Willy Tarreau [Fri, 1 Feb 2019 10:54:23 +0000 (11:54 +0100)]
BUG/MEDIUM: backend: always release the previous connection into its own target srv_list
There was a bug reported in issue #19 regarding the fact that haproxy
could mis-route requests to the wrong server. It turns out that when
switching to another server, the old connection was put back into the
srv_list corresponding to the stream's target instead of this connection's
target. Thus if this connection was later picked, it was pointing to the
wrong server.
The patch fixes this and also clarifies the assignment to srv_conn->target
so that it's clear we don't change it when picking it from the srv_list.
Olivier Houchard [Tue, 29 Jan 2019 14:20:16 +0000 (15:20 +0100)]
MINOR: debug: Add an option that causes random allocation failures.
When compiling with DEBUG_FAIL_ALLOC, add a new option, tune.fail-alloc,
that gives the percentage of chances an allocation fails.
This is useful to check that allocation failures are always handled
gracefully.
Willy Tarreau [Thu, 31 Jan 2019 09:42:05 +0000 (10:42 +0100)]
BUG/MEDIUM: mux-h2: properly consider the peer's advertised max-concurrent-streams
Till now we used to only rely on tune.h2.max-concurrent-streams but if
a peer advertises a lower limit this can cause streams to be reset or
even the conection to be killed. Let's respect the peer's value for
outgoing streams.
This patch should be backported to 1.9, though it depends on the following
ones :
BUG/MINOR: server: fix logic flaw in idle connection list management
MINOR: mux-h2: max-concurrent-streams should be unsigned
MINOR: mux-h2: make sure to only check concurrency limit on the frontend
MINOR: mux-h2: learn and store the peer's advertised MAX_CONCURRENT_STREAMS setting
Willy Tarreau [Thu, 31 Jan 2019 09:34:07 +0000 (10:34 +0100)]
MINOR: mux-h2: learn and store the peer's advertised MAX_CONCURRENT_STREAMS setting
We used not to take it into account because we only used the configured
parameter everywhere. This patch makes sure we can actually learn the
value advertised by the peer. We still enforce our own limit on top of
it however, to make sure we can actually limit resources or stream
concurrency in case of suboptimal server settings.
Willy Tarreau [Thu, 31 Jan 2019 09:31:51 +0000 (10:31 +0100)]
MINOR: mux-h2: make sure to only check concurrency limit on the frontend
h2_has_too_many_cs() was renamed to h2_frt_has_too_many_cs() to make it
clear it's only used to throttle the frontend connection, and the call
places were adjusted to only call this code on a front connection. In
practice it was already the case since the H2_CF_DEM_TOOMANY flag is
only set there. But now the ambiguity is removed.
Willy Tarreau [Sat, 26 Jan 2019 11:19:01 +0000 (12:19 +0100)]
BUG/MINOR: server: fix logic flaw in idle connection list management
With variable connection limits, it's not possible to accurately determine
whether the mux is still in use by comparing usage and max to be equal due
to the fact that one determines the capacity and the other one takes care
of the context. This can cause some connections to be dropped before they
reach their stream ID limit.
It seems it could also cause some connections to be terminated with
streams still alive if the limit was reduced to match the newly computed
avail_streams() value, though this cannot yet happen with existing muxes.
Instead let's switch to usage reports and simply check whether connections
are both unused and available before adding them to the idle list.
Willy Tarreau [Thu, 31 Jan 2019 18:12:48 +0000 (19:12 +0100)]
BUG/MEDIUM: mux-h2: do not close the connection on aborted streams
We used to rely on a hint that a shutw() or shutr() without data is an
indication that the upper layer had performed a tcp-request content reject
and really wanted to kill the connection, but sadly there is another
situation where this happens, which is failed keep-alive request to a
server. In this case the upper layer stream silently closes to let the
client retry. In our case this had the side effect of killing all the
connection.
Instead of relying on such hints, let's address the problem differently
and rely on information passed by the upper layers about the intent to
kill the connection. During shutr/shutw, this is detected because the
flag CS_FL_KILL_CONN is set on the connstream. Then only in this case
we send a GOAWAY(ENHANCE_YOUR_CALM), otherwise we only send the reset.
This makes sure that failed backend requests only fail frontend requests
and not the whole connections anymore.
This fix relies on the two previous patches adding SI_FL_KILL_CONN and
CS_FL_KILL_CONN as well as the fix for the connection close, and it must
be backported to 1.9 and 1.8, though the code in 1.8 could slightly differ
(cs is always valid) :
BUG/MEDIUM: mux-h2: wait for the mux buffer to be empty before closing the connection
MINOR: stream-int: add a new flag to mention that we want the connection to be killed
MINOR: connstream: have a new flag CS_FL_KILL_CONN to kill a connection
Willy Tarreau [Thu, 31 Jan 2019 18:02:43 +0000 (19:02 +0100)]
MINOR: stream-int: add a new flag to mention that we want the connection to be killed
The new flag SI_FL_KILL_CONN is now set by the rare actions which
deliberately want the whole connection (and not just the stream) to be
killed. This is only used for "tcp-request content reject",
"tcp-response content reject", "tcp-response content close" and
"http-request reject". The purpose is to desambiguate the close from
a regular shutdown. This will be used by the next patches.
