2 +--------------------------------------------------------------------+
3 | libmemcached - C/C++ Client Library for memcached |
4 +--------------------------------------------------------------------+
5 | Redistribution and use in source and binary forms, with or without |
6 | modification, are permitted under the terms of the BSD license. |
7 | You should have received a copy of the license in a bundled file |
8 | named LICENSE; in case you did not receive a copy you can review |
9 | the terms online at: https://opensource.org/licenses/BSD-3-Clause |
10 +--------------------------------------------------------------------+
11 | Copyright (c) 2006-2014 Brian Aker https://datadifferential.com/ |
12 | Copyright (c) 2020 Michael Wallner <mike@php.net> |
13 +--------------------------------------------------------------------+
16 #include "libmemcached/common.h"
18 #ifdef HAVE_SYS_SOCKET_H
19 # include <sys/socket.h>
22 void initialize_binary_request(memcached_instance_st
*server
,
23 protocol_binary_request_header
&header
) {
25 header
.request
.magic
= PROTOCOL_BINARY_REQ
;
26 header
.request
.opaque
= htons(server
->request_id
);
29 enum memc_read_or_write
{ MEM_READ
, MEM_WRITE
};
32 * Try to fill the input buffer for a server with as much
35 * @param instance the server to pack
37 static bool repack_input_buffer(memcached_instance_st
*instance
) {
38 if (instance
->read_ptr
!= instance
->read_buffer
) {
39 /* Move all of the data to the beginning of the buffer so
40 ** that we can fit more data into the buffer...
42 memmove(instance
->read_buffer
, instance
->read_ptr
, instance
->read_buffer_length
);
43 instance
->read_ptr
= instance
->read_buffer
;
46 /* There is room in the buffer, try to fill it! */
47 if (instance
->read_buffer_length
!= MEMCACHED_MAX_BUFFER
) {
49 /* Just try a single read to grab what's available */
51 if ((nr
= ::recv(instance
->fd
, instance
->read_ptr
+ instance
->read_buffer_length
,
52 MEMCACHED_MAX_BUFFER
- instance
->read_buffer_length
, MSG_NOSIGNAL
))
56 memcached_set_error(*instance
, MEMCACHED_CONNECTION_FAILURE
, MEMCACHED_AT
);
58 switch (get_socket_errno()) {
62 #if EWOULDBLOCK != EAGAIN
69 break; // No IO is fine, we can just move on
72 memcached_set_errno(*instance
, get_socket_errno(), MEMCACHED_AT
);
77 } else // We read data, append to our read buffer
79 instance
->read_buffer_length
+= size_t(nr
);
90 * If the we have callbacks connected to this server structure
91 * we may start process the input queue and fire the callbacks
92 * for the incomming messages. This function is _only_ called
93 * when the input buffer is full, so that we _know_ that we have
94 * at least _one_ message to process.
96 * @param instance the server to star processing iput messages for
97 * @return true if we processed anything, false otherwise
99 static bool process_input_buffer(memcached_instance_st
*instance
) {
101 ** We might be able to process some of the response messages if we
102 ** have a callback set up
104 if (instance
->root
->callbacks
!= NULL
) {
106 * We might have responses... try to read them out and fire
109 memcached_callback_st cb
= *instance
->root
->callbacks
;
111 memcached_set_processing_input((Memcached
*) instance
->root
, true);
113 char buffer
[MEMCACHED_DEFAULT_COMMAND_SIZE
];
114 Memcached
*root
= (Memcached
*) instance
->root
;
115 memcached_return_t error
= memcached_response(instance
, buffer
, sizeof(buffer
), &root
->result
);
117 memcached_set_processing_input(root
, false);
119 if (error
== MEMCACHED_SUCCESS
) {
120 for (unsigned int x
= 0; x
< cb
.number_of_callback
; x
++) {
121 error
= (*cb
.callback
[x
])(instance
->root
, &root
->result
, cb
.context
);
122 if (error
!= MEMCACHED_SUCCESS
) {
127 /* @todo what should I do with the error message??? */
129 /* @todo what should I do with other error messages?? */
136 static memcached_return_t
io_wait(memcached_instance_st
*instance
, const short events
) {
138 ** We are going to block on write, but at least on Solaris we might block
139 ** on write if we haven't read anything from our input buffer..
