| 1 | /* |
| 2 | * Copyright (c) 2015-2023 Apple Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | |
| 29 | /* |
| 30 | * Copyright (C) 2013-2014 Universita` di Pisa. All rights reserved. |
| 31 | * |
| 32 | * Redistribution and use in source and binary forms, with or without |
| 33 | * modification, are permitted provided that the following conditions |
| 34 | * are met: |
| 35 | * 1. Redistributions of source code must retain the above copyright |
| 36 | * notice, this list of conditions and the following disclaimer. |
| 37 | * 2. Redistributions in binary form must reproduce the above copyright |
| 38 | * notice, this list of conditions and the following disclaimer in the |
| 39 | * documentation and/or other materials provided with the distribution. |
| 40 | * |
| 41 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
| 42 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 43 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 44 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
| 45 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 46 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 47 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 48 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 49 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 50 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 51 | * SUCH DAMAGE. |
| 52 | */ |
| 53 | |
| 54 | #include <skywalk/os_skywalk_private.h> |
| 55 | #include <skywalk/nexus/flowswitch/nx_flowswitch.h> |
| 56 | #include <skywalk/nexus/flowswitch/fsw_var.h> |
| 57 | #include <sys/sdt.h> |
| 58 | |
| 59 | static void fsw_vp_na_dtor(struct nexus_adapter *); |
| 60 | static int fsw_vp_na_special(struct nexus_adapter *, |
| 61 | struct kern_channel *, struct chreq *, nxspec_cmd_t); |
| 62 | static struct nexus_vp_adapter *fsw_vp_na_alloc(zalloc_flags_t); |
| 63 | static void fsw_vp_na_free(struct nexus_adapter *); |
| 64 | static int fsw_vp_na_channel_event_notify(struct nexus_adapter *vpna, |
| 65 | struct __kern_channel_event *ev, uint16_t ev_len); |
| 66 | |
| 67 | static SKMEM_TYPE_DEFINE(na_vp_zone, struct nexus_vp_adapter); |
| 68 | |
| 69 | static uint16_t fsw_vpna_gencnt = 0; |
| 70 | |
| 71 | /* na_activate() callback for flow switch ports */ |
| 72 | int |
| 73 | fsw_vp_na_activate(struct nexus_adapter *na, na_activate_mode_t mode) |
| 74 | { |
| 75 | int ret = 0; |
| 76 | struct nexus_vp_adapter *vpna = (struct nexus_vp_adapter *)(void *)na; |
| 77 | struct nx_flowswitch *fsw = vpna->vpna_fsw; |
| 78 | |
| 79 | ASSERT(na->na_type == NA_FLOWSWITCH_VP); |
| 80 | |
| 81 | SK_DF(SK_VERB_FSW, "na \"%s\" (0x%llx) %s" , na->na_name, |
| 82 | SK_KVA(na), na_activate_mode2str(mode)); |
| 83 | |
| 84 | /* |
| 85 | * Persistent ports may be put in Skywalk mode |
| 86 | * before being attached to a FlowSwitch. |
| 87 | */ |
| 88 | FSW_WLOCK(fsw); |
| 89 | |
| 90 | os_atomic_inc(&fsw_vpna_gencnt, relaxed); |
| 91 | vpna->vpna_gencnt = fsw_vpna_gencnt; |
| 92 | |
| 93 | if (mode == NA_ACTIVATE_MODE_ON) { |
| 94 | os_atomic_or(&na->na_flags, NAF_ACTIVE, relaxed); |
