Finish v1, add some comments(docs)

readme.md

@@ -6,7 +6,7 @@ Waterwall!
 
 A simple core for tunneling and even direct user-server connections. based on high-performance, fully customizeable nodes for creating any kind of protocol without writing code for it.
 
-Currently on beta, to start please read the [docs](https://github.com/radkesvat/WaterWall/wiki). (The documents are currently in Persian language.)
+To start please read the [docs](https://github.com/radkesvat/WaterWall/wiki). (The documents are currently only in Persian language.)
 
 # Donation
 

tunnels/client/http2/http2_client.c

@@ -443,13 +443,13 @@ static void downStream(tunnel_t *self, context_t *c)
                 assert(false);
                 deleteHttp2Connection(con);
                 self->up->upStream(self->up, newFinContext(c->line));
+                reuseContextBuffer(c);
                 destroyContext(c);
                 return;
             }
 
             if (nghttp2_session_want_write(con->session) != 0)
             {
-
                 while (trySendRequest(self, con, 0, NULL))
                 {
                     if (! isAlive(c->line))

tunnels/server/http2/helpers.h

@@ -57,6 +57,8 @@ static void onStreamLineResumed(void *arg)
 static void onH2LinePaused(void *arg)
 {
     http2_server_con_state_t *con = (http2_server_con_state_t *) arg;
+    //  this kind of buffer actually reduces the speed instead of increasing it
+    // so its now disabled
     // ++(con->pause_counter);
     // if (con->pause_counter > 8)
     // {

tunnels/server/http2/http2_server.c

@@ -325,6 +325,7 @@ static void upStream(tunnel_t *self, context_t *c)
                 assert(false);
                 deleteHttp2Connection(con);
                 self->dw->downStream(self->dw, newFinContext(c->line));
+                reuseContextBuffer(c);
                 destroyContext(c);
                 return;
             }

ww/buffer_pool.h

@@ -21,7 +21,8 @@
 
     so the pool width is affected by ww memory profile
 
-    for performance reasons, this pool dose not inherit from generic_pool and 80% of the code is the same
+    for performance reasons, this pool dose not inherit from generic_pool, so 80% of the code is the same
+    but also it has its own differences ofcourse
     
 */
 

ww/buffer_stream.h

@@ -5,9 +5,12 @@
 #include <stdint.h>
 
 /*
-    buffers look like packets , you want a stream? you want to read the length you want?
-    then use buffer stream! put every buffer into this mix box and now you have a flow of bytes
-    which you can read but not consume, save and keep appending data to it untill it statisfy your protocol
+
+    This implements a simple container that holds buffers, some opmizations are also applied.
+
+    you can for example check byte index 1 or 5 of the buffers without concating them, then 
+    you'll be able to read only when your protocol is satisfied, the size you want
+    
 
 */
 

ww/cacert.h

@@ -1,8 +1,13 @@
 #ifndef CACERT_H
 #define CACERT_H
 
-// i don't have time to impleme... nah i just don't trust the user trusted certs :)
-// that also makes it much harder for the bad users to steal encrypted configs
+/*
+    Currently, waterwall uses its own CA and dose not trust user trusted certs
+    this also makes it much harder for the bad users to steal encrypted configs (IMO)
+
+    todo (hash verify), in order to make it even harder, do some hash verify of bytes
+*/
+
 
 extern unsigned char cacert_bytes[];
 extern unsigned int  cacert_len;

ww/idle_table.h

@@ -14,12 +14,14 @@
     you also can keep updating the item timeout
 
     The time checking has no cost and won't syscall at all, and the checking is synced by the
-    eventloop which by default wakes up every 100 ms. (debug note: current interval is set to 1s)
+    eventloop which by default wakes up every 100 ms. (debug note: current idletable interval is set to 1s)
 
     idle item is a threadlocal item, it belongs to the thread that created it
     and other threads must not change , remove or do anything to it
     because of that, tid parameter is required in order to find the item
 
+    -- valgrind unfriendly, since we required 64byte alignment, so it says "possibly lost"
+       but the pointer is saved in "memptr" field inside the object
 */
 
 struct idle_item_s;
@@ -38,6 +40,7 @@ struct idle_item_s
     uint8_t        tid;
     bool           removed;
 };
+
 typedef struct idle_table_s idle_table_t;
 
 idle_table_t *newIdleTable(hloop_t *loop);

ww/pipe_line.c

@@ -51,7 +51,7 @@ static void lock(pipe_line_t *pl)
     if (old_refc == 0)
     {
         // this should not happen, otherwise we must change memory order
-        // but i think its ok because threads synchronize around mutex
+        // but i think its ok because threads synchronize around the mutex in eventloop
         LOGF("PipeLine: thread-safety done incorrectly lock()");
         exit(1);
     }
@@ -66,7 +66,7 @@ static void unlock(pipe_line_t *pl)
         if (! atomic_load_explicit(&(pl->closed), memory_order_relaxed))
         {
             // this should not happen, otherwise we must change memory order
-            // but i think its ok because threads synchronize around mutex
+            // but i think its ok because threads synchronize around the mutex in eventloop
             LOGF("PipeLine: thread-safety done incorrectly unlock()");
             exit(1);
         }

ww/pipe_line.h

@@ -4,11 +4,17 @@
 
 /*
 
-
+    
+    <---------------  Down Line  ------------------>
+    
     ----------------------------|------------------|
                <- con ->           Pipe 1 thread 1 |
     ----------------------------|------------------|
 
+
+
+    <----------------  Up Line   ------------------>
+
     |------------------|----------------------------
     |  Pipe 1 thread 2              <- con ->
     |------------------|----------------------------
@@ -27,7 +33,8 @@
 
     i hope you don't use it, currently only used for halfduplex server since there were no other way...
 
-    -- valgrind unfriendly, mem align is not detected correctly , so it might log leaks
+    -- valgrind unfriendly, since we required 64byte alignment, so it says "possibly lost"
+       but the pointer is saved in "memptr" field inside the object
 
 */