2 * Copyright (c) 2014-2018, Matthew Macy <mmacy@mattmacy.io>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
8 * 1. Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
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.
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.
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
32 #include "opt_inet6.h"
34 #include "opt_sched.h"
36 #include <sys/param.h>
37 #include <sys/types.h>
39 #include <sys/eventhandler.h>
41 #include <sys/kernel.h>
44 #include <sys/mutex.h>
45 #include <sys/module.h>
51 #include <sys/socket.h>
52 #include <sys/sockio.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
55 #include <sys/taskqueue.h>
56 #include <sys/limits.h>
59 #include <net/if_var.h>
60 #include <net/if_types.h>
61 #include <net/if_media.h>
63 #include <net/ethernet.h>
64 #include <net/mp_ring.h>
67 #include <netinet/in.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/tcp_lro.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/if_ether.h>
72 #include <netinet/ip.h>
73 #include <netinet/ip6.h>
74 #include <netinet/tcp.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/netdump/netdump.h>
77 #include <netinet6/ip6_var.h>
79 #include <machine/bus.h>
80 #include <machine/in_cksum.h>
85 #include <dev/led/led.h>
86 #include <dev/pci/pcireg.h>
87 #include <dev/pci/pcivar.h>
88 #include <dev/pci/pci_private.h>
90 #include <net/iflib.h>
91 #include <net/iflib_private.h>
96 #include <dev/pci/pci_iov.h>
99 #include <sys/bitstring.h>
101 * enable accounting of every mbuf as it comes in to and goes out of
102 * iflib's software descriptor references
104 #define MEMORY_LOGGING 0
106 * Enable mbuf vectors for compressing long mbuf chains
111 * - Prefetching in tx cleaning should perhaps be a tunable. The distance ahead
112 * we prefetch needs to be determined by the time spent in m_free vis a vis
113 * the cost of a prefetch. This will of course vary based on the workload:
114 * - NFLX's m_free path is dominated by vm-based M_EXT manipulation which
115 * is quite expensive, thus suggesting very little prefetch.
116 * - small packet forwarding which is just returning a single mbuf to
117 * UMA will typically be very fast vis a vis the cost of a memory
124 * - private structures
125 * - iflib private utility functions
127 * - vlan registry and other exported functions
128 * - iflib public core functions
132 MALLOC_DEFINE(M_IFLIB, "iflib", "ifnet library");
135 typedef struct iflib_txq *iflib_txq_t;
137 typedef struct iflib_rxq *iflib_rxq_t;
139 typedef struct iflib_fl *iflib_fl_t;
143 static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid);
144 static void iflib_timer(void *arg);
146 typedef struct iflib_filter_info {
147 driver_filter_t *ifi_filter;
148 void *ifi_filter_arg;
149 struct grouptask *ifi_task;
151 } *iflib_filter_info_t;
156 * Pointer to hardware driver's softc
163 if_shared_ctx_t ifc_sctx;
164 struct if_softc_ctx ifc_softc_ctx;
166 struct sx ifc_ctx_sx;
167 struct mtx ifc_state_mtx;
169 iflib_txq_t ifc_txqs;
170 iflib_rxq_t ifc_rxqs;
171 uint32_t ifc_if_flags;
173 uint32_t ifc_max_fl_buf_size;
177 int ifc_watchdog_events;
178 struct cdev *ifc_led_dev;
179 struct resource *ifc_msix_mem;
181 struct if_irq ifc_legacy_irq;
182 struct grouptask ifc_admin_task;
183 struct grouptask ifc_vflr_task;
184 struct iflib_filter_info ifc_filter_info;
185 struct ifmedia ifc_media;
187 struct sysctl_oid *ifc_sysctl_node;
188 uint16_t ifc_sysctl_ntxqs;
189 uint16_t ifc_sysctl_nrxqs;
190 uint16_t ifc_sysctl_qs_eq_override;
191 uint16_t ifc_sysctl_rx_budget;
192 uint16_t ifc_sysctl_tx_abdicate;
194 qidx_t ifc_sysctl_ntxds[8];
195 qidx_t ifc_sysctl_nrxds[8];
196 struct if_txrx ifc_txrx;
197 #define isc_txd_encap ifc_txrx.ift_txd_encap
198 #define isc_txd_flush ifc_txrx.ift_txd_flush
199 #define isc_txd_credits_update ifc_txrx.ift_txd_credits_update
200 #define isc_rxd_available ifc_txrx.ift_rxd_available
201 #define isc_rxd_pkt_get ifc_txrx.ift_rxd_pkt_get
202 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
203 #define isc_rxd_flush ifc_txrx.ift_rxd_flush
204 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
205 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
206 #define isc_legacy_intr ifc_txrx.ift_legacy_intr
207 eventhandler_tag ifc_vlan_attach_event;
208 eventhandler_tag ifc_vlan_detach_event;
209 uint8_t ifc_mac[ETHER_ADDR_LEN];
210 char ifc_mtx_name[16];
215 iflib_get_softc(if_ctx_t ctx)
218 return (ctx->ifc_softc);
222 iflib_get_dev(if_ctx_t ctx)
225 return (ctx->ifc_dev);
229 iflib_get_ifp(if_ctx_t ctx)
232 return (ctx->ifc_ifp);
236 iflib_get_media(if_ctx_t ctx)
239 return (&ctx->ifc_media);
243 iflib_get_flags(if_ctx_t ctx)
245 return (ctx->ifc_flags);
249 iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN])
252 bcopy(mac, ctx->ifc_mac, ETHER_ADDR_LEN);
256 iflib_get_softc_ctx(if_ctx_t ctx)
259 return (&ctx->ifc_softc_ctx);
263 iflib_get_sctx(if_ctx_t ctx)
266 return (ctx->ifc_sctx);
269 #define IP_ALIGNED(m) ((((uintptr_t)(m)->m_data) & 0x3) == 0x2)
270 #define CACHE_PTR_INCREMENT (CACHE_LINE_SIZE/sizeof(void*))
271 #define CACHE_PTR_NEXT(ptr) ((void *)(((uintptr_t)(ptr)+CACHE_LINE_SIZE-1) & (CACHE_LINE_SIZE-1)))
273 #define LINK_ACTIVE(ctx) ((ctx)->ifc_link_state == LINK_STATE_UP)
274 #define CTX_IS_VF(ctx) ((ctx)->ifc_sctx->isc_flags & IFLIB_IS_VF)
276 typedef struct iflib_sw_rx_desc_array {
277 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */
278 struct mbuf **ifsd_m; /* pkthdr mbufs */
279 caddr_t *ifsd_cl; /* direct cluster pointer for rx */
280 bus_addr_t *ifsd_ba; /* bus addr of cluster for rx */
281 } iflib_rxsd_array_t;
283 typedef struct iflib_sw_tx_desc_array {
284 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */
285 bus_dmamap_t *ifsd_tso_map; /* bus_dma maps for TSO packet */
286 struct mbuf **ifsd_m; /* pkthdr mbufs */
290 /* magic number that should be high enough for any hardware */
291 #define IFLIB_MAX_TX_SEGS 128
292 #define IFLIB_RX_COPY_THRESH 128
293 #define IFLIB_MAX_RX_REFRESH 32
294 /* The minimum descriptors per second before we start coalescing */
295 #define IFLIB_MIN_DESC_SEC 16384
296 #define IFLIB_DEFAULT_TX_UPDATE_FREQ 16
297 #define IFLIB_QUEUE_IDLE 0
298 #define IFLIB_QUEUE_HUNG 1
299 #define IFLIB_QUEUE_WORKING 2
300 /* maximum number of txqs that can share an rx interrupt */
301 #define IFLIB_MAX_TX_SHARED_INTR 4
303 /* this should really scale with ring size - this is a fairly arbitrary value */
304 #define TX_BATCH_SIZE 32
306 #define IFLIB_RESTART_BUDGET 8
309 #define CSUM_OFFLOAD (CSUM_IP_TSO|CSUM_IP6_TSO|CSUM_IP| \
310 CSUM_IP_UDP|CSUM_IP_TCP|CSUM_IP_SCTP| \
311 CSUM_IP6_UDP|CSUM_IP6_TCP|CSUM_IP6_SCTP)
315 qidx_t ift_cidx_processed;
318 uint8_t ift_br_offset;
319 uint16_t ift_npending;
320 uint16_t ift_db_pending;
321 uint16_t ift_rs_pending;
323 uint8_t ift_txd_size[8];
324 uint64_t ift_processed;
325 uint64_t ift_cleaned;
326 uint64_t ift_cleaned_prev;
328 uint64_t ift_enqueued;
329 uint64_t ift_dequeued;
331 uint64_t ift_no_tx_dma_setup;
332 uint64_t ift_no_desc_avail;
333 uint64_t ift_mbuf_defrag_failed;
334 uint64_t ift_mbuf_defrag;
335 uint64_t ift_map_failed;
336 uint64_t ift_txd_encap_efbig;
337 uint64_t ift_pullups;
338 uint64_t ift_last_timer_tick;
341 struct mtx ift_db_mtx;
343 /* constant values */
345 struct ifmp_ring *ift_br;
346 struct grouptask ift_task;
349 struct callout ift_timer;
351 if_txsd_vec_t ift_sds;
354 uint8_t ift_update_freq;
355 struct iflib_filter_info ift_filter_info;
356 bus_dma_tag_t ift_buf_tag;
357 bus_dma_tag_t ift_tso_buf_tag;
358 iflib_dma_info_t ift_ifdi;
359 #define MTX_NAME_LEN 16
360 char ift_mtx_name[MTX_NAME_LEN];
361 char ift_db_mtx_name[MTX_NAME_LEN];
362 bus_dma_segment_t ift_segs[IFLIB_MAX_TX_SEGS] __aligned(CACHE_LINE_SIZE);
363 #ifdef IFLIB_DIAGNOSTICS
364 uint64_t ift_cpu_exec_count[256];
366 } __aligned(CACHE_LINE_SIZE);
373 uint8_t ifl_rxd_size;
375 uint64_t ifl_m_enqueued;
376 uint64_t ifl_m_dequeued;
377 uint64_t ifl_cl_enqueued;
378 uint64_t ifl_cl_dequeued;
382 bitstr_t *ifl_rx_bitmap;
386 uint16_t ifl_buf_size;
389 iflib_rxsd_array_t ifl_sds;
392 bus_dma_tag_t ifl_buf_tag;
393 iflib_dma_info_t ifl_ifdi;
394 uint64_t ifl_bus_addrs[IFLIB_MAX_RX_REFRESH] __aligned(CACHE_LINE_SIZE);
395 caddr_t ifl_vm_addrs[IFLIB_MAX_RX_REFRESH];
396 qidx_t ifl_rxd_idxs[IFLIB_MAX_RX_REFRESH];
397 } __aligned(CACHE_LINE_SIZE);
400 get_inuse(int size, qidx_t cidx, qidx_t pidx, uint8_t gen)
406 else if (pidx < cidx)
407 used = size - cidx + pidx;
408 else if (gen == 0 && pidx == cidx)
410 else if (gen == 1 && pidx == cidx)
418 #define TXQ_AVAIL(txq) (txq->ift_size - get_inuse(txq->ift_size, txq->ift_cidx, txq->ift_pidx, txq->ift_gen))
420 #define IDXDIFF(head, tail, wrap) \
421 ((head) >= (tail) ? (head) - (tail) : (wrap) - (tail) + (head))
424 /* If there is a separate completion queue -
425 * these are the cq cidx and pidx. Otherwise
432 uint8_t ifr_fl_offset;
438 uint8_t ifr_lro_enabled;
441 uint8_t ifr_txqid[IFLIB_MAX_TX_SHARED_INTR];
442 struct lro_ctrl ifr_lc;
443 struct grouptask ifr_task;
444 struct iflib_filter_info ifr_filter_info;
445 iflib_dma_info_t ifr_ifdi;
447 /* dynamically allocate if any drivers need a value substantially larger than this */
448 struct if_rxd_frag ifr_frags[IFLIB_MAX_RX_SEGS] __aligned(CACHE_LINE_SIZE);
449 #ifdef IFLIB_DIAGNOSTICS
450 uint64_t ifr_cpu_exec_count[256];
452 } __aligned(CACHE_LINE_SIZE);
454 typedef struct if_rxsd {
456 struct mbuf **ifsd_m;
461 /* multiple of word size */
463 #define PKT_INFO_SIZE 6
464 #define RXD_INFO_SIZE 5
465 #define PKT_TYPE uint64_t
467 #define PKT_INFO_SIZE 11
468 #define RXD_INFO_SIZE 8
469 #define PKT_TYPE uint32_t
471 #define PKT_LOOP_BOUND ((PKT_INFO_SIZE/3)*3)
472 #define RXD_LOOP_BOUND ((RXD_INFO_SIZE/4)*4)
474 typedef struct if_pkt_info_pad {
475 PKT_TYPE pkt_val[PKT_INFO_SIZE];
476 } *if_pkt_info_pad_t;
477 typedef struct if_rxd_info_pad {
478 PKT_TYPE rxd_val[RXD_INFO_SIZE];
479 } *if_rxd_info_pad_t;
481 CTASSERT(sizeof(struct if_pkt_info_pad) == sizeof(struct if_pkt_info));
482 CTASSERT(sizeof(struct if_rxd_info_pad) == sizeof(struct if_rxd_info));
486 pkt_info_zero(if_pkt_info_t pi)
488 if_pkt_info_pad_t pi_pad;
490 pi_pad = (if_pkt_info_pad_t)pi;
491 pi_pad->pkt_val[0] = 0; pi_pad->pkt_val[1] = 0; pi_pad->pkt_val[2] = 0;
492 pi_pad->pkt_val[3] = 0; pi_pad->pkt_val[4] = 0; pi_pad->pkt_val[5] = 0;
494 pi_pad->pkt_val[6] = 0; pi_pad->pkt_val[7] = 0; pi_pad->pkt_val[8] = 0;
495 pi_pad->pkt_val[9] = 0; pi_pad->pkt_val[10] = 0;
499 static device_method_t iflib_pseudo_methods[] = {
500 DEVMETHOD(device_attach, noop_attach),
501 DEVMETHOD(device_detach, iflib_pseudo_detach),
505 driver_t iflib_pseudodriver = {
506 "iflib_pseudo", iflib_pseudo_methods, sizeof(struct iflib_ctx),
510 rxd_info_zero(if_rxd_info_t ri)
512 if_rxd_info_pad_t ri_pad;
515 ri_pad = (if_rxd_info_pad_t)ri;
516 for (i = 0; i < RXD_LOOP_BOUND; i += 4) {
517 ri_pad->rxd_val[i] = 0;
518 ri_pad->rxd_val[i+1] = 0;
519 ri_pad->rxd_val[i+2] = 0;
520 ri_pad->rxd_val[i+3] = 0;
523 ri_pad->rxd_val[RXD_INFO_SIZE-1] = 0;
528 * Only allow a single packet to take up most 1/nth of the tx ring
530 #define MAX_SINGLE_PACKET_FRACTION 12
531 #define IF_BAD_DMA (bus_addr_t)-1
533 #define CTX_ACTIVE(ctx) ((if_getdrvflags((ctx)->ifc_ifp) & IFF_DRV_RUNNING))
535 #define CTX_LOCK_INIT(_sc) sx_init(&(_sc)->ifc_ctx_sx, "iflib ctx lock")
536 #define CTX_LOCK(ctx) sx_xlock(&(ctx)->ifc_ctx_sx)
537 #define CTX_UNLOCK(ctx) sx_xunlock(&(ctx)->ifc_ctx_sx)
538 #define CTX_LOCK_DESTROY(ctx) sx_destroy(&(ctx)->ifc_ctx_sx)
541 #define STATE_LOCK_INIT(_sc, _name) mtx_init(&(_sc)->ifc_state_mtx, _name, "iflib state lock", MTX_DEF)
542 #define STATE_LOCK(ctx) mtx_lock(&(ctx)->ifc_state_mtx)
543 #define STATE_UNLOCK(ctx) mtx_unlock(&(ctx)->ifc_state_mtx)
544 #define STATE_LOCK_DESTROY(ctx) mtx_destroy(&(ctx)->ifc_state_mtx)
548 #define CALLOUT_LOCK(txq) mtx_lock(&txq->ift_mtx)
549 #define CALLOUT_UNLOCK(txq) mtx_unlock(&txq->ift_mtx)
552 iflib_set_detach(if_ctx_t ctx)
555 ctx->ifc_flags |= IFC_IN_DETACH;
559 /* Our boot-time initialization hook */
560 static int iflib_module_event_handler(module_t, int, void *);
562 static moduledata_t iflib_moduledata = {
564 iflib_module_event_handler,
568 DECLARE_MODULE(iflib, iflib_moduledata, SI_SUB_INIT_IF, SI_ORDER_ANY);
569 MODULE_VERSION(iflib, 1);
571 MODULE_DEPEND(iflib, pci, 1, 1, 1);
572 MODULE_DEPEND(iflib, ether, 1, 1, 1);
574 TASKQGROUP_DEFINE(if_io_tqg, mp_ncpus, 1);
575 TASKQGROUP_DEFINE(if_config_tqg, 1, 1);
577 #ifndef IFLIB_DEBUG_COUNTERS
579 #define IFLIB_DEBUG_COUNTERS 1
581 #define IFLIB_DEBUG_COUNTERS 0
582 #endif /* !INVARIANTS */
585 static SYSCTL_NODE(_net, OID_AUTO, iflib, CTLFLAG_RD, 0,
586 "iflib driver parameters");
589 * XXX need to ensure that this can't accidentally cause the head to be moved backwards
591 static int iflib_min_tx_latency = 0;
592 SYSCTL_INT(_net_iflib, OID_AUTO, min_tx_latency, CTLFLAG_RW,
593 &iflib_min_tx_latency, 0, "minimize transmit latency at the possible expense of throughput");
594 static int iflib_no_tx_batch = 0;
595 SYSCTL_INT(_net_iflib, OID_AUTO, no_tx_batch, CTLFLAG_RW,
596 &iflib_no_tx_batch, 0, "minimize transmit latency at the possible expense of throughput");
599 #if IFLIB_DEBUG_COUNTERS
601 static int iflib_tx_seen;
602 static int iflib_tx_sent;
603 static int iflib_tx_encap;
604 static int iflib_rx_allocs;
605 static int iflib_fl_refills;
606 static int iflib_fl_refills_large;
607 static int iflib_tx_frees;
609 SYSCTL_INT(_net_iflib, OID_AUTO, tx_seen, CTLFLAG_RD,
610 &iflib_tx_seen, 0, "# tx mbufs seen");
611 SYSCTL_INT(_net_iflib, OID_AUTO, tx_sent, CTLFLAG_RD,
612 &iflib_tx_sent, 0, "# tx mbufs sent");
613 SYSCTL_INT(_net_iflib, OID_AUTO, tx_encap, CTLFLAG_RD,
614 &iflib_tx_encap, 0, "# tx mbufs encapped");
615 SYSCTL_INT(_net_iflib, OID_AUTO, tx_frees, CTLFLAG_RD,
616 &iflib_tx_frees, 0, "# tx frees");
617 SYSCTL_INT(_net_iflib, OID_AUTO, rx_allocs, CTLFLAG_RD,
618 &iflib_rx_allocs, 0, "# rx allocations");
619 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills, CTLFLAG_RD,
620 &iflib_fl_refills, 0, "# refills");
621 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills_large, CTLFLAG_RD,
622 &iflib_fl_refills_large, 0, "# large refills");
625 static int iflib_txq_drain_flushing;
626 static int iflib_txq_drain_oactive;
627 static int iflib_txq_drain_notready;
629 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_flushing, CTLFLAG_RD,
630 &iflib_txq_drain_flushing, 0, "# drain flushes");
631 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_oactive, CTLFLAG_RD,
632 &iflib_txq_drain_oactive, 0, "# drain oactives");
633 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_notready, CTLFLAG_RD,
634 &iflib_txq_drain_notready, 0, "# drain notready");
637 static int iflib_encap_load_mbuf_fail;
638 static int iflib_encap_pad_mbuf_fail;
639 static int iflib_encap_txq_avail_fail;
640 static int iflib_encap_txd_encap_fail;
642 SYSCTL_INT(_net_iflib, OID_AUTO, encap_load_mbuf_fail, CTLFLAG_RD,
643 &iflib_encap_load_mbuf_fail, 0, "# busdma load failures");
644 SYSCTL_INT(_net_iflib, OID_AUTO, encap_pad_mbuf_fail, CTLFLAG_RD,
645 &iflib_encap_pad_mbuf_fail, 0, "# runt frame pad failures");
646 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txq_avail_fail, CTLFLAG_RD,
647 &iflib_encap_txq_avail_fail, 0, "# txq avail failures");
648 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txd_encap_fail, CTLFLAG_RD,
649 &iflib_encap_txd_encap_fail, 0, "# driver encap failures");
651 static int iflib_task_fn_rxs;
652 static int iflib_rx_intr_enables;
653 static int iflib_fast_intrs;
654 static int iflib_rx_unavail;
655 static int iflib_rx_ctx_inactive;
656 static int iflib_rx_if_input;
657 static int iflib_rx_mbuf_null;
658 static int iflib_rxd_flush;
660 static int iflib_verbose_debug;
662 SYSCTL_INT(_net_iflib, OID_AUTO, task_fn_rx, CTLFLAG_RD,
663 &iflib_task_fn_rxs, 0, "# task_fn_rx calls");
664 SYSCTL_INT(_net_iflib, OID_AUTO, rx_intr_enables, CTLFLAG_RD,
665 &iflib_rx_intr_enables, 0, "# rx intr enables");
666 SYSCTL_INT(_net_iflib, OID_AUTO, fast_intrs, CTLFLAG_RD,
667 &iflib_fast_intrs, 0, "# fast_intr calls");
668 SYSCTL_INT(_net_iflib, OID_AUTO, rx_unavail, CTLFLAG_RD,
669 &iflib_rx_unavail, 0, "# times rxeof called with no available data");
670 SYSCTL_INT(_net_iflib, OID_AUTO, rx_ctx_inactive, CTLFLAG_RD,
671 &iflib_rx_ctx_inactive, 0, "# times rxeof called with inactive context");
672 SYSCTL_INT(_net_iflib, OID_AUTO, rx_if_input, CTLFLAG_RD,
673 &iflib_rx_if_input, 0, "# times rxeof called if_input");
674 SYSCTL_INT(_net_iflib, OID_AUTO, rx_mbuf_null, CTLFLAG_RD,
675 &iflib_rx_mbuf_null, 0, "# times rxeof got null mbuf");
676 SYSCTL_INT(_net_iflib, OID_AUTO, rxd_flush, CTLFLAG_RD,
677 &iflib_rxd_flush, 0, "# times rxd_flush called");
678 SYSCTL_INT(_net_iflib, OID_AUTO, verbose_debug, CTLFLAG_RW,
679 &iflib_verbose_debug, 0, "enable verbose debugging");
681 #define DBG_COUNTER_INC(name) atomic_add_int(&(iflib_ ## name), 1)
683 iflib_debug_reset(void)
685 iflib_tx_seen = iflib_tx_sent = iflib_tx_encap = iflib_rx_allocs =
686 iflib_fl_refills = iflib_fl_refills_large = iflib_tx_frees =
687 iflib_txq_drain_flushing = iflib_txq_drain_oactive =
688 iflib_txq_drain_notready =
689 iflib_encap_load_mbuf_fail = iflib_encap_pad_mbuf_fail =
690 iflib_encap_txq_avail_fail = iflib_encap_txd_encap_fail =
691 iflib_task_fn_rxs = iflib_rx_intr_enables = iflib_fast_intrs =
693 iflib_rx_ctx_inactive = iflib_rx_if_input =
694 iflib_rx_mbuf_null = iflib_rxd_flush = 0;
698 #define DBG_COUNTER_INC(name)
699 static void iflib_debug_reset(void) {}
702 #define IFLIB_DEBUG 0
704 static void iflib_tx_structures_free(if_ctx_t ctx);
705 static void iflib_rx_structures_free(if_ctx_t ctx);
706 static int iflib_queues_alloc(if_ctx_t ctx);
707 static int iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq);
708 static int iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget);
709 static int iflib_qset_structures_setup(if_ctx_t ctx);
710 static int iflib_msix_init(if_ctx_t ctx);
711 static int iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filterarg, int *rid, const char *str);
712 static void iflib_txq_check_drain(iflib_txq_t txq, int budget);
713 static uint32_t iflib_txq_can_drain(struct ifmp_ring *);
715 static void iflib_altq_if_start(if_t ifp);
716 static int iflib_altq_if_transmit(if_t ifp, struct mbuf *m);
718 static int iflib_register(if_ctx_t);
719 static void iflib_init_locked(if_ctx_t ctx);
720 static void iflib_add_device_sysctl_pre(if_ctx_t ctx);
721 static void iflib_add_device_sysctl_post(if_ctx_t ctx);
722 static void iflib_ifmp_purge(iflib_txq_t txq);
723 static void _iflib_pre_assert(if_softc_ctx_t scctx);
724 static void iflib_if_init_locked(if_ctx_t ctx);
725 static void iflib_free_intr_mem(if_ctx_t ctx);
726 #ifndef __NO_STRICT_ALIGNMENT
727 static struct mbuf * iflib_fixup_rx(struct mbuf *m);
730 NETDUMP_DEFINE(iflib);
733 #include <sys/selinfo.h>
734 #include <net/netmap.h>
735 #include <dev/netmap/netmap_kern.h>
737 MODULE_DEPEND(iflib, netmap, 1, 1, 1);
739 static int netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init);
742 * device-specific sysctl variables:
744 * iflib_crcstrip: 0: keep CRC in rx frames (default), 1: strip it.
