2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2012, 2015 Chelsio Communications, Inc.
6 * Written by: Navdeep Parhar <np@FreeBSD.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 #include "common/t4_hw.h"
37 #include "common/t4_msg.h"
38 #include "tom/t4_tls.h"
40 #define LISTEN_HASH_SIZE 32
43 * Min receive window. We want it to be large enough to accommodate receive
44 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance.
46 #define MIN_RCV_WND (24 * 1024U)
49 * Max receive window supported by HW in bytes. Only a small part of it can
50 * be set through option0, the rest needs to be set through RX_DATA_ACK.
52 #define MAX_RCV_WND ((1U << 27) - 1)
54 #define DDP_RSVD_WIN (16 * 1024U)
55 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */
57 #define USE_DDP_RX_FLOW_CONTROL
59 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod))
60 #define PPOD_SIZE (PPOD_SZ(1))
64 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */
65 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */
66 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */
67 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */
68 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */
69 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */
70 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */
71 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */
72 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */
73 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */
74 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */
75 TPF_KTLS = (1 << 11), /* send TLS records from KTLS */
76 TPF_INITIALIZED = (1 << 12), /* init_toepcb has been called */
80 DDP_OK = (1 << 0), /* OK to turn on DDP */
81 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
82 DDP_ON = (1 << 2), /* DDP is turned on */
83 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
84 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
85 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */
86 DDP_DEAD = (1 << 6), /* toepcb is shutting down */
90 struct offload_settings;
93 * Connection parameters for an offloaded connection. These are mostly (but not
94 * all) hardware TOE parameters.
109 int16_t txq_idx; /* ofld_txq = &sc->sge.ofld_txq[txq_idx] */
110 int16_t rxq_idx; /* ofld_rxq = &sc->sge.ofld_rxq[rxq_idx] */
113 uint16_t opt0_bufsize;
114 u_int sndbuf; /* controls TP tx pages */
117 struct ofld_tx_sdesc {
118 uint32_t plen; /* payload length */
119 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
120 void *iv_buffer; /* optional buffer holding IVs for TLS */
126 u_int pr_page_shift[4];
127 uint32_t pr_tag_mask; /* hardware tagmask for this region. */
128 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */
129 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */
130 u_int pr_alias_shift; /* shift this much for first alias bit. */
134 struct ppod_reservation {
135 struct ppod_region *prsv_pr;
136 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */
141 TAILQ_ENTRY(pageset) link;
145 int offset; /* offset in first page */
147 struct ppod_reservation prsv;
153 TAILQ_HEAD(pagesetq, pageset);
155 #define PS_PPODS_WRITTEN 0x0001 /* Page pods written to the card. */
166 struct ddp_buffer db[2];
167 TAILQ_HEAD(, pageset) cached_pagesets;
168 TAILQ_HEAD(, kaiocb) aiojobq;
172 int active_id; /* the currently active DDP buffer */
173 struct task requeue_task;
174 struct kaiocb *queueing;
180 struct inpcb *inp; /* backpointer to host stack's PCB */
181 u_int flags; /* miscellaneous flags */
182 TAILQ_ENTRY(toepcb) link; /* toep_list */
185 struct vi_info *vi; /* virtual interface */
186 struct sge_wrq *ofld_txq;
187 struct sge_ofld_rxq *ofld_rxq;
188 struct sge_wrq *ctrlq;
189 struct l2t_entry *l2te; /* L2 table entry used by this connection */
190 struct clip_entry *ce; /* CLIP table entry used by this tid */
191 int tid; /* Connection identifier */
193 /* tx credit handling */
194 u_int tx_total; /* total tx WR credits (in 16B units) */
195 u_int tx_credits; /* tx WR credits (in 16B units) available */
196 u_int tx_nocompl; /* tx WR credits since last compl request */
197 u_int plen_nocompl; /* payload since last compl request */
199 struct conn_params params;
203 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */
204 struct mbufq ulp_pdu_reclaimq;
207 struct tls_ofld_info tls;
209 TAILQ_HEAD(, kaiocb) aiotx_jobq;
210 struct task aiotx_task;
211 struct socket *aiotx_so;
213 /* Tx software descriptor */
218 struct ofld_tx_sdesc txsd[];
222 ulp_mode(struct toepcb *toep)
225 return (toep->params.ulp_mode);
228 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock)
229 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock)
230 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED)
233 * Compressed state for embryonic connections for a listener.
