[PATCH 0/11] pull request (net-next): ipsec-next 2024-11-15

[PATCH 0/11] pull request (net-next): ipsec-next 2024-11-15

[Steffen Klassert]

1) Add support for RFC 9611 per cpu xfrm state handling.

2) Add inbound and outbound xfrm state caches to speed up
   state lookups.

3) Convert xfrm to dscp_t. From Guillaume Nault.

4) Fix error handling in build_aevent.
   From Everest K.C.

5) Replace strncpy with strscpy_pad in copy_to_user_auth.
   From Daniel Yang.

6) Fix an uninitialized symbol during acquire state insertion.

Please pull or let me know if there are problems.

Thanks!

The following changes since commit ab101c553bc1f76a839163d1dc0d1e715ad6bb4e:

  neighbour: use kvzalloc()/kvfree() (2024-10-28 18:12:06 -0700)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next.git tags/ipsec-next-2024-11-15

for you to fetch changes up to a35672819f8d85e2ae38b80d40b923e3ef81e4ea:

  xfrm: Fix acquire state insertion. (2024-11-15 07:25:14 +0100)

----------------------------------------------------------------
ipsec-next-2024-11-15

----------------------------------------------------------------
Daniel Yang (1):
      xfrm: replace deprecated strncpy with strscpy_pad

Everest K.C (1):
      xfrm: Add error handling when nla_put_u32() returns an error

Guillaume Nault (4):
      xfrm: Convert xfrm_get_tos() to dscp_t.
      xfrm: Convert xfrm_bundle_create() to dscp_t.
      xfrm: Convert xfrm_dst_lookup() to dscp_t.
      xfrm: Convert struct xfrm_dst_lookup_params -> tos to dscp_t.

Steffen Klassert (6):
      xfrm: Add support for per cpu xfrm state handling.
      xfrm: Cache used outbound xfrm states at the policy.
      xfrm: Add an inbound percpu state cache.
      xfrm: Restrict percpu SA attribute to specific netlink message types
      Merge branch 'xfrm: Convert __xfrm4_dst_lookup() and its callers to dscp_t.'
      xfrm: Fix acquire state insertion.

 include/net/netns/xfrm.h  |   1 +
 include/net/xfrm.h        |  17 ++++-
 include/uapi/linux/xfrm.h |   2 +
 net/ipv4/esp4_offload.c   |   6 +-
 net/ipv4/xfrm4_policy.c   |   3 +-
 net/ipv6/esp6_offload.c   |   6 +-
 net/key/af_key.c          |   7 +-
 net/xfrm/xfrm_compat.c    |   6 +-
 net/xfrm/xfrm_input.c     |   2 +-
 net/xfrm/xfrm_policy.c    |  28 +++++---
 net/xfrm/xfrm_state.c     | 171 +++++++++++++++++++++++++++++++++++++++++++---
 net/xfrm/xfrm_user.c      |  75 ++++++++++++++++++--
 12 files changed, 284 insertions(+), 40 deletions(-)

[PATCH 01/11] xfrm: Add support for per cpu xfrm state handling.

[Steffen Klassert]

Currently all flows for a certain SA must be processed by the same
cpu to avoid packet reordering and lock contention of the xfrm
state lock.

To get rid of this limitation, the IETF standardized per cpu SAs
in RFC 9611. This patch implements the xfrm part of it.

We add the cpu as a lookup key for xfrm states and a config option
to generate acquire messages for each cpu.

With that, we can have on each cpu a SA with identical traffic selector
so that flows can be processed in parallel on all cpus.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Tested-by: Antony Antony <antony.antony@secunet.com>
Tested-by: Tobias Brunner <tobias@strongswan.org>
---
 include/net/xfrm.h        |  5 ++--
 include/uapi/linux/xfrm.h |  2 ++
 net/key/af_key.c          |  7 +++--
 net/xfrm/xfrm_compat.c    |  6 ++--
 net/xfrm/xfrm_state.c     | 58 +++++++++++++++++++++++++++++++--------
 net/xfrm/xfrm_user.c      | 56 ++++++++++++++++++++++++++++++++++---
 6 files changed, 112 insertions(+), 22 deletions(-)

diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index a0bdd58f401c..f5275618e744 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -188,6 +188,7 @@ struct xfrm_state {
 	refcount_t		refcnt;
 	spinlock_t		lock;
 
+	u32			pcpu_num;
 	struct xfrm_id		id;
 	struct xfrm_selector	sel;
 	struct xfrm_mark	mark;
@@ -1684,7 +1685,7 @@ struct xfrmk_spdinfo {
 	u32 spdhmcnt;
 };
 
-struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
+struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num);
 int xfrm_state_delete(struct xfrm_state *x);
 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync);
 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid);
@@ -1796,7 +1797,7 @@ int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack);
 int xfrm_alloc_spi(struct xfrm_state *x, u32 minspi, u32 maxspi,
 		   struct netlink_ext_ack *extack);
 struct xfrm_state *xfrm_find_acq(struct net *net, const struct xfrm_mark *mark,
-				 u8 mode, u32 reqid, u32 if_id, u8 proto,
+				 u8 mode, u32 reqid, u32 if_id, u32 pcpu_num, u8 proto,
 				 const xfrm_address_t *daddr,
 				 const xfrm_address_t *saddr, int create,
 				 unsigned short family);
diff --git a/include/uapi/linux/xfrm.h b/include/uapi/linux/xfrm.h
index f28701500714..d73a97e3030a 100644
--- a/include/uapi/linux/xfrm.h
+++ b/include/uapi/linux/xfrm.h
@@ -322,6 +322,7 @@ enum xfrm_attr_type_t {
 	XFRMA_MTIMER_THRESH,	/* __u32 in seconds for input SA */
 	XFRMA_SA_DIR,		/* __u8 */
 	XFRMA_NAT_KEEPALIVE_INTERVAL,	/* __u32 in seconds for NAT keepalive */
+	XFRMA_SA_PCPU,		/* __u32 */
 	__XFRMA_MAX
 
 #define XFRMA_OUTPUT_MARK XFRMA_SET_MARK	/* Compatibility */
@@ -437,6 +438,7 @@ struct xfrm_userpolicy_info {
 #define XFRM_POLICY_LOCALOK	1	/* Allow user to override global policy */
 	/* Automatically expand selector to include matching ICMP payloads. */
 #define XFRM_POLICY_ICMP	2
+#define XFRM_POLICY_CPU_ACQUIRE	4
 	__u8				share;
 };
 
diff --git a/net/key/af_key.c b/net/key/af_key.c
index f79fb99271ed..c56bb4f451e6 100644
--- a/net/key/af_key.c
+++ b/net/key/af_key.c
@@ -1354,7 +1354,7 @@ static int pfkey_getspi(struct sock *sk, struct sk_buff *skb, const struct sadb_
 	}
 
 	if (hdr->sadb_msg_seq) {
-		x = xfrm_find_acq_byseq(net, DUMMY_MARK, hdr->sadb_msg_seq);
+		x = xfrm_find_acq_byseq(net, DUMMY_MARK, hdr->sadb_msg_seq, UINT_MAX);
 		if (x && !xfrm_addr_equal(&x->id.daddr, xdaddr, family)) {
 			xfrm_state_put(x);
 			x = NULL;
@@ -1362,7 +1362,8 @@ static int pfkey_getspi(struct sock *sk, struct sk_buff *skb, const struct sadb_
 	}
 
 	if (!x)
-		x = xfrm_find_acq(net, &dummy_mark, mode, reqid, 0, proto, xdaddr, xsaddr, 1, family);
+		x = xfrm_find_acq(net, &dummy_mark, mode, reqid, 0, UINT_MAX,
+				  proto, xdaddr, xsaddr, 1, family);
 
 	if (x == NULL)
 		return -ENOENT;
@@ -1417,7 +1418,7 @@ static int pfkey_acquire(struct sock *sk, struct sk_buff *skb, const struct sadb
 	if (hdr->sadb_msg_seq == 0 || hdr->sadb_msg_errno == 0)
 		return 0;
 
