kvm tools, setup: Create private directory
[linux-2.6/next.git] / net / sctp / chunk.c
blob6c8556459a751b3e2faa6b0442b804396ff6de7e
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2003, 2004
4 * This file is part of the SCTP kernel implementation
6 * This file contains the code relating the chunk abstraction.
8 * This SCTP implementation is free software;
9 * you can redistribute it and/or modify it under the terms of
10 * the GNU General Public License as published by
11 * the Free Software Foundation; either version 2, or (at your option)
12 * any later version.
14 * This SCTP implementation is distributed in the hope that it
15 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
16 * ************************
17 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
18 * See the GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with GNU CC; see the file COPYING. If not, write to
22 * the Free Software Foundation, 59 Temple Place - Suite 330,
23 * Boston, MA 02111-1307, USA.
25 * Please send any bug reports or fixes you make to the
26 * email address(es):
27 * lksctp developers <lksctp-developers@lists.sourceforge.net>
29 * Or submit a bug report through the following website:
30 * http://www.sf.net/projects/lksctp
32 * Written or modified by:
33 * Jon Grimm <jgrimm@us.ibm.com>
34 * Sridhar Samudrala <sri@us.ibm.com>
36 * Any bugs reported given to us we will try to fix... any fixes shared will
37 * be incorporated into the next SCTP release.
40 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/net.h>
45 #include <linux/inet.h>
46 #include <linux/skbuff.h>
47 #include <linux/slab.h>
48 #include <net/sock.h>
49 #include <net/sctp/sctp.h>
50 #include <net/sctp/sm.h>
52 /* This file is mostly in anticipation of future work, but initially
53 * populate with fragment tracking for an outbound message.
56 /* Initialize datamsg from memory. */
57 static void sctp_datamsg_init(struct sctp_datamsg *msg)
59 atomic_set(&msg->refcnt, 1);
60 msg->send_failed = 0;
61 msg->send_error = 0;
62 msg->can_abandon = 0;
63 msg->can_delay = 1;
64 msg->expires_at = 0;
65 INIT_LIST_HEAD(&msg->chunks);
68 /* Allocate and initialize datamsg. */
69 SCTP_STATIC struct sctp_datamsg *sctp_datamsg_new(gfp_t gfp)
71 struct sctp_datamsg *msg;
72 msg = kmalloc(sizeof(struct sctp_datamsg), gfp);
73 if (msg) {
74 sctp_datamsg_init(msg);
75 SCTP_DBG_OBJCNT_INC(datamsg);
77 return msg;
80 void sctp_datamsg_free(struct sctp_datamsg *msg)
82 struct sctp_chunk *chunk;
84 /* This doesn't have to be a _safe vairant because
85 * sctp_chunk_free() only drops the refs.
87 list_for_each_entry(chunk, &msg->chunks, frag_list)
88 sctp_chunk_free(chunk);
90 sctp_datamsg_put(msg);
93 /* Final destructruction of datamsg memory. */
94 static void sctp_datamsg_destroy(struct sctp_datamsg *msg)
96 struct list_head *pos, *temp;
97 struct sctp_chunk *chunk;
98 struct sctp_sock *sp;
99 struct sctp_ulpevent *ev;
100 struct sctp_association *asoc = NULL;
101 int error = 0, notify;
103 /* If we failed, we may need to notify. */
104 notify = msg->send_failed ? -1 : 0;
106 /* Release all references. */
107 list_for_each_safe(pos, temp, &msg->chunks) {
108 list_del_init(pos);
109 chunk = list_entry(pos, struct sctp_chunk, frag_list);
110 /* Check whether we _really_ need to notify. */
111 if (notify < 0) {
112 asoc = chunk->asoc;
