1/*- 2 * Copyright (c) 2014-2017, Matthew Macy (mmacy@mattmacy.io) 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright notice, 9 * this list of conditions and the following disclaimer. 10 * 11 * 2. Neither the name of Matthew Macy nor the names of its 12 * contributors may be used to endorse or promote products derived from 13 * this software without specific prior written permission. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 16 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 19 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 20 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 21 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 22 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 23 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 24 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 25 * POSSIBILITY OF SUCH DAMAGE. 26 * 27 * $FreeBSD$ 28 */ 29#ifndef __IFLIB_H_ 30#define __IFLIB_H_ 31 32#include <sys/kobj.h> 33#include <sys/bus.h> 34#include <sys/cpuset.h> 35#include <machine/bus.h> 36#include <sys/nv.h> 37#include <sys/gtaskqueue.h> 38 39struct if_clone; 40 41/* 42 * The value type for indexing, limits max descriptors 43 * to 65535 can be conditionally redefined to uint32_t 44 * in the future if the need arises. 45 */ 46typedef uint16_t qidx_t; 47#define QIDX_INVALID 0xFFFF 48 49struct iflib_ctx; 50typedef struct iflib_ctx *if_ctx_t; 51struct if_shared_ctx; 52typedef struct if_shared_ctx *if_shared_ctx_t; 53struct if_int_delay_info; 54typedef struct if_int_delay_info *if_int_delay_info_t; 55struct if_pseudo; 56typedef struct if_pseudo *if_pseudo_t; 57 58/* 59 * File organization: 60 * - public structures 61 * - iflib accessors 62 * - iflib utility functions 63 * - iflib core functions 64 */ 65 66typedef struct if_rxd_frag { 67 uint8_t irf_flid; 68 qidx_t irf_idx; 69 uint16_t irf_len; 70} *if_rxd_frag_t; 71 72/* bnxt supports 64 with hardware LRO enabled */ 73#define IFLIB_MAX_RX_SEGS 64 74 75typedef struct if_rxd_info { 76 /* set by iflib */ 77 uint16_t iri_qsidx; /* qset index */ 78 uint16_t iri_vtag; /* vlan tag - if flag set */ 79 /* XXX redundant with the new irf_len field */ 80 uint16_t iri_len; /* packet length */ 81 qidx_t iri_cidx; /* consumer index of cq */ 82 if_t iri_ifp; /* driver may have >1 iface per softc */ 83 84 /* updated by driver */ 85 if_rxd_frag_t iri_frags; 86 uint32_t iri_flowid; /* RSS hash for packet */ 87 uint32_t iri_csum_flags; /* m_pkthdr csum flags */ 88 89 uint32_t iri_csum_data; /* m_pkthdr csum data */ 90 uint8_t iri_flags; /* mbuf flags for packet */ 91 uint8_t iri_nfrags; /* number of fragments in packet */ 92 uint8_t iri_rsstype; /* RSS hash type */ 93 uint8_t iri_pad; /* any padding in the received data */ 94} *if_rxd_info_t; 95 96typedef struct if_rxd_update { 97 uint64_t *iru_paddrs; 98 qidx_t *iru_idxs; 99 qidx_t iru_pidx; 100 uint16_t iru_qsidx; 101 uint16_t iru_count; 102 uint16_t iru_buf_size; 103 uint8_t iru_flidx; 104} *if_rxd_update_t; 