745 * During regular operations the CRC is stripped, but on some
746 * hardware reception of frames not multiple of 64 is slower,
747 * so using crcstrip=0 helps in benchmarks.
749 * iflib_rx_miss, iflib_rx_miss_bufs:
750 * count packets that might be missed due to lost interrupts.
752 SYSCTL_DECL(_dev_netmap);
754 * The xl driver by default strips CRCs and we do not override it.
757 int iflib_crcstrip = 1;
758 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_crcstrip,
759 CTLFLAG_RW, &iflib_crcstrip, 1, "strip CRC on rx frames");
761 int iflib_rx_miss, iflib_rx_miss_bufs;
762 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss,
763 CTLFLAG_RW, &iflib_rx_miss, 0, "potentially missed rx intr");
764 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss_bufs,
765 CTLFLAG_RW, &iflib_rx_miss_bufs, 0, "potentially missed rx intr bufs");
768 * Register/unregister. We are already under netmap lock.
769 * Only called on the first register or the last unregister.
772 iflib_netmap_register(struct netmap_adapter *na, int onoff)
774 struct ifnet *ifp = na->ifp;
775 if_ctx_t ctx = ifp->if_softc;
779 IFDI_INTR_DISABLE(ctx);
781 /* Tell the stack that the interface is no longer active */
782 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
785 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip);
787 /* enable or disable flags and callbacks in na and ifp */
789 nm_set_native_flags(na);
791 nm_clear_native_flags(na);
794 iflib_init_locked(ctx);
795 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); // XXX why twice ?
796 status = ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1;
798 nm_clear_native_flags(na);
804 netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init)
806 struct netmap_adapter *na = kring->na;
807 u_int const lim = kring->nkr_num_slots - 1;
808 u_int head = kring->rhead;
809 struct netmap_ring *ring = kring->ring;
811 struct if_rxd_update iru;
812 if_ctx_t ctx = rxq->ifr_ctx;
813 iflib_fl_t fl = &rxq->ifr_fl[0];
814 uint32_t refill_pidx, nic_i;
815 #if IFLIB_DEBUG_COUNTERS
819 if (nm_i == head && __predict_true(!init))
821 iru_init(&iru, rxq, 0 /* flid */);
822 map = fl->ifl_sds.ifsd_map;
823 refill_pidx = netmap_idx_k2n(kring, nm_i);
825 * IMPORTANT: we must leave one free slot in the ring,
826 * so move head back by one unit
828 head = nm_prev(head, lim);
830 DBG_COUNTER_INC(fl_refills);
831 while (nm_i != head) {
832 #if IFLIB_DEBUG_COUNTERS
834 DBG_COUNTER_INC(fl_refills_large);
836 for (int tmp_pidx = 0; tmp_pidx < IFLIB_MAX_RX_REFRESH && nm_i != head; tmp_pidx++) {
837 struct netmap_slot *slot = &ring->slot[nm_i];
838 void *addr = PNMB(na, slot, &fl->ifl_bus_addrs[tmp_pidx]);
839 uint32_t nic_i_dma = refill_pidx;
840 nic_i = netmap_idx_k2n(kring, nm_i);
842 MPASS(tmp_pidx < IFLIB_MAX_RX_REFRESH);
844 if (addr == NETMAP_BUF_BASE(na)) /* bad buf */
845 return netmap_ring_reinit(kring);
847 fl->ifl_vm_addrs[tmp_pidx] = addr;
848 if (__predict_false(init)) {
849 netmap_load_map(na, fl->ifl_buf_tag,
851 } else if (slot->flags & NS_BUF_CHANGED) {
852 /* buffer has changed, reload map */
853 netmap_reload_map(na, fl->ifl_buf_tag,
856 slot->flags &= ~NS_BUF_CHANGED;
858 nm_i = nm_next(nm_i, lim);
859 fl->ifl_rxd_idxs[tmp_pidx] = nic_i = nm_next(nic_i, lim);
860 if (nm_i != head && tmp_pidx < IFLIB_MAX_RX_REFRESH-1)
863 iru.iru_pidx = refill_pidx;
864 iru.iru_count = tmp_pidx+1;
865 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
867 for (int n = 0; n < iru.iru_count; n++) {
868 bus_dmamap_sync(fl->ifl_buf_tag, map[nic_i_dma],
869 BUS_DMASYNC_PREREAD);
870 /* XXX - change this to not use the netmap func*/
871 nic_i_dma = nm_next(nic_i_dma, lim);
875 kring->nr_hwcur = head;
877 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
878 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
879 if (__predict_true(nic_i != UINT_MAX)) {
880 ctx->isc_rxd_flush(ctx->ifc_softc, rxq->ifr_id, fl->ifl_id, nic_i);
881 DBG_COUNTER_INC(rxd_flush);
887 * Reconcile kernel and user view of the transmit ring.
889 * All information is in the kring.
890 * Userspace wants to send packets up to the one before kring->rhead,
891 * kernel knows kring->nr_hwcur is the first unsent packet.
893 * Here we push packets out (as many as possible), and possibly
894 * reclaim buffers from previously completed transmission.
896 * The caller (netmap) guarantees that there is only one instance
897 * running at any time. Any interference with other driver
898 * methods should be handled by the individual drivers.
901 iflib_netmap_txsync(struct netmap_kring *kring, int flags)
903 struct netmap_adapter *na = kring->na;
904 struct ifnet *ifp = na->ifp;
905 struct netmap_ring *ring = kring->ring;
906 u_int nm_i; /* index into the netmap kring */
907 u_int nic_i; /* index into the NIC ring */
909 u_int const lim = kring->nkr_num_slots - 1;
910 u_int const head = kring->rhead;
911 struct if_pkt_info pi;
914 * interrupts on every tx packet are expensive so request
915 * them every half ring, or where NS_REPORT is set
917 u_int report_frequency = kring->nkr_num_slots >> 1;
918 /* device-specific */
919 if_ctx_t ctx = ifp->if_softc;
920 iflib_txq_t txq = &ctx->ifc_txqs[kring->ring_id];
922 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
923 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
926 * First part: process new packets to send.
927 * nm_i is the current index in the netmap kring,
928 * nic_i is the corresponding index in the NIC ring.
930 * If we have packets to send (nm_i != head)
931 * iterate over the netmap ring, fetch length and update
932 * the corresponding slot in the NIC ring. Some drivers also
933 * need to update the buffer's physical address in the NIC slot
934 * even NS_BUF_CHANGED is not set (PNMB computes the addresses).
936 * The netmap_reload_map() calls is especially expensive,
937 * even when (as in this case) the tag is 0, so do only
938 * when the buffer has actually changed.
940 * If possible do not set the report/intr bit on all slots,
941 * but only a few times per ring or when NS_REPORT is set.
943 * Finally, on 10G and faster drivers, it might be useful
944 * to prefetch the next slot and txr entry.
947 nm_i = kring->nr_hwcur;
948 if (nm_i != head) { /* we have new packets to send */
950 pi.ipi_segs = txq->ift_segs;
951 pi.ipi_qsidx = kring->ring_id;
952 nic_i = netmap_idx_k2n(kring, nm_i);
954 __builtin_prefetch(&ring->slot[nm_i]);
955 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i]);
956 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i]);
958 for (n = 0; nm_i != head; n++) {
959 struct netmap_slot *slot = &ring->slot[nm_i];
960 u_int len = slot->len;
962 void *addr = PNMB(na, slot, &paddr);
963 int flags = (slot->flags & NS_REPORT ||
964 nic_i == 0 || nic_i == report_frequency) ?
967 /* device-specific */
969 pi.ipi_segs[0].ds_addr = paddr;
970 pi.ipi_segs[0].ds_len = len;
974 pi.ipi_flags = flags;
976 /* Fill the slot in the NIC ring. */
977 ctx->isc_txd_encap(ctx->ifc_softc, &pi);
978 DBG_COUNTER_INC(tx_encap);
980 /* prefetch for next round */
981 __builtin_prefetch(&ring->slot[nm_i + 1]);
982 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i + 1]);
983 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i + 1]);
985 NM_CHECK_ADDR_LEN(na, addr, len);
987 if (slot->flags & NS_BUF_CHANGED) {
988 /* buffer has changed, reload map */
989 netmap_reload_map(na, txq->ift_buf_tag,
990 txq->ift_sds.ifsd_map[nic_i], addr);
992 /* make sure changes to the buffer are synced */
993 bus_dmamap_sync(txq->ift_buf_tag,
994 txq->ift_sds.ifsd_map[nic_i],
995 BUS_DMASYNC_PREWRITE);
997 slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED);
998 nm_i = nm_next(nm_i, lim);
999 nic_i = nm_next(nic_i, lim);
1001 kring->nr_hwcur = nm_i;
1003 /* synchronize the NIC ring */
1004 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
1005 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1007 /* (re)start the tx unit up to slot nic_i (excluded) */
1008 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, nic_i);
1012 * Second part: reclaim buffers for completed transmissions.
1014 * If there are unclaimed buffers, attempt to reclaim them.
1015 * If none are reclaimed, and TX IRQs are not in use, do an initial
1016 * minimal delay, then trigger the tx handler which will spin in the
1019 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) {
1020 if (iflib_tx_credits_update(ctx, txq)) {
1021 /* some tx completed, increment avail */
1022 nic_i = txq->ift_cidx_processed;
1023 kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), lim);
1026 if (!(ctx->ifc_flags & IFC_NETMAP_TX_IRQ))
1027 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) {
1028 callout_reset_on(&txq->ift_timer, hz < 2000 ? 1 : hz / 1000,
1029 iflib_timer, txq, txq->ift_timer.c_cpu);
1035 * Reconcile kernel and user view of the receive ring.
1036 * Same as for the txsync, this routine must be efficient.
1037 * The caller guarantees a single invocations, but races against
1038 * the rest of the driver should be handled here.
1040 * On call, kring->rhead is the first packet that userspace wants
1041 * to keep, and kring->rcur is the wakeup point.
1042 * The kernel has previously reported packets up to kring->rtail.
1044 * If (flags & NAF_FORCE_READ) also check for incoming packets irrespective
1045 * of whether or not we received an interrupt.
1048 iflib_netmap_rxsync(struct netmap_kring *kring, int flags)
1050 struct netmap_adapter *na = kring->na;
1051 struct netmap_ring *ring = kring->ring;
1053 uint32_t nm_i; /* index into the netmap ring */
1054 uint32_t nic_i; /* index into the NIC ring */
1056 u_int const lim = kring->nkr_num_slots - 1;
1057 u_int const head = kring->rhead;
1058 int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR;
1059 struct if_rxd_info ri;
1061 struct ifnet *ifp = na->ifp;
1062 if_ctx_t ctx = ifp->if_softc;
1063 iflib_rxq_t rxq = &ctx->ifc_rxqs[kring->ring_id];
1065 return netmap_ring_reinit(kring);
1068 * XXX netmap_fl_refill() only ever (re)fills free list 0 so far.
1071 for (i = 0, fl = rxq->ifr_fl; i < rxq->ifr_nfl; i++, fl++) {
1072 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
1073 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
1077 * First part: import newly received packets.
1079 * nm_i is the index of the next free slot in the netmap ring,
1080 * nic_i is the index of the next received packet in the NIC ring,
1081 * and they may differ in case if_init() has been called while
1082 * in netmap mode. For the receive ring we have
1084 * nic_i = rxr->next_check;
1085 * nm_i = kring->nr_hwtail (previous)
1087 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size
1089 * rxr->next_check is set to 0 on a ring reinit
1091 if (netmap_no_pendintr || force_update) {
1092 int crclen = iflib_crcstrip ? 0 : 4;
1095 for (i = 0; i < rxq->ifr_nfl; i++) {
1096 fl = &rxq->ifr_fl[i];
1097 nic_i = fl->ifl_cidx;
1098 nm_i = netmap_idx_n2k(kring, nic_i);
1099 avail = ctx->isc_rxd_available(ctx->ifc_softc,
1100 rxq->ifr_id, nic_i, USHRT_MAX);
1101 for (n = 0; avail > 0; n++, avail--) {
1103 ri.iri_frags = rxq->ifr_frags;
1104 ri.iri_qsidx = kring->ring_id;
1105 ri.iri_ifp = ctx->ifc_ifp;
1106 ri.iri_cidx = nic_i;
1108 error = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri);
1109 ring->slot[nm_i].len = error ? 0 : ri.iri_len - crclen;
1110 ring->slot[nm_i].flags = 0;
1111 bus_dmamap_sync(fl->ifl_buf_tag,
1112 fl->ifl_sds.ifsd_map[nic_i], BUS_DMASYNC_POSTREAD);
1113 nm_i = nm_next(nm_i, lim);
1114 nic_i = nm_next(nic_i, lim);
1116 if (n) { /* update the state variables */
1117 if (netmap_no_pendintr && !force_update) {
1120 iflib_rx_miss_bufs += n;
1122 fl->ifl_cidx = nic_i;
1123 kring->nr_hwtail = nm_i;
1125 kring->nr_kflags &= ~NKR_PENDINTR;
1129 * Second part: skip past packets that userspace has released.
1130 * (kring->nr_hwcur to head excluded),
1131 * and make the buffers available for reception.
1132 * As usual nm_i is the index in the netmap ring,
1133 * nic_i is the index in the NIC ring, and
1134 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size
1136 /* XXX not sure how this will work with multiple free lists */
1137 nm_i = kring->nr_hwcur;
1139 return (netmap_fl_refill(rxq, kring, nm_i, false));
1143 iflib_netmap_intr(struct netmap_adapter *na, int onoff)
1145 struct ifnet *ifp = na->ifp;
1146 if_ctx_t ctx = ifp->if_softc;
1150 IFDI_INTR_ENABLE(ctx);
1152 IFDI_INTR_DISABLE(ctx);
1159 iflib_netmap_attach(if_ctx_t ctx)
1161 struct netmap_adapter na;
1162 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1164 bzero(&na, sizeof(na));
1166 na.ifp = ctx->ifc_ifp;
1167 na.na_flags = NAF_BDG_MAYSLEEP;
1168 MPASS(ctx->ifc_softc_ctx.isc_ntxqsets);
1169 MPASS(ctx->ifc_softc_ctx.isc_nrxqsets);
1171 na.num_tx_desc = scctx->isc_ntxd[0];
1172 na.num_rx_desc = scctx->isc_nrxd[0];
1173 na.nm_txsync = iflib_netmap_txsync;
1174 na.nm_rxsync = iflib_netmap_rxsync;
1175 na.nm_register = iflib_netmap_register;
1176 na.nm_intr = iflib_netmap_intr;
1177 na.num_tx_rings = ctx->ifc_softc_ctx.isc_ntxqsets;
1178 na.num_rx_rings = ctx->ifc_softc_ctx.isc_nrxqsets;
1179 return (netmap_attach(&na));
1183 iflib_netmap_txq_init(if_ctx_t ctx, iflib_txq_t txq)
1185 struct netmap_adapter *na = NA(ctx->ifc_ifp);
1186 struct netmap_slot *slot;
1188 slot = netmap_reset(na, NR_TX, txq->ift_id, 0);
1191 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxd[0]; i++) {
1194 * In netmap mode, set the map for the packet buffer.
1195 * NOTE: Some drivers (not this one) also need to set
1196 * the physical buffer address in the NIC ring.
1197 * netmap_idx_n2k() maps a nic index, i, into the corresponding
1198 * netmap slot index, si
1200 int si = netmap_idx_n2k(na->tx_rings[txq->ift_id], i);
1201 netmap_load_map(na, txq->ift_buf_tag, txq->ift_sds.ifsd_map[i],
1202 NMB(na, slot + si));
1207 iflib_netmap_rxq_init(if_ctx_t ctx, iflib_rxq_t rxq)
1209 struct netmap_adapter *na = NA(ctx->ifc_ifp);
1210 struct netmap_kring *kring = na->rx_rings[rxq->ifr_id];
1211 struct netmap_slot *slot;
1214 slot = netmap_reset(na, NR_RX, rxq->ifr_id, 0);
1217 nm_i = netmap_idx_n2k(kring, 0);
1218 netmap_fl_refill(rxq, kring, nm_i, true);
1222 iflib_netmap_timer_adjust(if_ctx_t ctx, iflib_txq_t txq, uint32_t *reset_on)
1224 struct netmap_kring *kring;
1227 txqid = txq->ift_id;
1228 kring = NA(ctx->ifc_ifp)->tx_rings[txqid];
1230 if (kring->nr_hwcur != nm_next(kring->nr_hwtail, kring->nkr_num_slots - 1)) {
1231 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
1232 BUS_DMASYNC_POSTREAD);
1233 if (ctx->isc_txd_credits_update(ctx->ifc_softc, txqid, false))
1234 netmap_tx_irq(ctx->ifc_ifp, txqid);
1235 if (!(ctx->ifc_flags & IFC_NETMAP_TX_IRQ)) {
1239 *reset_on = hz / 1000;
1244 #define iflib_netmap_detach(ifp) netmap_detach(ifp)
1247 #define iflib_netmap_txq_init(ctx, txq)
1248 #define iflib_netmap_rxq_init(ctx, rxq)
1249 #define iflib_netmap_detach(ifp)
1251 #define iflib_netmap_attach(ctx) (0)
1252 #define netmap_rx_irq(ifp, qid, budget) (0)
1253 #define netmap_tx_irq(ifp, qid) do {} while (0)
1254 #define iflib_netmap_timer_adjust(ctx, txq, reset_on)
1258 #if defined(__i386__) || defined(__amd64__)
1259 static __inline void
1262 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x));
1264 static __inline void
1265 prefetch2cachelines(void *x)
1267 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x));
1268 #if (CACHE_LINE_SIZE < 128)
1269 __asm volatile("prefetcht0 %0" :: "m" (*(((unsigned long *)x)+CACHE_LINE_SIZE/(sizeof(unsigned long)))));
1274 #define prefetch2cachelines(x)
1278 iflib_gen_mac(if_ctx_t ctx)
1282 char uuid[HOSTUUIDLEN+1];
1283 char buf[HOSTUUIDLEN+16];
1285 unsigned char digest[16];
1289 uuid[HOSTUUIDLEN] = 0;
1290 bcopy(td->td_ucred->cr_prison->pr_hostuuid, uuid, HOSTUUIDLEN);
1291 snprintf(buf, HOSTUUIDLEN+16, "%s-%s", uuid, device_get_nameunit(ctx->ifc_dev));
1293 * Generate a pseudo-random, deterministic MAC
1294 * address based on the UUID and unit number.
1295 * The FreeBSD Foundation OUI of 58-9C-FC is used.
1298 MD5Update(&mdctx, buf, strlen(buf));
1299 MD5Final(digest, &mdctx);
1310 iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid)
1314 fl = &rxq->ifr_fl[flid];
1315 iru->iru_paddrs = fl->ifl_bus_addrs;
1316 iru->iru_vaddrs = &fl->ifl_vm_addrs[0];
1317 iru->iru_idxs = fl->ifl_rxd_idxs;
1318 iru->iru_qsidx = rxq->ifr_id;
1319 iru->iru_buf_size = fl->ifl_buf_size;
1320 iru->iru_flidx = fl->ifl_id;
1324 _iflib_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int err)
1328 *(bus_addr_t *) arg = segs[0].ds_addr;
1332 iflib_dma_alloc_align(if_ctx_t ctx, int size, int align, iflib_dma_info_t dma, int mapflags)
1335 device_t dev = ctx->ifc_dev;
1337 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
1338 align, 0, /* alignment, bounds */
1339 BUS_SPACE_MAXADDR, /* lowaddr */
1340 BUS_SPACE_MAXADDR, /* highaddr */
1341 NULL, NULL, /* filter, filterarg */
1344 size, /* maxsegsize */
1345 BUS_DMA_ALLOCNOW, /* flags */
1346 NULL, /* lockfunc */
1351 "%s: bus_dma_tag_create failed: %d\n",
1356 err = bus_dmamem_alloc(dma->idi_tag, (void**) &dma->idi_vaddr,
1357 BUS_DMA_NOWAIT | BUS_DMA_COHERENT | BUS_DMA_ZERO, &dma->idi_map);
1360 "%s: bus_dmamem_alloc(%ju) failed: %d\n",
1361 __func__, (uintmax_t)size, err);
1365 dma->idi_paddr = IF_BAD_DMA;
1366 err = bus_dmamap_load(dma->idi_tag, dma->idi_map, dma->idi_vaddr,
1367 size, _iflib_dmamap_cb, &dma->idi_paddr, mapflags | BUS_DMA_NOWAIT);
1368 if (err || dma->idi_paddr == IF_BAD_DMA) {
1370 "%s: bus_dmamap_load failed: %d\n",
1375 dma->idi_size = size;
1379 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map);
1381 bus_dma_tag_destroy(dma->idi_tag);
1383 dma->idi_tag = NULL;
1389 iflib_dma_alloc(if_ctx_t ctx, int size, iflib_dma_info_t dma, int mapflags)
1391 if_shared_ctx_t sctx = ctx->ifc_sctx;
1393 KASSERT(sctx->isc_q_align != 0, ("alignment value not initialized"));
1395 return (iflib_dma_alloc_align(ctx, size, sctx->isc_q_align, dma, mapflags));
1399 iflib_dma_alloc_multi(if_ctx_t ctx, int *sizes, iflib_dma_info_t *dmalist, int mapflags, int count)
1402 iflib_dma_info_t *dmaiter;
1405 for (i = 0; i < count; i++, dmaiter++) {
1406 if ((err = iflib_dma_alloc(ctx, sizes[i], *dmaiter, mapflags)) != 0)
1410 iflib_dma_free_multi(dmalist, i);
1415 iflib_dma_free(iflib_dma_info_t dma)
1417 if (dma->idi_tag == NULL)
1419 if (dma->idi_paddr != IF_BAD_DMA) {
1420 bus_dmamap_sync(dma->idi_tag, dma->idi_map,
1421 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
1422 bus_dmamap_unload(dma->idi_tag, dma->idi_map);
1423 dma->idi_paddr = IF_BAD_DMA;
1425 if (dma->idi_vaddr != NULL) {
1426 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map);
1427 dma->idi_vaddr = NULL;
1429 bus_dma_tag_destroy(dma->idi_tag);
1430 dma->idi_tag = NULL;
1434 iflib_dma_free_multi(iflib_dma_info_t *dmalist, int count)
1437 iflib_dma_info_t *dmaiter = dmalist;
1439 for (i = 0; i < count; i++, dmaiter++)
1440 iflib_dma_free(*dmaiter);
1443 #ifdef EARLY_AP_STARTUP
1444 static const int iflib_started = 1;
1447 * We used to abuse the smp_started flag to decide if the queues have been
1448 * fully initialized (by late taskqgroup_adjust() calls in a SYSINIT()).
1449 * That gave bad races, since the SYSINIT() runs strictly after smp_started
1450 * is set. Run a SYSINIT() strictly after that to just set a usable
1454 static int iflib_started;
1457 iflib_record_started(void *arg)
1462 SYSINIT(iflib_record_started, SI_SUB_SMP + 1, SI_ORDER_FIRST,
1463 iflib_record_started, NULL);
1467 iflib_fast_intr(void *arg)
1469 iflib_filter_info_t info = arg;
1470 struct grouptask *gtask = info->ifi_task;
1472 return (FILTER_HANDLED);
1474 DBG_COUNTER_INC(fast_intrs);
1475 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED)
1476 return (FILTER_HANDLED);
1478 GROUPTASK_ENQUEUE(gtask);
1479 return (FILTER_HANDLED);
1483 iflib_fast_intr_rxtx(void *arg)
1485 iflib_filter_info_t info = arg;
1486 struct grouptask *gtask = info->ifi_task;
1488 iflib_rxq_t rxq = (iflib_rxq_t)info->ifi_ctx;
1495 return (FILTER_HANDLED);
1497 DBG_COUNTER_INC(fast_intrs);
1498 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED)
1499 return (FILTER_HANDLED);
1502 sc = ctx->ifc_softc;
1503 MPASS(rxq->ifr_ntxqirq);
1504 for (i = 0; i < rxq->ifr_ntxqirq; i++) {
1505 txqid = rxq->ifr_txqid[i];
1506 txq = &ctx->ifc_txqs[txqid];
1507 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
1508 BUS_DMASYNC_POSTREAD);
1509 if (!ctx->isc_txd_credits_update(sc, txqid, false)) {
1510 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txqid);
1513 GROUPTASK_ENQUEUE(&txq->ift_task);
1515 if (ctx->ifc_sctx->isc_flags & IFLIB_HAS_RXCQ)
1516 cidx = rxq->ifr_cq_cidx;
1518 cidx = rxq->ifr_fl[0].ifl_cidx;
1519 if (iflib_rxd_avail(ctx, rxq, cidx, 1))
1520 GROUPTASK_ENQUEUE(gtask);
1522 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id);
1523 DBG_COUNTER_INC(rx_intr_enables);
1525 return (FILTER_HANDLED);
1530 iflib_fast_intr_ctx(void *arg)
1532 iflib_filter_info_t info = arg;
1533 struct grouptask *gtask = info->ifi_task;
1536 return (FILTER_HANDLED);
1538 DBG_COUNTER_INC(fast_intrs);
1539 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED)
1540 return (FILTER_HANDLED);
1542 GROUPTASK_ENQUEUE(gtask);
1543 return (FILTER_HANDLED);
1547 _iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid,
1548 driver_filter_t filter, driver_intr_t handler, void *arg,
1552 struct resource *res;
1554 device_t dev = ctx->ifc_dev;
1557 if (ctx->ifc_flags & IFC_LEGACY)
1558 flags |= RF_SHAREABLE;
1561 res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &irq->ii_rid, flags);
1564 "failed to allocate IRQ for rid %d, name %s.\n", rid, name);
1568 KASSERT(filter == NULL || handler == NULL, ("filter and handler can't both be non-NULL"));
1569 rc = bus_setup_intr(dev, res, INTR_MPSAFE | INTR_TYPE_NET,
1570 filter, handler, arg, &tag);
1573 "failed to setup interrupt for rid %d, name %s: %d\n",
1574 rid, name ? name : "unknown", rc);
1577 bus_describe_intr(dev, res, tag, "%s", name);
1584 /*********************************************************************
1586 * Allocate DMA resources for TX buffers as well as memory for the TX
1587 * mbuf map. TX DMA maps (non-TSO/TSO) and TX mbuf map are kept in a
1588 * iflib_sw_tx_desc_array structure, storing all the information that
1589 * is needed to transmit a packet on the wire. This is called only
1590 * once at attach, setup is done every reset.
