Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 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 | |
Peter Maydell | 2744d92 | 2016-01-29 17:50:00 +0000 | [diff] [blame] | 26 | #include "qemu/osdep.h" |
Paolo Bonzini | 1422e32 | 2012-10-24 08:43:34 +0200 | [diff] [blame] | 27 | #include "tap_int.h" |
| 28 | #include "tap-linux.h" |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 29 | #include "net/tap.h" |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 30 | |
| 31 | #include <net/if.h> |
| 32 | #include <sys/ioctl.h> |
| 33 | |
Markus Armbruster | da34e65 | 2016-03-14 09:01:28 +0100 | [diff] [blame] | 34 | #include "qapi/error.h" |
Paolo Bonzini | 1de7afc | 2012-12-17 18:20:00 +0100 | [diff] [blame] | 35 | #include "qemu/error-report.h" |
Veronia Bahaa | f348b6d | 2016-03-20 19:16:19 +0200 | [diff] [blame] | 36 | #include "qemu/cutils.h" |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 37 | |
Michael Tokarev | 91ca60e | 2010-06-02 14:33:01 -0300 | [diff] [blame] | 38 | #define PATH_NET_TUN "/dev/net/tun" |
| 39 | |
Jason Wang | 264986e | 2013-01-30 19:12:34 +0800 | [diff] [blame] | 40 | int tap_open(char *ifname, int ifname_size, int *vnet_hdr, |
Markus Armbruster | 468dd82 | 2015-05-15 13:58:58 +0200 | [diff] [blame] | 41 | int vnet_hdr_required, int mq_required, Error **errp) |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 42 | { |
| 43 | struct ifreq ifr; |
| 44 | int fd, ret; |
Michael S. Tsirkin | 89e6d68 | 2012-11-12 09:13:04 +0200 | [diff] [blame] | 45 | int len = sizeof(struct virtio_net_hdr); |
Peter Lieven | d26e445 | 2013-02-25 10:17:08 +0100 | [diff] [blame] | 46 | unsigned int features; |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 47 | |
Michael Tokarev | 91ca60e | 2010-06-02 14:33:01 -0300 | [diff] [blame] | 48 | TFR(fd = open(PATH_NET_TUN, O_RDWR)); |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 49 | if (fd < 0) { |
Markus Armbruster | 47896e2 | 2015-05-15 13:58:59 +0200 | [diff] [blame] | 50 | error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN); |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 51 | return -1; |
| 52 | } |
| 53 | memset(&ifr, 0, sizeof(ifr)); |
| 54 | ifr.ifr_flags = IFF_TAP | IFF_NO_PI; |
| 55 | |
Kusanagi Kouichi | 1f149e7 | 2014-01-18 14:38:45 +0900 | [diff] [blame] | 56 | if (ioctl(fd, TUNGETFEATURES, &features) == -1) { |
Alistair Francis | 3dc6f86 | 2017-07-12 06:57:41 -0700 | [diff] [blame] | 57 | warn_report("TUNGETFEATURES failed: %s", strerror(errno)); |
Kusanagi Kouichi | 1f149e7 | 2014-01-18 14:38:45 +0900 | [diff] [blame] | 58 | features = 0; |
| 59 | } |
| 60 | |
| 61 | if (features & IFF_ONE_QUEUE) { |
Peter Lieven | d26e445 | 2013-02-25 10:17:08 +0100 | [diff] [blame] | 62 | ifr.ifr_flags |= IFF_ONE_QUEUE; |
| 63 | } |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 64 | |
Peter Lieven | d26e445 | 2013-02-25 10:17:08 +0100 | [diff] [blame] | 65 | if (*vnet_hdr) { |
Kusanagi Kouichi | 1f149e7 | 2014-01-18 14:38:45 +0900 | [diff] [blame] | 66 | if (features & IFF_VNET_HDR) { |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 67 | *vnet_hdr = 1; |
| 68 | ifr.ifr_flags |= IFF_VNET_HDR; |
Pierre Riteau | 6720b35 | 2009-11-25 18:49:34 +0000 | [diff] [blame] | 69 | } else { |
| 70 | *vnet_hdr = 0; |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 71 | } |
| 72 | |
| 73 | if (vnet_hdr_required && !*vnet_hdr) { |
Markus Armbruster | 47896e2 | 2015-05-15 13:58:59 +0200 | [diff] [blame] | 74 | error_setg(errp, "vnet_hdr=1 requested, but no kernel " |
