[RFC PATCH v3] net: sched: convert qdisc linked list to hashtable

From: Jiri Kosina
Date: Mon Jul 11 2016 - 10:03:06 EST


From: Jiri Kosina <jkosina@xxxxxxx>

Convert the per-device linked list into a hashtable. The primary
motivation for this change is that currently, we're not tracking all the
qdiscs in hierarchy (e.g. excluding default qdiscs), as the lookup
performed over the linked list by qdisc_match_from_root() is rather
expensive.

The ultimate goal is to get rid of hidden qdiscs completely, which will
bring much more determinism in user experience.

As we're adding hashtable.h include into generic netdevice.h, we have to
make sure HASH_SIZE macro is now non-conflicting with local definitions.

Signed-off-by: Jiri Kosina <jkosina@xxxxxxx>
---

v1 -> v2: fix up RCU hastable usage wrt. rtnl
fix compilation of .c files which define their own
HASH_SIZE that now oncflicts with the one from
hashtable.h (newly included via netdevice.h)

v2 -> v3: resolve HASH_SIZE identifier conflicts in a cleaner way
fix up the number of hash bucket bits (4 bits for 16 buckets)

include/linux/netdevice.h | 2 ++
include/net/pkt_sched.h | 4 ++--
include/net/sch_generic.h | 2 +-
net/core/dev.c | 1 +
net/ipv6/ip6_gre.c | 12 ++++++------
net/ipv6/ip6_tunnel.c | 10 +++++-----
net/ipv6/ip6_vti.c | 10 +++++-----
net/ipv6/sit.c | 10 +++++-----
net/sched/sch_api.c | 23 +++++++++++++----------
net/sched/sch_generic.c | 6 +++---
net/sched/sch_mq.c | 2 +-
net/sched/sch_mqprio.c | 2 +-
12 files changed, 45 insertions(+), 39 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index f45929c..0b5c172e 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -52,6 +52,7 @@
#include <uapi/linux/netdevice.h>
#include <uapi/linux/if_bonding.h>
#include <uapi/linux/pkt_cls.h>
+#include <linux/hashtable.h>

struct netpoll_info;
struct device;
@@ -1778,6 +1779,7 @@ struct net_device {
unsigned int num_tx_queues;
unsigned int real_num_tx_queues;
struct Qdisc *qdisc;
+ DECLARE_HASHTABLE (qdisc_hash, 4);
unsigned long tx_queue_len;
spinlock_t tx_global_lock;
int watchdog_timeo;
diff --git a/include/net/pkt_sched.h b/include/net/pkt_sched.h
index fea53f4..8ba11b4 100644
--- a/include/net/pkt_sched.h
+++ b/include/net/pkt_sched.h
@@ -90,8 +90,8 @@ int unregister_qdisc(struct Qdisc_ops *qops);
void qdisc_get_default(char *id, size_t len);
int qdisc_set_default(const char *id);

-void qdisc_list_add(struct Qdisc *q);
-void qdisc_list_del(struct Qdisc *q);
+void qdisc_hash_add(struct Qdisc *q);
+void qdisc_hash_del(struct Qdisc *q);
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle);
struct Qdisc *qdisc_lookup_class(struct net_device *dev, u32 handle);
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r,
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 62d5531..26f5cb3 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -67,7 +67,7 @@ struct Qdisc {
u32 limit;
const struct Qdisc_ops *ops;
struct qdisc_size_table __rcu *stab;
- struct list_head list;
+ struct hlist_node hash;
u32 handle;
u32 parent;
int (*reshape_fail)(struct sk_buff *skb,
diff --git a/net/core/dev.c b/net/core/dev.c
index 904ff43..edc1617 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -7511,6 +7511,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
INIT_LIST_HEAD(&dev->all_adj_list.lower);
INIT_LIST_HEAD(&dev->ptype_all);
INIT_LIST_HEAD(&dev->ptype_specific);
+ hash_init(dev->qdisc_hash);
dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
setup(dev);

diff --git a/net/ipv6/ip6_gre.c b/net/ipv6/ip6_gre.c
index fdc9de2..d3697a4 100644
--- a/net/ipv6/ip6_gre.c
+++ b/net/ipv6/ip6_gre.c
@@ -61,12 +61,12 @@ static bool log_ecn_error = true;
module_param(log_ecn_error, bool, 0644);
MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");

-#define HASH_SIZE_SHIFT 5
-#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
+#define IP6_GRE_HASH_SIZE_SHIFT 5
+#define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)

static int ip6gre_net_id __read_mostly;
struct ip6gre_net {
- struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
+ struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];

struct net_device *fb_tunnel_dev;
};
@@ -96,12 +96,12 @@ static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
will match fallback tunnel.
*/

-#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
+#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
static u32 HASH_ADDR(const struct in6_addr *addr)
{
u32 hash = ipv6_addr_hash(addr);

