1 /************************************************************************** 2 3 Copyright (c) 2007-2009, Chelsio Inc. 4 All rights reserved. 5 6 Redistribution and use in source and binary forms, with or without 7 modification, are permitted provided that the following conditions are met: 8 9 1. Redistributions of source code must retain the above copyright notice, 10 this list of conditions and the following disclaimer. 11 12 2. Neither the name of the Chelsio Corporation nor the names of its 13 contributors may be used to endorse or promote products derived from 14 this software without specific prior written permission. 15 16 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 17 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 20 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 POSSIBILITY OF SUCH DAMAGE. 27 28 ***************************************************************************/ 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include <sys/param.h> 34 #include <sys/systm.h> 35 #include <sys/kernel.h> 36 #include <sys/bus.h> 37 #include <sys/module.h> 38 #include <sys/pciio.h> 39 #include <sys/conf.h> 40 #include <machine/bus.h> 41 #include <machine/resource.h> 42 #include <sys/bus_dma.h> 43 #include <sys/ktr.h> 44 #include <sys/rman.h> 45 #include <sys/ioccom.h> 46 #include <sys/mbuf.h> 47 #include <sys/linker.h> 48 #include <sys/firmware.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/smp.h> 52 #include <sys/sysctl.h> 53 #include <sys/syslog.h> 54 #include <sys/queue.h> 55 #include <sys/taskqueue.h> 56 #include <sys/proc.h> 57 58 #include <net/bpf.h> 59 #include <net/ethernet.h> 60 #include <net/if.h> 61 #include <net/if_arp.h> 62 #include <net/if_dl.h> 63 #include <net/if_media.h> 64 #include <net/if_types.h> 65 #include <net/if_vlan_var.h> 66 67 #include <netinet/in_systm.h> 68 #include <netinet/in.h> 69 #include <netinet/if_ether.h> 70 #include <netinet/ip.h> 71 #include <netinet/ip.h> 72 #include <netinet/tcp.h> 73 #include <netinet/udp.h> 74 75 #include <dev/pci/pcireg.h> 76 #include <dev/pci/pcivar.h> 77 #include <dev/pci/pci_private.h> 78 79 #include <cxgb_include.h> 80 81 #ifdef PRIV_SUPPORTED 82 #include <sys/priv.h> 83 #endif 84 85 static int cxgb_setup_interrupts(adapter_t *); 86 static void cxgb_teardown_interrupts(adapter_t *); 87 static void cxgb_init(void *); 88 static int cxgb_init_locked(struct port_info *); 89 static int cxgb_uninit_locked(struct port_info *); 90 static int cxgb_uninit_synchronized(struct port_info *); 91 static int cxgb_ioctl(struct ifnet *, unsigned long, caddr_t); 92 static int cxgb_media_change(struct ifnet *); 93 static int cxgb_ifm_type(int); 94 static void cxgb_build_medialist(struct port_info *); 95 static void cxgb_media_status(struct ifnet *, struct ifmediareq *); 96 static int setup_sge_qsets(adapter_t *); 97 static void cxgb_async_intr(void *); 98 static void cxgb_tick_handler(void *, int); 99 static void cxgb_tick(void *); 100 static void link_check_callout(void *); 101 static void check_link_status(void *, int); 102 static void setup_rss(adapter_t *sc); 103 static int alloc_filters(struct adapter *); 104 static int setup_hw_filters(struct adapter *); 105 static int set_filter(struct adapter *, int, const struct filter_info *); 106 static inline void mk_set_tcb_field(struct cpl_set_tcb_field *, unsigned int, 107 unsigned int, u64, u64); 108 static inline void set_tcb_field_ulp(struct cpl_set_tcb_field *, unsigned int, 109 unsigned int, u64, u64); 110 111 /* Attachment glue for the PCI controller end of the device. Each port of 112 * the device is attached separately, as defined later. 113 */ 114 static int cxgb_controller_probe(device_t); 115 static int cxgb_controller_attach(device_t); 116 static int cxgb_controller_detach(device_t); 117 static void cxgb_free(struct adapter *); 118 static __inline void reg_block_dump(struct adapter *ap, uint8_t *buf, unsigned int start, 119 unsigned int end); 120 static void cxgb_get_regs(adapter_t *sc, struct ch_ifconf_regs *regs, uint8_t *buf); 121 static int cxgb_get_regs_len(void); 122 static int offload_open(struct port_info *pi); 123 static void touch_bars(device_t dev); 124 static int offload_close(struct t3cdev *tdev); 125 static void cxgb_update_mac_settings(struct port_info *p); 126 127 static device_method_t cxgb_controller_methods[] = { 128 DEVMETHOD(device_probe, cxgb_controller_probe), 129 DEVMETHOD(device_attach, cxgb_controller_attach), 130 DEVMETHOD(device_detach, cxgb_controller_detach), 131 132 DEVMETHOD_END 133 }; 134 135 static driver_t cxgb_controller_driver = { 136 "cxgbc", 137 cxgb_controller_methods, 138 sizeof(struct adapter) 139 }; 140 141 static devclass_t cxgb_controller_devclass; 142 DRIVER_MODULE(cxgbc, pci, cxgb_controller_driver, cxgb_controller_devclass, 0, 0); 143 144 /* 145 * Attachment glue for the ports. Attachment is done directly to the 146 * controller device. 147 */ 148 static int cxgb_port_probe(device_t); 149 static int cxgb_port_attach(device_t); 150 static int cxgb_port_detach(device_t); 151 152 static device_method_t cxgb_port_methods[] = { 153 DEVMETHOD(device_probe, cxgb_port_probe), 154 DEVMETHOD(device_attach, cxgb_port_attach), 155 DEVMETHOD(device_detach, cxgb_port_detach), 156 { 0, 0 } 157 }; 158 159 static driver_t cxgb_port_driver = { 160 "cxgb", 161 cxgb_port_methods, 162 0 163 }; 164 165 static d_ioctl_t cxgb_extension_ioctl; 166 static d_open_t cxgb_extension_open; 167 static d_close_t cxgb_extension_close; 168 169 static struct cdevsw cxgb_cdevsw = { 170 .d_version = D_VERSION, 171 .d_flags = 0, 172 .d_open = cxgb_extension_open, 173 .d_close = cxgb_extension_close, 174 .d_ioctl = cxgb_extension_ioctl, 175 .d_name = "cxgb", 176 }; 177 178 static devclass_t cxgb_port_devclass; 179 DRIVER_MODULE(cxgb, cxgbc, cxgb_port_driver, cxgb_port_devclass, 0, 0); 180 181 /* 182 * The driver uses the best interrupt scheme available on a platform in the 183 * order MSI-X, MSI, legacy pin interrupts. This parameter determines which 184 * of these schemes the driver may consider as follows: 185 * 186 * msi = 2: choose from among all three options 187 * msi = 1 : only consider MSI and pin interrupts 188 * msi = 0: force pin interrupts 189 */ 190 static int msi_allowed = 2; 191 192 TUNABLE_INT("hw.cxgb.msi_allowed", &msi_allowed); 193 SYSCTL_NODE(_hw, OID_AUTO, cxgb, CTLFLAG_RD, 0, "CXGB driver parameters"); 194 SYSCTL_INT(_hw_cxgb, OID_AUTO, msi_allowed, CTLFLAG_RDTUN, &msi_allowed, 0, 195 "MSI-X, MSI, INTx selector"); 196 197 /* 198 * The driver enables offload as a default. 199 * To disable it, use ofld_disable = 1. 200 */ 201 static int ofld_disable = 0; 202 TUNABLE_INT("hw.cxgb.ofld_disable", &ofld_disable); 203 SYSCTL_INT(_hw_cxgb, OID_AUTO, ofld_disable, CTLFLAG_RDTUN, &ofld_disable, 0, 204 "disable ULP offload"); 205 206 /* 207 * The driver uses an auto-queue algorithm by default. 208 * To disable it and force a single queue-set per port, use multiq = 0 209 */ 210 static int multiq = 1; 211 TUNABLE_INT("hw.cxgb.multiq", &multiq); 212 SYSCTL_INT(_hw_cxgb, OID_AUTO, multiq, CTLFLAG_RDTUN, &multiq, 0, 213 "use min(ncpus/ports, 8) queue-sets per port"); 214 215 /* 216 * By default the driver will not update the firmware unless 217 * it was compiled against a newer version 218 * 219 */ 220 static int force_fw_update = 0; 221 TUNABLE_INT("hw.cxgb.force_fw_update", &force_fw_update); 222 SYSCTL_INT(_hw_cxgb, OID_AUTO, force_fw_update, CTLFLAG_RDTUN, &force_fw_update, 0, 223 "update firmware even if up to date"); 224 225 int cxgb_use_16k_clusters = -1; 226 TUNABLE_INT("hw.cxgb.use_16k_clusters", &cxgb_use_16k_clusters); 227 SYSCTL_INT(_hw_cxgb, OID_AUTO, use_16k_clusters, CTLFLAG_RDTUN, 228 &cxgb_use_16k_clusters, 0, "use 16kB clusters for the jumbo queue "); 229 230 static int nfilters = -1; 231 TUNABLE_INT("hw.cxgb.nfilters", &nfilters); 232 SYSCTL_INT(_hw_cxgb, OID_AUTO, nfilters, CTLFLAG_RDTUN, 233 &nfilters, 0, "max number of entries in the filter table"); 234 235 enum { 236 MAX_TXQ_ENTRIES = 16384, 237 MAX_CTRL_TXQ_ENTRIES = 1024, 238 MAX_RSPQ_ENTRIES = 16384, 239 MAX_RX_BUFFERS = 16384, 240 MAX_RX_JUMBO_BUFFERS = 16384, 241 MIN_TXQ_ENTRIES = 4, 242 MIN_CTRL_TXQ_ENTRIES = 4, 243 MIN_RSPQ_ENTRIES = 32, 244 MIN_FL_ENTRIES = 32, 245 MIN_FL_JUMBO_ENTRIES = 32 246 }; 247 248 struct filter_info { 249 u32 sip; 250 u32 sip_mask; 251 u32 dip; 252 u16 sport; 253 u16 dport; 254 u32 vlan:12; 255 u32 vlan_prio:3; 256 u32 mac_hit:1; 257 u32 mac_idx:4; 258 u32 mac_vld:1; 259 u32 pkt_type:2; 260 u32 report_filter_id:1; 261 u32 pass:1; 262 u32 rss:1; 263 u32 qset:3; 264 u32 locked:1; 265 u32 valid:1; 266 }; 267 268 enum { FILTER_NO_VLAN_PRI = 7 }; 269 270 #define EEPROM_MAGIC 0x38E2F10C 271 272 #define PORT_MASK ((1 << MAX_NPORTS) - 1) 273 274 /* Table for probing the cards. The desc field isn't actually used */ 275 struct cxgb_ident { 276 uint16_t vendor; 277 uint16_t device; 278 int index; 279 char *desc; 280 } cxgb_identifiers[] = { 281 {PCI_VENDOR_ID_CHELSIO, 0x0020, 0, "PE9000"}, 282 {PCI_VENDOR_ID_CHELSIO, 0x0021, 1, "T302E"}, 283 {PCI_VENDOR_ID_CHELSIO, 0x0022, 2, "T310E"}, 284 {PCI_VENDOR_ID_CHELSIO, 0x0023, 3, "T320X"}, 285 {PCI_VENDOR_ID_CHELSIO, 0x0024, 1, "T302X"}, 286 {PCI_VENDOR_ID_CHELSIO, 0x0025, 3, "T320E"}, 287 {PCI_VENDOR_ID_CHELSIO, 0x0026, 2, "T310X"}, 288 {PCI_VENDOR_ID_CHELSIO, 0x0030, 2, "T3B10"}, 289 {PCI_VENDOR_ID_CHELSIO, 0x0031, 3, "T3B20"}, 290 {PCI_VENDOR_ID_CHELSIO, 0x0032, 1, "T3B02"}, 291 {PCI_VENDOR_ID_CHELSIO, 0x0033, 4, "T3B04"}, 292 {PCI_VENDOR_ID_CHELSIO, 0x0035, 6, "T3C10"}, 293 {PCI_VENDOR_ID_CHELSIO, 0x0036, 3, "S320E-CR"}, 294 {PCI_VENDOR_ID_CHELSIO, 0x0037, 7, "N320E-G2"}, 295 {0, 0, 0, NULL} 296 }; 297 298 static int set_eeprom(struct port_info *pi, const uint8_t *data, int len, int offset); 299 300 301 static __inline char 302 t3rev2char(struct adapter *adapter) 303 { 304 char rev = 'z'; 305 306 switch(adapter->params.rev) { 307 case T3_REV_A: 308 rev = 'a'; 309 break; 310 case T3_REV_B: 311 case T3_REV_B2: 312 rev = 'b'; 313 break; 314 case T3_REV_C: 315 rev = 'c'; 316 break; 317 } 318 return rev; 319 } 320 321 static struct cxgb_ident * 322 cxgb_get_ident(device_t dev) 323 { 324 struct cxgb_ident *id; 325 326 for (id = cxgb_identifiers; id->desc != NULL; id++) { 327 if ((id->vendor == pci_get_vendor(dev)) && 328 (id->device == pci_get_device(dev))) { 329 return (id); 330 } 331 } 332 return (NULL); 333 } 334 335 static const struct adapter_info * 336 cxgb_get_adapter_info(device_t dev) 337 { 338 struct cxgb_ident *id; 339 const struct adapter_info *ai; 340 341 id = cxgb_get_ident(dev); 342 if (id == NULL) 343 return (NULL); 344 345 ai = t3_get_adapter_info(id->index); 346 347 return (ai); 348 } 349 350 static int 351 cxgb_controller_probe(device_t dev) 352 { 353 const struct adapter_info *ai; 354 char *ports, buf[80]; 355 int nports; 356 357 ai = cxgb_get_adapter_info(dev); 358 if (ai == NULL) 359 return (ENXIO); 360 361 nports = ai->nports0 + ai->nports1; 362 if (nports == 1) 363 ports = "port"; 364 else 365 ports = "ports"; 366 367 snprintf(buf, sizeof(buf), "%s, %d %s", ai->desc, nports, ports); 368 device_set_desc_copy(dev, buf); 369 return (BUS_PROBE_DEFAULT); 370 } 371 372 #define FW_FNAME "cxgb_t3fw" 373 #define TPEEPROM_NAME "cxgb_t3%c_tp_eeprom" 374 #define TPSRAM_NAME "cxgb_t3%c_protocol_sram" 375 376 static int 377 upgrade_fw(adapter_t *sc) 378 { 379 const struct firmware *fw; 380 int status; 381 u32 vers; 382 383 if ((fw = firmware_get(FW_FNAME)) == NULL) { 384 device_printf(sc->dev, "Could not find firmware image %s\n", FW_FNAME); 385 return (ENOENT); 386 } else 387 device_printf(sc->dev, "installing firmware on card\n"); 388 status = t3_load_fw(sc, (const uint8_t *)fw->data, fw->datasize); 389 390 if (status != 0) { 391 device_printf(sc->dev, "failed to install firmware: %d\n", 392 status); 393 } else { 394 t3_get_fw_version(sc, &vers); 395 snprintf(&sc->fw_version[0], sizeof(sc->fw_version), "%d.%d.