-	x = xfrm_find_acq_byseq(net, DUMMY_MARK, hdr->sadb_msg_seq);
+	x = xfrm_find_acq_byseq(net, DUMMY_MARK, hdr->sadb_msg_seq, UINT_MAX);
 	if (x == NULL)
 		return 0;
 
diff --git a/net/xfrm/xfrm_compat.c b/net/xfrm/xfrm_compat.c
index 91357ccaf4af..5b9ee63e30b6 100644
--- a/net/xfrm/xfrm_compat.c
+++ b/net/xfrm/xfrm_compat.c
@@ -132,6 +132,7 @@ static const struct nla_policy compat_policy[XFRMA_MAX+1] = {
 	[XFRMA_MTIMER_THRESH]	= { .type = NLA_U32 },
 	[XFRMA_SA_DIR]          = NLA_POLICY_RANGE(NLA_U8, XFRM_SA_DIR_IN, XFRM_SA_DIR_OUT),
 	[XFRMA_NAT_KEEPALIVE_INTERVAL]	= { .type = NLA_U32 },
+	[XFRMA_SA_PCPU]		= { .type = NLA_U32 },
 };
 
 static struct nlmsghdr *xfrm_nlmsg_put_compat(struct sk_buff *skb,
@@ -282,9 +283,10 @@ static int xfrm_xlate64_attr(struct sk_buff *dst, const struct nlattr *src)
 	case XFRMA_MTIMER_THRESH:
 	case XFRMA_SA_DIR:
 	case XFRMA_NAT_KEEPALIVE_INTERVAL:
+	case XFRMA_SA_PCPU:
 		return xfrm_nla_cpy(dst, src, nla_len(src));
 	default:
-		BUILD_BUG_ON(XFRMA_MAX != XFRMA_NAT_KEEPALIVE_INTERVAL);
+		BUILD_BUG_ON(XFRMA_MAX != XFRMA_SA_PCPU);
 		pr_warn_once("unsupported nla_type %d\n", src->nla_type);
 		return -EOPNOTSUPP;
 	}
@@ -439,7 +441,7 @@ static int xfrm_xlate32_attr(void *dst, const struct nlattr *nla,
 	int err;
 
 	if (type > XFRMA_MAX) {
-		BUILD_BUG_ON(XFRMA_MAX != XFRMA_NAT_KEEPALIVE_INTERVAL);
+		BUILD_BUG_ON(XFRMA_MAX != XFRMA_SA_PCPU);
 		NL_SET_ERR_MSG(extack, "Bad attribute");
 		return -EOPNOTSUPP;
 	}
diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c
index 37478d36a8df..ebef07b80afa 100644
--- a/net/xfrm/xfrm_state.c
+++ b/net/xfrm/xfrm_state.c
@@ -679,6 +679,7 @@ struct xfrm_state *xfrm_state_alloc(struct net *net)
 		x->lft.hard_packet_limit = XFRM_INF;
 		x->replay_maxage = 0;
 		x->replay_maxdiff = 0;
+		x->pcpu_num = UINT_MAX;
 		spin_lock_init(&x->lock);
 	}
 	return x;
@@ -1155,6 +1156,12 @@ static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 			       struct xfrm_state **best, int *acq_in_progress,
 			       int *error)
 {
+	/* We need the cpu id just as a lookup key,
+	 * we don't require it to be stable.
+	 */
+	unsigned int pcpu_id = get_cpu();
+	put_cpu();
+
 	/* Resolution logic:
 	 * 1. There is a valid state with matching selector. Done.
 	 * 2. Valid state with inappropriate selector. Skip.
@@ -1174,13 +1181,18 @@ static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 							&fl->u.__fl_common))
 			return;
 
+		if (x->pcpu_num != UINT_MAX && x->pcpu_num != pcpu_id)
+			return;
+
 		if (!*best ||
+		    ((*best)->pcpu_num == UINT_MAX && x->pcpu_num == pcpu_id) ||
 		    (*best)->km.dying > x->km.dying ||
 		    ((*best)->km.dying == x->km.dying &&
 		     (*best)->curlft.add_time < x->curlft.add_time))
 			*best = x;
 	} else if (x->km.state == XFRM_STATE_ACQ) {
-		*acq_in_progress = 1;
+		if (!*best || x->pcpu_num == pcpu_id)
+			*acq_in_progress = 1;
 	} else if (x->km.state == XFRM_STATE_ERROR ||
 		   x->km.state == XFRM_STATE_EXPIRED) {
 		if ((!x->sel.family ||
@@ -1209,6 +1221,13 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 	unsigned short encap_family = tmpl->encap_family;
 	unsigned int sequence;
 	struct km_event c;
+	unsigned int pcpu_id;
+
+	/* We need the cpu id just as a lookup key,
+	 * we don't require it to be stable.
+	 */
+	pcpu_id = get_cpu();
+	put_cpu();
 
 	to_put = NULL;
 
@@ -1282,7 +1301,10 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 	}
 
 found:
-	x = best;
+	if (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) ||
+	    (best && (best->pcpu_num == pcpu_id)))
+		x = best;
+
 	if (!x && !error && !acquire_in_progress) {
 		if (tmpl->id.spi &&
 		    (x0 = __xfrm_state_lookup_all(net, mark, daddr,
@@ -1314,6 +1336,8 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 		x->if_id = if_id;
+		if ((pol->flags & XFRM_POLICY_CPU_ACQUIRE) && best)
+			x->pcpu_num = pcpu_id;
 
 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 		if (error) {
@@ -1392,6 +1416,11 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 			x = NULL;
 			error = -ESRCH;
 		}
+
+		/* Use the already installed 'fallback' while the CPU-specific
+		 * SA acquire is handled*/
+		if (best)
+			x = best;
 	}
 out:
 	if (x) {
@@ -1524,12 +1553,14 @@ static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 	unsigned int h;
 	u32 mark = xnew->mark.v & xnew->mark.m;
 	u32 if_id = xnew->if_id;
+	u32 cpu_id = xnew->pcpu_num;
 
