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 "qemu/osdep.h" 27 #include "tap_int.h" 28 #include "tap-linux.h" 29 #include "net/tap.h" 30 31 #include <net/if.h> 32 #include <sys/ioctl.h> 33 34 #include "sysemu/sysemu.h" 35 #include "qapi/error.h" 36 #include "qemu/error-report.h" 37 #include "qemu/cutils.h" 38 39 #define PATH_NET_TUN "/dev/net/tun" 40 41 int tap_open(char *ifname, int ifname_size, int *vnet_hdr, 42 int vnet_hdr_required, int mq_required, Error **errp) 43 { 44 struct ifreq ifr; 45 int fd, ret; 46 int len = sizeof(struct virtio_net_hdr); 47 unsigned int features; 48 49 TFR(fd = open(PATH_NET_TUN, O_RDWR)); 50 if (fd < 0) { 51 error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN); 52 return -1; 53 } 54 memset(&ifr, 0, sizeof(ifr)); 55 ifr.ifr_flags = IFF_TAP | IFF_NO_PI; 56 57 if (ioctl(fd, TUNGETFEATURES, &features) == -1) { 58 warn_report("TUNGETFEATURES failed: %s", strerror(errno)); 59 features = 0; 60 } 61 62 if (features & IFF_ONE_QUEUE) { 63 ifr.ifr_flags |= IFF_ONE_QUEUE; 64 } 65 66 if (*vnet_hdr) { 67 if (features & IFF_VNET_HDR) { 68 *vnet_hdr = 1; 69 ifr.ifr_flags |= IFF_VNET_HDR; 70 } else { 71 *vnet_hdr = 0; 72 } 73 74 if (vnet_hdr_required && !*vnet_hdr) { 75 error_setg(errp, "vnet_hdr=1 requested, but no kernel " 76 "support for IFF_VNET_HDR available"); 77 close(fd); 78 return -1; 79 } 80 /* 81 * Make sure vnet header size has the default value: for a persistent 82 * tap it might have been modified e.g. by another instance of qemu. 83 * Ignore errors since old kernels do not support this ioctl: in this 84 * case the header size implicitly has the correct value. 85 */ 86 ioctl(fd, TUNSETVNETHDRSZ, &len); 87 } 88 89 if (mq_required) { 90 if (!(features & IFF_MULTI_QUEUE)) { 91 error_setg(errp, "multiqueue required, but no kernel " 92 "support for IFF_MULTI_QUEUE available"); 93 close(fd); 94 return -1; 95 } else { 96 ifr.ifr_flags |= IFF_MULTI_QUEUE; 97 } 98 } 99 100 if (ifname[0] != '\0') 101 pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname); 102 else 103 pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d"); 104 ret = ioctl(fd, TUNSETIFF, (void *) &ifr); 105 if (ret != 0) { 106 if (ifname[0] != '\0') { 107 error_setg_errno(errp, errno, "could not configure %s (%s)", 108 PATH_NET_TUN, ifr.ifr_name); 109 } else { 110 error_setg_errno(errp, errno, "could not configure %s", 111 PATH_NET_TUN); 112 } 113 close(fd); 114 return -1; 115 } 116 pstrcpy(ifname, ifname_size, ifr.ifr_name); 117 fcntl(fd, F_SETFL, O_NONBLOCK); 118 return fd; 119 } 120 121 /* sndbuf implements a kind of flow control for tap. 122 * Unfortunately when it's enabled, and packets are sent 123 * to other guests on the same host, the receiver 124 * can lock up the transmitter indefinitely. 125 * 126 * To avoid packet loss, sndbuf should be set to a value lower than the tx 127 * queue capacity of any destination network interface. 128 * Ethernet NICs generally have txqueuelen=1000, so 1Mb is 129 * a good value, given a 1500 byte MTU. 130 */ 131 #define TAP_DEFAULT_SNDBUF 0 132 133 void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp) 134 { 135 int sndbuf; 136 137 sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF : 138 tap->sndbuf > INT_MAX ? INT_MAX : 139 tap->sndbuf; 140 141 if (!sndbuf) { 142 sndbuf = INT_MAX; 143 } 144 145 if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) { 146 error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed"); 147 } 148 } 149 150 int tap_probe_vnet_hdr(int fd) 151 { 152 struct ifreq ifr; 153 154 if (ioctl(fd, TUNGETIFF, &ifr) != 0) { 155 