summaryrefslogtreecommitdiff
path: root/src/common/socket.h
AgeCommit message (Collapse)AuthorFilesLines
2015-12-15Added GPL-compliant header to all sources and build scriptsHaru1-4/+20
Signed-off-by: Haru <haru@dotalux.com>
2015-10-11Changed various s_subnet arrays to VECTORsHaru1-7/+13
- sockt->lan_subnet was renamed to sockt->lan_subnets. - sockt->trusted_ip was renamed to sockt->trusted_ips. - sockt->allowed_ip was renamed to sockt->allowed_ips. - Convenience macros for checking IP ranges and subnets are provided (SUBNET_MATCH, APPLY_MASK). Signed-off-by: Haru <haru@dotalux.com>
2015-10-11Cleanup of the HPluginData implementation (second part)Haru1-2/+1
- Changed the hplugin_data_store's array into a VECTOR. Signed-off-by: Haru <haru@dotalux.com>
2015-10-11Cleanup of the HPluginData implementation (First part)Haru1-3/+2
- Several explicit casts are removed, to have a slightly better type-checking at compile time. - A destructor function is provided, to remove code duplication. Signed-off-by: Haru <haru@dotalux.com>
2015-08-15HPM compatibility improvementsHaru1-3/+3
Improved compatibility, portability and standards conformance. - Since it is not possible to portably and reliably re-use the core's symbols in plugins, symbols are no longer exported unless explicitly required, in the UNIX builds. This mimics the Windows behavior and adds HPM compatibility to OSes such as FreeBSD. Credits to Andrei Karas for making this possible. - For convenience, it is no longer necessary to call GET_SYMBOL, since the plugin will automatically import all the available symbols when it's loaded, depending on the included headers. - Plugins are now supposed to include the "common/hercules.h" header before including anything else. Incluing common/HPMi.h, common/cbasetypes.h or conf/core.h is no longer necessary, as those are guaranteed to be automatically included by hercules.h. - HPM API version bumped to 1.1. Signed-off-by: Haru <haru@dotalux.com>
2015-08-15Moved session array into the socket interfaceHaru1-17/+19
Signed-off-by: Haru <haru@dotalux.com>
2015-08-15Cleaned up socket interfaceHaru1-31/+11
- Replaced some macro calls with the proper interface syntax - Removed useless macros and workarounds API changes summary: - WFIFOSET() can now be safely used both inside and outside socket.c - RFIFOSKIP() can now be safely used both inside and outside socket.c - do_close() is now sockt->close() - flush_fifo() is now sockt->flush() - flush_fifos() is now sockt->flush_fifos() - getips() is now sockt->getips() - host2ip() is now sockt->host2ip() - ip2str() is now sockt->ip2str() - ntows() is now sockt->ntows() - make_connection() is now sockt->make_connection() - make_listen_bind() is now sockt->make_listen_bind() - realloc_fifo() is now sockt->realloc_fifo() - realloc_writefifo() is now sockt->realloc_writefifo() - session_isActive() is now sockt->session_is_active() - session_isValid() is now sockt->session_is_valid() - set_defaultparse() is now sockt->set_defaultparse() - set_eof() is now sockt->eof() - set_noblocking() is now sockt->set_noblocking() - str2ip() is now sockt->str2ip() Signed-off-by: Haru <haru@dotalux.com>
2015-08-11subnet.conf overhaulHaru1-0/+21
The subnet.conf system has been rewritten to offer greater flexibility, and to fix some issues that appeared with 838321a36c79e71117320154c9b611c99e93af03. It is now possible to enter, separately, LAN subnets: - `lan_subnets`: This is essentially the same feature present in the old subnet.conf. Each entry in this list defines a (LAN, private) subnet the server is in. Clients connecting from the same subnet, will be redirected to the LAN IP rather than the default public IP. The format has been simplified, and it only requires one IP and one subnet mask (as opposed to a character and a map server IP). - `allowed`: Allowed IPs are IP ranges a server (char to login or map to char) can connect from. Any attempt to connect as a server from an IP not included here, will fail. For convenience, a wildcard range (matching all possible IP addresses) has been provided (`0.0.0.0:0.0.0.0`), but it is very advisable to edit it to a more restrictive set. - `trusted`: Trusted IPs are IP ranges excluded from the IPban checks. This may be useful, for example, to exclude the server's own IP from ipbans, in case of false positives. Any IP ranges added to this list are also implicitly included in the allowed IP ranges.
