1 // SPDX-License-Identifier: GPL-2.0-or-later 1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* SCTP kernel implementation 2 /* SCTP kernel implementation
3 * (C) Copyright IBM Corp. 2003, 2004 3 * (C) Copyright IBM Corp. 2003, 2004
4 * 4 *
5 * This file is part of the SCTP kernel implem 5 * This file is part of the SCTP kernel implementation
6 * 6 *
7 * This file contains the code relating the ch 7 * This file contains the code relating the chunk abstraction.
8 * 8 *
9 * Please send any bug reports or fixes you ma 9 * Please send any bug reports or fixes you make to the
10 * email address(es): 10 * email address(es):
11 * lksctp developers <linux-sctp@vger.kerne 11 * lksctp developers <linux-sctp@vger.kernel.org>
12 * 12 *
13 * Written or modified by: 13 * Written or modified by:
14 * Jon Grimm <jgrimm@us.ibm.com 14 * Jon Grimm <jgrimm@us.ibm.com>
15 * Sridhar Samudrala <sri@us.ibm.com> 15 * Sridhar Samudrala <sri@us.ibm.com>
16 */ 16 */
17 17
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
19 19
20 #include <linux/types.h> 20 #include <linux/types.h>
21 #include <linux/kernel.h> 21 #include <linux/kernel.h>
22 #include <linux/net.h> 22 #include <linux/net.h>
23 #include <linux/inet.h> 23 #include <linux/inet.h>
24 #include <linux/skbuff.h> 24 #include <linux/skbuff.h>
25 #include <linux/slab.h> 25 #include <linux/slab.h>
26 #include <net/sock.h> 26 #include <net/sock.h>
27 #include <net/sctp/sctp.h> 27 #include <net/sctp/sctp.h>
28 #include <net/sctp/sm.h> 28 #include <net/sctp/sm.h>
29 29
30 /* This file is mostly in anticipation of futu 30 /* This file is mostly in anticipation of future work, but initially
31 * populate with fragment tracking for an outb 31 * populate with fragment tracking for an outbound message.
32 */ 32 */
33 33
34 /* Initialize datamsg from memory. */ 34 /* Initialize datamsg from memory. */
35 static void sctp_datamsg_init(struct sctp_data 35 static void sctp_datamsg_init(struct sctp_datamsg *msg)
36 { 36 {
37 refcount_set(&msg->refcnt, 1); 37 refcount_set(&msg->refcnt, 1);
38 msg->send_failed = 0; 38 msg->send_failed = 0;
39 msg->send_error = 0; 39 msg->send_error = 0;
40 msg->can_delay = 1; 40 msg->can_delay = 1;
41 msg->abandoned = 0; 41 msg->abandoned = 0;
42 msg->expires_at = 0; 42 msg->expires_at = 0;
43 INIT_LIST_HEAD(&msg->chunks); 43 INIT_LIST_HEAD(&msg->chunks);
44 } 44 }
45 45
46 /* Allocate and initialize datamsg. */ 46 /* Allocate and initialize datamsg. */
47 static struct sctp_datamsg *sctp_datamsg_new(g 47 static struct sctp_datamsg *sctp_datamsg_new(gfp_t gfp)
48 { 48 {
49 struct sctp_datamsg *msg; 49 struct sctp_datamsg *msg;
50 msg = kmalloc(sizeof(struct sctp_datam 50 msg = kmalloc(sizeof(struct sctp_datamsg), gfp);
51 if (msg) { 51 if (msg) {
52 sctp_datamsg_init(msg); 52 sctp_datamsg_init(msg);
53 SCTP_DBG_OBJCNT_INC(datamsg); 53 SCTP_DBG_OBJCNT_INC(datamsg);
54 } 54 }
55 return msg; 55 return msg;
56 } 56 }
57 57
58 void sctp_datamsg_free(struct sctp_datamsg *ms 58 void sctp_datamsg_free(struct sctp_datamsg *msg)
59 { 59 {
60 struct sctp_chunk *chunk; 60 struct sctp_chunk *chunk;
61 61
62 /* This doesn't have to be a _safe vai 62 /* This doesn't have to be a _safe vairant because
63 * sctp_chunk_free() only drops the re 63 * sctp_chunk_free() only drops the refs.
