[af_packet] Add new AF_PACKET driver for Linux

This code largely inspired by tap.c.  Allows for testing iPXE on real
NICs from within Linux.  For example:

  make bin-x86_64-linux/af_packet.linux
  valgrind ./bin-x86_64-linux/af_packet.linux --net af_packet,if=eth3

Tested as x86_64 and i386 binary.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/arch/x86/core/linux/linux_api.c

@@ -108,3 +108,42 @@ void * linux_mremap ( void *old_address, __kernel_size_t old_size,
 int linux_munmap ( void *addr, __kernel_size_t length ) {
 	return linux_syscall ( __NR_munmap, addr, length );
 }
+
+int linux_socket ( int domain, int type_, int protocol ) {
+#ifdef __NR_socket
+	return linux_syscall ( __NR_socket, domain, type_, protocol );
+#else
+#ifndef SOCKOP_socket
+# define SOCKOP_socket 1
+#endif
+	unsigned long sc_args[] = { domain, type_, protocol };
+	return linux_syscall ( __NR_socketcall, SOCKOP_socket, sc_args );
+#endif
+}
+
+int linux_bind ( int fd, const struct sockaddr *addr, socklen_t addrlen ) {
+#ifdef __NR_bind
+	return linux_syscall ( __NR_bind, fd, addr, addrlen );
+#else
+#ifndef SOCKOP_bind
+# define SOCKOP_bind 2
+#endif
+	unsigned long sc_args[] = { fd, (unsigned long)addr, addrlen };
+	return linux_syscall ( __NR_socketcall, SOCKOP_bind, sc_args );
+#endif
+}
+
+ssize_t linux_sendto ( int fd, const void *buf, size_t len, int flags,
+		       const struct sockaddr *daddr, socklen_t addrlen ) {
+#ifdef __NR_sendto
+	return linux_syscall ( __NR_sendto, fd, buf, len, flags,
+			       daddr, addrlen );
+#else
+#ifndef SOCKOP_sendto
+# define SOCKOP_sendto 11
+#endif
+	unsigned long sc_args[] = { fd, (unsigned long)buf, len,
+				    flags, (unsigned long)daddr, addrlen };
+	return linux_syscall ( __NR_socketcall, SOCKOP_sendto, sc_args );
+#endif
+}

