1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2015-2020 Amazon.com, Inc. or its affiliates. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 19 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 20 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 21 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 22 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 23 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 24 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 28 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 * 30 * $FreeBSD$ 31 * 32 */ 33 34 #ifndef ENA_H 35 #define ENA_H 36 37 #include "opt_rss.h" 38 39 #include "ena-com/ena_com.h" 40 #include "ena-com/ena_eth_com.h" 41 42 #define DRV_MODULE_VER_MAJOR 2 43 #define DRV_MODULE_VER_MINOR 5 44 #define DRV_MODULE_VER_SUBMINOR 0 45 46 #define DRV_MODULE_NAME "ena" 47 48 #ifndef DRV_MODULE_VERSION 49 #define DRV_MODULE_VERSION \ 50 __XSTRING(DRV_MODULE_VER_MAJOR) "." \ 51 __XSTRING(DRV_MODULE_VER_MINOR) "." \ 52 __XSTRING(DRV_MODULE_VER_SUBMINOR) 53 #endif 54 #define DEVICE_NAME "Elastic Network Adapter (ENA)" 55 #define DEVICE_DESC "ENA adapter" 56 57 /* Calculate DMA mask - width for ena cannot exceed 48, so it is safe */ 58 #define ENA_DMA_BIT_MASK(x) ((1ULL << (x)) - 1ULL) 59 60 /* 1 for AENQ + ADMIN */ 61 #define ENA_ADMIN_MSIX_VEC 1 62 #define ENA_MAX_MSIX_VEC(io_queues) (ENA_ADMIN_MSIX_VEC + (io_queues)) 63 64 #define ENA_REG_BAR 0 65 #define ENA_MEM_BAR 2 66 67 #define ENA_BUS_DMA_SEGS 32 68 69 #define ENA_DEFAULT_BUF_RING_SIZE 4096 70 71 #define ENA_DEFAULT_RING_SIZE 1024 72 #define ENA_MIN_RING_SIZE 256 73 74 /* 75 * Refill Rx queue when number of required descriptors is above 76 * QUEUE_SIZE / ENA_RX_REFILL_THRESH_DIVIDER or ENA_RX_REFILL_THRESH_PACKET 77 */ 78 #define ENA_RX_REFILL_THRESH_DIVIDER 8 79 #define ENA_RX_REFILL_THRESH_PACKET 256 80 81 #define ENA_IRQNAME_SIZE 40 82 83 #define ENA_PKT_MAX_BUFS 19 84 85 #define ENA_RX_RSS_TABLE_LOG_SIZE 7 86 #define ENA_RX_RSS_TABLE_SIZE (1 << ENA_RX_RSS_TABLE_LOG_SIZE) 87 88 #define ENA_HASH_KEY_SIZE 40 89 90 #define ENA_MAX_FRAME_LEN 10000 91 #define ENA_MIN_FRAME_LEN 60 92 93 #define ENA_TX_RESUME_THRESH (ENA_PKT_MAX_BUFS + 2) 94 95 #define DB_THRESHOLD 64 96 97 #define TX_COMMIT 32 98 /* 99 * TX budget for cleaning. It should be half of the RX budget to reduce amount 100 * of TCP retransmissions. 101 */ 102 #define TX_BUDGET 128 103 /* RX cleanup budget. -1 stands for infinity. */ 104 #define RX_BUDGET 256 105 /* 106 * How many times we can repeat cleanup in the io irq handling routine if the 107 * RX or TX budget was depleted. 