *net)
|
2007-12-05 22:14:28 -07:00
|
|
|
{
|
|
|
|
struct proc_dir_entry *pde;
|
|
|
|
|
|
|
|
pde = proc_net_fops_create(net, "rt_cache", S_IRUGO,
|
|
|
|
&rt_cache_seq_fops);
|
|
|
|
if (!pde)
|
|
|
|
goto err1;
|
|
|
|
|
2008-02-28 15:14:25 -07:00
|
|
|
pde = proc_create("rt_cache", S_IRUGO,
|
|
|
|
net->proc_net_stat, &rt_cpu_seq_fops);
|
2007-12-05 22:14:28 -07:00
|
|
|
if (!pde)
|
|
|
|
goto err2;
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
2009-11-25 16:40:35 -07:00
|
|
|
pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops);
|
2007-12-05 22:14:28 -07:00
|
|
|
if (!pde)
|
|
|
|
goto err3;
|
|
|
|
#endif
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
err3:
|
|
|
|
remove_proc_entry("rt_cache", net->proc_net_stat);
|
|
|
|
#endif
|
|
|
|
err2:
|
|
|
|
remove_proc_entry("rt_cache", net->proc_net);
|
|
|
|
err1:
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2008-02-28 21:51:18 -07:00
|
|
|
|
|
|
|
static void __net_exit ip_rt_do_proc_exit(struct net *net)
|
|
|
|
{
|
|
|
|
remove_proc_entry("rt_cache", net->proc_net_stat);
|
|
|
|
remove_proc_entry("rt_cache", net->proc_net);
|
|
|
|
remove_proc_entry("rt_acct", net->proc_net);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct pernet_operations ip_rt_proc_ops __net_initdata = {
|
|
|
|
.init = ip_rt_do_proc_init,
|
|
|
|
.exit = ip_rt_do_proc_exit,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init ip_rt_proc_init(void)
|
|
|
|
{
|
|
|
|
return register_pernet_subsys(&ip_rt_proc_ops);
|
|
|
|
}
|
|
|
|
|
2007-12-05 22:14:28 -07:00
|
|
|
#else
|
2008-02-28 21:51:18 -07:00
|
|
|
static inline int ip_rt_proc_init(void)
|
2007-12-05 22:14:28 -07:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif /* CONFIG_PROC_FS */
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static inline void rt_free(struct rtable *rt)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free);
|
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static inline void rt_drop(struct rtable *rt)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
ip_rt_put(rt);
|
|
|
|
call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free);
|
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static inline int rt_fast_clean(struct rtable *rth)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
/* Kill broadcast/multicast entries very aggresively, if they
|
|
|
|
collide in hash table with more useful entries */
|
|
|
|
return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
|
2007-02-09 17:19:26 -07:00
|
|
|
rth->fl.iif && rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static inline int rt_valuable(struct rtable *rth)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) ||
|
|
|
|
rth->u.dst.expires;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2)
|
|
|
|
{
|
|
|
|
unsigned long age;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
if (atomic_read(&rth->u.dst.__refcnt))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
ret = 1;
|
|
|
|
if (rth->u.dst.expires &&
|
|
|
|
time_after_eq(jiffies, rth->u.dst.expires))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
age = jiffies - rth->u.dst.lastuse;
|
|
|
|
ret = 0;
|
|
|
|
if ((age <= tmo1 && !rt_fast_clean(rth)) ||
|
|
|
|
(age <= tmo2 && rt_valuable(rth)))
|
|
|
|
goto out;
|
|
|
|
ret = 1;
|
|
|
|
out: return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Bits of score are:
|
|
|
|
* 31: very valuable
|
|
|
|
* 30: not quite useless
|
|
|
|
* 29..0: usage counter
|
|
|
|
*/
|
|
|
|
static inline u32 rt_score(struct rtable *rt)
|
|
|
|
{
|
|
|
|
u32 score = jiffies - rt->u.dst.lastuse;
|
|
|
|
|
|
|
|
score = ~score & ~(3<<30);
|
|
|
|
|
|
|
|
if (rt_valuable(rt))
|
|
|
|
score |= (1<<31);
|
|
|
|
|
|
|
|
if (!rt->fl.iif ||
|
|
|
|
!(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL)))
|
|
|
|
score |= (1<<30);
|
|
|
|
|
|
|
|
return score;
|
|
|
|
}
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
static inline bool rt_caching(const struct net *net)
|
|
|
|
{
|
|
|
|
return net->ipv4.current_rt_cache_rebuild_count <=
|
|
|
|
net->ipv4.sysctl_rt_cache_rebuild_count;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool compare_hash_inputs(const struct flowi *fl1,
|
|
|
|
const struct flowi *fl2)
|
|
|
|
{
|
|
|
|
return (__force u32)(((fl1->nl_u.ip4_u.daddr ^ fl2->nl_u.ip4_u.daddr) |
|
|
|
|
(fl1->nl_u.ip4_u.saddr ^ fl2->nl_u.ip4_u.saddr) |
|
|
|
|
(fl1->iif ^ fl2->iif)) == 0);
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
static inline int compare_keys(struct flowi *fl1, struct flowi *fl2)
|
|
|
|
{
|
2006-11-14 21:51:49 -07:00
|
|
|
return ((__force u32)((fl1->nl_u.ip4_u.daddr ^ fl2->nl_u.ip4_u.daddr) |
|
|
|
|
(fl1->nl_u.ip4_u.saddr ^ fl2->nl_u.ip4_u.saddr)) |
|
2006-11-09 16:20:38 -07:00
|
|
|
(fl1->mark ^ fl2->mark) |
|
2006-10-12 01:49:15 -06:00
|
|
|
(*(u16 *)&fl1->nl_u.ip4_u.tos ^
|
|
|
|
*(u16 *)&fl2->nl_u.ip4_u.tos) |
|
|
|
|
(fl1->oif ^ fl2->oif) |
|
|
|
|
(fl1->iif ^ fl2->iif)) == 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-01-23 00:50:25 -07:00
|
|
|
static inline int compare_netns(struct rtable *rt1, struct rtable *rt2)
|
|
|
|
{
|
2009-11-25 16:14:13 -07:00
|
|
|
return net_eq(dev_net(rt1->u.dst.dev), dev_net(rt2->u.dst.dev));
|
2008-01-23 00:50:25 -07:00
|
|
|
}
|
|
|
|
|
2008-07-05 20:04:32 -06:00
|
|
|
static inline int rt_is_expired(struct rtable *rth)
|
|
|
|
{
|
|
|
|
return rth->rt_genid != rt_genid(dev_net(rth->u.dst.dev));
|
|
|
|
}
|
|
|
|
|
2007-11-19 23:43:37 -07:00
|
|
|
/*
|
|
|
|
* Perform a full scan of hash table and free all entries.
|
|
|
|
* Can be called by a softirq or a process.
|
|
|
|
* In the later case, we want to be reschedule if necessary
|
|
|
|
*/
|
|
|
|
static void rt_do_flush(int process_context)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
struct rtable *rth, *next;
|
2008-07-05 20:06:12 -06:00
|
|
|
struct rtable * tail;
|
2007-11-19 23:43:37 -07:00
|
|
|
|
|
|
|
for (i = 0; i <= rt_hash_mask; i++) {
|
|
|
|
if (process_context && need_resched())
|
|
|
|
cond_resched();
|
|
|
|
rth = rt_hash_table[i].chain;
|
|
|
|
if (!rth)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
spin_lock_bh(rt_hash_lock_addr(i));
|
2008-07-05 20:06:12 -06:00
|
|
|
#ifdef CONFIG_NET_NS
|
|
|
|
{
|
|
|
|
struct rtable ** prev, * p;
|
|
|
|
|
|
|
|
rth = rt_hash_table[i].chain;
|
|
|
|
|
|
|
|
/* defer releasing the head of the list after spin_unlock */
|
|
|
|
for (tail = rth; tail; tail = tail->u.dst.rt_next)
|
|
|
|
if (!rt_is_expired(tail))
|
|
|
|
break;
|
|
|
|
if (rth != tail)
|
|
|
|
rt_hash_table[i].chain = tail;
|
|
|
|
|
|
|
|
/* call rt_free on entries after the tail requiring flush */
|
|
|
|
prev = &rt_hash_table[i].chain;
|
|
|
|
for (p = *prev; p; p = next) {
|
|
|
|
next = p->u.dst.rt_next;
|
|
|
|
if (!rt_is_expired(p)) {
|
|
|
|
prev = &p->u.dst.rt_next;
|
|
|
|
} else {
|
|
|
|
*prev = next;
|
|
|
|
rt_free(p);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#else
|
2007-11-19 23:43:37 -07:00
|
|
|
rth = rt_hash_table[i].chain;
|
|
|
|
rt_hash_table[i].chain = NULL;
|
2008-07-05 20:06:12 -06:00
|
|
|
tail = NULL;
|
|
|
|
#endif
|
2007-11-19 23:43:37 -07:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(i));
|
|
|
|
|
2008-07-05 20:06:12 -06:00
|
|
|
for (; rth != tail; rth = next) {
|
2007-11-19 23:43:37 -07:00
|
|
|
next = rth->u.dst.rt_next;
|
|
|
|
rt_free(rth);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
/*
|
|
|
|
* While freeing expired entries, we compute average chain length
|
|
|
|
* and standard deviation, using fixed-point arithmetic.
|
|
|
|
* This to have an estimation of rt_chain_length_max
|
|
|
|
* rt_chain_length_max = max(elasticity, AVG + 4*SD)
|
|
|
|
* We use 3 bits for frational part, and 29 (or 61) for magnitude.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#define FRACT_BITS 3
|
|
|
|
#define ONE (1UL << FRACT_BITS)
|
|
|
|
|
2007-11-19 23:43:37 -07:00
|
|
|
static void rt_check_expire(void)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2005-07-05 16:00:32 -06:00
|
|
|
static unsigned int rover;
|
|
|
|
unsigned int i = rover, goal;
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
struct rtable *rth, *aux, **rthp;
|
2009-05-19 12:54:22 -06:00
|
|
|
unsigned long samples = 0;
|
2008-10-27 13:28:25 -06:00
|
|
|
unsigned long sum = 0, sum2 = 0;
|
2009-06-11 14:10:07 -06:00
|
|
|
unsigned long delta;
|
2005-07-05 16:00:32 -06:00
|
|
|
u64 mult;
|
|
|
|
|
2009-06-11 14:10:07 -06:00
|
|
|
delta = jiffies - expires_ljiffies;
|
|
|
|
expires_ljiffies = jiffies;
|
|
|
|
mult = ((u64)delta) << rt_hash_log;
|
2005-07-05 16:00:32 -06:00
|
|
|
if (ip_rt_gc_timeout > 1)
|
|
|
|
do_div(mult, ip_rt_gc_timeout);
|
|
|
|
goal = (unsigned int)mult;
|
2007-09-15 11:55:54 -06:00
|
|
|
if (goal > rt_hash_mask)
|
|
|
|
goal = rt_hash_mask + 1;
|
2005-07-05 16:00:32 -06:00
|
|
|
for (; goal > 0; goal--) {
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long tmo = ip_rt_gc_timeout;
|
2009-05-19 12:54:22 -06:00
|
|
|
unsigned long length;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
i = (i + 1) & rt_hash_mask;
|
|
|
|
rthp = &rt_hash_table[i].chain;
|
|
|
|
|
2007-11-14 17:14:05 -07:00
|
|
|
if (need_resched())
|
|
|
|
cond_resched();
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
samples++;
|
|
|
|
|
2007-10-09 02:59:42 -06:00
|
|
|
if (*rthp == NULL)
|
2005-07-05 16:00:32 -06:00
|
|
|
continue;
|
2009-05-19 12:54:22 -06:00
|
|
|
length = 0;
|
2007-09-15 11:55:54 -06:00
|
|
|
spin_lock_bh(rt_hash_lock_addr(i));
|
2005-04-16 16:20:36 -06:00
|
|
|
while ((rth = *rthp) != NULL) {
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
prefetch(rth->u.dst.rt_next);
|
2008-07-05 20:04:32 -06:00
|
|
|
if (rt_is_expired(rth)) {
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
*rthp = rth->u.dst.rt_next;
|
|
|
|
rt_free(rth);
|
|
|
|
continue;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rth->u.dst.expires) {
|
|
|
|
/* Entry is expired even if it is in use */
|
2007-09-15 11:55:54 -06:00
|
|
|
if (time_before_eq(jiffies, rth->u.dst.expires)) {
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
nofree:
|
2005-04-16 16:20:36 -06:00
|
|
|
tmo >>= 1;
|
2007-02-09 17:19:26 -07:00
|
|
|
rthp = &rth->u.dst.rt_next;
|
2008-10-27 13:28:25 -06:00
|
|
|
/*
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
* We only count entries on
|
2008-10-27 13:28:25 -06:00
|
|
|
* a chain with equal hash inputs once
|
|
|
|
* so that entries for different QOS
|
|
|
|
* levels, and other non-hash input
|
|
|
|
* attributes don't unfairly skew
|
|
|
|
* the length computation
|
|
|
|
*/
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
for (aux = rt_hash_table[i].chain;;) {
|
|
|
|
if (aux == rth) {
|
|
|
|
length += ONE;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (compare_hash_inputs(&aux->fl, &rth->fl))
|
|
|
|
break;
|
|
|
|
aux = aux->u.dst.rt_next;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
|
|
|
}
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
} else if (!rt_may_expire(rth, tmo, ip_rt_gc_timeout))
|
|
|
|
goto nofree;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* Cleanup aged off entries. */
|
2007-02-09 17:19:26 -07:00
|
|
|
*rthp = rth->u.dst.rt_next;
|
2007-02-09 07:24:47 -07:00
|
|
|
rt_free(rth);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-09-15 11:55:54 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(i));
|
2008-10-27 13:28:25 -06:00
|
|
|
sum += length;
|
|
|
|
sum2 += length*length;
|
|
|
|
}
|
|
|
|
if (samples) {
|
|
|
|
unsigned long avg = sum / samples;
|
|
|
|
unsigned long sd = int_sqrt(sum2 / samples - avg*avg);
|
|
|
|
rt_chain_length_max = max_t(unsigned long,
|
|
|
|
ip_rt_gc_elasticity,
|
|
|
|
(avg + 4*sd) >> FRACT_BITS);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
rover = i;
|
2007-11-19 23:43:37 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* rt_worker_func() is run in process context.
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
* we call rt_check_expire() to scan part of the hash table
|
2007-11-19 23:43:37 -07:00
|
|
|
*/
|
|
|
|
static void rt_worker_func(struct work_struct *work)
|
|
|
|
{
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
rt_check_expire();
|
2007-09-15 11:55:54 -06:00
|
|
|
schedule_delayed_work(&expires_work, ip_rt_gc_interval);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
/*
|
|
|
|
* Pertubation of rt_genid by a small quantity [1..256]
|
|
|
|
* Using 8 bits of shuffling ensure we can call rt_cache_invalidate()
|
|
|
|
* many times (2^24) without giving recent rt_genid.
|
|
|
|
* Jenkins hash is strong enough that litle changes of rt_genid are OK.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2008-07-05 20:03:31 -06:00
|
|
|
static void rt_cache_invalidate(struct net *net)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
unsigned char shuffle;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
get_random_bytes(&shuffle, sizeof(shuffle));
|
2008-07-05 20:04:32 -06:00
|
|
|
atomic_add(shuffle + 1U, &net->ipv4.rt_genid);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
/*
|
|
|
|
* delay < 0 : invalidate cache (fast : entries will be deleted later)
|
|
|
|
* delay >= 0 : invalidate & flush cache (can be long)
|
|
|
|
*/
|
2008-07-05 20:00:44 -06:00
|
|
|
void rt_cache_flush(struct net *net, int delay)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-07-05 20:03:31 -06:00
|
|
|
rt_cache_invalidate(net);
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
if (delay >= 0)
|
|
|
|
rt_do_flush(!in_softirq());
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2009-11-29 08:45:58 -07:00
|
|
|
/* Flush previous cache invalidated entries from the cache */
|
|
|
|
void rt_cache_flush_batch(void)
|
|
|
|
{
|
|
|
|
rt_do_flush(!in_softirq());
|
|
|
|
}
|
|
|
|
|
2007-11-19 23:43:37 -07:00
|
|
|
/*
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
* We change rt_genid and let gc do the cleanup
|
2007-11-19 23:43:37 -07:00
|
|
|
*/
|
2008-07-05 20:02:59 -06:00
|
|
|
static void rt_secret_rebuild(unsigned long __net)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-07-05 20:02:59 -06:00
|
|
|
struct net *net = (struct net *)__net;
|
2008-07-05 20:03:31 -06:00
|
|
|
rt_cache_invalidate(net);
|
2008-07-05 20:02:59 -06:00
|
|
|
mod_timer(&net->ipv4.rt_secret_timer, jiffies + ip_rt_secret_interval);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
static void rt_secret_rebuild_oneshot(struct net *net)
|
|
|
|
{
|
|
|
|
del_timer_sync(&net->ipv4.rt_secret_timer);
|
|
|
|
rt_cache_invalidate(net);
|
|
|
|
if (ip_rt_secret_interval) {
|
|
|
|
net->ipv4.rt_secret_timer.expires += ip_rt_secret_interval;
|
|
|
|
add_timer(&net->ipv4.rt_secret_timer);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rt_emergency_hash_rebuild(struct net *net)
|
|
|
|
{
|
|
|
|
if (net_ratelimit()) {
|
|
|
|
printk(KERN_WARNING "Route hash chain too long!\n");
|
|
|
|
printk(KERN_WARNING "Adjust your secret_interval!\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
rt_secret_rebuild_oneshot(net);
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
Short description of GC goals.
|
|
|
|
|
|
|
|
We want to build algorithm, which will keep routing cache
|
|
|
|
at some equilibrium point, when number of aged off entries
|
|
|
|
is kept approximately equal to newly generated ones.
|
|
|
|
|
|
|
|
Current expiration strength is variable "expire".
|
|
|
|
We try to adjust it dynamically, so that if networking
|
|
|
|
is idle expires is large enough to keep enough of warm entries,
|
|
|
|
and when load increases it reduces to limit cache size.
|
|
|
|
*/
|
|
|
|
|
2008-01-18 04:56:57 -07:00
|
|
|
static int rt_garbage_collect(struct dst_ops *ops)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
static unsigned long expire = RT_GC_TIMEOUT;
|
|
|
|
static unsigned long last_gc;
|
|
|
|
static int rover;
|
|
|
|
static int equilibrium;
|
|
|
|
struct rtable *rth, **rthp;
|
|
|
|
unsigned long now = jiffies;
|
|
|
|
int goal;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Garbage collection is pretty expensive,
|
|
|
|
* do not make it too frequently.