105 106#define IPI_TX_INTR 0x1 /* send an interrupt when this packet is sent */ 107#define IPI_TX_IPV4 0x2 /* ethertype IPv4 */ 108#define IPI_TX_IPV6 0x4 /* ethertype IPv6 */ 109 110typedef struct if_pkt_info { 111 bus_dma_segment_t *ipi_segs; /* physical addresses */ 112 uint32_t ipi_len; /* packet length */ 113 uint16_t ipi_qsidx; /* queue set index */ 114 qidx_t ipi_nsegs; /* number of segments */ 115 116 qidx_t ipi_ndescs; /* number of descriptors used by encap */ 117 uint16_t ipi_flags; /* iflib per-packet flags */ 118 qidx_t ipi_pidx; /* start pidx for encap */ 119 qidx_t ipi_new_pidx; /* next available pidx post-encap */ 120 /* offload handling */ 121 uint8_t ipi_ehdrlen; /* ether header length */ 122 uint8_t ipi_ip_hlen; /* ip header length */ 123 uint8_t ipi_tcp_hlen; /* tcp header length */ 124 uint8_t ipi_ipproto; /* ip protocol */ 125 126 uint32_t ipi_csum_flags; /* packet checksum flags */ 127 uint16_t ipi_tso_segsz; /* tso segment size */ 128 uint16_t ipi_vtag; /* VLAN tag */ 129 uint16_t ipi_etype; /* ether header type */ 130 uint8_t ipi_tcp_hflags; /* tcp header flags */ 131 uint8_t ipi_mflags; /* packet mbuf flags */ 132 133 uint32_t ipi_tcp_seq; /* tcp seqno */ 134 uint32_t ipi_tcp_sum; /* tcp csum */ 135} *if_pkt_info_t; 136 137typedef struct if_irq { 138 struct resource *ii_res; 139 int ii_rid; 140 void *ii_tag; 141} *if_irq_t; 142 143struct if_int_delay_info { 144 if_ctx_t iidi_ctx; /* Back-pointer to the iflib ctx (softc) */ 145 int iidi_offset; /* Register offset to read/write */ 146 int iidi_value; /* Current value in usecs */ 147 struct sysctl_oid *iidi_oidp; 148 struct sysctl_req *iidi_req; 149}; 150 151typedef enum { 152 IFLIB_INTR_LEGACY, 153 IFLIB_INTR_MSI, 154 IFLIB_INTR_MSIX 155} iflib_intr_mode_t; 156 157/* 158 * This really belongs in pciio.h or some place more general 159 * but this is the only consumer for now. 160 */ 161typedef struct pci_vendor_info { 162 uint32_t pvi_vendor_id; 163 uint32_t pvi_device_id; 164 uint32_t pvi_subvendor_id; 165 uint32_t pvi_subdevice_id; 166 uint32_t pvi_rev_id; 167 uint32_t pvi_class_mask; 168 caddr_t pvi_name; 169} pci_vendor_info_t; 170 171#define PVID(vendor, devid, name) {vendor, devid, 0, 0, 0, 0, name} 172#define PVID_OEM(vendor, devid, svid, sdevid, revid, name) {vendor, devid, svid, sdevid, revid, 0, name} 173#define PVID_END {0, 0, 0, 0, 0, 0, NULL} 174 175#define IFLIB_PNP_DESCR "U32:vendor;U32:device;U32:subvendor;U32:subdevice;" \ 176 "U32:revision;U32:class;D:#" 177#define IFLIB_PNP_INFO(b, u, t) \ 178 MODULE_PNP_INFO(IFLIB_PNP_DESCR, b, u, t, nitems(t) - 1) 179 180typedef struct if_txrx { 181 int (*ift_txd_encap) (void *, if_pkt_info_t); 182 void (*ift_txd_flush) (void *, uint16_t, qidx_t pidx); 183 int (*ift_txd_credits_update) (void *, uint16_t qsidx, bool clear); 184 185 int (*ift_rxd_available) (void *, uint16_t qsidx, qidx_t pidx, qidx_t budget); 186 int (*ift_rxd_pkt_get) (void *, if_rxd_info_t ri); 187 void (*ift_rxd_refill) (void * , if_rxd_update_t