Fix DNS-over-HTTP(S) implementation issues that arise under heavy query load. Optimize resource usage for :iscman:`named` instances that accept queries over DNS-over-HTTP(S). Previously, :iscman:`named` would process all incoming HTTP/2 data at once, which could overwhelm the server, especially when dealing with clients that send requests but don't wait for responses. That has been fixed. Now, :iscman:`named` handles HTTP/2 data in smaller chunks and throttles reading until the remote side reads the response data. It also throttles clients that send too many requests at once. Additionally, :iscman:`named` now carefully processes data sent by some clients, which can be considered "flooding." It logs these clients and drops connections from them. :gl:`#4795` In some cases, :iscman:`named` could leave DNS-over-HTTP(S) connections in the `CLOSE_WAIT` state indefinitely. That also has been fixed. ISC would like to thank JF Billaud for thoroughly investigating the issue and verifying the fix. :gl:`#5083` Vulnerability: CVE-2024-12705 Resolves: RHEL-76868
1419 lines
44 KiB
Diff
1419 lines
44 KiB
Diff
From 05f63912f7a9246ccf1c59ee0b93782b5a576dbf Mon Sep 17 00:00:00 2001
|
|
From: Artem Boldariev <artem@isc.org>
|
|
Date: Wed, 4 Sep 2024 18:53:35 +0300
|
|
Subject: [PATCH] Implement TCP manual read timer control functionality
|
|
|
|
This commit adds a manual TCP read timer control mode which is
|
|
supposed to override automatic resetting of the timer when any data is
|
|
received. That can be accomplished by
|
|
`isc__nmhandle_set_manual_timer()`.
|
|
|
|
This functionality is supposed to be used by multilevel networking
|
|
transports which require finer grained control over the read
|
|
timer (TLS Stream, DoH).
|
|
|
|
The commit is essentially an implementation of the functionality from
|
|
newer versions of BIND.
|
|
|
|
(cherry picked from commit a67b3255426160c55faa9d1d3aa6e7579439b2b4)
|
|
|
|
Implement TLS manual read timer control functionality
|
|
|
|
This commit adds a manual TLS read timer control mode which is
|
|
supposed to override automatic resetting of the timer when any data is
|
|
received.
|
|
|
|
It both depends and complements similar functionality in TCP.
|
|
|
|
(cherry picked from commit 13d521fa5fe7102733db15079fd49c7ccf5b0648)
|
|
|
|
Add isc__nm_async_run()
|
|
|
|
This commit adds isc__nm_async_run() which is very similar to
|
|
isc_async_run() in newer versions of BIND: it allows calling a
|
|
callback asynchronously.
|
|
|
|
Potentially, it can be used to replace some other async operations in
|
|
other networking code, in particular the delayed I/O calls in TLS a
|
|
TCP DNS transports to name a few and remove quiet a lot of code, but
|
|
it we are unlikely to do that for the strictly maintenance only
|
|
branch, so it is protected with DoH-related #ifdefs.
|
|
|
|
It is implemented in a "universal" way mainly because doing it in the
|
|
specific code requires the same amount of code and is not simpler.
|
|
|
|
(cherry picked from commit 125bfd71d3c8d0ad23477867f1e41a2392e03138)
|
|
|
|
DoH: process data chunk by chunk instead of all at once
|
|
|
|
Initially, our DNS-over-HTTP(S) implementation would try to process as
|
|
much incoming data from the network as possible. However, that might
|
|
be undesirable as we might create too many streams (each effectively
|
|
backed by a ns_client_t object). That is too forgiving as it might
|
|
overwhelm the server and trash its memory allocator, causing high CPU
|
|
and memory usage.
|
|
|
|
Instead of doing that, we resort to processing incoming data using a
|
|
chunk-by-chunk processing strategy. That is, we split data into small
|
|
chunks (currently 256 bytes) and process each of them
|
|
asynchronously. However, we can process more than one chunk at
|
|
once (up to 4 currently), given that the number of HTTP/2 streams has
|
|
not increased while processing a chunk.
|
|
|
|
That alone is not enough, though. In addition to the above, we should
|
|
limit the number of active streams: these streams for which we have
|
|
received a request and started processing it (the ones for which a
|
|
read callback was called), as it is perfectly fine to have more opened
|
|
streams than active ones. In the case we have reached or surpassed the
|
|
limit of active streams, we stop reading AND processing the data from
|
|
the remote peer. The number of active streams is effectively decreased
|
|
only when responses associated with the active streams are sent to the
|
|
remote peer.
|
|
|
|
Overall, this strategy is very similar to the one used for other
|
|
stream-based DNS transports like TCP and TLS.
|
|
|
|
(cherry picked from commit 9846f395ad79bb50a5fa5ca6ab97ef904b3be35a)
|
|
(cherry picked from commit 11a2956dce6f983d2bfcb532f5719791845b06ab)
|
|
|
|
DoH: floodding clients detection
|
|
|
|
This commit adds logic to make code better protected against clients
|
|
that send valid HTTP/2 data that is useless from a DNS server
|
|
perspective.
|
|
|
|
Firstly, it adds logic that protects against clients who send too
|
|
little useful (=DNS) data. We achieve that by adding a check that
|
|
eventually detects such clients with a nonfavorable useful to
|
|
processed data ratio after the initial grace period. The grace period
|
|
is limited to processing 128 KiB of data, which should be enough for
|
|
sending the largest possible DNS message in a GET request and then
|
|
some. This is the main safety belt that would detect even flooding
|
|
clients that initially behave well in order to fool the checks server.
|
|
|
|
Secondly, in addition to the above, we introduce additional checks to
|
|
detect outright misbehaving clients earlier:
|
|
|
|
The code will treat clients that open too many streams (50) without
|
|
sending any data for processing as flooding ones; The clients that
|
|
managed to send 1.5 KiB of data without opening a single stream or
|
|
submitting at least some DNS data will be treated as flooding ones.
|
|
Of course, the behaviour described above is nothing else but
|
|
heuristical checks, so they can never be perfect. At the same time,
|
|
they should be reasonable enough not to drop any valid clients,
|
|
realatively easy to implement, and have negligible computational
|
|
overhead.
|
|
|
|
(cherry picked from commit 3425e4b1d04746520931e93ac7ef5979fd6b54fd)
|
|
(cherry picked from commit ee42514be2cc56f5d31fdb9d0c6a838e0d372654)
|
|
|
|
DoH: introduce manual read timer control
|
|
|
|
This commit introduces manual read timer control as used by StreamDNS
|
|
and its underlying transports. Before that, DoH code would rely on the
|
|
timer control provided by TCP, which would reset the timer any time
|
|
some data arrived. Now, the timer is restarted only when a full DNS
|
|
message is processed in line with other DNS transports.
|
|
|
|
That change is required because we should not stop the timer when
|
|
reading from the network is paused due to throttling. We need a way to
|
|
drop timed-out clients, particularly those who refuse to read the data
|
|
we send.
|
|
|
|
(cherry picked from commit 609a41517b1631c320876a41c43c68c9a0ee0f9f)
|
|
(cherry picked from commit 796708775d178adc5256ce03956d134c9fd38a33)
|
|
|
|
DoH: reduce excessive bad request logging
|
|
|
|
We started using isc_nm_bad_request() more actively throughout
|
|
codebase. In the case of HTTP/2 it can lead to a large count of
|
|
useless "Bad Request" messages in the BIND log, as often we attempt to
|
|
send such request over effectively finished HTTP/2 sessions.
|
|
|
|
This commit fixes that.
|
|
|
|
(cherry picked from commit 937b5f8349a6a5e15af254a53a659e39c7c1d179)
|
|
(cherry picked from commit 550b692343e398e0debe18ddb5908b4ccb183b91)
|
|
---
|
|
lib/isc/netmgr/http.c | 448 +++++++++++++++++++++++++++++++++---
|
|
lib/isc/netmgr/netmgr-int.h | 81 ++++++-
|
|
lib/isc/netmgr/netmgr.c | 80 ++++++-
|
|
lib/isc/netmgr/tcp.c | 26 ++-
|
|
lib/isc/netmgr/tlsstream.c | 137 +++++++++--
|
|
5 files changed, 721 insertions(+), 51 deletions(-)
|
|
|
|
diff --git a/lib/isc/netmgr/http.c b/lib/isc/netmgr/http.c
|
|
index 1d4b82fd0ec..2002848c519 100644
|
|
--- a/lib/isc/netmgr/http.c
|
|
+++ b/lib/isc/netmgr/http.c
|
|
@@ -85,6 +85,37 @@
|
|
|
|
#define INITIAL_DNS_MESSAGE_BUFFER_SIZE (512)
|
|
|
|
+/*
|
|
+ * The value should be small enough to not allow a server to open too
|
|
+ * many streams at once. It should not be too small either because
|
|
+ * the incoming data will be split into too many chunks with each of
|
|
+ * them processed asynchronously.