140 ** Try to purge the input buffer if we don't do any flow control in the
141 ** application layer (just sending a lot of data etc)
142 ** The test is moved down in the purge function to avoid duplication of
145 if (events
& POLLOUT
) {
146 if (memcached_purge(instance
) == false) {
147 return MEMCACHED_FAILURE
;
152 fds
.fd
= instance
->fd
;
156 if (fds
.events
& POLLOUT
) /* write */ {
157 instance
->io_wait_count
.write
++;
159 instance
->io_wait_count
.read
++;
162 if (instance
->root
->poll_timeout
163 == 0) // Mimic 0 causes timeout behavior (not all platforms do this)
165 return memcached_set_error(*instance
, MEMCACHED_TIMEOUT
, MEMCACHED_AT
,
166 memcached_literal_param("poll_timeout() was set to zero"));
170 while (--loop_max
) // While loop is for ERESTART or EINTR
172 int active_fd
= poll(&fds
, 1, instance
->root
->poll_timeout
);
174 if (active_fd
>= 1) {
175 assert_msg(active_fd
== 1, "poll() returned an unexpected number of active file descriptors");
176 if (fds
.revents
& POLLIN
or fds
.revents
& POLLOUT
) {
177 return MEMCACHED_SUCCESS
;
180 if (fds
.revents
& POLLHUP
) {
181 return memcached_set_error(*instance
, MEMCACHED_CONNECTION_FAILURE
, MEMCACHED_AT
,
182 memcached_literal_param("poll() detected hang up"));
185 if (fds
.revents
& POLLERR
) {
186 int local_errno
= EINVAL
;
188 socklen_t len
= sizeof(err
);
189 if (getsockopt(instance
->fd
, SOL_SOCKET
, SO_ERROR
, (char *) &err
, &len
) == 0) {
190 if (err
== 0) // treat this as EINTR
196 memcached_quit_server(instance
, true);
197 return memcached_set_errno(*instance
, local_errno
, MEMCACHED_AT
,
198 memcached_literal_param("poll() returned POLLHUP"));
201 return memcached_set_error(
202 *instance
, MEMCACHED_FAILURE
, MEMCACHED_AT
,
203 memcached_literal_param("poll() returned a value that was not dealt with"));
206 if (active_fd
== 0) {
207 return memcached_set_error(*instance
, MEMCACHED_TIMEOUT
, MEMCACHED_AT
,
208 memcached_literal_param("No active_fd were found"));
211 // Only an error should result in this code being called.
212 int local_errno
= get_socket_errno(); // We cache in case memcached_quit_server() modifies errno
213 assert_msg(active_fd
== -1, "poll() returned an unexpected value");
214 switch (local_errno
) {
223 memcached_set_error(*instance
, MEMCACHED_MEMORY_ALLOCATION_FAILURE
, MEMCACHED_AT
);
227 memcached_set_error(*instance
, MEMCACHED_MEMORY_ALLOCATION_FAILURE
, MEMCACHED_AT
,
228 memcached_literal_param(
229 "RLIMIT_NOFILE exceeded, or if OSX the timeout value was invalid"));
233 memcached_set_errno(*instance
, local_errno
, MEMCACHED_AT
, memcached_literal_param("poll"));
239 memcached_quit_server(instance
, true);
241 if (memcached_has_error(instance
)) {
242 return memcached_instance_error_return(instance
);
245 return memcached_set_error(
246 *instance
, MEMCACHED_CONNECTION_FAILURE
, MEMCACHED_AT
,
247 memcached_literal_param("number of attempts to call io_wait() failed"));
250 static bool io_flush(memcached_instance_st
*instance
, const bool with_flush
,
251 memcached_return_t
&error
) {
253 ** We might want to purge the input buffer if we haven't consumed
254 ** any output yet... The test for the limits is the purge is inline
255 ** in the purge function to avoid duplicating the logic..