src/drivers/linux/af_packet.c

@@ -0,0 +1,325 @@
+/*
+ * Copyright (C) 2016 David Decotigny <ddecotig@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#include <errno.h>
+#include <string.h>
+#include <stdio.h>
+#include <linux_api.h>
+#include <ipxe/list.h>
+#include <ipxe/linux.h>
+#include <ipxe/malloc.h>
+#include <ipxe/device.h>
+#include <ipxe/netdevice.h>
+#include <ipxe/iobuf.h>
+#include <ipxe/ethernet.h>
+#include <ipxe/settings.h>
+#include <ipxe/socket.h>
+
+/* This hack prevents pre-2.6.32 headers from redefining struct sockaddr */
+#define __GLIBC__ 2
+#include <linux/socket.h>
+#include <linux/if.h>
+#include <linux/if_ether.h>
+#include <linux/if_packet.h>
+#undef __GLIBC__
+#include <byteswap.h>
+
+/* linux-specifc syscall params */
+#define LINUX_AF_PACKET 17
+#define LINUX_SOCK_RAW 3
+#define LINUX_SIOCGIFINDEX 0x8933
+#define LINUX_SIOCGIFHWADDR 0x8927
+
+#define RX_BUF_SIZE 1536
+
+/** @file
+ *
+ * The AF_PACKET driver.
+ *
+ * Bind to an existing linux network interface.
+ */
+
+struct af_packet_nic {
+	/** Linux network interface name */
+	char * ifname;
+	/** Packet socket descriptor */
+	int fd;
+	/** ifindex */
+	int ifindex;
+};
+
+/** Open the linux interface */
+static int af_packet_nic_open ( struct net_device * netdev )
+{
+	struct af_packet_nic * nic = netdev->priv;
+	struct sockaddr_ll socket_address;
+	struct ifreq if_data;
+	int ret;
+
+	nic->fd = linux_socket(LINUX_AF_PACKET, LINUX_SOCK_RAW,
+			       htons(ETH_P_ALL));
+	if (nic->fd < 0) {
+		DBGC(nic, "af_packet %p socket(AF_PACKET) = %d (%s)\n",
+		     nic, nic->fd, linux_strerror(linux_errno));
+		return nic->fd;
+	}
+
+	/* resolve ifindex of ifname */
+	memset(&if_data, 0, sizeof(if_data));
+	strncpy(if_data.ifr_name, nic->ifname, sizeof(if_data.ifr_name));
+	ret = linux_ioctl(nic->fd, LINUX_SIOCGIFINDEX, &if_data);
+	if (ret < 0) {
+		DBGC(nic, "af_packet %p ioctl(SIOCGIFINDEX) = %d (%s)\n",
+		     nic, ret, linux_strerror(linux_errno));
+		linux_close(nic->fd);
+		return ret;
+	}
+
+	nic->ifindex = if_data.ifr_ifindex;
+
+	/* bind to interface */
+	memset(&socket_address, 0, sizeof(socket_address));
+	socket_address.sll_family = LINUX_AF_PACKET;
+	socket_address.sll_ifindex = nic->ifindex;
+	socket_address.sll_protocol = htons(ETH_P_ALL);
+	ret = linux_bind(nic->fd, (void *) &socket_address,
+			 sizeof(socket_address));
+	if (ret == -1) {
+		DBGC(nic, "af_packet %p bind() = %d (%s)\n",
+		     nic, ret, linux_strerror(linux_errno));
+		linux_close(nic->fd);
+		return ret;
+	}
+
+	/* Set nonblocking mode to make af_packet_nic_poll() easier */
+	ret = linux_fcntl(nic->fd, F_SETFL, O_NONBLOCK);
+	if (ret != 0) {
+		DBGC(nic, "af_packet %p fcntl(%d, ...) = %d (%s)\n",
+		     nic, nic->fd, ret, linux_strerror(linux_errno));
+		linux_close(nic->fd);
+		return ret;
+	}
+
+	return 0;
+}
+
+/** Close the packet socket */
+static void af_packet_nic_close ( struct net_device *netdev )
+{
+	struct af_packet_nic * nic = netdev->priv;
+	linux_close(nic->fd);
+}
+
+/**
+ * Transmit an ethernet packet.
+ *
+ * The packet can be written to the socket and marked as complete immediately.
+ */
+static int af_packet_nic_transmit ( struct net_device *netdev,
+				    struct io_buffer *iobuf )
+{
+	struct af_packet_nic * nic = netdev->priv;
+	struct sockaddr_ll socket_address;
+	const struct ethhdr * eh;
+	int rc;
+
+	memset(&socket_address, 0, sizeof(socket_address));
+	socket_address.sll_family = LINUX_AF_PACKET;
+	socket_address.sll_ifindex = nic->ifindex;
+	socket_address.sll_halen = ETH_ALEN;
+
+	eh = iobuf->data;
+	memcpy(socket_address.sll_addr, eh->h_dest, ETH_ALEN);
+
+	rc = linux_sendto(nic->fd, iobuf->data, iobuf->tail - iobuf->data,
+			  0, (struct sockaddr *)&socket_address,
+			  sizeof(socket_address));
+
+	DBGC2(nic, "af_packet %p wrote %d bytes\n", nic, rc);
+	netdev_tx_complete(netdev, iobuf);
+
+	return 0;
+}
+
+/** Poll for new packets */
+static void af_packet_nic_poll ( struct net_device *netdev )
+{
+	struct af_packet_nic * nic = netdev->priv;
+	struct pollfd pfd;
+	struct io_buffer * iobuf;
+	int r;
+
+	pfd.fd = nic->fd;
+	pfd.events = POLLIN;
+	if (linux_poll(&pfd, 1, 0) == -1) {
+		DBGC(nic, "af_packet %p poll failed (%s)\n",
+		     nic, linux_strerror(linux_errno));
+		return;
+	}
+	if ((pfd.revents & POLLIN) == 0)
+		return;
+
+	/* At this point we know there is at least one new packet to be read */
+
+	iobuf = alloc_iob(RX_BUF_SIZE);
+	if (! iobuf)
+		goto allocfail;
+
+	while ((r = linux_read(nic->fd, iobuf->data, RX_BUF_SIZE)) > 0) {
+		DBGC2(nic, "af_packet %p read %d bytes\n", nic, r);
+
+		iob_put(iobuf, r);
+		netdev_rx(netdev, iobuf);
+
+		iobuf = alloc_iob(RX_BUF_SIZE);
+		if (! iobuf)
+			goto allocfail;
+	}
+
+	free_iob(iobuf);
+	return;
+
+allocfail:
+	DBGC(nic, "af_packet %p alloc_iob failed\n", nic);
+}
+
+/**
+ * Set irq.
+ *
+ * Not used on linux, provide a dummy implementation.
+ */
+static void af_packet_nic_irq ( struct net_device *netdev, int enable )
+{
+	struct af_packet_nic *nic = netdev->priv;
+
+	DBGC(nic, "af_packet %p irq enable = %d\n", nic, enable);
+}
+
+
+static int af_packet_update_properties ( struct net_device *netdev )
+{
+	struct af_packet_nic *nic = netdev->priv;
+	struct ifreq if_data;
+	int ret;
+
+	/* retrieve default MAC address */
+	int fd = linux_socket(LINUX_AF_PACKET, LINUX_SOCK_RAW, 0);
+	if (fd < 0) {
+		DBGC(nic, "af_packet %p cannot create raw socket (%s)\n",
+		     nic, linux_strerror(linux_errno));
+		return fd;
+	}
+
+	/* retrieve host's MAC address */
+	memset(&if_data, 0, sizeof(if_data));
+	strncpy(if_data.ifr_name, nic->ifname, sizeof(if_data.ifr_name));
+	ret = linux_ioctl(fd, LINUX_SIOCGIFHWADDR, &if_data);
+	if (ret < 0) {
+		DBGC(nic, "af_packet %p cannot get mac addr (%s)\n",
+		     nic, linux_strerror(linux_errno));
+		linux_close(fd);
+		return ret;
+	}
+
+	linux_close(fd);
+	/* struct sockaddr = { u16 family, u8 pad[14] (equiv. sa_data) }; */
+	memcpy(netdev->ll_addr, if_data.ifr_hwaddr.pad, ETH_ALEN);
+	return 0;
+}
+
+/** AF_PACKET operations */
+static struct net_device_operations af_packet_nic_operations = {
+	.open		= af_packet_nic_open,
+	.close		= af_packet_nic_close,
+	.transmit	= af_packet_nic_transmit,
+	.poll		= af_packet_nic_poll,
+	.irq		= af_packet_nic_irq,
+};
+
+/** Handle a device request for the af_packet driver */
+static int af_packet_nic_probe ( struct linux_device *device,
+				 struct linux_device_request *request )
+{
+	struct linux_setting *if_setting;
+	struct net_device *netdev;
+	struct af_packet_nic *nic;
+	int rc;
+
+	netdev = alloc_etherdev(sizeof(*nic));
+	if (! netdev)
+		return -ENOMEM;
+
+	netdev_init(netdev, &af_packet_nic_operations);
+	nic = netdev->priv;
+	linux_set_drvdata(device, netdev);
+	netdev->dev = &device->dev;
+
+	memset(nic, 0, sizeof(*nic));
+
+	/* Look for the mandatory if setting */
+	if_setting = linux_find_setting("if", &request->settings);
+
+	/* No if setting */
+	if (! if_setting) {
+		printf("af_packet missing a mandatory if setting\n");
+		rc = -EINVAL;
+		goto err_settings;
+	}
+
+	nic->ifname = if_setting->value;
+	snprintf ( device->dev.name, sizeof ( device->dev.name ), "%s",
+		   nic->ifname );
+	device->dev.desc.bus_type = BUS_TYPE_TAP;
+	af_packet_update_properties(netdev);
+	if_setting->applied = 1;
+
+	/* Apply rest of the settings */
+	linux_apply_settings(&request->settings, &netdev->settings.settings);
+
+	/* Register network device */
+	if ((rc = register_netdev(netdev)) != 0)
+		goto err_register;
+
+	netdev_link_up(netdev);
+
+	return 0;
+
+err_settings:
+	unregister_netdev(netdev);
+err_register:
+	netdev_nullify(netdev);
+	netdev_put(netdev);
+	return rc;
+}
+
+/** Remove the device */
+static void af_packet_nic_remove ( struct linux_device *device )
+{
+	struct net_device *netdev = linux_get_drvdata(device);
+	unregister_netdev(netdev);
+	netdev_nullify(netdev);
+	netdev_put(netdev);
+}
+
+/** AF_PACKET linux_driver */
+struct linux_driver af_packet_nic_driver __linux_driver = {
+	.name = "af_packet",
+	.probe = af_packet_nic_probe,
+	.remove = af_packet_nic_remove,
+	.can_probe = 1,
+};

