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;
174 uint32_t ifc_rx_mbuf_sz;
178 int ifc_watchdog_events;
179 struct cdev *ifc_led_dev;
180 struct resource *ifc_msix_mem;
182 struct if_irq ifc_legacy_irq;
183 struct grouptask ifc_admin_task;
184 struct grouptask ifc_vflr_task;
185 struct iflib_filter_info ifc_filter_info;
186 struct ifmedia ifc_media;
188 struct sysctl_oid *ifc_sysctl_node;
189 uint16_t ifc_sysctl_ntxqs;
190 uint16_t ifc_sysctl_nrxqs;
191 uint16_t ifc_sysctl_qs_eq_override;
192 uint16_t ifc_sysctl_rx_budget;
193 uint16_t ifc_sysctl_tx_abdicate;
195 qidx_t ifc_sysctl_ntxds[8];
196 qidx_t ifc_sysctl_nrxds[8];
197 struct if_txrx ifc_txrx;
198 #define isc_txd_encap ifc_txrx.ift_txd_encap
199 #define isc_txd_flush ifc_txrx.ift_txd_flush
200 #define isc_txd_credits_update ifc_txrx.ift_txd_credits_update
201 #define isc_rxd_available ifc_txrx.ift_rxd_available
202 #define isc_rxd_pkt_get ifc_txrx.ift_rxd_pkt_get
203 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
204 #define isc_rxd_flush ifc_txrx.ift_rxd_flush
205 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
206 #define isc_rxd_refill ifc_txrx.ift_rxd_refill
207 #define isc_legacy_intr ifc_txrx.ift_legacy_intr
208 eventhandler_tag ifc_vlan_attach_event;
209 eventhandler_tag ifc_vlan_detach_event;
210 uint8_t ifc_mac[ETHER_ADDR_LEN];
211 char ifc_mtx_name[16];
216 iflib_get_softc(if_ctx_t ctx)
219 return (ctx->ifc_softc);
223 iflib_get_dev(if_ctx_t ctx)
226 return (ctx->ifc_dev);
230 iflib_get_ifp(if_ctx_t ctx)
233 return (ctx->ifc_ifp);
237 iflib_get_media(if_ctx_t ctx)
240 return (&ctx->ifc_media);
244 iflib_get_flags(if_ctx_t ctx)
246 return (ctx->ifc_flags);
250 iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN])
253 bcopy(mac, ctx->ifc_mac, ETHER_ADDR_LEN);
257 iflib_get_softc_ctx(if_ctx_t ctx)
260 return (&ctx->ifc_softc_ctx);
264 iflib_get_sctx(if_ctx_t ctx)
267 return (ctx->ifc_sctx);
270 #define IP_ALIGNED(m) ((((uintptr_t)(m)->m_data) & 0x3) == 0x2)
271 #define CACHE_PTR_INCREMENT (CACHE_LINE_SIZE/sizeof(void*))
272 #define CACHE_PTR_NEXT(ptr) ((void *)(((uintptr_t)(ptr)+CACHE_LINE_SIZE-1) & (CACHE_LINE_SIZE-1)))
274 #define LINK_ACTIVE(ctx) ((ctx)->ifc_link_state == LINK_STATE_UP)
275 #define CTX_IS_VF(ctx) ((ctx)->ifc_sctx->isc_flags & IFLIB_IS_VF)
277 typedef struct iflib_sw_rx_desc_array {
278 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */
279 struct mbuf **ifsd_m; /* pkthdr mbufs */
280 caddr_t *ifsd_cl; /* direct cluster pointer for rx */
281 bus_addr_t *ifsd_ba; /* bus addr of cluster for rx */
282 } iflib_rxsd_array_t;
284 typedef struct iflib_sw_tx_desc_array {
285 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */
286 bus_dmamap_t *ifsd_tso_map; /* bus_dma maps for TSO packet */
287 struct mbuf **ifsd_m; /* pkthdr mbufs */
291 /* magic number that should be high enough for any hardware */
292 #define IFLIB_MAX_TX_SEGS 128
293 #define IFLIB_RX_COPY_THRESH 128
294 #define IFLIB_MAX_RX_REFRESH 32
295 /* The minimum descriptors per second before we start coalescing */
296 #define IFLIB_MIN_DESC_SEC 16384
297 #define IFLIB_DEFAULT_TX_UPDATE_FREQ 16
298 #define IFLIB_QUEUE_IDLE 0
299 #define IFLIB_QUEUE_HUNG 1
300 #define IFLIB_QUEUE_WORKING 2
301 /* maximum number of txqs that can share an rx interrupt */
302 #define IFLIB_MAX_TX_SHARED_INTR 4
304 /* this should really scale with ring size - this is a fairly arbitrary value */
305 #define TX_BATCH_SIZE 32
307 #define IFLIB_RESTART_BUDGET 8
310 #define CSUM_OFFLOAD (CSUM_IP_TSO|CSUM_IP6_TSO|CSUM_IP| \
311 CSUM_IP_UDP|CSUM_IP_TCP|CSUM_IP_SCTP| \
312 CSUM_IP6_UDP|CSUM_IP6_TCP|CSUM_IP6_SCTP)
316 qidx_t ift_cidx_processed;
319 uint8_t ift_br_offset;
320 uint16_t ift_npending;
321 uint16_t ift_db_pending;
322 uint16_t ift_rs_pending;
324 uint8_t ift_txd_size[8];
325 uint64_t ift_processed;
326 uint64_t ift_cleaned;
327 uint64_t ift_cleaned_prev;
329 uint64_t ift_enqueued;
330 uint64_t ift_dequeued;
332 uint64_t ift_no_tx_dma_setup;
333 uint64_t ift_no_desc_avail;
334 uint64_t ift_mbuf_defrag_failed;
335 uint64_t ift_mbuf_defrag;
336 uint64_t ift_map_failed;
337 uint64_t ift_txd_encap_efbig;
338 uint64_t ift_pullups;
339 uint64_t ift_last_timer_tick;
342 struct mtx ift_db_mtx;
344 /* constant values */
346 struct ifmp_ring *ift_br;
347 struct grouptask ift_task;
350 struct callout ift_timer;
352 if_txsd_vec_t ift_sds;
355 uint8_t ift_update_freq;
356 struct iflib_filter_info ift_filter_info;
357 bus_dma_tag_t ift_buf_tag;
358 bus_dma_tag_t ift_tso_buf_tag;
359 iflib_dma_info_t ift_ifdi;
360 #define MTX_NAME_LEN 16
361 char ift_mtx_name[MTX_NAME_LEN];
362 char ift_db_mtx_name[MTX_NAME_LEN];
363 bus_dma_segment_t ift_segs[IFLIB_MAX_TX_SEGS] __aligned(CACHE_LINE_SIZE);
364 #ifdef IFLIB_DIAGNOSTICS
365 uint64_t ift_cpu_exec_count[256];
367 } __aligned(CACHE_LINE_SIZE);
374 uint8_t ifl_rxd_size;
376 uint64_t ifl_m_enqueued;
377 uint64_t ifl_m_dequeued;
378 uint64_t ifl_cl_enqueued;
379 uint64_t ifl_cl_dequeued;
383 bitstr_t *ifl_rx_bitmap;
387 uint16_t ifl_buf_size;
390 iflib_rxsd_array_t ifl_sds;
393 bus_dma_tag_t ifl_buf_tag;
394 iflib_dma_info_t ifl_ifdi;
395 uint64_t ifl_bus_addrs[IFLIB_MAX_RX_REFRESH] __aligned(CACHE_LINE_SIZE);
396 caddr_t ifl_vm_addrs[IFLIB_MAX_RX_REFRESH];
397 qidx_t ifl_rxd_idxs[IFLIB_MAX_RX_REFRESH];
398 } __aligned(CACHE_LINE_SIZE);
401 get_inuse(int size, qidx_t cidx, qidx_t pidx, uint8_t gen)
407 else if (pidx < cidx)
408 used = size - cidx + pidx;
409 else if (gen == 0 && pidx == cidx)
411 else if (gen == 1 && pidx == cidx)
419 #define TXQ_AVAIL(txq) (txq->ift_size - get_inuse(txq->ift_size, txq->ift_cidx, txq->ift_pidx, txq->ift_gen))
421 #define IDXDIFF(head, tail, wrap) \
422 ((head) >= (tail) ? (head) - (tail) : (wrap) - (tail) + (head))
425 /* If there is a separate completion queue -
426 * these are the cq cidx and pidx. Otherwise
433 uint8_t ifr_fl_offset;
439 uint8_t ifr_lro_enabled;
442 uint8_t ifr_txqid[IFLIB_MAX_TX_SHARED_INTR];
443 struct lro_ctrl ifr_lc;
444 struct grouptask ifr_task;
445 struct iflib_filter_info ifr_filter_info;
446 iflib_dma_info_t ifr_ifdi;
448 /* dynamically allocate if any drivers need a value substantially larger than this */
449 struct if_rxd_frag ifr_frags[IFLIB_MAX_RX_SEGS] __aligned(CACHE_LINE_SIZE);
450 #ifdef IFLIB_DIAGNOSTICS
451 uint64_t ifr_cpu_exec_count[256];
453 } __aligned(CACHE_LINE_SIZE);
455 typedef struct if_rxsd {
457 struct mbuf **ifsd_m;
462 /* multiple of word size */
464 #define PKT_INFO_SIZE 6
465 #define RXD_INFO_SIZE 5
466 #define PKT_TYPE uint64_t
468 #define PKT_INFO_SIZE 11
469 #define RXD_INFO_SIZE 8
470 #define PKT_TYPE uint32_t
472 #define PKT_LOOP_BOUND ((PKT_INFO_SIZE/3)*3)
473 #define RXD_LOOP_BOUND ((RXD_INFO_SIZE/4)*4)
475 typedef struct if_pkt_info_pad {
476 PKT_TYPE pkt_val[PKT_INFO_SIZE];
477 } *if_pkt_info_pad_t;
478 typedef struct if_rxd_info_pad {
479 PKT_TYPE rxd_val[RXD_INFO_SIZE];
480 } *if_rxd_info_pad_t;
482 CTASSERT(sizeof(struct if_pkt_info_pad) == sizeof(struct if_pkt_info));
483 CTASSERT(sizeof(struct if_rxd_info_pad) == sizeof(struct if_rxd_info));
487 pkt_info_zero(if_pkt_info_t pi)
489 if_pkt_info_pad_t pi_pad;
491 pi_pad = (if_pkt_info_pad_t)pi;
492 pi_pad->pkt_val[0] = 0; pi_pad->pkt_val[1] = 0; pi_pad->pkt_val[2] = 0;
493 pi_pad->pkt_val[3] = 0; pi_pad->pkt_val[4] = 0; pi_pad->pkt_val[5] = 0;
495 pi_pad->pkt_val[6] = 0; pi_pad->pkt_val[7] = 0; pi_pad->pkt_val[8] = 0;
496 pi_pad->pkt_val[9] = 0; pi_pad->pkt_val[10] = 0;
500 static device_method_t iflib_pseudo_methods[] = {
501 DEVMETHOD(device_attach, noop_attach),
502 DEVMETHOD(device_detach, iflib_pseudo_detach),
506 driver_t iflib_pseudodriver = {
507 "iflib_pseudo", iflib_pseudo_methods, sizeof(struct iflib_ctx),
511 rxd_info_zero(if_rxd_info_t ri)
513 if_rxd_info_pad_t ri_pad;
516 ri_pad = (if_rxd_info_pad_t)ri;
517 for (i = 0; i < RXD_LOOP_BOUND; i += 4) {
518 ri_pad->rxd_val[i] = 0;
519 ri_pad->rxd_val[i+1] = 0;
520 ri_pad->rxd_val[i+2] = 0;
521 ri_pad->rxd_val[i+3] = 0;
524 ri_pad->rxd_val[RXD_INFO_SIZE-1] = 0;
529 * Only allow a single packet to take up most 1/nth of the tx ring
531 #define MAX_SINGLE_PACKET_FRACTION 12
532 #define IF_BAD_DMA (bus_addr_t)-1
534 #define CTX_ACTIVE(ctx) ((if_getdrvflags((ctx)->ifc_ifp) & IFF_DRV_RUNNING))
536 #define CTX_LOCK_INIT(_sc) sx_init(&(_sc)->ifc_ctx_sx, "iflib ctx lock")
537 #define CTX_LOCK(ctx) sx_xlock(&(ctx)->ifc_ctx_sx)
538 #define CTX_UNLOCK(ctx) sx_xunlock(&(ctx)->ifc_ctx_sx)
539 #define CTX_LOCK_DESTROY(ctx) sx_destroy(&(ctx)->ifc_ctx_sx)
542 #define STATE_LOCK_INIT(_sc, _name) mtx_init(&(_sc)->ifc_state_mtx, _name, "iflib state lock", MTX_DEF)
543 #define STATE_LOCK(ctx) mtx_lock(&(ctx)->ifc_state_mtx)
544 #define STATE_UNLOCK(ctx) mtx_unlock(&(ctx)->ifc_state_mtx)
545 #define STATE_LOCK_DESTROY(ctx) mtx_destroy(&(ctx)->ifc_state_mtx)
549 #define CALLOUT_LOCK(txq) mtx_lock(&txq->ift_mtx)
550 #define CALLOUT_UNLOCK(txq) mtx_unlock(&txq->ift_mtx)
553 iflib_set_detach(if_ctx_t ctx)
556 ctx->ifc_flags |= IFC_IN_DETACH;
560 /* Our boot-time initialization hook */
561 static int iflib_module_event_handler(module_t, int, void *);
563 static moduledata_t iflib_moduledata = {
565 iflib_module_event_handler,
569 DECLARE_MODULE(iflib, iflib_moduledata, SI_SUB_INIT_IF, SI_ORDER_ANY);
570 MODULE_VERSION(iflib, 1);
572 MODULE_DEPEND(iflib, pci, 1, 1, 1);
573 MODULE_DEPEND(iflib, ether, 1, 1, 1);
575 TASKQGROUP_DEFINE(if_io_tqg, mp_ncpus, 1);
576 TASKQGROUP_DEFINE(if_config_tqg, 1, 1);
578 #ifndef IFLIB_DEBUG_COUNTERS
580 #define IFLIB_DEBUG_COUNTERS 1
582 #define IFLIB_DEBUG_COUNTERS 0
583 #endif /* !INVARIANTS */
586 static SYSCTL_NODE(_net, OID_AUTO, iflib, CTLFLAG_RD, 0,
587 "iflib driver parameters");
590 * XXX need to ensure that this can't accidentally cause the head to be moved backwards
592 static int iflib_min_tx_latency = 0;
593 SYSCTL_INT(_net_iflib, OID_AUTO, min_tx_latency, CTLFLAG_RW,
594 &iflib_min_tx_latency, 0, "minimize transmit latency at the possible expense of throughput");
595 static int iflib_no_tx_batch = 0;
596 SYSCTL_INT(_net_iflib, OID_AUTO, no_tx_batch, CTLFLAG_RW,
597 &iflib_no_tx_batch, 0, "minimize transmit latency at the possible expense of throughput");
600 #if IFLIB_DEBUG_COUNTERS
602 static int iflib_tx_seen;
603 static int iflib_tx_sent;
604 static int iflib_tx_encap;
605 static int iflib_rx_allocs;
606 static int iflib_fl_refills;
607 static int iflib_fl_refills_large;
608 static int iflib_tx_frees;
610 SYSCTL_INT(_net_iflib, OID_AUTO, tx_seen, CTLFLAG_RD,
611 &iflib_tx_seen, 0, "# tx mbufs seen");
612 SYSCTL_INT(_net_iflib, OID_AUTO, tx_sent, CTLFLAG_RD,
613 &iflib_tx_sent, 0, "# tx mbufs sent");
614 SYSCTL_INT(_net_iflib, OID_AUTO, tx_encap, CTLFLAG_RD,
615 &iflib_tx_encap, 0, "# tx mbufs encapped");
616 SYSCTL_INT(_net_iflib, OID_AUTO, tx_frees, CTLFLAG_RD,
617 &iflib_tx_frees, 0, "# tx frees");
618 SYSCTL_INT(_net_iflib, OID_AUTO, rx_allocs, CTLFLAG_RD,
619 &iflib_rx_allocs, 0, "# rx allocations");
620 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills, CTLFLAG_RD,
621 &iflib_fl_refills, 0, "# refills");
622 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills_large, CTLFLAG_RD,
623 &iflib_fl_refills_large, 0, "# large refills");
626 static int iflib_txq_drain_flushing;
627 static int iflib_txq_drain_oactive;
628 static int iflib_txq_drain_notready;
630 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_flushing, CTLFLAG_RD,
631 &iflib_txq_drain_flushing, 0, "# drain flushes");
632 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_oactive, CTLFLAG_RD,
633 &iflib_txq_drain_oactive, 0, "# drain oactives");
634 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_notready, CTLFLAG_RD,
635 &iflib_txq_drain_notready, 0, "# drain notready");
638 static int iflib_encap_load_mbuf_fail;
639 static int iflib_encap_pad_mbuf_fail;
640 static int iflib_encap_txq_avail_fail;
641 static int iflib_encap_txd_encap_fail;
643 SYSCTL_INT(_net_iflib, OID_AUTO, encap_load_mbuf_fail, CTLFLAG_RD,
644 &iflib_encap_load_mbuf_fail, 0, "# busdma load failures");
645 SYSCTL_INT(_net_iflib, OID_AUTO, encap_pad_mbuf_fail, CTLFLAG_RD,
646 &iflib_encap_pad_mbuf_fail, 0, "# runt frame pad failures");
647 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txq_avail_fail, CTLFLAG_RD,
648 &iflib_encap_txq_avail_fail, 0, "# txq avail failures");
649 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txd_encap_fail, CTLFLAG_RD,
650 &iflib_encap_txd_encap_fail, 0, "# driver encap failures");
652 static int iflib_task_fn_rxs;
653 static int iflib_rx_intr_enables;
654 static int iflib_fast_intrs;
655 static int iflib_rx_unavail;
656 static int iflib_rx_ctx_inactive;
657 static int iflib_rx_if_input;
658 static int iflib_rx_mbuf_null;
659 static int iflib_rxd_flush;
661 static int iflib_verbose_debug;
663 SYSCTL_INT(_net_iflib, OID_AUTO, task_fn_rx, CTLFLAG_RD,
664 &iflib_task_fn_rxs, 0, "# task_fn_rx calls");
665 SYSCTL_INT(_net_iflib, OID_AUTO, rx_intr_enables, CTLFLAG_RD,
666 &iflib_rx_intr_enables, 0, "# rx intr enables");
667 SYSCTL_INT(_net_iflib, OID_AUTO, fast_intrs, CTLFLAG_RD,
668 &iflib_fast_intrs, 0, "# fast_intr calls");
669 SYSCTL_INT(_net_iflib, OID_AUTO, rx_unavail, CTLFLAG_RD,
670 &iflib_rx_unavail, 0, "# times rxeof called with no available data");
671 SYSCTL_INT(_net_iflib, OID_AUTO, rx_ctx_inactive, CTLFLAG_RD,
672 &iflib_rx_ctx_inactive, 0, "# times rxeof called with inactive context");
673 SYSCTL_INT(_net_iflib, OID_AUTO, rx_if_input, CTLFLAG_RD,
674 &iflib_rx_if_input, 0, "# times rxeof called if_input");
675 SYSCTL_INT(_net_iflib, OID_AUTO, rx_mbuf_null, CTLFLAG_RD,
676 &iflib_rx_mbuf_null, 0, "# times rxeof got null mbuf");
677 SYSCTL_INT(_net_iflib, OID_AUTO, rxd_flush, CTLFLAG_RD,
678 &iflib_rxd_flush, 0, "# times rxd_flush called");
679 SYSCTL_INT(_net_iflib, OID_AUTO, verbose_debug, CTLFLAG_RW,
680 &iflib_verbose_debug, 0, "enable verbose debugging");
682 #define DBG_COUNTER_INC(name) atomic_add_int(&(iflib_ ## name), 1)
684 iflib_debug_reset(void)
686 iflib_tx_seen = iflib_tx_sent = iflib_tx_encap = iflib_rx_allocs =
687 iflib_fl_refills = iflib_fl_refills_large = iflib_tx_frees =
688 iflib_txq_drain_flushing = iflib_txq_drain_oactive =
689 iflib_txq_drain_notready =
690 iflib_encap_load_mbuf_fail = iflib_encap_pad_mbuf_fail =
691 iflib_encap_txq_avail_fail = iflib_encap_txd_encap_fail =
692 iflib_task_fn_rxs = iflib_rx_intr_enables = iflib_fast_intrs =
694 iflib_rx_ctx_inactive = iflib_rx_if_input =
695 iflib_rx_mbuf_null = iflib_rxd_flush = 0;
699 #define DBG_COUNTER_INC(name)
700 static void iflib_debug_reset(void) {}
703 #define IFLIB_DEBUG 0
705 static void iflib_tx_structures_free(if_ctx_t ctx);
706 static void iflib_rx_structures_free(if_ctx_t ctx);
707 static int iflib_queues_alloc(if_ctx_t ctx);
708 static int iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq);
709 static int iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget);
710 static int iflib_qset_structures_setup(if_ctx_t ctx);
711 static int iflib_msix_init(if_ctx_t ctx);
712 static int iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filterarg, int *rid, const char *str);
713 static void iflib_txq_check_drain(iflib_txq_t txq, int budget);
714 static uint32_t iflib_txq_can_drain(struct ifmp_ring *);
716 static void iflib_altq_if_start(if_t ifp);
717 static int iflib_altq_if_transmit(if_t ifp, struct mbuf *m);
719 static int iflib_register(if_ctx_t);
720 static void iflib_init_locked(if_ctx_t ctx);
721 static void iflib_add_device_sysctl_pre(if_ctx_t ctx);
722 static void iflib_add_device_sysctl_post(if_ctx_t ctx);
723 static void iflib_ifmp_purge(iflib_txq_t txq);
724 static void _iflib_pre_assert(if_softc_ctx_t scctx);
725 static void iflib_if_init_locked(if_ctx_t ctx);
726 static void iflib_free_intr_mem(if_ctx_t ctx);
727 #ifndef __NO_STRICT_ALIGNMENT
728 static struct mbuf * iflib_fixup_rx(struct mbuf *m);
731 NETDUMP_DEFINE(iflib);
734 #include <sys/selinfo.h>
735 #include <net/netmap.h>
736 #include <dev/netmap/netmap_kern.h>
738 MODULE_DEPEND(iflib, netmap, 1, 1, 1);
740 static int netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init);
743 * device-specific sysctl variables:
745 * iflib_crcstrip: 0: keep CRC in rx frames (default), 1: strip it.
746 * During regular operations the CRC is stripped, but on some
747 * hardware reception of frames not multiple of 64 is slower,
748 * so using crcstrip=0 helps in benchmarks.
750 * iflib_rx_miss, iflib_rx_miss_bufs:
751 * count packets that might be missed due to lost interrupts.
753 SYSCTL_DECL(_dev_netmap);
755 * The xl driver by default strips CRCs and we do not override it.
758 int iflib_crcstrip = 1;
759 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_crcstrip,
760 CTLFLAG_RW, &iflib_crcstrip, 1, "strip CRC on rx frames");
762 int iflib_rx_miss, iflib_rx_miss_bufs;
763 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss,
764 CTLFLAG_RW, &iflib_rx_miss, 0, "potentially missed rx intr");
765 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss_bufs,
766 CTLFLAG_RW, &iflib_rx_miss_bufs, 0, "potentially missed rx intr bufs");
769 * Register/unregister. We are already under netmap lock.
770 * Only called on the first register or the last unregister.
773 iflib_netmap_register(struct netmap_adapter *na, int onoff)
775 struct ifnet *ifp = na->ifp;
776 if_ctx_t ctx = ifp->if_softc;
780 IFDI_INTR_DISABLE(ctx);
782 /* Tell the stack that the interface is no longer active */
783 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
786 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip);
788 /* enable or disable flags and callbacks in na and ifp */
790 nm_set_native_flags(na);
792 nm_clear_native_flags(na);
795 iflib_init_locked(ctx);
796 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); // XXX why twice ?