2015-06-19Cleaned up some #includesHaru1-2/+0
Signed-off-by: Haru <haru@dotalux.com>
2015-06-19Removed ".." from include directivesHaru1-3/+3
- Include directives are now directory-independent. - This will allow building plugins from other directories in future. Signed-off-by: Haru <haru@dotalux.com>
2015-01-12Blocked compilation of plugins that use unavailable functionsHaru1-0/+2
- Rather than failing at runtime, plugins that try to access non-interfaced, unavailable functions or variables, will now show an error at compile-time. Signed-off-by: Haru <haru@dotalux.com>
2014-11-16Whitespace cleanup (no code changes)Haru1-1/+1
This includes, and is not limited to: mixed or wrong indentation, excess whitespace (horizontal and vertical), misalignment, trailing spaces. Signed-off-by: Haru <haru@dotalux.com>
2014-07-11Fixed reserved __identifier violationsHaru1-5/+5
- Complies with CERT DCL37-C - Fixes issue #293 (special thanks to elfring) Signed-off-by: Haru <haru@dotalux.com>
2014-05-10Re-commit of "Fixed order of includes in all source files"Haru1-6/+6
This reverts commit 94657284973f4037596bae468ebfbee5c217e02b.
2014-05-10Revert "Fixed order of includes in all source files"panikon1-6/+6
This reverts commit b6b3f58795288701d0e162d43fa6f0a47af913b3. Fixes issue 8184 http://hercules.ws/board/tracker/issue-8184-cart-related/
2014-05-09Fixed order of includes in all source filesHaru1-6/+6
- Changed order according to the (upcoming) code style guidelines. - Fixes several issues caused by missing headers when their include order is changed or in plugins. Signed-off-by: Haru <haru@dotalux.com>
2014-04-17Whitespace cleanupHaru1-1/+1
Replaced some tabs in the middle of the line with spaces. Thanks to KeiKun for pointing out #ifdef<tab>_COMMON_SYSINFO_H_ and making me realize how many of them were there. Signed-off-by: Haru <haru@dotalux.com>
2014-02-02Introducing HPM Datacheckshennetsind1-3/+3
http://hercules.ws/board/topic/4283-introducing-hpm-datacheck/ Signed-off-by: shennetsind <ind@henn.et>
2014-01-18Socket interfacedshennetsind1-61/+81
We designed this one with having as little changes required elsewhere in mind, thus the way most socket-related things are called has not changed. Plugins may now take a greater advantage of the socket features. Signed-off-by: shennetsind <ind@henn.et>
2013-11-19Sanitized and improved several macros through the codeHaru1-10/+11
- Sanitized all potentially unsafe macros (related eA:15259) - Improved some function-like macros to evaluate their argument only once and keep it in a temporary variable. This improves performance in the damage calculation related code. Signed-off-by: Haru <haru@dotalux.com>
2013-10-26Changed 'tick' variables to 64 bitHaru1-2/+1
- This fixes an issue with timers that stop working after about 24-49 days when the tick overflows (note that this may happen much earlier than that, and at hard to predict times, on some systems) - Updated the RDTSC help message in the configure script to also warn users about issues with SpeedStep enabled systems. - On Windows, tick() still has a resolution of 10~15ms (or even as low as 100ms on some systems). A TODO comment (thanks, Ai4rei) was added for a follow-up patch, as I want this one to be as small as possible) - Note: on Windows versions earlier than 6.x (Vista, Server 2008), the tick overflow issue is NOT fixed, since they don't support the function used to retrieve a 64 bit tick. This isn't a big issue, since those platforms are already - or going soon to be - out of their extended support period, and it's already advisable to upgrade, for other reasons. If you're the unfortunate user of such a system, it is recommended that you reboot your machine at least once every 49 days for Hercules to work reliably. - Note: To clear some doubts, since I've already been asked, this has absolutely NOTHING to do with 32/64 bit CPUs or OSes. It's all about a variable's size, not the size of registers of your CPU, and your 32bit CPU will be able to handle this just fine. Signed-off-by: Haru <haru@dotalux.com>
2013-08-08HPM Updateshennetsind1-1/+6
- Custom Packet Support - Custom Data Struct Support (currently append-able to map_session_data and socket_data) - Char Server Support - Login Server Support http://hercules.ws/board/topic/1934-hercules-plugin-manager-update/ Documentation will soon be updated in http://hercules.ws/wiki/HPM Signed-off-by: shennetsind <ind@henn.et>
2013-07-08Follow up ea5a413cdf7eec37141a7fe124cad38d66464ac4malufett1-1/+0
-where it breaks some skills & in game behaviors.
