Update Linux to v5.10.109
Sourced from [1]
[1] https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.10.109.tar.xz
Change-Id: I19bca9fc6762d4e63bcf3e4cba88bbe560d9c76c
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
diff --git a/drivers/net/vrf.c b/drivers/net/vrf.c
index 9b626c1..8ab0b5a 100644
--- a/drivers/net/vrf.c
+++ b/drivers/net/vrf.c
@@ -21,6 +21,7 @@
#include <net/rtnetlink.h>
#include <linux/u64_stats_sync.h>
#include <linux/hashtable.h>
+#include <linux/spinlock_types.h>
#include <linux/inetdevice.h>
#include <net/arp.h>
@@ -33,14 +34,79 @@
#include <net/l3mdev.h>
#include <net/fib_rules.h>
#include <net/netns/generic.h>
+#include <net/netfilter/nf_conntrack.h>
#define DRV_NAME "vrf"
-#define DRV_VERSION "1.0"
+#define DRV_VERSION "1.1"
#define FIB_RULE_PREF 1000 /* default preference for FIB rules */
+#define HT_MAP_BITS 4
+#define HASH_INITVAL ((u32)0xcafef00d)
+
+struct vrf_map {
+ DECLARE_HASHTABLE(ht, HT_MAP_BITS);
+ spinlock_t vmap_lock;
+
+ /* shared_tables:
+ * count how many distinct tables do not comply with the strict mode
+ * requirement.
+ * shared_tables value must be 0 in order to enable the strict mode.
+ *
+ * example of the evolution of shared_tables:
+ * | time
+ * add vrf0 --> table 100 shared_tables = 0 | t0
+ * add vrf1 --> table 101 shared_tables = 0 | t1
+ * add vrf2 --> table 100 shared_tables = 1 | t2
+ * add vrf3 --> table 100 shared_tables = 1 | t3
+ * add vrf4 --> table 101 shared_tables = 2 v t4
+ *
+ * shared_tables is a "step function" (or "staircase function")
+ * and it is increased by one when the second vrf is associated to a
+ * table.
+ *
+ * at t2, vrf0 and vrf2 are bound to table 100: shared_tables = 1.
+ *
+ * at t3, another dev (vrf3) is bound to the same table 100 but the
+ * value of shared_tables is still 1.
+ * This means that no matter how many new vrfs will register on the
+ * table 100, the shared_tables will not increase (considering only
+ * table 100).
+ *
+ * at t4, vrf4 is bound to table 101, and shared_tables = 2.
+ *
+ * Looking at the value of shared_tables we can immediately know if
+ * the strict_mode can or cannot be enforced. Indeed, strict_mode
+ * can be enforced iff shared_tables = 0.
+ *
+ * Conversely, shared_tables is decreased when a vrf is de-associated
+ * from a table with exactly two associated vrfs.
+ */
+ u32 shared_tables;
+
+ bool strict_mode;
+};
+
+struct vrf_map_elem {
+ struct hlist_node hnode;
+ struct list_head vrf_list; /* VRFs registered to this table */
+
+ u32 table_id;
+ int users;
+ int ifindex;
+};
+
static unsigned int vrf_net_id;
+/* per netns vrf data */
+struct netns_vrf {
+ /* protected by rtnl lock */
+ bool add_fib_rules;
+
+ struct vrf_map vmap;
+ struct ctl_table_header *ctl_hdr;
+};
+
struct net_vrf {
struct rtable __rcu *rth;
struct rt6_info __rcu *rt6;
@@ -48,6 +114,9 @@
struct fib6_table *fib6_table;
