1/*-
2 * Copyright (c) 2012 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 *
| 1/*-
2 * Copyright (c) 2012 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 *
|
28 *
29 */
30
31#ifndef __T4_TOM_H__
32#define __T4_TOM_H__
33
34#define KTR_CXGBE KTR_SPARE3
35#define LISTEN_HASH_SIZE 32
36
37/*
38 * Min receive window. We want it to be large enough to accommodate receive
39 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance.
40 */
41#define MIN_RCV_WND (24 * 1024U)
42
43/*
44 * Max receive window supported by HW in bytes. Only a small part of it can
45 * be set through option0, the rest needs to be set through RX_DATA_ACK.
46 */
47#define MAX_RCV_WND ((1U << 27) - 1)
48
49#define DDP_RSVD_WIN (16 * 1024U)
50#define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */
51
52#define M_DDP M_PROTO1
53
54#define USE_DDP_RX_FLOW_CONTROL
55
56/* TOE PCB flags */
57enum {
58 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */
59 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */
60 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */
61 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */
62 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */
63 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */
64 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */
65 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */
66 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */
67 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */
68 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */
| 28 *
29 */
30
31#ifndef __T4_TOM_H__
32#define __T4_TOM_H__
33
34#define KTR_CXGBE KTR_SPARE3
35#define LISTEN_HASH_SIZE 32
36
37/*
38 * Min receive window. We want it to be large enough to accommodate receive
39 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance.
40 */
41#define MIN_RCV_WND (24 * 1024U)
42
43/*
44 * Max receive window supported by HW in bytes. Only a small part of it can
45 * be set through option0, the rest needs to be set through RX_DATA_ACK.
46 */
47#define MAX_RCV_WND ((1U << 27) - 1)
48
49#define DDP_RSVD_WIN (16 * 1024U)
50#define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */
51
52#define M_DDP M_PROTO1
53
54#define USE_DDP_RX_FLOW_CONTROL
55
56/* TOE PCB flags */
57enum {
58 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */
59 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */
60 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */
61 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */
62 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */
63 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */
64 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */
65 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */
66 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */
67 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */
68 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */
|
69};
70
71enum {
72 DDP_OK = (1 << 0), /* OK to turn on DDP */
73 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
74 DDP_ON = (1 << 2), /* DDP is turned on */
75 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
76 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
77};
78
79struct ofld_tx_sdesc {
80 uint32_t plen; /* payload length */
81 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
82};
83
84struct ppod_region {
85 TAILQ_ENTRY(ppod_region) link;
86 int used; /* # of pods used by this region */
87 int free; /* # of contiguous pods free right after this region */
88};
89
90struct ddp_buffer {
91 uint32_t tag; /* includes color, page pod addr, and DDP page size */
92 int nppods;
93 int offset;
94 int len;
95 struct ppod_region ppod_region;
96 int npages;
97 vm_page_t *pages;
98};
99
100struct toepcb {
101 TAILQ_ENTRY(toepcb) link; /* toep_list */
102 unsigned int flags; /* miscellaneous flags */
103 struct tom_data *td;
104 struct inpcb *inp; /* backpointer to host stack's PCB */
105 struct port_info *port; /* physical port */
106 struct sge_wrq *ofld_txq;
107 struct sge_ofld_rxq *ofld_rxq;
108 struct sge_wrq *ctrlq;
109 struct l2t_entry *l2te; /* L2 table entry used by this connection */
110 int tid; /* Connection identifier */
111 unsigned int tx_credits;/* tx WR credits (in 16 byte units) remaining */
112 unsigned int sb_cc; /* last noted value of so_rcv->sb_cc */
113 int rx_credits; /* rx credits (in bytes) to be returned to hw */
114
115 unsigned int ulp_mode; /* ULP mode */
116
117 unsigned int ddp_flags;
118 struct ddp_buffer *db[2];
119 time_t ddp_disabled;
120 uint8_t ddp_score;
121
122 /* Tx software descriptor */
123 uint8_t txsd_total;
124 uint8_t txsd_pidx;
125 uint8_t txsd_cidx;
126 uint8_t txsd_avail;
127 struct ofld_tx_sdesc txsd[];
128};
129
130struct flowc_tx_params {
131 uint32_t snd_nxt;
132 uint32_t rcv_nxt;
133 unsigned int snd_space;
134 unsigned int mss;
135};
136
137#define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */
138#define DDP_LOW_SCORE 1
139#define DDP_HIGH_SCORE 3
140
141static inline void
142set_tcpddp_ulp_mode(struct toepcb *toep)
143{
144
145 toep->ulp_mode = ULP_MODE_TCPDDP;
146 toep->ddp_flags = DDP_OK;
147 toep->ddp_score = DDP_LOW_SCORE;
148}
149
150/*
151 * Compressed state for embryonic connections for a listener. Barely fits in
152 * 64B, try not to grow it further.