%d", 396 G_FW_VERSION_MAJOR(vers), G_FW_VERSION_MINOR(vers), 397 G_FW_VERSION_MICRO(vers)); 398 } 399 400 firmware_put(fw, FIRMWARE_UNLOAD); 401 402 return (status); 403 } 404 405 /* 406 * The cxgb_controller_attach function is responsible for the initial 407 * bringup of the device. Its responsibilities include: 408 * 409 * 1. Determine if the device supports MSI or MSI-X. 410 * 2. Allocate bus resources so that we can access the Base Address Register 411 * 3. Create and initialize mutexes for the controller and its control 412 * logic such as SGE and MDIO. 413 * 4. Call hardware specific setup routine for the adapter as a whole. 414 * 5. Allocate the BAR for doing MSI-X. 415 * 6. Setup the line interrupt iff MSI-X is not supported. 416 * 7. Create the driver's taskq. 417 * 8. Start one task queue service thread. 418 * 9. Check if the firmware and SRAM are up-to-date. They will be 419 * auto-updated later (before FULL_INIT_DONE), if required. 420 * 10. Create a child device for each MAC (port) 421 * 11. Initialize T3 private state. 422 * 12. Trigger the LED 423 * 13. Setup offload iff supported. 424 * 14. Reset/restart the tick callout. 425 * 15. Attach sysctls 426 * 427 * NOTE: Any modification or deviation from this list MUST be reflected in 428 * the above comment. Failure to do so will result in problems on various 429 * error conditions including link flapping. 430 */ 431 static int 432 cxgb_controller_attach(device_t dev) 433 { 434 device_t child; 435 const struct adapter_info *ai; 436 struct adapter *sc; 437 int i, error = 0; 438 uint32_t vers; 439 int port_qsets = 1; 440 int msi_needed, reg; 441 char buf[80]; 442 443 sc = device_get_softc(dev); 444 sc->dev = dev; 445 sc->msi_count = 0; 446 ai = cxgb_get_adapter_info(dev); 447 448 /* find the PCIe link width and set max read request to 4KB*/ 449 if (pci_find_cap(dev, PCIY_EXPRESS, ®) == 0) { 450 uint16_t lnk; 451 452 lnk = pci_read_config(dev, reg + PCIR_EXPRESS_LINK_STA, 2); 453 sc->link_width = (lnk & PCIM_LINK_STA_WIDTH) >> 4; 454 if (sc->link_width < 8 && 455 (ai->caps & SUPPORTED_10000baseT_Full)) { 456 device_printf(sc->dev, 457 "PCIe x%d Link, expect reduced performance\n", 458 sc->link_width); 459 } 460 461 pci_set_max_read_req(dev, 4096); 462 } 463 464 touch_bars(dev); 465 pci_enable_busmaster(dev); 466 /* 467 * Allocate the registers and make them available to the driver. 468 * The registers that we care about for NIC mode are in BAR 0 469 */ 470 sc->regs_rid = PCIR_BAR(0); 471 if ((sc->regs_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 472 &sc->regs_rid, RF_ACTIVE)) == NULL) { 473 device_printf(dev, "Cannot allocate BAR region 0\n"); 474 return (ENXIO); 475 } 476 477 snprintf(sc->lockbuf, ADAPTER_LOCK_NAME_LEN, "cxgb controller lock %d", 478 device_get_unit(dev)); 479 ADAPTER_LOCK_INIT(sc, sc->lockbuf); 480 481 snprintf(sc->reglockbuf, ADAPTER_LOCK_NAME_LEN, "SGE reg lock %d", 482 device_get_unit(dev)); 483 snprintf(sc->mdiolockbuf, ADAPTER_LOCK_NAME_LEN, "cxgb mdio lock %d", 484 device_get_unit(dev)); 485 snprintf(sc->elmerlockbuf, ADAPTER_LOCK_NAME_LEN, "cxgb elmer lock %d", 486 device_get_unit(dev)); 487 488 MTX_INIT(&sc->sge.reg_lock, sc->reglockbuf, NULL, MTX_SPIN); 489 MTX_INIT(&sc->mdio_lock, sc->mdiolockbuf, NULL, MTX_DEF); 490 MTX_INIT(&sc->elmer_lock, sc->elmerlockbuf, NULL, MTX_DEF); 491 492 sc->bt = rman_get_bustag(sc->regs_res); 493 sc->bh = rman_get_bushandle(sc->regs_res); 494 sc->mmio_len = rman_get_size(sc->regs_res); 495 496 for (i = 0; i < MAX_NPORTS; i++) 497 sc->port[i].adapter = sc; 498 499 if (t3_prep_adapter(sc, ai, 1) < 0) { 500 printf("prep adapter failed\n"); 501 error = ENODEV; 502 goto out; 503 } 504 505 sc->udbs_rid = PCIR_BAR(2); 506 sc->udbs_res = NULL; 507 if (is_offload(sc) && 508 ((sc->udbs_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 509 &sc->udbs_rid, RF_ACTIVE)) == NULL)) { 510 device_printf(dev, "Cannot allocate BAR region 1\n"); 511 error = ENXIO; 512 goto out; 513 } 514 515 /* Allocate the BAR for doing MSI-X. If it succeeds, try to allocate 516 * enough messages for the queue sets. If that fails, try falling 517 * back to MSI. If that fails, then try falling back to the legacy 518 * interrupt pin model. 519 */ 520 sc->msix_regs_rid = 0x20; 521 if ((msi_allowed >= 2) && 522 (sc->msix_regs_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 523 &sc->msix_regs_rid, RF_ACTIVE)) != NULL) { 524 525 if (multiq) 526 port_qsets = min(SGE_QSETS/sc->params.nports, mp_ncpus); 527 msi_needed = sc->msi_count = sc->params.nports * port_qsets + 1; 528 529 if (pci_msix_count(dev) == 0 || 530 (error = pci_alloc_msix(dev, &sc->msi_count)) != 0 || 531 sc->msi_count != msi_needed) { 532 device_printf(dev, "alloc msix failed - " 533 "msi_count=%d, msi_needed=%d, err=%d; " 534 "will try MSI\n", sc->msi_count, 535 msi_needed, error); 536 sc->msi_count = 0; 537 port_qsets = 1; 538 pci_release_msi(dev); 539 bus_release_resource(dev, SYS_RES_MEMORY, 540 sc->msix_regs_rid, sc->msix_regs_res); 541 sc->msix_regs_res = NULL; 542 } else { 543 sc->flags |= USING_MSIX; 544 sc->cxgb_intr = cxgb_async_intr; 545 device_printf(dev, 546 "using MSI-X interrupts (%u vectors)\n", 547 sc->msi_count); 548 } 549 } 550 551 if ((msi_allowed >= 1) && (sc->msi_count == 0)) { 552 sc->msi_count = 1; 553 if ((error = pci_alloc_msi(dev, &sc->msi_count)) != 0) { 554 device_printf(dev, "alloc msi failed - " 555 "err=%d; will try INTx\n", error); 556 sc->msi_count = 0; 557 port_qsets = 1; 558 pci_release_msi(dev); 559 } else { 560 sc->flags |= USING_MSI; 561 sc->cxgb_intr = t3_intr_msi; 562 device_printf(dev, "using MSI interrupts\n"); 563 } 564 } 565 if (sc->msi_count == 0) { 566 device_printf(dev, "using line interrupts\n"); 567 sc->cxgb_intr = t3b_intr; 568 } 569 570 /* Create a private taskqueue thread for handling driver events */ 571 sc->tq = taskqueue_create("cxgb_taskq", M_NOWAIT, 572 taskqueue_thread_enqueue, &sc->tq); 573 if (sc->tq == NULL) { 574 device_printf(dev, "failed to allocate controller task queue\n"); 575 goto out; 576 } 577 578 taskqueue_start_threads(&sc->tq, 1, PI_NET, "%s taskq", 579 device_get_nameunit(dev)); 580 TASK_INIT(&sc->tick_task, 0, cxgb_tick_handler, sc); 581 582 583 /* Create a periodic callout for checking adapter status */ 584 callout_init(&sc->cxgb_tick_ch, TRUE); 585 586 if (t3_check_fw_version(sc) < 0 || force_fw_update) { 587 /* 588 * Warn user that a firmware update will be attempted in init. 589 */ 590 device_printf(dev, "firmware needs to be updated to version %d.%d.%d\n", 591 FW_VERSION_MAJOR, FW_VERSION_MINOR, FW_VERSION_MICRO); 592 sc->flags &= ~FW_UPTODATE; 593 } else { 594 sc->flags |= FW_UPTODATE; 595 } 596 597 if (t3_check_tpsram_version(sc) < 0) { 598 /* 599 * Warn user that a firmware update will be attempted in init. 600 */ 601 device_printf(dev, "SRAM needs to be updated to version %c-%d.%d.%d\n", 602 t3rev2char(sc), TP_VERSION_MAJOR, TP_VERSION_MINOR, TP_VERSION_MICRO); 603 sc->flags &= ~TPS_UPTODATE; 604 } else { 605 sc->flags |= TPS_UPTODATE; 606 } 607 608 /* 609 * Create a child device for each MAC. The ethernet attachment 610 * will be done in these children. 611 */ 612 for (i = 0; i < (sc)->params.nports; i++) { 613 struct port_info *pi; 614 615 if ((child = device_add_child(dev, "cxgb", -1)) == NULL) { 616 device_printf(dev, "failed to add child port\n"); 617 error = EINVAL; 618 goto out; 619 } 620 pi = &sc->port[i]; 621 pi->adapter = sc; 622 pi->nqsets = port_qsets; 623 pi->first_qset = i*port_qsets; 624 pi->port_id = i; 625 pi->tx_chan = i >= ai->nports0; 626 pi->txpkt_intf = pi->tx_chan ? 2 * (i - ai->nports0) + 1 : 2 * i; 627 sc->rxpkt_map[pi->txpkt_intf] = i; 628 sc->port[i].tx_chan = i >= ai->nports0; 629 sc->portdev[i] = child; 630 device_set_softc(child, pi); 631 } 632 if ((error = bus_generic_attach(dev)) != 0) 633 goto out; 634 635 /* initialize sge private state */ 636 t3_sge_init_adapter(sc); 637 638 t3_led_ready(sc); 639 640 cxgb_offload_init(); 641 if (is_offload(sc)) { 642 setbit(&sc->registered_device_map, OFFLOAD_DEVMAP_BIT); 643 cxgb_adapter_ofld(sc); 644 } 645 error = t3_get_fw_version(sc, &vers); 646 if (error) 647 goto out; 648 649 snprintf(&sc->fw_version[0], sizeof(sc->fw_version), "%d.%d.%d", 650 G_FW_VERSION_MAJOR(vers), G_FW_VERSION_MINOR(vers), 651 G_FW_VERSION_MICRO(vers)); 652 653 snprintf(buf, sizeof(buf), "%s %sNIC\t E/C: %s S/N: %s", 654 ai->desc, is_offload(sc) ? "R" : "", 655 sc->params.vpd.ec, sc->params.vpd.sn); 656 device_set_desc_copy(dev, buf); 657 658 snprintf(&sc->port_types[0], sizeof(sc->port_types), "%x%x%x%x", 659 sc->params.vpd.port_type[0], sc->params.vpd.port_type[1], 660 sc->params.vpd.port_type[2], sc->params.vpd.port_type[3]); 661 662 device_printf(sc->dev, "Firmware Version %s\n", &sc->fw_version[0]); 663 callout_reset(&sc->cxgb_tick_ch, hz, cxgb_tick, sc); 664 t3_add_attach_sysctls(sc); 665 out: 666 if (error) 667 cxgb_free(sc); 668 669 return (error); 670 } 671 672 /* 673 * The cxgb_controller_detach routine is called with the device is 674 * unloaded from the system. 675 */ 676 677 static int 678 cxgb_controller_detach(device_t dev) 679 { 680 struct adapter *sc; 681 682 sc = device_get_softc(dev); 683 684 cxgb_free(sc); 685 686 return (0); 687 } 688 689 /* 690 * The cxgb_free() is called by the cxgb_controller_detach() routine 691 * to tear down the structures that were built up in 692 * cxgb_controller_attach(), and should be the final piece of work 693 * done when fully unloading the driver. 694 * 695 * 696 * 1. Shutting down the threads started by the cxgb_controller_attach() 697 * routine. 698 * 2. Stopping the lower level device and all callouts (cxgb_down_locked()). 699 * 3. Detaching all of the port devices created during the 700 * cxgb_controller_attach() routine. 701 * 4. Removing the device children created via cxgb_controller_attach(). 702 * 5. Releasing PCI resources associated with the device. 703 * 6. Turning off the offload support, iff it was turned on. 704 * 7. Destroying the mutexes created in cxgb_controller_attach(). 705 * 706 */ 707 static void 708 cxgb_free(struct adapter *sc) 709 { 710 int i, nqsets = 0; 711 712 ADAPTER_LOCK(sc); 713 sc->flags |= CXGB_SHUTDOWN; 714 ADAPTER_UNLOCK(sc); 715 716 /* 717 * Make sure all child devices are gone. 718 */ 719 bus_generic_detach(sc->dev); 720 for (i = 0; i < (sc)->params.nports; i++) { 721 if (sc->portdev[i] && 722 device_delete_child(sc->dev, sc->portdev[i]) != 0) 723 device_printf(sc->dev, "failed to delete child port\n"); 724 nqsets += sc->port[i].nqsets; 725 } 726 727 /* 728 * At this point, it is as if cxgb_port_detach has run on all ports, and 729 * cxgb_down has run on the adapter. All interrupts have been silenced, 730 * all open devices have been closed. 731 */ 732 KASSERT(sc->open_device_map == 0, ("%s: device(s) still open (%x)", 733 __func__, sc->open_device_map)); 734 for (i = 0; i < sc->params.nports; i++) { 735 KASSERT(sc->port[i].ifp == NULL, ("%s: port %i undead!", 736 __func__, i)); 737 } 738 739 /* 740 * Finish off the adapter's callouts. 741 */ 742 callout_drain(&sc->cxgb_tick_ch); 743 callout_drain(&sc->sge_timer_ch); 744 745 /* 746 * Release resources grabbed under FULL_INIT_DONE by cxgb_up. The 747 * sysctls are cleaned up by the kernel linker. 748 */ 749 if (sc->flags & FULL_INIT_DONE) { 750 t3_free_sge_resources(sc, nqsets); 751 sc->flags &= ~FULL_INIT_DONE; 752 } 753 754 /* 755 * Release all interrupt resources. 756 */ 757 cxgb_teardown_interrupts(sc); 758 if (sc->flags & (USING_MSI | USING_MSIX)) { 759 device_printf(sc->dev, "releasing msi message(s)\n"); 760 pci_release_msi(sc->dev); 761 } else { 762 device_printf(sc->dev, "no msi message to release\n"); 763 } 764 765 if (sc->msix_regs_res != NULL) { 766 bus_release_resource(sc->dev, SYS_RES_MEMORY, sc->msix_regs_rid, 767 sc->msix_regs_res); 768 } 769 770 /* 771 * Free the adapter's taskqueue. 