Willy Tarreau [Thu, 31 Jan 2019 17:48:20 +0000 (18:48 +0100)]
BUG/MEDIUM: mux-h2: wait for the mux buffer to be empty before closing the connection
When finishing to respond on a stream, a shutw() is called (resulting
in either an end of stream or RST), then h2_detach() is called, and may
decide to kill the connection is a number of conditions are satisfied.
Actually one of these conditions is that a GOAWAY frame was already sent
or attempted to be sent. This one is wrong, because it can happen in at
least these two situations :
- a shutw() sends a GOAWAY to obey tcp-request content reject
- a graceful shutdown is pending
In both cases, the connection will be aborted with the mux buffer holding
some data. In case of a strong abort the client will not see the GOAWAY or
RST and might want to try again, which is counter-productive. In case of
the graceful shutdown, it could result in truncated data. It looks like a
valid candidate for the issue reported here :
Willy Tarreau [Thu, 31 Jan 2019 17:58:06 +0000 (18:58 +0100)]
BUG/MINOR: stream: don't close the front connection when facing a backend error
In 1.5-dev13, a bug was introduced by commit e3224e870 ("BUG/MINOR:
session: ensure that we don't retry connection if some data were sent").
If a connection error is reported after some data were sent (and lost),
we used to accidently mark the front connection as being in error instead
of only the back one because the two direction flags were applied to the
same channel. This case is extremely rare with raw connections but can
happen a bit more often with multiplexed streams. This will result in
the error not being correctly reported to the client.
This patch can be backported to all supported versions.
Olivier Houchard [Thu, 31 Jan 2019 18:31:19 +0000 (19:31 +0100)]
BUG/MEDIUM: connections: Don't forget to remove CO_FL_SESS_IDLE.
If we're adding a connection to the server orphan idle list, don't forget
to remove the CO_FL_SESS_IDLE flag, or we will assume later it's still
attached to a session.
BUG/MEDIUM: mux-h1: Don't add "transfer-encoding" if message-body is forbidden
When a HTTP/1.1 or above response is emitted to a client, if the flag
H1_MF_XFER_LEN is set whereas H1_MF_CLEN and H1_MF_CHNK are not, the header
"transfer-encoding" is added. It is a way to make HTX chunking consistent with
H2. But we must exclude all cases where the message-body is explicitly forbidden
by the RFC:
* for all 1XX, 204 and 304 responses
* for any responses to HEAD requests
* for 200 responses to CONNECT requests
For these 3 cases, the flag H1_MF_XFER_LEN is set but H1_MF_CLEN and H1_MF_CHNK
not. And the header "transfer-encoding" must not be added.
See issue #27 on github for details about the bug.
This bug was introduced by 355b203 commit which prevented the peer
addresses to be parsed for the local peer of a "peers" section.
When adding "parse_addr" boolean parameter to parse_server(), this commit
missed the case where the syntax with "peer" keyword should still be
supported in addition to the new syntax with "server"+"bind" keyword.
Willy Tarreau [Thu, 31 Jan 2019 06:23:00 +0000 (07:23 +0100)]
MINOR: mux-h2: consistently rely on the htx variable to detect the mode
In h2_frt_transfer_data(), we support both HTX and legacy modes. The
HTX mode is detected from the proxy option and sets a valid pointer
into the htx variable. Better rely on this variable in all the function
rather than testing the option again. This way the code is clearer and
even the compiler knows this pointer is valid when it's dereferenced.
This should be backported to 1.9 if the b_is_null() patch is backported.
Willy Tarreau [Thu, 31 Jan 2019 06:31:18 +0000 (07:31 +0100)]
MINOR: htx: never check for null htx pointer in htx_is_{,not_}empty()
The previous patch clarifies the fact that the htx pointer is never null
along all the code. This test for a null will never match, didn't catch
the pointer 1 before the fix for b_is_null(), but it confuses the compiler
letting it think that any dereferences made to this pointer after this
test could actually mean we're dereferencing a null. Let's now drop this
test. This saves us from having to add impossible tests everywhere to
avoid the warning.
This should be backported to 1.9 if the b_is_null() patch is backported.
Willy Tarreau [Thu, 31 Jan 2019 06:21:42 +0000 (07:21 +0100)]
DOC: htx: make it clear that htxbuf() and htx_from_buf() always return valid pointers
Update the comments above htxbuf() and htx_from_buf() to make it clear
that they always return valid htx pointers so that callers know they do
not have to test them. This is only true after the fix on b_is_null()
which was the only known corner case.