797 status = ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1;
799 nm_clear_native_flags(na);
805 netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init)
807 struct netmap_adapter *na = kring->na;
808 u_int const lim = kring->nkr_num_slots - 1;
809 u_int head = kring->rhead;
810 struct netmap_ring *ring = kring->ring;
812 struct if_rxd_update iru;
813 if_ctx_t ctx = rxq->ifr_ctx;
814 iflib_fl_t fl = &rxq->ifr_fl[0];
815 uint32_t refill_pidx, nic_i;
816 #if IFLIB_DEBUG_COUNTERS
820 if (nm_i == head && __predict_true(!init))
822 iru_init(&iru, rxq, 0 /* flid */);
823 map = fl->ifl_sds.ifsd_map;
824 refill_pidx = netmap_idx_k2n(kring, nm_i);
826 * IMPORTANT: we must leave one free slot in the ring,
827 * so move head back by one unit
829 head = nm_prev(head, lim);
831 DBG_COUNTER_INC(fl_refills);
832 while (nm_i != head) {
833 #if IFLIB_DEBUG_COUNTERS
835 DBG_COUNTER_INC(fl_refills_large);
837 for (int tmp_pidx = 0; tmp_pidx < IFLIB_MAX_RX_REFRESH && nm_i != head; tmp_pidx++) {
838 struct netmap_slot *slot = &ring->slot[nm_i];
839 void *addr = PNMB(na, slot, &fl->ifl_bus_addrs[tmp_pidx]);
840 uint32_t nic_i_dma = refill_pidx;
841 nic_i = netmap_idx_k2n(kring, nm_i);
843 MPASS(tmp_pidx < IFLIB_MAX_RX_REFRESH);
845 if (addr == NETMAP_BUF_BASE(na)) /* bad buf */
846 return netmap_ring_reinit(kring);
848 fl->ifl_vm_addrs[tmp_pidx] = addr;
849 if (__predict_false(init) && map) {
850 netmap_load_map(na, fl->ifl_ifdi->idi_tag, map[nic_i], addr);
851 } else if (map && (slot->flags & NS_BUF_CHANGED)) {
852 /* buffer has changed, reload map */
853 netmap_reload_map(na, fl->ifl_ifdi->idi_tag, map[nic_i], addr);
855 slot->flags &= ~NS_BUF_CHANGED;
857 nm_i = nm_next(nm_i, lim);
858 fl->ifl_rxd_idxs[tmp_pidx] = nic_i = nm_next(nic_i, lim);
859 if (nm_i != head && tmp_pidx < IFLIB_MAX_RX_REFRESH-1)
862 iru.iru_pidx = refill_pidx;
863 iru.iru_count = tmp_pidx+1;
864 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
870 for (int n = 0; n < iru.iru_count; n++) {
871 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, map[nic_i_dma],
872 BUS_DMASYNC_PREREAD);
873 /* XXX - change this to not use the netmap func*/
874 nic_i_dma = nm_next(nic_i_dma, lim);
878 kring->nr_hwcur = head;
881 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
882 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
883 if (__predict_true(nic_i != UINT_MAX)) {
884 ctx->isc_rxd_flush(ctx->ifc_softc, rxq->ifr_id, fl->ifl_id, nic_i);
885 DBG_COUNTER_INC(rxd_flush);
891 * Reconcile kernel and user view of the transmit ring.
893 * All information is in the kring.
894 * Userspace wants to send packets up to the one before kring->rhead,
895 * kernel knows kring->nr_hwcur is the first unsent packet.
897 * Here we push packets out (as many as possible), and possibly
898 * reclaim buffers from previously completed transmission.
900 * The caller (netmap) guarantees that there is only one instance
901 * running at any time. Any interference with other driver
902 * methods should be handled by the individual drivers.
905 iflib_netmap_txsync(struct netmap_kring *kring, int flags)
907 struct netmap_adapter *na = kring->na;
908 struct ifnet *ifp = na->ifp;
909 struct netmap_ring *ring = kring->ring;
910 u_int nm_i; /* index into the netmap kring */
911 u_int nic_i; /* index into the NIC ring */
913 u_int const lim = kring->nkr_num_slots - 1;
914 u_int const head = kring->rhead;
915 struct if_pkt_info pi;
918 * interrupts on every tx packet are expensive so request
919 * them every half ring, or where NS_REPORT is set
921 u_int report_frequency = kring->nkr_num_slots >> 1;
922 /* device-specific */
923 if_ctx_t ctx = ifp->if_softc;
924 iflib_txq_t txq = &ctx->ifc_txqs[kring->ring_id];
926 bus_dmamap_sync(txq->ift_buf_tag, txq->ift_ifdi->idi_map,
927 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
930 * First part: process new packets to send.
931 * nm_i is the current index in the netmap kring,
932 * nic_i is the corresponding index in the NIC ring.
934 * If we have packets to send (nm_i != head)
935 * iterate over the netmap ring, fetch length and update
936 * the corresponding slot in the NIC ring. Some drivers also
937 * need to update the buffer's physical address in the NIC slot
938 * even NS_BUF_CHANGED is not set (PNMB computes the addresses).
940 * The netmap_reload_map() calls is especially expensive,
941 * even when (as in this case) the tag is 0, so do only
942 * when the buffer has actually changed.
944 * If possible do not set the report/intr bit on all slots,
945 * but only a few times per ring or when NS_REPORT is set.
947 * Finally, on 10G and faster drivers, it might be useful
948 * to prefetch the next slot and txr entry.
951 nm_i = kring->nr_hwcur;
952 if (nm_i != head) { /* we have new packets to send */
954 pi.ipi_segs = txq->ift_segs;
955 pi.ipi_qsidx = kring->ring_id;
956 nic_i = netmap_idx_k2n(kring, nm_i);
958 __builtin_prefetch(&ring->slot[nm_i]);
959 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i]);
960 if (txq->ift_sds.ifsd_map)
961 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i]);
963 for (n = 0; nm_i != head; n++) {
964 struct netmap_slot *slot = &ring->slot[nm_i];
965 u_int len = slot->len;
967 void *addr = PNMB(na, slot, &paddr);
968 int flags = (slot->flags & NS_REPORT ||
969 nic_i == 0 || nic_i == report_frequency) ?
972 /* device-specific */
974 pi.ipi_segs[0].ds_addr = paddr;
975 pi.ipi_segs[0].ds_len = len;
979 pi.ipi_flags = flags;
981 /* Fill the slot in the NIC ring. */
982 ctx->isc_txd_encap(ctx->ifc_softc, &pi);
983 DBG_COUNTER_INC(tx_encap);
985 /* prefetch for next round */
986 __builtin_prefetch(&ring->slot[nm_i + 1]);
987 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i + 1]);
988 if (txq->ift_sds.ifsd_map) {
989 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i + 1]);
991 NM_CHECK_ADDR_LEN(na, addr, len);
993 if (slot->flags & NS_BUF_CHANGED) {
994 /* buffer has changed, reload map */
995 netmap_reload_map(na, txq->ift_buf_tag,
996 txq->ift_sds.ifsd_map[nic_i], addr);
998 /* make sure changes to the buffer are synced */
999 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_sds.ifsd_map[nic_i],
1000 BUS_DMASYNC_PREWRITE);
1002 slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED);
1003 nm_i = nm_next(nm_i, lim);
1004 nic_i = nm_next(nic_i, lim);
1006 kring->nr_hwcur = nm_i;
1008 /* synchronize the NIC ring */
1009 bus_dmamap_sync(txq->ift_buf_tag, txq->ift_ifdi->idi_map,
1010 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1012 /* (re)start the tx unit up to slot nic_i (excluded) */
1013 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, nic_i);
1017 * Second part: reclaim buffers for completed transmissions.
1019 * If there are unclaimed buffers, attempt to reclaim them.
1020 * If none are reclaimed, and TX IRQs are not in use, do an initial
1021 * minimal delay, then trigger the tx handler which will spin in the
1024 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) {
1025 if (iflib_tx_credits_update(ctx, txq)) {
1026 /* some tx completed, increment avail */
1027 nic_i = txq->ift_cidx_processed;
1028 kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), lim);
1031 if (!(ctx->ifc_flags & IFC_NETMAP_TX_IRQ))
1032 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) {
1033 callout_reset_on(&txq->ift_timer, hz < 2000 ? 1 : hz / 1000,
1034 iflib_timer, txq, txq->ift_timer.c_cpu);
1040 * Reconcile kernel and user view of the receive ring.
1041 * Same as for the txsync, this routine must be efficient.
1042 * The caller guarantees a single invocations, but races against
1043 * the rest of the driver should be handled here.
1045 * On call, kring->rhead is the first packet that userspace wants
1046 * to keep, and kring->rcur is the wakeup point.
1047 * The kernel has previously reported packets up to kring->rtail.
1049 * If (flags & NAF_FORCE_READ) also check for incoming packets irrespective
1050 * of whether or not we received an interrupt.
1053 iflib_netmap_rxsync(struct netmap_kring *kring, int flags)
1055 struct netmap_adapter *na = kring->na;
1056 struct netmap_ring *ring = kring->ring;
1057 uint32_t nm_i; /* index into the netmap ring */
1058 uint32_t nic_i; /* index into the NIC ring */
1060 u_int const lim = kring->nkr_num_slots - 1;
1061 u_int const head = kring->rhead;
1062 int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR;
1063 struct if_rxd_info ri;
1065 struct ifnet *ifp = na->ifp;
1066 if_ctx_t ctx = ifp->if_softc;
1067 iflib_rxq_t rxq = &ctx->ifc_rxqs[kring->ring_id];
1068 iflib_fl_t fl = rxq->ifr_fl;
1070 return netmap_ring_reinit(kring);
1072 /* XXX check sync modes */
1073 for (i = 0, fl = rxq->ifr_fl; i < rxq->ifr_nfl; i++, fl++) {
1074 if (fl->ifl_sds.ifsd_map == NULL)
1076 bus_dmamap_sync(rxq->ifr_fl[i].ifl_buf_tag,
1077 fl->ifl_ifdi->idi_map,
1078 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
1081 * First part: import newly received packets.
1083 * nm_i is the index of the next free slot in the netmap ring,
1084 * nic_i is the index of the next received packet in the NIC ring,
1085 * and they may differ in case if_init() has been called while
1086 * in netmap mode. For the receive ring we have
1088 * nic_i = rxr->next_check;
1089 * nm_i = kring->nr_hwtail (previous)
1091 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size
1093 * rxr->next_check is set to 0 on a ring reinit
1095 if (netmap_no_pendintr || force_update) {
1096 int crclen = iflib_crcstrip ? 0 : 4;
1099 for (i = 0; i < rxq->ifr_nfl; i++) {
1100 fl = &rxq->ifr_fl[i];
1101 nic_i = fl->ifl_cidx;
1102 nm_i = netmap_idx_n2k(kring, nic_i);
1103 avail = iflib_rxd_avail(ctx, rxq, nic_i, USHRT_MAX);
1104 for (n = 0; avail > 0; n++, avail--) {
1106 ri.iri_frags = rxq->ifr_frags;
1107 ri.iri_qsidx = kring->ring_id;
1108 ri.iri_ifp = ctx->ifc_ifp;
1109 ri.iri_cidx = nic_i;
1111 error = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri);
1112 ring->slot[nm_i].len = error ? 0 : ri.iri_len - crclen;
1113 ring->slot[nm_i].flags = 0;
1114 bus_dmamap_sync(fl->ifl_ifdi->idi_tag,
1115 fl->ifl_sds.ifsd_map[nic_i], BUS_DMASYNC_POSTREAD);
1116 nm_i = nm_next(nm_i, lim);
1117 nic_i = nm_next(nic_i, lim);
1119 if (n) { /* update the state variables */
1120 if (netmap_no_pendintr && !force_update) {
1123 iflib_rx_miss_bufs += n;
1125 fl->ifl_cidx = nic_i;
1126 kring->nr_hwtail = nm_i;
1128 kring->nr_kflags &= ~NKR_PENDINTR;
1132 * Second part: skip past packets that userspace has released.
1133 * (kring->nr_hwcur to head excluded),
1134 * and make the buffers available for reception.
1135 * As usual nm_i is the index in the netmap ring,
1136 * nic_i is the index in the NIC ring, and
1137 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size
1139 /* XXX not sure how this will work with multiple free lists */
1140 nm_i = kring->nr_hwcur;
1142 return (netmap_fl_refill(rxq, kring, nm_i, false));
1146 iflib_netmap_intr(struct netmap_adapter *na, int onoff)
1148 struct ifnet *ifp = na->ifp;
1149 if_ctx_t ctx = ifp->if_softc;
1153 IFDI_INTR_ENABLE(ctx);
1155 IFDI_INTR_DISABLE(ctx);
1162 iflib_netmap_attach(if_ctx_t ctx)
1164 struct netmap_adapter na;
1165 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1167 bzero(&na, sizeof(na));
1169 na.ifp = ctx->ifc_ifp;
1170 na.na_flags = NAF_BDG_MAYSLEEP;
1171 MPASS(ctx->ifc_softc_ctx.isc_ntxqsets);
1172 MPASS(ctx->ifc_softc_ctx.isc_nrxqsets);
1174 na.num_tx_desc = scctx->isc_ntxd[0];
1175 na.num_rx_desc = scctx->isc_nrxd[0];
1176 na.nm_txsync = iflib_netmap_txsync;
1177 na.nm_rxsync = iflib_netmap_rxsync;
1178 na.nm_register = iflib_netmap_register;
1179 na.nm_intr = iflib_netmap_intr;
1180 na.num_tx_rings = ctx->ifc_softc_ctx.isc_ntxqsets;
1181 na.num_rx_rings = ctx->ifc_softc_ctx.isc_nrxqsets;
1182 return (netmap_attach(&na));
1186 iflib_netmap_txq_init(if_ctx_t ctx, iflib_txq_t txq)
1188 struct netmap_adapter *na = NA(ctx->ifc_ifp);
1189 struct netmap_slot *slot;
1191 slot = netmap_reset(na, NR_TX, txq->ift_id, 0);
1194 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxd[0]; i++) {
1197 * In netmap mode, set the map for the packet buffer.
1198 * NOTE: Some drivers (not this one) also need to set
1199 * the physical buffer address in the NIC ring.
1200 * netmap_idx_n2k() maps a nic index, i, into the corresponding
1201 * netmap slot index, si
1203 int si = netmap_idx_n2k(na->tx_rings[txq->ift_id], i);
1204 netmap_load_map(na, txq->ift_buf_tag, txq->ift_sds.ifsd_map[i],
1205 NMB(na, slot + si));
1210 iflib_netmap_rxq_init(if_ctx_t ctx, iflib_rxq_t rxq)
1212 struct netmap_adapter *na = NA(ctx->ifc_ifp);
1213 struct netmap_kring *kring = na->rx_rings[rxq->ifr_id];
1214 struct netmap_slot *slot;
1217 slot = netmap_reset(na, NR_RX, rxq->ifr_id, 0);
1220 nm_i = netmap_idx_n2k(kring, 0);
1221 netmap_fl_refill(rxq, kring, nm_i, true);
1225 iflib_netmap_timer_adjust(if_ctx_t ctx, uint16_t txqid, uint32_t *reset_on)
1227 struct netmap_kring *kring;
1229 kring = NA(ctx->ifc_ifp)->tx_rings[txqid];
1231 if (kring->nr_hwcur != nm_next(kring->nr_hwtail, kring->nkr_num_slots - 1)) {
1232 if (ctx->isc_txd_credits_update(ctx->ifc_softc, txqid, false))
1233 netmap_tx_irq(ctx->ifc_ifp, txqid);
1234 if (!(ctx->ifc_flags & IFC_NETMAP_TX_IRQ)) {
1238 *reset_on = hz / 1000;
1243 #define iflib_netmap_detach(ifp) netmap_detach(ifp)
1246 #define iflib_netmap_txq_init(ctx, txq)
1247 #define iflib_netmap_rxq_init(ctx, rxq)
1248 #define iflib_netmap_detach(ifp)
1250 #define iflib_netmap_attach(ctx) (0)
1251 #define netmap_rx_irq(ifp, qid, budget) (0)
1252 #define netmap_tx_irq(ifp, qid) do {} while (0)
1253 #define iflib_netmap_timer_adjust(ctx, txqid, reset_on)
1257 #if defined(__i386__) || defined(__amd64__)
1258 static __inline void
1261 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x));
1263 static __inline void
1264 prefetch2cachelines(void *x)
1266 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x));
1267 #if (CACHE_LINE_SIZE < 128)
1268 __asm volatile("prefetcht0 %0" :: "m" (*(((unsigned long *)x)+CACHE_LINE_SIZE/(sizeof(unsigned long)))));
1273 #define prefetch2cachelines(x)
1277 iflib_gen_mac(if_ctx_t ctx)
1281 char uuid[HOSTUUIDLEN+1];
1282 char buf[HOSTUUIDLEN+16];
1284 unsigned char digest[16];
1288 uuid[HOSTUUIDLEN] = 0;
1289 bcopy(td->td_ucred->cr_prison->pr_hostuuid, uuid, HOSTUUIDLEN);
1290 snprintf(buf, HOSTUUIDLEN+16, "%s-%s", uuid, device_get_nameunit(ctx->ifc_dev));
1292 * Generate a pseudo-random, deterministic MAC
1293 * address based on the UUID and unit number.
1294 * The FreeBSD Foundation OUI of 58-9C-FC is used.
1297 MD5Update(&mdctx, buf, strlen(buf));
1298 MD5Final(digest, &mdctx);
1309 iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid)
1313 fl = &rxq->ifr_fl[flid];
1314 iru->iru_paddrs = fl->ifl_bus_addrs;
1315 iru->iru_vaddrs = &fl->ifl_vm_addrs[0];
1316 iru->iru_idxs = fl->ifl_rxd_idxs;
1317 iru->iru_qsidx = rxq->ifr_id;
1318 iru->iru_buf_size = fl->ifl_buf_size;
1319 iru->iru_flidx = fl->ifl_id;
1323 _iflib_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int err)
1327 *(bus_addr_t *) arg = segs[0].ds_addr;
1331 iflib_dma_alloc_align(if_ctx_t ctx, int size, int align, iflib_dma_info_t dma, int mapflags)
1334 device_t dev = ctx->ifc_dev;
1336 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
1337 align, 0, /* alignment, bounds */
1338 BUS_SPACE_MAXADDR, /* lowaddr */
1339 BUS_SPACE_MAXADDR, /* highaddr */
1340 NULL, NULL, /* filter, filterarg */
1343 size, /* maxsegsize */
1344 BUS_DMA_ALLOCNOW, /* flags */
1345 NULL, /* lockfunc */
1350 "%s: bus_dma_tag_create failed: %d\n",
1355 err = bus_dmamem_alloc(dma->idi_tag, (void**) &dma->idi_vaddr,
1356 BUS_DMA_NOWAIT | BUS_DMA_COHERENT | BUS_DMA_ZERO, &dma->idi_map);
1359 "%s: bus_dmamem_alloc(%ju) failed: %d\n",
1360 __func__, (uintmax_t)size, err);
1364 dma->idi_paddr = IF_BAD_DMA;
1365 err = bus_dmamap_load(dma->idi_tag, dma->idi_map, dma->idi_vaddr,
1366 size, _iflib_dmamap_cb, &dma->idi_paddr, mapflags | BUS_DMA_NOWAIT);
1367 if (err || dma->idi_paddr == IF_BAD_DMA) {
1369 "%s: bus_dmamap_load failed: %d\n",
1374 dma->idi_size = size;
1378 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map);
1380 bus_dma_tag_destroy(dma->idi_tag);
1382 dma->idi_tag = NULL;
1388 iflib_dma_alloc(if_ctx_t ctx, int size, iflib_dma_info_t dma, int mapflags)
1390 if_shared_ctx_t sctx = ctx->ifc_sctx;
1392 KASSERT(sctx->isc_q_align != 0, ("alignment value not initialized"));
1394 return (iflib_dma_alloc_align(ctx, size, sctx->isc_q_align, dma, mapflags));
1398 iflib_dma_alloc_multi(if_ctx_t ctx, int *sizes, iflib_dma_info_t *dmalist, int mapflags, int count)
1401 iflib_dma_info_t *dmaiter;
1404 for (i = 0; i < count; i++, dmaiter++) {
1405 if ((err = iflib_dma_alloc(ctx, sizes[i], *dmaiter, mapflags)) != 0)
1409 iflib_dma_free_multi(dmalist, i);
1414 iflib_dma_free(iflib_dma_info_t dma)
1416 if (dma->idi_tag == NULL)
1418 if (dma->idi_paddr != IF_BAD_DMA) {
1419 bus_dmamap_sync(dma->idi_tag, dma->idi_map,
1420 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
1421 bus_dmamap_unload(dma->idi_tag, dma->idi_map);
1422 dma->idi_paddr = IF_BAD_DMA;
1424 if (dma->idi_vaddr != NULL) {
1425 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map);
1426 dma->idi_vaddr = NULL;
1428 bus_dma_tag_destroy(dma->idi_tag);
1429 dma->idi_tag = NULL;
1433 iflib_dma_free_multi(iflib_dma_info_t *dmalist, int count)
1436 iflib_dma_info_t *dmaiter = dmalist;
1438 for (i = 0; i < count; i++, dmaiter++)
1439 iflib_dma_free(*dmaiter);
1442 #ifdef EARLY_AP_STARTUP
1443 static const int iflib_started = 1;
1446 * We used to abuse the smp_started flag to decide if the queues have been
1447 * fully initialized (by late taskqgroup_adjust() calls in a SYSINIT()).
1448 * That gave bad races, since the SYSINIT() runs strictly after smp_started
1449 * is set. Run a SYSINIT() strictly after that to just set a usable
1453 static int iflib_started;
1456 iflib_record_started(void *arg)
1461 SYSINIT(iflib_record_started, SI_SUB_SMP + 1, SI_ORDER_FIRST,
1462 iflib_record_started, NULL);
1466 iflib_fast_intr(void *arg)
1468 iflib_filter_info_t info = arg;
1469 struct grouptask *gtask = info->ifi_task;
1473 return (FILTER_STRAY);
1475 DBG_COUNTER_INC(fast_intrs);
1476 if (info->ifi_filter != NULL) {
1477 result = info->ifi_filter(info->ifi_filter_arg);
1478 if ((result & FILTER_SCHEDULE_THREAD) == 0)
1482 GROUPTASK_ENQUEUE(gtask);
1483 return (FILTER_HANDLED);
1487 iflib_fast_intr_rxtx(void *arg)
1489 iflib_filter_info_t info = arg;
1490 struct grouptask *gtask = info->ifi_task;
1491 iflib_rxq_t rxq = (iflib_rxq_t)info->ifi_ctx;
1492 if_ctx_t ctx = NULL;;
1493 int i, cidx, result;
1496 return (FILTER_STRAY);
1498 DBG_COUNTER_INC(fast_intrs);
1499 if (info->ifi_filter != NULL) {
1500 result = info->ifi_filter(info->ifi_filter_arg);
1501 if ((result & FILTER_SCHEDULE_THREAD) == 0)
1505 MPASS(rxq->ifr_ntxqirq);
1506 for (i = 0; i < rxq->ifr_ntxqirq; i++) {
1507 qidx_t txqid = rxq->ifr_txqid[i];
1511 bus_dmamap_sync(rxq->ifr_ifdi->idi_tag, rxq->ifr_ifdi->idi_map,
1512 BUS_DMASYNC_POSTREAD);
1513 if (!ctx->isc_txd_credits_update(ctx->ifc_softc, txqid, false)) {
1514 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txqid);
1517 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task);
1519 if (ctx->ifc_sctx->isc_flags & IFLIB_HAS_RXCQ)
1520 cidx = rxq->ifr_cq_cidx;
1522 cidx = rxq->ifr_fl[0].ifl_cidx;
1523 if (iflib_rxd_avail(ctx, rxq, cidx, 1))
1524 GROUPTASK_ENQUEUE(gtask);
1526 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id);
1527 DBG_COUNTER_INC(rx_intr_enables);
1529 return (FILTER_HANDLED);
1534 iflib_fast_intr_ctx(void *arg)
1536 iflib_filter_info_t info = arg;
1537 struct grouptask *gtask = info->ifi_task;
1541 return (FILTER_STRAY);
1543 DBG_COUNTER_INC(fast_intrs);
1544 if (info->ifi_filter != NULL) {
1545 result = info->ifi_filter(info->ifi_filter_arg);
1546 if ((result & FILTER_SCHEDULE_THREAD) == 0)
1550 GROUPTASK_ENQUEUE(gtask);
1551 return (FILTER_HANDLED);
1555 _iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid,
1556 driver_filter_t filter, driver_intr_t handler, void *arg,
1560 struct resource *res;
1562 device_t dev = ctx->ifc_dev;
1565 if (ctx->ifc_flags & IFC_LEGACY)
1566 flags |= RF_SHAREABLE;
1569 res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &irq->ii_rid, flags);
1572 "failed to allocate IRQ for rid %d, name %s.\n", rid, name);
1576 KASSERT(filter == NULL || handler == NULL, ("filter and handler can't both be non-NULL"));
1577 rc = bus_setup_intr(dev, res, INTR_MPSAFE | INTR_TYPE_NET,
1578 filter, handler, arg, &tag);
1581 "failed to setup interrupt for rid %d, name %s: %d\n",
1582 rid, name ? name : "unknown", rc);
1585 bus_describe_intr(dev, res, tag, "%s", name);
1592 /*********************************************************************
1594 * Allocate DMA resources for TX buffers as well as memory for the TX
1595 * mbuf map. TX DMA maps (non-TSO/TSO) and TX mbuf map are kept in a
1596 * iflib_sw_tx_desc_array structure, storing all the information that
1597 * is needed to transmit a packet on the wire. This is called only
1598 * once at attach, setup is done every reset.
1600 **********************************************************************/
1602 iflib_txsd_alloc(iflib_txq_t txq)
1604 if_ctx_t ctx = txq->ift_ctx;
1605 if_shared_ctx_t sctx = ctx->ifc_sctx;
1606 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1607 device_t dev = ctx->ifc_dev;
1608 bus_size_t tsomaxsize;
1609 int err, nsegments, ntsosegments;
1612 nsegments = scctx->isc_tx_nsegments;
1613 ntsosegments = scctx->isc_tx_tso_segments_max;
1614 tsomaxsize = scctx->isc_tx_tso_size_max;
1615 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_VLAN_MTU)
1616 tsomaxsize += sizeof(struct ether_vlan_header);
1617 MPASS(scctx->isc_ntxd[0] > 0);
1618 MPASS(scctx->isc_ntxd[txq->ift_br_offset] > 0);
1619 MPASS(nsegments > 0);
1620 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) {
1621 MPASS(ntsosegments > 0);
1622 MPASS(sctx->isc_tso_maxsize >= tsomaxsize);
1626 * Set up DMA tags for TX buffers.
