Merge tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-nextgrafted

Pull networking updates from Jakub Kicinski:
 "Core:

   - Add dedicated kmem_cache for typical/small skb->head, avoid having
     to access struct page at kfree time, and improve memory use.

   - Introduce sysctl to set default RPS configuration for new netdevs.

   - Define Netlink protocol specification format which can be used to
     describe messages used by each family and auto-generate parsers.
     Add tools for generating kernel data structures and uAPI headers.

   - Expose all net/core sysctls inside netns.

   - Remove 4s sleep in netpoll if carrier is instantly detected on
     boot.

   - Add configurable limit of MDB entries per port, and port-vlan.

   - Continue populating drop reasons throughout the stack.

   - Retire a handful of legacy Qdiscs and classifiers.

  Protocols:

   - Support IPv4 big TCP (TSO frames larger than 64kB).

   - Add IP_LOCAL_PORT_RANGE socket option, to control local port range
     on socket by socket basis.

   - Track and report in procfs number of MPTCP sockets used.

   - Support mixing IPv4 and IPv6 flows in the in-kernel MPTCP path
     manager.

   - IPv6: don't check net.ipv6.route.max_size and rely on garbage
     collection to free memory (similarly to IPv4).

   - Support Penultimate Segment Pop (PSP) flavor in SRv6 (RFC8986).

   - ICMP: add per-rate limit counters.

   - Add support for user scanning requests in ieee802154.

   - Remove static WEP support.

   - Support minimal Wi-Fi 7 Extremely High Throughput (EHT) rate
     reporting.

   - WiFi 7 EHT channel puncturing support (client & AP).

  BPF:

   - Add a rbtree data structure following the "next-gen data structure"
     precedent set by recently added linked list, that is, by using
     kfunc + kptr instead of adding a new BPF map type.

   - Expose XDP hints via kfuncs with initial support for RX hash and
     timestamp metadata.

   - Add BPF_F_NO_TUNNEL_KEY extension to bpf_skb_set_tunnel_key to
     better support decap on GRE tunnel devices not operating in collect
     metadata.

   - Improve x86 JIT's codegen for PROBE_MEM runtime error checks.

   - Remove the need for trace_printk_lock for bpf_trace_printk and
     bpf_trace_vprintk helpers.

   - Extend libbpf's bpf_tracing.h support for tracing arguments of
     kprobes/uprobes and syscall as a special case.

   - Significantly reduce the search time for module symbols by
     livepatch and BPF.

   - Enable cpumasks to be used as kptrs, which is useful for tracing
     programs tracking which tasks end up running on which CPUs in
     different time intervals.

   - Add support for BPF trampoline on s390x and riscv64.

   - Add capability to export the XDP features supported by the NIC.

   - Add __bpf_kfunc tag for marking kernel functions as kfuncs.

   - Add cgroup.memory=nobpf kernel parameter option to disable BPF
     memory accounting for container environments.

  Netfilter:

   - Remove the CLUSTERIP target. It has been marked as obsolete for
     years, and we still have WARN splats wrt races of the out-of-band
     /proc interface installed by this target.

   - Add 'destroy' commands to nf_tables. They are identical to the
     existing 'delete' commands, but do not return an error if the
     referenced object (set, chain, rule...) did not exist.

  Driver API:

   - Improve cpumask_local_spread() locality to help NICs set the right
     IRQ affinity on AMD platforms.

   - Separate C22 and C45 MDIO bus transactions more clearly.

   - Introduce new DCB table to control DSCP rewrite on egress.

   - Support configuration of Physical Layer Collision Avoidance (PLCA)
     Reconciliation Sublayer (RS) (802.3cg-2019). Modern version of
     shared medium Ethernet.

   - Support for MAC Merge layer (IEEE 802.3-2018 clause 99). Allowing
     preemption of low priority frames by high priority frames.

   - Add support for controlling MACSec offload using netlink SET.

   - Rework devlink instance refcounts to allow registration and
     de-registration under the instance lock. Split the code into
     multiple files, drop some of the unnecessarily granular locks and
     factor out common parts of netlink operation handling.

   - Add TX frame aggregation parameters (for USB drivers).

   - Add a new attr TCA_EXT_WARN_MSG to report TC (offload) warning
     messages with notifications for debug.