1592 **********************************************************************/
1594 iflib_txsd_alloc(iflib_txq_t txq)
1596 if_ctx_t ctx = txq->ift_ctx;
1597 if_shared_ctx_t sctx = ctx->ifc_sctx;
1598 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1599 device_t dev = ctx->ifc_dev;
1600 bus_size_t tsomaxsize;
1601 int err, nsegments, ntsosegments;
1604 nsegments = scctx->isc_tx_nsegments;
1605 ntsosegments = scctx->isc_tx_tso_segments_max;
1606 tsomaxsize = scctx->isc_tx_tso_size_max;
1607 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_VLAN_MTU)
1608 tsomaxsize += sizeof(struct ether_vlan_header);
1609 MPASS(scctx->isc_ntxd[0] > 0);
1610 MPASS(scctx->isc_ntxd[txq->ift_br_offset] > 0);
1611 MPASS(nsegments > 0);
1612 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) {
1613 MPASS(ntsosegments > 0);
1614 MPASS(sctx->isc_tso_maxsize >= tsomaxsize);
1618 * Set up DMA tags for TX buffers.
1620 if ((err = bus_dma_tag_create(bus_get_dma_tag(dev),
1621 1, 0, /* alignment, bounds */
1622 BUS_SPACE_MAXADDR, /* lowaddr */
1623 BUS_SPACE_MAXADDR, /* highaddr */
1624 NULL, NULL, /* filter, filterarg */
1625 sctx->isc_tx_maxsize, /* maxsize */
1626 nsegments, /* nsegments */
1627 sctx->isc_tx_maxsegsize, /* maxsegsize */
1629 NULL, /* lockfunc */
1630 NULL, /* lockfuncarg */
1631 &txq->ift_buf_tag))) {
1632 device_printf(dev,"Unable to allocate TX DMA tag: %d\n", err);
1633 device_printf(dev,"maxsize: %ju nsegments: %d maxsegsize: %ju\n",
1634 (uintmax_t)sctx->isc_tx_maxsize, nsegments, (uintmax_t)sctx->isc_tx_maxsegsize);
1637 tso = (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) != 0;
1638 if (tso && (err = bus_dma_tag_create(bus_get_dma_tag(dev),
1639 1, 0, /* alignment, bounds */
1640 BUS_SPACE_MAXADDR, /* lowaddr */
1641 BUS_SPACE_MAXADDR, /* highaddr */
1642 NULL, NULL, /* filter, filterarg */
1643 tsomaxsize, /* maxsize */
1644 ntsosegments, /* nsegments */
1645 sctx->isc_tso_maxsegsize,/* maxsegsize */
1647 NULL, /* lockfunc */
1648 NULL, /* lockfuncarg */
1649 &txq->ift_tso_buf_tag))) {
1650 device_printf(dev, "Unable to allocate TSO TX DMA tag: %d\n",
1655 /* Allocate memory for the TX mbuf map. */
1656 if (!(txq->ift_sds.ifsd_m =
1657 (struct mbuf **) malloc(sizeof(struct mbuf *) *
1658 scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1659 device_printf(dev, "Unable to allocate TX mbuf map memory\n");
1665 * Create the DMA maps for TX buffers.
1667 if ((txq->ift_sds.ifsd_map = (bus_dmamap_t *)malloc(
1668 sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset],
1669 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
1671 "Unable to allocate TX buffer DMA map memory\n");
1675 if (tso && (txq->ift_sds.ifsd_tso_map = (bus_dmamap_t *)malloc(
1676 sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset],
1677 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
1679 "Unable to allocate TSO TX buffer map memory\n");
1683 for (int i = 0; i < scctx->isc_ntxd[txq->ift_br_offset]; i++) {
1684 err = bus_dmamap_create(txq->ift_buf_tag, 0,
1685 &txq->ift_sds.ifsd_map[i]);
1687 device_printf(dev, "Unable to create TX DMA map\n");
1692 err = bus_dmamap_create(txq->ift_tso_buf_tag, 0,
1693 &txq->ift_sds.ifsd_tso_map[i]);
1695 device_printf(dev, "Unable to create TSO TX DMA map\n");
1701 /* We free all, it handles case where we are in the middle */
1702 iflib_tx_structures_free(ctx);
1707 iflib_txsd_destroy(if_ctx_t ctx, iflib_txq_t txq, int i)
1712 if (txq->ift_sds.ifsd_map != NULL)
1713 map = txq->ift_sds.ifsd_map[i];
1715 bus_dmamap_sync(txq->ift_buf_tag, map, BUS_DMASYNC_POSTWRITE);
1716 bus_dmamap_unload(txq->ift_buf_tag, map);
1717 bus_dmamap_destroy(txq->ift_buf_tag, map);
1718 txq->ift_sds.ifsd_map[i] = NULL;
1722 if (txq->ift_sds.ifsd_tso_map != NULL)
1723 map = txq->ift_sds.ifsd_tso_map[i];
1725 bus_dmamap_sync(txq->ift_tso_buf_tag, map,
1726 BUS_DMASYNC_POSTWRITE);
1727 bus_dmamap_unload(txq->ift_tso_buf_tag, map);
1728 bus_dmamap_destroy(txq->ift_tso_buf_tag, map);
1729 txq->ift_sds.ifsd_tso_map[i] = NULL;
1734 iflib_txq_destroy(iflib_txq_t txq)
1736 if_ctx_t ctx = txq->ift_ctx;
1738 for (int i = 0; i < txq->ift_size; i++)
1739 iflib_txsd_destroy(ctx, txq, i);
1740 if (txq->ift_sds.ifsd_map != NULL) {
1741 free(txq->ift_sds.ifsd_map, M_IFLIB);
1742 txq->ift_sds.ifsd_map = NULL;
1744 if (txq->ift_sds.ifsd_tso_map != NULL) {
1745 free(txq->ift_sds.ifsd_tso_map, M_IFLIB);
1746 txq->ift_sds.ifsd_tso_map = NULL;
1748 if (txq->ift_sds.ifsd_m != NULL) {
1749 free(txq->ift_sds.ifsd_m, M_IFLIB);
1750 txq->ift_sds.ifsd_m = NULL;
1752 if (txq->ift_buf_tag != NULL) {
1753 bus_dma_tag_destroy(txq->ift_buf_tag);
1754 txq->ift_buf_tag = NULL;
1756 if (txq->ift_tso_buf_tag != NULL) {
1757 bus_dma_tag_destroy(txq->ift_tso_buf_tag);
1758 txq->ift_tso_buf_tag = NULL;
1763 iflib_txsd_free(if_ctx_t ctx, iflib_txq_t txq, int i)
1767 mp = &txq->ift_sds.ifsd_m[i];
1771 if (txq->ift_sds.ifsd_map != NULL) {
1772 bus_dmamap_sync(txq->ift_buf_tag,
1773 txq->ift_sds.ifsd_map[i], BUS_DMASYNC_POSTWRITE);
1774 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[i]);
1776 if (txq->ift_sds.ifsd_tso_map != NULL) {
1777 bus_dmamap_sync(txq->ift_tso_buf_tag,
1778 txq->ift_sds.ifsd_tso_map[i], BUS_DMASYNC_POSTWRITE);
1779 bus_dmamap_unload(txq->ift_tso_buf_tag,
1780 txq->ift_sds.ifsd_tso_map[i]);
1783 DBG_COUNTER_INC(tx_frees);
1788 iflib_txq_setup(iflib_txq_t txq)
1790 if_ctx_t ctx = txq->ift_ctx;
1791 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1792 if_shared_ctx_t sctx = ctx->ifc_sctx;
1793 iflib_dma_info_t di;
1796 /* Set number of descriptors available */
1797 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
1798 /* XXX make configurable */
1799 txq->ift_update_freq = IFLIB_DEFAULT_TX_UPDATE_FREQ;
1802 txq->ift_cidx_processed = 0;
1803 txq->ift_pidx = txq->ift_cidx = txq->ift_npending = 0;
1804 txq->ift_size = scctx->isc_ntxd[txq->ift_br_offset];
1806 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++)
1807 bzero((void *)di->idi_vaddr, di->idi_size);
1809 IFDI_TXQ_SETUP(ctx, txq->ift_id);
1810 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++)
1811 bus_dmamap_sync(di->idi_tag, di->idi_map,
1812 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1816 /*********************************************************************
1818 * Allocate DMA resources for RX buffers as well as memory for the RX
1819 * mbuf map, direct RX cluster pointer map and RX cluster bus address
1820 * map. RX DMA map, RX mbuf map, direct RX cluster pointer map and
1821 * RX cluster map are kept in a iflib_sw_rx_desc_array structure.
1822 * Since we use use one entry in iflib_sw_rx_desc_array per received
1823 * packet, the maximum number of entries we'll need is equal to the
1824 * number of hardware receive descriptors that we've allocated.
1826 **********************************************************************/
1828 iflib_rxsd_alloc(iflib_rxq_t rxq)
1830 if_ctx_t ctx = rxq->ifr_ctx;
1831 if_shared_ctx_t sctx = ctx->ifc_sctx;
1832 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1833 device_t dev = ctx->ifc_dev;
1837 MPASS(scctx->isc_nrxd[0] > 0);
1838 MPASS(scctx->isc_nrxd[rxq->ifr_fl_offset] > 0);
1841 for (int i = 0; i < rxq->ifr_nfl; i++, fl++) {
1842 fl->ifl_size = scctx->isc_nrxd[rxq->ifr_fl_offset]; /* this isn't necessarily the same */
1843 /* Set up DMA tag for RX buffers. */
1844 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
1845 1, 0, /* alignment, bounds */
1846 BUS_SPACE_MAXADDR, /* lowaddr */
1847 BUS_SPACE_MAXADDR, /* highaddr */
1848 NULL, NULL, /* filter, filterarg */
1849 sctx->isc_rx_maxsize, /* maxsize */
1850 sctx->isc_rx_nsegments, /* nsegments */
1851 sctx->isc_rx_maxsegsize, /* maxsegsize */
1853 NULL, /* lockfunc */
1858 "Unable to allocate RX DMA tag: %d\n", err);
1862 /* Allocate memory for the RX mbuf map. */
1863 if (!(fl->ifl_sds.ifsd_m =
1864 (struct mbuf **) malloc(sizeof(struct mbuf *) *
1865 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1867 "Unable to allocate RX mbuf map memory\n");
1872 /* Allocate memory for the direct RX cluster pointer map. */
1873 if (!(fl->ifl_sds.ifsd_cl =
1874 (caddr_t *) malloc(sizeof(caddr_t) *
1875 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1877 "Unable to allocate RX cluster map memory\n");
1882 /* Allocate memory for the RX cluster bus address map. */
1883 if (!(fl->ifl_sds.ifsd_ba =
1884 (bus_addr_t *) malloc(sizeof(bus_addr_t) *
1885 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1887 "Unable to allocate RX bus address map memory\n");
1893 * Create the DMA maps for RX buffers.
1895 if (!(fl->ifl_sds.ifsd_map =
1896 (bus_dmamap_t *) malloc(sizeof(bus_dmamap_t) * scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1898 "Unable to allocate RX buffer DMA map memory\n");
1902 for (int i = 0; i < scctx->isc_nrxd[rxq->ifr_fl_offset]; i++) {
1903 err = bus_dmamap_create(fl->ifl_buf_tag, 0,
1904 &fl->ifl_sds.ifsd_map[i]);
1906 device_printf(dev, "Unable to create RX buffer DMA map\n");
1914 iflib_rx_structures_free(ctx);
1920 * Internal service routines
1923 struct rxq_refill_cb_arg {
1925 bus_dma_segment_t seg;
1930 _rxq_refill_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
1932 struct rxq_refill_cb_arg *cb_arg = arg;
1934 cb_arg->error = error;
1935 cb_arg->seg = segs[0];
1936 cb_arg->nseg = nseg;
1940 * rxq_refill - refill an rxq free-buffer list
1941 * @ctx: the iflib context
1942 * @rxq: the free-list to refill
1943 * @n: the number of new buffers to allocate
1945 * (Re)populate an rxq free-buffer list with up to @n new packet buffers.
1946 * The caller must assure that @n does not exceed the queue's capacity.
1949 _iflib_fl_refill(if_ctx_t ctx, iflib_fl_t fl, int count)
1951 struct if_rxd_update iru;
1952 struct rxq_refill_cb_arg cb_arg;
1956 bus_dmamap_t *sd_map;
1957 bus_addr_t bus_addr, *sd_ba;
1958 int err, frag_idx, i, idx, n, pidx;
1961 sd_m = fl->ifl_sds.ifsd_m;
1962 sd_map = fl->ifl_sds.ifsd_map;
1963 sd_cl = fl->ifl_sds.ifsd_cl;
1964 sd_ba = fl->ifl_sds.ifsd_ba;
1965 pidx = fl->ifl_pidx;
1967 frag_idx = fl->ifl_fragidx;
1968 credits = fl->ifl_credits;
1973 MPASS(credits + n <= fl->ifl_size);
1975 if (pidx < fl->ifl_cidx)
1976 MPASS(pidx + n <= fl->ifl_cidx);
1977 if (pidx == fl->ifl_cidx && (credits < fl->ifl_size))
1978 MPASS(fl->ifl_gen == 0);
1979 if (pidx > fl->ifl_cidx)
1980 MPASS(n <= fl->ifl_size - pidx + fl->ifl_cidx);
1982 DBG_COUNTER_INC(fl_refills);
1984 DBG_COUNTER_INC(fl_refills_large);
1985 iru_init(&iru, fl->ifl_rxq, fl->ifl_id);
1988 * We allocate an uninitialized mbuf + cluster, mbuf is
1989 * initialized after rx.
1991 * If the cluster is still set then we know a minimum sized packet was received
1993 bit_ffc_at(fl->ifl_rx_bitmap, frag_idx, fl->ifl_size,
1996 bit_ffc(fl->ifl_rx_bitmap, fl->ifl_size, &frag_idx);
1997 MPASS(frag_idx >= 0);
1998 if ((cl = sd_cl[frag_idx]) == NULL) {
1999 if ((cl = m_cljget(NULL, M_NOWAIT, fl->ifl_buf_size)) == NULL)
2003 MPASS(sd_map != NULL);
2004 err = bus_dmamap_load(fl->ifl_buf_tag, sd_map[frag_idx],
2005 cl, fl->ifl_buf_size, _rxq_refill_cb, &cb_arg,
2007 if (err != 0 || cb_arg.error) {
2011 if (fl->ifl_zone == zone_pack)
2012 uma_zfree(fl->ifl_zone, cl);
2016 sd_ba[frag_idx] = bus_addr = cb_arg.seg.ds_addr;
2017 sd_cl[frag_idx] = cl;
2019 fl->ifl_cl_enqueued++;
2022 bus_addr = sd_ba[frag_idx];
2024 bus_dmamap_sync(fl->ifl_buf_tag, sd_map[frag_idx],
2025 BUS_DMASYNC_PREREAD);
2027 MPASS(sd_m[frag_idx] == NULL);
2028 if ((m = m_gethdr(M_NOWAIT, MT_NOINIT)) == NULL) {
2032 bit_set(fl->ifl_rx_bitmap, frag_idx);
2034 fl->ifl_m_enqueued++;
2037 DBG_COUNTER_INC(rx_allocs);
2038 fl->ifl_rxd_idxs[i] = frag_idx;
2039 fl->ifl_bus_addrs[i] = bus_addr;
2040 fl->ifl_vm_addrs[i] = cl;
2043 MPASS(credits <= fl->ifl_size);
2044 if (++idx == fl->ifl_size) {
2048 if (n == 0 || i == IFLIB_MAX_RX_REFRESH) {
2049 iru.iru_pidx = pidx;
2051 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
2055 fl->ifl_credits = credits;
2060 iru.iru_pidx = pidx;
2062 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
2064 fl->ifl_credits = credits;
2066 DBG_COUNTER_INC(rxd_flush);
2067 if (fl->ifl_pidx == 0)
2068 pidx = fl->ifl_size - 1;
2070 pidx = fl->ifl_pidx - 1;
2072 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
2073 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2074 ctx->isc_rxd_flush(ctx->ifc_softc, fl->ifl_rxq->ifr_id, fl->ifl_id, pidx);
2075 fl->ifl_fragidx = frag_idx;
2078 static __inline void
2079 __iflib_fl_refill_lt(if_ctx_t ctx, iflib_fl_t fl, int max)
2081 /* we avoid allowing pidx to catch up with cidx as it confuses ixl */
2082 int32_t reclaimable = fl->ifl_size - fl->ifl_credits - 1;
2084 int32_t delta = fl->ifl_size - get_inuse(fl->ifl_size, fl->ifl_cidx, fl->ifl_pidx, fl->ifl_gen) - 1;
2087 MPASS(fl->ifl_credits <= fl->ifl_size);
2088 MPASS(reclaimable == delta);
2090 if (reclaimable > 0)
2091 _iflib_fl_refill(ctx, fl, min(max, reclaimable));
2095 iflib_in_detach(if_ctx_t ctx)
2099 in_detach = !!(ctx->ifc_flags & IFC_IN_DETACH);
2105 iflib_fl_bufs_free(iflib_fl_t fl)
2107 iflib_dma_info_t idi = fl->ifl_ifdi;
2108 bus_dmamap_t sd_map;
2111 for (i = 0; i < fl->ifl_size; i++) {
2112 struct mbuf **sd_m = &fl->ifl_sds.ifsd_m[i];
2113 caddr_t *sd_cl = &fl->ifl_sds.ifsd_cl[i];
2115 if (*sd_cl != NULL) {
2116 sd_map = fl->ifl_sds.ifsd_map[i];
2117 bus_dmamap_sync(fl->ifl_buf_tag, sd_map,
2118 BUS_DMASYNC_POSTREAD);
2119 bus_dmamap_unload(fl->ifl_buf_tag, sd_map);
2121 uma_zfree(fl->ifl_zone, *sd_cl);
2122 // XXX: Should this get moved out?
2123 if (iflib_in_detach(fl->ifl_rxq->ifr_ctx))
2124 bus_dmamap_destroy(fl->ifl_buf_tag, sd_map);
2125 if (*sd_m != NULL) {
2126 m_init(*sd_m, M_NOWAIT, MT_DATA, 0);
2127 uma_zfree(zone_mbuf, *sd_m);
2130 MPASS(*sd_cl == NULL);
2131 MPASS(*sd_m == NULL);
2134 fl->ifl_m_dequeued++;
2135 fl->ifl_cl_dequeued++;
2141 for (i = 0; i < fl->ifl_size; i++) {
2142 MPASS(fl->ifl_sds.ifsd_cl[i] == NULL);
2143 MPASS(fl->ifl_sds.ifsd_m[i] == NULL);
2147 * Reset free list values
2149 fl->ifl_credits = fl->ifl_cidx = fl->ifl_pidx = fl->ifl_gen = fl->ifl_fragidx = 0;
2150 bzero(idi->idi_vaddr, idi->idi_size);
2153 /*********************************************************************
2155 * Initialize a receive ring and its buffers.
2157 **********************************************************************/
2159 iflib_fl_setup(iflib_fl_t fl)
2161 iflib_rxq_t rxq = fl->ifl_rxq;
2162 if_ctx_t ctx = rxq->ifr_ctx;
2163 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2165 bit_nclear(fl->ifl_rx_bitmap, 0, fl->ifl_size - 1);
2167 ** Free current RX buffer structs and their mbufs
2169 iflib_fl_bufs_free(fl);
2170 /* Now replenish the mbufs */
2171 MPASS(fl->ifl_credits == 0);
2173 * XXX don't set the max_frame_size to larger
2174 * than the hardware can handle
2176 if (sctx->isc_max_frame_size <= 2048)
2177 fl->ifl_buf_size = MCLBYTES;
2178 #ifndef CONTIGMALLOC_WORKS
2180 fl->ifl_buf_size = MJUMPAGESIZE;
2182 else if (sctx->isc_max_frame_size <= 4096)
2183 fl->ifl_buf_size = MJUMPAGESIZE;
2184 else if (sctx->isc_max_frame_size <= 9216)
2185 fl->ifl_buf_size = MJUM9BYTES;
2187 fl->ifl_buf_size = MJUM16BYTES;
2189 if (fl->ifl_buf_size > ctx->ifc_max_fl_buf_size)
2190 ctx->ifc_max_fl_buf_size = fl->ifl_buf_size;
2191 fl->ifl_cltype = m_gettype(fl->ifl_buf_size);
2192 fl->ifl_zone = m_getzone(fl->ifl_buf_size);
2195 /* avoid pre-allocating zillions of clusters to an idle card
2196 * potentially speeding up attach
2198 _iflib_fl_refill(ctx, fl, min(128, fl->ifl_size));
2199 MPASS(min(128, fl->ifl_size) == fl->ifl_credits);
2200 if (min(128, fl->ifl_size) != fl->ifl_credits)
2206 MPASS(fl->ifl_ifdi != NULL);
2207 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
2208 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2212 /*********************************************************************
2214 * Free receive ring data structures
2216 **********************************************************************/
2218 iflib_rx_sds_free(iflib_rxq_t rxq)
2223 if (rxq->ifr_fl != NULL) {
2224 for (i = 0; i < rxq->ifr_nfl; i++) {
2225 fl = &rxq->ifr_fl[i];
2226 if (fl->ifl_buf_tag != NULL) {
2227 if (fl->ifl_sds.ifsd_map != NULL) {
2228 for (j = 0; j < fl->ifl_size; j++) {
2229 if (fl->ifl_sds.ifsd_map[j] ==
2234 fl->ifl_sds.ifsd_map[j],
2235 BUS_DMASYNC_POSTREAD);
2238 fl->ifl_sds.ifsd_map[j]);
2241 bus_dma_tag_destroy(fl->ifl_buf_tag);
2242 fl->ifl_buf_tag = NULL;
2244 free(fl->ifl_sds.ifsd_m, M_IFLIB);
2245 free(fl->ifl_sds.ifsd_cl, M_IFLIB);
2246 free(fl->ifl_sds.ifsd_ba, M_IFLIB);
2247 free(fl->ifl_sds.ifsd_map, M_IFLIB);
2248 fl->ifl_sds.ifsd_m = NULL;
2249 fl->ifl_sds.ifsd_cl = NULL;
2250 fl->ifl_sds.ifsd_ba = NULL;
2251 fl->ifl_sds.ifsd_map = NULL;
2253 free(rxq->ifr_fl, M_IFLIB);
2255 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
2260 * MI independent logic
2264 iflib_timer(void *arg)
2266 iflib_txq_t txq = arg;
2267 if_ctx_t ctx = txq->ift_ctx;
2268 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2269 uint64_t this_tick = ticks;
2270 uint32_t reset_on = hz / 2;
2272 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))
2275 ** Check on the state of the TX queue(s), this
2276 ** can be done without the lock because its RO
2277 ** and the HUNG state will be static if set.
