First shot on chunked transfers

opentracker.c

@@ -79,6 +79,7 @@ static void defaul_signal_handlers( void ) {
   sigaddset (&signal_mask, SIGPIPE);
   sigaddset (&signal_mask, SIGHUP);
   sigaddset (&signal_mask, SIGINT);
+  sigaddset (&signal_mask, SIGALRM);
   pthread_sigmask (SIG_BLOCK, &signal_mask, NULL);
 }
 
@@ -90,7 +91,7 @@ static void install_signal_handlers( void ) {
   sa.sa_handler = signal_handler;
   sigemptyset(&sa.sa_mask);
   sa.sa_flags = SA_RESTART;
-  if ((sigaction(SIGINT, &sa, NULL) == -1))
+  if ((sigaction(SIGINT, &sa, NULL) == -1) || (sigaction(SIGALRM, &sa, NULL) == -1) )
     panic( "install_signal_handlers" );
 
   sigaddset (&signal_mask, SIGINT);
@@ -208,15 +209,23 @@ static void handle_read( const int64 sock, struct ot_workstruct *ws ) {
 static void handle_write( const int64 sock ) {
   struct http_data* cookie=io_getcookie( sock );
   size_t i;
+  int    chunked = 0;
 
   /* Look for the first io_batch still containing bytes to write */
-  if( cookie )
-    for( i = 0; i < cookie->batches; ++i )
+  if( cookie ) {
+    if( cookie->flag & STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER )
+      chunked = 1;
+
+    for( i = 0; i < cookie->batches; ++i ) {
+      fprintf(stderr, "handle_write inspects batch %d of %d (bytes left: %d)\n", i, cookie->batches, cookie->batch[i].bytesleft);
       if( cookie->batch[i].bytesleft ) {
         int64 res = iob_send( sock, cookie->batch + i );
 
-        if( res == -3 )
-          break;
+        fprintf(stderr, "handle_write yields res %lld when trying to iob_send\n", res);
+        if( res == -3 ) {
+          handle_dead( sock );
+          return;
+        }
 
         if( !cookie->batch[i].bytesleft )
           continue;
@@ -224,8 +233,17 @@ static void handle_write( const int64 sock ) {
         if( res == -1 || res > 0 || i < cookie->batches - 1 )
           return;
       }
+    }
+  }
 
+  /* In a chunked transfer after all batches accumulated have been sent, wait for the next one */
+  if( chunked ) {
+//fprintf( stderr, "handle_write is STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER => dont want write on sock %lld\n", sock);
+    //io_dontwantwrite( sock );
+  } else {
+fprintf( stderr, "handle_write is STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER => handle dead on sock %lld\n", sock);
     handle_dead( sock );
+  }
 }
 
 static void handle_accept( const int64 serversocket ) {
@@ -266,7 +284,7 @@ static void * server_mainloop( void * args ) {
   struct ot_workstruct ws;
   time_t next_timeout_check = g_now_seconds + OT_CLIENT_TIMEOUT_CHECKINTERVAL;
   struct iovec *iovector;
-  int    iovec_entries;
+  int    iovec_entries, is_partial;
 
   (void)args;
 
@@ -305,8 +323,8 @@ static void * server_mainloop( void * args ) {
         handle_read( sock, &ws );
     }
 
-    while( ( sock = mutex_workqueue_popresult( &iovec_entries, &iovector ) ) != -1 )
-      http_sendiovecdata( sock, &ws, iovec_entries, iovector );
+    while( ( sock = mutex_workqueue_popresult( &iovec_entries, &iovector, &is_partial ) ) != -1 )
+      http_sendiovecdata( sock, &ws, iovec_entries, iovector, is_partial );
 
     while( ( sock = io_canwrite( ) ) != -1 )
       handle_write( sock );
@@ -318,9 +336,6 @@ static void * server_mainloop( void * args ) {
     }
 
     livesync_ticker();
-
-    /* Enforce setting the clock */
-    signal_handler( SIGALRM );
   }
   return 0;
 }

ot_fullscrape.c

@@ -36,9 +36,9 @@
 #define OT_SCRAPE_MAXENTRYLEN 256
 
 /* Forward declaration */
-static void fullscrape_make( int *iovec_entries, struct iovec **iovector, ot_tasktype mode );
+static void fullscrape_make( int taskid, ot_tasktype mode);
 #ifdef WANT_COMPRESSION_GZIP
-static void fullscrape_make_gzip( int *iovec_entries, struct iovec **iovector, ot_tasktype mode );
+static void fullscrape_make_gzip( int taskid, ot_tasktype mode);
 #endif
 
