In the previous patch ("MEDIUM: mux-h2: Block client data on server side
waiting tunnel establishment"), we added a way to block client data for not
fully established tunnel on the server side. This one closes the stream with
an ERR_CANCEL erorr if there are some pending tunneled data while the tunnel
was aborted. This may happen on the client side if a non-empty DATA frame or
an empty DATA frame without the ES flag is received. This may also happen on
the server side if there is a DATA htx block. However in this last case, we
first wait the response is fully forwarded.
This patch contributes to fix the tunnel mode between the H1 and the H2
muxes.
On the server side, when a tunnel is not fully established, we must block
tunneled data, waiting for the server response. It is mandatory because the
server may refuse the tunnel. This happens when a DATA htx block is
processed in tunnel mode (H2_SF_BODY_TUNNEL flag set) but before the
response HEADERS frame is received (H2_SF_HEADERS_RCVD flag no set). In this
case, the H2_SF_BLK_MBUSY flag is set to mark the stream as busy. This flag
is removed when the tunnel is fully established or aborted.
This patch contributes to fix the tunnel mode between the H1 and the H2
muxes.
H2_SF_BODY_TUNNEL and H2_SF_TUNNEL_ABRT flags are added to properly handle
the tunnel mode in the H2 mux. The first one is used to detect tunnel
establishment or fully established tunnel. The second one is used to abort a
tunnel attempt. It is the first commit having as a goal to fix tunnel
establishment between H1 and H2 muxes.
There is a subtlety in h2_rcv_buf(). CS_FL_EOS flag is added on the
conn-stream when ES is received on a tunneled stream. It really reflects the
conn-stream state and is mandatory for next commits.
The H1C_F_WAIT_OPPOSITE flag is now splitted in 2 flags, H1C_F_WAIT_INPUT
and H1C_F_WAIT_OUTPUT, depending on the side is waiting. The change is a
prerequisite to fix the tunnel mode management in HTTP muxes.
H1C_F_WAIT_INPUT must be used to bloc the output side and to wait for an
event from the input side. H1C_F_WAIT_OUTPUT does the opposite. It bloc the
input side and wait for an event from the output side.
Responses with no C-L and T-E headers are no longer switched in TUNNEL mode
and remains in DATA mode instead. The H1 and FCGI muxes are updated
accordingly. This change reflects the real message state. It is not a true
tunnel. Data received are still part of the message.
It is not a bug. However, this message may be backported after some
observation period (at least as far as 2.2).
As stated in the RFC7540, section 8.1.1, the HTTP/2 removes support for the
101 informational status code. Thus a PROTOCOL_ERROR is now returned to the
server if a 101-switching-protocols response is received. Thus, the server
connection is aborted.
This patch may be backported as far as 2.0.
A 101-switching-protocols response must contain a Connection header with the
Upgrade option. And this response must only be received from a server if the
client explicitly requested a protocol upgrade. Thus, the request must also
contain a Connection header with the Upgrade option. If not, a
502-bad-gateway response is returned to the client. This way, a tunnel is
only established if both sides are agree.
It is closer to what the RFC says, but it remains a bit flexible because
there is no check on the Upgrade header itself. However, that's probably
enough to ensure a tunnel is not established when not requested.
This one is not tagged as a bug. But it may be backported, at least to
2.3. It relies on :
* MINOR: htx/http-ana: Save info about Upgrade option in the Connection header
Add an HTX start-line flag and its counterpart into the HTTP message to
track the presence of the Upgrade option into the Connection header. This
way, without parsing the Connection header again, it will be easy to know if
a client asks for a protocol upgrade and if the server agrees to do so. It
will also be easy to perform some conformance checks when a
101-switching-protocols is received.
It is the second part and the most important of the fix.