#endif
u32 tb_id;
+
+ struct list_head me_list; /* entry in vrf_map_elem */
+ int ifindex;
};
struct pcpu_dstats {
@@ -103,6 +172,214 @@
}
}
+static struct vrf_map *netns_vrf_map(struct net *net)
+{
+ struct netns_vrf *nn_vrf = net_generic(net, vrf_net_id);
+
+ return &nn_vrf->vmap;
+}
+
+static struct vrf_map *netns_vrf_map_by_dev(struct net_device *dev)
+{
+ return netns_vrf_map(dev_net(dev));
+}
+
+static int vrf_map_elem_get_vrf_ifindex(struct vrf_map_elem *me)
+{
+ struct list_head *me_head = &me->vrf_list;
+ struct net_vrf *vrf;
+
+ if (list_empty(me_head))
+ return -ENODEV;
+
+ vrf = list_first_entry(me_head, struct net_vrf, me_list);
+
+ return vrf->ifindex;
+}
+
+static struct vrf_map_elem *vrf_map_elem_alloc(gfp_t flags)
+{
+ struct vrf_map_elem *me;
+
+ me = kmalloc(sizeof(*me), flags);
+ if (!me)
+ return NULL;
+
+ return me;
+}
+
+static void vrf_map_elem_free(struct vrf_map_elem *me)
+{
+ kfree(me);
+}
+
+static void vrf_map_elem_init(struct vrf_map_elem *me, int table_id,
+ int ifindex, int users)
+{
+ me->table_id = table_id;
+ me->ifindex = ifindex;
+ me->users = users;
+ INIT_LIST_HEAD(&me->vrf_list);
+}
+
+static struct vrf_map_elem *vrf_map_lookup_elem(struct vrf_map *vmap,
+ u32 table_id)
+{
+ struct vrf_map_elem *me;
+ u32 key;
+
+ key = jhash_1word(table_id, HASH_INITVAL);
+ hash_for_each_possible(vmap->ht, me, hnode, key) {
+ if (me->table_id == table_id)
+ return me;
+ }
+
+ return NULL;
+}
+
+static void vrf_map_add_elem(struct vrf_map *vmap, struct vrf_map_elem *me)
+{
+ u32 table_id = me->table_id;
+ u32 key;
+
+ key = jhash_1word(table_id, HASH_INITVAL);
+ hash_add(vmap->ht, &me->hnode, key);
+}
+
+static void vrf_map_del_elem(struct vrf_map_elem *me)
+{
+ hash_del(&me->hnode);
+}
+
+static void vrf_map_lock(struct vrf_map *vmap) __acquires(&vmap->vmap_lock)
+{
+ spin_lock(&vmap->vmap_lock);
+}
+
+static void vrf_map_unlock(struct vrf_map *vmap) __releases(&vmap->vmap_lock)
+{
+ spin_unlock(&vmap->vmap_lock);
+}
+
+/* called with rtnl lock held */
+static int
+vrf_map_register_dev(struct net_device *dev, struct netlink_ext_ack *extack)
+{
+ struct vrf_map *vmap = netns_vrf_map_by_dev(dev);
+ struct net_vrf *vrf = netdev_priv(dev);
+ struct vrf_map_elem *new_me, *me;
+ u32 table_id = vrf->tb_id;
+ bool free_new_me = false;
+ int users;
+ int res;
+
+ /* we pre-allocate elements used in the spin-locked section (so that we
+ * keep the spinlock as short as possibile).
+ */
+ new_me = vrf_map_elem_alloc(GFP_KERNEL);
+ if (!new_me)
+ return -ENOMEM;
+
+ vrf_map_elem_init(new_me, table_id, dev->ifindex, 0);
+
+ vrf_map_lock(vmap);
+
+ me = vrf_map_lookup_elem(vmap, table_id);
+ if (!me) {
+ me = new_me;
+ vrf_map_add_elem(vmap, me);
+ goto link_vrf;
+ }
+
+ /* we already have an entry in the vrf_map, so it means there is (at
+ * least) a vrf registered on the specific table.