 /* Converter function from memory to human readable hex strings
@@ -49,9 +49,6 @@ static char*to_hex(char*d,uint8_t*s){char*m="0123456789ABCDEF";char *t=d;char*e=
    It grabs tasks from mutex_tasklist and delivers results back
 */
 static void * fullscrape_worker( void * args ) {
-  int iovec_entries;
-  struct iovec *iovector;
-
   (void) args;
 
   while( g_opentracker_running ) {
@@ -59,12 +56,11 @@ static void * fullscrape_worker( void * args ) {
     ot_taskid   taskid   = mutex_workqueue_poptask( &tasktype );
 #ifdef WANT_COMPRESSION_GZIP
     if (tasktype & TASK_FLAG_GZIP)
-        fullscrape_make_gzip( &iovec_entries, &iovector, tasktype );
+      fullscrape_make_gzip( taskid, tasktype );
     else
 #endif
-        fullscrape_make( &iovec_entries, &iovector, tasktype );
-    if( mutex_workqueue_pushresult( taskid, iovec_entries, iovector ) )
-      iovec_free( &iovec_entries, &iovector );
+      fullscrape_make( taskid, tasktype );
+    mutex_workqueue_pushchunked( taskid, NULL );
   }
   return NULL;
 }
@@ -123,14 +119,13 @@ static char * fullscrape_write_one( ot_tasktype mode, char *r, ot_torrent *torre
     return r;
 }
 
-static void fullscrape_make( int *iovec_entries, struct iovec **iovector, ot_tasktype mode ) {
+static void fullscrape_make( int taskid, ot_tasktype mode ) {
   int      bucket;
   char    *r, *re;
+  struct iovec iovector = { NULL, 0 };
 
   /* Setup return vector... */
-  *iovec_entries = 0;
-  *iovector = NULL;
-  r = iovec_increase( iovec_entries, iovector, OT_SCRAPE_CHUNK_SIZE );
+  r = iovector.iov_base = malloc( OT_SCRAPE_CHUNK_SIZE );
   if( !r )
     return;
 
@@ -152,8 +147,14 @@ static void fullscrape_make( int *iovec_entries, struct iovec **iovector, ot_tas
       r = fullscrape_write_one( mode, r, torrents+i, &torrents[i].hash );
 
       if( r > re) {
+        iovector.iov_len = r - (char *)iovector.iov_base;
+
+        if (mutex_workqueue_pushchunked(taskid, &iovector) ) {
+          free(iovector.iov_base);
+          return mutex_bucket_unlock( bucket, 0 );
+        }
         /* Allocate a fresh output buffer at the end of our buffers list */
-        r = iovec_fix_increase_or_free( iovec_entries, iovector, r, OT_SCRAPE_CHUNK_SIZE );
+        r = iovector.iov_base = malloc( OT_SCRAPE_CHUNK_SIZE );
         if( !r )
           return mutex_bucket_unlock( bucket, 0 );
 
@@ -174,7 +175,9 @@ static void fullscrape_make( int *iovec_entries, struct iovec **iovector, ot_tas
     r += sprintf( r, "ee" );
 
   /* Release unused memory in current output buffer */
-  iovec_fixlast( iovec_entries, iovector, r );
+  iovector.iov_len = r - (char *)iovector.iov_base;
+  if( mutex_workqueue_pushchunked(taskid, &iovector) )
+    free(iovector.iov_base);
 }
 
 #ifdef WANT_COMPRESSION_GZIP

ot_http.c

@@ -121,9 +121,10 @@ ssize_t http_issue_error( const int64 sock, struct ot_workstruct *ws, int code )
   return ws->reply_size = -2;
 }
 
-ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iovec_entries, struct iovec *iovector ) {
+ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iovec_entries, struct iovec *iovector, int is_partial ) {
   struct http_data *cookie = io_getcookie( sock );
   char *header;
+  const char *encoding = "";
   int i;
   size_t header_size, size = iovec_length( &iovec_entries, (const struct iovec **)&iovector );
   tai6464 t;
@@ -140,10 +141,9 @@ ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iove
   /* If we came here, wait for the answer is over */
   cookie->flag &= ~STRUCT_HTTP_FLAG_WAITINGFORTASK;
 
-  /* Our answers never are 0 vectors. Return an error. */
-  if( !iovec_entries ) {
-    HTTPERROR_500;
-  }
+fprintf(stderr, "http_sendiovecdata sending %d iovec entries found cookie->batch == %p\n", iovec_entries, cookie->batch);
+
+  if( iovec_entries ) {
 