64 */ 64 */
65 list_for_each_entry(chunk, &msg->chunk 65 list_for_each_entry(chunk, &msg->chunks, frag_list)
66 sctp_chunk_free(chunk); 66 sctp_chunk_free(chunk);
67 67
68 sctp_datamsg_put(msg); 68 sctp_datamsg_put(msg);
69 } 69 }
70 70
71 /* Final destructruction of datamsg memory. */ 71 /* Final destructruction of datamsg memory. */
72 static void sctp_datamsg_destroy(struct sctp_d 72 static void sctp_datamsg_destroy(struct sctp_datamsg *msg)
73 { 73 {
74 struct sctp_association *asoc = NULL; 74 struct sctp_association *asoc = NULL;
75 struct list_head *pos, *temp; 75 struct list_head *pos, *temp;
76 struct sctp_chunk *chunk; 76 struct sctp_chunk *chunk;
77 struct sctp_ulpevent *ev; 77 struct sctp_ulpevent *ev;
78 int error, sent; 78 int error, sent;
79 79
80 /* Release all references. */ 80 /* Release all references. */
81 list_for_each_safe(pos, temp, &msg->ch 81 list_for_each_safe(pos, temp, &msg->chunks) {
82 list_del_init(pos); 82 list_del_init(pos);
83 chunk = list_entry(pos, struct 83 chunk = list_entry(pos, struct sctp_chunk, frag_list);
84 84
85 if (!msg->send_failed) { 85 if (!msg->send_failed) {
86 sctp_chunk_put(chunk); 86 sctp_chunk_put(chunk);
87 continue; 87 continue;
88 } 88 }
89 89
90 asoc = chunk->asoc; 90 asoc = chunk->asoc;
91 error = msg->send_error ?: aso 91 error = msg->send_error ?: asoc->outqueue.error;
92 sent = chunk->has_tsn ? SCTP_D 92 sent = chunk->has_tsn ? SCTP_DATA_SENT : SCTP_DATA_UNSENT;
93 93
94 if (sctp_ulpevent_type_enabled 94 if (sctp_ulpevent_type_enabled(asoc->subscribe,
95 95 SCTP_SEND_FAILED)) {
96 ev = sctp_ulpevent_mak 96 ev = sctp_ulpevent_make_send_failed(asoc, chunk, sent,
97 97 error, GFP_ATOMIC);
98 if (ev) 98 if (ev)
99 asoc->stream.s 99 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
100 } 100 }
101 101
102 if (sctp_ulpevent_type_enabled 102 if (sctp_ulpevent_type_enabled(asoc->subscribe,
103 103 SCTP_SEND_FAILED_EVENT)) {
104 ev = sctp_ulpevent_mak 104 ev = sctp_ulpevent_make_send_failed_event(asoc, chunk,
105 105 sent, error,
106 106 GFP_ATOMIC);
107 if (ev) 107 if (ev)
108 asoc->stream.s 108 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
109 } 109 }
110 110
111 sctp_chunk_put(chunk); 111 sctp_chunk_put(chunk);
112 } 112 }
113 113
114 SCTP_DBG_OBJCNT_DEC(datamsg); 114 SCTP_DBG_OBJCNT_DEC(datamsg);
115 kfree(msg); 115 kfree(msg);
116 } 116 }
117 117
118 /* Hold a reference. */ 118 /* Hold a reference. */
119 static void sctp_datamsg_hold(struct sctp_data 119 static void sctp_datamsg_hold(struct sctp_datamsg *msg)
120 { 120 {
121 refcount_inc(&msg->refcnt); 121 refcount_inc(&msg->refcnt);
122 } 122 }
123 123
124 /* Release a reference. */ 124 /* Release a reference. */
125 void sctp_datamsg_put(struct sctp_datamsg *msg 125 void sctp_datamsg_put(struct sctp_datamsg *msg)
126 { 126 {
127 if (refcount_dec_and_test(&msg->refcnt 127 if (refcount_dec_and_test(&msg->refcnt))
128 sctp_datamsg_destroy(msg); 128 sctp_datamsg_destroy(msg);
129 } 129 }
130 130
131 /* Assign a chunk to this datamsg. */ 131 /* Assign a chunk to this datamsg. */
132 static void sctp_datamsg_assign(struct sctp_da 132 static void sctp_datamsg_assign(struct sctp_datamsg *msg, struct sctp_chunk *chunk)
