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/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     int mq_required = 0;
  45
  46     TFR(fd = open(PATH_NET_TUN, O_RDWR));
  47     if (fd < 0) {
  48         error_report("could not open %s: %m", PATH_NET_TUN);
  49         return -1;
  50     }
  51     memset(&ifr, 0, sizeof(ifr));
  52     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
  53
  54     if (*vnet_hdr) {
  55         unsigned int features;
  56
  57         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
  58             features & IFF_VNET_HDR) {
  59             *vnet_hdr = 1;
  60             ifr.ifr_flags |= IFF_VNET_HDR;
  61         } else {
  62             *vnet_hdr = 0;
  63         }
  64
  65         if (vnet_hdr_required && !*vnet_hdr) {
  66             error_report("vnet_hdr=1 requested, but no kernel "
  67                          "support for IFF_VNET_HDR available");
  68             close(fd);
  69             return -1;
  70         }
  71         /*
  72          * Make sure vnet header size has the default value: for a persistent
  73          * tap it might have been modified e.g. by another instance of qemu.
  74          * Ignore errors since old kernels do not support this ioctl: in this
  75          * case the header size implicitly has the correct value.
  76          */
  77         ioctl(fd, TUNSETVNETHDRSZ, &len);
  78     }
  79
  80     if (mq_required) {
  81         unsigned int features;
  82
  83         if ((ioctl(fd, TUNGETFEATURES, &features) != 0) ||
  84             !(features & IFF_MULTI_QUEUE)) {
  85             error_report("multiqueue required, but no kernel "
  86                          "support for IFF_MULTI_QUEUE available");
  87             close(fd);
  88             return -1;
  89         } else {
  90             ifr.ifr_flags |= IFF_MULTI_QUEUE;
  91         }
  92     }
  93
  94     if (ifname[0] != '\0')
  95         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
  96     else
  97         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
  98     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
  99     if (ret != 0) {
 100         if (ifname[0] != '\0') {
 101             error_report("could not configure %s (%s): %m", PATH_NET_TUN, ifr.ifr_name);
 102         } else {
 103             error_report("could not configure %s: %m", PATH_NET_TUN);
 104         }
 105         close(fd);
 106         return -1;
 107     }
 108     pstrcpy(ifname, ifname_size, ifr.ifr_name);
 109     fcntl(fd, F_SETFL, O_NONBLOCK);
 110     return fd;
 111 }
 112
 113 /* sndbuf implements a kind of flow control for tap.
 114  * Unfortunately when it's enabled, and packets are sent
 115  * to other guests on the same host, the receiver
 116  * can lock up the transmitter indefinitely.
 117  *
 118  * To avoid packet loss, sndbuf should be set to a value lower than the tx
 119  * queue capacity of any destination network interface.
 120  * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
 121  * a good value, given a 1500 byte MTU.
 122  */
 123 #define TAP_DEFAULT_SNDBUF 0
 124
 125 int tap_set_sndbuf(int fd, const NetdevTapOptions *tap)
 126 {
 127     int sndbuf;
 128
 129     sndbuf = !tap->has_sndbuf       ? TAP_DEFAULT_SNDBUF :
 130              tap->sndbuf > INT_MAX  ? INT_MAX :
 131              tap->sndbuf;
 132
 133     if (!sndbuf) {
 134         sndbuf = INT_MAX;
 135     }
 136
 137     if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
 138         error_report("TUNSETSNDBUF ioctl failed: %s", strerror(errno));
 139         return -1;
 140     }
 141     return 0;
 142 }
 143
 144 int tap_probe_vnet_hdr(int fd)
 145 {
 146     struct ifreq ifr;
 147
 148     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 149         error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
 150         return 0;
 151     }
 152
 153     return ifr.ifr_flags & IFF_VNET_HDR;
 154 }
 155
 156 int tap_probe_has_ufo(int fd)
 157 {
 158     unsigned offload;
 159
 160     offload = TUN_F_CSUM | TUN_F_UFO;
 161
 162     if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
 163         return 0;
 164
 165     return 1;
 166 }
 167
 168 /* Verify that we can assign given length */
 169 int tap_probe_vnet_hdr_len(int fd, int len)
 170 {
 171     int orig;
 172     if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
 173         return 0;
 174     }
 175     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 176         return 0;
 177     }
 178     /* Restore original length: we can't handle failure. */
 179     if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
 180         fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 181                 strerror(errno));
 182         abort();
 183         return -errno;
 184     }
 185     return 1;
 186 }
 187
 188 void tap_fd_set_vnet_hdr_len(int fd, int len)
 189 {
 190     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 191         fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 192                 strerror(errno));
 193         abort();
 194     }
 195 }
 196
 197 void tap_fd_set_offload(int fd, int csum, int tso4,
 198                         int tso6, int ecn, int ufo)
 199 {
 200     unsigned int offload = 0;
 201
 202     /* Check if our kernel supports TUNSETOFFLOAD */
 203     if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
 204         return;
 205     }
 206
 207     if (csum) {
 208         offload |= TUN_F_CSUM;
 209         if (tso4)
 210             offload |= TUN_F_TSO4;
 211         if (tso6)
 212             offload |= TUN_F_TSO6;
 213         if ((tso4 || tso6) && ecn)
 214             offload |= TUN_F_TSO_ECN;
 215         if (ufo)
 216             offload |= TUN_F_UFO;
 217     }
 218
 219     if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 220         offload &= ~TUN_F_UFO;
 221         if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 222             fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 223                     strerror(errno));
 224         }
 225     }
 226 }
 227
 228 /* Enable a specific queue of tap. */
 229 int tap_fd_enable(int fd)
 230 {
 231     struct ifreq ifr;
 232     int ret;
 233
 234     memset(&ifr, 0, sizeof(ifr));
 235
 236     ifr.ifr_flags = IFF_ATTACH_QUEUE;
 237     ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
 238
 239     if (ret != 0) {
 240         error_report("could not enable queue");
 241     }
 242
 243     return ret;
 244 }
 245
 246 /* Disable a specific queue of tap/ */
 247 int tap_fd_disable(int fd)
 248 {
 249     struct ifreq ifr;
 250     int ret;
 251
 252     memset(&ifr, 0, sizeof(ifr));
 253
 254     ifr.ifr_flags = IFF_DETACH_QUEUE;
 255     ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
 256
 257     if (ret != 0) {
 258         error_report("could not disable queue");
 259     }
 260
 261     return ret;
 262 }
 263