2279 if (this_tick - txq->ift_last_timer_tick >= hz / 2) {
2280 txq->ift_last_timer_tick = this_tick;
2281 IFDI_TIMER(ctx, txq->ift_id);
2282 if ((txq->ift_qstatus == IFLIB_QUEUE_HUNG) &&
2283 ((txq->ift_cleaned_prev == txq->ift_cleaned) ||
2284 (sctx->isc_pause_frames == 0)))
2287 if (ifmp_ring_is_stalled(txq->ift_br))
2288 txq->ift_qstatus = IFLIB_QUEUE_HUNG;
2289 txq->ift_cleaned_prev = txq->ift_cleaned;
2292 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP)
2293 iflib_netmap_timer_adjust(ctx, txq, &reset_on);
2295 /* handle any laggards */
2296 if (txq->ift_db_pending)
2297 GROUPTASK_ENQUEUE(&txq->ift_task);
2299 sctx->isc_pause_frames = 0;
2300 if (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)
2301 callout_reset_on(&txq->ift_timer, reset_on, iflib_timer, txq, txq->ift_timer.c_cpu);
2304 device_printf(ctx->ifc_dev, "TX(%d) desc avail = %d, pidx = %d\n",
2305 txq->ift_id, TXQ_AVAIL(txq), txq->ift_pidx);
2307 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2308 ctx->ifc_flags |= (IFC_DO_WATCHDOG|IFC_DO_RESET);
2309 iflib_admin_intr_deferred(ctx);
2314 iflib_init_locked(if_ctx_t ctx)
2316 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2317 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2318 if_t ifp = ctx->ifc_ifp;
2322 int i, j, tx_ip_csum_flags, tx_ip6_csum_flags;
2325 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2326 IFDI_INTR_DISABLE(ctx);
2328 tx_ip_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP);
2329 tx_ip6_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_SCTP);
2330 /* Set hardware offload abilities */
2331 if_clearhwassist(ifp);
2332 if (if_getcapenable(ifp) & IFCAP_TXCSUM)
2333 if_sethwassistbits(ifp, tx_ip_csum_flags, 0);
2334 if (if_getcapenable(ifp) & IFCAP_TXCSUM_IPV6)
2335 if_sethwassistbits(ifp, tx_ip6_csum_flags, 0);
2336 if (if_getcapenable(ifp) & IFCAP_TSO4)
2337 if_sethwassistbits(ifp, CSUM_IP_TSO, 0);
2338 if (if_getcapenable(ifp) & IFCAP_TSO6)
2339 if_sethwassistbits(ifp, CSUM_IP6_TSO, 0);
2341 for (i = 0, txq = ctx->ifc_txqs; i < sctx->isc_ntxqsets; i++, txq++) {
2343 callout_stop(&txq->ift_timer);
2344 CALLOUT_UNLOCK(txq);
2345 iflib_netmap_txq_init(ctx, txq);
2348 i = if_getdrvflags(ifp);
2351 MPASS(if_getdrvflags(ifp) == i);
2352 for (i = 0, rxq = ctx->ifc_rxqs; i < sctx->isc_nrxqsets; i++, rxq++) {
2353 /* XXX this should really be done on a per-queue basis */
2354 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
2355 MPASS(rxq->ifr_id == i);
2356 iflib_netmap_rxq_init(ctx, rxq);
2359 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) {
2360 if (iflib_fl_setup(fl)) {
2361 device_printf(ctx->ifc_dev, "freelist setup failed - check cluster settings\n");
2367 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
2368 IFDI_INTR_ENABLE(ctx);
2369 txq = ctx->ifc_txqs;
2370 for (i = 0; i < sctx->isc_ntxqsets; i++, txq++)
2371 callout_reset_on(&txq->ift_timer, hz/2, iflib_timer, txq,
2372 txq->ift_timer.c_cpu);
2376 iflib_media_change(if_t ifp)
2378 if_ctx_t ctx = if_getsoftc(ifp);
2382 if ((err = IFDI_MEDIA_CHANGE(ctx)) == 0)
2383 iflib_init_locked(ctx);
2389 iflib_media_status(if_t ifp, struct ifmediareq *ifmr)
2391 if_ctx_t ctx = if_getsoftc(ifp);
2394 IFDI_UPDATE_ADMIN_STATUS(ctx);
2395 IFDI_MEDIA_STATUS(ctx, ifmr);
2400 iflib_stop(if_ctx_t ctx)
2402 iflib_txq_t txq = ctx->ifc_txqs;
2403 iflib_rxq_t rxq = ctx->ifc_rxqs;
2404 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2405 if_shared_ctx_t sctx = ctx->ifc_sctx;
2406 iflib_dma_info_t di;
2410 /* Tell the stack that the interface is no longer active */
2411 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2413 IFDI_INTR_DISABLE(ctx);
2418 iflib_debug_reset();
2419 /* Wait for current tx queue users to exit to disarm watchdog timer. */
2420 for (i = 0; i < scctx->isc_ntxqsets; i++, txq++) {
2421 /* make sure all transmitters have completed before proceeding XXX */
2424 callout_stop(&txq->ift_timer);
2425 CALLOUT_UNLOCK(txq);
2427 /* clean any enqueued buffers */
2428 iflib_ifmp_purge(txq);
2429 /* Free any existing tx buffers. */
2430 for (j = 0; j < txq->ift_size; j++) {
2431 iflib_txsd_free(ctx, txq, j);
2433 txq->ift_processed = txq->ift_cleaned = txq->ift_cidx_processed = 0;
2434 txq->ift_in_use = txq->ift_gen = txq->ift_cidx = txq->ift_pidx = txq->ift_no_desc_avail = 0;
2435 txq->ift_closed = txq->ift_mbuf_defrag = txq->ift_mbuf_defrag_failed = 0;
2436 txq->ift_no_tx_dma_setup = txq->ift_txd_encap_efbig = txq->ift_map_failed = 0;
2437 txq->ift_pullups = 0;
2438 ifmp_ring_reset_stats(txq->ift_br);
2439 for (j = 0, di = txq->ift_ifdi; j < sctx->isc_ntxqs; j++, di++)
2440 bzero((void *)di->idi_vaddr, di->idi_size);
2442 for (i = 0; i < scctx->isc_nrxqsets; i++, rxq++) {
2443 /* make sure all transmitters have completed before proceeding XXX */
2445 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
2446 for (j = 0, di = rxq->ifr_ifdi; j < sctx->isc_nrxqs; j++, di++)
2447 bzero((void *)di->idi_vaddr, di->idi_size);
2448 /* also resets the free lists pidx/cidx */
2449 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
2450 iflib_fl_bufs_free(fl);
2454 static inline caddr_t
2455 calc_next_rxd(iflib_fl_t fl, int cidx)
2459 caddr_t start, end, cur, next;
2461 nrxd = fl->ifl_size;
2462 size = fl->ifl_rxd_size;
2463 start = fl->ifl_ifdi->idi_vaddr;
2465 if (__predict_false(size == 0))
2467 cur = start + size*cidx;
2468 end = start + size*nrxd;
2469 next = CACHE_PTR_NEXT(cur);
2470 return (next < end ? next : start);
2474 prefetch_pkts(iflib_fl_t fl, int cidx)
2477 int nrxd = fl->ifl_size;
2481 nextptr = (cidx + CACHE_PTR_INCREMENT) & (nrxd-1);
2482 prefetch(&fl->ifl_sds.ifsd_m[nextptr]);
2483 prefetch(&fl->ifl_sds.ifsd_cl[nextptr]);
2484 next_rxd = calc_next_rxd(fl, cidx);
2486 prefetch(fl->ifl_sds.ifsd_m[(cidx + 1) & (nrxd-1)]);
2487 prefetch(fl->ifl_sds.ifsd_m[(cidx + 2) & (nrxd-1)]);
2488 prefetch(fl->ifl_sds.ifsd_m[(cidx + 3) & (nrxd-1)]);
2489 prefetch(fl->ifl_sds.ifsd_m[(cidx + 4) & (nrxd-1)]);
2490 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 1) & (nrxd-1)]);
2491 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 2) & (nrxd-1)]);
2492 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 3) & (nrxd-1)]);
2493 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 4) & (nrxd-1)]);
2497 rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, int unload, if_rxsd_t sd)
2505 flid = irf->irf_flid;
2506 cidx = irf->irf_idx;
2507 fl = &rxq->ifr_fl[flid];
2509 sd->ifsd_cidx = cidx;
2510 sd->ifsd_m = &fl->ifl_sds.ifsd_m[cidx];
2511 sd->ifsd_cl = &fl->ifl_sds.ifsd_cl[cidx];
2514 fl->ifl_m_dequeued++;
2516 if (rxq->ifr_ctx->ifc_flags & IFC_PREFETCH)
2517 prefetch_pkts(fl, cidx);
2518 next = (cidx + CACHE_PTR_INCREMENT) & (fl->ifl_size-1);
2519 prefetch(&fl->ifl_sds.ifsd_map[next]);
2520 map = fl->ifl_sds.ifsd_map[cidx];
2521 next = (cidx + CACHE_LINE_SIZE) & (fl->ifl_size-1);
2523 /* not valid assert if bxe really does SGE from non-contiguous elements */
2524 MPASS(fl->ifl_cidx == cidx);
2525 bus_dmamap_sync(fl->ifl_buf_tag, map, BUS_DMASYNC_POSTREAD);
2527 bus_dmamap_unload(fl->ifl_buf_tag, map);
2528 fl->ifl_cidx = (fl->ifl_cidx + 1) & (fl->ifl_size-1);
2529 if (__predict_false(fl->ifl_cidx == 0))
2531 bit_clear(fl->ifl_rx_bitmap, cidx);
2534 static struct mbuf *
2535 assemble_segments(iflib_rxq_t rxq, if_rxd_info_t ri, if_rxsd_t sd)
2537 int i, padlen , flags;
2538 struct mbuf *m, *mh, *mt;
2544 rxd_frag_to_sd(rxq, &ri->iri_frags[i], TRUE, sd);
2546 MPASS(*sd->ifsd_cl != NULL);
2547 MPASS(*sd->ifsd_m != NULL);
2549 /* Don't include zero-length frags */
2550 if (ri->iri_frags[i].irf_len == 0) {
2551 /* XXX we can save the cluster here, but not the mbuf */
2552 m_init(*sd->ifsd_m, M_NOWAIT, MT_DATA, 0);
2553 m_free(*sd->ifsd_m);
2560 flags = M_PKTHDR|M_EXT;
2562 padlen = ri->iri_pad;
2567 /* assuming padding is only on the first fragment */
2571 *sd->ifsd_cl = NULL;
2573 /* Can these two be made one ? */
2574 m_init(m, M_NOWAIT, MT_DATA, flags);
2575 m_cljset(m, cl, sd->ifsd_fl->ifl_cltype);
2577 * These must follow m_init and m_cljset
2579 m->m_data += padlen;
2580 ri->iri_len -= padlen;
2581 m->m_len = ri->iri_frags[i].irf_len;
2582 } while (++i < ri->iri_nfrags);
2588 * Process one software descriptor
2590 static struct mbuf *
2591 iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri)
2596 /* should I merge this back in now that the two paths are basically duplicated? */
2597 if (ri->iri_nfrags == 1 &&
2598 ri->iri_frags[0].irf_len <= MIN(IFLIB_RX_COPY_THRESH, MHLEN)) {
2599 rxd_frag_to_sd(rxq, &ri->iri_frags[0], FALSE, &sd);
2602 m_init(m, M_NOWAIT, MT_DATA, M_PKTHDR);
2603 #ifndef __NO_STRICT_ALIGNMENT
2607 memcpy(m->m_data, *sd.ifsd_cl, ri->iri_len);
2608 m->m_len = ri->iri_frags[0].irf_len;
2610 m = assemble_segments(rxq, ri, &sd);
2612 m->m_pkthdr.len = ri->iri_len;
2613 m->m_pkthdr.rcvif = ri->iri_ifp;
2614 m->m_flags |= ri->iri_flags;
2615 m->m_pkthdr.ether_vtag = ri->iri_vtag;
2616 m->m_pkthdr.flowid = ri->iri_flowid;
2617 M_HASHTYPE_SET(m, ri->iri_rsstype);
2618 m->m_pkthdr.csum_flags = ri->iri_csum_flags;
2619 m->m_pkthdr.csum_data = ri->iri_csum_data;
2623 #if defined(INET6) || defined(INET)
2625 iflib_get_ip_forwarding(struct lro_ctrl *lc, bool *v4, bool *v6)
2627 CURVNET_SET(lc->ifp->if_vnet);
2629 *v6 = VNET(ip6_forwarding);
2632 *v4 = VNET(ipforwarding);
2638 * Returns true if it's possible this packet could be LROed.
2639 * if it returns false, it is guaranteed that tcp_lro_rx()
2640 * would not return zero.
2643 iflib_check_lro_possible(struct mbuf *m, bool v4_forwarding, bool v6_forwarding)
2645 struct ether_header *eh;
2648 eh = mtod(m, struct ether_header *);
2649 eh_type = ntohs(eh->ether_type);
2652 case ETHERTYPE_IPV6:
2653 return !v6_forwarding;
2657 return !v4_forwarding;
2665 iflib_get_ip_forwarding(struct lro_ctrl *lc __unused, bool *v4 __unused, bool *v6 __unused)
2671 iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
2673 if_ctx_t ctx = rxq->ifr_ctx;
2674 if_shared_ctx_t sctx = ctx->ifc_sctx;
2675 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2678 struct if_rxd_info ri;
2679 int err, budget_left, rx_bytes, rx_pkts;
2683 bool v4_forwarding, v6_forwarding, lro_possible;
2686 * XXX early demux data packets so that if_input processing only handles
2687 * acks in interrupt context
2689 struct mbuf *m, *mh, *mt, *mf;
2691 lro_possible = v4_forwarding = v6_forwarding = false;
2695 rx_pkts = rx_bytes = 0;
2696 if (sctx->isc_flags & IFLIB_HAS_RXCQ)
2697 cidxp = &rxq->ifr_cq_cidx;
2699 cidxp = &rxq->ifr_fl[0].ifl_cidx;
2700 if ((avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget)) == 0) {
2701 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
2702 __iflib_fl_refill_lt(ctx, fl, budget + 8);
2703 DBG_COUNTER_INC(rx_unavail);
2707 for (budget_left = budget; budget_left > 0 && avail > 0;) {
2708 if (__predict_false(!CTX_ACTIVE(ctx))) {
2709 DBG_COUNTER_INC(rx_ctx_inactive);
2713 * Reset client set fields to their default values
2716 ri.iri_qsidx = rxq->ifr_id;
2717 ri.iri_cidx = *cidxp;
2719 ri.iri_frags = rxq->ifr_frags;
2720 err = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri);
2724 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
2725 *cidxp = ri.iri_cidx;
2726 /* Update our consumer index */
2727 /* XXX NB: shurd - check if this is still safe */
2728 while (rxq->ifr_cq_cidx >= scctx->isc_nrxd[0]) {
2729 rxq->ifr_cq_cidx -= scctx->isc_nrxd[0];
2730 rxq->ifr_cq_gen = 0;
2732 /* was this only a completion queue message? */
2733 if (__predict_false(ri.iri_nfrags == 0))
2736 MPASS(ri.iri_nfrags != 0);
2737 MPASS(ri.iri_len != 0);
2739 /* will advance the cidx on the corresponding free lists */
2740 m = iflib_rxd_pkt_get(rxq, &ri);
2743 if (avail == 0 && budget_left)
2744 avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget_left);
2746 if (__predict_false(m == NULL)) {
2747 DBG_COUNTER_INC(rx_mbuf_null);
2750 /* imm_pkt: -- cxgb */
2758 /* make sure that we can refill faster than drain */
2759 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
2760 __iflib_fl_refill_lt(ctx, fl, budget + 8);
2762 lro_enabled = (if_getcapenable(ifp) & IFCAP_LRO);
2764 iflib_get_ip_forwarding(&rxq->ifr_lc, &v4_forwarding, &v6_forwarding);
2766 while (mh != NULL) {
2769 m->m_nextpkt = NULL;
2770 #ifndef __NO_STRICT_ALIGNMENT
2771 if (!IP_ALIGNED(m) && (m = iflib_fixup_rx(m)) == NULL)
2774 rx_bytes += m->m_pkthdr.len;
2776 #if defined(INET6) || defined(INET)
2778 if (!lro_possible) {
2779 lro_possible = iflib_check_lro_possible(m, v4_forwarding, v6_forwarding);
2780 if (lro_possible && mf != NULL) {
2781 ifp->if_input(ifp, mf);
2782 DBG_COUNTER_INC(rx_if_input);
2786 if ((m->m_pkthdr.csum_flags & (CSUM_L4_CALC|CSUM_L4_VALID)) ==
2787 (CSUM_L4_CALC|CSUM_L4_VALID)) {
2788 if (lro_possible && tcp_lro_rx(&rxq->ifr_lc, m, 0) == 0)
2794 ifp->if_input(ifp, m);
2795 DBG_COUNTER_INC(rx_if_input);
2806 ifp->if_input(ifp, mf);
2807 DBG_COUNTER_INC(rx_if_input);
2810 if_inc_counter(ifp, IFCOUNTER_IBYTES, rx_bytes);
2811 if_inc_counter(ifp, IFCOUNTER_IPACKETS, rx_pkts);
2814 * Flush any outstanding LRO work
2816 #if defined(INET6) || defined(INET)
2817 tcp_lro_flush_all(&rxq->ifr_lc);
2821 return (iflib_rxd_avail(ctx, rxq, *cidxp, 1));
2824 ctx->ifc_flags |= IFC_DO_RESET;
2825 iflib_admin_intr_deferred(ctx);
2830 #define TXD_NOTIFY_COUNT(txq) (((txq)->ift_size / (txq)->ift_update_freq)-1)
2831 static inline qidx_t
2832 txq_max_db_deferred(iflib_txq_t txq, qidx_t in_use)
2834 qidx_t notify_count = TXD_NOTIFY_COUNT(txq);
2835 qidx_t minthresh = txq->ift_size / 8;
2836 if (in_use > 4*minthresh)
2837 return (notify_count);
2838 if (in_use > 2*minthresh)
2839 return (notify_count >> 1);
2840 if (in_use > minthresh)
2841 return (notify_count >> 3);
2845 static inline qidx_t
2846 txq_max_rs_deferred(iflib_txq_t txq)
2848 qidx_t notify_count = TXD_NOTIFY_COUNT(txq);
2849 qidx_t minthresh = txq->ift_size / 8;
2850 if (txq->ift_in_use > 4*minthresh)
2851 return (notify_count);
2852 if (txq->ift_in_use > 2*minthresh)
2853 return (notify_count >> 1);
2854 if (txq->ift_in_use > minthresh)
2855 return (notify_count >> 2);
2859 #define M_CSUM_FLAGS(m) ((m)->m_pkthdr.csum_flags)
2860 #define M_HAS_VLANTAG(m) (m->m_flags & M_VLANTAG)
2862 #define TXQ_MAX_DB_DEFERRED(txq, in_use) txq_max_db_deferred((txq), (in_use))
2863 #define TXQ_MAX_RS_DEFERRED(txq) txq_max_rs_deferred(txq)
2864 #define TXQ_MAX_DB_CONSUMED(size) (size >> 4)
2866 /* forward compatibility for cxgb */
2867 #define FIRST_QSET(ctx) 0
2868 #define NTXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_ntxqsets)
2869 #define NRXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_nrxqsets)
2870 #define QIDX(ctx, m) ((((m)->m_pkthdr.flowid & ctx->ifc_softc_ctx.isc_rss_table_mask) % NTXQSETS(ctx)) + FIRST_QSET(ctx))
2871 #define DESC_RECLAIMABLE(q) ((int)((q)->ift_processed - (q)->ift_cleaned - (q)->ift_ctx->ifc_softc_ctx.isc_tx_nsegments))
2873 /* XXX we should be setting this to something other than zero */
2874 #define RECLAIM_THRESH(ctx) ((ctx)->ifc_sctx->isc_tx_reclaim_thresh)
2875 #define MAX_TX_DESC(ctx) max((ctx)->ifc_softc_ctx.isc_tx_tso_segments_max, \
2876 (ctx)->ifc_softc_ctx.isc_tx_nsegments)
2879 iflib_txd_db_check(if_ctx_t ctx, iflib_txq_t txq, int ring, qidx_t in_use)
2885 max = TXQ_MAX_DB_DEFERRED(txq, in_use);
2886 if (ring || txq->ift_db_pending >= max) {
2887 dbval = txq->ift_npending ? txq->ift_npending : txq->ift_pidx;
2888 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
2889 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2890 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, dbval);
2891 txq->ift_db_pending = txq->ift_npending = 0;
2899 print_pkt(if_pkt_info_t pi)
2901 printf("pi len: %d qsidx: %d nsegs: %d ndescs: %d flags: %x pidx: %d\n",
2902 pi->ipi_len, pi->ipi_qsidx, pi->ipi_nsegs, pi->ipi_ndescs, pi->ipi_flags, pi->ipi_pidx);
2903 printf("pi new_pidx: %d csum_flags: %lx tso_segsz: %d mflags: %x vtag: %d\n",
2904 pi->ipi_new_pidx, pi->ipi_csum_flags, pi->ipi_tso_segsz, pi->ipi_mflags, pi->ipi_vtag);
2905 printf("pi etype: %d ehdrlen: %d ip_hlen: %d ipproto: %d\n",
2906 pi->ipi_etype, pi->ipi_ehdrlen, pi->ipi_ip_hlen, pi->ipi_ipproto);
2910 #define IS_TSO4(pi) ((pi)->ipi_csum_flags & CSUM_IP_TSO)
2911 #define IS_TX_OFFLOAD4(pi) ((pi)->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP_TSO))
2912 #define IS_TSO6(pi) ((pi)->ipi_csum_flags & CSUM_IP6_TSO)
2913 #define IS_TX_OFFLOAD6(pi) ((pi)->ipi_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_TSO))
2916 iflib_parse_header(iflib_txq_t txq, if_pkt_info_t pi, struct mbuf **mp)
2918 if_shared_ctx_t sctx = txq->ift_ctx->ifc_sctx;
2919 struct ether_vlan_header *eh;
2923 if ((sctx->isc_flags & IFLIB_NEED_SCRATCH) &&
2924 M_WRITABLE(m) == 0) {
2925 if ((m = m_dup(m, M_NOWAIT)) == NULL) {
2929 DBG_COUNTER_INC(tx_frees);
2935 * Determine where frame payload starts.
2936 * Jump over vlan headers if already present,
2937 * helpful for QinQ too.
2939 if (__predict_false(m->m_len < sizeof(*eh))) {
2941 if (__predict_false((m = m_pullup(m, sizeof(*eh))) == NULL))
2944 eh = mtod(m, struct ether_vlan_header *);
2945 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
2946 pi->ipi_etype = ntohs(eh->evl_proto);
2947 pi->ipi_ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
2949 pi->ipi_etype = ntohs(eh->evl_encap_proto);
2950 pi->ipi_ehdrlen = ETHER_HDR_LEN;
2953 switch (pi->ipi_etype) {
2958 struct ip *ip = NULL;
2959 struct tcphdr *th = NULL;
2962 minthlen = min(m->m_pkthdr.len, pi->ipi_ehdrlen + sizeof(*ip) + sizeof(*th));
2963 if (__predict_false(m->m_len < minthlen)) {
2965 * if this code bloat is causing too much of a hit
2966 * move it to a separate function and mark it noinline
2968 if (m->m_len == pi->ipi_ehdrlen) {
2971 if (n->m_len >= sizeof(*ip)) {
2972 ip = (struct ip *)n->m_data;
2973 if (n->m_len >= (ip->ip_hl << 2) + sizeof(*th))
2974 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
2977 if (__predict_false((m = m_pullup(m, minthlen)) == NULL))
2979 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
2983 if (__predict_false((m = m_pullup(m, minthlen)) == NULL))
2985 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
2986 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th))
2987 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
2990 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
2991 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th))
2992 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
2994 pi->ipi_ip_hlen = ip->ip_hl << 2;
2995 pi->ipi_ipproto = ip->ip_p;
2996 pi->ipi_flags |= IPI_TX_IPV4;
2998 /* TCP checksum offload may require TCP header length */
2999 if (IS_TX_OFFLOAD4(pi)) {
3000 if (__predict_true(pi->ipi_ipproto == IPPROTO_TCP)) {
3001 if (__predict_false(th == NULL)) {
3003 if (__predict_false((m = m_pullup(m, (ip->ip_hl << 2) + sizeof(*th))) == NULL))
3005 th = (struct tcphdr *)((caddr_t)ip + pi->ipi_ip_hlen);
3007 pi->ipi_tcp_hflags = th->th_flags;
3008 pi->ipi_tcp_hlen = th->th_off << 2;
3009 pi->ipi_tcp_seq = th->th_seq;
3012 if (__predict_false(ip->ip_p != IPPROTO_TCP))
3015 * TSO always requires hardware checksum offload.
3017 pi->ipi_csum_flags |= (CSUM_IP_TCP | CSUM_IP);
3018 th->th_sum = in_pseudo(ip->ip_src.s_addr,
3019 ip->ip_dst.s_addr, htons(IPPROTO_TCP));
3020 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz;
3021 if (sctx->isc_flags & IFLIB_TSO_INIT_IP) {
3023 ip->ip_len = htons(pi->ipi_ip_hlen + pi->ipi_tcp_hlen + pi->ipi_tso_segsz);
3027 if ((sctx->isc_flags & IFLIB_NEED_ZERO_CSUM) && (pi->ipi_csum_flags & CSUM_IP))
3034 case ETHERTYPE_IPV6:
3036 struct ip6_hdr *ip6 = (struct ip6_hdr *)(m->m_data + pi->ipi_ehdrlen);
3038 pi->ipi_ip_hlen = sizeof(struct ip6_hdr);
3040 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) {
3042 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) == NULL))
3045 th = (struct tcphdr *)((caddr_t)ip6 + pi->ipi_ip_hlen);
3047 /* XXX-BZ this will go badly in case of ext hdrs. */
3048 pi->ipi_ipproto = ip6->ip6_nxt;
3049 pi->ipi_flags |= IPI_TX_IPV6;
3051 /* TCP checksum offload may require TCP header length */
3052 if (IS_TX_OFFLOAD6(pi)) {
3053 if (pi->ipi_ipproto == IPPROTO_TCP) {
3054 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) {
3056 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) == NULL))
3059 pi->ipi_tcp_hflags = th->th_flags;
3060 pi->ipi_tcp_hlen = th->th_off << 2;
3061 pi->ipi_tcp_seq = th->th_seq;
3064 if (__predict_false(ip6->ip6_nxt != IPPROTO_TCP))
3067 * TSO always requires hardware checksum offload.