153 */
154struct synq_entry {
155 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */
156 int flags; /* same as toepcb's tp_flags */
157 int tid;
158 struct listen_ctx *lctx; /* backpointer to listen ctx */
159 struct mbuf *syn;
160 uint32_t iss;
161 uint32_t ts;
162 volatile uintptr_t wr;
163 volatile u_int refcnt;
164 uint16_t l2e_idx;
165 uint16_t rcv_bufsize;
166};
167
168/* listen_ctx flags */
169#define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
170
171struct listen_ctx {
172 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
173 volatile int refcount;
174 int stid;
175 int flags;
176 struct inpcb *inp; /* listening socket's inp */
177 struct sge_wrq *ctrlq;
178 struct sge_ofld_rxq *ofld_rxq;
179 TAILQ_HEAD(, synq_entry) synq;
180};
181
182TAILQ_HEAD(ppod_head, ppod_region);
183
184struct tom_data {
185 struct toedev tod;
186
187 /* toepcb's associated with this TOE device */
188 struct mtx toep_list_lock;
189 TAILQ_HEAD(, toepcb) toep_list;
190
191 struct mtx lctx_hash_lock;
192 LIST_HEAD(, listen_ctx) *listen_hash;
193 u_long listen_mask;
194 int lctx_count; /* # of lctx in the hash table */
195
196 struct mtx ppod_lock;
197 int nppods;
198 int nppods_free; /* # of available ppods */
199 int nppods_free_head; /* # of available ppods at the begining */
200 struct ppod_head ppods;
201};
202
203static inline struct tom_data *
204tod_td(struct toedev *tod)
205{
206
207 return (member2struct(tom_data, tod, tod));
208}
209
210static inline struct adapter *
211td_adapter(struct tom_data *td)
212{
213
214 return (td->tod.tod_softc);
215}
216
217/* t4_tom.c */
218struct toepcb *alloc_toepcb(struct port_info *, int, int, int);
219void free_toepcb(struct toepcb *);
220void offload_socket(struct socket *, struct toepcb *);
221void undo_offload_socket(struct socket *);
222void final_cpl_received(struct toepcb *);
223void insert_tid(struct adapter *, int, void *);
224void *lookup_tid(struct adapter *, int);
225void update_tid(struct adapter *, int, void *);
226void remove_tid(struct adapter *, int);
227void release_tid(struct adapter *, int, struct sge_wrq *);
228int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int);
229u_long select_rcv_wnd(struct socket *);
230int select_rcv_wscale(void);
231uint64_t calc_opt0(struct socket *, struct port_info *, struct l2t_entry *,
232 int, int, int, int);
233uint32_t select_ntuple(struct port_info *, struct l2t_entry *, uint32_t);
234
235/* t4_connect.c */
236void t4_init_connect_cpl_handlers(struct adapter *);
237int t4_connect(struct toedev *, struct socket *, struct rtentry *,
238 struct sockaddr *);
239
240/* t4_listen.c */
241void t4_init_listen_cpl_handlers(struct adapter *);
242int t4_listen_start(struct toedev *, struct tcpcb *);
243int t4_listen_stop(struct toedev *, struct tcpcb *);
244void t4_syncache_added(struct toedev *, void *);
245void t4_syncache_removed(struct toedev *, void *);
246int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
247int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
248 struct mbuf *);
249int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
250 struct mbuf *);
251void t4_offload_socket(struct toedev *, void *, struct socket *);
252
253/* t4_cpl_io.c */
254void t4_init_cpl_io_handlers(struct adapter *);
255void t4_uninit_cpl_io_handlers(struct adapter *);
