2 * Copyright (c) 2008-2012 Niels Provos and Nick Mathewson
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
12 * 3. The name of the author may not be used to endorse or promote products
13 * derived from this software without specific prior written permission.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 #ifndef BUFFEREVENT_INTERNAL_H_INCLUDED_
27 #define BUFFEREVENT_INTERNAL_H_INCLUDED_
33 #include "event2/event-config.h"
34 #include "event2/event_struct.h"
35 #include "evconfig-private.h"
36 #include "event2/util.h"
37 #include "defer-internal.h"
38 #include "evthread-internal.h"
39 #include "event2/thread.h"
40 #include "ratelim-internal.h"
41 #include "event2/bufferevent_struct.h"
43 #include "ipv6-internal.h"
47 #ifdef EVENT__HAVE_NETINET_IN_H
48 #include <netinet/in.h>
50 #ifdef EVENT__HAVE_NETINET_IN6_H
51 #include <netinet/in6.h>
54 /* These flags are reasons that we might be declining to actually enable
55 reading or writing on a bufferevent.
58 /* On a all bufferevents, for reading: used when we have read up to the
61 On a filtering bufferevent, for writing: used when the underlying
62 bufferevent's write buffer has been filled up to its watermark
65 #define BEV_SUSPEND_WM 0x01
66 /* On a base bufferevent: when we have emptied a bandwidth buckets */
67 #define BEV_SUSPEND_BW 0x02
68 /* On a base bufferevent: when we have emptied the group's bandwidth bucket. */
69 #define BEV_SUSPEND_BW_GROUP 0x04
70 /* On a socket bufferevent: can't do any operations while we're waiting for
71 * name lookup to finish. */
72 #define BEV_SUSPEND_LOOKUP 0x08
73 /* On a base bufferevent, for reading: used when a filter has choked this
74 * (underlying) bufferevent because it has stopped reading from it. */
75 #define BEV_SUSPEND_FILT_READ 0x10
77 typedef ev_uint16_t bufferevent_suspend_flags
;
79 struct bufferevent_rate_limit_group
{
80 /** List of all members in the group */
81 LIST_HEAD(rlim_group_member_list
, bufferevent_private
) members
;
82 /** Current limits for the group. */
83 struct ev_token_bucket rate_limit
;
84 struct ev_token_bucket_cfg rate_limit_cfg
;
86 /** True iff we don't want to read from any member of the group.until
87 * the token bucket refills. */
88 unsigned read_suspended
: 1;
89 /** True iff we don't want to write from any member of the group.until
90 * the token bucket refills. */
91 unsigned write_suspended
: 1;
92 /** True iff we were unable to suspend one of the bufferevents in the
93 * group for reading the last time we tried, and we should try
95 unsigned pending_unsuspend_read
: 1;
96 /** True iff we were unable to suspend one of the bufferevents in the
97 * group for writing the last time we tried, and we should try
99 unsigned pending_unsuspend_write
: 1;
102 /** Total number of bytes read or written in this group since last
104 ev_uint64_t total_read
;
105 ev_uint64_t total_written
;
108 /** The number of bufferevents in the group. */
111 /** The smallest number of bytes that any member of the group should
112 * be limited to read or write at a time. */
113 ev_ssize_t min_share
;
114 ev_ssize_t configured_min_share
;
116 /** Timeout event that goes off once a tick, when the bucket is ready
118 struct event master_refill_event
;
120 /** Seed for weak random number generator. Protected by 'lock' */
121 struct evutil_weakrand_state weakrand_seed
;
123 /** Lock to protect the members of this group. This lock should nest
124 * within every bufferevent lock: if you are holding this lock, do
125 * not assume you can lock another bufferevent. */
129 /** Fields for rate-limiting a single bufferevent. */
130 struct bufferevent_rate_limit
{
131 /* Linked-list elements for storing this bufferevent_private in a
134 * Note that this field is supposed to be protected by the group
136 LIST_ENTRY(bufferevent_private
) next_in_group
;
137 /** The rate-limiting group for this bufferevent, or NULL if it is
138 * only rate-limited on its own. */
139 struct bufferevent_rate_limit_group
*group
;
141 /* This bufferevent's current limits. */
142 struct ev_token_bucket limit
;
143 /* Pointer to the rate-limit configuration for this bufferevent.