1628 if ((err = bus_dma_tag_create(bus_get_dma_tag(dev),
1629 1, 0, /* alignment, bounds */
1630 BUS_SPACE_MAXADDR, /* lowaddr */
1631 BUS_SPACE_MAXADDR, /* highaddr */
1632 NULL, NULL, /* filter, filterarg */
1633 sctx->isc_tx_maxsize, /* maxsize */
1634 nsegments, /* nsegments */
1635 sctx->isc_tx_maxsegsize, /* maxsegsize */
1637 NULL, /* lockfunc */
1638 NULL, /* lockfuncarg */
1639 &txq->ift_buf_tag))) {
1640 device_printf(dev,"Unable to allocate TX DMA tag: %d\n", err);
1641 device_printf(dev,"maxsize: %ju nsegments: %d maxsegsize: %ju\n",
1642 (uintmax_t)sctx->isc_tx_maxsize, nsegments, (uintmax_t)sctx->isc_tx_maxsegsize);
1645 tso = (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) != 0;
1646 if (tso && (err = bus_dma_tag_create(bus_get_dma_tag(dev),
1647 1, 0, /* alignment, bounds */
1648 BUS_SPACE_MAXADDR, /* lowaddr */
1649 BUS_SPACE_MAXADDR, /* highaddr */
1650 NULL, NULL, /* filter, filterarg */
1651 tsomaxsize, /* maxsize */
1652 ntsosegments, /* nsegments */
1653 sctx->isc_tso_maxsegsize,/* maxsegsize */
1655 NULL, /* lockfunc */
1656 NULL, /* lockfuncarg */
1657 &txq->ift_tso_buf_tag))) {
1658 device_printf(dev, "Unable to allocate TSO TX DMA tag: %d\n",
1663 /* Allocate memory for the TX mbuf map. */
1664 if (!(txq->ift_sds.ifsd_m =
1665 (struct mbuf **) malloc(sizeof(struct mbuf *) *
1666 scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1667 device_printf(dev, "Unable to allocate TX mbuf map memory\n");
1673 * Create the DMA maps for TX buffers.
1675 if ((txq->ift_sds.ifsd_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 TX buffer DMA map memory\n");
1683 if (tso && (txq->ift_sds.ifsd_tso_map = (bus_dmamap_t *)malloc(
1684 sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset],
1685 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
1687 "Unable to allocate TSO TX buffer map memory\n");
1691 for (int i = 0; i < scctx->isc_ntxd[txq->ift_br_offset]; i++) {
1692 err = bus_dmamap_create(txq->ift_buf_tag, 0,
1693 &txq->ift_sds.ifsd_map[i]);
1695 device_printf(dev, "Unable to create TX DMA map\n");
1700 err = bus_dmamap_create(txq->ift_tso_buf_tag, 0,
1701 &txq->ift_sds.ifsd_tso_map[i]);
1703 device_printf(dev, "Unable to create TSO TX DMA map\n");
1709 /* We free all, it handles case where we are in the middle */
1710 iflib_tx_structures_free(ctx);
1715 iflib_txsd_destroy(if_ctx_t ctx, iflib_txq_t txq, int i)
1720 if (txq->ift_sds.ifsd_map != NULL)
1721 map = txq->ift_sds.ifsd_map[i];
1723 bus_dmamap_sync(txq->ift_buf_tag, map, BUS_DMASYNC_POSTWRITE);
1724 bus_dmamap_unload(txq->ift_buf_tag, map);
1725 bus_dmamap_destroy(txq->ift_buf_tag, map);
1726 txq->ift_sds.ifsd_map[i] = NULL;
1730 if (txq->ift_sds.ifsd_tso_map != NULL)
1731 map = txq->ift_sds.ifsd_tso_map[i];
1733 bus_dmamap_sync(txq->ift_tso_buf_tag, map,
1734 BUS_DMASYNC_POSTWRITE);
1735 bus_dmamap_unload(txq->ift_tso_buf_tag, map);
1736 bus_dmamap_destroy(txq->ift_tso_buf_tag, map);
1737 txq->ift_sds.ifsd_tso_map[i] = NULL;
1742 iflib_txq_destroy(iflib_txq_t txq)
1744 if_ctx_t ctx = txq->ift_ctx;
1746 for (int i = 0; i < txq->ift_size; i++)
1747 iflib_txsd_destroy(ctx, txq, i);
1748 if (txq->ift_sds.ifsd_map != NULL) {
1749 free(txq->ift_sds.ifsd_map, M_IFLIB);
1750 txq->ift_sds.ifsd_map = NULL;
1752 if (txq->ift_sds.ifsd_tso_map != NULL) {
1753 free(txq->ift_sds.ifsd_tso_map, M_IFLIB);
1754 txq->ift_sds.ifsd_tso_map = NULL;
1756 if (txq->ift_sds.ifsd_m != NULL) {
1757 free(txq->ift_sds.ifsd_m, M_IFLIB);
1758 txq->ift_sds.ifsd_m = NULL;
1760 if (txq->ift_buf_tag != NULL) {
1761 bus_dma_tag_destroy(txq->ift_buf_tag);
1762 txq->ift_buf_tag = NULL;
1764 if (txq->ift_tso_buf_tag != NULL) {
1765 bus_dma_tag_destroy(txq->ift_tso_buf_tag);
1766 txq->ift_tso_buf_tag = NULL;
1771 iflib_txsd_free(if_ctx_t ctx, iflib_txq_t txq, int i)
1775 mp = &txq->ift_sds.ifsd_m[i];
1779 if (txq->ift_sds.ifsd_map != NULL) {
1780 bus_dmamap_sync(txq->ift_buf_tag,
1781 txq->ift_sds.ifsd_map[i], BUS_DMASYNC_POSTWRITE);
1782 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[i]);
1784 if (txq->ift_sds.ifsd_tso_map != NULL) {
1785 bus_dmamap_sync(txq->ift_tso_buf_tag,
1786 txq->ift_sds.ifsd_tso_map[i], BUS_DMASYNC_POSTWRITE);
1787 bus_dmamap_unload(txq->ift_tso_buf_tag,
1788 txq->ift_sds.ifsd_tso_map[i]);
1791 DBG_COUNTER_INC(tx_frees);
1796 iflib_txq_setup(iflib_txq_t txq)
1798 if_ctx_t ctx = txq->ift_ctx;
1799 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1800 if_shared_ctx_t sctx = ctx->ifc_sctx;
1801 iflib_dma_info_t di;
1804 /* Set number of descriptors available */
1805 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
1806 /* XXX make configurable */
1807 txq->ift_update_freq = IFLIB_DEFAULT_TX_UPDATE_FREQ;
1810 txq->ift_cidx_processed = 0;
1811 txq->ift_pidx = txq->ift_cidx = txq->ift_npending = 0;
1812 txq->ift_size = scctx->isc_ntxd[txq->ift_br_offset];
1814 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++)
1815 bzero((void *)di->idi_vaddr, di->idi_size);
1817 IFDI_TXQ_SETUP(ctx, txq->ift_id);
1818 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++)
1819 bus_dmamap_sync(di->idi_tag, di->idi_map,
1820 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1824 /*********************************************************************
1826 * Allocate DMA resources for RX buffers as well as memory for the RX
1827 * mbuf map, direct RX cluster pointer map and RX cluster bus address
1828 * map. RX DMA map, RX mbuf map, direct RX cluster pointer map and
1829 * RX cluster map are kept in a iflib_sw_rx_desc_array structure.
1830 * Since we use use one entry in iflib_sw_rx_desc_array per received
1831 * packet, the maximum number of entries we'll need is equal to the
1832 * number of hardware receive descriptors that we've allocated.
1834 **********************************************************************/
1836 iflib_rxsd_alloc(iflib_rxq_t rxq)
1838 if_ctx_t ctx = rxq->ifr_ctx;
1839 if_shared_ctx_t sctx = ctx->ifc_sctx;
1840 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
1841 device_t dev = ctx->ifc_dev;
1845 MPASS(scctx->isc_nrxd[0] > 0);
1846 MPASS(scctx->isc_nrxd[rxq->ifr_fl_offset] > 0);
1849 for (int i = 0; i < rxq->ifr_nfl; i++, fl++) {
1850 fl->ifl_size = scctx->isc_nrxd[rxq->ifr_fl_offset]; /* this isn't necessarily the same */
1851 /* Set up DMA tag for RX buffers. */
1852 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
1853 1, 0, /* alignment, bounds */
1854 BUS_SPACE_MAXADDR, /* lowaddr */
1855 BUS_SPACE_MAXADDR, /* highaddr */
1856 NULL, NULL, /* filter, filterarg */
1857 sctx->isc_rx_maxsize, /* maxsize */
1858 sctx->isc_rx_nsegments, /* nsegments */
1859 sctx->isc_rx_maxsegsize, /* maxsegsize */
1861 NULL, /* lockfunc */
1866 "Unable to allocate RX DMA tag: %d\n", err);
1870 /* Allocate memory for the RX mbuf map. */
1871 if (!(fl->ifl_sds.ifsd_m =
1872 (struct mbuf **) malloc(sizeof(struct mbuf *) *
1873 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1875 "Unable to allocate RX mbuf map memory\n");
1880 /* Allocate memory for the direct RX cluster pointer map. */
1881 if (!(fl->ifl_sds.ifsd_cl =
1882 (caddr_t *) malloc(sizeof(caddr_t) *
1883 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1885 "Unable to allocate RX cluster map memory\n");
1890 /* Allocate memory for the RX cluster bus address map. */
1891 if (!(fl->ifl_sds.ifsd_ba =
1892 (bus_addr_t *) malloc(sizeof(bus_addr_t) *
1893 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1895 "Unable to allocate RX bus address map memory\n");
1901 * Create the DMA maps for RX buffers.
1903 if (!(fl->ifl_sds.ifsd_map =
1904 (bus_dmamap_t *) malloc(sizeof(bus_dmamap_t) * scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) {
1906 "Unable to allocate RX buffer DMA map memory\n");
1910 for (int i = 0; i < scctx->isc_nrxd[rxq->ifr_fl_offset]; i++) {
1911 err = bus_dmamap_create(fl->ifl_buf_tag, 0,
1912 &fl->ifl_sds.ifsd_map[i]);
1914 device_printf(dev, "Unable to create RX buffer DMA map\n");
1922 iflib_rx_structures_free(ctx);
1928 * Internal service routines
1931 struct rxq_refill_cb_arg {
1933 bus_dma_segment_t seg;
1938 _rxq_refill_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
1940 struct rxq_refill_cb_arg *cb_arg = arg;
1942 cb_arg->error = error;
1943 cb_arg->seg = segs[0];
1944 cb_arg->nseg = nseg;
1948 * rxq_refill - refill an rxq free-buffer list
1949 * @ctx: the iflib context
1950 * @rxq: the free-list to refill
1951 * @n: the number of new buffers to allocate
1953 * (Re)populate an rxq free-buffer list with up to @n new packet buffers.
1954 * The caller must assure that @n does not exceed the queue's capacity.
1957 _iflib_fl_refill(if_ctx_t ctx, iflib_fl_t fl, int count)
1959 struct if_rxd_update iru;
1960 struct rxq_refill_cb_arg cb_arg;
1964 bus_dmamap_t *sd_map;
1965 bus_addr_t bus_addr, *sd_ba;
1966 int err, frag_idx, i, idx, n, pidx;
1969 sd_m = fl->ifl_sds.ifsd_m;
1970 sd_map = fl->ifl_sds.ifsd_map;
1971 sd_cl = fl->ifl_sds.ifsd_cl;
1972 sd_ba = fl->ifl_sds.ifsd_ba;
1973 pidx = fl->ifl_pidx;
1975 frag_idx = fl->ifl_fragidx;
1976 credits = fl->ifl_credits;
1981 MPASS(credits + n <= fl->ifl_size);
1983 if (pidx < fl->ifl_cidx)
1984 MPASS(pidx + n <= fl->ifl_cidx);
1985 if (pidx == fl->ifl_cidx && (credits < fl->ifl_size))
1986 MPASS(fl->ifl_gen == 0);
1987 if (pidx > fl->ifl_cidx)
1988 MPASS(n <= fl->ifl_size - pidx + fl->ifl_cidx);
1990 DBG_COUNTER_INC(fl_refills);
1992 DBG_COUNTER_INC(fl_refills_large);
1993 iru_init(&iru, fl->ifl_rxq, fl->ifl_id);
1996 * We allocate an uninitialized mbuf + cluster, mbuf is
1997 * initialized after rx.
1999 * If the cluster is still set then we know a minimum sized packet was received
2001 bit_ffc_at(fl->ifl_rx_bitmap, frag_idx, fl->ifl_size,
2004 bit_ffc(fl->ifl_rx_bitmap, fl->ifl_size, &frag_idx);
2005 MPASS(frag_idx >= 0);
2006 if ((cl = sd_cl[frag_idx]) == NULL) {
2007 if ((cl = m_cljget(NULL, M_NOWAIT, fl->ifl_buf_size)) == NULL)
2011 MPASS(sd_map != NULL);
2012 err = bus_dmamap_load(fl->ifl_buf_tag, sd_map[frag_idx],
2013 cl, fl->ifl_buf_size, _rxq_refill_cb, &cb_arg,
2015 if (err != 0 || cb_arg.error) {
2019 if (fl->ifl_zone == zone_pack)
2020 uma_zfree(fl->ifl_zone, cl);
2024 bus_dmamap_sync(fl->ifl_buf_tag, sd_map[frag_idx],
2025 BUS_DMASYNC_PREREAD);
2026 sd_ba[frag_idx] = bus_addr = cb_arg.seg.ds_addr;
2027 sd_cl[frag_idx] = cl;
2029 fl->ifl_cl_enqueued++;
2032 bus_addr = sd_ba[frag_idx];
2035 MPASS(sd_m[frag_idx] == NULL);
2036 if ((m = m_gethdr(M_NOWAIT, MT_NOINIT)) == NULL) {
2040 bit_set(fl->ifl_rx_bitmap, frag_idx);
2042 fl->ifl_m_enqueued++;
2045 DBG_COUNTER_INC(rx_allocs);
2046 fl->ifl_rxd_idxs[i] = frag_idx;
2047 fl->ifl_bus_addrs[i] = bus_addr;
2048 fl->ifl_vm_addrs[i] = cl;
2051 MPASS(credits <= fl->ifl_size);
2052 if (++idx == fl->ifl_size) {
2056 if (n == 0 || i == IFLIB_MAX_RX_REFRESH) {
2057 iru.iru_pidx = pidx;
2059 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
2063 fl->ifl_credits = credits;
2068 iru.iru_pidx = pidx;
2070 ctx->isc_rxd_refill(ctx->ifc_softc, &iru);
2072 fl->ifl_credits = credits;
2074 DBG_COUNTER_INC(rxd_flush);
2075 if (fl->ifl_pidx == 0)
2076 pidx = fl->ifl_size - 1;
2078 pidx = fl->ifl_pidx - 1;
2080 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
2081 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2082 ctx->isc_rxd_flush(ctx->ifc_softc, fl->ifl_rxq->ifr_id, fl->ifl_id, pidx);
2083 fl->ifl_fragidx = frag_idx;
2086 static __inline void
2087 __iflib_fl_refill_lt(if_ctx_t ctx, iflib_fl_t fl, int max)
2089 /* we avoid allowing pidx to catch up with cidx as it confuses ixl */
2090 int32_t reclaimable = fl->ifl_size - fl->ifl_credits - 1;
2092 int32_t delta = fl->ifl_size - get_inuse(fl->ifl_size, fl->ifl_cidx, fl->ifl_pidx, fl->ifl_gen) - 1;
2095 MPASS(fl->ifl_credits <= fl->ifl_size);
2096 MPASS(reclaimable == delta);
2098 if (reclaimable > 0)
2099 _iflib_fl_refill(ctx, fl, min(max, reclaimable));
2103 iflib_in_detach(if_ctx_t ctx)
2107 in_detach = !!(ctx->ifc_flags & IFC_IN_DETACH);
2113 iflib_fl_bufs_free(iflib_fl_t fl)
2115 iflib_dma_info_t idi = fl->ifl_ifdi;
2116 bus_dmamap_t sd_map;
2119 for (i = 0; i < fl->ifl_size; i++) {
2120 struct mbuf **sd_m = &fl->ifl_sds.ifsd_m[i];
2121 caddr_t *sd_cl = &fl->ifl_sds.ifsd_cl[i];
2123 if (*sd_cl != NULL) {
2124 sd_map = fl->ifl_sds.ifsd_map[i];
2125 bus_dmamap_sync(fl->ifl_buf_tag, sd_map,
2126 BUS_DMASYNC_POSTREAD);
2127 bus_dmamap_unload(fl->ifl_buf_tag, sd_map);
2129 uma_zfree(fl->ifl_zone, *sd_cl);
2130 // XXX: Should this get moved out?
2131 if (iflib_in_detach(fl->ifl_rxq->ifr_ctx))
2132 bus_dmamap_destroy(fl->ifl_buf_tag, sd_map);
2133 if (*sd_m != NULL) {
2134 m_init(*sd_m, M_NOWAIT, MT_DATA, 0);
2135 uma_zfree(zone_mbuf, *sd_m);
2138 MPASS(*sd_cl == NULL);
2139 MPASS(*sd_m == NULL);
2142 fl->ifl_m_dequeued++;
2143 fl->ifl_cl_dequeued++;
2149 for (i = 0; i < fl->ifl_size; i++) {
2150 MPASS(fl->ifl_sds.ifsd_cl[i] == NULL);
2151 MPASS(fl->ifl_sds.ifsd_m[i] == NULL);
2155 * Reset free list values
2157 fl->ifl_credits = fl->ifl_cidx = fl->ifl_pidx = fl->ifl_gen = fl->ifl_fragidx = 0;
2158 bzero(idi->idi_vaddr, idi->idi_size);
2161 /*********************************************************************
2163 * Initialize a receive ring and its buffers.
2165 **********************************************************************/
2167 iflib_fl_setup(iflib_fl_t fl)
2169 iflib_rxq_t rxq = fl->ifl_rxq;
2170 if_ctx_t ctx = rxq->ifr_ctx;
2172 bit_nclear(fl->ifl_rx_bitmap, 0, fl->ifl_size - 1);
2174 ** Free current RX buffer structs and their mbufs
2176 iflib_fl_bufs_free(fl);
2177 /* Now replenish the mbufs */
2178 MPASS(fl->ifl_credits == 0);
2179 fl->ifl_buf_size = ctx->ifc_rx_mbuf_sz;
2180 if (fl->ifl_buf_size > ctx->ifc_max_fl_buf_size)
2181 ctx->ifc_max_fl_buf_size = fl->ifl_buf_size;
2182 fl->ifl_cltype = m_gettype(fl->ifl_buf_size);
2183 fl->ifl_zone = m_getzone(fl->ifl_buf_size);
2186 /* avoid pre-allocating zillions of clusters to an idle card
2187 * potentially speeding up attach
2189 _iflib_fl_refill(ctx, fl, min(128, fl->ifl_size));
2190 MPASS(min(128, fl->ifl_size) == fl->ifl_credits);
2191 if (min(128, fl->ifl_size) != fl->ifl_credits)
2197 MPASS(fl->ifl_ifdi != NULL);
2198 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
2199 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2203 /*********************************************************************
2205 * Free receive ring data structures
2207 **********************************************************************/
2209 iflib_rx_sds_free(iflib_rxq_t rxq)
2214 if (rxq->ifr_fl != NULL) {
2215 for (i = 0; i < rxq->ifr_nfl; i++) {
2216 fl = &rxq->ifr_fl[i];
2217 if (fl->ifl_buf_tag != NULL) {
2218 if (fl->ifl_sds.ifsd_map != NULL) {
2219 for (j = 0; j < fl->ifl_size; j++) {
2220 if (fl->ifl_sds.ifsd_map[j] ==
2225 fl->ifl_sds.ifsd_map[j],
2226 BUS_DMASYNC_POSTREAD);
2229 fl->ifl_sds.ifsd_map[j]);
2232 bus_dma_tag_destroy(fl->ifl_buf_tag);
2233 fl->ifl_buf_tag = NULL;
2235 free(fl->ifl_sds.ifsd_m, M_IFLIB);
2236 free(fl->ifl_sds.ifsd_cl, M_IFLIB);
2237 free(fl->ifl_sds.ifsd_ba, M_IFLIB);
2238 free(fl->ifl_sds.ifsd_map, M_IFLIB);
2239 fl->ifl_sds.ifsd_m = NULL;
2240 fl->ifl_sds.ifsd_cl = NULL;
2241 fl->ifl_sds.ifsd_ba = NULL;
2242 fl->ifl_sds.ifsd_map = NULL;
2244 free(rxq->ifr_fl, M_IFLIB);
2246 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
2251 * MI independent logic
2255 iflib_timer(void *arg)
2257 iflib_txq_t txq = arg;
2258 if_ctx_t ctx = txq->ift_ctx;
2259 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2260 uint64_t this_tick = ticks;
2261 uint32_t reset_on = hz / 2;
2263 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))
2266 ** Check on the state of the TX queue(s), this
2267 ** can be done without the lock because its RO
2268 ** and the HUNG state will be static if set.
