pull request (net-next): ipsec-next 2017-12-22

PATCH New Steffen Klassert 2 participants 11 messages Dec 22, 2017 – Dec 27, 2017

Subsystems (6)

Networking Network drivers Documentation Traffic control (tc) +2 more subsystems

lore

Share & tools

lore

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #0 lore raw

1) Separate ESP handling from segmentation for GRO packets. This unifies the IPsec GSO and non GSO codepath.

2) Add asynchronous callbacks for xfrm on layer 2. This adds the necessary infrastructure to core networking.

3) Allow to use the layer2 IPsec GSO codepath for software crypto, all infrastructure is there now.

4) Also allow IPsec GSO with software crypto for local sockets.

5) Don't require synchronous crypto fallback on IPsec offloading, it is not needed anymore.

6) Check for xdo_dev_state_free and only call it if implemented. From Shannon Nelson.

7) Check for the required add and delete functions when a driver registers xdo_dev_ops. From Shannon Nelson.

8) Define xfrmdev_ops only with offload config.
From Shannon Nelson.

9) Update the xfrm stats documentation.
From Shannon Nelson.

Please pull or let me know if there are problems.

Thanks!

The following changes since commit f39a5c01c3d24f2f61ec9d8c7d7e81f9aca506ce:

Merge branch 'nfp-flower-add-Geneve-tunnel-support' (2017-12-19 14:52:13 -0500)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next.git master

for you to fetch changes up to 1a4bb1d14f7c0c4df418d08eb8e24d1c0e54b06a:

xfrm: update the stats documentation (2017-12-22 06:45:48 +0100)

----------------------------------------------------------------
Shannon Nelson (4):
xfrm: check for xdo_dev_state_free
xfrm: check for xdo_dev_ops add and delete
xfrm: wrap xfrmdev_ops with offload config
xfrm: update the stats documentation

Steffen Klassert (5):
xfrm: Separate ESP handling from segmentation for GRO packets.
net: Add asynchronous callbacks for xfrm on layer 2.
xfrm: Allow to use the layer2 IPsec GSO codepath for software crypto.
xfrm: Allow IPsec GSO with software crypto for local sockets.
esp: Don't require synchronous crypto fallback on offloading anymore.

 Documentation/networking/xfrm_proc.txt |  20 ++--
 include/linux/netdevice.h              |   8 +-
 include/net/xfrm.h                     |  29 ++++-
 net/core/dev.c                         |  19 ++--
 net/ipv4/esp4.c                        |  36 ++++---
 net/ipv4/esp4_offload.c                |  73 ++++---------
 net/ipv4/xfrm4_mode_tunnel.c           |   5 +-
 net/ipv6/esp6.c                        |  36 ++++---
 net/ipv6/esp6_offload.c                |  80 +++++---------
 net/ipv6/xfrm6_mode_tunnel.c           |   5 +-
 net/packet/af_packet.c                 |   3 +-
 net/sched/sch_generic.c                |  16 ++-
 net/xfrm/xfrm_device.c                 | 186 ++++++++++++++++++++++++++++-----
 13 files changed, 326 insertions(+), 190 deletions(-)

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #1 lore raw

PATCH 1/9 xfrm: Separate ESP handling from segmentation for GRO packets.

We change the ESP GSO handlers to only segment the packets. The ESP handling and encryption is defered to validate_xmit_xfrm() where this is done for non GRO packets too. This makes the code more robust and prepares for asynchronous crypto handling.

Signed-off-by Steffen Klassert <[email protected]>

---
 include/net/xfrm.h           |  6 +--
 net/core/dev.c               |  5 +--
 net/ipv4/esp4_offload.c      | 73 +++++++++++--------------------------
 net/ipv4/xfrm4_mode_tunnel.c |  5 +--
 net/ipv6/esp6_offload.c      | 80 ++++++++++++----------------------------
 net/ipv6/xfrm6_mode_tunnel.c |  5 +--
 net/xfrm/xfrm_device.c       | 87 +++++++++++++++++++++++++++++++++++++++-----
 7 files changed, 129 insertions(+), 132 deletions(-)

include/net/xfrm.h +3 -3

index 1ec0c4760646..df7f3d0ac4a1 100644
@@ -1888,7 +1888,7 @@ static inline struct xfrm_offload *xfrm_offload(struct sk_buff *skb)
 void __net_init xfrm_dev_init(void);
 