133 { 133 {
134 sctp_datamsg_hold(msg); 134 sctp_datamsg_hold(msg);
135 chunk->msg = msg; 135 chunk->msg = msg;
136 } 136 }
137 137
138 138
139 /* A data chunk can have a maximum payload of 139 /* A data chunk can have a maximum payload of (2^16 - 20). Break
140 * down any such message into smaller chunks. 140 * down any such message into smaller chunks. Opportunistically, fragment
141 * the chunks down to the current MTU constrai 141 * the chunks down to the current MTU constraints. We may get refragmented
142 * later if the PMTU changes, but it is _much 142 * later if the PMTU changes, but it is _much better_ to fragment immediately
143 * with a reasonable guess than always doing o 143 * with a reasonable guess than always doing our fragmentation on the
144 * soft-interrupt. 144 * soft-interrupt.
145 */ 145 */
146 struct sctp_datamsg *sctp_datamsg_from_user(st 146 struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
147 st 147 struct sctp_sndrcvinfo *sinfo,
148 st 148 struct iov_iter *from)
149 { 149 {
150 size_t len, first_len, max_data, remai 150 size_t len, first_len, max_data, remaining;
151 size_t msg_len = iov_iter_count(from); 151 size_t msg_len = iov_iter_count(from);
152 struct sctp_shared_key *shkey = NULL; 152 struct sctp_shared_key *shkey = NULL;
153 struct list_head *pos, *temp; 153 struct list_head *pos, *temp;
154 struct sctp_chunk *chunk; 154 struct sctp_chunk *chunk;
155 struct sctp_datamsg *msg; 155 struct sctp_datamsg *msg;
156 int err; 156 int err;
157 157
158 msg = sctp_datamsg_new(GFP_KERNEL); 158 msg = sctp_datamsg_new(GFP_KERNEL);
159 if (!msg) 159 if (!msg)
160 return ERR_PTR(-ENOMEM); 160 return ERR_PTR(-ENOMEM);
161 161
162 /* Note: Calculate this outside of the 162 /* Note: Calculate this outside of the loop, so that all fragments
163 * have the same expiration. 163 * have the same expiration.
164 */ 164 */
165 if (asoc->peer.prsctp_capable && sinfo 165 if (asoc->peer.prsctp_capable && sinfo->sinfo_timetolive &&
166 (SCTP_PR_TTL_ENABLED(sinfo->sinfo_ 166 (SCTP_PR_TTL_ENABLED(sinfo->sinfo_flags) ||
167 !SCTP_PR_POLICY(sinfo->sinfo_flag 167 !SCTP_PR_POLICY(sinfo->sinfo_flags)))
168 msg->expires_at = jiffies + 168 msg->expires_at = jiffies +
169 msecs_to_jif 169 msecs_to_jiffies(sinfo->sinfo_timetolive);
170 170
171 /* This is the biggest possible DATA c 171 /* This is the biggest possible DATA chunk that can fit into
172 * the packet 172 * the packet
173 */ 173 */
174 max_data = asoc->frag_point; 174 max_data = asoc->frag_point;
175 if (unlikely(!max_data)) { 175 if (unlikely(!max_data)) {
176 max_data = sctp_min_frag_point 176 max_data = sctp_min_frag_point(sctp_sk(asoc->base.sk),
177 177 sctp_datachk_len(&asoc->stream));
178 pr_warn_ratelimited("%s: asoc: 178 pr_warn_ratelimited("%s: asoc:%p frag_point is zero, forcing max_data to default minimum (%zu)",
179 __func__, 179 __func__, asoc, max_data);
180 } 180 }
181 181
182 /* If the peer requested that we authe 182 /* If the peer requested that we authenticate DATA chunks
183 * we need to account for bundling of 183 * we need to account for bundling of the AUTH chunks along with
184 * DATA. 184 * DATA.