 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 		if (x->props.family	== family &&
 		    x->props.reqid	== reqid &&
 		    x->if_id		== if_id &&
+		    x->pcpu_num		== cpu_id &&
 		    (mark & x->mark.m) == x->mark.v &&
 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
@@ -1552,7 +1583,7 @@ EXPORT_SYMBOL(xfrm_state_insert);
 static struct xfrm_state *__find_acq_core(struct net *net,
 					  const struct xfrm_mark *m,
 					  unsigned short family, u8 mode,
-					  u32 reqid, u32 if_id, u8 proto,
+					  u32 reqid, u32 if_id, u32 pcpu_num, u8 proto,
 					  const xfrm_address_t *daddr,
 					  const xfrm_address_t *saddr,
 					  int create)
@@ -1569,6 +1600,7 @@ static struct xfrm_state *__find_acq_core(struct net *net,
 		    x->id.spi       != 0 ||
 		    x->id.proto	    != proto ||
 		    (mark & x->mark.m) != x->mark.v ||
+		    x->pcpu_num != pcpu_num ||
 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
 			continue;
@@ -1602,6 +1634,7 @@ static struct xfrm_state *__find_acq_core(struct net *net,
 			break;
 		}
 
+		x->pcpu_num = pcpu_num;
 		x->km.state = XFRM_STATE_ACQ;
 		x->id.proto = proto;
 		x->props.family = family;
@@ -1630,7 +1663,7 @@ static struct xfrm_state *__find_acq_core(struct net *net,
 	return x;
 }
 
-static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
+static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num);
 
 int xfrm_state_add(struct xfrm_state *x)
 {
@@ -1656,7 +1689,7 @@ int xfrm_state_add(struct xfrm_state *x)
 	}
 
 	if (use_spi && x->km.seq) {
-		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
+		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq, x->pcpu_num);
 		if (x1 && ((x1->id.proto != x->id.proto) ||
 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
 			to_put = x1;
@@ -1666,7 +1699,7 @@ int xfrm_state_add(struct xfrm_state *x)
 
 	if (use_spi && !x1)
 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
-				     x->props.reqid, x->if_id, x->id.proto,
+				     x->props.reqid, x->if_id, x->pcpu_num, x->id.proto,
 				     &x->id.daddr, &x->props.saddr, 0);
 
 	__xfrm_state_bump_genids(x);
@@ -1791,6 +1824,7 @@ static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
 	x->props.flags = orig->props.flags;
 	x->props.extra_flags = orig->props.extra_flags;
 
+	x->pcpu_num = orig->pcpu_num;
 	x->if_id = orig->if_id;
 	x->tfcpad = orig->tfcpad;
 	x->replay_maxdiff = orig->replay_maxdiff;
@@ -2066,13 +2100,14 @@ EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
 
 struct xfrm_state *
 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
-	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
+	      u32 if_id, u32 pcpu_num, u8 proto, const xfrm_address_t *daddr,
 	      const xfrm_address_t *saddr, int create, unsigned short family)
 {
 	struct xfrm_state *x;
 
 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
-	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
+	x = __find_acq_core(net, mark, family, mode, reqid, if_id, pcpu_num,
+			    proto, daddr, saddr, create);
 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 
 	return x;
@@ -2207,7 +2242,7 @@ xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
 
 /* Silly enough, but I'm lazy to build resolution list */
 
-static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
+static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
 {
 	unsigned int h = xfrm_seq_hash(net, seq);
 	struct xfrm_state *x;
@@ -2215,6 +2250,7 @@ static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 s
 	hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
 		if (x->km.seq == seq &&
 		    (mark & x->mark.m) == x->mark.v &&
+		    x->pcpu_num == pcpu_num &&
 		    x->km.state == XFRM_STATE_ACQ) {
 			xfrm_state_hold(x);
 			return x;
@@ -2224,12 +2260,12 @@ static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 s
 	return NULL;
 }
 
-struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
+struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
 {
 	struct xfrm_state *x;
 
 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
-	x = __xfrm_find_acq_byseq(net, mark, seq);
+	x = __xfrm_find_acq_byseq(net, mark, seq, pcpu_num);
 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 	return x;
 }
diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c
index e3b8ce89831a..e4d448950d05 100644
--- a/net/xfrm/xfrm_user.c
+++ b/net/xfrm/xfrm_user.c
@@ -460,6 +460,12 @@ static int verify_newsa_info(struct xfrm_usersa_info *p,
 		}
 	}
 
+	if (!sa_dir && attrs[XFRMA_SA_PCPU]) {
+		NL_SET_ERR_MSG(extack, "SA_PCPU only supported with SA_DIR");
+		err = -EINVAL;
+		goto out;
+	}
+
 out:
 	return err;
 }
@@ -841,6 +847,12 @@ static struct xfrm_state *xfrm_state_construct(struct net *net,
 		x->nat_keepalive_interval =
 			nla_get_u32(attrs[XFRMA_NAT_KEEPALIVE_INTERVAL]);
 
+	if (attrs[XFRMA_SA_PCPU]) {
+		x->pcpu_num = nla_get_u32(attrs[XFRMA_SA_PCPU]);
+		if (x->pcpu_num >= num_possible_cpus())
+			goto error;
+	}
+
 	err = __xfrm_init_state(x, false, attrs[XFRMA_OFFLOAD_DEV], extack);
 	if (err)
 		goto error;
@@ -1296,6 +1308,11 @@ static int copy_to_user_state_extra(struct xfrm_state *x,
 		if (ret)
 			goto out;
 	}
+	if (x->pcpu_num != UINT_MAX) {
+		ret = nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num);
+		if (ret)
+			goto out;
+	}
 	if (x->dir)
 		ret = nla_put_u8(skb, XFRMA_SA_DIR, x->dir);
 
@@ -1700,6 +1717,7 @@ static int xfrm_alloc_userspi(struct sk_buff *skb, struct nlmsghdr *nlh,
 	u32 mark;
 	struct xfrm_mark m;
 	u32 if_id = 0;
+	u32 pcpu_num = UINT_MAX;
 
 	p = nlmsg_data(nlh);
 	err = verify_spi_info(p->info.id.proto, p->min, p->max, extack);
@@ -1716,8 +1734,16 @@ static int xfrm_alloc_userspi(struct sk_buff *skb, struct nlmsghdr *nlh,
 	if (attrs[XFRMA_IF_ID])
 		if_id = nla_get_u32(attrs[XFRMA_IF_ID]);
 
+	if (attrs[XFRMA_SA_PCPU]) {
+		pcpu_num = nla_get_u32(attrs[XFRMA_SA_PCPU]);
+		if (pcpu_num >= num_possible_cpus()) {
+			err = -EINVAL;
+			goto out_noput;
+		}
+	}
+
 	if (p->info.seq) {
-		x = xfrm_find_acq_byseq(net, mark, p->info.seq);
+		x = xfrm_find_acq_byseq(net, mark, p->info.seq, pcpu_num);
 		if (x && !xfrm_addr_equal(&x->id.daddr, daddr, family)) {
 			xfrm_state_put(x);
 			x = NULL;
@@ -1726,7 +1752,7 @@ static int xfrm_alloc_userspi(struct sk_buff *skb, struct nlmsghdr *nlh,
 
 	if (!x)
 		x = xfrm_find_acq(net, &m, p->info.mode, p->info.reqid,
-				  if_id, p->info.id.proto, daddr,
+				  if_id, pcpu_num, p->info.id.proto, daddr,
 				  &p->info.saddr, 1,
 				  family);
 	err = -ENOENT;
@@ -2526,7 +2552,8 @@ static inline unsigned int xfrm_aevent_msgsize(struct xfrm_state *x)
 	       + nla_total_size(sizeof(struct xfrm_mark))
 	       + nla_total_size(4) /* XFRM_AE_RTHR */
 	       + nla_total_size(4) /* XFRM_AE_ETHR */
-	       + nla_total_size(sizeof(x->dir)); /* XFRMA_SA_DIR */
+	       + nla_total_size(sizeof(x->dir)) /* XFRMA_SA_DIR */
+	       + nla_total_size(4); /* XFRMA_SA_PCPU */
 }
 
 static int build_aevent(struct sk_buff *skb, struct xfrm_state *x, const struct km_event *c)
@@ -2582,6 +2609,8 @@ static int build_aevent(struct sk_buff *skb, struct xfrm_state *x, const struct
 	err = xfrm_if_id_put(skb, x->if_id);
 	if (err)
 		goto out_cancel;
+	if (x->pcpu_num != UINT_MAX)
+		err = nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num);
 
 	if (x->dir) {
 		err = nla_put_u8(skb, XFRMA_SA_DIR, x->dir);
@@ -2852,6 +2881,13 @@ static int xfrm_add_acquire(struct sk_buff *skb, struct nlmsghdr *nlh,
 
 	xfrm_mark_get(attrs, &mark);
 