258 WATCHPOINT_ASSERT(instance
->fd
!= INVALID_SOCKET
);
260 if (memcached_purge(instance
) == false) {
264 char *local_write_ptr
= instance
->write_buffer
;
265 size_t write_length
= instance
->write_buffer_offset
;
267 error
= MEMCACHED_SUCCESS
;
269 WATCHPOINT_ASSERT(instance
->fd
!= INVALID_SOCKET
);
271 /* Looking for memory overflows */
273 if (write_length
== MEMCACHED_MAX_BUFFER
)
274 WATCHPOINT_ASSERT(instance
->write_buffer
== local_write_ptr
);
275 WATCHPOINT_ASSERT((instance
->write_buffer
+ MEMCACHED_MAX_BUFFER
)
276 >= (local_write_ptr
+ write_length
));
279 while (write_length
) {
280 WATCHPOINT_ASSERT(instance
->fd
!= INVALID_SOCKET
);
281 WATCHPOINT_ASSERT(write_length
> 0);
285 flags
= MSG_NOSIGNAL
;
287 flags
= MSG_NOSIGNAL
| MSG_MORE
;
290 ssize_t sent_length
= ::send(instance
->fd
, local_write_ptr
, write_length
, flags
);
291 int local_errno
= get_socket_errno(); // We cache in case memcached_quit_server() modifies errno
293 if (sent_length
== SOCKET_ERROR
) {
294 #if 0 // @todo I should look at why we hit this bit of code hard frequently
295 WATCHPOINT_ERRNO(get_socket_errno());
296 WATCHPOINT_NUMBER(get_socket_errno());
298 switch (get_socket_errno()) {
302 #if EWOULDBLOCK != EAGAIN
307 * We may be blocked on write because the input buffer
308 * is full. Let's check if we have room in our input
309 * buffer for more data and retry the write before
312 if (repack_input_buffer(instance
) or process_input_buffer(instance
)) {
316 memcached_return_t rc
= io_wait(instance
, POLLOUT
);
317 if (memcached_success(rc
)) {
319 } else if (rc
== MEMCACHED_TIMEOUT
) {
323 memcached_quit_server(instance
, true);
324 error
= memcached_set_errno(*instance
, local_errno
, MEMCACHED_AT
);
330 memcached_quit_server(instance
, true);
331 error
= memcached_set_errno(*instance
, local_errno
, MEMCACHED_AT
);
332 WATCHPOINT_ASSERT(instance
->fd
== INVALID_SOCKET
);
337 instance
->io_bytes_sent
+= uint32_t(sent_length
);
339 local_write_ptr
+= sent_length
;
340 write_length
-= uint32_t(sent_length
);
343 WATCHPOINT_ASSERT(write_length
== 0);
344 instance
->write_buffer_offset
= 0;
349 memcached_return_t
memcached_io_wait_for_write(memcached_instance_st
*instance
) {
350 return io_wait(instance
, POLLOUT
);
353 memcached_return_t
memcached_io_wait_for_read(memcached_instance_st
*instance
) {
354 return io_wait(instance
, POLLIN
);
357 static memcached_return_t
_io_fill(memcached_instance_st
*instance
) {
360 data_read
= ::recv(instance
->fd
, instance
->read_buffer
, MEMCACHED_MAX_BUFFER
, MSG_NOSIGNAL
);
361 int local_errno
= get_socket_errno(); // We cache in case memcached_quit_server() modifies errno
363 if (data_read
== SOCKET_ERROR
) {
364 switch (get_socket_errno()) {
365 case EINTR
: // We just retry
368 case ETIMEDOUT
: // OSX
369 #if EWOULDBLOCK != EAGAIN
377 memcached_return_t io_wait_ret
;
378 if (memcached_success(io_wait_ret
= io_wait(instance
, POLLIN
))) {
387 case ENOTCONN
: // Programmer Error
388 WATCHPOINT_ASSERT(0);
391 WATCHPOINT_ASSERT(0);
394 assert_msg(instance
->fd
!= INVALID_SOCKET
, "Programmer error, invalid socket");
400 memcached_quit_server(instance
, true);
401 memcached_set_errno(*instance
, local_errno
, MEMCACHED_AT
);
405 return memcached_instance_error_return(instance
);
406 } else if (data_read
== 0) {
408 EOF. Any data received so far is incomplete
409 so discard it. This always reads by byte in case of TCP
410 and protocol enforcement happens at memcached_response()
411 looking for '\n'. We do not care for UDB which requests 8 bytes
412 at once. Generally, this means that connection went away. Since
413 for blocking I/O we do not return 0 and for non-blocking case
414 it will return EGAIN if data is not immediatly available.