185 */ 185 */
186 if (sctp_auth_send_cid(SCTP_CID_DATA, 186 if (sctp_auth_send_cid(SCTP_CID_DATA, asoc)) {
187 struct sctp_hmac *hmac_desc = 187 struct sctp_hmac *hmac_desc = sctp_auth_asoc_get_hmac(asoc);
188 188
189 if (hmac_desc) 189 if (hmac_desc)
190 max_data -= SCTP_PAD4( 190 max_data -= SCTP_PAD4(sizeof(struct sctp_auth_chunk) +
191 191 hmac_desc->hmac_len);
192 192
193 if (sinfo->sinfo_tsn && 193 if (sinfo->sinfo_tsn &&
194 sinfo->sinfo_ssn != asoc-> 194 sinfo->sinfo_ssn != asoc->active_key_id) {
195 shkey = sctp_auth_get_ 195 shkey = sctp_auth_get_shkey(asoc, sinfo->sinfo_ssn);
196 if (!shkey) { 196 if (!shkey) {
197 err = -EINVAL; 197 err = -EINVAL;
198 goto errout; 198 goto errout;
199 } 199 }
200 } else { 200 } else {
201 shkey = asoc->shkey; 201 shkey = asoc->shkey;
202 } 202 }
203 } 203 }
204 204
205 /* Set first_len and then account for 205 /* Set first_len and then account for possible bundles on first frag */
206 first_len = max_data; 206 first_len = max_data;
207 207
208 /* Check to see if we have a pending S 208 /* Check to see if we have a pending SACK and try to let it be bundled
209 * with this message. Do this if we d 209 * with this message. Do this if we don't have any data queued already.
210 * To check that, look at out_qlen and 210 * To check that, look at out_qlen and retransmit list.
211 * NOTE: we will not reduce to account 211 * NOTE: we will not reduce to account for SACK, if the message would
212 * not have been fragmented. 212 * not have been fragmented.
213 */ 213 */
214 if (timer_pending(&asoc->timers[SCTP_E 214 if (timer_pending(&asoc->timers[SCTP_EVENT_TIMEOUT_SACK]) &&
215 asoc->outqueue.out_qlen == 0 && 215 asoc->outqueue.out_qlen == 0 &&
216 list_empty(&asoc->outqueue.retrans 216 list_empty(&asoc->outqueue.retransmit) &&
217 msg_len > max_data) 217 msg_len > max_data)
218 first_len -= SCTP_PAD4(sizeof( 218 first_len -= SCTP_PAD4(sizeof(struct sctp_sack_chunk));
219 219
220 /* Encourage Cookie-ECHO bundling. */ 220 /* Encourage Cookie-ECHO bundling. */
221 if (asoc->state < SCTP_STATE_COOKIE_EC 221 if (asoc->state < SCTP_STATE_COOKIE_ECHOED)
222 first_len -= SCTP_ARBITRARY_CO 222 first_len -= SCTP_ARBITRARY_COOKIE_ECHO_LEN;
223 223
224 /* Account for a different sized first 224 /* Account for a different sized first fragment */
225 if (msg_len >= first_len) { 225 if (msg_len >= first_len) {
226 msg->can_delay = 0; 226 msg->can_delay = 0;
227 if (msg_len > first_len) 227 if (msg_len > first_len)
228 SCTP_INC_STATS(asoc->b 228 SCTP_INC_STATS(asoc->base.net,
229 SCTP_MI 229 SCTP_MIB_FRAGUSRMSGS);
230 } else { 230 } else {
231 /* Which may be the only one.. 231 /* Which may be the only one... */
232 first_len = msg_len; 232 first_len = msg_len;
233 } 233 }
234 234
235 /* Create chunks for all DATA chunks. 235 /* Create chunks for all DATA chunks. */
236 for (remaining = msg_len; remaining; r 236 for (remaining = msg_len; remaining; remaining -= len) {
237 u8 frag = SCTP_DATA_MIDDLE_FRA 237 u8 frag = SCTP_DATA_MIDDLE_FRAG;
238 238
239 if (remaining == msg_len) { 239 if (remaining == msg_len) {
240 /* First frag, which m 240 /* First frag, which may also be the last */
241 frag |= SCTP_DATA_FIRS 241 frag |= SCTP_DATA_FIRST_FRAG;
242 len = first_len; 242 len = first_len;
243 } else { 243 } else {
244 /* Middle frags */ 244 /* Middle frags */
245 len = max_data; 245 len = max_data;
246 } 246 }
247 247
248 if (len >= remaining) { 248 if (len >= remaining) {
249 /* Last frag, which ma 249 /* Last frag, which may also be the first */
250 len = remaining; 250 len = remaining;
251 frag |= SCTP_DATA_LAST 251 frag |= SCTP_DATA_LAST_FRAG;
252 252
253 /* The application req 253 /* The application requests to set the I-bit of the
254 * last DATA chunk of 254 * last DATA chunk of a user message when providing
255 * the user message to 255 * the user message to the SCTP implementation.