   - Allow offloading of UDP NEW connections via act_ct.

   - Add support for per action HW stats in TC.

   - Support hardware miss to TC action (continue processing in SW from
     a specific point in the action chain).

   - Warn if old Wireless Extension user space interface is used with
     modern cfg80211/mac80211 drivers. Do not support Wireless
     Extensions for Wi-Fi 7 devices at all. Everyone should switch to
     using nl80211 interface instead.

   - Improve the CAN bit timing configuration. Use extack to return
     error messages directly to user space, update the SJW handling,
     including the definition of a new default value that will benefit
     CAN-FD controllers, by increasing their oscillator tolerance.

  New hardware / drivers:

   - Ethernet:
      - nVidia BlueField-3 support (control traffic driver)
      - Ethernet support for imx93 SoCs
      - Motorcomm yt8531 gigabit Ethernet PHY
      - onsemi NCN26000 10BASE-T1S PHY (with support for PLCA)
      - Microchip LAN8841 PHY (incl. cable diagnostics and PTP)
      - Amlogic gxl MDIO mux

   - WiFi:
      - RealTek RTL8188EU (rtl8xxxu)
      - Qualcomm Wi-Fi 7 devices (ath12k)

   - CAN:
      - Renesas R-Car V4H

  Drivers:

   - Bluetooth:
      - Set Per Platform Antenna Gain (PPAG) for Intel controllers.

   - Ethernet NICs:
      - Intel (1G, igc):
         - support TSN / Qbv / packet scheduling features of i226 model
      - Intel (100G, ice):
         - use GNSS subsystem instead of TTY
         - multi-buffer XDP support
         - extend support for GPIO pins to E823 devices
      - nVidia/Mellanox:
         - update the shared buffer configuration on PFC commands
         - implement PTP adjphase function for HW offset control
         - TC support for Geneve and GRE with VF tunnel offload
         - more efficient crypto key management method
         - multi-port eswitch support
      - Netronome/Corigine:
         - add DCB IEEE support
         - support IPsec offloading for NFP3800
      - Freescale/NXP (enetc):
         - support XDP_REDIRECT for XDP non-linear buffers
         - improve reconfig, avoid link flap and waiting for idle
         - support MAC Merge layer
      - Other NICs:
         - sfc/ef100: add basic devlink support for ef100
         - ionic: rx_push mode operation (writing descriptors via MMIO)
         - bnxt: use the auxiliary bus abstraction for RDMA
         - r8169: disable ASPM and reset bus in case of tx timeout
         - cpsw: support QSGMII mode for J721e CPSW9G
         - cpts: support pulse-per-second output
         - ngbe: add an mdio bus driver
         - usbnet: optimize usbnet_bh() by avoiding unnecessary queuing
         - r8152: handle devices with FW with NCM support
         - amd-xgbe: support 10Mbps, 2.5GbE speeds and rx-adaptation
         - virtio-net: support multi buffer XDP
         - virtio/vsock: replace virtio_vsock_pkt with sk_buff
         - tsnep: XDP support

   - Ethernet high-speed switches:
      - nVidia/Mellanox (mlxsw):
         - add support for latency TLV (in FW control messages)
      - Microchip (sparx5):
         - separate explicit and implicit traffic forwarding rules, make
           the implicit rules always active
         - add support for egress DSCP rewrite
         - IS0 VCAP support (Ingress Classification)
         - IS2 VCAP filters (protos, L3 addrs, L4 ports, flags, ToS
           etc.)
         - ES2 VCAP support (Egress Access Control)
         - support for Per-Stream Filtering and Policing (802.1Q,
           8.6.5.1)

   - Ethernet embedded switches:
      - Marvell (mv88e6xxx):
         - add MAB (port auth) offload support
         - enable PTP receive for mv88e6390
      - NXP (ocelot):
         - support MAC Merge layer
         - support for the the vsc7512 internal copper phys
      - Microchip:
         - lan9303: convert to PHYLINK
         - lan966x: support TC flower filter statistics
         - lan937x: PTP support for KSZ9563/KSZ8563 and LAN937x
         - lan937x: support Credit Based Shaper configuration
         - ksz9477: support Energy Efficient Ethernet
      - other:
         - qca8k: convert to regmap read/write API, use bulk operations
         - rswitch: Improve TX timestamp accuracy

   - Intel WiFi (iwlwifi):
      - EHT (Wi-Fi 7) rate reporting
      - STEP equalizer support: transfer some STEP (connection to radio
        on platforms with integrated wifi) related parameters from the
        BIOS to the firmware.