108 */ 109 #define CLEAN_BUDGET 8 110 111 #define RX_IRQ_INTERVAL 20 112 #define TX_IRQ_INTERVAL 50 113 114 #define ENA_MIN_MTU 128 115 116 #define ENA_TSO_MAXSIZE 65536 117 118 #define ENA_MMIO_DISABLE_REG_READ BIT(0) 119 120 #define ENA_TX_RING_IDX_NEXT(idx, ring_size) (((idx) + 1) & ((ring_size) - 1)) 121 122 #define ENA_RX_RING_IDX_NEXT(idx, ring_size) (((idx) + 1) & ((ring_size) - 1)) 123 124 #define ENA_IO_TXQ_IDX(q) (2 * (q)) 125 #define ENA_IO_RXQ_IDX(q) (2 * (q) + 1) 126 #define ENA_IO_TXQ_IDX_TO_COMBINED_IDX(q) ((q) / 2) 127 #define ENA_IO_RXQ_IDX_TO_COMBINED_IDX(q) (((q) - 1) / 2) 128 129 #define ENA_MGMNT_IRQ_IDX 0 130 #define ENA_IO_IRQ_FIRST_IDX 1 131 #define ENA_IO_IRQ_IDX(q) (ENA_IO_IRQ_FIRST_IDX + (q)) 132 133 #define ENA_MAX_NO_INTERRUPT_ITERATIONS 3 134 135 /* 136 * ENA device should send keep alive msg every 1 sec. 137 * We wait for 6 sec just to be on the safe side. 138 */ 139 #define DEFAULT_KEEP_ALIVE_TO (SBT_1S * 6) 140 141 /* Time in jiffies before concluding the transmitter is hung. */ 142 #define DEFAULT_TX_CMP_TO (SBT_1S * 5) 143 144 /* Number of queues to check for missing queues per timer tick */ 145 #define DEFAULT_TX_MONITORED_QUEUES (4) 146 147 /* Max number of timeouted packets before device reset */ 148 #define DEFAULT_TX_CMP_THRESHOLD (128) 149 150 /* 151 * Supported PCI vendor and devices IDs 152 */ 153 #define PCI_VENDOR_ID_AMAZON 0x1d0f 154 155 #define PCI_DEV_ID_ENA_PF 0x0ec2 156 #define PCI_DEV_ID_ENA_PF_RSERV0 0x1ec2 157 #define PCI_DEV_ID_ENA_VF 0xec20 158 #define PCI_DEV_ID_ENA_VF_RSERV0 0xec21 159 160 /* 161 * Flags indicating current ENA driver state 162 */ 163 enum ena_flags_t { 164 ENA_FLAG_DEVICE_RUNNING, 165 ENA_FLAG_DEV_UP, 166 ENA_FLAG_LINK_UP, 167 ENA_FLAG_MSIX_ENABLED, 168 ENA_FLAG_TRIGGER_RESET, 169 ENA_FLAG_ONGOING_RESET, 170 ENA_FLAG_DEV_UP_BEFORE_RESET, 171 ENA_FLAG_RSS_ACTIVE, 172 ENA_FLAGS_NUMBER = ENA_FLAG_RSS_ACTIVE 173 }; 174 175 BITSET_DEFINE(_ena_state, ENA_FLAGS_NUMBER); 176 typedef struct _ena_state ena_state_t; 177 178 #define ENA_FLAG_ZERO(adapter) \ 179 BIT_ZERO(ENA_FLAGS_NUMBER, &(adapter)->flags) 180 #define ENA_FLAG_ISSET(bit, adapter) \ 181 BIT_ISSET(ENA_FLAGS_NUMBER, (bit), &(adapter)->flags) 182 #define ENA_FLAG_SET_ATOMIC(bit, adapter) \ 183 BIT_SET_ATOMIC(ENA_FLAGS_NUMBER, (bit), &(adapter)->flags) 184 #define ENA_FLAG_CLEAR_ATOMIC(bit, adapter) \ 185 BIT_CLR_ATOMIC(ENA_FLAGS_NUMBER, (bit), &(adapter)->flags) 186 187 struct msix_entry { 188 int entry; 189 int vector; 190 }; 191 192 typedef struct _ena_vendor_info_t { 193 uint16_t vendor_id; 194 uint16_t device_id; 195 unsigned int index; 196 } ena_vendor_info_t; 197 198 struct ena_irq { 199 /* Interrupt resources */ 200 struct resource *res; 201 driver_filter_t *handler; 202 void *data; 203 void *cookie; 204 unsigned int vector; 205 bool requested; 206 #ifdef RSS 207 int cpu; 208 #endif 209 char name[ENA_IRQNAME_SIZE]; 210 }; 211 212 struct ena_que { 213 struct ena_adapter *adapter; 214 struct ena_ring *tx_ring; 215 struct ena_ring *rx_ring; 216 217 struct task cleanup_task; 218 struct taskqueue *cleanup_tq; 219 220 uint32_t id; 221 #ifdef RSS 222 int cpu; 223 cpuset_t cpu_mask; 224 #endif 225 int domain; 226 struct sysctl_oid *oid; 227 }; 228 229 struct