416 memcached_quit_server(instance
, true);
417 return memcached_set_error(
418 *instance
, MEMCACHED_CONNECTION_FAILURE
, MEMCACHED_AT
,
419 memcached_literal_param("::rec() returned zero, server has disconnected"));
421 instance
->io_wait_count
._bytes_read
+= data_read
;
422 } while (data_read
<= 0);
424 instance
->io_bytes_sent
= 0;
425 instance
->read_buffer_length
= (size_t) data_read
;
426 instance
->read_ptr
= instance
->read_buffer
;
428 return MEMCACHED_SUCCESS
;
431 memcached_return_t
memcached_io_read(memcached_instance_st
*instance
, void *buffer
, size_t length
,
433 assert(memcached_is_udp(instance
->root
) == false);
436 "Programmer error, memcached_io_read() recieved an invalid Instance"); // Programmer error
437 char *buffer_ptr
= static_cast<char *>(buffer
);
439 if (instance
->fd
== INVALID_SOCKET
) {
441 assert_msg(int(instance
->state
) <= int(MEMCACHED_SERVER_STATE_ADDRINFO
), "Programmer error, invalid socket state");
443 return MEMCACHED_CONNECTION_FAILURE
;
447 if (instance
->read_buffer_length
== 0) {
448 memcached_return_t io_fill_ret
;
449 if (memcached_fatal(io_fill_ret
= _io_fill(instance
))) {
457 (length
> instance
->read_buffer_length
) ? instance
->read_buffer_length
: length
;
459 memcpy(buffer_ptr
, instance
->read_ptr
, difference
);
460 length
-= difference
;
461 instance
->read_ptr
+= difference
;
462 instance
->read_buffer_length
-= difference
;
463 buffer_ptr
+= difference
;
465 *buffer_ptr
= *instance
->read_ptr
;
466 instance
->read_ptr
++;
467 instance
->read_buffer_length
--;
473 nread
= ssize_t(buffer_ptr
- (char *) buffer
);
475 return MEMCACHED_SUCCESS
;
478 memcached_return_t
memcached_io_slurp(memcached_instance_st
*instance
) {
479 assert_msg(instance
, "Programmer error, invalid Instance");
480 assert(memcached_is_udp(instance
->root
) == false);
482 if (instance
->fd
== INVALID_SOCKET
) {
483 assert_msg(int(instance
->state
) <= int(MEMCACHED_SERVER_STATE_ADDRINFO
),
484 "Invalid socket state");
485 return MEMCACHED_CONNECTION_FAILURE
;
489 char buffer
[MEMCACHED_MAX_BUFFER
];
491 data_read
= ::recv(instance
->fd
, instance
->read_buffer
, sizeof(buffer
), MSG_NOSIGNAL
);
492 if (data_read
== SOCKET_ERROR
) {
493 switch (get_socket_errno()) {
494 case EINTR
: // We just retry
497 case ETIMEDOUT
: // OSX
498 #if EWOULDBLOCK != EAGAIN
505 if (memcached_success(io_wait(instance
, POLLIN
))) {
508 return MEMCACHED_IN_PROGRESS
;
512 case ENOTCONN
: // Programmer Error
517 assert_msg(instance
->fd
!= INVALID_SOCKET
, "Invalid socket state");
523 return MEMCACHED_CONNECTION_FAILURE
; // We want this!