   - Qualcomm 802.11ax WiFi (ath11k):
      - IPQ5018 support
      - Fine Timing Measurement (FTM) responder role support
      - channel 177 support

   - MediaTek WiFi (mt76):
      - per-PHY LED support
      - mt7996: EHT (Wi-Fi 7) support
      - Wireless Ethernet Dispatch (WED) reset support
      - switch to using page pool allocator

   - RealTek WiFi (rtw89):
      - support new version of Bluetooth co-existance

   - Mobile:
      - rmnet: support TX aggregation"

* tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1872 commits)
  page_pool: add a comment explaining the fragment counter usage
  net: ethtool: fix __ethtool_dev_mm_supported() implementation
  ethtool: pse-pd: Fix double word in comments
  xsk: add linux/vmalloc.h to xsk.c
  sefltests: netdevsim: wait for devlink instance after netns removal
  selftest: fib_tests: Always cleanup before exit
  net/mlx5e: Align IPsec ASO result memory to be as required by hardware
  net/mlx5e: TC, Set CT miss to the specific ct action instance
  net/mlx5e: Rename CHAIN_TO_REG to MAPPED_OBJ_TO_REG
  net/mlx5: Refactor tc miss handling to a single function
  net/mlx5: Kconfig: Make tc offload depend on tc skb extension
  net/sched: flower: Support hardware miss to tc action
  net/sched: flower: Move filter handle initialization earlier
  net/sched: cls_api: Support hardware miss to tc action
  net/sched: Rename user cookie and act cookie
  sfc: fix builds without CONFIG_RTC_LIB
  sfc: clean up some inconsistent indentings
  net/mlx4_en: Introduce flexible array to silence overflow warning
  net: lan966x: Fix possible deadlock inside PTP
  net/ulp: Remove redundant ->clone() test in inet_clone_ulp().
  ...

