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 */
78 DDP_OK = (1 << 0), /* OK to turn on DDP */
79 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
80 DDP_ON = (1 << 2), /* DDP is turned on */
81 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
82 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
83 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */
84 DDP_DEAD = (1 << 6), /* toepcb is shutting down */
88 struct offload_settings;
91 * Connection parameters for an offloaded connection. These are mostly (but not
92 * all) hardware TOE parameters.
107 int16_t txq_idx; /* ofld_txq = &sc->sge.ofld_txq[txq_idx] */
108 int16_t rxq_idx; /* ofld_rxq = &sc->sge.ofld_rxq[rxq_idx] */
111 uint16_t opt0_bufsize;
112 u_int sndbuf; /* controls TP tx pages */
115 struct ofld_tx_sdesc {
116 uint32_t plen; /* payload length */
117 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
118 void *iv_buffer; /* optional buffer holding IVs for TLS */
124 u_int pr_page_shift[4];
125 uint32_t pr_tag_mask; /* hardware tagmask for this region. */
126 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */
127 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */
128 u_int pr_alias_shift; /* shift this much for first alias bit. */
132 struct ppod_reservation {
133 struct ppod_region *prsv_pr;
134 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */
139 TAILQ_ENTRY(pageset) link;
143 int offset; /* offset in first page */
145 struct ppod_reservation prsv;
151 TAILQ_HEAD(pagesetq, pageset);
153 #define PS_PPODS_WRITTEN 0x0001 /* Page pods written to the card. */
164 struct ddp_buffer db[2];
165 TAILQ_HEAD(, pageset) cached_pagesets;
166 TAILQ_HEAD(, kaiocb) aiojobq;
170 int active_id; /* the currently active DDP buffer */
171 struct task requeue_task;
172 struct kaiocb *queueing;
177 TAILQ_ENTRY(toepcb) link; /* toep_list */
178 u_int flags; /* miscellaneous flags */
181 struct inpcb *inp; /* backpointer to host stack's PCB */
183 struct vi_info *vi; /* virtual interface */
184 struct sge_wrq *ofld_txq;
185 struct sge_ofld_rxq *ofld_rxq;
186 struct sge_wrq *ctrlq;
187 struct l2t_entry *l2te; /* L2 table entry used by this connection */
188 struct clip_entry *ce; /* CLIP table entry used by this tid */
189 int tid; /* Connection identifier */
191 /* tx credit handling */
192 u_int tx_total; /* total tx WR credits (in 16B units) */
193 u_int tx_credits; /* tx WR credits (in 16B units) available */
194 u_int tx_nocompl; /* tx WR credits since last compl request */
195 u_int plen_nocompl; /* payload since last compl request */
197 struct conn_params params;
201 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */
202 struct mbufq ulp_pdu_reclaimq;
205 struct tls_ofld_info tls;
207 TAILQ_HEAD(, kaiocb) aiotx_jobq;
208 struct task aiotx_task;
209 struct socket *aiotx_so;
211 /* Tx software descriptor */
216 struct ofld_tx_sdesc txsd[];
220 ulp_mode(struct toepcb *toep)
223 return (toep->params.ulp_mode);
226 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock)
227 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock)
228 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED)
231 * Compressed state for embryonic connections for a listener.
234 struct listen_ctx *lctx; /* backpointer to listen ctx */
236 int flags; /* same as toepcb's tp_flags */
237 volatile int ok_to_respond;
238 volatile u_int refcnt;
243 __be16 tcp_opt; /* from cpl_pass_establish */
246 struct conn_params params;
249 /* listen_ctx flags */
250 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
253 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
254 volatile int refcount;
256 struct stid_region stid_region;
258 struct inpcb *inp; /* listening socket's inp */
260 struct sge_wrq *ctrlq;
261 struct sge_ofld_rxq *ofld_rxq;
262 struct clip_entry *ce;
265 /* tcb_histent flags */
266 #define TE_RPL_PENDING 1
269 /* bits in one 8b tcb_histent sample. */
270 #define TS_RTO (1 << 0)
271 #define TS_DUPACKS (1 << 1)
272 #define TS_FASTREXMT (1 << 2)
273 #define TS_SND_BACKLOGGED (1 << 3)
274 #define TS_CWND_LIMITED (1 << 4)
275 #define TS_ECN_ECE (1 << 5)
276 #define TS_ECN_CWR (1 << 6)
277 #define TS_RESERVED (1 << 7) /* Unused. */
281 struct callout te_callout;
282 uint64_t te_tcb[TCB_SIZE / sizeof(uint64_t)];
283 struct adapter *te_adapter;
287 uint8_t te_sample[100];
293 /* toepcb's associated with this TOE device */
294 struct mtx toep_list_lock;
295 TAILQ_HEAD(, toepcb) toep_list;
297 struct mtx lctx_hash_lock;
298 LIST_HEAD(, listen_ctx) *listen_hash;
300 int lctx_count; /* # of lctx in the hash table */
302 struct ppod_region pr;
304 struct rwlock tcb_history_lock __aligned(CACHE_LINE_SIZE);
305 struct tcb_histent **tcb_history;
306 int dupack_threshold;
308 /* WRs that will not be sent to the chip because L2 resolution failed */
309 struct mtx unsent_wr_lock;
310 STAILQ_HEAD(, wrqe) unsent_wr_list;
311 struct task reclaim_wr_resources;
314 static inline struct tom_data *
315 tod_td(struct toedev *tod)
318 return (__containerof(tod, struct tom_data, tod));
321 static inline struct adapter *