- return hash_32(hash, HASH_SIZE_SHIFT);
+ return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
}

#define tunnels_r_l tunnels[3]
@@ -1089,7 +1089,7 @@ static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)

for (prio = 0; prio < 4; prio++) {
int h;
- for (h = 0; h < HASH_SIZE; h++) {
+ for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
struct ip6_tnl *t;

t = rtnl_dereference(ign->tunnels[prio][h]);
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index 7b0481e..2050217 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -64,8 +64,8 @@ MODULE_LICENSE("GPL");
MODULE_ALIAS_RTNL_LINK("ip6tnl");
MODULE_ALIAS_NETDEV("ip6tnl0");

-#define HASH_SIZE_SHIFT 5
-#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
+#define IP6_TUNNEL_HASH_SIZE_SHIFT 5
+#define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)

static bool log_ecn_error = true;
module_param(log_ecn_error, bool, 0644);
@@ -75,7 +75,7 @@ static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
{
u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);

- return hash_32(hash, HASH_SIZE_SHIFT);
+ return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
}

static int ip6_tnl_dev_init(struct net_device *dev);
@@ -87,7 +87,7 @@ struct ip6_tnl_net {
/* the IPv6 tunnel fallback device */
struct net_device *fb_tnl_dev;
/* lists for storing tunnels in use */
- struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
+ struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
struct ip6_tnl __rcu *tnls_wc[1];
struct ip6_tnl __rcu **tnls[2];
};
@@ -2031,7 +2031,7 @@ static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
if (dev->rtnl_link_ops == &ip6_link_ops)
unregister_netdevice_queue(dev, &list);

- for (h = 0; h < HASH_SIZE; h++) {
+ for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
t = rtnl_dereference(ip6n->tnls_r_l[h]);
while (t) {
/* If dev is in the same netns, it has already
diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
index d90a11f..cc7e058 100644
--- a/net/ipv6/ip6_vti.c
+++ b/net/ipv6/ip6_vti.c
@@ -50,14 +50,14 @@
#include <net/net_namespace.h>
#include <net/netns/generic.h>

-#define HASH_SIZE_SHIFT 5
-#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
+#define IP6_VTI_HASH_SIZE_SHIFT 5
+#define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)

static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
{
u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);

- return hash_32(hash, HASH_SIZE_SHIFT);
+ return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
}

static int vti6_dev_init(struct net_device *dev);
@@ -69,7 +69,7 @@ struct vti6_net {
/* the vti6 tunnel fallback device */
struct net_device *fb_tnl_dev;
/* lists for storing tunnels in use */
- struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
+ struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
struct ip6_tnl __rcu *tnls_wc[1];
struct ip6_tnl __rcu **tnls[2];
};
@@ -1040,7 +1040,7 @@ static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
struct ip6_tnl *t;
LIST_HEAD(list);

- for (h = 0; h < HASH_SIZE; h++) {
+ for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
t = rtnl_dereference(ip6n->tnls_r_l[h]);
while (t) {
unregister_netdevice_queue(t->dev, &list);
diff --git a/net/ipv6/sit.c b/net/ipv6/sit.c
index 0a5a255..94dd0f0 100644
--- a/net/ipv6/sit.c
+++ b/net/ipv6/sit.c
@@ -62,7 +62,7 @@
For comments look at net/ipv4/ip_gre.c --ANK
*/

-#define HASH_SIZE 16
+#define IP6_SIT_HASH_SIZE 16
#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)

static bool log_ecn_error = true;
@@ -78,9 +78,9 @@ static struct rtnl_link_ops sit_link_ops __read_mostly;

static int sit_net_id __read_mostly;
struct sit_net {
- struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
- struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
- struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
+ struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
+ struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
+ struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
struct ip_tunnel __rcu *tunnels_wc[1];
struct ip_tunnel __rcu **tunnels[4];

@@ -1773,7 +1773,7 @@ static void __net_exit sit_destroy_tunnels(struct net *net,

for (prio = 1; prio < 4; prio++) {
int h;
- for (h = 0; h < HASH_SIZE; h++) {
+ for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
struct ip_tunnel *t;

t = rtnl_dereference(sitn->tunnels[prio][h]);
diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c
index ddf047d..c093d32 100644
--- a/net/sched/sch_api.c
+++ b/net/sched/sch_api.c
@@ -29,6 +29,7 @@
#include <linux/hrtimer.h>
#include <linux/lockdep.h>
#include <linux/slab.h>
+#include <linux/hashtable.h>

#include <net/net_namespace.h>
#include <net/sock.h>
@@ -265,33 +266,33 @@ static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle)
root->handle == handle)
return root;

- list_for_each_entry_rcu(q, &root->list, list) {
+ hash_for_each_possible_rcu(qdisc_dev(root)->qdisc_hash, q, hash, handle) {
if (q->handle == handle)
return q;
}
return NULL;
}