| 75 | "support for IFF_VNET_HDR available"); |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 76 | close(fd); |
| 77 | return -1; |
| 78 | } |
Michael S. Tsirkin | 89e6d68 | 2012-11-12 09:13:04 +0200 | [diff] [blame] | 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); |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 86 | } |
| 87 | |
Jason Wang | 94fdc6d | 2013-01-30 19:12:31 +0800 | [diff] [blame] | 88 | if (mq_required) { |
Kusanagi Kouichi | 1f149e7 | 2014-01-18 14:38:45 +0900 | [diff] [blame] | 89 | if (!(features & IFF_MULTI_QUEUE)) { |
Markus Armbruster | 47896e2 | 2015-05-15 13:58:59 +0200 | [diff] [blame] | 90 | error_setg(errp, "multiqueue required, but no kernel " |
| 91 | "support for IFF_MULTI_QUEUE available"); |
Jason Wang | 94fdc6d | 2013-01-30 19:12:31 +0800 | [diff] [blame] | 92 | close(fd); |
| 93 | return -1; |
| 94 | } else { |
| 95 | ifr.ifr_flags |= IFF_MULTI_QUEUE; |
| 96 | } |
| 97 | } |
| 98 | |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 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) { |
Luiz Capitulino | 93a7320 | 2011-10-14 15:05:10 -0300 | [diff] [blame] | 105 | if (ifname[0] != '\0') { |
Markus Armbruster | 47896e2 | 2015-05-15 13:58:59 +0200 | [diff] [blame] | 106 | error_setg_errno(errp, errno, "could not configure %s (%s)", |
| 107 | PATH_NET_TUN, ifr.ifr_name); |
Luiz Capitulino | 93a7320 | 2011-10-14 15:05:10 -0300 | [diff] [blame] | 108 | } else { |
Markus Armbruster | 47896e2 | 2015-05-15 13:58:59 +0200 | [diff] [blame] | 109 | error_setg_errno(errp, errno, "could not configure %s", |
| 110 | PATH_NET_TUN); |
Luiz Capitulino | 93a7320 | 2011-10-14 15:05:10 -0300 | [diff] [blame] | 111 | } |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 112 | close(fd); |
| 113 | return -1; |
| 114 | } |
| 115 | pstrcpy(ifname, ifname_size, ifr.ifr_name); |
Marc-André Lureau | 22e135f | 2022-04-25 11:56:42 +0400 | [diff] [blame] | 116 | g_unix_set_fd_nonblocking(fd, true, NULL); |
Mark McLoughlin | c28b1c1 | 2009-10-22 17:49:12 +0100 | [diff] [blame] | 117 | return fd; |
| 118 | } |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 119 | |
Michael S. Tsirkin | f157ed2 | 2011-02-01 14:25:40 +0200 | [diff] [blame] | 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. |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 127 | * Ethernet NICs generally have txqueuelen=1000, so 1Mb is |
Michael S. Tsirkin | f157ed2 | 2011-02-01 14:25:40 +0200 | [diff] [blame] | 128 | * a good value, given a 1500 byte MTU. |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 129 | */ |
Michael S. Tsirkin | f157ed2 | 2011-02-01 14:25:40 +0200 | [diff] [blame] | 130 | #define TAP_DEFAULT_SNDBUF 0 |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 131 | |
Markus Armbruster | 80b832c | 2015-05-15 13:58:55 +0200 | [diff] [blame] | 132 | void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp) |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 133 | { |
| 134 | int sndbuf; |
| 135 | |
Laszlo Ersek | 08c573a | 2012-07-17 16:17:19 +0200 | [diff] [blame] | 136 | sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF : |
| 137 | tap->sndbuf > INT_MAX ? INT_MAX : |
| 138 | tap->sndbuf; |
| 139 | |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 140 | if (!sndbuf) { |
| 141 | sndbuf = INT_MAX; |
| 142 | } |
| 143 | |
Laszlo Ersek | 08c573a | 2012-07-17 16:17:19 +0200 | [diff] [blame] | 144 | if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) { |
Markus Armbruster | 80b832c | 2015-05-15 13:58:55 +0200 | [diff] [blame] | 145 | error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed"); |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 146 | } |
