For HTX streams, info about the HTX structure is now dumped for the request and
the response channels in "show sess all" command.
The patch may be backported to 1.9.
Now the H2 mux will parse and encode the HTX trailers blocks and send
the corresponding HEADERS frame. Since these blocks contain pure H1
trailers which may be fragmented on line boundaries, if first needs
to collect all of them, parse them using the H1 parser, build a list
and finally encode all of them at once once the EOM is met. Note that
this HEADERS frame always carries the end-of-headers and end-of-stream
flags.
This was tested using the helloworld examples from the grpc project,
as well as with the h2c tools. It doesn't seem possible at the moment
to test tailers using varnishtest though.
Currently the H1 headers parser works for either a request or a response
because it starts from the start line. It is also able to resume its
processing when it was interrupted, but in this case it doesn't update
the list.
Make it support a new flag, H1_MF_HDRS_ONLY so that the caller can
indicate it's only interested in the headers list and not the start
line. This will be convenient to parse H1 trailers.
We want to make sure we won't emit another empty DATA frame if we meet
HTX_BLK_EOM after and end of stream was already sent. For now it cannot
happen as far as HTX is respected, but with trailers it may become
ambiguous.
Recent commit 4710d20 ("BUG/MEDIUM: mux-h1: make HTX chunking
consistent with H2") tried to address chunking inconsistencies between
H1/HTX/H2 and has enforced it on every outgoing message carrying
H1_MF_XFER_LEN without H1_MF_CLEN nor H1_MF_CHNK. But it also does it
on requests, which is not appropriate since a request by default
doesn't have a message body unless explicitly mentioned. Also make
sure we only do this on HTTP/1.1 messages.
The problem is to guarantee the highest level of compatibility between
H1/H1, H1/H2, H2/H1 in each direction regarding the lack of content-
length. We have this truth table (a star '*' indicates which one can
pass trailers) :
H1 client -> H1 server :
request:
CL=0 TE=0 XL=1 -> CL=0 TE=0
CL=0 TE=1 XL=1 -> CL=0 TE=1 *
CL=1 TE=0 XL=1 -> CL=1 TE=0
CL=1 TE=1 XL=1 -> CL=1 TE=1 *
response:
CL=0 TE=0 XL=0 -> CL=0 TE=0
CL=0 TE=1 XL=1 -> CL=0 TE=1 *
CL=1 TE=0 XL=1 -> CL=1 TE=0
CL=1 TE=1 XL=1 -> CL=1 TE=1 *
H2 client -> H1 server : (H2 messages always carry XFER_LEN)
request:
CL=0 XL=1 -> CL=0 TE=0
CL=1 XL=1 -> CL=1 TE=0
response:
CL=0 TE=0 XL=0 -> CL=0
CL=0 TE=1 XL=1 -> CL=0 *
CL=1 TE=0 XL=1 -> CL=1
CL=1 TE=1 XL=1 -> CL=1 *
H1 client -> H2 server : (H2 messages always carry XFER_LEN)
request:
CL=0 TE=0 XL=1 -> CL=0
CL=0 TE=1 XL=1 -> CL=0 *
CL=1 TE=0 XL=1 -> CL=1
CL=1 TE=1 XL=1 -> CL=1 *
response:
CL=0 XL=1 -> CL=0 TE=1 *
CL=1 XL=1 -> CL=1 TE=0
For H1 client to H2 server, it will be possible to rely on the presence
of "TE: trailers" in the H1 request to automatically switch to chunks
in the response, and be able to pass trailers at the end. For now this
check is not implemented so an H2 response missing a content-length to
an H1 request will always have a transfer-encoding header added and
trailers will be forwarded if any.
This patch depends on previous commit "MINOR: mux-h1: parse the
content-length header on output and set H1_MF_CLEN" to work properly.
Since the aforementioned commit is scheduled for backport to 1.9 this
commit must also be backported to 1.9.
The H1_MF_CLEN flag is needed to figure whether a content-length header is
present or not when producing a request, so let's check it on output just
like we already check the transfer-encoding header.
This function is usable to transform a list of H2 header fields to a
HTX trailers block. It takes care of rejecting forbidden headers and
pseudo-headers when performing the conversion. It also emits the
trailing CRLF that is currently needed in the HTX trailers block.