iru); 188 void (*ift_rxd_flush) (void *, uint16_t qsidx, uint8_t flidx, qidx_t pidx); 189 int (*ift_legacy_intr) (void *); 190} *if_txrx_t; 191 192typedef struct if_softc_ctx { 193 int isc_vectors; 194 int isc_nrxqsets; 195 int isc_ntxqsets; 196 uint8_t isc_min_tx_latency; /* disable doorbell update batching */ 197 uint8_t isc_rx_mvec_enable; /* generate mvecs on rx */ 198 uint32_t isc_txrx_budget_bytes_max; 199 int isc_msix_bar; /* can be model specific - initialize in attach_pre */ 200 int isc_tx_nsegments; /* can be model specific - initialize in attach_pre */ 201 int isc_ntxd[8]; 202 int isc_nrxd[8]; 203 204 uint32_t isc_txqsizes[8]; 205 uint32_t isc_rxqsizes[8]; 206 /* is there such thing as a descriptor that is more than 248 bytes ? */ 207 uint8_t isc_txd_size[8]; 208 uint8_t isc_rxd_size[8]; 209 210 int isc_tx_tso_segments_max; 211 int isc_tx_tso_size_max; 212 int isc_tx_tso_segsize_max; 213 int isc_tx_csum_flags; 214 int isc_capabilities; 215 int isc_capenable; 216 int isc_rss_table_size; 217 int isc_rss_table_mask; 218 int isc_nrxqsets_max; 219 int isc_ntxqsets_max; 220 uint32_t isc_tx_qdepth; 221 222 iflib_intr_mode_t isc_intr; 223 uint16_t isc_rxd_buf_size[8]; /* set at init time by driver, 0 224 means use iflib-calculated size 225 based on isc_max_frame_size */ 226 uint16_t isc_max_frame_size; /* set at init time by driver */ 227 uint16_t isc_min_frame_size; /* set at init time by driver, only used if 228 IFLIB_NEED_ETHER_PAD is set. */ 229 uint32_t isc_pause_frames; /* set by driver for iflib_timer to detect */ 230 pci_vendor_info_t isc_vendor_info; /* set by iflib prior to attach_pre */ 231 int isc_disable_msix; 232 if_txrx_t isc_txrx; 233} *if_softc_ctx_t; 234 235/* 236 * Initialization values for device 237 */ 238struct if_shared_ctx { 239 unsigned isc_magic; 240 driver_t *isc_driver; 241 bus_size_t isc_q_align; 242 bus_size_t isc_tx_maxsize; 243 bus_size_t isc_tx_maxsegsize; 244 bus_size_t isc_tso_maxsize; 245 bus_size_t isc_tso_maxsegsize; 246 bus_size_t isc_rx_maxsize; 247 bus_size_t isc_rx_maxsegsize; 248 int isc_rx_nsegments; 249 int isc_admin_intrcnt; /* # of admin/link interrupts */ 250 251 /* fields necessary for probe */ 252 pci_vendor_info_t *isc_vendor_info; 253 const char *isc_driver_version; 254 /* optional function to transform the read values to match the table*/ 255 void (*isc_parse_devinfo) (uint16_t *device_id, uint16_t *subvendor_id, 256 uint16_t *subdevice_id, uint16_t *rev_id); 257 int isc_nrxd_min[8]; 258 int isc_nrxd_default[8]; 259 int isc_nrxd_max[8]; 260 int isc_ntxd_min[8]; 261 int isc_ntxd_default[8]; 262 int isc_ntxd_max[8]; 263 264 /* actively used during operation */ 265 int isc_nfl __aligned(CACHE_LINE_SIZE); 266 int isc_ntxqs; /* # of tx queues per tx qset - usually 1 */ 267 int isc_nrxqs; /* # of rx queues per rx qset - intel 1, chelsio 2, broadcom 3 */ 268 int isc_rx_process_limit; 269 int isc_tx_reclaim_thresh; 270 int isc_flags; 271 const char *isc_name; 272}; 273 274typedef struct