Mark McLoughlin | 15ac913 | 2009-10-22 17:49:13 +0100 | [diff] [blame] | 147 | } |
Mark McLoughlin | dc69004 | 2009-10-22 17:49:14 +0100 | [diff] [blame] | 148 | |
Daniel P. Berrange | e7b347d | 2020-07-07 20:45:15 +0200 | [diff] [blame] | 149 | int tap_probe_vnet_hdr(int fd, Error **errp) |
Mark McLoughlin | dc69004 | 2009-10-22 17:49:14 +0100 | [diff] [blame] | 150 | { |
| 151 | struct ifreq ifr; |
Peter Foley | e29919c | 2022-01-14 13:08:59 +0800 | [diff] [blame] | 152 | memset(&ifr, 0, sizeof(ifr)); |
Mark McLoughlin | dc69004 | 2009-10-22 17:49:14 +0100 | [diff] [blame] | 153 | |
| 154 | if (ioctl(fd, TUNGETIFF, &ifr) != 0) { |
Daniel P. Berrange | e7b347d | 2020-07-07 20:45:15 +0200 | [diff] [blame] | 155 | /* TUNGETIFF is available since kernel v2.6.27 */ |
| 156 | error_setg_errno(errp, errno, |
| 157 | "Unable to query TUNGETIFF on FD %d", fd); |
| 158 | return -1; |
Mark McLoughlin | dc69004 | 2009-10-22 17:49:14 +0100 | [diff] [blame] | 159 | } |
| 160 | |
| 161 | return ifr.ifr_flags & IFF_VNET_HDR; |
| 162 | } |
Mark McLoughlin | 1faac1f | 2009-10-22 17:49:15 +0100 | [diff] [blame] | 163 | |
Mark McLoughlin | 9c28271 | 2009-10-22 17:49:16 +0100 | [diff] [blame] | 164 | int tap_probe_has_ufo(int fd) |
| 165 | { |
| 166 | unsigned offload; |
| 167 | |
| 168 | offload = TUN_F_CSUM | TUN_F_UFO; |
| 169 | |
| 170 | if (ioctl(fd, TUNSETOFFLOAD, offload) < 0) |
| 171 | return 0; |
| 172 | |
| 173 | return 1; |
| 174 | } |
| 175 | |
Michael S. Tsirkin | 445d892 | 2010-07-16 11:16:06 +0300 | [diff] [blame] | 176 | /* Verify that we can assign given length */ |
| 177 | int tap_probe_vnet_hdr_len(int fd, int len) |
| 178 | { |
| 179 | int orig; |
| 180 | if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) { |
| 181 | return 0; |
| 182 | } |
| 183 | if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { |
| 184 | return 0; |
| 185 | } |
| 186 | /* Restore original length: we can't handle failure. */ |
| 187 | if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) { |
| 188 | fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n", |
| 189 | strerror(errno)); |
Jason Wang | 28a6589 | 2013-01-30 19:12:21 +0800 | [diff] [blame] | 190 | abort(); |
Michael S. Tsirkin | 445d892 | 2010-07-16 11:16:06 +0300 | [diff] [blame] | 191 | return -errno; |
| 192 | } |
| 193 | return 1; |
| 194 | } |
| 195 | |
| 196 | void tap_fd_set_vnet_hdr_len(int fd, int len) |
| 197 | { |
| 198 | if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { |
| 199 | fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n", |
| 200 | strerror(errno)); |
Jason Wang | 28a6589 | 2013-01-30 19:12:21 +0800 | [diff] [blame] | 201 | abort(); |
Michael S. Tsirkin | 445d892 | 2010-07-16 11:16:06 +0300 | [diff] [blame] | 202 | } |
| 203 | } |
| 204 | |
Greg Kurz | c80cd6b | 2015-06-17 15:23:44 +0200 | [diff] [blame] | 205 | int tap_fd_set_vnet_le(int fd, int is_le) |
| 206 | { |
| 207 | int arg = is_le ? 1 : 0; |
| 208 | |
| 209 | if (!ioctl(fd, TUNSETVNETLE, &arg)) { |
| 210 | return 0; |
| 211 | } |
| 212 | |
| 213 | /* Check if our kernel supports TUNSETVNETLE */ |
| 214 | if (errno == EINVAL) { |
| 215 | return -errno; |
| 216 | } |
| 217 | |
John Snow | 594fd21 | 2015-06-29 16:56:26 -0400 | [diff] [blame] | 218 | error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno)); |
Greg Kurz | c80cd6b | 2015-06-17 15:23:44 +0200 | [diff] [blame] | 219 | abort(); |
| 220 | } |
| 221 | |
| 222 | int tap_fd_set_vnet_be(int fd, int is_be) |
| 223 | { |