256void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
257void send_flowc_wr(struct toepcb *, struct flowc_tx_params *);
258void send_reset(struct adapter *, struct toepcb *, uint32_t);
259void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
260void t4_rcvd(struct toedev *, struct tcpcb *);
261int t4_tod_output(struct toedev *, struct tcpcb *);
262int t4_send_fin(struct toedev *, struct tcpcb *);
263int t4_send_rst(struct toedev *, struct tcpcb *);
264void t4_set_tcb_field(struct adapter *, struct toepcb *, uint16_t, uint64_t,
265 uint64_t);
266
267/* t4_ddp.c */
268void t4_init_ddp(struct adapter *, struct tom_data *);
269void t4_uninit_ddp(struct adapter *, struct tom_data *);
270int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
271 struct mbuf **, struct mbuf **, int *);
272void enable_ddp(struct adapter *, struct toepcb *toep);
273void release_ddp_resources(struct toepcb *toep);
274#endif
| 70};
71
72enum {
73 DDP_OK = (1 << 0), /* OK to turn on DDP */
74 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
75 DDP_ON = (1 << 2), /* DDP is turned on */
76 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
77 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
78};
79
80struct ofld_tx_sdesc {
81 uint32_t plen; /* payload length */
82 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
83};
84
85struct ppod_region {
86 TAILQ_ENTRY(ppod_region) link;
87 int used; /* # of pods used by this region */
88 int free; /* # of contiguous pods free right after this region */
89};
90
91struct ddp_buffer {
92 uint32_t tag; /* includes color, page pod addr, and DDP page size */
93 int nppods;
94 int offset;
95 int len;
96 struct ppod_region ppod_region;
97 int npages;
98 vm_page_t *pages;
99};
100
101struct toepcb {
102 TAILQ_ENTRY(toepcb) link; /* toep_list */
103 unsigned int flags; /* miscellaneous flags */
104 struct tom_data *td;
105 struct inpcb *inp; /* backpointer to host stack's PCB */
106 struct port_info *port; /* physical port */
107 struct sge_wrq *ofld_txq;
108 struct sge_ofld_rxq *ofld_rxq;
109 struct sge_wrq *ctrlq;
110 struct l2t_entry *l2te; /* L2 table entry used by this connection */
111 int tid; /* Connection identifier */
112 unsigned int tx_credits;/* tx WR credits (in 16 byte units) remaining */
113 unsigned int sb_cc; /* last noted value of so_rcv->sb_cc */
114 int rx_credits; /* rx credits (in bytes) to be returned to hw */
115
116 unsigned int ulp_mode; /* ULP mode */
117
118 unsigned int ddp_flags;
119 struct ddp_buffer *db[2];
120 time_t ddp_disabled;
121 uint8_t ddp_score;
122
123 /* Tx software descriptor */
124 uint8_t txsd_total;
125 uint8_t txsd_pidx;
126 uint8_t txsd_cidx;
127 uint8_t txsd_avail;
128 struct ofld_tx_sdesc txsd[];
129};
130
131struct flowc_tx_params {
132 uint32_t snd_nxt;
133 uint32_t rcv_nxt;
134 unsigned int snd_space;
135 unsigned int mss;
136};
137
138#define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */
139#define DDP_LOW_SCORE 1
140#define DDP_HIGH_SCORE 3
141
142static inline void
143set_tcpddp_ulp_mode(struct toepcb *toep)
144{
145
146 toep->ulp_mode = ULP_MODE_TCPDDP;
147 toep->ddp_flags = DDP_OK;
148 toep->ddp_score = DDP_LOW_SCORE;
149}
150
151/*
152 * Compressed state for embryonic connections for a listener. Barely fits in
153 * 64B, try not to grow it further.