1 files changed, 599 insertions, 0 deletions

diff --git a/drivers/misc/sgi-xp/xpnet.c b/drivers/misc/sgi-xp/xpnet.c
new file mode 100644
index 000000000..2396ba3b0
--- /dev/null
+++ b/drivers/misc/sgi-xp/xpnet.c
@@ -0,0 +1,599 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * (C) Copyright 2020 Hewlett Packard Enterprise Development LP
+ * Copyright (C) 1999-2009 Silicon Graphics, Inc. All rights reserved.
+ */
+
+/*
+ * Cross Partition Network Interface (XPNET) support
+ *
+ *	XPNET provides a virtual network layered on top of the Cross
+ *	Partition communication layer.
+ *
+ *	XPNET provides direct point-to-point and broadcast-like support
+ *	for an ethernet-like device.  The ethernet broadcast medium is
+ *	replaced with a point-to-point message structure which passes
+ *	pointers to a DMA-capable block that a remote partition should
+ *	retrieve and pass to the upper level networking layer.
+ *
+ */
+
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include "xp.h"
+
+/*
+ * The message payload transferred by XPC.
+ *
+ * buf_pa is the physical address where the DMA should pull from.
+ *
+ * NOTE: for performance reasons, buf_pa should _ALWAYS_ begin on a
+ * cacheline boundary.  To accomplish this, we record the number of
+ * bytes from the beginning of the first cacheline to the first useful
+ * byte of the skb (leadin_ignore) and the number of bytes from the
+ * last useful byte of the skb to the end of the last cacheline
+ * (tailout_ignore).
+ *
+ * size is the number of bytes to transfer which includes the skb->len
+ * (useful bytes of the senders skb) plus the leadin and tailout
+ */
+struct xpnet_message {
+	u16 version;		/* Version for this message */
+	u16 embedded_bytes;	/* #of bytes embedded in XPC message */
+	u32 magic;		/* Special number indicating this is xpnet */
+	unsigned long buf_pa;	/* phys address of buffer to retrieve */
+	u32 size;		/* #of bytes in buffer */
+	u8 leadin_ignore;	/* #of bytes to ignore at the beginning */
+	u8 tailout_ignore;	/* #of bytes to ignore at the end */
+	unsigned char data;	/* body of small packets */
+};
+
+/*
+ * Determine the size of our message, the cacheline aligned size,
+ * and then the number of message will request from XPC.
+ *
+ * XPC expects each message to exist in an individual cacheline.
+ */
+#define XPNET_MSG_SIZE		XPC_MSG_PAYLOAD_MAX_SIZE
+#define XPNET_MSG_DATA_MAX	\
+		(XPNET_MSG_SIZE - offsetof(struct xpnet_message, data))
+#define XPNET_MSG_NENTRIES	(PAGE_SIZE / XPC_MSG_MAX_SIZE)
+
+#define XPNET_MAX_KTHREADS	(XPNET_MSG_NENTRIES + 1)
+#define XPNET_MAX_IDLE_KTHREADS	(XPNET_MSG_NENTRIES + 1)
+
+/*
+ * Version number of XPNET implementation. XPNET can always talk to versions
+ * with same major #, and never talk to versions with a different version.
+ */
+#define _XPNET_VERSION(_major, _minor)	(((_major) << 4) | (_minor))
+#define XPNET_VERSION_MAJOR(_v)		((_v) >> 4)
+#define XPNET_VERSION_MINOR(_v)		((_v) & 0xf)
+
+#define	XPNET_VERSION _XPNET_VERSION(1, 0)	/* version 1.0 */
+#define	XPNET_VERSION_EMBED _XPNET_VERSION(1, 1)	/* version 1.1 */
+#define XPNET_MAGIC	0x88786984	/* "XNET" */
+
+#define XPNET_VALID_MSG(_m)						     \
+   ((XPNET_VERSION_MAJOR(_m->version) == XPNET_VERSION_MAJOR(XPNET_VERSION)) \
+    && (msg->magic == XPNET_MAGIC))
+
+#define XPNET_DEVICE_NAME		"xp0"
+
+/*
+ * When messages are queued with xpc_send_notify, a kmalloc'd buffer
+ * of the following type is passed as a notification cookie.  When the
+ * notification function is called, we use the cookie to decide
+ * whether all outstanding message sends have completed.  The skb can
+ * then be released.
+ */
+struct xpnet_pending_msg {