| 95 | } |
| 96 | |
| 97 | ret = fsw_port_na_activate(fsw, vpna, mode); |
| 98 | if (ret != 0) { |
| 99 | SK_DF(SK_VERB_FSW, "na \"%s\" (0x%llx) %s err(%d)" , |
| 100 | na->na_name, SK_KVA(na), na_activate_mode2str(mode), ret); |
| 101 | if (mode == NA_ACTIVATE_MODE_ON) { |
| 102 | os_atomic_andnot(&na->na_flags, NAF_ACTIVE, relaxed); |
| 103 | } |
| 104 | goto done; |
| 105 | } |
| 106 | |
| 107 | if (mode == NA_ACTIVATE_MODE_DEFUNCT || |
| 108 | mode == NA_ACTIVATE_MODE_OFF) { |
| 109 | struct skmem_arena_nexus *arn = skmem_arena_nexus(ar: na->na_arena); |
| 110 | |
| 111 | if (mode == NA_ACTIVATE_MODE_OFF) { |
| 112 | os_atomic_andnot(&na->na_flags, NAF_ACTIVE, relaxed); |
| 113 | } |
| 114 | |
| 115 | AR_LOCK(na->na_arena); |
| 116 | if (na->na_type == NA_FLOWSWITCH_VP && |
| 117 | arn->arn_stats_obj != NULL) { |
| 118 | fsw_fold_stats(fsw, |
| 119 | data: arn->arn_stats_obj, type: na->na_stats_type); |
| 120 | } |
| 121 | AR_UNLOCK(na->na_arena); |
| 122 | |
| 123 | enum txrx t; |
| 124 | uint32_t i; |
| 125 | struct __nx_stats_channel_errors stats; |
| 126 | for_all_rings(t) { |
| 127 | for (i = 0; i < na_get_nrings(na, t); i++) { |
| 128 | stats.nxs_cres = |
| 129 | &NAKR(na, t)[i].ckr_err_stats; |
| 130 | fsw_fold_stats(fsw, data: &stats, |
| 131 | type: NEXUS_STATS_TYPE_CHAN_ERRORS); |
| 132 | } |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | done: |
| 137 | FSW_WUNLOCK(fsw); |
| 138 | return ret; |
| 139 | } |
| 140 | |
| 141 | /* na_dtor callback for ephemeral flow switch ports */ |
| 142 | static void |
| 143 | fsw_vp_na_dtor(struct nexus_adapter *na) |
| 144 | { |
| 145 | struct nexus_vp_adapter *vpna = (struct nexus_vp_adapter *)(void *)na; |
| 146 | struct nx_flowswitch *fsw = vpna->vpna_fsw; |
| 147 | |
| 148 | SK_LOCK_ASSERT_HELD(); |
| 149 | ASSERT(na->na_type == NA_FLOWSWITCH_VP); |
| 150 | |
| 151 | SK_DF(SK_VERB_FSW, "na \"%s\" (0x%llx)" , na->na_name, SK_KVA(na)); |
| 152 | |
| 153 | if (fsw != NULL) { |
| 154 | FSW_WLOCK(fsw); |
| 155 | fsw_port_free(fsw, vpna, nx_port: vpna->vpna_nx_port, FALSE); |
| 156 | FSW_WUNLOCK(fsw); |
| 157 | } |
| 158 | } |
| 159 | |
| 160 | /* |
| 161 | * na_krings_create callback for flow switch ports. |
| 162 | * Calls the standard na_kr_create(), then adds leases on rx |
| 163 | * rings and bdgfwd on tx rings. |
| 164 | */ |
| 165 | int |
| 166 | fsw_vp_na_krings_create(struct nexus_adapter *na, struct kern_channel *ch) |
| 167 | { |
| 168 | ASSERT(na->na_type == NA_FLOWSWITCH_VP); |
| 169 | |
| 170 | return na_rings_mem_setup(na, FALSE, ch); |
| 171 | } |
| 172 | |
| 173 | |
| 174 | /* na_krings_delete callback for flow switch ports. */ |
| 175 | void |
| 176 | fsw_vp_na_krings_delete(struct nexus_adapter *na, struct kern_channel *ch, |
| 177 | boolean_t defunct) |
| 178 | { |
| 179 | ASSERT(na->na_type == NA_FLOWSWITCH_VP); |
| 180 | |
| 181 | na_rings_mem_teardown(na, ch, defunct); |
| 182 | } |
| 183 | |
| 184 | /* na_txsync callback for flow switch ports */ |
| 185 | int |
| 186 | fsw_vp_na_txsync(struct __kern_channel_ring *kring, struct proc *p, |