ena_calc_queue_size_ctx { 230 struct ena_com_dev_get_features_ctx *get_feat_ctx; 231 struct ena_com_dev *ena_dev; 232 device_t pdev; 233 uint32_t tx_queue_size; 234 uint32_t rx_queue_size; 235 uint32_t max_tx_queue_size; 236 uint32_t max_rx_queue_size; 237 uint16_t max_tx_sgl_size; 238 uint16_t max_rx_sgl_size; 239 }; 240 241 #ifdef DEV_NETMAP 242 struct ena_netmap_tx_info { 243 uint32_t socket_buf_idx[ENA_PKT_MAX_BUFS]; 244 bus_dmamap_t map_seg[ENA_PKT_MAX_BUFS]; 245 unsigned int sockets_used; 246 }; 247 #endif 248 249 struct ena_tx_buffer { 250 struct mbuf *mbuf; 251 /* # of ena desc for this specific mbuf 252 * (includes data desc and metadata desc) */ 253 unsigned int tx_descs; 254 /* # of buffers used by this mbuf */ 255 unsigned int num_of_bufs; 256 257 bus_dmamap_t dmamap; 258 259 /* Used to detect missing tx packets */ 260 struct bintime timestamp; 261 bool print_once; 262 263 #ifdef DEV_NETMAP 264 struct ena_netmap_tx_info nm_info; 265 #endif /* DEV_NETMAP */ 266 267 struct ena_com_buf bufs[ENA_PKT_MAX_BUFS]; 268 } __aligned(CACHE_LINE_SIZE); 269 270 struct ena_rx_buffer { 271 struct mbuf *mbuf; 272 bus_dmamap_t map; 273 struct ena_com_buf ena_buf; 274 #ifdef DEV_NETMAP 275 uint32_t netmap_buf_idx; 276 #endif /* DEV_NETMAP */ 277 } __aligned(CACHE_LINE_SIZE); 278 279 struct ena_stats_tx { 280 counter_u64_t cnt; 281 counter_u64_t bytes; 282 counter_u64_t prepare_ctx_err; 283 counter_u64_t dma_mapping_err; 284 counter_u64_t doorbells; 285 counter_u64_t missing_tx_comp; 286 counter_u64_t bad_req_id; 287 counter_u64_t collapse; 288 counter_u64_t collapse_err; 289 counter_u64_t queue_wakeup; 290 counter_u64_t queue_stop; 291 counter_u64_t llq_buffer_copy; 292 counter_u64_t unmask_interrupt_num; 293 }; 294 295 struct ena_stats_rx { 296 counter_u64_t cnt; 297 counter_u64_t bytes; 298 counter_u64_t refil_partial; 299 counter_u64_t csum_bad; 300 counter_u64_t mjum_alloc_fail; 301 counter_u64_t mbuf_alloc_fail; 302 counter_u64_t dma_mapping_err; 303 counter_u64_t bad_desc_num; 304 counter_u64_t bad_req_id; 305 counter_u64_t empty_rx_ring; 306 counter_u64_t csum_good; 307 }; 308 309 struct ena_ring { 310 /* Holds the empty requests for TX/RX out of order completions */ 311 union { 312 uint16_t *free_tx_ids; 313 uint16_t *free_rx_ids; 314 }; 315 struct ena_com_dev *ena_dev; 316 struct ena_adapter *adapter; 317 struct ena_com_io_cq *ena_com_io_cq; 318 struct ena_com_io_sq *ena_com_io_sq; 319 320 uint16_t qid; 321 322 /* Determines if device will use LLQ or normal mode for TX */ 323 enum ena_admin_placement_policy_type tx_mem_queue_type; 324 union { 325 /* The maximum length the driver can push to the device (For LLQ) */ 326 uint8_t tx_max_header_size; 327 /* The maximum (and default) mbuf size for the Rx descriptor. */ 328 uint16_t rx_mbuf_sz; 329 330 }; 331 332 bool first_interrupt; 333 uint16_t no_interrupt_event_cnt; 334 335 struct ena_com_rx_buf_info ena_bufs[ENA_PKT_MAX_BUFS]; 