+	if (attrs[XFRMA_SA_PCPU]) {
+		x->pcpu_num = nla_get_u32(attrs[XFRMA_SA_PCPU]);
+		err = -EINVAL;
+		if (x->pcpu_num >= num_possible_cpus())
+			goto free_state;
+	}
+
 	err = verify_newpolicy_info(&ua->policy, extack);
 	if (err)
 		goto free_state;
@@ -3182,6 +3218,7 @@ const struct nla_policy xfrma_policy[XFRMA_MAX+1] = {
 	[XFRMA_MTIMER_THRESH]   = { .type = NLA_U32 },
 	[XFRMA_SA_DIR]          = NLA_POLICY_RANGE(NLA_U8, XFRM_SA_DIR_IN, XFRM_SA_DIR_OUT),
 	[XFRMA_NAT_KEEPALIVE_INTERVAL] = { .type = NLA_U32 },
+	[XFRMA_SA_PCPU]		= { .type = NLA_U32 },
 };
 EXPORT_SYMBOL_GPL(xfrma_policy);
 
@@ -3348,7 +3385,8 @@ static inline unsigned int xfrm_expire_msgsize(void)
 {
 	return NLMSG_ALIGN(sizeof(struct xfrm_user_expire)) +
 	       nla_total_size(sizeof(struct xfrm_mark)) +
-	       nla_total_size(sizeof_field(struct xfrm_state, dir));
+	       nla_total_size(sizeof_field(struct xfrm_state, dir)) +
+	       nla_total_size(4); /* XFRMA_SA_PCPU */
 }
 
 static int build_expire(struct sk_buff *skb, struct xfrm_state *x, const struct km_event *c)
@@ -3374,6 +3412,11 @@ static int build_expire(struct sk_buff *skb, struct xfrm_state *x, const struct
 	err = xfrm_if_id_put(skb, x->if_id);
 	if (err)
 		return err;
+	if (x->pcpu_num != UINT_MAX) {
+		err = nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num);
+		if (err)
+			return err;
+	}
 
 	if (x->dir) {
 		err = nla_put_u8(skb, XFRMA_SA_DIR, x->dir);
@@ -3481,6 +3524,8 @@ static inline unsigned int xfrm_sa_len(struct xfrm_state *x)
 	}
 	if (x->if_id)
 		l += nla_total_size(sizeof(x->if_id));
+	if (x->pcpu_num)
+		l += nla_total_size(sizeof(x->pcpu_num));
 
 	/* Must count x->lastused as it may become non-zero behind our back. */
 	l += nla_total_size_64bit(sizeof(u64));
@@ -3587,6 +3632,7 @@ static inline unsigned int xfrm_acquire_msgsize(struct xfrm_state *x,
 	       + nla_total_size(sizeof(struct xfrm_user_tmpl) * xp->xfrm_nr)
 	       + nla_total_size(sizeof(struct xfrm_mark))
 	       + nla_total_size(xfrm_user_sec_ctx_size(x->security))
+	       + nla_total_size(4) /* XFRMA_SA_PCPU */
 	       + userpolicy_type_attrsize();
 }
 
@@ -3623,6 +3669,8 @@ static int build_acquire(struct sk_buff *skb, struct xfrm_state *x,
 		err = xfrm_if_id_put(skb, xp->if_id);
 	if (!err && xp->xdo.dev)
 		err = copy_user_offload(&xp->xdo, skb);
+	if (!err && x->pcpu_num != UINT_MAX)
+		err = nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num);
 	if (err) {
 		nlmsg_cancel(skb, nlh);
 		return err;
-- 
2.34.1

[PATCH 02/11] xfrm: Cache used outbound xfrm states at the policy.

[Steffen Klassert]

Now that we can have percpu xfrm states, the number of active
states might increase. To get a better lookup performance,
we cache the used xfrm states at the policy for outbound
IPsec traffic.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Tested-by: Antony Antony <antony.antony@secunet.com>
Tested-by: Tobias Brunner <tobias@strongswan.org>
---
 include/net/xfrm.h     |  4 +++
 net/xfrm/xfrm_policy.c | 12 +++++++++
 net/xfrm/xfrm_state.c  | 55 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 71 insertions(+)

diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index f5275618e744..0b394c5fb5f3 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -184,6 +184,7 @@ struct xfrm_state {
 	};
 	struct hlist_node	byspi;
 	struct hlist_node	byseq;
+	struct hlist_node	state_cache;
 
 	refcount_t		refcnt;
 	spinlock_t		lock;
@@ -537,6 +538,7 @@ struct xfrm_policy_queue {
  *	@xp_net: network namespace the policy lives in
  *	@bydst: hlist node for SPD hash table or rbtree list
  *	@byidx: hlist node for index hash table
+ *	@state_cache_list: hlist head for policy cached xfrm states
  *	@lock: serialize changes to policy structure members
  *	@refcnt: reference count, freed once it reaches 0
  *	@pos: kernel internal tie-breaker to determine age of policy
@@ -567,6 +569,8 @@ struct xfrm_policy {
 	struct hlist_node	bydst;
 	struct hlist_node	byidx;
 
+	struct hlist_head	state_cache_list;
+
 	/* This lock only affects elements except for entry. */
 	rwlock_t		lock;
 	refcount_t		refcnt;
diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index a2ea9dbac90b..8a1b83191a6c 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -434,6 +434,7 @@ struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
 	if (policy) {
 		write_pnet(&policy->xp_net, net);
 		INIT_LIST_HEAD(&policy->walk.all);
+		INIT_HLIST_HEAD(&policy->state_cache_list);
 		INIT_HLIST_NODE(&policy->bydst);
 		INIT_HLIST_NODE(&policy->byidx);
 		rwlock_init(&policy->lock);
@@ -475,6 +476,9 @@ EXPORT_SYMBOL(xfrm_policy_destroy);
 
 static void xfrm_policy_kill(struct xfrm_policy *policy)
 {
+	struct net *net = xp_net(policy);
+	struct xfrm_state *x;
+
 	xfrm_dev_policy_delete(policy);
 
 	write_lock_bh(&policy->lock);
@@ -490,6 +494,13 @@ static void xfrm_policy_kill(struct xfrm_policy *policy)
 	if (del_timer(&policy->timer))
 		xfrm_pol_put(policy);
 