3069 pi->ipi_csum_flags |= CSUM_IP6_TCP;
3070 th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0);
3071 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz;
3078 pi->ipi_csum_flags &= ~CSUM_OFFLOAD;
3079 pi->ipi_ip_hlen = 0;
3088 * If dodgy hardware rejects the scatter gather chain we've handed it
3089 * we'll need to remove the mbuf chain from ifsg_m[] before we can add the
3092 static __noinline struct mbuf *
3093 iflib_remove_mbuf(iflib_txq_t txq)
3096 struct mbuf *m, **ifsd_m;
3098 ifsd_m = txq->ift_sds.ifsd_m;
3099 ntxd = txq->ift_size;
3100 pidx = txq->ift_pidx & (ntxd - 1);
3101 ifsd_m = txq->ift_sds.ifsd_m;
3103 ifsd_m[pidx] = NULL;
3104 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[pidx]);
3105 if (txq->ift_sds.ifsd_tso_map != NULL)
3106 bus_dmamap_unload(txq->ift_tso_buf_tag,
3107 txq->ift_sds.ifsd_tso_map[pidx]);
3109 txq->ift_dequeued++;
3114 static inline caddr_t
3115 calc_next_txd(iflib_txq_t txq, int cidx, uint8_t qid)
3119 caddr_t start, end, cur, next;
3121 ntxd = txq->ift_size;
3122 size = txq->ift_txd_size[qid];
3123 start = txq->ift_ifdi[qid].idi_vaddr;
3125 if (__predict_false(size == 0))
3127 cur = start + size*cidx;
3128 end = start + size*ntxd;
3129 next = CACHE_PTR_NEXT(cur);
3130 return (next < end ? next : start);
3134 * Pad an mbuf to ensure a minimum ethernet frame size.
3135 * min_frame_size is the frame size (less CRC) to pad the mbuf to
3137 static __noinline int
3138 iflib_ether_pad(device_t dev, struct mbuf **m_head, uint16_t min_frame_size)
3141 * 18 is enough bytes to pad an ARP packet to 46 bytes, and
3142 * and ARP message is the smallest common payload I can think of
3144 static char pad[18]; /* just zeros */
3146 struct mbuf *new_head;
3148 if (!M_WRITABLE(*m_head)) {
3149 new_head = m_dup(*m_head, M_NOWAIT);
3150 if (new_head == NULL) {
3152 device_printf(dev, "cannot pad short frame, m_dup() failed");
3153 DBG_COUNTER_INC(encap_pad_mbuf_fail);
3154 DBG_COUNTER_INC(tx_frees);
3161 for (n = min_frame_size - (*m_head)->m_pkthdr.len;
3162 n > 0; n -= sizeof(pad))
3163 if (!m_append(*m_head, min(n, sizeof(pad)), pad))
3168 device_printf(dev, "cannot pad short frame\n");
3169 DBG_COUNTER_INC(encap_pad_mbuf_fail);
3170 DBG_COUNTER_INC(tx_frees);
3178 iflib_encap(iflib_txq_t txq, struct mbuf **m_headp)
3181 if_shared_ctx_t sctx;
3182 if_softc_ctx_t scctx;
3183 bus_dma_tag_t buf_tag;
3184 bus_dma_segment_t *segs;
3185 struct mbuf *m_head, **ifsd_m;
3188 struct if_pkt_info pi;
3190 int err, nsegs, ndesc, max_segs, pidx, cidx, next, ntxd;
3193 sctx = ctx->ifc_sctx;
3194 scctx = &ctx->ifc_softc_ctx;
3195 segs = txq->ift_segs;
3196 ntxd = txq->ift_size;
3201 * If we're doing TSO the next descriptor to clean may be quite far ahead
3203 cidx = txq->ift_cidx;
3204 pidx = txq->ift_pidx;
3205 if (ctx->ifc_flags & IFC_PREFETCH) {
3206 next = (cidx + CACHE_PTR_INCREMENT) & (ntxd-1);
3207 if (!(ctx->ifc_flags & IFLIB_HAS_TXCQ)) {
3208 next_txd = calc_next_txd(txq, cidx, 0);
3212 /* prefetch the next cache line of mbuf pointers and flags */
3213 prefetch(&txq->ift_sds.ifsd_m[next]);
3214 prefetch(&txq->ift_sds.ifsd_map[next]);
3215 next = (cidx + CACHE_LINE_SIZE) & (ntxd-1);
3217 map = txq->ift_sds.ifsd_map[pidx];
3218 ifsd_m = txq->ift_sds.ifsd_m;
3220 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3221 buf_tag = txq->ift_tso_buf_tag;
3222 max_segs = scctx->isc_tx_tso_segments_max;
3223 map = txq->ift_sds.ifsd_tso_map[pidx];
3224 MPASS(buf_tag != NULL);
3225 MPASS(max_segs > 0);
3227 buf_tag = txq->ift_buf_tag;
3228 max_segs = scctx->isc_tx_nsegments;
3229 map = txq->ift_sds.ifsd_map[pidx];
3231 if ((sctx->isc_flags & IFLIB_NEED_ETHER_PAD) &&
3232 __predict_false(m_head->m_pkthdr.len < scctx->isc_min_frame_size)) {
3233 err = iflib_ether_pad(ctx->ifc_dev, m_headp, scctx->isc_min_frame_size);
3235 DBG_COUNTER_INC(encap_txd_encap_fail);
3242 pi.ipi_mflags = (m_head->m_flags & (M_VLANTAG|M_BCAST|M_MCAST));
3244 pi.ipi_qsidx = txq->ift_id;
3245 pi.ipi_len = m_head->m_pkthdr.len;
3246 pi.ipi_csum_flags = m_head->m_pkthdr.csum_flags;
3247 pi.ipi_vtag = (m_head->m_flags & M_VLANTAG) ? m_head->m_pkthdr.ether_vtag : 0;
3249 /* deliberate bitwise OR to make one condition */
3250 if (__predict_true((pi.ipi_csum_flags | pi.ipi_vtag))) {
3251 if (__predict_false((err = iflib_parse_header(txq, &pi, m_headp)) != 0)) {
3252 DBG_COUNTER_INC(encap_txd_encap_fail);
3259 err = bus_dmamap_load_mbuf_sg(buf_tag, map, m_head, segs, &nsegs,
3262 if (__predict_false(err)) {
3265 /* try collapse once and defrag once */
3267 m_head = m_collapse(*m_headp, M_NOWAIT, max_segs);
3268 /* try defrag if collapsing fails */
3273 txq->ift_mbuf_defrag++;
3274 m_head = m_defrag(*m_headp, M_NOWAIT);
3277 if (__predict_false(m_head == NULL))
3283 txq->ift_no_tx_dma_setup++;
3286 txq->ift_no_tx_dma_setup++;
3288 DBG_COUNTER_INC(tx_frees);
3292 txq->ift_map_failed++;
3293 DBG_COUNTER_INC(encap_load_mbuf_fail);
3294 DBG_COUNTER_INC(encap_txd_encap_fail);
3297 ifsd_m[pidx] = m_head;
3299 * XXX assumes a 1 to 1 relationship between segments and
3300 * descriptors - this does not hold true on all drivers, e.g.
3303 if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
3304 txq->ift_no_desc_avail++;
3305 bus_dmamap_unload(buf_tag, map);
3306 DBG_COUNTER_INC(encap_txq_avail_fail);
3307 DBG_COUNTER_INC(encap_txd_encap_fail);
3308 if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
3309 GROUPTASK_ENQUEUE(&txq->ift_task);
3313 * On Intel cards we can greatly reduce the number of TX interrupts
3314 * we see by only setting report status on every Nth descriptor.
3315 * However, this also means that the driver will need to keep track
3316 * of the descriptors that RS was set on to check them for the DD bit.
3318 txq->ift_rs_pending += nsegs + 1;
3319 if (txq->ift_rs_pending > TXQ_MAX_RS_DEFERRED(txq) ||
3320 iflib_no_tx_batch || (TXQ_AVAIL(txq) - nsegs) <= MAX_TX_DESC(ctx) + 2) {
3321 pi.ipi_flags |= IPI_TX_INTR;
3322 txq->ift_rs_pending = 0;
3326 pi.ipi_nsegs = nsegs;
3328 MPASS(pidx >= 0 && pidx < txq->ift_size);
3332 if ((err = ctx->isc_txd_encap(ctx->ifc_softc, &pi)) == 0) {
3333 bus_dmamap_sync(buf_tag, map, BUS_DMASYNC_PREWRITE);
3334 DBG_COUNTER_INC(tx_encap);
3335 MPASS(pi.ipi_new_pidx < txq->ift_size);
3337 ndesc = pi.ipi_new_pidx - pi.ipi_pidx;
3338 if (pi.ipi_new_pidx < pi.ipi_pidx) {
3339 ndesc += txq->ift_size;
3343 * drivers can need as many as
3346 MPASS(ndesc <= pi.ipi_nsegs + 2);
3347 MPASS(pi.ipi_new_pidx != pidx);
3349 txq->ift_in_use += ndesc;
3352 * We update the last software descriptor again here because there may
3353 * be a sentinel and/or there may be more mbufs than segments
3355 txq->ift_pidx = pi.ipi_new_pidx;
3356 txq->ift_npending += pi.ipi_ndescs;
3358 *m_headp = m_head = iflib_remove_mbuf(txq);
3360 txq->ift_txd_encap_efbig++;
3369 * err can't possibly be non-zero here, so we don't neet to test it
3370 * to see if we need to DBG_COUNTER_INC(encap_txd_encap_fail).
3375 txq->ift_mbuf_defrag_failed++;
3376 txq->ift_map_failed++;
3378 DBG_COUNTER_INC(tx_frees);
3380 DBG_COUNTER_INC(encap_txd_encap_fail);
3385 iflib_tx_desc_free(iflib_txq_t txq, int n)
3387 uint32_t qsize, cidx, mask, gen;
3388 struct mbuf *m, **ifsd_m;
3391 cidx = txq->ift_cidx;
3393 qsize = txq->ift_size;
3395 ifsd_m = txq->ift_sds.ifsd_m;
3396 do_prefetch = (txq->ift_ctx->ifc_flags & IFC_PREFETCH);
3400 prefetch(ifsd_m[(cidx + 3) & mask]);
3401 prefetch(ifsd_m[(cidx + 4) & mask]);
3403 if ((m = ifsd_m[cidx]) != NULL) {
3404 prefetch(&ifsd_m[(cidx + CACHE_PTR_INCREMENT) & mask]);
3405 if (m->m_pkthdr.csum_flags & CSUM_TSO) {
3406 bus_dmamap_sync(txq->ift_tso_buf_tag,
3407 txq->ift_sds.ifsd_tso_map[cidx],
3408 BUS_DMASYNC_POSTWRITE);
3409 bus_dmamap_unload(txq->ift_tso_buf_tag,
3410 txq->ift_sds.ifsd_tso_map[cidx]);
3412 bus_dmamap_sync(txq->ift_buf_tag,
3413 txq->ift_sds.ifsd_map[cidx],
3414 BUS_DMASYNC_POSTWRITE);
3415 bus_dmamap_unload(txq->ift_buf_tag,
3416 txq->ift_sds.ifsd_map[cidx]);
3418 /* XXX we don't support any drivers that batch packets yet */
3419 MPASS(m->m_nextpkt == NULL);
3421 ifsd_m[cidx] = NULL;
3423 txq->ift_dequeued++;
3425 DBG_COUNTER_INC(tx_frees);
3427 if (__predict_false(++cidx == qsize)) {
3432 txq->ift_cidx = cidx;
3437 iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh)
3440 if_ctx_t ctx = txq->ift_ctx;
3442 KASSERT(thresh >= 0, ("invalid threshold to reclaim"));
3443 MPASS(thresh /*+ MAX_TX_DESC(txq->ift_ctx) */ < txq->ift_size);
3446 * Need a rate-limiting check so that this isn't called every time
3448 iflib_tx_credits_update(ctx, txq);
3449 reclaim = DESC_RECLAIMABLE(txq);
3451 if (reclaim <= thresh /* + MAX_TX_DESC(txq->ift_ctx) */) {
3453 if (iflib_verbose_debug) {
3454 printf("%s processed=%ju cleaned=%ju tx_nsegments=%d reclaim=%d thresh=%d\n", __FUNCTION__,
3455 txq->ift_processed, txq->ift_cleaned, txq->ift_ctx->ifc_softc_ctx.isc_tx_nsegments,
3462 iflib_tx_desc_free(txq, reclaim);
3463 txq->ift_cleaned += reclaim;
3464 txq->ift_in_use -= reclaim;
3469 static struct mbuf **
3470 _ring_peek_one(struct ifmp_ring *r, int cidx, int offset, int remaining)
3473 struct mbuf **items;
3476 next = (cidx + CACHE_PTR_INCREMENT) & (size-1);
3477 items = __DEVOLATILE(struct mbuf **, &r->items[0]);
3479 prefetch(items[(cidx + offset) & (size-1)]);
3480 if (remaining > 1) {
3481 prefetch2cachelines(&items[next]);
3482 prefetch2cachelines(items[(cidx + offset + 1) & (size-1)]);
3483 prefetch2cachelines(items[(cidx + offset + 2) & (size-1)]);
3484 prefetch2cachelines(items[(cidx + offset + 3) & (size-1)]);
3486 return (__DEVOLATILE(struct mbuf **, &r->items[(cidx + offset) & (size-1)]));
3490 iflib_txq_check_drain(iflib_txq_t txq, int budget)
3493 ifmp_ring_check_drainage(txq->ift_br, budget);
3497 iflib_txq_can_drain(struct ifmp_ring *r)
3499 iflib_txq_t txq = r->cookie;
3500 if_ctx_t ctx = txq->ift_ctx;
3502 if (TXQ_AVAIL(txq) > MAX_TX_DESC(ctx) + 2)
3504 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
3505 BUS_DMASYNC_POSTREAD);
3506 return (ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id,
3511 iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
3513 iflib_txq_t txq = r->cookie;
3514 if_ctx_t ctx = txq->ift_ctx;
3515 struct ifnet *ifp = ctx->ifc_ifp;
3516 struct mbuf **mp, *m;
3517 int i, count, consumed, pkt_sent, bytes_sent, mcast_sent, avail;
3518 int reclaimed, err, in_use_prev, desc_used;
3519 bool do_prefetch, ring, rang;
3521 if (__predict_false(!(if_getdrvflags(ifp) & IFF_DRV_RUNNING) ||
3522 !LINK_ACTIVE(ctx))) {
3523 DBG_COUNTER_INC(txq_drain_notready);
3526 reclaimed = iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
3527 rang = iflib_txd_db_check(ctx, txq, reclaimed, txq->ift_in_use);
3528 avail = IDXDIFF(pidx, cidx, r->size);
3529 if (__predict_false(ctx->ifc_flags & IFC_QFLUSH)) {
3530 DBG_COUNTER_INC(txq_drain_flushing);
3531 for (i = 0; i < avail; i++) {
3532 if (__predict_true(r->items[(cidx + i) & (r->size-1)] != (void *)txq))
3533 m_free(r->items[(cidx + i) & (r->size-1)]);
3534 r->items[(cidx + i) & (r->size-1)] = NULL;
3539 if (__predict_false(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE)) {
3540 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3542 callout_stop(&txq->ift_timer);
3543 CALLOUT_UNLOCK(txq);
3544 DBG_COUNTER_INC(txq_drain_oactive);
3548 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3549 consumed = mcast_sent = bytes_sent = pkt_sent = 0;
3550 count = MIN(avail, TX_BATCH_SIZE);
3552 if (iflib_verbose_debug)
3553 printf("%s avail=%d ifc_flags=%x txq_avail=%d ", __FUNCTION__,
3554 avail, ctx->ifc_flags, TXQ_AVAIL(txq));
3556 do_prefetch = (ctx->ifc_flags & IFC_PREFETCH);
3557 avail = TXQ_AVAIL(txq);
3559 for (desc_used = i = 0; i < count && avail > MAX_TX_DESC(ctx) + 2; i++) {
3560 int rem = do_prefetch ? count - i : 0;
3562 mp = _ring_peek_one(r, cidx, i, rem);
3563 MPASS(mp != NULL && *mp != NULL);
3564 if (__predict_false(*mp == (struct mbuf *)txq)) {
3569 in_use_prev = txq->ift_in_use;
3570 err = iflib_encap(txq, mp);
3571 if (__predict_false(err)) {
3572 /* no room - bail out */
3576 /* we can't send this packet - skip it */
3582 DBG_COUNTER_INC(tx_sent);
3583 bytes_sent += m->m_pkthdr.len;
3584 mcast_sent += !!(m->m_flags & M_MCAST);
3585 avail = TXQ_AVAIL(txq);
3587 txq->ift_db_pending += (txq->ift_in_use - in_use_prev);
3588 desc_used += (txq->ift_in_use - in_use_prev);
3589 ETHER_BPF_MTAP(ifp, m);
3590 if (__predict_false(!(ifp->if_drv_flags & IFF_DRV_RUNNING)))
3592 rang = iflib_txd_db_check(ctx, txq, false, in_use_prev);
3595 /* deliberate use of bitwise or to avoid gratuitous short-circuit */
3596 ring = rang ? false : (iflib_min_tx_latency | err) || (TXQ_AVAIL(txq) < MAX_TX_DESC(ctx));
3597 iflib_txd_db_check(ctx, txq, ring, txq->ift_in_use);
3598 if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
3599 if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
3601 if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent);
3603 if (iflib_verbose_debug)
3604 printf("consumed=%d\n", consumed);
3610 iflib_txq_drain_always(struct ifmp_ring *r)
3616 iflib_txq_drain_free(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
3624 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3626 callout_stop(&txq->ift_timer);
3627 CALLOUT_UNLOCK(txq);
3629 avail = IDXDIFF(pidx, cidx, r->size);
3630 for (i = 0; i < avail; i++) {
3631 mp = _ring_peek_one(r, cidx, i, avail - i);
3632 if (__predict_false(*mp == (struct mbuf *)txq))
3635 DBG_COUNTER_INC(tx_frees);
3637 MPASS(ifmp_ring_is_stalled(r) == 0);
3642 iflib_ifmp_purge(iflib_txq_t txq)
3644 struct ifmp_ring *r;
3647 r->drain = iflib_txq_drain_free;
3648 r->can_drain = iflib_txq_drain_always;
3650 ifmp_ring_check_drainage(r, r->size);
3652 r->drain = iflib_txq_drain;
3653 r->can_drain = iflib_txq_can_drain;
3657 _task_fn_tx(void *context)
3659 iflib_txq_t txq = context;
3660 if_ctx_t ctx = txq->ift_ctx;
3661 #if defined(ALTQ) || defined(DEV_NETMAP)
3662 if_t ifp = ctx->ifc_ifp;
3664 int abdicate = ctx->ifc_sysctl_tx_abdicate;
3666 #ifdef IFLIB_DIAGNOSTICS
3667 txq->ift_cpu_exec_count[curcpu]++;
3669 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))
3672 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
3673 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
3674 BUS_DMASYNC_POSTREAD);
3675 if (ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, false))
3676 netmap_tx_irq(ifp, txq->ift_id);
3677 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id);
3682 if (ALTQ_IS_ENABLED(&ifp->if_snd))
3683 iflib_altq_if_start(ifp);
3685 if (txq->ift_db_pending)
3686 ifmp_ring_enqueue(txq->ift_br, (void **)&txq, 1, TX_BATCH_SIZE, abdicate);
3688 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3690 * When abdicating, we always need to check drainage, not just when we don't enqueue
3693 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3694 if (ctx->ifc_flags & IFC_LEGACY)
3695 IFDI_INTR_ENABLE(ctx);
3700 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id);
3701 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver"));
3706 _task_fn_rx(void *context)
3708 iflib_rxq_t rxq = context;
3709 if_ctx_t ctx = rxq->ifr_ctx;
3713 #ifdef IFLIB_DIAGNOSTICS
3714 rxq->ifr_cpu_exec_count[curcpu]++;
3716 DBG_COUNTER_INC(task_fn_rxs);
3717 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)))
3721 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP) {
3723 if (netmap_rx_irq(ctx->ifc_ifp, rxq->ifr_id, &work)) {
3728 budget = ctx->ifc_sysctl_rx_budget;
3730 budget = 16; /* XXX */
3731 if (more == false || (more = iflib_rxeof(rxq, budget)) == false) {
3732 if (ctx->ifc_flags & IFC_LEGACY)
3733 IFDI_INTR_ENABLE(ctx);
3738 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id);
3739 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver"));
3740 DBG_COUNTER_INC(rx_intr_enables);
3743 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)))
3746 GROUPTASK_ENQUEUE(&rxq->ifr_task);
3750 _task_fn_admin(void *context)
3752 if_ctx_t ctx = context;
3753 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
3756 bool oactive, running, do_reset, do_watchdog, in_detach;
3757 uint32_t reset_on = hz / 2;
3760 running = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING);
3761 oactive = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE);
3762 do_reset = (ctx->ifc_flags & IFC_DO_RESET);
3763 do_watchdog = (ctx->ifc_flags & IFC_DO_WATCHDOG);
3764 in_detach = (ctx->ifc_flags & IFC_IN_DETACH);
3765 ctx->ifc_flags &= ~(IFC_DO_RESET|IFC_DO_WATCHDOG);
3768 if ((!running && !oactive) && !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN))
3774 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) {
3776 callout_stop(&txq->ift_timer);
3777 CALLOUT_UNLOCK(txq);
3780 ctx->ifc_watchdog_events++;
3781 IFDI_WATCHDOG_RESET(ctx);
3783 IFDI_UPDATE_ADMIN_STATUS(ctx);
3784 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) {
3787 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP)
3788 iflib_netmap_timer_adjust(ctx, txq, &reset_on);
3790 callout_reset_on(&txq->ift_timer, reset_on, iflib_timer, txq, txq->ift_timer.c_cpu);
3792 IFDI_LINK_INTR_ENABLE(ctx);
3794 iflib_if_init_locked(ctx);
3797 if (LINK_ACTIVE(ctx) == 0)
3799 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++)
3800 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET);
3805 _task_fn_iov(void *context)
3807 if_ctx_t ctx = context;
3809 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) &&
3810 !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN))
3814 IFDI_VFLR_HANDLE(ctx);
3819 iflib_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
3822 if_int_delay_info_t info;
3825 info = (if_int_delay_info_t)arg1;
3826 ctx = info->iidi_ctx;
3827 info->iidi_req = req;
3828 info->iidi_oidp = oidp;
3830 err = IFDI_SYSCTL_INT_DELAY(ctx, info);
3835 /*********************************************************************
3839 **********************************************************************/
3842 iflib_if_init_locked(if_ctx_t ctx)
3845 iflib_init_locked(ctx);
3850 iflib_if_init(void *arg)
3855 iflib_if_init_locked(ctx);
3860 iflib_if_transmit(if_t ifp, struct mbuf *m)
3862 if_ctx_t ctx = if_getsoftc(ifp);
3866 int abdicate = ctx->ifc_sysctl_tx_abdicate;
3868 if (__predict_false((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || !LINK_ACTIVE(ctx))) {
3869 DBG_COUNTER_INC(tx_frees);
3874 MPASS(m->m_nextpkt == NULL);
3875 /* ALTQ-enabled interfaces always use queue 0. */
3877 if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m) && !ALTQ_IS_ENABLED(&ifp->if_snd))
3878 qidx = QIDX(ctx, m);
3880 * XXX calculate buf_ring based on flowid (divvy up bits?)
3882 txq = &ctx->ifc_txqs[qidx];
3884 #ifdef DRIVER_BACKPRESSURE
3885 if (txq->ift_closed) {
3887 next = m->m_nextpkt;
3888 m->m_nextpkt = NULL;
3890 DBG_COUNTER_INC(tx_frees);
3902 next = next->m_nextpkt;
3903 } while (next != NULL);
3905 if (count > nitems(marr))
3906 if ((mp = malloc(count*sizeof(struct mbuf *), M_IFLIB, M_NOWAIT)) == NULL) {
3907 /* XXX check nextpkt */
3909 /* XXX simplify for now */
3910 DBG_COUNTER_INC(tx_frees);
3913 for (next = m, i = 0; next != NULL; i++) {
3915 next = next->m_nextpkt;
3916 mp[i]->m_nextpkt = NULL;
3919 DBG_COUNTER_INC(tx_seen);
3920 err = ifmp_ring_enqueue(txq->ift_br, (void **)&m, 1, TX_BATCH_SIZE, abdicate);
3923 GROUPTASK_ENQUEUE(&txq->ift_task);
3926 GROUPTASK_ENQUEUE(&txq->ift_task);
3927 /* support forthcoming later */
3928 #ifdef DRIVER_BACKPRESSURE
3929 txq->ift_closed = TRUE;
3931 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3933 DBG_COUNTER_INC(tx_frees);
3941 * The overall approach to integrating iflib with ALTQ is to continue to use
3942 * the iflib mp_ring machinery between the ALTQ queue(s) and the hardware
3943 * ring. Technically, when using ALTQ, queueing to an intermediate mp_ring
3944 * is redundant/unnecessary, but doing so minimizes the amount of
3945 * ALTQ-specific code required in iflib. It is assumed that the overhead of
3946 * redundantly queueing to an intermediate mp_ring is swamped by the
3947 * performance limitations inherent in using ALTQ.