 #ifdef CONFIG_XFRM_OFFLOAD
-int validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features);
+struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features);
 int xfrm_dev_state_add(struct net *net, struct xfrm_state *x,
 		       struct xfrm_user_offload *xuo);
 bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x);

@@ -1929,9 +1929,9 @@ static inline void xfrm_dev_state_free(struct xfrm_state *x)
 	}
 }
 #else
-static inline int validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
+static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
 {
-	return 0;
+	return skb;
 }
 
 static inline int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, struct xfrm_user_offload *xuo)

net/core/dev.c +2 -3

index c7db39926769..fb7a24a373d1 100644
@@ -3083,9 +3083,6 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		    __skb_linearize(skb))
 			goto out_kfree_skb;
 
-		if (validate_xmit_xfrm(skb, features))
-			goto out_kfree_skb;
-
 		/* If packet is not checksummed and device does not
 		 * support checksumming for this protocol, complete
 		 * checksumming here.

@@ -3102,6 +3099,8 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		}
 	}
 
+	skb = validate_xmit_xfrm(skb, features);
+
 	return skb;
 
 out_kfree_skb:

net/ipv4/esp4_offload.c +21 ~~-52~~

125 lines not on this page · raw message

net/ipv4/xfrm4_mode_tunnel.c +1 -4

index 7d885a44dc9d..8affc6d83d58 100644
@@ -105,18 +105,15 @@ static struct sk_buff *xfrm4_mode_tunnel_gso_segment(struct xfrm_state *x,
 {
 	__skb_push(skb, skb->mac_len);
 	return skb_mac_gso_segment(skb, features);
-
 }
 
 static void xfrm4_mode_tunnel_xmit(struct xfrm_state *x, struct sk_buff *skb)
 {
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
-	if (xo->flags & XFRM_GSO_SEGMENT) {
-		skb->network_header = skb->network_header - x->props.header_len;
+	if (xo->flags & XFRM_GSO_SEGMENT)
 		skb->transport_header = skb->network_header +
 					sizeof(struct iphdr);
-	}
 
 	skb_reset_mac_len(skb);
 	pskb_pull(skb, skb->mac_len + x->props.header_len);

net/ipv6/esp6_offload.c +24 ~~-56~~

139 lines not on this page · raw message

net/ipv6/xfrm6_mode_tunnel.c +1 -4

index e66b94f46532..4e12859bc2ee 100644
@@ -105,17 +105,14 @@ static struct sk_buff *xfrm6_mode_tunnel_gso_segment(struct xfrm_state *x,
 {
 	__skb_push(skb, skb->mac_len);
 	return skb_mac_gso_segment(skb, features);
-
 }
 
 static void xfrm6_mode_tunnel_xmit(struct xfrm_state *x, struct sk_buff *skb)
 {
 	struct xfrm_offload *xo = xfrm_offload(skb);
 
-	if (xo->flags & XFRM_GSO_SEGMENT) {
-		skb->network_header = skb->network_header - x->props.header_len;
+	if (xo->flags & XFRM_GSO_SEGMENT)
 		skb->transport_header = skb->network_header + sizeof(struct ipv6hdr);
-	}
 
 	skb_reset_mac_len(skb);
 	pskb_pull(skb, skb->mac_len + x->props.header_len);

net/xfrm/xfrm_device.c +77 ~~-10~~

109 lines not on this page · raw message

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #2 lore raw

PATCH 2/9 net: Add asynchronous callbacks for xfrm on layer 2.

In reply to Steffen Klassert

This patch implements asynchronous crypto callbacks and a backlog handler that can be used when IPsec is done at layer 2 in the TX path. It also extends the skb validate functions so that we can update the driver transmit return codes based on async crypto operation or to indicate that we queued the packet in a backlog queue.