+	/* XXX: Flush state cache */
+	spin_lock_bh(&net->xfrm.xfrm_state_lock);
+	hlist_for_each_entry_rcu(x, &policy->state_cache_list, state_cache) {
+		hlist_del_init_rcu(&x->state_cache);
+	}
+	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
+
 	xfrm_pol_put(policy);
 }
 
@@ -3275,6 +3286,7 @@ struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
 		dst_release(dst);
 		dst = dst_orig;
 	}
+
 ok:
 	xfrm_pols_put(pols, drop_pols);
 	if (dst && dst->xfrm &&
diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c
index ebef07b80afa..a2047825f6c8 100644
--- a/net/xfrm/xfrm_state.c
+++ b/net/xfrm/xfrm_state.c
@@ -665,6 +665,7 @@ struct xfrm_state *xfrm_state_alloc(struct net *net)
 		refcount_set(&x->refcnt, 1);
 		atomic_set(&x->tunnel_users, 0);
 		INIT_LIST_HEAD(&x->km.all);
+		INIT_HLIST_NODE(&x->state_cache);
 		INIT_HLIST_NODE(&x->bydst);
 		INIT_HLIST_NODE(&x->bysrc);
 		INIT_HLIST_NODE(&x->byspi);
@@ -744,12 +745,15 @@ int __xfrm_state_delete(struct xfrm_state *x)
 
 	if (x->km.state != XFRM_STATE_DEAD) {
 		x->km.state = XFRM_STATE_DEAD;
+
 		spin_lock(&net->xfrm.xfrm_state_lock);
 		list_del(&x->km.all);
 		hlist_del_rcu(&x->bydst);
 		hlist_del_rcu(&x->bysrc);
 		if (x->km.seq)
 			hlist_del_rcu(&x->byseq);
+		if (!hlist_unhashed(&x->state_cache))
+			hlist_del_rcu(&x->state_cache);
 		if (x->id.spi)
 			hlist_del_rcu(&x->byspi);
 		net->xfrm.state_num--;
@@ -1222,6 +1226,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 	unsigned int sequence;
 	struct km_event c;
 	unsigned int pcpu_id;
+	bool cached = false;
 
 	/* We need the cpu id just as a lookup key,
 	 * we don't require it to be stable.
@@ -1234,6 +1239,46 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 	sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
 
 	rcu_read_lock();
+	hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
+		if (x->props.family == encap_family &&
+		    x->props.reqid == tmpl->reqid &&
+		    (mark & x->mark.m) == x->mark.v &&
+		    x->if_id == if_id &&
+		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
+		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
+		    tmpl->mode == x->props.mode &&
+		    tmpl->id.proto == x->id.proto &&
+		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
+			xfrm_state_look_at(pol, x, fl, encap_family,
+					   &best, &acquire_in_progress, &error);
+	}
+
+	if (best)
+		goto cached;
+
+	hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
+		if (x->props.family == encap_family &&
+		    x->props.reqid == tmpl->reqid &&
+		    (mark & x->mark.m) == x->mark.v &&
+		    x->if_id == if_id &&
+		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
+		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
+		    tmpl->mode == x->props.mode &&
+		    tmpl->id.proto == x->id.proto &&
+		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
+			xfrm_state_look_at(pol, x, fl, family,
+					   &best, &acquire_in_progress, &error);
+	}
+
+cached:
+	cached = true;
+	if (best)
+		goto found;
+	else if (error)
+		best = NULL;
+	else if (acquire_in_progress) /* XXX: acquire_in_progress should not happen */
+		WARN_ON(1);
+
 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
 #ifdef CONFIG_XFRM_OFFLOAD
@@ -1383,6 +1428,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 			XFRM_STATE_INSERT(bysrc, &x->bysrc,
 					  net->xfrm.state_bysrc + h,
 					  x->xso.type);
+			INIT_HLIST_NODE(&x->state_cache);
 			if (x->id.spi) {
 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 				XFRM_STATE_INSERT(byspi, &x->byspi,
@@ -1431,6 +1477,15 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 	} else {
 		*err = acquire_in_progress ? -EAGAIN : error;
 	}
+
+	if (x && x->km.state == XFRM_STATE_VALID && !cached &&
+	    (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) || x->pcpu_num == pcpu_id)) {
+		spin_lock_bh(&net->xfrm.xfrm_state_lock);
+		if (hlist_unhashed(&x->state_cache))
+			hlist_add_head_rcu(&x->state_cache, &pol->state_cache_list);
+		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
+	}
+
 	rcu_read_unlock();
 	if (to_put)
 		xfrm_state_put(to_put);
-- 
2.34.1

[PATCH 03/11] xfrm: Add an inbound percpu state cache.

[Steffen Klassert]

Now that we can have percpu xfrm states, the number of active
states might increase. To get a better lookup performance,
we add a percpu cache to cache the used inbound xfrm states.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Tested-by: Antony Antony <antony.antony@secunet.com>
Tested-by: Tobias Brunner <tobias@strongswan.org>
---
 include/net/netns/xfrm.h |  1 +
 include/net/xfrm.h       |  5 ++++
 net/ipv4/esp4_offload.c  |  6 ++---
 net/ipv6/esp6_offload.c  |  6 ++---
 net/xfrm/xfrm_input.c    |  2 +-
 net/xfrm/xfrm_state.c    | 57 ++++++++++++++++++++++++++++++++++++++++
 6 files changed, 70 insertions(+), 7 deletions(-)

diff --git a/include/net/netns/xfrm.h b/include/net/netns/xfrm.h
index ae60d6664095..23dd647fe024 100644
--- a/include/net/netns/xfrm.h
+++ b/include/net/netns/xfrm.h
@@ -43,6 +43,7 @@ struct netns_xfrm {
 	struct hlist_head	__rcu *state_bysrc;
 	struct hlist_head	__rcu *state_byspi;
 	struct hlist_head	__rcu *state_byseq;
+	struct hlist_head	 __percpu *state_cache_input;
 	unsigned int		state_hmask;
 	unsigned int		state_num;
 	struct work_struct	state_hash_work;
diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index 0b394c5fb5f3..2b87999bd5aa 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -185,6 +185,7 @@ struct xfrm_state {
 	struct hlist_node	byspi;
 	struct hlist_node	byseq;
 	struct hlist_node	state_cache;
+	struct hlist_node	state_cache_input;
 
 	refcount_t		refcnt;
 	spinlock_t		lock;
@@ -1650,6 +1651,10 @@ int xfrm_state_update(struct xfrm_state *x);
 struct xfrm_state *xfrm_state_lookup(struct net *net, u32 mark,
 				     const xfrm_address_t *daddr, __be32 spi,
 				     u8 proto, unsigned short family);
+struct xfrm_state *xfrm_input_state_lookup(struct net *net, u32 mark,
+					   const xfrm_address_t *daddr,
+					   __be32 spi, u8 proto,
+					   unsigned short family);
 struct xfrm_state *xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 					    const xfrm_address_t *daddr,
 					    const xfrm_address_t *saddr,
diff --git a/net/ipv4/esp4_offload.c b/net/ipv4/esp4_offload.c
index 80c4ea0e12f4..e0d94270da28 100644
--- a/net/ipv4/esp4_offload.c
+++ b/net/ipv4/esp4_offload.c
@@ -53,9 +53,9 @@ static struct sk_buff *esp4_gro_receive(struct list_head *head,
 		if (sp->len == XFRM_MAX_DEPTH)
 			goto out_reset;
 