772 */ 773 if (sc->tq != NULL) { 774 taskqueue_free(sc->tq); 775 sc->tq = NULL; 776 } 777 778 if (is_offload(sc)) { 779 clrbit(&sc->registered_device_map, OFFLOAD_DEVMAP_BIT); 780 cxgb_adapter_unofld(sc); 781 } 782 783 #ifdef notyet 784 if (sc->flags & CXGB_OFLD_INIT) 785 cxgb_offload_deactivate(sc); 786 #endif 787 free(sc->filters, M_DEVBUF); 788 t3_sge_free(sc); 789 790 cxgb_offload_exit(); 791 792 if (sc->udbs_res != NULL) 793 bus_release_resource(sc->dev, SYS_RES_MEMORY, sc->udbs_rid, 794 sc->udbs_res); 795 796 if (sc->regs_res != NULL) 797 bus_release_resource(sc->dev, SYS_RES_MEMORY, sc->regs_rid, 798 sc->regs_res); 799 800 MTX_DESTROY(&sc->mdio_lock); 801 MTX_DESTROY(&sc->sge.reg_lock); 802 MTX_DESTROY(&sc->elmer_lock); 803 ADAPTER_LOCK_DEINIT(sc); 804 } 805 806 /** 807 * setup_sge_qsets - configure SGE Tx/Rx/response queues 808 * @sc: the controller softc 809 * 810 * Determines how many sets of SGE queues to use and initializes them. 811 * We support multiple queue sets per port if we have MSI-X, otherwise 812 * just one queue set per port. 813 */ 814 static int 815 setup_sge_qsets(adapter_t *sc) 816 { 817 int i, j, err, irq_idx = 0, qset_idx = 0; 818 u_int ntxq = SGE_TXQ_PER_SET; 819 820 if ((err = t3_sge_alloc(sc)) != 0) { 821 device_printf(sc->dev, "t3_sge_alloc returned %d\n", err); 822 return (err); 823 } 824 825 if (sc->params.rev > 0 && !(sc->flags & USING_MSI)) 826 irq_idx = -1; 827 828 for (i = 0; i < (sc)->params.nports; i++) { 829 struct port_info *pi = &sc->port[i]; 830 831 for (j = 0; j < pi->nqsets; j++, qset_idx++) { 832 err = t3_sge_alloc_qset(sc, qset_idx, (sc)->params.nports, 833 (sc->flags & USING_MSIX) ? qset_idx + 1 : irq_idx, 834 &sc->params.sge.qset[qset_idx], ntxq, pi); 835 if (err) { 836 t3_free_sge_resources(sc, qset_idx); 837 device_printf(sc->dev, 838 "t3_sge_alloc_qset failed with %d\n", err); 839 return (err); 840 } 841 } 842 } 843 844 return (0); 845 } 846 847 static void 848 cxgb_teardown_interrupts(adapter_t *sc) 849 { 850 int i; 851 852 for (i = 0; i < SGE_QSETS; i++) { 853 if (sc->msix_intr_tag[i] == NULL) { 854 855 /* Should have been setup fully or not at all */ 856 KASSERT(sc->msix_irq_res[i] == NULL && 857 sc->msix_irq_rid[i] == 0, 858 ("%s: half-done interrupt (%d).", __func__, i)); 859 860 continue; 861 } 862 863 bus_teardown_intr(sc->dev, sc->msix_irq_res[i], 864 sc->msix_intr_tag[i]); 865 bus_release_resource(sc->dev, SYS_RES_IRQ, sc->msix_irq_rid[i], 866 sc->msix_irq_res[i]); 867 868 sc->msix_irq_res[i] = sc->msix_intr_tag[i] = NULL; 869 sc->msix_irq_rid[i] = 0; 870 } 871 872 if (sc->intr_tag) { 873 KASSERT(sc->irq_res != NULL, 874 ("%s: half-done interrupt.", __func__)); 875 876 bus_teardown_intr(sc->dev, sc->irq_res, sc->intr_tag); 877 bus_release_resource(sc->dev, SYS_RES_IRQ, sc->irq_rid, 878 sc->irq_res); 879 880 sc->irq_res = sc->intr_tag = NULL; 881 sc->irq_rid = 0; 882 } 883 } 884 885 static int 886 cxgb_setup_interrupts(adapter_t *sc) 887 { 888 struct resource *res; 889 void *tag; 890 int i, rid, err, intr_flag = sc->flags & (USING_MSI | USING_MSIX); 891 892 sc->irq_rid = intr_flag ? 1 : 0; 893 sc->irq_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, &sc->irq_rid, 894 RF_SHAREABLE | RF_ACTIVE); 895 if (sc->irq_res == NULL) { 896 device_printf(sc->dev, "Cannot allocate interrupt (%x, %u)\n", 897 intr_flag, sc->irq_rid); 898 err = EINVAL; 899 sc->irq_rid = 0; 900 } else { 901 err = bus_setup_intr(sc->dev, sc->irq_res, 902 INTR_MPSAFE | INTR_TYPE_NET, NULL, 903 sc->cxgb_intr, sc, &sc->intr_tag); 904 905 if (err) { 906 device_printf(sc->dev, 907 "Cannot set up interrupt (%x, %u, %d)\n", 908 intr_flag, sc->irq_rid, err); 909 bus_release_resource(sc->dev, SYS_RES_IRQ, sc->irq_rid, 910 sc->irq_res); 911 sc->irq_res = sc->intr_tag = NULL; 912 sc->irq_rid = 0; 913 } 914 } 915 916 /* That's all for INTx or MSI */ 917 if (!(intr_flag & USING_MSIX) || err) 918 return (err); 919 920 for (i = 0; i < sc->msi_count - 1; i++) { 921 rid = i + 2; 922 res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, &rid, 923 RF_SHAREABLE | RF_ACTIVE); 924 if (res == NULL) { 925 device_printf(sc->dev, "Cannot allocate interrupt " 926 "for message %d\n", rid); 927 err = EINVAL; 928 break; 929 } 930 931 err = bus_setup_intr(sc->dev, res, INTR_MPSAFE | INTR_TYPE_NET, 932 NULL, t3_intr_msix, &sc->sge.qs[i], &tag); 933 if (err) { 934 device_printf(sc->dev, "Cannot set up interrupt " 935 "for message %d (%d)\n", rid, err); 936 bus_release_resource(sc->dev, SYS_RES_IRQ, rid, res); 937 break; 938 } 939 940 sc->msix_irq_rid[i] = rid; 941 sc->msix_irq_res[i] = res; 942 sc->msix_intr_tag[i] = tag; 943 } 944 945 if (err) 946 cxgb_teardown_interrupts(sc); 947 948 return (err); 949 } 950 951 952 static int 953 cxgb_port_probe(device_t dev) 954 { 955 struct port_info *p; 956 char buf[80]; 957 const char *desc; 958 959 p = device_get_softc(dev); 960 desc = p->phy.desc; 961 snprintf(buf, sizeof(buf), "Port %d %s", p->port_id, desc); 962 device_set_desc_copy(dev, buf); 963 return (0); 964 } 965 966 967 static int 968 cxgb_makedev(struct port_info *pi) 969 { 970 971 pi->port_cdev = make_dev(&cxgb_cdevsw, pi->ifp->if_dunit, 972 UID_ROOT, GID_WHEEL, 0600, "%s", if_name(pi->ifp)); 973 974 if (pi->port_cdev == NULL) 975 return (ENOMEM); 976 977 pi->port_cdev->si_drv1 = (void *)pi; 978 979 return (0); 980 } 981 982 #define CXGB_CAP (IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_HWCSUM | \ 983 IFCAP_VLAN_HWCSUM | IFCAP_TSO | IFCAP_JUMBO_MTU | IFCAP_LRO | \ 984 IFCAP_VLAN_HWTSO | IFCAP_LINKSTATE) 985 #define CXGB_CAP_ENABLE CXGB_CAP 986 987 static int 988 cxgb_port_attach(device_t dev) 989 { 990 struct port_info *p; 991 struct ifnet *ifp; 992 int err; 993 struct adapter *sc; 994 995 p = device_get_softc(dev); 996 sc = p->adapter; 997 snprintf(p->lockbuf, PORT_NAME_LEN, "cxgb port lock %d:%d", 998 device_get_unit(device_get_parent(dev)), p->port_id); 999 PORT_LOCK_INIT(p, p->lockbuf); 1000 1001 callout_init(&p->link_check_ch, CALLOUT_MPSAFE); 1002 TASK_INIT(&p->link_check_task, 0, check_link_status, p); 1003 1004 /* Allocate an ifnet object and set it up */ 1005 ifp = p->ifp = if_alloc(IFT_ETHER); 1006 if (ifp == NULL) { 1007 device_printf(dev, "Cannot allocate ifnet\n"); 1008 return (ENOMEM); 1009 } 1010 1011 if_initname(ifp, device_get_name(dev), device_get_unit(dev)); 1012 ifp->if_init = cxgb_init; 1013 ifp->if_softc = p; 1014 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 1015 ifp->if_ioctl = cxgb_ioctl; 1016 ifp->if_transmit = cxgb_transmit; 1017 ifp->if_qflush = cxgb_qflush; 1018 1019 ifp->if_capabilities = CXGB_CAP; 1020 ifp->if_capenable = CXGB_CAP_ENABLE; 1021 ifp->if_hwassist = CSUM_TCP | CSUM_UDP | CSUM_IP | CSUM_TSO; 1022 1023 /* 1024 * Disable TSO on 4-port - it isn't supported by the firmware. 1025 */ 1026 if (sc->params.nports > 2) { 1027 ifp->if_capabilities &= ~(IFCAP_TSO | IFCAP_VLAN_HWTSO); 1028 ifp->if_capenable &= ~(IFCAP_TSO | IFCAP_VLAN_HWTSO); 1029 ifp->if_hwassist &= ~CSUM_TSO; 1030 } 1031 1032 ether_ifattach(ifp, p->hw_addr); 1033 1034 #ifdef DEFAULT_JUMBO 1035 if (sc->params.nports <= 2) 1036 ifp->if_mtu = ETHERMTU_JUMBO; 1037 #endif 1038 if ((err = cxgb_makedev(p)) != 0) { 1039 printf("makedev failed %d\n", err); 1040 return (err); 1041 } 1042 1043 /* Create a list of media supported by this port */ 1044 ifmedia_init(&p->media, IFM_IMASK, cxgb_media_change, 1045 cxgb_media_status); 1046 cxgb_build_medialist(p); 1047 1048 t3_sge_init_port(p); 1049 1050 return (err); 1051 } 1052 1053 /* 1054 * cxgb_port_detach() is called via the device_detach methods when 1055 * cxgb_free() calls the bus_generic_detach. It is responsible for 1056 * removing the device from the view of the kernel, i.e. from all 1057 * interfaces lists etc. This routine is only called when the driver is 1058 * being unloaded, not when the link goes down. 1059 */ 1060 static int 1061 cxgb_port_detach(device_t dev) 1062 { 1063 struct port_info *p; 1064 struct adapter *sc; 1065 int i; 1066 1067 p = device_get_softc(dev); 1068 sc = p->adapter; 1069 1070 /* Tell cxgb_ioctl and if_init that the port is going away */ 1071 ADAPTER_LOCK(sc); 1072 SET_DOOMED(p); 1073 wakeup(&sc->flags); 1074 while (IS_BUSY(sc)) 1075 mtx_sleep(&sc->flags, &sc->lock, 0, "cxgbdtch", 0); 1076 SET_BUSY(sc); 1077 ADAPTER_UNLOCK(sc); 1078 1079 if (p->port_cdev != NULL) 1080 destroy_dev(p->port_cdev); 1081 1082 cxgb_uninit_synchronized(p); 1083 ether_ifdetach(p->ifp); 1084 1085 for (i = p->first_qset; i < p->first_qset + p->nqsets; i++) { 1086 struct sge_qset *qs = &sc->sge.qs[i]; 1087 struct sge_txq *txq = &qs->txq[TXQ_ETH]; 1088 1089 callout_drain(&txq->txq_watchdog); 1090 callout_drain(&txq->txq_timer); 1091 } 1092 1093 PORT_LOCK_DEINIT(p); 1094 if_free(p->ifp); 1095 p->ifp = NULL; 1096 1097 ADAPTER_LOCK(sc); 1098 CLR_BUSY(sc); 1099 wakeup_one(&sc->flags); 1100 ADAPTER_UNLOCK(sc); 1101 return (0); 1102 } 1103 1104 void 1105 t3_fatal_err(struct adapter *sc) 1106 { 1107 u_int fw_status[4]; 1108 1109 if (sc->flags & FULL_INIT_DONE) { 1110 t3_sge_stop(sc); 1111 t3_write_reg(sc, A_XGM_TX_CTRL, 0); 1112 t3_write_reg(sc, A_XGM_RX_CTRL, 0); 1113 t3_write_reg(sc, XGM_REG(A_XGM_TX_CTRL, 1), 0); 1114 t3_write_reg(sc, XGM_REG(A_XGM_RX_CTRL, 1), 0); 1115 t3_intr_disable(sc); 1116 } 1117 device_printf(sc->dev,"encountered fatal error, operation suspended\n"); 1118 if (!t3_cim_ctl_blk_read(sc, 0xa0, 4, fw_status)) 1119 device_printf(sc->dev, "FW_ status: 0x%x, 0x%x, 0x%x, 0x%x\n", 1120 fw_status[0], fw_status[1], fw_status[2], fw_status[3]); 1121 } 1122 1123 int 1124 t3_os_find_pci_capability(adapter_t *sc, int cap) 1125 { 1126 device_t dev; 1127 struct pci_devinfo *dinfo; 1128 pcicfgregs *cfg; 1129 uint32_t status; 1130 uint8_t ptr; 1131 1132 dev = sc->dev; 1133 dinfo = device_get_ivars(dev); 1134 cfg = &dinfo->cfg; 1135 1136 status = pci_read_config(dev, PCIR_STATUS, 2); 1137 if (!(status & PCIM_STATUS_CAPPRESENT)) 1138 return (0); 1139 1140 switch (cfg->hdrtype & PCIM_HDRTYPE) { 1141 case 0: 1142 case 1: 1143 ptr = PCIR_CAP_PTR; 1144 break; 1145 case 2: 1146 ptr = PCIR_CAP_PTR_2; 1147 break; 1148 default: 1149 return (0); 1150 break; 1151 } 1152 ptr = pci_read_config(dev, ptr, 1); 1153 1154 while (ptr != 0) { 1155 if (pci_read_config(dev, ptr + PCICAP_ID, 1) == cap) 1156 return (ptr); 1157 ptr = pci_read_config(dev, ptr + PCICAP_NEXTPTR, 1); 1158 } 1159 1160 return (0); 1161 } 1162 1163 int 1164 t3_os_pci_save_state(struct adapter *sc) 1165 { 1166 device_t dev; 1167 struct pci_devinfo *dinfo; 1168 1169 dev = sc->dev; 1170 dinfo = device_get_ivars(dev); 1171 1172 pci_cfg_save(dev, dinfo, 0); 1173 return (0); 1174 } 1175 1176 int 1177 t3_os_pci_restore_state(struct adapter *sc) 1178 { 1179 device_t dev; 1180 struct pci_devinfo *dinfo; 1181 1182 dev = sc->dev; 1183 dinfo = device_get_ivars(dev); 1184 1185 pci_cfg_restore(dev, dinfo); 1186 return (0); 1187 } 1188 1189 /** 1190 * t3_os_link_changed - handle link status changes 1191 * @sc: the adapter associated with the link change 1192 * @port_id: the port index whose link status has changed 1193 * @link_status: the new status of the link 1194 * @speed: the new speed setting 1195 * @duplex: the new duplex setting 1196 * @fc: the new flow-control setting 1197 * 1198 * This is the OS-dependent handler for link status changes. The OS 1199 * neutral handler takes care of most of the processing for these events, 1200 * then calls this handler for any OS-specific processing. 1201 */ 1202 void 1203 t3_os_link_changed(adapter_t *adapter, int port_id, int link_status, int speed, 1204 int duplex, int fc, int mac_was_reset) 1205 { 1206 struct port_info *pi = &adapter->port[port_id]; 1207 struct ifnet *ifp = pi->ifp; 1208 1209 /* no race with detach, so ifp should always be good */ 1210 KASSERT(ifp, ("%s: if detached.", __func__)); 1211 1212 /* Reapply mac settings if they were lost due to a reset */ 1213 if (mac_was_reset) { 1214 PORT_LOCK(pi); 1215 cxgb_update_mac_settings(pi); 1216 PORT_UNLOCK(pi); 1217 } 1218 1219 if (link_status) { 1220 ifp->if_baudrate = IF_Mbps(speed); 1221 if_link_state_change(ifp, LINK_STATE_UP); 1222 } else 1223 if_link_state_change(ifp, LINK_STATE_DOWN); 1224 } 1225 1226 /** 1227 * t3_os_phymod_changed - handle PHY module changes 1228 * @phy: the PHY reporting the module change 1229 * @mod_type: new module type 1230 * 1231 * This is the OS-dependent handler for PHY module changes. It is 1232 * invoked when a PHY module is removed or inserted for any OS-specific 1233 * processing. 