net/tap-linux.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  * Copyright (c) 2009 Red Hat, Inc.
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a copy
   8  * of this software and associated documentation files (the "Software"), to deal
   9  * in the Software without restriction, including without limitation the rights
  10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  11  * copies of the Software, and to permit persons to whom the Software is
  12  * furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included in
  15  * all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  23  * THE SOFTWARE.
  24  */
  25
  26 #include "net/tap.h"
  27 #include "net/tap-linux.h"
  28
  29 #include <net/if.h>
  30 #include <sys/ioctl.h>
  31
  32 #include "sysemu.h"
  33 #include "qemu-common.h"
  34 #include "qemu-error.h"
  35
  36 #define PATH_NET_TUN "/dev/net/tun"
  37
  38 int tap_open(char *ifname, int ifname_size, int *vnet_hdr, int vnet_hdr_required)
  39 {
  40     struct ifreq ifr;
  41     int fd, ret;
  42     int len = sizeof(struct virtio_net_hdr);
  43
  44     TFR(fd = open(PATH_NET_TUN, O_RDWR));
  45     if (fd < 0) {
  46         error_report("could not open %s: %m", PATH_NET_TUN);
  47         return -1;
  48     }
  49     memset(&ifr, 0, sizeof(ifr));
  50     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
  51
  52     if (*vnet_hdr) {
  53         unsigned int features;
  54
  55         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
  56             features & IFF_VNET_HDR) {
  57             *vnet_hdr = 1;
  58             ifr.ifr_flags |= IFF_VNET_HDR;
  59         } else {
  60             *vnet_hdr = 0;
  61         }
  62
  63         if (vnet_hdr_required && !*vnet_hdr) {
  64             error_report("vnet_hdr=1 requested, but no kernel "
  65                          "support for IFF_VNET_HDR available");
  66             close(fd);
  67             return -1;
  68         }
  69         /*
  70          * Make sure vnet header size has the default value: for a persistent
  71          * tap it might have been modified e.g. by another instance of qemu.
  72          * Ignore errors since old kernels do not support this ioctl: in this
  73          * case the header size implicitly has the correct value.
  74          */
  75         ioctl(fd, TUNSETVNETHDRSZ, &len);
  76     }
  77
  78     if (ifname[0] != '\0')
  79         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
  80     else
  81         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
  82     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
  83     if (ret != 0) {
  84         if (ifname[0] != '\0') {
  85             error_report("could not configure %s (%s): %m", PATH_NET_TUN, ifr.ifr_name);
  86         } else {
  87             error_report("could not configure %s: %m", PATH_NET_TUN);
  88         }
  89         close(fd);
  90         return -1;
  91     }
  92     pstrcpy(ifname, ifname_size, ifr.ifr_name);
  93     fcntl(fd, F_SETFL, O_NONBLOCK);
  94     return fd;
  95 }
  96
  97 /* sndbuf implements a kind of flow control for tap.
  98  * Unfortunately when it's enabled, and packets are sent
  99  * to other guests on the same host, the receiver
 100  * can lock up the transmitter indefinitely.
 101  *
 102  * To avoid packet loss, sndbuf should be set to a value lower than the tx
 103  * queue capacity of any destination network interface.
 104  * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
 105  * a good value, given a 1500 byte MTU.
 106  */
 107 #define TAP_DEFAULT_SNDBUF 0
 108
 109 int tap_set_sndbuf(int fd, const NetdevTapOptions *tap)
 110 {
 111     int sndbuf;
 112
 113     sndbuf = !tap->has_sndbuf       ? TAP_DEFAULT_SNDBUF :
 114              tap->sndbuf > INT_MAX  ? INT_MAX :
 115              tap->sndbuf;
 116
 117     if (!sndbuf) {
 118         sndbuf = INT_MAX;
 119     }
 120
 121     if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
 122         error_report("TUNSETSNDBUF ioctl failed: %s", strerror(errno));
 123         return -1;
 124     }
 125     return 0;
 126 }
 127
 128 int tap_probe_vnet_hdr(int fd)
 129 {
 130     struct ifreq ifr;
 131
 132     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 133         error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
 134         return 0;
 135     }
 136
 137     return ifr.ifr_flags & IFF_VNET_HDR;
 138 }
 139
 140 int tap_probe_has_ufo(int fd)
 141 {
 142     unsigned offload;
 143
 144     offload = TUN_F_CSUM | TUN_F_UFO;
 145
 146     if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
 147         return 0;
 148
 149     return 1;
 150 }
 151
 152 /* Verify that we can assign given length */
 153 int tap_probe_vnet_hdr_len(int fd, int len)
 154 {
 155     int orig;
 156     if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
 157         return 0;
 158     }
 159     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 160         return 0;
 161     }
 162     /* Restore original length: we can't handle failure. */
 163     if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
 164         fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 165                 strerror(errno));
 166         assert(0);
 167         return -errno;
 168     }
 169     return 1;
 170 }
 171
 172 void tap_fd_set_vnet_hdr_len(int fd, int len)
 173 {
 174     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 175         fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 176                 strerror(errno));
 177         assert(0);
 178     }
 179 }
 180
 181 void tap_fd_set_offload(int fd, int csum, int tso4,
 182                         int tso6, int ecn, int ufo)
 183 {
 184     unsigned int offload = 0;
 185
 186     /* Check if our kernel supports TUNSETOFFLOAD */
 187     if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
 188         return;
 189     }
 190
 191     if (csum) {
 192         offload |= TUN_F_CSUM;
 193         if (tso4)
 194             offload |= TUN_F_TSO4;
 195         if (tso6)
 196             offload |= TUN_F_TSO6;
 197         if ((tso4 || tso6) && ecn)
 198             offload |= TUN_F_TSO_ECN;
 199         if (ufo)
 200             offload |= TUN_F_UFO;
 201     }
 202
 203     if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 204         offload &= ~TUN_F_UFO;
 205         if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 206             fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 207                     strerror(errno));
 208         }
 209     }
 210 }