256 */ 256 */
257 if ((sinfo->sinfo_flag 257 if ((sinfo->sinfo_flags & SCTP_EOF) ||
258 (sinfo->sinfo_flag 258 (sinfo->sinfo_flags & SCTP_SACK_IMMEDIATELY))
259 frag |= SCTP_D 259 frag |= SCTP_DATA_SACK_IMM;
260 } 260 }
261 261
262 chunk = asoc->stream.si->make_ 262 chunk = asoc->stream.si->make_datafrag(asoc, sinfo, len, frag,
263 263 GFP_KERNEL);
264 if (!chunk) { 264 if (!chunk) {
265 err = -ENOMEM; 265 err = -ENOMEM;
266 goto errout; 266 goto errout;
267 } 267 }
268 268
269 err = sctp_user_addto_chunk(ch 269 err = sctp_user_addto_chunk(chunk, len, from);
270 if (err < 0) 270 if (err < 0)
271 goto errout_chunk_free 271 goto errout_chunk_free;
272 272
273 chunk->shkey = shkey; 273 chunk->shkey = shkey;
274 274
275 /* Put the chunk->skb back int 275 /* Put the chunk->skb back into the form expected by send. */
276 __skb_pull(chunk->skb, (__u8 * 276 __skb_pull(chunk->skb, (__u8 *)chunk->chunk_hdr -
277 chunk-> 277 chunk->skb->data);
278 278
279 sctp_datamsg_assign(msg, chunk 279 sctp_datamsg_assign(msg, chunk);
280 list_add_tail(&chunk->frag_lis 280 list_add_tail(&chunk->frag_list, &msg->chunks);
281 } 281 }
282 282
283 return msg; 283 return msg;
284 284
285 errout_chunk_free: 285 errout_chunk_free:
286 sctp_chunk_free(chunk); 286 sctp_chunk_free(chunk);
287 287
288 errout: 288 errout:
289 list_for_each_safe(pos, temp, &msg->ch 289 list_for_each_safe(pos, temp, &msg->chunks) {
290 list_del_init(pos); 290 list_del_init(pos);
291 chunk = list_entry(pos, struct 291 chunk = list_entry(pos, struct sctp_chunk, frag_list);
292 sctp_chunk_free(chunk); 292 sctp_chunk_free(chunk);
293 } 293 }
294 sctp_datamsg_put(msg); 294 sctp_datamsg_put(msg);
295 295
296 return ERR_PTR(err); 296 return ERR_PTR(err);
297 } 297 }
298 298
299 /* Check whether this message has expired. */ 299 /* Check whether this message has expired. */
300 int sctp_chunk_abandoned(struct sctp_chunk *ch 300 int sctp_chunk_abandoned(struct sctp_chunk *chunk)
301 { 301 {
302 if (!chunk->asoc->peer.prsctp_capable) 302 if (!chunk->asoc->peer.prsctp_capable)
303 return 0; 303 return 0;
304 304
305 if (chunk->msg->abandoned) 305 if (chunk->msg->abandoned)
306 return 1; 306 return 1;
307 307
308 if (!chunk->has_tsn && 308 if (!chunk->has_tsn &&
309 !(chunk->chunk_hdr->flags & SCTP_D 309 !(chunk->chunk_hdr->flags & SCTP_DATA_FIRST_FRAG))
310 return 0; 310 return 0;
311 311
312 if (SCTP_PR_TTL_ENABLED(chunk->sinfo.s 312 if (SCTP_PR_TTL_ENABLED(chunk->sinfo.sinfo_flags) &&