This should be backported to 1.9 if the b_is_null() patch is backported.
Olivier Houchard [Tue, 29 Jan 2019 18:10:02 +0000 (19:10 +0100)]
BUG/MEDIUM: buffer: Make sure b_is_null handles buffers waiting for allocation.
In b_is_null(), make sure we return 1 if the buffer is waiting for its
allocation, as users assume there's memory allocated if b_is_null() returns
0.
The indirect impact of not having this was that htxbuf() would not match
b_is_null() for a buffer waiting for an allocation, and would thus return
the value 1 for the htx pointer, causing various crashes under low memory
condition.
Note that this patch makes gcc versions 6 and above report two null-deref
warnings in proto_htx.c since htx_is_empty() continues to check for a null
pointer without knowing that this is protected by the test on b_is_null().
This is addressed by the following patches.
Tim Duesterhus [Wed, 30 Jan 2019 22:46:04 +0000 (23:46 +0100)]
DOC: compression: Update the reasons for disabled compression
- Update the list of status codes to include 201 - 203.
- Remove the fact about the temporary workaround for chunked responses
(this is verified using reg-test compression/h00000.vtc).
- Add malformed ETags
This commit should be backported together with b229f018eedef4d18571ce6da23d8e153249a836
the changes should be correct until 1.7 at the very least, possibly older.
Willy Tarreau [Wed, 30 Jan 2019 10:44:07 +0000 (11:44 +0100)]
BUG/MINOR: mux-h2: make sure request trailers on aborted streams don't break the connection
We used to respond a connection error in case we received a trailers
frame on a closed stream, but it's a problem to do this if the error
was caused by a reset because the sender has not yet received it and
is just a victim of the timing. Thus we must not close the connection
in this case.
This patch may be backported to 1.9 but then it requires the following
previous ones :
MINOR: h2: add a generic frame checker
MEDIUM: mux-h2: check the frame validity before considering the stream state
CLEANUP: mux-h2: remove stream ID and frame length checks from the frame parsers
Willy Tarreau [Wed, 30 Jan 2019 14:39:55 +0000 (15:39 +0100)]
CLEANUP: mux-h2: remove stream ID and frame length checks from the frame parsers
It's not convenient to have such structural checks mixed with the ones
related to the stream state. Let's remove all these basic tests that are
already covered once for all when reading the frame header.
Willy Tarreau [Wed, 30 Jan 2019 14:11:03 +0000 (15:11 +0100)]
MEDIUM: mux-h2: check the frame validity before considering the stream state
There are some uneasy situation where it's difficult to validate a frame's
format without being in an appropriate state. This patch makes sure that
each frame passes through h2_frame_check() before being checked in the
context of the stream's state. This makes sure we can always return a GOAWAY
for protocol violations even if we can't process the frame.
Willy Tarreau [Wed, 30 Jan 2019 14:09:21 +0000 (15:09 +0100)]
MINOR: h2: add a generic frame checker
The new function h2_frame_check() checks the protocol limits for the
received frame (length, ID, direction) and returns a verdict made of
a connection error code. The purpose is to be able to validate any
frame regardless of the state and the ability to call the frame handler,
and to emit a GOAWAY early in this case.
Willy Tarreau [Wed, 30 Jan 2019 15:55:48 +0000 (16:55 +0100)]
BUG/MINOR: mux-h2: make sure response HEADERS are not received in other states than OPEN and HLOC
RFC7540#5.1 states that these are the only states allowing any frame
type. For response HEADERS frames, we cannot accept that they are
delivered on idle streams of course, so we're left with these two
states only. It is important to test this so that we can remove the
generic CLOSE_STREAM test for such frames in the main loop.
This must be backported to 1.9 (1.8 doesn't have response HEADERS).
Willy Tarreau [Wed, 30 Jan 2019 15:58:30 +0000 (16:58 +0100)]
BUG/MEDIUM: mux-h2: do not abort HEADERS frame before decoding them
If a response HEADERS frame arrives on a closed connection (due to a
client abort sending an RST_STREAM), it's currently immediately rejected
with an RST_STREAM, like any other frame. This is incorrect, as HEADERS
frames must first be decoded to keep the HPACK decoder synchronized,
possibly breaking subsequent responses.
This patch excludes HEADERS/CONTINUATION/PUSH_PROMISE frames from the
central closed state test and leaves to the respective frame parsers
the responsibility to decode the frame then send RST_STREAM.
This fix must be backported to 1.9. 1.8 is not directly impacted since
it doesn't have response HEADERS nor trailers thus cannot recover from
such situations anyway.
Willy Tarreau [Wed, 30 Jan 2019 18:20:09 +0000 (19:20 +0100)]
BUG/MEDIUM: mux-h2: make sure never to send GOAWAY on too old streams
The H2 spec requires to send GOAWAY when the client sends a frame after
it has already closed using END_STREAM. Here the corresponding case was
the fallback of a series of tests on the stream state, but it unfortunately
also catches old closed streams which we don't know anymore. Thus any late
packet after we've sent an RST_STREAM will trigger this GOAWAY and break
other streams on the connection.