+ */
+ free_new_me = true;
+ if (vmap->strict_mode) {
+ /* vrfs cannot share the same table */
+ NL_SET_ERR_MSG(extack, "Table is used by another VRF");
+ res = -EBUSY;
+ goto unlock;
+ }
+
+link_vrf:
+ users = ++me->users;
+ if (users == 2)
+ ++vmap->shared_tables;
+
+ list_add(&vrf->me_list, &me->vrf_list);
+
+ res = 0;
+
+unlock:
+ vrf_map_unlock(vmap);
+
+ /* clean-up, if needed */
+ if (free_new_me)
+ vrf_map_elem_free(new_me);
+
+ return res;
+}
+
+/* called with rtnl lock held */
+static void vrf_map_unregister_dev(struct net_device *dev)
+{
+ struct vrf_map *vmap = netns_vrf_map_by_dev(dev);
+ struct net_vrf *vrf = netdev_priv(dev);
+ u32 table_id = vrf->tb_id;
+ struct vrf_map_elem *me;
+ int users;
+
+ vrf_map_lock(vmap);
+
+ me = vrf_map_lookup_elem(vmap, table_id);
+ if (!me)
+ goto unlock;
+
+ list_del(&vrf->me_list);
+
+ users = --me->users;
+ if (users == 1) {
+ --vmap->shared_tables;
+ } else if (users == 0) {
+ vrf_map_del_elem(me);
+
+ /* no one will refer to this element anymore */
+ vrf_map_elem_free(me);
+ }
+
+unlock:
+ vrf_map_unlock(vmap);
+}
+
+/* return the vrf device index associated with the table_id */
+static int vrf_ifindex_lookup_by_table_id(struct net *net, u32 table_id)
+{
+ struct vrf_map *vmap = netns_vrf_map(net);
+ struct vrf_map_elem *me;
+ int ifindex;
+
+ vrf_map_lock(vmap);
+
+ if (!vmap->strict_mode) {
+ ifindex = -EPERM;
+ goto unlock;
+ }
+
+ me = vrf_map_lookup_elem(vmap, table_id);
+ if (!me) {
+ ifindex = -ENODEV;
+ goto unlock;
+ }
+
+ ifindex = vrf_map_elem_get_vrf_ifindex(me);
+
+unlock:
+ vrf_map_unlock(vmap);
+
+ return ifindex;
+}
+
/* by default VRF devices do not have a qdisc and are expected
* to be created with only a single queue.
*/
@@ -147,12 +424,26 @@
return NETDEV_TX_OK;
}
+static void vrf_nf_set_untracked(struct sk_buff *skb)
+{
+ if (skb_get_nfct(skb) == 0)
+ nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
+}
+
+static void vrf_nf_reset_ct(struct sk_buff *skb)
+{
+ if (skb_get_nfct(skb) == IP_CT_UNTRACKED)
+ nf_reset_ct(skb);
+}
+
#if IS_ENABLED(CONFIG_IPV6)
static int vrf_ip6_local_out(struct net *net, struct sock *sk,
struct sk_buff *skb)
{
int err;
+ vrf_nf_reset_ct(skb);
+
err = nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net,
sk, skb, NULL, skb_dst(skb)->dev, dst_output);
@@ -206,6 +497,7 @@
/* strip the ethernet header added for pass through VRF device */
__skb_pull(skb, skb_network_offset(skb));
+ memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
ret = vrf_ip6_local_out(net, skb->sk, skb);
if (unlikely(net_xmit_eval(ret)))
dev->stats.tx_errors++;
@@ -232,6 +524,8 @@
{
int err;
+ vrf_nf_reset_ct(skb);
+
err = nf_hook(NFPROTO_IPV4, NF_INET_LOCAL_OUT, net, sk,
skb, NULL, skb_dst(skb)->dev, dst_output);
if (likely(err == 1))
@@ -287,6 +581,7 @@
RT_SCOPE_LINK);
}
+ memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
ret = vrf_ip_local_out(dev_net(skb_dst(skb)->dev), skb->sk, skb);
if (unlikely(net_xmit_eval(ret)))
vrf_dev->stats.tx_errors++;
@@ -351,8 +646,7 @@
skb_pull(skb, ETH_HLEN);
}
- /* reset skb device */
- nf_reset_ct(skb);
+ vrf_nf_reset_ct(skb);
}
#if IS_ENABLED(CONFIG_IPV6)
@@ -366,7 +660,7 @@
struct neighbour *neigh;
int ret;
- nf_reset_ct(skb);
+ vrf_nf_reset_ct(skb);
skb->protocol = htons(ETH_P_IPV6);
skb->dev = dev;
@@ -497,6 +791,8 @@
if (rt6_need_strict(&ipv6_hdr(skb)->daddr))
return skb;
+ vrf_nf_set_untracked(skb);
+
if (qdisc_tx_is_default(vrf_dev) ||
IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED)
return vrf_ip6_out_direct(vrf_dev, sk, skb);
@@ -528,7 +824,7 @@