113 if (msg->send_error)
114 error = msg->send_error;
115 else
116 error = asoc->outqueue.error;
118 sp = sctp_sk(asoc->base.sk);
119 notify = sctp_ulpevent_type_enabled(SCTP_SEND_FAILED,
120 &sp->subscribe);
123 /* Generate a SEND FAILED event only if enabled. */
124 if (notify > 0) {
125 int sent;
126 if (chunk->has_tsn)
127 sent = SCTP_DATA_SENT;
128 else
129 sent = SCTP_DATA_UNSENT;
131 ev = sctp_ulpevent_make_send_failed(asoc, chunk, sent,
132 error, GFP_ATOMIC);
133 if (ev)
134 sctp_ulpq_tail_event(&asoc->ulpq, ev);
137 sctp_chunk_put(chunk);
140 SCTP_DBG_OBJCNT_DEC(datamsg);
141 kfree(msg);
144 /* Hold a reference. */
145 static void sctp_datamsg_hold(struct sctp_datamsg *msg)
147 atomic_inc(&msg->refcnt);
150 /* Release a reference. */
151 void sctp_datamsg_put(struct sctp_datamsg *msg)
153 if (atomic_dec_and_test(&msg->refcnt))
154 sctp_datamsg_destroy(msg);
157 /* Assign a chunk to this datamsg. */
158 static void sctp_datamsg_assign(struct sctp_datamsg *msg, struct sctp_chunk *chunk)
160 sctp_datamsg_hold(msg);
161 chunk->msg = msg;
165 /* A data chunk can have a maximum payload of (2^16 - 20). Break
166 * down any such message into smaller chunks. Opportunistically, fragment
167 * the chunks down to the current MTU constraints. We may get refragmented
168 * later if the PMTU changes, but it is _much better_ to fragment immediately
169 * with a reasonable guess than always doing our fragmentation on the
170 * soft-interrupt.
172 struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
173 struct sctp_sndrcvinfo *sinfo,
174 struct msghdr *msgh, int msg_len)
176 int max, whole, i, offset, over, err;
177 int len, first_len;
178 int max_data;
179 struct sctp_chunk *chunk;
180 struct sctp_datamsg *msg;
181 struct list_head *pos, *temp;
182 __u8 frag;
184 msg = sctp_datamsg_new(GFP_KERNEL);
185 if (!msg)
186 return NULL;
188 /* Note: Calculate this outside of the loop, so that all fragments
189 * have the same expiration.
191 if (sinfo->sinfo_timetolive) {
192 /* sinfo_timetolive is in milliseconds */
193 msg->expires_at = jiffies +
194 msecs_to_jiffies(sinfo->sinfo_timetolive);
195 msg->can_abandon = 1;
196 SCTP_DEBUG_PRINTK("%s: msg:%p expires_at: %ld jiffies:%ld\n",
197 __func__, msg, msg->expires_at, jiffies);
200 /* This is the biggest possible DATA chunk that can fit into
201 * the packet
203 max_data = asoc->pathmtu -
204 sctp_sk(asoc->base.sk)->pf->af->net_header_len -
205 sizeof(struct sctphdr) - sizeof(struct sctp_data_chunk);
207 max = asoc->frag_point;
208 /* If the the peer requested that we authenticate DATA chunks
209 * we need to accound for bundling of the AUTH chunks along with
210 * DATA.
212 if (sctp_auth_send_cid(SCTP_CID_DATA, asoc)) {
213 struct sctp_hmac *hmac_desc = sctp_auth_asoc_get_hmac(asoc);
215 if (hmac_desc)
216 max_data -= WORD_ROUND(sizeof(sctp_auth_chunk_t) +
217 hmac_desc->hmac_len);
220 /* Now, check if we need to reduce our max */
221 if (max > max_data)
222 max = max_data;
224 whole = 0;
225 first_len = max;
227 /* Check to see if we have a pending SACK and try to let it be bundled
228 * with this message. Do this if we don't have any data queued already.
229 * To check that, look at out_qlen and retransmit list.
230 * NOTE: we will not reduce to account for SACK, if the message would
231 * not have been fragmented.