This can happen when launching two tabs in a browser targetting the same
slow page through an H2-to-H2 proxy, and pressing Escape to stop one of
them. The other one gets an error when the page finally responds (and it
generally retries), and the logs in the middle indicate SD-- flags since
the late response was cancelled.
This patch takes care to only send GOAWAY on streams we still know.
Willy Tarreau [Wed, 30 Jan 2019 18:28:32 +0000 (19:28 +0100)]
BUG/MEDIUM: mux-h2: fix two half-closed to closed transitions
When receiving a HEADERS or DATA frame with END_STREAM set, we would
inconditionally switch to half-closed(remote). This is wrong because we
could already have been in half-closed(local) and need to switch to closed.
This happens in the following situations :
- receipt of the end of a client upload after we've already responded
(e.g. redirects to POST requests)
- receipt of a response on the backend side after we've already finished
sending the request (most common case).
This may possibly have caused some streams to stay longer than needed
at the end of a transfer, though this is not apparent in tests.
Willy Tarreau [Wed, 30 Jan 2019 15:11:20 +0000 (16:11 +0100)]
BUG/MEDIUM: mux-h2: wake up flow-controlled streams on initial window update
When a settings frame updates the initial window, all affected streams's
window is updated as well. However the streams are not put back into the
send list if they were already blocked on flow control. The effect is that
such a stream will only be woken up by a WINDOW_UPDATE message but not by
a SETTINGS changing the initial window size. This can be verified with
h2spec's test http2/6.9.2/1 which occasionally fails without this patch.
It is unclear whether this situation is really met in field, but the
fix is trivial, it consists in adding each unblocked streams to the
wait list as is done for the window updates.
This fix must be backported to 1.9. For 1.8 the patch needs quite
a few adaptations. It's better to copy-paste the code block from
h2c_handle_window_update() adding the stream to the send_list when
its mws is > 0.
Willy Tarreau [Wed, 30 Jan 2019 14:42:44 +0000 (15:42 +0100)]
CLEANUP: mux-h2: remove misleading leftover test on h2s' nullity
The WINDOW_UPDATE and DATA frame handlers used to still have a check on
h2s to return either h2s_error() or h2c_error(). This is a leftover from
the early code. The h2s cannot be null there anymore as it has already
been dereferenced before reaching these locations.
Tim Duesterhus [Tue, 29 Jan 2019 15:38:56 +0000 (16:38 +0100)]
BUG/MEDIUM: compression: Rewrite strong ETags
RFC 7232 section 2.3.3 states:
> Note: Content codings are a property of the representation data,
> so a strong entity-tag for a content-encoded representation has to
> be distinct from the entity tag of an unencoded representation to
> prevent potential conflicts during cache updates and range
> requests. In contrast, transfer codings (Section 4 of [RFC7230])
> apply only during message transfer and do not result in distinct
> entity-tags.
Thus a strong ETag must be changed when compressing. Usually this is done
by converting it into a weak ETag, which represents a semantically, but not
byte-by-byte identical response. A conversion to a weak ETag still allows
If-None-Match to work.
This should be backported to 1.9 and might be backported to every supported
branch with compression.
Olivier Houchard [Tue, 29 Jan 2019 18:11:16 +0000 (19:11 +0100)]
BUG/MEDIUM: servers: Close the connection if we failed to install the mux.
If we failed to install the mux, just close the connection, or
conn_fd_handler() will be called for the FD, and crash as soon as it attempts
to access the mux' wake method.
Olivier Houchard [Tue, 29 Jan 2019 15:05:02 +0000 (16:05 +0100)]
BUG/MEDIUM: servers: Don't add an incomplete conn to the server idle list.
If we failed to add the connection to the session, only try to add it back
to the server idle list if it has a mux, otherwise the connection is
incomplete and unusable.
Olivier Houchard [Tue, 29 Jan 2019 14:50:38 +0000 (15:50 +0100)]
BUG/MEDIUM: servers: Only destroy a conn_stream we just allocated.
In connect_server(), if we failed to add the connection to the session,
only destroy the conn_stream if we just allocated it, otherwise it may
have been allocated outside connect_server(), along with a connection which
has its destination address set.
Also use si_release_endpoint() instead of cs_destroy(), to make sure the
stream_interface doesn't reference it anymore.
Willy Tarreau [Tue, 29 Jan 2019 17:33:26 +0000 (18:33 +0100)]
BUG/MEDIUM: mux-h2: only close connection on request frames on closed streams
A subtle bug was introduced with H2 on the backend. RFC7540 states that
an attempt to create a stream on an ID not higher than the max known is
a connection error. This was translated into rejecting HEADERS frames
for closed streams. But with H2 on the backend, if the client aborts
and causes an RST_STREAM to be emitted, the stream is effectively closed,
and if/once the server responds, it starts by emitting a HEADERS frame
with this ID thus it is interpreted as a connection error.