236 struct listen_ctx *lctx; /* backpointer to listen ctx */
238 int flags; /* same as toepcb's tp_flags */
239 volatile int ok_to_respond;
240 volatile u_int refcnt;
246 __be16 tcp_opt; /* from cpl_pass_establish */
249 struct conn_params params;
252 /* listen_ctx flags */
253 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
256 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
257 volatile int refcount;
259 struct stid_region stid_region;
261 struct inpcb *inp; /* listening socket's inp */
263 struct sge_wrq *ctrlq;
264 struct sge_ofld_rxq *ofld_rxq;
265 struct clip_entry *ce;
268 /* tcb_histent flags */
269 #define TE_RPL_PENDING 1
272 /* bits in one 8b tcb_histent sample. */
273 #define TS_RTO (1 << 0)
274 #define TS_DUPACKS (1 << 1)
275 #define TS_FASTREXMT (1 << 2)
276 #define TS_SND_BACKLOGGED (1 << 3)
277 #define TS_CWND_LIMITED (1 << 4)
278 #define TS_ECN_ECE (1 << 5)
279 #define TS_ECN_CWR (1 << 6)
280 #define TS_RESERVED (1 << 7) /* Unused. */
284 struct callout te_callout;
285 uint64_t te_tcb[TCB_SIZE / sizeof(uint64_t)];
286 struct adapter *te_adapter;
290 uint8_t te_sample[100];
296 /* toepcb's associated with this TOE device */
297 struct mtx toep_list_lock;
298 TAILQ_HEAD(, toepcb) toep_list;
300 struct mtx lctx_hash_lock;
301 LIST_HEAD(, listen_ctx) *listen_hash;
303 int lctx_count; /* # of lctx in the hash table */
305 struct ppod_region pr;
307 struct rwlock tcb_history_lock __aligned(CACHE_LINE_SIZE);
308 struct tcb_histent **tcb_history;
309 int dupack_threshold;
311 /* WRs that will not be sent to the chip because L2 resolution failed */
312 struct mtx unsent_wr_lock;
313 STAILQ_HEAD(, wrqe) unsent_wr_list;
314 struct task reclaim_wr_resources;
317 static inline struct tom_data *
318 tod_td(struct toedev *tod)
321 return (__containerof(tod, struct tom_data, tod));
324 static inline struct adapter *
325 td_adapter(struct tom_data *td)
328 return (td->tod.tod_softc);
332 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode)
336 m->m_pkthdr.PH_per.eight[0] = ulp_submode;
339 static inline uint8_t
340 mbuf_ulp_submode(struct mbuf *m)
344 return (m->m_pkthdr.PH_per.eight[0]);
348 struct toepcb *alloc_toepcb(struct vi_info *, int);
349 int init_toepcb(struct vi_info *, struct toepcb *);
350 struct toepcb *hold_toepcb(struct toepcb *);
351 void free_toepcb(struct toepcb *);
352 void offload_socket(struct socket *, struct toepcb *);
353 void undo_offload_socket(struct socket *);
354 void final_cpl_received(struct toepcb *);
355 void insert_tid(struct adapter *, int, void *, int);
356 void *lookup_tid(struct adapter *, int);
357 void update_tid(struct adapter *, int, void *);
358 void remove_tid(struct adapter *, int, int);
359 u_long select_rcv_wnd(struct socket *);
360 int select_rcv_wscale(void);
361 void init_conn_params(struct vi_info *, struct offload_settings *,
362 struct in_conninfo *, struct socket *, const struct tcp_options *, int16_t,
363 struct conn_params *cp);
364 __be64 calc_options0(struct vi_info *, struct conn_params *);
365 __be32 calc_options2(struct vi_info *, struct conn_params *);
366 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *);
367 int negative_advice(int);
368 int add_tid_to_history(struct adapter *, u_int);
371 void t4_init_connect_cpl_handlers(void);
372 void t4_uninit_connect_cpl_handlers(void);
373 int t4_connect(struct toedev *, struct socket *, struct nhop_object *,
375 void act_open_failure_cleanup(struct adapter *, u_int, u_int);
378 void t4_init_listen_cpl_handlers(void);