154 */
155struct synq_entry {
156 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */
157 int flags; /* same as toepcb's tp_flags */
158 int tid;
159 struct listen_ctx *lctx; /* backpointer to listen ctx */
160 struct mbuf *syn;
161 uint32_t iss;
162 uint32_t ts;
163 volatile uintptr_t wr;
164 volatile u_int refcnt;
165 uint16_t l2e_idx;
166 uint16_t rcv_bufsize;
167};
168
169/* listen_ctx flags */
170#define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
171
172struct listen_ctx {
173 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
174 volatile int refcount;
175 int stid;
176 int flags;
177 struct inpcb *inp; /* listening socket's inp */
178 struct sge_wrq *ctrlq;
179 struct sge_ofld_rxq *ofld_rxq;
180 TAILQ_HEAD(, synq_entry) synq;
181};
182
183TAILQ_HEAD(ppod_head, ppod_region);
184
185struct tom_data {
186 struct toedev tod;
187
188 /* toepcb's associated with this TOE device */
189 struct mtx toep_list_lock;
190 TAILQ_HEAD(, toepcb) toep_list;
191
192 struct mtx lctx_hash_lock;
193 LIST_HEAD(, listen_ctx) *listen_hash;
194 u_long listen_mask;
195 int lctx_count; /* # of lctx in the hash table */
196
197 struct mtx ppod_lock;
198 int nppods;
199 int nppods_free; /* # of available ppods */
200 int nppods_free_head; /* # of available ppods at the begining */
201 struct ppod_head ppods;
202};
203
204static inline struct tom_data *
205tod_td(struct toedev *tod)
206{
207
208 return (member2struct(tom_data, tod, tod));
209}
210
211static inline struct adapter *
212td_adapter(struct tom_data *td)
213{
214
215 return (td->tod.tod_softc);
216}
217
218/* t4_tom.c */
219struct toepcb *alloc_toepcb(struct port_info *, int, int, int);
220void free_toepcb(struct toepcb *);
221void offload_socket(struct socket *, struct toepcb *);
222void undo_offload_socket(struct socket *);
223void final_cpl_received(struct toepcb *);
224void insert_tid(struct adapter *, int, void *);
225void *lookup_tid(struct adapter *, int);
226void update_tid(struct adapter *, int, void *);
227void remove_tid(struct adapter *, int);
228void release_tid(struct adapter *, int, struct sge_wrq *);
229int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int);
230u_long select_rcv_wnd(struct socket *);
231int select_rcv_wscale(void);
232uint64_t calc_opt0(struct socket *, struct port_info *, struct l2t_entry *,
233 int, int, int, int);
234uint32_t select_ntuple(struct port_info *, struct l2t_entry *, uint32_t);
235
236/* t4_connect.c */
237void t4_init_connect_cpl_handlers(struct adapter *);
238int t4_connect(struct toedev *, struct socket *, struct rtentry *,
239 struct sockaddr *);
240
241/* t4_listen.c */
242void t4_init_listen_cpl_handlers(struct adapter *);
243int t4_listen_start(struct toedev *, struct tcpcb *);
244int t4_listen_stop(struct toedev *, struct tcpcb *);
245void t4_syncache_added(struct toedev *, void *);
246void t4_syncache_removed(struct toedev *, void *);
247int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
248int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
249 struct mbuf *);
250int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
251 struct mbuf *);
252void t4_offload_socket(struct toedev *, void *, struct socket *);
253
254/* t4_cpl_io.c */
255void t4_init_cpl_io_handlers(struct adapter *);
256void t4_uninit_cpl_io_handlers(struct adapter *);
257void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
258void send_flowc_wr(struct toepcb *, struct flowc_tx_params *);
259void send_reset(struct adapter *, struct toepcb *, uint32_t);
260void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
261void t4_rcvd(struct toedev *, struct tcpcb *);
262int t4_tod_output(struct toedev *, struct tcpcb *);
263int t4_send_fin(struct toedev *, struct tcpcb *);
264int t4_send_rst(struct toedev *, struct tcpcb *);
265void t4_set_tcb_field(struct adapter *, struct toepcb *, uint16_t, uint64_t,
266 uint64_t);
267
268/* t4_ddp.c */
269void t4_init_ddp(struct adapter *, struct tom_data *);
270void t4_uninit_ddp(struct adapter *, struct tom_data *);
271int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
272 struct mbuf **, struct mbuf **, int *);
273void enable_ddp(struct adapter *, struct toepcb *toep);
274void release_ddp_resources(struct toepcb *toep);
275#endif
|