This test must of course consider the side the mux is installed on and
not take this for a connection error on responses.
The effect is that an aborted stream on an outgoing H2 connection, for
example due to a client stopping a transfer with option abortonclose
set, would lead to an abort of all other streams. In the logs, this
appears as one or several CD-- line(s) followed by one or several SD--
lines which are victims.
Thanks to Luke Seelenbinder for reporting this problem and providing
enough elements to help understanding how to reproduce it.
Willy Tarreau [Tue, 29 Jan 2019 16:45:23 +0000 (17:45 +0100)]
BUILD/MINOR: peers: shut up a build warning introduced during last cleanup
A new warning appears when building at -O0 since commit 3f0fb9df6 ("MINOR:
peers: move "hello" message treatment code to reduce the size of the I/O
handler."), it is related to the fact that proto_len is initialized from
strlen() which is not a constant. Let's replace it with sizeof-1 instead
and also mark the variable as static since it's useless outside of the file.
Willy Tarreau [Tue, 29 Jan 2019 10:08:06 +0000 (11:08 +0100)]
CLEANUP: peers: factor the error handling code in peer_treet_updatemsg()
The error handling code was extremely repetitive and error-prone due
to the numerous copy-pastes, some involving unlocks or free. Let's
factor this out. The code could be further simplified, but 12 locations
were already cleaned without taking risks.
MINOR: peers: move peer initializations code to reduce the size of the I/O handler.
Implements two new functions to init peer flags and other stuff after
having accepted or connected them with the peer I/O handler so that
to reduce its size.
MINOR: peers: move "hello" message treatment code to reduce the size of the I/O handler.
This patch implements three functions to read and parse the three
line of a "hello" peer protocol message so that to call them from the
peer I/O handler and reduce its size.
When implementing peer_recv_msg() we added the statements reached with
a "goto imcomplete" at the end of this function. This statements
are executed only when co_getblk() returns something <0. So they
are useless for now on, and may be safely removed. The following
section wich was responsible of sending any peer protocol messages
were reached only when co_getblk() returned 0 (no more message to
read). In this case we replace the "goto impcomplete" statement by
a "goto send_msgs" to reach this only when peer_recv_msg() returns 0.
MINOR: peers: move send code to reduce the size of the I/O handler.
This patch extracts the code responsible of sending peer protocol
messages from the peer I/O handler to create a new function and to
reduce the size of this handler.
MINOR: peers: move messages treatment code to reduce the size of the I/O handler.
Extract the code of the peer I/O handler responsible of treating
any peer protocol message to create peer_treat_awaited_msg() function.
Also rename peer_recv_updatemsg() to peer_treat_updatemsg() as this
function only parse a stick-table update message already received
by peer_recv_msg().
MINOR: peers: Move ack, switch and definition receive code to reduce the size of the I/O handler.
Implement three new functions to treat peer acks, switch and
definition messages extracting the code from the big swich-case
of the peer I/O handler to give more chances to this latter to be
readable.
MINOR: peers: Add new functions to send code and reduce the I/O handler.
This patch reduces the size of the peer I/O handler implementing
a new function named peer_send_updatemsg() which uses the already
implement peer_prepare_updatemsg(), then ci_putblk().
Reuse the code used to implement peer_send_(ack|swith)msg() function
especially the more generic function peer_send_msg().
Implements peer_send_*msg() functions for switch and ack messages which call the
already defined peer_prepare_*msg() before calling ci_putblk().
These two new functions are used at three places in the peer_io_handler().
Willy Tarreau [Mon, 28 Jan 2019 15:33:35 +0000 (16:33 +0100)]
BUG/MEDIUM: backend: always call si_detach_endpoint() on async connection failure
In case an asynchronous connection (ALPN) succeeds but the mux fails to
attach, we must release the stream interface's endpoint, otherwise we
leave the stream interface with an endpoint pointing to a freed connection
with si_ops == si_conn_ops, and sess_update_st_cer() calls si_shutw() on
it, causing it to crash.
Olivier Houchard [Mon, 28 Jan 2019 14:33:15 +0000 (15:33 +0100)]
BUG/MEDIUM: servers: Attempt to reuse an unfinished connection on retry.
In connect_server(), if the previous connection failed, but had an alpn, no
mux was created, and thus the stream_interface's endpoint would be the
connection. In this case, instead of forgetting about it, and overriding
the stream_interface's endpoint later, try to reuse the connection, or the
connection will still be in the session's connection list, and will reference
to a stream that was probably destroyed.
Willy Tarreau [Sun, 27 Jan 2019 16:41:27 +0000 (17:41 +0100)]
BUG/MINOR: task: fix possibly missed event in inter-thread wakeups
There's a very small but existing uncertainty window when waking another
thread up where it is possible for task_wakeup() not to wake the other
task up because it's still running while this once is in the process of
finishing and loses its TASK_RUNNING flag. In this case the wakeup will
be missed.
