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 "tap_int.h"
  27 #include "tap-linux.h"
  28 #include "net/tap.h"
  29
  30 #include <net/if.h>
  31 #include <sys/ioctl.h>
  32
  33 #include "sysemu.h"
  34 #include "qemu-common.h"
  35 #include "qemu/error-report.h"
  36
  37 #define PATH_NET_TUN "/dev/net/tun"
  38
  39 int tap_open(char *ifname, int ifname_size, int *vnet_hdr, int vnet_hdr_required)
  40 {
  41     struct ifreq ifr;
  42     int fd, ret;
  43     int len = sizeof(struct virtio_net_hdr);
  44
  45     TFR(fd = open(PATH_NET_TUN, O_RDWR));
  46     if (fd < 0) {
  47         error_report("could not open %s: %m", PATH_NET_TUN);
  48         return -1;
  49     }
  50     memset(&ifr, 0, sizeof(ifr));
  51     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
  52
  53     if (*vnet_hdr) {
  54         unsigned int features;
  55
  56         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
  57             features & IFF_VNET_HDR) {
  58             *vnet_hdr = 1;
  59             ifr.ifr_flags |= IFF_VNET_HDR;
  60         } else {
  61             *vnet_hdr = 0;
  62         }
  63
  64         if (vnet_hdr_required && !*vnet_hdr) {
  65             error_report("vnet_hdr=1 requested, but no kernel "
  66                          "support for IFF_VNET_HDR available");
  67             close(fd);
  68             return -1;
  69         }
  70         /*
  71          * Make sure vnet header size has the default value: for a persistent
  72          * tap it might have been modified e.g. by another instance of qemu.
  73          * Ignore errors since old kernels do not support this ioctl: in this
  74          * case the header size implicitly has the correct value.
  75          */
  76         ioctl(fd, TUNSETVNETHDRSZ, &len);
  77     }
  78
  79     if (ifname[0] != '\0')
  80         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
  81     else
  82         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
  83     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
  84     if (ret != 0) {
  85         if (ifname[0] != '\0') {
  86             error_report("could not configure %s (%s): %m", PATH_NET_TUN, ifr.ifr_name);
  87         } else {
  88             error_report("could not configure %s: %m", PATH_NET_TUN);
  89         }
  90         close(fd);
  91         return -1;
  92     }
  93     pstrcpy(ifname, ifname_size, ifr.ifr_name);
  94     fcntl(fd, F_SETFL, O_NONBLOCK);
  95     return fd;
  96 }
  97
  98 /* sndbuf implements a kind of flow control for tap.
  99  * Unfortunately when it's enabled, and packets are sent
 100  * to other guests on the same host, the receiver
 101  * can lock up the transmitter indefinitely.
 102  *
 103  * To avoid packet loss, sndbuf should be set to a value lower than the tx
 104  * queue capacity of any destination network interface.
 105  * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
 106  * a good value, given a 1500 byte MTU.
 107  */
 108 #define TAP_DEFAULT_SNDBUF 0
 109
 110 int tap_set_sndbuf(int fd, const NetdevTapOptions *tap)
 111 {
 112     int sndbuf;
 113
 114     sndbuf = !tap->has_sndbuf       ? TAP_DEFAULT_SNDBUF :
 115              tap->sndbuf > INT_MAX  ? INT_MAX :
 116              tap->sndbuf;
 117
 118     if (!sndbuf) {
 119         sndbuf = INT_MAX;
 120     }
 121
 122     if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
 123         error_report("TUNSETSNDBUF ioctl failed: %s", strerror(errno));
 124         return -1;
 125     }
 126     return 0;
 127 }
 128
 129 int tap_probe_vnet_hdr(int fd)
 130 {
 131     struct ifreq ifr;
 132
 133     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 134         error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
 135         return 0;
 136     }
 137
 138     return ifr.ifr_flags & IFF_VNET_HDR;
 139 }
 140
 141 int tap_probe_has_ufo(int fd)
 142 {
 143     unsigned offload;
 144
 145     offload = TUN_F_CSUM | TUN_F_UFO;
 146
 147     if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
 148         return 0;
 149
 150     return 1;
 151 }
 152
 153 /* Verify that we can assign given length */
 154 int tap_probe_vnet_hdr_len(int fd, int len)
 155 {
 156     int orig;
 157     if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
 158         return 0;
 159     }
 160     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 161         return 0;
 162     }
 163     /* Restore original length: we can't handle failure. */
 164     if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
 165         fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 166                 strerror(errno));
 167         assert(0);
 168         return -errno;
 169     }
 170     return 1;
 171 }
 172
 173 void tap_fd_set_vnet_hdr_len(int fd, int len)
 174 {
 175     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 176         fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 177                 strerror(errno));
 178         assert(0);
 179     }
 180 }
 181
 182 void tap_fd_set_offload(int fd, int csum, int tso4,
 183                         int tso6, int ecn, int ufo)
 184 {
 185     unsigned int offload = 0;
 186
 187     /* Check if our kernel supports TUNSETOFFLOAD */
 188     if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
 189         return;
 190     }
 191
 192     if (csum) {
 193         offload |= TUN_F_CSUM;
 194         if (tso4)
 195             offload |= TUN_F_TSO4;
 196         if (tso6)
 197             offload |= TUN_F_TSO6;
 198         if ((tso4 || tso6) && ecn)
 199             offload |= TUN_F_TSO_ECN;
 200         if (ufo)
 201             offload |= TUN_F_UFO;
 202     }
 203
 204     if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 205         offload &= ~TUN_F_UFO;
 206         if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 207             fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 208                     strerror(errno));
 209         }
 210     }
 211 }