2013-07-05Implemented real-time server stats (in and out data and ram usage), made by ↵Matheus Macabu1-0/+1
Ai4rei; thanks! Another follow up, now to fix maximum value of chatdori setting; And fixed a bug where warping to a disable map would cause the map-server to crash. Signed-off-by: Matheus Macabu <mkbu95@gmail.com>
2013-06-28BG Queue Testingshennetsind1-1/+1
http://hercules.ws/board/topic/1302-bg-queue-debug/ Signed-off-by: shennetsind <ind@henn.et>
2013-05-30Memory Slasher - May 30 Patchshennetsind1-3/+11
http://hercules.ws/board/topic/928-memory-slasher-may-30-patch/ Signed-off-by: shennetsind <ind@henn.et>
2013-04-01Introducing socket_datasyncshennetsind1-0/+3
a user-friendly way to notify users when their servers are out of sync, e.g. http://hercules.ws/board/topic/328-intif-guild-data-size-error/ http://hercules.ws/board/topic/315-error-chrif-authok-data-size-mismatch-52224-52228/ Signed-off-by: shennetsind <ind@henn.et>
2012-12-05- Undid r16968: SVN Replaced with source:/trunk/src/@16966 (tid:74924).brianluau1-37/+38
[16969:16991/trunk/src/] will be re-committed in the next 24 hours. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16992 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-11-25Applied AStyle code formating as discussed on tid:74602.greenboxal21-38/+37
Removed /SAFESEH option from MSVC11 projects. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16968 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-10-31Fixed bugreport:6779 dropped between-server ping timers, replaced by a much ↵shennetsind1-0/+1
more reliable and performance-efficient on-demand flagging. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16854 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-09-16Introducing rAthena's anonymous data collector. This aims at improving ↵shennetsind1-1/+1
rAthena's overall features by letting we know which options are most and least used. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16785 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-06-03cleaned up windows header usage (added one central file to use windows / ↵blacksirius1-2/+1
winapi specific stuff, to be able to set the appropriate flags before including it correctly) Note: this may screw up mingw or cygwin building, ill fix it later... git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16219 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-02-19- reverted part of r15609 since apparently the issue is purely limited to ↵lordttseven1-1/+0
ZC_SKILLINFO_LIST, and this change breaks big storages git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@15613 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-02-18Follow up r15609, fixed gcc warningsshennetsind1-1/+1
git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@15610 54d463be-8e91-2dee-dedb-b68131a5f0ec
2012-02-18* adjusted socket_max_client_packet to a more accurate (if guessed) value to ↵lordttseven1-0/+1
prevent undefined client behavior (may only affect more recent clients?) + added a workaround for too large ZC_SKILLINFO_LIST packets resulting from all_skill group permission by sending excess skills one by one (bugreport:5348 and bugreport:5349) git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@15609 54d463be-8e91-2dee-dedb-b68131a5f0ec
2011-06-30Removed extra guards around cbasetypes.h include lines, since they give no ↵ultramage1-2/+0
performance improvement. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@14876 54d463be-8e91-2dee-dedb-b68131a5f0ec
2011-01-31* Various accumulated cleanups and fixes.ai4rei1-0/+2
- Improved the compile speed for files which include common/socket.h on windows builds (related r10471). - Moved FIFOSIZE_SERVERLINK define from common/mmo.h to common/socket.h (since it is a server connection FIFO size setting). - Fixed script command 'areamobuseskill' canceling monster's skill cast before it determined it's new target, which could be 'none' (bugreport:3272, since r13897). - Added a protection against attempts to read an empty backup / write to a full backup in skill_dance_switch as per TODO from r11347. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@14690 54d463be-8e91-2dee-dedb-b68131a5f0ec
2011-01-28* Resolving allow/deny IP rules not working (bugreport:2632).ai4rei1-0/+1
- Fixed 'mask' being filled with 'ip' when standard mask was specified (since r9647). - Fixed 'ip' and 'mask' (bit mask) being stored in wrong byte order (network order instead of host order) (since r10162). git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@14683 54d463be-8e91-2dee-dedb-b68131a5f0ec
2011-01-23* Some cleaning here and there.ai4rei1-6/+0
- Fixed a typo in basejob_baseclass_upper sample script (since r3893). - Fixed (for sake of consistency) usage of 'sd' before corresponding nullpo check in party_invite (bugreport:2752). - Fixed atcommand config reading would not cap level for charcommands, when using atcommand level as fallback (bugreport:2961, since r13409). - Removed orphaned clif_mob_hp declaration from clif.h (bugreport:2788, since r2092). - Removed unnecessary look-up in itemdb_available macro, as dummy_item is considered unavailable. - Removed leftover shortlist linked-list struct (followup to r10507). git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@14680 54d463be-8e91-2dee-dedb-b68131a5f0ec