     /* Prepare space for http header */
     header = malloc( SUCCESS_HTTP_HEADER_LENGTH + SUCCESS_HTTP_HEADER_LENGTH_CONTENT_ENCODING );
@@ -153,11 +153,19 @@ ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iove
     }
 
     if( cookie->flag & STRUCT_HTTP_FLAG_GZIP )
-    header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Encoding: gzip\r\nContent-Length: %zd\r\n\r\n", size );
+      encoding = "Content-Encoding: gzip\r\n";
     else if( cookie->flag & STRUCT_HTTP_FLAG_BZIP2 )
-    header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Encoding: bzip2\r\nContent-Length: %zd\r\n\r\n", size );
-  else
-    header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\nContent-Length: %zd\r\n\r\n", size );
+      encoding = "Content-Encoding: bzip2\r\n";
+
+    if( !(cookie->flag & STRUCT_HTTP_FLAG_CHUNKED) )
+      header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\n%sContent-Length: %zd\r\n\r\n", encoding, size );
+    else {
+      if ( !(cookie->flag & STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER )) {
+        header_size = sprintf( header, "HTTP/1.0 200 OK\r\nContent-Type: text/plain\r\n%sTransfer-Encoding: chunked\r\n\r\n%zx\r\n", encoding, size );
+        cookie->flag |= STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER;
+      } else
+        header_size = sprintf( header, "%zx\r\n", size );
+    }
 
     if (!cookie->batch ) {
       cookie->batch = malloc( sizeof(io_batch) );
@@ -172,22 +180,33 @@ ssize_t http_sendiovecdata( const int64 sock, struct ot_workstruct *ws, int iove
 
       /* If the current batch's limit is reached, try to reallocate a new batch to work on */
       if( current->bytesleft > OT_BATCH_LIMIT ) {
-        io_batch * new_batch = realloc( current, (cookie->batches + 1) * sizeof(io_batch) );
+fprintf(stderr, "http_sendiovecdata found batch above limit: %zd\n", current->bytesleft);
+        io_batch * new_batch = realloc( cookie->batch, (cookie->batches + 1) * sizeof(io_batch) );
         if( new_batch ) {
-            cookie->batches++;
-            current = cookie->batch = new_batch;
+          cookie->batch = new_batch;
+          current = cookie->batch + cookie->batches++;
           memset( current, 0, sizeof(io_batch) );
         }
       }
-
+fprintf(stderr, "http_sendiovecdata calling iob_addbuf_free with %zd\n", iovector[i].iov_len);
       iob_addbuf_free( current, iovector[i].iov_base, iovector[i].iov_len );
     }
     free( iovector );
+    if ( cookie->flag & STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER )
+      iob_addbuf(cookie->batch + cookie->batches - 1, "\r\n", 2);
+  }
+
+  if ((cookie->flag & STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER) && cookie->batch && !is_partial) {
+fprintf(stderr, "http_sendiovecdata adds a terminating 0 size buffer to batch\n");
+    iob_addbuf(cookie->batch + cookie->batches - 1, "0\r\n\r\n", 5);
+    cookie->flag &= ~STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER;
+  }
 
   /* writeable sockets timeout after 10 minutes */
   taia_now( &t ); taia_addsec( &t, &t, OT_CLIENT_TIMEOUT_SEND );
   io_timeout( sock, t );
   io_dontwantread( sock );
+fprintf (stderr, "http_sendiovecdata marks socket %lld as wantwrite\n", sock);
   io_wantwrite( sock );
   return 0;
 }
@@ -254,7 +273,7 @@ static const ot_keywords keywords_format[] =
 #endif
 #endif
     /* Pass this task to the worker thread */
-    cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK;
+    cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK | STRUCT_HTTP_FLAG_CHUNKED;
 
     /* Clients waiting for us should not easily timeout */
     taia_uint( &t, 0 ); io_timeout( sock, t );
@@ -325,7 +344,7 @@ static ssize_t http_handle_fullscrape( const int64 sock, struct ot_workstruct *w
 #endif
 