| 187 | uint32_t flags) |
| 188 | { |
| 189 | #pragma unused(flags) |
| 190 | struct nexus_vp_adapter *vpna = VPNA(KRNA(kring)); |
| 191 | struct nx_flowswitch *fsw = vpna->vpna_fsw; |
| 192 | int error = 0; |
| 193 | |
| 194 | /* |
| 195 | * Flush packets if and only if the ring isn't in drop mode, |
| 196 | * and if the adapter is currently attached to a nexus port; |
| 197 | * otherwise we drop them. |
| 198 | */ |
| 199 | if (__probable(!KR_DROP(kring) && fsw != NULL)) { |
| 200 | fsw_ring_flush(fsw, skring: kring, p); |
| 201 | } else { |
| 202 | int dropped_pkts; |
| 203 | /* packets between khead to rhead have been dropped */ |
| 204 | dropped_pkts = kring->ckr_rhead - kring->ckr_khead; |
| 205 | if (dropped_pkts < 0) { |
| 206 | dropped_pkts += kring->ckr_num_slots; |
| 207 | } |
| 208 | if (fsw != NULL) { |
| 209 | STATS_INC(&fsw->fsw_stats, FSW_STATS_DST_RING_DROPMODE); |
| 210 | STATS_ADD(&fsw->fsw_stats, FSW_STATS_DROP, |
| 211 | dropped_pkts); |
| 212 | } |
| 213 | /* we're dropping; claim all */ |
| 214 | slot_idx_t sidx = kring->ckr_khead; |
| 215 | while (sidx != kring->ckr_rhead) { |
| 216 | struct __kern_slot_desc *ksd = KR_KSD(kring, sidx); |
| 217 | if (KSD_VALID_METADATA(ksd)) { |
| 218 | struct __kern_packet *pkt = ksd->sd_pkt; |
| 219 | (void) KR_SLOT_DETACH_METADATA(kring, ksd); |
| 220 | pp_free_packet_single(pkt); |
| 221 | } |
| 222 | sidx = SLOT_NEXT(i: sidx, lim: kring->ckr_lim); |
| 223 | } |
| 224 | kring->ckr_khead = kring->ckr_rhead; |
| 225 | kring->ckr_ktail = SLOT_PREV(i: kring->ckr_rhead, lim: kring->ckr_lim); |
| 226 | error = ENODEV; |
| 227 | SK_ERR("kr \"%s\" (0x%llx) krflags 0x%b in drop mode (err %d)" , |
| 228 | kring->ckr_name, SK_KVA(kring), kring->ckr_flags, |
| 229 | CKRF_BITS, error); |
| 230 | } |
| 231 | |
| 232 | SK_DF(SK_VERB_FSW | SK_VERB_SYNC | SK_VERB_TX, |
| 233 | "%s(%d) kr \"%s\" (0x%llx) krflags 0x%b ring %u flags 0x%x" , |
| 234 | sk_proc_name_address(p), sk_proc_pid(p), kring->ckr_name, |
| 235 | SK_KVA(kring), kring->ckr_flags, CKRF_BITS, kring->ckr_ring_id, |
| 236 | flags); |
| 237 | |
| 238 | return error; |
| 239 | } |
| 240 | |
| 241 | /* |
| 242 | * na_rxsync callback for flow switch ports. We're already protected |
| 243 | * against concurrent calls from userspace. |
| 244 | */ |
| 245 | int |
| 246 | fsw_vp_na_rxsync(struct __kern_channel_ring *kring, struct proc *p, |
| 247 | uint32_t flags) |
| 248 | { |
| 249 | #pragma unused(p, flags) |
| 250 | slot_idx_t head, khead_prev; |
| 251 | |
| 252 | head = kring->ckr_rhead; |
| 253 | ASSERT(head <= kring->ckr_lim); |
| 254 | |
| 255 | /* First part, import newly received packets. */ |
| 256 | /* actually nothing to do here, they are already in the kring */ |
| 257 | |
| 258 | /* Second part, skip past packets that userspace has released. */ |
| 259 | khead_prev = kring->ckr_khead; |
| 260 | kring->ckr_khead = head; |
| 261 | |
| 262 | /* ensure global visibility */ |
| 263 | os_atomic_thread_fence(seq_cst); |
| 264 | |
| 265 | SK_DF(SK_VERB_FSW | SK_VERB_SYNC | SK_VERB_RX, |
| 266 | "%s(%d) kr \"%s\" (0x%llx) krflags 0x%b ring %u " |