static int vrf_rt6_create(struct net_device *dev)
{
- int flags = DST_HOST | DST_NOPOLICY | DST_NOXFRM;
+ int flags = DST_NOPOLICY | DST_NOXFRM;
struct net_vrf *vrf = netdev_priv(dev);
struct net *net = dev_net(dev);
struct rt6_info *rt6;
@@ -584,7 +880,7 @@
bool is_v6gw = false;
int ret = -EINVAL;
- nf_reset_ct(skb);
+ vrf_nf_reset_ct(skb);
/* Be paranoid, rather than too clever. */
if (unlikely(skb_headroom(skb) < hh_len && dev->header_ops)) {
@@ -733,6 +1029,8 @@
ipv4_is_lbcast(ip_hdr(skb)->daddr))
return skb;
+ vrf_nf_set_untracked(skb);
+
if (qdisc_tx_is_default(vrf_dev) ||
IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
return vrf_ip_out_direct(vrf_dev, sk, skb);
@@ -787,7 +1085,7 @@
return -ENOMEM;
/* create a dst for routing packets out through a VRF device */
- rth = rt_dst_alloc(dev, 0, RTN_UNICAST, 1, 1, 0);
+ rth = rt_dst_alloc(dev, 0, RTN_UNICAST, 1, 1);
if (!rth)
return -ENOMEM;
@@ -910,6 +1208,7 @@
/* similarly, oper state is irrelevant; set to up to avoid confusion */
dev->operstate = IF_OPER_UP;
+ netdev_lockdep_set_classes(dev);
return 0;
out_rth:
@@ -1036,8 +1335,6 @@
bool need_strict = rt6_need_strict(&ipv6_hdr(skb)->daddr);
bool is_ndisc = ipv6_ndisc_frame(skb);
- nf_reset_ct(skb);
-
/* loopback, multicast & non-ND link-local traffic; do not push through
* packet taps again. Reset pkt_type for upper layers to process skb.
* For strict packets with a source LLA, determine the dst using the
@@ -1094,8 +1391,6 @@
skb->skb_iif = vrf_dev->ifindex;
IPCB(skb)->flags |= IPSKB_L3SLAVE;
- nf_reset_ct(skb);
-
if (ipv4_is_multicast(ip_hdr(skb)->daddr))
goto out;
@@ -1372,6 +1667,8 @@
netdev_for_each_lower_dev(dev, port_dev, iter)
vrf_del_slave(dev, port_dev);
+ vrf_map_unregister_dev(dev);
+
unregister_netdevice_queue(dev, head);
}
@@ -1380,6 +1677,7 @@
struct netlink_ext_ack *extack)
{
struct net_vrf *vrf = netdev_priv(dev);
+ struct netns_vrf *nn_vrf;
bool *add_fib_rules;
struct net *net;
int err;
@@ -1402,11 +1700,26 @@
if (err)
goto out;
+ /* mapping between table_id and vrf;
+ * note: such binding could not be done in the dev init function
+ * because dev->ifindex id is not available yet.
+ */
+ vrf->ifindex = dev->ifindex;
+
+ err = vrf_map_register_dev(dev, extack);
+ if (err) {
+ unregister_netdevice(dev);
+ goto out;
+ }
+
net = dev_net(dev);
- add_fib_rules = net_generic(net, vrf_net_id);
+ nn_vrf = net_generic(net, vrf_net_id);
+
+ add_fib_rules = &nn_vrf->add_fib_rules;
if (*add_fib_rules) {
err = vrf_add_fib_rules(dev);
if (err) {
+ vrf_map_unregister_dev(dev);
unregister_netdevice(dev);
goto out;
}
@@ -1493,20 +1806,164 @@
.notifier_call = vrf_device_event,
};
-/* Initialize per network namespace state */
-static int __net_init vrf_netns_init(struct net *net)
+static int vrf_map_init(struct vrf_map *vmap)
{
- bool *add_fib_rules = net_generic(net, vrf_net_id);
+ spin_lock_init(&vmap->vmap_lock);
+ hash_init(vmap->ht);
- *add_fib_rules = true;
+ vmap->strict_mode = false;
return 0;
}
+#ifdef CONFIG_SYSCTL
+static bool vrf_strict_mode(struct vrf_map *vmap)
+{
+ bool strict_mode;
+
+ vrf_map_lock(vmap);
+ strict_mode = vmap->strict_mode;
+ vrf_map_unlock(vmap);
+
+ return strict_mode;
+}
+
+static int vrf_strict_mode_change(struct vrf_map *vmap, bool new_mode)
+{
+ bool *cur_mode;
+ int res = 0;
+
+ vrf_map_lock(vmap);
+
+ cur_mode = &vmap->strict_mode;
+ if (*cur_mode == new_mode)
+ goto unlock;
+
+ if (*cur_mode) {
+ /* disable strict mode */
+ *cur_mode = false;
+ } else {
+ if (vmap->shared_tables) {
+ /* we cannot allow strict_mode because there are some
+ * vrfs that share one or more tables.