-void qdisc_list_add(struct Qdisc *q)
+void qdisc_hash_add(struct Qdisc *q)
{
if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) {
struct Qdisc *root = qdisc_dev(q)->qdisc;

WARN_ON_ONCE(root == &noop_qdisc);
ASSERT_RTNL();
- list_add_tail_rcu(&q->list, &root->list);
+ hash_add_rcu(qdisc_dev(q)->qdisc_hash, &q->hash, q->handle);
}
}
-EXPORT_SYMBOL(qdisc_list_add);
+EXPORT_SYMBOL(qdisc_hash_add);

-void qdisc_list_del(struct Qdisc *q)
+void qdisc_hash_del(struct Qdisc *q)
{
if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) {
ASSERT_RTNL();
- list_del_rcu(&q->list);
+ hash_del_rcu(&q->hash);
}
}
-EXPORT_SYMBOL(qdisc_list_del);
+EXPORT_SYMBOL(qdisc_hash_del);

struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
{
@@ -1004,7 +1005,7 @@ qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue,
goto err_out4;
}

- qdisc_list_add(sch);
+ qdisc_hash_add(sch);

return sch;
}
@@ -1440,6 +1441,7 @@ static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb,
{
int ret = 0, q_idx = *q_idx_p;
struct Qdisc *q;
+ int b;

if (!root)
return 0;
@@ -1454,7 +1456,7 @@ static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb,
goto done;
q_idx++;
}
- list_for_each_entry(q, &root->list, list) {
+ hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
if (q_idx < s_q_idx) {
q_idx++;
continue;
@@ -1771,6 +1773,7 @@ static int tc_dump_tclass_root(struct Qdisc *root, struct sk_buff *skb,
int *t_p, int s_t)
{
struct Qdisc *q;
+ int b;

if (!root)
return 0;
@@ -1778,7 +1781,7 @@ static int tc_dump_tclass_root(struct Qdisc *root, struct sk_buff *skb,
if (tc_dump_tclass_qdisc(root, skb, tcm, cb, t_p, s_t) < 0)
return -1;

- list_for_each_entry(q, &root->list, list) {
+ hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0)
return -1;
}
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index f9e0e9c..7efc923 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -378,7 +378,6 @@ struct Qdisc noop_qdisc = {
.dequeue = noop_dequeue,
.flags = TCQ_F_BUILTIN,
.ops = &noop_qdisc_ops,
- .list = LIST_HEAD_INIT(noop_qdisc.list),
.q.lock = __SPIN_LOCK_UNLOCKED(noop_qdisc.q.lock),
.dev_queue = &noop_netdev_queue,
.busylock = __SPIN_LOCK_UNLOCKED(noop_qdisc.busylock),
@@ -565,7 +564,6 @@ struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
sch = (struct Qdisc *) QDISC_ALIGN((unsigned long) p);
sch->padded = (char *) sch - (char *) p;
}
- INIT_LIST_HEAD(&sch->list);
skb_queue_head_init(&sch->q);

spin_lock_init(&sch->busylock);
@@ -645,7 +643,7 @@ void qdisc_destroy(struct Qdisc *qdisc)
return;

#ifdef CONFIG_NET_SCHED
- qdisc_list_del(qdisc);
+ qdisc_hash_del(qdisc);

qdisc_put_stab(rtnl_dereference(qdisc->stab));
#endif
@@ -732,6 +730,8 @@ static void attach_default_qdiscs(struct net_device *dev)
qdisc->ops->attach(qdisc);
}
}
+ if (dev->qdisc)
+ qdisc_hash_add(dev->qdisc);
}

static void transition_one_qdisc(struct net_device *dev,
diff --git a/net/sched/sch_mq.c b/net/sched/sch_mq.c
index 56a77b8..3bee15d 100644
--- a/net/sched/sch_mq.c
+++ b/net/sched/sch_mq.c
@@ -88,7 +88,7 @@ static void mq_attach(struct Qdisc *sch)
qdisc_destroy(old);
#ifdef CONFIG_NET_SCHED
if (ntx < dev->real_num_tx_queues)
- qdisc_list_add(qdisc);
+ qdisc_hash_add(qdisc);
#endif

}
diff --git a/net/sched/sch_mqprio.c b/net/sched/sch_mqprio.c
index b8002ce..dbfb3a5 100644
--- a/net/sched/sch_mqprio.c
+++ b/net/sched/sch_mqprio.c
@@ -182,7 +182,7 @@ static void mqprio_attach(struct Qdisc *sch)
if (old)
qdisc_destroy(old);
if (ntx < dev->real_num_tx_queues)
- qdisc_list_add(qdisc);
+ qdisc_hash_add(qdisc);
}
kfree(priv->qdiscs);
priv->qdiscs = NULL;

--
Jiri Kosina
SUSE Labs