1234 */ 1235 void t3_os_phymod_changed(struct adapter *adap, int port_id) 1236 { 1237 static const char *mod_str[] = { 1238 NULL, "SR", "LR", "LRM", "TWINAX", "TWINAX-L", "unknown" 1239 }; 1240 struct port_info *pi = &adap->port[port_id]; 1241 int mod = pi->phy.modtype; 1242 1243 if (mod != pi->media.ifm_cur->ifm_data) 1244 cxgb_build_medialist(pi); 1245 1246 if (mod == phy_modtype_none) 1247 if_printf(pi->ifp, "PHY module unplugged\n"); 1248 else { 1249 KASSERT(mod < ARRAY_SIZE(mod_str), 1250 ("invalid PHY module type %d", mod)); 1251 if_printf(pi->ifp, "%s PHY module inserted\n", mod_str[mod]); 1252 } 1253 } 1254 1255 void 1256 t3_os_set_hw_addr(adapter_t *adapter, int port_idx, u8 hw_addr[]) 1257 { 1258 1259 /* 1260 * The ifnet might not be allocated before this gets called, 1261 * as this is called early on in attach by t3_prep_adapter 1262 * save the address off in the port structure 1263 */ 1264 if (cxgb_debug) 1265 printf("set_hw_addr on idx %d addr %6D\n", port_idx, hw_addr, ":"); 1266 bcopy(hw_addr, adapter->port[port_idx].hw_addr, ETHER_ADDR_LEN); 1267 } 1268 1269 /* 1270 * Programs the XGMAC based on the settings in the ifnet. These settings 1271 * include MTU, MAC address, mcast addresses, etc. 1272 */ 1273 static void 1274 cxgb_update_mac_settings(struct port_info *p) 1275 { 1276 struct ifnet *ifp = p->ifp; 1277 struct t3_rx_mode rm; 1278 struct cmac *mac = &p->mac; 1279 int mtu, hwtagging; 1280 1281 PORT_LOCK_ASSERT_OWNED(p); 1282 1283 bcopy(IF_LLADDR(ifp), p->hw_addr, ETHER_ADDR_LEN); 1284 1285 mtu = ifp->if_mtu; 1286 if (ifp->if_capenable & IFCAP_VLAN_MTU) 1287 mtu += ETHER_VLAN_ENCAP_LEN; 1288 1289 hwtagging = (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0; 1290 1291 t3_mac_set_mtu(mac, mtu); 1292 t3_set_vlan_accel(p->adapter, 1 << p->tx_chan, hwtagging); 1293 t3_mac_set_address(mac, 0, p->hw_addr); 1294 t3_init_rx_mode(&rm, p); 1295 t3_mac_set_rx_mode(mac, &rm); 1296 } 1297 1298 1299 static int 1300 await_mgmt_replies(struct adapter *adap, unsigned long init_cnt, 1301 unsigned long n) 1302 { 1303 int attempts = 5; 1304 1305 while (adap->sge.qs[0].rspq.offload_pkts < init_cnt + n) { 1306 if (!--attempts) 1307 return (ETIMEDOUT); 1308 t3_os_sleep(10); 1309 } 1310 return 0; 1311 } 1312 1313 static int 1314 init_tp_parity(struct adapter *adap) 1315 { 1316 int i; 1317 struct mbuf *m; 1318 struct cpl_set_tcb_field *greq; 1319 unsigned long cnt = adap->sge.qs[0].rspq.offload_pkts; 1320 1321 t3_tp_set_offload_mode(adap, 1); 1322 1323 for (i = 0; i < 16; i++) { 1324 struct cpl_smt_write_req *req; 1325 1326 m = m_gethdr(M_WAITOK, MT_DATA); 1327 req = mtod(m, struct cpl_smt_write_req *); 1328 m->m_len = m->m_pkthdr.len = sizeof(*req); 1329 memset(req, 0, sizeof(*req)); 1330 req->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 1331 OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_SMT_WRITE_REQ, i)); 1332 req->iff = i; 1333 t3_mgmt_tx(adap, m); 1334 } 1335 1336 for (i = 0; i < 2048; i++) { 1337 struct cpl_l2t_write_req *req; 1338 1339 m = m_gethdr(M_WAITOK, MT_DATA); 1340 req = mtod(m, struct cpl_l2t_write_req *); 1341 m->m_len = m->m_pkthdr.len = sizeof(*req); 1342 memset(req, 0, sizeof(*req)); 1343 req->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 1344 OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_L2T_WRITE_REQ, i)); 1345 req->params = htonl(V_L2T_W_IDX(i)); 1346 t3_mgmt_tx(adap, m); 1347 } 1348 1349 for (i = 0; i < 2048; i++) { 1350 struct cpl_rte_write_req *req; 1351 1352 m = m_gethdr(M_WAITOK, MT_DATA); 1353 req = mtod(m, struct cpl_rte_write_req *); 1354 m->m_len = m->m_pkthdr.len = sizeof(*req); 1355 memset(req, 0, sizeof(*req)); 1356 req->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 1357 OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_RTE_WRITE_REQ, i)); 1358 req->l2t_idx = htonl(V_L2T_W_IDX(i)); 1359 t3_mgmt_tx(adap, m); 1360 } 1361 1362 m = m_gethdr(M_WAITOK, MT_DATA); 1363 greq = mtod(m, struct cpl_set_tcb_field *); 1364 m->m_len = m->m_pkthdr.len = sizeof(*greq); 1365 memset(greq, 0, sizeof(*greq)); 1366 greq->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 1367 OPCODE_TID(greq) = htonl(MK_OPCODE_TID(CPL_SET_TCB_FIELD, 0)); 1368 greq->mask = htobe64(1); 1369 t3_mgmt_tx(adap, m); 1370 1371 i = await_mgmt_replies(adap, cnt, 16 + 2048 + 2048 + 1); 1372 t3_tp_set_offload_mode(adap, 0); 1373 return (i); 1374 } 1375 1376 /** 1377 * setup_rss - configure Receive Side Steering (per-queue connection demux) 1378 * @adap: the adapter 1379 * 1380 * Sets up RSS to distribute packets to multiple receive queues. We 1381 * configure the RSS CPU lookup table to distribute to the number of HW 1382 * receive queues, and the response queue lookup table to narrow that 1383 * down to the response queues actually configured for each port. 1384 * We always configure the RSS mapping for two ports since the mapping 1385 * table has plenty of entries. 1386 */ 1387 static void 1388 setup_rss(adapter_t *adap) 1389 { 1390 int i; 1391 u_int nq[2]; 1392 uint8_t cpus[SGE_QSETS + 1]; 1393 uint16_t rspq_map[RSS_TABLE_SIZE]; 1394 1395 for (i = 0; i < SGE_QSETS; ++i) 1396 cpus[i] = i; 1397 cpus[SGE_QSETS] = 0xff; 1398 1399 nq[0] = nq[1] = 0; 1400 for_each_port(adap, i) { 1401 const struct port_info *pi = adap2pinfo(adap, i); 1402 1403 nq[pi->tx_chan] += pi->nqsets; 1404 } 1405 for (i = 0; i < RSS_TABLE_SIZE / 2; ++i) { 1406 rspq_map[i] = nq[0] ? i % nq[0] : 0; 1407 rspq_map[i + RSS_TABLE_SIZE / 2] = nq[1] ? i % nq[1] + nq[0] : 0; 1408 } 1409 1410 /* Calculate the reverse RSS map table */ 1411 for (i = 0; i < SGE_QSETS; ++i) 1412 adap->rrss_map[i] = 0xff; 1413 for (i = 0; i < RSS_TABLE_SIZE; ++i) 1414 if (adap->rrss_map[rspq_map[i]] == 0xff) 1415 adap->rrss_map[rspq_map[i]] = i; 1416 1417 t3_config_rss(adap, F_RQFEEDBACKENABLE | F_TNLLKPEN | F_TNLMAPEN | 1418 F_TNLPRTEN | F_TNL2TUPEN | F_TNL4TUPEN | F_OFDMAPEN | 1419 F_RRCPLMAPEN | V_RRCPLCPUSIZE(6) | F_HASHTOEPLITZ, 1420 cpus, rspq_map); 1421 1422 } 1423 1424 /* 1425 * Sends an mbuf to an offload queue driver 1426 * after dealing with any active network taps. 1427 */ 1428 static inline int 1429 offload_tx(struct t3cdev *tdev, struct mbuf *m) 1430 { 1431 int ret; 1432 1433 ret = t3_offload_tx(tdev, m); 1434 return (ret); 1435 } 1436 1437 static int 1438 write_smt_entry(struct adapter *adapter, int idx) 1439 { 1440 struct port_info *pi = &adapter->port[idx]; 1441 struct cpl_smt_write_req *req; 1442 struct mbuf *m; 1443 1444 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) 1445 return (ENOMEM); 1446 1447 req = mtod(m, struct cpl_smt_write_req *); 1448 m->m_pkthdr.len = m->m_len = sizeof(struct cpl_smt_write_req); 1449 1450 req->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 1451 OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_SMT_WRITE_REQ, idx)); 1452 req->mtu_idx = NMTUS - 1; /* should be 0 but there's a T3 bug */ 1453 req->iff = idx; 1454 memset(req->src_mac1, 0, sizeof(req->src_mac1)); 1455 memcpy(req->src_mac0, pi->hw_addr, ETHER_ADDR_LEN); 1456 1457 m_set_priority(m, 1); 1458 1459 offload_tx(&adapter->tdev, m); 1460 1461 return (0); 1462 } 1463 1464 static int 1465 init_smt(struct adapter *adapter) 1466 { 1467 int i; 1468 1469 for_each_port(adapter, i) 1470 write_smt_entry(adapter, i); 1471 return 0; 1472 } 1473 1474 static void 1475 init_port_mtus(adapter_t *adapter) 1476 { 1477 unsigned int mtus = ETHERMTU | (ETHERMTU << 16); 1478 1479 t3_write_reg(adapter, A_TP_MTU_PORT_TABLE, mtus); 1480 } 1481 1482 static void 1483 send_pktsched_cmd(struct adapter *adap, int sched, int qidx, int lo, 1484 int hi, int port) 1485 { 1486 struct mbuf *m; 1487 struct mngt_pktsched_wr *req; 1488 1489 m = m_gethdr(M_DONTWAIT, MT_DATA); 1490 if (m) { 1491 req = mtod(m, struct mngt_pktsched_wr *); 1492 req->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_MNGT)); 1493 req->mngt_opcode = FW_MNGTOPCODE_PKTSCHED_SET; 1494 req->sched = sched; 1495 req->idx = qidx; 1496 req->min = lo; 1497 req->max = hi; 1498 req->binding = port; 1499 m->m_len = m->m_pkthdr.len = sizeof(*req); 1500 t3_mgmt_tx(adap, m); 1501 } 1502 } 1503 1504 static void 1505 bind_qsets(adapter_t *sc) 1506 { 1507 int i, j; 1508 1509 for (i = 0; i < (sc)->params.nports; ++i) { 1510 const struct port_info *pi = adap2pinfo(sc, i); 1511 1512 for (j = 0; j < pi->nqsets; ++j) { 1513 send_pktsched_cmd(sc, 1, pi->first_qset + j, -1, 1514 -1, pi->tx_chan); 1515 1516 } 1517 } 1518 } 1519 1520 static void 1521 update_tpeeprom(struct adapter *adap) 1522 { 1523 const struct firmware *tpeeprom; 1524 1525 uint32_t version; 1526 unsigned int major, minor; 1527 int ret, len; 1528 char rev, name[32]; 1529 1530 t3_seeprom_read(adap, TP_SRAM_OFFSET, &version); 1531 1532 major = G_TP_VERSION_MAJOR(version); 1533 minor = G_TP_VERSION_MINOR(version); 1534 if (major == TP_VERSION_MAJOR && minor == TP_VERSION_MINOR) 1535 return; 1536 1537 rev = t3rev2char(adap); 1538 snprintf(name, sizeof(name), TPEEPROM_NAME, rev); 1539 1540 tpeeprom = firmware_get(name); 1541 if (tpeeprom == NULL) { 1542 device_printf(adap->dev, 1543 "could not load TP EEPROM: unable to load %s\n", 1544 name); 1545 return; 1546 } 1547 1548 len = tpeeprom->datasize - 4; 1549 1550 ret = t3_check_tpsram(adap, tpeeprom->data, tpeeprom->datasize); 1551 if (ret) 1552 goto release_tpeeprom; 1553 1554 if (len != TP_SRAM_LEN) { 1555 device_printf(adap->dev, 1556 "%s length is wrong len=%d expected=%d\n", name, 1557 len, TP_SRAM_LEN); 1558 return; 1559 } 1560 1561 ret = set_eeprom(&adap->port[0], tpeeprom->data, tpeeprom->datasize, 1562 TP_SRAM_OFFSET); 1563 1564 if (!ret) { 1565 device_printf(adap->dev, 1566 "Protocol SRAM image updated in EEPROM to %d.%d.%d\n", 1567 TP_VERSION_MAJOR, TP_VERSION_MINOR, TP_VERSION_MICRO); 1568 } else 1569 device_printf(adap->dev, 1570 "Protocol SRAM image update in EEPROM failed\n"); 1571 1572 release_tpeeprom: 1573 firmware_put(tpeeprom, FIRMWARE_UNLOAD); 1574 1575 return; 1576 } 1577 1578 static int 1579 update_tpsram(struct adapter *adap) 1580 { 1581 const struct firmware *tpsram; 1582 int ret; 1583 char rev, name[32]; 1584 1585 rev = t3rev2char(adap); 1586 snprintf(name, sizeof(name), TPSRAM_NAME, rev); 1587 1588 update_tpeeprom(adap); 1589 1590 tpsram = firmware_get(name); 1591 if (tpsram == NULL){ 1592 device_printf(adap->dev, "could not load TP SRAM\n"); 1593 return (EINVAL); 1594 } else 1595 device_printf(adap->dev, "updating TP SRAM\n"); 1596 1597 ret = t3_check_tpsram(adap, tpsram->data, tpsram->datasize); 1598 if (ret) 1599 goto release_tpsram; 1600 1601 ret = t3_set_proto_sram(adap, tpsram->data); 1602 if (ret) 1603 device_printf(adap->dev, "loading protocol SRAM failed\n"); 1604 1605 release_tpsram: 1606 firmware_put(tpsram, FIRMWARE_UNLOAD); 1607 1608 return ret; 1609 } 1610 1611 /** 1612 * cxgb_up - enable the adapter 1613 * @adap: adapter being enabled 1614 * 1615 * Called when the first port is enabled, this function performs the 1616 * actions necessary to make an adapter operational, such as completing 1617 * the initialization of HW modules, and enabling interrupts. 