-		x = xfrm_state_lookup(dev_net(skb->dev), skb->mark,
-				      (xfrm_address_t *)&ip_hdr(skb)->daddr,
-				      spi, IPPROTO_ESP, AF_INET);
+		x = xfrm_input_state_lookup(dev_net(skb->dev), skb->mark,
+					    (xfrm_address_t *)&ip_hdr(skb)->daddr,
+					    spi, IPPROTO_ESP, AF_INET);
 
 		if (unlikely(x && x->dir && x->dir != XFRM_SA_DIR_IN)) {
 			/* non-offload path will record the error and audit log */
diff --git a/net/ipv6/esp6_offload.c b/net/ipv6/esp6_offload.c
index 919ebfabbe4e..7b41fb4f00b5 100644
--- a/net/ipv6/esp6_offload.c
+++ b/net/ipv6/esp6_offload.c
@@ -80,9 +80,9 @@ static struct sk_buff *esp6_gro_receive(struct list_head *head,
 		if (sp->len == XFRM_MAX_DEPTH)
 			goto out_reset;
 
-		x = xfrm_state_lookup(dev_net(skb->dev), skb->mark,
-				      (xfrm_address_t *)&ipv6_hdr(skb)->daddr,
-				      spi, IPPROTO_ESP, AF_INET6);
+		x = xfrm_input_state_lookup(dev_net(skb->dev), skb->mark,
+					    (xfrm_address_t *)&ipv6_hdr(skb)->daddr,
+					    spi, IPPROTO_ESP, AF_INET6);
 
 		if (unlikely(x && x->dir && x->dir != XFRM_SA_DIR_IN)) {
 			/* non-offload path will record the error and audit log */
diff --git a/net/xfrm/xfrm_input.c b/net/xfrm/xfrm_input.c
index 749e7eea99e4..841a60a6fbfe 100644
--- a/net/xfrm/xfrm_input.c
+++ b/net/xfrm/xfrm_input.c
@@ -572,7 +572,7 @@ int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
 			goto drop;
 		}
 
-		x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
+		x = xfrm_input_state_lookup(net, mark, daddr, spi, nexthdr, family);
 		if (x == NULL) {
 			secpath_reset(skb);
 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c
index a2047825f6c8..e3266a5d4f90 100644
--- a/net/xfrm/xfrm_state.c
+++ b/net/xfrm/xfrm_state.c
@@ -754,6 +754,9 @@ int __xfrm_state_delete(struct xfrm_state *x)
 			hlist_del_rcu(&x->byseq);
 		if (!hlist_unhashed(&x->state_cache))
 			hlist_del_rcu(&x->state_cache);
+		if (!hlist_unhashed(&x->state_cache_input))
+			hlist_del_rcu(&x->state_cache_input);
+
 		if (x->id.spi)
 			hlist_del_rcu(&x->byspi);
 		net->xfrm.state_num--;
@@ -1106,6 +1109,52 @@ static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 	return NULL;
 }
 
+struct xfrm_state *xfrm_input_state_lookup(struct net *net, u32 mark,
+					   const xfrm_address_t *daddr,
+					   __be32 spi, u8 proto,
+					   unsigned short family)
+{
+	struct hlist_head *state_cache_input;
+	struct xfrm_state *x = NULL;
+	int cpu = get_cpu();
+
+	state_cache_input =  per_cpu_ptr(net->xfrm.state_cache_input, cpu);
+
+	rcu_read_lock();
+	hlist_for_each_entry_rcu(x, state_cache_input, state_cache_input) {
+		if (x->props.family != family ||
+		    x->id.spi       != spi ||
+		    x->id.proto     != proto ||
+		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
+			continue;
+
+		if ((mark & x->mark.m) != x->mark.v)
+			continue;
+		if (!xfrm_state_hold_rcu(x))
+			continue;
+		goto out;
+	}
+
+	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
+
+	if (x && x->km.state == XFRM_STATE_VALID) {
+		spin_lock_bh(&net->xfrm.xfrm_state_lock);
+		if (hlist_unhashed(&x->state_cache_input)) {
+			hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
+		} else {
+			hlist_del_rcu(&x->state_cache_input);
+			hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
+		}
+		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
+	}
+
+out:
+	rcu_read_unlock();
+	put_cpu();
+	return x;
+}
+EXPORT_SYMBOL(xfrm_input_state_lookup);
+
 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 						     const xfrm_address_t *daddr,
 						     const xfrm_address_t *saddr,
@@ -3079,6 +3128,11 @@ int __net_init xfrm_state_init(struct net *net)
 	net->xfrm.state_byseq = xfrm_hash_alloc(sz);
 	if (!net->xfrm.state_byseq)
 		goto out_byseq;
+
+	net->xfrm.state_cache_input = alloc_percpu(struct hlist_head);
+	if (!net->xfrm.state_cache_input)
+		goto out_state_cache_input;
+
 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
 
 	net->xfrm.state_num = 0;
@@ -3088,6 +3142,8 @@ int __net_init xfrm_state_init(struct net *net)
 			       &net->xfrm.xfrm_state_lock);
 	return 0;
 
+out_state_cache_input:
+	xfrm_hash_free(net->xfrm.state_byseq, sz);
 out_byseq:
 	xfrm_hash_free(net->xfrm.state_byspi, sz);
 out_byspi:
@@ -3117,6 +3173,7 @@ void xfrm_state_fini(struct net *net)
 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
 	xfrm_hash_free(net->xfrm.state_bydst, sz);
+	free_percpu(net->xfrm.state_cache_input);
 }
 
 #ifdef CONFIG_AUDITSYSCALL
-- 
2.34.1

[PATCH 04/11] xfrm: Restrict percpu SA attribute to specific netlink message types

[Steffen Klassert]

Reject the usage of XFRMA_SA_PCPU in xfrm netlink messages when
it's not applicable.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Tested-by: Antony Antony <antony.antony@secunet.com>
Tested-by: Tobias Brunner <tobias@strongswan.org>
---
 net/xfrm/xfrm_user.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c
index e4d448950d05..b6ce2b3c6b87 100644
--- a/net/xfrm/xfrm_user.c
+++ b/net/xfrm/xfrm_user.c
@@ -3282,6 +3282,20 @@ static int xfrm_reject_unused_attr(int type, struct nlattr **attrs,
 		}
 	}
 
+	if (attrs[XFRMA_SA_PCPU]) {
+		switch (type) {
+		case XFRM_MSG_NEWSA:
+		case XFRM_MSG_UPDSA:
+		case XFRM_MSG_ALLOCSPI:
+		case XFRM_MSG_ACQUIRE:
+
+			break;
+		default:
+			NL_SET_ERR_MSG(extack, "Invalid attribute SA_PCPU");
+			return -EINVAL;
+		}
+	}
+
 	return 0;
 }
 
-- 
2.34.1

[PATCH 05/11] xfrm: Convert xfrm_get_tos() to dscp_t.

[Steffen Klassert]

From: Guillaume Nault <gnault@redhat.com>

Return a dscp_t variable to prepare for the future conversion of
xfrm_bundle_create() to dscp_t.

