diff options
author | brain <brain@e03df62e-2008-0410-955e-edbf42e46eb7> | 2006-07-22 01:06:55 +0000 |
---|---|---|
committer | brain <brain@e03df62e-2008-0410-955e-edbf42e46eb7> | 2006-07-22 01:06:55 +0000 |
commit | c4adf4b98d3d89739883263cd9ec49544c55311d (patch) | |
tree | bc11a6cc065beb649f32e50a33282a65d725ee6b /src | |
parent | 1484a054870bdfe94346057053d5c8e48a708232 (diff) |
Huge ass comment about how this works
git-svn-id: http://svn.inspircd.org/repository/trunk/inspircd@4488 e03df62e-2008-0410-955e-edbf42e46eb7
Diffstat (limited to 'src')
-rw-r--r-- | src/modules/extra/m_mysql.cpp | 35 |
1 files changed, 35 insertions, 0 deletions
diff --git a/src/modules/extra/m_mysql.cpp b/src/modules/extra/m_mysql.cpp index 34ca1ec17..57083f60a 100644 --- a/src/modules/extra/m_mysql.cpp +++ b/src/modules/extra/m_mysql.cpp @@ -33,6 +33,41 @@ using namespace std; /* $LinkerFlags: `mysql_config --libs_r` `perl ../mysql_rpath.pl` */ +/* THE NONBLOCKING MYSQL API! + * + * MySQL provides no nonblocking (asyncronous) API of its own, and its developers recommend + * that instead, you should thread your program. This is what i've done here to allow for + * asyncronous SQL requests via mysql. The way this works is as follows: + * + * The module spawns a thread via pthreads, and performs its mysql queries in this thread, + * using a queue with priorities. There is a mutex on either end which prevents two threads + * adjusting the queue at the same time, and crashing the ircd. Every 50 milliseconds, the + * worker thread wakes up, and checks if there is a request at the head of its queue. + * If there is, it processes this request, blocking the worker thread but leaving the ircd + * thread to go about its business as usual. During this period, the ircd thread is able + * to insert futher pending requests into the queue. + * + * Once the processing of a request is complete, it is removed from the incoming queue to + * an outgoing queue, and initialized as a 'response'. The worker thread then signals the + * ircd thread (via a loopback socket) of the fact a result is available, by sending the + * connection ID through the connection. + * + * The ircd thread then mutexes the queue once more, reads the outbound response off the head + * of the queue, and sends it on its way to the original calling module. + * + * XXX: You might be asking "why doesnt he just send the response from within the worker thread?" + * The answer to this is simple. The majority of InspIRCd, and in fact most ircd's are not + * threadsafe. This module is designed to be threadsafe and is careful with its use of threads, + * however, if we were to call a module's OnRequest even from within a thread which was not the + * one the module was originally instantiated upon, there is a chance of all hell breaking loose + * if a module is ever put in a re-enterant state (stack corruption could occur, crashes, data + * corruption, and worse, so DONT think about it until the day comes when InspIRCd is 100% + * gauranteed threadsafe!) + * + * For a diagram of this system please see http://www.inspircd.org/wiki/Mysql2 + */ + + class SQLConnection; class Notifier; |