2270 if (this_tick - txq->ift_last_timer_tick >= hz / 2) {
2271 txq->ift_last_timer_tick = this_tick;
2272 IFDI_TIMER(ctx, txq->ift_id);
2273 if ((txq->ift_qstatus == IFLIB_QUEUE_HUNG) &&
2274 ((txq->ift_cleaned_prev == txq->ift_cleaned) ||
2275 (sctx->isc_pause_frames == 0)))
2278 if (ifmp_ring_is_stalled(txq->ift_br))
2279 txq->ift_qstatus = IFLIB_QUEUE_HUNG;
2280 txq->ift_cleaned_prev = txq->ift_cleaned;
2283 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP)
2284 iflib_netmap_timer_adjust(ctx, txq->ift_id, &reset_on);
2286 /* handle any laggards */
2287 if (txq->ift_db_pending)
2288 GROUPTASK_ENQUEUE(&txq->ift_task);
2290 sctx->isc_pause_frames = 0;
2291 if (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)
2292 callout_reset_on(&txq->ift_timer, reset_on, iflib_timer, txq, txq->ift_timer.c_cpu);
2295 device_printf(ctx->ifc_dev, "TX(%d) desc avail = %d, pidx = %d\n",
2296 txq->ift_id, TXQ_AVAIL(txq), txq->ift_pidx);
2298 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2299 ctx->ifc_flags |= (IFC_DO_WATCHDOG|IFC_DO_RESET);
2300 iflib_admin_intr_deferred(ctx);
2305 iflib_calc_rx_mbuf_sz(if_ctx_t ctx)
2307 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2310 * XXX don't set the max_frame_size to larger
2311 * than the hardware can handle
2313 if (sctx->isc_max_frame_size <= MCLBYTES)
2314 ctx->ifc_rx_mbuf_sz = MCLBYTES;
2316 ctx->ifc_rx_mbuf_sz = MJUMPAGESIZE;
2320 iflib_get_rx_mbuf_sz(if_ctx_t ctx)
2322 return (ctx->ifc_rx_mbuf_sz);
2326 iflib_init_locked(if_ctx_t ctx)
2328 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
2329 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2330 if_t ifp = ctx->ifc_ifp;
2334 int i, j, tx_ip_csum_flags, tx_ip6_csum_flags;
2337 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2338 IFDI_INTR_DISABLE(ctx);
2340 tx_ip_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP);
2341 tx_ip6_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_SCTP);
2342 /* Set hardware offload abilities */
2343 if_clearhwassist(ifp);
2344 if (if_getcapenable(ifp) & IFCAP_TXCSUM)
2345 if_sethwassistbits(ifp, tx_ip_csum_flags, 0);
2346 if (if_getcapenable(ifp) & IFCAP_TXCSUM_IPV6)
2347 if_sethwassistbits(ifp, tx_ip6_csum_flags, 0);
2348 if (if_getcapenable(ifp) & IFCAP_TSO4)
2349 if_sethwassistbits(ifp, CSUM_IP_TSO, 0);
2350 if (if_getcapenable(ifp) & IFCAP_TSO6)
2351 if_sethwassistbits(ifp, CSUM_IP6_TSO, 0);
2353 for (i = 0, txq = ctx->ifc_txqs; i < sctx->isc_ntxqsets; i++, txq++) {
2355 callout_stop(&txq->ift_timer);
2356 CALLOUT_UNLOCK(txq);
2357 iflib_netmap_txq_init(ctx, txq);
2361 * Calculate a suitable Rx mbuf size prior to calling IFDI_INIT, so
2362 * that drivers can use the value when setting up the hardware receive
2365 iflib_calc_rx_mbuf_sz(ctx);
2368 i = if_getdrvflags(ifp);
2371 MPASS(if_getdrvflags(ifp) == i);
2372 for (i = 0, rxq = ctx->ifc_rxqs; i < sctx->isc_nrxqsets; i++, rxq++) {
2373 /* XXX this should really be done on a per-queue basis */
2374 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
2375 MPASS(rxq->ifr_id == i);
2376 iflib_netmap_rxq_init(ctx, rxq);
2379 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) {
2380 if (iflib_fl_setup(fl)) {
2381 device_printf(ctx->ifc_dev, "freelist setup failed - check cluster settings\n");
2387 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
2388 IFDI_INTR_ENABLE(ctx);
2389 txq = ctx->ifc_txqs;
2390 for (i = 0; i < sctx->isc_ntxqsets; i++, txq++)
2391 callout_reset_on(&txq->ift_timer, hz/2, iflib_timer, txq,
2392 txq->ift_timer.c_cpu);
2396 iflib_media_change(if_t ifp)
2398 if_ctx_t ctx = if_getsoftc(ifp);
2402 if ((err = IFDI_MEDIA_CHANGE(ctx)) == 0)
2403 iflib_init_locked(ctx);
2409 iflib_media_status(if_t ifp, struct ifmediareq *ifmr)
2411 if_ctx_t ctx = if_getsoftc(ifp);
2414 IFDI_UPDATE_ADMIN_STATUS(ctx);
2415 IFDI_MEDIA_STATUS(ctx, ifmr);
2420 iflib_stop(if_ctx_t ctx)
2422 iflib_txq_t txq = ctx->ifc_txqs;
2423 iflib_rxq_t rxq = ctx->ifc_rxqs;
2424 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2425 if_shared_ctx_t sctx = ctx->ifc_sctx;
2426 iflib_dma_info_t di;
2430 /* Tell the stack that the interface is no longer active */
2431 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
2433 IFDI_INTR_DISABLE(ctx);
2438 iflib_debug_reset();
2439 /* Wait for current tx queue users to exit to disarm watchdog timer. */
2440 for (i = 0; i < scctx->isc_ntxqsets; i++, txq++) {
2441 /* make sure all transmitters have completed before proceeding XXX */
2444 callout_stop(&txq->ift_timer);
2445 CALLOUT_UNLOCK(txq);
2447 /* clean any enqueued buffers */
2448 iflib_ifmp_purge(txq);
2449 /* Free any existing tx buffers. */
2450 for (j = 0; j < txq->ift_size; j++) {
2451 iflib_txsd_free(ctx, txq, j);
2453 txq->ift_processed = txq->ift_cleaned = txq->ift_cidx_processed = 0;
2454 txq->ift_in_use = txq->ift_gen = txq->ift_cidx = txq->ift_pidx = txq->ift_no_desc_avail = 0;
2455 txq->ift_closed = txq->ift_mbuf_defrag = txq->ift_mbuf_defrag_failed = 0;
2456 txq->ift_no_tx_dma_setup = txq->ift_txd_encap_efbig = txq->ift_map_failed = 0;
2457 txq->ift_pullups = 0;
2458 ifmp_ring_reset_stats(txq->ift_br);
2459 for (j = 0, di = txq->ift_ifdi; j < sctx->isc_ntxqs; j++, di++)
2460 bzero((void *)di->idi_vaddr, di->idi_size);
2462 for (i = 0; i < scctx->isc_nrxqsets; i++, rxq++) {
2463 /* make sure all transmitters have completed before proceeding XXX */
2465 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
2466 for (j = 0, di = rxq->ifr_ifdi; j < sctx->isc_nrxqs; j++, di++)
2467 bzero((void *)di->idi_vaddr, di->idi_size);
2468 /* also resets the free lists pidx/cidx */
2469 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
2470 iflib_fl_bufs_free(fl);
2474 static inline caddr_t
2475 calc_next_rxd(iflib_fl_t fl, int cidx)
2479 caddr_t start, end, cur, next;
2481 nrxd = fl->ifl_size;
2482 size = fl->ifl_rxd_size;
2483 start = fl->ifl_ifdi->idi_vaddr;
2485 if (__predict_false(size == 0))
2487 cur = start + size*cidx;
2488 end = start + size*nrxd;
2489 next = CACHE_PTR_NEXT(cur);
2490 return (next < end ? next : start);
2494 prefetch_pkts(iflib_fl_t fl, int cidx)
2497 int nrxd = fl->ifl_size;
2501 nextptr = (cidx + CACHE_PTR_INCREMENT) & (nrxd-1);
2502 prefetch(&fl->ifl_sds.ifsd_m[nextptr]);
2503 prefetch(&fl->ifl_sds.ifsd_cl[nextptr]);
2504 next_rxd = calc_next_rxd(fl, cidx);
2506 prefetch(fl->ifl_sds.ifsd_m[(cidx + 1) & (nrxd-1)]);
2507 prefetch(fl->ifl_sds.ifsd_m[(cidx + 2) & (nrxd-1)]);
2508 prefetch(fl->ifl_sds.ifsd_m[(cidx + 3) & (nrxd-1)]);
2509 prefetch(fl->ifl_sds.ifsd_m[(cidx + 4) & (nrxd-1)]);
2510 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 1) & (nrxd-1)]);
2511 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 2) & (nrxd-1)]);
2512 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 3) & (nrxd-1)]);
2513 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 4) & (nrxd-1)]);
2517 rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, int unload, if_rxsd_t sd)
2522 iflib_dma_info_t di;
2526 flid = irf->irf_flid;
2527 cidx = irf->irf_idx;
2528 fl = &rxq->ifr_fl[flid];
2530 sd->ifsd_cidx = cidx;
2531 sd->ifsd_m = &fl->ifl_sds.ifsd_m[cidx];
2532 sd->ifsd_cl = &fl->ifl_sds.ifsd_cl[cidx];
2535 fl->ifl_m_dequeued++;
2537 if (rxq->ifr_ctx->ifc_flags & IFC_PREFETCH)
2538 prefetch_pkts(fl, cidx);
2539 next = (cidx + CACHE_PTR_INCREMENT) & (fl->ifl_size-1);
2540 prefetch(&fl->ifl_sds.ifsd_map[next]);
2541 map = fl->ifl_sds.ifsd_map[cidx];
2543 next = (cidx + CACHE_LINE_SIZE) & (fl->ifl_size-1);
2545 /* not valid assert if bxe really does SGE from non-contiguous elements */
2546 MPASS(fl->ifl_cidx == cidx);
2547 bus_dmamap_sync(fl->ifl_buf_tag, map, BUS_DMASYNC_POSTREAD);
2549 bus_dmamap_unload(fl->ifl_buf_tag, map);
2550 fl->ifl_cidx = (fl->ifl_cidx + 1) & (fl->ifl_size-1);
2551 if (__predict_false(fl->ifl_cidx == 0))
2553 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map,
2554 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2555 bit_clear(fl->ifl_rx_bitmap, cidx);
2558 static struct mbuf *
2559 assemble_segments(iflib_rxq_t rxq, if_rxd_info_t ri, if_rxsd_t sd)
2561 int i, padlen , flags;
2562 struct mbuf *m, *mh, *mt;
2568 rxd_frag_to_sd(rxq, &ri->iri_frags[i], TRUE, sd);
2570 MPASS(*sd->ifsd_cl != NULL);
2571 MPASS(*sd->ifsd_m != NULL);
2573 /* Don't include zero-length frags */
2574 if (ri->iri_frags[i].irf_len == 0) {
2575 /* XXX we can save the cluster here, but not the mbuf */
2576 m_init(*sd->ifsd_m, M_NOWAIT, MT_DATA, 0);
2577 m_free(*sd->ifsd_m);
2584 flags = M_PKTHDR|M_EXT;
2586 padlen = ri->iri_pad;
2591 /* assuming padding is only on the first fragment */
2595 *sd->ifsd_cl = NULL;
2597 /* Can these two be made one ? */
2598 m_init(m, M_NOWAIT, MT_DATA, flags);
2599 m_cljset(m, cl, sd->ifsd_fl->ifl_cltype);
2601 * These must follow m_init and m_cljset
2603 m->m_data += padlen;
2604 ri->iri_len -= padlen;
2605 m->m_len = ri->iri_frags[i].irf_len;
2606 } while (++i < ri->iri_nfrags);
2612 * Process one software descriptor
2614 static struct mbuf *
2615 iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri)
2620 /* should I merge this back in now that the two paths are basically duplicated? */
2621 if (ri->iri_nfrags == 1 &&
2622 ri->iri_frags[0].irf_len <= MIN(IFLIB_RX_COPY_THRESH, MHLEN)) {
2623 rxd_frag_to_sd(rxq, &ri->iri_frags[0], FALSE, &sd);
2626 m_init(m, M_NOWAIT, MT_DATA, M_PKTHDR);
2627 #ifndef __NO_STRICT_ALIGNMENT
2631 memcpy(m->m_data, *sd.ifsd_cl, ri->iri_len);
2632 bus_dmamap_sync(rxq->ifr_fl->ifl_buf_tag,
2633 rxq->ifr_fl->ifl_sds.ifsd_map[ri->iri_frags[0].irf_idx],
2634 BUS_DMASYNC_PREREAD);
2635 m->m_len = ri->iri_frags[0].irf_len;
2637 m = assemble_segments(rxq, ri, &sd);
2639 m->m_pkthdr.len = ri->iri_len;
2640 m->m_pkthdr.rcvif = ri->iri_ifp;
2641 m->m_flags |= ri->iri_flags;
2642 m->m_pkthdr.ether_vtag = ri->iri_vtag;
2643 m->m_pkthdr.flowid = ri->iri_flowid;
2644 M_HASHTYPE_SET(m, ri->iri_rsstype);
2645 m->m_pkthdr.csum_flags = ri->iri_csum_flags;
2646 m->m_pkthdr.csum_data = ri->iri_csum_data;
2650 #if defined(INET6) || defined(INET)
2652 iflib_get_ip_forwarding(struct lro_ctrl *lc, bool *v4, bool *v6)
2654 CURVNET_SET(lc->ifp->if_vnet);
2656 *v6 = VNET(ip6_forwarding);
2659 *v4 = VNET(ipforwarding);
2665 * Returns true if it's possible this packet could be LROed.
2666 * if it returns false, it is guaranteed that tcp_lro_rx()
2667 * would not return zero.
2670 iflib_check_lro_possible(struct mbuf *m, bool v4_forwarding, bool v6_forwarding)
2672 struct ether_header *eh;
2675 eh = mtod(m, struct ether_header *);
2676 eh_type = ntohs(eh->ether_type);
2679 case ETHERTYPE_IPV6:
2680 return !v6_forwarding;
2684 return !v4_forwarding;
2692 iflib_get_ip_forwarding(struct lro_ctrl *lc __unused, bool *v4 __unused, bool *v6 __unused)
2698 iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
2700 if_ctx_t ctx = rxq->ifr_ctx;
2701 if_shared_ctx_t sctx = ctx->ifc_sctx;
2702 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
2703 iflib_dma_info_t di;
2706 struct if_rxd_info ri;
2707 int err, budget_left, rx_bytes, rx_pkts;
2711 bool v4_forwarding, v6_forwarding, lro_possible;
2714 * XXX early demux data packets so that if_input processing only handles
2715 * acks in interrupt context
2717 struct mbuf *m, *mh, *mt, *mf;
2719 lro_possible = v4_forwarding = v6_forwarding = false;
2723 rx_pkts = rx_bytes = 0;
2724 if (sctx->isc_flags & IFLIB_HAS_RXCQ)
2725 cidxp = &rxq->ifr_cq_cidx;
2727 cidxp = &rxq->ifr_fl[0].ifl_cidx;
2728 if ((avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget)) == 0) {
2729 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
2730 __iflib_fl_refill_lt(ctx, fl, budget + 8);
2731 DBG_COUNTER_INC(rx_unavail);
2735 for (budget_left = budget; budget_left > 0 && avail > 0;) {
2736 if (__predict_false(!CTX_ACTIVE(ctx))) {
2737 DBG_COUNTER_INC(rx_ctx_inactive);
2741 * Reset client set fields to their default values
2744 ri.iri_qsidx = rxq->ifr_id;
2745 ri.iri_cidx = *cidxp;
2747 ri.iri_frags = rxq->ifr_frags;
2748 di = rxq->ifr_fl[rxq->ifr_frags[0].irf_flid].ifl_ifdi;
2749 bus_dmamap_sync(di->idi_tag, di->idi_map,
2750 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2751 err = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri);
2755 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
2756 *cidxp = ri.iri_cidx;
2757 /* Update our consumer index */
2758 /* XXX NB: shurd - check if this is still safe */
2759 while (rxq->ifr_cq_cidx >= scctx->isc_nrxd[0]) {
2760 rxq->ifr_cq_cidx -= scctx->isc_nrxd[0];
2761 rxq->ifr_cq_gen = 0;
2763 /* was this only a completion queue message? */
2764 if (__predict_false(ri.iri_nfrags == 0))
2767 MPASS(ri.iri_nfrags != 0);
2768 MPASS(ri.iri_len != 0);
2770 /* will advance the cidx on the corresponding free lists */
2771 m = iflib_rxd_pkt_get(rxq, &ri);
2774 if (avail == 0 && budget_left)
2775 avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget_left);
2777 if (__predict_false(m == NULL)) {
2778 DBG_COUNTER_INC(rx_mbuf_null);
2781 /* imm_pkt: -- cxgb */
2789 /* make sure that we can refill faster than drain */
2790 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
2791 __iflib_fl_refill_lt(ctx, fl, budget + 8);
2793 lro_enabled = (if_getcapenable(ifp) & IFCAP_LRO);
2795 iflib_get_ip_forwarding(&rxq->ifr_lc, &v4_forwarding, &v6_forwarding);
2797 while (mh != NULL) {
2800 m->m_nextpkt = NULL;
2801 #ifndef __NO_STRICT_ALIGNMENT
2802 if (!IP_ALIGNED(m) && (m = iflib_fixup_rx(m)) == NULL)
2805 rx_bytes += m->m_pkthdr.len;
2807 #if defined(INET6) || defined(INET)
2809 if (!lro_possible) {
2810 lro_possible = iflib_check_lro_possible(m, v4_forwarding, v6_forwarding);
2811 if (lro_possible && mf != NULL) {
2812 ifp->if_input(ifp, mf);
2813 DBG_COUNTER_INC(rx_if_input);
2817 if ((m->m_pkthdr.csum_flags & (CSUM_L4_CALC|CSUM_L4_VALID)) ==
2818 (CSUM_L4_CALC|CSUM_L4_VALID)) {
2819 if (lro_possible && tcp_lro_rx(&rxq->ifr_lc, m, 0) == 0)
2825 ifp->if_input(ifp, m);
2826 DBG_COUNTER_INC(rx_if_input);
2837 ifp->if_input(ifp, mf);
2838 DBG_COUNTER_INC(rx_if_input);
2841 if_inc_counter(ifp, IFCOUNTER_IBYTES, rx_bytes);
2842 if_inc_counter(ifp, IFCOUNTER_IPACKETS, rx_pkts);
2845 * Flush any outstanding LRO work
2847 #if defined(INET6) || defined(INET)
2848 tcp_lro_flush_all(&rxq->ifr_lc);
2852 return (iflib_rxd_avail(ctx, rxq, *cidxp, 1));
2855 ctx->ifc_flags |= IFC_DO_RESET;
2856 iflib_admin_intr_deferred(ctx);
2861 #define TXD_NOTIFY_COUNT(txq) (((txq)->ift_size / (txq)->ift_update_freq)-1)
2862 static inline qidx_t
2863 txq_max_db_deferred(iflib_txq_t txq, qidx_t in_use)
2865 qidx_t notify_count = TXD_NOTIFY_COUNT(txq);
2866 qidx_t minthresh = txq->ift_size / 8;
2867 if (in_use > 4*minthresh)
2868 return (notify_count);
2869 if (in_use > 2*minthresh)
2870 return (notify_count >> 1);
2871 if (in_use > minthresh)
2872 return (notify_count >> 3);
2876 static inline qidx_t
2877 txq_max_rs_deferred(iflib_txq_t txq)
2879 qidx_t notify_count = TXD_NOTIFY_COUNT(txq);
2880 qidx_t minthresh = txq->ift_size / 8;
2881 if (txq->ift_in_use > 4*minthresh)
2882 return (notify_count);
2883 if (txq->ift_in_use > 2*minthresh)
2884 return (notify_count >> 1);
2885 if (txq->ift_in_use > minthresh)
2886 return (notify_count >> 2);
2890 #define M_CSUM_FLAGS(m) ((m)->m_pkthdr.csum_flags)
2891 #define M_HAS_VLANTAG(m) (m->m_flags & M_VLANTAG)
2893 #define TXQ_MAX_DB_DEFERRED(txq, in_use) txq_max_db_deferred((txq), (in_use))
2894 #define TXQ_MAX_RS_DEFERRED(txq) txq_max_rs_deferred(txq)
2895 #define TXQ_MAX_DB_CONSUMED(size) (size >> 4)
2897 /* forward compatibility for cxgb */
2898 #define FIRST_QSET(ctx) 0
2899 #define NTXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_ntxqsets)
2900 #define NRXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_nrxqsets)
2901 #define QIDX(ctx, m) ((((m)->m_pkthdr.flowid & ctx->ifc_softc_ctx.isc_rss_table_mask) % NTXQSETS(ctx)) + FIRST_QSET(ctx))
2902 #define DESC_RECLAIMABLE(q) ((int)((q)->ift_processed - (q)->ift_cleaned - (q)->ift_ctx->ifc_softc_ctx.isc_tx_nsegments))
2904 /* XXX we should be setting this to something other than zero */
2905 #define RECLAIM_THRESH(ctx) ((ctx)->ifc_sctx->isc_tx_reclaim_thresh)
2906 #define MAX_TX_DESC(ctx) max((ctx)->ifc_softc_ctx.isc_tx_tso_segments_max, \
2907 (ctx)->ifc_softc_ctx.isc_tx_nsegments)
2910 iflib_txd_db_check(if_ctx_t ctx, iflib_txq_t txq, int ring, qidx_t in_use)
2916 max = TXQ_MAX_DB_DEFERRED(txq, in_use);
2917 if (ring || txq->ift_db_pending >= max) {
2918 dbval = txq->ift_npending ? txq->ift_npending : txq->ift_pidx;
2919 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, dbval);
2920 txq->ift_db_pending = txq->ift_npending = 0;
2928 print_pkt(if_pkt_info_t pi)
2930 printf("pi len: %d qsidx: %d nsegs: %d ndescs: %d flags: %x pidx: %d\n",
2931 pi->ipi_len, pi->ipi_qsidx, pi->ipi_nsegs, pi->ipi_ndescs, pi->ipi_flags, pi->ipi_pidx);
2932 printf("pi new_pidx: %d csum_flags: %lx tso_segsz: %d mflags: %x vtag: %d\n",
2933 pi->ipi_new_pidx, pi->ipi_csum_flags, pi->ipi_tso_segsz, pi->ipi_mflags, pi->ipi_vtag);
2934 printf("pi etype: %d ehdrlen: %d ip_hlen: %d ipproto: %d\n",
2935 pi->ipi_etype, pi->ipi_ehdrlen, pi->ipi_ip_hlen, pi->ipi_ipproto);
2939 #define IS_TSO4(pi) ((pi)->ipi_csum_flags & CSUM_IP_TSO)
2940 #define IS_TX_OFFLOAD4(pi) ((pi)->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP_TSO))
2941 #define IS_TSO6(pi) ((pi)->ipi_csum_flags & CSUM_IP6_TSO)
2942 #define IS_TX_OFFLOAD6(pi) ((pi)->ipi_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_TSO))
2945 iflib_parse_header(iflib_txq_t txq, if_pkt_info_t pi, struct mbuf **mp)
2947 if_shared_ctx_t sctx = txq->ift_ctx->ifc_sctx;
2948 struct ether_vlan_header *eh;
2952 if ((sctx->isc_flags & IFLIB_NEED_SCRATCH) &&
2953 M_WRITABLE(m) == 0) {
2954 if ((m = m_dup(m, M_NOWAIT)) == NULL) {
2958 DBG_COUNTER_INC(tx_frees);
2964 * Determine where frame payload starts.
2965 * Jump over vlan headers if already present,
2966 * helpful for QinQ too.
2968 if (__predict_false(m->m_len < sizeof(*eh))) {
2970 if (__predict_false((m = m_pullup(m, sizeof(*eh))) == NULL))
2973 eh = mtod(m, struct ether_vlan_header *);
2974 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
2975 pi->ipi_etype = ntohs(eh->evl_proto);
2976 pi->ipi_ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
2978 pi->ipi_etype = ntohs(eh->evl_encap_proto);
2979 pi->ipi_ehdrlen = ETHER_HDR_LEN;
2982 switch (pi->ipi_etype) {
2987 struct ip *ip = NULL;
2988 struct tcphdr *th = NULL;
2991 minthlen = min(m->m_pkthdr.len, pi->ipi_ehdrlen + sizeof(*ip) + sizeof(*th));
2992 if (__predict_false(m->m_len < minthlen)) {
2994 * if this code bloat is causing too much of a hit
2995 * move it to a separate function and mark it noinline
2997 if (m->m_len == pi->ipi_ehdrlen) {
3000 if (n->m_len >= sizeof(*ip)) {
3001 ip = (struct ip *)n->m_data;
3002 if (n->m_len >= (ip->ip_hl << 2) + sizeof(*th))
3003 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
3006 if (__predict_false((m = m_pullup(m, minthlen)) == NULL))
3008 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
3012 if (__predict_false((m = m_pullup(m, minthlen)) == NULL))
3014 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
3015 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th))
3016 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
3019 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen);
3020 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th))
3021 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
3023 pi->ipi_ip_hlen = ip->ip_hl << 2;
3024 pi->ipi_ipproto = ip->ip_p;
3025 pi->ipi_flags |= IPI_TX_IPV4;
3027 /* TCP checksum offload may require TCP header length */
3028 if (IS_TX_OFFLOAD4(pi)) {
3029 if (__predict_true(pi->ipi_ipproto == IPPROTO_TCP)) {
3030 if (__predict_false(th == NULL)) {
3032 if (__predict_false((m = m_pullup(m, (ip->ip_hl << 2) + sizeof(*th))) == NULL))
3034 th = (struct tcphdr *)((caddr_t)ip + pi->ipi_ip_hlen);
3036 pi->ipi_tcp_hflags = th->th_flags;
3037 pi->ipi_tcp_hlen = th->th_off << 2;
3038 pi->ipi_tcp_seq = th->th_seq;
3041 if (__predict_false(ip->ip_p != IPPROTO_TCP))
3044 * TSO always requires hardware checksum offload.
3046 pi->ipi_csum_flags |= (CSUM_IP_TCP | CSUM_IP);
3047 th->th_sum = in_pseudo(ip->ip_src.s_addr,
3048 ip->ip_dst.s_addr, htons(IPPROTO_TCP));
3049 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz;
3050 if (sctx->isc_flags & IFLIB_TSO_INIT_IP) {
3052 ip->ip_len = htons(pi->ipi_ip_hlen + pi->ipi_tcp_hlen + pi->ipi_tso_segsz);
3056 if ((sctx->isc_flags & IFLIB_NEED_ZERO_CSUM) && (pi->ipi_csum_flags & CSUM_IP))
3063 case ETHERTYPE_IPV6:
3065 struct ip6_hdr *ip6 = (struct ip6_hdr *)(m->m_data + pi->ipi_ehdrlen);
3067 pi->ipi_ip_hlen = sizeof(struct ip6_hdr);
3069 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) {
3071 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) == NULL))
3074 th = (struct tcphdr *)((caddr_t)ip6 + pi->ipi_ip_hlen);
3076 /* XXX-BZ this will go badly in case of ext hdrs. */
3077 pi->ipi_ipproto = ip6->ip6_nxt;
3078 pi->ipi_flags |= IPI_TX_IPV6;
3080 /* TCP checksum offload may require TCP header length */
3081 if (IS_TX_OFFLOAD6(pi)) {
3082 if (pi->ipi_ipproto == IPPROTO_TCP) {
3083 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) {
3085 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) == NULL))
3088 pi->ipi_tcp_hflags = th->th_flags;
3089 pi->ipi_tcp_hlen = th->th_off << 2;
3090 pi->ipi_tcp_seq = th->th_seq;
3093 if (__predict_false(ip6->ip6_nxt != IPPROTO_TCP))
3096 * TSO always requires hardware checksum offload.
3098 pi->ipi_csum_flags |= CSUM_IP6_TCP;
3099 th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0);
3100 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz;
3107 pi->ipi_csum_flags &= ~CSUM_OFFLOAD;
3108 pi->ipi_ip_hlen = 0;
3117 * If dodgy hardware rejects the scatter gather chain we've handed it
3118 * we'll need to remove the mbuf chain from ifsg_m[] before we can add the
3121 static __noinline struct mbuf *
3122 iflib_remove_mbuf(iflib_txq_t txq)
3125 struct mbuf *m, **ifsd_m;
3127 ifsd_m = txq->ift_sds.ifsd_m;
3128 ntxd = txq->ift_size;
3129 pidx = txq->ift_pidx & (ntxd - 1);
3130 ifsd_m = txq->ift_sds.ifsd_m;
3132 ifsd_m[pidx] = NULL;
3133 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[pidx]);
3134 if (txq->ift_sds.ifsd_tso_map != NULL)
3135 bus_dmamap_unload(txq->ift_tso_buf_tag,
3136 txq->ift_sds.ifsd_tso_map[pidx]);
3138 txq->ift_dequeued++;
3143 static inline caddr_t
3144 calc_next_txd(iflib_txq_t txq, int cidx, uint8_t qid)
3148 caddr_t start, end, cur, next;
3150 ntxd = txq->ift_size;
3151 size = txq->ift_txd_size[qid];
3152 start = txq->ift_ifdi[qid].idi_vaddr;
3154 if (__predict_false(size == 0))
3156 cur = start + size*cidx;
3157 end = start + size*ntxd;
3158 next = CACHE_PTR_NEXT(cur);
3159 return (next < end ? next : start);
3163 * Pad an mbuf to ensure a minimum ethernet frame size.