   /* Pass this task to the worker thread */
-  cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK;
+  cookie->flag |= STRUCT_HTTP_FLAG_WAITINGFORTASK | STRUCT_HTTP_FLAG_CHUNKED;
   /* Clients waiting for us should not easily timeout */
   taia_uint( &t, 0 ); io_timeout( sock, t );
   fullscrape_deliver( sock, TASK_FULLSCRAPE | format );

ot_http.h

@@ -9,7 +9,9 @@
 typedef enum {
   STRUCT_HTTP_FLAG_WAITINGFORTASK      = 1,
   STRUCT_HTTP_FLAG_GZIP                = 2,
-  STRUCT_HTTP_FLAG_BZIP2          = 4
+  STRUCT_HTTP_FLAG_BZIP2               = 4,
+  STRUCT_HTTP_FLAG_CHUNKED             = 8,
+  STRUCT_HTTP_FLAG_CHUNKED_IN_TRANSFER = 16
 } STRUCT_HTTP_FLAG;
 
 struct http_data {
@@ -21,7 +23,7 @@ struct http_data {
 };
 
 ssize_t http_handle_request( const int64 s, struct ot_workstruct *ws );
-ssize_t http_sendiovecdata( const int64 s, struct ot_workstruct *ws, int iovec_entries, struct iovec *iovector );
+ssize_t http_sendiovecdata( const int64 s, struct ot_workstruct *ws, int iovec_entries, struct iovec *iovector, int is_partial );
 ssize_t http_issue_error( const int64 s, struct ot_workstruct *ws, int code );
 
 extern char   *g_stats_path;

ot_iovec.c

@@ -35,6 +35,26 @@ void *iovec_increase( int *iovec_entries, struct iovec **iovector, size_t new_al
   return new_data;
 }
 
+void *iovec_append( int *iovec_entries, struct iovec **iovector, struct iovec *append_iovector) {
+  int new_entries = *iovec_entries + 1;
+  struct iovec *new_vec = realloc( *iovector, new_entries * sizeof( struct iovec ) );
+  if( !new_vec )
+    return NULL;
+
+  /* Take over data from appended iovec */
+  new_vec[*iovec_entries].iov_base = append_iovector->iov_base;
+  new_vec[*iovec_entries].iov_len  = append_iovector->iov_len;
+
+  append_iovector->iov_base = NULL;
+  append_iovector->iov_len = 0;
+
+  *iovector = new_vec;
+  *iovec_entries = new_entries;
+
+  return new_vec;
+}
+
+
 void iovec_free( int *iovec_entries, struct iovec **iovector ) {
   int i;
   for( i=0; i<*iovec_entries; ++i )
@@ -64,7 +84,6 @@ void *iovec_fix_increase_or_free( int *iovec_entries, struct iovec **iovector, v
   return new_data;
 }
 
-
 size_t iovec_length( const int *iovec_entries, const struct iovec **iovector ) {
   size_t length = 0;
   int i;

ot_iovec.h

@@ -9,6 +9,7 @@
 #include <sys/uio.h>
 
 void  *iovec_increase( int *iovec_entries, struct iovec **iovector, size_t new_alloc );
+void  *iovec_append( int *iovec_entries, struct iovec **iovector, struct iovec *append_iovector );
 void   iovec_fixlast( int *iovec_entries, struct iovec **iovector, void *last_ptr );
 void   iovec_free( int *iovec_entries, struct iovec **iovector );
 

ot_mutex.c

@@ -17,6 +17,7 @@
 
 /* Opentracker */
 #include "trackerlogic.h"
+#include "ot_iovec.h"
 #include "ot_mutex.h"
 #include "ot_stats.h"
 
@@ -194,23 +195,66 @@ int mutex_workqueue_pushresult( ot_taskid taskid, int iovec_entries, struct iove
   return task ? 0 : -1;
 }
 
-int64 mutex_workqueue_popresult( int *iovec_entries, struct iovec ** iovec ) {
+int mutex_workqueue_pushchunked(ot_taskid taskid, struct iovec *iovec) {
+  struct ot_task * task;
+  const char byte = 'o';
+
+  /* Want exclusive access to tasklist */
+  pthread_mutex_lock( &tasklist_mutex );
+
+  for (task = tasklist; task; task = task->next)
+    if (task->taskid == taskid) {
+      if( iovec ) {
+fprintf(stderr, "mutex_workqueue_pushchunked pushing on taskid %d\n", taskid);
+        if (!iovec_append(&task->iovec_entries, &task->iovec, iovec) )
+          return -1;
+        task->tasktype = TASK_DONE_PARTIAL;
+      } else {
+fprintf(stderr, "mutex_workqueue_pushchunked finished taskid %d\n", taskid);
+        task->tasktype = TASK_DONE;
+      }
+      break;
+    }
+
+  /* Release lock */
+  pthread_mutex_unlock( &tasklist_mutex );
+
+  io_trywrite( g_self_pipe[1], &byte, 1 );
+if(!task)
+fprintf(stderr, "mutex_workqueue_pushchunked taskid %d not found\n", taskid);
+
+  /* Indicate whether the worker has to throw away results */
+  return task ? 0 : -1;
+}
+
+
+int64 mutex_workqueue_popresult( int *iovec_entries, struct iovec ** iovec, int *is_partial ) {
   struct ot_task ** task;
   int64 sock = -1;
 