|
|
+ */
|
|
+#define INCOMING_DATA_CHUNK_SIZE (256)
|
|
+
|
|
+/*
|
|
+ * Often processing a chunk does not change the number of streams. In
|
|
+ * that case we can process more than once, but we still should have a
|
|
+ * hard limit on that.
|
|
+ */
|
|
+#define INCOMING_DATA_MAX_CHUNKS_AT_ONCE (4)
|
|
+
|
|
+/*
|
|
+ * These constants define the grace period to help detect flooding clients.
|
|
+ *
|
|
+ * The first one defines how much data can be processed before opening
|
|
+ * a first stream and received at least some useful (=DNS) data.
|
|
+ *
|
|
+ * The second one defines how much data from a client we read before
|
|
+ * trying to drop a clients who sends not enough useful data.
|
|
+ *
|
|
+ * The third constant defines how many streams we agree to process
|
|
+ * before checking if there was at least one DNS request received.
|
|
+ */
|
|
+#define INCOMING_DATA_INITIAL_STREAM_SIZE (1536)
|
|
+#define INCOMING_DATA_GRACE_SIZE (MAX_ALLOWED_DATA_IN_HEADERS)
|
|
+#define MAX_STREAMS_BEFORE_FIRST_REQUEST (50)
|
|
+
|
|
typedef struct isc_nm_http_response_status {
|
|
size_t code;
|
|
size_t content_length;
|
|
@@ -143,6 +174,7 @@ struct isc_nm_http_session {
|
|
ISC_LIST(http_cstream_t) cstreams;
|
|
ISC_LIST(isc_nmsocket_h2_t) sstreams;
|
|
size_t nsstreams;
|
|
+ uint64_t total_opened_sstreams;
|
|
|
|
isc_nmhandle_t *handle;
|
|
isc_nmhandle_t *client_httphandle;
|
|
@@ -155,6 +187,18 @@ struct isc_nm_http_session {
|
|
|
|
isc__nm_http_pending_callbacks_t pending_write_callbacks;
|
|
isc_buffer_t *pending_write_data;
|
|
+
|
|
+ /*
|
|
+ * The statistical values below are for usage on server-side
|
|
+ * only. They are meant to detect clients that are taking too many
|
|
+ * resources from the server.
|
|
+ */
|
|
+ uint64_t received; /* How many requests have been received. */
|
|
+ uint64_t submitted; /* How many responses were submitted to send */
|
|
+ uint64_t processed; /* How many responses were processed. */
|
|
+
|
|
+ uint64_t processed_incoming_data;
|
|
+ uint64_t processed_useful_data; /* DNS data */
|
|
};
|
|
|
|
typedef enum isc_http_error_responses {
|
|
@@ -177,6 +221,7 @@ typedef struct isc_http_send_req {
|
|
void *cbarg;
|
|
isc_buffer_t *pending_write_data;
|
|
isc__nm_http_pending_callbacks_t pending_write_callbacks;
|
|
+ uint64_t submitted;
|
|
} isc_http_send_req_t;
|
|
|
|
#define HTTP_ENDPOINTS_MAGIC ISC_MAGIC('H', 'T', 'E', 'P')
|
|
@@ -189,10 +234,26 @@ static bool
|
|
http_send_outgoing(isc_nm_http_session_t *session, isc_nmhandle_t *httphandle,
|
|
isc_nm_cb_t cb, void *cbarg);
|
|
|
|
+static void
|
|
+http_log_flooding_peer(isc_nm_http_session_t *session);
|
|
+
|
|
+static bool
|
|
+http_is_flooding_peer(isc_nm_http_session_t *session);
|
|
+
|
|
+static ssize_t
|
|
+http_process_input_data(isc_nm_http_session_t *session,
|
|
+ isc_buffer_t *input_data);
|
|
+
|
|
+static inline bool
|
|
+http_too_many_active_streams(isc_nm_http_session_t *session);
|
|
+
|
|
static void
|
|
http_do_bio(isc_nm_http_session_t *session, isc_nmhandle_t *send_httphandle,
|
|
isc_nm_cb_t send_cb, void *send_cbarg);
|
|
|
|
+static void
|
|
+http_do_bio_async(isc_nm_http_session_t *session);
|
|
+
|
|
static void
|
|
failed_httpstream_read_cb(isc_nmsocket_t *sock, isc_result_t result,
|
|
isc_nm_http_session_t *session);
|
|
@@ -494,6 +555,16 @@ finish_http_session(isc_nm_http_session_t *session) {
|
|
if (!session->closed) {
|
|
session->closed = true;
|
|
isc_nm_cancelread(session->handle);
|
|
+ isc__nmsocket_timer_stop(session->handle->sock);
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * Free any unprocessed incoming data in order to not process
|
|
+ * it during indirect calls to http_do_bio() that might happen
|
|
+ * when calling the failed callbacks.
|
|
+ */
|
|
+ if (session->buf != NULL) {
|
|
+ isc_buffer_free(&session->buf);
|
|
}
|
|
|
|
if (session->client) {
|
|
@@ -567,6 +638,7 @@ on_server_data_chunk_recv_callback(int32_t stream_id, const uint8_t *data,
|
|
if (new_bufsize <= MAX_DNS_MESSAGE_SIZE &&
|
|
new_bufsize <= h2->content_length)
|
|
{
|
|
+ session->processed_useful_data += len;
|
|
isc_buffer_putmem(&h2->rbuf, data, len);
|
|
break;
|
|
}
|
|
@@ -615,6 +687,9 @@ call_unlink_cstream_readcb(http_cstream_t *cstream,
|
|
isc_buffer_usedregion(cstream->rbuf, &read_data);
|
|
cstream->read_cb(session->client_httphandle, result, &read_data,
|
|
cstream->read_cbarg);
|
|
+ if (result == ISC_R_SUCCESS) {
|
|
+ isc__nmsocket_timer_restart(session->handle->sock);
|
|
+ }
|
|
put_http_cstream(session->mctx, cstream);
|
|
}
|
|
|
|
@@ -656,6 +731,9 @@ on_server_stream_close_callback(int32_t stream_id,
|
|
|
|
ISC_LIST_UNLINK(session->sstreams, &sock->h2, link);
|
|
session->nsstreams--;
|
|
+ if (sock->h2.request_received) {
|
|
+ session->submitted++;
|
|
+ }
|
|
|
|
/*
|
|
* By making a call to isc__nmsocket_prep_destroy(), we ensure that
|
|
@@ -967,6 +1045,182 @@ client_submit_request(isc_nm_http_session_t *session, http_cstream_t *stream) {
|
|
return (ISC_R_SUCCESS);
|
|
}
|
|
|
|
+static ssize_t
|
|
+http_process_input_data(isc_nm_http_session_t *session,
|
|
+ isc_buffer_t *input_data) {
|
|
+ ssize_t readlen = 0;
|
|
+ ssize_t processed = 0;
|
|
+ isc_region_t chunk = { 0 };
|
|
+ size_t before, after;
|
|
+ size_t i;
|
|
+
|
|
+ REQUIRE(VALID_HTTP2_SESSION(session));
|
|
+ REQUIRE(input_data != NULL);
|
|
+
|
|
+ if (!http_session_active(session)) {
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * For clients that initiate request themselves just process
|
|
+ * everything.
|
|
+ */
|
|
+ if (session->client) {
|
|
+ isc_buffer_remainingregion(input_data, &chunk);
|
|
+ if (chunk.length == 0) {
|
|
+ return 0;
|
|
+ }
|
|
+
|
|
+ readlen = nghttp2_session_mem_recv(session->ngsession,
|
|
+ chunk.base, chunk.length);
|
|
+
|
|
+ if (readlen >= 0) {
|
|
+ isc_buffer_forward(input_data, readlen);
|
|
+ session->processed_incoming_data += readlen;
|
|
+ }
|
|
+
|
|
+ return readlen;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * If no streams are created during processing, we might process
|
|
+ * more than one chunk at a time. Still we should not overdo that
|
|
+ * to avoid processing too much data at once as such behaviour is
|
|
+ * known for trashing the memory allocator at times.