+	struct sk_buff *skb;
+	atomic_t use_count;
+};
+
+static struct net_device *xpnet_device;
+
+/*
+ * When we are notified of other partitions activating, we add them to
+ * our bitmask of partitions to which we broadcast.
+ */
+static unsigned long *xpnet_broadcast_partitions;
+/* protect above */
+static DEFINE_SPINLOCK(xpnet_broadcast_lock);
+
+/*
+ * Since the Block Transfer Engine (BTE) is being used for the transfer
+ * and it relies upon cache-line size transfers, we need to reserve at
+ * least one cache-line for head and tail alignment.  The BTE is
+ * limited to 8MB transfers.
+ *
+ * Testing has shown that changing MTU to greater than 64KB has no effect
+ * on TCP as the two sides negotiate a Max Segment Size that is limited
+ * to 64K.  Other protocols May use packets greater than this, but for
+ * now, the default is 64KB.
+ */
+#define XPNET_MAX_MTU (0x800000UL - L1_CACHE_BYTES)
+/* 68 comes from min TCP+IP+MAC header */
+#define XPNET_MIN_MTU 68
+/* 32KB has been determined to be the ideal */
+#define XPNET_DEF_MTU (0x8000UL)
+
+/*
+ * The partid is encapsulated in the MAC address beginning in the following
+ * octet and it consists of two octets.
+ */
+#define XPNET_PARTID_OCTET	2
+
+/* Define the XPNET debug device structures to be used with dev_dbg() et al */
+
+static struct device_driver xpnet_dbg_name = {
+	.name = "xpnet"
+};
+
+static struct device xpnet_dbg_subname = {
+	.init_name = "",	/* set to "" */
+	.driver = &xpnet_dbg_name
+};
+
+static struct device *xpnet = &xpnet_dbg_subname;
+
+/*
+ * Packet was recevied by XPC and forwarded to us.
+ */
+static void
+xpnet_receive(short partid, int channel, struct xpnet_message *msg)
+{
+	struct sk_buff *skb;
+	void *dst;
+	enum xp_retval ret;
+
+	if (!XPNET_VALID_MSG(msg)) {
+		/*
+		 * Packet with a different XPC version.  Ignore.
+		 */
+		xpc_received(partid, channel, (void *)msg);
+
+		xpnet_device->stats.rx_errors++;
+
+		return;
+	}
+	dev_dbg(xpnet, "received 0x%lx, %d, %d, %d\n", msg->buf_pa, msg->size,
+		msg->leadin_ignore, msg->tailout_ignore);
+
+	/* reserve an extra cache line */
+	skb = dev_alloc_skb(msg->size + L1_CACHE_BYTES);
+	if (!skb) {
+		dev_err(xpnet, "failed on dev_alloc_skb(%d)\n",
+			msg->size + L1_CACHE_BYTES);
+
+		xpc_received(partid, channel, (void *)msg);
+
+		xpnet_device->stats.rx_errors++;
+
+		return;
+	}
+
+	/*
+	 * The allocated skb has some reserved space.
+	 * In order to use xp_remote_memcpy(), we need to get the
+	 * skb->data pointer moved forward.
+	 */
+	skb_reserve(skb, (L1_CACHE_BYTES - ((u64)skb->data &
+					    (L1_CACHE_BYTES - 1)) +
+			  msg->leadin_ignore));
+
+	/*
+	 * Update the tail pointer to indicate data actually
+	 * transferred.
+	 */
+	skb_put(skb, (msg->size - msg->leadin_ignore - msg->tailout_ignore));
+
+	/*
+	 * Move the data over from the other side.
+	 */
+	if ((XPNET_VERSION_MINOR(msg->version) == 1) &&
+	    (msg->embedded_bytes != 0)) {
+		dev_dbg(xpnet, "copying embedded message. memcpy(0x%p, 0x%p, "
+			"%lu)\n", skb->data, &msg->data,
+			(size_t)msg->embedded_bytes);
+
+		skb_copy_to_linear_data(skb, &msg->data,
+					(size_t)msg->embedded_bytes);
+	} else {
+		dst = (void *)((u64)skb->data & ~(L1_CACHE_BYTES - 1));
+		dev_dbg(xpnet, "transferring buffer to the skb->data area;\n\t"
+			"xp_remote_memcpy(0x%p, 0x%p, %u)\n", dst,
+					  (void *)msg->buf_pa, msg->size);
+
+		ret = xp_remote_memcpy(xp_pa(dst), msg->buf_pa, msg->size);
+		if (ret != xpSuccess) {
+			/*
+			 * !!! Need better way of cleaning skb.  Currently skb
+			 * !!! appears in_use and we can't just call
+			 * !!! dev_kfree_skb.
+			 */
+			dev_err(xpnet, "xp_remote_memcpy(0x%p, 0x%p, 0x%x) "