The problem is that we have a single flag to store 3 states, since the
transition from running to sleeping isn't atomic. Thus we need to have
another flag to cover this part. This patch introduces TASK_QUEUED to
mention that the task is already in the run queue, running or not. This
bit will be removed while TASK_RUNNING is kept once dequeued, and will
be used when removing TASK_RUNNING to check if the task has been requeued.
It might be possible to slightly improve this but the occurrence rate
is quite low and we don't really need to complexify the scheduler to
optimize for a rare case.
The impact with the current code is very low since we have few inter-
thread wakeups. Most of them are caused by checks killing sessions.
Willy Tarreau [Mon, 28 Jan 2019 05:40:19 +0000 (06:40 +0100)]
BUG/MINOR: mux-h2: do not report available outgoing streams after GOAWAY
The calculation of available outgoing H2 streams was improved by commit d64a3ebe6 ("BUG/MINOR: mux-h2: always check the stream ID limit in
h2_avail_streams()"), but it still is incorrect because RFC7540#6.8
specifically forbids the creation of new streams after a GOAWAY frame
was received. Thus we must not mark the connection as available anymore
in order to be able to handle a graceful shutdown.
Willy Tarreau [Sun, 27 Jan 2019 17:34:12 +0000 (18:34 +0100)]
BUG/MINOR: listener: always fill the source address for accepted socketpairs
The source address was not set but passed down the chain to the upper
layer's accept() calls. Let's initialize it like other UNIX sockets in
this case. At the moment it should not have any impact since socketpairs
are only usable for the master CLI.
Willy Tarreau [Sat, 26 Jan 2019 12:35:03 +0000 (13:35 +0100)]
MINOR: threads: make MAX_THREADS configurable at build time
There's some value in being able to limit MAX_THREADS, either to save
precious resources in embedded environments, or to protect certain
deployments against accidently incorrect settings.
With this patch, if MAX_THREADS is defined at build time, it will be
used. However, given that LONGBITS is not a macro but is defined
according to sizeof(long), we can't check the value range at build
time and instead we need to perform the check at early boot time.
However, the compiler is able to optimize away the constant comparisons
and doesn't even emit the check code when values are correct.
The output message regarding threading support was improved to report
the number of threads.
Willy Tarreau [Thu, 24 Jan 2019 10:49:37 +0000 (11:49 +0100)]
BUG/MINOR: mux-h2: always compare content-length to the sum of DATA frames
This is mandated by RFC7541#8.1.2.6. Till now we didn't have a copy of
the content-length header field. But now that it's already parsed, it's
easy to add the check.
The reg-test was updated to match the new behaviour as the previous one
expected unadvertised data to be silently discarded.
This should be backported to 1.9 along with previous patch (MEDIUM: h2:
always parse and deduplicate the content-length header) after it has got
a bit more exposure.
Willy Tarreau [Thu, 24 Jan 2019 10:33:02 +0000 (11:33 +0100)]
MEDIUM: h2: always parse and deduplicate the content-length header
The header used to be parsed only in HTX but not in legacy. And even in
HTX mode, the value was dropped. Let's always parse it and report the
parsed value back so that we'll be able to store it in the streams.
Willy Tarreau [Wed, 23 Jan 2019 15:29:53 +0000 (16:29 +0100)]
MINOR: stream: don't wait before retrying after a failed connection reuse
When a connection reuse fails, we must not wait before retrying, as most
likely the issue is related to the reused connection and not to the server
itself.
This should be backported to 1.9, though it depends on previous patches
dealing with SI_ST_CON for connection reuse.
Willy Tarreau [Wed, 23 Jan 2019 14:15:09 +0000 (15:15 +0100)]
MEDIUM: stream-int: always mark pending outgoing SI_ST_CON
Before the first send() attempt, we should be in SI_ST_CON, not
SI_ST_EST, since we have not yet attempted to send and we are
allowed to retry. This is particularly important with complex
outgoing muxes which can fail during the first send attempt (e.g.
failed stream ID allocation).
It only requires that sess_update_st_con_tcp() knows about this
possibility, as we must not forcefully close a reused connection
when facing an error in this case, this will be handled later.
This may be backported to 1.9 with care after some observation period.
Willy Tarreau [Wed, 23 Jan 2019 09:25:10 +0000 (10:25 +0100)]
MINOR: mux-h2: always consider a server's max-reuse parameter
This parameter allows to limit the number of successive requests sent
on a connection. Let's compare it to the number of streams already sent
on the connection to decide if the connection may still appear in the
idle list or not. This may be used to help certain servers work around
resource leaks, and also helps dealing with the issue of the GOAWAY in
flight which requires to set a usage limit on the client to be reliable.