Joint work with: Aviv Heller <[email protected]>

Signed-off-by Steffen Klassert <[email protected]>

---
 include/linux/netdevice.h |   6 ++-
 include/net/xfrm.h        |  22 ++++++++--
 net/core/dev.c            |  16 +++++---
 net/ipv4/esp4.c           |  24 +++++++++--
 net/ipv6/esp6.c           |  24 +++++++++--
 net/packet/af_packet.c    |   3 +-
 net/sched/sch_generic.c   |  16 +++++++-
 net/xfrm/xfrm_device.c    | 100 +++++++++++++++++++++++++++++++++++++---------
 8 files changed, 175 insertions(+), 36 deletions(-)

include/linux/netdevice.h +4 -2

index cc4ce7456e38..c82d207ebc97 100644
@@ -2793,7 +2793,9 @@ struct softnet_data {
 	struct Qdisc		*output_queue;
 	struct Qdisc		**output_queue_tailp;
 	struct sk_buff		*completion_queue;
-
+#ifdef CONFIG_XFRM_OFFLOAD
+	struct sk_buff_head	xfrm_backlog;
+#endif
 #ifdef CONFIG_RPS
 	/* input_queue_head should be written by cpu owning this struct,
 	 * and only read by other cpus. Worth using a cache line.

@@ -3325,7 +3327,7 @@ int dev_get_phys_port_id(struct net_device *dev,
 int dev_get_phys_port_name(struct net_device *dev,
 			   char *name, size_t len);
 int dev_change_proto_down(struct net_device *dev, bool proto_down);
-struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
+struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
 				    struct netdev_queue *txq, int *ret);

include/net/xfrm.h +19 -3

index df7f3d0ac4a1..2517c4f7781a 100644
@@ -1051,6 +1051,7 @@ struct xfrm_offload {
 #define	XFRM_GSO_SEGMENT	16
 #define	XFRM_GRO		32
 #define	XFRM_ESP_NO_TRAILER	64
+#define	XFRM_DEV_RESUME		128
 
 	__u32			status;
 #define CRYPTO_SUCCESS				1

@@ -1874,21 +1875,28 @@ static inline struct xfrm_state *xfrm_input_state(struct sk_buff *skb)
 {
 	return skb->sp->xvec[skb->sp->len - 1];
 }
+#endif
+
 static inline struct xfrm_offload *xfrm_offload(struct sk_buff *skb)
 {
+#ifdef CONFIG_XFRM
 	struct sec_path *sp = skb->sp;
 
 	if (!sp || !sp->olen || sp->len != sp->olen)
 		return NULL;
 
 	return &sp->ovec[sp->olen - 1];
-}
+#else
+	return NULL;
 #endif
+}
 
 void __net_init xfrm_dev_init(void);
 
 #ifdef CONFIG_XFRM_OFFLOAD
-struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features);
+void xfrm_dev_resume(struct sk_buff *skb);
+void xfrm_dev_backlog(struct softnet_data *sd);
+struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again);
 int xfrm_dev_state_add(struct net *net, struct xfrm_state *x,
 		       struct xfrm_user_offload *xuo);
 bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x);

@@ -1929,7 +1937,15 @@ static inline void xfrm_dev_state_free(struct xfrm_state *x)
 	}
 }
 #else
-static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features)
+static inline void xfrm_dev_resume(struct sk_buff *skb)
+{
+}
+
+static inline void xfrm_dev_backlog(struct softnet_data *sd)
+{
+}
+
+static inline struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again)
 {
 	return skb;
 }

net/core/dev.c +11 -5

index fb7a24a373d1..821dd8cb7169 100644
@@ -3059,7 +3059,7 @@ int skb_csum_hwoffload_help(struct sk_buff *skb,
 }
 EXPORT_SYMBOL(skb_csum_hwoffload_help);
 
-static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
+static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev, bool *again)
 {
 	netdev_features_t features;

@@ -3099,7 +3099,7 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 		}
 	}
 
-	skb = validate_xmit_xfrm(skb, features);
+	skb = validate_xmit_xfrm(skb, features, again);
 
 	return skb;

@@ -3110,7 +3110,7 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
 	return NULL;
 }
 
-struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
+struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again)
 {
 	struct sk_buff *next, *head = NULL, *tail;

@@ -3121,7 +3121,7 @@ struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *d
 		/* in case skb wont be segmented, point to itself */
 		skb->prev = skb;
 