|
|
|
|
*/
|
|
|
|
|
|
|
|
RT_CACHE_STAT_INC(gc_total);
|
|
|
|
|
|
|
|
if (now - last_gc < ip_rt_gc_min_interval &&
|
|
|
|
atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) {
|
|
|
|
RT_CACHE_STAT_INC(gc_ignored);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Calculate number of entries, which we want to expire now. */
|
|
|
|
goal = atomic_read(&ipv4_dst_ops.entries) -
|
|
|
|
(ip_rt_gc_elasticity << rt_hash_log);
|
|
|
|
if (goal <= 0) {
|
|
|
|
if (equilibrium < ipv4_dst_ops.gc_thresh)
|
|
|
|
equilibrium = ipv4_dst_ops.gc_thresh;
|
|
|
|
goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium;
|
|
|
|
if (goal > 0) {
|
2007-12-21 02:49:07 -07:00
|
|
|
equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1);
|
2005-04-16 16:20:36 -06:00
|
|
|
goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* We are in dangerous area. Try to reduce cache really
|
|
|
|
* aggressively.
|
|
|
|
*/
|
2007-12-21 02:49:07 -07:00
|
|
|
goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1);
|
2005-04-16 16:20:36 -06:00
|
|
|
equilibrium = atomic_read(&ipv4_dst_ops.entries) - goal;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (now - last_gc >= ip_rt_gc_min_interval)
|
|
|
|
last_gc = now;
|
|
|
|
|
|
|
|
if (goal <= 0) {
|
|
|
|
equilibrium += goal;
|
|
|
|
goto work_done;
|
|
|
|
}
|
|
|
|
|
|
|
|
do {
|
|
|
|
int i, k;
|
|
|
|
|
|
|
|
for (i = rt_hash_mask, k = rover; i >= 0; i--) {
|
|
|
|
unsigned long tmo = expire;
|
|
|
|
|
|
|
|
k = (k + 1) & rt_hash_mask;
|
|
|
|
rthp = &rt_hash_table[k].chain;
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_lock_bh(rt_hash_lock_addr(k));
|
2005-04-16 16:20:36 -06:00
|
|
|
while ((rth = *rthp) != NULL) {
|
2008-07-05 20:04:32 -06:00
|
|
|
if (!rt_is_expired(rth) &&
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
!rt_may_expire(rth, tmo, expire)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
tmo >>= 1;
|
2007-02-09 17:19:26 -07:00
|
|
|
rthp = &rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
|
|
|
}
|
2007-02-09 17:19:26 -07:00
|
|
|
*rthp = rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
rt_free(rth);
|
|
|
|
goal--;
|
|
|
|
}
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(k));
|
2005-04-16 16:20:36 -06:00
|
|
|
if (goal <= 0)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
rover = k;
|
|
|
|
|
|
|
|
if (goal <= 0)
|
|
|
|
goto work_done;
|
|
|
|
|
|
|
|
/* Goal is not achieved. We stop process if:
|
|
|
|
|
|
|
|
- if expire reduced to zero. Otherwise, expire is halfed.
|
|
|
|
- if table is not full.
|
|
|
|
- if we are called from interrupt.
|
|
|
|
- jiffies check is just fallback/debug loop breaker.
|
|
|
|
We will not spin here for long time in any case.
|
|
|
|
*/
|
|
|
|
|
|
|
|
RT_CACHE_STAT_INC(gc_goal_miss);
|
|
|
|
|
|
|
|
if (expire == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
expire >>= 1;
|
|
|
|
#if RT_CACHE_DEBUG >= 2
|
|
|
|
printk(KERN_DEBUG "expire>> %u %d %d %d\n", expire,
|
|
|
|
atomic_read(&ipv4_dst_ops.entries), goal, i);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size)
|
|
|
|
goto out;
|
|
|
|
} while (!in_softirq() && time_before_eq(jiffies, now));
|
|
|
|
|
|
|
|
if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size)
|
|
|
|
goto out;
|
|
|
|
if (net_ratelimit())
|
|
|
|
printk(KERN_WARNING "dst cache overflow\n");
|
|
|
|
RT_CACHE_STAT_INC(gc_dst_overflow);
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
work_done:
|
|
|
|
expire += ip_rt_gc_min_interval;
|
|
|
|
if (expire > ip_rt_gc_timeout ||
|
|
|
|
atomic_read(&ipv4_dst_ops.entries) < ipv4_dst_ops.gc_thresh)
|
|
|
|
expire = ip_rt_gc_timeout;
|
|
|
|
#if RT_CACHE_DEBUG >= 2
|
|
|
|
printk(KERN_DEBUG "expire++ %u %d %d %d\n", expire,
|
|
|
|
atomic_read(&ipv4_dst_ops.entries), goal, rover);
|
|
|
|
#endif
|
|
|
|
out: return 0;
|
|
|
|
}
|
|
|
|
|
2009-06-01 23:14:27 -06:00
|
|
|
static int rt_intern_hash(unsigned hash, struct rtable *rt,
|
|
|
|
struct rtable **rp, struct sk_buff *skb)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
struct rtable *rth, **rthp;
|
|
|
|
unsigned long now;
|
|
|
|
struct rtable *cand, **candp;
|
|
|
|
u32 min_score;
|
|
|
|
int chain_length;
|
|
|
|
int attempts = !in_softirq();
|
|
|
|
|
|
|
|
restart:
|
|
|
|
chain_length = 0;
|
|
|
|
min_score = ~(u32)0;
|
|
|
|
cand = NULL;
|
|
|
|
candp = NULL;
|
|
|
|
now = jiffies;
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
if (!rt_caching(dev_net(rt->u.dst.dev))) {
|
ipv4: fix NULL pointer + success return in route lookup path
Don't drop route if we're not caching
I recently got a report of an oops on a route lookup. Maxime was
testing what would happen if route caching was turned off (doing so by setting
making rt_caching always return 0), and found that it triggered an oops. I
looked at it and found that the problem stemmed from the fact that the route
lookup routines were returning success from their lookup paths (which is good),
but never set the **rp pointer to anything (which is bad). This happens because
in rt_intern_hash, if rt_caching returns false, we call rt_drop and return 0.
This almost emulates slient success. What we should be doing is assigning *rp =
rt and _not_ dropping the route. This way, during slow path lookups, when we
create a new route cache entry, we don't immediately discard it, rather we just
don't add it into the cache hash table, but we let this one lookup use it for
the purpose of this route request. Maxime has tested and reports it prevents
the oops. There is still a subsequent routing issue that I'm looking into
further, but I'm confident that, even if its related to this same path, this
patch makes sense to take.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-20 02:15:16 -06:00
|
|
|
/*
|
|
|
|
* If we're not caching, just tell the caller we
|
|
|
|
* were successful and don't touch the route. The
|
|
|
|
* caller hold the sole reference to the cache entry, and
|
|
|
|
* it will be released when the caller is done with it.
|
|
|
|
* If we drop it here, the callers have no way to resolve routes
|
|
|
|
* when we're not caching. Instead, just point *rp at rt, so
|
|
|
|
* the caller gets a single use out of the route
|
ipv4 routing: Ensure that route cache entries are usable and reclaimable with caching is off
When route caching is disabled (rt_caching returns false), We still use route
cache entries that are created and passed into rt_intern_hash once. These
routes need to be made usable for the one call path that holds a reference to
them, and they need to be reclaimed when they're finished with their use. To be
made usable, they need to be associated with a neighbor table entry (which they
currently are not), otherwise iproute_finish2 just discards the packet, since we
don't know which L2 peer to send the packet to. To do this binding, we need to
follow the path a bit higher up in rt_intern_hash, which calls
arp_bind_neighbour, but not assign the route entry to the hash table.
Currently, if caching is off, we simply assign the route to the rp pointer and
are reutrn success. This patch associates us with a neighbor entry first.
Secondly, we need to make sure that any single use routes like this are known to
the garbage collector when caching is off. If caching is off, and we try to
hash in a route, it will leak when its refcount reaches zero. To avoid this,
this patch calls rt_free on the route cache entry passed into rt_intern_hash.
This places us on the gc list for the route cache garbage collector, so that
when its refcount reaches zero, it will be reclaimed (Thanks to Alexey for this
suggestion).
I've tested this on a local system here, and with these patches in place, I'm
able to maintain routed connectivity to remote systems, even if I set
/proc/sys/net/ipv4/rt_cache_rebuild_count to -1, which forces rt_caching to
return false.
Signed-off-by: Neil Horman <nhorman@redhat.com>
Reported-by: Jarek Poplawski <jarkao2@gmail.com>
Reported-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-22 04:18:53 -06:00
|
|
|
* Note that we do rt_free on this new route entry, so that
|
|
|
|
* once its refcount hits zero, we are still able to reap it
|
|
|
|
* (Thanks Alexey)
|
|
|
|
* Note also the rt_free uses call_rcu. We don't actually
|
|
|
|
* need rcu protection here, this is just our path to get
|
|
|
|
* on the route gc list.
|
ipv4: fix NULL pointer + success return in route lookup path
Don't drop route if we're not caching
I recently got a report of an oops on a route lookup. Maxime was
testing what would happen if route caching was turned off (doing so by setting
making rt_caching always return 0), and found that it triggered an oops. I
looked at it and found that the problem stemmed from the fact that the route
lookup routines were returning success from their lookup paths (which is good),
but never set the **rp pointer to anything (which is bad). This happens because
in rt_intern_hash, if rt_caching returns false, we call rt_drop and return 0.
This almost emulates slient success. What we should be doing is assigning *rp =
rt and _not_ dropping the route. This way, during slow path lookups, when we
create a new route cache entry, we don't immediately discard it, rather we just
don't add it into the cache hash table, but we let this one lookup use it for
the purpose of this route request. Maxime has tested and reports it prevents
the oops. There is still a subsequent routing issue that I'm looking into
further, but I'm confident that, even if its related to this same path, this
patch makes sense to take.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-20 02:15:16 -06:00
|
|
|
*/
|
ipv4 routing: Ensure that route cache entries are usable and reclaimable with caching is off
When route caching is disabled (rt_caching returns false), We still use route
cache entries that are created and passed into rt_intern_hash once. These
routes need to be made usable for the one call path that holds a reference to
them, and they need to be reclaimed when they're finished with their use. To be
made usable, they need to be associated with a neighbor table entry (which they
currently are not), otherwise iproute_finish2 just discards the packet, since we
don't know which L2 peer to send the packet to. To do this binding, we need to
follow the path a bit higher up in rt_intern_hash, which calls
arp_bind_neighbour, but not assign the route entry to the hash table.
Currently, if caching is off, we simply assign the route to the rp pointer and
are reutrn success. This patch associates us with a neighbor entry first.
Secondly, we need to make sure that any single use routes like this are known to
the garbage collector when caching is off. If caching is off, and we try to
hash in a route, it will leak when its refcount reaches zero. To avoid this,
this patch calls rt_free on the route cache entry passed into rt_intern_hash.
This places us on the gc list for the route cache garbage collector, so that
when its refcount reaches zero, it will be reclaimed (Thanks to Alexey for this
suggestion).
I've tested this on a local system here, and with these patches in place, I'm
able to maintain routed connectivity to remote systems, even if I set
/proc/sys/net/ipv4/rt_cache_rebuild_count to -1, which forces rt_caching to
return false.
Signed-off-by: Neil Horman <nhorman@redhat.com>
Reported-by: Jarek Poplawski <jarkao2@gmail.com>
Reported-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-22 04:18:53 -06:00
|
|
|
|
|
|
|
if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) {
|
|
|
|
int err = arp_bind_neighbour(&rt->u.dst);
|
|
|
|
if (err) {
|
|
|
|
if (net_ratelimit())
|
|
|
|
printk(KERN_WARNING
|
|
|
|
"Neighbour table failure & not caching routes.\n");
|
|
|
|
rt_drop(rt);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
rt_free(rt);
|
|
|
|
goto skip_hashing;
|
2008-10-27 13:28:25 -06:00
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
rthp = &rt_hash_table[hash].chain;
|
|
|
|
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_lock_bh(rt_hash_lock_addr(hash));
|
2005-04-16 16:20:36 -06:00
|
|
|
while ((rth = *rthp) != NULL) {
|
2008-07-05 20:04:32 -06:00
|
|
|
if (rt_is_expired(rth)) {
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
*rthp = rth->u.dst.rt_next;
|
|
|
|
rt_free(rth);
|
|
|
|
continue;
|
|
|
|
}
|
2008-01-23 00:50:25 -07:00
|
|
|
if (compare_keys(&rth->fl, &rt->fl) && compare_netns(rth, rt)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Put it first */
|
2007-02-09 17:19:26 -07:00
|
|
|
*rthp = rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* Since lookup is lockfree, the deletion
|
|
|
|
* must be visible to another weakly ordered CPU before
|
|
|
|
* the insertion at the start of the hash chain.
|
|
|
|
*/
|
2007-02-09 17:19:26 -07:00
|
|
|
rcu_assign_pointer(rth->u.dst.rt_next,
|
2005-04-16 16:20:36 -06:00
|
|
|
rt_hash_table[hash].chain);
|
|
|
|
/*
|
|
|
|
* Since lookup is lockfree, the update writes
|
|
|
|
* must be ordered for consistency on SMP.
|
|
|
|
*/
|
|
|
|
rcu_assign_pointer(rt_hash_table[hash].chain, rth);
|
|
|
|
|
2007-11-10 22:28:34 -07:00
|
|
|
dst_use(&rth->u.dst, now);
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(hash));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rt_drop(rt);
|
2009-06-01 23:14:27 -06:00
|
|
|
if (rp)
|
|
|
|
*rp = rth;
|
|
|
|
else
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_set(skb, &rth->u.dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!atomic_read(&rth->u.dst.__refcnt)) {
|
|
|
|
u32 score = rt_score(rth);
|
|
|
|
|
|
|
|
if (score <= min_score) {
|
|
|
|
cand = rth;
|
|
|
|
candp = rthp;
|
|
|
|
min_score = score;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
chain_length++;
|
|
|
|
|
2007-02-09 17:19:26 -07:00
|
|
|
rthp = &rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
if (cand) {
|
|
|
|
/* ip_rt_gc_elasticity used to be average length of chain
|
|
|
|
* length, when exceeded gc becomes really aggressive.
|
|
|
|
*
|
|
|
|
* The second limit is less certain. At the moment it allows
|
|
|
|
* only 2 entries per bucket. We will see.
|
|
|
|
*/
|
|
|
|
if (chain_length > ip_rt_gc_elasticity) {
|
2007-02-09 17:19:26 -07:00
|
|
|
*candp = cand->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
rt_free(cand);
|
|
|
|
}
|
2008-10-27 13:28:25 -06:00
|
|
|
} else {
|
|
|
|
if (chain_length > rt_chain_length_max) {
|
|
|
|
struct net *net = dev_net(rt->u.dst.dev);
|
|
|
|
int num = ++net->ipv4.current_rt_cache_rebuild_count;
|
|
|
|
if (!rt_caching(dev_net(rt->u.dst.dev))) {
|
|
|
|
printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n",
|
|
|
|
rt->u.dst.dev->name, num);
|
|
|
|
}
|
|
|
|
rt_emergency_hash_rebuild(dev_net(rt->u.dst.dev));
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Try to bind route to arp only if it is output
|
|
|
|
route or unicast forwarding path.
|
|
|
|
*/
|
|
|
|
if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) {
|
|
|
|
int err = arp_bind_neighbour(&rt->u.dst);
|
|
|
|
if (err) {
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(hash));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (err != -ENOBUFS) {
|
|
|
|
rt_drop(rt);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Neighbour tables are full and nothing
|
|
|
|
can be released. Try to shrink route cache,
|
|
|
|
it is most likely it holds some neighbour records.
|
|
|
|
*/
|
|
|
|
if (attempts-- > 0) {
|
|
|
|
int saved_elasticity = ip_rt_gc_elasticity;
|
|
|
|
int saved_int = ip_rt_gc_min_interval;
|
|
|
|
ip_rt_gc_elasticity = 1;
|
|
|
|
ip_rt_gc_min_interval = 0;
|
2008-01-18 04:56:57 -07:00
|
|
|
rt_garbage_collect(&ipv4_dst_ops);
|
2005-04-16 16:20:36 -06:00
|
|
|
ip_rt_gc_min_interval = saved_int;
|
|
|
|
ip_rt_gc_elasticity = saved_elasticity;
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (net_ratelimit())
|
|
|
|
printk(KERN_WARNING "Neighbour table overflow.\n");
|
|
|
|
rt_drop(rt);
|
|
|
|
return -ENOBUFS;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
rt->u.dst.rt_next = rt_hash_table[hash].chain;
|
2008-10-27 13:28:25 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
#if RT_CACHE_DEBUG >= 2
|
2007-02-09 17:19:26 -07:00
|
|
|
if (rt->u.dst.rt_next) {
|
2005-04-16 16:20:36 -06:00
|
|
|
struct rtable *trt;
|
ipv4 routing: Ensure that route cache entries are usable and reclaimable with caching is off
When route caching is disabled (rt_caching returns false), We still use route
cache entries that are created and passed into rt_intern_hash once. These
routes need to be made usable for the one call path that holds a reference to
them, and they need to be reclaimed when they're finished with their use. To be
made usable, they need to be associated with a neighbor table entry (which they
currently are not), otherwise iproute_finish2 just discards the packet, since we
don't know which L2 peer to send the packet to. To do this binding, we need to
follow the path a bit higher up in rt_intern_hash, which calls
arp_bind_neighbour, but not assign the route entry to the hash table.
Currently, if caching is off, we simply assign the route to the rp pointer and
are reutrn success. This patch associates us with a neighbor entry first.
Secondly, we need to make sure that any single use routes like this are known to
the garbage collector when caching is off. If caching is off, and we try to
hash in a route, it will leak when its refcount reaches zero. To avoid this,
this patch calls rt_free on the route cache entry passed into rt_intern_hash.
This places us on the gc list for the route cache garbage collector, so that
when its refcount reaches zero, it will be reclaimed (Thanks to Alexey for this
suggestion).
I've tested this on a local system here, and with these patches in place, I'm
able to maintain routed connectivity to remote systems, even if I set
/proc/sys/net/ipv4/rt_cache_rebuild_count to -1, which forces rt_caching to
return false.