144 * Can be shared. XXX reference-count this? */
145 struct ev_token_bucket_cfg
*cfg
;
147 /* Timeout event used when one this bufferevent's buckets are
149 struct event refill_bucket_event
;
152 /** Parts of the bufferevent structure that are shared among all bufferevent
153 * types, but not exposed in bufferevent_struct.h. */
154 struct bufferevent_private
{
155 /** The underlying bufferevent structure. */
156 struct bufferevent bev
;
158 /** Evbuffer callback to enforce watermarks on input. */
159 struct evbuffer_cb_entry
*read_watermarks_cb
;
161 /** If set, we should free the lock when we free the bufferevent. */
162 unsigned own_lock
: 1;
164 /** Flag: set if we have deferred callbacks and a read callback is
166 unsigned readcb_pending
: 1;
167 /** Flag: set if we have deferred callbacks and a write callback is
169 unsigned writecb_pending
: 1;
170 /** Flag: set if we are currently busy connecting. */
171 unsigned connecting
: 1;
172 /** Flag: set if a connect failed prematurely; this is a hack for
173 * getting around the bufferevent abstraction. */
174 unsigned connection_refused
: 1;
175 /** Set to the events pending if we have deferred callbacks and
176 * an events callback is pending. */
177 short eventcb_pending
;
179 /** If set, read is suspended until one or more conditions are over.
180 * The actual value here is a bitfield of those conditions; see the
181 * BEV_SUSPEND_* flags above. */
182 bufferevent_suspend_flags read_suspended
;
184 /** If set, writing is suspended until one or more conditions are over.
185 * The actual value here is a bitfield of those conditions; see the
186 * BEV_SUSPEND_* flags above. */
187 bufferevent_suspend_flags write_suspended
;
189 /** Set to the current socket errno if we have deferred callbacks and
190 * an events callback is pending. */
193 /** The DNS error code for bufferevent_socket_connect_hostname */
196 /** Used to implement deferred callbacks */
197 struct event_callback deferred
;
199 /** The options this bufferevent was constructed with */
200 enum bufferevent_options options
;
202 /** Current reference count for this bufferevent. */
205 /** Lock for this bufferevent. Shared by the inbuf and the outbuf.
206 * If NULL, locking is disabled. */
209 /** No matter how big our bucket gets, don't try to read more than this
210 * much in a single read operation. */
211 ev_ssize_t max_single_read
;
213 /** No matter how big our bucket gets, don't try to write more than this
214 * much in a single write operation. */
215 ev_ssize_t max_single_write
;
217 /** Rate-limiting information for this bufferevent */
218 struct bufferevent_rate_limit
*rate_limiting
;
220 /* Saved conn_addr, to extract IP address from it.
222 * Because some servers may reset/close connection without waiting clients,
223 * in that case we can't extract IP address even in close_cb.
224 * So we need to save it, just after we connected to remote server, or
225 * after resolving (to avoid extra dns requests during retrying, since UDP
228 struct sockaddr_in6 in6
;
229 struct sockaddr_in in
;
233 /** Possible operations for a control callback. */
234 enum bufferevent_ctrl_op
{
237 BEV_CTRL_GET_UNDERLYING
,
241 /** Possible data types for a control callback */
242 union bufferevent_ctrl_data
{
248 Implementation table for a bufferevent: holds function pointers and other
249 information to make the various bufferevent types work.
251 struct bufferevent_ops
{
252 /** The name of the bufferevent's type. */
254 /** At what offset into the implementation type will we find a
255 bufferevent structure?
257 Example: if the type is implemented as
258 struct bufferevent_x {
260 struct bufferevent bev;
262 then mem_offset should be offsetof(struct bufferevent_x, bev)
266 /** Enables one or more of EV_READ|EV_WRITE on a bufferevent. Does
267 not need to adjust the 'enabled' field. Returns 0 on success, -1
270 int (*enable
)(struct bufferevent
*, short);
272 /** Disables one or more of EV_READ|EV_WRITE on a bufferevent. Does
273 not need to adjust the 'enabled' field. Returns 0 on success, -1
276 int (*disable
)(struct bufferevent
*, short);
278 /** Detatches the bufferevent from related data structures. Called as
279 * soon as its reference count reaches 0. */
280 void (*unlink
)(struct bufferevent
*);
282 /** Free any storage and deallocate any extra data or structures used
283 in this implementation. Called when the bufferevent is
286 void (*destruct
)(struct bufferevent
*);
288 /** Called when the timeouts on the bufferevent have changed.*/
289 int (*adj_timeouts