|
|
+ */
|
|
+ for (before = after = session->nsstreams, i = 0;
|
|
+ after <= before && i < INCOMING_DATA_MAX_CHUNKS_AT_ONCE;
|
|
+ after = session->nsstreams, i++)
|
|
+ {
|
|
+ const uint64_t active_streams =
|
|
+ (session->received - session->processed);
|
|
+
|
|
+ /*
|
|
+ * If there are non completed send requests in flight -let's
|
|
+ * not process any incoming data, as it could lead to piling
|
|
+ * up too much send data in send buffers. With many clients
|
|
+ * connected it can lead to excessive memory consumption on
|
|
+ * the server instance.
|
|
+ */
|
|
+ if (session->sending > 0) {
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * If we have reached the maximum number of streams used, we
|
|
+ * might stop processing for now, as nghttp2 will happily
|
|
+ * consume as much data as possible.
|
|
+ */
|
|
+ if (session->nsstreams >= session->max_concurrent_streams &&
|
|
+ active_streams > 0)
|
|
+ {
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ if (http_too_many_active_streams(session)) {
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ isc_buffer_remainingregion(input_data, &chunk);
|
|
+ if (chunk.length == 0) {
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ chunk.length = ISC_MIN(chunk.length, INCOMING_DATA_CHUNK_SIZE);
|
|
+
|
|
+ readlen = nghttp2_session_mem_recv(session->ngsession,
|
|
+ chunk.base, chunk.length);
|
|
+
|
|
+ if (readlen >= 0) {
|
|
+ isc_buffer_forward(input_data, readlen);
|
|
+ session->processed_incoming_data += readlen;
|
|
+ processed += readlen;
|
|
+ } else {
|
|
+ isc_buffer_clear(input_data);
|
|
+ return readlen;
|
|
+ }
|
|
+ }
|
|
+
|
|
+ return processed;
|
|
+}
|
|
+
|
|
+static void
|
|
+http_log_flooding_peer(isc_nm_http_session_t *session) {
|
|
+ const int log_level = ISC_LOG_DEBUG(1);
|
|
+ if (session->handle != NULL && isc_log_wouldlog(isc_lctx, log_level)) {
|
|
+ char client_sabuf[ISC_SOCKADDR_FORMATSIZE];
|
|
+ char local_sabuf[ISC_SOCKADDR_FORMATSIZE];
|
|
+
|
|
+ isc_sockaddr_format(&session->handle->sock->peer, client_sabuf,
|
|
+ sizeof(client_sabuf));
|
|
+ isc_sockaddr_format(&session->handle->sock->iface, local_sabuf,
|
|
+ sizeof(local_sabuf));
|
|
+ isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL,
|
|
+ ISC_LOGMODULE_NETMGR, log_level,
|
|
+ "Dropping a flooding HTTP/2 peer "
|
|
+ "%s (on %s) - processed: %" PRIu64
|
|
+ " bytes, of them useful: %" PRIu64 "",
|
|
+ client_sabuf, local_sabuf,
|
|
+ session->processed_incoming_data,
|
|
+ session->processed_useful_data);
|
|
+ }
|
|
+}
|
|
+
|
|
+static bool
|
|
+http_is_flooding_peer(isc_nm_http_session_t *session) {
|
|
+ if (session->client) {
|
|
+ return false;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * A flooding client can try to open a lot of streams before
|
|
+ * submitting a request. Let's drop such clients.
|
|
+ */
|
|
+ if (session->received == 0 &&
|
|
+ session->total_opened_sstreams > MAX_STREAMS_BEFORE_FIRST_REQUEST)
|
|
+ {
|
|
+ return true;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * We have processed enough data to open at least one stream and
|
|
+ * get some useful data.
|
|
+ */
|
|
+ if (session->processed_incoming_data >
|
|
+ INCOMING_DATA_INITIAL_STREAM_SIZE &&
|
|
+ (session->total_opened_sstreams == 0 ||
|
|
+ session->processed_useful_data == 0))
|
|
+ {
|
|
+ return true;
|
|
+ }
|
|
+
|
|
+ if (session->processed_incoming_data < INCOMING_DATA_GRACE_SIZE) {
|
|
+ return false;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * The overhead of DoH per DNS message can be minimum 160-180
|
|
+ * bytes. We should allow more for extra information that can be
|
|
+ * included in headers, so let's use 256 bytes. Minimum DNS
|
|
+ * message size is 12 bytes. So, (256+12)/12=22. Even that can be
|
|
+ * too restricting for some edge cases, but should be good enough
|
|
+ * for any practical purposes. Not to mention that HTTP/2 may
|
|
+ * include legitimate data that is completely useless for DNS
|
|
+ * purposes...
|
|
+ *
|
|
+ * Anyway, at that point we should have processed enough requests
|
|
+ * for such clients (if any).
|
|
+ */
|
|
+ if (session->processed_useful_data == 0 ||
|
|
+ (session->processed_incoming_data /
|
|
+ session->processed_useful_data) > 22)
|
|
+ {
|
|
+ return true;
|
|
+ }
|
|
+
|
|
+ return false;
|
|
+}
|
|
+
|
|
/*
|
|
* Read callback from TLS socket.
|
|
*/
|
|
@@ -976,6 +1230,7 @@ http_readcb(isc_nmhandle_t *handle, isc_result_t result, isc_region_t *region,
|
|
isc_nm_http_session_t *session = (isc_nm_http_session_t *)data;
|
|
isc_nm_http_session_t *tmpsess = NULL;
|
|
ssize_t readlen;
|
|
+ isc_buffer_t input;
|
|
|
|
REQUIRE(VALID_HTTP2_SESSION(session));
|
|
|
|
@@ -994,11 +1249,17 @@ http_readcb(isc_nmhandle_t *handle, isc_result_t result, isc_region_t *region,
|
|
goto done;
|
|
}
|
|
|
|
- readlen = nghttp2_session_mem_recv(session->ngsession, region->base,
|
|
- region->length);
|
|
+ isc_buffer_init(&input, region->base, region->length);
|
|
+ isc_buffer_add(&input, region->length);
|
|
+
|
|
+ readlen = http_process_input_data(session, &input);
|
|
if (readlen < 0) {
|
|
failed_read_cb(ISC_R_UNEXPECTED, session);
|
|
goto done;
|
|
+ } else if (http_is_flooding_peer(session)) {
|
|
+ http_log_flooding_peer(session);
|
|
+ failed_read_cb(ISC_R_RANGE, session);
|
|
+ goto done;
|
|
}
|
|
|
|
if ((size_t)readlen < region->length) {
|
|
@@ -1011,11 +1272,12 @@ http_readcb(isc_nmhandle_t *handle, isc_result_t result, isc_region_t *region,
|
|
isc_buffer_putmem(session->buf, region->base + readlen,
|
|
unread_size);
|
|
isc_nm_pauseread(session->handle);
|
|
+ http_do_bio_async(session);
|
|
+ } else {
|
|
+ /* We might have something to receive or send, do IO */
|
|
+ http_do_bio(session, NULL, NULL, NULL);
|
|
}
|
|
|
|
- /* We might have something to receive or send, do IO */
|
|
- http_do_bio(session, NULL, NULL, NULL);
|
|
-
|
|
done:
|
|
isc__nm_httpsession_detach(&tmpsess);
|
|
}
|
|
@@ -1053,14 +1315,18 @@ http_writecb(isc_nmhandle_t *handle, isc_result_t result, void *arg) {
|
|
}
|
|
|
|
isc_buffer_free(&req->pending_write_data);
|
|
+ session->processed += req->submitted;
|
|
isc_mem_put(session->mctx, req, sizeof(*req));
|
|
|
|
session->sending--;
|
|
- http_do_bio(session, NULL, NULL, NULL);
|
|
- isc_nmhandle_detach(&transphandle);
|
|
- if (result != ISC_R_SUCCESS && session->sending == 0) {
|
|
+
|
|
+ if (result == ISC_R_SUCCESS) {
|
|
+ http_do_bio(session, NULL, NULL, NULL);
|
|
+ } else {
|
|
finish_http_session(session);
|
|
}
|
|
+ isc_nmhandle_detach(&transphandle);
|
|
+
|
|
isc__nm_httpsession_detach(&session);
|
|
}
|
|
|
|
@@ -1206,7 +1472,9 @@ http_send_outgoing(isc_nm_http_session_t *session, isc_nmhandle_t *httphandle,
|
|
*send = (isc_http_send_req_t){ .pending_write_data =
|
|
session->pending_write_data,
|
|
.cb = cb,
|
|
- .cbarg = cbarg };
|
|
+ .cbarg = cbarg,
|
|
+ .submitted = session->submitted };
|
|
+ session->submitted = 0;
|
|
session->pending_write_data = NULL;
|
|
move_pending_send_callbacks(session, send);
|
|
|
|
@@ -1227,6 +1495,27 @@ nothing_to_send:
|
|
return (false);
|
|
}
|
|
|
|
+static inline bool
|
|
+http_too_many_active_streams(isc_nm_http_session_t *session) {
|
|
+ const uint64_t active_streams = session->received - session->processed;
|
|
+ const uint64_t max_active_streams = ISC_MIN(
|
|
+ STREAM_CLIENTS_PER_CONN, session->max_concurrent_streams);
|
|
+
|
|
+ if (session->client) {
|
|
+ return false;
|
|
+ }
|
|
+
|
|
+ /*
|
|
+ * Do not process incoming data if there are too many active DNS
|
|
+ * clients (streams) per connection.