Signed-off-by: Neil Horman <nhorman@redhat.com>
Reported-by: Jarek Poplawski <jarkao2@gmail.com>
Reported-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-22 04:18:53 -06:00
|
|
|
printk(KERN_DEBUG "rt_cache @%02x: %pI4",
|
|
|
|
hash, &rt->rt_dst);
|
2007-02-09 17:19:26 -07:00
|
|
|
for (trt = rt->u.dst.rt_next; trt; trt = trt->u.dst.rt_next)
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(" . %pI4", &trt->rt_dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
printk("\n");
|
|
|
|
}
|
|
|
|
#endif
|
2008-10-16 15:18:29 -06:00
|
|
|
/*
|
|
|
|
* Since lookup is lockfree, we must make sure
|
|
|
|
* previous writes to rt are comitted to memory
|
|
|
|
* before making rt visible to other CPUS.
|
|
|
|
*/
|
net: fix rtable leak in net/ipv4/route.c
Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
Quoted here because its a perfect one :
begin_of_quotation
2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
patch has at least one critical flaw, and another problem.
rt_intern_hash calculates rthi pointer, which is later used for new entry
insertion. The same loop calculates cand pointer which is used to clean the
list. If the pointers are the same, rtable leak occurs, as first the cand is
removed then the new entry is appended to it.
This leak leads to unregister_netdevice problem (usage count > 0).
Another problem of the patch is that it tries to insert the entries in certain
order, to facilitate counting of entries distinct by all but QoS parameters.
Unfortunately, referencing an existing rtable entry moves it to list beginning,
to speed up further lookups, so the carefully built order is destroyed.
For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
it will also destroy the ordering.
end_of_quotation
Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
(net: implement emergency route cache rebulds when gc_elasticity is exceeded)
Trying to keep dst_entries ordered is too complex and breaks the fact that
order should depend on the frequency of use for garbage collection.
A possible fix is to make rt_intern_hash() simpler, and only makes
rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
entries order. The added loop is running on cache hot data, while cpu
is prefetching next object, so should be unnoticied.
Reported-and-analyzed-by: Alexander V. Lukyanov <lav@yar.ru>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-19 14:14:28 -06:00
|
|
|
rcu_assign_pointer(rt_hash_table[hash].chain, rt);
|
2008-10-27 13:28:25 -06:00
|
|
|
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(hash));
|
ipv4: fix NULL pointer + success return in route lookup path
Don't drop route if we're not caching
I recently got a report of an oops on a route lookup. Maxime was
testing what would happen if route caching was turned off (doing so by setting
making rt_caching always return 0), and found that it triggered an oops. I
looked at it and found that the problem stemmed from the fact that the route
lookup routines were returning success from their lookup paths (which is good),
but never set the **rp pointer to anything (which is bad). This happens because
in rt_intern_hash, if rt_caching returns false, we call rt_drop and return 0.
This almost emulates slient success. What we should be doing is assigning *rp =
rt and _not_ dropping the route. This way, during slow path lookups, when we
create a new route cache entry, we don't immediately discard it, rather we just
don't add it into the cache hash table, but we let this one lookup use it for
the purpose of this route request. Maxime has tested and reports it prevents
the oops. There is still a subsequent routing issue that I'm looking into
further, but I'm confident that, even if its related to this same path, this
patch makes sense to take.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-20 02:15:16 -06:00
|
|
|
|
ipv4 routing: Ensure that route cache entries are usable and reclaimable with caching is off
When route caching is disabled (rt_caching returns false), We still use route
cache entries that are created and passed into rt_intern_hash once. These
routes need to be made usable for the one call path that holds a reference to
them, and they need to be reclaimed when they're finished with their use. To be
made usable, they need to be associated with a neighbor table entry (which they
currently are not), otherwise iproute_finish2 just discards the packet, since we
don't know which L2 peer to send the packet to. To do this binding, we need to
follow the path a bit higher up in rt_intern_hash, which calls
arp_bind_neighbour, but not assign the route entry to the hash table.
Currently, if caching is off, we simply assign the route to the rp pointer and
are reutrn success. This patch associates us with a neighbor entry first.
Secondly, we need to make sure that any single use routes like this are known to
the garbage collector when caching is off. If caching is off, and we try to
hash in a route, it will leak when its refcount reaches zero. To avoid this,
this patch calls rt_free on the route cache entry passed into rt_intern_hash.
This places us on the gc list for the route cache garbage collector, so that
when its refcount reaches zero, it will be reclaimed (Thanks to Alexey for this
suggestion).
I've tested this on a local system here, and with these patches in place, I'm
able to maintain routed connectivity to remote systems, even if I set
/proc/sys/net/ipv4/rt_cache_rebuild_count to -1, which forces rt_caching to
return false.
Signed-off-by: Neil Horman <nhorman@redhat.com>
Reported-by: Jarek Poplawski <jarkao2@gmail.com>
Reported-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-06-22 04:18:53 -06:00
|
|
|
skip_hashing:
|
2009-06-01 23:14:27 -06:00
|
|
|
if (rp)
|
|
|
|
*rp = rt;
|
|
|
|
else
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_set(skb, &rt->u.dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void rt_bind_peer(struct rtable *rt, int create)
|
|
|
|
{
|
|
|
|
static DEFINE_SPINLOCK(rt_peer_lock);
|
|
|
|
struct inet_peer *peer;
|
|
|
|
|
|
|
|
peer = inet_getpeer(rt->rt_dst, create);
|
|
|
|
|
|
|
|
spin_lock_bh(&rt_peer_lock);
|
|
|
|
if (rt->peer == NULL) {
|
|
|
|
rt->peer = peer;
|
|
|
|
peer = NULL;
|
|
|
|
}
|
|
|
|
spin_unlock_bh(&rt_peer_lock);
|
|
|
|
if (peer)
|
|
|
|
inet_putpeer(peer);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Peer allocation may fail only in serious out-of-memory conditions. However
|
|
|
|
* we still can generate some output.
|
|
|
|
* Random ID selection looks a bit dangerous because we have no chances to
|
|
|
|
* select ID being unique in a reasonable period of time.
|
|
|
|
* But broken packet identifier may be better than no packet at all.
|
|
|
|
*/
|
|
|
|
static void ip_select_fb_ident(struct iphdr *iph)
|
|
|
|
{
|
|
|
|
static DEFINE_SPINLOCK(ip_fb_id_lock);
|
|
|
|
static u32 ip_fallback_id;
|
|
|
|
u32 salt;
|
|
|
|
|
|
|
|
spin_lock_bh(&ip_fb_id_lock);
|
2006-09-26 23:15:01 -06:00
|
|
|
salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr);
|
2005-04-16 16:20:36 -06:00
|
|
|
iph->id = htons(salt & 0xFFFF);
|
|
|
|
ip_fallback_id = salt;
|
|
|
|
spin_unlock_bh(&ip_fb_id_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more)
|
|
|
|
{
|
|
|
|
struct rtable *rt = (struct rtable *) dst;
|
|
|
|
|
|
|
|
if (rt) {
|
|
|
|
if (rt->peer == NULL)
|
|
|
|
rt_bind_peer(rt, 1);
|
|
|
|
|
|
|
|
/* If peer is attached to destination, it is never detached,
|
|
|
|
so that we need not to grab a lock to dereference it.
|
|
|
|
*/
|
|
|
|
if (rt->peer) {
|
|
|
|
iph->id = htons(inet_getid(rt->peer, more));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
} else
|
2007-02-09 07:24:47 -07:00
|
|
|
printk(KERN_DEBUG "rt_bind_peer(0) @%p\n",
|
2005-04-19 23:39:42 -06:00
|
|
|
__builtin_return_address(0));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
ip_select_fb_ident(iph);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rt_del(unsigned hash, struct rtable *rt)
|
|
|
|
{
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
struct rtable **rthp, *aux;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
rthp = &rt_hash_table[hash].chain;
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_lock_bh(rt_hash_lock_addr(hash));
|
2005-04-16 16:20:36 -06:00
|
|
|
ip_rt_put(rt);
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
while ((aux = *rthp) != NULL) {
|
2008-07-05 20:04:32 -06:00
|
|
|
if (aux == rt || rt_is_expired(aux)) {
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
*rthp = aux->u.dst.rt_next;
|
|
|
|
rt_free(aux);
|
|
|
|
continue;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
rthp = &aux->u.dst.rt_next;
|
|
|
|
}
|
2005-07-05 15:55:24 -06:00
|
|
|
spin_unlock_bh(rt_hash_lock_addr(hash));
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-09-26 22:25:43 -06:00
|
|
|
void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw,
|
|
|
|
__be32 saddr, struct net_device *dev)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int i, k;
|
|
|
|
struct in_device *in_dev = in_dev_get(dev);
|
|
|
|
struct rtable *rth, **rthp;
|
2006-09-26 22:25:43 -06:00
|
|
|
__be32 skeys[2] = { saddr, 0 };
|
2005-04-16 16:20:36 -06:00
|
|
|
int ikeys[2] = { dev->ifindex, 0 };
|
2006-07-30 21:43:36 -06:00
|
|
|
struct netevent_redirect netevent;
|
2008-02-28 21:50:06 -07:00
|
|
|
struct net *net;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (!in_dev)
|
|
|
|
return;
|
|
|
|
|
2008-03-25 06:47:49 -06:00
|
|
|
net = dev_net(dev);
|
2009-11-23 11:41:23 -07:00
|
|
|
if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
|
|
|
|
ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
|
|
|
|
ipv4_is_zeronet(new_gw))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto reject_redirect;
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
if (!rt_caching(net))
|
|
|
|
goto reject_redirect;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!IN_DEV_SHARED_MEDIA(in_dev)) {
|
|
|
|
if (!inet_addr_onlink(in_dev, new_gw, old_gw))
|
|
|
|
goto reject_redirect;
|
|
|
|
if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
|
|
|
|
goto reject_redirect;
|
|
|
|
} else {
|
2008-02-28 21:50:06 -07:00
|
|
|
if (inet_addr_type(net, new_gw) != RTN_UNICAST)
|
2005-04-16 16:20:36 -06:00
|
|
|
goto reject_redirect;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < 2; i++) {
|
|
|
|
for (k = 0; k < 2; k++) {
|
2008-07-05 20:04:09 -06:00
|
|
|
unsigned hash = rt_hash(daddr, skeys[i], ikeys[k],
|
2008-07-05 20:04:32 -06:00
|
|
|
rt_genid(net));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rthp=&rt_hash_table[hash].chain;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
while ((rth = rcu_dereference(*rthp)) != NULL) {
|
|
|
|
struct rtable *rt;
|
|
|
|
|
|
|
|
if (rth->fl.fl4_dst != daddr ||
|
|
|
|
rth->fl.fl4_src != skeys[i] ||
|
|
|
|
rth->fl.oif != ikeys[k] ||
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
rth->fl.iif != 0 ||
|
2008-07-05 20:04:32 -06:00
|
|
|
rt_is_expired(rth) ||
|
2008-03-25 12:57:35 -06:00
|
|
|
!net_eq(dev_net(rth->u.dst.dev), net)) {
|
2007-02-09 17:19:26 -07:00
|
|
|
rthp = &rth->u.dst.rt_next;
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (rth->rt_dst != daddr ||
|
|
|
|
rth->rt_src != saddr ||
|
|
|
|
rth->u.dst.error ||
|
|
|
|
rth->rt_gateway != old_gw ||
|
|
|
|
rth->u.dst.dev != dev)
|
|
|
|
break;
|
|
|
|
|
|
|
|
dst_hold(&rth->u.dst);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
rt = dst_alloc(&ipv4_dst_ops);
|
|
|
|
if (rt == NULL) {
|
|
|
|
ip_rt_put(rth);
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy all the information. */
|
|
|
|
*rt = *rth;
|
|
|
|
rt->u.dst.__use = 1;
|
|
|
|
atomic_set(&rt->u.dst.__refcnt, 1);
|
|
|
|
rt->u.dst.child = NULL;
|
|
|
|
if (rt->u.dst.dev)
|
|
|
|
dev_hold(rt->u.dst.dev);
|
|
|
|
if (rt->idev)
|
|
|
|
in_dev_hold(rt->idev);
|
|
|
|
rt->u.dst.obsolete = 0;
|
|
|
|
rt->u.dst.lastuse = jiffies;
|
|
|
|
rt->u.dst.path = &rt->u.dst;
|
|
|
|
rt->u.dst.neighbour = NULL;
|
|
|
|
rt->u.dst.hh = NULL;
|
2008-10-28 14:24:06 -06:00
|
|
|
#ifdef CONFIG_XFRM
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.xfrm = NULL;
|
2008-10-28 14:24:06 -06:00
|
|
|
#endif
|
2008-07-05 20:04:32 -06:00
|
|
|
rt->rt_genid = rt_genid(net);
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->rt_flags |= RTCF_REDIRECTED;
|
|
|
|
|
|
|
|
/* Gateway is different ... */
|
|
|
|
rt->rt_gateway = new_gw;
|
|
|
|
|
|
|
|
/* Redirect received -> path was valid */
|
|
|
|
dst_confirm(&rth->u.dst);
|
|
|
|
|
|
|
|
if (rt->peer)
|
|
|
|
atomic_inc(&rt->peer->refcnt);
|
|
|
|
|
|
|
|
if (arp_bind_neighbour(&rt->u.dst) ||
|
|
|
|
!(rt->u.dst.neighbour->nud_state &
|
|
|
|
NUD_VALID)) {
|
|
|
|
if (rt->u.dst.neighbour)
|
|
|
|
neigh_event_send(rt->u.dst.neighbour, NULL);
|
|
|
|
ip_rt_put(rth);
|
|
|
|
rt_drop(rt);
|
|
|
|
goto do_next;
|
|
|
|
}
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2006-07-30 21:43:36 -06:00
|
|
|
netevent.old = &rth->u.dst;
|
|
|
|
netevent.new = &rt->u.dst;
|
2007-02-09 07:24:47 -07:00
|
|
|
call_netevent_notifiers(NETEVENT_REDIRECT,
|
|
|
|
&netevent);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rt_del(hash, rth);
|
2009-06-01 23:14:27 -06:00
|
|
|
if (!rt_intern_hash(hash, rt, &rt, NULL))
|
2005-04-16 16:20:36 -06:00
|
|
|
ip_rt_put(rt);
|
|
|
|
goto do_next;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
do_next:
|
|
|
|
;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
return;
|
|
|
|
|
|
|
|
reject_redirect:
|
|
|
|
#ifdef CONFIG_IP_ROUTE_VERBOSE
|
|
|
|
if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n"
|
|
|
|
" Advised path = %pI4 -> %pI4\n",
|
|
|
|
&old_gw, dev->name, &new_gw,
|
|
|
|
&saddr, &daddr);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
|
|
|
|
{
|
2008-03-05 19:30:47 -07:00
|
|
|
struct rtable *rt = (struct rtable *)dst;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct dst_entry *ret = dst;
|
|
|
|
|
|
|
|
if (rt) {
|
|
|
|
if (dst->obsolete) {
|
|
|
|
ip_rt_put(rt);
|
|
|
|
ret = NULL;
|
|
|
|
} else if ((rt->rt_flags & RTCF_REDIRECTED) ||
|
|
|
|
rt->u.dst.expires) {
|
2006-09-26 22:26:19 -06:00
|
|
|
unsigned hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src,
|
2008-07-05 20:04:09 -06:00
|
|
|
rt->fl.oif,
|
2008-07-05 20:04:32 -06:00
|
|
|
rt_genid(dev_net(dst->dev)));
|
2005-04-16 16:20:36 -06:00
|
|
|
#if RT_CACHE_DEBUG >= 1
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_DEBUG "ipv4_negative_advice: redirect to %pI4/%02x dropped\n",
|
|
|
|
&rt->rt_dst, rt->fl.fl4_tos);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
|
|
|
rt_del(hash, rt);
|
|
|
|
ret = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Algorithm:
|
|
|
|
* 1. The first ip_rt_redirect_number redirects are sent
|
|
|
|
* with exponential backoff, then we stop sending them at all,
|
|
|
|
* assuming that the host ignores our redirects.
|
|
|
|
* 2. If we did not see packets requiring redirects
|
|
|
|
* during ip_rt_redirect_silence, we assume that the host
|
|
|
|
* forgot redirected route and start to send redirects again.
|
|
|
|
*
|
|
|
|
* This algorithm is much cheaper and more intelligent than dumb load limiting
|
|
|
|
* in icmp.c.
|
|
|
|
*
|
|
|
|
* NOTE. Do not forget to inhibit load limiting for redirects (redundant)
|
|
|
|
* and "frag. need" (breaks PMTU discovery) in icmp.c.
|
|
|
|
*/
|
|
|
|
|
|
|
|
void ip_rt_send_redirect(struct sk_buff *skb)
|
|
|
|
{
|
2009-06-01 23:14:27 -06:00
|
|
|
struct rtable *rt = skb_rtable(skb);
|
2009-08-29 00:52:01 -06:00
|
|
|
struct in_device *in_dev;
|
|
|
|
int log_martians;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2009-08-29 00:52:01 -06:00
|
|
|
rcu_read_lock();
|
|
|
|
in_dev = __in_dev_get_rcu(rt->u.dst.dev);
|
|
|
|
if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
|
|
|
|
rcu_read_unlock();
|
2005-04-16 16:20:36 -06:00
|
|
|
return;
|
2009-08-29 00:52:01 -06:00
|
|
|
}
|
|
|
|
log_martians = IN_DEV_LOG_MARTIANS(in_dev);
|
|
|
|
rcu_read_unlock();
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* No redirected packets during ip_rt_redirect_silence;
|
|
|
|
* reset the algorithm.
|
|
|
|
*/
|
|
|
|
if (time_after(jiffies, rt->u.dst.rate_last + ip_rt_redirect_silence))
|
|
|
|
rt->u.dst.rate_tokens = 0;
|
|
|
|
|
|
|
|
/* Too many ignored redirects; do not send anything
|
|
|
|
* set u.dst.rate_last to the last seen redirected packet.
|
|
|
|
*/
|
|
|
|
if (rt->u.dst.rate_tokens >= ip_rt_redirect_number) {
|
|
|
|
rt->u.dst.rate_last = jiffies;
|
2009-08-29 00:52:01 -06:00
|
|
|
return;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Check for load limit; set rate_last to the latest sent
|
|
|
|
* redirect.