)(struct bufferevent
*);
291 /** Called to flush data. */
292 int (*flush
)(struct bufferevent
*, short, enum bufferevent_flush_mode
);
294 /** Called to access miscellaneous fields. */
295 int (*ctrl
)(struct bufferevent
*, enum bufferevent_ctrl_op
, union bufferevent_ctrl_data
*);
299 extern const struct bufferevent_ops bufferevent_ops_socket
;
300 extern const struct bufferevent_ops bufferevent_ops_filter
;
301 extern const struct bufferevent_ops bufferevent_ops_pair
;
303 #define BEV_IS_SOCKET(bevp) ((bevp)->be_ops == &bufferevent_ops_socket)
304 #define BEV_IS_FILTER(bevp) ((bevp)->be_ops == &bufferevent_ops_filter)
305 #define BEV_IS_PAIR(bevp) ((bevp)->be_ops == &bufferevent_ops_pair)
308 extern const struct bufferevent_ops bufferevent_ops_async
;
309 #define BEV_IS_ASYNC(bevp) ((bevp)->be_ops == &bufferevent_ops_async)
311 #define BEV_IS_ASYNC(bevp) 0
314 /** Initialize the shared parts of a bufferevent. */
315 int bufferevent_init_common_(struct bufferevent_private
*, struct event_base
*, const struct bufferevent_ops
*, enum bufferevent_options options
);
317 /** For internal use: temporarily stop all reads on bufev, until the conditions
318 * in 'what' are over. */
319 void bufferevent_suspend_read_(struct bufferevent
*bufev
, bufferevent_suspend_flags what
);
320 /** For internal use: clear the conditions 'what' on bufev, and re-enable
321 * reading if there are no conditions left. */
322 void bufferevent_unsuspend_read_(struct bufferevent
*bufev
, bufferevent_suspend_flags what
);
324 /** For internal use: temporarily stop all writes on bufev, until the conditions
325 * in 'what' are over. */
326 void bufferevent_suspend_write_(struct bufferevent
*bufev
, bufferevent_suspend_flags what
);
327 /** For internal use: clear the conditions 'what' on bufev, and re-enable
328 * writing if there are no conditions left. */
329 void bufferevent_unsuspend_write_(struct bufferevent
*bufev
, bufferevent_suspend_flags what
);
331 #define bufferevent_wm_suspend_read(b) \
332 bufferevent_suspend_read_((b), BEV_SUSPEND_WM)
333 #define bufferevent_wm_unsuspend_read(b) \
334 bufferevent_unsuspend_read_((b), BEV_SUSPEND_WM)
337 Disable a bufferevent. Equivalent to bufferevent_disable(), but
338 first resets 'connecting' flag to force EV_WRITE down for sure.
340 XXXX this method will go away in the future; try not to add new users.
341 See comment in evhttp_connection_reset_() for discussion.
343 @param bufev the bufferevent to be disabled
344 @param event any combination of EV_READ | EV_WRITE.
345 @return 0 if successful, or -1 if an error occurred
346 @see bufferevent_disable()
348 int bufferevent_disable_hard_(struct bufferevent
*bufev
, short event
);
350 /** Internal: Set up locking on a bufferevent. If lock is set, use it.
351 * Otherwise, use a new lock. */
352 int bufferevent_enable_locking_(struct bufferevent
*bufev
, void *lock
);
353 /** Internal: backwards compat macro for the now public function
354 * Increment the reference count on bufev. */
355 #define bufferevent_incref_(bufev) bufferevent_incref(bufev)
356 /** Internal: Lock bufev and increase its reference count.
357 * unlocking it otherwise. */
358 void bufferevent_incref_and_lock_(struct bufferevent
*bufev
);
359 /** Internal: backwards compat macro for the now public function
360 * Decrement the reference count on bufev. Returns 1 if it freed
362 #define bufferevent_decref_(bufev) bufferevent_decref(bufev)
364 /** Internal: Drop the reference count on bufev, freeing as necessary, and
365 * unlocking it otherwise. Returns 1 if it freed the bufferevent. */
366 int bufferevent_decref_and_unlock_(struct bufferevent
*bufev
);
368 /** Internal: If callbacks are deferred and we have a read callback, schedule
369 * a readcb. Otherwise just run the readcb. Ignores watermarks. */
370 void bufferevent_run_readcb_(struct bufferevent
*bufev
, int options
);
371 /** Internal: If callbacks are deferred and we have a write callback, schedule
372 * a writecb. Otherwise just run the writecb. Ignores watermarks. */
373 void bufferevent_run_writecb_(struct bufferevent
*bufev
, int options
);
374 /** Internal: If callbacks are deferred and we have an eventcb, schedule
375 * it to run with events "what". Otherwise just run the eventcb.
376 * See bufferevent_trigger_event for meaning of "options". */
377 void bufferevent_run_eventcb_(struct bufferevent
*bufev
, short what
, int options
);
379 /** Internal: Run or schedule (if deferred or options contain
380 * BEV_TRIG_DEFER_CALLBACKS) I/O callbacks specified in iotype.
