April 2011 - tor-commits - lists.torproject.org

[tor/master] Add a missing cast to silence the compiler
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 8d81831d715adf9cb374100e23cc2b118efe3101 Author: Sebastian Hahn <sebastian(a)torproject.org> Date: Mon Mar 28 18:20:50 2011 +0200 Add a missing cast to silence the compiler --- src/or/connection.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/or/connection.c b/src/or/connection.c index 084237d..6e7bbd5 100644 --- a/src/or/connection.c +++ b/src/or/connection.c @@ -2790,7 +2790,7 @@ connection_handle_write_impl(connection_t *conn, int force) edge_connection_t *edge_conn = TO_EDGE_CONN(conn); /* Check for overflow: */ if (PREDICT_LIKELY(UINT32_MAX - edge_conn->n_written > n_written)) - edge_conn->n_written += n_written; + edge_conn->n_written += (int)n_written; else edge_conn->n_written = UINT32_MAX; }

1 0

[tor/master] Fix handling of StreamID exhaustion.
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 432734279d3688fafb466a23f43585ff509ff693 Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 17:57:15 2011 -0400 Fix handling of StreamID exhaustion. Since svn r1475/git 5b6099e8 in tor-0.0.6, we have responded to an exhaustion of all 65535 stream IDs on a circuit by marking that circuit for close. That's not the right response. Instead, we should mark the circuit as "too dirty for new circuits". Of course in reality this isn't really right either. If somebody has managed to cram 65535 streams onto a circuit, the circuit is probably not going to work well for any of those streams, so maybe we should be limiting the number of streams on an origin circuit concurrently. Also, closing the stream in this case is probably the wrong thing to do as well, but fixing that can also wait. --- changes/full_ap_circuits | 6 ++++++ src/or/connection_edge.c | 18 ++++++++++++++---- 2 files changed, 20 insertions(+), 4 deletions(-) diff --git a/changes/full_ap_circuits b/changes/full_ap_circuits new file mode 100644 index 0000000..379a1a1 --- /dev/null +++ b/changes/full_ap_circuits @@ -0,0 +1,6 @@ + o Minor bugfixes + - When a client finds that an origin circuit has run out of 16-bit + stream IDs, we now mark it as unusable for new streams. + Previously, we would try to close the entire circuit. Bugfix on + Tor version 0.0.6. + diff --git a/src/or/connection_edge.c b/src/or/connection_edge.c index af0cfbe..72e2c8a 100644 --- a/src/or/connection_edge.c +++ b/src/or/connection_edge.c @@ -2164,9 +2164,14 @@ connection_ap_handshake_send_begin(edge_connection_t *ap_conn) ap_conn->stream_id = get_unique_stream_id_by_circ(circ); if (ap_conn->stream_id==0) { + /* XXXX023 Instead of closing this stream, we should make it get + * retried on another circuit. */ connection_mark_unattached_ap(ap_conn, END_STREAM_REASON_INTERNAL); - /*XXXX022 _close_ the circuit because it's full? That sounds dumb. */ - circuit_mark_for_close(TO_CIRCUIT(circ), END_CIRC_REASON_RESOURCELIMIT); + + /* Mark this circuit "unusable for new streams". */ + /* XXXX023 this is a kludgy way to do this. */ + tor_assert(circ->_base.timestamp_dirty); + circ->_base.timestamp_dirty -= get_options()->MaxCircuitDirtiness; return -1; } @@ -2224,9 +2229,14 @@ connection_ap_handshake_send_resolve(edge_connection_t *ap_conn) ap_conn->stream_id = get_unique_stream_id_by_circ(circ); if (ap_conn->stream_id==0) { + /* XXXX023 Instead of closing this stream, we should make it get + * retried on another circuit. */ connection_mark_unattached_ap(ap_conn, END_STREAM_REASON_INTERNAL); - /*XXXX022 _close_ the circuit because it's full? That sounds dumb. */ - circuit_mark_for_close(TO_CIRCUIT(circ), END_CIRC_REASON_RESOURCELIMIT); + + /* Mark this circuit "unusable for new streams". */ + /* XXXX023 this is a kludgy way to do this. */ + tor_assert(circ->_base.timestamp_dirty); + circ->_base.timestamp_dirty -= get_options()->MaxCircuitDirtiness; return -1; }

1 0

[tor/master] Merge remote-tracking branch 'public/xxx_fixups' into maint-0.2.2
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit ba0cd8094f8e6ae0113ad69958d9d0973bb1f2c3 Merge: 118d8ff fc64783 Author: Nick Mathewson <nickm(a)torproject.org> Date: Thu Apr 7 12:03:04 2011 -0400 Merge remote-tracking branch 'public/xxx_fixups' into maint-0.2.2 Conflicts: src/or/or.h changes/bug539_removal | 6 +++ changes/connect_err_reporting | 6 +++ changes/count_overflow | 5 +++ changes/full_ap_circuits | 6 +++ changes/kill_ftime | 7 ++++ changes/noroute | 5 +++ src/common/address.c | 20 ++++++----- src/common/compat_libevent.c | 17 +++------ src/common/memarea.c | 7 +++- src/common/util.c | 77 ----------------------------------------- src/common/util.h | 10 ----- src/or/buffers.c | 6 ++-- src/or/circuitbuild.c | 9 +++-- src/or/circuitlist.c | 3 ++ src/or/circuituse.c | 5 ++- src/or/config.c | 6 ++-- src/or/connection.c | 36 ++++++++++++------- src/or/connection_edge.c | 30 +++++++++++----- src/or/connection_or.c | 6 +++- src/or/directory.c | 37 ++++++++------------ src/or/dirserv.c | 11 ++++-- src/or/dns.c | 2 +- src/or/dnsserv.c | 2 +- src/or/eventdns.c | 2 +- src/or/geoip.c | 5 ++- src/or/networkstatus.c | 11 ++++-- src/or/or.h | 12 +++++-- src/or/reasons.c | 8 +---- src/or/relay.c | 2 + src/or/rendclient.c | 4 +- src/or/rendcommon.c | 2 +- src/or/rephist.c | 2 +- src/or/routerlist.c | 18 +++++----- src/or/routerparse.c | 2 +- 34 files changed, 184 insertions(+), 203 deletions(-) diff --cc src/or/or.h index 3cadd31,e44c626..1688a08 --- a/src/or/or.h +++ b/src/or/or.h @@@ -2126,9 -2126,16 +2126,15 @@@ typedef struct circuit_t * length ONIONSKIN_CHALLENGE_LEN. */ char *n_conn_onionskin; - time_t timestamp_created; /**< When was this circuit created? */ + struct timeval timestamp_created; /**< When was the circuit created? */ - time_t timestamp_dirty; /**< When the circuit was first used, or 0 if the - * circuit is clean. */ + /** When the circuit was first used, or 0 if the circuit is clean. + * + * XXXX023 Note that some code will artifically adjust this value backward + * in time in order to indicate that a circuit shouldn't be used for new + * streams, but that it can stay alive as long as it has streams on it. + * That's a kludge we should fix. + */ + time_t timestamp_dirty; - struct timeval highres_created; /**< When exactly was the circuit created? */ uint16_t marked_for_close; /**< Should we close this circuit at the end of * the main loop? (If true, holds the line number

1 0

[tor/master] Fixup tor_addr_to_sockaddr return convention
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 41380fa3b3bc3647699116866afda4286ba90f8b Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 15:41:19 2011 -0400 Fixup tor_addr_to_sockaddr return convention --- src/common/address.c | 18 ++++++++++-------- src/or/dns.c | 2 +- 2 files changed, 11 insertions(+), 9 deletions(-) diff --git a/src/common/address.c b/src/common/address.c index 0046d2d..a4f8d34 100644 --- a/src/common/address.c +++ b/src/common/address.c @@ -50,11 +50,13 @@ #include <assert.h> /** Convert the tor_addr_t in a, with port in port, into a - * socklen object in *sa_out of object size len. If not enough - * room is free, or on error, return -1. Else return the length of the - * sockaddr. */ -/* XXXX021 This returns socklen_t. socklen_t is sometimes unsigned. This - * function claims to return -1 sometimes. Problematic! */ + * sockaddr object in *sa_out of object size len. If not enough + * room is available in sa_out, or on error, return 0 On success, return + * the length of the sockaddr. + * + * Interface note: ordinarly, we return -1 for error. We can't do that here, + * since socklen is unsigned on some platforms. + **/ socklen_t tor_addr_to_sockaddr(const tor_addr_t *a, uint16_t port, @@ -65,7 +67,7 @@ tor_addr_to_sockaddr(const tor_addr_t *a, if (family == AF_INET) { struct sockaddr_in *sin; if (len < (int)sizeof(struct sockaddr_in)) - return -1; + return 0; sin = (struct sockaddr_in *)sa_out; memset(sin, 0, sizeof(struct sockaddr_in)); #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN @@ -78,7 +80,7 @@ tor_addr_to_sockaddr(const tor_addr_t *a, } else if (family == AF_INET6) { struct sockaddr_in6 *sin6; if (len < (int)sizeof(struct sockaddr_in6)) - return -1; + return 0; sin6 = (struct sockaddr_in6 *)sa_out; memset(sin6, 0, sizeof(struct sockaddr_in6)); #ifdef HAVE_STRUCT_SOCKADDR_IN6_SIN6_LEN @@ -89,7 +91,7 @@ tor_addr_to_sockaddr(const tor_addr_t *a, memcpy(&sin6->sin6_addr, tor_addr_to_in6(a), sizeof(struct in6_addr)); return sizeof(struct sockaddr_in6); } else { - return -1; + return 0; } } diff --git a/src/or/dns.c b/src/or/dns.c index dcccd13..61c8f32 100644 --- a/src/or/dns.c +++ b/src/or/dns.c @@ -1206,7 +1206,7 @@ configure_nameservers(int force) struct sockaddr_storage ss; socklen = tor_addr_to_sockaddr(&addr, 0, (struct sockaddr *)&ss, sizeof(ss)); - if (socklen < 0) { + if (socklen <= 0) { log_warn(LD_BUG, "Couldn't convert outbound bind address to sockaddr." " Ignoring."); } else {

1 0

[tor/master] Remove the "fuzzy time" code
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 444e46d96d05c3e12eca02769512c74f2eb16b48 Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 15:27:06 2011 -0400 Remove the "fuzzy time" code It was the start of a neat idea, but it only got used in 3 places, none of which really needed it. --- changes/kill_ftime | 7 ++++ src/common/util.c | 77 ------------------------------------------------ src/common/util.h | 10 ------ src/or/networkstatus.c | 6 +++- src/or/routerlist.c | 4 +- 5 files changed, 14 insertions(+), 90 deletions(-) diff --git a/changes/kill_ftime b/changes/kill_ftime new file mode 100644 index 0000000..47f4769 --- /dev/null +++ b/changes/kill_ftime @@ -0,0 +1,7 @@ + o Code simplification and refactoring + - Remove the old 'fuzzy time' logic. It was supposed to be used + for handling calculations where we have a known amount of clock + skew and an allowed amount of unknown skew. But we only used it + in three places, and we never adjusted the known/unknown skew + values. This is still something we might want to do someday, + but if we do, we'll want to do it differently. diff --git a/src/common/util.c b/src/common/util.c index abd87ea..38c0ad0 100644 --- a/src/common/util.c +++ b/src/common/util.c @@ -1511,83 +1511,6 @@ update_approx_time(time_t now) #endif /* ===== - * Fuzzy time - * XXXX022 Use this consistently or rip most of it out. - * ===== */ - -/* In a perfect world, everybody would run NTP, and NTP would be perfect, so - * if we wanted to know "Is the current time before time X?" we could just say - * "time(NULL) < X". - * - * But unfortunately, many users are running Tor in an imperfect world, on - * even more imperfect computers. Hence, we need to track time oddly. We - * model the user's computer as being "skewed" from accurate time by - * -ftime_skew seconds, such that our best guess of the current time is - * time(NULL)+ftime_skew. We also assume that our measurements of time may - * have up to ftime_slop seconds of inaccuracy; IOW, our window of - * estimate for the current time is now + ftime_skew +/- ftime_slop. - */ -/** Our current estimate of our skew, such that we think the current time is - * closest to time(NULL)+ftime_skew. */ -static int ftime_skew = 0; -/** Tolerance during time comparisons, in seconds. */ -static int ftime_slop = 60; -/** Set the largest amount of sloppiness we'll allow in fuzzy time - * comparisons. */ -void -ftime_set_maximum_sloppiness(int seconds) -{ - tor_assert(seconds >= 0); - ftime_slop = seconds; -} -/** Set the amount by which we believe our system clock to differ from - * real time. */ -void -ftime_set_estimated_skew(int seconds) -{ - ftime_skew = seconds; -} -#if 0 -void -ftime_get_window(time_t now, ftime_t *ft_out) -{ - ft_out->earliest = now + ftime_skew - ftime_slop; - ft_out->latest = now + ftime_skew + ftime_slop; -} -#endif -/** Return true iff we think that now might be after when. */ -int -ftime_maybe_after(time_t now, time_t when) -{ - /* It may be after when iff the latest possible current time is after when */ - return (now + ftime_skew + ftime_slop) >= when; -} -/** Return true iff we think that now might be before when. */ -int -ftime_maybe_before(time_t now, time_t when) -{ - /* It may be before when iff the earliest possible current time is before */ - return (now + ftime_skew - ftime_slop) < when; -} -/** Return true if we think that now is definitely after when. */ -int -ftime_definitely_after(time_t now, time_t when) -{ - /* It is definitely after when if the earliest time it could be is still - * after when. */ - return (now + ftime_skew - ftime_slop) >= when; -} -/** Return true if we think that now is definitely before when. - */ -int -ftime_definitely_before(time_t now, time_t when) -{ - /* It is definitely before when if the latest time it could be is still - * before when. */ - return (now + ftime_skew + ftime_slop) < when; -} - -/* ===== * Rate limiting * ===== */ diff --git a/src/common/util.h b/src/common/util.h index 3736237..6b54856 100644 --- a/src/common/util.h +++ b/src/common/util.h @@ -247,16 +247,6 @@ time_t approx_time(void); void update_approx_time(time_t now); #endif -/* Fuzzy time. */ -void ftime_set_maximum_sloppiness(int seconds); -void ftime_set_estimated_skew(int seconds); -/* typedef struct ftime_t { time_t earliest; time_t latest; } ftime_t; */ -/* void ftime_get_window(time_t now, ftime_t *ft_out); */ -int ftime_maybe_after(time_t now, time_t when); -int ftime_maybe_before(time_t now, time_t when); -int ftime_definitely_after(time_t now, time_t when); -int ftime_definitely_before(time_t now, time_t when); - /* Rate-limiter */ /** A ratelim_t remembers how often an event is occurring, and how often diff --git a/src/or/networkstatus.c b/src/or/networkstatus.c index 687ac03..155dffe 100644 --- a/src/or/networkstatus.c +++ b/src/or/networkstatus.c @@ -1764,7 +1764,11 @@ networkstatus_set_current_consensus(const char *consensus, write_str_to_file(consensus_fname, consensus, 0); } - if (ftime_definitely_before(now, current_consensus->valid_after)) { +/** If a consensus appears more than this many seconds before its declared + * valid-after time, declare that our clock is skewed. */ +#define EARLY_CONSENSUS_NOTICE_SKEW 60 + + if (now < current_consensus->valid_after - EARLY_CONSENSUS_NOTICE_SKEW) { char tbuf[ISO_TIME_LEN+1]; char dbuf[64]; long delta = now - current_consensus->valid_after; diff --git a/src/or/routerlist.c b/src/or/routerlist.c index 4421d5c..2dca578 100644 --- a/src/or/routerlist.c +++ b/src/or/routerlist.c @@ -328,7 +328,7 @@ trusted_dirs_remove_old_certs(void) time_t cert_published; if (newest == cert) continue; - expired = ftime_definitely_after(now, cert->expires); + expired = now > cert->expires; cert_published = cert->cache_info.published_on; /* Store expired certs for 48 hours after a newer arrives; */ @@ -520,7 +520,7 @@ authority_certs_fetch_missing(networkstatus_t *status, time_t now) continue; cl = get_cert_list(ds->v3_identity_digest); SMARTLIST_FOREACH(cl->certs, authority_cert_t *, cert, { - if (!ftime_definitely_after(now, cert->expires)) { + if (now < cert->expires) { /* It's not expired, and we weren't looking for something to * verify a consensus with. Call it done. */ download_status_reset(&cl->dl_status);

1 0

[tor/master] Clean up a comment-conversation about bad libevent version/method combos
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 88bb40d8f87077e2f954b6de779ac52c637c69ec Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 15:32:45 2011 -0400 Clean up a comment-conversation about bad libevent version/method combos --- src/common/compat_libevent.c | 17 ++++++----------- 1 files changed, 6 insertions(+), 11 deletions(-) diff --git a/src/common/compat_libevent.c b/src/common/compat_libevent.c index 3efc56d..3ad9be1 100644 --- a/src/common/compat_libevent.c +++ b/src/common/compat_libevent.c @@ -330,17 +330,12 @@ tor_check_libevent_version(const char *m, int server, version = tor_get_libevent_version(&v); - /* XXX Would it be worthwhile disabling the methods that we know - * are buggy, rather than just warning about them and then proceeding - * to use them? If so, we should probably not wrap this whole thing - * in HAVE_EVENT_GET_VERSION and HAVE_EVENT_GET_METHOD. -RD */ - /* XXXX The problem is that it's not trivial to get libevent to change it's - * method once it's initialized, and it's not trivial to tell what method it - * will use without initializing it. I guess we could preemptively disable - * buggy libevent modes based on the version _before_ initializing it, - * though, but then there's no good way (afaict) to warn "I would have used - * kqueue, but instead I'm using select." -NM */ - /* XXXX022 revist the above; it is fixable now. */ + /* It would be better to disable known-buggy methods rather than warning + * about them. But the problem is that with older versions of Libevent, + * it's not trivial to get them to change their methods once they're + * initialized... and with newer versions of Libevent, they aren't actually + * broken. But we should revisit this if we ever find a post-1.4 version + * of Libevent where we need to disable a given method. */ if (!strcmp(m, "kqueue")) { if (version < V_OLD(1,1,'b')) buggy = 1;

1 0

[tor/master] Clarify some documentation and comments wrt resetting OR token buckets
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit d38030381b04263f76ce7f8ed2d6ceefc4ce363f Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 17:21:16 2011 -0400 Clarify some documentation and comments wrt resetting OR token buckets --- src/or/connection_or.c | 6 +++++- src/or/networkstatus.c | 3 ++- 2 files changed, 7 insertions(+), 2 deletions(-) diff --git a/src/or/connection_or.c b/src/or/connection_or.c index ff863fd..4b932ec 100644 --- a/src/or/connection_or.c +++ b/src/or/connection_or.c @@ -353,6 +353,9 @@ connection_or_digest_is_known_relay(const char *id_digest) * per-conn limits that are big enough they'll never matter. But if it's * not a known relay, first check if we set PerConnBwRate/Burst, then * check if the consensus sets them, else default to 'big enough'. + * + * If reset is true, set the bucket to be full. Otherwise, just + * clip the bucket if it happens to be too full. */ static void connection_or_update_token_buckets_helper(or_connection_t *conn, int reset, @@ -392,7 +395,8 @@ connection_or_update_token_buckets_helper(or_connection_t *conn, int reset, } /** Either our set of relays or our per-conn rate limits have changed. - * Go through all the OR connections and update their token buckets. */ + * Go through all the OR connections and update their token buckets to make + * sure they don't exceed their maximum values. */ void connection_or_update_token_buckets(smartlist_t *conns, or_options_t *options) { diff --git a/src/or/networkstatus.c b/src/or/networkstatus.c index 18cb477..4f6fe15 100644 --- a/src/or/networkstatus.c +++ b/src/or/networkstatus.c @@ -1747,7 +1747,8 @@ networkstatus_set_current_consensus(const char *consensus, routerstatus_list_update_named_server_map(); cell_ewma_set_scale_factor(options, current_consensus); - /* XXX022 where is the right place to put this call? */ + /* XXXX023 this call might be unnecessary here: can changing the + * current consensus really alter our view of any OR's rate limits? */ connection_or_update_token_buckets(get_connection_array(), options); circuit_build_times_new_consensus_params(&circ_times, current_consensus);

1 0

[tor/master] Remove workaround code for bug539
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 550749555cd57a1f82fe6c08e866ae14456ed439 Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 17:11:04 2011 -0400 Remove workaround code for bug539 We fixed bug 539 (where directories would say "503" but send data anyway) back in 0.2.0.16-alpha/0.1.2.19. Because most directory versions were affected, we added workaround to make sure that we examined the contents of 503-replies to make sure there wasn't any data for them to find. But now that such routers are nonexistent, we can remove this code. (Even if somebody fired up an 0.1.2.19 directory cache today, it would still be fine to ignore data in its erroneous 503 replies.) --- changes/bug539_removal | 6 ++++++ src/or/directory.c | 33 +++++++++++++-------------------- 2 files changed, 19 insertions(+), 20 deletions(-) diff --git a/changes/bug539_removal b/changes/bug539_removal new file mode 100644 index 0000000..dbff43d --- /dev/null +++ b/changes/bug539_removal @@ -0,0 +1,6 @@ + o Removed code + - Removed workaround code to handle directory responses from + servers that had bug 539 (they would send HTTP status 503 + responses _and_ send a body too). Since only server versions + before 0.2.0.16-alpha/0.1.2.19 were affected, there is no longer + reason to keep the workaround in place. diff --git a/src/or/directory.c b/src/or/directory.c index 1347c8b..8f33a60 100644 --- a/src/or/directory.c +++ b/src/or/directory.c @@ -1539,26 +1539,19 @@ connection_dir_client_reached_eof(dir_connection_t *conn) (void) skewed; /* skewed isn't used yet. */ if (status_code == 503) { - if (body_len < 16) { - routerstatus_t *rs; - trusted_dir_server_t *ds; - log_info(LD_DIR,"Received http status code %d (%s) from server " - "'%s:%d'. I'll try again soon.", - status_code, escaped(reason), conn->_base.address, - conn->_base.port); - if ((rs = router_get_consensus_status_by_id(conn->identity_digest))) - rs->last_dir_503_at = now; - if ((ds = router_get_trusteddirserver_by_digest(conn->identity_digest))) - ds->fake_status.last_dir_503_at = now; + routerstatus_t *rs; + trusted_dir_server_t *ds; + log_info(LD_DIR,"Received http status code %d (%s) from server " + "'%s:%d'. I'll try again soon.", + status_code, escaped(reason), conn->_base.address, + conn->_base.port); + if ((rs = router_get_consensus_status_by_id(conn->identity_digest))) + rs->last_dir_503_at = now; + if ((ds = router_get_trusteddirserver_by_digest(conn->identity_digest))) + ds->fake_status.last_dir_503_at = now; - tor_free(body); tor_free(headers); tor_free(reason); - return -1; - } - /* XXXX022 Remove this once every server with bug 539 is obsolete. */ - log_info(LD_DIR, "Server at '%s:%d' sent us a 503 response, but included " - "a body anyway. We'll pretend it gave us a 200.", - conn->_base.address, conn->_base.port); - status_code = 200; + tor_free(body); tor_free(headers); tor_free(reason); + return -1; } plausible = body_is_plausible(body, body_len, conn->_base.purpose);

1 0

[tor/master] Fix some 'impossible' overflow bugs in byte counting
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit dddd333a80ee2e9bb731cb3c127ace3741d49673 Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 16:45:25 2011 -0400 Fix some 'impossible' overflow bugs in byte counting The first was genuinely impossible, I think: it could only happen when the amount we read differed from the amount we wanted to read by more than INT_MAX. The second is just very unlikely: it would give incorrect results to the controller if you somehow wrote or read more than 4GB on one edge conn in one second. That one is a bugfix on 0.1.2.8-beta. --- changes/count_overflow | 5 +++++ src/or/connection.c | 36 ++++++++++++++++++++++-------------- 2 files changed, 27 insertions(+), 14 deletions(-) diff --git a/changes/count_overflow b/changes/count_overflow new file mode 100644 index 0000000..f302ff2 --- /dev/null +++ b/changes/count_overflow @@ -0,0 +1,5 @@ + o Minor bugfixes: + - Correctly handle an "impossible" overflow cases in connection + byte counting, where we write or read more than 4GB on an edge + connection in single second. Bugfix on 0.1.2.8-beta. + diff --git a/src/or/connection.c b/src/or/connection.c index 7fa6cd9..084237d 100644 --- a/src/or/connection.c +++ b/src/or/connection.c @@ -51,7 +51,7 @@ static int connection_finished_flushing(connection_t *conn); static int connection_flushed_some(connection_t *conn); static int connection_finished_connecting(connection_t *conn); static int connection_reached_eof(connection_t *conn); -static int connection_read_to_buf(connection_t *conn, int *max_to_read, +static int connection_read_to_buf(connection_t *conn, ssize_t *max_to_read, int *socket_error); static int connection_process_inbuf(connection_t *conn, int package_partial); static void client_check_address_changed(int sock); @@ -2338,7 +2338,7 @@ connection_bucket_should_increase(int bucket, or_connection_t *conn) static int connection_handle_read_impl(connection_t *conn) { - int max_to_read=-1, try_to_read; + ssize_t max_to_read=-1, try_to_read; size_t before, n_read = 0; int socket_error = 0; @@ -2456,7 +2456,8 @@ connection_handle_read(connection_t *conn) * Return -1 if we want to break conn, else return 0. */ static int -connection_read_to_buf(connection_t *conn, int *max_to_read, int *socket_error) +connection_read_to_buf(connection_t *conn, ssize_t *max_to_read, + int *socket_error) { int result; ssize_t at_most = *max_to_read; @@ -2574,15 +2575,19 @@ connection_read_to_buf(connection_t *conn, int *max_to_read, int *socket_error) n_read = (size_t) result; } - if (n_read > 0) { /* change *max_to_read */ - /*XXXX022 check for overflow*/ - *max_to_read = (int)(at_most - n_read); - } + if (n_read > 0) { + /* change *max_to_read */ + *max_to_read = at_most - n_read; - if (conn->type == CONN_TYPE_AP) { - edge_connection_t *edge_conn = TO_EDGE_CONN(conn); - /*XXXX022 check for overflow*/ - edge_conn->n_read += (int)n_read; + /* Update edge_conn->n_read */ + if (conn->type == CONN_TYPE_AP) { + edge_connection_t *edge_conn = TO_EDGE_CONN(conn); + /* Check for overflow: */ + if (PREDICT_LIKELY(UINT32_MAX - edge_conn->n_read > n_read)) + edge_conn->n_read += (int)n_read; + else + edge_conn->n_read = UINT32_MAX; + } } connection_buckets_decrement(conn, approx_time(), n_read, n_written); @@ -2781,10 +2786,13 @@ connection_handle_write_impl(connection_t *conn, int force) n_written = (size_t) result; } - if (conn->type == CONN_TYPE_AP) { + if (n_written && conn->type == CONN_TYPE_AP) { edge_connection_t *edge_conn = TO_EDGE_CONN(conn); - /*XXXX022 check for overflow.*/ - edge_conn->n_written += (int)n_written; + /* Check for overflow: */ + if (PREDICT_LIKELY(UINT32_MAX - edge_conn->n_written > n_written)) + edge_conn->n_written += n_written; + else + edge_conn->n_written = UINT32_MAX; } connection_buckets_decrement(conn, approx_time(), n_read, n_written);

1 0

[tor/master] Look at the right errno when sending reason for connect() failure
by nickm＠torproject.org 07 Apr '11

07 Apr '11

commit 6a5b94de6c51478ca638ec03c40a6d3c27f2d674 Author: Nick Mathewson <nickm(a)torproject.org> Date: Fri Mar 25 16:14:42 2011 -0400 Look at the right errno when sending reason for connect() failure In afe414 (tor-0.1.0.1-rc~173), when we moved to connection_edge_end_errno(), we used it in handling errors from connection_connect(). That's not so good, since by the time connection_connect() returns, the socket is no longer set, and we're supposed to be looking at the socket_errno return value from connection_connect() instead. So do what we should've done, and look at the socket_errno value that we get from connection_connect(). --- changes/connect_err_reporting | 6 ++++++ src/or/connection_edge.c | 8 ++++---- 2 files changed, 10 insertions(+), 4 deletions(-) diff --git a/changes/connect_err_reporting b/changes/connect_err_reporting new file mode 100644 index 0000000..61a46b6 --- /dev/null +++ b/changes/connect_err_reporting @@ -0,0 +1,6 @@ + o Minor bugfixes: + - Be more careful about reporting the correct error from a failed + connect() operation. Under some circumstances, it was possible to + look at an incorrect value for errno when sending the end reason. + Bugfix on Tor-0.1.0.1-rc. + diff --git a/src/or/connection_edge.c b/src/or/connection_edge.c index 82b7139..c75c06b 100644 --- a/src/or/connection_edge.c +++ b/src/or/connection_edge.c @@ -2823,13 +2823,13 @@ connection_exit_connect(edge_connection_t *edge_conn) log_debug(LD_EXIT,"about to try connecting"); switch (connection_connect(conn, conn->address, addr, port, &socket_error)) { - case -1: - /* XXX022 use socket_error below rather than trying to piece things - * together from the current errno, which may have been clobbered. */ - connection_edge_end_errno(edge_conn); + case -1: { + int reason = errno_to_stream_end_reason(socket_error); + connection_edge_end(edge_conn, reason); circuit_detach_stream(circuit_get_by_edge_conn(edge_conn), edge_conn); connection_free(conn); return; + } case 0: conn->state = EXIT_CONN_STATE_CONNECTING;

1 0