|
|
|
|
*/
|
2006-12-18 01:26:35 -07:00
|
|
|
if (rt->u.dst.rate_tokens == 0 ||
|
|
|
|
time_after(jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
(rt->u.dst.rate_last +
|
|
|
|
(ip_rt_redirect_load << rt->u.dst.rate_tokens)))) {
|
|
|
|
icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway);
|
|
|
|
rt->u.dst.rate_last = jiffies;
|
|
|
|
++rt->u.dst.rate_tokens;
|
|
|
|
#ifdef CONFIG_IP_ROUTE_VERBOSE
|
2009-08-29 00:52:01 -06:00
|
|
|
if (log_martians &&
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.rate_tokens == ip_rt_redirect_number &&
|
|
|
|
net_ratelimit())
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n",
|
|
|
|
&rt->rt_src, rt->rt_iif,
|
|
|
|
&rt->rt_dst, &rt->rt_gateway);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ip_error(struct sk_buff *skb)
|
|
|
|
{
|
2009-06-01 23:14:27 -06:00
|
|
|
struct rtable *rt = skb_rtable(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long now;
|
|
|
|
int code;
|
|
|
|
|
|
|
|
switch (rt->u.dst.error) {
|
|
|
|
case EINVAL:
|
|
|
|
default:
|
|
|
|
goto out;
|
|
|
|
case EHOSTUNREACH:
|
|
|
|
code = ICMP_HOST_UNREACH;
|
|
|
|
break;
|
|
|
|
case ENETUNREACH:
|
|
|
|
code = ICMP_NET_UNREACH;
|
2008-07-16 21:20:11 -06:00
|
|
|
IP_INC_STATS_BH(dev_net(rt->u.dst.dev),
|
|
|
|
IPSTATS_MIB_INNOROUTES);
|
2005-04-16 16:20:36 -06:00
|
|
|
break;
|
|
|
|
case EACCES:
|
|
|
|
code = ICMP_PKT_FILTERED;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
now = jiffies;
|
|
|
|
rt->u.dst.rate_tokens += now - rt->u.dst.rate_last;
|
|
|
|
if (rt->u.dst.rate_tokens > ip_rt_error_burst)
|
|
|
|
rt->u.dst.rate_tokens = ip_rt_error_burst;
|
|
|
|
rt->u.dst.rate_last = now;
|
|
|
|
if (rt->u.dst.rate_tokens >= ip_rt_error_cost) {
|
|
|
|
rt->u.dst.rate_tokens -= ip_rt_error_cost;
|
|
|
|
icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
out: kfree_skb(skb);
|
|
|
|
return 0;
|
2007-02-09 07:24:47 -07:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The last two values are not from the RFC but
|
|
|
|
* are needed for AMPRnet AX.25 paths.
|
|
|
|
*/
|
|
|
|
|
2005-11-29 17:21:38 -07:00
|
|
|
static const unsigned short mtu_plateau[] =
|
2005-04-16 16:20:36 -06:00
|
|
|
{32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 };
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static inline unsigned short guess_mtu(unsigned short old_mtu)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int i;
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++)
|
|
|
|
if (old_mtu > mtu_plateau[i])
|
|
|
|
return mtu_plateau[i];
|
|
|
|
return 68;
|
|
|
|
}
|
|
|
|
|
2008-01-23 00:50:25 -07:00
|
|
|
unsigned short ip_rt_frag_needed(struct net *net, struct iphdr *iph,
|
2008-04-29 04:32:25 -06:00
|
|
|
unsigned short new_mtu,
|
|
|
|
struct net_device *dev)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-04-29 04:32:25 -06:00
|
|
|
int i, k;
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned short old_mtu = ntohs(iph->tot_len);
|
|
|
|
struct rtable *rth;
|
2008-04-29 04:32:25 -06:00
|
|
|
int ikeys[2] = { dev->ifindex, 0 };
|
2006-09-26 23:15:01 -06:00
|
|
|
__be32 skeys[2] = { iph->saddr, 0, };
|
|
|
|
__be32 daddr = iph->daddr;
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned short est_mtu = 0;
|
|
|
|
|
2008-04-29 04:32:25 -06:00
|
|
|
for (k = 0; k < 2; k++) {
|
|
|
|
for (i = 0; i < 2; i++) {
|
2008-07-05 20:04:09 -06:00
|
|
|
unsigned hash = rt_hash(daddr, skeys[i], ikeys[k],
|
2008-07-05 20:04:32 -06:00
|
|
|
rt_genid(net));
|
2008-04-29 04:32:25 -06:00
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
for (rth = rcu_dereference(rt_hash_table[hash].chain); rth;
|
|
|
|
rth = rcu_dereference(rth->u.dst.rt_next)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned short mtu = new_mtu;
|
|
|
|
|
2008-04-29 04:32:25 -06:00
|
|
|
if (rth->fl.fl4_dst != daddr ||
|
|
|
|
rth->fl.fl4_src != skeys[i] ||
|
|
|
|
rth->rt_dst != daddr ||
|
|
|
|
rth->rt_src != iph->saddr ||
|
|
|
|
rth->fl.oif != ikeys[k] ||
|
|
|
|
rth->fl.iif != 0 ||
|
|
|
|
dst_metric_locked(&rth->u.dst, RTAX_MTU) ||
|
|
|
|
!net_eq(dev_net(rth->u.dst.dev), net) ||
|
2008-07-26 18:51:06 -06:00
|
|
|
rt_is_expired(rth))
|
2008-04-29 04:32:25 -06:00
|
|
|
continue;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (new_mtu < 68 || new_mtu >= old_mtu) {
|
|
|
|
|
|
|
|
/* BSD 4.2 compatibility hack :-( */
|
|
|
|
if (mtu == 0 &&
|
2008-08-06 03:33:49 -06:00
|
|
|
old_mtu >= dst_mtu(&rth->u.dst) &&
|
2005-04-16 16:20:36 -06:00
|
|
|
old_mtu >= 68 + (iph->ihl << 2))
|
|
|
|
old_mtu -= iph->ihl << 2;
|
|
|
|
|
|
|
|
mtu = guess_mtu(old_mtu);
|
|
|
|
}
|
2008-08-06 03:33:49 -06:00
|
|
|
if (mtu <= dst_mtu(&rth->u.dst)) {
|
|
|
|
if (mtu < dst_mtu(&rth->u.dst)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
dst_confirm(&rth->u.dst);
|
|
|
|
if (mtu < ip_rt_min_pmtu) {
|
|
|
|
mtu = ip_rt_min_pmtu;
|
|
|
|
rth->u.dst.metrics[RTAX_LOCK-1] |=
|
|
|
|
(1 << RTAX_MTU);
|
|
|
|
}
|
|
|
|
rth->u.dst.metrics[RTAX_MTU-1] = mtu;
|
|
|
|
dst_set_expires(&rth->u.dst,
|
|
|
|
ip_rt_mtu_expires);
|
|
|
|
}
|
|
|
|
est_mtu = mtu;
|
|
|
|
}
|
|
|
|
}
|
2008-04-29 04:32:25 -06:00
|
|
|
rcu_read_unlock();
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
return est_mtu ? : new_mtu;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
|
|
|
|
{
|
2008-08-06 03:33:49 -06:00
|
|
|
if (dst_mtu(dst) > mtu && mtu >= 68 &&
|
2005-04-16 16:20:36 -06:00
|
|
|
!(dst_metric_locked(dst, RTAX_MTU))) {
|
|
|
|
if (mtu < ip_rt_min_pmtu) {
|
|
|
|
mtu = ip_rt_min_pmtu;
|
|
|
|
dst->metrics[RTAX_LOCK-1] |= (1 << RTAX_MTU);
|
|
|
|
}
|
|
|
|
dst->metrics[RTAX_MTU-1] = mtu;
|
|
|
|
dst_set_expires(dst, ip_rt_mtu_expires);
|
2006-07-30 21:43:36 -06:00
|
|
|
call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ipv4_dst_destroy(struct dst_entry *dst)
|
|
|
|
{
|
|
|
|
struct rtable *rt = (struct rtable *) dst;
|
|
|
|
struct inet_peer *peer = rt->peer;
|
|
|
|
struct in_device *idev = rt->idev;
|
|
|
|
|
|
|
|
if (peer) {
|
|
|
|
rt->peer = NULL;
|
|
|
|
inet_putpeer(peer);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (idev) {
|
|
|
|
rt->idev = NULL;
|
|
|
|
in_dev_put(idev);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
|
|
|
|
int how)
|
|
|
|
{
|
|
|
|
struct rtable *rt = (struct rtable *) dst;
|
|
|
|
struct in_device *idev = rt->idev;
|
2008-03-25 06:47:49 -06:00
|
|
|
if (dev != dev_net(dev)->loopback_dev && idev && idev->dev == dev) {
|
2007-12-07 01:38:10 -07:00
|
|
|
struct in_device *loopback_idev =
|
2008-03-25 06:47:49 -06:00
|
|
|
in_dev_get(dev_net(dev)->loopback_dev);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (loopback_idev) {
|
|
|
|
rt->idev = loopback_idev;
|
|
|
|
in_dev_put(idev);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ipv4_link_failure(struct sk_buff *skb)
|
|
|
|
{
|
|
|
|
struct rtable *rt;
|
|
|
|
|
|
|
|
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
|
|
|
|
|
2009-06-01 23:14:27 -06:00
|
|
|
rt = skb_rtable(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt)
|
|
|
|
dst_set_expires(&rt->u.dst, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ip_rt_bug(struct sk_buff *skb)
|
|
|
|
{
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n",
|
|
|
|
&ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
|
2005-04-16 16:20:36 -06:00
|
|
|
skb->dev ? skb->dev->name : "?");
|
|
|
|
kfree_skb(skb);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
We do not cache source address of outgoing interface,
|
|
|
|
because it is used only by IP RR, TS and SRR options,
|
|
|
|
so that it out of fast path.
|
|
|
|
|
|
|
|
BTW remember: "addr" is allowed to be not aligned
|
|
|
|
in IP options!
|
|
|
|
*/
|
|
|
|
|
|
|
|
void ip_rt_get_source(u8 *addr, struct rtable *rt)
|
|
|
|
{
|
2006-09-26 22:27:54 -06:00
|
|
|
__be32 src;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct fib_result res;
|
|
|
|
|
|
|
|
if (rt->fl.iif == 0)
|
|
|
|
src = rt->rt_src;
|
2008-03-25 06:47:49 -06:00
|
|
|
else if (fib_lookup(dev_net(rt->u.dst.dev), &rt->fl, &res) == 0) {
|
2005-04-16 16:20:36 -06:00
|
|
|
src = FIB_RES_PREFSRC(res);
|
|
|
|
fib_res_put(&res);
|
|
|
|
} else
|
|
|
|
src = inet_select_addr(rt->u.dst.dev, rt->rt_gateway,
|
|
|
|
RT_SCOPE_UNIVERSE);
|
|
|
|
memcpy(addr, &src, 4);
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
static void set_class_tag(struct rtable *rt, u32 tag)
|
|
|
|
{
|
|
|
|
if (!(rt->u.dst.tclassid & 0xFFFF))
|
|
|
|
rt->u.dst.tclassid |= tag & 0xFFFF;
|
|
|
|
if (!(rt->u.dst.tclassid & 0xFFFF0000))
|
|
|
|
rt->u.dst.tclassid |= tag & 0xFFFF0000;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
static void rt_set_nexthop(struct rtable *rt, struct fib_result *res, u32 itag)
|
|
|
|
{
|
|
|
|
struct fib_info *fi = res->fi;
|
|
|
|
|
|
|
|
if (fi) {
|
|
|
|
if (FIB_RES_GW(*res) &&
|
|
|
|
FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
|
|
|
|
rt->rt_gateway = FIB_RES_GW(*res);
|
|
|
|
memcpy(rt->u.dst.metrics, fi->fib_metrics,
|
|
|
|
sizeof(rt->u.dst.metrics));
|
|
|
|
if (fi->fib_mtu == 0) {
|
|
|
|
rt->u.dst.metrics[RTAX_MTU-1] = rt->u.dst.dev->mtu;
|
2008-05-04 23:12:43 -06:00
|
|
|
if (dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->rt_gateway != rt->rt_dst &&
|
|
|
|
rt->u.dst.dev->mtu > 576)
|
|
|
|
rt->u.dst.metrics[RTAX_MTU-1] = 576;
|
|
|
|
}
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
rt->u.dst.tclassid = FIB_RES_NH(*res).nh_tclassid;
|
|
|
|
#endif
|
|
|
|
} else
|
|
|
|
rt->u.dst.metrics[RTAX_MTU-1]= rt->u.dst.dev->mtu;
|
|
|
|
|
2008-05-04 23:14:42 -06:00
|
|
|
if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.metrics[RTAX_HOPLIMIT-1] = sysctl_ip_default_ttl;
|
2008-08-06 03:33:49 -06:00
|
|
|
if (dst_mtu(&rt->u.dst) > IP_MAX_MTU)
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.metrics[RTAX_MTU-1] = IP_MAX_MTU;
|
2008-05-04 23:14:42 -06:00
|
|
|
if (dst_metric(&rt->u.dst, RTAX_ADVMSS) == 0)
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.metrics[RTAX_ADVMSS-1] = max_t(unsigned int, rt->u.dst.dev->mtu - 40,
|
|
|
|
ip_rt_min_advmss);
|
2008-05-04 23:14:42 -06:00
|
|
|
if (dst_metric(&rt->u.dst, RTAX_ADVMSS) > 65535 - 40)
|
2005-04-16 16:20:36 -06:00
|
|
|
rt->u.dst.metrics[RTAX_ADVMSS-1] = 65535 - 40;
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
#ifdef CONFIG_IP_MULTIPLE_TABLES
|
|
|
|
set_class_tag(rt, fib_rules_tclass(res));
|
|
|
|
#endif
|
|
|
|
set_class_tag(rt, itag);
|
|
|
|
#endif
|
2007-02-09 07:24:47 -07:00
|
|
|
rt->rt_type = res->type;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-09-26 22:25:20 -06:00
|
|
|
static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
|
2005-04-16 16:20:36 -06:00
|
|
|
u8 tos, struct net_device *dev, int our)
|
|
|
|
{
|
|
|
|
unsigned hash;
|
|
|
|
struct rtable *rth;
|
2006-09-26 22:27:54 -06:00
|
|
|
__be32 spec_dst;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct in_device *in_dev = in_dev_get(dev);
|
|
|
|
u32 itag = 0;
|
|
|
|
|
|
|
|
/* Primary sanity checks. */
|
|
|
|
|
|
|
|
if (in_dev == NULL)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2008-01-21 04:18:08 -07:00
|
|
|
if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
|
2007-12-16 14:45:43 -07:00
|
|
|
ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto e_inval;
|
|
|
|
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_zeronet(saddr)) {
|
|
|
|
if (!ipv4_is_local_multicast(daddr))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto e_inval;
|
|
|
|
spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK);
|
|
|
|
} else if (fib_validate_source(saddr, 0, tos, 0,
|
2009-10-17 20:12:33 -06:00
|
|
|
dev, &spec_dst, &itag, 0) < 0)
|
2005-04-16 16:20:36 -06:00
|
|
|
goto e_inval;
|
|
|
|
|
|
|
|
rth = dst_alloc(&ipv4_dst_ops);
|
|
|
|
if (!rth)
|
|
|
|
goto e_nobufs;
|
|
|
|
|
|
|
|
rth->u.dst.output= ip_rt_bug;
|
|
|
|
|
|
|
|
atomic_set(&rth->u.dst.__refcnt, 1);
|
|
|
|
rth->u.dst.flags= DST_HOST;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(in_dev, NOPOLICY))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOPOLICY;
|
|
|
|
rth->fl.fl4_dst = daddr;
|
|
|
|
rth->rt_dst = daddr;
|
|
|
|
rth->fl.fl4_tos = tos;
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark = skb->mark;
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->fl.fl4_src = saddr;
|
|
|
|
rth->rt_src = saddr;
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
rth->u.dst.tclassid = itag;
|
|
|
|
#endif
|
|
|
|
rth->rt_iif =
|
|
|
|
rth->fl.iif = dev->ifindex;
|
2007-09-26 23:10:56 -06:00
|
|
|
rth->u.dst.dev = init_net.loopback_dev;
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_hold(rth->u.dst.dev);
|
|
|
|
rth->idev = in_dev_get(rth->u.dst.dev);
|
|
|
|
rth->fl.oif = 0;
|
|
|
|
rth->rt_gateway = daddr;
|
|
|
|
rth->rt_spec_dst= spec_dst;
|
2008-07-05 20:04:32 -06:00
|
|
|
rth->rt_genid = rt_genid(dev_net(dev));
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->rt_flags = RTCF_MULTICAST;
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
rth->rt_type = RTN_MULTICAST;
|
2005-04-16 16:20:36 -06:00
|
|
|
if (our) {
|
|
|
|
rth->u.dst.input= ip_local_deliver;
|
|
|
|
rth->rt_flags |= RTCF_LOCAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_IP_MROUTE
|
2007-12-16 14:45:43 -07:00
|
|
|
if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.input = ip_mr_input;
|
|
|
|
#endif
|
|
|
|
RT_CACHE_STAT_INC(in_slow_mc);
|
|
|
|
|
|
|
|
in_dev_put(in_dev);
|
2008-07-05 20:04:32 -06:00
|
|
|
hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev)));
|
2009-06-01 23:14:27 -06:00
|
|
|
return rt_intern_hash(hash, rth, NULL, skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
e_nobufs:
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
return -ENOBUFS;
|
|
|
|
|
|
|
|
e_inval:
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void ip_handle_martian_source(struct net_device *dev,
|
|
|
|
struct in_device *in_dev,
|
|
|
|
struct sk_buff *skb,
|
2006-09-26 22:25:20 -06:00
|
|
|
__be32 daddr,
|
|
|
|
__be32 saddr)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
RT_CACHE_STAT_INC(in_martian_src);
|
|
|
|
#ifdef CONFIG_IP_ROUTE_VERBOSE
|
|
|
|
if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
|
|
|
|
/*
|
|
|
|
* RFC1812 recommendation, if source is martian,
|
|
|
|
* the only hint is MAC header.