+ */
+ res = -EBUSY;
+ goto unlock;
+ }
+
+ /* no tables are shared among vrfs, so we can go back
+ * to 1:1 association between a vrf with its table.
+ */
+ *cur_mode = true;
+ }
+
+unlock:
+ vrf_map_unlock(vmap);
+
+ return res;
+}
+
+static int vrf_shared_table_handler(struct ctl_table *table, int write,
+ void *buffer, size_t *lenp, loff_t *ppos)
+{
+ struct net *net = (struct net *)table->extra1;
+ struct vrf_map *vmap = netns_vrf_map(net);
+ int proc_strict_mode = 0;
+ struct ctl_table tmp = {
+ .procname = table->procname,
+ .data = &proc_strict_mode,
+ .maxlen = sizeof(int),
+ .mode = table->mode,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ };
+ int ret;
+
+ if (!write)
+ proc_strict_mode = vrf_strict_mode(vmap);
+
+ ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+
+ if (write && ret == 0)
+ ret = vrf_strict_mode_change(vmap, (bool)proc_strict_mode);
+
+ return ret;
+}
+
+static const struct ctl_table vrf_table[] = {
+ {
+ .procname = "strict_mode",
+ .data = NULL,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = vrf_shared_table_handler,
+ /* set by the vrf_netns_init */
+ .extra1 = NULL,
+ },
+ { },
+};
+
+static int vrf_netns_init_sysctl(struct net *net, struct netns_vrf *nn_vrf)
+{
+ struct ctl_table *table;
+
+ table = kmemdup(vrf_table, sizeof(vrf_table), GFP_KERNEL);
+ if (!table)
+ return -ENOMEM;
+
+ /* init the extra1 parameter with the reference to current netns */
+ table[0].extra1 = net;
+
+ nn_vrf->ctl_hdr = register_net_sysctl(net, "net/vrf", table);
+ if (!nn_vrf->ctl_hdr) {
+ kfree(table);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static void vrf_netns_exit_sysctl(struct net *net)
+{
+ struct netns_vrf *nn_vrf = net_generic(net, vrf_net_id);
+ struct ctl_table *table;
+
+ table = nn_vrf->ctl_hdr->ctl_table_arg;
+ unregister_net_sysctl_table(nn_vrf->ctl_hdr);
+ kfree(table);
+}
+#else
+static int vrf_netns_init_sysctl(struct net *net, struct netns_vrf *nn_vrf)
+{
+ return 0;
+}
+
+static void vrf_netns_exit_sysctl(struct net *net)
+{
+}
+#endif
+
+/* Initialize per network namespace state */
+static int __net_init vrf_netns_init(struct net *net)
+{
+ struct netns_vrf *nn_vrf = net_generic(net, vrf_net_id);
+
+ nn_vrf->add_fib_rules = true;
+ vrf_map_init(&nn_vrf->vmap);
+
+ return vrf_netns_init_sysctl(net, nn_vrf);
+}
+
+static void __net_exit vrf_netns_exit(struct net *net)
+{
+ vrf_netns_exit_sysctl(net);
+}
+
static struct pernet_operations vrf_net_ops __net_initdata = {
.init = vrf_netns_init,
+ .exit = vrf_netns_exit,
.id = &vrf_net_id,
- .size = sizeof(bool),
+ .size = sizeof(struct netns_vrf),
};
static int __init vrf_init_module(void)
@@ -1519,14 +1976,24 @@
if (rc < 0)
goto error;
+ rc = l3mdev_table_lookup_register(L3MDEV_TYPE_VRF,
+ vrf_ifindex_lookup_by_table_id);
+ if (rc < 0)
+ goto unreg_pernet;
+
rc = rtnl_link_register(&vrf_link_ops);
- if (rc < 0) {
- unregister_pernet_subsys(&vrf_net_ops);
- goto error;
- }
+ if (rc < 0)
+ goto table_lookup_unreg;
return 0;
+table_lookup_unreg:
+ l3mdev_table_lookup_unregister(L3MDEV_TYPE_VRF,
+ vrf_ifindex_lookup_by_table_id);
+
+unreg_pernet:
+ unregister_pernet_subsys(&vrf_net_ops);
+
error:
unregister_netdevice_notifier(&vrf_notifier_block);
return rc;