3164 * min_frame_size is the frame size (less CRC) to pad the mbuf to
3166 static __noinline int
3167 iflib_ether_pad(device_t dev, struct mbuf **m_head, uint16_t min_frame_size)
3170 * 18 is enough bytes to pad an ARP packet to 46 bytes, and
3171 * and ARP message is the smallest common payload I can think of
3173 static char pad[18]; /* just zeros */
3175 struct mbuf *new_head;
3177 if (!M_WRITABLE(*m_head)) {
3178 new_head = m_dup(*m_head, M_NOWAIT);
3179 if (new_head == NULL) {
3181 device_printf(dev, "cannot pad short frame, m_dup() failed");
3182 DBG_COUNTER_INC(encap_pad_mbuf_fail);
3183 DBG_COUNTER_INC(tx_frees);
3190 for (n = min_frame_size - (*m_head)->m_pkthdr.len;
3191 n > 0; n -= sizeof(pad))
3192 if (!m_append(*m_head, min(n, sizeof(pad)), pad))
3197 device_printf(dev, "cannot pad short frame\n");
3198 DBG_COUNTER_INC(encap_pad_mbuf_fail);
3199 DBG_COUNTER_INC(tx_frees);
3207 iflib_encap(iflib_txq_t txq, struct mbuf **m_headp)
3210 if_shared_ctx_t sctx;
3211 if_softc_ctx_t scctx;
3212 bus_dma_tag_t buf_tag;
3213 bus_dma_segment_t *segs;
3214 struct mbuf *m_head, **ifsd_m;
3217 struct if_pkt_info pi;
3219 int err, nsegs, ndesc, max_segs, pidx, cidx, next, ntxd;
3222 sctx = ctx->ifc_sctx;
3223 scctx = &ctx->ifc_softc_ctx;
3224 segs = txq->ift_segs;
3225 ntxd = txq->ift_size;
3230 * If we're doing TSO the next descriptor to clean may be quite far ahead
3232 cidx = txq->ift_cidx;
3233 pidx = txq->ift_pidx;
3234 if (ctx->ifc_flags & IFC_PREFETCH) {
3235 next = (cidx + CACHE_PTR_INCREMENT) & (ntxd-1);
3236 if (!(ctx->ifc_flags & IFLIB_HAS_TXCQ)) {
3237 next_txd = calc_next_txd(txq, cidx, 0);
3241 /* prefetch the next cache line of mbuf pointers and flags */
3242 prefetch(&txq->ift_sds.ifsd_m[next]);
3243 prefetch(&txq->ift_sds.ifsd_map[next]);
3244 next = (cidx + CACHE_LINE_SIZE) & (ntxd-1);
3246 map = txq->ift_sds.ifsd_map[pidx];
3247 ifsd_m = txq->ift_sds.ifsd_m;
3249 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3250 buf_tag = txq->ift_tso_buf_tag;
3251 max_segs = scctx->isc_tx_tso_segments_max;
3252 map = txq->ift_sds.ifsd_tso_map[pidx];
3253 MPASS(buf_tag != NULL);
3254 MPASS(max_segs > 0);
3256 buf_tag = txq->ift_buf_tag;
3257 max_segs = scctx->isc_tx_nsegments;
3258 map = txq->ift_sds.ifsd_map[pidx];
3260 if ((sctx->isc_flags & IFLIB_NEED_ETHER_PAD) &&
3261 __predict_false(m_head->m_pkthdr.len < scctx->isc_min_frame_size)) {
3262 err = iflib_ether_pad(ctx->ifc_dev, m_headp, scctx->isc_min_frame_size);
3264 DBG_COUNTER_INC(encap_txd_encap_fail);
3271 pi.ipi_mflags = (m_head->m_flags & (M_VLANTAG|M_BCAST|M_MCAST));
3273 pi.ipi_qsidx = txq->ift_id;
3274 pi.ipi_len = m_head->m_pkthdr.len;
3275 pi.ipi_csum_flags = m_head->m_pkthdr.csum_flags;
3276 pi.ipi_vtag = (m_head->m_flags & M_VLANTAG) ? m_head->m_pkthdr.ether_vtag : 0;
3278 /* deliberate bitwise OR to make one condition */
3279 if (__predict_true((pi.ipi_csum_flags | pi.ipi_vtag))) {
3280 if (__predict_false((err = iflib_parse_header(txq, &pi, m_headp)) != 0)) {
3281 DBG_COUNTER_INC(encap_txd_encap_fail);
3288 err = bus_dmamap_load_mbuf_sg(buf_tag, map, m_head, segs, &nsegs,
3291 if (__predict_false(err)) {
3294 /* try collapse once and defrag once */
3296 m_head = m_collapse(*m_headp, M_NOWAIT, max_segs);
3297 /* try defrag if collapsing fails */
3302 txq->ift_mbuf_defrag++;
3303 m_head = m_defrag(*m_headp, M_NOWAIT);
3306 * remap should never be >1 unless bus_dmamap_load_mbuf_sg
3307 * failed to map an mbuf that was run through m_defrag
3310 if (__predict_false(m_head == NULL || remap > 1))
3317 txq->ift_no_tx_dma_setup++;
3320 txq->ift_no_tx_dma_setup++;
3322 DBG_COUNTER_INC(tx_frees);
3326 txq->ift_map_failed++;
3327 DBG_COUNTER_INC(encap_load_mbuf_fail);
3328 DBG_COUNTER_INC(encap_txd_encap_fail);
3331 ifsd_m[pidx] = m_head;
3333 * XXX assumes a 1 to 1 relationship between segments and
3334 * descriptors - this does not hold true on all drivers, e.g.
3337 if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
3338 txq->ift_no_desc_avail++;
3339 bus_dmamap_unload(buf_tag, map);
3340 DBG_COUNTER_INC(encap_txq_avail_fail);
3341 DBG_COUNTER_INC(encap_txd_encap_fail);
3342 if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
3343 GROUPTASK_ENQUEUE(&txq->ift_task);
3347 * On Intel cards we can greatly reduce the number of TX interrupts
3348 * we see by only setting report status on every Nth descriptor.
3349 * However, this also means that the driver will need to keep track
3350 * of the descriptors that RS was set on to check them for the DD bit.
3352 txq->ift_rs_pending += nsegs + 1;
3353 if (txq->ift_rs_pending > TXQ_MAX_RS_DEFERRED(txq) ||
3354 iflib_no_tx_batch || (TXQ_AVAIL(txq) - nsegs) <= MAX_TX_DESC(ctx) + 2) {
3355 pi.ipi_flags |= IPI_TX_INTR;
3356 txq->ift_rs_pending = 0;
3360 pi.ipi_nsegs = nsegs;
3362 MPASS(pidx >= 0 && pidx < txq->ift_size);
3366 bus_dmamap_sync(buf_tag, map, BUS_DMASYNC_PREWRITE);
3367 if ((err = ctx->isc_txd_encap(ctx->ifc_softc, &pi)) == 0) {
3368 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
3369 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3370 DBG_COUNTER_INC(tx_encap);
3371 MPASS(pi.ipi_new_pidx < txq->ift_size);
3373 ndesc = pi.ipi_new_pidx - pi.ipi_pidx;
3374 if (pi.ipi_new_pidx < pi.ipi_pidx) {
3375 ndesc += txq->ift_size;
3379 * drivers can need as many as
3382 MPASS(ndesc <= pi.ipi_nsegs + 2);
3383 MPASS(pi.ipi_new_pidx != pidx);
3385 txq->ift_in_use += ndesc;
3388 * We update the last software descriptor again here because there may
3389 * be a sentinel and/or there may be more mbufs than segments
3391 txq->ift_pidx = pi.ipi_new_pidx;
3392 txq->ift_npending += pi.ipi_ndescs;
3394 *m_headp = m_head = iflib_remove_mbuf(txq);
3396 txq->ift_txd_encap_efbig++;
3405 * err can't possibly be non-zero here, so we don't neet to test it
3406 * to see if we need to DBG_COUNTER_INC(encap_txd_encap_fail).
3411 txq->ift_mbuf_defrag_failed++;
3412 txq->ift_map_failed++;
3414 DBG_COUNTER_INC(tx_frees);
3416 DBG_COUNTER_INC(encap_txd_encap_fail);
3421 iflib_tx_desc_free(iflib_txq_t txq, int n)
3423 uint32_t qsize, cidx, mask, gen;
3424 struct mbuf *m, **ifsd_m;
3427 cidx = txq->ift_cidx;
3429 qsize = txq->ift_size;
3431 ifsd_m = txq->ift_sds.ifsd_m;
3432 do_prefetch = (txq->ift_ctx->ifc_flags & IFC_PREFETCH);
3436 prefetch(ifsd_m[(cidx + 3) & mask]);
3437 prefetch(ifsd_m[(cidx + 4) & mask]);
3439 if ((m = ifsd_m[cidx]) != NULL) {
3440 prefetch(&ifsd_m[(cidx + CACHE_PTR_INCREMENT) & mask]);
3441 if (m->m_pkthdr.csum_flags & CSUM_TSO) {
3442 bus_dmamap_sync(txq->ift_tso_buf_tag,
3443 txq->ift_sds.ifsd_tso_map[cidx],
3444 BUS_DMASYNC_POSTWRITE);
3445 bus_dmamap_unload(txq->ift_tso_buf_tag,
3446 txq->ift_sds.ifsd_tso_map[cidx]);
3448 bus_dmamap_sync(txq->ift_buf_tag,
3449 txq->ift_sds.ifsd_map[cidx],
3450 BUS_DMASYNC_POSTWRITE);
3451 bus_dmamap_unload(txq->ift_buf_tag,
3452 txq->ift_sds.ifsd_map[cidx]);
3454 /* XXX we don't support any drivers that batch packets yet */
3455 MPASS(m->m_nextpkt == NULL);
3457 ifsd_m[cidx] = NULL;
3459 txq->ift_dequeued++;
3461 DBG_COUNTER_INC(tx_frees);
3463 if (__predict_false(++cidx == qsize)) {
3468 txq->ift_cidx = cidx;
3473 iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh)
3476 if_ctx_t ctx = txq->ift_ctx;
3478 KASSERT(thresh >= 0, ("invalid threshold to reclaim"));
3479 MPASS(thresh /*+ MAX_TX_DESC(txq->ift_ctx) */ < txq->ift_size);
3482 * Need a rate-limiting check so that this isn't called every time
3484 iflib_tx_credits_update(ctx, txq);
3485 reclaim = DESC_RECLAIMABLE(txq);
3487 if (reclaim <= thresh /* + MAX_TX_DESC(txq->ift_ctx) */) {
3489 if (iflib_verbose_debug) {
3490 printf("%s processed=%ju cleaned=%ju tx_nsegments=%d reclaim=%d thresh=%d\n", __FUNCTION__,
3491 txq->ift_processed, txq->ift_cleaned, txq->ift_ctx->ifc_softc_ctx.isc_tx_nsegments,
3498 iflib_tx_desc_free(txq, reclaim);
3499 txq->ift_cleaned += reclaim;
3500 txq->ift_in_use -= reclaim;
3505 static struct mbuf **
3506 _ring_peek_one(struct ifmp_ring *r, int cidx, int offset, int remaining)
3509 struct mbuf **items;
3512 next = (cidx + CACHE_PTR_INCREMENT) & (size-1);
3513 items = __DEVOLATILE(struct mbuf **, &r->items[0]);
3515 prefetch(items[(cidx + offset) & (size-1)]);
3516 if (remaining > 1) {
3517 prefetch2cachelines(&items[next]);
3518 prefetch2cachelines(items[(cidx + offset + 1) & (size-1)]);
3519 prefetch2cachelines(items[(cidx + offset + 2) & (size-1)]);
3520 prefetch2cachelines(items[(cidx + offset + 3) & (size-1)]);
3522 return (__DEVOLATILE(struct mbuf **, &r->items[(cidx + offset) & (size-1)]));
3526 iflib_txq_check_drain(iflib_txq_t txq, int budget)
3529 ifmp_ring_check_drainage(txq->ift_br, budget);
3533 iflib_txq_can_drain(struct ifmp_ring *r)
3535 iflib_txq_t txq = r->cookie;
3536 if_ctx_t ctx = txq->ift_ctx;
3538 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
3539 BUS_DMASYNC_POSTREAD);
3540 return ((TXQ_AVAIL(txq) > MAX_TX_DESC(ctx) + 2) ||
3541 ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, false));
3545 iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
3547 iflib_txq_t txq = r->cookie;
3548 if_ctx_t ctx = txq->ift_ctx;
3549 struct ifnet *ifp = ctx->ifc_ifp;
3550 struct mbuf **mp, *m;
3551 int i, count, consumed, pkt_sent, bytes_sent, mcast_sent, avail;
3552 int reclaimed, err, in_use_prev, desc_used;
3553 bool do_prefetch, ring, rang;
3555 if (__predict_false(!(if_getdrvflags(ifp) & IFF_DRV_RUNNING) ||
3556 !LINK_ACTIVE(ctx))) {
3557 DBG_COUNTER_INC(txq_drain_notready);
3560 reclaimed = iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
3561 rang = iflib_txd_db_check(ctx, txq, reclaimed, txq->ift_in_use);
3562 avail = IDXDIFF(pidx, cidx, r->size);
3563 if (__predict_false(ctx->ifc_flags & IFC_QFLUSH)) {
3564 DBG_COUNTER_INC(txq_drain_flushing);
3565 for (i = 0; i < avail; i++) {
3566 if (__predict_true(r->items[(cidx + i) & (r->size-1)] != (void *)txq))
3567 m_free(r->items[(cidx + i) & (r->size-1)]);
3568 r->items[(cidx + i) & (r->size-1)] = NULL;
3573 if (__predict_false(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE)) {
3574 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3576 callout_stop(&txq->ift_timer);
3577 CALLOUT_UNLOCK(txq);
3578 DBG_COUNTER_INC(txq_drain_oactive);
3582 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3583 consumed = mcast_sent = bytes_sent = pkt_sent = 0;
3584 count = MIN(avail, TX_BATCH_SIZE);
3586 if (iflib_verbose_debug)
3587 printf("%s avail=%d ifc_flags=%x txq_avail=%d ", __FUNCTION__,
3588 avail, ctx->ifc_flags, TXQ_AVAIL(txq));
3590 do_prefetch = (ctx->ifc_flags & IFC_PREFETCH);
3591 avail = TXQ_AVAIL(txq);
3593 for (desc_used = i = 0; i < count && avail > MAX_TX_DESC(ctx) + 2; i++) {
3594 int rem = do_prefetch ? count - i : 0;
3596 mp = _ring_peek_one(r, cidx, i, rem);
3597 MPASS(mp != NULL && *mp != NULL);
3598 if (__predict_false(*mp == (struct mbuf *)txq)) {
3603 in_use_prev = txq->ift_in_use;
3604 err = iflib_encap(txq, mp);
3605 if (__predict_false(err)) {
3606 /* no room - bail out */
3610 /* we can't send this packet - skip it */
3616 DBG_COUNTER_INC(tx_sent);
3617 bytes_sent += m->m_pkthdr.len;
3618 mcast_sent += !!(m->m_flags & M_MCAST);
3619 avail = TXQ_AVAIL(txq);
3621 txq->ift_db_pending += (txq->ift_in_use - in_use_prev);
3622 desc_used += (txq->ift_in_use - in_use_prev);
3623 ETHER_BPF_MTAP(ifp, m);
3624 if (__predict_false(!(ifp->if_drv_flags & IFF_DRV_RUNNING)))
3626 rang = iflib_txd_db_check(ctx, txq, false, in_use_prev);
3629 /* deliberate use of bitwise or to avoid gratuitous short-circuit */
3630 ring = rang ? false : (iflib_min_tx_latency | err) || (TXQ_AVAIL(txq) < MAX_TX_DESC(ctx));
3631 iflib_txd_db_check(ctx, txq, ring, txq->ift_in_use);
3632 if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
3633 if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
3635 if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent);
3637 if (iflib_verbose_debug)
3638 printf("consumed=%d\n", consumed);
3644 iflib_txq_drain_always(struct ifmp_ring *r)
3650 iflib_txq_drain_free(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
3658 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
3660 callout_stop(&txq->ift_timer);
3661 CALLOUT_UNLOCK(txq);
3663 avail = IDXDIFF(pidx, cidx, r->size);
3664 for (i = 0; i < avail; i++) {
3665 mp = _ring_peek_one(r, cidx, i, avail - i);
3666 if (__predict_false(*mp == (struct mbuf *)txq))
3669 DBG_COUNTER_INC(tx_frees);
3671 MPASS(ifmp_ring_is_stalled(r) == 0);
3676 iflib_ifmp_purge(iflib_txq_t txq)
3678 struct ifmp_ring *r;
3681 r->drain = iflib_txq_drain_free;
3682 r->can_drain = iflib_txq_drain_always;
3684 ifmp_ring_check_drainage(r, r->size);
3686 r->drain = iflib_txq_drain;
3687 r->can_drain = iflib_txq_can_drain;
3691 _task_fn_tx(void *context)
3693 iflib_txq_t txq = context;
3694 if_ctx_t ctx = txq->ift_ctx;
3695 struct ifnet *ifp = ctx->ifc_ifp;
3696 int abdicate = ctx->ifc_sysctl_tx_abdicate;
3698 #ifdef IFLIB_DIAGNOSTICS
3699 txq->ift_cpu_exec_count[curcpu]++;
3701 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))
3703 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
3704 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
3705 BUS_DMASYNC_POSTREAD);
3706 if (ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, false))
3707 netmap_tx_irq(ifp, txq->ift_id);
3708 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id);
3712 if (ALTQ_IS_ENABLED(&ifp->if_snd))
3713 iflib_altq_if_start(ifp);
3715 if (txq->ift_db_pending)
3716 ifmp_ring_enqueue(txq->ift_br, (void **)&txq, 1, TX_BATCH_SIZE, abdicate);
3718 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3720 * When abdicating, we always need to check drainage, not just when we don't enqueue
3723 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3724 if (ctx->ifc_flags & IFC_LEGACY)
3725 IFDI_INTR_ENABLE(ctx);
3730 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id);
3731 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver"));
3736 _task_fn_rx(void *context)
3738 iflib_rxq_t rxq = context;
3739 if_ctx_t ctx = rxq->ifr_ctx;
3743 #ifdef IFLIB_DIAGNOSTICS
3744 rxq->ifr_cpu_exec_count[curcpu]++;
3746 DBG_COUNTER_INC(task_fn_rxs);
3747 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)))
3751 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP) {
3753 if (netmap_rx_irq(ctx->ifc_ifp, rxq->ifr_id, &work)) {
3758 budget = ctx->ifc_sysctl_rx_budget;
3760 budget = 16; /* XXX */
3761 if (more == false || (more = iflib_rxeof(rxq, budget)) == false) {
3762 if (ctx->ifc_flags & IFC_LEGACY)
3763 IFDI_INTR_ENABLE(ctx);
3768 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id);
3769 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver"));
3770 DBG_COUNTER_INC(rx_intr_enables);
3773 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)))
3776 GROUPTASK_ENQUEUE(&rxq->ifr_task);
3780 _task_fn_admin(void *context)
3782 if_ctx_t ctx = context;
3783 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx;
3786 bool oactive, running, do_reset, do_watchdog, in_detach;
3787 uint32_t reset_on = hz / 2;
3790 running = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING);
3791 oactive = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE);
3792 do_reset = (ctx->ifc_flags & IFC_DO_RESET);
3793 do_watchdog = (ctx->ifc_flags & IFC_DO_WATCHDOG);
3794 in_detach = (ctx->ifc_flags & IFC_IN_DETACH);
3795 ctx->ifc_flags &= ~(IFC_DO_RESET|IFC_DO_WATCHDOG);
3798 if ((!running && !oactive) && !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN))
3804 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) {
3806 callout_stop(&txq->ift_timer);
3807 CALLOUT_UNLOCK(txq);
3810 ctx->ifc_watchdog_events++;
3811 IFDI_WATCHDOG_RESET(ctx);
3813 IFDI_UPDATE_ADMIN_STATUS(ctx);
3814 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) {
3817 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP)
3818 iflib_netmap_timer_adjust(ctx, txq->ift_id, &reset_on);
3820 callout_reset_on(&txq->ift_timer, reset_on, iflib_timer, txq, txq->ift_timer.c_cpu);
3822 IFDI_LINK_INTR_ENABLE(ctx);
3824 iflib_if_init_locked(ctx);
3827 if (LINK_ACTIVE(ctx) == 0)
3829 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++)
3830 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET);
3835 _task_fn_iov(void *context)
3837 if_ctx_t ctx = context;
3839 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) &&
3840 !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN))
3844 IFDI_VFLR_HANDLE(ctx);
3849 iflib_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
3852 if_int_delay_info_t info;
3855 info = (if_int_delay_info_t)arg1;
3856 ctx = info->iidi_ctx;
3857 info->iidi_req = req;
3858 info->iidi_oidp = oidp;
3860 err = IFDI_SYSCTL_INT_DELAY(ctx, info);
3865 /*********************************************************************
3869 **********************************************************************/
3872 iflib_if_init_locked(if_ctx_t ctx)
3875 iflib_init_locked(ctx);
3880 iflib_if_init(void *arg)
3885 iflib_if_init_locked(ctx);
3890 iflib_if_transmit(if_t ifp, struct mbuf *m)
3892 if_ctx_t ctx = if_getsoftc(ifp);
3896 int abdicate = ctx->ifc_sysctl_tx_abdicate;
3898 if (__predict_false((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || !LINK_ACTIVE(ctx))) {
3899 DBG_COUNTER_INC(tx_frees);
3904 MPASS(m->m_nextpkt == NULL);
3905 /* ALTQ-enabled interfaces always use queue 0. */
3907 if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m) && !ALTQ_IS_ENABLED(&ifp->if_snd))
3908 qidx = QIDX(ctx, m);
3910 * XXX calculate buf_ring based on flowid (divvy up bits?)
3912 txq = &ctx->ifc_txqs[qidx];
3914 #ifdef DRIVER_BACKPRESSURE
3915 if (txq->ift_closed) {
3917 next = m->m_nextpkt;
3918 m->m_nextpkt = NULL;
3920 DBG_COUNTER_INC(tx_frees);
3932 next = next->m_nextpkt;
3933 } while (next != NULL);
3935 if (count > nitems(marr))
3936 if ((mp = malloc(count*sizeof(struct mbuf *), M_IFLIB, M_NOWAIT)) == NULL) {
3937 /* XXX check nextpkt */
3939 /* XXX simplify for now */
3940 DBG_COUNTER_INC(tx_frees);
3943 for (next = m, i = 0; next != NULL; i++) {
3945 next = next->m_nextpkt;
3946 mp[i]->m_nextpkt = NULL;
3949 DBG_COUNTER_INC(tx_seen);
3950 err = ifmp_ring_enqueue(txq->ift_br, (void **)&m, 1, TX_BATCH_SIZE, abdicate);
3953 GROUPTASK_ENQUEUE(&txq->ift_task);
3956 GROUPTASK_ENQUEUE(&txq->ift_task);
3957 /* support forthcoming later */
3958 #ifdef DRIVER_BACKPRESSURE
3959 txq->ift_closed = TRUE;
3961 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
3963 DBG_COUNTER_INC(tx_frees);
3971 * The overall approach to integrating iflib with ALTQ is to continue to use
3972 * the iflib mp_ring machinery between the ALTQ queue(s) and the hardware
3973 * ring. Technically, when using ALTQ, queueing to an intermediate mp_ring
3974 * is redundant/unnecessary, but doing so minimizes the amount of
3975 * ALTQ-specific code required in iflib. It is assumed that the overhead of
3976 * redundantly queueing to an intermediate mp_ring is swamped by the
3977 * performance limitations inherent in using ALTQ.
3979 * When ALTQ support is compiled in, all iflib drivers will use a transmit
3980 * routine, iflib_altq_if_transmit(), that checks if ALTQ is enabled for the
3981 * given interface. If ALTQ is enabled for an interface, then all
3982 * transmitted packets for that interface will be submitted to the ALTQ
3983 * subsystem via IFQ_ENQUEUE(). We don't use the legacy if_transmit()
3984 * implementation because it uses IFQ_HANDOFF(), which will duplicatively
3985 * update stats that the iflib machinery handles, and which is sensitve to
3986 * the disused IFF_DRV_OACTIVE flag. Additionally, iflib_altq_if_start()
3987 * will be installed as the start routine for use by ALTQ facilities that
3988 * need to trigger queue drains on a scheduled basis.