iflib_dma_info { 275 bus_addr_t idi_paddr; 276 caddr_t idi_vaddr; 277 bus_dma_tag_t idi_tag; 278 bus_dmamap_t idi_map; 279 uint32_t idi_size; 280} *iflib_dma_info_t; 281 282#define IFLIB_MAGIC 0xCAFEF00D 283 284typedef enum { 285 IFLIB_INTR_RX, 286 IFLIB_INTR_TX, 287 IFLIB_INTR_RXTX, 288 IFLIB_INTR_ADMIN, 289 IFLIB_INTR_IOV, 290} iflib_intr_type_t; 291 292#ifndef ETH_ADDR_LEN 293#define ETH_ADDR_LEN 6 294#endif 295 296 297/* 298 * Interface has a separate completion queue for RX 299 */ 300#define IFLIB_HAS_RXCQ 0x01 301/* 302 * Driver has already allocated vectors 303 */ 304#define IFLIB_SKIP_MSIX 0x02 305/* 306 * Interface is a virtual function 307 */ 308#define IFLIB_IS_VF 0x04 309/* 310 * Interface has a separate completion queue for TX 311 */ 312#define IFLIB_HAS_TXCQ 0x08 313/* 314 * Interface does checksum in place 315 */ 316#define IFLIB_NEED_SCRATCH 0x10 317/* 318 * Interface doesn't expect in_pseudo for th_sum 319 */ 320#define IFLIB_TSO_INIT_IP 0x20 321/* 322 * Interface doesn't align IP header 323 */ 324#define IFLIB_DO_RX_FIXUP 0x40 325/* 326 * Driver needs csum zeroed for offloading 327 */ 328#define IFLIB_NEED_ZERO_CSUM 0x80 329/* 330 * Driver needs frames padded to some minimum length 331 */ 332#define IFLIB_NEED_ETHER_PAD 0x100 333/* 334 * Packets can be freed immediately after encap 335 */ 336#define IFLIB_TXD_ENCAP_PIO 0x00200 337/* 338 * Use RX completion handler 339 */ 340#define IFLIB_RX_COMPLETION 0x00400 341/* 342 * Skip refilling cluster free lists 343 */ 344#define IFLIB_SKIP_CLREFILL 0x00800 345/* 346 * Don't reset on hang 347 */ 348#define IFLIB_NO_HANG_RESET 0x01000 349/* 350 * Don't need/want most of the niceties of 351 * queue management 352 */ 353#define IFLIB_PSEUDO 0x02000 354/* 355 * No DMA support needed / wanted 356 */ 357#define IFLIB_VIRTUAL 0x04000 358/* 359 * autogenerate a MAC address 360 */ 361#define IFLIB_GEN_MAC 0x08000 362/* 363 * Interface needs admin task to ignore interface up/down status 364 */ 365#define IFLIB_ADMIN_ALWAYS_RUN 0x10000 366/* 367 * When using a single hardware interrupt for the interface, only process RX 368 * interrupts instead of doing combined RX/TX processing. 369 */ 370#define IFLIB_SINGLE_IRQ_RX_ONLY 0x40000 371/* 372 * Don't need/want most of the niceties of 373 * emulating ethernet 374 */ 375#define IFLIB_PSEUDO_ETHER 0x80000 376 377 378/* 379 * These enum values are used in iflib_needs_restart to indicate to iflib 380 * functions whether or not the interface needs restarting when certain events 381 * happen. 382 */ 383enum iflib_restart_event { 384 IFLIB_RESTART_VLAN_CONFIG, 385}; 386 387/* 388 * field accessors 389 */ 390void *iflib_get_softc(if_ctx_t ctx); 391 392device_t iflib_get_dev(if_ctx_t ctx); 393 394if_t iflib_get_ifp(if_ctx_t ctx); 395 396struct ifmedia *iflib_get_media(if_ctx_t ctx); 397 398if_softc_ctx_t iflib_get_softc_ctx(if_ctx_t ctx); 399if_shared_ctx_t iflib_get_sctx(if_ctx_t ctx); 400 401void iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN]); 