|
|
|
|
*/
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n",
|
|
|
|
&daddr, &saddr, dev->name);
|
2007-03-19 16:33:04 -06:00
|
|
|
if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
int i;
|
2007-03-19 16:33:04 -06:00
|
|
|
const unsigned char *p = skb_mac_header(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
printk(KERN_WARNING "ll header: ");
|
|
|
|
for (i = 0; i < dev->hard_header_len; i++, p++) {
|
|
|
|
printk("%02x", *p);
|
|
|
|
if (i < (dev->hard_header_len - 1))
|
|
|
|
printk(":");
|
|
|
|
}
|
|
|
|
printk("\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static int __mkroute_input(struct sk_buff *skb,
|
|
|
|
struct fib_result *res,
|
|
|
|
struct in_device *in_dev,
|
|
|
|
__be32 daddr, __be32 saddr, u32 tos,
|
|
|
|
struct rtable **result)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
|
|
|
|
struct rtable *rth;
|
|
|
|
int err;
|
|
|
|
struct in_device *out_dev;
|
|
|
|
unsigned flags = 0;
|
2006-09-26 22:28:14 -06:00
|
|
|
__be32 spec_dst;
|
|
|
|
u32 itag;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* get a working reference to the output device */
|
|
|
|
out_dev = in_dev_get(FIB_RES_DEV(*res));
|
|
|
|
if (out_dev == NULL) {
|
|
|
|
if (net_ratelimit())
|
|
|
|
printk(KERN_CRIT "Bug in ip_route_input" \
|
|
|
|
"_slow(). Please, report\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-02-09 07:24:47 -07:00
|
|
|
err = fib_validate_source(saddr, daddr, tos, FIB_RES_OIF(*res),
|
2009-10-17 20:12:33 -06:00
|
|
|
in_dev->dev, &spec_dst, &itag, skb->mark);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (err < 0) {
|
2007-02-09 07:24:47 -07:00
|
|
|
ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
|
2005-04-16 16:20:36 -06:00
|
|
|
saddr);
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
err = -EINVAL;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
flags |= RTCF_DIRECTSRC;
|
|
|
|
|
2008-06-03 17:36:01 -06:00
|
|
|
if (out_dev == in_dev && err &&
|
2005-04-16 16:20:36 -06:00
|
|
|
(IN_DEV_SHARED_MEDIA(out_dev) ||
|
|
|
|
inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res))))
|
|
|
|
flags |= RTCF_DOREDIRECT;
|
|
|
|
|
|
|
|
if (skb->protocol != htons(ETH_P_IP)) {
|
|
|
|
/* Not IP (i.e. ARP). Do not create route, if it is
|
|
|
|
* invalid for proxy arp. DNAT routes are always valid.
|
|
|
|
*/
|
2008-01-09 01:18:24 -07:00
|
|
|
if (out_dev == in_dev) {
|
2005-04-16 16:20:36 -06:00
|
|
|
err = -EINVAL;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
rth = dst_alloc(&ipv4_dst_ops);
|
|
|
|
if (!rth) {
|
|
|
|
err = -ENOBUFS;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2005-09-08 14:34:47 -06:00
|
|
|
atomic_set(&rth->u.dst.__refcnt, 1);
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags= DST_HOST;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(in_dev, NOPOLICY))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOPOLICY;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(out_dev, NOXFRM))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOXFRM;
|
|
|
|
rth->fl.fl4_dst = daddr;
|
|
|
|
rth->rt_dst = daddr;
|
|
|
|
rth->fl.fl4_tos = tos;
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark = skb->mark;
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->fl.fl4_src = saddr;
|
|
|
|
rth->rt_src = saddr;
|
|
|
|
rth->rt_gateway = daddr;
|
|
|
|
rth->rt_iif =
|
|
|
|
rth->fl.iif = in_dev->dev->ifindex;
|
|
|
|
rth->u.dst.dev = (out_dev)->dev;
|
|
|
|
dev_hold(rth->u.dst.dev);
|
|
|
|
rth->idev = in_dev_get(rth->u.dst.dev);
|
|
|
|
rth->fl.oif = 0;
|
|
|
|
rth->rt_spec_dst= spec_dst;
|
|
|
|
|
|
|
|
rth->u.dst.input = ip_forward;
|
|
|
|
rth->u.dst.output = ip_output;
|
2008-07-05 20:04:32 -06:00
|
|
|
rth->rt_genid = rt_genid(dev_net(rth->u.dst.dev));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rt_set_nexthop(rth, res, itag);
|
|
|
|
|
|
|
|
rth->rt_flags = flags;
|
|
|
|
|
|
|
|
*result = rth;
|
|
|
|
err = 0;
|
|
|
|
cleanup:
|
|
|
|
/* release the working reference to the output device */
|
|
|
|
in_dev_put(out_dev);
|
|
|
|
return err;
|
2007-02-09 07:24:47 -07:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static int ip_mkroute_input(struct sk_buff *skb,
|
|
|
|
struct fib_result *res,
|
|
|
|
const struct flowi *fl,
|
|
|
|
struct in_device *in_dev,
|
|
|
|
__be32 daddr, __be32 saddr, u32 tos)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2005-06-22 23:10:23 -06:00
|
|
|
struct rtable* rth = NULL;
|
2005-04-16 16:20:36 -06:00
|
|
|
int err;
|
|
|
|
unsigned hash;
|
|
|
|
|
|
|
|
#ifdef CONFIG_IP_ROUTE_MULTIPATH
|
|
|
|
if (res->fi && res->fi->fib_nhs > 1 && fl->oif == 0)
|
|
|
|
fib_select_multipath(fl, res);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/* create a routing cache entry */
|
|
|
|
err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
/* put it into the cache */
|
2008-07-05 20:04:32 -06:00
|
|
|
hash = rt_hash(daddr, saddr, fl->iif,
|
|
|
|
rt_genid(dev_net(rth->u.dst.dev)));
|
2009-06-01 23:14:27 -06:00
|
|
|
return rt_intern_hash(hash, rth, NULL, skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* NOTE. We drop all the packets that has local source
|
|
|
|
* addresses, because every properly looped back packet
|
|
|
|
* must have correct destination already attached by output routine.
|
|
|
|
*
|
|
|
|
* Such approach solves two big problems:
|
|
|
|
* 1. Not simplex devices are handled properly.
|
|
|
|
* 2. IP spoofing attempts are filtered with 100% of guarantee.
|
|
|
|
*/
|
|
|
|
|
2006-09-26 22:25:20 -06:00
|
|
|
static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
|
2005-04-16 16:20:36 -06:00
|
|
|
u8 tos, struct net_device *dev)
|
|
|
|
{
|
|
|
|
struct fib_result res;
|
|
|
|
struct in_device *in_dev = in_dev_get(dev);
|
|
|
|
struct flowi fl = { .nl_u = { .ip4_u =
|
|
|
|
{ .daddr = daddr,
|
|
|
|
.saddr = saddr,
|
|
|
|
.tos = tos,
|
|
|
|
.scope = RT_SCOPE_UNIVERSE,
|
|
|
|
} },
|
2006-11-09 16:20:38 -07:00
|
|
|
.mark = skb->mark,
|
2005-04-16 16:20:36 -06:00
|
|
|
.iif = dev->ifindex };
|
|
|
|
unsigned flags = 0;
|
|
|
|
u32 itag = 0;
|
|
|
|
struct rtable * rth;
|
|
|
|
unsigned hash;
|
2006-09-26 22:25:20 -06:00
|
|
|
__be32 spec_dst;
|
2005-04-16 16:20:36 -06:00
|
|
|
int err = -EINVAL;
|
|
|
|
int free_res = 0;
|
2008-03-25 06:47:49 -06:00
|
|
|
struct net * net = dev_net(dev);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* IP on this device is disabled. */
|
|
|
|
|
|
|
|
if (!in_dev)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* Check for the most weird martians, which can be not detected
|
|
|
|
by fib_lookup.
|
|
|
|
*/
|
|
|
|
|
2008-01-21 04:18:08 -07:00
|
|
|
if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
|
2007-12-16 14:45:43 -07:00
|
|
|
ipv4_is_loopback(saddr))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto martian_source;
|
|
|
|
|
2006-09-26 23:15:01 -06:00
|
|
|
if (daddr == htonl(0xFFFFFFFF) || (saddr == 0 && daddr == 0))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto brd_input;
|
|
|
|
|
|
|
|
/* Accept zero addresses only to limited broadcast;
|
|
|
|
* I even do not know to fix it or not. Waiting for complains :-)
|
|
|
|
*/
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_zeronet(saddr))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto martian_source;
|
|
|
|
|
2008-01-21 04:18:08 -07:00
|
|
|
if (ipv4_is_lbcast(daddr) || ipv4_is_zeronet(daddr) ||
|
2007-12-16 14:45:43 -07:00
|
|
|
ipv4_is_loopback(daddr))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto martian_destination;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now we are ready to route packet.
|
|
|
|
*/
|
2008-01-21 18:34:35 -07:00
|
|
|
if ((err = fib_lookup(net, &fl, &res)) != 0) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!IN_DEV_FORWARD(in_dev))
|
2005-06-28 14:06:23 -06:00
|
|
|
goto e_hostunreach;
|
2005-04-16 16:20:36 -06:00
|
|
|
goto no_route;
|
|
|
|
}
|
|
|
|
free_res = 1;
|
|
|
|
|
|
|
|
RT_CACHE_STAT_INC(in_slow_tot);
|
|
|
|
|
|
|
|
if (res.type == RTN_BROADCAST)
|
|
|
|
goto brd_input;
|
|
|
|
|
|
|
|
if (res.type == RTN_LOCAL) {
|
|
|
|
int result;
|
|
|
|
result = fib_validate_source(saddr, daddr, tos,
|
2008-01-21 18:34:35 -07:00
|
|
|
net->loopback_dev->ifindex,
|
2009-10-17 20:12:33 -06:00
|
|
|
dev, &spec_dst, &itag, skb->mark);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (result < 0)
|
|
|
|
goto martian_source;
|
|
|
|
if (result)
|
|
|
|
flags |= RTCF_DIRECTSRC;
|
|
|
|
spec_dst = daddr;
|
|
|
|
goto local_input;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!IN_DEV_FORWARD(in_dev))
|
2005-06-28 14:06:23 -06:00
|
|
|
goto e_hostunreach;
|
2005-04-16 16:20:36 -06:00
|
|
|
if (res.type != RTN_UNICAST)
|
|
|
|
goto martian_destination;
|
|
|
|
|
|
|
|
err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos);
|
|
|
|
done:
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
if (free_res)
|
|
|
|
fib_res_put(&res);
|
|
|
|
out: return err;
|
|
|
|
|
|
|
|
brd_input:
|
|
|
|
if (skb->protocol != htons(ETH_P_IP))
|
|
|
|
goto e_inval;
|
|
|
|
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_zeronet(saddr))
|
2005-04-16 16:20:36 -06:00
|
|
|
spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK);
|
|
|
|
else {
|
|
|
|
err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst,
|
2009-10-17 20:12:33 -06:00
|
|
|
&itag, skb->mark);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (err < 0)
|
|
|
|
goto martian_source;
|
|
|
|
if (err)
|
|
|
|
flags |= RTCF_DIRECTSRC;
|
|
|
|
}
|
|
|
|
flags |= RTCF_BROADCAST;
|
|
|
|
res.type = RTN_BROADCAST;
|
|
|
|
RT_CACHE_STAT_INC(in_brd);
|
|
|
|
|
|
|
|
local_input:
|
|
|
|
rth = dst_alloc(&ipv4_dst_ops);
|
|
|
|
if (!rth)
|
|
|
|
goto e_nobufs;
|
|
|
|
|
|
|
|
rth->u.dst.output= ip_rt_bug;
|
2008-07-05 20:04:32 -06:00
|
|
|
rth->rt_genid = rt_genid(net);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
atomic_set(&rth->u.dst.__refcnt, 1);
|
|
|
|
rth->u.dst.flags= DST_HOST;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(in_dev, NOPOLICY))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOPOLICY;
|
|
|
|
rth->fl.fl4_dst = daddr;
|
|
|
|
rth->rt_dst = daddr;
|
|
|
|
rth->fl.fl4_tos = tos;
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark = skb->mark;
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->fl.fl4_src = saddr;
|
|
|
|
rth->rt_src = saddr;
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
rth->u.dst.tclassid = itag;
|
|
|
|
#endif
|
|
|
|
rth->rt_iif =
|
|
|
|
rth->fl.iif = dev->ifindex;
|
2008-01-21 18:34:35 -07:00
|
|
|
rth->u.dst.dev = net->loopback_dev;
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_hold(rth->u.dst.dev);
|
|
|
|
rth->idev = in_dev_get(rth->u.dst.dev);
|
|
|
|
rth->rt_gateway = daddr;
|
|
|
|
rth->rt_spec_dst= spec_dst;
|
|
|
|
rth->u.dst.input= ip_local_deliver;
|
|
|
|
rth->rt_flags = flags|RTCF_LOCAL;
|
|
|
|
if (res.type == RTN_UNREACHABLE) {
|
|
|
|
rth->u.dst.input= ip_error;
|
|
|
|
rth->u.dst.error= -err;
|
|
|
|
rth->rt_flags &= ~RTCF_LOCAL;
|
|
|
|
}
|
|
|
|
rth->rt_type = res.type;
|
2008-07-05 20:04:32 -06:00
|
|
|
hash = rt_hash(daddr, saddr, fl.iif, rt_genid(net));
|
2009-06-01 23:14:27 -06:00
|
|
|
err = rt_intern_hash(hash, rth, NULL, skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
goto done;
|
|
|
|
|
|
|
|
no_route:
|
|
|
|
RT_CACHE_STAT_INC(in_no_route);
|
|
|
|
spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
|
|
|
|
res.type = RTN_UNREACHABLE;
|
2007-12-07 02:07:24 -07:00
|
|
|
if (err == -ESRCH)
|
|
|
|
err = -ENETUNREACH;
|
2005-04-16 16:20:36 -06:00
|
|
|
goto local_input;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do not cache martian addresses: they should be logged (RFC1812)
|
|
|
|
*/
|
|
|
|
martian_destination:
|
|
|
|
RT_CACHE_STAT_INC(in_martian_dst);
|
|
|
|
#ifdef CONFIG_IP_ROUTE_VERBOSE
|
|
|
|
if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
|
2008-10-31 01:53:57 -06:00
|
|
|
printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n",
|
|
|
|
&daddr, &saddr, dev->name);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
2005-06-28 14:06:23 -06:00
|
|
|
|
|
|
|
e_hostunreach:
|
2007-02-09 07:24:47 -07:00
|
|
|
err = -EHOSTUNREACH;
|
|
|
|
goto done;
|
2005-06-28 14:06:23 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
e_inval:
|
|
|
|
err = -EINVAL;
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
e_nobufs:
|
|
|
|
err = -ENOBUFS;
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
martian_source:
|
|
|
|
ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
|
|
|
|
goto e_inval;
|
|
|
|
}
|
|
|
|
|
2006-09-26 22:25:20 -06:00
|
|
|
int ip_route_input(struct sk_buff *skb, __be32 daddr, __be32 saddr,
|
2005-04-16 16:20:36 -06:00
|
|
|
u8 tos, struct net_device *dev)
|
|
|
|
{
|
|
|
|
struct rtable * rth;
|
|
|
|
unsigned hash;
|
|
|
|
int iif = dev->ifindex;
|
2008-01-23 00:50:25 -07:00
|
|
|
struct net *net;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-03-25 06:47:49 -06:00
|
|
|
net = dev_net(dev);
|
2008-10-27 13:28:25 -06:00
|
|
|
|
|
|
|
if (!rt_caching(net))
|
|
|
|
goto skip_cache;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
tos &= IPTOS_RT_MASK;
|
2008-07-05 20:04:32 -06:00
|
|
|
hash = rt_hash(daddr, saddr, iif, rt_genid(net));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
for (rth = rcu_dereference(rt_hash_table[hash].chain); rth;
|
2007-02-09 17:19:26 -07:00
|
|
|
rth = rcu_dereference(rth->u.dst.rt_next)) {
|
2008-04-10 05:00:28 -06:00
|
|
|
if (((rth->fl.fl4_dst ^ daddr) |
|
|
|
|
(rth->fl.fl4_src ^ saddr) |
|
|
|
|
(rth->fl.iif ^ iif) |
|
|
|
|
rth->fl.oif |
|
|
|
|
(rth->fl.fl4_tos ^ tos)) == 0 &&
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark == skb->mark &&
|
2008-03-25 12:57:35 -06:00
|
|
|
net_eq(dev_net(rth->u.dst.dev), net) &&
|
2008-07-05 20:04:32 -06:00
|
|
|
!rt_is_expired(rth)) {
|
2007-11-10 22:28:34 -07:00
|
|
|
dst_use(&rth->u.dst, jiffies);
|
2005-04-16 16:20:36 -06:00
|
|
|
RT_CACHE_STAT_INC(in_hit);
|
|
|
|
rcu_read_unlock();
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_set(skb, &rth->u.dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
RT_CACHE_STAT_INC(in_hlist_search);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
skip_cache:
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Multicast recognition logic is moved from route cache to here.
|
|
|
|
The problem was that too many Ethernet cards have broken/missing
|
|
|
|
hardware multicast filters :-( As result the host on multicasting
|
|
|
|
network acquires a lot of useless route cache entries, sort of
|
|
|
|
SDR messages from all the world. Now we try to get rid of them.
|
|
|
|
Really, provided software IP multicast filter is organized
|
|
|
|
reasonably (at least, hashed), it does not result in a slowdown
|
|
|
|
comparing with route cache reject entries.
|
|
|
|
Note, that multicast routers are not affected, because
|
|
|
|
route cache entry is created eventually.
|
|
|
|
*/
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_multicast(daddr)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
struct in_device *in_dev;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
2005-10-03 15:35:55 -06:00
|
|
|
if ((in_dev = __in_dev_get_rcu(dev)) != NULL) {
|
2005-04-16 16:20:36 -06:00
|
|
|
int our = ip_check_mc(in_dev, daddr, saddr,
|
2007-04-20 23:47:35 -06:00
|
|
|
ip_hdr(skb)->protocol);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (our
|
|
|
|
#ifdef CONFIG_IP_MROUTE
|
2009-11-23 11:41:23 -07:00
|
|
|
||
|
|
|
|
(!ipv4_is_local_multicast(daddr) &&
|
|
|
|
IN_DEV_MFORWARD(in_dev))
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
2009-11-23 11:41:23 -07:00
|
|
|
) {
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_unlock();
|
|
|
|
return ip_route_input_mc(skb, daddr, saddr,
|
|
|
|
tos, dev, our);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
return ip_route_input_slow(skb, daddr, saddr, tos, dev);
|
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static int __mkroute_output(struct rtable **result,
|
|
|
|
struct fib_result *res,
|
|
|
|
const struct flowi *fl,
|
|
|
|
const struct flowi *oldflp,
|
|
|
|
struct net_device *dev_out,
|
|
|
|
unsigned flags)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
struct rtable *rth;
|
|
|
|
struct in_device *in_dev;
|
|
|
|
u32 tos = RT_FL_TOS(oldflp);
|
|
|
|
int err = 0;
|
|
|
|
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_loopback(fl->fl4_src) && !(dev_out->flags&IFF_LOOPBACK))
|
2005-04-16 16:20:36 -06:00
|
|
|
return -EINVAL;
|
|
|
|
|
2006-09-26 23:15:01 -06:00
|
|
|
if (fl->fl4_dst == htonl(0xFFFFFFFF))
|
2005-04-16 16:20:36 -06:00
|
|
|
res->type = RTN_BROADCAST;
|
2007-12-16 14:45:43 -07:00
|
|
|
else if (ipv4_is_multicast(fl->fl4_dst))
|
2005-04-16 16:20:36 -06:00
|
|
|
res->type = RTN_MULTICAST;
|
2008-01-21 04:18:08 -07:00
|
|
|
else if (ipv4_is_lbcast(fl->fl4_dst) || ipv4_is_zeronet(fl->fl4_dst))
|
2005-04-16 16:20:36 -06:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (dev_out->flags & IFF_LOOPBACK)
|
|
|
|
flags |= RTCF_LOCAL;
|
|
|
|
|
|
|
|
/* get work reference to inet device */
|
|
|
|
in_dev = in_dev_get(dev_out);
|
|
|
|
if (!in_dev)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (res->type == RTN_BROADCAST) {
|
|
|
|
flags |= RTCF_BROADCAST | RTCF_LOCAL;
|
|
|
|
if (res->fi) {
|
|
|
|
fib_info_put(res->fi);
|
|
|
|
res->fi = NULL;
|
|
|
|
}
|
|
|
|
} else if (res->type == RTN_MULTICAST) {
|
|
|
|
flags |= RTCF_MULTICAST|RTCF_LOCAL;
|
2007-02-09 07:24:47 -07:00
|
|
|
if (!ip_check_mc(in_dev, oldflp->fl4_dst, oldflp->fl4_src,
|
2005-04-16 16:20:36 -06:00
|
|
|
oldflp->proto))
|
|
|
|
flags &= ~RTCF_LOCAL;
|
|
|
|
/* If multicast route do not exist use
|
|
|
|
default one, but do not gateway in this case.