| 267 | "kh %u (was %u) rh %u flags 0x%x" , sk_proc_name_address(p), |
| 268 | sk_proc_pid(p), kring->ckr_name, SK_KVA(kring), kring->ckr_flags, |
| 269 | CKRF_BITS, kring->ckr_ring_id, kring->ckr_khead, khead_prev, |
| 270 | kring->ckr_rhead, flags); |
| 271 | |
| 272 | return 0; |
| 273 | } |
| 274 | |
| 275 | static int |
| 276 | fsw_vp_na_special(struct nexus_adapter *na, struct kern_channel *ch, |
| 277 | struct chreq *chr, nxspec_cmd_t spec_cmd) |
| 278 | { |
| 279 | int error = 0; |
| 280 | |
| 281 | SK_LOCK_ASSERT_HELD(); |
| 282 | ASSERT(na->na_type == NA_FLOWSWITCH_VP); |
| 283 | |
| 284 | /* |
| 285 | * fsw_vp_na_attach() must have created this adapter |
| 286 | * exclusively for kernel (NAF_KERNEL); leave this alone. |
| 287 | */ |
| 288 | ASSERT(NA_KERNEL_ONLY(na)); |
| 289 | |
| 290 | switch (spec_cmd) { |
| 291 | case NXSPEC_CMD_CONNECT: |
| 292 | ASSERT(!(na->na_flags & NAF_SPEC_INIT)); |
| 293 | ASSERT(na->na_channels == 0); |
| 294 | |
| 295 | error = na_bind_channel(na, ch, chr); |
| 296 | if (error != 0) { |
| 297 | goto done; |
| 298 | } |
| 299 | |
| 300 | os_atomic_or(&na->na_flags, NAF_SPEC_INIT, relaxed); |
| 301 | break; |
| 302 | |
| 303 | case NXSPEC_CMD_DISCONNECT: |
| 304 | ASSERT(na->na_channels > 0); |
| 305 | ASSERT(na->na_flags & NAF_SPEC_INIT); |
| 306 | os_atomic_andnot(&na->na_flags, NAF_SPEC_INIT, relaxed); |
| 307 | |
| 308 | na_unbind_channel(ch); |
| 309 | break; |
| 310 | |
| 311 | case NXSPEC_CMD_START: |
| 312 | na_kr_drop(na, FALSE); |
| 313 | break; |
| 314 | |
| 315 | case NXSPEC_CMD_STOP: |
| 316 | na_kr_drop(na, TRUE); |
| 317 | break; |
| 318 | |
| 319 | default: |
| 320 | error = EINVAL; |
| 321 | break; |
| 322 | } |
| 323 | |
| 324 | done: |
| 325 | SK_DF(error ? SK_VERB_ERROR : SK_VERB_FSW, |
| 326 | "ch 0x%llx na \"%s\" (0x%llx) nx 0x%llx spec_cmd %u (err %d)" , |
| 327 | SK_KVA(ch), na->na_name, SK_KVA(na), SK_KVA(ch->ch_nexus), |
| 328 | spec_cmd, error); |
| 329 | |
| 330 | return error; |
| 331 | } |
| 332 | |
| 333 | /* |
| 334 | * Create a nexus_vp_adapter that describes a flow switch port. |
| 335 | */ |
| 336 | int |
| 337 | fsw_vp_na_create(struct kern_nexus *nx, struct chreq *chr, struct proc *p, |
| 338 | struct nexus_vp_adapter **ret) |
| 339 | { |
| 340 | struct nxprov_params *nxp = NX_PROV(nx)->nxprov_params; |
| 341 | struct nx_flowswitch *fsw = NX_FSW_PRIVATE(nx); |
| 342 | struct nexus_vp_adapter *vpna; |
| 343 | struct nexus_adapter *na; |
| 344 | int error; |
| 345 | |
| 346 | SK_LOCK_ASSERT_HELD(); |
| 347 | |
| 348 | if ((chr->cr_mode & CHMODE_KERNEL) != 0) { |
| 349 | SK_ERR("VP adapter can't be used by kernel" ); |
| 350 | return ENOTSUP; |
| 351 | } |
| 352 | if ((chr->cr_mode & CHMODE_USER_PACKET_POOL) == 0) { |
| 353 | SK_ERR("user packet pool required" ); |
| 354 | return EINVAL; |
| 355 | } |
| 356 | |
| 357 | vpna = fsw_vp_na_alloc(Z_WAITOK); |
| 358 | |
| 359 | ASSERT(vpna->vpna_up.na_type == NA_FLOWSWITCH_VP); |
| 360 | ASSERT(vpna->vpna_up.na_free == fsw_vp_na_free); |
| 361 | |
| 362 | na = &vpna->vpna_up; |