src/include/ipxe/errfile.h

@@ -194,6 +194,7 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 #define ERRFILE_pciea		     ( ERRFILE_DRIVER | 0x00c00000 )
 #define ERRFILE_axge		     ( ERRFILE_DRIVER | 0x00c10000 )
 #define ERRFILE_thunderx	     ( ERRFILE_DRIVER | 0x00c20000 )
+#define ERRFILE_af_packet	     ( ERRFILE_DRIVER | 0x00c30000 )
 
 #define ERRFILE_aoe			( ERRFILE_NET | 0x00000000 )
 #define ERRFILE_arp			( ERRFILE_NET | 0x00010000 )

src/include/linux_api.h

@@ -46,6 +46,8 @@ typedef __kernel_loff_t loff_t;
 #include <linux/poll.h>
 typedef unsigned long nfds_t;
 typedef uint32_t useconds_t;
+typedef uint32_t socklen_t;
+struct sockaddr;
 #define MAP_FAILED ( ( void * ) -1 )
 #define SEEK_SET 0
 
@@ -68,6 +70,11 @@ extern void * linux_mmap ( void *addr, __kernel_size_t length, int prot,
 extern void * linux_mremap ( void *old_address, __kernel_size_t old_size,
 			     __kernel_size_t new_size, int flags );
 extern int linux_munmap ( void *addr, __kernel_size_t length );
+extern int linux_socket ( int domain, int type_, int protocol );
+extern int linux_bind ( int fd, const struct sockaddr *addr,
+			socklen_t addrlen );
+extern ssize_t linux_sendto ( int fd, const void *buf, size_t len, int flags,
+			      const struct sockaddr *daddr, socklen_t addrlen );
 
 extern const char * linux_strerror ( int errnum );