|
|
|
|
Yes, it is hack.
|
|
|
|
*/
|
|
|
|
if (res->fi && res->prefixlen < 4) {
|
|
|
|
fib_info_put(res->fi);
|
|
|
|
res->fi = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
rth = dst_alloc(&ipv4_dst_ops);
|
|
|
|
if (!rth) {
|
|
|
|
err = -ENOBUFS;
|
|
|
|
goto cleanup;
|
2007-02-09 07:24:47 -07:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-09-08 14:34:47 -06:00
|
|
|
atomic_set(&rth->u.dst.__refcnt, 1);
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags= DST_HOST;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(in_dev, NOXFRM))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOXFRM;
|
2007-06-05 00:34:44 -06:00
|
|
|
if (IN_DEV_CONF_GET(in_dev, NOPOLICY))
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.flags |= DST_NOPOLICY;
|
|
|
|
|
|
|
|
rth->fl.fl4_dst = oldflp->fl4_dst;
|
|
|
|
rth->fl.fl4_tos = tos;
|
|
|
|
rth->fl.fl4_src = oldflp->fl4_src;
|
|
|
|
rth->fl.oif = oldflp->oif;
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark = oldflp->mark;
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->rt_dst = fl->fl4_dst;
|
|
|
|
rth->rt_src = fl->fl4_src;
|
|
|
|
rth->rt_iif = oldflp->oif ? : dev_out->ifindex;
|
2007-02-09 07:24:47 -07:00
|
|
|
/* get references to the devices that are to be hold by the routing
|
2005-04-16 16:20:36 -06:00
|
|
|
cache entry */
|
|
|
|
rth->u.dst.dev = dev_out;
|
|
|
|
dev_hold(dev_out);
|
|
|
|
rth->idev = in_dev_get(dev_out);
|
|
|
|
rth->rt_gateway = fl->fl4_dst;
|
|
|
|
rth->rt_spec_dst= fl->fl4_src;
|
|
|
|
|
|
|
|
rth->u.dst.output=ip_output;
|
2008-07-05 20:04:32 -06:00
|
|
|
rth->rt_genid = rt_genid(dev_net(dev_out));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
RT_CACHE_STAT_INC(out_slow_tot);
|
|
|
|
|
|
|
|
if (flags & RTCF_LOCAL) {
|
|
|
|
rth->u.dst.input = ip_local_deliver;
|
|
|
|
rth->rt_spec_dst = fl->fl4_dst;
|
|
|
|
}
|
|
|
|
if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
|
|
|
|
rth->rt_spec_dst = fl->fl4_src;
|
2007-02-09 07:24:47 -07:00
|
|
|
if (flags & RTCF_LOCAL &&
|
2005-04-16 16:20:36 -06:00
|
|
|
!(dev_out->flags & IFF_LOOPBACK)) {
|
|
|
|
rth->u.dst.output = ip_mc_output;
|
|
|
|
RT_CACHE_STAT_INC(out_slow_mc);
|
|
|
|
}
|
|
|
|
#ifdef CONFIG_IP_MROUTE
|
|
|
|
if (res->type == RTN_MULTICAST) {
|
|
|
|
if (IN_DEV_MFORWARD(in_dev) &&
|
2007-12-16 14:45:43 -07:00
|
|
|
!ipv4_is_local_multicast(oldflp->fl4_dst)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
rth->u.dst.input = ip_mr_input;
|
|
|
|
rth->u.dst.output = ip_mc_output;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
rt_set_nexthop(rth, res, 0);
|
|
|
|
|
|
|
|
rth->rt_flags = flags;
|
|
|
|
|
|
|
|
*result = rth;
|
|
|
|
cleanup:
|
|
|
|
/* release work reference to inet device */
|
|
|
|
in_dev_put(in_dev);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2008-04-10 02:52:09 -06:00
|
|
|
static int ip_mkroute_output(struct rtable **rp,
|
|
|
|
struct fib_result *res,
|
|
|
|
const struct flowi *fl,
|
|
|
|
const struct flowi *oldflp,
|
|
|
|
struct net_device *dev_out,
|
|
|
|
unsigned flags)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2005-06-22 23:10:23 -06:00
|
|
|
struct rtable *rth = NULL;
|
2005-04-16 16:20:36 -06:00
|
|
|
int err = __mkroute_output(&rth, res, fl, oldflp, dev_out, flags);
|
|
|
|
unsigned hash;
|
|
|
|
if (err == 0) {
|
2008-07-05 20:04:09 -06:00
|
|
|
hash = rt_hash(oldflp->fl4_dst, oldflp->fl4_src, oldflp->oif,
|
2008-07-05 20:04:32 -06:00
|
|
|
rt_genid(dev_net(dev_out)));
|
2009-06-01 23:14:27 -06:00
|
|
|
err = rt_intern_hash(hash, rth, rp, NULL);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Major route resolver routine.
|
|
|
|
*/
|
|
|
|
|
2008-01-22 23:06:19 -07:00
|
|
|
static int ip_route_output_slow(struct net *net, struct rtable **rp,
|
|
|
|
const struct flowi *oldflp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
u32 tos = RT_FL_TOS(oldflp);
|
|
|
|
struct flowi fl = { .nl_u = { .ip4_u =
|
|
|
|
{ .daddr = oldflp->fl4_dst,
|
|
|
|
.saddr = oldflp->fl4_src,
|
|
|
|
.tos = tos & IPTOS_RT_MASK,
|
|
|
|
.scope = ((tos & RTO_ONLINK) ?
|
|
|
|
RT_SCOPE_LINK :
|
|
|
|
RT_SCOPE_UNIVERSE),
|
|
|
|
} },
|
2006-11-09 16:20:38 -07:00
|
|
|
.mark = oldflp->mark,
|
2008-01-22 23:06:19 -07:00
|
|
|
.iif = net->loopback_dev->ifindex,
|
2005-04-16 16:20:36 -06:00
|
|
|
.oif = oldflp->oif };
|
|
|
|
struct fib_result res;
|
|
|
|
unsigned flags = 0;
|
|
|
|
struct net_device *dev_out = NULL;
|
|
|
|
int free_res = 0;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
|
|
res.fi = NULL;
|
|
|
|
#ifdef CONFIG_IP_MULTIPLE_TABLES
|
|
|
|
res.r = NULL;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
if (oldflp->fl4_src) {
|
|
|
|
err = -EINVAL;
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_multicast(oldflp->fl4_src) ||
|
2008-01-21 04:18:08 -07:00
|
|
|
ipv4_is_lbcast(oldflp->fl4_src) ||
|
2007-12-16 14:45:43 -07:00
|
|
|
ipv4_is_zeronet(oldflp->fl4_src))
|
2005-04-16 16:20:36 -06:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* I removed check for oif == dev_out->oif here.
|
|
|
|
It was wrong for two reasons:
|
2008-01-22 23:04:30 -07:00
|
|
|
1. ip_dev_find(net, saddr) can return wrong iface, if saddr
|
|
|
|
is assigned to multiple interfaces.
|
2005-04-16 16:20:36 -06:00
|
|
|
2. Moreover, we are allowed to send packets with saddr
|
|
|
|
of another iface. --ANK
|
|
|
|
*/
|
|
|
|
|
2009-11-23 11:41:23 -07:00
|
|
|
if (oldflp->oif == 0 &&
|
|
|
|
(ipv4_is_multicast(oldflp->fl4_dst) ||
|
|
|
|
oldflp->fl4_dst == htonl(0xFFFFFFFF))) {
|
2008-10-01 08:28:28 -06:00
|
|
|
/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
|
|
|
|
dev_out = ip_dev_find(net, oldflp->fl4_src);
|
|
|
|
if (dev_out == NULL)
|
|
|
|
goto out;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Special hack: user can direct multicasts
|
|
|
|
and limited broadcast via necessary interface
|
|
|
|
without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
|
|
|
|
This hack is not just for fun, it allows
|
|
|
|
vic,vat and friends to work.
|
|
|
|
They bind socket to loopback, set ttl to zero
|
|
|
|
and expect that it will work.
|
|
|
|
From the viewpoint of routing cache they are broken,
|
|
|
|
because we are not allowed to build multicast path
|
|
|
|
with loopback source addr (look, routing cache
|
|
|
|
cannot know, that ttl is zero, so that packet
|
|
|
|
will not leave this host and route is valid).
|
|
|
|
Luckily, this hack is good workaround.
|
|
|
|
*/
|
|
|
|
|
|
|
|
fl.oif = dev_out->ifindex;
|
|
|
|
goto make_route;
|
|
|
|
}
|
2008-10-01 08:28:28 -06:00
|
|
|
|
|
|
|
if (!(oldflp->flags & FLOWI_FLAG_ANYSRC)) {
|
|
|
|
/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
|
|
|
|
dev_out = ip_dev_find(net, oldflp->fl4_src);
|
|
|
|
if (dev_out == NULL)
|
|
|
|
goto out;
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_put(dev_out);
|
2008-10-01 08:28:28 -06:00
|
|
|
dev_out = NULL;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
if (oldflp->oif) {
|
2008-01-22 23:06:19 -07:00
|
|
|
dev_out = dev_get_by_index(net, oldflp->oif);
|
2005-04-16 16:20:36 -06:00
|
|
|
err = -ENODEV;
|
|
|
|
if (dev_out == NULL)
|
|
|
|
goto out;
|
2005-10-03 15:35:55 -06:00
|
|
|
|
|
|
|
/* RACE: Check return value of inet_select_addr instead. */
|
|
|
|
if (__in_dev_get_rtnl(dev_out) == NULL) {
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_put(dev_out);
|
|
|
|
goto out; /* Wrong error code */
|
|
|
|
}
|
|
|
|
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_local_multicast(oldflp->fl4_dst) ||
|
|
|
|
oldflp->fl4_dst == htonl(0xFFFFFFFF)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!fl.fl4_src)
|
|
|
|
fl.fl4_src = inet_select_addr(dev_out, 0,
|
|
|
|
RT_SCOPE_LINK);
|
|
|
|
goto make_route;
|
|
|
|
}
|
|
|
|
if (!fl.fl4_src) {
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_multicast(oldflp->fl4_dst))
|
2005-04-16 16:20:36 -06:00
|
|
|
fl.fl4_src = inet_select_addr(dev_out, 0,
|
|
|
|
fl.fl4_scope);
|
|
|
|
else if (!oldflp->fl4_dst)
|
|
|
|
fl.fl4_src = inet_select_addr(dev_out, 0,
|
|
|
|
RT_SCOPE_HOST);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!fl.fl4_dst) {
|
|
|
|
fl.fl4_dst = fl.fl4_src;
|
|
|
|
if (!fl.fl4_dst)
|
|
|
|
fl.fl4_dst = fl.fl4_src = htonl(INADDR_LOOPBACK);
|
|
|
|
if (dev_out)
|
|
|
|
dev_put(dev_out);
|
2008-01-22 23:06:19 -07:00
|
|
|
dev_out = net->loopback_dev;
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_hold(dev_out);
|
2008-01-22 23:06:19 -07:00
|
|
|
fl.oif = net->loopback_dev->ifindex;
|
2005-04-16 16:20:36 -06:00
|
|
|
res.type = RTN_LOCAL;
|
|
|
|
flags |= RTCF_LOCAL;
|
|
|
|
goto make_route;
|
|
|
|
}
|
|
|
|
|
2008-01-22 23:06:19 -07:00
|
|
|
if (fib_lookup(net, &fl, &res)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
res.fi = NULL;
|
|
|
|
if (oldflp->oif) {
|
|
|
|
/* Apparently, routing tables are wrong. Assume,
|
|
|
|
that the destination is on link.
|
|
|
|
|
|
|
|
WHY? DW.
|
|
|
|
Because we are allowed to send to iface
|
|
|
|
even if it has NO routes and NO assigned
|
|
|
|
addresses. When oif is specified, routing
|
|
|
|
tables are looked up with only one purpose:
|
|
|
|
to catch if destination is gatewayed, rather than
|
|
|
|
direct. Moreover, if MSG_DONTROUTE is set,
|
|
|
|
we send packet, ignoring both routing tables
|
|
|
|
and ifaddr state. --ANK
|
|
|
|
|
|
|
|
|
|
|
|
We could make it even if oif is unknown,
|
|
|
|
likely IPv6, but we do not.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (fl.fl4_src == 0)
|
|
|
|
fl.fl4_src = inet_select_addr(dev_out, 0,
|
|
|
|
RT_SCOPE_LINK);
|
|
|
|
res.type = RTN_UNICAST;
|
|
|
|
goto make_route;
|
|
|
|
}
|
|
|
|
if (dev_out)
|
|
|
|
dev_put(dev_out);
|
|
|
|
err = -ENETUNREACH;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
free_res = 1;
|
|
|
|
|
|
|
|
if (res.type == RTN_LOCAL) {
|
|
|
|
if (!fl.fl4_src)
|
|
|
|
fl.fl4_src = fl.fl4_dst;
|
|
|
|
if (dev_out)
|
|
|
|
dev_put(dev_out);
|
2008-01-22 23:06:19 -07:00
|
|
|
dev_out = net->loopback_dev;
|
2005-04-16 16:20:36 -06:00
|
|
|
dev_hold(dev_out);
|
|
|
|
fl.oif = dev_out->ifindex;
|
|
|
|
if (res.fi)
|
|
|
|
fib_info_put(res.fi);
|
|
|
|
res.fi = NULL;
|
|
|
|
flags |= RTCF_LOCAL;
|
|
|
|
goto make_route;
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_IP_ROUTE_MULTIPATH
|
|
|
|
if (res.fi->fib_nhs > 1 && fl.oif == 0)
|
|
|
|
fib_select_multipath(&fl, &res);
|
|
|
|
else
|
|
|
|
#endif
|
|
|
|
if (!res.prefixlen && res.type == RTN_UNICAST && !fl.oif)
|
2008-01-22 23:06:19 -07:00
|
|
|
fib_select_default(net, &fl, &res);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (!fl.fl4_src)
|
|
|
|
fl.fl4_src = FIB_RES_PREFSRC(res);
|
|
|
|
|
|
|
|
if (dev_out)
|
|
|
|
dev_put(dev_out);
|
|
|
|
dev_out = FIB_RES_DEV(res);
|
|
|
|
dev_hold(dev_out);
|
|
|
|
fl.oif = dev_out->ifindex;
|
|
|
|
|
|
|
|
|
|
|
|
make_route:
|
|
|
|
err = ip_mkroute_output(rp, &res, &fl, oldflp, dev_out, flags);
|
|
|
|
|
|
|
|
|
|
|
|
if (free_res)
|
|
|
|
fib_res_put(&res);
|
|
|
|
if (dev_out)
|
|
|
|
dev_put(dev_out);
|
|
|
|
out: return err;
|
|
|
|
}
|
|
|
|
|
2008-01-22 23:06:48 -07:00
|
|
|
int __ip_route_output_key(struct net *net, struct rtable **rp,
|
|
|
|
const struct flowi *flp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
unsigned hash;
|
|
|
|
struct rtable *rth;
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
if (!rt_caching(net))
|
|
|
|
goto slow_output;
|
|
|
|
|
2008-07-05 20:04:32 -06:00
|
|
|
hash = rt_hash(flp->fl4_dst, flp->fl4_src, flp->oif, rt_genid(net));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rcu_read_lock_bh();
|
|
|
|
for (rth = rcu_dereference(rt_hash_table[hash].chain); rth;
|
2007-02-09 17:19:26 -07:00
|
|
|
rth = rcu_dereference(rth->u.dst.rt_next)) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rth->fl.fl4_dst == flp->fl4_dst &&
|
|
|
|
rth->fl.fl4_src == flp->fl4_src &&
|
|
|
|
rth->fl.iif == 0 &&
|
|
|
|
rth->fl.oif == flp->oif &&
|
2006-11-09 16:20:38 -07:00
|
|
|
rth->fl.mark == flp->mark &&
|
2005-04-16 16:20:36 -06:00
|
|
|
!((rth->fl.fl4_tos ^ flp->fl4_tos) &
|
2008-01-23 00:50:25 -07:00
|
|
|
(IPTOS_RT_MASK | RTO_ONLINK)) &&
|
2008-03-25 12:57:35 -06:00
|
|
|
net_eq(dev_net(rth->u.dst.dev), net) &&
|
2008-07-05 20:04:32 -06:00
|
|
|
!rt_is_expired(rth)) {
|
2007-11-10 22:28:34 -07:00
|
|
|
dst_use(&rth->u.dst, jiffies);
|
2005-04-16 16:20:36 -06:00
|
|
|
RT_CACHE_STAT_INC(out_hit);
|
|
|
|
rcu_read_unlock_bh();
|
|
|
|
*rp = rth;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