3949 * When ALTQ support is compiled in, all iflib drivers will use a transmit
3950 * routine, iflib_altq_if_transmit(), that checks if ALTQ is enabled for the
3951 * given interface. If ALTQ is enabled for an interface, then all
3952 * transmitted packets for that interface will be submitted to the ALTQ
3953 * subsystem via IFQ_ENQUEUE(). We don't use the legacy if_transmit()
3954 * implementation because it uses IFQ_HANDOFF(), which will duplicatively
3955 * update stats that the iflib machinery handles, and which is sensitve to
3956 * the disused IFF_DRV_OACTIVE flag. Additionally, iflib_altq_if_start()
3957 * will be installed as the start routine for use by ALTQ facilities that
3958 * need to trigger queue drains on a scheduled basis.
3962 iflib_altq_if_start(if_t ifp)
3964 struct ifaltq *ifq = &ifp->if_snd;
3968 IFQ_DEQUEUE_NOLOCK(ifq, m);
3970 iflib_if_transmit(ifp, m);
3971 IFQ_DEQUEUE_NOLOCK(ifq, m);
3977 iflib_altq_if_transmit(if_t ifp, struct mbuf *m)
3981 if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
3982 IFQ_ENQUEUE(&ifp->if_snd, m, err);
3984 iflib_altq_if_start(ifp);
3986 err = iflib_if_transmit(ifp, m);
3993 iflib_if_qflush(if_t ifp)
3995 if_ctx_t ctx = if_getsoftc(ifp);
3996 iflib_txq_t txq = ctx->ifc_txqs;
4000 ctx->ifc_flags |= IFC_QFLUSH;
4002 for (i = 0; i < NTXQSETS(ctx); i++, txq++)
4003 while (!(ifmp_ring_is_idle(txq->ift_br) || ifmp_ring_is_stalled(txq->ift_br)))
4004 iflib_txq_check_drain(txq, 0);
4006 ctx->ifc_flags &= ~IFC_QFLUSH;
4010 * When ALTQ is enabled, this will also take care of purging the
4017 #define IFCAP_FLAGS (IFCAP_HWCSUM_IPV6 | IFCAP_HWCSUM | IFCAP_LRO | \
4018 IFCAP_TSO | IFCAP_VLAN_HWTAGGING | IFCAP_HWSTATS | \
4019 IFCAP_VLAN_MTU | IFCAP_VLAN_HWFILTER | \
4020 IFCAP_VLAN_HWTSO | IFCAP_VLAN_HWCSUM)
4023 iflib_if_ioctl(if_t ifp, u_long command, caddr_t data)
4025 if_ctx_t ctx = if_getsoftc(ifp);
4026 struct ifreq *ifr = (struct ifreq *)data;
4027 #if defined(INET) || defined(INET6)
4028 struct ifaddr *ifa = (struct ifaddr *)data;
4030 bool avoid_reset = FALSE;
4031 int err = 0, reinit = 0, bits;
4036 if (ifa->ifa_addr->sa_family == AF_INET)
4040 if (ifa->ifa_addr->sa_family == AF_INET6)
4044 ** Calling init results in link renegotiation,
4045 ** so we avoid doing it when possible.
4048 if_setflagbits(ifp, IFF_UP,0);
4049 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
4052 if (!(if_getflags(ifp) & IFF_NOARP))
4053 arp_ifinit(ifp, ifa);
4056 err = ether_ioctl(ifp, command, data);
4060 if (ifr->ifr_mtu == if_getmtu(ifp)) {
4064 bits = if_getdrvflags(ifp);
4065 /* stop the driver and free any clusters before proceeding */
4068 if ((err = IFDI_MTU_SET(ctx, ifr->ifr_mtu)) == 0) {
4070 if (ifr->ifr_mtu > ctx->ifc_max_fl_buf_size)
4071 ctx->ifc_flags |= IFC_MULTISEG;
4073 ctx->ifc_flags &= ~IFC_MULTISEG;
4075 err = if_setmtu(ifp, ifr->ifr_mtu);
4077 iflib_init_locked(ctx);
4079 if_setdrvflags(ifp, bits);
4085 if (if_getflags(ifp) & IFF_UP) {
4086 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4087 if ((if_getflags(ifp) ^ ctx->ifc_if_flags) &
4088 (IFF_PROMISC | IFF_ALLMULTI)) {
4089 err = IFDI_PROMISC_SET(ctx, if_getflags(ifp));
4093 } else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4096 ctx->ifc_if_flags = if_getflags(ifp);
4101 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4103 IFDI_INTR_DISABLE(ctx);
4104 IFDI_MULTI_SET(ctx);
4105 IFDI_INTR_ENABLE(ctx);
4111 IFDI_MEDIA_SET(ctx);
4116 err = ifmedia_ioctl(ifp, ifr, &ctx->ifc_media, command);
4120 struct ifi2creq i2c;
4122 err = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
4125 if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) {
4129 if (i2c.len > sizeof(i2c.data)) {
4134 if ((err = IFDI_I2C_REQ(ctx, &i2c)) == 0)
4135 err = copyout(&i2c, ifr_data_get_ptr(ifr),
4141 int mask, setmask, oldmask;
4143 oldmask = if_getcapenable(ifp);
4144 mask = ifr->ifr_reqcap ^ oldmask;
4145 mask &= ctx->ifc_softc_ctx.isc_capabilities;
4148 setmask |= mask & (IFCAP_TOE4|IFCAP_TOE6);
4150 setmask |= (mask & IFCAP_FLAGS);
4151 setmask |= (mask & IFCAP_WOL);
4154 * If any RX csum has changed, change all the ones that
4155 * are supported by the driver.
4157 if (setmask & (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) {
4158 setmask |= ctx->ifc_softc_ctx.isc_capabilities &
4159 (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6);
4163 * want to ensure that traffic has stopped before we change any of the flags
4167 bits = if_getdrvflags(ifp);
4168 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL)
4171 if_togglecapenable(ifp, setmask);
4173 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL)
4174 iflib_init_locked(ctx);
4176 if_setdrvflags(ifp, bits);
4183 case SIOCGPRIVATE_0:
4187 err = IFDI_PRIV_IOCTL(ctx, command, data);
4191 err = ether_ioctl(ifp, command, data);
4200 iflib_if_get_counter(if_t ifp, ift_counter cnt)
4202 if_ctx_t ctx = if_getsoftc(ifp);
4204 return (IFDI_GET_COUNTER(ctx, cnt));
4207 /*********************************************************************
4209 * OTHER FUNCTIONS EXPORTED TO THE STACK
4211 **********************************************************************/
4214 iflib_vlan_register(void *arg, if_t ifp, uint16_t vtag)
4216 if_ctx_t ctx = if_getsoftc(ifp);
4218 if ((void *)ctx != arg)
4221 if ((vtag == 0) || (vtag > 4095))
4225 IFDI_VLAN_REGISTER(ctx, vtag);
4226 /* Re-init to load the changes */
4227 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
4228 iflib_if_init_locked(ctx);
4233 iflib_vlan_unregister(void *arg, if_t ifp, uint16_t vtag)
4235 if_ctx_t ctx = if_getsoftc(ifp);
4237 if ((void *)ctx != arg)
4240 if ((vtag == 0) || (vtag > 4095))
4244 IFDI_VLAN_UNREGISTER(ctx, vtag);
4245 /* Re-init to load the changes */
4246 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
4247 iflib_if_init_locked(ctx);
4252 iflib_led_func(void *arg, int onoff)
4257 IFDI_LED_FUNC(ctx, onoff);
4261 /*********************************************************************
4263 * BUS FUNCTION DEFINITIONS
4265 **********************************************************************/
4268 iflib_device_probe(device_t dev)
4270 pci_vendor_info_t *ent;
4272 uint16_t pci_vendor_id, pci_device_id;
4273 uint16_t pci_subvendor_id, pci_subdevice_id;
4274 uint16_t pci_rev_id;
4275 if_shared_ctx_t sctx;
4277 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC)
4280 pci_vendor_id = pci_get_vendor(dev);
4281 pci_device_id = pci_get_device(dev);
4282 pci_subvendor_id = pci_get_subvendor(dev);
4283 pci_subdevice_id = pci_get_subdevice(dev);
4284 pci_rev_id = pci_get_revid(dev);
4285 if (sctx->isc_parse_devinfo != NULL)
4286 sctx->isc_parse_devinfo(&pci_device_id, &pci_subvendor_id, &pci_subdevice_id, &pci_rev_id);
4288 ent = sctx->isc_vendor_info;
4289 while (ent->pvi_vendor_id != 0) {
4290 if (pci_vendor_id != ent->pvi_vendor_id) {
4294 if ((pci_device_id == ent->pvi_device_id) &&
4295 ((pci_subvendor_id == ent->pvi_subvendor_id) ||
4296 (ent->pvi_subvendor_id == 0)) &&
4297 ((pci_subdevice_id == ent->pvi_subdevice_id) ||
4298 (ent->pvi_subdevice_id == 0)) &&
4299 ((pci_rev_id == ent->pvi_rev_id) ||
4300 (ent->pvi_rev_id == 0))) {
4302 device_set_desc_copy(dev, ent->pvi_name);
4303 /* this needs to be changed to zero if the bus probing code
4304 * ever stops re-probing on best match because the sctx
4305 * may have its values over written by register calls
4306 * in subsequent probes
4308 return (BUS_PROBE_DEFAULT);
4316 iflib_reset_qvalues(if_ctx_t ctx)
4318 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
4319 if_shared_ctx_t sctx = ctx->ifc_sctx;
4320 device_t dev = ctx->ifc_dev;
4323 scctx->isc_txrx_budget_bytes_max = IFLIB_MAX_TX_BYTES;
4324 scctx->isc_tx_qdepth = IFLIB_DEFAULT_TX_QDEPTH;
4326 * XXX sanity check that ntxd & nrxd are a power of 2
4328 if (ctx->ifc_sysctl_ntxqs != 0)
4329 scctx->isc_ntxqsets = ctx->ifc_sysctl_ntxqs;
4330 if (ctx->ifc_sysctl_nrxqs != 0)
4331 scctx->isc_nrxqsets = ctx->ifc_sysctl_nrxqs;
4333 for (i = 0; i < sctx->isc_ntxqs; i++) {
4334 if (ctx->ifc_sysctl_ntxds[i] != 0)
4335 scctx->isc_ntxd[i] = ctx->ifc_sysctl_ntxds[i];
4337 scctx->isc_ntxd[i] = sctx->isc_ntxd_default[i];
4340 for (i = 0; i < sctx->isc_nrxqs; i++) {
4341 if (ctx->ifc_sysctl_nrxds[i] != 0)
4342 scctx->isc_nrxd[i] = ctx->ifc_sysctl_nrxds[i];
4344 scctx->isc_nrxd[i] = sctx->isc_nrxd_default[i];
4347 for (i = 0; i < sctx->isc_nrxqs; i++) {
4348 if (scctx->isc_nrxd[i] < sctx->isc_nrxd_min[i]) {
4349 device_printf(dev, "nrxd%d: %d less than nrxd_min %d - resetting to min\n",
4350 i, scctx->isc_nrxd[i], sctx->isc_nrxd_min[i]);
4351 scctx->isc_nrxd[i] = sctx->isc_nrxd_min[i];
4353 if (scctx->isc_nrxd[i] > sctx->isc_nrxd_max[i]) {
4354 device_printf(dev, "nrxd%d: %d greater than nrxd_max %d - resetting to max\n",
4355 i, scctx->isc_nrxd[i], sctx->isc_nrxd_max[i]);
4356 scctx->isc_nrxd[i] = sctx->isc_nrxd_max[i];
4360 for (i = 0; i < sctx->isc_ntxqs; i++) {
4361 if (scctx->isc_ntxd[i] < sctx->isc_ntxd_min[i]) {
4362 device_printf(dev, "ntxd%d: %d less than ntxd_min %d - resetting to min\n",
4363 i, scctx->isc_ntxd[i], sctx->isc_ntxd_min[i]);
4364 scctx->isc_ntxd[i] = sctx->isc_ntxd_min[i];
4366 if (scctx->isc_ntxd[i] > sctx->isc_ntxd_max[i]) {
4367 device_printf(dev, "ntxd%d: %d greater than ntxd_max %d - resetting to max\n",
4368 i, scctx->isc_ntxd[i], sctx->isc_ntxd_max[i]);
4369 scctx->isc_ntxd[i] = sctx->isc_ntxd_max[i];
4375 iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ctxp)
4380 if_softc_ctx_t scctx;
4386 ctx = malloc(sizeof(* ctx), M_IFLIB, M_WAITOK|M_ZERO);
4389 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK|M_ZERO);
4390 device_set_softc(dev, ctx);
4391 ctx->ifc_flags |= IFC_SC_ALLOCATED;
4394 ctx->ifc_sctx = sctx;
4396 ctx->ifc_softc = sc;
4398 if ((err = iflib_register(ctx)) != 0) {
4399 device_printf(dev, "iflib_register failed %d\n", err);
4402 iflib_add_device_sysctl_pre(ctx);
4404 scctx = &ctx->ifc_softc_ctx;
4407 iflib_reset_qvalues(ctx);
4409 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) {
4410 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err);
4413 _iflib_pre_assert(scctx);
4414 ctx->ifc_txrx = *scctx->isc_txrx;
4417 MPASS(scctx->isc_capabilities);
4418 if (scctx->isc_capabilities & IFCAP_TXCSUM)
4419 MPASS(scctx->isc_tx_csum_flags);
4422 if_setcapabilities(ifp, scctx->isc_capabilities | IFCAP_HWSTATS);
4423 if_setcapenable(ifp, scctx->isc_capenable | IFCAP_HWSTATS);
4425 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets))
4426 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max;
4427 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets))
4428 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max;
4430 main_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0;
4431 main_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0;
4433 /* XXX change for per-queue sizes */
4434 device_printf(dev, "Using %d tx descriptors and %d rx descriptors\n",
4435 scctx->isc_ntxd[main_txq], scctx->isc_nrxd[main_rxq]);
4436 for (i = 0; i < sctx->isc_nrxqs; i++) {
4437 if (!powerof2(scctx->isc_nrxd[i])) {
4438 /* round down instead? */
4439 device_printf(dev, "# rx descriptors must be a power of 2\n");
4441 goto fail_iflib_detach;
4444 for (i = 0; i < sctx->isc_ntxqs; i++) {
4445 if (!powerof2(scctx->isc_ntxd[i])) {
4447 "# tx descriptors must be a power of 2");
4449 goto fail_iflib_detach;
4453 if (scctx->isc_tx_nsegments > scctx->isc_ntxd[main_txq] /
4454 MAX_SINGLE_PACKET_FRACTION)
4455 scctx->isc_tx_nsegments = max(1, scctx->isc_ntxd[main_txq] /
4456 MAX_SINGLE_PACKET_FRACTION);
4457 if (scctx->isc_tx_tso_segments_max > scctx->isc_ntxd[main_txq] /
4458 MAX_SINGLE_PACKET_FRACTION)
4459 scctx->isc_tx_tso_segments_max = max(1,
4460 scctx->isc_ntxd[main_txq] / MAX_SINGLE_PACKET_FRACTION);
4462 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */
4463 if (if_getcapabilities(ifp) & IFCAP_TSO) {
4465 * The stack can't handle a TSO size larger than IP_MAXPACKET,
4468 if_sethwtsomax(ifp, min(scctx->isc_tx_tso_size_max,
4471 * Take maximum number of m_pullup(9)'s in iflib_parse_header()
4472 * into account. In the worst case, each of these calls will
4473 * add another mbuf and, thus, the requirement for another DMA
4474 * segment. So for best performance, it doesn't make sense to
4475 * advertize a maximum of TSO segments that typically will
4476 * require defragmentation in iflib_encap().
4478 if_sethwtsomaxsegcount(ifp, scctx->isc_tx_tso_segments_max - 3);
4479 if_sethwtsomaxsegsize(ifp, scctx->isc_tx_tso_segsize_max);
4481 if (scctx->isc_rss_table_size == 0)
4482 scctx->isc_rss_table_size = 64;
4483 scctx->isc_rss_table_mask = scctx->isc_rss_table_size-1;
4485 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx);
4486 /* XXX format name */
4487 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx,
4488 NULL, NULL, "admin");
4490 /* Set up cpu set. If it fails, use the set of all CPUs. */
4491 if (bus_get_cpus(dev, INTR_CPUS, sizeof(ctx->ifc_cpus), &ctx->ifc_cpus) != 0) {
4492 device_printf(dev, "Unable to fetch CPU list\n");
4493 CPU_COPY(&all_cpus, &ctx->ifc_cpus);
4495 MPASS(CPU_COUNT(&ctx->ifc_cpus) > 0);
4498 ** Now set up MSI or MSI-X, should return us the number of supported
4499 ** vectors (will be 1 for a legacy interrupt and MSI).
4501 if (sctx->isc_flags & IFLIB_SKIP_MSIX) {
4502 msix = scctx->isc_vectors;
4503 } else if (scctx->isc_msix_bar != 0)
4505 * The simple fact that isc_msix_bar is not 0 does not mean we
4506 * we have a good value there that is known to work.
4508 msix = iflib_msix_init(ctx);
4510 scctx->isc_vectors = 1;
4511 scctx->isc_ntxqsets = 1;
4512 scctx->isc_nrxqsets = 1;
4513 scctx->isc_intr = IFLIB_INTR_LEGACY;
4516 /* Get memory for the station queues */
4517 if ((err = iflib_queues_alloc(ctx))) {
4518 device_printf(dev, "Unable to allocate queue memory\n");
4519 goto fail_intr_free;
4522 if ((err = iflib_qset_structures_setup(ctx)))
4526 * Group taskqueues aren't properly set up until SMP is started,
4527 * so we disable interrupts until we can handle them post
4530 * XXX: disabling interrupts doesn't actually work, at least for
4531 * the non-MSI case. When they occur before SI_SUB_SMP completes,
4532 * we do null handling and depend on this not causing too large an
4535 IFDI_INTR_DISABLE(ctx);
4536 if (msix > 1 && (err = IFDI_MSIX_INTR_ASSIGN(ctx, msix)) != 0) {
4537 device_printf(dev, "IFDI_MSIX_INTR_ASSIGN failed %d\n", err);
4542 if (scctx->isc_intr == IFLIB_INTR_MSI) {
4546 if ((err = iflib_legacy_setup(ctx, ctx->isc_legacy_intr, ctx->ifc_softc, &rid, "irq0")) != 0) {
4547 device_printf(dev, "iflib_legacy_setup failed %d\n", err);
4552 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4554 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4555 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4560 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4561 * This must appear after the call to ether_ifattach() because
4562 * ether_ifattach() sets if_hdrlen to the default value.
4564 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4565 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
4567 if ((err = iflib_netmap_attach(ctx))) {
4568 device_printf(ctx->ifc_dev, "netmap attach failed: %d\n", err);
4573 NETDUMP_SET(ctx->ifc_ifp, iflib);
4575 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4576 iflib_add_device_sysctl_post(ctx);
4577 ctx->ifc_flags |= IFC_INIT_DONE;
4582 ether_ifdetach(ctx->ifc_ifp);
4584 iflib_free_intr_mem(ctx);
4586 iflib_tx_structures_free(ctx);
4587 iflib_rx_structures_free(ctx);
4593 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4594 free(ctx->ifc_softc, M_IFLIB);
4600 iflib_pseudo_register(device_t dev, if_shared_ctx_t sctx, if_ctx_t *ctxp,
4601 struct iflib_cloneattach_ctx *clctx)
4606 if_softc_ctx_t scctx;
4612 ctx = malloc(sizeof(*ctx), M_IFLIB, M_WAITOK|M_ZERO);
4613 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK|M_ZERO);
4614 ctx->ifc_flags |= IFC_SC_ALLOCATED;
4615 if (sctx->isc_flags & (IFLIB_PSEUDO|IFLIB_VIRTUAL))
4616 ctx->ifc_flags |= IFC_PSEUDO;
4618 ctx->ifc_sctx = sctx;
4619 ctx->ifc_softc = sc;
4622 if ((err = iflib_register(ctx)) != 0) {
4623 device_printf(dev, "%s: iflib_register failed %d\n", __func__, err);
4626 iflib_add_device_sysctl_pre(ctx);
4628 scctx = &ctx->ifc_softc_ctx;
4632 * XXX sanity check that ntxd & nrxd are a power of 2
4634 iflib_reset_qvalues(ctx);
4636 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) {
4637 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err);
4640 if (sctx->isc_flags & IFLIB_GEN_MAC)
4642 if ((err = IFDI_CLONEATTACH(ctx, clctx->cc_ifc, clctx->cc_name,
4643 clctx->cc_params)) != 0) {
4644 device_printf(dev, "IFDI_CLONEATTACH failed %d\n", err);
4647 ifmedia_add(&ctx->ifc_media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
4648 ifmedia_add(&ctx->ifc_media, IFM_ETHER | IFM_AUTO, 0, NULL);
4649 ifmedia_set(&ctx->ifc_media, IFM_ETHER | IFM_AUTO);
4652 MPASS(scctx->isc_capabilities);
4653 if (scctx->isc_capabilities & IFCAP_TXCSUM)
4654 MPASS(scctx->isc_tx_csum_flags);
4657 if_setcapabilities(ifp, scctx->isc_capabilities | IFCAP_HWSTATS | IFCAP_LINKSTATE);
4658 if_setcapenable(ifp, scctx->isc_capenable | IFCAP_HWSTATS | IFCAP_LINKSTATE);
4660 ifp->if_flags |= IFF_NOGROUP;
4661 if (sctx->isc_flags & IFLIB_PSEUDO) {
4662 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4664 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4665 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4671 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4672 * This must appear after the call to ether_ifattach() because
4673 * ether_ifattach() sets if_hdrlen to the default value.
4675 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4676 if_setifheaderlen(ifp,
4677 sizeof(struct ether_vlan_header));
4679 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4680 iflib_add_device_sysctl_post(ctx);
4681 ctx->ifc_flags |= IFC_INIT_DONE;
4684 _iflib_pre_assert(scctx);
4685 ctx->ifc_txrx = *scctx->isc_txrx;
4687 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets))
4688 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max;
4689 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets))
4690 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max;
4692 main_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0;
4693 main_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0;
4695 /* XXX change for per-queue sizes */
4696 device_printf(dev, "Using %d tx descriptors and %d rx descriptors\n",
4697 scctx->isc_ntxd[main_txq], scctx->isc_nrxd[main_rxq]);
4698 for (i = 0; i < sctx->isc_nrxqs; i++) {
4699 if (!powerof2(scctx->isc_nrxd[i])) {
4700 /* round down instead? */
4701 device_printf(dev, "# rx descriptors must be a power of 2\n");
4703 goto fail_iflib_detach;
4706 for (i = 0; i < sctx->isc_ntxqs; i++) {
4707 if (!powerof2(scctx->isc_ntxd[i])) {
4709 "# tx descriptors must be a power of 2");
4711 goto fail_iflib_detach;
4715 if (scctx->isc_tx_nsegments > scctx->isc_ntxd[main_txq] /
4716 MAX_SINGLE_PACKET_FRACTION)
4717 scctx->isc_tx_nsegments = max(1, scctx->isc_ntxd[main_txq] /
4718 MAX_SINGLE_PACKET_FRACTION);
4719 if (scctx->isc_tx_tso_segments_max > scctx->isc_ntxd[main_txq] /
4720 MAX_SINGLE_PACKET_FRACTION)
4721 scctx->isc_tx_tso_segments_max = max(1,
4722 scctx->isc_ntxd[main_txq] / MAX_SINGLE_PACKET_FRACTION);
4724 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */
4725 if (if_getcapabilities(ifp) & IFCAP_TSO) {
4727 * The stack can't handle a TSO size larger than IP_MAXPACKET,
4730 if_sethwtsomax(ifp, min(scctx->isc_tx_tso_size_max,
4733 * Take maximum number of m_pullup(9)'s in iflib_parse_header()
4734 * into account. In the worst case, each of these calls will
4735 * add another mbuf and, thus, the requirement for another DMA
4736 * segment. So for best performance, it doesn't make sense to
4737 * advertize a maximum of TSO segments that typically will
4738 * require defragmentation in iflib_encap().
4740 if_sethwtsomaxsegcount(ifp, scctx->isc_tx_tso_segments_max - 3);
4741 if_sethwtsomaxsegsize(ifp, scctx->isc_tx_tso_segsize_max);
4743 if (scctx->isc_rss_table_size == 0)
4744 scctx->isc_rss_table_size = 64;
4745 scctx->isc_rss_table_mask = scctx->isc_rss_table_size-1;
4747 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx);
4748 /* XXX format name */
4749 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx,
4750 NULL, NULL, "admin");
4752 /* XXX --- can support > 1 -- but keep it simple for now */
4753 scctx->isc_intr = IFLIB_INTR_LEGACY;
4755 /* Get memory for the station queues */
4756 if ((err = iflib_queues_alloc(ctx))) {
4757 device_printf(dev, "Unable to allocate queue memory\n");
4758 goto fail_iflib_detach;
4761 if ((err = iflib_qset_structures_setup(ctx))) {
4762 device_printf(dev, "qset structure setup failed %d\n", err);
4767 * XXX What if anything do we want to do about interrupts?
4769 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4770 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4771 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4776 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4777 * This must appear after the call to ether_ifattach() because
4778 * ether_ifattach() sets if_hdrlen to the default value.