error_report("TUNGETIFF ioctl() failed: %s", strerror(errno)); 156 return 0; 157 } 158 159 return ifr.ifr_flags & IFF_VNET_HDR; 160 } 161 162 int tap_probe_has_ufo(int fd) 163 { 164 unsigned offload; 165 166 offload = TUN_F_CSUM | TUN_F_UFO; 167 168 if (ioctl(fd, TUNSETOFFLOAD, offload) < 0) 169 return 0; 170 171 return 1; 172 } 173 174 /* Verify that we can assign given length */ 175 int tap_probe_vnet_hdr_len(int fd, int len) 176 { 177 int orig; 178 if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) { 179 return 0; 180 } 181 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { 182 return 0; 183 } 184 /* Restore original length: we can't handle failure. */ 185 if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) { 186 fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n", 187 strerror(errno)); 188 abort(); 189 return -errno; 190 } 191 return 1; 192 } 193 194 void tap_fd_set_vnet_hdr_len(int fd, int len) 195 { 196 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { 197 fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n", 198 strerror(errno)); 199 abort(); 200 } 201 } 202 203 int tap_fd_set_vnet_le(int fd, int is_le) 204 { 205 int arg = is_le ? 1 : 0; 206 207 if (!ioctl(fd, TUNSETVNETLE, &arg)) { 208 return 0; 209 } 210 211 /* Check if our kernel supports TUNSETVNETLE */ 212 if (errno == EINVAL) { 213 return -errno; 214 } 215 216 error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno)); 217 abort(); 218 } 219 220 int tap_fd_set_vnet_be(int fd, int is_be) 221 { 222 int arg = is_be ? 1 : 0; 223 224 if (!ioctl(fd, TUNSETVNETBE, &arg)) { 225 return 0; 226 } 227 228 /* Check if our kernel supports TUNSETVNETBE */ 229 if (errno == EINVAL) { 230 return -errno; 231 } 232 233 error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno)); 234 abort(); 235 } 236 237 void tap_fd_set_offload(int fd, int csum, int tso4, 238 int tso6, int ecn, int ufo) 239 { 240 unsigned int offload = 0; 241 242 /* Check if our kernel supports TUNSETOFFLOAD */ 243 if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) { 244 return; 245 } 246 247 if (csum) { 248 offload |= TUN_F_CSUM; 249 if (tso4) 250 offload |= TUN_F_TSO4; 251 if (tso6) 252 offload |= TUN_F_TSO6; 253 if ((tso4 || tso6) && ecn) 254 offload |= TUN_F_TSO_ECN; 255 if (ufo) 256 offload |= TUN_F_UFO; 257 } 258 259 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { 260 offload &= ~TUN_F_UFO; 261 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { 262 fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n", 263 strerror(errno)); 264 } 265 } 266 } 267 268 /* Enable a specific queue of tap. */ 269 int tap_fd_enable(int fd) 270 { 271 struct ifreq ifr; 272 int ret; 273 274 memset(&ifr, 0, sizeof(ifr)); 275 276 ifr.ifr_flags = IFF_ATTACH_QUEUE; 277 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); 278 279 if (ret != 0) { 280 error_report("could not enable queue"); 281 } 282 283 return ret; 284 } 285 286 /* Disable a specific queue of tap/ */ 287 int tap_fd_disable(int fd) 288 { 289 struct ifreq ifr; 290 int ret; 291 292 memset(&ifr, 0, sizeof(ifr)); 293 294 ifr.ifr_flags = IFF_DETACH_QUEUE; 295 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); 296 297 if (ret != 0) { 298 error_report("could not disable queue"); 299 } 300 301 return ret; 302 } 303 304 int tap_fd_get_ifname(int fd, char *ifname) 305 { 306 struct ifreq ifr; 307 308 if (ioctl(fd, TUNGETIFF, &ifr) != 0) { 309 error_report("TUNGETIFF ioctl() failed: %s", 310 strerror(errno)); 311 return -1; 312 } 313 314 pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name); 315 return 0; 316 } 317