Willy Tarreau [Wed, 23 Jan 2019 09:21:49 +0000 (10:21 +0100)]
MINOR: server: add a max-reuse parameter
Some servers may wish to limit the total number of requests they execute
over a connection because some of their components might leak resources.
In HTTP/1 it was easy, they just had to emit a "connection: close" header
field with the last response. In HTTP/2, it's less easy because the info
is not always shared with the component dealing with the H2 protocol and
it could be harder to advertise a GOAWAY with a stream limit.
This patch provides a solution to this by adding a new "max-reuse" parameter
to the server keyword. This parameter indicates how many times an idle
connection may be reused for new requests. The information is made available
and the underlying muxes will be able to use it at will.
Willy Tarreau [Thu, 24 Jan 2019 17:22:19 +0000 (18:22 +0100)]
BUG/MEDIUM: backend: never try to attach to a mux having no more stream available
The code dealing with idle connections used to check the number of streams
available on the connection only to unlink the connection from the idle
list. But this still resulted in too many streams reusing the same connection
when they were already attached to it.
We must detect that there is no more room and refrain from using this
connection at all, and instead fall back to the no-reuse case. Ideally
we should try to search among other idle connections, but for a backport
let's stay safe.
Willy Tarreau [Thu, 24 Jan 2019 16:08:28 +0000 (17:08 +0100)]
BUG/MINOR: mux-h2: refuse to allocate a stream with too high an ID
One of the reasons for the excessive number of aborted requests when a
server sets a limit on the highest stream ID is that we don't check
this limit while allocating a new stream.
This patch does this at two locations :
- when a backend stream is allocated, we verify that there are still
IDs left ;
- when the ID is assigned, we verify that it's not higher than the
advertised limit.
Willy Tarreau [Wed, 23 Jan 2019 09:22:21 +0000 (10:22 +0100)]
BUG/MINOR: mux-h2: always check the stream ID limit in h2_avail_streams()
This function is used to decide whether to put an idle connection back
into the idle pool. While it considers the limit in number of concurrent
requests, it does not consider the limit in number of streams, so if a
server announces a low limit in a GOAWAY frame, it will be ignored.
However there is a caveat : since we assign the stream IDs when sending
them, we have a number of allocated streams which max_id doesn't take
care of. This can be addressed by adding a new nb_reserved count on each
connection to keep track of the ID-less streams.
This patch makes sure we take care of the remaining number of streams
if such a limit was announced, or of the number of streams before the
highest ID. Now it is possible to accurately know how many streams
can be allocated, and the number of failed outgoing streams has dropped
in half.
Willy Tarreau [Wed, 23 Jan 2019 14:18:19 +0000 (15:18 +0100)]
BUG/MINOR: stream: take care of synchronous errors when trying to send
We currently detect a number of situations where we have to immediately
deal with a state change, but we failed to consider the case of the
synchronous error reported on the stream-interface. We definitely do not
want to have to wait for a timeout to handle this one, especially at the
beginning of the connection when it can lead to an immediate retry.
Willy Tarreau [Thu, 24 Jan 2019 09:47:10 +0000 (10:47 +0100)]
BUG/MINOR: hpack: return a compression error on invalid table size updates
RFC7541#6.3 mandates that an error is reported when a dynamic table size
update announces a size larger than the one configured with settings. This
is tested by h2spec using test "hpack/6.3/1".
This must be backported to 1.9 and possibly 1.8 as well.
Willy Tarreau [Thu, 24 Jan 2019 09:02:24 +0000 (10:02 +0100)]
BUG/MINOR: mux-h2: make it possible to set the error code on an already closed stream
When sending RST_STREAM in response to a frame delivered on an already
closed stream, we used not to be able to update the error code and
deliver an RST_STREAM with a wrong code (e.g. H2_ERR_CANCEL). Let's
always allow to update the code so that RST_STREAM is always sent
with the appropriate error code (most often H2_ERR_STREAM_CLOSED).
This should be backported to 1.9 and possibly to 1.8.
Willy Tarreau [Thu, 24 Jan 2019 08:43:32 +0000 (09:43 +0100)]
BUG/MINOR: mux-h2: headers-type frames in HREM are always a connection error
There are incompatible MUST statements in the HTTP/2 specification. Some
require a stream error and others a connection error for the same situation.
As discussed in the thread below, let's always apply the connection error
when relevant (headers-like frame in half-closed(remote)) :
Willy Tarreau [Thu, 24 Jan 2019 08:36:53 +0000 (09:36 +0100)]
BUG/MINOR: mux-h2: CONTINUATION in closed state must always return GOAWAY
Since we now support CONTINUATION frames, we must take care of properly
aborting the connection when they are sent on a closed stream. By default
we'd get a stream error which is not sufficient since the compression
context is modified and unrecoverable.
Willy Tarreau [Thu, 24 Jan 2019 08:31:40 +0000 (09:31 +0100)]
MINOR: h2: declare new sets of frame types
This patch adds H2_FT_HDR_MASK to group all frame types carrying headers
information, and H2_FT_LATE_MASK to group frame types allowed to arrive
after a stream was closed.