|
|
+ */
|
|
+ if (active_streams >= max_active_streams) {
|
|
+ return true;
|
|
+ }
|
|
+
|
|
+ return false;
|
|
+}
|
|
+
|
|
static void
|
|
http_do_bio(isc_nm_http_session_t *session, isc_nmhandle_t *send_httphandle,
|
|
isc_nm_cb_t send_cb, void *send_cbarg) {
|
|
@@ -1242,59 +1531,140 @@ http_do_bio(isc_nm_http_session_t *session, isc_nmhandle_t *send_httphandle,
|
|
finish_http_session(session);
|
|
}
|
|
return;
|
|
- } else if (nghttp2_session_want_read(session->ngsession) == 0 &&
|
|
- nghttp2_session_want_write(session->ngsession) == 0 &&
|
|
- session->pending_write_data == NULL)
|
|
- {
|
|
- session->closing = true;
|
|
+ }
|
|
+
|
|
+ if (send_cb != NULL) {
|
|
+ INSIST(VALID_NMHANDLE(send_httphandle));
|
|
+ (void)http_send_outgoing(session, send_httphandle, send_cb,
|
|
+ send_cbarg);
|
|
+ return;
|
|
+ }
|
|
+
|
|
+ INSIST(send_httphandle == NULL);
|
|
+ INSIST(send_cb == NULL);
|
|
+ INSIST(send_cbarg == NULL);
|
|
+
|
|
+ if (session->pending_write_data != NULL && session->sending == 0) {
|
|
+ (void)http_send_outgoing(session, NULL, NULL, NULL);
|
|
return;
|
|
}
|
|
|
|
if (nghttp2_session_want_read(session->ngsession) != 0) {
|
|
if (!session->reading) {
|
|
/* We have not yet started reading from this handle */
|
|
+ isc__nmsocket_timer_start(session->handle->sock);
|
|
isc_nm_read(session->handle, http_readcb, session);
|
|
session->reading = true;
|
|
} else if (session->buf != NULL) {
|
|
size_t remaining =
|
|
isc_buffer_remaininglength(session->buf);
|
|
/* Leftover data in the buffer, use it */
|
|
- size_t readlen = nghttp2_session_mem_recv(
|
|
- session->ngsession,
|
|
- isc_buffer_current(session->buf), remaining);
|
|
+ size_t remaining_after = 0;
|
|
+ ssize_t readlen = 0;
|
|
+ isc_nm_http_session_t *tmpsess = NULL;
|
|
|
|
- if (readlen == remaining) {
|
|
+ /*
|
|
+ * Let's ensure that HTTP/2 session and its associated
|
|
+ * data will not go "out of scope" too early.
|
|
+ */
|
|
+ isc__nm_httpsession_attach(session, &tmpsess);
|
|
+
|
|
+ readlen = http_process_input_data(session,
|
|
+ session->buf);
|
|
+
|
|
+ remaining_after =
|
|
+ isc_buffer_remaininglength(session->buf);
|
|
+
|
|
+ if (readlen < 0) {
|
|
+ failed_read_cb(ISC_R_UNEXPECTED, session);
|
|
+ } else if (http_is_flooding_peer(session)) {
|
|
+ http_log_flooding_peer(session);
|
|
+ failed_read_cb(ISC_R_RANGE, session);
|
|
+ } else if ((size_t)readlen == remaining) {
|
|
isc_buffer_free(&session->buf);
|
|
+ http_do_bio(session, NULL, NULL, NULL);
|
|
+ } else if (remaining_after > 0 &&
|
|
+ remaining_after < remaining)
|
|
+ {
|
|
+ /*
|
|
+ * We have processed a part of the data, now
|
|
+ * let's delay processing of whatever is left
|
|
+ * here. We want it to be an async operation so
|
|
+ * that we will:
|
|
+ *
|
|
+ * a) let other things run;
|
|
+ * b) have finer grained control over how much
|
|
+ * data is processed at once, because nghttp2
|
|
+ * would happily consume as much data we pass to
|
|
+ * it and that could overwhelm the server.
|
|
+ */
|
|
+ http_do_bio_async(session);
|
|
} else {
|
|
- isc_buffer_forward(session->buf, readlen);
|
|
+ (void)http_send_outgoing(session, NULL, NULL,
|
|
+ NULL);
|
|
}
|
|
|
|
- http_do_bio(session, send_httphandle, send_cb,
|
|
- send_cbarg);
|
|
+ isc__nm_httpsession_detach(&tmpsess);
|
|
return;
|
|
} else {
|
|
/* Resume reading, it's idempotent, wait for more */
|
|
isc_nm_resumeread(session->handle);
|
|
+ isc__nmsocket_timer_start(session->handle->sock);
|
|
}
|
|
} else {
|
|
/* We don't want more data, stop reading for now */
|
|
isc_nm_pauseread(session->handle);
|
|
}
|
|
|
|
- if (send_cb != NULL) {
|
|
- INSIST(VALID_NMHANDLE(send_httphandle));
|
|
- (void)http_send_outgoing(session, send_httphandle, send_cb,
|
|
- send_cbarg);
|
|
- } else {
|
|
- INSIST(send_httphandle == NULL);
|
|
- INSIST(send_cb == NULL);
|
|
- INSIST(send_cbarg == NULL);
|
|
- (void)http_send_outgoing(session, NULL, NULL, NULL);
|
|
+ /* we might have some data to send after processing */
|
|
+ (void)http_send_outgoing(session, NULL, NULL, NULL);
|
|
+
|
|
+ if (nghttp2_session_want_read(session->ngsession) == 0 &&
|
|
+ nghttp2_session_want_write(session->ngsession) == 0 &&
|
|
+ session->pending_write_data == NULL)
|
|
+ {
|
|
+ session->closing = true;
|
|
+ isc_nm_pauseread(session->handle);
|
|
+ if (session->sending == 0) {
|
|
+ finish_http_session(session);
|
|
+ }
|
|
}
|
|
|
|
return;
|
|
}
|
|
|
|
+static void
|
|
+http_do_bio_async_cb(void *arg) {
|
|
+ isc_nm_http_session_t *session = arg;
|
|
+
|
|
+ REQUIRE(VALID_HTTP2_SESSION(session));
|
|
+
|
|
+ if (session->handle != NULL &&
|
|
+ !isc__nmsocket_closing(session->handle->sock))
|
|
+ {
|
|
+ http_do_bio(session, NULL, NULL, NULL);
|
|
+ }
|
|
+
|
|
+ isc__nm_httpsession_detach(&session);
|
|
+}
|
|
+
|
|
+static void
|
|
+http_do_bio_async(isc_nm_http_session_t *session) {
|
|
+ isc_nm_http_session_t *tmpsess = NULL;
|
|
+
|
|
+ REQUIRE(VALID_HTTP2_SESSION(session));
|
|
+
|
|
+ if (session->handle == NULL ||
|
|
+ isc__nmsocket_closing(session->handle->sock))
|
|
+ {
|
|
+ return;
|
|
+ }
|
|
+ isc__nm_httpsession_attach(session, &tmpsess);
|
|
+ isc__nm_async_run(
|
|
+ &session->handle->sock->mgr->workers[session->handle->sock->tid],
|
|
+ http_do_bio_async_cb, tmpsess);
|
|
+}
|
|
+
|
|
static isc_result_t
|
|
get_http_cstream(isc_nmsocket_t *sock, http_cstream_t **streamp) {
|
|