1618 */ 1619 static int 1620 cxgb_up(struct adapter *sc) 1621 { 1622 int err = 0; 1623 unsigned int mxf = t3_mc5_size(&sc->mc5) - MC5_MIN_TIDS; 1624 1625 KASSERT(sc->open_device_map == 0, ("%s: device(s) already open (%x)", 1626 __func__, sc->open_device_map)); 1627 1628 if ((sc->flags & FULL_INIT_DONE) == 0) { 1629 1630 ADAPTER_LOCK_ASSERT_NOTOWNED(sc); 1631 1632 if ((sc->flags & FW_UPTODATE) == 0) 1633 if ((err = upgrade_fw(sc))) 1634 goto out; 1635 1636 if ((sc->flags & TPS_UPTODATE) == 0) 1637 if ((err = update_tpsram(sc))) 1638 goto out; 1639 1640 if (is_offload(sc) && nfilters != 0) { 1641 sc->params.mc5.nservers = 0; 1642 1643 if (nfilters < 0) 1644 sc->params.mc5.nfilters = mxf; 1645 else 1646 sc->params.mc5.nfilters = min(nfilters, mxf); 1647 } 1648 1649 err = t3_init_hw(sc, 0); 1650 if (err) 1651 goto out; 1652 1653 t3_set_reg_field(sc, A_TP_PARA_REG5, 0, F_RXDDPOFFINIT); 1654 t3_write_reg(sc, A_ULPRX_TDDP_PSZ, V_HPZ0(PAGE_SHIFT - 12)); 1655 1656 err = setup_sge_qsets(sc); 1657 if (err) 1658 goto out; 1659 1660 alloc_filters(sc); 1661 setup_rss(sc); 1662 1663 t3_intr_clear(sc); 1664 err = cxgb_setup_interrupts(sc); 1665 if (err) 1666 goto out; 1667 1668 t3_add_configured_sysctls(sc); 1669 sc->flags |= FULL_INIT_DONE; 1670 } 1671 1672 t3_intr_clear(sc); 1673 t3_sge_start(sc); 1674 t3_intr_enable(sc); 1675 1676 if (sc->params.rev >= T3_REV_C && !(sc->flags & TP_PARITY_INIT) && 1677 is_offload(sc) && init_tp_parity(sc) == 0) 1678 sc->flags |= TP_PARITY_INIT; 1679 1680 if (sc->flags & TP_PARITY_INIT) { 1681 t3_write_reg(sc, A_TP_INT_CAUSE, F_CMCACHEPERR | F_ARPLUTPERR); 1682 t3_write_reg(sc, A_TP_INT_ENABLE, 0x7fbfffff); 1683 } 1684 1685 if (!(sc->flags & QUEUES_BOUND)) { 1686 bind_qsets(sc); 1687 setup_hw_filters(sc); 1688 sc->flags |= QUEUES_BOUND; 1689 } 1690 1691 t3_sge_reset_adapter(sc); 1692 out: 1693 return (err); 1694 } 1695 1696 /* 1697 * Called when the last open device is closed. Does NOT undo all of cxgb_up's 1698 * work. Specifically, the resources grabbed under FULL_INIT_DONE are released 1699 * during controller_detach, not here. 1700 */ 1701 static void 1702 cxgb_down(struct adapter *sc) 1703 { 1704 t3_sge_stop(sc); 1705 t3_intr_disable(sc); 1706 } 1707 1708 static int 1709 offload_open(struct port_info *pi) 1710 { 1711 struct adapter *sc = pi->adapter; 1712 struct t3cdev *tdev = &sc->tdev; 1713 1714 setbit(&sc->open_device_map, OFFLOAD_DEVMAP_BIT); 1715 1716 t3_tp_set_offload_mode(sc, 1); 1717 tdev->lldev = pi->ifp; 1718 init_port_mtus(sc); 1719 t3_load_mtus(sc, sc->params.mtus, sc->params.a_wnd, sc->params.b_wnd, 1720 sc->params.rev == 0 ? sc->port[0].ifp->if_mtu : 0xffff); 1721 init_smt(sc); 1722 cxgb_add_clients(tdev); 1723 1724 return (0); 1725 } 1726 1727 static int 1728 offload_close(struct t3cdev *tdev) 1729 { 1730 struct adapter *adapter = tdev2adap(tdev); 1731 1732 if (!isset(&adapter->open_device_map, OFFLOAD_DEVMAP_BIT)) 1733 return (0); 1734 1735 /* Call back all registered clients */ 1736 cxgb_remove_clients(tdev); 1737 1738 tdev->lldev = NULL; 1739 cxgb_set_dummy_ops(tdev); 1740 t3_tp_set_offload_mode(adapter, 0); 1741 1742 clrbit(&adapter->open_device_map, OFFLOAD_DEVMAP_BIT); 1743 1744 return (0); 1745 } 1746 1747 /* 1748 * if_init for cxgb ports. 1749 */ 1750 static void 1751 cxgb_init(void *arg) 1752 { 1753 struct port_info *p = arg; 1754 struct adapter *sc = p->adapter; 1755 1756 ADAPTER_LOCK(sc); 1757 cxgb_init_locked(p); /* releases adapter lock */ 1758 ADAPTER_LOCK_ASSERT_NOTOWNED(sc); 1759 } 1760 1761 static int 1762 cxgb_init_locked(struct port_info *p) 1763 { 1764 struct adapter *sc = p->adapter; 1765 struct ifnet *ifp = p->ifp; 1766 struct cmac *mac = &p->mac; 1767 int i, rc = 0, may_sleep = 0, gave_up_lock = 0; 1768 1769 ADAPTER_LOCK_ASSERT_OWNED(sc); 1770 1771 while (!IS_DOOMED(p) && IS_BUSY(sc)) { 1772 gave_up_lock = 1; 1773 if (mtx_sleep(&sc->flags, &sc->lock, PCATCH, "cxgbinit", 0)) { 1774 rc = EINTR; 1775 goto done; 1776 } 1777 } 1778 if (IS_DOOMED(p)) { 1779 rc = ENXIO; 1780 goto done; 1781 } 1782 KASSERT(!IS_BUSY(sc), ("%s: controller busy.", __func__)); 1783 1784 /* 1785 * The code that runs during one-time adapter initialization can sleep 1786 * so it's important not to hold any locks across it. 1787 */ 1788 may_sleep = sc->flags & FULL_INIT_DONE ? 0 : 1; 1789 1790 if (may_sleep) { 1791 SET_BUSY(sc); 1792 gave_up_lock = 1; 1793 ADAPTER_UNLOCK(sc); 1794 } 1795 1796 if (sc->open_device_map == 0) { 1797 if ((rc = cxgb_up(sc)) != 0) 1798 goto done; 1799 1800 if (is_offload(sc) && !ofld_disable && offload_open(p)) 1801 log(LOG_WARNING, 1802 "Could not initialize offload capabilities\n"); 1803 } 1804 1805 PORT_LOCK(p); 1806 if (isset(&sc->open_device_map, p->port_id) && 1807 (ifp->if_drv_flags & IFF_DRV_RUNNING)) { 1808 PORT_UNLOCK(p); 1809 goto done; 1810 } 1811 t3_port_intr_enable(sc, p->port_id); 1812 if (!mac->multiport) 1813 t3_mac_init(mac); 1814 cxgb_update_mac_settings(p); 1815 t3_link_start(&p->phy, mac, &p->link_config); 1816 t3_mac_enable(mac, MAC_DIRECTION_RX | MAC_DIRECTION_TX); 1817 ifp->if_drv_flags |= IFF_DRV_RUNNING; 1818 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 1819 PORT_UNLOCK(p); 1820 1821 for (i = p->first_qset; i < p->first_qset + p->nqsets; i++) { 1822 struct sge_qset *qs = &sc->sge.qs[i]; 1823 struct sge_txq *txq = &qs->txq[TXQ_ETH]; 1824 1825 callout_reset_on(&txq->txq_watchdog, hz, cxgb_tx_watchdog, qs, 1826 txq->txq_watchdog.c_cpu); 1827 } 1828 1829 /* all ok */ 1830 setbit(&sc->open_device_map, p->port_id); 1831 callout_reset(&p->link_check_ch, 1832 p->phy.caps & SUPPORTED_LINK_IRQ ? hz * 3 : hz / 4, 1833 link_check_callout, p); 1834 1835 done: 1836 if (may_sleep) { 1837 ADAPTER_LOCK(sc); 1838 KASSERT(IS_BUSY(sc), ("%s: controller not busy.", __func__)); 1839 CLR_BUSY(sc); 1840 } 1841 if (gave_up_lock) 1842 wakeup_one(&sc->flags); 1843 ADAPTER_UNLOCK(sc); 1844 return (rc); 1845 } 1846 1847 static int 1848 cxgb_uninit_locked(struct port_info *p) 1849 { 1850 struct adapter *sc = p->adapter; 1851 int rc; 1852 1853 ADAPTER_LOCK_ASSERT_OWNED(sc); 1854 1855 while (!IS_DOOMED(p) && IS_BUSY(sc)) { 1856 if (mtx_sleep(&sc->flags, &sc->lock, PCATCH, "cxgbunin", 0)) { 1857 rc = EINTR; 1858 goto done; 1859 } 1860 } 1861 if (IS_DOOMED(p)) { 1862 rc = ENXIO; 1863 goto done; 1864 } 1865 KASSERT(!IS_BUSY(sc), ("%s: controller busy.", __func__)); 1866 SET_BUSY(sc); 1867 ADAPTER_UNLOCK(sc); 1868 1869 rc = cxgb_uninit_synchronized(p); 1870 1871 ADAPTER_LOCK(sc); 1872 KASSERT(IS_BUSY(sc), ("%s: controller not busy.", __func__)); 1873 CLR_BUSY(sc); 1874 wakeup_one(&sc->flags); 1875 done: 1876 ADAPTER_UNLOCK(sc); 1877 return (rc); 1878 } 1879 1880 /* 1881 * Called on "ifconfig down", and from port_detach 1882 */ 1883 static int 1884 cxgb_uninit_synchronized(struct port_info *pi) 1885 { 1886 struct adapter *sc = pi->adapter; 1887 struct ifnet *ifp = pi->ifp; 1888 1889 /* 1890 * taskqueue_drain may cause a deadlock if the adapter lock is held. 1891 */ 1892 ADAPTER_LOCK_ASSERT_NOTOWNED(sc); 1893 1894 /* 1895 * Clear this port's bit from the open device map, and then drain all 1896 * the tasks that can access/manipulate this port's port_info or ifp. 1897 * We disable this port's interrupts here and so the slow/ext 1898 * interrupt tasks won't be enqueued. The tick task will continue to 1899 * be enqueued every second but the runs after this drain will not see 1900 * this port in the open device map. 1901 * 1902 * A well behaved task must take open_device_map into account and ignore 1903 * ports that are not open. 1904 */ 1905 clrbit(&sc->open_device_map, pi->port_id); 1906 t3_port_intr_disable(sc, pi->port_id); 1907 taskqueue_drain(sc->tq, &sc->slow_intr_task); 1908 taskqueue_drain(sc->tq, &sc->tick_task); 1909 1910 callout_drain(&pi->link_check_ch); 1911 taskqueue_drain(sc->tq, &pi->link_check_task); 1912 1913 PORT_LOCK(pi); 1914 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); 1915 1916 /* disable pause frames */ 1917 t3_set_reg_field(sc, A_XGM_TX_CFG + pi->mac.offset, F_TXPAUSEEN, 0); 1918 1919 /* Reset RX FIFO HWM */ 1920 t3_set_reg_field(sc, A_XGM_RXFIFO_CFG + pi->mac.offset, 1921 V_RXFIFOPAUSEHWM(M_RXFIFOPAUSEHWM), 0); 1922 1923 DELAY(100 * 1000); 1924 1925 /* Wait for TXFIFO empty */ 1926 t3_wait_op_done(sc, A_XGM_TXFIFO_CFG + pi->mac.offset, 1927 F_TXFIFO_EMPTY, 1, 20, 5); 1928 1929 DELAY(100 * 1000); 1930 t3_mac_disable(&pi->mac, MAC_DIRECTION_RX); 1931 1932 1933 pi->phy.ops->power_down(&pi->phy, 1); 1934 1935 PORT_UNLOCK(pi); 1936 1937 pi->link_config.link_ok = 0; 1938 t3_os_link_changed(sc, pi->port_id, 0, 0, 0, 0, 0); 1939 1940 if ((sc->open_device_map & PORT_MASK) == 0) 1941 offload_close(&sc->tdev); 1942 1943 if (sc->open_device_map == 0) 1944 cxgb_down(pi->adapter); 1945 1946 return (0); 1947 } 1948 1949 /* 1950 * Mark lro enabled or disabled in all qsets for this port 1951 */ 1952 static int 1953 cxgb_set_lro(struct port_info *p, int enabled) 1954 { 1955 int i; 1956 struct adapter *adp = p->adapter; 1957 struct sge_qset *q; 1958 1959 for (i = 0; i < p->nqsets; i++) { 1960 q = &adp->sge.qs[p->first_qset + i]; 1961 q->lro.enabled = (enabled != 0); 1962 } 1963 return (0); 1964 } 1965 1966 static int 1967 cxgb_ioctl(struct ifnet *ifp, unsigned long command, caddr_t data) 1968 { 1969 struct port_info *p = ifp->if_softc; 1970 struct adapter *sc = p->adapter; 1971 struct ifreq *ifr = (struct ifreq *)data; 1972 int flags, error = 0, mtu; 1973 uint32_t mask; 1974 1975 switch (command) { 1976 case SIOCSIFMTU: 1977 ADAPTER_LOCK(sc); 1978 error = IS_DOOMED(p) ? ENXIO : (IS_BUSY(sc) ? EBUSY : 0); 1979 if (error) { 1980 fail: 1981 ADAPTER_UNLOCK(sc); 1982 return (error); 1983 } 1984 1985 mtu = ifr->ifr_mtu; 1986 if ((mtu < ETHERMIN) || (mtu > ETHERMTU_JUMBO)) { 1987 error = EINVAL; 1988 } else { 1989 ifp->if_mtu = mtu; 1990 PORT_LOCK(p); 1991 cxgb_update_mac_settings(p); 1992 PORT_UNLOCK(p); 1993 } 1994 ADAPTER_UNLOCK(sc); 1995 break; 1996 case SIOCSIFFLAGS: 1997 ADAPTER_LOCK(sc); 1998 if (IS_DOOMED(p)) { 1999 error = ENXIO; 2000 goto fail; 2001 } 2002 if (ifp->if_flags & IFF_UP) { 2003 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 2004 flags = p->if_flags; 2005 if (((ifp->if_flags ^ flags) & IFF_PROMISC) || 2006 ((ifp->if_flags ^ flags) & IFF_ALLMULTI)) { 2007 if (IS_BUSY(sc)) { 2008 error = EBUSY; 2009 goto fail; 2010 } 2011 PORT_LOCK(p); 2012 cxgb_update_mac_settings(p); 2013 PORT_UNLOCK(p); 2014 } 2015 ADAPTER_UNLOCK(sc); 2016 } else 2017 error = cxgb_init_locked(p); 2018 p->if_flags = ifp->if_flags; 2019 } else if (ifp->if_drv_flags & IFF_DRV_RUNNING) 2020 error = cxgb_uninit_locked(p); 2021 else 2022 ADAPTER_UNLOCK(sc); 2023 2024 ADAPTER_LOCK_ASSERT_NOTOWNED(sc); 2025 break; 2026 case SIOCADDMULTI: 2027 case SIOCDELMULTI: 2028 ADAPTER_LOCK(sc); 2029 error = IS_DOOMED(p) ? ENXIO : (IS_BUSY(sc) ? EBUSY : 0); 2030 if (error) 2031 goto fail; 2032 2033 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 2034 PORT_LOCK(p); 2035 cxgb_update_mac_settings(p); 2036 PORT_UNLOCK(p); 2037 } 2038 ADAPTER_UNLOCK(sc); 2039 2040 break; 2041 case SIOCSIFCAP: 2042 ADAPTER_LOCK(sc); 2043 error = IS_DOOMED(p) ? ENXIO : (IS_BUSY(sc) ? EBUSY : 0); 2044 if (error) 2045 goto fail; 2046 2047 mask = ifr->ifr_reqcap ^ ifp->if_capenable; 2048 if (mask & IFCAP_TXCSUM) { 2049 ifp->if_capenable ^= IFCAP_TXCSUM; 2050 ifp->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP); 2051 2052 if (IFCAP_TSO & ifp->if_capenable && 2053 !(IFCAP_TXCSUM & ifp->if_capenable)) { 2054 ifp->if_capenable &= ~IFCAP_TSO; 2055 ifp->if_hwassist &= ~CSUM_TSO; 2056 if_printf(ifp, 2057 "tso disabled due to -txcsum.\n"); 2058 } 2059 } 2060 if (mask & IFCAP_RXCSUM) 2061 ifp->if_capenable ^= IFCAP_RXCSUM; 2062 if (mask & IFCAP_TSO) { 2063 ifp->if_capenable ^= IFCAP_TSO; 2064 2065 if (IFCAP_TSO & ifp->if_capenable) { 2066 if (IFCAP_TXCSUM & ifp->if_capenable) 2067 ifp->if_hwassist |= CSUM_TSO; 2068 else { 2069 ifp->if_capenable &= ~IFCAP_TSO; 2070 ifp->if_hwassist &= ~CSUM_TSO; 2071 if_printf(ifp, 2072 "enable txcsum first.