| 363 | (void) snprintf(na->na_name, count: sizeof(na->na_name), "fsw_%s[%u]_%s.%d" , |
| 364 | fsw->fsw_ifp ? if_name(fsw->fsw_ifp) : "??" , chr->cr_port, |
| 365 | proc_best_name(p), proc_pid(p)); |
| 366 | na->na_name[sizeof(na->na_name) - 1] = '\0'; |
| 367 | uuid_generate_random(out: na->na_uuid); |
| 368 | |
| 369 | /* |
| 370 | * Verify upper bounds; for all cases including user pipe nexus, |
| 371 | * as well as flow switch-based ones, the parameters must have |
| 372 | * already been validated by corresponding nxdom_prov_params() |
| 373 | * function defined by each domain. The user pipe nexus would |
| 374 | * be checking against the flow switch's parameters there. |
| 375 | */ |
| 376 | na_set_nrings(na, t: NR_TX, v: nxp->nxp_tx_rings); |
| 377 | na_set_nrings(na, t: NR_RX, v: nxp->nxp_rx_rings); |
| 378 | /* |
| 379 | * If the packet pool is configured to be multi-buflet, then we |
| 380 | * need 2 pairs of alloc/free rings(for packet and buflet). |
| 381 | */ |
| 382 | na_set_nrings(na, t: NR_A, v: ((nxp->nxp_max_frags > 1) && |
| 383 | (sk_channel_buflet_alloc != 0)) ? 2 : 1); |
| 384 | na_set_nslots(na, t: NR_TX, v: nxp->nxp_tx_slots); |
| 385 | na_set_nslots(na, t: NR_RX, v: nxp->nxp_rx_slots); |
| 386 | na_set_nslots(na, t: NR_A, NX_FSW_AFRINGSIZE); |
| 387 | ASSERT(na_get_nrings(na, NR_TX) <= NX_DOM(nx)->nxdom_tx_rings.nb_max); |
| 388 | ASSERT(na_get_nrings(na, NR_RX) <= NX_DOM(nx)->nxdom_rx_rings.nb_max); |
| 389 | ASSERT(na_get_nslots(na, NR_TX) <= NX_DOM(nx)->nxdom_tx_slots.nb_max); |
| 390 | ASSERT(na_get_nslots(na, NR_RX) <= NX_DOM(nx)->nxdom_rx_slots.nb_max); |
| 391 | |
| 392 | os_atomic_or(&na->na_flags, NAF_USER_PKT_POOL, relaxed); |
| 393 | |
| 394 | if (chr->cr_mode & CHMODE_LOW_LATENCY) { |
| 395 | os_atomic_or(&na->na_flags, NAF_LOW_LATENCY, relaxed); |
| 396 | } |
| 397 | |
| 398 | if (chr->cr_mode & CHMODE_EVENT_RING) { |
| 399 | na_set_nrings(na, t: NR_EV, NX_FSW_EVENT_RING_NUM); |
| 400 | na_set_nslots(na, t: NR_EV, NX_FSW_EVENT_RING_SIZE); |
| 401 | os_atomic_or(&na->na_flags, NAF_EVENT_RING, relaxed); |
| 402 | na->na_channel_event_notify = fsw_vp_na_channel_event_notify; |
| 403 | } |
| 404 | if (nxp->nxp_max_frags > 1 && fsw->fsw_tso_mode != FSW_TSO_MODE_NONE) { |
| 405 | na_set_nrings(na, t: NR_LBA, v: 1); |
| 406 | na_set_nslots(na, t: NR_LBA, NX_FSW_AFRINGSIZE); |
| 407 | } |
| 408 | vpna->vpna_nx_port = chr->cr_port; |
| 409 | na->na_dtor = fsw_vp_na_dtor; |
| 410 | na->na_activate = fsw_vp_na_activate; |
| 411 | na->na_txsync = fsw_vp_na_txsync; |
| 412 | na->na_rxsync = fsw_vp_na_rxsync; |
| 413 | na->na_krings_create = fsw_vp_na_krings_create; |
| 414 | na->na_krings_delete = fsw_vp_na_krings_delete; |
| 415 | na->na_special = fsw_vp_na_special; |
| 416 | |
| 417 | *(nexus_stats_type_t *)(uintptr_t)&na->na_stats_type = |
| 418 | NEXUS_STATS_TYPE_FSW; |
| 419 | |
| 420 | /* other fields are set in the common routine */ |
| 421 | na_attach_common(na, nx, &nx_fsw_prov_s); |
| 422 | |
| 423 | if ((error = NX_DOM_PROV(nx)->nxdom_prov_mem_new(NX_DOM_PROV(nx), |
| 424 | nx, na)) != 0) { |