When forwarding an H2 request to an H1 server, if the request doesn't
have a content-length header field, it is chunked. In this case it is
possible to send trailers to the server, which is what this patch does.
If the transfer is performed without chunking, then the trailers are
silently discarded.
This function is usable to transform a list of H2 header fields to a
H1 trailers block. It takes care of rejecting forbidden headers and
pseudo-headers when performing the conversion.
This is not exactly a bug but a long-time design limitation. We used not
to decode trailers in H2, resulting in broken connections each time a
trailer was sent, since it was impossible to keep the HPACK decompressor
synchronized. Now that the sequencing of operations permits it, we must
make sure to at least properly decode them.
What we try to do is to identify if a HEADERS frame was already seen and
use this indication to know if it's a headers or a trailers. For this,
h2c_decode_headers() checks if the stream indicates that a HEADERS frame
was already received. If so, it decodes it and emits the trailing
0 CRLF CRLF in case of H1, or the HTX_EOD + HTX_EOM blocks in case of HTX,
to terminate the data stream.
The trailers contents are still deleted for now but the request works, and
the connection remains synchronized and usable for subsequent streams.
The correctness may be tested using a simple config and h2spec :
h2spec -o 1000 -v -t -S -k -h 127.0.0.1 -p 4443 generic/4/4
This should definitely be backported to 1.9 given the low impact for the
benefit. However it cannot be backported to 1.8 since the operations cannot
be resumed. The following patches are also needed with this one :
MINOR: mux-h2: make h2c_decode_headers() return a status, not a count
MINOR: mux-h2: add a new dummy stream : h2_error_stream
MEDIUM: mux-h2: make h2c_decode_headers() support recoverable errors
BUG/MINOR: mux-h2: detect when the HTX EOM block cannot be added after headers
MINOR: mux-h2: check for too many streams only for idle streams
MINOR: mux-h2: set H2_SF_HEADERS_RCVD when a HEADERS frame was decoded
The HEADERS frame parser checks if we still have too many streams, but
this should only be done for idle streams, otherwise it would prevent
us from processing trailer frames.
In h2c_frt_handle_headers() and h2c_bck_handle_headers() we have an unused
error path made of the strm_err label, while send_rst is used to emit an
RST upon stream error after forcing the stream to h2_refused_stream. Let's
remove this unused strm_err block now.
In h2c_frt_handle_headers(), we test the stream for SS_ERROR just after
setting it to SS_OPEN, this makes no sense and creates confusion in the
error path. Remove this misleading test.
In case we receive a very large HEADERS frame which doesn't leave enough
room to place the EOM block after the decoded headers, we must fail the
stream. This test was missing, resulting in the loss of the EOM, possibly
leaving the stream waiting for a time-out.
Note that we also clear h2c->dfl here so that we don't attempt to clear
it twice when going back to the demux.
If this is backported to 1.9, it also requires that the following patches
are backported as well :
MINOR: mux-h2: make h2c_decode_headers() return a status, not a count
MINOR: mux-h2: add a new dummy stream : h2_error_stream
MEDIUM: mux-h2: make h2c_decode_headers() support recoverable errors
When a decoding error is recoverable, we should emit a stream error and
not a connection error. This patch does this by carefully checking the
connection state before deciding to send a connection error. If only the
stream is in error, an RST_STREAM is sent.
This function used to return a byte count for the output produced, or
zero on failure. Not only this value is not used differently than a
boolean, but it prevents us from returning stream errors when a frame
cannot be extracted because it's too large, or from parsing a frame
and producing nothing on output.
This patch modifies its API to return <0 on errors, 0 on inability to
proceed, or >0 on success, irrelevant to the amount of output data.
The mux h1 mainly depends on the stream to handle errors and timeouts. But,
there is one unhandled case. If a timeout occurred when some outgoing data are
blocked in the output buffer, the stream is detached and the mux waits infinitly
the data are gone before closing the connection. To fix the bug, a task has been
added on the mux to handle connection timeouts. For now, a expiration date is
set only when some outgoing data are blocked. And if a stream is still attached
when the mux's task timed out, an error flag is set on the mux but the
connection is not closed immediatly. We assume the stream will hit the same
timeout just after.
This patch must be backported to 1.9.
In the function htx_end_request, the flag SI_FL_NOHALF must be set on the server
side once the request is in the state HTTP_MSG_DONE. But the response state was
checked before and the flag was only set when the response was also in the state
HTTP_MSG_DONE. Of course, it is not desirable.