-		skb = validate_xmit_skb(skb, dev);
+		skb = validate_xmit_skb(skb, dev, again);
 		if (!skb)
 			continue;

@@ -3448,6 +3448,7 @@ static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
 	struct netdev_queue *txq;
 	struct Qdisc *q;
 	int rc = -ENOMEM;
+	bool again = false;
 
 	skb_reset_mac_header(skb);

@@ -3509,7 +3510,7 @@ static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
 				     XMIT_RECURSION_LIMIT))
 				goto recursion_alert;
 
-			skb = validate_xmit_skb(skb, dev);
+			skb = validate_xmit_skb(skb, dev, &again);
 			if (!skb)
 				goto out;

@@ -4193,6 +4194,8 @@ static __latent_entropy void net_tx_action(struct softirq_action *h)
 				spin_unlock(root_lock);
 		}
 	}
+
+	xfrm_dev_backlog(sd);
 }
 
 #if IS_ENABLED(CONFIG_BRIDGE) && IS_ENABLED(CONFIG_ATM_LANE)

@@ -8874,6 +8877,9 @@ static int __init net_dev_init(void)
 
 		skb_queue_head_init(&sd->input_pkt_queue);
 		skb_queue_head_init(&sd->process_queue);
+#ifdef CONFIG_XFRM_OFFLOAD
+		skb_queue_head_init(&sd->xfrm_backlog);
+#endif
 		INIT_LIST_HEAD(&sd->poll_list);
 		sd->output_queue_tailp = &sd->output_queue;
 #ifdef CONFIG_RPS

net/ipv4/esp4.c +21 -3

index d57aa64fa7c7..7948833dc204 100644
@@ -121,14 +121,32 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
 static void esp_output_done(struct crypto_async_request *base, int err)
 {
 	struct sk_buff *skb = base->data;
+	struct xfrm_offload *xo = xfrm_offload(skb);
 	void *tmp;
-	struct dst_entry *dst = skb_dst(skb);
-	struct xfrm_state *x = dst->xfrm;
+	struct xfrm_state *x;
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME))
+		x = skb->sp->xvec[skb->sp->len - 1];
+	else
+		x = skb_dst(skb)->xfrm;
 
 	tmp = ESP_SKB_CB(skb)->tmp;
 	esp_ssg_unref(x, tmp);
 	kfree(tmp);
-	xfrm_output_resume(skb, err);
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME)) {
+		if (err) {
+			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+			kfree_skb(skb);
+			return;
+		}
+
+		skb_push(skb, skb->data - skb_mac_header(skb));
+		secpath_reset(skb);
+		xfrm_dev_resume(skb);
+	} else {
+		xfrm_output_resume(skb, err);
+	}
 }
 
 /* Move ESP header back into place. */

net/ipv6/esp6.c +21 -3

index a902ff8f59be..08a424fa8009 100644
@@ -141,14 +141,32 @@ static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
 static void esp_output_done(struct crypto_async_request *base, int err)
 {
 	struct sk_buff *skb = base->data;
+	struct xfrm_offload *xo = xfrm_offload(skb);
 	void *tmp;
-	struct dst_entry *dst = skb_dst(skb);
-	struct xfrm_state *x = dst->xfrm;
+	struct xfrm_state *x;
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME))
+		x = skb->sp->xvec[skb->sp->len - 1];
+	else
+		x = skb_dst(skb)->xfrm;
 
 	tmp = ESP_SKB_CB(skb)->tmp;
 	esp_ssg_unref(x, tmp);
 	kfree(tmp);
-	xfrm_output_resume(skb, err);
+
+	if (xo && (xo->flags & XFRM_DEV_RESUME)) {
+		if (err) {
+			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
+			kfree_skb(skb);
+			return;
+		}
+
+		skb_push(skb, skb->data - skb_mac_header(skb));
+		secpath_reset(skb);
+		xfrm_dev_resume(skb);
+	} else {
+		xfrm_output_resume(skb, err);
+	}
 }
 