RT_CACHE_STAT_INC(out_hlist_search);
|
|
|
|
}
|
|
|
|
rcu_read_unlock_bh();
|
|
|
|
|
2008-10-27 13:28:25 -06:00
|
|
|
slow_output:
|
2008-01-22 23:06:48 -07:00
|
|
|
return ip_route_output_slow(net, rp, flp);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2005-08-09 21:12:12 -06:00
|
|
|
EXPORT_SYMBOL_GPL(__ip_route_output_key);
|
|
|
|
|
2007-05-24 19:17:54 -06:00
|
|
|
static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct dst_ops ipv4_dst_blackhole_ops = {
|
|
|
|
.family = AF_INET,
|
2009-02-01 01:45:17 -07:00
|
|
|
.protocol = cpu_to_be16(ETH_P_IP),
|
2007-05-24 19:17:54 -06:00
|
|
|
.destroy = ipv4_dst_destroy,
|
|
|
|
.check = ipv4_dst_check,
|
|
|
|
.update_pmtu = ipv4_rt_blackhole_update_pmtu,
|
2008-01-30 21:07:45 -07:00
|
|
|
.entries = ATOMIC_INIT(0),
|
2007-05-24 19:17:54 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
|
2008-07-05 20:04:32 -06:00
|
|
|
static int ipv4_dst_blackhole(struct net *net, struct rtable **rp, struct flowi *flp)
|
2007-05-24 19:17:54 -06:00
|
|
|
{
|
|
|
|
struct rtable *ort = *rp;
|
|
|
|
struct rtable *rt = (struct rtable *)
|
|
|
|
dst_alloc(&ipv4_dst_blackhole_ops);
|
|
|
|
|
|
|
|
if (rt) {
|
|
|
|
struct dst_entry *new = &rt->u.dst;
|
|
|
|
|
|
|
|
atomic_set(&new->__refcnt, 1);
|
|
|
|
new->__use = 1;
|
2007-11-13 22:34:06 -07:00
|
|
|
new->input = dst_discard;
|
|
|
|
new->output = dst_discard;
|
2007-05-24 19:17:54 -06:00
|
|
|
memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
|
|
|
|
|
|
|
|
new->dev = ort->u.dst.dev;
|
|
|
|
if (new->dev)
|
|
|
|
dev_hold(new->dev);
|
|
|
|
|
|
|
|
rt->fl = ort->fl;
|
|
|
|
|
|
|
|
rt->idev = ort->idev;
|
|
|
|
if (rt->idev)
|
|
|
|
in_dev_hold(rt->idev);
|
2008-07-05 20:04:32 -06:00
|
|
|
rt->rt_genid = rt_genid(net);
|
2007-05-24 19:17:54 -06:00
|
|
|
rt->rt_flags = ort->rt_flags;
|
|
|
|
rt->rt_type = ort->rt_type;
|
|
|
|
rt->rt_dst = ort->rt_dst;
|
|
|
|
rt->rt_src = ort->rt_src;
|
|
|
|
rt->rt_iif = ort->rt_iif;
|
|
|
|
rt->rt_gateway = ort->rt_gateway;
|
|
|
|
rt->rt_spec_dst = ort->rt_spec_dst;
|
|
|
|
rt->peer = ort->peer;
|
|
|
|
if (rt->peer)
|
|
|
|
atomic_inc(&rt->peer->refcnt);
|
|
|
|
|
|
|
|
dst_free(new);
|
|
|
|
}
|
|
|
|
|
|
|
|
dst_release(&(*rp)->u.dst);
|
|
|
|
*rp = rt;
|
|
|
|
return (rt ? 0 : -ENOMEM);
|
|
|
|
}
|
|
|
|
|
2008-01-22 23:07:10 -07:00
|
|
|
int ip_route_output_flow(struct net *net, struct rtable **rp, struct flowi *flp,
|
|
|
|
struct sock *sk, int flags)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2008-01-22 23:07:10 -07:00
|
|
|
if ((err = __ip_route_output_key(net, rp, flp)) != 0)
|
2005-04-16 16:20:36 -06:00
|
|
|
return err;
|
|
|
|
|
|
|
|
if (flp->proto) {
|
|
|
|
if (!flp->fl4_src)
|
|
|
|
flp->fl4_src = (*rp)->rt_src;
|
|
|
|
if (!flp->fl4_dst)
|
|
|
|
flp->fl4_dst = (*rp)->rt_dst;
|
2008-11-25 18:35:18 -07:00
|
|
|
err = __xfrm_lookup(net, (struct dst_entry **)rp, flp, sk,
|
2007-12-12 19:48:58 -07:00
|
|
|
flags ? XFRM_LOOKUP_WAIT : 0);
|
2007-05-24 19:17:54 -06:00
|
|
|
if (err == -EREMOTE)
|
2008-07-05 20:04:32 -06:00
|
|
|
err = ipv4_dst_blackhole(net, rp, flp);
|
2007-05-24 19:17:54 -06:00
|
|
|
|
|
|
|
return err;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-08-09 21:12:12 -06:00
|
|
|
EXPORT_SYMBOL_GPL(ip_route_output_flow);
|
|
|
|
|
2008-01-22 23:07:34 -07:00
|
|
|
int ip_route_output_key(struct net *net, struct rtable **rp, struct flowi *flp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-01-22 23:07:34 -07:00
|
|
|
return ip_route_output_flow(net, rp, flp, NULL, 0);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2009-01-21 21:56:23 -07:00
|
|
|
static int rt_fill_info(struct net *net,
|
|
|
|
struct sk_buff *skb, u32 pid, u32 seq, int event,
|
2005-06-18 23:54:12 -06:00
|
|
|
int nowait, unsigned int flags)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2009-06-01 23:14:27 -06:00
|
|
|
struct rtable *rt = skb_rtable(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
struct rtmsg *r;
|
2006-08-17 19:15:17 -06:00
|
|
|
struct nlmsghdr *nlh;
|
2006-11-27 10:27:07 -07:00
|
|
|
long expires;
|
|
|
|
u32 id = 0, ts = 0, tsage = 0, error;
|
2006-08-17 19:15:17 -06:00
|
|
|
|
|
|
|
nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags);
|
|
|
|
if (nlh == NULL)
|
2007-02-01 00:16:40 -07:00
|
|
|
return -EMSGSIZE;
|
2006-08-17 19:15:17 -06:00
|
|
|
|
|
|
|
r = nlmsg_data(nlh);
|
2005-04-16 16:20:36 -06:00
|
|
|
r->rtm_family = AF_INET;
|
|
|
|
r->rtm_dst_len = 32;
|
|
|
|
r->rtm_src_len = 0;
|
|
|
|
r->rtm_tos = rt->fl.fl4_tos;
|
|
|
|
r->rtm_table = RT_TABLE_MAIN;
|
2006-08-17 19:15:17 -06:00
|
|
|
NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN);
|
2005-04-16 16:20:36 -06:00
|
|
|
r->rtm_type = rt->rt_type;
|
|
|
|
r->rtm_scope = RT_SCOPE_UNIVERSE;
|
|
|
|
r->rtm_protocol = RTPROT_UNSPEC;
|
|
|
|
r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
|
|
|
|
if (rt->rt_flags & RTCF_NOTIFY)
|
|
|
|
r->rtm_flags |= RTM_F_NOTIFY;
|
2006-08-17 19:15:17 -06:00
|
|
|
|
2006-09-26 23:15:25 -06:00
|
|
|
NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst);
|
2006-08-17 19:15:17 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt->fl.fl4_src) {
|
|
|
|
r->rtm_src_len = 32;
|
2006-09-26 23:15:25 -06:00
|
|
|
NLA_PUT_BE32(skb, RTA_SRC, rt->fl.fl4_src);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
if (rt->u.dst.dev)
|
2006-08-17 19:15:17 -06:00
|
|
|
NLA_PUT_U32(skb, RTA_OIF, rt->u.dst.dev->ifindex);
|
2005-04-16 16:20:36 -06:00
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
|
|
|
if (rt->u.dst.tclassid)
|
2006-08-17 19:15:17 -06:00
|
|
|
NLA_PUT_U32(skb, RTA_FLOW, rt->u.dst.tclassid);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
|
|
|
if (rt->fl.iif)
|
2006-09-26 23:15:25 -06:00
|
|
|
NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
else if (rt->rt_src != rt->fl.fl4_src)
|
2006-09-26 23:15:25 -06:00
|
|
|
NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src);
|
2006-08-17 19:15:17 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt->rt_dst != rt->rt_gateway)
|
2006-09-26 23:15:25 -06:00
|
|
|
NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway);
|
2006-08-17 19:15:17 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
|
2006-08-17 19:15:17 -06:00
|
|
|
goto nla_put_failure;
|
|
|
|
|
2006-11-27 10:27:07 -07:00
|
|
|
error = rt->u.dst.error;
|
|
|
|
expires = rt->u.dst.expires ? rt->u.dst.expires - jiffies : 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt->peer) {
|
2009-11-12 02:33:09 -07:00
|
|
|
id = atomic_read(&rt->peer->ip_id_count) & 0xffff;
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt->peer->tcp_ts_stamp) {
|
2006-11-27 10:27:07 -07:00
|
|
|
ts = rt->peer->tcp_ts;
|
2007-03-04 17:12:44 -07:00
|
|
|
tsage = get_seconds() - rt->peer->tcp_ts_stamp;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
2006-08-17 19:15:17 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rt->fl.iif) {
|
|
|
|
#ifdef CONFIG_IP_MROUTE
|
2006-09-26 23:15:01 -06:00
|
|
|
__be32 dst = rt->rt_dst;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-12-16 14:45:43 -07:00
|
|
|
if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) &&
|
2009-01-21 21:56:23 -07:00
|
|
|
IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
|
|
|
|
int err = ipmr_get_route(net, skb, r, nowait);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (err <= 0) {
|
|
|
|
if (!nowait) {
|
|
|
|
if (err == 0)
|
|
|
|
return 0;
|
2006-08-17 19:15:17 -06:00
|
|
|
goto nla_put_failure;
|
2005-04-16 16:20:36 -06:00
|
|
|
} else {
|
|
|
|
if (err == -EMSGSIZE)
|
2006-08-17 19:15:17 -06:00
|
|
|
goto nla_put_failure;
|
2006-11-27 10:27:07 -07:00
|
|
|
error = err;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
#endif
|
2006-08-17 19:15:17 -06:00
|
|
|
NLA_PUT_U32(skb, RTA_IIF, rt->fl.iif);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-11-27 10:27:07 -07:00
|
|
|
if (rtnl_put_cacheinfo(skb, &rt->u.dst, id, ts, tsage,
|
|
|
|
expires, error) < 0)
|
|
|
|
goto nla_put_failure;
|
2006-08-17 19:15:17 -06:00
|
|
|
|
|
|
|
return nlmsg_end(skb, nlh);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-08-17 19:15:17 -06:00
|
|
|
nla_put_failure:
|
2007-02-01 00:16:40 -07:00
|
|
|
nlmsg_cancel(skb, nlh);
|
|
|
|
return -EMSGSIZE;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-03-22 12:55:17 -06:00
|
|
|
static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-03-25 11:26:21 -06:00
|
|
|
struct net *net = sock_net(in_skb->sk);
|
2006-08-17 19:15:44 -06:00
|
|
|
struct rtmsg *rtm;
|
|
|
|
struct nlattr *tb[RTA_MAX+1];
|
2005-04-16 16:20:36 -06:00
|
|
|
struct rtable *rt = NULL;
|
2006-09-26 22:25:20 -06:00
|
|
|
__be32 dst = 0;
|
|
|
|
__be32 src = 0;
|
|
|
|
u32 iif;
|
2006-08-17 19:15:44 -06:00
|
|
|
int err;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct sk_buff *skb;
|
|
|
|
|
2006-08-17 19:15:44 -06:00
|
|
|
err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy);
|
|
|
|
if (err < 0)
|
|
|
|
goto errout;
|
|
|
|
|
|
|
|
rtm = nlmsg_data(nlh);
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
|
2006-08-17 19:15:44 -06:00
|
|
|
if (skb == NULL) {
|
|
|
|
err = -ENOBUFS;
|
|
|
|
goto errout;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* Reserve room for dummy headers, this skb can pass
|
|
|
|
through good chunk of routing engine.
|
|
|
|
*/
|
2007-03-19 16:30:44 -06:00
|
|
|
skb_reset_mac_header(skb);
|
2007-04-10 21:45:18 -06:00
|
|
|
skb_reset_network_header(skb);
|
2006-04-17 18:27:11 -06:00
|
|
|
|
|
|
|
/* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
|
2007-04-20 23:47:35 -06:00
|
|
|
ip_hdr(skb)->protocol = IPPROTO_ICMP;
|
2005-04-16 16:20:36 -06:00
|
|
|
skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr));
|
|
|
|
|
2006-09-26 23:15:25 -06:00
|
|
|
src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0;
|
|
|
|
dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0;
|
2006-08-17 19:15:44 -06:00
|
|
|
iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (iif) {
|
2006-08-17 19:15:44 -06:00
|
|
|
struct net_device *dev;
|
|
|
|
|
2008-02-28 21:52:04 -07:00
|
|
|
dev = __dev_get_by_index(net, iif);
|
2006-08-17 19:15:44 -06:00
|
|
|
if (dev == NULL) {
|
|
|
|
err = -ENODEV;
|
|
|
|
goto errout_free;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
skb->protocol = htons(ETH_P_IP);
|
|
|
|
skb->dev = dev;
|
|
|
|
local_bh_disable();
|
|
|
|
err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev);
|
|
|
|
local_bh_enable();
|
2006-08-17 19:15:44 -06:00
|
|
|
|
2009-06-01 23:14:27 -06:00
|
|
|
rt = skb_rtable(skb);
|
2006-08-17 19:15:44 -06:00
|
|
|
if (err == 0 && rt->u.dst.error)
|
2005-04-16 16:20:36 -06:00
|
|
|
err = -rt->u.dst.error;
|
|
|
|
} else {
|
2006-08-17 19:15:44 -06:00
|
|
|
struct flowi fl = {
|
|
|
|
.nl_u = {
|
|
|
|
.ip4_u = {
|
|
|
|
.daddr = dst,
|
|
|
|
.saddr = src,
|
|
|
|
.tos = rtm->rtm_tos,
|
|
|
|
},
|
|
|
|
},
|
|
|
|
.oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0,
|
|
|
|
};
|
2008-02-28 21:52:04 -07:00
|
|
|
err = ip_route_output_key(net, &rt, &fl);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2006-08-17 19:15:44 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (err)
|
2006-08-17 19:15:44 -06:00
|
|
|
goto errout_free;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_set(skb, &rt->u.dst);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rtm->rtm_flags & RTM_F_NOTIFY)
|
|
|
|
rt->rt_flags |= RTCF_NOTIFY;
|
|
|
|
|
2009-01-21 21:56:23 -07:00
|
|
|
err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
|
2008-02-28 21:52:04 -07:00
|
|
|
RTM_NEWROUTE, 0, 0);
|
2006-08-17 19:15:44 -06:00
|
|
|
if (err <= 0)
|
|
|
|
goto errout_free;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-02-28 21:52:04 -07:00
|
|
|
err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
|
2006-08-17 19:15:44 -06:00
|
|
|
errout:
|
2006-08-15 01:30:25 -06:00
|
|
|
return err;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-08-17 19:15:44 -06:00
|
|
|
errout_free:
|
2005-04-16 16:20:36 -06:00
|
|
|
kfree_skb(skb);
|
2006-08-17 19:15:44 -06:00
|
|
|
goto errout;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb)
|
|
|
|
{
|
|
|
|
struct rtable *rt;
|
|
|
|
int h, s_h;
|
|
|
|
int idx, s_idx;
|
2008-02-28 21:52:04 -07:00
|
|
|
struct net *net;
|
|
|
|
|
2008-03-25 11:26:21 -06:00
|
|
|
net = sock_net(skb->sk);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
s_h = cb->args[0];
|
[IPV4] ROUTE: ip_rt_dump() is unecessary slow
I noticed "ip route list cache x.y.z.t" can be *very* slow.