4780 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4781 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
4783 /* XXX handle more than one queue */
4784 for (i = 0; i < scctx->isc_nrxqsets; i++)
4785 IFDI_RX_CLSET(ctx, 0, i, ctx->ifc_rxqs[i].ifr_fl[0].ifl_sds.ifsd_cl);
4789 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4790 iflib_add_device_sysctl_post(ctx);
4791 ctx->ifc_flags |= IFC_INIT_DONE;
4794 ether_ifdetach(ctx->ifc_ifp);
4796 iflib_tx_structures_free(ctx);
4797 iflib_rx_structures_free(ctx);
4801 free(ctx->ifc_softc, M_IFLIB);
4807 iflib_pseudo_deregister(if_ctx_t ctx)
4809 if_t ifp = ctx->ifc_ifp;
4813 struct taskqgroup *tqg;
4816 /* Unregister VLAN events */
4817 if (ctx->ifc_vlan_attach_event != NULL)
4818 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event);
4819 if (ctx->ifc_vlan_detach_event != NULL)
4820 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event);
4822 ether_ifdetach(ifp);
4823 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/
4824 CTX_LOCK_DESTROY(ctx);
4825 /* XXX drain any dependent tasks */
4826 tqg = qgroup_if_io_tqg;
4827 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) {
4828 callout_drain(&txq->ift_timer);
4829 if (txq->ift_task.gt_uniq != NULL)
4830 taskqgroup_detach(tqg, &txq->ift_task);
4832 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) {
4833 if (rxq->ifr_task.gt_uniq != NULL)
4834 taskqgroup_detach(tqg, &rxq->ifr_task);
4836 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
4837 free(fl->ifl_rx_bitmap, M_IFLIB);
4839 tqg = qgroup_if_config_tqg;
4840 if (ctx->ifc_admin_task.gt_uniq != NULL)
4841 taskqgroup_detach(tqg, &ctx->ifc_admin_task);
4842 if (ctx->ifc_vflr_task.gt_uniq != NULL)
4843 taskqgroup_detach(tqg, &ctx->ifc_vflr_task);
4847 iflib_tx_structures_free(ctx);
4848 iflib_rx_structures_free(ctx);
4849 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4850 free(ctx->ifc_softc, M_IFLIB);
4856 iflib_device_attach(device_t dev)
4859 if_shared_ctx_t sctx;
4861 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC)
4864 pci_enable_busmaster(dev);
4866 return (iflib_device_register(dev, NULL, sctx, &ctx));
4870 iflib_device_deregister(if_ctx_t ctx)
4872 if_t ifp = ctx->ifc_ifp;
4875 device_t dev = ctx->ifc_dev;
4877 struct taskqgroup *tqg;
4880 /* Make sure VLANS are not using driver */
4881 if (if_vlantrunkinuse(ifp)) {
4882 device_printf(dev, "Vlan in use, detach first\n");
4886 if (!CTX_IS_VF(ctx) && pci_iov_detach(dev) != 0) {
4887 device_printf(dev, "SR-IOV in use; detach first.\n");
4893 ctx->ifc_flags |= IFC_IN_DETACH;
4900 /* Unregister VLAN events */
4901 if (ctx->ifc_vlan_attach_event != NULL)
4902 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event);
4903 if (ctx->ifc_vlan_detach_event != NULL)
4904 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event);
4906 iflib_netmap_detach(ifp);
4907 ether_ifdetach(ifp);
4908 if (ctx->ifc_led_dev != NULL)
4909 led_destroy(ctx->ifc_led_dev);
4910 /* XXX drain any dependent tasks */
4911 tqg = qgroup_if_io_tqg;
4912 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) {
4913 callout_drain(&txq->ift_timer);
4914 if (txq->ift_task.gt_uniq != NULL)
4915 taskqgroup_detach(tqg, &txq->ift_task);
4917 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) {
4918 if (rxq->ifr_task.gt_uniq != NULL)
4919 taskqgroup_detach(tqg, &rxq->ifr_task);
4921 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
4922 free(fl->ifl_rx_bitmap, M_IFLIB);
4924 tqg = qgroup_if_config_tqg;
4925 if (ctx->ifc_admin_task.gt_uniq != NULL)
4926 taskqgroup_detach(tqg, &ctx->ifc_admin_task);
4927 if (ctx->ifc_vflr_task.gt_uniq != NULL)
4928 taskqgroup_detach(tqg, &ctx->ifc_vflr_task);
4933 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/
4934 CTX_LOCK_DESTROY(ctx);
4935 device_set_softc(ctx->ifc_dev, NULL);
4936 iflib_free_intr_mem(ctx);
4938 bus_generic_detach(dev);
4941 iflib_tx_structures_free(ctx);
4942 iflib_rx_structures_free(ctx);
4943 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4944 free(ctx->ifc_softc, M_IFLIB);
4945 STATE_LOCK_DESTROY(ctx);
4951 iflib_free_intr_mem(if_ctx_t ctx)
4954 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) {
4955 iflib_irq_free(ctx, &ctx->ifc_legacy_irq);
4957 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) {
4958 pci_release_msi(ctx->ifc_dev);
4960 if (ctx->ifc_msix_mem != NULL) {
4961 bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY,
4962 rman_get_rid(ctx->ifc_msix_mem), ctx->ifc_msix_mem);
4963 ctx->ifc_msix_mem = NULL;
4968 iflib_device_detach(device_t dev)
4970 if_ctx_t ctx = device_get_softc(dev);
4972 return (iflib_device_deregister(ctx));
4976 iflib_device_suspend(device_t dev)
4978 if_ctx_t ctx = device_get_softc(dev);
4984 return bus_generic_suspend(dev);
4987 iflib_device_shutdown(device_t dev)
4989 if_ctx_t ctx = device_get_softc(dev);
4995 return bus_generic_suspend(dev);
5000 iflib_device_resume(device_t dev)
5002 if_ctx_t ctx = device_get_softc(dev);
5003 iflib_txq_t txq = ctx->ifc_txqs;
5007 iflib_if_init_locked(ctx);
5009 for (int i = 0; i < NTXQSETS(ctx); i++, txq++)
5010 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET);
5012 return (bus_generic_resume(dev));
5016 iflib_device_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
5019 if_ctx_t ctx = device_get_softc(dev);
5022 error = IFDI_IOV_INIT(ctx, num_vfs, params);
5029 iflib_device_iov_uninit(device_t dev)
5031 if_ctx_t ctx = device_get_softc(dev);
5034 IFDI_IOV_UNINIT(ctx);
5039 iflib_device_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
5042 if_ctx_t ctx = device_get_softc(dev);
5045 error = IFDI_IOV_VF_ADD(ctx, vfnum, params);
5051 /*********************************************************************
5053 * MODULE FUNCTION DEFINITIONS
5055 **********************************************************************/
5058 * - Start a fast taskqueue thread for each core
5059 * - Start a taskqueue for control operations
5062 iflib_module_init(void)
5068 iflib_module_event_handler(module_t mod, int what, void *arg)
5074 if ((err = iflib_module_init()) != 0)
5080 return (EOPNOTSUPP);
5086 /*********************************************************************
5088 * PUBLIC FUNCTION DEFINITIONS
5089 * ordered as in iflib.h
5091 **********************************************************************/
5095 _iflib_assert(if_shared_ctx_t sctx)
5097 MPASS(sctx->isc_tx_maxsize);
5098 MPASS(sctx->isc_tx_maxsegsize);
5100 MPASS(sctx->isc_rx_maxsize);
5101 MPASS(sctx->isc_rx_nsegments);
5102 MPASS(sctx->isc_rx_maxsegsize);
5104 MPASS(sctx->isc_nrxd_min[0]);
5105 MPASS(sctx->isc_nrxd_max[0]);
5106 MPASS(sctx->isc_nrxd_default[0]);
5107 MPASS(sctx->isc_ntxd_min[0]);
5108 MPASS(sctx->isc_ntxd_max[0]);
5109 MPASS(sctx->isc_ntxd_default[0]);
5113 _iflib_pre_assert(if_softc_ctx_t scctx)
5116 MPASS(scctx->isc_txrx->ift_txd_encap);
5117 MPASS(scctx->isc_txrx->ift_txd_flush);
5118 MPASS(scctx->isc_txrx->ift_txd_credits_update);
5119 MPASS(scctx->isc_txrx->ift_rxd_available);
5120 MPASS(scctx->isc_txrx->ift_rxd_pkt_get);
5121 MPASS(scctx->isc_txrx->ift_rxd_refill);
5122 MPASS(scctx->isc_txrx->ift_rxd_flush);
5126 iflib_register(if_ctx_t ctx)
5128 if_shared_ctx_t sctx = ctx->ifc_sctx;
5129 driver_t *driver = sctx->isc_driver;
5130 device_t dev = ctx->ifc_dev;
5133 _iflib_assert(sctx);
5136 STATE_LOCK_INIT(ctx, device_get_nameunit(ctx->ifc_dev));
5137 ifp = ctx->ifc_ifp = if_alloc(IFT_ETHER);
5139 device_printf(dev, "can not allocate ifnet structure\n");
5144 * Initialize our context's device specific methods
5146 kobj_init((kobj_t) ctx, (kobj_class_t) driver);
5147 kobj_class_compile((kobj_class_t) driver);
5150 if_initname(ifp, device_get_name(dev), device_get_unit(dev));
5151 if_setsoftc(ifp, ctx);
5152 if_setdev(ifp, dev);
5153 if_setinitfn(ifp, iflib_if_init);
5154 if_setioctlfn(ifp, iflib_if_ioctl);
5156 if_setstartfn(ifp, iflib_altq_if_start);
5157 if_settransmitfn(ifp, iflib_altq_if_transmit);
5158 if_setsendqready(ifp);
5160 if_settransmitfn(ifp, iflib_if_transmit);
5162 if_setqflushfn(ifp, iflib_if_qflush);
5163 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
5165 ctx->ifc_vlan_attach_event =
5166 EVENTHANDLER_REGISTER(vlan_config, iflib_vlan_register, ctx,
5167 EVENTHANDLER_PRI_FIRST);
5168 ctx->ifc_vlan_detach_event =
5169 EVENTHANDLER_REGISTER(vlan_unconfig, iflib_vlan_unregister, ctx,
5170 EVENTHANDLER_PRI_FIRST);
5172 ifmedia_init(&ctx->ifc_media, IFM_IMASK,
5173 iflib_media_change, iflib_media_status);
5180 iflib_queues_alloc(if_ctx_t ctx)
5182 if_shared_ctx_t sctx = ctx->ifc_sctx;
5183 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
5184 device_t dev = ctx->ifc_dev;
5185 int nrxqsets = scctx->isc_nrxqsets;
5186 int ntxqsets = scctx->isc_ntxqsets;
5189 iflib_fl_t fl = NULL;
5190 int i, j, cpu, err, txconf, rxconf;
5191 iflib_dma_info_t ifdip;
5192 uint32_t *rxqsizes = scctx->isc_rxqsizes;
5193 uint32_t *txqsizes = scctx->isc_txqsizes;
5194 uint8_t nrxqs = sctx->isc_nrxqs;
5195 uint8_t ntxqs = sctx->isc_ntxqs;
5196 int nfree_lists = sctx->isc_nfl ? sctx->isc_nfl : 1;
5200 KASSERT(ntxqs > 0, ("number of queues per qset must be at least 1"));
5201 KASSERT(nrxqs > 0, ("number of queues per qset must be at least 1"));
5203 /* Allocate the TX ring struct memory */
5204 if (!(ctx->ifc_txqs =
5205 (iflib_txq_t) malloc(sizeof(struct iflib_txq) *
5206 ntxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) {
5207 device_printf(dev, "Unable to allocate TX ring memory\n");
5212 /* Now allocate the RX */
5213 if (!(ctx->ifc_rxqs =
5214 (iflib_rxq_t) malloc(sizeof(struct iflib_rxq) *
5215 nrxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) {
5216 device_printf(dev, "Unable to allocate RX ring memory\n");
5221 txq = ctx->ifc_txqs;
5222 rxq = ctx->ifc_rxqs;
5225 * XXX handle allocation failure
5227 for (txconf = i = 0, cpu = CPU_FIRST(); i < ntxqsets; i++, txconf++, txq++, cpu = CPU_NEXT(cpu)) {
5228 /* Set up some basics */
5230 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * ntxqs,
5231 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
5233 "Unable to allocate TX DMA info memory\n");
5237 txq->ift_ifdi = ifdip;
5238 for (j = 0; j < ntxqs; j++, ifdip++) {
5239 if (iflib_dma_alloc(ctx, txqsizes[j], ifdip, 0)) {
5241 "Unable to allocate TX descriptors\n");
5245 txq->ift_txd_size[j] = scctx->isc_txd_size[j];
5246 bzero((void *)ifdip->idi_vaddr, txqsizes[j]);
5250 if (sctx->isc_flags & IFLIB_HAS_TXCQ) {
5251 txq->ift_br_offset = 1;
5253 txq->ift_br_offset = 0;
5256 txq->ift_timer.c_cpu = cpu;
5258 if (iflib_txsd_alloc(txq)) {
5259 device_printf(dev, "Critical Failure setting up TX buffers\n");
5264 /* Initialize the TX lock */
5265 snprintf(txq->ift_mtx_name, MTX_NAME_LEN, "%s:tx(%d):callout",
5266 device_get_nameunit(dev), txq->ift_id);
5267 mtx_init(&txq->ift_mtx, txq->ift_mtx_name, NULL, MTX_DEF);
5268 callout_init_mtx(&txq->ift_timer, &txq->ift_mtx, 0);
5270 snprintf(txq->ift_db_mtx_name, MTX_NAME_LEN, "%s:tx(%d):db",
5271 device_get_nameunit(dev), txq->ift_id);
5273 err = ifmp_ring_alloc(&txq->ift_br, 2048, txq, iflib_txq_drain,
5274 iflib_txq_can_drain, M_IFLIB, M_WAITOK);
5276 /* XXX free any allocated rings */
5277 device_printf(dev, "Unable to allocate buf_ring\n");
5282 for (rxconf = i = 0; i < nrxqsets; i++, rxconf++, rxq++) {
5283 /* Set up some basics */
5285 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * nrxqs,
5286 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
5288 "Unable to allocate RX DMA info memory\n");
5293 rxq->ifr_ifdi = ifdip;
5294 /* XXX this needs to be changed if #rx queues != #tx queues */
5295 rxq->ifr_ntxqirq = 1;
5296 rxq->ifr_txqid[0] = i;
5297 for (j = 0; j < nrxqs; j++, ifdip++) {
5298 if (iflib_dma_alloc(ctx, rxqsizes[j], ifdip, 0)) {
5300 "Unable to allocate RX descriptors\n");
5304 bzero((void *)ifdip->idi_vaddr, rxqsizes[j]);
5308 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
5309 rxq->ifr_fl_offset = 1;
5311 rxq->ifr_fl_offset = 0;
5313 rxq->ifr_nfl = nfree_lists;
5315 (iflib_fl_t) malloc(sizeof(struct iflib_fl) * nfree_lists, M_IFLIB, M_NOWAIT | M_ZERO))) {
5316 device_printf(dev, "Unable to allocate free list memory\n");
5321 for (j = 0; j < nfree_lists; j++) {
5322 fl[j].ifl_rxq = rxq;
5324 fl[j].ifl_ifdi = &rxq->ifr_ifdi[j + rxq->ifr_fl_offset];
5325 fl[j].ifl_rxd_size = scctx->isc_rxd_size[j];
5327 /* Allocate receive buffers for the ring */
5328 if (iflib_rxsd_alloc(rxq)) {
5330 "Critical Failure setting up receive buffers\n");
5335 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
5336 fl->ifl_rx_bitmap = bit_alloc(fl->ifl_size, M_IFLIB,
5341 vaddrs = malloc(sizeof(caddr_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK);
5342 paddrs = malloc(sizeof(uint64_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK);
5343 for (i = 0; i < ntxqsets; i++) {
5344 iflib_dma_info_t di = ctx->ifc_txqs[i].ift_ifdi;
5346 for (j = 0; j < ntxqs; j++, di++) {
5347 vaddrs[i*ntxqs + j] = di->idi_vaddr;
5348 paddrs[i*ntxqs + j] = di->idi_paddr;
5351 if ((err = IFDI_TX_QUEUES_ALLOC(ctx, vaddrs, paddrs, ntxqs, ntxqsets)) != 0) {
5352 device_printf(ctx->ifc_dev,
5353 "Unable to allocate device TX queue\n");
5354 iflib_tx_structures_free(ctx);
5355 free(vaddrs, M_IFLIB);
5356 free(paddrs, M_IFLIB);
5359 free(vaddrs, M_IFLIB);
5360 free(paddrs, M_IFLIB);
5363 vaddrs = malloc(sizeof(caddr_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK);
5364 paddrs = malloc(sizeof(uint64_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK);
5365 for (i = 0; i < nrxqsets; i++) {
5366 iflib_dma_info_t di = ctx->ifc_rxqs[i].ifr_ifdi;
5368 for (j = 0; j < nrxqs; j++, di++) {
5369 vaddrs[i*nrxqs + j] = di->idi_vaddr;
5370 paddrs[i*nrxqs + j] = di->idi_paddr;
5373 if ((err = IFDI_RX_QUEUES_ALLOC(ctx, vaddrs, paddrs, nrxqs, nrxqsets)) != 0) {
5374 device_printf(ctx->ifc_dev,
5375 "Unable to allocate device RX queue\n");
5376 iflib_tx_structures_free(ctx);
5377 free(vaddrs, M_IFLIB);
5378 free(paddrs, M_IFLIB);
5381 free(vaddrs, M_IFLIB);
5382 free(paddrs, M_IFLIB);
5386 /* XXX handle allocation failure changes */
5390 if (ctx->ifc_rxqs != NULL)
5391 free(ctx->ifc_rxqs, M_IFLIB);
5392 ctx->ifc_rxqs = NULL;
5393 if (ctx->ifc_txqs != NULL)
5394 free(ctx->ifc_txqs, M_IFLIB);
5395 ctx->ifc_txqs = NULL;
5401 iflib_tx_structures_setup(if_ctx_t ctx)
5403 iflib_txq_t txq = ctx->ifc_txqs;
5406 for (i = 0; i < NTXQSETS(ctx); i++, txq++)
5407 iflib_txq_setup(txq);
5413 iflib_tx_structures_free(if_ctx_t ctx)
5415 iflib_txq_t txq = ctx->ifc_txqs;
5416 if_shared_ctx_t sctx = ctx->ifc_sctx;
5419 for (i = 0; i < NTXQSETS(ctx); i++, txq++) {
5420 iflib_txq_destroy(txq);
5421 for (j = 0; j < sctx->isc_ntxqs; j++)
5422 iflib_dma_free(&txq->ift_ifdi[j]);
5424 free(ctx->ifc_txqs, M_IFLIB);
5425 ctx->ifc_txqs = NULL;
5426 IFDI_QUEUES_FREE(ctx);
5429 /*********************************************************************
5431 * Initialize all receive rings.
5433 **********************************************************************/
5435 iflib_rx_structures_setup(if_ctx_t ctx)
5437 iflib_rxq_t rxq = ctx->ifc_rxqs;
5439 #if defined(INET6) || defined(INET)
5443 for (q = 0; q < ctx->ifc_softc_ctx.isc_nrxqsets; q++, rxq++) {
5444 #if defined(INET6) || defined(INET)
5445 tcp_lro_free(&rxq->ifr_lc);
5446 if ((err = tcp_lro_init_args(&rxq->ifr_lc, ctx->ifc_ifp,
5447 TCP_LRO_ENTRIES, min(1024,
5448 ctx->ifc_softc_ctx.isc_nrxd[rxq->ifr_fl_offset]))) != 0) {
5449 device_printf(ctx->ifc_dev, "LRO Initialization failed!\n");
5452 rxq->ifr_lro_enabled = TRUE;
5454 IFDI_RXQ_SETUP(ctx, rxq->ifr_id);
5457 #if defined(INET6) || defined(INET)
5460 * Free RX software descriptors allocated so far, we will only handle
5461 * the rings that completed, the failing case will have
5462 * cleaned up for itself. 'q' failed, so its the terminus.
5464 rxq = ctx->ifc_rxqs;
5465 for (i = 0; i < q; ++i, rxq++) {
5466 iflib_rx_sds_free(rxq);
5467 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
5473 /*********************************************************************
5475 * Free all receive rings.
5477 **********************************************************************/
5479 iflib_rx_structures_free(if_ctx_t ctx)
5481 iflib_rxq_t rxq = ctx->ifc_rxqs;
5483 for (int i = 0; i < ctx->ifc_softc_ctx.isc_nrxqsets; i++, rxq++) {
5484 iflib_rx_sds_free(rxq);
5486 free(ctx->ifc_rxqs, M_IFLIB);
5487 ctx->ifc_rxqs = NULL;
5491 iflib_qset_structures_setup(if_ctx_t ctx)
5496 * It is expected that the caller takes care of freeing queues if this
5499 if ((err = iflib_tx_structures_setup(ctx)) != 0) {
5500 device_printf(ctx->ifc_dev, "iflib_tx_structures_setup failed: %d\n", err);
5504 if ((err = iflib_rx_structures_setup(ctx)) != 0)
5505 device_printf(ctx->ifc_dev, "iflib_rx_structures_setup failed: %d\n", err);
5511 iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid,
5512 driver_filter_t filter, void *filter_arg, driver_intr_t handler, void *arg, const char *name)
5515 return (_iflib_irq_alloc(ctx, irq, rid, filter, handler, arg, name));
5520 find_nth(if_ctx_t ctx, int qid)
5523 int i, cpuid, eqid, count;
5525 CPU_COPY(&ctx->ifc_cpus, &cpus);
5526 count = CPU_COUNT(&cpus);
5528 /* clear up to the qid'th bit */
5529 for (i = 0; i < eqid; i++) {
5530 cpuid = CPU_FFS(&cpus);
5532 CPU_CLR(cpuid-1, &cpus);
5534 cpuid = CPU_FFS(&cpus);
5540 extern struct cpu_group *cpu_top; /* CPU topology */
5543 find_child_with_core(int cpu, struct cpu_group *grp)
5547 if (grp->cg_children == 0)
5550 MPASS(grp->cg_child);
5551 for (i = 0; i < grp->cg_children; i++) {
5552 if (CPU_ISSET(cpu, &grp->cg_child[i].cg_mask))
5560 * Find the nth "close" core to the specified core
5561 * "close" is defined as the deepest level that shares
5562 * at least an L2 cache. With threads, this will be
5563 * threads on the same core. If the sahred cache is L3
5564 * or higher, simply returns the same core.