http_cstream_t *cstream = sock->h2.connect.cstream;
|
|
@@ -1424,6 +1794,7 @@ transport_connect_cb(isc_nmhandle_t *handle, isc_result_t result, void *cbarg) {
|
|
}
|
|
|
|
http_transpost_tcp_nodelay(handle);
|
|
+ isc__nmhandle_set_manual_timer(session->handle, true);
|
|
|
|
http_call_connect_cb(http_sock, session, result);
|
|
|
|
@@ -1670,6 +2041,7 @@ server_on_begin_headers_callback(nghttp2_session *ngsession,
|
|
socket->tid = session->handle->sock->tid;
|
|
ISC_LINK_INIT(&socket->h2, link);
|
|
ISC_LIST_APPEND(session->sstreams, &socket->h2, link);
|
|
+ session->total_opened_sstreams++;
|
|
|
|
nghttp2_session_set_stream_user_data(ngsession, frame->hd.stream_id,
|
|
socket);
|
|
@@ -1730,6 +2102,8 @@ server_handle_path_header(isc_nmsocket_t *socket, const uint8_t *value,
|
|
socket->mgr->mctx, dns_value,
|
|
dns_value_len,
|
|
&socket->h2.query_data_len);
|
|
+ socket->h2.session->processed_useful_data +=
|
|
+ dns_value_len;
|
|
} else {
|
|
socket->h2.query_too_large = true;
|
|
return (ISC_HTTP_ERROR_PAYLOAD_TOO_LARGE);
|
|
@@ -2038,6 +2412,12 @@ server_call_cb(isc_nmsocket_t *socket, const isc_result_t result,
|
|
handle = isc__nmhandle_get(socket, NULL, NULL);
|
|
if (result != ISC_R_SUCCESS) {
|
|
data = NULL;
|
|
+ } else if (socket->h2.session->handle != NULL) {
|
|
+ isc__nmsocket_timer_restart(socket->h2.session->handle->sock);
|
|
+ }
|
|
+ if (result == ISC_R_SUCCESS) {
|
|
+ socket->h2.request_received = true;
|
|
+ socket->h2.session->received++;
|
|
}
|
|
socket->h2.cb(handle, result, data, socket->h2.cbarg);
|
|
isc_nmhandle_detach(&handle);
|
|
@@ -2054,6 +2434,12 @@ isc__nm_http_bad_request(isc_nmhandle_t *handle) {
|
|
REQUIRE(!atomic_load(&sock->client));
|
|
REQUIRE(VALID_HTTP2_SESSION(sock->h2.session));
|
|
|
|
+ if (sock->h2.response_submitted ||
|
|
+ !http_session_active(sock->h2.session))
|
|
+ {
|
|
+ return;
|
|
+ }
|
|
+
|
|
(void)server_send_error_response(ISC_HTTP_ERROR_BAD_REQUEST,
|
|
sock->h2.session->ngsession, sock);
|
|
}
|
|
@@ -2475,6 +2861,8 @@ httplisten_acceptcb(isc_nmhandle_t *handle, isc_result_t result, void *cbarg) {
|
|
isc__nmsocket_attach(httplistensock, &session->serversocket);
|
|
server_send_connection_header(session);
|
|
|
|
+ isc__nmhandle_set_manual_timer(session->handle, true);
|
|
+
|
|
/* TODO H2 */
|
|
http_do_bio(session, NULL, NULL, NULL);
|
|
return (ISC_R_SUCCESS);
|
|
diff --git a/lib/isc/netmgr/netmgr-int.h b/lib/isc/netmgr/netmgr-int.h
|
|
index cc635e3f5f9..79e970a2f6d 100644
|
|
--- a/lib/isc/netmgr/netmgr-int.h
|
|
+++ b/lib/isc/netmgr/netmgr-int.h
|
|
@@ -337,6 +337,7 @@ typedef enum isc__netievent_type {
|
|
netievent_privilegedtask,
|
|
|
|
netievent_settlsctx,
|
|
+ netievent_asyncrun,
|
|
|
|
/*
|
|
* event type values higher than this will be treated
|
|
@@ -708,6 +709,42 @@ typedef struct isc__netievent__tlsctx {
|
|
}
|
|
|
|
#ifdef HAVE_LIBNGHTTP2
|
|
+typedef void (*isc__nm_asyncrun_cb_t)(void *);
|
|
+
|
|
+typedef struct isc__netievent__asyncrun {
|
|
+ isc__netievent_type type;
|
|
+ ISC_LINK(isc__netievent_t) link;
|
|
+ isc__nm_asyncrun_cb_t cb;
|
|
+ void *cbarg;
|
|
+} isc__netievent__asyncrun_t;
|
|
+
|
|
+#define NETIEVENT_ASYNCRUN_TYPE(type) \
|
|
+ typedef isc__netievent__asyncrun_t isc__netievent_##type##_t;
|
|
+
|
|
+#define NETIEVENT_ASYNCRUN_DECL(type) \
|
|
+ isc__netievent_##type##_t *isc__nm_get_netievent_##type( \
|
|
+ isc_nm_t *nm, isc__nm_asyncrun_cb_t cb, void *cbarg); \
|
|
+ void isc__nm_put_netievent_##type(isc_nm_t *nm, \
|
|
+ isc__netievent_##type##_t *ievent);
|
|
+
|
|
+#define NETIEVENT_ASYNCRUN_DEF(type) \
|
|
+ isc__netievent_##type##_t *isc__nm_get_netievent_##type( \
|
|
+ isc_nm_t *nm, isc__nm_asyncrun_cb_t cb, void *cbarg) { \
|
|
+ isc__netievent_##type##_t *ievent = \
|
|
+ isc__nm_get_netievent(nm, netievent_##type); \
|
|
+ ievent->cb = cb; \
|
|
+ ievent->cbarg = cbarg; \
|
|
+ \
|
|
+ return (ievent); \
|
|
+ } \
|
|
+ \
|
|
+ void isc__nm_put_netievent_##type(isc_nm_t *nm, \
|
|
+ isc__netievent_##type##_t *ievent) { \
|
|
+ ievent->cb = NULL; \
|
|
+ ievent->cbarg = NULL; \
|
|
+ isc__nm_put_netievent(nm, ievent); \
|
|
+ }
|
|
+
|
|
typedef struct isc__netievent__http_eps {
|
|
NETIEVENT__SOCKET;
|
|
isc_nm_http_endpoints_t *endpoints;
|
|
@@ -752,6 +789,7 @@ typedef union {
|
|
isc__netievent_tlsconnect_t nitc;
|
|
isc__netievent__tlsctx_t nitls;
|
|
#ifdef HAVE_LIBNGHTTP2
|
|
+ isc__netievent__asyncrun_t niasync;
|
|
isc__netievent__http_eps_t nihttpeps;
|
|
#endif /* HAVE_LIBNGHTTP2 */
|
|
} isc__netievent_storage_t;
|
|
@@ -944,6 +982,7 @@ typedef struct isc_nmsocket_h2 {
|
|
|
|
isc_nm_http_endpoints_t *peer_endpoints;
|
|
|
|
+ bool request_received;
|
|
bool response_submitted;
|
|
struct {
|
|
char *uri;
|
|
@@ -1228,6 +1267,7 @@ struct isc_nmsocket {
|
|
|
|
isc_barrier_t barrier;
|
|
bool barrier_initialised;
|
|
+ atomic_bool manual_read_timer;
|
|
#ifdef NETMGR_TRACE
|
|
void *backtrace[TRACE_SIZE];
|
|
int backtrace_size;
|
|
@@ -1546,6 +1586,9 @@ isc__nm_tcp_settimeout(isc_nmhandle_t *handle, uint32_t timeout);
|
|
* Set the read timeout for the TCP socket associated with 'handle'.
|
|
*/
|
|
|
|
+void
|
|
+isc__nmhandle_tcp_set_manual_timer(isc_nmhandle_t *handle, const bool manual);
|
|
+
|
|
void
|
|
isc__nm_async_tcpconnect(isc__networker_t *worker, isc__netievent_t *ev0);
|
|
void
|
|
@@ -1788,6 +1831,9 @@ isc__nm_tls_cleartimeout(isc_nmhandle_t *handle);
|
|
* around.