336 337 struct ena_que *que; 338 struct lro_ctrl lro; 339 340 uint16_t next_to_use; 341 uint16_t next_to_clean; 342 343 union { 344 struct ena_tx_buffer *tx_buffer_info; /* contex of tx packet */ 345 struct ena_rx_buffer *rx_buffer_info; /* contex of rx packet */ 346 }; 347 int ring_size; /* number of tx/rx_buffer_info's entries */ 348 349 struct buf_ring *br; /* only for TX */ 350 uint32_t buf_ring_size; 351 352 struct mtx ring_mtx; 353 char mtx_name[16]; 354 355 struct { 356 struct task enqueue_task; 357 struct taskqueue *enqueue_tq; 358 }; 359 360 union { 361 struct ena_stats_tx tx_stats; 362 struct ena_stats_rx rx_stats; 363 }; 364 365 union { 366 int empty_rx_queue; 367 /* For Tx ring to indicate if it's running or not */ 368 bool running; 369 }; 370 371 /* How many packets are sent in one Tx loop, used for doorbells */ 372 uint32_t acum_pkts; 373 374 /* Used for LLQ */ 375 uint8_t *push_buf_intermediate_buf; 376 377 #ifdef DEV_NETMAP 378 bool initialized; 379 #endif /* DEV_NETMAP */ 380 } __aligned(CACHE_LINE_SIZE); 381 382 struct ena_stats_dev { 383 counter_u64_t wd_expired; 384 counter_u64_t interface_up; 385 counter_u64_t interface_down; 386 counter_u64_t admin_q_pause; 387 }; 388 389 struct ena_hw_stats { 390 counter_u64_t rx_packets; 391 counter_u64_t tx_packets; 392 393 counter_u64_t rx_bytes; 394 counter_u64_t tx_bytes; 395 396 counter_u64_t rx_drops; 397 counter_u64_t tx_drops; 398 }; 399 400 /* Board specific private data structure */ 401 struct ena_adapter { 402 struct ena_com_dev *ena_dev; 403 404 /* OS defined structs */ 405 if_t ifp; 406 device_t pdev; 407 struct ifmedia media; 408 409 /* OS resources */ 410 struct resource *memory; 411 struct resource *registers; 412 struct resource *msix; 413 int msix_rid; 414 415 /* MSI-X */ 416 struct msix_entry *msix_entries; 417 int msix_vecs; 418 419 /* DMA tags used throughout the driver adapter for Tx and Rx */ 420 bus_dma_tag_t tx_buf_tag; 421 bus_dma_tag_t rx_buf_tag; 422 int dma_width; 423 424 uint32_t max_mtu; 425 426 uint32_t num_io_queues; 427 uint32_t max_num_io_queues; 428 429 uint32_t requested_tx_ring_size; 430 uint32_t requested_rx_ring_size; 431 432 uint32_t max_tx_ring_size; 433 uint32_t max_rx_ring_size; 434 435 uint16_t max_tx_sgl_size; 436 uint16_t max_rx_sgl_size; 437 438 uint32_t tx_offload_cap; 439 440 uint32_t buf_ring_size; 441 442 /* RSS*/ 443 int first_bind; 444 struct ena_indir *rss_indir; 445 446 uint8_t mac_addr[ETHER_ADDR_LEN]; 447 /* mdio and phy*/ 448 449 ena_state_t flags; 450 451 /* Queue will represent one TX and one RX ring */ 452 struct ena_que que[ENA_MAX_NUM_IO_QUEUES] 453 __aligned(CACHE_LINE_SIZE); 454 455 /* TX */ 456 struct ena_ring tx_ring[ENA_MAX_NUM_IO_QUEUES] 457 __aligned(CACHE_LINE_SIZE); 458 459 /* RX */ 460 struct ena_ring rx_ring[ENA_MAX_NUM_IO_QUEUES] 461 __aligned(CACHE_LINE_SIZE); 462 463 struct ena_irq irq_tbl[ENA_MAX_MSIX_VEC(ENA_MAX_NUM_IO_QUEUES)]; 464 