From: Roger Dingledine <arma@torproject.org>
Date: Mon, 10 Mar 2003 22:40:02 +0000 (+0000)
Subject: pack into data cells more intelligently
X-Git-Tag: tor-0.0.2pre8~295
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=fb2f4a0418d1bf16d28a14b4dc80c417e61917e8;p=thirdparty%2Ftor.git

pack into data cells more intelligently


svn:r173
---

diff --git a/src/or/buffers.c b/src/or/buffers.c
index 5bdc8e73eb..424e2fe9e6 100644
--- a/src/or/buffers.c
+++ b/src/or/buffers.c
@@ -52,9 +52,14 @@ int read_to_buf(int s, int at_most, char **buf, int *buflen, int *buf_datalen, i
 
   if(!options.LinkPadding && at_most > 10*sizeof(cell_t)) {
     /* if no linkpadding: do a rudimentary round-robin so one
-     * connection can't hog an outgoing connection
+     * connection can't hog a thickpipe
+     */
+    at_most = 10*(CELL_PAYLOAD_SIZE - TOPIC_HEADER_SIZE);
+    /* XXX this still isn't perfect. now we read 10 data payloads per read --
+     * but if we're reading from a connection that speaks cells, we always
+     * read a partial cell from the network and can't process it yet. Good
+     * enough for now though. (And maybe best, to stress our code more.)
      */
-    at_most = 10*sizeof(cell_t); /* FIXME should be 10* size of usable payload */
   }
 
 //  log(LOG_DEBUG,"read_to_buf(): reading at most %d bytes.",at_most);