+				"returned error=0x%x\n", dst,
+				(void *)msg->buf_pa, msg->size, ret);
+
+			xpc_received(partid, channel, (void *)msg);
+
+			xpnet_device->stats.rx_errors++;
+
+			return;
+		}
+	}
+
+	dev_dbg(xpnet, "<skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n", (void *)skb->head,
+		(void *)skb->data, skb_tail_pointer(skb), skb_end_pointer(skb),
+		skb->len);
+
+	skb->protocol = eth_type_trans(skb, xpnet_device);
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+	dev_dbg(xpnet, "passing skb to network layer\n"
+		"\tskb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n",
+		(void *)skb->head, (void *)skb->data, skb_tail_pointer(skb),
+		skb_end_pointer(skb), skb->len);
+
+	xpnet_device->stats.rx_packets++;
+	xpnet_device->stats.rx_bytes += skb->len + ETH_HLEN;
+
+	netif_rx(skb);
+	xpc_received(partid, channel, (void *)msg);
+}
+
+/*
+ * This is the handler which XPC calls during any sort of change in
+ * state or message reception on a connection.
+ */
+static void
+xpnet_connection_activity(enum xp_retval reason, short partid, int channel,
+			  void *data, void *key)
+{
+	DBUG_ON(partid < 0 || partid >= xp_max_npartitions);
+	DBUG_ON(channel != XPC_NET_CHANNEL);
+
+	switch (reason) {
+	case xpMsgReceived:	/* message received */
+		DBUG_ON(data == NULL);
+
+		xpnet_receive(partid, channel, (struct xpnet_message *)data);
+		break;
+
+	case xpConnected:	/* connection completed to a partition */
+		spin_lock_bh(&xpnet_broadcast_lock);
+		__set_bit(partid, xpnet_broadcast_partitions);
+		spin_unlock_bh(&xpnet_broadcast_lock);
+
+		netif_carrier_on(xpnet_device);
+
+		dev_dbg(xpnet, "%s connected to partition %d\n",
+			xpnet_device->name, partid);
+		break;
+
+	default:
+		spin_lock_bh(&xpnet_broadcast_lock);
+		__clear_bit(partid, xpnet_broadcast_partitions);
+		spin_unlock_bh(&xpnet_broadcast_lock);
+
+		if (bitmap_empty(xpnet_broadcast_partitions,
+				 xp_max_npartitions)) {
+			netif_carrier_off(xpnet_device);
+		}
+
+		dev_dbg(xpnet, "%s disconnected from partition %d\n",
+			xpnet_device->name, partid);
+		break;
+	}
+}
+
+static int
+xpnet_dev_open(struct net_device *dev)
+{
+	enum xp_retval ret;
+
+	dev_dbg(xpnet, "calling xpc_connect(%d, 0x%p, NULL, %ld, %ld, %ld, "
+		"%ld)\n", XPC_NET_CHANNEL, xpnet_connection_activity,
+		(unsigned long)XPNET_MSG_SIZE,
+		(unsigned long)XPNET_MSG_NENTRIES,
+		(unsigned long)XPNET_MAX_KTHREADS,
+		(unsigned long)XPNET_MAX_IDLE_KTHREADS);
+
+	ret = xpc_connect(XPC_NET_CHANNEL, xpnet_connection_activity, NULL,
+			  XPNET_MSG_SIZE, XPNET_MSG_NENTRIES,
+			  XPNET_MAX_KTHREADS, XPNET_MAX_IDLE_KTHREADS);
+	if (ret != xpSuccess) {
+		dev_err(xpnet, "ifconfig up of %s failed on XPC connect, "
+			"ret=%d\n", dev->name, ret);
+
+		return -ENOMEM;
+	}
+
+	dev_dbg(xpnet, "ifconfig up of %s; XPC connected\n", dev->name);
+
+	return 0;
+}
+
+static int
+xpnet_dev_stop(struct net_device *dev)
+{
+	xpc_disconnect(XPC_NET_CHANNEL);
+
+	dev_dbg(xpnet, "ifconfig down of %s; XPC disconnected\n", dev->name);
+
+	return 0;
+}
+
+/*
+ * Notification that the other end has received the message and
+ * DMA'd the skb information.  At this point, they are done with
+ * our side.  When all recipients are done processing, we
+ * release the skb and then release our pending message structure.
+ */
+static void
+xpnet_send_completed(enum xp_retval reason, short partid, int channel,
+		     void *__qm)
+{
+	struct xpnet_pending_msg *queued_msg = (struct xpnet_pending_msg *)__qm;
+
+	DBUG_ON(queued_msg == NULL);
+
+	dev_dbg(xpnet, "message to %d notified with reason %d\n",
+		partid, reason);
+
+	if (atomic_dec_return(&queued_msg->use_count) == 0) {