381 * Must already hold the bufev lock. Honors watermarks unless
382 * BEV_TRIG_IGNORE_WATERMARKS is in options. */
383 static inline void bufferevent_trigger_nolock_(struct bufferevent
*bufev
, short iotype
, int options
);
385 /* Making this inline since all of the common-case calls to this function in
386 * libevent use constant arguments. */
388 bufferevent_trigger_nolock_(struct bufferevent
*bufev
, short iotype
, int options
)
390 if ((iotype
& EV_READ
) && ((options
& BEV_TRIG_IGNORE_WATERMARKS
) ||
391 evbuffer_get_length(bufev
->input
) >= bufev
->wm_read
.low
))
392 bufferevent_run_readcb_(bufev
, options
);
393 if ((iotype
& EV_WRITE
) && ((options
& BEV_TRIG_IGNORE_WATERMARKS
) ||
394 evbuffer_get_length(bufev
->output
) <= bufev
->wm_write
.low
))
395 bufferevent_run_writecb_(bufev
, options
);
398 /** Internal: Add the event 'ev' with timeout tv, unless tv is set to 0, in
399 * which case add ev with no timeout. */
400 int bufferevent_add_event_(struct event
*ev
, const struct timeval
*tv
);
403 * These next functions implement timeouts for bufferevents that aren't doing
404 * anything else with ev_read and ev_write, to handle timeouts.
406 /** Internal use: Set up the ev_read and ev_write callbacks so that
407 * the other "generic_timeout" functions will work on it. Call this from
408 * the constructor function. */
409 void bufferevent_init_generic_timeout_cbs_(struct bufferevent
*bev
);
410 /** Internal use: Add or delete the generic timeout events as appropriate.
411 * (If an event is enabled and a timeout is set, we add the event. Otherwise
412 * we delete it.) Call this from anything that changes the timeout values,
413 * that enabled EV_READ or EV_WRITE, or that disables EV_READ or EV_WRITE. */
414 int bufferevent_generic_adj_timeouts_(struct bufferevent
*bev
);
415 int bufferevent_generic_adj_existing_timeouts_(struct bufferevent
*bev
);
417 enum bufferevent_options
bufferevent_get_options_(struct bufferevent
*bev
);
419 const struct sockaddr
*
420 bufferevent_socket_get_conn_address_(struct bufferevent
*bev
);
422 /** Internal use: We have just successfully read data into an inbuf, so
423 * reset the read timeout (if any). */
424 #define BEV_RESET_GENERIC_READ_TIMEOUT(bev) \
426 if (evutil_timerisset(&(bev)->timeout_read)) \
427 event_add(&(bev)->ev_read, &(bev)->timeout_read); \
429 /** Internal use: We have just successfully written data from an inbuf, so
430 * reset the read timeout (if any). */
431 #define BEV_RESET_GENERIC_WRITE_TIMEOUT(bev) \
433 if (evutil_timerisset(&(bev)->timeout_write)) \
434 event_add(&(bev)->ev_write, &(bev)->timeout_write); \
436 #define BEV_DEL_GENERIC_READ_TIMEOUT(bev) \
437 event_del(&(bev)->ev_read)
438 #define BEV_DEL_GENERIC_WRITE_TIMEOUT(bev) \
439 event_del(&(bev)->ev_write)
442 /** Internal: Given a bufferevent, return its corresponding
443 * bufferevent_private. */
444 #define BEV_UPCAST(b) EVUTIL_UPCAST((b), struct bufferevent_private, bev)
446 #ifdef EVENT__DISABLE_THREAD_SUPPORT
447 #define BEV_LOCK(b) EVUTIL_NIL_STMT_
448 #define BEV_UNLOCK(b) EVUTIL_NIL_STMT_
450 /** Internal: Grab the lock (if any) on a bufferevent */
451 #define BEV_LOCK(b) do { \
452 struct bufferevent_private *locking = BEV_UPCAST(b); \
453 EVLOCK_LOCK(locking->lock, 0); \
456 /** Internal: Release the lock (if any) on a bufferevent */
457 #define BEV_UNLOCK(b) do { \
458 struct bufferevent_private *locking = BEV_UPCAST(b); \
459 EVLOCK_UNLOCK(locking->lock, 0); \
464 /* ==== For rate-limiting. */
466 int bufferevent_decrement_write_buckets_(struct bufferevent_private
*bev
,
468 int bufferevent_decrement_read_buckets_(struct bufferevent_private
*bev
,
470 ev_ssize_t
bufferevent_get_read_max_(struct bufferevent_private
*bev
);
471 ev_ssize_t
bufferevent_get_write_max_(struct bufferevent_private
*bev
);
473 int bufferevent_ratelim_init_(struct bufferevent_private
*bev
);
480 #endif /* BUFFEREVENT_INTERNAL_H_INCLUDED_ */