| 224 | int arg = is_be ? 1 : 0; |
| 225 | |
| 226 | if (!ioctl(fd, TUNSETVNETBE, &arg)) { |
| 227 | return 0; |
| 228 | } |
| 229 | |
| 230 | /* Check if our kernel supports TUNSETVNETBE */ |
| 231 | if (errno == EINVAL) { |
| 232 | return -errno; |
| 233 | } |
| 234 | |
John Snow | 594fd21 | 2015-06-29 16:56:26 -0400 | [diff] [blame] | 235 | error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno)); |
Greg Kurz | c80cd6b | 2015-06-17 15:23:44 +0200 | [diff] [blame] | 236 | abort(); |
| 237 | } |
| 238 | |
Mark McLoughlin | 1faac1f | 2009-10-22 17:49:15 +0100 | [diff] [blame] | 239 | void tap_fd_set_offload(int fd, int csum, int tso4, |
| 240 | int tso6, int ecn, int ufo) |
| 241 | { |
| 242 | unsigned int offload = 0; |
| 243 | |
Pierre Riteau | 2e50326 | 2009-11-25 18:49:35 +0000 | [diff] [blame] | 244 | /* Check if our kernel supports TUNSETOFFLOAD */ |
| 245 | if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) { |
| 246 | return; |
| 247 | } |
| 248 | |
Mark McLoughlin | 1faac1f | 2009-10-22 17:49:15 +0100 | [diff] [blame] | 249 | if (csum) { |
| 250 | offload |= TUN_F_CSUM; |
| 251 | if (tso4) |
| 252 | offload |= TUN_F_TSO4; |
| 253 | if (tso6) |
| 254 | offload |= TUN_F_TSO6; |
| 255 | if ((tso4 || tso6) && ecn) |
| 256 | offload |= TUN_F_TSO_ECN; |
| 257 | if (ufo) |
| 258 | offload |= TUN_F_UFO; |
| 259 | } |
| 260 | |
| 261 | if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { |
| 262 | offload &= ~TUN_F_UFO; |
| 263 | if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { |
| 264 | fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n", |
| 265 | strerror(errno)); |
| 266 | } |
| 267 | } |
| 268 | } |
Jason Wang | 94fdc6d | 2013-01-30 19:12:31 +0800 | [diff] [blame] | 269 | |
| 270 | /* Enable a specific queue of tap. */ |
| 271 | int tap_fd_enable(int fd) |
| 272 | { |
| 273 | struct ifreq ifr; |
| 274 | int ret; |
| 275 | |
| 276 | memset(&ifr, 0, sizeof(ifr)); |
| 277 | |
| 278 | ifr.ifr_flags = IFF_ATTACH_QUEUE; |
| 279 | ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); |
| 280 | |
| 281 | if (ret != 0) { |
| 282 | error_report("could not enable queue"); |
| 283 | } |
| 284 | |
| 285 | return ret; |
| 286 | } |
| 287 | |
| 288 | /* Disable a specific queue of tap/ */ |
| 289 | int tap_fd_disable(int fd) |
| 290 | { |
| 291 | struct ifreq ifr; |
| 292 | int ret; |
| 293 | |
| 294 | memset(&ifr, 0, sizeof(ifr)); |
| 295 | |
| 296 | ifr.ifr_flags = IFF_DETACH_QUEUE; |
| 297 | ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); |
| 298 | |
| 299 | if (ret != 0) { |
| 300 | error_report("could not disable queue"); |
| 301 | } |
| 302 | |
| 303 | return ret; |
| 304 | } |
| 305 | |
Jason Wang | e5dc0b4 | 2013-01-30 19:12:33 +0800 | [diff] [blame] | 306 | int tap_fd_get_ifname(int fd, char *ifname) |
| 307 | { |
| 308 | struct ifreq ifr; |
| 309 | |
| 310 | if (ioctl(fd, TUNGETIFF, &ifr) != 0) { |
| 311 | error_report("TUNGETIFF ioctl() failed: %s", |
| 312 | strerror(errno)); |
| 313 | return -1; |
| 314 | } |
| 315 | |
| 316 | pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name); |
| 317 | return 0; |
| 318 | } |
Andrew Melnychenko | 8f364e3 | 2021-05-14 14:48:30 +0300 | [diff] [blame] | 319 | |
| 320 | int tap_fd_set_steering_ebpf(int fd, int prog_fd) |
| 321 | { |
| 322 | if (ioctl(fd, TUNSETSTEERINGEBPF, (void *) &prog_fd) != 0) { |
| 323 | error_report("Issue while setting TUNSETSTEERINGEBPF:" |
| 324 | " %s with fd: %d, prog_fd: %d", |
| 325 | strerror(errno), fd, prog_fd); |
| 326 | |
| 327 | return -1; |
| 328 | } |
| 329 | |
| 330 | return 0; |
| 331 | } |