526 } while (data_read
> 0);
528 return MEMCACHED_CONNECTION_FAILURE
;
531 static bool _io_write(memcached_instance_st
*instance
, const void *buffer
, size_t length
,
532 bool with_flush
, size_t &written
) {
533 assert(instance
->fd
!= INVALID_SOCKET
);
534 assert(memcached_is_udp(instance
->root
) == false);
536 const char *buffer_ptr
= static_cast<const char *>(buffer
);
538 const size_t original_length
= length
;
542 size_t buffer_end
= MEMCACHED_MAX_BUFFER
;
543 size_t should_write
= buffer_end
- instance
->write_buffer_offset
;
544 should_write
= (should_write
< length
) ? should_write
: length
;
546 write_ptr
= instance
->write_buffer
+ instance
->write_buffer_offset
;
547 memcpy(write_ptr
, buffer_ptr
, should_write
);
548 instance
->write_buffer_offset
+= should_write
;
549 buffer_ptr
+= should_write
;
550 length
-= should_write
;
552 if (instance
->write_buffer_offset
== buffer_end
) {
553 WATCHPOINT_ASSERT(instance
->fd
!= INVALID_SOCKET
);
555 memcached_return_t rc
;
556 if (io_flush(instance
, with_flush
, rc
) == false) {
557 written
= original_length
- length
;
564 memcached_return_t rc
;
565 WATCHPOINT_ASSERT(instance
->fd
!= INVALID_SOCKET
);
566 if (io_flush(instance
, with_flush
, rc
) == false) {
567 written
= original_length
- length
;
572 written
= original_length
- length
;
577 bool memcached_io_write(memcached_instance_st
*instance
) {
579 return _io_write(instance
, NULL
, 0, true, written
);
582 ssize_t
memcached_io_write(memcached_instance_st
*instance
, const void *buffer
, const size_t length
,
583 const bool with_flush
) {
586 if (_io_write(instance
, buffer
, length
, with_flush
, written
) == false) {
590 return ssize_t(written
);
593 bool memcached_io_writev(memcached_instance_st
*instance
, libmemcached_io_vector_st vector
[],
594 const size_t number_of
, const bool with_flush
) {
595 ssize_t complete_total
= 0;
598 for (size_t x
= 0; x
< number_of
; x
++, vector
++) {
599 complete_total
+= vector
->length
;
600 if (vector
->length
) {
602 if ((_io_write(instance
, vector
->buffer
, vector
->length
, false, written
)) == false) {
610 if (memcached_io_write(instance
) == false) {
615 return (complete_total
== total
);
618 void memcached_instance_st::start_close_socket() {
619 if (fd
!= INVALID_SOCKET
) {
620 shutdown(fd
, SHUT_WR
);
621 options
.is_shutting_down
= true;
625 void memcached_instance_st::reset_socket() {
626 if (fd
!= INVALID_SOCKET
) {
627 (void) closesocket(fd
);
632 void memcached_instance_st::close_socket() {
633 if (fd
!= INVALID_SOCKET
) {
634 int shutdown_options
= SHUT_RD
;
635 if (options
.is_shutting_down
== false) {
636 shutdown_options
= SHUT_RDWR
;
639 /* in case of death shutdown to avoid blocking at close() */
640 if (shutdown(fd
, shutdown_options
) == SOCKET_ERROR
and get_socket_errno() != ENOTCONN
) {
641 WATCHPOINT_NUMBER(fd
);
642 WATCHPOINT_ERRNO(get_socket_errno());
643 WATCHPOINT_ASSERT(get_socket_errno());
647 state
= MEMCACHED_SERVER_STATE_NEW
;
650 state
= MEMCACHED_SERVER_STATE_NEW
;
653 write_buffer_offset
= size_t(root
and memcached_is_udp(root
) ? UDP_DATAGRAM_HEADER_LENGTH
: 0);
654 read_buffer_length
= 0;
655 read_ptr
= read_buffer
;
656 options
.is_shutting_down
= false;
657 memcached_server_response_reset(this);
659 // We reset the version so that if we end up talking to a different server
660 // we don't have stale server version information.
