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