313 time_after(jiffies, chunk->msg->ex 313 time_after(jiffies, chunk->msg->expires_at)) {
314 struct sctp_stream_out *stream 314 struct sctp_stream_out *streamout =
315 SCTP_SO(&chunk->asoc-> 315 SCTP_SO(&chunk->asoc->stream,
316 chunk->sinfo.s 316 chunk->sinfo.sinfo_stream);
317 317
318 if (chunk->sent_count) { 318 if (chunk->sent_count) {
319 chunk->asoc->abandoned 319 chunk->asoc->abandoned_sent[SCTP_PR_INDEX(TTL)]++;
320 streamout->ext->abando 320 streamout->ext->abandoned_sent[SCTP_PR_INDEX(TTL)]++;
321 } else { 321 } else {
322 chunk->asoc->abandoned 322 chunk->asoc->abandoned_unsent[SCTP_PR_INDEX(TTL)]++;
323 streamout->ext->abando 323 streamout->ext->abandoned_unsent[SCTP_PR_INDEX(TTL)]++;
324 } 324 }
325 chunk->msg->abandoned = 1; 325 chunk->msg->abandoned = 1;
326 return 1; 326 return 1;
327 } else if (SCTP_PR_RTX_ENABLED(chunk-> 327 } else if (SCTP_PR_RTX_ENABLED(chunk->sinfo.sinfo_flags) &&
328 chunk->sent_count > chunk-> 328 chunk->sent_count > chunk->sinfo.sinfo_timetolive) {
329 struct sctp_stream_out *stream 329 struct sctp_stream_out *streamout =
330 SCTP_SO(&chunk->asoc-> 330 SCTP_SO(&chunk->asoc->stream,
331 chunk->sinfo.s 331 chunk->sinfo.sinfo_stream);
332 332
333 chunk->asoc->abandoned_sent[SC 333 chunk->asoc->abandoned_sent[SCTP_PR_INDEX(RTX)]++;
334 streamout->ext->abandoned_sent 334 streamout->ext->abandoned_sent[SCTP_PR_INDEX(RTX)]++;
335 chunk->msg->abandoned = 1; 335 chunk->msg->abandoned = 1;
336 return 1; 336 return 1;
337 } else if (!SCTP_PR_POLICY(chunk->sinf 337 } else if (!SCTP_PR_POLICY(chunk->sinfo.sinfo_flags) &&
338 chunk->msg->expires_at && 338 chunk->msg->expires_at &&
339 time_after(jiffies, chunk-> 339 time_after(jiffies, chunk->msg->expires_at)) {
340 chunk->msg->abandoned = 1; 340 chunk->msg->abandoned = 1;
341 return 1; 341 return 1;
342 } 342 }
343 /* PRIO policy is processed by sendmsg 343 /* PRIO policy is processed by sendmsg, not here */
344 344
345 return 0; 345 return 0;
346 } 346 }
347 347
348 /* This chunk (and consequently entire message 348 /* This chunk (and consequently entire message) has failed in its sending. */
349 void sctp_chunk_fail(struct sctp_chunk *chunk, 349 void sctp_chunk_fail(struct sctp_chunk *chunk, int error)
350 { 350 {
351 chunk->msg->send_failed = 1; 351 chunk->msg->send_failed = 1;
352 chunk->msg->send_error = error; 352 chunk->msg->send_error = error;
353 } 353 }
354 354
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.