While there, rename the function "xfrm_get_dscp", to align its name
with the new return type.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_policy.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index 8a1b83191a6c..51a071a79016 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -2587,10 +2587,10 @@ xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
 
 }
 
-static int xfrm_get_tos(const struct flowi *fl, int family)
+static dscp_t xfrm_get_dscp(const struct flowi *fl, int family)
 {
 	if (family == AF_INET)
-		return fl->u.ip4.flowi4_tos & INET_DSCP_MASK;
+		return inet_dsfield_to_dscp(fl->u.ip4.flowi4_tos);
 
 	return 0;
 }
@@ -2684,7 +2684,7 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
 
 	xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
 
-	tos = xfrm_get_tos(fl, family);
+	tos = inet_dscp_to_dsfield(xfrm_get_dscp(fl, family));
 
 	dst_hold(dst);
 
-- 
2.34.1

[PATCH 06/11] xfrm: Convert xfrm_bundle_create() to dscp_t.

[Steffen Klassert]

From: Guillaume Nault <gnault@redhat.com>

Use a dscp_t variable to store the result of xfrm_get_dscp().
This prepares for the future conversion of xfrm_dst_lookup().

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_policy.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index 51a071a79016..ecb989347bd4 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -2678,13 +2678,13 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
 	int header_len = 0;
 	int nfheader_len = 0;
 	int trailer_len = 0;
-	int tos;
 	int family = policy->selector.family;
 	xfrm_address_t saddr, daddr;
+	dscp_t dscp;
 
 	xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
 
-	tos = inet_dscp_to_dsfield(xfrm_get_dscp(fl, family));
+	dscp = xfrm_get_dscp(fl, family);
 
 	dst_hold(dst);
 
@@ -2732,7 +2732,8 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
 				family = xfrm[i]->props.family;
 
 			oif = fl->flowi_oif ? : fl->flowi_l3mdev;
-			dst = xfrm_dst_lookup(xfrm[i], tos, oif,
+			dst = xfrm_dst_lookup(xfrm[i],
+					      inet_dscp_to_dsfield(dscp), oif,
 					      &saddr, &daddr, family, mark);
 			err = PTR_ERR(dst);
 			if (IS_ERR(dst))
-- 
2.34.1

[PATCH 07/11] xfrm: Convert xfrm_dst_lookup() to dscp_t.

[Steffen Klassert]

From: Guillaume Nault <gnault@redhat.com>

Pass a dscp_t variable to xfrm_dst_lookup(), instead of an int, to
prevent accidental setting of ECN bits in ->flowi4_tos.

Only xfrm_bundle_create() actually calls xfrm_dst_lookup(). Since it
already has a dscp_t variable to pass as parameter, we only need to
remove the inet_dscp_to_dsfield() conversion.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_policy.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index ecb989347bd4..7e3e10fb9ca0 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -289,7 +289,7 @@ struct dst_entry *__xfrm_dst_lookup(int family,
 EXPORT_SYMBOL(__xfrm_dst_lookup);
 
 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
-						int tos, int oif,
+						dscp_t dscp, int oif,
 						xfrm_address_t *prev_saddr,
 						xfrm_address_t *prev_daddr,
 						int family, u32 mark)
@@ -312,7 +312,7 @@ static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
 	params.net = net;
 	params.saddr = saddr;
 	params.daddr = daddr;
-	params.tos = tos;
+	params.tos = inet_dscp_to_dsfield(dscp);
 	params.oif = oif;
 	params.mark = mark;
 	params.ipproto = x->id.proto;
@@ -2732,9 +2732,8 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
 				family = xfrm[i]->props.family;
 
 			oif = fl->flowi_oif ? : fl->flowi_l3mdev;
-			dst = xfrm_dst_lookup(xfrm[i],
-					      inet_dscp_to_dsfield(dscp), oif,
-					      &saddr, &daddr, family, mark);
+			dst = xfrm_dst_lookup(xfrm[i], dscp, oif, &saddr,
+					      &daddr, family, mark);
 			err = PTR_ERR(dst);
 			if (IS_ERR(dst))
 				goto put_states;
-- 
2.34.1

[PATCH 08/11] xfrm: Convert struct xfrm_dst_lookup_params -> tos to dscp_t.

[Steffen Klassert]

From: Guillaume Nault <gnault@redhat.com>

Add type annotation to the "tos" field of struct xfrm_dst_lookup_params,
to ensure that the ECN bits aren't mistakenly taken into account when
doing route lookups. Rename that field (tos -> dscp) to make that
change explicit.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 include/net/xfrm.h      | 3 ++-
 net/ipv4/xfrm4_policy.c | 3 ++-
 net/xfrm/xfrm_policy.c  | 2 +-
 3 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index 2b87999bd5aa..32c09e85a64c 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -19,6 +19,7 @@
 
 #include <net/sock.h>
 #include <net/dst.h>
+#include <net/inet_dscp.h>
 #include <net/ip.h>
 #include <net/route.h>
 #include <net/ipv6.h>
@@ -354,7 +355,7 @@ void xfrm_if_unregister_cb(void);
 
 struct xfrm_dst_lookup_params {
 	struct net *net;
-	int tos;
+	dscp_t dscp;
 	int oif;
 	xfrm_address_t *saddr;
 	xfrm_address_t *daddr;
diff --git a/net/ipv4/xfrm4_policy.c b/net/ipv4/xfrm4_policy.c
index 7e1c2faed1ff..7fb6205619e7 100644
--- a/net/ipv4/xfrm4_policy.c
+++ b/net/ipv4/xfrm4_policy.c
@@ -14,6 +14,7 @@
 #include <linux/inetdevice.h>
 #include <net/dst.h>
 #include <net/xfrm.h>
+#include <net/inet_dscp.h>
 #include <net/ip.h>
 #include <net/l3mdev.h>
 
@@ -24,7 +25,7 @@ static struct dst_entry *__xfrm4_dst_lookup(struct flowi4 *fl4,
 
 	memset(fl4, 0, sizeof(*fl4));
 	fl4->daddr = params->daddr->a4;
-	fl4->flowi4_tos = params->tos;
+	fl4->flowi4_tos = inet_dscp_to_dsfield(params->dscp);
 	fl4->flowi4_l3mdev = l3mdev_master_ifindex_by_index(params->net,
 							    params->oif);
 	fl4->flowi4_mark = params->mark;
diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index 7e3e10fb9ca0..4408c11c0835 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -312,7 +312,7 @@ static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
 	params.net = net;
 	params.saddr = saddr;
 	params.daddr = daddr;
-	params.tos = inet_dscp_to_dsfield(dscp);
+	params.dscp = dscp;
 	params.oif = oif;
 	params.mark = mark;
 	params.ipproto = x->id.proto;
-- 
2.34.1

[PATCH 09/11] xfrm: Add error handling when nla_put_u32() returns an error

[Steffen Klassert]

From: "Everest K.C" <everestkc@everestkc.com.np>

Error handling is missing when call to nla_put_u32() fails.
Handle the error when the call to nla_put_u32() returns an error.