 /* Move ESP header back into place. */

net/packet/af_packet.c +2 -1

index da215e5c1399..ee7aa0ba3a67 100644
@@ -247,12 +247,13 @@ static int packet_direct_xmit(struct sk_buff *skb)
 	struct sk_buff *orig_skb = skb;
 	struct netdev_queue *txq;
 	int ret = NETDEV_TX_BUSY;
+	bool again = false;
 
 	if (unlikely(!netif_running(dev) ||
 		     !netif_carrier_ok(dev)))
 		goto drop;
 
-	skb = validate_xmit_skb_list(skb, dev);
+	skb = validate_xmit_skb_list(skb, dev, &again);
 	if (skb != orig_skb)
 		goto drop;

net/sched/sch_generic.c +15 -1

index 876fab2604b8..f9a8761f0ff2 100644
@@ -32,6 +32,7 @@
 #include <net/pkt_sched.h>
 #include <net/dst.h>
 #include <trace/events/qdisc.h>
+#include <net/xfrm.h>
 
 /* Qdisc to use by default */
 const struct Qdisc_ops *default_qdisc_ops = &pfifo_fast_ops;

@@ -230,6 +231,8 @@ static struct sk_buff *dequeue_skb(struct Qdisc *q, bool *validate,
 
 		/* skb in gso_skb were already validated */
 		*validate = false;
+		if (xfrm_offload(skb))
+			*validate = true;
 		/* check the reason of requeuing without tx lock first */
 		txq = skb_get_tx_queue(txq->dev, skb);
 		if (!netif_xmit_frozen_or_stopped(txq)) {

@@ -285,6 +288,7 @@ bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
 		     spinlock_t *root_lock, bool validate)
 {
 	int ret = NETDEV_TX_BUSY;
+	bool again = false;
 
 	/* And release qdisc */
 	if (root_lock)

@@ -292,7 +296,17 @@ bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
 
 	/* Note that we validate skb (GSO, checksum, ...) outside of locks */
 	if (validate)
-		skb = validate_xmit_skb_list(skb, dev);
+		skb = validate_xmit_skb_list(skb, dev, &again);
+
+#ifdef CONFIG_XFRM_OFFLOAD
+	if (unlikely(again)) {
+		if (root_lock)
+			spin_lock(root_lock);
+
+		dev_requeue_skb(skb, q);
+		return false;
+	}
+#endif
 
 	if (likely(skb)) {
 		HARD_TX_LOCK(dev, txq, smp_processor_id());

net/xfrm/xfrm_device.c +82 ~~-18~~

165 lines not on this page · raw message

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #3 lore raw

PATCH 3/9 xfrm: Allow to use the layer2 IPsec GSO codepath for software crypto.

In reply to Steffen Klassert

We now have support for asynchronous crypto operations in the layer 2 TX path. This was the missing part to allow the GSO codepath for software crypto, so allow this codepath now.

Signed-off-by Steffen Klassert <[email protected]>

---
 net/xfrm/xfrm_device.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

net/xfrm/xfrm_device.c +2 -2

index fc8ab9f71127..20a96181867a 100644
@@ -202,8 +202,8 @@ bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x)
 	if (!x->type_offload || x->encap)
 		return false;
 
-	if ((x->xso.offload_handle && (dev == xfrm_dst_path(dst)->dev)) &&
-	     !xdst->child->xfrm && x->type->get_mtu) {
+	if ((!dev || (x->xso.offload_handle && (dev == xfrm_dst_path(dst)->dev))) &&
+	     (!xdst->child->xfrm && x->type->get_mtu)) {
 		mtu = x->type->get_mtu(x, xdst->child_mtu_cached);
 
 		if (skb->len <= mtu)

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #4 lore raw

PATCH 4/9 xfrm: Allow IPsec GSO with software crypto for local sockets.

In reply to Steffen Klassert

With support of async crypto operations in the GSO codepath we have everything in place to allow GSO for local sockets. This patch enables the GSO codepath.