| 425 | ASSERT(na->na_arena == NULL); |
| 426 | goto err; |
| 427 | } |
| 428 | ASSERT(na->na_arena != NULL); |
| 429 | |
| 430 | *(uint32_t *)(uintptr_t)&na->na_flowadv_max = nxp->nxp_flowadv_max; |
| 431 | ASSERT(na->na_flowadv_max == 0 || |
| 432 | skmem_arena_nexus(na->na_arena)->arn_flowadv_obj != NULL); |
| 433 | |
| 434 | #if SK_LOG |
| 435 | uuid_string_t uuidstr; |
| 436 | SK_DF(SK_VERB_FSW, "na_name: \"%s\"" , na->na_name); |
| 437 | SK_DF(SK_VERB_FSW, " UUID: %s" , sk_uuid_unparse(na->na_uuid, |
| 438 | uuidstr)); |
| 439 | SK_DF(SK_VERB_FSW, " nx: 0x%llx (\"%s\":\"%s\")" , |
| 440 | SK_KVA(na->na_nx), NX_DOM(na->na_nx)->nxdom_name, |
| 441 | NX_DOM_PROV(na->na_nx)->nxdom_prov_name); |
| 442 | SK_DF(SK_VERB_FSW, " flags: 0x%b" , na->na_flags, NAF_BITS); |
| 443 | SK_DF(SK_VERB_FSW, " stats_type: %u" , na->na_stats_type); |
| 444 | SK_DF(SK_VERB_FSW, " flowadv_max: %u" , na->na_flowadv_max); |
| 445 | SK_DF(SK_VERB_FSW, " rings: tx %u rx %u af %u" , |
| 446 | na_get_nrings(na, NR_TX), na_get_nrings(na, NR_RX), |
| 447 | na_get_nrings(na, NR_A)); |
| 448 | SK_DF(SK_VERB_FSW, " slots: tx %u rx %u af %u" , |
| 449 | na_get_nslots(na, NR_TX), na_get_nslots(na, NR_RX), |
| 450 | na_get_nslots(na, NR_A)); |
| 451 | #if CONFIG_NEXUS_USER_PIPE |
| 452 | SK_DF(SK_VERB_FSW, " next_pipe: %u" , na->na_next_pipe); |
| 453 | SK_DF(SK_VERB_FSW, " max_pipes: %u" , na->na_max_pipes); |
| 454 | #endif /* CONFIG_NEXUS_USER_PIPE */ |
| 455 | SK_DF(SK_VERB_FSW, " nx_port: %d" , (int)vpna->vpna_nx_port); |
| 456 | #endif /* SK_LOG */ |
| 457 | |
| 458 | *ret = vpna; |
| 459 | na_retain_locked(na: &vpna->vpna_up); |
| 460 | |
| 461 | return 0; |
| 462 | |
| 463 | err: |
| 464 | if (na->na_arena != NULL) { |
| 465 | skmem_arena_release(na->na_arena); |
| 466 | na->na_arena = NULL; |
| 467 | } |
| 468 | NA_FREE(&vpna->vpna_up); |
| 469 | return error; |
| 470 | } |
| 471 | |
| 472 | static struct nexus_vp_adapter * |
| 473 | fsw_vp_na_alloc(zalloc_flags_t how) |
| 474 | { |
| 475 | struct nexus_vp_adapter *vpna; |
| 476 | |
| 477 | _CASSERT(offsetof(struct nexus_vp_adapter, vpna_up) == 0); |
| 478 | |
| 479 | vpna = zalloc_flags(na_vp_zone, how | Z_ZERO); |
| 480 | if (vpna) { |
| 481 | vpna->vpna_up.na_type = NA_FLOWSWITCH_VP; |
| 482 | vpna->vpna_up.na_free = fsw_vp_na_free; |
| 483 | } |
| 484 | return vpna; |
| 485 | } |
| 486 | |
| 487 | static void |
| 488 | fsw_vp_na_free(struct nexus_adapter *na) |
| 489 | { |
| 490 | struct nexus_vp_adapter *vpna = (struct nexus_vp_adapter *)(void *)na; |
| 491 | |
| 492 | ASSERT(vpna->vpna_up.na_refcount == 0); |
| 493 | SK_DF(SK_VERB_MEM, "vpna 0x%llx FREE" , SK_KVA(vpna)); |
| 494 | bzero(s: vpna, n: sizeof(*vpna)); |
| 495 | zfree(na_vp_zone, vpna); |
| 496 | } |
| 497 | |
| 498 | void |
| 499 | fsw_vp_channel_error_stats_fold(struct fsw_stats *fs, |
| 500 | struct __nx_stats_channel_errors *es) |
| 501 | { |
| 502 | STATS_ADD(fs, FSW_STATS_CHAN_ERR_UPP_ALLOC, |
| 503 | es->nxs_cres->cres_pkt_alloc_failures); |
| 504 | } |
| 505 | |
| 506 | SK_NO_INLINE_ATTRIBUTE |
| 507 | static struct __kern_packet * |