While strace-ing -T it I also noticed that first part of route cache
is fetched quite fast :
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202
GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3772 <0.000047>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\234\0\0\0\30\0\2\0\254i\
202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3736 <0.000042>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\204\0\0\0\30\0\2\0\254i\
202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3740 <0.000055>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\234\0\0\0\30\0\2\0\254i\
202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3712 <0.000043>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\204\0\0\0\30\0\2\0\254i\
202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3732 <0.000053>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202
GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3708 <0.000052>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202
GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3680 <0.000041>
while the part at the end of the table is more expensive:
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\204\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3656 <0.003857>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\204\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3772 <0.003891>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3712 <0.003765>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3700 <0.003879>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3676 <0.003797>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"p\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\2\0\2\0"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3724 <0.003856>
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\234\0\0\0\30\0\2\0\254i\202GXm\0\0\2 \0\376\0\0\1\0\2"..., 16384}], msg_controllen=0, msg_flags=0}, 0) = 3736 <0.003848>
The following patch corrects this performance/latency problem,
removing quadratic behavior.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-01-07 22:52:14 -07:00
|
|
|
if (s_h < 0)
|
|
|
|
s_h = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
s_idx = idx = cb->args[1];
|
2008-08-28 02:11:25 -06:00
|
|
|
for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) {
|
|
|
|
if (!rt_hash_table[h].chain)
|
|
|
|
continue;
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_lock_bh();
|
|
|
|
for (rt = rcu_dereference(rt_hash_table[h].chain), idx = 0; rt;
|
2007-02-09 17:19:26 -07:00
|
|
|
rt = rcu_dereference(rt->u.dst.rt_next), idx++) {
|
2008-03-25 12:57:35 -06:00
|
|
|
if (!net_eq(dev_net(rt->u.dst.dev), net) || idx < s_idx)
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
2008-07-05 20:04:32 -06:00
|
|
|
if (rt_is_expired(rt))
|
[IPV4] route cache: Introduce rt_genid for smooth cache invalidation
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
2008-01-31 18:05:09 -07:00
|
|
|
continue;
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_set(skb, dst_clone(&rt->u.dst));
|
2009-01-21 21:56:23 -07:00
|
|
|
if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid,
|
2007-02-09 07:24:47 -07:00
|
|
|
cb->nlh->nlmsg_seq, RTM_NEWROUTE,
|
2005-06-18 23:54:12 -06:00
|
|
|
1, NLM_F_MULTI) <= 0) {
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_drop(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
rcu_read_unlock_bh();
|
|
|
|
goto done;
|
|
|
|
}
|
2009-06-01 23:19:30 -06:00
|
|
|
skb_dst_drop(skb);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
rcu_read_unlock_bh();
|
|
|
|
}
|
|
|
|
|
|
|
|
done:
|
|
|
|
cb->args[0] = h;
|
|
|
|
cb->args[1] = idx;
|
|
|
|
return skb->len;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ip_rt_multicast_event(struct in_device *in_dev)
|
|
|
|
{
|
2008-07-05 20:00:44 -06:00
|
|
|
rt_cache_flush(dev_net(in_dev->dev), 0);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_SYSCTL
|
2008-07-08 04:05:28 -06:00
|
|
|
static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write,
|
2009-09-23 16:57:19 -06:00
|
|
|
void __user *buffer,
|
2005-04-16 16:20:36 -06:00
|
|
|
size_t *lenp, loff_t *ppos)
|
|
|
|
{
|
|
|
|
if (write) {
|
2008-07-05 20:02:06 -06:00
|
|
|
int flush_delay;
|
2008-07-08 04:05:28 -06:00
|
|
|
ctl_table ctl;
|
2008-07-05 20:02:33 -06:00
|
|
|
struct net *net;
|
2008-07-05 20:02:06 -06:00
|
|
|
|
2008-07-08 04:05:28 -06:00
|
|
|
memcpy(&ctl, __ctl, sizeof(ctl));
|
|
|
|
ctl.data = &flush_delay;
|
2009-09-23 16:57:19 -06:00
|
|
|
proc_dointvec(&ctl, write, buffer, lenp, ppos);
|
2008-07-05 20:02:06 -06:00
|
|
|
|
2008-07-08 04:05:28 -06:00
|
|
|
net = (struct net *)__ctl->extra1;
|
2008-07-05 20:02:33 -06:00
|
|
|
rt_cache_flush(net, flush_delay);
|
2005-04-16 16:20:36 -06:00
|
|
|
return 0;
|
2007-02-09 07:24:47 -07:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2008-08-15 14:44:31 -06:00
|
|
|
static void rt_secret_reschedule(int old)
|
|
|
|
{
|
|
|
|
struct net *net;
|
|
|
|
int new = ip_rt_secret_interval;
|
|
|
|
int diff = new - old;
|
|
|
|
|
|
|
|
if (!diff)
|
|
|
|
return;
|
|
|
|
|
|
|
|
rtnl_lock();
|
|
|
|
for_each_net(net) {
|
|
|
|
int deleted = del_timer_sync(&net->ipv4.rt_secret_timer);
|
|
|
|
|
|
|
|
if (!new)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (deleted) {
|
|
|
|
long time = net->ipv4.rt_secret_timer.expires - jiffies;
|
|
|
|
|
|
|
|
if (time <= 0 || (time += diff) <= 0)
|
|
|
|
time = 0;
|
|
|
|
|
|
|
|
net->ipv4.rt_secret_timer.expires = time;
|
|
|
|
} else
|
|
|
|
net->ipv4.rt_secret_timer.expires = new;
|
|
|
|
|
|
|
|
net->ipv4.rt_secret_timer.expires += jiffies;
|
|
|
|
add_timer(&net->ipv4.rt_secret_timer);
|
|
|
|
}
|
|
|
|
rtnl_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ipv4_sysctl_rt_secret_interval(ctl_table *ctl, int write,
|
|
|
|
void __user *buffer, size_t *lenp,
|
|
|
|
loff_t *ppos)
|
|
|
|
{
|
|
|
|
int old = ip_rt_secret_interval;
|
2009-09-23 16:57:19 -06:00
|
|
|
int ret = proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
|
2008-08-15 14:44:31 -06:00
|
|
|
|
|
|
|
rt_secret_reschedule(old);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-07-27 01:59:33 -06:00
|
|
|
static ctl_table ipv4_route_table[] = {
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
.procname = "gc_thresh",
|
|
|
|
.data = &ipv4_dst_ops.gc_thresh,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "max_size",
|
|
|
|
.data = &ip_rt_max_size,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
/* Deprecated. Use gc_min_interval_ms */
|
2007-02-09 07:24:47 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
.procname = "gc_min_interval",
|
|
|
|
.data = &ip_rt_gc_min_interval,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec_jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "gc_min_interval_ms",
|
|
|
|
.data = &ip_rt_gc_min_interval,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec_ms_jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "gc_timeout",
|
|
|
|
.data = &ip_rt_gc_timeout,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec_jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "gc_interval",
|
|
|
|
.data = &ip_rt_gc_interval,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec_jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "redirect_load",
|
|
|
|
.data = &ip_rt_redirect_load,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "redirect_number",
|
|
|
|
.data = &ip_rt_redirect_number,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "redirect_silence",
|
|
|
|
.data = &ip_rt_redirect_silence,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "error_cost",
|
|
|
|
.data = &ip_rt_error_cost,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "error_burst",
|
|
|
|
.data = &ip_rt_error_burst,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "gc_elasticity",
|
|
|
|
.data = &ip_rt_gc_elasticity,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "mtu_expires",
|
|
|
|
.data = &ip_rt_mtu_expires,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec_jiffies,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "min_pmtu",
|
|
|
|
.data = &ip_rt_min_pmtu,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "min_adv_mss",
|
|
|
|
.data = &ip_rt_min_advmss,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = proc_dointvec,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.procname = "secret_interval",
|
|
|
|
.data = &ip_rt_secret_interval,
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0644,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = ipv4_sysctl_rt_secret_interval,
|
2005-04-16 16:20:36 -06:00
|
|
|
},
|
2009-11-05 14:32:03 -07:00
|
|
|
{ }
|
2005-04-16 16:20:36 -06:00
|
|
|
};
|
2008-07-05 20:02:33 -06:00
|
|
|
|
2008-08-25 16:17:44 -06:00
|
|
|
static struct ctl_table empty[1];
|
|
|
|
|
|
|
|
static struct ctl_table ipv4_skeleton[] =
|
|
|
|
{
|
2009-11-05 14:32:03 -07:00
|
|
|
{ .procname = "route",
|
2008-08-27 03:35:18 -06:00
|
|
|
.mode = 0555, .child = ipv4_route_table},
|
2009-11-05 14:32:03 -07:00
|
|
|
{ .procname = "neigh",
|
2008-08-27 03:35:18 -06:00
|
|
|
.mode = 0555, .child = empty},
|
2008-08-25 16:17:44 -06:00
|
|
|
{ }
|
|
|
|
};
|
|
|
|
|
|
|
|
static __net_initdata struct ctl_path ipv4_path[] = {
|
2009-11-05 14:32:03 -07:00
|
|
|
{ .procname = "net", },
|
|
|
|
{ .procname = "ipv4", },
|
2008-07-05 20:02:33 -06:00
|
|
|
{ },
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct ctl_table ipv4_route_flush_table[] = {
|
|
|
|
{
|
|
|
|
.procname = "flush",
|
|
|
|
.maxlen = sizeof(int),
|
|
|
|
.mode = 0200,
|
2008-11-03 19:21:05 -07:00
|
|
|
.proc_handler = ipv4_sysctl_rtcache_flush,
|
2008-07-05 20:02:33 -06:00
|
|
|
},
|
2009-11-05 14:32:03 -07:00
|
|
|
{ },
|
2008-07-05 20:02:33 -06:00
|
|
|
};
|
|
|
|
|
2008-08-25 16:17:44 -06:00
|
|
|
static __net_initdata struct ctl_path ipv4_route_path[] = {
|
2009-11-05 14:32:03 -07:00
|
|
|
{ .procname = "net", },
|
|
|
|
{ .procname = "ipv4", },
|
|
|
|
{ .procname = "route", },
|
2008-08-25 16:17:44 -06:00
|
|
|
{ },
|
|
|
|
};
|
|
|
|
|
2008-07-05 20:02:33 -06:00
|
|
|
static __net_init int sysctl_route_net_init(struct net *net)
|
|
|
|
{
|
|
|
|
struct ctl_table *tbl;
|
|
|
|
|
|
|
|
tbl = ipv4_route_flush_table;
|
2009-11-25 16:14:13 -07:00
|
|
|
if (!net_eq(net, &init_net)) {
|
2008-07-05 20:02:33 -06:00
|
|
|
tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
|
|
|
|
if (tbl == NULL)
|
|
|
|
goto err_dup;
|
|
|
|
}
|
|
|
|
tbl[0].extra1 = net;
|
|
|
|
|
|
|
|
net->ipv4.route_hdr =
|
|
|
|
register_net_sysctl_table(net, ipv4_route_path, tbl);
|
|
|
|
if (net->ipv4.route_hdr == NULL)
|
|
|
|
goto err_reg;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err_reg:
|
|
|
|
if (tbl != ipv4_route_flush_table)
|
|
|
|
kfree(tbl);
|
|
|
|
err_dup:
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
static __net_exit void sysctl_route_net_exit(struct net *net)
|
|
|
|
{
|
|
|
|
struct ctl_table *tbl;
|
|
|
|
|
|
|
|
tbl = net->ipv4.route_hdr->ctl_table_arg;
|
|
|
|
unregister_net_sysctl_table(net->ipv4.route_hdr);
|
|
|
|
BUG_ON(tbl == ipv4_route_flush_table);
|
|
|
|
kfree(tbl);
|
|
|
|
}
|
|
|
|
|
|
|
|
static __net_initdata struct pernet_operations sysctl_route_ops = {
|
|
|
|
.init = sysctl_route_net_init,
|
|
|
|
.exit = sysctl_route_net_exit,
|
|
|
|
};
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
|
|
|
|
2008-07-05 20:02:59 -06:00
|
|
|
|
|
|
|
static __net_init int rt_secret_timer_init(struct net *net)
|
|
|
|
{
|
2008-07-05 20:04:32 -06:00
|
|
|
atomic_set(&net->ipv4.rt_genid,
|
|
|
|
(int) ((num_physpages ^ (num_physpages>>8)) ^
|
|
|
|
(jiffies ^ (jiffies >> 7))));
|
|
|
|
|
2008-07-05 20:02:59 -06:00
|
|
|
net->ipv4.rt_secret_timer.function = rt_secret_rebuild;
|
|
|
|
net->ipv4.rt_secret_timer.data = (unsigned long)net;
|
|
|
|
init_timer_deferrable(&net->ipv4.rt_secret_timer);
|
|
|
|
|
2008-08-15 14:44:31 -06:00
|
|
|
if (ip_rt_secret_interval) {
|
|
|
|
net->ipv4.rt_secret_timer.expires =
|
|
|
|
jiffies + net_random() % ip_rt_secret_interval +
|
|
|
|
ip_rt_secret_interval;
|
|
|
|
add_timer(&net->ipv4.rt_secret_timer);
|
|
|
|
}
|
2008-07-05 20:02:59 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static __net_exit void rt_secret_timer_exit(struct net *net)
|
|
|
|
{
|
|
|
|
del_timer_sync(&net->ipv4.rt_secret_timer);
|
|
|
|
}
|
|
|
|
|
|
|
|
static __net_initdata struct pernet_operations rt_secret_timer_ops = {
|
|
|
|
.init = rt_secret_timer_init,
|
|
|
|
.exit = rt_secret_timer_exit,
|
|
|
|
};
|
|
|
|
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
2007-11-16 04:32:10 -07:00
|
|
|
struct ip_rt_acct *ip_rt_acct __read_mostly;
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif /* CONFIG_NET_CLS_ROUTE */
|
|
|
|
|
|
|
|
static __initdata unsigned long rhash_entries;
|
|
|
|
static int __init set_rhash_entries(char *str)
|
|
|
|
{
|
|
|
|
if (!str)
|
|
|
|
return 0;
|
|
|
|
rhash_entries = simple_strtoul(str, &str, 0);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("rhash_entries=", set_rhash_entries);
|
|
|
|
|
|
|
|
int __init ip_rt_init(void)
|
|
|
|
{
|
2005-07-05 15:58:19 -06:00
|
|
|
int rc = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
#ifdef CONFIG_NET_CLS_ROUTE
|
2009-02-25 06:07:33 -07:00
|
|
|
ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!ip_rt_acct)
|
|
|
|
panic("IP: failed to allocate ip_rt_acct\n");
|
|
|
|
#endif
|
|
|
|
|
2006-08-26 20:25:52 -06:00
|
|
|
ipv4_dst_ops.kmem_cachep =
|
|
|
|
kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
|
2007-07-19 19:11:58 -06:00
|
|
|
SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-05-24 19:17:54 -06:00
|
|
|
ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;
|
|
|
|
|
2005-07-05 15:58:19 -06:00
|
|
|
rt_hash_table = (struct rt_hash_bucket *)
|
|
|
|
alloc_large_system_hash("IP route cache",
|
|
|
|
sizeof(struct rt_hash_bucket),
|
|
|
|
rhash_entries,
|
2009-09-21 18:03:05 -06:00
|
|
|
(totalram_pages >= 128 * 1024) ?
|
[IPV4] tcp/route: Another look at hash table sizes
The tcp_ehash hash table gets too big on systems with really big memory.
It is worse on systems with pages larger than 4KB. It wastes memory that
could be better used. It also makes the netstat command slow because reading
/proc/net/tcp and /proc/net/tcp6 needs to go through the full hash table.
The default value should not be larger for larger page sizes. It seems
that the effect of page size is an unintended error dating back a long
time. I also wonder if the default value really should be a larger
fraction of memory for systems with more memory. While systems with
really big ram can afford more space for hash tables, it is not clear to
me that they benefit from increasing the allocation ratio for this table.
The amount of memory allocated is determined by net/ipv4/tcp.c:tcp_init and
mm/page_alloc.c:alloc_large_system_hash.
tcp_init calls alloc_large_system_hash passing parameters-
bucketsize=sizeof(struct tcp_ehash_bucket)
numentries=thash_entries
scale=(num_physpages >= 128 * 1024) ? (25-PAGE_SHIFT) : (27-PAGE_SHIFT)
limit=0
On i386, PAGE_SHIFT is 12 for a page size of 4K
On ia64, PAGE_SHIFT defaults to 14 for a page size of 16K
The num_physpages test above makes the allocation take a larger fraction
of the total memory on systems with larger memory. The threshold size
for a i386 system is 512MB. For an ia64 system with 16KB pages the
threshold is 2GB.
For smaller memory systems-
On i386, scale = (27 - 12) = 15
On ia64, scale = (27 - 14) = 13
For larger memory systems-
On i386, scale = (25 - 12) = 13
On ia64, scale = (25 - 14) = 11
For the rest of this discussion, I'll just track the larger memory case.
The default behavior has numentries=thash_entries=0, so the allocated
size is determined by either scale or by the default limit of 1/16 of
total memory.
In alloc_large_system_hash-
| numentries = (flags & HASH_HIGHMEM) ? nr_all_pages : nr_kernel_pages;
| numentries += (1UL << (20 - PAGE_SHIFT)) - 1;
| numentries >>= 20 - PAGE_SHIFT;
| numentries <<= 20 - PAGE_SHIFT;
At this point, numentries is pages for all of memory, rounded up to the
nearest megabyte boundary.
| /* limit to 1 bucket per 2^scale bytes of low memory */
| if (scale > PAGE_SHIFT)
| numentries >>= (scale - PAGE_SHIFT);
| else
| numentries <<= (PAGE_SHIFT - scale);
On i386, numentries >>= (13 - 12), so numentries is 1/8196 of
bytes of total memory.
On ia64, numentries <<= (14 - 11), so numentries is 1/2048 of
bytes of total memory.
| log2qty = long_log2(numentries);
|
| do {
| size = bucketsize << log2qty;
bucketsize is 16, so size is 16 times numentries, rounded
down to a power of two.
On i386, size is 1/512 of bytes of total memory.
On ia64, size is 1/128 of bytes of total memory.
For smaller systems the results are
On i386, size is 1/2048 of bytes of total memory.
On ia64, size is 1/512 of bytes of total memory.
The large page effect can be removed by just replacing
the use of PAGE_SHIFT with a constant of 12 in the calls to
alloc_large_system_hash. That makes them more like the other uses of
that function from fs/inode.c and fs/dcache.c
Signed-off-by: David S. Miller <davem@davemloft.net>
2005-11-29 17:12:55 -07:00
|
|
|
15 : 17,
|
2006-08-07 21:44:22 -06:00
|
|
|
0,
|
2005-07-05 15:58:19 -06:00
|
|
|
&rt_hash_log,
|
|
|
|
&rt_hash_mask,
|
2009-04-27 06:42:24 -06:00
|
|
|
rhash_entries ? 0 : 512 * 1024);
|
2005-07-05 15:55:24 -06:00
|
|
|
memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket));
|
|
|
|
rt_hash_lock_init();
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1);
|
|
|
|
ip_rt_max_size = (rt_hash_mask + 1) * 16;
|
|
|
|
|
|
|
|
devinet_init();
|
|
|
|
ip_fib_init();
|
|
|
|
|
|
|
|
/* All the timers, started at system startup tend
|
|
|
|
to synchronize. Perturb it a bit.
|
|
|
|
*/
|
2009-06-11 14:10:07 -06:00
|
|
|
INIT_DELAYED_WORK_DEFERRABLE(&expires_work, rt_worker_func);
|
|
|
|
expires_ljiffies = jiffies;
|
2007-09-15 11:55:54 -06:00
|
|
|
schedule_delayed_work(&expires_work,
|
|
|
|
net_random() % ip_rt_gc_interval + ip_rt_gc_interval);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-07-05 20:02:59 -06:00
|
|
|
if (register_pernet_subsys(&rt_secret_timer_ops))
|
|
|
|
printk(KERN_ERR "Unable to setup rt_secret_timer\n");
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-02-28 21:51:18 -07:00
|
|
|
if (ip_rt_proc_init())
|
2007-12-05 22:14:28 -07:00
|
|
|
printk(KERN_ERR "Unable to create route proc files\n");
|
2005-04-16 16:20:36 -06:00
|
|
|
#ifdef CONFIG_XFRM
|
|
|
|
xfrm_init();
|
2009-07-30 19:52:15 -06:00
|
|
|
xfrm4_init(ip_rt_max_size);
|
2005-04-16 16:20:36 -06:00
|
|
|
#endif
|
2007-03-22 12:55:17 -06:00
|
|
|
rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL);
|
|
|
|
|
2008-07-05 20:02:33 -06:00
|
|
|
#ifdef CONFIG_SYSCTL
|
|
|
|
register_pernet_subsys(&sysctl_route_ops);
|
|
|
|
#endif
|
2005-04-16 16:20:36 -06:00
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2008-07-30 04:32:52 -06:00
|
|
|
#ifdef CONFIG_SYSCTL
|
2008-07-27 01:59:33 -06:00
|
|
|
/*
|
|
|
|
* We really need to sanitize the damn ipv4 init order, then all
|
|
|
|
* this nonsense will go away.
|
|
|
|
*/
|
|
|
|
void __init ip_static_sysctl_init(void)
|
|
|
|
{
|
2008-08-25 16:17:44 -06:00
|
|
|
register_sysctl_paths(ipv4_path, ipv4_skeleton);
|
2008-07-27 01:59:33 -06:00
|
|
|
}
|
2008-07-30 04:32:52 -06:00
|
|
|
#endif
|
2008-07-27 01:59:33 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
EXPORT_SYMBOL(__ip_select_ident);
|
|
|
|
EXPORT_SYMBOL(ip_route_input);
|
|
|
|
EXPORT_SYMBOL(ip_route_output_key);
|