465 /* Timer service */ 466 struct callout timer_service; 467 sbintime_t keep_alive_timestamp; 468 uint32_t next_monitored_tx_qid; 469 struct task reset_task; 470 struct taskqueue *reset_tq; 471 int wd_active; 472 sbintime_t keep_alive_timeout; 473 sbintime_t missing_tx_timeout; 474 uint32_t missing_tx_max_queues; 475 uint32_t missing_tx_threshold; 476 bool disable_meta_caching; 477 478 uint16_t eni_metrics_sample_interval; 479 uint16_t eni_metrics_sample_interval_cnt; 480 481 /* Statistics */ 482 struct ena_stats_dev dev_stats; 483 struct ena_hw_stats hw_stats; 484 struct ena_admin_eni_stats eni_metrics; 485 486 enum ena_regs_reset_reason_types reset_reason; 487 }; 488 489 #define ENA_RING_MTX_LOCK(_ring) mtx_lock(&(_ring)->ring_mtx) 490 #define ENA_RING_MTX_TRYLOCK(_ring) mtx_trylock(&(_ring)->ring_mtx) 491 #define ENA_RING_MTX_UNLOCK(_ring) mtx_unlock(&(_ring)->ring_mtx) 492 #define ENA_RING_MTX_ASSERT(_ring) \ 493 mtx_assert(&(_ring)->ring_mtx, MA_OWNED) 494 495 #define ENA_LOCK_INIT() \ 496 sx_init(&ena_global_lock, "ENA global lock") 497 #define ENA_LOCK_DESTROY() sx_destroy(&ena_global_lock) 498 #define ENA_LOCK_LOCK() sx_xlock(&ena_global_lock) 499 #define ENA_LOCK_UNLOCK() sx_unlock(&ena_global_lock) 500 #define ENA_LOCK_ASSERT() sx_assert(&ena_global_lock, SA_XLOCKED) 501 502 #define ENA_TIMER_INIT(_adapter) \ 503 callout_init(&(_adapter)->timer_service, true) 504 #define ENA_TIMER_DRAIN(_adapter) \ 505 callout_drain(&(_adapter)->timer_service) 506 #define ENA_TIMER_RESET(_adapter) \ 507 callout_reset_sbt(&(_adapter)->timer_service, SBT_1S, SBT_1S, \ 508 ena_timer_service, (void*)(_adapter), 0) 509 510 #define clamp_t(type, _x, min, max) min_t(type, max_t(type, _x, min), max) 511 #define clamp_val(val, lo, hi) clamp_t(__typeof(val), val, lo, hi) 512 513 extern struct sx ena_global_lock; 514 515 static inline int ena_mbuf_count(struct mbuf *mbuf) 516 { 517 int count = 1; 518 519 while ((mbuf = mbuf->m_next) != NULL) 520 ++count; 521 522 return count; 523 } 524 525 int ena_up(struct ena_adapter *adapter); 526 void ena_down(struct ena_adapter *adapter); 527 int ena_restore_device(struct ena_adapter *adapter); 528 void ena_destroy_device(struct ena_adapter *adapter, bool graceful); 529 int ena_refill_rx_bufs(struct ena_ring *rx_ring, uint32_t num); 530 int ena_update_buf_ring_size(struct ena_adapter *adapter, 531 uint32_t new_buf_ring_size); 532 int ena_update_queue_size(struct ena_adapter *adapter, uint32_t new_tx_size, 533 uint32_t new_rx_size); 534 int ena_update_io_queue_nb(struct ena_adapter *adapter, uint32_t new_num); 535 536 static inline void 537 ena_trigger_reset(struct ena_adapter *adapter, 538 enum ena_regs_reset_reason_types reset_reason) 539 { 540 if (likely(!ENA_FLAG_ISSET(ENA_FLAG_TRIGGER_RESET, adapter))) { 541 adapter->reset_reason = reset_reason; 542 ENA_FLAG_SET_ATOMIC(ENA_FLAG_TRIGGER_RESET, adapter); 543 } 544 } 545 546 #endif /* !(ENA_H) */ 547