This patch must be backported to 1.9.
Since a long time, the expiration date of a stream is only updated in
process_stream(). It is calculated, among others, using the channels expiration
dates for reads and writes (.rex and .wex values). But these values are updated
by the stream-interface. So when this happens at the connection layer, the
update is only done if the stream's task is woken up. Otherwise, the stream
expiration date is not immediatly updated. This leads to unexpected
behaviours. Time to time, users reported that the wrong timeout was hitted or
the wrong termination state was reported. This is partly because of this
bug.
Recently, we observed some blocked sessions for a while when big objects are
served from the cache applet. It seems only concern the clients not reading the
response. Because delivered objects are big, not all data can be sent. And
because delivered objects are big, data are fast forwarded (from the input to
the output with no stream wakeup). So in such situation, the stream expiration
date is never updated and no timeout is hitted. The session remains blocked
while the client remains connected.
This bug exists at least since HAProxy 1.5. But recent changes on the connection
layer make it more visible. It must be backported from 1.9 to 1.6. And with more
pain it should be backported to 1.5.
When transfering from H1 to H1, chunking is always indicated by the
presence of the Transfer-encoding header field. But when a message
comes from H2 there is no such header and only HTX_SL_F_XFER_LEN
ought to be relied on. This one will also result in H1_MF_XFER_LEN
to be set, just like transfer-encoding, so let's always rely on
this latter flag to detect the need for chunking (when CLEN is not
here) and automatically add the transfer-encoding header if it was
not present, as reported by H1_MF_CHNK.
This must be backported to 1.9.
The H1 mux needs to store some information regarding the states that
were met (EOD, trailers, etc) for each direction but currently uses
only one set of flags. This results in failures when both the request
and the response use chunked-encoding because some elements are believed
to have been met already and a trailing 0 CRLF or just a CRLF may be
missing at the end.
The solution here consists in splitting these flags per direction, one
set for input processing and another set for output processing. Only
two flags were affected so this is not a big deal.
This needs to be backported to 1.9.
In h2c_decode_headers() we update the buffer's length according to the
amount of data produced (outlen). But in case of HTX this outlen value
is not a quantity, just an indicator of success, resulting in the buffer
being added one extra byte and temporarily showing .data > .size, which
is wrong. Fortunately this is overridden when leaving the function by
htx_to_buf() so the impact only exists in step-by-step debugging, but
it definitely needs to be fixed.
This must be backported to 1.9.
When dealing with a server's H2 response, we used to set the
end-of-stream flag on the conn_stream and the stream before parsing
the response, which is incorrect since we can fail to process this
response by lack of room, buffer or anything. The extend of this problem
is still limited to a few rare cases, but with trailers it will cause a
systematic failure.
This fix must be backported to 1.9.
This function handles response HEADERS frames, it is not responsible
for creating new streams thus it must not check if we've reached the
stream count limit, otherwise it could lead to some undesired pauses
which bring no benefit.
This must be backported to 1.9.
If we exit this function because some data are pending in the rxbuf, we
currently don't indicate any blocking flag, which will prevent the operation
from being attempted again. Let's set H2_CF_DEM_SFULL in this case to indicate
there's not enough room in the stream buffer so that the operation may be
attempted again once we make room. It seems that this issue cannot be
triggered right now but it definitely will with trailers.
This fix should be backported to 1.9 for completeness.
h2c_restart_reading() is used at various place to resume processing of
demux data, but this one refrains from doing so if the mux is already
subscribed for receiving. It just happens that even if some incoming
frame processing is interrupted, the mux is always subscribed for
receiving, so this condition alone is not enough, it must be combined
with the fact that the demux buffer is empty, otherwise some resume
events are lost. This typically happens when we refrain from processing
some incoming data due to missing room in the stream's rxbuf, and want
to resume in h2c_rcv_buf(). It will become even more visible with trailers
since these ones want to have an empty rxbuf before proceeding.
This must be backported to 1.9.
In h2c_decode_headers() we mistakenly check for H2_F_DATA_END_STREAM
while we should check for H2_F_HEADERS_END_STREAM. Both have the same
value (1) but better stick to the correct flag.