Signed-off-by Steffen Klassert <[email protected]>

---
 include/net/xfrm.h | 2 ++
 1 file changed, 2 insertions(+)

include/net/xfrm.h +2 -0

index 2517c4f7781a..357764a2bb4e 100644
@@ -1910,6 +1910,8 @@ static inline bool xfrm_dst_offload_ok(struct dst_entry *dst)
 		return false;
 
 	xdst = (struct xfrm_dst *) dst;
+	if (!x->xso.offload_handle && !xdst->child->xfrm)
+		return true;
 	if (x->xso.offload_handle && (x->xso.dev == xfrm_dst_path(dst)->dev) &&
 	    !xdst->child->xfrm)
 		return true;

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #5 lore raw

PATCH 5/9 esp: Don't require synchronous crypto fallback on offloading anymore.

In reply to Steffen Klassert

We support asynchronous crypto on layer 2 ESP now. So no need to force synchronous crypto fallback on offloading anymore.

Signed-off-by Steffen Klassert <[email protected]>

---
 net/ipv4/esp4.c | 12 ++----------
 net/ipv6/esp6.c | 12 ++----------
 2 files changed, 4 insertions(+), 20 deletions(-)

net/ipv4/esp4.c +2 ~~-10~~

index 7948833dc204..6f00e43120a8 100644
@@ -843,17 +843,13 @@ static int esp_init_aead(struct xfrm_state *x)
 	char aead_name[CRYPTO_MAX_ALG_NAME];
 	struct crypto_aead *aead;
 	int err;
-	u32 mask = 0;
 
 	err = -ENAMETOOLONG;
 	if (snprintf(aead_name, CRYPTO_MAX_ALG_NAME, "%s(%s)",
 		     x->geniv, x->aead->alg_name) >= CRYPTO_MAX_ALG_NAME)
 		goto error;
 
-	if (x->xso.offload_handle)
-		mask |= CRYPTO_ALG_ASYNC;
-
-	aead = crypto_alloc_aead(aead_name, 0, mask);
+	aead = crypto_alloc_aead(aead_name, 0, 0);
 	err = PTR_ERR(aead);
 	if (IS_ERR(aead))
 		goto error;

@@ -883,7 +879,6 @@ static int esp_init_authenc(struct xfrm_state *x)
 	char authenc_name[CRYPTO_MAX_ALG_NAME];
 	unsigned int keylen;
 	int err;
-	u32 mask = 0;
 
 	err = -EINVAL;
 	if (!x->ealg)

@@ -909,10 +904,7 @@ static int esp_init_authenc(struct xfrm_state *x)
 			goto error;
 	}
 
-	if (x->xso.offload_handle)
-		mask |= CRYPTO_ALG_ASYNC;
-
-	aead = crypto_alloc_aead(authenc_name, 0, mask);
+	aead = crypto_alloc_aead(authenc_name, 0, 0);
 	err = PTR_ERR(aead);
 	if (IS_ERR(aead))
 		goto error;

net/ipv6/esp6.c +2 ~~-10~~

index 08a424fa8009..7c888c6e53a9 100644
@@ -752,17 +752,13 @@ static int esp_init_aead(struct xfrm_state *x)
 	char aead_name[CRYPTO_MAX_ALG_NAME];
 	struct crypto_aead *aead;
 	int err;
-	u32 mask = 0;
 
 	err = -ENAMETOOLONG;
 	if (snprintf(aead_name, CRYPTO_MAX_ALG_NAME, "%s(%s)",
 		     x->geniv, x->aead->alg_name) >= CRYPTO_MAX_ALG_NAME)
 		goto error;
 
-	if (x->xso.offload_handle)
-		mask |= CRYPTO_ALG_ASYNC;
-
-	aead = crypto_alloc_aead(aead_name, 0, mask);
+	aead = crypto_alloc_aead(aead_name, 0, 0);
 	err = PTR_ERR(aead);
 	if (IS_ERR(aead))
 		goto error;

@@ -792,7 +788,6 @@ static int esp_init_authenc(struct xfrm_state *x)
 	char authenc_name[CRYPTO_MAX_ALG_NAME];
 	unsigned int keylen;
 	int err;
-	u32 mask = 0;
 
 	err = -EINVAL;
 	if (!x->ealg)

@@ -818,10 +813,7 @@ static int esp_init_authenc(struct xfrm_state *x)
 			goto error;
 	}
 
-	if (x->xso.offload_handle)
-		mask |= CRYPTO_ALG_ASYNC;
-
-	aead = crypto_alloc_aead(authenc_name, 0, mask);
+	aead = crypto_alloc_aead(authenc_name, 0, 0);
 	err = PTR_ERR(aead);
 	if (IS_ERR(aead))
 		goto error;

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #6 lore raw

PATCH 6/9 xfrm: check for xdo_dev_state_free

In reply to Steffen Klassert

From: Shannon Nelson <[email protected]>

The current XFRM code assumes that we've implemented the xdo_dev_state_free() callback, even if it is meaningless to the driver. This patch adds a check for it before calling, as done in other APIs, to prevent a NULL function pointer kernel crash.