379 void t4_uninit_listen_cpl_handlers(void);
380 int t4_listen_start(struct toedev *, struct tcpcb *);
381 int t4_listen_stop(struct toedev *, struct tcpcb *);
382 void t4_syncache_added(struct toedev *, void *);
383 void t4_syncache_removed(struct toedev *, void *);
384 int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
385 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
387 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
389 void t4_offload_socket(struct toedev *, void *, struct socket *);
390 void synack_failure_cleanup(struct adapter *, int);
393 void aiotx_init_toep(struct toepcb *);
394 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *);
395 void t4_init_cpl_io_handlers(void);
396 void t4_uninit_cpl_io_handlers(void);
397 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
398 void send_flowc_wr(struct toepcb *, struct tcpcb *);
399 void send_reset(struct adapter *, struct toepcb *, uint32_t);
400 int send_rx_credits(struct adapter *, struct toepcb *, int);
401 void send_rx_modulate(struct adapter *, struct toepcb *);
402 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
403 int t4_close_conn(struct adapter *, struct toepcb *);
404 void t4_rcvd(struct toedev *, struct tcpcb *);
405 void t4_rcvd_locked(struct toedev *, struct tcpcb *);
406 int t4_tod_output(struct toedev *, struct tcpcb *);
407 int t4_send_fin(struct toedev *, struct tcpcb *);
408 int t4_send_rst(struct toedev *, struct tcpcb *);
409 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *,
410 uint16_t, uint64_t, uint64_t, int, int);
411 void t4_push_frames(struct adapter *, struct toepcb *, int);
412 void t4_push_pdus(struct adapter *, struct toepcb *, int);
415 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int,
417 void t4_free_ppod_region(struct ppod_region *);
418 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *);
419 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int,
420 struct ppod_reservation *);
421 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int,
423 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid,
424 struct ppod_reservation *, vm_offset_t, int);
425 void t4_free_page_pods(struct ppod_reservation *);
426 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
427 struct mbuf **, struct mbuf **, int *);
428 int t4_aio_queue_ddp(struct socket *, struct kaiocb *);
429 void t4_ddp_mod_load(void);
430 void t4_ddp_mod_unload(void);
431 void ddp_assert_empty(struct toepcb *);
432 void ddp_init_toep(struct toepcb *);
433 void ddp_uninit_toep(struct toepcb *);
434 void ddp_queue_toep(struct toepcb *);
435 void release_ddp_resources(struct toepcb *toep);
436 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t);
437 void handle_ddp_indicate(struct toepcb *);
438 void insert_ddp_data(struct toepcb *, uint32_t);
439 const struct offload_settings *lookup_offload_policy(struct adapter *, int,
440 struct mbuf *, uint16_t, struct inpcb *);
443 bool can_tls_offload(struct adapter *);
444 int t4_ctloutput_tls(struct socket *, struct sockopt *);
445 void t4_push_tls_records(struct adapter *, struct toepcb *, int);
446 void t4_push_ktls(struct adapter *, struct toepcb *, int);
447 void t4_tls_mod_load(void);
448 void t4_tls_mod_unload(void);
449 void tls_establish(struct toepcb *);
450 void tls_init_toep(struct toepcb *);
451 int tls_rx_key(struct toepcb *);
452 void tls_stop_handshake_timer(struct toepcb *);
453 int tls_tx_key(struct toepcb *);
454 void tls_uninit_toep(struct toepcb *);
455 int tls_alloc_ktls(struct toepcb *, struct ktls_session *);