1 /*-
2 * Copyright (c) 2012, 2015 Chelsio Communications, Inc.
3 * All rights reserved.
4 * Written by: Navdeep Parhar <np@FreeBSD.org>
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * 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 AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 *
27 * $FreeBSD: stable/10/sys/dev/cxgbe/tom/t4_tom.h 318804 2017-05-24 20:01:12Z np $
28 *
29 */
30
31 #ifndef __T4_TOM_H__
32 #define __T4_TOM_H__
33 #include <sys/vmem.h>
34
35 /*
36 * Inline version of mbufq for use on 10.x. Borrowed from
37 * sys/cam/ctl/ctl_ha.c.
38 */
39 struct mbufq {
40 struct mbuf *head;
41 struct mbuf *tail;
42 };
43
44 #define LISTEN_HASH_SIZE 32
45
46 /*
47 * Min receive window. We want it to be large enough to accommodate receive
48 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance.
49 */
50 #define MIN_RCV_WND (24 * 1024U)
51
52 /*
53 * Max receive window supported by HW in bytes. Only a small part of it can
54 * be set through option0, the rest needs to be set through RX_DATA_ACK.
55 */
56 #define MAX_RCV_WND ((1U << 27) - 1)
57
58 #define DDP_RSVD_WIN (16 * 1024U)
59 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */
60
61 #define USE_DDP_RX_FLOW_CONTROL
62
63 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod))
64 #define PPOD_SIZE (PPOD_SZ(1))
65
66 /* TOE PCB flags */
67 enum {
68 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */
69 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */
70 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */
71 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */
72 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */
73 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */
74 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */
75 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */
76 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */
77 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */
78 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */
79 TPF_SYNQE_EXPANDED = (1 << 11), /* toepcb ready, tid context updated */
80 TPF_SYNQE_HAS_L2TE = (1 << 12), /* we've replied to PASS_ACCEPT_REQ */
81 };
82
83 enum {
84 DDP_OK = (1 << 0), /* OK to turn on DDP */
85 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
86 DDP_ON = (1 << 2), /* DDP is turned on */
87 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
88 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
89 };
90
91 struct ofld_tx_sdesc {
92 uint32_t plen; /* payload length */
93 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
94 };
95
96 struct ppod_region {
97 u_int pr_start;
98 u_int pr_len;
99 u_int pr_page_shift[4];
100 uint32_t pr_tag_mask; /* hardware tagmask for this region. */
101 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */
102 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */
103 u_int pr_alias_shift; /* shift this much for first alias bit. */
104 vmem_t *pr_arena;
105 };
106
107 struct ppod_reservation {
108 struct ppod_region *prsv_pr;
109 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */
110 u_int prsv_nppods;
111 };
112
113 struct ddp_buffer {
114 int offset;
115 int len;
116 int npages;
117 vm_page_t *pages;
118 struct ppod_reservation prsv;
119 };
120
121 struct toepcb {
122 TAILQ_ENTRY(toepcb) link; /* toep_list */
123 u_int flags; /* miscellaneous flags */
124 struct tom_data *td;
125 struct inpcb *inp; /* backpointer to host stack's PCB */
126 struct vnet *vnet;
127 struct vi_info *vi; /* virtual interface */
128 struct sge_wrq *ofld_txq;
129 struct sge_ofld_rxq *ofld_rxq;
130 struct sge_wrq *ctrlq;
131 struct l2t_entry *l2te; /* L2 table entry used by this connection */
132 struct clip_entry *ce; /* CLIP table entry used by this tid */
133 int tid; /* Connection identifier */
134
135 /* tx credit handling */
136 u_int tx_total; /* total tx WR credits (in 16B units) */
137 u_int tx_credits; /* tx WR credits (in 16B units) available */
138 u_int tx_nocompl; /* tx WR credits since last compl request */
139 u_int plen_nocompl; /* payload since last compl request */
140
141 /* rx credit handling */
142 u_int sb_cc; /* last noted value of so_rcv->sb_cc */
143 int rx_credits; /* rx credits (in bytes) to be returned to hw */
144
145 u_int ulp_mode; /* ULP mode */
146 void *ulpcb;
147 void *ulpcb2;
148 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */
149 struct mbufq ulp_pdu_reclaimq;
150
151 u_int ddp_flags;
152 struct ddp_buffer *db[2];
153 time_t ddp_disabled;
154 uint8_t ddp_score;
155
156 /* Tx software descriptor */
157 uint8_t txsd_total;
158 uint8_t txsd_pidx;
159 uint8_t txsd_cidx;
160 uint8_t txsd_avail;
161 struct ofld_tx_sdesc txsd[];
162 };
163
164 struct flowc_tx_params {
165 uint32_t snd_nxt;
166 uint32_t rcv_nxt;
167 unsigned int snd_space;
168 unsigned int mss;
169 };
170
171 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */
172 #define DDP_LOW_SCORE 1
173 #define DDP_HIGH_SCORE 3
174
175 /*
176 * Compressed state for embryonic connections for a listener. Barely fits in
177 * 64B, try not to grow it further.