|
|
*/
|
|
|
|
+void
|
|
+isc__nmhandle_tls_set_manual_timer(isc_nmhandle_t *handle, const bool manual);
|
|
+
|
|
const char *
|
|
isc__nm_tls_verify_tls_peer_result_string(const isc_nmhandle_t *handle);
|
|
|
|
@@ -1805,6 +1851,15 @@ void
|
|
isc__nmhandle_tls_setwritetimeout(isc_nmhandle_t *handle,
|
|
uint64_t write_timeout);
|
|
|
|
+bool
|
|
+isc__nmsocket_tls_timer_running(isc_nmsocket_t *sock);
|
|
+
|
|
+void
|
|
+isc__nmsocket_tls_timer_restart(isc_nmsocket_t *sock);
|
|
+
|
|
+void
|
|
+isc__nmsocket_tls_timer_stop(isc_nmsocket_t *sock);
|
|
+
|
|
void
|
|
isc__nm_http_stoplistening(isc_nmsocket_t *sock);
|
|
|
|
@@ -1897,7 +1952,10 @@ void
|
|
isc__nm_http_set_max_streams(isc_nmsocket_t *listener,
|
|
const uint32_t max_concurrent_streams);
|
|
|
|
-#endif
|
|
+void
|
|
+isc__nm_async_asyncrun(isc__networker_t *worker, isc__netievent_t *ev0);
|
|
+
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
|
|
void
|
|
isc__nm_async_settlsctx(isc__networker_t *worker, isc__netievent_t *ev0);
|
|
@@ -2093,6 +2151,8 @@ NETIEVENT_SOCKET_TYPE(tlsdnscycle);
|
|
NETIEVENT_SOCKET_REQ_TYPE(httpsend);
|
|
NETIEVENT_SOCKET_TYPE(httpclose);
|
|
NETIEVENT_SOCKET_HTTP_EPS_TYPE(httpendpoints);
|
|
+
|
|
+NETIEVENT_ASYNCRUN_TYPE(asyncrun);
|
|
#endif /* HAVE_LIBNGHTTP2 */
|
|
|
|
NETIEVENT_SOCKET_REQ_TYPE(tcpconnect);
|
|
@@ -2167,6 +2227,8 @@ NETIEVENT_SOCKET_DECL(tlsdnscycle);
|
|
NETIEVENT_SOCKET_REQ_DECL(httpsend);
|
|
NETIEVENT_SOCKET_DECL(httpclose);
|
|
NETIEVENT_SOCKET_HTTP_EPS_DECL(httpendpoints);
|
|
+
|
|
+NETIEVENT_ASYNCRUN_DECL(asyncrun);
|
|
#endif /* HAVE_LIBNGHTTP2 */
|
|
|
|
NETIEVENT_SOCKET_REQ_DECL(tcpconnect);
|
|
@@ -2283,3 +2345,20 @@ isc__nmsocket_writetimeout_cb(void *data, isc_result_t eresult);
|
|
|
|
void
|
|
isc__nmsocket_log_tls_session_reuse(isc_nmsocket_t *sock, isc_tls_t *tls);
|
|
+
|
|
+void
|
|
+isc__nmhandle_set_manual_timer(isc_nmhandle_t *handle, const bool manual);
|
|
+/*
|
|
+ * Set manual read timer control mode - so that it will not get reset
|
|
+ * automatically on read nor get started when read is initiated.
|
|
+ */
|
|
+
|
|
+#if HAVE_LIBNGHTTP2
|
|
+void
|
|
+isc__nm_async_run(isc__networker_t *worker, isc__nm_asyncrun_cb_t cb,
|
|
+ void *cbarg);
|
|
+/*
|
|
+ * Call the given callback asynchronously by the give network manager
|
|
+ * worker, pass the given argument to it.
|
|
+ */
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
diff --git a/lib/isc/netmgr/netmgr.c b/lib/isc/netmgr/netmgr.c
|
|
index a42ca90e8d2..8df6dc07408 100644
|
|
--- a/lib/isc/netmgr/netmgr.c
|
|
+++ b/lib/isc/netmgr/netmgr.c
|
|
@@ -998,6 +998,8 @@ process_netievent(isc__networker_t *worker, isc__netievent_t *ievent) {
|
|
NETIEVENT_CASE(httpsend);
|
|
NETIEVENT_CASE(httpclose);
|
|
NETIEVENT_CASE(httpendpoints);
|
|
+
|
|
+ NETIEVENT_CASE(asyncrun);
|
|
#endif
|
|
NETIEVENT_CASE(settlsctx);
|
|
NETIEVENT_CASE(sockstop);
|
|
@@ -1116,6 +1118,8 @@ NETIEVENT_SOCKET_DEF(tlsdnsshutdown);
|
|
NETIEVENT_SOCKET_REQ_DEF(httpsend);
|
|
NETIEVENT_SOCKET_DEF(httpclose);
|
|
NETIEVENT_SOCKET_HTTP_EPS_DEF(httpendpoints);
|
|
+
|
|
+NETIEVENT_ASYNCRUN_DEF(asyncrun);
|
|
#endif /* HAVE_LIBNGHTTP2 */
|
|
|
|
NETIEVENT_SOCKET_REQ_DEF(tcpconnect);
|
|
@@ -1627,6 +1631,7 @@ isc___nmsocket_init(isc_nmsocket_t *sock, isc_nm_t *mgr, isc_nmsocket_type type,
|
|
atomic_init(&sock->keepalive, false);
|
|
atomic_init(&sock->connected, false);
|
|
atomic_init(&sock->timedout, false);
|
|
+ atomic_init(&sock->manual_read_timer, false);
|
|
|
|
atomic_init(&sock->active_child_connections, 0);
|
|
|
|
@@ -2136,6 +2141,15 @@ void
|
|
isc__nmsocket_timer_restart(isc_nmsocket_t *sock) {
|
|
REQUIRE(VALID_NMSOCK(sock));
|
|
|
|
+ switch (sock->type) {
|
|
+#if HAVE_LIBNGHTTP2
|
|
+ case isc_nm_tlssocket:
|
|
+ return isc__nmsocket_tls_timer_restart(sock);
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
+ default:
|
|
+ break;
|
|
+ }
|
|
+
|
|
if (uv_is_closing((uv_handle_t *)&sock->read_timer)) {
|
|
return;
|
|
}
|
|
@@ -2170,7 +2184,16 @@ bool
|
|
isc__nmsocket_timer_running(isc_nmsocket_t *sock) {
|
|
REQUIRE(VALID_NMSOCK(sock));
|
|
|
|
- return (uv_is_active((uv_handle_t *)&sock->read_timer));
|
|
+ switch (sock->type) {
|
|
+#if HAVE_LIBNGHTTP2
|
|
+ case isc_nm_tlssocket:
|
|
+ return isc__nmsocket_tls_timer_running(sock);
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
+ default:
|
|
+ break;
|
|
+ }
|
|
+
|
|
+ return uv_is_active((uv_handle_t *)&sock->read_timer);
|
|
}
|
|
|
|
void
|
|
@@ -2190,6 +2213,15 @@ isc__nmsocket_timer_stop(isc_nmsocket_t *sock) {
|
|
|
|
REQUIRE(VALID_NMSOCK(sock));
|
|
|
|
+ switch (sock->type) {
|
|
+#if HAVE_LIBNGHTTP2
|
|
+ case isc_nm_tlssocket:
|
|
+ return isc__nmsocket_tls_timer_stop(sock);
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
+ default:
|
|
+ break;
|
|
+ }
|
|
+
|
|
/* uv_timer_stop() is idempotent, no need to check if running */
|
|
|
|
r = uv_timer_stop(&sock->read_timer);
|
|
@@ -3946,6 +3978,52 @@ isc__nmsocket_log_tls_session_reuse(isc_nmsocket_t *sock, isc_tls_t *tls) {
|
|
client_sabuf, local_sabuf);
|
|
}
|
|
|
|
+void
|
|
+isc__nmhandle_set_manual_timer(isc_nmhandle_t *handle, const bool manual) {
|
|
+ REQUIRE(VALID_NMHANDLE(handle));
|
|
+ REQUIRE(VALID_NMSOCK(handle->sock));
|
|
+
|
|
+ isc_nmsocket_t *sock = handle->sock;
|
|
+
|
|
+ switch (sock->type) {
|
|
+ case isc_nm_tcpsocket:
|
|
+ isc__nmhandle_tcp_set_manual_timer(handle, manual);
|
|
+ return;
|
|
+#if HAVE_LIBNGHTTP2
|
|
+ case isc_nm_tlssocket:
|
|
+ isc__nmhandle_tls_set_manual_timer(handle, manual);
|
|
+ return;
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
+ default:
|
|
+ break;
|
|
+ };
|
|
+
|
|
+ UNREACHABLE();
|
|
+}
|
|
+
|
|
+#if HAVE_LIBNGHTTP2
|
|
+void
|
|
+isc__nm_async_run(isc__networker_t *worker, isc__nm_asyncrun_cb_t cb,