Willy Tarreau [Thu, 24 Jan 2019 09:26:47 +0000 (10:26 +0100)]
BUG/MEDIUM: mux-h2: properly abort on trailers decoding errors
There was an incomplete test in h2c_frt_handle_headers() resulting
in negative return values from h2c_decode_headers() not being taken
as errors. The effect is that the stream is then aborted on timeout
only.
Willy Tarreau [Wed, 23 Jan 2019 16:33:06 +0000 (17:33 +0100)]
BUG/MEDIUM: backend: also remove from idle list muxes that have no more room
The current test consists in removing muxes which report that they're going
to assign their last available stream, but a mux may already be saturated
without having passed in this situation at all. This is what happens with
mux_h2 when receiving a GOAWAY frame informing the mux about the ID of the
last stream the other end is willing to process. The limit suddenly changes
from near infinite to 0. Currently what happens is that such a mux remains
in the idle list for a long time and refuses all new streams. Now at least
it will only fail a single stream in a retryable way. A future improvement
should consist in trying to pick another connection from the idle list.
Willy Tarreau [Wed, 23 Jan 2019 13:39:41 +0000 (14:39 +0100)]
BUG/MAJOR: mux-h2: don't destroy the stream on failed allocation in h2_snd_buf()
In case we cannot allocate a stream ID for an outgoing stream, the stream
will be aborted. The problem is that we also release it and it will be
destroyed again by the application detecting the error, leading to a NULL
dereference in h2_shutr() and h2_shutw(). Let's only mark the error on the
CS and let the rest of the code handle the close.
Willy Tarreau [Wed, 23 Jan 2019 19:43:53 +0000 (20:43 +0100)]
BUG/MINOR: mux-h1: avoid copying output over itself in zero-copy
It's almost funny but one side effect of the latest zero-copy changes made
to mux-h1 resulted in the temporary buffer being copied over itself at the
exact same location. This has no impact except slowing down operations and
irritating valgrind. The cause is an incorrect pointer check after the
alignment optimizations were made.
BUG/MINOR: mux-h1: Apply the reserve on the channel's buffer only
There is no reason to use the reserve to limit the amount of data of the input
buffer that we can parse at a time. The only important thing is to keep the
reserve free in the channel's buffer.
BUG/MEDIUM: mux-h2/htx: Respect the channel's reserve
When data are pushed in the channel's buffer, in h2_rcv_buf(), the mux-h2 must
respect the reserve if the flag CO_RFL_KEEP_RSV is set. In HTX, because the
stream-interface always sees the buffer as full, there is no other way to know
the reserve must be respected.
BUG/MINOR: proto-htx: Return an error if all headers cannot be received at once
When an HTX stream is waiting for a request or a response, it reports an error
(400 for the request or 502 for the response) if a parsing error is reported by
the mux (HTX_FL_PARSING_ERROR). The mux-h1 uses this error, among other things,
when the headers are too big to be analyzed at once. But the mux-h2 doesn't. So
the stream must also report an error if the multiplexer is unable to emit all
headers at once. The multiplexers must always emit all the headers at once
otherwise it is an error.
There are 2 ways to detect this error:
* The end-of-headers marker was not received yet _AND_ the HTX message is not
empty.
* The end-of-headers marker was not received yet _AND_ the multiplexer have
some data to emit but it is waiting for more space in the channel's buffer.
Note the mux-h2 is buggy for now when HTX is enabled. It does not respect the
reserve. So there is no way to hit this bug.
Dirkjan Bussink [Mon, 21 Jan 2019 17:35:03 +0000 (09:35 -0800)]
BUG/MEDIUM: ssl: Fix handling of TLS 1.3 KeyUpdate messages
In OpenSSL 1.1.1 TLS 1.3 KeyUpdate messages will trigger the callback
that is used to verify renegotiation is disabled. This means that these
KeyUpdate messages fail. In OpenSSL 1.1.1 a better mechanism is
available with the SSL_OP_NO_RENEGOTIATION flag that disables any TLS
1.2 and earlier negotiation.
So if this SSL_OP_NO_RENEGOTIATION flag is available, instead of having
a manual check, trust OpenSSL and disable the check. This means that TLS
1.3 KeyUpdate messages will work properly.
Reported-By: Adam Langley <agl@imperialviolet.org>
BUG/MINOR: check: Wake the check task if the check is finished in wake_srv_chk()
With tcp-check, the result of the check is set by the function tcpcheck_main()
from the I/O layer. So it is important to wake up the check task to handle the
result and finish the check. Otherwise, we will wait the task timeout to handle
the result of a tcp-check, delaying the next check by as much.
This patch also fixes a problem about email alerts reported by PiBa-NL (Pieter)
on the ML [1] on all versions since the 1.6. So this patch must be backported
from 1.9 to 1.6.