[or-cvs] [tor/master 3/3] Make circuit_resume_edge_reading_helper treat streams more fairly.
nickm at torproject.org
nickm at torproject.org
Mon Sep 13 22:59:21 UTC 2010
Author: Nick Mathewson <nickm at torproject.org>
Date: Mon, 13 Sep 2010 17:04:13 -0400
Subject: Make circuit_resume_edge_reading_helper treat streams more fairly.
Commit: 424ca963ad7b08822b82491fbeba05438ddf5da2
Previously[*], the function would start with the first stream on the
circuit, and let it package as many cells as it wanted before
proceeding to the next stream in turn. If a circuit had many live
streams that all wanted to package data, the oldest would get
preference, and the newest would get ignored.
Now, we figure out how many cells we're willing to send per stream,
and try to allocate them fairly.
Roger diagnosed this in the comments for bug 1298.
[*] This bug has existed since before the first-ever public release
of Tor. It was added by r152 of Tor on 26 Jan 2003, which was
the first commit to implement streams (then called "topics").
This is not the oldest bug to be fixed in 0.2.2.x: that honor
goes to the windowing bug in r54, which got fixed in e50b7768 by
Roger with diagnosis by Karsten. This is, however, the most
long-lived bug to be fixed in 0.2.2.x: the r54 bug was fixed
2580 days after it was introduced, whereas I am writing this
commit message 2787 days after r152.
---
changes/bug1937 | 10 +++++
src/or/relay.c | 100 ++++++++++++++++++++++++++++++++++++++++++++++--------
2 files changed, 95 insertions(+), 15 deletions(-)
create mode 100644 changes/bug1937
diff --git a/changes/bug1937 b/changes/bug1937
new file mode 100644
index 0000000..b1fed6b
--- /dev/null
+++ b/changes/bug1937
@@ -0,0 +1,10 @@
+ o Major bugfixes
+ - When receiving a circuit-level SENDME for a blocked circuit, try
+ to package cells fairly from all the streams that had previously
+ been blocked on that circuit. Previously, we had started with
+ the oldest stream, and allowed each stream to potentially
+ exhaust the circuit's package window. This gave older streams
+ on any given circuit priority over newer ones. Fixes bug 1937.
+ Detected by Camilo Viecco. This bug was introduced before the
+ first Tor release, in svn commit r152: it is the new winner of
+ the longest-lived bug prize.
diff --git a/src/or/relay.c b/src/or/relay.c
index 7d3d780..125b2f7 100644
--- a/src/or/relay.c
+++ b/src/or/relay.c
@@ -57,6 +57,13 @@ static int circuit_queue_streams_are_blocked(circuit_t *circ);
static struct timeval cached_time_hires = {0, 0};
+/** Stop reading on edge connections when we have this many cells
+ * waiting on the appropriate queue. */
+#define CELL_QUEUE_HIGHWATER_SIZE 256
+/** Start reading from edge connections again when we get down to this many
+ * cells. */
+#define CELL_QUEUE_LOWWATER_SIZE 64
+
static void
tor_gettimeofday_cached(struct timeval *tv)
{
@@ -1471,31 +1478,101 @@ circuit_resume_edge_reading(circuit_t *circ, crypt_path_t *layer_hint)
* of a linked list of edge streams that should each be considered.
*/
static int
-circuit_resume_edge_reading_helper(edge_connection_t *conn,
+circuit_resume_edge_reading_helper(edge_connection_t *first_conn,
circuit_t *circ,
crypt_path_t *layer_hint)
{
- for ( ; conn; conn=conn->next_stream) {
- if (conn->_base.marked_for_close)
+ edge_connection_t *conn;
+ int n_streams, n_streams_left;
+ int packaged_this_round;
+ int cells_on_queue;
+ int cells_per_conn;
+
+ /* How many cells do we have space for? It will be the minimum of
+ * the number needed to exhaust the package window, and the minimum
+ * needed to fill the cell queue. */
+ int max_to_package = circ->package_window;
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ cells_on_queue = circ->n_conn_cells.n;
+ } else {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ cells_on_queue = or_circ->p_conn_cells.n;
+ }
+ if (CELL_QUEUE_HIGHWATER_SIZE - cells_on_queue < max_to_package)
+ max_to_package = CELL_QUEUE_HIGHWATER_SIZE - cells_on_queue;
+
+ /* Count how many non-marked streams there are that have anything on
+ * their inbuf, and enable reading on all of the connections. */
+ n_streams = 0;
+ for (conn=first_conn; conn; conn=conn->next_stream) {
+ if (conn->_base.marked_for_close || conn->package_window <= 0)
continue;
- if ((!layer_hint && conn->package_window > 0) ||
- (layer_hint && conn->package_window > 0 &&
- conn->cpath_layer == layer_hint)) {
+ if (!layer_hint || conn->cpath_layer == layer_hint) {
connection_start_reading(TO_CONN(conn));
+
+ if (buf_datalen(conn->_base.inbuf) > 0)
+ ++n_streams;
+ }
+ }
+
+ if (n_streams == 0) /* avoid divide-by-zero */
+ return 0;
+
+ again:
+
+ /* ??? turn this into a ceildiv function? */
+ cells_per_conn = (max_to_package + n_streams - 1 ) / n_streams;
+
+ packaged_this_round = 0;
+ n_streams_left = 0;
+
+ /* Iterate over all connections. Package up to cells_per_conn cells on
+ * each. Update packaged_this_round with the total number of cells
+ * packaged, and n_streams_left with the number that still have data to
+ * package.
+ */
+ for (conn=first_conn; conn; conn=conn->next_stream) {
+ if (conn->_base.marked_for_close || conn->package_window <= 0)
+ continue;
+ if (!layer_hint || conn->cpath_layer == layer_hint) {
+ int n = cells_per_conn, r;
/* handle whatever might still be on the inbuf */
- if (connection_edge_package_raw_inbuf(conn, 1, NULL)<0) {
- /* (We already sent an end cell if possible) */
+ r = connection_edge_package_raw_inbuf(conn, 1, &n);
+
+ /* Note how many we packaged */
+ packaged_this_round += (cells_per_conn-n);
+
+ if (r<0) {
+ /* Problem while packaging. (We already sent an end cell if
+ * possible) */
connection_mark_for_close(TO_CONN(conn));
continue;
}
+ /* If there's still data to read, we'll be coming back to this stream. */
+ if (buf_datalen(conn->_base.inbuf))
+ ++n_streams_left;
+
/* If the circuit won't accept any more data, return without looking
* at any more of the streams. Any connections that should be stopped
* have already been stopped by connection_edge_package_raw_inbuf. */
if (circuit_consider_stop_edge_reading(circ, layer_hint))
return -1;
+ /* XXXX should we also stop immediately if we fill up the cell queue?
+ * Probably. */
}
}
+
+ /* If we made progress, and we are willing to package more, and there are
+ * any streams left that want to package stuff... try again!
+ */
+ if (packaged_this_round && packaged_this_round < max_to_package &&
+ n_streams_left) {
+ max_to_package -= packaged_this_round;
+ n_streams = n_streams_left;
+ goto again;
+ }
+
return 0;
}
@@ -1564,13 +1641,6 @@ circuit_consider_sending_sendme(circuit_t *circ, crypt_path_t *layer_hint)
}
}
-/** Stop reading on edge connections when we have this many cells
- * waiting on the appropriate queue. */
-#define CELL_QUEUE_HIGHWATER_SIZE 256
-/** Start reading from edge connections again when we get down to this many
- * cells. */
-#define CELL_QUEUE_LOWWATER_SIZE 64
-
#ifdef ACTIVE_CIRCUITS_PARANOIA
#define assert_active_circuits_ok_paranoid(conn) \
assert_active_circuits_ok(conn)
--
1.7.1
More information about the tor-commits
mailing list