2010-11-21* Added 64-bit variants of the socket and buffer I/O macros.ai4rei1-0/+4
git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@14479 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-12-17* Added flag.server to indicate interserver socketsultramage1-5/+8
- replaces the previous way (setting 'client_addr' to 0) git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@11930 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-10-17* Merged do_sendrecv() and do_parse() into do_sockets()ultramage1-9/+13
* Fixed subnet check message displaying incorrect ip addresses * client_addr will now be properly set to 0 for server connections * Removed socket code that attempts to cope with code bugs at runtime * Removed outdated copyright and version number from startup logo git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@11503 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-09-20* Merged the tmpsql branch:FlavioJS1-2/+2
- Abstraction for the sql code (sql.c/h). - New configure script and makefiles. - Restored txt zeny logging code. (r10814) - Rewrote mapserver's sql code - itemdb, mobdb, mapreg, logs. (r10814) - Fixed a precedence issue (&& and ) in char_sql/char.c. (r10833) - Improved db reading code a bit for consistency. (r11077) - Added separate atcommand for mail deletion. (r11077) - Corrected a few messages that said "new" instead of "unread". (r11077) - Broadcast (*) messages now use "*" as the target's name (not ""). (r11077) - Moved StringBuf code from utils.c/h to strlib.c/h. (r11084 r11117) - Some misc login server cleanups (reformatting etc). (r11136) - Corrected/modified some header entries. (r11141 r11147 11148) - Adjusted VS project files. (r11147) - Adjusted the way the sql charserver does item saving. (r11192) - Corrected usage of reserved keyword 'friend' in mmo.h. (r11192) git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@11245 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-05-28* RFIFOREST returns 0 when the session is eof (input data is implicitly ↵FlavioJS1-1/+1
discarted). * Reworked the player fame update: - would crash if an invalid pos was received - the wrong player could be updated on certain conditions git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10640 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-05-21- Removed the TURBO socket mechanism - it was in fact just using a variable ↵ultramage1-9/+0
to cache a few arithmetic operations (premature optimization) - Since the fifohead mess is gone, fixed leftover bad code structuring git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10599 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-05-09* Adapted the shortlist to use a static array instead of a linked list and ↵FlavioJS1-2/+3
enabled it. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10507 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-05-09* Added Buuyo-Tama's shortlist for send/eof sockets (defined out for now). FlavioJS1-0/+22
* Replaced toupper/tolower in ladmin by TOUPPER/TOLOWER defines. Shortlist: It's a list of sockets that have data to send and/or are ready for eof processing. It aims to reduce the amount of time spent on do_sendrecv, where it was spending ~13.5% of execution time on a server with 1k users at WoE. thanks to Buuyo-tama for the profile info and code git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10506 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-05-05* Removed the dependency on winsock.h for definitions, it's all winsock2.h now.FlavioJS1-4/+5
* Minor documentation/re-coding in chat.c. * More work on ticket #41. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10471 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-04-28* Tried and fixed md5 password encryptionultramage1-0/+2
- login_sql was doing stupid things like re-using one pre-generated key for all connections, and escaping binary md5 data (destroying it) - added missing code to login_txt when creating new accounts, showing that storing passwords as md5 hashes never really worked for txt - removed PASSWORDENC's conditional compilation effect, now just a flag - greatly simplified md5 password checking code, credits to eapp * login server will now bind only after finishing the init phase git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10384 54d463be-8e91-2dee-dedb-b68131a5f0ec
2007-04-20- Added ntows() to properly handle that client port number deficiencyultramage1-0/+1
- Adjusted the packet_db array for a more consistent look - Reorganized some entries in packet_db.txt - Fixed a small scripts_custom.conf typo git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@10296 54d463be-8e91-2dee-dedb-b68131a5f0ec