+  *is_partial = 0;
+
   /* Want exclusive access to tasklist */
   pthread_mutex_lock( &tasklist_mutex );
 
   for (task = &tasklist; *task; task = &((*task)->next))
-    if ((*task)->tasktype == TASK_DONE) {
+    if (((*task)->tasktype & TASK_CLASS_MASK ) == TASK_DONE) {
       struct ot_task *ptask = *task;
-
+fprintf(stderr, "Got task %d type %d with %d entries\n", (*task)->taskid, (*task)->tasktype, ptask->iovec_entries);
       *iovec_entries = ptask->iovec_entries;
       *iovec         = ptask->iovec;
       sock           = ptask->sock;
 
+      if ((*task)->tasktype == TASK_DONE) {
         *task = ptask->next;
         free( ptask );
+      } else {
+        ptask->iovec_entries = 0;
+        ptask->iovec = NULL;
+        *is_partial = 1;
+        /* Prevent task from showing up immediately again unless new data was added */
+        (*task)->tasktype = TASK_FULLSCRAPE;
+      }
       break;
     }
 

ot_mutex.h

@@ -54,9 +54,11 @@ typedef enum {
   TASK_DMEM                        = 0x0300,
 
   TASK_DONE                        = 0x0f00,
+  TASK_DONE_PARTIAL                = 0x0f01,
 
   TASK_FLAG_GZIP                   = 0x1000,
   TASK_FLAG_BZIP2                  = 0x2000,
+  TASK_FLAG_CHUNKED                = 0x4000,
 
   TASK_TASK_MASK                   = 0x0fff,
   TASK_CLASS_MASK                  = 0x0f00,
@@ -70,6 +72,7 @@ void      mutex_workqueue_canceltask( int64 sock );
 void      mutex_workqueue_pushsuccess( ot_taskid taskid );
 ot_taskid mutex_workqueue_poptask( ot_tasktype *tasktype );
 int       mutex_workqueue_pushresult( ot_taskid taskid, int iovec_entries, struct iovec *iovector );
-int64     mutex_workqueue_popresult( int *iovec_entries, struct iovec ** iovector );
+int       mutex_workqueue_pushchunked(ot_taskid taskid, struct iovec *iovec);
+int64     mutex_workqueue_popresult( int *iovec_entries, struct iovec ** iovector, int *is_partial );
 
 #endif

trackerlogic.c

@@ -517,6 +517,29 @@ size_t   peer_size_from_peer6(ot_peer6 *peer) {
   return OT_PEER_SIZE4;
 }
 
+void trackerlogic_add_random_torrents(size_t amount) {
+  struct ot_workstruct ws;
+  memset( &ws, 0, sizeof(ws) );
+
+  ws.inbuf=malloc(G_INBUF_SIZE);
+  ws.outbuf=malloc(G_OUTBUF_SIZE);
+  ws.reply=ws.outbuf;
+  ws.hash=ws.inbuf;
+
+  while( amount-- ) {
+    arc4random_buf(ws.hash, sizeof(ot_hash));
+    arc4random_buf(&ws.peer, sizeof(ws.peer));
+
+    OT_PEERFLAG(ws.peer) &= PEER_FLAG_SEEDING | PEER_FLAG_COMPLETED | PEER_FLAG_STOPPED;
+
+    add_peer_to_torrent_and_return_peers( FLAG_TCP, &ws, 1 );
+  }
+
+  free(ws.inbuf);
+  free(ws.outbuf);
+}
+
+
 void exerr( char * message ) {
   fprintf( stderr, "%s\n", message );
   exit( 111 );

trackerlogic.h

@@ -190,6 +190,7 @@ size_t  remove_peer_from_torrent( PROTO_FLAG proto, struct ot_workstruct *ws );
 size_t  return_tcp_scrape_for_torrent( ot_hash const *hash_list, int amount, char *reply );
 size_t  return_udp_scrape_for_torrent( ot_hash const hash, char *reply );
 void    add_torrent_from_saved_state( ot_hash const hash, ot_time base, size_t down_count );
+void    trackerlogic_add_random_torrents(size_t amount);
 
 /* torrent iterator */
 void iterate_all_torrents( int (*for_each)( ot_torrent* torrent, uintptr_t data ), uintptr_t data );