Signed-off-by Shannon Nelson <[email protected]>

Signed-off-by Steffen Klassert <[email protected]>

---
 include/net/xfrm.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

include/net/xfrm.h +2 -1

index 357764a2bb4e..079ea9455bcd 100644
@@ -1933,7 +1933,8 @@ static inline void xfrm_dev_state_free(struct xfrm_state *x)
 	 struct net_device *dev = xso->dev;
 
 	if (dev && dev->xfrmdev_ops) {
-		dev->xfrmdev_ops->xdo_dev_state_free(x);
+		if (dev->xfrmdev_ops->xdo_dev_state_free)
+			dev->xfrmdev_ops->xdo_dev_state_free(x);
 		xso->dev = NULL;
 		dev_put(dev);
 	}

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #7 lore raw

PATCH 7/9 xfrm: check for xdo_dev_ops add and delete

In reply to Steffen Klassert

From: Shannon Nelson <[email protected]>

This adds a check for the required add and delete functions up front at registration time to be sure both are defined.

Since both the features check and the registration check are looking at the same things, break out the check for both to call.

Lastly, for some reason the feature check was setting xfrmdev_ops to NULL if the NETIF_F_HW_ESP bit was missing, which would probably surprise the driver later if the driver turned its NETIF_F_HW_ESP bit back on. We shouldn't be messing with the driver's callback list, so we stop doing that with this patch.

Signed-off-by Shannon Nelson <[email protected]>

Signed-off-by Steffen Klassert <[email protected]>

---
 net/xfrm/xfrm_device.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

net/xfrm/xfrm_device.c +18 ~~-13~~

index 20a96181867a..75982506617b 100644
@@ -273,17 +273,31 @@ void xfrm_dev_backlog(struct softnet_data *sd)
 }
 #endif
 
-static int xfrm_dev_register(struct net_device *dev)
+static int xfrm_api_check(struct net_device *dev)
 {
-	if ((dev->features & NETIF_F_HW_ESP) && !dev->xfrmdev_ops)
-		return NOTIFY_BAD;
+#ifdef CONFIG_XFRM_OFFLOAD
 	if ((dev->features & NETIF_F_HW_ESP_TX_CSUM) &&
 	    !(dev->features & NETIF_F_HW_ESP))
 		return NOTIFY_BAD;
 
+	if ((dev->features & NETIF_F_HW_ESP) &&
+	    (!(dev->xfrmdev_ops &&
+	       dev->xfrmdev_ops->xdo_dev_state_add &&
+	       dev->xfrmdev_ops->xdo_dev_state_delete)))
+		return NOTIFY_BAD;
+#else
+	if (dev->features & (NETIF_F_HW_ESP | NETIF_F_HW_ESP_TX_CSUM))
+		return NOTIFY_BAD;
+#endif
+
 	return NOTIFY_DONE;
 }
 
+static int xfrm_dev_register(struct net_device *dev)
+{
+	return xfrm_api_check(dev);
+}
+
 static int xfrm_dev_unregister(struct net_device *dev)
 {
 	xfrm_policy_cache_flush();

@@ -292,16 +306,7 @@ static int xfrm_dev_unregister(struct net_device *dev)
 
 static int xfrm_dev_feat_change(struct net_device *dev)
 {
-	if ((dev->features & NETIF_F_HW_ESP) && !dev->xfrmdev_ops)
-		return NOTIFY_BAD;
-	else if (!(dev->features & NETIF_F_HW_ESP))
-		dev->xfrmdev_ops = NULL;
-
-	if ((dev->features & NETIF_F_HW_ESP_TX_CSUM) &&
-	    !(dev->features & NETIF_F_HW_ESP))
-		return NOTIFY_BAD;
-
-	return NOTIFY_DONE;
+	return xfrm_api_check(dev);
 }
 
 static int xfrm_dev_down(struct net_device *dev)

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #8 lore raw

PATCH 8/9 xfrm: wrap xfrmdev_ops with offload config

In reply to Steffen Klassert

From: Shannon Nelson <[email protected]>

There's no reason to define netdev->xfrmdev_ops if the offload facility is not CONFIG'd in.