+		dev_dbg(xpnet, "all acks for skb->head=-x%p\n",
+			(void *)queued_msg->skb->head);
+
+		dev_kfree_skb_any(queued_msg->skb);
+		kfree(queued_msg);
+	}
+}
+
+static void
+xpnet_send(struct sk_buff *skb, struct xpnet_pending_msg *queued_msg,
+	   u64 start_addr, u64 end_addr, u16 embedded_bytes, int dest_partid)
+{
+	u8 msg_buffer[XPNET_MSG_SIZE];
+	struct xpnet_message *msg = (struct xpnet_message *)&msg_buffer;
+	u16 msg_size = sizeof(struct xpnet_message);
+	enum xp_retval ret;
+
+	msg->embedded_bytes = embedded_bytes;
+	if (unlikely(embedded_bytes != 0)) {
+		msg->version = XPNET_VERSION_EMBED;
+		dev_dbg(xpnet, "calling memcpy(0x%p, 0x%p, 0x%lx)\n",
+			&msg->data, skb->data, (size_t)embedded_bytes);
+		skb_copy_from_linear_data(skb, &msg->data,
+					  (size_t)embedded_bytes);
+		msg_size += embedded_bytes - 1;
+	} else {
+		msg->version = XPNET_VERSION;
+	}
+	msg->magic = XPNET_MAGIC;
+	msg->size = end_addr - start_addr;
+	msg->leadin_ignore = (u64)skb->data - start_addr;
+	msg->tailout_ignore = end_addr - (u64)skb_tail_pointer(skb);
+	msg->buf_pa = xp_pa((void *)start_addr);
+
+	dev_dbg(xpnet, "sending XPC message to %d:%d\n"
+		"msg->buf_pa=0x%lx, msg->size=%u, "
+		"msg->leadin_ignore=%u, msg->tailout_ignore=%u\n",
+		dest_partid, XPC_NET_CHANNEL, msg->buf_pa, msg->size,
+		msg->leadin_ignore, msg->tailout_ignore);
+
+	atomic_inc(&queued_msg->use_count);
+
+	ret = xpc_send_notify(dest_partid, XPC_NET_CHANNEL, XPC_NOWAIT, msg,
+			      msg_size, xpnet_send_completed, queued_msg);
+	if (unlikely(ret != xpSuccess))
+		atomic_dec(&queued_msg->use_count);
+}
+
+/*
+ * Network layer has formatted a packet (skb) and is ready to place it
+ * "on the wire".  Prepare and send an xpnet_message to all partitions
+ * which have connected with us and are targets of this packet.
+ *
+ * MAC-NOTE:  For the XPNET driver, the MAC address contains the
+ * destination partid.  If the destination partid octets are 0xffff,
+ * this packet is to be broadcast to all connected partitions.
+ */
+static netdev_tx_t
+xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct xpnet_pending_msg *queued_msg;
+	u64 start_addr, end_addr;
+	short dest_partid;
+	u16 embedded_bytes = 0;
+
+	dev_dbg(xpnet, ">skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n", (void *)skb->head,
+		(void *)skb->data, skb_tail_pointer(skb), skb_end_pointer(skb),
+		skb->len);
+
+	if (skb->data[0] == 0x33) {
+		dev_kfree_skb(skb);
+		return NETDEV_TX_OK;	/* nothing needed to be done */
+	}
+
+	/*
+	 * The xpnet_pending_msg tracks how many outstanding
+	 * xpc_send_notifies are relying on this skb.  When none
+	 * remain, release the skb.
+	 */
+	queued_msg = kmalloc(sizeof(struct xpnet_pending_msg), GFP_ATOMIC);
+	if (queued_msg == NULL) {
+		dev_warn(xpnet, "failed to kmalloc %ld bytes; dropping "
+			 "packet\n", sizeof(struct xpnet_pending_msg));
+
+		dev->stats.tx_errors++;
+		dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	/* get the beginning of the first cacheline and end of last */
+	start_addr = ((u64)skb->data & ~(L1_CACHE_BYTES - 1));
+	end_addr = L1_CACHE_ALIGN((u64)skb_tail_pointer(skb));
+
+	/* calculate how many bytes to embed in the XPC message */
+	if (unlikely(skb->len <= XPNET_MSG_DATA_MAX)) {
+		/* skb->data does fit so embed */
+		embedded_bytes = skb->len;
+	}
+
+	/*
+	 * Since the send occurs asynchronously, we set the count to one
+	 * and begin sending.  Any sends that happen to complete before
+	 * we are done sending will not free the skb.  We will be left
+	 * with that task during exit.  This also handles the case of
+	 * a packet destined for a partition which is no longer up.
+	 */