\n"); 2073 error = EAGAIN; 2074 } 2075 } else 2076 ifp->if_hwassist &= ~CSUM_TSO; 2077 } 2078 if (mask & IFCAP_LRO) { 2079 ifp->if_capenable ^= IFCAP_LRO; 2080 2081 /* Safe to do this even if cxgb_up not called yet */ 2082 cxgb_set_lro(p, ifp->if_capenable & IFCAP_LRO); 2083 } 2084 if (mask & IFCAP_VLAN_HWTAGGING) { 2085 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; 2086 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 2087 PORT_LOCK(p); 2088 cxgb_update_mac_settings(p); 2089 PORT_UNLOCK(p); 2090 } 2091 } 2092 if (mask & IFCAP_VLAN_MTU) { 2093 ifp->if_capenable ^= IFCAP_VLAN_MTU; 2094 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 2095 PORT_LOCK(p); 2096 cxgb_update_mac_settings(p); 2097 PORT_UNLOCK(p); 2098 } 2099 } 2100 if (mask & IFCAP_VLAN_HWTSO) 2101 ifp->if_capenable ^= IFCAP_VLAN_HWTSO; 2102 if (mask & IFCAP_VLAN_HWCSUM) 2103 ifp->if_capenable ^= IFCAP_VLAN_HWCSUM; 2104 2105 #ifdef VLAN_CAPABILITIES 2106 VLAN_CAPABILITIES(ifp); 2107 #endif 2108 ADAPTER_UNLOCK(sc); 2109 break; 2110 case SIOCSIFMEDIA: 2111 case SIOCGIFMEDIA: 2112 error = ifmedia_ioctl(ifp, ifr, &p->media, command); 2113 break; 2114 default: 2115 error = ether_ioctl(ifp, command, data); 2116 } 2117 2118 return (error); 2119 } 2120 2121 static int 2122 cxgb_media_change(struct ifnet *ifp) 2123 { 2124 return (EOPNOTSUPP); 2125 } 2126 2127 /* 2128 * Translates phy->modtype to the correct Ethernet media subtype. 2129 */ 2130 static int 2131 cxgb_ifm_type(int mod) 2132 { 2133 switch (mod) { 2134 case phy_modtype_sr: 2135 return (IFM_10G_SR); 2136 case phy_modtype_lr: 2137 return (IFM_10G_LR); 2138 case phy_modtype_lrm: 2139 return (IFM_10G_LRM); 2140 case phy_modtype_twinax: 2141 return (IFM_10G_TWINAX); 2142 case phy_modtype_twinax_long: 2143 return (IFM_10G_TWINAX_LONG); 2144 case phy_modtype_none: 2145 return (IFM_NONE); 2146 case phy_modtype_unknown: 2147 return (IFM_UNKNOWN); 2148 } 2149 2150 KASSERT(0, ("%s: modtype %d unknown", __func__, mod)); 2151 return (IFM_UNKNOWN); 2152 } 2153 2154 /* 2155 * Rebuilds the ifmedia list for this port, and sets the current media. 2156 */ 2157 static void 2158 cxgb_build_medialist(struct port_info *p) 2159 { 2160 struct cphy *phy = &p->phy; 2161 struct ifmedia *media = &p->media; 2162 int mod = phy->modtype; 2163 int m = IFM_ETHER | IFM_FDX; 2164 2165 PORT_LOCK(p); 2166 2167 ifmedia_removeall(media); 2168 if (phy->caps & SUPPORTED_TP && phy->caps & SUPPORTED_Autoneg) { 2169 /* Copper (RJ45) */ 2170 2171 if (phy->caps & SUPPORTED_10000baseT_Full) 2172 ifmedia_add(media, m | IFM_10G_T, mod, NULL); 2173 2174 if (phy->caps & SUPPORTED_1000baseT_Full) 2175 ifmedia_add(media, m | IFM_1000_T, mod, NULL); 2176 2177 if (phy->caps & SUPPORTED_100baseT_Full) 2178 ifmedia_add(media, m | IFM_100_TX, mod, NULL); 2179 2180 if (phy->caps & SUPPORTED_10baseT_Full) 2181 ifmedia_add(media, m | IFM_10_T, mod, NULL); 2182 2183 ifmedia_add(media, IFM_ETHER | IFM_AUTO, mod, NULL); 2184 ifmedia_set(media, IFM_ETHER | IFM_AUTO); 2185 2186 } else if (phy->caps & SUPPORTED_TP) { 2187 /* Copper (CX4) */ 2188 2189 KASSERT(phy->caps & SUPPORTED_10000baseT_Full, 2190 ("%s: unexpected cap 0x%x", __func__, phy->caps)); 2191 2192 ifmedia_add(media, m | IFM_10G_CX4, mod, NULL); 2193 ifmedia_set(media, m | IFM_10G_CX4); 2194 2195 } else if (phy->caps & SUPPORTED_FIBRE && 2196 phy->caps & SUPPORTED_10000baseT_Full) { 2197 /* 10G optical (but includes SFP+ twinax) */ 2198 2199 m |= cxgb_ifm_type(mod); 2200 if (IFM_SUBTYPE(m) == IFM_NONE) 2201 m &= ~IFM_FDX; 2202 2203 ifmedia_add(media, m, mod, NULL); 2204 ifmedia_set(media, m); 2205 2206 } else if (phy->caps & SUPPORTED_FIBRE && 2207 phy->caps & SUPPORTED_1000baseT_Full) { 2208 /* 1G optical */ 2209 2210 /* XXX: Lie and claim to be SX, could actually be any 1G-X */ 2211 ifmedia_add(media, m | IFM_1000_SX, mod, NULL); 2212 ifmedia_set(media, m | IFM_1000_SX); 2213 2214 } else { 2215 KASSERT(0, ("%s: don't know how to handle 0x%x.", __func__, 2216 phy->caps)); 2217 } 2218 2219 PORT_UNLOCK(p); 2220 } 2221 2222 static void 2223 cxgb_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) 2224 { 2225 struct port_info *p = ifp->if_softc; 2226 struct ifmedia_entry *cur = p->media.ifm_cur; 2227 int speed = p->link_config.speed; 2228 2229 if (cur->ifm_data != p->phy.modtype) { 2230 cxgb_build_medialist(p); 2231 cur = p->media.ifm_cur; 2232 } 2233 2234 ifmr->ifm_status = IFM_AVALID; 2235 if (!p->link_config.link_ok) 2236 return; 2237 2238 ifmr->ifm_status |= IFM_ACTIVE; 2239 2240 /* 2241 * active and current will differ iff current media is autoselect. That 2242 * can happen only for copper RJ45. 2243 */ 2244 if (IFM_SUBTYPE(cur->ifm_media) != IFM_AUTO) 2245 return; 2246 KASSERT(p->phy.caps & SUPPORTED_TP && p->phy.caps & SUPPORTED_Autoneg, 2247 ("%s: unexpected PHY caps 0x%x", __func__, p->phy.caps)); 2248 2249 ifmr->ifm_active = IFM_ETHER | IFM_FDX; 2250 if (speed == SPEED_10000) 2251 ifmr->ifm_active |= IFM_10G_T; 2252 else if (speed == SPEED_1000) 2253 ifmr->ifm_active |= IFM_1000_T; 2254 else if (speed == SPEED_100) 2255 ifmr->ifm_active |= IFM_100_TX; 2256 else if (speed == SPEED_10) 2257 ifmr->ifm_active |= IFM_10_T; 2258 else 2259 KASSERT(0, ("%s: link up but speed unknown (%u)", __func__, 2260 speed)); 2261 } 2262 2263 static void 2264 cxgb_async_intr(void *data) 2265 { 2266 adapter_t *sc = data; 2267 2268 t3_write_reg(sc, A_PL_INT_ENABLE0, 0); 2269 (void) t3_read_reg(sc, A_PL_INT_ENABLE0); 2270 taskqueue_enqueue(sc->tq, &sc->slow_intr_task); 2271 } 2272 2273 static void 2274 link_check_callout(void *arg) 2275 { 2276 struct port_info *pi = arg; 2277 struct adapter *sc = pi->adapter; 2278 2279 if (!isset(&sc->open_device_map, pi->port_id)) 2280 return; 2281 2282 taskqueue_enqueue(sc->tq, &pi->link_check_task); 2283 } 2284 2285 static void 2286 check_link_status(void *arg, int pending) 2287 { 2288 struct port_info *pi = arg; 2289 struct adapter *sc = pi->adapter; 2290 2291 if (!isset(&sc->open_device_map, pi->port_id)) 2292 return; 2293 2294 t3_link_changed(sc, pi->port_id); 2295 2296 if (pi->link_fault || !(pi->phy.caps & SUPPORTED_LINK_IRQ)) 2297 callout_reset(&pi->link_check_ch, hz, link_check_callout, pi); 2298 } 2299 2300 void 2301 t3_os_link_intr(struct port_info *pi) 2302 { 2303 /* 2304 * Schedule a link check in the near future. If the link is flapping 2305 * rapidly we'll keep resetting the callout and delaying the check until 2306 * things stabilize a bit. 2307 */ 2308 callout_reset(&pi->link_check_ch, hz / 4, link_check_callout, pi); 2309 } 2310 2311 static void 2312 check_t3b2_mac(struct adapter *sc) 2313 { 2314 int i; 2315 2316 if (sc->flags & CXGB_SHUTDOWN) 2317 return; 2318 2319 for_each_port(sc, i) { 2320 struct port_info *p = &sc->port[i]; 2321 int status; 2322 #ifdef INVARIANTS 2323 struct ifnet *ifp = p->ifp; 2324 #endif 2325 2326 if (!isset(&sc->open_device_map, p->port_id) || p->link_fault || 2327 !p->link_config.link_ok) 2328 continue; 2329 2330 KASSERT(ifp->if_drv_flags & IFF_DRV_RUNNING, 2331 ("%s: state mismatch (drv_flags %x, device_map %x)", 2332 __func__, ifp->if_drv_flags, sc->open_device_map)); 2333 2334 PORT_LOCK(p); 2335 status = t3b2_mac_watchdog_task(&p->mac); 2336 if (status == 1) 2337 p->mac.stats.num_toggled++; 2338 else if (status == 2) { 2339 struct cmac *mac = &p->mac; 2340 2341 cxgb_update_mac_settings(p); 2342 t3_link_start(&p->phy, mac, &p->link_config); 2343 t3_mac_enable(mac, MAC_DIRECTION_RX | MAC_DIRECTION_TX); 2344 t3_port_intr_enable(sc, p->port_id); 2345 p->mac.stats.num_resets++; 2346 } 2347 PORT_UNLOCK(p); 2348 } 2349 } 2350 2351 static void 2352 cxgb_tick(void *arg) 2353 { 2354 adapter_t *sc = (adapter_t *)arg; 2355 2356 if (sc->flags & CXGB_SHUTDOWN) 2357 return; 2358 2359 taskqueue_enqueue(sc->tq, &sc->tick_task); 2360 callout_reset(&sc->cxgb_tick_ch, hz, cxgb_tick, sc); 2361 } 2362 2363 static void 2364 cxgb_tick_handler(void *arg, int count) 2365 { 2366 adapter_t *sc = (adapter_t *)arg; 2367 const struct adapter_params *p = &sc->params; 2368 int i; 2369 uint32_t cause, reset; 2370 2371 if (sc->flags & CXGB_SHUTDOWN || !(sc->flags & FULL_INIT_DONE)) 2372 return; 2373 2374 if (p->rev == T3_REV_B2 && p->nports < 4 && sc->open_device_map) 2375 check_t3b2_mac(sc); 2376 2377 cause = t3_read_reg(sc, A_SG_INT_CAUSE) & (F_RSPQSTARVE | F_FLEMPTY); 2378 if (cause) { 2379 struct sge_qset *qs = &sc->sge.qs[0]; 2380 uint32_t mask, v; 2381 2382 v = t3_read_reg(sc, A_SG_RSPQ_FL_STATUS) & ~0xff00; 2383 2384 mask = 1; 2385 for (i = 0; i < SGE_QSETS; i++) { 2386 if (v & mask) 2387 qs[i].rspq.starved++; 2388 mask <<= 1; 2389 } 2390 2391 mask <<= SGE_QSETS; /* skip RSPQXDISABLED */ 2392 2393 for (i = 0; i < SGE_QSETS * 2; i++) { 2394 if (v & mask) { 2395 qs[i / 2].fl[i % 2].empty++; 2396 } 2397 mask <<= 1; 2398 } 2399 2400 /* clear */ 2401 t3_write_reg(sc, A_SG_RSPQ_FL_STATUS, v); 2402 t3_write_reg(sc, A_SG_INT_CAUSE, cause); 2403 } 2404 2405 for (i = 0; i < sc->params.nports; i++) { 2406 struct port_info *pi = &sc->port[i]; 2407 struct ifnet *ifp = pi->ifp; 2408 struct cmac *mac = &pi->mac; 2409 struct mac_stats *mstats = &mac->stats; 2410 int drops, j; 2411 2412 if (!isset(&sc->open_device_map, pi->port_id)) 2413 continue; 2414 2415 PORT_LOCK(pi); 2416 t3_mac_update_stats(mac); 2417 PORT_UNLOCK(pi); 2418 2419 ifp->if_opackets = mstats->tx_frames; 2420 ifp->if_ipackets = mstats->rx_frames; 2421 ifp->if_obytes = mstats->tx_octets; 2422 ifp->if_ibytes = mstats->rx_octets; 2423 ifp->if_omcasts = mstats->tx_mcast_frames; 2424 ifp->if_imcasts = mstats->rx_mcast_frames; 2425 ifp->if_collisions = mstats->tx_total_collisions; 2426 ifp->if_iqdrops = mstats->rx_cong_drops; 2427 2428 drops = 0; 2429 for (j = pi->first_qset; j < pi->first_qset + pi->nqsets; j++) 2430 drops += sc->sge.qs[j].txq[TXQ_ETH].txq_mr->br_drops; 2431 ifp->if_snd.ifq_drops = drops; 2432 2433 ifp->if_oerrors = 2434 mstats->tx_excess_collisions + 2435 mstats->tx_underrun + 2436 mstats->tx_len_errs + 2437 mstats->tx_mac_internal_errs + 2438 mstats->tx_excess_deferral + 2439 mstats->tx_fcs_errs; 2440 ifp->if_ierrors = 2441 mstats->rx_jabber + 2442 mstats->rx_data_errs + 2443 mstats->rx_sequence_errs + 2444 mstats->rx_runt + 2445 mstats->rx_too_long + 2446 mstats->rx_mac_internal_errs + 2447 mstats->rx_short + 2448 mstats->rx_fcs_errs; 2449 2450 if (mac->multiport) 2451 continue; 2452 2453 /* Count rx fifo overflows, once per second */ 2454 cause = t3_read_reg(sc, A_XGM_INT_CAUSE + mac->offset); 2455 reset = 0; 2456 if (cause & F_RXFIFO_OVERFLOW) { 2457 mac->stats.rx_fifo_ovfl++; 2458 reset |= F_RXFIFO_OVERFLOW; 2459 } 2460 t3_write_reg(sc, A_XGM_INT_CAUSE + mac->offset, reset); 2461 } 2462 } 2463 2464 static void 2465 touch_bars(device_t dev) 2466 { 2467 /* 2468 * Don't enable yet 2469 */ 2470 #if !defined(__LP64__) && 0 2471 u32 v; 2472 2473 pci_read_config_dword(pdev, PCI_BASE_ADDRESS_1, &v); 2474 pci_write_config_dword(pdev, PCI_BASE_ADDRESS_1, v); 2475 pci_read_config_dword(pdev, PCI_BASE_ADDRESS_3, &v); 2476 pci_write_config_dword(pdev, PCI_BASE_ADDRESS_3, v); 2477 pci_read_config_dword(pdev, PCI_BASE_ADDRESS_5, &v); 2478 pci_write_config_dword(pdev, PCI_BASE_ADDRESS_5, v); 2479 #endif 2480 } 2481 2482 static int 2483 set_eeprom(struct port_info *pi, const uint8_t *data, int len, int offset) 2484 { 2485 uint8_t *buf; 2486 int err = 0; 2487 u32 aligned_offset, aligned_len, *p; 2488 struct adapter *adapter = pi->adapter; 2489 2490 2491 aligned_offset = offset & ~3; 2492 aligned_len = (len + (offset & 3) + 3) & ~3; 2493 2494 if (aligned_offset != offset || aligned_len != len) { 2495 buf = malloc(aligned_len, M_DEVBUF, M_WAITOK|M_ZERO); 2496 if (!buf) 2497 return (ENOMEM); 2498 err = t3_seeprom_read(adapter, aligned_offset, (u32 *)buf); 2499 if (!err && aligned_len > 4) 2500 err = t3_seeprom_read(adapter, 2501 aligned_offset + aligned_len - 4, 2502 (u32 *)&buf[aligned_len - 4]); 2503 if (err) 2504 goto out; 2505 memcpy(buf + (offset & 3), data, len); 2506 } else 2507 buf = (uint8_t *)(uintptr_t)data; 2508 2509 err = t3_seeprom_wp(adapter, 0); 2510 if (err) 2511 goto out; 2512 2513 for (p = (u32 *)buf; !err && aligned_len; aligned_len -= 4, p++) { 2514 err = t3_seeprom_write(adapter, aligned_offset, *p); 2515 aligned_offset += 4; 2516 } 2517 2518 if (!err) 2519 err = t3_seeprom_wp(adapter, 1); 2520 out: 2521 if (buf != data) 2522 free(buf, M_DEVBUF); 2523 return err; 2524 } 2525 2526 2527 static int 2528 in_range(int val, int lo, int hi) 2529 { 2530 return val < 0 || (val <= hi && val >= lo); 2531 } 2532 2533 static int 2534 cxgb_extension_open(struct cdev *dev, int flags, int fmp, struct thread *td) 2535 { 2536 return (0); 2537 } 2538 2539 static int 2540 cxgb_extension_close(struct cdev *dev, int flags, int fmt, struct thread *td) 2541 { 2542 return (0); 2543 } 2544 2545 static int 2546 cxgb_extension_ioctl(struct cdev *dev, unsigned long cmd, caddr_t data, 2547 int fflag, struct thread *td) 2548 { 2549 int mmd, error = 0; 2550 struct port_info *pi = dev->si_drv1; 2551 adapter_t *sc = pi->adapter; 2552 2553 #ifdef PRIV_SUPPORTED 2554 if (priv_check(td, PRIV_DRIVER)) { 2555 if (cxgb_debug) 2556 printf("user does not have access to privileged ioctls\n"); 2557 return (EPERM); 2558 } 2559 #else 2560 if (suser(td)) { 2561 if (cxgb_debug) 2562 printf("user does not have access to privileged ioctls\n"); 2563 return (EPERM); 2564 } 2565 #endif 2566 2567 switch (cmd) { 2568 case CHELSIO_GET_MIIREG: { 2569 uint32_t val; 2570 struct cphy *phy = &pi->phy; 2571 struct ch_mii_data *mid = (struct ch_mii_data *)data; 2572 2573 if (!phy->mdio_read) 2574 return (EOPNOTSUPP); 2575 if (is_10G(sc)) { 2576 mmd = mid->phy_id >> 8; 2577 if (!mmd) 2578 mmd = MDIO_DEV_PCS; 2579 else if (mmd > MDIO_DEV_VEND2) 2580 return (EINVAL); 2581 2582 error = phy->mdio_read(sc, mid->phy_id & 0x1f, mmd, 2583 mid->reg_num, &val); 2584 } else 2585 error = phy->mdio_read(sc, mid->phy_id & 0x1f, 0, 2586 mid->reg_num & 0x1f, &val); 2587 if (error == 0) 2588 mid->val_out = val; 2589 break; 2590 } 2591 case CHELSIO_SET_MIIREG: { 2592 struct cphy *phy = &pi->phy; 2593 struct ch_mii_data *mid = (struct ch_mii_data *)data; 2594 2595 if (!phy->mdio_write) 2596 return (EOPNOTSUPP); 2597 if (is_10G(sc)) { 2598 mmd = mid->phy_id >> 8; 2599 if (!mmd) 2600 mmd = MDIO_DEV_PCS; 2601 else if (mmd > MDIO_DEV_VEND2) 2602 return (EINVAL); 2603 2604 error = phy->mdio_write(sc, mid->phy_id & 0x1f, 2605 mmd, mid->reg_num, mid->val_in); 2606 } else 2607 error = phy->mdio_write(sc, mid->phy_id & 0x1f, 0, 2608 mid->reg_num & 0x1f, 2609 mid->val_in); 2610 break; 2611 } 2612 case CHELSIO_SETREG: { 2613 struct ch_reg *edata = (struct ch_reg *)data; 2614 if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len) 2615 return (EFAULT); 2616 t3_write_reg(sc, edata->addr, edata->val); 2617 break; 2618 } 2619 case CHELSIO_GETREG: { 2620 struct ch_reg *edata = (struct ch_reg *)data; 2621 if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len) 2622 return (EFAULT); 2623 edata->val = t3_read_reg(sc, edata->addr); 2624 break; 2625 } 2626 case CHELSIO_GET_SGE_CONTEXT: { 2627 struct ch_cntxt *ecntxt = (struct ch_cntxt *)data; 2628 mtx_lock_spin(&sc->sge.reg_lock); 2629 switch (ecntxt->cntxt_type) { 2630 case CNTXT_TYPE_EGRESS: 2631 error = -t3_sge_read_ecntxt(sc, ecntxt->cntxt_id, 2632 ecntxt->data); 2633 break; 2634 case CNTXT_TYPE_FL: 2635 error = -t3_sge_read_fl(sc, ecntxt->cntxt_id, 2636 ecntxt->data); 2637 break; 2638 case CNTXT_TYPE_RSP: 2639 error = -t3_sge_read_rspq(sc, ecntxt->cntxt_id, 2640 ecntxt->data); 2641 break; 2642 case CNTXT_TYPE_CQ: 2643 error = -t3_sge_read_cq(sc, ecntxt->cntxt_id, 2644 ecntxt->data); 2645 break; 2646 default: 2647 error = EINVAL; 2648 break; 2649 } 2650 mtx_unlock_spin(&sc->sge.reg_lock); 2651 break; 2652 } 2653 case CHELSIO_GET_SGE_DESC: { 2654 struct ch_desc *edesc = (struct ch_desc *)data; 2655 int ret; 2656 if (edesc->queue_num >= SGE_QSETS * 6) 2657 return (EINVAL); 2658 ret = t3_get_desc(&sc->sge.qs[edesc->queue_num / 6], 2659 edesc->queue_num % 6, edesc->idx, edesc->data); 2660 if (ret < 0) 2661 return (EINVAL); 2662 edesc->size = ret; 2663 break; 2664 } 2665 case CHELSIO_GET_QSET_PARAMS: { 2666 struct qset_params *q; 2667 struct ch_qset_params *t = (struct ch_qset_params *)data; 2668 int q1 = pi->first_qset; 2669 int nqsets = pi->nqsets; 2670 int i; 2671 2672 if (t->qset_idx >= nqsets) 2673 return EINVAL; 2674 2675 i = q1 + t->qset_idx; 2676 q = &sc->params.sge.qset[i]; 2677 t->rspq_size = q->rspq_size; 2678 t->txq_size[0] = q->txq_size[0]; 2679 t->txq_size[1] = q->txq_size[1]; 2680 t->txq_size[2] = q->txq_size[2]; 2681 t->fl_size[0] = q->fl_size; 2682 t->fl_size[1] = q->jumbo_size; 2683 t->polling = q->polling; 2684 t->lro = q->lro; 2685 t->intr_lat = q->coalesce_usecs; 2686 t->cong_thres = q->cong_thres; 2687 t->qnum = i; 2688 2689 if ((sc->flags & FULL_INIT_DONE) == 0) 2690 t->vector = 0; 2691 else if (sc->flags & USING_MSIX) 2692 t->vector = rman_get_start(sc->msix_irq_res[i]); 2693 else 2694 t->vector = rman_get_start(sc->irq_res); 2695 2696 break; 2697 } 2698 case CHELSIO_GET_QSET_NUM: { 2699 struct ch_reg *edata = (struct ch_reg *)data; 2700 edata->val = pi->nqsets; 2701 break; 2702 } 2703 case CHELSIO_LOAD_FW: { 2704 uint8_t *fw_data; 2705 uint32_t vers; 2706 struct ch_mem_range *t = (struct ch_mem_range *)data; 2707 2708 /* 2709 * You're allowed to load a firmware only before FULL_INIT_DONE 2710 * 2711 * FW_UPTODATE is also set so the rest of the initialization 2712 * will not overwrite what was loaded here. This gives you the 2713 * flexibility to load any firmware (and maybe shoot yourself in 2714 * the foot). 2715 */ 2716 2717 ADAPTER_LOCK(sc); 2718 if (sc->open_device_map || sc->flags & FULL_INIT_DONE) { 2719 ADAPTER_UNLOCK(sc); 2720 return (EBUSY); 2721 } 2722 2723 fw_data = malloc(t->len, M_DEVBUF, M_NOWAIT); 2724 if (!fw_data) 2725 error = ENOMEM; 2726 else 2727 error = copyin(t->buf, fw_data, t->len); 2728 2729 if (!error) 2730 error = -t3_load_fw(sc, fw_data, t->len); 2731 2732 if (t3_get_fw_version(sc, &vers) == 0) { 2733 snprintf(&sc->fw_version[0], sizeof(sc->fw_version), 2734 "%d.%d.%d", G_FW_VERSION_MAJOR(vers), 2735 G_FW_VERSION_MINOR(vers), G_FW_VERSION_MICRO(vers)); 2736 } 2737 2738 if (!error) 2739 sc->flags |= FW_UPTODATE; 2740 2741 free(fw_data, M_DEVBUF); 2742 ADAPTER_UNLOCK(sc); 2743 break; 2744 } 2745 case CHELSIO_LOAD_BOOT: { 2746 uint8_t *boot_data; 2747 struct ch_mem_range *t = (struct ch_mem_range *)data; 2748 2749 boot_data = malloc(t->len, M_DEVBUF, M_NOWAIT); 2750 if (!boot_data) 2751 return ENOMEM; 2752 2753 error = copyin(t->buf, boot_data, t->len); 2754 if (!error) 2755 error = -t3_load_boot(sc, boot_data, t->len); 2756 2757 free(boot_data, M_DEVBUF); 2758 break; 2759 } 2760 case CHELSIO_GET_PM: { 2761 struct ch_pm *m = (struct ch_pm *)data; 2762 struct tp_params *p = &sc->params.tp; 2763 2764 if (!is_offload(sc)) 2765 return (EOPNOTSUPP); 2766 2767 m->tx_pg_sz = p->tx_pg_size; 2768 m->tx_num_pg = p->tx_num_pgs; 2769 m->rx_pg_sz = p->rx_pg_size; 2770 m->rx_num_pg = p->rx_num_pgs; 2771 m->pm_total = p->pmtx_size + p->chan_rx_size * p->nchan; 2772 2773 break; 2774 } 2775 case CHELSIO_SET_PM: { 2776 struct ch_pm *m = (struct ch_pm *)data; 2777 struct tp_params *p = &sc->params.tp; 2778 2779 if (!is_offload(sc)) 2780 return (EOPNOTSUPP); 2781 if (sc->flags & FULL_INIT_DONE) 2782 return (EBUSY); 2783 2784 if (!m->rx_pg_sz || (m->rx_pg_sz & (m->rx_pg_sz - 1)) || 2785 !m->tx_pg_sz || (m->tx_pg_sz & (m->tx_pg_sz - 1))) 2786 return (EINVAL); /* not power of 2 */ 2787 if (!(m->rx_pg_sz & 0x14000)) 2788 return (EINVAL); /* not 16KB or 64KB */ 2789 if (!(m->tx_pg_sz & 0x1554000)) 2790 return (EINVAL); 2791 if (m->tx_num_pg == -1) 2792 m->tx_num_pg = p->tx_num_pgs; 2793 if (m->rx_num_pg == -1) 2794 m->rx_num_pg = p->rx_num_pgs; 2795 if (m->tx_num_pg % 24 || m->rx_num_pg % 24) 2796 return (EINVAL); 2797 if (m->rx_num_pg * m->rx_pg_sz > p->chan_rx_size || 2798 m->tx_num_pg * m->tx_pg_sz > p->chan_tx_size) 2799 return (EINVAL); 2800 2801 p->rx_pg_size = m->rx_pg_sz; 2802 p->tx_pg_size = m->tx_pg_sz; 2803 p->rx_num_pgs = m->rx_num_pg; 2804 p->tx_num_pgs = m->tx_num_pg; 2805 break; 2806 } 2807 case CHELSIO_SETMTUTAB: { 2808 struct ch_mtus *m = (struct ch_mtus *)data; 2809 int i; 2810 2811 if (!is_offload(sc)) 2812 return (EOPNOTSUPP); 2813 if (offload_running(sc)) 2814 return (EBUSY); 2815 if (m->nmtus != NMTUS) 2816 return (EINVAL); 2817 if (m->mtus[0] < 81) /* accommodate SACK */ 2818 return (EINVAL); 2819 2820 /* 2821 * MTUs must be in ascending order 2822 */ 2823 for (i = 1; i < NMTUS; ++i) 2824 if (m->mtus[i] < m->mtus[i - 1]) 2825 return (EINVAL); 2826 2827 memcpy(sc->params.mtus, m->mtus, sizeof(sc->params.mtus)); 2828 break; 2829 } 2830 case CHELSIO_GETMTUTAB: { 2831 struct ch_mtus *m = (struct ch_mtus *)data; 2832 2833 if (!is_offload(sc)) 2834 return (EOPNOTSUPP); 2835 2836 memcpy(m->mtus, sc->params.mtus, sizeof(m->mtus)); 2837 m->nmtus = NMTUS; 2838 break; 2839 } 2840 case CHELSIO_GET_MEM: { 2841 struct ch_mem_range *t = (struct ch_mem_range *)data; 2842 struct mc7 *mem; 2843 uint8_t *useraddr; 2844 u64 buf[32]; 2845 2846 /* 2847 * Use these to avoid modifying len/addr in the return 2848 * struct 2849 */ 2850 uint32_t len = t->len, addr = t->addr; 2851 2852 if (!is_offload(sc)) 2853 return (EOPNOTSUPP); 2854 if (!(sc->flags & FULL_INIT_DONE)) 2855 return (EIO); /* need the memory controllers */ 2856 if ((addr & 0x7) || (len & 0x7)) 2857 return (EINVAL); 2858 if (t->mem_id == MEM_CM) 2859 mem = &sc->cm; 2860 else if (t->mem_id == MEM_PMRX) 2861 mem = &sc->pmrx; 2862 else if (t->mem_id == MEM_PMTX) 2863 mem = &sc->pmtx; 2864 else 2865 return (EINVAL); 2866 2867 /* 2868 * Version scheme: 2869 * bits 0..9: chip version 2870 * bits 10..15: chip revision 2871 */ 2872 t->version = 3 | (sc->params.rev << 10); 2873 2874 /* 2875 * Read 256 bytes at a time as len can be large and we don't 2876 * want to use huge intermediate buffers. 2877 */ 2878 useraddr = (uint8_t *)t->buf; 2879 while (len) { 2880 unsigned int chunk = min(len, sizeof(buf)); 2881 2882 error = t3_mc7_bd_read(mem, addr / 8, chunk / 8, buf); 2883 if (error) 2884 return (-error); 2885 if (copyout(buf, useraddr, chunk)) 2886 return (EFAULT); 2887 useraddr += chunk; 2888 addr += chunk; 2889 len -= chunk; 2890 } 2891 break; 2892 } 2893 case CHELSIO_READ_TCAM_WORD: { 2894 struct ch_tcam_word *t = (struct ch_tcam_word *)data; 2895 2896 if (!is_offload(sc)) 2897 return (EOPNOTSUPP); 2898 if (!(sc->flags & FULL_INIT_DONE)) 2899 return (EIO); /* need MC5 */ 2900 return -t3_read_mc5_range(&sc->mc5, t->addr, 1, t->buf); 2901 break; 2902 } 2903 case CHELSIO_SET_TRACE_FILTER: { 2904 struct ch_trace *t = (struct ch_trace *)data; 2905 const struct trace_params *tp; 2906 2907 tp = (const struct trace_params *)&t->sip; 2908 if (t->config_tx) 2909 t3_config_trace_filter(sc, tp, 0, t->invert_match, 2910 t->trace_tx); 2911 if (t->config_rx) 2912 t3_config_trace_filter(sc, tp, 1, t->invert_match, 2913 t->trace_rx); 2914 break; 2915 } 2916 case CHELSIO_SET_PKTSCHED: { 2917 struct ch_pktsched_params *p = (struct ch_pktsched_params *)data; 2918 if (sc->open_device_map == 0) 2919 return (EAGAIN); 2920 send_pktsched_cmd(sc, p->sched, p->idx, p->min, p->max, 2921 p->binding); 2922 break; 2923 } 2924 case CHELSIO_IFCONF_GETREGS: { 2925 struct ch_ifconf_regs *regs = (struct ch_ifconf_regs *)data; 2926 int reglen = cxgb_get_regs_len(); 2927 uint8_t *buf = malloc(reglen, M_DEVBUF, M_NOWAIT); 2928 if (buf == NULL) { 2929 return (ENOMEM); 2930 } 2931 if (regs->len > reglen) 2932 regs->len = reglen; 2933 else if (regs->len < reglen) 2934 error = ENOBUFS; 2935 2936 if (!error) { 2937 cxgb_get_regs(sc, regs, buf); 2938 error = copyout(buf, regs->data, reglen); 2939 } 2940 free(buf, M_DEVBUF); 2941 2942 break; 2943 } 2944 case CHELSIO_SET_HW_SCHED: { 2945 struct ch_hw_sched *t = (struct ch_hw_sched *)data; 2946 unsigned int ticks_per_usec = core_ticks_per_usec(sc); 2947 2948 if ((sc->flags & FULL_INIT_DONE) == 0) 2949 return (EAGAIN); /* need TP to be initialized */ 2950 if (t->sched >= NTX_SCHED || !in_range(t->mode, 0, 1) || 2951 !in_range(t->channel, 0, 1) || 2952 !in_range(t->kbps, 0, 10000000) || 2953 !in_range(t->class_ipg, 0, 10000 * 65535 / ticks_per_usec) || 2954 !in_range(t->flow_ipg, 0, 2955 dack_ticks_to_usec(sc, 0x7ff))) 2956 return (EINVAL); 2957 2958 if (t->kbps >= 0) { 2959 error = t3_config_sched(sc, t->kbps, t->sched); 2960 if (error < 0) 2961 return (-error); 2962 } 2963 if (t->class_ipg >= 0) 2964 t3_set_sched_ipg(sc, t->sched, t->class_ipg); 2965 if (t->flow_ipg >= 0) { 2966 t->flow_ipg *= 1000; /* us -> ns */ 2967 t3_set_pace_tbl(sc, &t->flow_ipg, t->sched, 1); 2968 } 2969 if (t->mode >= 0) { 2970 int bit = 1 << (S_TX_MOD_TIMER_MODE + t->sched); 2971 2972 t3_set_reg_field(sc, A_TP_TX_MOD_QUEUE_REQ_MAP, 2973 bit, t->mode ? bit : 0); 2974 } 2975 if (t->channel >= 0) 2976 t3_set_reg_field(sc, A_TP_TX_MOD_QUEUE_REQ_MAP, 2977 1 << t->sched, t->channel << t->sched); 2978 break; 2979 } 2980 case CHELSIO_GET_EEPROM: { 2981 int i; 2982 struct ch_eeprom *e = (struct ch_eeprom *)data; 2983 uint8_t *buf = malloc(EEPROMSIZE, M_DEVBUF, M_NOWAIT); 2984 2985 if (buf == NULL) { 2986 return (ENOMEM); 2987 } 2988 e->magic = EEPROM_MAGIC; 2989 for (i = e->offset & ~3; !error && i < e->offset + e->len; i += 4) 2990 error = -t3_seeprom_read(sc, i, (uint32_t *)&buf[i]); 2991 2992 if (!error) 2993 error = copyout(buf + e->offset, e->data, e->len); 2994 2995 free(buf, M_DEVBUF); 2996 break; 2997 } 2998 case CHELSIO_CLEAR_STATS: { 2999 if (!