From: Eric Dumazet <eric.dumazet@gmail•com>
To: "David S . Miller" <davem@davemloft•net>,
Jakub Kicinski <kuba@kernel•org>
Cc: David Ahern <dsahern@kernel•org>, netdev <netdev@vger•kernel.org>,
Eric Dumazet <edumazet@google•com>,
Eric Dumazet <eric.dumazet@gmail•com>
Subject: [PATCH v2 net 1/2] ipv4: avoid quadratic behavior in netns dismantle
Date: Wed, 19 Jan 2022 02:04:12 -0800 [thread overview]
Message-ID: <20220119100413.4077866-2-eric.dumazet@gmail.com> (raw)
In-Reply-To: <20220119100413.4077866-1-eric.dumazet@gmail.com>
From: Eric Dumazet <edumazet@google•com>
net/ipv4/fib_semantics.c uses an hash table of 256 slots,
keyed by device ifindexes: fib_info_devhash[DEVINDEX_HASHSIZE]
Problem is that with network namespaces, devices tend
to use the same ifindex.
lo device for instance has a fixed ifindex of one,
for all network namespaces.
This means that hosts with thousands of netns spend
a lot of time looking at some hash buckets with thousands
of elements, notably at netns dismantle.
Simply add a per netns perturbation (net_hash_mix())
to spread elements more uniformely.
Also change fib_devindex_hashfn() to use more entropy.
Fixes: aa79e66eee5d ("net: Make ifindex generation per-net namespace")
Signed-off-by: Eric Dumazet <edumazet@google•com>
Reviewed-by: David Ahern <dsahern@kernel•org>
---
net/ipv4/fib_semantics.c | 36 +++++++++++++++++-------------------
1 file changed, 17 insertions(+), 19 deletions(-)
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index 45619c005b8dddd7ccd5c7029efa4ed69b6ce1de..9813949da10493de36b9db797b6a5d94fd9bd3b1 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -29,6 +29,7 @@
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/netlink.h>
+#include <linux/hash.h>
#include <net/arp.h>
#include <net/ip.h>
@@ -319,11 +320,15 @@ static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
static inline unsigned int fib_devindex_hashfn(unsigned int val)
{
- unsigned int mask = DEVINDEX_HASHSIZE - 1;
+ return hash_32(val, DEVINDEX_HASHBITS);
+}
- return (val ^
- (val >> DEVINDEX_HASHBITS) ^
- (val >> (DEVINDEX_HASHBITS * 2))) & mask;
+static struct hlist_head *
+fib_info_devhash_bucket(const struct net_device *dev)
+{
+ u32 val = net_hash_mix(dev_net(dev)) ^ dev->ifindex;
+
+ return &fib_info_devhash[fib_devindex_hashfn(val)];
}
static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
@@ -433,12 +438,11 @@ int ip_fib_check_default(__be32 gw, struct net_device *dev)
{
struct hlist_head *head;
struct fib_nh *nh;
- unsigned int hash;
spin_lock(&fib_info_lock);
- hash = fib_devindex_hashfn(dev->ifindex);
- head = &fib_info_devhash[hash];
+ head = fib_info_devhash_bucket(dev);
+
hlist_for_each_entry(nh, head, nh_hash) {
if (nh->fib_nh_dev == dev &&
nh->fib_nh_gw4 == gw &&
@@ -1609,12 +1613,10 @@ struct fib_info *fib_create_info(struct fib_config *cfg,
} else {
change_nexthops(fi) {
struct hlist_head *head;
- unsigned int hash;
if (!nexthop_nh->fib_nh_dev)
continue;
- hash = fib_devindex_hashfn(nexthop_nh->fib_nh_dev->ifindex);
- head = &fib_info_devhash[hash];
+ head = fib_info_devhash_bucket(nexthop_nh->fib_nh_dev);
hlist_add_head(&nexthop_nh->nh_hash, head);
} endfor_nexthops(fi)
}
@@ -1966,8 +1968,7 @@ void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig)
void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
{
- unsigned int hash = fib_devindex_hashfn(dev->ifindex);
- struct hlist_head *head = &fib_info_devhash[hash];
+ struct hlist_head *head = fib_info_devhash_bucket(dev);
struct fib_nh *nh;
hlist_for_each_entry(nh, head, nh_hash) {
@@ -1986,12 +1987,11 @@ void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
*/
int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
{
- int ret = 0;
- int scope = RT_SCOPE_NOWHERE;
+ struct hlist_head *head = fib_info_devhash_bucket(dev);
struct fib_info *prev_fi = NULL;
- unsigned int hash = fib_devindex_hashfn(dev->ifindex);
- struct hlist_head *head = &fib_info_devhash[hash];
+ int scope = RT_SCOPE_NOWHERE;
struct fib_nh *nh;
+ int ret = 0;
if (force)
scope = -1;
@@ -2136,7 +2136,6 @@ static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
{
struct fib_info *prev_fi;
- unsigned int hash;
struct hlist_head *head;
struct fib_nh *nh;
int ret;
@@ -2152,8 +2151,7 @@ int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
}
prev_fi = NULL;
- hash = fib_devindex_hashfn(dev->ifindex);
- head = &fib_info_devhash[hash];
+ head = fib_info_devhash_bucket(dev);
ret = 0;
hlist_for_each_entry(nh, head, nh_hash) {
--
2.34.1.703.g22d0c6ccf7-goog
next prev parent reply other threads:[~2022-01-19 10:04 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-01-19 10:04 [PATCH v2 net 0/2] ipv4: avoid pathological hash tables Eric Dumazet
2022-01-19 10:04 ` Eric Dumazet [this message]
2022-01-19 10:04 ` [PATCH v2 net 2/2] ipv4: add net_hash_mix() dispersion to fib_info_laddrhash keys Eric Dumazet
2022-01-19 15:40 ` David Ahern
2022-01-19 18:24 ` [PATCH v2 net 0/2] ipv4: avoid pathological hash tables Jakub Kicinski
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20220119100413.4077866-2-eric.dumazet@gmail.com \
--to=eric.dumazet@gmail$(echo .)com \
--cc=davem@davemloft$(echo .)net \
--cc=dsahern@kernel$(echo .)org \
--cc=edumazet@google$(echo .)com \
--cc=kuba@kernel$(echo .)org \
--cc=netdev@vger$(echo .)kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox