git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Denton Liu <liu.denton@gmail.com>
To: Git Mailing List <git@vger.kernel.org>
Cc: Jeff King <peff@peff.net>, Force Charlie <charlieio@outlook.com>
Subject: [PATCH 6/6] remote-curl: ensure last packet is a flush
Date: Wed, 13 May 2020 14:04:58 -0400	[thread overview]
Message-ID: <7a689da2bb820f70d9e668d656b088af2297d456.1589393036.git.liu.denton@gmail.com> (raw)
In-Reply-To: <cover.1589393036.git.liu.denton@gmail.com>

Currently, remote-curl acts as a proxy and blindly forwards packets
between an HTTP server and fetch-pack. In the case of a stateless RPC
connection where the connection is terminated before the transaction is
complete, remote-curl will blindly forward the packets before waiting on
more input from fetch-pack. Meanwhile, fetch-pack will read the
transaction and continue reading, expecting more input to continue the
transaction. This results in a deadlock between the two processes.

This can be seen in the following command which does not terminate:

	$ git -c protocol.version=2 clone https://github.com/git/git.git --shallow-since=20151012
	Cloning into 'git'...

whereas the v1 version does terminate as expected:

	$ git -c protocol.version=1 clone https://github.com/git/git.git --shallow-since=20151012
	Cloning into 'git'...
	fatal: the remote end hung up unexpectedly

Instead of blindly forwarding packets, raise a flag when a flush packet
is encountered. Ensure that the last packet sent is a flush packet
otherwise error out, breaking the deadlock.

This is not a complete solution to the problem, however. It is possible
that a flush packet could be sent in the middle of a message and the
connection could die immediately after. Then, remote-curl would not
error out and fetch-pack would still be in the middle of a transaction
and they would enter deadlock. A complete solution would involve
reframing the stateless-connect protocol, possibly by introducing
another control packet ("0002"?) as a stateless request separator
packet which is always sent at the end of post_rpc().

Although this is not a complete solution, it is better than nothing and
it resolves the reported issue for now.

Reported-by: Force Charlie <charlieio@outlook.com>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Denton Liu <liu.denton@gmail.com>
---

Notes:
    I wish there were some way to insert a timeout on the test case so that
    we don't block forever in case we regress.

 remote-curl.c          |  9 +++++++++
 t/t5702-protocol-v2.sh | 17 +++++++++++++++++
 2 files changed, 26 insertions(+)

diff --git a/remote-curl.c b/remote-curl.c
index 8b740354e5..aab17851be 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -684,6 +684,7 @@ struct rpc_in_data {
 	struct active_request_slot *slot;
 	struct strbuf len_buf;
 	int remaining;
+	int last_flush;
 };
 
 /*
@@ -707,6 +708,8 @@ static size_t rpc_in(char *ptr, size_t eltsize,
 		data->rpc->any_written = 1;
 
 	while (unwritten) {
+		data->last_flush = 0;
+
 		if (!data->remaining) {
 			int digits_remaining = 4 - data->len_buf.len;
 			if (digits_remaining > unwritten)
@@ -720,6 +723,8 @@ static size_t rpc_in(char *ptr, size_t eltsize,
 				if (data->remaining < 0) {
 					die(_("remote-curl: bad line length character: %.4s"), data->len_buf.buf);
 				} else if (data->remaining <= 1) {
+					if (!data->remaining)
+						data->last_flush = 1;
 					data->remaining = 0;
 				} else if (data->remaining < 4) {
 					die(_("remote-curl: bad line length %d"), data->remaining);
@@ -960,6 +965,7 @@ static int post_rpc(struct rpc_state *rpc, int flush_received)
 	rpc_in_data.slot = slot;
 	strbuf_init(&rpc_in_data.len_buf, 4);
 	rpc_in_data.remaining = 0;
+	rpc_in_data.last_flush = 0;
 	curl_easy_setopt(slot->curl, CURLOPT_FILE, &rpc_in_data);
 	curl_easy_setopt(slot->curl, CURLOPT_FAILONERROR, 0);
 
@@ -979,6 +985,9 @@ static int post_rpc(struct rpc_state *rpc, int flush_received)
 	if (rpc_in_data.remaining)
 		err = error(_("%d bytes are still expected"), rpc_in_data.remaining);
 
+	if (!rpc_in_data.last_flush)
+		err = error(_("last packet was not a flush"));
+
 	curl_slist_free_all(headers);
 	free(gzip_body);
 	return err;
diff --git a/t/t5702-protocol-v2.sh b/t/t5702-protocol-v2.sh
index 5039e66dc4..4570d0746c 100755
--- a/t/t5702-protocol-v2.sh
+++ b/t/t5702-protocol-v2.sh
@@ -586,6 +586,23 @@ test_expect_success 'clone with http:// using protocol v2' '
 	! grep "Send header: Transfer-Encoding: chunked" log
 '
 
+test_expect_success 'clone with http:// using protocol v2 and invalid parameters' '
+	test_when_finished "rm -f log" &&
+
+	test_must_fail env GIT_TRACE_PACKET="$(pwd)/log" GIT_TRACE_CURL="$(pwd)/log" \
+		git -c protocol.version=2 \
+		clone --shallow-since=20151012 "$HTTPD_URL/smart/http_parent" http_child_invalid 2>err &&
+
+	# Client requested to use protocol v2
+	grep "Git-Protocol: version=2" log &&
+	# Server responded using protocol v2
+	grep "git< version 2" log &&
+	# Verify that we errored out in the expected way
+	test_i18ngrep "last packet was not a flush" err &&
+	# Verify that the chunked encoding sending codepath is NOT exercised
+	! grep "Send header: Transfer-Encoding: chunked" log
+'
+
 test_expect_success 'clone big repository with http:// using protocol v2' '
 	test_when_finished "rm -f log" &&
 
-- 
2.26.2.706.g87896c9627


  parent reply	other threads:[~2020-05-13 18:05 UTC|newest]

Thread overview: 62+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-13 18:04 [PATCH 0/6] remote-curl: partial fix for a deadlock with stateless rpc Denton Liu
2020-05-13 18:04 ` [PATCH 1/6] remote-curl: fix typo Denton Liu
2020-05-13 18:04 ` [PATCH 2/6] remote-curl: remove label indentation Denton Liu
2020-05-13 18:04 ` [PATCH 3/6] transport: combine common cases with a fallthrough Denton Liu
2020-05-13 23:14   ` Eric Sunshine
2020-05-18  9:18     ` Denton Liu
2020-05-18 17:43       ` Eric Sunshine
2020-05-13 18:04 ` [PATCH 4/6] pkt-line: extern packet_length() Denton Liu
2020-05-13 23:23   ` Eric Sunshine
2020-05-15 20:56   ` Jeff King
2020-05-15 20:57     ` Jeff King
2020-05-13 18:04 ` [PATCH 5/6] remote-curl: error on incomplete packet Denton Liu
2020-05-15 21:38   ` Jeff King
2020-05-18  9:08     ` Denton Liu
2020-05-18 15:49       ` Jeff King
2020-05-13 18:04 ` Denton Liu [this message]
2020-05-15 21:02   ` [PATCH 6/6] remote-curl: ensure last packet is a flush Denton Liu
2020-05-15 21:41     ` Jeff King
2020-05-18 16:34       ` Junio C Hamano
2020-05-18 16:52         ` Jeff King
2020-05-18 21:00           ` Jeff King
2020-05-18 15:47 ` [PATCH v2 0/7] remote-curl: fix deadlocks when remote server disconnects Denton Liu
2020-05-18 15:47   ` [PATCH v2 1/7] remote-curl: fix typo Denton Liu
2020-05-18 15:47   ` [PATCH v2 2/7] remote-curl: remove label indentation Denton Liu
2020-05-18 18:37     ` Junio C Hamano
2020-05-18 15:47   ` [PATCH v2 3/7] transport: extract common fetch_pack() call Denton Liu
2020-05-18 18:40     ` Junio C Hamano
2020-05-18 15:47   ` [PATCH v2 4/7] pkt-line: extern packet_length() Denton Liu
2020-05-18 16:04     ` Jeff King
2020-05-18 17:50       ` Eric Sunshine
2020-05-18 20:08         ` Jeff King
2020-05-18 18:44       ` Junio C Hamano
2020-05-18 15:47   ` [PATCH v2 5/7] remote-curl: error on incomplete packet Denton Liu
2020-05-18 16:22     ` Jeff King
2020-05-18 16:51       ` Denton Liu
2020-05-18 15:47   ` [PATCH v2 6/7] pkt-line: PACKET_READ_RESPONSE_END Denton Liu
2020-05-18 15:47   ` [PATCH v2 7/7] stateless-connect: send response end packet Denton Liu
2020-05-18 16:43     ` Jeff King
2020-05-18 17:12       ` Denton Liu
2020-05-18 17:26         ` Jeff King
2020-05-18 16:50   ` [PATCH v2 0/7] remote-curl: fix deadlocks when remote server disconnects Jeff King
2020-05-18 17:36     ` Denton Liu
2020-05-18 20:58       ` Jeff King
2020-05-18 22:52         ` Junio C Hamano
2020-05-19  2:38           ` Jeff King
2020-05-18 19:36     ` Junio C Hamano
2020-05-19 10:53   ` [PATCH v3 " Denton Liu
2020-05-19 10:53     ` [PATCH v3 1/7] remote-curl: fix typo Denton Liu
2020-05-19 10:53     ` [PATCH v3 2/7] remote-curl: remove label indentation Denton Liu
2020-05-19 10:53     ` [PATCH v3 3/7] transport: extract common fetch_pack() call Denton Liu
2020-05-19 10:53     ` [PATCH v3 4/7] pkt-line: extern packet_length() Denton Liu
2020-05-19 16:23       ` Eric Sunshine
2020-05-19 10:53     ` [PATCH v3 5/7] remote-curl: error on incomplete packet Denton Liu
2020-05-19 10:53     ` [PATCH v3 6/7] pkt-line: define PACKET_READ_RESPONSE_END Denton Liu
2020-05-19 10:54     ` [PATCH v3 7/7] stateless-connect: send response end packet Denton Liu
2020-05-19 18:40     ` [PATCH v3 0/7] remote-curl: fix deadlocks when remote server disconnects Jeff King
2020-05-19 21:14       ` Denton Liu
2020-05-19 20:51     ` [PATCH v3 8/7] fixup! pkt-line: extern packet_length() Denton Liu
2020-05-22 13:33     ` [PATCH v3 9/9] fixup! remote-curl: error on incomplete packet Denton Liu
2020-05-22 15:54       ` Jeff King
2020-05-22 16:05         ` Denton Liu
2020-05-22 16:31           ` Jeff King

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=7a689da2bb820f70d9e668d656b088af2297d456.1589393036.git.liu.denton@gmail.com \
    --to=liu.denton@gmail.com \
    --cc=charlieio@outlook.com \
    --cc=git@vger.kernel.org \
    --cc=peff@peff.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).