| 508 | nx_fsw_alloc_packet(struct kern_pbufpool *pp, uint32_t sz, kern_packet_t *php) |
| 509 | { |
| 510 | kern_packet_t ph; |
| 511 | ph = pp_alloc_packet_by_size(pp, sz, SKMEM_NOSLEEP); |
| 512 | if (__improbable(ph == 0)) { |
| 513 | DTRACE_SKYWALK2(alloc__fail, struct kern_pbufpool *, |
| 514 | pp, size_t, sz); |
| 515 | return NULL; |
| 516 | } |
| 517 | if (php != NULL) { |
| 518 | *php = ph; |
| 519 | } |
| 520 | return SK_PTR_ADDR_KPKT(ph); |
| 521 | } |
| 522 | |
| 523 | SK_NO_INLINE_ATTRIBUTE |
| 524 | static void |
| 525 | nx_fsw_free_packet(struct __kern_packet *pkt) |
| 526 | { |
| 527 | pp_free_packet_single(pkt); |
| 528 | } |
| 529 | |
| 530 | static int |
| 531 | fsw_vp_na_channel_event_notify(struct nexus_adapter *vpna, |
| 532 | struct __kern_channel_event *ev, uint16_t ev_len) |
| 533 | { |
| 534 | int err; |
| 535 | char *baddr; |
| 536 | kern_packet_t ph; |
| 537 | kern_buflet_t buf; |
| 538 | sk_protect_t protect; |
| 539 | kern_channel_slot_t slot; |
| 540 | struct __kern_packet *vpna_pkt = NULL; |
| 541 | struct __kern_channel_event_metadata *emd; |
| 542 | struct __kern_channel_ring *ring = &vpna->na_event_rings[0]; |
| 543 | struct fsw_stats *fs = &((struct nexus_vp_adapter *)(vpna))->vpna_fsw->fsw_stats; |
| 544 | |
| 545 | if (__probable(ev->ev_type == CHANNEL_EVENT_PACKET_TRANSMIT_STATUS)) { |
| 546 | STATS_INC(fs, FSW_STATS_EV_RECV_TX_STATUS); |
| 547 | } |
| 548 | if (__improbable(ev->ev_type == CHANNEL_EVENT_PACKET_TRANSMIT_EXPIRED)) { |
| 549 | STATS_INC(fs, FSW_STATS_EV_RECV_TX_EXPIRED); |
| 550 | } |
| 551 | STATS_INC(fs, FSW_STATS_EV_RECV); |
| 552 | |
| 553 | if (__improbable(!NA_IS_ACTIVE(vpna))) { |
| 554 | STATS_INC(fs, FSW_STATS_EV_DROP_NA_INACTIVE); |
| 555 | err = ENXIO; |
| 556 | goto error; |
| 557 | } |
| 558 | if (__improbable(NA_IS_DEFUNCT(vpna))) { |
| 559 | STATS_INC(fs, FSW_STATS_EV_DROP_NA_DEFUNCT); |
| 560 | err = ENXIO; |
| 561 | goto error; |
| 562 | } |
| 563 | if (!NA_CHANNEL_EVENT_ATTACHED(vpna)) { |
| 564 | STATS_INC(fs, FSW_STATS_EV_DROP_KEVENT_INACTIVE); |
| 565 | err = ENXIO; |
| 566 | goto error; |
| 567 | } |
| 568 | if (__improbable(KR_DROP(ring))) { |
| 569 | STATS_INC(fs, FSW_STATS_EV_DROP_KRDROP_MODE); |
| 570 | err = ENXIO; |
| 571 | goto error; |
| 572 | } |
| 573 | |
| 574 | vpna_pkt = nx_fsw_alloc_packet(pp: ring->ckr_pp, sz: ev_len, php: &ph); |
| 575 | if (__improbable(vpna_pkt == NULL)) { |
| 576 | STATS_INC(fs, FSW_STATS_EV_DROP_NOMEM_PKT); |
| 577 | err = ENOMEM; |
| 578 | goto error; |
| 579 | } |
| 580 | buf = __packet_get_next_buflet(ph, NULL); |
| 581 | baddr = __buflet_get_data_address(buf); |
| 582 | emd = (struct __kern_channel_event_metadata *)(void *)baddr; |
| 583 | emd->emd_etype = ev->ev_type; |
| 584 | emd->emd_nevents = 1; |
| 585 | bcopy(src: ev, dst: (baddr + __KERN_CHANNEL_EVENT_OFFSET), n: ev_len); |
| 586 | err = __buflet_set_data_length(buf, |
| 587 | dlen: (ev_len + __KERN_CHANNEL_EVENT_OFFSET)); |
| 588 | VERIFY(err == 0); |
| 589 | err = __packet_finalize(ph); |
| 590 | VERIFY(err == 0); |
| 591 | kr_enter(ring, TRUE); |