661 major_version
= minor_version
= micro_version
= UINT8_MAX
;
664 memcached_instance_st
*memcached_io_get_readable_server(Memcached
*memc
, memcached_return_t
&) {
665 #define MAX_SERVERS_TO_POLL 100
666 struct pollfd fds
[MAX_SERVERS_TO_POLL
];
667 nfds_t host_index
= 0;
669 for (uint32_t x
= 0; x
< memcached_server_count(memc
) and host_index
< MAX_SERVERS_TO_POLL
; ++x
) {
670 memcached_instance_st
*instance
= memcached_instance_fetch(memc
, x
);
672 if (instance
->read_buffer_length
> 0) /* I have data in the buffer */ {
676 if (instance
->response_count() > 0) {
677 fds
[host_index
].events
= POLLIN
;
678 fds
[host_index
].revents
= 0;
679 fds
[host_index
].fd
= instance
->fd
;
684 if (host_index
< 2) {
685 /* We have 0 or 1 server with pending events.. */
686 for (uint32_t x
= 0; x
< memcached_server_count(memc
); ++x
) {
687 memcached_instance_st
*instance
= memcached_instance_fetch(memc
, x
);
689 if (instance
->response_count() > 0) {
697 int error
= poll(fds
, host_index
, memc
->poll_timeout
);
700 memcached_set_errno(*memc
, get_socket_errno(), MEMCACHED_AT
);
706 for (nfds_t x
= 0; x
< host_index
; ++x
) {
707 if (fds
[x
].revents
& POLLIN
) {
708 for (uint32_t y
= 0; y
< memcached_server_count(memc
); ++y
) {
709 memcached_instance_st
*instance
= memcached_instance_fetch(memc
, y
);
711 if (instance
->fd
== fds
[x
].fd
) {
723 Eventually we will just kill off the server with the problem.
725 void memcached_io_reset(memcached_instance_st
*instance
) {
726 memcached_quit_server(instance
, true);
730 * Read a given number of bytes from the server and place it into a specific
731 * buffer. Reset the IO channel on this server if an error occurs.
733 memcached_return_t
memcached_safe_read(memcached_instance_st
*instance
, void *dta
,
736 char *data
= static_cast<char *>(dta
);
738 while (offset
< size
) {
740 memcached_return_t rc
;
743 memcached_continue(rc
= memcached_io_read(instance
, data
+ offset
, size
- offset
, nread
))) {
746 if (memcached_failed(rc
)) {
750 offset
+= size_t(nread
);
753 return MEMCACHED_SUCCESS
;
756 memcached_return_t
memcached_io_readline(memcached_instance_st
*instance
, char *buffer_ptr
,
757 size_t size
, size_t &total_nr
) {
759 bool line_complete
= false;
761 while (line_complete
== false) {
762 if (instance
->read_buffer_length
== 0) {
764 * We don't have any data in the buffer, so let's fill the read
765 * buffer. Call the standard read function to avoid duplicating
769 memcached_return_t rc
= memcached_io_read(instance
, buffer_ptr
, 1, nread
);
770 if (memcached_failed(rc
) and rc
== MEMCACHED_IN_PROGRESS
) {
771 memcached_quit_server(instance
, true);
772 return memcached_set_error(*instance
, rc
, MEMCACHED_AT
);
773 } else if (memcached_failed(rc
)) {
777 if (*buffer_ptr
== '\n') {
778 line_complete
= true;
785 /* Now let's look in the buffer and copy as we go! */
786 while (instance
->read_buffer_length
and total_nr
< size
and line_complete
== false) {
787 *buffer_ptr
= *instance
->read_ptr
;
788 if (*buffer_ptr
== '\n') {
789 line_complete
= true;
791 --instance
->read_buffer_length
;
792 ++instance
->read_ptr
;
797 if (total_nr
== size
) {
798 return MEMCACHED_PROTOCOL_ERROR
;
802 return MEMCACHED_SUCCESS
;