178 */
179 struct synq_entry {
180 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */
181 int flags; /* same as toepcb's tp_flags */
182 int tid;
183 struct listen_ctx *lctx; /* backpointer to listen ctx */
184 struct mbuf *syn;
185 uint32_t iss;
186 uint32_t ts;
187 volatile uintptr_t wr;
188 volatile u_int refcnt;
189 uint16_t l2e_idx;
190 uint16_t rcv_bufsize;
191 };
192
193 /* listen_ctx flags */
194 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
195
196 struct listen_ctx {
197 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
198 volatile int refcount;
199 int stid;
200 struct stid_region stid_region;
201 int flags;
202 struct inpcb *inp; /* listening socket's inp */
203 struct vnet *vnet;
204 struct sge_wrq *ctrlq;
205 struct sge_ofld_rxq *ofld_rxq;
206 struct clip_entry *ce;
207 TAILQ_HEAD(, synq_entry) synq;
208 };
209
210 struct clip_entry {
211 TAILQ_ENTRY(clip_entry) link;
212 struct in6_addr lip; /* local IPv6 address */
213 u_int refcount;
214 };
215
216 TAILQ_HEAD(clip_head, clip_entry);
217 struct tom_data {
218 struct toedev tod;
219
220 /* toepcb's associated with this TOE device */
221 struct mtx toep_list_lock;
222 TAILQ_HEAD(, toepcb) toep_list;
223
224 struct mtx lctx_hash_lock;
225 LIST_HEAD(, listen_ctx) *listen_hash;
226 u_long listen_mask;
227 int lctx_count; /* # of lctx in the hash table */
228
229 struct ppod_region pr;
230
231 struct mtx clip_table_lock;
232 struct clip_head clip_table;
233 int clip_gen;
234
235 /* WRs that will not be sent to the chip because L2 resolution failed */
236 struct mtx unsent_wr_lock;
237 STAILQ_HEAD(, wrqe) unsent_wr_list;
238 struct task reclaim_wr_resources;
239 };
240
241 static inline struct tom_data *
tod_td(struct toedev * tod)242 tod_td(struct toedev *tod)
243 {
244
245 return (__containerof(tod, struct tom_data, tod));
246 }
247
248 static inline struct adapter *
td_adapter(struct tom_data * td)249 td_adapter(struct tom_data *td)
250 {
251
252 return (td->tod.tod_softc);
253 }
254
255 /*
256 * XXX: Don't define these for the iWARP driver on 10 due to differences
257 * in LinuxKPI.