233 if (timer_pending(&asoc->timers[SCTP_EVENT_TIMEOUT_SACK]) &&
234 asoc->outqueue.out_qlen == 0 &&
235 list_empty(&asoc->outqueue.retransmit) &&
236 msg_len > max)
237 max_data -= WORD_ROUND(sizeof(sctp_sack_chunk_t));
239 /* Encourage Cookie-ECHO bundling. */
240 if (asoc->state < SCTP_STATE_COOKIE_ECHOED)
241 max_data -= SCTP_ARBITRARY_COOKIE_ECHO_LEN;
243 /* Now that we adjusted completely, reset first_len */
244 if (first_len > max_data)
245 first_len = max_data;
247 /* Account for a different sized first fragment */
248 if (msg_len >= first_len) {
249 msg_len -= first_len;
250 whole = 1;
251 msg->can_delay = 0;
254 /* How many full sized? How many bytes leftover? */
255 whole += msg_len / max;
256 over = msg_len % max;
257 offset = 0;
259 if ((whole > 1) || (whole && over))
260 SCTP_INC_STATS_USER(SCTP_MIB_FRAGUSRMSGS);
262 /* Create chunks for all the full sized DATA chunks. */
263 for (i=0, len=first_len; i < whole; i++) {
264 frag = SCTP_DATA_MIDDLE_FRAG;
266 if (0 == i)
267 frag |= SCTP_DATA_FIRST_FRAG;
269 if ((i == (whole - 1)) && !over) {
270 frag |= SCTP_DATA_LAST_FRAG;
272 /* The application requests to set the I-bit of the
273 * last DATA chunk of a user message when providing
274 * the user message to the SCTP implementation.
276 if ((sinfo->sinfo_flags & SCTP_EOF) ||
277 (sinfo->sinfo_flags & SCTP_SACK_IMMEDIATELY))
278 frag |= SCTP_DATA_SACK_IMM;
281 chunk = sctp_make_datafrag_empty(asoc, sinfo, len, frag, 0);
283 if (!chunk)
284 goto errout;
285 err = sctp_user_addto_chunk(chunk, offset, len, msgh->msg_iov);
286 if (err < 0)
287 goto errout;
289 offset += len;
291 /* Put the chunk->skb back into the form expected by send. */
292 __skb_pull(chunk->skb, (__u8 *)chunk->chunk_hdr
293 - (__u8 *)chunk->skb->data);
295 sctp_datamsg_assign(msg, chunk);
296 list_add_tail(&chunk->frag_list, &msg->chunks);
298 /* The first chunk, the first chunk was likely short
299 * to allow bundling, so reset to full size.
301 if (0 == i)
302 len = max;
305 /* .. now the leftover bytes. */
306 if (over) {
307 if (!whole)
308 frag = SCTP_DATA_NOT_FRAG;
309 else
310 frag = SCTP_DATA_LAST_FRAG;
312 if ((sinfo->sinfo_flags & SCTP_EOF) ||
313 (sinfo->sinfo_flags & SCTP_SACK_IMMEDIATELY))
314 frag |= SCTP_DATA_SACK_IMM;
316 chunk = sctp_make_datafrag_empty(asoc, sinfo, over, frag, 0);
318 if (!chunk)
319 goto errout;
321 err = sctp_user_addto_chunk(chunk, offset, over,msgh->msg_iov);
323 /* Put the chunk->skb back into the form expected by send. */
324 __skb_pull(chunk->skb, (__u8 *)chunk->chunk_hdr
325 - (__u8 *)chunk->skb->data);
326 if (err < 0)
327 goto errout;
329 sctp_datamsg_assign(msg, chunk);
330 list_add_tail(&chunk->frag_list, &msg->chunks);
333 return msg;
335 errout:
336 list_for_each_safe(pos, temp, &msg->chunks) {
337 list_del_init(pos);
338 chunk = list_entry(pos, struct sctp_chunk, frag_list);
339 sctp_chunk_free(chunk);
341 sctp_datamsg_put(msg);
342 return NULL;
345 /* Check whether this message has expired. */
346 int sctp_chunk_abandoned(struct sctp_chunk *chunk)
348 struct sctp_datamsg *msg = chunk->msg;
350 if (!msg->can_abandon)
351 return 0;
353 if (time_after(jiffies, msg->expires_at))
354 return 1;
356 return 0;
359 /* This chunk (and consequently entire message) has failed in its sending. */
360 void sctp_chunk_fail(struct sctp_chunk *chunk, int error)
362 chunk->msg->send_failed = 1;
363 chunk->msg->send_error = error;