|
|
+ void *cbarg) {
|
|
+ isc__netievent__asyncrun_t *ievent = NULL;
|
|
+ REQUIRE(worker != NULL);
|
|
+ REQUIRE(cb != NULL);
|
|
+
|
|
+ ievent = isc__nm_get_netievent_asyncrun(worker->mgr, cb, cbarg);
|
|
+ isc__nm_enqueue_ievent(worker, (isc__netievent_t *)ievent);
|
|
+}
|
|
+
|
|
+void
|
|
+isc__nm_async_asyncrun(isc__networker_t *worker, isc__netievent_t *ev0) {
|
|
+ isc__netievent_asyncrun_t *ievent = (isc__netievent_asyncrun_t *)ev0;
|
|
+
|
|
+ UNUSED(worker);
|
|
+
|
|
+ ievent->cb(ievent->cbarg);
|
|
+}
|
|
+
|
|
+#endif /* HAVE_LIBNGHTTP2 */
|
|
+
|
|
#ifdef NETMGR_TRACE
|
|
/*
|
|
* Dump all active sockets in netmgr. We output to stderr
|
|
diff --git a/lib/isc/netmgr/tcp.c b/lib/isc/netmgr/tcp.c
|
|
index 37d44bd9c84..925bc85e028 100644
|
|
--- a/lib/isc/netmgr/tcp.c
|
|
+++ b/lib/isc/netmgr/tcp.c
|
|
@@ -784,7 +784,9 @@ isc__nm_async_tcpstartread(isc__networker_t *worker, isc__netievent_t *ev0) {
|
|
return;
|
|
}
|
|
|
|
- isc__nmsocket_timer_start(sock);
|
|
+ if (!atomic_load(&sock->manual_read_timer)) {
|
|
+ isc__nmsocket_timer_start(sock);
|
|
+ }
|
|
}
|
|
|
|
void
|
|
@@ -822,7 +824,9 @@ isc__nm_async_tcppauseread(isc__networker_t *worker, isc__netievent_t *ev0) {
|
|
REQUIRE(sock->tid == isc_nm_tid());
|
|
UNUSED(worker);
|
|
|
|
- isc__nmsocket_timer_stop(sock);
|
|
+ if (!atomic_load(&sock->manual_read_timer)) {
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
+ }
|
|
isc__nm_stop_reading(sock);
|
|
}
|
|
|
|
@@ -931,8 +935,10 @@ isc__nm_tcp_read_cb(uv_stream_t *stream, ssize_t nread, const uv_buf_t *buf) {
|
|
}
|
|
}
|
|
|
|
- /* The timer will be updated */
|
|
- isc__nmsocket_timer_restart(sock);
|
|
+ if (!atomic_load(&sock->manual_read_timer)) {
|
|
+ /* The timer will be updated */
|
|
+ isc__nmsocket_timer_restart(sock);
|
|
+ }
|
|
}
|
|
|
|
free:
|
|
@@ -1521,3 +1527,15 @@ isc__nm_tcp_listener_nactive(isc_nmsocket_t *listener) {
|
|
INSIST(nactive >= 0);
|
|
return (nactive);
|
|
}
|
|
+
|
|
+void
|
|
+isc__nmhandle_tcp_set_manual_timer(isc_nmhandle_t *handle, const bool manual) {
|
|
+ isc_nmsocket_t *sock;
|
|
+
|
|
+ REQUIRE(VALID_NMHANDLE(handle));
|
|
+ sock = handle->sock;
|
|
+ REQUIRE(VALID_NMSOCK(sock));
|
|
+ REQUIRE(sock->type == isc_nm_tcpsocket);
|
|
+
|
|
+ atomic_store(&sock->manual_read_timer, manual);
|
|
+}
|
|
diff --git a/lib/isc/netmgr/tlsstream.c b/lib/isc/netmgr/tlsstream.c
|
|
index 4fef5985b25..3d78df6a4d1 100644
|
|
--- a/lib/isc/netmgr/tlsstream.c
|
|
+++ b/lib/isc/netmgr/tlsstream.c
|
|
@@ -60,6 +60,12 @@ tls_error_to_result(const int tls_err, const int tls_state, isc_tls_t *tls) {
|
|
}
|
|
}
|
|
|
|
+static void
|
|
+tls_read_start(isc_nmsocket_t *sock);
|
|
+
|
|
+static void
|
|
+tls_read_stop(isc_nmsocket_t *sock);
|
|
+
|
|
static void
|
|
tls_failed_read_cb(isc_nmsocket_t *sock, const isc_result_t result);
|
|
|
|
@@ -203,8 +209,13 @@ tls_failed_read_cb(isc_nmsocket_t *sock, const isc_result_t result) {
|
|
tls_call_connect_cb(sock, handle, result);
|
|
isc__nmsocket_clearcb(sock);
|
|
isc_nmhandle_detach(&handle);
|
|
- } else if (sock->recv_cb != NULL && sock->statichandle != NULL &&
|
|
- (sock->recv_read || result == ISC_R_TIMEDOUT))
|
|
+ goto do_destroy;
|
|
+ }
|
|
+
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
+
|
|
+ if (sock->recv_cb != NULL && sock->statichandle != NULL &&
|
|
+ (sock->recv_read || result == ISC_R_TIMEDOUT))
|
|
{
|
|
isc__nm_uvreq_t *req = NULL;
|
|
INSIST(VALID_NMHANDLE(sock->statichandle));
|
|
@@ -218,13 +229,13 @@ tls_failed_read_cb(isc_nmsocket_t *sock, const isc_result_t result) {
|
|
}
|
|
isc__nm_readcb(sock, req, result);
|
|
if (result == ISC_R_TIMEDOUT &&
|
|
- (sock->outerhandle == NULL ||
|
|
- isc__nmsocket_timer_running(sock->outerhandle->sock)))
|
|
+ isc__nmsocket_timer_running(sock))
|
|
{
|
|
destroy = false;
|
|
}
|
|
}
|
|
|
|
+do_destroy:
|
|
if (destroy) {
|
|
isc__nmsocket_prep_destroy(sock);
|
|
}
|
|
@@ -344,6 +355,8 @@ tls_try_handshake(isc_nmsocket_t *sock, isc_result_t *presult) {
|
|
INSIST(sock->statichandle == NULL);
|
|
isc__nmsocket_log_tls_session_reuse(sock, sock->tlsstream.tls);
|
|
tlshandle = isc__nmhandle_get(sock, &sock->peer, &sock->iface);
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
+ tls_read_stop(sock);
|
|
|
|
if (isc__nm_closing(sock)) {
|
|
result = ISC_R_SHUTTINGDOWN;
|
|
@@ -437,6 +450,7 @@ tls_do_bio(isc_nmsocket_t *sock, isc_region_t *received_data,
|
|
sock->tlsstream.state = TLS_HANDSHAKE;
|
|
rv = tls_try_handshake(sock, NULL);
|
|
INSIST(SSL_is_init_finished(sock->tlsstream.tls) == 0);
|
|
+ isc__nmsocket_timer_restart(sock);
|
|
} else if (sock->tlsstream.state == TLS_CLOSED) {
|
|
return;
|
|
} else { /* initialised and doing I/O */
|
|
@@ -502,6 +516,7 @@ tls_do_bio(isc_nmsocket_t *sock, isc_region_t *received_data,
|
|
!atomic_load(&sock->readpaused) &&
|
|
sock->statichandle != NULL && !finish)
|
|
{
|
|
+ bool was_new_data = false;
|
|
uint8_t recv_buf[TLS_BUF_SIZE];
|
|
INSIST(sock->tlsstream.state > TLS_HANDSHAKE);
|
|
while ((rv = SSL_read_ex(sock->tlsstream.tls, recv_buf,
|
|
@@ -510,7 +525,7 @@ tls_do_bio(isc_nmsocket_t *sock, isc_region_t *received_data,
|
|
isc_region_t region;
|
|
region = (isc_region_t){ .base = &recv_buf[0],
|
|
.length = len };
|
|
-
|
|
+ was_new_data = true;
|
|
INSIST(VALID_NMHANDLE(sock->statichandle));
|
|
sock->recv_cb(sock->statichandle, ISC_R_SUCCESS,
|
|
®ion, sock->recv_cbarg);
|
|
@@ -547,8 +562,29 @@ tls_do_bio(isc_nmsocket_t *sock, isc_region_t *received_data,
|
|
break;
|
|
}
|
|
}
|
|
+
|
|
+ if (was_new_data && !sock->manual_read_timer) {
|
|
+ /*
|
|
+ * Some data has been decrypted, it is the right
|
|
+ * time to stop the read timer as it will be
|
|
+ * restarted on the next read attempt.