3992 iflib_altq_if_start(if_t ifp)
3994 struct ifaltq *ifq = &ifp->if_snd;
3998 IFQ_DEQUEUE_NOLOCK(ifq, m);
4000 iflib_if_transmit(ifp, m);
4001 IFQ_DEQUEUE_NOLOCK(ifq, m);
4007 iflib_altq_if_transmit(if_t ifp, struct mbuf *m)
4011 if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
4012 IFQ_ENQUEUE(&ifp->if_snd, m, err);
4014 iflib_altq_if_start(ifp);
4016 err = iflib_if_transmit(ifp, m);
4023 iflib_if_qflush(if_t ifp)
4025 if_ctx_t ctx = if_getsoftc(ifp);
4026 iflib_txq_t txq = ctx->ifc_txqs;
4030 ctx->ifc_flags |= IFC_QFLUSH;
4032 for (i = 0; i < NTXQSETS(ctx); i++, txq++)
4033 while (!(ifmp_ring_is_idle(txq->ift_br) || ifmp_ring_is_stalled(txq->ift_br)))
4034 iflib_txq_check_drain(txq, 0);
4036 ctx->ifc_flags &= ~IFC_QFLUSH;
4040 * When ALTQ is enabled, this will also take care of purging the
4047 #define IFCAP_FLAGS (IFCAP_HWCSUM_IPV6 | IFCAP_HWCSUM | IFCAP_LRO | \
4048 IFCAP_TSO | IFCAP_VLAN_HWTAGGING | IFCAP_HWSTATS | \
4049 IFCAP_VLAN_MTU | IFCAP_VLAN_HWFILTER | \
4050 IFCAP_VLAN_HWTSO | IFCAP_VLAN_HWCSUM)
4053 iflib_if_ioctl(if_t ifp, u_long command, caddr_t data)
4055 if_ctx_t ctx = if_getsoftc(ifp);
4056 struct ifreq *ifr = (struct ifreq *)data;
4057 #if defined(INET) || defined(INET6)
4058 struct ifaddr *ifa = (struct ifaddr *)data;
4060 bool avoid_reset = FALSE;
4061 int err = 0, reinit = 0, bits;
4066 if (ifa->ifa_addr->sa_family == AF_INET)
4070 if (ifa->ifa_addr->sa_family == AF_INET6)
4074 ** Calling init results in link renegotiation,
4075 ** so we avoid doing it when possible.
4078 if_setflagbits(ifp, IFF_UP,0);
4079 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
4082 if (!(if_getflags(ifp) & IFF_NOARP))
4083 arp_ifinit(ifp, ifa);
4086 err = ether_ioctl(ifp, command, data);
4090 if (ifr->ifr_mtu == if_getmtu(ifp)) {
4094 bits = if_getdrvflags(ifp);
4095 /* stop the driver and free any clusters before proceeding */
4098 if ((err = IFDI_MTU_SET(ctx, ifr->ifr_mtu)) == 0) {
4100 if (ifr->ifr_mtu > ctx->ifc_max_fl_buf_size)
4101 ctx->ifc_flags |= IFC_MULTISEG;
4103 ctx->ifc_flags &= ~IFC_MULTISEG;
4105 err = if_setmtu(ifp, ifr->ifr_mtu);
4107 iflib_init_locked(ctx);
4109 if_setdrvflags(ifp, bits);
4115 if (if_getflags(ifp) & IFF_UP) {
4116 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4117 if ((if_getflags(ifp) ^ ctx->ifc_if_flags) &
4118 (IFF_PROMISC | IFF_ALLMULTI)) {
4119 err = IFDI_PROMISC_SET(ctx, if_getflags(ifp));
4123 } else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4126 ctx->ifc_if_flags = if_getflags(ifp);
4131 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
4133 IFDI_INTR_DISABLE(ctx);
4134 IFDI_MULTI_SET(ctx);
4135 IFDI_INTR_ENABLE(ctx);
4141 IFDI_MEDIA_SET(ctx);
4146 err = ifmedia_ioctl(ifp, ifr, &ctx->ifc_media, command);
4150 struct ifi2creq i2c;
4152 err = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
4155 if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) {
4159 if (i2c.len > sizeof(i2c.data)) {
4164 if ((err = IFDI_I2C_REQ(ctx, &i2c)) == 0)
4165 err = copyout(&i2c, ifr_data_get_ptr(ifr),
4171 int mask, setmask, oldmask;
4173 oldmask = if_getcapenable(ifp);
4174 mask = ifr->ifr_reqcap ^ oldmask;
4175 mask &= ctx->ifc_softc_ctx.isc_capabilities;
4178 setmask |= mask & (IFCAP_TOE4|IFCAP_TOE6);
4180 setmask |= (mask & IFCAP_FLAGS);
4181 setmask |= (mask & IFCAP_WOL);
4184 * If any RX csum has changed, change all the ones that
4185 * are supported by the driver.
4187 if (setmask & (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) {
4188 setmask |= ctx->ifc_softc_ctx.isc_capabilities &
4189 (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6);
4193 * want to ensure that traffic has stopped before we change any of the flags
4197 bits = if_getdrvflags(ifp);
4198 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL)
4201 if_togglecapenable(ifp, setmask);
4203 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL)
4204 iflib_init_locked(ctx);
4206 if_setdrvflags(ifp, bits);
4213 case SIOCGPRIVATE_0:
4217 err = IFDI_PRIV_IOCTL(ctx, command, data);
4221 err = ether_ioctl(ifp, command, data);
4230 iflib_if_get_counter(if_t ifp, ift_counter cnt)
4232 if_ctx_t ctx = if_getsoftc(ifp);
4234 return (IFDI_GET_COUNTER(ctx, cnt));
4237 /*********************************************************************
4239 * OTHER FUNCTIONS EXPORTED TO THE STACK
4241 **********************************************************************/
4244 iflib_vlan_register(void *arg, if_t ifp, uint16_t vtag)
4246 if_ctx_t ctx = if_getsoftc(ifp);
4248 if ((void *)ctx != arg)
4251 if ((vtag == 0) || (vtag > 4095))
4255 IFDI_VLAN_REGISTER(ctx, vtag);
4256 /* Re-init to load the changes */
4257 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
4258 iflib_if_init_locked(ctx);
4263 iflib_vlan_unregister(void *arg, if_t ifp, uint16_t vtag)
4265 if_ctx_t ctx = if_getsoftc(ifp);
4267 if ((void *)ctx != arg)
4270 if ((vtag == 0) || (vtag > 4095))
4274 IFDI_VLAN_UNREGISTER(ctx, vtag);
4275 /* Re-init to load the changes */
4276 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
4277 iflib_if_init_locked(ctx);
4282 iflib_led_func(void *arg, int onoff)
4287 IFDI_LED_FUNC(ctx, onoff);
4291 /*********************************************************************
4293 * BUS FUNCTION DEFINITIONS
4295 **********************************************************************/
4298 iflib_device_probe(device_t dev)
4300 pci_vendor_info_t *ent;
4302 uint16_t pci_vendor_id, pci_device_id;
4303 uint16_t pci_subvendor_id, pci_subdevice_id;
4304 uint16_t pci_rev_id;
4305 if_shared_ctx_t sctx;
4307 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC)
4310 pci_vendor_id = pci_get_vendor(dev);
4311 pci_device_id = pci_get_device(dev);
4312 pci_subvendor_id = pci_get_subvendor(dev);
4313 pci_subdevice_id = pci_get_subdevice(dev);
4314 pci_rev_id = pci_get_revid(dev);
4315 if (sctx->isc_parse_devinfo != NULL)
4316 sctx->isc_parse_devinfo(&pci_device_id, &pci_subvendor_id, &pci_subdevice_id, &pci_rev_id);
4318 ent = sctx->isc_vendor_info;
4319 while (ent->pvi_vendor_id != 0) {
4320 if (pci_vendor_id != ent->pvi_vendor_id) {
4324 if ((pci_device_id == ent->pvi_device_id) &&
4325 ((pci_subvendor_id == ent->pvi_subvendor_id) ||
4326 (ent->pvi_subvendor_id == 0)) &&
4327 ((pci_subdevice_id == ent->pvi_subdevice_id) ||
4328 (ent->pvi_subdevice_id == 0)) &&
4329 ((pci_rev_id == ent->pvi_rev_id) ||
4330 (ent->pvi_rev_id == 0))) {
4332 device_set_desc_copy(dev, ent->pvi_name);
4333 /* this needs to be changed to zero if the bus probing code
4334 * ever stops re-probing on best match because the sctx
4335 * may have its values over written by register calls
4336 * in subsequent probes
4338 return (BUS_PROBE_DEFAULT);
4346 iflib_reset_qvalues(if_ctx_t ctx)
4348 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
4349 if_shared_ctx_t sctx = ctx->ifc_sctx;
4350 device_t dev = ctx->ifc_dev;
4353 scctx->isc_txrx_budget_bytes_max = IFLIB_MAX_TX_BYTES;
4354 scctx->isc_tx_qdepth = IFLIB_DEFAULT_TX_QDEPTH;
4356 * XXX sanity check that ntxd & nrxd are a power of 2
4358 if (ctx->ifc_sysctl_ntxqs != 0)
4359 scctx->isc_ntxqsets = ctx->ifc_sysctl_ntxqs;
4360 if (ctx->ifc_sysctl_nrxqs != 0)
4361 scctx->isc_nrxqsets = ctx->ifc_sysctl_nrxqs;
4363 for (i = 0; i < sctx->isc_ntxqs; i++) {
4364 if (ctx->ifc_sysctl_ntxds[i] != 0)
4365 scctx->isc_ntxd[i] = ctx->ifc_sysctl_ntxds[i];
4367 scctx->isc_ntxd[i] = sctx->isc_ntxd_default[i];
4370 for (i = 0; i < sctx->isc_nrxqs; i++) {
4371 if (ctx->ifc_sysctl_nrxds[i] != 0)
4372 scctx->isc_nrxd[i] = ctx->ifc_sysctl_nrxds[i];
4374 scctx->isc_nrxd[i] = sctx->isc_nrxd_default[i];
4377 for (i = 0; i < sctx->isc_nrxqs; i++) {
4378 if (scctx->isc_nrxd[i] < sctx->isc_nrxd_min[i]) {
4379 device_printf(dev, "nrxd%d: %d less than nrxd_min %d - resetting to min\n",
4380 i, scctx->isc_nrxd[i], sctx->isc_nrxd_min[i]);
4381 scctx->isc_nrxd[i] = sctx->isc_nrxd_min[i];
4383 if (scctx->isc_nrxd[i] > sctx->isc_nrxd_max[i]) {
4384 device_printf(dev, "nrxd%d: %d greater than nrxd_max %d - resetting to max\n",
4385 i, scctx->isc_nrxd[i], sctx->isc_nrxd_max[i]);
4386 scctx->isc_nrxd[i] = sctx->isc_nrxd_max[i];
4390 for (i = 0; i < sctx->isc_ntxqs; i++) {
4391 if (scctx->isc_ntxd[i] < sctx->isc_ntxd_min[i]) {
4392 device_printf(dev, "ntxd%d: %d less than ntxd_min %d - resetting to min\n",
4393 i, scctx->isc_ntxd[i], sctx->isc_ntxd_min[i]);
4394 scctx->isc_ntxd[i] = sctx->isc_ntxd_min[i];
4396 if (scctx->isc_ntxd[i] > sctx->isc_ntxd_max[i]) {
4397 device_printf(dev, "ntxd%d: %d greater than ntxd_max %d - resetting to max\n",
4398 i, scctx->isc_ntxd[i], sctx->isc_ntxd_max[i]);
4399 scctx->isc_ntxd[i] = sctx->isc_ntxd_max[i];
4405 iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ctxp)
4410 if_softc_ctx_t scctx;
4416 ctx = malloc(sizeof(* ctx), M_IFLIB, M_WAITOK|M_ZERO);
4419 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK|M_ZERO);
4420 device_set_softc(dev, ctx);
4421 ctx->ifc_flags |= IFC_SC_ALLOCATED;
4424 ctx->ifc_sctx = sctx;
4426 ctx->ifc_softc = sc;
4428 if ((err = iflib_register(ctx)) != 0) {
4429 device_printf(dev, "iflib_register failed %d\n", err);
4432 iflib_add_device_sysctl_pre(ctx);
4434 scctx = &ctx->ifc_softc_ctx;
4437 iflib_reset_qvalues(ctx);
4439 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) {
4440 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err);
4443 _iflib_pre_assert(scctx);
4444 ctx->ifc_txrx = *scctx->isc_txrx;
4447 if (scctx->isc_capabilities & IFCAP_TXCSUM)
4448 MPASS(scctx->isc_tx_csum_flags);
4451 if_setcapabilities(ifp, scctx->isc_capabilities | IFCAP_HWSTATS);
4452 if_setcapenable(ifp, scctx->isc_capenable | IFCAP_HWSTATS);
4454 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets))
4455 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max;
4456 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets))
4457 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max;
4459 main_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0;
4460 main_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0;
4462 /* XXX change for per-queue sizes */
4463 device_printf(dev, "Using %d tx descriptors and %d rx descriptors\n",
4464 scctx->isc_ntxd[main_txq], scctx->isc_nrxd[main_rxq]);
4465 for (i = 0; i < sctx->isc_nrxqs; i++) {
4466 if (!powerof2(scctx->isc_nrxd[i])) {
4467 /* round down instead? */
4468 device_printf(dev, "# rx descriptors must be a power of 2\n");
4470 goto fail_iflib_detach;
4473 for (i = 0; i < sctx->isc_ntxqs; i++) {
4474 if (!powerof2(scctx->isc_ntxd[i])) {
4476 "# tx descriptors must be a power of 2");
4478 goto fail_iflib_detach;
4482 if (scctx->isc_tx_nsegments > scctx->isc_ntxd[main_txq] /
4483 MAX_SINGLE_PACKET_FRACTION)
4484 scctx->isc_tx_nsegments = max(1, scctx->isc_ntxd[main_txq] /
4485 MAX_SINGLE_PACKET_FRACTION);
4486 if (scctx->isc_tx_tso_segments_max > scctx->isc_ntxd[main_txq] /
4487 MAX_SINGLE_PACKET_FRACTION)
4488 scctx->isc_tx_tso_segments_max = max(1,
4489 scctx->isc_ntxd[main_txq] / MAX_SINGLE_PACKET_FRACTION);
4491 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */
4492 if (if_getcapabilities(ifp) & IFCAP_TSO) {
4494 * The stack can't handle a TSO size larger than IP_MAXPACKET,
4497 if_sethwtsomax(ifp, min(scctx->isc_tx_tso_size_max,
4500 * Take maximum number of m_pullup(9)'s in iflib_parse_header()
4501 * into account. In the worst case, each of these calls will
4502 * add another mbuf and, thus, the requirement for another DMA
4503 * segment. So for best performance, it doesn't make sense to
4504 * advertize a maximum of TSO segments that typically will
4505 * require defragmentation in iflib_encap().
4507 if_sethwtsomaxsegcount(ifp, scctx->isc_tx_tso_segments_max - 3);
4508 if_sethwtsomaxsegsize(ifp, scctx->isc_tx_tso_segsize_max);
4510 if (scctx->isc_rss_table_size == 0)
4511 scctx->isc_rss_table_size = 64;
4512 scctx->isc_rss_table_mask = scctx->isc_rss_table_size-1;
4514 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx);
4515 /* XXX format name */
4516 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx, -1, "admin");
4518 /* Set up cpu set. If it fails, use the set of all CPUs. */
4519 if (bus_get_cpus(dev, INTR_CPUS, sizeof(ctx->ifc_cpus), &ctx->ifc_cpus) != 0) {
4520 device_printf(dev, "Unable to fetch CPU list\n");
4521 CPU_COPY(&all_cpus, &ctx->ifc_cpus);
4523 MPASS(CPU_COUNT(&ctx->ifc_cpus) > 0);
4526 ** Now set up MSI or MSI-X, should return us the number of supported
4527 ** vectors (will be 1 for a legacy interrupt and MSI).
4529 if (sctx->isc_flags & IFLIB_SKIP_MSIX) {
4530 msix = scctx->isc_vectors;
4531 } else if (scctx->isc_msix_bar != 0)
4533 * The simple fact that isc_msix_bar is not 0 does not mean we
4534 * we have a good value there that is known to work.
4536 msix = iflib_msix_init(ctx);
4538 scctx->isc_vectors = 1;
4539 scctx->isc_ntxqsets = 1;
4540 scctx->isc_nrxqsets = 1;
4541 scctx->isc_intr = IFLIB_INTR_LEGACY;
4544 /* Get memory for the station queues */
4545 if ((err = iflib_queues_alloc(ctx))) {
4546 device_printf(dev, "Unable to allocate queue memory\n");
4547 goto fail_intr_free;
4550 if ((err = iflib_qset_structures_setup(ctx)))
4554 * Group taskqueues aren't properly set up until SMP is started,
4555 * so we disable interrupts until we can handle them post
4558 * XXX: disabling interrupts doesn't actually work, at least for
4559 * the non-MSI case. When they occur before SI_SUB_SMP completes,
4560 * we do null handling and depend on this not causing too large an
4563 IFDI_INTR_DISABLE(ctx);
4564 if (msix > 1 && (err = IFDI_MSIX_INTR_ASSIGN(ctx, msix)) != 0) {
4565 device_printf(dev, "IFDI_MSIX_INTR_ASSIGN failed %d\n", err);
4570 if (scctx->isc_intr == IFLIB_INTR_MSI) {
4574 if ((err = iflib_legacy_setup(ctx, ctx->isc_legacy_intr, ctx->ifc_softc, &rid, "irq0")) != 0) {
4575 device_printf(dev, "iflib_legacy_setup failed %d\n", err);
4580 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4582 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4583 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4588 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4589 * This must appear after the call to ether_ifattach() because
4590 * ether_ifattach() sets if_hdrlen to the default value.
4592 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4593 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
4595 if ((err = iflib_netmap_attach(ctx))) {
4596 device_printf(ctx->ifc_dev, "netmap attach failed: %d\n", err);
4601 NETDUMP_SET(ctx->ifc_ifp, iflib);
4603 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4604 iflib_add_device_sysctl_post(ctx);
4605 ctx->ifc_flags |= IFC_INIT_DONE;
4610 ether_ifdetach(ctx->ifc_ifp);
4612 iflib_free_intr_mem(ctx);
4614 iflib_tx_structures_free(ctx);
4615 iflib_rx_structures_free(ctx);
4621 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4622 free(ctx->ifc_softc, M_IFLIB);
4628 iflib_pseudo_register(device_t dev, if_shared_ctx_t sctx, if_ctx_t *ctxp,
4629 struct iflib_cloneattach_ctx *clctx)
4634 if_softc_ctx_t scctx;
4640 ctx = malloc(sizeof(*ctx), M_IFLIB, M_WAITOK|M_ZERO);
4641 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK|M_ZERO);
4642 ctx->ifc_flags |= IFC_SC_ALLOCATED;
4643 if (sctx->isc_flags & (IFLIB_PSEUDO|IFLIB_VIRTUAL))
4644 ctx->ifc_flags |= IFC_PSEUDO;
4646 ctx->ifc_sctx = sctx;
4647 ctx->ifc_softc = sc;
4650 if ((err = iflib_register(ctx)) != 0) {
4651 device_printf(dev, "%s: iflib_register failed %d\n", __func__, err);
4654 iflib_add_device_sysctl_pre(ctx);
4656 scctx = &ctx->ifc_softc_ctx;
4660 * XXX sanity check that ntxd & nrxd are a power of 2
4662 iflib_reset_qvalues(ctx);
4664 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) {
4665 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err);
4668 if (sctx->isc_flags & IFLIB_GEN_MAC)
4670 if ((err = IFDI_CLONEATTACH(ctx, clctx->cc_ifc, clctx->cc_name,
4671 clctx->cc_params)) != 0) {
4672 device_printf(dev, "IFDI_CLONEATTACH failed %d\n", err);
4675 ifmedia_add(&ctx->ifc_media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
4676 ifmedia_add(&ctx->ifc_media, IFM_ETHER | IFM_AUTO, 0, NULL);
4677 ifmedia_set(&ctx->ifc_media, IFM_ETHER | IFM_AUTO);
4680 if (scctx->isc_capabilities & IFCAP_TXCSUM)
4681 MPASS(scctx->isc_tx_csum_flags);
4684 if_setcapabilities(ifp, scctx->isc_capabilities | IFCAP_HWSTATS | IFCAP_LINKSTATE);
4685 if_setcapenable(ifp, scctx->isc_capenable | IFCAP_HWSTATS | IFCAP_LINKSTATE);
4687 ifp->if_flags |= IFF_NOGROUP;
4688 if (sctx->isc_flags & IFLIB_PSEUDO) {
4689 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4691 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4692 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4698 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4699 * This must appear after the call to ether_ifattach() because
4700 * ether_ifattach() sets if_hdrlen to the default value.
4702 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4703 if_setifheaderlen(ifp,
4704 sizeof(struct ether_vlan_header));
4706 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4707 iflib_add_device_sysctl_post(ctx);
4708 ctx->ifc_flags |= IFC_INIT_DONE;
4711 _iflib_pre_assert(scctx);
4712 ctx->ifc_txrx = *scctx->isc_txrx;
4714 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets))
4715 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max;
4716 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets))
4717 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max;
4719 main_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0;
4720 main_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0;
4722 /* XXX change for per-queue sizes */
4723 device_printf(dev, "Using %d tx descriptors and %d rx descriptors\n",
4724 scctx->isc_ntxd[main_txq], scctx->isc_nrxd[main_rxq]);
4725 for (i = 0; i < sctx->isc_nrxqs; i++) {
4726 if (!powerof2(scctx->isc_nrxd[i])) {
4727 /* round down instead? */
4728 device_printf(dev, "# rx descriptors must be a power of 2\n");
4730 goto fail_iflib_detach;
4733 for (i = 0; i < sctx->isc_ntxqs; i++) {
4734 if (!powerof2(scctx->isc_ntxd[i])) {
4736 "# tx descriptors must be a power of 2");
4738 goto fail_iflib_detach;
4742 if (scctx->isc_tx_nsegments > scctx->isc_ntxd[main_txq] /
4743 MAX_SINGLE_PACKET_FRACTION)
4744 scctx->isc_tx_nsegments = max(1, scctx->isc_ntxd[main_txq] /
4745 MAX_SINGLE_PACKET_FRACTION);
4746 if (scctx->isc_tx_tso_segments_max > scctx->isc_ntxd[main_txq] /
4747 MAX_SINGLE_PACKET_FRACTION)
4748 scctx->isc_tx_tso_segments_max = max(1,
4749 scctx->isc_ntxd[main_txq] / MAX_SINGLE_PACKET_FRACTION);
4751 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */
4752 if (if_getcapabilities(ifp) & IFCAP_TSO) {
4754 * The stack can't handle a TSO size larger than IP_MAXPACKET,
4757 if_sethwtsomax(ifp, min(scctx->isc_tx_tso_size_max,
4760 * Take maximum number of m_pullup(9)'s in iflib_parse_header()
4761 * into account. In the worst case, each of these calls will
4762 * add another mbuf and, thus, the requirement for another DMA
4763 * segment. So for best performance, it doesn't make sense to
4764 * advertize a maximum of TSO segments that typically will
4765 * require defragmentation in iflib_encap().
4767 if_sethwtsomaxsegcount(ifp, scctx->isc_tx_tso_segments_max - 3);
4768 if_sethwtsomaxsegsize(ifp, scctx->isc_tx_tso_segsize_max);
4770 if (scctx->isc_rss_table_size == 0)
4771 scctx->isc_rss_table_size = 64;
4772 scctx->isc_rss_table_mask = scctx->isc_rss_table_size-1;
4774 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx);
4775 /* XXX format name */
4776 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx, -1, "admin");
4778 /* XXX --- can support > 1 -- but keep it simple for now */
4779 scctx->isc_intr = IFLIB_INTR_LEGACY;
4781 /* Get memory for the station queues */
4782 if ((err = iflib_queues_alloc(ctx))) {
4783 device_printf(dev, "Unable to allocate queue memory\n");
4784 goto fail_iflib_detach;
4787 if ((err = iflib_qset_structures_setup(ctx))) {
4788 device_printf(dev, "qset structure setup failed %d\n", err);
4793 * XXX What if anything do we want to do about interrupts?
4795 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac);
4796 if ((err = IFDI_ATTACH_POST(ctx)) != 0) {
4797 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err);
4802 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported.
4803 * This must appear after the call to ether_ifattach() because
4804 * ether_ifattach() sets if_hdrlen to the default value.