Signed-off-by Shannon Nelson <[email protected]>

Signed-off-by Steffen Klassert <[email protected]>

---
 include/linux/netdevice.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

include/linux/netdevice.h +1 -1

index c82d207ebc97..352066e4eeef 100644
@@ -1726,7 +1726,7 @@ struct net_device {
 	const struct ndisc_ops *ndisc_ops;
 #endif
 
-#ifdef CONFIG_XFRM
+#ifdef CONFIG_XFRM_OFFLOAD
 	const struct xfrmdev_ops *xfrmdev_ops;
 #endif

Signature · 1 line

2.14.1

Steffen Klassert

author maintainer

2017-12-22 11:24 UTC #9 lore raw

PATCH 9/9 xfrm: update the stats documentation

In reply to Steffen Klassert

From: Shannon Nelson <[email protected]>

Add a couple of stats that aren't in the documentation file and rework the top description to be a little more readable.

Signed-off-by Shannon Nelson <[email protected]>

Signed-off-by Steffen Klassert <[email protected]>

---
 Documentation/networking/xfrm_proc.txt | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

Documentation/networking/xfrm_proc.txt +14 -6

index d0d8bafa9016..2eae619ab67b 100644
@@ -5,13 +5,15 @@ Masahide NAKAMURA <[email protected]>
 
 Transformation Statistics
 -------------------------
-xfrm_proc is a statistics shown factor dropped by transformation
-for developer.
-It is a counter designed from current transformation source code
-and defined like linux private MIB.
 
-Inbound statistics
-~~~~~~~~~~~~~~~~~~
+The xfrm_proc code is a set of statistics showing numbers of packets
+dropped by the transformation code and why.  These counters are defined
+as part of the linux private MIB.  These counters can be viewed in
+/proc/net/xfrm_stat.
+
+
+Inbound errors
+~~~~~~~~~~~~~~
 XfrmInError:
 	All errors which is not matched others
 XfrmInBufferError:

@@ -46,6 +48,10 @@ XfrmInPolBlock:
 	Policy discards
 XfrmInPolError:
 	Policy error
+XfrmAcquireError:
+	State hasn't been fully acquired before use
+XfrmFwdHdrError:
+	Forward routing of a packet is not allowed
 
 Outbound errors
 ~~~~~~~~~~~~~~~

@@ -72,3 +78,5 @@ XfrmOutPolDead:
 	Policy is dead
 XfrmOutPolError:
 	Policy error
+XfrmOutStateInvalid:
+	State is invalid, perhaps expired

Signature · 1 line

2.14.1

David Miller

maintainer

2017-12-27 16:59 UTC #10 lore raw

In reply to Steffen Klassert

From: Steffen Klassert <[email protected]>
Date: Fri, 22 Dec 2017 12:24:30 +0100

Quoted 20 lines

1) Separate ESP handling from segmentation for GRO packets. This unifies the IPsec GSO and non GSO codepath.
2) Add asynchronous callbacks for xfrm on layer 2. This adds the necessary infrastructure to core networking.
3) Allow to use the layer2 IPsec GSO codepath for software crypto, all infrastructure is there now.
4) Also allow IPsec GSO with software crypto for local sockets.
5) Don't require synchronous crypto fallback on IPsec offloading, it is not needed anymore.
6) Check for xdo_dev_state_free and only call it if implemented. From Shannon Nelson.
7) Check for the required add and delete functions when a driver registers xdo_dev_ops. From Shannon Nelson.

Quoted 7 lines

8) Define xfrmdev_ops only with offload config.
From Shannon Nelson.
9) Update the xfrm stats documentation.
From Shannon Nelson.
Please pull or let me know if there are problems.

Pulled, thanks!