402void iflib_request_reset(if_ctx_t ctx); 403uint8_t iflib_in_detach(if_ctx_t ctx); 404 405uint32_t iflib_get_rx_mbuf_sz(if_ctx_t ctx); 406 407/* 408 * If the driver can plug cleanly in to newbus use these 409 */ 410int iflib_device_probe(device_t); 411int iflib_device_attach(device_t); 412int iflib_device_detach(device_t); 413int iflib_device_suspend(device_t); 414int iflib_device_resume(device_t); 415int iflib_device_shutdown(device_t); 416 417/* 418 * Use this instead of iflib_device_probe if the driver should report 419 * BUS_PROBE_VENDOR instead of BUS_PROBE_DEFAULT. (For example, an out-of-tree 420 * driver based on iflib). 421 */ 422int iflib_device_probe_vendor(device_t); 423 424 425int iflib_device_iov_init(device_t, uint16_t, const nvlist_t *); 426void iflib_device_iov_uninit(device_t); 427int iflib_device_iov_add_vf(device_t, uint16_t, const nvlist_t *); 428 429/* 430 * If the driver can't plug cleanly in to newbus 431 * use these 432 */ 433int iflib_device_register(device_t dev, void *softc, if_shared_ctx_t sctx, if_ctx_t *ctxp); 434int iflib_device_deregister(if_ctx_t); 435 436 437 438int iflib_irq_alloc(if_ctx_t, if_irq_t, int, driver_filter_t, void *filter_arg, driver_intr_t, void *arg, const char *name); 439int iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid, 440 iflib_intr_type_t type, driver_filter_t *filter, 441 void *filter_arg, int qid, const char *name); 442void iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, const char *name); 443 444void iflib_irq_free(if_ctx_t ctx, if_irq_t irq); 445 446void iflib_io_tqg_attach(struct grouptask *gt, void *uniq, int cpu, char *name); 447 448void iflib_config_gtask_init(void *ctx, struct grouptask *gtask, 449 gtask_fn_t *fn, const char *name); 450 451void iflib_config_gtask_deinit(struct grouptask *gtask); 452 453 454 455void iflib_tx_intr_deferred(if_ctx_t ctx, int txqid); 456void iflib_rx_intr_deferred(if_ctx_t ctx, int rxqid); 457void iflib_admin_intr_deferred(if_ctx_t ctx); 458void iflib_iov_intr_deferred(if_ctx_t ctx); 459 460 461void iflib_link_state_change(if_ctx_t ctx, int linkstate, uint64_t baudrate); 462 463int iflib_dma_alloc(if_ctx_t ctx, int size, iflib_dma_info_t dma, int mapflags); 464int iflib_dma_alloc_align(if_ctx_t ctx, int size, int align, iflib_dma_info_t dma, int mapflags); 465void iflib_dma_free(iflib_dma_info_t dma); 466 467int iflib_dma_alloc_multi(if_ctx_t ctx, int *sizes, iflib_dma_info_t *dmalist, int mapflags, int count); 468 469void iflib_dma_free_multi(iflib_dma_info_t *dmalist, int count); 470 471 472struct sx *iflib_ctx_lock_get(if_ctx_t); 473struct mtx *iflib_qset_lock_get(if_ctx_t, uint16_t); 474 475void iflib_led_create(if_ctx_t ctx); 476 477void iflib_add_int_delay_sysctl(if_ctx_t, const char *, const char *, 478 if_int_delay_info_t, int, int); 479 480/* 481 * Pseudo device support 482 */ 483if_pseudo_t iflib_clone_register(if_shared_ctx_t); 484void iflib_clone_deregister(if_pseudo_t); 485#endif /* __IFLIB_H_ */ 486