258 */
259 #ifndef _LINUX_TYPES_H_
260 static inline void
set_mbuf_ulp_submode(struct mbuf * m,uint8_t ulp_submode)261 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode)
262 {
263
264 M_ASSERTPKTHDR(m);
265 m->m_pkthdr.PH_per.eigth[0] = ulp_submode;
266 }
267
268 static inline uint8_t
mbuf_ulp_submode(struct mbuf * m)269 mbuf_ulp_submode(struct mbuf *m)
270 {
271
272 M_ASSERTPKTHDR(m);
273 return (m->m_pkthdr.PH_per.eigth[0]);
274 }
275 #endif
276
277 /* t4_tom.c */
278 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int);
279 void free_toepcb(struct toepcb *);
280 void offload_socket(struct socket *, struct toepcb *);
281 void undo_offload_socket(struct socket *);
282 void final_cpl_received(struct toepcb *);
283 void insert_tid(struct adapter *, int, void *, int);
284 void *lookup_tid(struct adapter *, int);
285 void update_tid(struct adapter *, int, void *);
286 void remove_tid(struct adapter *, int, int);
287 void release_tid(struct adapter *, int, struct sge_wrq *);
288 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int);
289 u_long select_rcv_wnd(struct socket *);
290 int select_rcv_wscale(void);
291 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *,
292 int, int, int, int);
293 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *);
294 void set_tcpddp_ulp_mode(struct toepcb *);
295 int negative_advice(int);
296 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *,
297 struct clip_entry *);
298 void release_lip(struct tom_data *, struct clip_entry *);
299
300 /* t4_connect.c */
301 void t4_init_connect_cpl_handlers(void);
302 void t4_uninit_connect_cpl_handlers(void);
303 int t4_connect(struct toedev *, struct socket *, struct rtentry *,
304 struct sockaddr *);
305 void act_open_failure_cleanup(struct adapter *, u_int, u_int);
306
307 /* t4_listen.c */
308 void t4_init_listen_cpl_handlers(void);
309 void t4_uninit_listen_cpl_handlers(void);
310 int t4_listen_start(struct toedev *, struct tcpcb *);
311 int t4_listen_stop(struct toedev *, struct tcpcb *);
312 void t4_syncache_added(struct toedev *, void *);
313 void t4_syncache_removed(struct toedev *, void *);
314 int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
315 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
316 struct mbuf *);
317 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
318 struct mbuf *);
319 void t4_offload_socket(struct toedev *, void *, struct socket *);
320
321 /* t4_cpl_io.c */
322 void t4_init_cpl_io_handlers(void);
323 void t4_uninit_cpl_io_handlers(void);
324 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
325 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *);
326 void send_reset(struct adapter *, struct toepcb *, uint32_t);
327 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
328 void t4_rcvd(struct toedev *, struct tcpcb *);
329 int t4_tod_output(struct toedev *, struct tcpcb *);
330 int t4_send_fin(struct toedev *, struct tcpcb *);
331 int t4_send_rst(struct toedev *, struct tcpcb *);
332 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, int, uint16_t,
333 uint64_t, uint64_t, int, int, int);
334 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop);
335 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop);
336 int do_set_tcb_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *);
337
338 /* t4_ddp.c */
339 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int,
340 const char *);
341 void t4_free_ppod_region(struct ppod_region *);
342 int t4_alloc_page_pods_for_db(struct ppod_region *, struct ddp_buffer *);
343 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int,
344 struct ppod_reservation *);
345 int t4_write_page_pods_for_db(struct adapter *, struct sge_wrq *, int,
346 struct ddp_buffer *);
347 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid,
348 struct ppod_reservation *, vm_offset_t, int);
349 void t4_free_page_pods(struct ppod_reservation *);
350 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
351 struct mbuf **, struct mbuf **, int *);
352 int t4_ddp_mod_load(void);
353 void t4_ddp_mod_unload(void);
354 void enable_ddp(struct adapter *, struct toepcb *toep);
355 void release_ddp_resources(struct toepcb *toep);
356 void handle_ddp_close(struct toepcb *, struct tcpcb *, struct sockbuf *,
357 uint32_t);
358 void insert_ddp_data(struct toepcb *, uint32_t);
359
360 #endif
361