+	atomic_set(&queued_msg->use_count, 1);
+	queued_msg->skb = skb;
+
+	if (skb->data[0] == 0xff) {
+		/* we are being asked to broadcast to all partitions */
+		for_each_set_bit(dest_partid, xpnet_broadcast_partitions,
+			     xp_max_npartitions) {
+
+			xpnet_send(skb, queued_msg, start_addr, end_addr,
+				   embedded_bytes, dest_partid);
+		}
+	} else {
+		dest_partid = (short)skb->data[XPNET_PARTID_OCTET + 1];
+		dest_partid |= (short)skb->data[XPNET_PARTID_OCTET + 0] << 8;
+
+		if (dest_partid >= 0 &&
+		    dest_partid < xp_max_npartitions &&
+		    test_bit(dest_partid, xpnet_broadcast_partitions) != 0) {
+
+			xpnet_send(skb, queued_msg, start_addr, end_addr,
+				   embedded_bytes, dest_partid);
+		}
+	}
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += skb->len;
+
+	if (atomic_dec_return(&queued_msg->use_count) == 0) {
+		dev_kfree_skb(skb);
+		kfree(queued_msg);
+	}
+
+	return NETDEV_TX_OK;
+}
+
+/*
+ * Deal with transmit timeouts coming from the network layer.
+ */
+static void
+xpnet_dev_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	dev->stats.tx_errors++;
+}
+
+static const struct net_device_ops xpnet_netdev_ops = {
+	.ndo_open		= xpnet_dev_open,
+	.ndo_stop		= xpnet_dev_stop,
+	.ndo_start_xmit		= xpnet_dev_hard_start_xmit,
+	.ndo_tx_timeout		= xpnet_dev_tx_timeout,
+	.ndo_set_mac_address 	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+};
+
+static int __init
+xpnet_init(void)
+{
+	u8 addr[ETH_ALEN];
+	int result;
+
+	if (!is_uv_system())
+		return -ENODEV;
+
+	dev_info(xpnet, "registering network device %s\n", XPNET_DEVICE_NAME);
+
+	xpnet_broadcast_partitions = bitmap_zalloc(xp_max_npartitions,
+						   GFP_KERNEL);
+	if (xpnet_broadcast_partitions == NULL)
+		return -ENOMEM;
+
+	/*
+	 * use ether_setup() to init the majority of our device
+	 * structure and then override the necessary pieces.
+	 */
+	xpnet_device = alloc_netdev(0, XPNET_DEVICE_NAME, NET_NAME_UNKNOWN,
+				    ether_setup);
+	if (xpnet_device == NULL) {
+		bitmap_free(xpnet_broadcast_partitions);
+		return -ENOMEM;
+	}
+
+	netif_carrier_off(xpnet_device);
+
+	xpnet_device->netdev_ops = &xpnet_netdev_ops;
+	xpnet_device->mtu = XPNET_DEF_MTU;
+	xpnet_device->min_mtu = XPNET_MIN_MTU;
+	xpnet_device->max_mtu = XPNET_MAX_MTU;
+
+	memset(addr, 0, sizeof(addr));
+	/*
+	 * Multicast assumes the LSB of the first octet is set for multicast
+	 * MAC addresses.  We chose the first octet of the MAC to be unlikely
+	 * to collide with any vendor's officially issued MAC.
+	 */
+	addr[0] = 0x02;     /* locally administered, no OUI */
+
+	addr[XPNET_PARTID_OCTET + 1] = xp_partition_id;
+	addr[XPNET_PARTID_OCTET + 0] = (xp_partition_id >> 8);
+	eth_hw_addr_set(xpnet_device, addr);
+
+	/*
+	 * ether_setup() sets this to a multicast device.  We are
+	 * really not supporting multicast at this time.
+	 */
+	xpnet_device->flags &= ~IFF_MULTICAST;
+
+	/*
+	 * No need to checksum as it is a DMA transfer.  The BTE will
+	 * report an error if the data is not retrievable and the
+	 * packet will be dropped.
+	 */
+	xpnet_device->features = NETIF_F_HW_CSUM;
+
+	result = register_netdev(xpnet_device);
+	if (result != 0) {
+		free_netdev(xpnet_device);
+		bitmap_free(xpnet_broadcast_partitions);
+	}
+
+	return result;
+}
+
+module_init(xpnet_init);
+
+static void __exit
+xpnet_exit(void)
+{
+	dev_info(xpnet, "unregistering network device %s\n",
+		 xpnet_device[0].name);
+
+	unregister_netdev(xpnet_device);
+	free_netdev(xpnet_device);
+	bitmap_free(xpnet_broadcast_partitions);
+}
+
+module_exit(xpnet_exit);
+
+MODULE_AUTHOR("Silicon Graphics, Inc.");
+MODULE_DESCRIPTION("Cross Partition Network adapter (XPNET)");
+MODULE_LICENSE("GPL");