Previously we scheduled the transmission of response HEADERS using
priority tree in the belief that it allows more better utilization of
bandwidth for prioritized streams. But to reduce the overhead of
reconstruction of priority queue when connection level flow control
window is depleted, we just don't check priority tree in this case.
This means that response HEADERS frames are not sent even though they
are not flow controlled. This could waste bandwidth. To improve this
situation, we stop scheduling response HEADERS with priority tree for
now. Now they are just sent in the order they submitted. The
response body DATA continued to be scheduled with priority tree as
before.
return session_inflate_handle_invalid_stream(...) case is for streams
for INITIAL state, but this is rare case. In general, we'd like to
reduce RST_STREAM transmission, and it is suffice to ignore this frame
for now.
Previously, we only updated stream's weight field when only weight was
changed by PRIORITY frame. If stream is queued, it would be better to
actually reschedule it based on new weight. This could be especially
useful if weight is increased.
Previously, we updated descendant_last_cycle in
nghttp2_stream_reschedule, which is called after non-zero DATA frame.
But this was not optimal since we still had old descendant_last_cycle,
and new stream was scheduled based on it. Now descendant_last_cycle
is updated in nghttp2_stream_next_outbound_item, which is called when
stream with highest priority is selected from queue. And new stream
is scheduled based on it. This commit also removes 0-reset of
descendant_last_cycle and cycle in nghttp2_stream_reschedule. This
could help making them lower, so that they are not overflow. But
there is a pattern that it doesn't work, and we are not sure they are
really useful at this moment.
Previously, stream object for pushed resource was not created during
nghttp2_submit_push_promise(). It was created just before
nghttp2_before_frame_send_callback was called for that PUSH_PROMISE
frame. This means that application could not call
nghttp2_submit_response for the pushed resource before
nghttp2_before_frame_send_callback was called. This could be solved
by callback chaining, but for web server with back pressure from
backend stream, it is a bit unnecessarily hard to use.
This commit changes nghttp2_submit_push_promise() behaviour so that
stream object is created during that call. It makes application call
nghttp2_submit_response right after successful
nghttp2_submit_push_promise call.
Previously, nghttp2_session_end_request_headers_received assumes
stream is still writable (in other words, local endpoint has not sent
END_STREAM). But this assumption is false, because application can
send response in nghttp2_on_begin_frame_callback. Probably, this
assumption was made before the callback was introduced. This commit
addresses this issue. Since all
nghttp2_session_end_*_headers_received functions are identical, we
refactored them into one function.
Because of the nature of heap based priority queue, and our compare
function, streams with the same weight and same parent are handled in
the reverse order they pushed to the queue. To allow stream pushed
earlier to be served first, secondary key "seq" is introduced to break
a tie. "seq" is monotonically increased integer per parent stream,
and it is assigned to stream when it is pused to the queue, and gets
incremented.
Previously, nghttp2_session_find_stream(session, 0) returned NULL
despite the fact that documentation said that it should return root
stream. Now it is corrected, and it returns root stream as
documented.
The added API is nghttp2_session_change_stream_priority(). This
provides the same functionality to re-prioritize stream when PRIORITY
frame. is received, but we do it without PRIORITY frame. This could
be useful for server to change pushed stream's priority silently.
When stream is removed from tree, stream is either closed, or its
remote flow control window is depleted. In the latter case, we
schedule this stream as fast as possible if its remote window gets
positive, since it did not sent anything in its turn. To achieve
this, reset last_writelen to 0 when stream is removed from tree.
It has no usecase at the moment. It is most likely that applications
know the flags when it submitted extension frame, no need to modify it
later. Possibly feature bloat.
For clients, CANCEL is more appropriate for both incoming/outgoing
streams. For servers, CANCEL is appropriate for its pushed stream
(outgoing), but REFUSED_STREAM is more appropriate for incoming
stream.
To validate actual response body length against the value declared in
content-length response header field, we first check request method.
If request method is HEAD, respose body must be 0 regardless of the
value in content-length. nghttp2_session_upgrade() has no parameter
to indicate the request method is HEAD, so we failed to validate
response body if HEAD is used with HTTP Upgrade. New
nghttp2_session_upgrade2() accepts new parameter to indicate that
request method is HEAD or not to fix this issue. Although, this issue
affects client side only, we deprecate nghttp2_session_upgrade() in
favor of nghttp2_session_upgrade2() for both client and server side.
By default, we check the length of response body matches
content-length. For HEAD request, this is not necessarily true, so we
sniff request method, and if it is HEAD, make sure that response body
length is 0. But this does not work for HTTP Upgrade, since
nghttp2_session_upgrade() has no parameter to tell the request method
was HEAD. This commit disables this response body length validation
for the stream upgraded by HTTP Upgrade. We will add new version of
nghttp2_session_upgrade with the parameter to pass the request method
information so that we can handle this situation properly.
This function is useful for the client application to know that there
is a chance that request can be sent. If this function returns 0,
there is zero chance to make a request.
This commit also set error_code passed to
nghttp2_on_stream_close_callback to NGHTTP2_REFUSED_STREAM if request
is not sent.
The encoder is not required to send dynamic table size update if the
table size is not changed from the previous value after accepting new
maximum value.
This will improve performance since we can avoid indirect call of
internal functions. The downside is we now require libnghttp2 static
library to run unit tests.
Application can utilize this framework to send/receive user defined
extension frames. These frames are expected not to change existing
protocol behaviour.
If application returns NGHTTP2_ERR_PAUSE from send_data_callback, it
means application processed all data, but wants to make
nghttp2_session_mem_send or nghttp2_session_send return immediately.
This is useful if application writes to fixed sized buffers, and there
is no room to write more data.
This change adds new return error code from nghttp2_session_mem_recv
and nghttp2_session_recv functions, namely NGHTTP2_ERR_FLOODED. It is
fatal error, and is returned when flooding was detected.
If it is called through libnghttp2 internally, name/value pairs are
all NULL-terminated. But it is one of public API, and we cannot
expect that applications always make NULL-terminated string for
name/value pairs.
RFC 7540 does not enforce any limit on the number of incoming reserved
streams (in RFC 7540 terms, streams in reserved (remote) state). This
only affects client side, since only server can push streams.
Malicious server can push arbitrary number of streams, and make
client's memory exhausted. The new option,
nghttp2_set_max_reserved_remote_streams, can set the maximum number of
such incoming streams to avoid possible memory exhaustion. If this
option is set, and pushed streams are automatically closed on
reception, without calling user provided callback, if they exceed the
given limit. The default value is 200. If session is configured as
server side, this option has no effect. Server can control the number
of streams to push.
The intention of this stream API is give server application about
stream dependency information, so that it can utilize it for better
scheduling of stream processing. We have no plan to add object
oriented API based on stream object.
We now use priority queue per stream, which contains the stream which
has ready to send a frame, or one of its descendants have a frame to
send. We maintain invariant that if a stream is queued, then its
ancestors are also queued (except for root). When we re-schedule
stream after transmission, we re-schedule all ancestors, so that
streams on the other path can get a chance to send. This is basically
the same mechanism h2o project uses, but there are differences in the
details.
Previously, the number of stream in one dependency tree (not including
root) is limited to 120. This is due to the fact that we use
recursive calls to traverse trees. Now we replaced recursive calls
with loop, we can remove this limitation. Also now all streams are
descendant of root stream, rather than linked list of individual
subtree root.
Tatsuhiro Tsujikawa
Alexis La Goutte