The error was reported by Coverity Scan.
Report:
CID 1601525: (#1 of 1): Unused value (UNUSED_VALUE)
returned_value: Assigning value from nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num)
to err here, but that stored value is overwritten before it can be used

Fixes: 1ddf9916ac09 ("xfrm: Add support for per cpu xfrm state handling.")
Signed-off-by: Everest K.C. <everestkc@everestkc.com.np>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_user.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c
index b6ce2b3c6b87..fab18b85af53 100644
--- a/net/xfrm/xfrm_user.c
+++ b/net/xfrm/xfrm_user.c
@@ -2609,8 +2609,11 @@ static int build_aevent(struct sk_buff *skb, struct xfrm_state *x, const struct
 	err = xfrm_if_id_put(skb, x->if_id);
 	if (err)
 		goto out_cancel;
-	if (x->pcpu_num != UINT_MAX)
+	if (x->pcpu_num != UINT_MAX) {
 		err = nla_put_u32(skb, XFRMA_SA_PCPU, x->pcpu_num);
+		if (err)
+			goto out_cancel;
+	}
 
 	if (x->dir) {
 		err = nla_put_u8(skb, XFRMA_SA_DIR, x->dir);
-- 
2.34.1

[PATCH 10/11] xfrm: replace deprecated strncpy with strscpy_pad

[Steffen Klassert]

From: Daniel Yang <danielyangkang@gmail.com>

The function strncpy is deprecated since it does not guarantee the
destination buffer is NULL terminated. Recommended replacement is
strscpy. The padded version was used to remain consistent with the other
strscpy_pad usage in the modified function.

Signed-off-by: Daniel Yang <danielyangkang@gmail.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_user.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c
index fab18b85af53..6b0800c7c75e 100644
--- a/net/xfrm/xfrm_user.c
+++ b/net/xfrm/xfrm_user.c
@@ -1101,7 +1101,7 @@ static int copy_to_user_auth(struct xfrm_algo_auth *auth, struct sk_buff *skb)
 	if (!nla)
 		return -EMSGSIZE;
 	algo = nla_data(nla);
-	strncpy(algo->alg_name, auth->alg_name, sizeof(algo->alg_name));
+	strscpy_pad(algo->alg_name, auth->alg_name, sizeof(algo->alg_name));
 
 	if (redact_secret && auth->alg_key_len)
 		memset(algo->alg_key, 0, (auth->alg_key_len + 7) / 8);
-- 
2.34.1

[PATCH 11/11] xfrm: Fix acquire state insertion.

[Steffen Klassert]

A recent commit jumped over the dst hash computation and
left the symbol uninitialized. Fix this by explicitly
computing the dst hash before it is used.

Fixes: 0045e3d80613 ("xfrm: Cache used outbound xfrm states at the policy.")
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/xfrm/xfrm_state.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c
index e3266a5d4f90..67ca7ac955a3 100644
--- a/net/xfrm/xfrm_state.c
+++ b/net/xfrm/xfrm_state.c
@@ -1470,6 +1470,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 			x->km.state = XFRM_STATE_ACQ;
 			x->dir = XFRM_SA_DIR_OUT;
 			list_add(&x->km.all, &net->xfrm.state_all);
+			h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 			XFRM_STATE_INSERT(bydst, &x->bydst,
 					  net->xfrm.state_bydst + h,
 					  x->xso.type);
-- 
2.34.1

Re: [PATCH 01/11] xfrm: Add support for per cpu xfrm state handling.

[Jakub Kicinski]

On Fri, 15 Nov 2024 09:33:33 +0100 Steffen Klassert wrote:
> +	/* We need the cpu id just as a lookup key,
> +	 * we don't require it to be stable.
> +	 */
> +	pcpu_id = get_cpu();
> +	put_cpu();

Why not smp_processor_id() ?

> +	if (attrs[XFRMA_SA_PCPU]) {
> +		x->pcpu_num = nla_get_u32(attrs[XFRMA_SA_PCPU]);
> +		if (x->pcpu_num >= num_possible_cpus())
> +			goto error;
> +	}

cpu ids can be sparse, shouldn't it be checking if the CPU is online ?

Re: [PATCH 01/11] xfrm: Add support for per cpu xfrm state handling.

[Steffen Klassert]

On Fri, Nov 15, 2024 at 06:09:08PM -0800, Jakub Kicinski wrote:
> On Fri, 15 Nov 2024 09:33:33 +0100 Steffen Klassert wrote:
> > +	/* We need the cpu id just as a lookup key,
> > +	 * we don't require it to be stable.
> > +	 */
> > +	pcpu_id = get_cpu();
> > +	put_cpu();
> 
> Why not smp_processor_id() ?

This might be executed in preemptable code sections,
smp_processor_id() will throw a warning if that happens.
Maybe raw_smp_processor_id() can be used instead here,
but was not sure if that's the right thing.

> 
> > +	if (attrs[XFRMA_SA_PCPU]) {
> > +		x->pcpu_num = nla_get_u32(attrs[XFRMA_SA_PCPU]);
> > +		if (x->pcpu_num >= num_possible_cpus())
> > +			goto error;
> > +	}
> 
> cpu ids can be sparse, shouldn't it be checking if the CPU is online ?

I thought about that. But then we must wait for an IKE negotiation
before we can use a fresh booted cpu. If we pre-negotiate a SA,
the cpu can be used right away. I depends a bit on the usecase what's
preferred here.

Re: [PATCH 01/11] xfrm: Add support for per cpu xfrm state handling.

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Steffen Klassert <steffen.klassert@secunet.com>:

On Fri, 15 Nov 2024 09:33:33 +0100 you wrote:
> Currently all flows for a certain SA must be processed by the same
> cpu to avoid packet reordering and lock contention of the xfrm
> state lock.
> 
> To get rid of this limitation, the IETF standardized per cpu SAs
> in RFC 9611. This patch implements the xfrm part of it.
> 
> [...]

Here is the summary with links:
  - [01/11] xfrm: Add support for per cpu xfrm state handling.
    https://git.kernel.org/netdev/net-next/c/1ddf9916ac09
  - [02/11] xfrm: Cache used outbound xfrm states at the policy.
    https://git.kernel.org/netdev/net-next/c/0045e3d80613
  - [03/11] xfrm: Add an inbound percpu state cache.
    https://git.kernel.org/netdev/net-next/c/81a331a0e72d
  - [04/11] xfrm: Restrict percpu SA attribute to specific netlink message types
    https://git.kernel.org/netdev/net-next/c/83dfce38c49f
  - [05/11] xfrm: Convert xfrm_get_tos() to dscp_t.
    https://git.kernel.org/netdev/net-next/c/766f532089af
  - [06/11] xfrm: Convert xfrm_bundle_create() to dscp_t.
    https://git.kernel.org/netdev/net-next/c/01f61cbfc8b2
  - [07/11] xfrm: Convert xfrm_dst_lookup() to dscp_t.
    https://git.kernel.org/netdev/net-next/c/3021a2a3403d
  - [08/11] xfrm: Convert struct xfrm_dst_lookup_params -> tos to dscp_t.
    https://git.kernel.org/netdev/net-next/c/e57dfaa4b0a7
  - [09/11] xfrm: Add error handling when nla_put_u32() returns an error
    https://git.kernel.org/netdev/net-next/c/9d287e70c51f
  - [10/11] xfrm: replace deprecated strncpy with strscpy_pad
    https://git.kernel.org/netdev/net-next/c/9e1a6db68e3c
  - [11/11] xfrm: Fix acquire state insertion.
    https://git.kernel.org/netdev/net-next/c/a35672819f8d

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html