4806 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU)
4807 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
4809 /* XXX handle more than one queue */
4810 for (i = 0; i < scctx->isc_nrxqsets; i++)
4811 IFDI_RX_CLSET(ctx, 0, i, ctx->ifc_rxqs[i].ifr_fl[0].ifl_sds.ifsd_cl);
4815 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
4816 iflib_add_device_sysctl_post(ctx);
4817 ctx->ifc_flags |= IFC_INIT_DONE;
4821 ether_ifdetach(ctx->ifc_ifp);
4823 iflib_tx_structures_free(ctx);
4824 iflib_rx_structures_free(ctx);
4830 free(ctx->ifc_softc, M_IFLIB);
4836 iflib_pseudo_deregister(if_ctx_t ctx)
4838 if_t ifp = ctx->ifc_ifp;
4842 struct taskqgroup *tqg;
4845 /* Unregister VLAN events */
4846 if (ctx->ifc_vlan_attach_event != NULL)
4847 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event);
4848 if (ctx->ifc_vlan_detach_event != NULL)
4849 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event);
4851 ether_ifdetach(ifp);
4852 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/
4853 CTX_LOCK_DESTROY(ctx);
4854 /* XXX drain any dependent tasks */
4855 tqg = qgroup_if_io_tqg;
4856 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) {
4857 callout_drain(&txq->ift_timer);
4858 if (txq->ift_task.gt_uniq != NULL)
4859 taskqgroup_detach(tqg, &txq->ift_task);
4861 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) {
4862 if (rxq->ifr_task.gt_uniq != NULL)
4863 taskqgroup_detach(tqg, &rxq->ifr_task);
4865 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
4866 free(fl->ifl_rx_bitmap, M_IFLIB);
4868 tqg = qgroup_if_config_tqg;
4869 if (ctx->ifc_admin_task.gt_uniq != NULL)
4870 taskqgroup_detach(tqg, &ctx->ifc_admin_task);
4871 if (ctx->ifc_vflr_task.gt_uniq != NULL)
4872 taskqgroup_detach(tqg, &ctx->ifc_vflr_task);
4876 iflib_tx_structures_free(ctx);
4877 iflib_rx_structures_free(ctx);
4878 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4879 free(ctx->ifc_softc, M_IFLIB);
4885 iflib_device_attach(device_t dev)
4888 if_shared_ctx_t sctx;
4890 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC)
4893 pci_enable_busmaster(dev);
4895 return (iflib_device_register(dev, NULL, sctx, &ctx));
4899 iflib_device_deregister(if_ctx_t ctx)
4901 if_t ifp = ctx->ifc_ifp;
4904 device_t dev = ctx->ifc_dev;
4906 struct taskqgroup *tqg;
4909 /* Make sure VLANS are not using driver */
4910 if (if_vlantrunkinuse(ifp)) {
4911 device_printf(dev, "Vlan in use, detach first\n");
4915 if (!CTX_IS_VF(ctx) && pci_iov_detach(dev) != 0) {
4916 device_printf(dev, "SR-IOV in use; detach first.\n");
4922 ctx->ifc_flags |= IFC_IN_DETACH;
4929 /* Unregister VLAN events */
4930 if (ctx->ifc_vlan_attach_event != NULL)
4931 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event);
4932 if (ctx->ifc_vlan_detach_event != NULL)
4933 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event);
4935 iflib_netmap_detach(ifp);
4936 ether_ifdetach(ifp);
4937 if (ctx->ifc_led_dev != NULL)
4938 led_destroy(ctx->ifc_led_dev);
4939 /* XXX drain any dependent tasks */
4940 tqg = qgroup_if_io_tqg;
4941 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) {
4942 callout_drain(&txq->ift_timer);
4943 if (txq->ift_task.gt_uniq != NULL)
4944 taskqgroup_detach(tqg, &txq->ift_task);
4946 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) {
4947 if (rxq->ifr_task.gt_uniq != NULL)
4948 taskqgroup_detach(tqg, &rxq->ifr_task);
4950 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
4951 free(fl->ifl_rx_bitmap, M_IFLIB);
4953 tqg = qgroup_if_config_tqg;
4954 if (ctx->ifc_admin_task.gt_uniq != NULL)
4955 taskqgroup_detach(tqg, &ctx->ifc_admin_task);
4956 if (ctx->ifc_vflr_task.gt_uniq != NULL)
4957 taskqgroup_detach(tqg, &ctx->ifc_vflr_task);
4962 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/
4963 CTX_LOCK_DESTROY(ctx);
4964 device_set_softc(ctx->ifc_dev, NULL);
4965 iflib_free_intr_mem(ctx);
4967 bus_generic_detach(dev);
4970 iflib_tx_structures_free(ctx);
4971 iflib_rx_structures_free(ctx);
4972 if (ctx->ifc_flags & IFC_SC_ALLOCATED)
4973 free(ctx->ifc_softc, M_IFLIB);
4974 STATE_LOCK_DESTROY(ctx);
4980 iflib_free_intr_mem(if_ctx_t ctx)
4983 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) {
4984 iflib_irq_free(ctx, &ctx->ifc_legacy_irq);
4986 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) {
4987 pci_release_msi(ctx->ifc_dev);
4989 if (ctx->ifc_msix_mem != NULL) {
4990 bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY,
4991 rman_get_rid(ctx->ifc_msix_mem), ctx->ifc_msix_mem);
4992 ctx->ifc_msix_mem = NULL;
4997 iflib_device_detach(device_t dev)
4999 if_ctx_t ctx = device_get_softc(dev);
5001 return (iflib_device_deregister(ctx));
5005 iflib_device_suspend(device_t dev)
5007 if_ctx_t ctx = device_get_softc(dev);
5013 return bus_generic_suspend(dev);
5016 iflib_device_shutdown(device_t dev)
5018 if_ctx_t ctx = device_get_softc(dev);
5024 return bus_generic_suspend(dev);
5029 iflib_device_resume(device_t dev)
5031 if_ctx_t ctx = device_get_softc(dev);
5032 iflib_txq_t txq = ctx->ifc_txqs;
5036 iflib_if_init_locked(ctx);
5038 for (int i = 0; i < NTXQSETS(ctx); i++, txq++)
5039 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET);
5041 return (bus_generic_resume(dev));
5045 iflib_device_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
5048 if_ctx_t ctx = device_get_softc(dev);
5051 error = IFDI_IOV_INIT(ctx, num_vfs, params);
5058 iflib_device_iov_uninit(device_t dev)
5060 if_ctx_t ctx = device_get_softc(dev);
5063 IFDI_IOV_UNINIT(ctx);
5068 iflib_device_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
5071 if_ctx_t ctx = device_get_softc(dev);
5074 error = IFDI_IOV_VF_ADD(ctx, vfnum, params);
5080 /*********************************************************************
5082 * MODULE FUNCTION DEFINITIONS
5084 **********************************************************************/
5087 * - Start a fast taskqueue thread for each core
5088 * - Start a taskqueue for control operations
5091 iflib_module_init(void)
5097 iflib_module_event_handler(module_t mod, int what, void *arg)
5103 if ((err = iflib_module_init()) != 0)
5109 return (EOPNOTSUPP);
5115 /*********************************************************************
5117 * PUBLIC FUNCTION DEFINITIONS
5118 * ordered as in iflib.h
5120 **********************************************************************/
5124 _iflib_assert(if_shared_ctx_t sctx)
5126 MPASS(sctx->isc_tx_maxsize);
5127 MPASS(sctx->isc_tx_maxsegsize);
5129 MPASS(sctx->isc_rx_maxsize);
5130 MPASS(sctx->isc_rx_nsegments);
5131 MPASS(sctx->isc_rx_maxsegsize);
5133 MPASS(sctx->isc_nrxd_min[0]);
5134 MPASS(sctx->isc_nrxd_max[0]);
5135 MPASS(sctx->isc_nrxd_default[0]);
5136 MPASS(sctx->isc_ntxd_min[0]);
5137 MPASS(sctx->isc_ntxd_max[0]);
5138 MPASS(sctx->isc_ntxd_default[0]);
5142 _iflib_pre_assert(if_softc_ctx_t scctx)
5145 MPASS(scctx->isc_txrx->ift_txd_encap);
5146 MPASS(scctx->isc_txrx->ift_txd_flush);
5147 MPASS(scctx->isc_txrx->ift_txd_credits_update);
5148 MPASS(scctx->isc_txrx->ift_rxd_available);
5149 MPASS(scctx->isc_txrx->ift_rxd_pkt_get);
5150 MPASS(scctx->isc_txrx->ift_rxd_refill);
5151 MPASS(scctx->isc_txrx->ift_rxd_flush);
5155 iflib_register(if_ctx_t ctx)
5157 if_shared_ctx_t sctx = ctx->ifc_sctx;
5158 driver_t *driver = sctx->isc_driver;
5159 device_t dev = ctx->ifc_dev;
5162 _iflib_assert(sctx);
5165 STATE_LOCK_INIT(ctx, device_get_nameunit(ctx->ifc_dev));
5166 ifp = ctx->ifc_ifp = if_alloc(IFT_ETHER);
5168 device_printf(dev, "can not allocate ifnet structure\n");
5173 * Initialize our context's device specific methods
5175 kobj_init((kobj_t) ctx, (kobj_class_t) driver);
5176 kobj_class_compile((kobj_class_t) driver);
5179 if_initname(ifp, device_get_name(dev), device_get_unit(dev));
5180 if_setsoftc(ifp, ctx);
5181 if_setdev(ifp, dev);
5182 if_setinitfn(ifp, iflib_if_init);
5183 if_setioctlfn(ifp, iflib_if_ioctl);
5185 if_setstartfn(ifp, iflib_altq_if_start);
5186 if_settransmitfn(ifp, iflib_altq_if_transmit);
5187 if_setsendqready(ifp);
5189 if_settransmitfn(ifp, iflib_if_transmit);
5191 if_setqflushfn(ifp, iflib_if_qflush);
5192 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
5194 ctx->ifc_vlan_attach_event =
5195 EVENTHANDLER_REGISTER(vlan_config, iflib_vlan_register, ctx,
5196 EVENTHANDLER_PRI_FIRST);
5197 ctx->ifc_vlan_detach_event =
5198 EVENTHANDLER_REGISTER(vlan_unconfig, iflib_vlan_unregister, ctx,
5199 EVENTHANDLER_PRI_FIRST);
5201 ifmedia_init(&ctx->ifc_media, IFM_IMASK,
5202 iflib_media_change, iflib_media_status);
5209 iflib_queues_alloc(if_ctx_t ctx)
5211 if_shared_ctx_t sctx = ctx->ifc_sctx;
5212 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
5213 device_t dev = ctx->ifc_dev;
5214 int nrxqsets = scctx->isc_nrxqsets;
5215 int ntxqsets = scctx->isc_ntxqsets;
5218 iflib_fl_t fl = NULL;
5219 int i, j, cpu, err, txconf, rxconf;
5220 iflib_dma_info_t ifdip;
5221 uint32_t *rxqsizes = scctx->isc_rxqsizes;
5222 uint32_t *txqsizes = scctx->isc_txqsizes;
5223 uint8_t nrxqs = sctx->isc_nrxqs;
5224 uint8_t ntxqs = sctx->isc_ntxqs;
5225 int nfree_lists = sctx->isc_nfl ? sctx->isc_nfl : 1;
5229 KASSERT(ntxqs > 0, ("number of queues per qset must be at least 1"));
5230 KASSERT(nrxqs > 0, ("number of queues per qset must be at least 1"));
5232 /* Allocate the TX ring struct memory */
5233 if (!(ctx->ifc_txqs =
5234 (iflib_txq_t) malloc(sizeof(struct iflib_txq) *
5235 ntxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) {
5236 device_printf(dev, "Unable to allocate TX ring memory\n");
5241 /* Now allocate the RX */
5242 if (!(ctx->ifc_rxqs =
5243 (iflib_rxq_t) malloc(sizeof(struct iflib_rxq) *
5244 nrxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) {
5245 device_printf(dev, "Unable to allocate RX ring memory\n");
5250 txq = ctx->ifc_txqs;
5251 rxq = ctx->ifc_rxqs;
5254 * XXX handle allocation failure
5256 for (txconf = i = 0, cpu = CPU_FIRST(); i < ntxqsets; i++, txconf++, txq++, cpu = CPU_NEXT(cpu)) {
5257 /* Set up some basics */
5259 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * ntxqs,
5260 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
5262 "Unable to allocate TX DMA info memory\n");
5266 txq->ift_ifdi = ifdip;
5267 for (j = 0; j < ntxqs; j++, ifdip++) {
5268 if (iflib_dma_alloc(ctx, txqsizes[j], ifdip, 0)) {
5270 "Unable to allocate TX descriptors\n");
5274 txq->ift_txd_size[j] = scctx->isc_txd_size[j];
5275 bzero((void *)ifdip->idi_vaddr, txqsizes[j]);
5279 if (sctx->isc_flags & IFLIB_HAS_TXCQ) {
5280 txq->ift_br_offset = 1;
5282 txq->ift_br_offset = 0;
5285 txq->ift_timer.c_cpu = cpu;
5287 if (iflib_txsd_alloc(txq)) {
5288 device_printf(dev, "Critical Failure setting up TX buffers\n");
5293 /* Initialize the TX lock */
5294 snprintf(txq->ift_mtx_name, MTX_NAME_LEN, "%s:tx(%d):callout",
5295 device_get_nameunit(dev), txq->ift_id);
5296 mtx_init(&txq->ift_mtx, txq->ift_mtx_name, NULL, MTX_DEF);
5297 callout_init_mtx(&txq->ift_timer, &txq->ift_mtx, 0);
5299 snprintf(txq->ift_db_mtx_name, MTX_NAME_LEN, "%s:tx(%d):db",
5300 device_get_nameunit(dev), txq->ift_id);
5302 err = ifmp_ring_alloc(&txq->ift_br, 2048, txq, iflib_txq_drain,
5303 iflib_txq_can_drain, M_IFLIB, M_WAITOK);
5305 /* XXX free any allocated rings */
5306 device_printf(dev, "Unable to allocate buf_ring\n");
5311 for (rxconf = i = 0; i < nrxqsets; i++, rxconf++, rxq++) {
5312 /* Set up some basics */
5314 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * nrxqs,
5315 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) {
5317 "Unable to allocate RX DMA info memory\n");
5322 rxq->ifr_ifdi = ifdip;
5323 /* XXX this needs to be changed if #rx queues != #tx queues */
5324 rxq->ifr_ntxqirq = 1;
5325 rxq->ifr_txqid[0] = i;
5326 for (j = 0; j < nrxqs; j++, ifdip++) {
5327 if (iflib_dma_alloc(ctx, rxqsizes[j], ifdip, 0)) {
5329 "Unable to allocate RX descriptors\n");
5333 bzero((void *)ifdip->idi_vaddr, rxqsizes[j]);
5337 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
5338 rxq->ifr_fl_offset = 1;
5340 rxq->ifr_fl_offset = 0;
5342 rxq->ifr_nfl = nfree_lists;
5344 (iflib_fl_t) malloc(sizeof(struct iflib_fl) * nfree_lists, M_IFLIB, M_NOWAIT | M_ZERO))) {
5345 device_printf(dev, "Unable to allocate free list memory\n");
5350 for (j = 0; j < nfree_lists; j++) {
5351 fl[j].ifl_rxq = rxq;
5353 fl[j].ifl_ifdi = &rxq->ifr_ifdi[j + rxq->ifr_fl_offset];
5354 fl[j].ifl_rxd_size = scctx->isc_rxd_size[j];
5356 /* Allocate receive buffers for the ring */
5357 if (iflib_rxsd_alloc(rxq)) {
5359 "Critical Failure setting up receive buffers\n");
5364 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++)
5365 fl->ifl_rx_bitmap = bit_alloc(fl->ifl_size, M_IFLIB,
5370 vaddrs = malloc(sizeof(caddr_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK);
5371 paddrs = malloc(sizeof(uint64_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK);
5372 for (i = 0; i < ntxqsets; i++) {
5373 iflib_dma_info_t di = ctx->ifc_txqs[i].ift_ifdi;
5375 for (j = 0; j < ntxqs; j++, di++) {
5376 vaddrs[i*ntxqs + j] = di->idi_vaddr;
5377 paddrs[i*ntxqs + j] = di->idi_paddr;
5380 if ((err = IFDI_TX_QUEUES_ALLOC(ctx, vaddrs, paddrs, ntxqs, ntxqsets)) != 0) {
5381 device_printf(ctx->ifc_dev,
5382 "Unable to allocate device TX queue\n");
5383 iflib_tx_structures_free(ctx);
5384 free(vaddrs, M_IFLIB);
5385 free(paddrs, M_IFLIB);
5388 free(vaddrs, M_IFLIB);
5389 free(paddrs, M_IFLIB);
5392 vaddrs = malloc(sizeof(caddr_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK);
5393 paddrs = malloc(sizeof(uint64_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK);
5394 for (i = 0; i < nrxqsets; i++) {
5395 iflib_dma_info_t di = ctx->ifc_rxqs[i].ifr_ifdi;
5397 for (j = 0; j < nrxqs; j++, di++) {
5398 vaddrs[i*nrxqs + j] = di->idi_vaddr;
5399 paddrs[i*nrxqs + j] = di->idi_paddr;
5402 if ((err = IFDI_RX_QUEUES_ALLOC(ctx, vaddrs, paddrs, nrxqs, nrxqsets)) != 0) {
5403 device_printf(ctx->ifc_dev,
5404 "Unable to allocate device RX queue\n");
5405 iflib_tx_structures_free(ctx);
5406 free(vaddrs, M_IFLIB);
5407 free(paddrs, M_IFLIB);
5410 free(vaddrs, M_IFLIB);
5411 free(paddrs, M_IFLIB);
5415 /* XXX handle allocation failure changes */
5419 if (ctx->ifc_rxqs != NULL)
5420 free(ctx->ifc_rxqs, M_IFLIB);
5421 ctx->ifc_rxqs = NULL;
5422 if (ctx->ifc_txqs != NULL)
5423 free(ctx->ifc_txqs, M_IFLIB);
5424 ctx->ifc_txqs = NULL;
5430 iflib_tx_structures_setup(if_ctx_t ctx)
5432 iflib_txq_t txq = ctx->ifc_txqs;
5435 for (i = 0; i < NTXQSETS(ctx); i++, txq++)
5436 iflib_txq_setup(txq);
5442 iflib_tx_structures_free(if_ctx_t ctx)
5444 iflib_txq_t txq = ctx->ifc_txqs;
5445 if_shared_ctx_t sctx = ctx->ifc_sctx;
5448 for (i = 0; i < NTXQSETS(ctx); i++, txq++) {
5449 iflib_txq_destroy(txq);
5450 for (j = 0; j < sctx->isc_ntxqs; j++)
5451 iflib_dma_free(&txq->ift_ifdi[j]);
5453 free(ctx->ifc_txqs, M_IFLIB);
5454 ctx->ifc_txqs = NULL;
5455 IFDI_QUEUES_FREE(ctx);
5458 /*********************************************************************
5460 * Initialize all receive rings.
5462 **********************************************************************/
5464 iflib_rx_structures_setup(if_ctx_t ctx)
5466 iflib_rxq_t rxq = ctx->ifc_rxqs;
5468 #if defined(INET6) || defined(INET)
5472 for (q = 0; q < ctx->ifc_softc_ctx.isc_nrxqsets; q++, rxq++) {
5473 #if defined(INET6) || defined(INET)
5474 tcp_lro_free(&rxq->ifr_lc);
5475 if ((err = tcp_lro_init_args(&rxq->ifr_lc, ctx->ifc_ifp,
5476 TCP_LRO_ENTRIES, min(1024,
5477 ctx->ifc_softc_ctx.isc_nrxd[rxq->ifr_fl_offset]))) != 0) {
5478 device_printf(ctx->ifc_dev, "LRO Initialization failed!\n");
5481 rxq->ifr_lro_enabled = TRUE;
5483 IFDI_RXQ_SETUP(ctx, rxq->ifr_id);
5486 #if defined(INET6) || defined(INET)
5489 * Free RX software descriptors allocated so far, we will only handle
5490 * the rings that completed, the failing case will have
5491 * cleaned up for itself. 'q' failed, so its the terminus.
5493 rxq = ctx->ifc_rxqs;
5494 for (i = 0; i < q; ++i, rxq++) {
5495 iflib_rx_sds_free(rxq);
5496 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0;
5502 /*********************************************************************
5504 * Free all receive rings.
5506 **********************************************************************/
5508 iflib_rx_structures_free(if_ctx_t ctx)
5510 iflib_rxq_t rxq = ctx->ifc_rxqs;
5512 for (int i = 0; i < ctx->ifc_softc_ctx.isc_nrxqsets; i++, rxq++) {
5513 iflib_rx_sds_free(rxq);
5515 free(ctx->ifc_rxqs, M_IFLIB);
5516 ctx->ifc_rxqs = NULL;
5520 iflib_qset_structures_setup(if_ctx_t ctx)
5525 * It is expected that the caller takes care of freeing queues if this
5528 if ((err = iflib_tx_structures_setup(ctx)) != 0) {
5529 device_printf(ctx->ifc_dev, "iflib_tx_structures_setup failed: %d\n", err);
5533 if ((err = iflib_rx_structures_setup(ctx)) != 0)
5534 device_printf(ctx->ifc_dev, "iflib_rx_structures_setup failed: %d\n", err);
5540 iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid,
5541 driver_filter_t filter, void *filter_arg, driver_intr_t handler, void *arg, const char *name)
5544 return (_iflib_irq_alloc(ctx, irq, rid, filter, handler, arg, name));
5549 find_nth(if_ctx_t ctx, int qid)
5552 int i, cpuid, eqid, count;
5554 CPU_COPY(&ctx->ifc_cpus, &cpus);
5555 count = CPU_COUNT(&cpus);
5557 /* clear up to the qid'th bit */
5558 for (i = 0; i < eqid; i++) {
5559 cpuid = CPU_FFS(&cpus);
5561 CPU_CLR(cpuid-1, &cpus);
5563 cpuid = CPU_FFS(&cpus);
5569 extern struct cpu_group *cpu_top; /* CPU topology */
5572 find_child_with_core(int cpu, struct cpu_group *grp)
5576 if (grp->cg_children == 0)
5579 MPASS(grp->cg_child);
5580 for (i = 0; i < grp->cg_children; i++) {
5581 if (CPU_ISSET(cpu, &grp->cg_child[i].cg_mask))
5589 * Find the nth "close" core to the specified core
5590 * "close" is defined as the deepest level that shares
5591 * at least an L2 cache. With threads, this will be
5592 * threads on the same core. If the sahred cache is L3
5593 * or higher, simply returns the same core.