| 592 | protect = sk_sync_protect(); |
| 593 | slot = kern_channel_get_next_slot(kring: ring, NULL, NULL); |
| 594 | if (slot == NULL) { |
| 595 | sk_sync_unprotect(protect); |
| 596 | kr_exit(ring); |
| 597 | STATS_INC(fs, FSW_STATS_EV_DROP_KRSPACE); |
| 598 | err = ENOSPC; |
| 599 | goto error; |
| 600 | } |
| 601 | err = kern_channel_slot_attach_packet(ring, slot, packet: ph); |
| 602 | VERIFY(err == 0); |
| 603 | vpna_pkt = NULL; |
| 604 | kern_channel_advance_slot(kring: ring, slot); |
| 605 | sk_sync_unprotect(protect); |
| 606 | kr_exit(ring); |
| 607 | kern_channel_event_notify(&vpna->na_tx_rings[0]); |
| 608 | STATS_INC(fs, NETIF_STATS_EV_SENT); |
| 609 | return 0; |
| 610 | |
| 611 | error: |
| 612 | ASSERT(err != 0); |
| 613 | if (vpna_pkt != NULL) { |
| 614 | nx_fsw_free_packet(pkt: vpna_pkt); |
| 615 | } |
| 616 | STATS_INC(fs, FSW_STATS_EV_DROP); |
| 617 | return err; |
| 618 | } |
| 619 | |
| 620 | static inline struct nexus_adapter * |
| 621 | fsw_find_port_vpna(struct nx_flowswitch *fsw, uint32_t nx_port_id) |
| 622 | { |
| 623 | struct kern_nexus *nx = fsw->fsw_nx; |
| 624 | struct nexus_adapter *na = NULL; |
| 625 | nexus_port_t port; |
| 626 | uint16_t gencnt; |
| 627 | |
| 628 | PKT_DECOMPOSE_NX_PORT_ID(nx_port_id, port, gencnt); |
| 629 | |
| 630 | if (port < FSW_VP_USER_MIN) { |
| 631 | SK_ERR("non VPNA port" ); |
| 632 | return NULL; |
| 633 | } |
| 634 | |
| 635 | if (__improbable(!nx_port_is_valid(nx, port))) { |
| 636 | SK_ERR("%s[%d] port no longer valid" , |
| 637 | if_name(fsw->fsw_ifp), port); |
| 638 | return NULL; |
| 639 | } |
| 640 | |
| 641 | na = nx_port_get_na(nx, port); |
| 642 | if (na != NULL && VPNA(na)->vpna_gencnt != gencnt) { |
| 643 | return NULL; |
| 644 | } |
| 645 | return na; |
| 646 | } |
| 647 | |
| 648 | errno_t |
| 649 | fsw_vp_na_channel_event(struct nx_flowswitch *fsw, uint32_t nx_port_id, |
| 650 | struct __kern_channel_event *event, uint16_t event_len) |
| 651 | { |
| 652 | int err = 0; |
| 653 | struct nexus_adapter *fsw_vpna; |
| 654 | |
| 655 | SK_DF(SK_VERB_EVENTS, "%s[%d] ev: %p ev_len: %hu " |
| 656 | "ev_type: %u ev_flags: %u _reserved: %hu ev_dlen: %hu" , |
| 657 | if_name(fsw->fsw_ifp), nx_port_id, event, event_len, |
| 658 | event->ev_type, event->ev_flags, event->_reserved, event->ev_dlen); |
| 659 | |
| 660 | FSW_RLOCK(fsw); |
| 661 | struct fsw_stats *fs = &fsw->fsw_stats; |
| 662 | |
| 663 | fsw_vpna = fsw_find_port_vpna(fsw, nx_port_id); |
| 664 | if (__improbable(fsw_vpna == NULL)) { |
| 665 | err = ENXIO; |
| 666 | STATS_INC(fs, FSW_STATS_EV_DROP_DEMUX_ERR); |
| 667 | goto error; |
| 668 | } |
| 669 | if (__improbable(fsw_vpna->na_channel_event_notify == NULL)) { |
| 670 | err = ENOTSUP; |
| 671 | STATS_INC(fs, FSW_STATS_EV_DROP_EV_VPNA_NOTSUP); |
| 672 | goto error; |
| 673 | } |
| 674 | err = fsw_vpna->na_channel_event_notify(fsw_vpna, event, event_len); |
| 675 | FSW_RUNLOCK(fsw); |
| 676 | return err; |
| 677 | |
| 678 | error: |
| 679 | STATS_INC(fs, FSW_STATS_EV_DROP); |
| 680 | FSW_RUNLOCK(fsw); |
| 681 | return err; |
| 682 | } |
| 683 | |