When an HTX structure is defragmented, it is possible to retrieve the new block
corresponding to an old one. This is useful to do a defrag during a loop on
blocks, to be sure to continue looping on the good block. But, instead of
returning the address of the new block in the HTX structure, the one in the
temporary structure used to do the defrag was returned, leading to unexpected
behaviours.
This patch must be backported to 1.9.
This bug exists in the HTX code and in the legacy one. When the body length is
unknown, the applet hangs. For the legacy code, it hangs because the end of the
cached object is not correctly handled and the applet is never recalled. For the
HTX code, only the begining of the response (the 1st buffer) is sent then the
applet hangs. To work in HTX, The fast forwarding must be correctly handled.
This patch must be backported to 1.9.
[cf: the patch adding the function channel_add_input must be backported with
this one. It does not exist in 1.8 because only responses with a C-L are cached.]
This way we are sure the channel state is always correctly upadated, especially
the amount of data directly forwarded. For the stats applet, it is not a bug
because the fast forwarding is never used (the response is chunked and the HTX
extra field is always set to 0).
This patch must be backported to 1.9.
This function must be called when new incoming data are pushed in the channel's
buffer. It updates the channel state and take care of the fast forwarding by
consuming right amount of data and decrementing "->to_forward" accordingly when
necessary. In fact, this patch just moves a part of ci_putblk in a dedicated
function.
This patch must be backported to 1.9.
With the new ability to log to a terminal, it's convenient to be able
to use "log stdout" in a config file, except that it now results in
setting the terminal to non-blocking mode, breaking every utility
relying on stdin afterwards. Since the only reason for logging to a
terminal is to debug, do not set the FD to non-blocking mode when it's
a terminal.
This fix must be backported to 1.9.
Application-Layer Protocol Negotiation (ALPN, RFC7301) is a TLS
extension which allows a client to present a preference for which
protocols it wishes to connect to, when a single port supports multiple
multiple application protocols.
It allows a transparent proxy to take a decision based on the beginning
of an SSL/TLS stream without deciphering it.
The new fetch "req.ssl_alpn" extracts the ALPN protocol names that may
be present in the ClientHello message.
Instead of keeping track of the number of connections we're responsible for,
keep track of the number of connections we're responsible for that we are
currently considering idling (ie that we are not using, they may be in use
by other sessions), that way we can actually reuse connections when we have
more connections than the max configured.
When connecting to a server, and reusing a connection, always attempt to give
the owner of the previous session one of its own connections, so that one
session won't be responsible for too many connections.
This should be backported to 1.9.
When creating a new outgoing connection, if we're using ALPN and waiting
for the handshake completion to choose the mux, and for some reason the
handshake failed, add the SI_FL_ERR flag to the stream_interface, so that
process_streams() knows the connection failed, and can attempt to retry,
instead of just hanging.
This should be backported to 1.9.
When a session adds a connection to its connection list, we used to remove
connections for an another server if there were not enough room for our
server. This can't work, because those lists are now the list of connections
we're responsible for, not just the idle connections.
To fix this, allow for an unlimited number of servers, instead of using
an array, we're now using a linked list.
To access the first element of the list, correctly use LIST_ELEM(), or we
end up getting the head of the list, instead of getting the first connection.
This should be backported to 1.9.
In connect_server(), if we looked for an usable connection and failed to
find one, srv_conn won't be NULL at the end of list_for_each_entry(), but
will point to the head of a list, which is not a pointer to a struct
connection, so explicitely set it to NULL.
This should be backported to 1.9.
If, for some reason we failed to allocate a conn_stream when reusing an
existing connection, set srv_conn to NULL, so that we fail later, instead
of pretending all is right. This ends up giving a stream_interface with
no endpoint, and so the stream will never end.
This should be backported to 1.9.
In h2_detach(), don't add the connection to the idle list if nb_streams
is at the max. This can happen if we already closed that stream before, so
its slot became available and was used by another stream.
This should be backported to 1.9.
When we use htx and http-request auth rules, we need to send WWW-Authenticate
with a 401 and Proxy-Authenticate with a 407. We only sent Proxy-Authenticate
regardless of status, with htx enabled.
To be backported to 1.9.
In connect_server(), don't attempt to reuse the old connection if it's
targetting a different server than the one we're supposed to access, or
we will never be able to connect to a server if the first one we tried failed.
This should be backported to 1.9.
