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,
  40              int vnet_hdr_required, int mq_required)
  41 {
  42     struct ifreq ifr;
  43     int fd, ret;
  44     int len = sizeof(struct virtio_net_hdr);
  45     unsigned int features;
  46
  47     TFR(fd = open(PATH_NET_TUN, O_RDWR));
  48     if (fd < 0) {
  49         error_report("could not open %s: %m", PATH_NET_TUN);
  50         return -1;
  51     }
  52     memset(&ifr, 0, sizeof(ifr));
  53     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
  54
  55     if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
  56         features & IFF_ONE_QUEUE) {
  57         ifr.ifr_flags |= IFF_ONE_QUEUE;
  58     }
  59
  60     if (*vnet_hdr) {
  61         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
  62             features & IFF_VNET_HDR) {
  63             *vnet_hdr = 1;
  64             ifr.ifr_flags |= IFF_VNET_HDR;
  65         } else {
  66             *vnet_hdr = 0;
  67         }
  68
  69         if (vnet_hdr_required && !*vnet_hdr) {
  70             error_report("vnet_hdr=1 requested, but no kernel "
  71                          "support for IFF_VNET_HDR available");
  72             close(fd);
  73             return -1;
  74         }
  75         /*
  76          * Make sure vnet header size has the default value: for a persistent
  77          * tap it might have been modified e.g. by another instance of qemu.
  78          * Ignore errors since old kernels do not support this ioctl: in this
  79          * case the header size implicitly has the correct value.
  80          */
  81         ioctl(fd, TUNSETVNETHDRSZ, &len);
  82     }
  83
  84     if (mq_required) {
  85         if ((ioctl(fd, TUNGETFEATURES, &features) != 0) ||
  86             !(features & IFF_MULTI_QUEUE)) {
  87             error_report("multiqueue required, but no kernel "
  88                          "support for IFF_MULTI_QUEUE available");
  89             close(fd);
  90             return -1;
  91         } else {
  92             ifr.ifr_flags |= IFF_MULTI_QUEUE;
  93         }
  94     }
  95
  96     if (ifname[0] != '\0')
  97         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
  98     else
  99         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
 100     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
 101     if (ret != 0) {
 102         if (ifname[0] != '\0') {
 103             error_report("could not configure %s (%s): %m", PATH_NET_TUN, ifr.ifr_name);
 104         } else {
 105             error_report("could not configure %s: %m", PATH_NET_TUN);
 106         }
 107         close(fd);
 108         return -1;
 109     }
 110     pstrcpy(ifname, ifname_size, ifr.ifr_name);
 111     fcntl(fd, F_SETFL, O_NONBLOCK);
 112     return fd;
 113 }
 114
 115 /* sndbuf implements a kind of flow control for tap.
 116  * Unfortunately when it's enabled, and packets are sent
 117  * to other guests on the same host, the receiver
 118  * can lock up the transmitter indefinitely.
 119  *
 120  * To avoid packet loss, sndbuf should be set to a value lower than the tx
 121  * queue capacity of any destination network interface.
 122  * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
 123  * a good value, given a 1500 byte MTU.
 124  */
 125 #define TAP_DEFAULT_SNDBUF 0
 126
 127 int tap_set_sndbuf(int fd, const NetdevTapOptions *tap)
 128 {
 129     int sndbuf;
 130
 131     sndbuf = !tap->has_sndbuf       ? TAP_DEFAULT_SNDBUF :
 132              tap->sndbuf > INT_MAX  ? INT_MAX :
 133              tap->sndbuf;
 134
 135     if (!sndbuf) {
 136         sndbuf = INT_MAX;
 137     }
 138
 139     if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
 140         error_report("TUNSETSNDBUF ioctl failed: %s", strerror(errno));
 141         return -1;
 142     }
 143     return 0;
 144 }
 145
 146 int tap_probe_vnet_hdr(int fd)
 147 {
 148     struct ifreq ifr;
 149
 150     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 151         error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
 152         return 0;
 153     }
 154
 155     return ifr.ifr_flags & IFF_VNET_HDR;
 156 }
 157
 158 int tap_probe_has_ufo(int fd)
 159 {
 160     unsigned offload;
 161
 162     offload = TUN_F_CSUM | TUN_F_UFO;
 163
 164     if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
 165         return 0;
 166
 167     return 1;
 168 }
 169
 170 /* Verify that we can assign given length */
 171 int tap_probe_vnet_hdr_len(int fd, int len)
 172 {
 173     int orig;
 174     if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
 175         return 0;
 176     }
 177     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 178         return 0;
 179     }
 180     /* Restore original length: we can't handle failure. */
 181     if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
 182         fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 183                 strerror(errno));
 184         abort();
 185         return -errno;
 186     }
 187     return 1;
 188 }
 189
 190 void tap_fd_set_vnet_hdr_len(int fd, int len)
 191 {
 192     if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
 193         fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
 194                 strerror(errno));
 195         abort();
 196     }
 197 }
 198
 199 void tap_fd_set_offload(int fd, int csum, int tso4,
 200                         int tso6, int ecn, int ufo)
 201 {
 202     unsigned int offload = 0;
 203
 204     /* Check if our kernel supports TUNSETOFFLOAD */
 205     if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
 206         return;
 207     }
 208
 209     if (csum) {
 210         offload |= TUN_F_CSUM;
 211         if (tso4)
 212             offload |= TUN_F_TSO4;
 213         if (tso6)
 214             offload |= TUN_F_TSO6;
 215         if ((tso4 || tso6) && ecn)
 216             offload |= TUN_F_TSO_ECN;
 217         if (ufo)
 218             offload |= TUN_F_UFO;
 219     }
 220
 221     if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 222         offload &= ~TUN_F_UFO;
 223         if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
 224             fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 225                     strerror(errno));
 226         }
 227     }
 228 }
 229
 230 /* Enable a specific queue of tap. */
 231 int tap_fd_enable(int fd)
 232 {
 233     struct ifreq ifr;
 234     int ret;
 235
 236     memset(&ifr, 0, sizeof(ifr));
 237
 238     ifr.ifr_flags = IFF_ATTACH_QUEUE;
 239     ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
 240
 241     if (ret != 0) {
 242         error_report("could not enable queue");
 243     }
 244
 245     return ret;
 246 }
 247
 248 /* Disable a specific queue of tap/ */
 249 int tap_fd_disable(int fd)
 250 {
 251     struct ifreq ifr;
 252     int ret;
 253
 254     memset(&ifr, 0, sizeof(ifr));
 255
 256     ifr.ifr_flags = IFF_DETACH_QUEUE;
 257     ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
 258
 259     if (ret != 0) {
 260         error_report("could not disable queue");
 261     }
 262
 263     return ret;
 264 }
 265
 266 int tap_fd_get_ifname(int fd, char *ifname)
 267 {
 268     struct ifreq ifr;
 269
 270     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 271         error_report("TUNGETIFF ioctl() failed: %s",
 272                      strerror(errno));
 273         return -1;
 274     }
 275
 276     pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name);
 277     return 0;
 278 }