322 td_adapter(struct tom_data *td)
325 return (td->tod.tod_softc);
329 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode)
333 m->m_pkthdr.PH_per.eight[0] = ulp_submode;
336 static inline uint8_t
337 mbuf_ulp_submode(struct mbuf *m)
341 return (m->m_pkthdr.PH_per.eight[0]);
345 struct toepcb *alloc_toepcb(struct vi_info *, int);
346 int init_toepcb(struct vi_info *, struct toepcb *);
347 struct toepcb *hold_toepcb(struct toepcb *);
348 void free_toepcb(struct toepcb *);
349 void offload_socket(struct socket *, struct toepcb *);
350 void undo_offload_socket(struct socket *);
351 void final_cpl_received(struct toepcb *);
352 void insert_tid(struct adapter *, int, void *, int);
353 void *lookup_tid(struct adapter *, int);
354 void update_tid(struct adapter *, int, void *);
355 void remove_tid(struct adapter *, int, int);
356 u_long select_rcv_wnd(struct socket *);
357 int select_rcv_wscale(void);
358 void init_conn_params(struct vi_info *, struct offload_settings *,
359 struct in_conninfo *, struct socket *, const struct tcp_options *, int16_t,
360 struct conn_params *cp);
361 __be64 calc_options0(struct vi_info *, struct conn_params *);
362 __be32 calc_options2(struct vi_info *, struct conn_params *);
363 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *);
364 int negative_advice(int);
365 int add_tid_to_history(struct adapter *, u_int);
368 void t4_init_connect_cpl_handlers(void);
369 void t4_uninit_connect_cpl_handlers(void);
370 int t4_connect(struct toedev *, struct socket *, struct rtentry *,
372 void act_open_failure_cleanup(struct adapter *, u_int, u_int);
375 void t4_init_listen_cpl_handlers(void);
376 void t4_uninit_listen_cpl_handlers(void);
377 int t4_listen_start(struct toedev *, struct tcpcb *);
378 int t4_listen_stop(struct toedev *, struct tcpcb *);
379 void t4_syncache_added(struct toedev *, void *);
380 void t4_syncache_removed(struct toedev *, void *);
381 int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
382 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
384 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
386 void t4_offload_socket(struct toedev *, void *, struct socket *);
387 void synack_failure_cleanup(struct adapter *, int);
390 void aiotx_init_toep(struct toepcb *);
391 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *);
392 void t4_init_cpl_io_handlers(void);
393 void t4_uninit_cpl_io_handlers(void);
394 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
395 void send_flowc_wr(struct toepcb *, struct tcpcb *);
396 void send_reset(struct adapter *, struct toepcb *, uint32_t);
397 int send_rx_credits(struct adapter *, struct toepcb *, int);
398 void send_rx_modulate(struct adapter *, struct toepcb *);
399 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
400 int t4_close_conn(struct adapter *, struct toepcb *);
401 void t4_rcvd(struct toedev *, struct tcpcb *);
402 void t4_rcvd_locked(struct toedev *, struct tcpcb *);
403 int t4_tod_output(struct toedev *, struct tcpcb *);
404 int t4_send_fin(struct toedev *, struct tcpcb *);
405 int t4_send_rst(struct toedev *, struct tcpcb *);
406 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *,
407 uint16_t, uint64_t, uint64_t, int, int);
408 void t4_push_frames(struct adapter *, struct toepcb *, int);
409 void t4_push_pdus(struct adapter *, struct toepcb *, int);
412 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int,
414 void t4_free_ppod_region(struct ppod_region *);
415 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *);
416 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int,
417 struct ppod_reservation *);
418 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int,
420 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid,
421 struct ppod_reservation *, vm_offset_t, int);
422 void t4_free_page_pods(struct ppod_reservation *);
423 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
424 struct mbuf **, struct mbuf **, int *);
425 int t4_aio_queue_ddp(struct socket *, struct kaiocb *);
426 void t4_ddp_mod_load(void);
427 void t4_ddp_mod_unload(void);
428 void ddp_assert_empty(struct toepcb *);
429 void ddp_init_toep(struct toepcb *);
430 void ddp_uninit_toep(struct toepcb *);
431 void ddp_queue_toep(struct toepcb *);
432 void release_ddp_resources(struct toepcb *toep);
433 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t);
434 void handle_ddp_indicate(struct toepcb *);
435 void insert_ddp_data(struct toepcb *, uint32_t);
436 const struct offload_settings *lookup_offload_policy(struct adapter *, int,
437 struct mbuf *, uint16_t, struct inpcb *);
440 bool can_tls_offload(struct adapter *);
441 int t4_ctloutput_tls(struct socket *, struct sockopt *);
442 void t4_push_tls_records(struct adapter *, struct toepcb *, int);
443 void t4_tls_mod_load(void);
444 void t4_tls_mod_unload(void);
445 void tls_establish(struct toepcb *);
446 void tls_init_toep(struct toepcb *);
447 int tls_rx_key(struct toepcb *);
448 void tls_stop_handshake_timer(struct toepcb *);
449 int tls_tx_key(struct toepcb *);
450 void tls_uninit_toep(struct toepcb *);