|
|
+ */
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
+ }
|
|
}
|
|
}
|
|
+
|
|
+ /*
|
|
+ * Setting 'finish' to 'true' means that we are about to close the
|
|
+ * TLS stream (we intend to send TLS shutdown message to the
|
|
+ * remote side). After that no new data can be received, so we
|
|
+ * should stop the timer regardless of the
|
|
+ * 'sock->manual_read_timer' value.
|
|
+ */
|
|
+ if (finish) {
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
+ }
|
|
+
|
|
errno = 0;
|
|
tls_status = SSL_get_error(sock->tlsstream.tls, rv);
|
|
saved_errno = errno;
|
|
@@ -601,14 +637,7 @@ tls_do_bio(isc_nmsocket_t *sock, isc_region_t *received_data,
|
|
return;
|
|
}
|
|
|
|
- INSIST(VALID_NMHANDLE(sock->outerhandle));
|
|
-
|
|
- if (sock->tlsstream.reading) {
|
|
- isc_nm_resumeread(sock->outerhandle);
|
|
- } else if (sock->tlsstream.state == TLS_HANDSHAKE) {
|
|
- sock->tlsstream.reading = true;
|
|
- isc_nm_read(sock->outerhandle, tls_readcb, sock);
|
|
- }
|
|
+ tls_read_start(sock);
|
|
return;
|
|
default:
|
|
result = tls_error_to_result(tls_status, sock->tlsstream.state,
|
|
@@ -743,6 +772,7 @@ tlslisten_acceptcb(isc_nmhandle_t *handle, isc_result_t result, void *cbarg) {
|
|
RUNTIME_CHECK(result == ISC_R_SUCCESS);
|
|
/* TODO: catch failure code, detach tlssock, and log the error */
|
|
|
|
+ isc__nmhandle_set_manual_timer(tlssock->outerhandle, true);
|
|
tls_do_bio(tlssock, NULL, NULL, false);
|
|
return (result);
|
|
}
|
|
@@ -898,6 +928,29 @@ isc__nm_tls_read(isc_nmhandle_t *handle, isc_nm_recv_cb_t cb, void *cbarg) {
|
|
(isc__netievent_t *)ievent);
|
|
}
|
|
|
|
+static void
|
|
+tls_read_start(isc_nmsocket_t *sock) {
|
|
+ INSIST(VALID_NMHANDLE(sock->outerhandle));
|
|
+
|
|
+ if (sock->tlsstream.reading) {
|
|
+ isc_nm_resumeread(sock->outerhandle);
|
|
+ } else if (sock->tlsstream.state == TLS_HANDSHAKE) {
|
|
+ sock->tlsstream.reading = true;
|
|
+ isc_nm_read(sock->outerhandle, tls_readcb, sock);
|
|
+ }
|
|
+
|
|
+ if (!sock->manual_read_timer) {
|
|
+ isc__nmsocket_timer_start(sock);
|
|
+ }
|
|
+}
|
|
+
|
|
+static void
|
|
+tls_read_stop(isc_nmsocket_t *sock) {
|
|
+ if (sock->outerhandle != NULL) {
|
|
+ isc_nm_pauseread(sock->outerhandle);
|
|
+ }
|
|
+}
|
|
+
|
|
void
|
|
isc__nm_tls_pauseread(isc_nmhandle_t *handle) {
|
|
REQUIRE(VALID_NMHANDLE(handle));
|
|
@@ -906,9 +959,11 @@ isc__nm_tls_pauseread(isc_nmhandle_t *handle) {
|
|
if (atomic_compare_exchange_strong(&handle->sock->readpaused,
|
|
&(bool){ false }, true))
|
|
{
|
|
- if (handle->sock->outerhandle != NULL) {
|
|
- isc_nm_pauseread(handle->sock->outerhandle);
|
|
+ if (!atomic_load(&handle->sock->manual_read_timer)) {
|
|
+ isc__nmsocket_timer_stop(handle->sock);
|
|
}
|
|
+
|
|
+ tls_read_stop(handle->sock);
|
|
}
|
|
}
|
|
|
|
@@ -937,6 +992,7 @@ tls_close_direct(isc_nmsocket_t *sock) {
|
|
* external references, we can close everything.
|
|
*/
|
|
if (sock->outerhandle != NULL) {
|
|
+ isc__nmsocket_timer_stop(sock);
|
|
isc_nm_pauseread(sock->outerhandle);
|
|
isc__nmsocket_clearcb(sock->outerhandle->sock);
|
|
isc_nmhandle_detach(&sock->outerhandle);
|
|
@@ -1085,6 +1141,7 @@ tcp_connected(isc_nmhandle_t *handle, isc_result_t result, void *cbarg) {
|
|
*/
|
|
handle->sock->tlsstream.tlssocket = tlssock;
|
|
|
|
+ isc__nmhandle_set_manual_timer(tlssock->outerhandle, true);
|
|
tls_do_bio(tlssock, NULL, NULL, false);
|
|
return;
|
|
error:
|
|
@@ -1251,6 +1308,44 @@ isc__nmhandle_tls_setwritetimeout(isc_nmhandle_t *handle,
|
|
}
|
|
}
|
|
|
|
+bool
|
|
+isc__nmsocket_tls_timer_running(isc_nmsocket_t *sock) {
|
|
+ REQUIRE(VALID_NMSOCK(sock));
|
|
+ REQUIRE(sock->type == isc_nm_tlssocket);
|
|
+
|
|
+ if (sock->outerhandle != NULL) {
|
|
+ INSIST(VALID_NMHANDLE(sock->outerhandle));
|
|
+ REQUIRE(VALID_NMSOCK(sock->outerhandle->sock));
|
|
+ return isc__nmsocket_timer_running(sock->outerhandle->sock);
|
|
+ }
|
|
+
|
|
+ return false;
|
|
+}
|
|
+
|
|
+void
|
|
+isc__nmsocket_tls_timer_restart(isc_nmsocket_t *sock) {
|
|
+ REQUIRE(VALID_NMSOCK(sock));
|
|
+ REQUIRE(sock->type == isc_nm_tlssocket);
|
|
+
|
|
+ if (sock->outerhandle != NULL) {
|
|
+ INSIST(VALID_NMHANDLE(sock->outerhandle));
|
|
+ REQUIRE(VALID_NMSOCK(sock->outerhandle->sock));
|
|
+ isc__nmsocket_timer_restart(sock->outerhandle->sock);
|
|
+ }
|
|
+}
|
|
+
|
|
+void
|
|
+isc__nmsocket_tls_timer_stop(isc_nmsocket_t *sock) {
|
|
+ REQUIRE(VALID_NMSOCK(sock));
|
|
+ REQUIRE(sock->type == isc_nm_tlssocket);
|
|
+
|
|
+ if (sock->outerhandle != NULL) {
|
|
+ INSIST(VALID_NMHANDLE(sock->outerhandle));
|
|
+ REQUIRE(VALID_NMSOCK(sock->outerhandle->sock));
|
|
+ isc__nmsocket_timer_stop(sock->outerhandle->sock);
|
|
+ }
|
|
+}
|
|
+
|
|
const char *
|
|
isc__nm_tls_verify_tls_peer_result_string(const isc_nmhandle_t *handle) {
|
|
isc_nmsocket_t *sock = NULL;
|
|
@@ -1351,3 +1446,15 @@ tls_try_shutdown(isc_tls_t *tls, const bool force) {
|
|
(void)SSL_shutdown(tls);
|
|
}
|
|
}
|
|
+
|
|
+void
|
|
+isc__nmhandle_tls_set_manual_timer(isc_nmhandle_t *handle, const bool manual) {
|
|
+ isc_nmsocket_t *sock;
|
|
+
|
|
+ REQUIRE(VALID_NMHANDLE(handle));
|
|
+ sock = handle->sock;
|
|
+ REQUIRE(VALID_NMSOCK(sock));
|
|
+ REQUIRE(sock->type == isc_nm_tlssocket);
|
|
+
|
|
+ atomic_store(&sock->manual_read_timer, manual);
|
|
+}
|
|
--
|
|
2.48.1
|
|
|