5567 find_close_core(int cpu, int core_offset)
5569 struct cpu_group *grp;
5578 while ((i = find_child_with_core(cpu, grp)) != -1) {
5579 /* If the child only has one cpu, don't descend */
5580 if (grp->cg_child[i].cg_count <= 1)
5582 grp = &grp->cg_child[i];
5585 /* If they don't share at least an L2 cache, use the same CPU */
5586 if (grp->cg_level > CG_SHARE_L2 || grp->cg_level == CG_SHARE_NONE)
5590 CPU_COPY(&grp->cg_mask, &cs);
5592 /* Add the selected CPU offset to core offset. */
5593 for (i = 0; (fcpu = CPU_FFS(&cs)) != 0; i++) {
5594 if (fcpu - 1 == cpu)
5596 CPU_CLR(fcpu - 1, &cs);
5602 CPU_COPY(&grp->cg_mask, &cs);
5603 for (i = core_offset % grp->cg_count; i > 0; i--) {
5604 MPASS(CPU_FFS(&cs));
5605 CPU_CLR(CPU_FFS(&cs) - 1, &cs);
5607 MPASS(CPU_FFS(&cs));
5608 return CPU_FFS(&cs) - 1;
5612 find_close_core(int cpu, int core_offset __unused)
5619 get_core_offset(if_ctx_t ctx, iflib_intr_type_t type, int qid)
5623 /* TX queues get cores which share at least an L2 cache with the corresponding RX queue */
5624 /* XXX handle multiple RX threads per core and more than two core per L2 group */
5625 return qid / CPU_COUNT(&ctx->ifc_cpus) + 1;
5627 case IFLIB_INTR_RXTX:
5628 /* RX queues get the specified core */
5629 return qid / CPU_COUNT(&ctx->ifc_cpus);
5635 #define get_core_offset(ctx, type, qid) CPU_FIRST()
5636 #define find_close_core(cpuid, tid) CPU_FIRST()
5637 #define find_nth(ctx, gid) CPU_FIRST()
5640 /* Just to avoid copy/paste */
5642 iflib_irq_set_affinity(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type,
5643 int qid, struct grouptask *gtask, struct taskqgroup *tqg, void *uniq,
5647 int err, cpuid, tid;
5650 cpuid = find_nth(ctx, qid);
5651 tid = get_core_offset(ctx, type, qid);
5653 cpuid = find_close_core(cpuid, tid);
5654 err = taskqgroup_attach_cpu(tqg, gtask, uniq, cpuid, dev, irq->ii_res,
5657 device_printf(dev, "taskqgroup_attach_cpu failed %d\n", err);
5661 if (cpuid > ctx->ifc_cpuid_highest)
5662 ctx->ifc_cpuid_highest = cpuid;
5668 iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid,
5669 iflib_intr_type_t type, driver_filter_t *filter,
5670 void *filter_arg, int qid, const char *name)
5673 struct grouptask *gtask;
5674 struct taskqgroup *tqg;
5675 iflib_filter_info_t info;
5678 driver_filter_t *intr_fast;
5681 info = &ctx->ifc_filter_info;
5685 /* XXX merge tx/rx for netmap? */
5687 q = &ctx->ifc_txqs[qid];
5688 info = &ctx->ifc_txqs[qid].ift_filter_info;
5689 gtask = &ctx->ifc_txqs[qid].ift_task;
5690 tqg = qgroup_if_io_tqg;
5692 intr_fast = iflib_fast_intr;
5693 GROUPTASK_INIT(gtask, 0, fn, q);
5694 ctx->ifc_flags |= IFC_NETMAP_TX_IRQ;
5697 q = &ctx->ifc_rxqs[qid];
5698 info = &ctx->ifc_rxqs[qid].ifr_filter_info;
5699 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5700 tqg = qgroup_if_io_tqg;
5702 intr_fast = iflib_fast_intr;
5703 GROUPTASK_INIT(gtask, 0, fn, q);
5705 case IFLIB_INTR_RXTX:
5706 q = &ctx->ifc_rxqs[qid];
5707 info = &ctx->ifc_rxqs[qid].ifr_filter_info;
5708 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5709 tqg = qgroup_if_io_tqg;
5711 intr_fast = iflib_fast_intr_rxtx;
5712 GROUPTASK_INIT(gtask, 0, fn, q);
5714 case IFLIB_INTR_ADMIN:
5717 info = &ctx->ifc_filter_info;
5718 gtask = &ctx->ifc_admin_task;
5719 tqg = qgroup_if_config_tqg;
5720 fn = _task_fn_admin;
5721 intr_fast = iflib_fast_intr_ctx;
5724 panic("unknown net intr type");
5727 info->ifi_filter = filter;
5728 info->ifi_filter_arg = filter_arg;
5729 info->ifi_task = gtask;
5733 err = _iflib_irq_alloc(ctx, irq, rid, intr_fast, NULL, info, name);
5735 device_printf(dev, "_iflib_irq_alloc failed %d\n", err);
5738 if (type == IFLIB_INTR_ADMIN)
5742 err = iflib_irq_set_affinity(ctx, irq, type, qid, gtask, tqg,
5747 taskqgroup_attach(tqg, gtask, q, dev, irq->ii_res, name);
5754 iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, const char *name)
5756 struct grouptask *gtask;
5757 struct taskqgroup *tqg;
5764 q = &ctx->ifc_txqs[qid];
5765 gtask = &ctx->ifc_txqs[qid].ift_task;
5766 tqg = qgroup_if_io_tqg;
5770 q = &ctx->ifc_rxqs[qid];
5771 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5772 tqg = qgroup_if_io_tqg;
5775 case IFLIB_INTR_IOV:
5777 gtask = &ctx->ifc_vflr_task;
5778 tqg = qgroup_if_config_tqg;
5782 panic("unknown net intr type");
5784 GROUPTASK_INIT(gtask, 0, fn, q);
5786 err = iflib_irq_set_affinity(ctx, irq, type, qid, gtask, tqg,
5789 taskqgroup_attach(tqg, gtask, q, ctx->ifc_dev,
5792 taskqgroup_attach(tqg, gtask, q, NULL, NULL, name);
5797 iflib_irq_free(if_ctx_t ctx, if_irq_t irq)
5801 bus_teardown_intr(ctx->ifc_dev, irq->ii_res, irq->ii_tag);
5804 bus_release_resource(ctx->ifc_dev, SYS_RES_IRQ,
5805 rman_get_rid(irq->ii_res), irq->ii_res);
5809 iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filter_arg, int *rid, const char *name)
5811 iflib_txq_t txq = ctx->ifc_txqs;
5812 iflib_rxq_t rxq = ctx->ifc_rxqs;
5813 if_irq_t irq = &ctx->ifc_legacy_irq;
5814 iflib_filter_info_t info;
5816 struct grouptask *gtask;
5817 struct resource *res;
5818 struct taskqgroup *tqg;
5824 q = &ctx->ifc_rxqs[0];
5825 info = &rxq[0].ifr_filter_info;
5826 gtask = &rxq[0].ifr_task;
5827 tqg = qgroup_if_io_tqg;
5828 tqrid = irq->ii_rid = *rid;
5831 ctx->ifc_flags |= IFC_LEGACY;
5832 info->ifi_filter = filter;
5833 info->ifi_filter_arg = filter_arg;
5834 info->ifi_task = gtask;
5835 info->ifi_ctx = ctx;
5838 /* We allocate a single interrupt resource */
5839 if ((err = _iflib_irq_alloc(ctx, irq, tqrid, iflib_fast_intr_ctx, NULL, info, name)) != 0)
5841 GROUPTASK_INIT(gtask, 0, fn, q);
5843 taskqgroup_attach(tqg, gtask, q, dev, res, name);
5845 GROUPTASK_INIT(&txq->ift_task, 0, _task_fn_tx, txq);
5846 taskqgroup_attach(qgroup_if_io_tqg, &txq->ift_task, txq, dev, res,
5852 iflib_led_create(if_ctx_t ctx)
5855 ctx->ifc_led_dev = led_create(iflib_led_func, ctx,
5856 device_get_nameunit(ctx->ifc_dev));
5860 iflib_tx_intr_deferred(if_ctx_t ctx, int txqid)
5863 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task);
5867 iflib_rx_intr_deferred(if_ctx_t ctx, int rxqid)
5870 GROUPTASK_ENQUEUE(&ctx->ifc_rxqs[rxqid].ifr_task);
5874 iflib_admin_intr_deferred(if_ctx_t ctx)
5877 struct grouptask *gtask;
5879 gtask = &ctx->ifc_admin_task;
5880 MPASS(gtask != NULL && gtask->gt_taskqueue != NULL);
5883 GROUPTASK_ENQUEUE(&ctx->ifc_admin_task);
5887 iflib_iov_intr_deferred(if_ctx_t ctx)
5890 GROUPTASK_ENQUEUE(&ctx->ifc_vflr_task);
5894 iflib_io_tqg_attach(struct grouptask *gt, void *uniq, int cpu, char *name)
5897 taskqgroup_attach_cpu(qgroup_if_io_tqg, gt, uniq, cpu, NULL, NULL,
5902 iflib_config_gtask_init(void *ctx, struct grouptask *gtask, gtask_fn_t *fn,
5906 GROUPTASK_INIT(gtask, 0, fn, ctx);
5907 taskqgroup_attach(qgroup_if_config_tqg, gtask, gtask, NULL, NULL,
5912 iflib_config_gtask_deinit(struct grouptask *gtask)
5915 taskqgroup_detach(qgroup_if_config_tqg, gtask);
5919 iflib_link_state_change(if_ctx_t ctx, int link_state, uint64_t baudrate)
5921 if_t ifp = ctx->ifc_ifp;
5922 iflib_txq_t txq = ctx->ifc_txqs;
5924 if_setbaudrate(ifp, baudrate);
5925 if (baudrate >= IF_Gbps(10)) {
5927 ctx->ifc_flags |= IFC_PREFETCH;
5930 /* If link down, disable watchdog */
5931 if ((ctx->ifc_link_state == LINK_STATE_UP) && (link_state == LINK_STATE_DOWN)) {
5932 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxqsets; i++, txq++)
5933 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
5935 ctx->ifc_link_state = link_state;
5936 if_link_state_change(ifp, link_state);
5940 iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq)
5944 int credits_pre = txq->ift_cidx_processed;
5947 if (ctx->isc_txd_credits_update == NULL)
5950 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
5951 BUS_DMASYNC_POSTREAD);
5952 if ((credits = ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, true)) == 0)
5955 txq->ift_processed += credits;
5956 txq->ift_cidx_processed += credits;
5958 MPASS(credits_pre + credits == txq->ift_cidx_processed);
5959 if (txq->ift_cidx_processed >= txq->ift_size)
5960 txq->ift_cidx_processed -= txq->ift_size;
5965 iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget)
5970 for (i = 0, fl = &rxq->ifr_fl[0]; i < rxq->ifr_nfl; i++, fl++)
5971 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
5972 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
5973 return (ctx->isc_rxd_available(ctx->ifc_softc, rxq->ifr_id, cidx,
5978 iflib_add_int_delay_sysctl(if_ctx_t ctx, const char *name,
5979 const char *description, if_int_delay_info_t info,
5980 int offset, int value)
5982 info->iidi_ctx = ctx;
5983 info->iidi_offset = offset;
5984 info->iidi_value = value;
5985 SYSCTL_ADD_PROC(device_get_sysctl_ctx(ctx->ifc_dev),
5986 SYSCTL_CHILDREN(device_get_sysctl_tree(ctx->ifc_dev)),
5987 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
5988 info, 0, iflib_sysctl_int_delay, "I", description);
5992 iflib_ctx_lock_get(if_ctx_t ctx)
5995 return (&ctx->ifc_ctx_sx);
5999 iflib_msix_init(if_ctx_t ctx)
6001 device_t dev = ctx->ifc_dev;
6002 if_shared_ctx_t sctx = ctx->ifc_sctx;
6003 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
6004 int vectors, queues, rx_queues, tx_queues, queuemsgs, msgs;
6005 int iflib_num_tx_queues, iflib_num_rx_queues;
6006 int err, admincnt, bar;
6008 iflib_num_tx_queues = ctx->ifc_sysctl_ntxqs;
6009 iflib_num_rx_queues = ctx->ifc_sysctl_nrxqs;
6012 device_printf(dev, "msix_init qsets capped at %d\n",
6013 imax(scctx->isc_ntxqsets, scctx->isc_nrxqsets));
6015 bar = ctx->ifc_softc_ctx.isc_msix_bar;
6016 admincnt = sctx->isc_admin_intrcnt;
6017 /* Override by tuneable */
6018 if (scctx->isc_disable_msix)
6021 /* First try MSI-X */
6022 if ((msgs = pci_msix_count(dev)) == 0) {
6024 device_printf(dev, "MSI-X not supported or disabled\n");
6028 * bar == -1 => "trust me I know what I'm doing"
6029 * Some drivers are for hardware that is so shoddily
6030 * documented that no one knows which bars are which
6031 * so the developer has to map all bars. This hack
6032 * allows shoddy garbage to use MSI-X in this framework.
6035 ctx->ifc_msix_mem = bus_alloc_resource_any(dev,
6036 SYS_RES_MEMORY, &bar, RF_ACTIVE);
6037 if (ctx->ifc_msix_mem == NULL) {
6038 device_printf(dev, "Unable to map MSI-X table\n");
6043 /* use only 1 qset in debug mode */
6044 queuemsgs = min(msgs - admincnt, 1);
6046 queuemsgs = msgs - admincnt;
6049 queues = imin(queuemsgs, rss_getnumbuckets());
6053 queues = imin(CPU_COUNT(&ctx->ifc_cpus), queues);
6056 "intr CPUs: %d queue msgs: %d admincnt: %d\n",
6057 CPU_COUNT(&ctx->ifc_cpus), queuemsgs, admincnt);
6059 /* If we're doing RSS, clamp at the number of RSS buckets */
6060 if (queues > rss_getnumbuckets())
6061 queues = rss_getnumbuckets();
6063 if (iflib_num_rx_queues > 0 && iflib_num_rx_queues < queuemsgs - admincnt)
6064 rx_queues = iflib_num_rx_queues;
6068 if (rx_queues > scctx->isc_nrxqsets)
6069 rx_queues = scctx->isc_nrxqsets;
6072 * We want this to be all logical CPUs by default
6074 if (iflib_num_tx_queues > 0 && iflib_num_tx_queues < queues)
6075 tx_queues = iflib_num_tx_queues;
6077 tx_queues = mp_ncpus;
6079 if (tx_queues > scctx->isc_ntxqsets)
6080 tx_queues = scctx->isc_ntxqsets;
6082 if (ctx->ifc_sysctl_qs_eq_override == 0) {
6084 if (tx_queues != rx_queues)
6086 "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n",
6087 min(rx_queues, tx_queues), min(rx_queues, tx_queues));
6089 tx_queues = min(rx_queues, tx_queues);
6090 rx_queues = min(rx_queues, tx_queues);
6093 device_printf(dev, "Using %d rx queues %d tx queues\n",
6094 rx_queues, tx_queues);
6096 vectors = rx_queues + admincnt;
6097 if ((err = pci_alloc_msix(dev, &vectors)) == 0) {
6098 device_printf(dev, "Using MSI-X interrupts with %d vectors\n",
6100 scctx->isc_vectors = vectors;
6101 scctx->isc_nrxqsets = rx_queues;
6102 scctx->isc_ntxqsets = tx_queues;
6103 scctx->isc_intr = IFLIB_INTR_MSIX;
6108 "failed to allocate %d MSI-X vectors, err: %d - using MSI\n",
6110 bus_release_resource(dev, SYS_RES_MEMORY, bar,
6112 ctx->ifc_msix_mem = NULL;
6115 vectors = pci_msi_count(dev);
6116 scctx->isc_nrxqsets = 1;
6117 scctx->isc_ntxqsets = 1;
6118 scctx->isc_vectors = vectors;
6119 if (vectors == 1 && pci_alloc_msi(dev, &vectors) == 0) {
6120 device_printf(dev,"Using an MSI interrupt\n");
6121 scctx->isc_intr = IFLIB_INTR_MSI;
6123 scctx->isc_vectors = 1;
6124 device_printf(dev,"Using a Legacy interrupt\n");
6125 scctx->isc_intr = IFLIB_INTR_LEGACY;
6131 static const char *ring_states[] = { "IDLE", "BUSY", "STALLED", "ABDICATED" };
6134 mp_ring_state_handler(SYSCTL_HANDLER_ARGS)
6137 uint16_t *state = ((uint16_t *)oidp->oid_arg1);
6139 const char *ring_state = "UNKNOWN";
6142 rc = sysctl_wire_old_buffer(req, 0);
6146 sb = sbuf_new_for_sysctl(NULL, NULL, 80, req);
6151 ring_state = ring_states[state[3]];
6153 sbuf_printf(sb, "pidx_head: %04hd pidx_tail: %04hd cidx: %04hd state: %s",
6154 state[0], state[1], state[2], ring_state);
6155 rc = sbuf_finish(sb);
6160 enum iflib_ndesc_handler {
6166 mp_ndesc_handler(SYSCTL_HANDLER_ARGS)
6168 if_ctx_t ctx = (void *)arg1;
6169 enum iflib_ndesc_handler type = arg2;
6170 char buf[256] = {0};
6175 MPASS(type == IFLIB_NTXD_HANDLER || type == IFLIB_NRXD_HANDLER);
6179 case IFLIB_NTXD_HANDLER:
6180 ndesc = ctx->ifc_sysctl_ntxds;
6182 nqs = ctx->ifc_sctx->isc_ntxqs;
6184 case IFLIB_NRXD_HANDLER:
6185 ndesc = ctx->ifc_sysctl_nrxds;
6187 nqs = ctx->ifc_sctx->isc_nrxqs;
6190 panic("unhandled type");
6195 for (i=0; i<8; i++) {
6200 sprintf(strchr(buf, 0), "%d", ndesc[i]);
6203 rc = sysctl_handle_string(oidp, buf, sizeof(buf), req);
6204 if (rc || req->newptr == NULL)
6207 for (i = 0, next = buf, p = strsep(&next, " ,"); i < 8 && p;
6208 i++, p = strsep(&next, " ,")) {
6209 ndesc[i] = strtoul(p, NULL, 10);
6215 #define NAME_BUFLEN 32
6217 iflib_add_device_sysctl_pre(if_ctx_t ctx)
6219 device_t dev = iflib_get_dev(ctx);
6220 struct sysctl_oid_list *child, *oid_list;
6221 struct sysctl_ctx_list *ctx_list;
6222 struct sysctl_oid *node;
6224 ctx_list = device_get_sysctl_ctx(dev);
6225 child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
6226 ctx->ifc_sysctl_node = node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "iflib",
6227 CTLFLAG_RD, NULL, "IFLIB fields");
6228 oid_list = SYSCTL_CHILDREN(node);
6230 SYSCTL_ADD_STRING(ctx_list, oid_list, OID_AUTO, "driver_version",
6231 CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version, 0,
6234 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs",
6235 CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0,
6236 "# of txqs to use, 0 => use default #");
6237 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_nrxqs",
6238 CTLFLAG_RWTUN, &ctx->ifc_sysctl_nrxqs, 0,
6239 "# of rxqs to use, 0 => use default #");
6240 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_qs_enable",
6241 CTLFLAG_RWTUN, &ctx->ifc_sysctl_qs_eq_override, 0,
6242 "permit #txq != #rxq");
6243 SYSCTL_ADD_INT(ctx_list, oid_list, OID_AUTO, "disable_msix",
6244 CTLFLAG_RWTUN, &ctx->ifc_softc_ctx.isc_disable_msix, 0,
6245 "disable MSI-X (default 0)");
6246 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "rx_budget",
6247 CTLFLAG_RWTUN, &ctx->ifc_sysctl_rx_budget, 0,
6248 "set the rx budget");
6249 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "tx_abdicate",
6250 CTLFLAG_RWTUN, &ctx->ifc_sysctl_tx_abdicate, 0,
6251 "cause tx to abdicate instead of running to completion");
6253 /* XXX change for per-queue sizes */
6254 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_ntxds",
6255 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NTXD_HANDLER,
6256 mp_ndesc_handler, "A",
6257 "list of # of tx descriptors to use, 0 = use default #");
6258 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_nrxds",
6259 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NRXD_HANDLER,
6260 mp_ndesc_handler, "A",
6261 "list of # of rx descriptors to use, 0 = use default #");
6265 iflib_add_device_sysctl_post(if_ctx_t ctx)
6267 if_shared_ctx_t sctx = ctx->ifc_sctx;
6268 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
6269 device_t dev = iflib_get_dev(ctx);
6270 struct sysctl_oid_list *child;
6271 struct sysctl_ctx_list *ctx_list;
6276 char namebuf[NAME_BUFLEN];
6278 struct sysctl_oid *queue_node, *fl_node, *node;
6279 struct sysctl_oid_list *queue_list, *fl_list;
6280 ctx_list = device_get_sysctl_ctx(dev);
6282 node = ctx->ifc_sysctl_node;
6283 child = SYSCTL_CHILDREN(node);
6285 if (scctx->isc_ntxqsets > 100)
6287 else if (scctx->isc_ntxqsets > 10)
6291 for (i = 0, txq = ctx->ifc_txqs; i < scctx->isc_ntxqsets; i++, txq++) {
6292 snprintf(namebuf, NAME_BUFLEN, qfmt, i);
6293 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf,
6294 CTLFLAG_RD, NULL, "Queue Name");
6295 queue_list = SYSCTL_CHILDREN(queue_node);
6297 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_dequeued",
6299 &txq->ift_dequeued, "total mbufs freed");
6300 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_enqueued",
6302 &txq->ift_enqueued, "total mbufs enqueued");
6304 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag",
6306 &txq->ift_mbuf_defrag, "# of times m_defrag was called");
6307 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "m_pullups",
6309 &txq->ift_pullups, "# of times m_pullup was called");
6310 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag_failed",
6312 &txq->ift_mbuf_defrag_failed, "# of times m_defrag failed");
6313 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_desc_avail",
6315 &txq->ift_no_desc_avail, "# of times no descriptors were available");
6316 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "tx_map_failed",
6318 &txq->ift_map_failed, "# of times dma map failed");
6319 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txd_encap_efbig",
6321 &txq->ift_txd_encap_efbig, "# of times txd_encap returned EFBIG");
6322 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_tx_dma_setup",
6324 &txq->ift_no_tx_dma_setup, "# of times map failed for other than EFBIG");
6325 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_pidx",
6327 &txq->ift_pidx, 1, "Producer Index");
6328 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx",
6330 &txq->ift_cidx, 1, "Consumer Index");
6331 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx_processed",
6333 &txq->ift_cidx_processed, 1, "Consumer Index seen by credit update");
6334 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_in_use",
6336 &txq->ift_in_use, 1, "descriptors in use");
6337 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_processed",
6339 &txq->ift_processed, "descriptors procesed for clean");
6340 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_cleaned",
6342 &txq->ift_cleaned, "total cleaned");
6343 SYSCTL_ADD_PROC(ctx_list, queue_list, OID_AUTO, "ring_state",
6344 CTLTYPE_STRING | CTLFLAG_RD, __DEVOLATILE(uint64_t *, &txq->ift_br->state),
6345 0, mp_ring_state_handler, "A", "soft ring state");
6346 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_enqueues",
6347 CTLFLAG_RD, &txq->ift_br->enqueues,
6348 "# of enqueues to the mp_ring for this queue");
6349 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_drops",
6350 CTLFLAG_RD, &txq->ift_br->drops,
6351 "# of drops in the mp_ring for this queue");
6352 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_starts",
6353 CTLFLAG_RD, &txq->ift_br->starts,
6354 "# of normal consumer starts in the mp_ring for this queue");
6355 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_stalls",
6356 CTLFLAG_RD, &txq->ift_br->stalls,
6357 "# of consumer stalls in the mp_ring for this queue");
6358 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_restarts",
6359 CTLFLAG_RD, &txq->ift_br->restarts,
6360 "# of consumer restarts in the mp_ring for this queue");
6361 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_abdications",
6362 CTLFLAG_RD, &txq->ift_br->abdications,
6363 "# of consumer abdications in the mp_ring for this queue");
6366 if (scctx->isc_nrxqsets > 100)
6368 else if (scctx->isc_nrxqsets > 10)
6372 for (i = 0, rxq = ctx->ifc_rxqs; i < scctx->isc_nrxqsets; i++, rxq++) {
6373 snprintf(namebuf, NAME_BUFLEN, qfmt, i);
6374 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf,
6375 CTLFLAG_RD, NULL, "Queue Name");
6376 queue_list = SYSCTL_CHILDREN(queue_node);
6377 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
6378 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_pidx",
6380 &rxq->ifr_cq_pidx, 1, "Producer Index");
6381 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_cidx",
6383 &rxq->ifr_cq_cidx, 1, "Consumer Index");
6386 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) {
6387 snprintf(namebuf, NAME_BUFLEN, "rxq_fl%d", j);
6388 fl_node = SYSCTL_ADD_NODE(ctx_list, queue_list, OID_AUTO, namebuf,
6389 CTLFLAG_RD, NULL, "freelist Name");
6390 fl_list = SYSCTL_CHILDREN(fl_node);
6391 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "pidx",
6393 &fl->ifl_pidx, 1, "Producer Index");
6394 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "cidx",
6396 &fl->ifl_cidx, 1, "Consumer Index");
6397 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "credits",
6399 &fl->ifl_credits, 1, "credits available");
6401 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_enqueued",
6403 &fl->ifl_m_enqueued, "mbufs allocated");
6404 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_dequeued",
6406 &fl->ifl_m_dequeued, "mbufs freed");
6407 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_enqueued",
6409 &fl->ifl_cl_enqueued, "clusters allocated");
6410 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_dequeued",
6412 &fl->ifl_cl_dequeued, "clusters freed");
6421 iflib_request_reset(if_ctx_t ctx)
6425 ctx->ifc_flags |= IFC_DO_RESET;
6429 #ifndef __NO_STRICT_ALIGNMENT
6430 static struct mbuf *
6431 iflib_fixup_rx(struct mbuf *m)
6435 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) {
6436 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len);
6437 m->m_data += ETHER_HDR_LEN;
6440 MGETHDR(n, M_NOWAIT, MT_DATA);
6445 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN);
6446 m->m_data += ETHER_HDR_LEN;
6447 m->m_len -= ETHER_HDR_LEN;
6448 n->m_len = ETHER_HDR_LEN;
6449 M_MOVE_PKTHDR(n, m);
6458 iflib_netdump_init(struct ifnet *ifp, int *nrxr, int *ncl, int *clsize)
6462 ctx = if_getsoftc(ifp);
6464 *nrxr = NRXQSETS(ctx);
6465 *ncl = ctx->ifc_rxqs[0].ifr_fl->ifl_size;
6466 *clsize = ctx->ifc_rxqs[0].ifr_fl->ifl_buf_size;
6471 iflib_netdump_event(struct ifnet *ifp, enum netdump_ev event)
6474 if_softc_ctx_t scctx;
6479 ctx = if_getsoftc(ifp);
6480 scctx = &ctx->ifc_softc_ctx;
6484 for (i = 0; i < scctx->isc_nrxqsets; i++) {
6485 rxq = &ctx->ifc_rxqs[i];
6486 for (j = 0; j < rxq->ifr_nfl; j++) {
6488 fl->ifl_zone = m_getzone(fl->ifl_buf_size);
6491 iflib_no_tx_batch = 1;
6499 iflib_netdump_transmit(struct ifnet *ifp, struct mbuf *m)
6505 ctx = if_getsoftc(ifp);
6506 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
6510 txq = &ctx->ifc_txqs[0];
6511 error = iflib_encap(txq, &m);
6513 (void)iflib_txd_db_check(ctx, txq, true, txq->ift_in_use);
6518 iflib_netdump_poll(struct ifnet *ifp, int count)
6521 if_softc_ctx_t scctx;
6525 ctx = if_getsoftc(ifp);
6526 scctx = &ctx->ifc_softc_ctx;
6528 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
6532 txq = &ctx->ifc_txqs[0];
6533 (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
6535 for (i = 0; i < scctx->isc_nrxqsets; i++)
6536 (void)iflib_rxeof(&ctx->ifc_rxqs[i], 16 /* XXX */);
6539 #endif /* NETDUMP */