5596 find_close_core(int cpu, int core_offset)
5598 struct cpu_group *grp;
5607 while ((i = find_child_with_core(cpu, grp)) != -1) {
5608 /* If the child only has one cpu, don't descend */
5609 if (grp->cg_child[i].cg_count <= 1)
5611 grp = &grp->cg_child[i];
5614 /* If they don't share at least an L2 cache, use the same CPU */
5615 if (grp->cg_level > CG_SHARE_L2 || grp->cg_level == CG_SHARE_NONE)
5619 CPU_COPY(&grp->cg_mask, &cs);
5621 /* Add the selected CPU offset to core offset. */
5622 for (i = 0; (fcpu = CPU_FFS(&cs)) != 0; i++) {
5623 if (fcpu - 1 == cpu)
5625 CPU_CLR(fcpu - 1, &cs);
5631 CPU_COPY(&grp->cg_mask, &cs);
5632 for (i = core_offset % grp->cg_count; i > 0; i--) {
5633 MPASS(CPU_FFS(&cs));
5634 CPU_CLR(CPU_FFS(&cs) - 1, &cs);
5636 MPASS(CPU_FFS(&cs));
5637 return CPU_FFS(&cs) - 1;
5641 find_close_core(int cpu, int core_offset __unused)
5648 get_core_offset(if_ctx_t ctx, iflib_intr_type_t type, int qid)
5652 /* TX queues get cores which share at least an L2 cache with the corresponding RX queue */
5653 /* XXX handle multiple RX threads per core and more than two core per L2 group */
5654 return qid / CPU_COUNT(&ctx->ifc_cpus) + 1;
5656 case IFLIB_INTR_RXTX:
5657 /* RX queues get the specified core */
5658 return qid / CPU_COUNT(&ctx->ifc_cpus);
5664 #define get_core_offset(ctx, type, qid) CPU_FIRST()
5665 #define find_close_core(cpuid, tid) CPU_FIRST()
5666 #define find_nth(ctx, gid) CPU_FIRST()
5669 /* Just to avoid copy/paste */
5671 iflib_irq_set_affinity(if_ctx_t ctx, int irq, iflib_intr_type_t type, int qid,
5672 struct grouptask *gtask, struct taskqgroup *tqg, void *uniq, const char *name)
5677 cpuid = find_nth(ctx, qid);
5678 tid = get_core_offset(ctx, type, qid);
5680 cpuid = find_close_core(cpuid, tid);
5681 err = taskqgroup_attach_cpu(tqg, gtask, uniq, cpuid, irq, name);
5683 device_printf(ctx->ifc_dev, "taskqgroup_attach_cpu failed %d\n", err);
5687 if (cpuid > ctx->ifc_cpuid_highest)
5688 ctx->ifc_cpuid_highest = cpuid;
5694 iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid,
5695 iflib_intr_type_t type, driver_filter_t *filter,
5696 void *filter_arg, int qid, const char *name)
5698 struct grouptask *gtask;
5699 struct taskqgroup *tqg;
5700 iflib_filter_info_t info;
5703 driver_filter_t *intr_fast;
5706 info = &ctx->ifc_filter_info;
5710 /* XXX merge tx/rx for netmap? */
5712 q = &ctx->ifc_txqs[qid];
5713 info = &ctx->ifc_txqs[qid].ift_filter_info;
5714 gtask = &ctx->ifc_txqs[qid].ift_task;
5715 tqg = qgroup_if_io_tqg;
5717 intr_fast = iflib_fast_intr;
5718 GROUPTASK_INIT(gtask, 0, fn, q);
5719 ctx->ifc_flags |= IFC_NETMAP_TX_IRQ;
5722 q = &ctx->ifc_rxqs[qid];
5723 info = &ctx->ifc_rxqs[qid].ifr_filter_info;
5724 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5725 tqg = qgroup_if_io_tqg;
5727 intr_fast = iflib_fast_intr;
5728 GROUPTASK_INIT(gtask, 0, fn, q);
5730 case IFLIB_INTR_RXTX:
5731 q = &ctx->ifc_rxqs[qid];
5732 info = &ctx->ifc_rxqs[qid].ifr_filter_info;
5733 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5734 tqg = qgroup_if_io_tqg;
5736 intr_fast = iflib_fast_intr_rxtx;
5737 GROUPTASK_INIT(gtask, 0, fn, q);
5739 case IFLIB_INTR_ADMIN:
5742 info = &ctx->ifc_filter_info;
5743 gtask = &ctx->ifc_admin_task;
5744 tqg = qgroup_if_config_tqg;
5745 fn = _task_fn_admin;
5746 intr_fast = iflib_fast_intr_ctx;
5749 panic("unknown net intr type");
5752 info->ifi_filter = filter;
5753 info->ifi_filter_arg = filter_arg;
5754 info->ifi_task = gtask;
5757 err = _iflib_irq_alloc(ctx, irq, rid, intr_fast, NULL, info, name);
5759 device_printf(ctx->ifc_dev, "_iflib_irq_alloc failed %d\n", err);
5762 if (type == IFLIB_INTR_ADMIN)
5766 err = iflib_irq_set_affinity(ctx, rman_get_start(irq->ii_res), type, qid, gtask, tqg, q, name);
5770 taskqgroup_attach(tqg, gtask, q, rman_get_start(irq->ii_res), name);
5777 iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, const char *name)
5779 struct grouptask *gtask;
5780 struct taskqgroup *tqg;
5788 q = &ctx->ifc_txqs[qid];
5789 gtask = &ctx->ifc_txqs[qid].ift_task;
5790 tqg = qgroup_if_io_tqg;
5793 irq_num = rman_get_start(irq->ii_res);
5796 q = &ctx->ifc_rxqs[qid];
5797 gtask = &ctx->ifc_rxqs[qid].ifr_task;
5798 tqg = qgroup_if_io_tqg;
5801 irq_num = rman_get_start(irq->ii_res);
5803 case IFLIB_INTR_IOV:
5805 gtask = &ctx->ifc_vflr_task;
5806 tqg = qgroup_if_config_tqg;
5810 panic("unknown net intr type");
5812 GROUPTASK_INIT(gtask, 0, fn, q);
5813 if (irq_num != -1) {
5814 err = iflib_irq_set_affinity(ctx, irq_num, type, qid, gtask, tqg, q, name);
5816 taskqgroup_attach(tqg, gtask, q, irq_num, name);
5819 taskqgroup_attach(tqg, gtask, q, irq_num, name);
5824 iflib_irq_free(if_ctx_t ctx, if_irq_t irq)
5828 bus_teardown_intr(ctx->ifc_dev, irq->ii_res, irq->ii_tag);
5831 bus_release_resource(ctx->ifc_dev, SYS_RES_IRQ,
5832 rman_get_rid(irq->ii_res), irq->ii_res);
5836 iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filter_arg, int *rid, const char *name)
5838 iflib_txq_t txq = ctx->ifc_txqs;
5839 iflib_rxq_t rxq = ctx->ifc_rxqs;
5840 if_irq_t irq = &ctx->ifc_legacy_irq;
5841 iflib_filter_info_t info;
5842 struct grouptask *gtask;
5843 struct taskqgroup *tqg;
5849 q = &ctx->ifc_rxqs[0];
5850 info = &rxq[0].ifr_filter_info;
5851 gtask = &rxq[0].ifr_task;
5852 tqg = qgroup_if_io_tqg;
5853 tqrid = irq->ii_rid = *rid;
5856 ctx->ifc_flags |= IFC_LEGACY;
5857 info->ifi_filter = filter;
5858 info->ifi_filter_arg = filter_arg;
5859 info->ifi_task = gtask;
5860 info->ifi_ctx = ctx;
5862 /* We allocate a single interrupt resource */
5863 if ((err = _iflib_irq_alloc(ctx, irq, tqrid, iflib_fast_intr_ctx, NULL, info, name)) != 0)
5865 GROUPTASK_INIT(gtask, 0, fn, q);
5866 taskqgroup_attach(tqg, gtask, q, rman_get_start(irq->ii_res), name);
5868 GROUPTASK_INIT(&txq->ift_task, 0, _task_fn_tx, txq);
5869 taskqgroup_attach(qgroup_if_io_tqg, &txq->ift_task, txq, rman_get_start(irq->ii_res), "tx");
5874 iflib_led_create(if_ctx_t ctx)
5877 ctx->ifc_led_dev = led_create(iflib_led_func, ctx,
5878 device_get_nameunit(ctx->ifc_dev));
5882 iflib_tx_intr_deferred(if_ctx_t ctx, int txqid)
5885 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task);
5889 iflib_rx_intr_deferred(if_ctx_t ctx, int rxqid)
5892 GROUPTASK_ENQUEUE(&ctx->ifc_rxqs[rxqid].ifr_task);
5896 iflib_admin_intr_deferred(if_ctx_t ctx)
5899 struct grouptask *gtask;
5901 gtask = &ctx->ifc_admin_task;
5902 MPASS(gtask != NULL && gtask->gt_taskqueue != NULL);
5905 GROUPTASK_ENQUEUE(&ctx->ifc_admin_task);
5909 iflib_iov_intr_deferred(if_ctx_t ctx)
5912 GROUPTASK_ENQUEUE(&ctx->ifc_vflr_task);
5916 iflib_io_tqg_attach(struct grouptask *gt, void *uniq, int cpu, char *name)
5919 taskqgroup_attach_cpu(qgroup_if_io_tqg, gt, uniq, cpu, -1, name);
5923 iflib_config_gtask_init(void *ctx, struct grouptask *gtask, gtask_fn_t *fn,
5927 GROUPTASK_INIT(gtask, 0, fn, ctx);
5928 taskqgroup_attach(qgroup_if_config_tqg, gtask, gtask, -1, name);
5932 iflib_config_gtask_deinit(struct grouptask *gtask)
5935 taskqgroup_detach(qgroup_if_config_tqg, gtask);
5939 iflib_link_state_change(if_ctx_t ctx, int link_state, uint64_t baudrate)
5941 if_t ifp = ctx->ifc_ifp;
5942 iflib_txq_t txq = ctx->ifc_txqs;
5944 if_setbaudrate(ifp, baudrate);
5945 if (baudrate >= IF_Gbps(10)) {
5947 ctx->ifc_flags |= IFC_PREFETCH;
5950 /* If link down, disable watchdog */
5951 if ((ctx->ifc_link_state == LINK_STATE_UP) && (link_state == LINK_STATE_DOWN)) {
5952 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxqsets; i++, txq++)
5953 txq->ift_qstatus = IFLIB_QUEUE_IDLE;
5955 ctx->ifc_link_state = link_state;
5956 if_link_state_change(ifp, link_state);
5960 iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq)
5964 int credits_pre = txq->ift_cidx_processed;
5967 if (ctx->isc_txd_credits_update == NULL)
5970 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map,
5971 BUS_DMASYNC_POSTREAD);
5972 if ((credits = ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, true)) == 0)
5975 txq->ift_processed += credits;
5976 txq->ift_cidx_processed += credits;
5978 MPASS(credits_pre + credits == txq->ift_cidx_processed);
5979 if (txq->ift_cidx_processed >= txq->ift_size)
5980 txq->ift_cidx_processed -= txq->ift_size;
5985 iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget)
5988 return (ctx->isc_rxd_available(ctx->ifc_softc, rxq->ifr_id, cidx,
5993 iflib_add_int_delay_sysctl(if_ctx_t ctx, const char *name,
5994 const char *description, if_int_delay_info_t info,
5995 int offset, int value)
5997 info->iidi_ctx = ctx;
5998 info->iidi_offset = offset;
5999 info->iidi_value = value;
6000 SYSCTL_ADD_PROC(device_get_sysctl_ctx(ctx->ifc_dev),
6001 SYSCTL_CHILDREN(device_get_sysctl_tree(ctx->ifc_dev)),
6002 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
6003 info, 0, iflib_sysctl_int_delay, "I", description);
6007 iflib_ctx_lock_get(if_ctx_t ctx)
6010 return (&ctx->ifc_ctx_sx);
6014 iflib_msix_init(if_ctx_t ctx)
6016 device_t dev = ctx->ifc_dev;
6017 if_shared_ctx_t sctx = ctx->ifc_sctx;
6018 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
6019 int vectors, queues, rx_queues, tx_queues, queuemsgs, msgs;
6020 int iflib_num_tx_queues, iflib_num_rx_queues;
6021 int err, admincnt, bar;
6023 iflib_num_tx_queues = ctx->ifc_sysctl_ntxqs;
6024 iflib_num_rx_queues = ctx->ifc_sysctl_nrxqs;
6027 device_printf(dev, "msix_init qsets capped at %d\n",
6028 imax(scctx->isc_ntxqsets, scctx->isc_nrxqsets));
6030 bar = ctx->ifc_softc_ctx.isc_msix_bar;
6031 admincnt = sctx->isc_admin_intrcnt;
6032 /* Override by tuneable */
6033 if (scctx->isc_disable_msix)
6036 /* First try MSI-X */
6037 if ((msgs = pci_msix_count(dev)) == 0) {
6039 device_printf(dev, "MSI-X not supported or disabled\n");
6043 * bar == -1 => "trust me I know what I'm doing"
6044 * Some drivers are for hardware that is so shoddily
6045 * documented that no one knows which bars are which
6046 * so the developer has to map all bars. This hack
6047 * allows shoddy garbage to use MSI-X in this framework.
6050 ctx->ifc_msix_mem = bus_alloc_resource_any(dev,
6051 SYS_RES_MEMORY, &bar, RF_ACTIVE);
6052 if (ctx->ifc_msix_mem == NULL) {
6053 device_printf(dev, "Unable to map MSI-X table\n");
6058 /* use only 1 qset in debug mode */
6059 queuemsgs = min(msgs - admincnt, 1);
6061 queuemsgs = msgs - admincnt;
6064 queues = imin(queuemsgs, rss_getnumbuckets());
6068 queues = imin(CPU_COUNT(&ctx->ifc_cpus), queues);
6071 "intr CPUs: %d queue msgs: %d admincnt: %d\n",
6072 CPU_COUNT(&ctx->ifc_cpus), queuemsgs, admincnt);
6074 /* If we're doing RSS, clamp at the number of RSS buckets */
6075 if (queues > rss_getnumbuckets())
6076 queues = rss_getnumbuckets();
6078 if (iflib_num_rx_queues > 0 && iflib_num_rx_queues < queuemsgs - admincnt)
6079 rx_queues = iflib_num_rx_queues;
6083 if (rx_queues > scctx->isc_nrxqsets)
6084 rx_queues = scctx->isc_nrxqsets;
6087 * We want this to be all logical CPUs by default
6089 if (iflib_num_tx_queues > 0 && iflib_num_tx_queues < queues)
6090 tx_queues = iflib_num_tx_queues;
6092 tx_queues = mp_ncpus;
6094 if (tx_queues > scctx->isc_ntxqsets)
6095 tx_queues = scctx->isc_ntxqsets;
6097 if (ctx->ifc_sysctl_qs_eq_override == 0) {
6099 if (tx_queues != rx_queues)
6101 "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n",
6102 min(rx_queues, tx_queues), min(rx_queues, tx_queues));
6104 tx_queues = min(rx_queues, tx_queues);
6105 rx_queues = min(rx_queues, tx_queues);
6108 device_printf(dev, "Using %d rx queues %d tx queues\n",
6109 rx_queues, tx_queues);
6111 vectors = rx_queues + admincnt;
6112 if ((err = pci_alloc_msix(dev, &vectors)) == 0) {
6113 device_printf(dev, "Using MSI-X interrupts with %d vectors\n",
6115 scctx->isc_vectors = vectors;
6116 scctx->isc_nrxqsets = rx_queues;
6117 scctx->isc_ntxqsets = tx_queues;
6118 scctx->isc_intr = IFLIB_INTR_MSIX;
6123 "failed to allocate %d MSI-X vectors, err: %d - using MSI\n",
6125 bus_release_resource(dev, SYS_RES_MEMORY, bar,
6127 ctx->ifc_msix_mem = NULL;
6130 vectors = pci_msi_count(dev);
6131 scctx->isc_nrxqsets = 1;
6132 scctx->isc_ntxqsets = 1;
6133 scctx->isc_vectors = vectors;
6134 if (vectors == 1 && pci_alloc_msi(dev, &vectors) == 0) {
6135 device_printf(dev,"Using an MSI interrupt\n");
6136 scctx->isc_intr = IFLIB_INTR_MSI;
6138 scctx->isc_vectors = 1;
6139 device_printf(dev,"Using a Legacy interrupt\n");
6140 scctx->isc_intr = IFLIB_INTR_LEGACY;
6146 static const char *ring_states[] = { "IDLE", "BUSY", "STALLED", "ABDICATED" };
6149 mp_ring_state_handler(SYSCTL_HANDLER_ARGS)
6152 uint16_t *state = ((uint16_t *)oidp->oid_arg1);
6154 const char *ring_state = "UNKNOWN";
6157 rc = sysctl_wire_old_buffer(req, 0);
6161 sb = sbuf_new_for_sysctl(NULL, NULL, 80, req);
6166 ring_state = ring_states[state[3]];
6168 sbuf_printf(sb, "pidx_head: %04hd pidx_tail: %04hd cidx: %04hd state: %s",
6169 state[0], state[1], state[2], ring_state);
6170 rc = sbuf_finish(sb);
6175 enum iflib_ndesc_handler {
6181 mp_ndesc_handler(SYSCTL_HANDLER_ARGS)
6183 if_ctx_t ctx = (void *)arg1;
6184 enum iflib_ndesc_handler type = arg2;
6185 char buf[256] = {0};
6190 MPASS(type == IFLIB_NTXD_HANDLER || type == IFLIB_NRXD_HANDLER);
6194 case IFLIB_NTXD_HANDLER:
6195 ndesc = ctx->ifc_sysctl_ntxds;
6197 nqs = ctx->ifc_sctx->isc_ntxqs;
6199 case IFLIB_NRXD_HANDLER:
6200 ndesc = ctx->ifc_sysctl_nrxds;
6202 nqs = ctx->ifc_sctx->isc_nrxqs;
6205 panic("unhandled type");
6210 for (i=0; i<8; i++) {
6215 sprintf(strchr(buf, 0), "%d", ndesc[i]);
6218 rc = sysctl_handle_string(oidp, buf, sizeof(buf), req);
6219 if (rc || req->newptr == NULL)
6222 for (i = 0, next = buf, p = strsep(&next, " ,"); i < 8 && p;
6223 i++, p = strsep(&next, " ,")) {
6224 ndesc[i] = strtoul(p, NULL, 10);
6230 #define NAME_BUFLEN 32
6232 iflib_add_device_sysctl_pre(if_ctx_t ctx)
6234 device_t dev = iflib_get_dev(ctx);
6235 struct sysctl_oid_list *child, *oid_list;
6236 struct sysctl_ctx_list *ctx_list;
6237 struct sysctl_oid *node;
6239 ctx_list = device_get_sysctl_ctx(dev);
6240 child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
6241 ctx->ifc_sysctl_node = node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "iflib",
6242 CTLFLAG_RD, NULL, "IFLIB fields");
6243 oid_list = SYSCTL_CHILDREN(node);
6245 SYSCTL_ADD_CONST_STRING(ctx_list, oid_list, OID_AUTO, "driver_version",
6246 CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version,
6249 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs",
6250 CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0,
6251 "# of txqs to use, 0 => use default #");
6252 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_nrxqs",
6253 CTLFLAG_RWTUN, &ctx->ifc_sysctl_nrxqs, 0,
6254 "# of rxqs to use, 0 => use default #");
6255 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_qs_enable",
6256 CTLFLAG_RWTUN, &ctx->ifc_sysctl_qs_eq_override, 0,
6257 "permit #txq != #rxq");
6258 SYSCTL_ADD_INT(ctx_list, oid_list, OID_AUTO, "disable_msix",
6259 CTLFLAG_RWTUN, &ctx->ifc_softc_ctx.isc_disable_msix, 0,
6260 "disable MSI-X (default 0)");
6261 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "rx_budget",
6262 CTLFLAG_RWTUN, &ctx->ifc_sysctl_rx_budget, 0,
6263 "set the rx budget");
6264 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "tx_abdicate",
6265 CTLFLAG_RWTUN, &ctx->ifc_sysctl_tx_abdicate, 0,
6266 "cause tx to abdicate instead of running to completion");
6268 /* XXX change for per-queue sizes */
6269 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_ntxds",
6270 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NTXD_HANDLER,
6271 mp_ndesc_handler, "A",
6272 "list of # of tx descriptors to use, 0 = use default #");
6273 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_nrxds",
6274 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NRXD_HANDLER,
6275 mp_ndesc_handler, "A",
6276 "list of # of rx descriptors to use, 0 = use default #");
6280 iflib_add_device_sysctl_post(if_ctx_t ctx)
6282 if_shared_ctx_t sctx = ctx->ifc_sctx;
6283 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx;
6284 device_t dev = iflib_get_dev(ctx);
6285 struct sysctl_oid_list *child;
6286 struct sysctl_ctx_list *ctx_list;
6291 char namebuf[NAME_BUFLEN];
6293 struct sysctl_oid *queue_node, *fl_node, *node;
6294 struct sysctl_oid_list *queue_list, *fl_list;
6295 ctx_list = device_get_sysctl_ctx(dev);
6297 node = ctx->ifc_sysctl_node;
6298 child = SYSCTL_CHILDREN(node);
6300 if (scctx->isc_ntxqsets > 100)
6302 else if (scctx->isc_ntxqsets > 10)
6306 for (i = 0, txq = ctx->ifc_txqs; i < scctx->isc_ntxqsets; i++, txq++) {
6307 snprintf(namebuf, NAME_BUFLEN, qfmt, i);
6308 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf,
6309 CTLFLAG_RD, NULL, "Queue Name");
6310 queue_list = SYSCTL_CHILDREN(queue_node);
6312 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_dequeued",
6314 &txq->ift_dequeued, "total mbufs freed");
6315 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_enqueued",
6317 &txq->ift_enqueued, "total mbufs enqueued");
6319 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag",
6321 &txq->ift_mbuf_defrag, "# of times m_defrag was called");
6322 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "m_pullups",
6324 &txq->ift_pullups, "# of times m_pullup was called");
6325 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag_failed",
6327 &txq->ift_mbuf_defrag_failed, "# of times m_defrag failed");
6328 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_desc_avail",
6330 &txq->ift_no_desc_avail, "# of times no descriptors were available");
6331 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "tx_map_failed",
6333 &txq->ift_map_failed, "# of times dma map failed");
6334 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txd_encap_efbig",
6336 &txq->ift_txd_encap_efbig, "# of times txd_encap returned EFBIG");
6337 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_tx_dma_setup",
6339 &txq->ift_no_tx_dma_setup, "# of times map failed for other than EFBIG");
6340 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_pidx",
6342 &txq->ift_pidx, 1, "Producer Index");
6343 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx",
6345 &txq->ift_cidx, 1, "Consumer Index");
6346 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx_processed",
6348 &txq->ift_cidx_processed, 1, "Consumer Index seen by credit update");
6349 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_in_use",
6351 &txq->ift_in_use, 1, "descriptors in use");
6352 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_processed",
6354 &txq->ift_processed, "descriptors procesed for clean");
6355 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_cleaned",
6357 &txq->ift_cleaned, "total cleaned");
6358 SYSCTL_ADD_PROC(ctx_list, queue_list, OID_AUTO, "ring_state",
6359 CTLTYPE_STRING | CTLFLAG_RD, __DEVOLATILE(uint64_t *, &txq->ift_br->state),
6360 0, mp_ring_state_handler, "A", "soft ring state");
6361 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_enqueues",
6362 CTLFLAG_RD, &txq->ift_br->enqueues,
6363 "# of enqueues to the mp_ring for this queue");
6364 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_drops",
6365 CTLFLAG_RD, &txq->ift_br->drops,
6366 "# of drops in the mp_ring for this queue");
6367 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_starts",
6368 CTLFLAG_RD, &txq->ift_br->starts,
6369 "# of normal consumer starts in the mp_ring for this queue");
6370 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_stalls",
6371 CTLFLAG_RD, &txq->ift_br->stalls,
6372 "# of consumer stalls in the mp_ring for this queue");
6373 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_restarts",
6374 CTLFLAG_RD, &txq->ift_br->restarts,
6375 "# of consumer restarts in the mp_ring for this queue");
6376 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_abdications",
6377 CTLFLAG_RD, &txq->ift_br->abdications,
6378 "# of consumer abdications in the mp_ring for this queue");
6381 if (scctx->isc_nrxqsets > 100)
6383 else if (scctx->isc_nrxqsets > 10)
6387 for (i = 0, rxq = ctx->ifc_rxqs; i < scctx->isc_nrxqsets; i++, rxq++) {
6388 snprintf(namebuf, NAME_BUFLEN, qfmt, i);
6389 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf,
6390 CTLFLAG_RD, NULL, "Queue Name");
6391 queue_list = SYSCTL_CHILDREN(queue_node);
6392 if (sctx->isc_flags & IFLIB_HAS_RXCQ) {
6393 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_pidx",
6395 &rxq->ifr_cq_pidx, 1, "Producer Index");
6396 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_cidx",
6398 &rxq->ifr_cq_cidx, 1, "Consumer Index");
6401 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) {
6402 snprintf(namebuf, NAME_BUFLEN, "rxq_fl%d", j);
6403 fl_node = SYSCTL_ADD_NODE(ctx_list, queue_list, OID_AUTO, namebuf,
6404 CTLFLAG_RD, NULL, "freelist Name");
6405 fl_list = SYSCTL_CHILDREN(fl_node);
6406 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "pidx",
6408 &fl->ifl_pidx, 1, "Producer Index");
6409 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "cidx",
6411 &fl->ifl_cidx, 1, "Consumer Index");
6412 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "credits",
6414 &fl->ifl_credits, 1, "credits available");
6416 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_enqueued",
6418 &fl->ifl_m_enqueued, "mbufs allocated");
6419 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_dequeued",
6421 &fl->ifl_m_dequeued, "mbufs freed");
6422 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_enqueued",
6424 &fl->ifl_cl_enqueued, "clusters allocated");
6425 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_dequeued",
6427 &fl->ifl_cl_dequeued, "clusters freed");
6436 iflib_request_reset(if_ctx_t ctx)
6440 ctx->ifc_flags |= IFC_DO_RESET;
6444 #ifndef __NO_STRICT_ALIGNMENT
6445 static struct mbuf *
6446 iflib_fixup_rx(struct mbuf *m)
6450 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) {
6451 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len);
6452 m->m_data += ETHER_HDR_LEN;
6455 MGETHDR(n, M_NOWAIT, MT_DATA);
6460 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN);
6461 m->m_data += ETHER_HDR_LEN;
6462 m->m_len -= ETHER_HDR_LEN;
6463 n->m_len = ETHER_HDR_LEN;
6464 M_MOVE_PKTHDR(n, m);
6473 iflib_netdump_init(struct ifnet *ifp, int *nrxr, int *ncl, int *clsize)
6477 ctx = if_getsoftc(ifp);
6479 *nrxr = NRXQSETS(ctx);
6480 *ncl = ctx->ifc_rxqs[0].ifr_fl->ifl_size;
6481 *clsize = ctx->ifc_rxqs[0].ifr_fl->ifl_buf_size;
6486 iflib_netdump_event(struct ifnet *ifp, enum netdump_ev event)
6489 if_softc_ctx_t scctx;
6494 ctx = if_getsoftc(ifp);
6495 scctx = &ctx->ifc_softc_ctx;
6499 for (i = 0; i < scctx->isc_nrxqsets; i++) {
6500 rxq = &ctx->ifc_rxqs[i];
6501 for (j = 0; j < rxq->ifr_nfl; j++) {
6503 fl->ifl_zone = m_getzone(fl->ifl_buf_size);
6506 iflib_no_tx_batch = 1;
6514 iflib_netdump_transmit(struct ifnet *ifp, struct mbuf *m)
6520 ctx = if_getsoftc(ifp);
6521 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
6525 txq = &ctx->ifc_txqs[0];
6526 error = iflib_encap(txq, &m);
6528 (void)iflib_txd_db_check(ctx, txq, true, txq->ift_in_use);
6533 iflib_netdump_poll(struct ifnet *ifp, int count)
6536 if_softc_ctx_t scctx;
6540 ctx = if_getsoftc(ifp);
6541 scctx = &ctx->ifc_softc_ctx;
6543 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
6547 txq = &ctx->ifc_txqs[0];
6548 (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
6550 for (i = 0; i < scctx->isc_nrxqsets; i++)
6551 (void)iflib_rxeof(&ctx->ifc_rxqs[i], 16 /* XXX */);
6554 #endif /* NETDUMP */