(sc->flags & FULL_INIT_DONE)) 3000 return EAGAIN; 3001 3002 PORT_LOCK(pi); 3003 t3_mac_update_stats(&pi->mac); 3004 memset(&pi->mac.stats, 0, sizeof(pi->mac.stats)); 3005 PORT_UNLOCK(pi); 3006 break; 3007 } 3008 case CHELSIO_GET_UP_LA: { 3009 struct ch_up_la *la = (struct ch_up_la *)data; 3010 uint8_t *buf = malloc(LA_BUFSIZE, M_DEVBUF, M_NOWAIT); 3011 if (buf == NULL) { 3012 return (ENOMEM); 3013 } 3014 if (la->bufsize < LA_BUFSIZE) 3015 error = ENOBUFS; 3016 3017 if (!error) 3018 error = -t3_get_up_la(sc, &la->stopped, &la->idx, 3019 &la->bufsize, buf); 3020 if (!error) 3021 error = copyout(buf, la->data, la->bufsize); 3022 3023 free(buf, M_DEVBUF); 3024 break; 3025 } 3026 case CHELSIO_GET_UP_IOQS: { 3027 struct ch_up_ioqs *ioqs = (struct ch_up_ioqs *)data; 3028 uint8_t *buf = malloc(IOQS_BUFSIZE, M_DEVBUF, M_NOWAIT); 3029 uint32_t *v; 3030 3031 if (buf == NULL) { 3032 return (ENOMEM); 3033 } 3034 if (ioqs->bufsize < IOQS_BUFSIZE) 3035 error = ENOBUFS; 3036 3037 if (!error) 3038 error = -t3_get_up_ioqs(sc, &ioqs->bufsize, buf); 3039 3040 if (!error) { 3041 v = (uint32_t *)buf; 3042 3043 ioqs->ioq_rx_enable = *v++; 3044 ioqs->ioq_tx_enable = *v++; 3045 ioqs->ioq_rx_status = *v++; 3046 ioqs->ioq_tx_status = *v++; 3047 3048 error = copyout(v, ioqs->data, ioqs->bufsize); 3049 } 3050 3051 free(buf, M_DEVBUF); 3052 break; 3053 } 3054 case CHELSIO_SET_FILTER: { 3055 struct ch_filter *f = (struct ch_filter *)data;; 3056 struct filter_info *p; 3057 unsigned int nfilters = sc->params.mc5.nfilters; 3058 3059 if (!is_offload(sc)) 3060 return (EOPNOTSUPP); /* No TCAM */ 3061 if (!(sc->flags & FULL_INIT_DONE)) 3062 return (EAGAIN); /* mc5 not setup yet */ 3063 if (nfilters == 0) 3064 return (EBUSY); /* TOE will use TCAM */ 3065 3066 /* sanity checks */ 3067 if (f->filter_id >= nfilters || 3068 (f->val.dip && f->mask.dip != 0xffffffff) || 3069 (f->val.sport && f->mask.sport != 0xffff) || 3070 (f->val.dport && f->mask.dport != 0xffff) || 3071 (f->val.vlan && f->mask.vlan != 0xfff) || 3072 (f->val.vlan_prio && 3073 f->mask.vlan_prio != FILTER_NO_VLAN_PRI) || 3074 (f->mac_addr_idx != 0xffff && f->mac_addr_idx > 15) || 3075 f->qset >= SGE_QSETS || 3076 sc->rrss_map[f->qset] >= RSS_TABLE_SIZE) 3077 return (EINVAL); 3078 3079 /* Was allocated with M_WAITOK */ 3080 KASSERT(sc->filters, ("filter table NULL\n")); 3081 3082 p = &sc->filters[f->filter_id]; 3083 if (p->locked) 3084 return (EPERM); 3085 3086 bzero(p, sizeof(*p)); 3087 p->sip = f->val.sip; 3088 p->sip_mask = f->mask.sip; 3089 p->dip = f->val.dip; 3090 p->sport = f->val.sport; 3091 p->dport = f->val.dport; 3092 p->vlan = f->mask.vlan ? f->val.vlan : 0xfff; 3093 p->vlan_prio = f->mask.vlan_prio ? (f->val.vlan_prio & 6) : 3094 FILTER_NO_VLAN_PRI; 3095 p->mac_hit = f->mac_hit; 3096 p->mac_vld = f->mac_addr_idx != 0xffff; 3097 p->mac_idx = f->mac_addr_idx; 3098 p->pkt_type = f->proto; 3099 p->report_filter_id = f->want_filter_id; 3100 p->pass = f->pass; 3101 p->rss = f->rss; 3102 p->qset = f->qset; 3103 3104 error = set_filter(sc, f->filter_id, p); 3105 if (error == 0) 3106 p->valid = 1; 3107 break; 3108 } 3109 case CHELSIO_DEL_FILTER: { 3110 struct ch_filter *f = (struct ch_filter *)data; 3111 struct filter_info *p; 3112 unsigned int nfilters = sc->params.mc5.nfilters; 3113 3114 if (!is_offload(sc)) 3115 return (EOPNOTSUPP); 3116 if (!(sc->flags & FULL_INIT_DONE)) 3117 return (EAGAIN); 3118 if (nfilters == 0 || sc->filters == NULL) 3119 return (EINVAL); 3120 if (f->filter_id >= nfilters) 3121 return (EINVAL); 3122 3123 p = &sc->filters[f->filter_id]; 3124 if (p->locked) 3125 return (EPERM); 3126 if (!p->valid) 3127 return (EFAULT); /* Read "Bad address" as "Bad index" */ 3128 3129 bzero(p, sizeof(*p)); 3130 p->sip = p->sip_mask = 0xffffffff; 3131 p->vlan = 0xfff; 3132 p->vlan_prio = FILTER_NO_VLAN_PRI; 3133 p->pkt_type = 1; 3134 error = set_filter(sc, f->filter_id, p); 3135 break; 3136 } 3137 case CHELSIO_GET_FILTER: { 3138 struct ch_filter *f = (struct ch_filter *)data; 3139 struct filter_info *p; 3140 unsigned int i, nfilters = sc->params.mc5.nfilters; 3141 3142 if (!is_offload(sc)) 3143 return (EOPNOTSUPP); 3144 if (!(sc->flags & FULL_INIT_DONE)) 3145 return (EAGAIN); 3146 if (nfilters == 0 || sc->filters == NULL) 3147 return (EINVAL); 3148 3149 i = f->filter_id == 0xffffffff ? 0 : f->filter_id + 1; 3150 for (; i < nfilters; i++) { 3151 p = &sc->filters[i]; 3152 if (!p->valid) 3153 continue; 3154 3155 bzero(f, sizeof(*f)); 3156 3157 f->filter_id = i; 3158 f->val.sip = p->sip; 3159 f->mask.sip = p->sip_mask; 3160 f->val.dip = p->dip; 3161 f->mask.dip = p->dip ? 0xffffffff : 0; 3162 f->val.sport = p->sport; 3163 f->mask.sport = p->sport ? 0xffff : 0; 3164 f->val.dport = p->dport; 3165 f->mask.dport = p->dport ? 0xffff : 0; 3166 f->val.vlan = p->vlan == 0xfff ? 0 : p->vlan; 3167 f->mask.vlan = p->vlan == 0xfff ? 0 : 0xfff; 3168 f->val.vlan_prio = p->vlan_prio == FILTER_NO_VLAN_PRI ? 3169 0 : p->vlan_prio; 3170 f->mask.vlan_prio = p->vlan_prio == FILTER_NO_VLAN_PRI ? 3171 0 : FILTER_NO_VLAN_PRI; 3172 f->mac_hit = p->mac_hit; 3173 f->mac_addr_idx = p->mac_vld ? p->mac_idx : 0xffff; 3174 f->proto = p->pkt_type; 3175 f->want_filter_id = p->report_filter_id; 3176 f->pass = p->pass; 3177 f->rss = p->rss; 3178 f->qset = p->qset; 3179 3180 break; 3181 } 3182 3183 if (i == nfilters) 3184 f->filter_id = 0xffffffff; 3185 break; 3186 } 3187 default: 3188 return (EOPNOTSUPP); 3189 break; 3190 } 3191 3192 return (error); 3193 } 3194 3195 static __inline void 3196 reg_block_dump(struct adapter *ap, uint8_t *buf, unsigned int start, 3197 unsigned int end) 3198 { 3199 uint32_t *p = (uint32_t *)(buf + start); 3200 3201 for ( ; start <= end; start += sizeof(uint32_t)) 3202 *p++ = t3_read_reg(ap, start); 3203 } 3204 3205 #define T3_REGMAP_SIZE (3 * 1024) 3206 static int 3207 cxgb_get_regs_len(void) 3208 { 3209 return T3_REGMAP_SIZE; 3210 } 3211 3212 static void 3213 cxgb_get_regs(adapter_t *sc, struct ch_ifconf_regs *regs, uint8_t *buf) 3214 { 3215 3216 /* 3217 * Version scheme: 3218 * bits 0..9: chip version 3219 * bits 10..15: chip revision 3220 * bit 31: set for PCIe cards 3221 */ 3222 regs->version = 3 | (sc->params.rev << 10) | (is_pcie(sc) << 31); 3223 3224 /* 3225 * We skip the MAC statistics registers because they are clear-on-read. 3226 * Also reading multi-register stats would need to synchronize with the 3227 * periodic mac stats accumulation. Hard to justify the complexity. 3228 */ 3229 memset(buf, 0, cxgb_get_regs_len()); 3230 reg_block_dump(sc, buf, 0, A_SG_RSPQ_CREDIT_RETURN); 3231 reg_block_dump(sc, buf, A_SG_HI_DRB_HI_THRSH, A_ULPRX_PBL_ULIMIT); 3232 reg_block_dump(sc, buf, A_ULPTX_CONFIG, A_MPS_INT_CAUSE); 3233 reg_block_dump(sc, buf, A_CPL_SWITCH_CNTRL, A_CPL_MAP_TBL_DATA); 3234 reg_block_dump(sc, buf, A_SMB_GLOBAL_TIME_CFG, A_XGM_SERDES_STAT3); 3235 reg_block_dump(sc, buf, A_XGM_SERDES_STATUS0, 3236 XGM_REG(A_XGM_SERDES_STAT3, 1)); 3237 reg_block_dump(sc, buf, XGM_REG(A_XGM_SERDES_STATUS0, 1), 3238 XGM_REG(A_XGM_RX_SPI4_SOP_EOP_CNT, 1)); 3239 } 3240 3241 static int 3242 alloc_filters(struct adapter *sc) 3243 { 3244 struct filter_info *p; 3245 unsigned int nfilters = sc->params.mc5.nfilters; 3246 3247 if (nfilters == 0) 3248 return (0); 3249 3250 p = malloc(sizeof(*p) * nfilters, M_DEVBUF, M_WAITOK | M_ZERO); 3251 sc->filters = p; 3252 3253 p = &sc->filters[nfilters - 1]; 3254 p->vlan = 0xfff; 3255 p->vlan_prio = FILTER_NO_VLAN_PRI; 3256 p->pass = p->rss = p->valid = p->locked = 1; 3257 3258 return (0); 3259 } 3260 3261 static int 3262 setup_hw_filters(struct adapter *sc) 3263 { 3264 int i, rc; 3265 unsigned int nfilters = sc->params.mc5.nfilters; 3266 3267 if (!sc->filters) 3268 return (0); 3269 3270 t3_enable_filters(sc); 3271 3272 for (i = rc = 0; i < nfilters && !rc; i++) { 3273 if (sc->filters[i].locked) 3274 rc = set_filter(sc, i, &sc->filters[i]); 3275 } 3276 3277 return (rc); 3278 } 3279 3280 static int 3281 set_filter(struct adapter *sc, int id, const struct filter_info *f) 3282 { 3283 int len; 3284 struct mbuf *m; 3285 struct ulp_txpkt *txpkt; 3286 struct work_request_hdr *wr; 3287 struct cpl_pass_open_req *oreq; 3288 struct cpl_set_tcb_field *sreq; 3289 3290 len = sizeof(*wr) + sizeof(*oreq) + 2 * sizeof(*sreq); 3291 KASSERT(len <= MHLEN, ("filter request too big for an mbuf")); 3292 3293 id += t3_mc5_size(&sc->mc5) - sc->params.mc5.nroutes - 3294 sc->params.mc5.nfilters; 3295 3296 m = m_gethdr(M_WAITOK, MT_DATA); 3297 m->m_len = m->m_pkthdr.len = len; 3298 bzero(mtod(m, char *), len); 3299 3300 wr = mtod(m, struct work_request_hdr *); 3301 wr->wrh_hi = htonl(V_WR_OP(FW_WROPCODE_BYPASS) | F_WR_ATOMIC); 3302 3303 oreq = (struct cpl_pass_open_req *)(wr + 1); 3304 txpkt = (struct ulp_txpkt *)oreq; 3305 txpkt->cmd_dest = htonl(V_ULPTX_CMD(ULP_TXPKT)); 3306 txpkt->len = htonl(V_ULPTX_NFLITS(sizeof(*oreq) / 8)); 3307 OPCODE_TID(oreq) = htonl(MK_OPCODE_TID(CPL_PASS_OPEN_REQ, id)); 3308 oreq->local_port = htons(f->dport); 3309 oreq->peer_port = htons(f->sport); 3310 oreq->local_ip = htonl(f->dip); 3311 oreq->peer_ip = htonl(f->sip); 3312 oreq->peer_netmask = htonl(f->sip_mask); 3313 oreq->opt0h = 0; 3314 oreq->opt0l = htonl(F_NO_OFFLOAD); 3315 oreq->opt1 = htonl(V_MAC_MATCH_VALID(f->mac_vld) | 3316 V_CONN_POLICY(CPL_CONN_POLICY_FILTER) | 3317 V_VLAN_PRI(f->vlan_prio >> 1) | 3318 V_VLAN_PRI_VALID(f->vlan_prio != FILTER_NO_VLAN_PRI) | 3319 V_PKT_TYPE(f->pkt_type) | V_OPT1_VLAN(f->vlan) | 3320 V_MAC_MATCH(f->mac_idx | (f->mac_hit << 4))); 3321 3322 sreq = (struct cpl_set_tcb_field *)(oreq + 1); 3323 set_tcb_field_ulp(sreq, id, 1, 0x1800808000ULL, 3324 (f->report_filter_id << 15) | (1 << 23) | 3325 ((u64)f->pass << 35) | ((u64)!f->rss << 36)); 3326 set_tcb_field_ulp(sreq + 1, id, 0, 0xffffffff, (2 << 19) | 1); 3327 t3_mgmt_tx(sc, m); 3328 3329 if (f->pass && !f->rss) { 3330 len = sizeof(*sreq); 3331 m = m_gethdr(M_WAITOK, MT_DATA); 3332 m->m_len = m->m_pkthdr.len = len; 3333 bzero(mtod(m, char *), len); 3334 sreq = mtod(m, struct cpl_set_tcb_field *); 3335 sreq->wr.wrh_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD)); 3336 mk_set_tcb_field(sreq, id, 25, 0x3f80000, 3337 (u64)sc->rrss_map[f->qset] << 19); 3338 t3_mgmt_tx(sc, m); 3339 } 3340 return 0; 3341 } 3342 3343 static inline void 3344 mk_set_tcb_field(struct cpl_set_tcb_field *req, unsigned int tid, 3345 unsigned int word, u64 mask, u64 val) 3346 { 3347 OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_SET_TCB_FIELD, tid)); 3348 req->reply = V_NO_REPLY(1); 3349 req->cpu_idx = 0; 3350 req->word = htons(word); 3351 req->mask = htobe64(mask); 3352 req->val = htobe64(val); 3353 } 3354 3355 static inline void 3356 set_tcb_field_ulp(struct cpl_set_tcb_field *req, unsigned int tid, 3357 unsigned int word, u64 mask, u64 val) 3358 { 3359 struct ulp_txpkt *txpkt = (struct ulp_txpkt *)req; 3360 3361 txpkt->cmd_dest = htonl(V_ULPTX_CMD(ULP_TXPKT)); 3362 txpkt->len = htonl(V_ULPTX_NFLITS(sizeof(*req) / 8)); 3363 mk_set_tcb_field(req, tid, word, mask, val); 3364 } 3365