Since the mux-h1 refactoring, and more specifically since the commit
c4bfa59f1 ("MAJOR: mux-h1: Create the client stream as later as possible"),
the upgrade from a TCP client connection to H1 is broken. Indeed, now the H1
mux is responsible to create the frontend conn-stream once the request
headers are fully received. But, to properly support TCP to H1 upgrades, we
must inherit from the existing conn-stream. To do so, if the conn-stream
already exists when the client H1 connection is created, we create a H1
stream in ST_ATTACHED state, but not ST_READY, and the conn-stream is
attached to it. Because the ST_READY state is not set, no data are xferred
to the data layer when h1_rcv_buf() is called and shutdowns are inhibited
except on client aborts. This way, the request is parsed the same way than
for a classical H1 connection. Once the request headers are fully received
and parsed, the data stream is upgraded and the ST_READY state is set.
A tricky case appears when an H2 upgrade is performed because the H2 preface
is matched. In this case, the conn-stream must be detached and destroyed
before switching to the H2 mux and releasing the current H1 mux. We must
also take care to detach and destroy the conn-stream when a timeout
occurres.
This patch relies on the following series of patches :
* BUG/MEDIUM: stream: Don't immediatly ack the TCP to H1 upgrades
* MEDIUM: http-ana: Do nothing in wait-for-request analyzer if not htx
* MINOR: stream: Add a function to validate TCP to H1 upgrades
* MEDIUM: mux-h1: Add ST_READY state for the H1 connections
* MINOR: mux-h1: Wake up instead of subscribe for reads after H1C creation
* MINOR: mux-h1: Try to wake up data layer first before calling its wake callback
* MINOR: stream-int: Take care of EOS in the SI wake callback function
* BUG/MINOR: stream: Don't update counters when TCP to H2 upgrades are performed
This fix is specific for 2.4. No backport needed.
Instead of switching the stream to HTX mode, the request channel is only
reset (the request buffer is xferred to the mux) and the SF_IGNORE flag is
set on the stream. This flag prevent any processing in case of abort. Once
the upgrade confirmed, the flag is removed, in stream_upgrade_from_cs().
It is only the first part of the fix. The next one ("BUG/MAJOR: mux-h1:
Properly handle TCP to H1 upgrades") is also required. Both rely on the
following series of patches :
* MEDIUM: http-ana: Do nothing in wait-for-request analyzer if not htx
* MINOR: stream: Add a function to validate TCP to H1 upgrades
* MEDIUM: mux-h1: Add ST_READY state for the H1 connections
* MINOR: mux-h1: Wake up instead of subscribe for reads after H1C creation
* MINOR: mux-h1: Try to wake up data layer first before calling its wake callback
* MINOR: stream-int: Take care of EOS in the SI wake callback function
* BUG/MINOR: stream: Don't update counters when TCP to H2 upgrades are performed
This fix is specific for 2.4. No backport needed.
If http_wait_for_request() analyzer is called with a non-htx stream, nothing
is performed and we return immediatly. For now, it is totally unexpected.
But it will be true during TCP to H1 upgrades, once fixed. Indeed, there
will be a transition period during these upgrades. First the mux will be
upgraded and the not the stream, and finally the stream will be upgraded by
the mux once ready. In the meantime, the stream will still be in raw
mode. Nothing will be performed in wait-for-request analyzer because it will
be the mux responsibility to handle errors.
This patch is required to fix the TCP to H1 upgrades.
TCP to H1 upgrades are buggy for now. When such upgrade is performed, a
crash is experienced. The bug is the result of the recent H1 mux
refactoring, and more specifically because of the commit c4bfa59f1 ("MAJOR:
mux-h1: Create the client stream as later as possible"). Indeed, now the H1
mux is responsible to create the frontend conn-stream once the request
headers are fully received. Thus the TCP to H1 upgrade is a problem because
the frontend conn-stream already exists.
To fix the bug, we must keep this conn-stream and the associate stream and
use it in the H1 mux. To do so, the upgrade will be performed in two
steps. First, the mux is upgraded from mux-pt to mux-h1. Then, the mux-h1
performs the stream upgrade, once the request headers are fully received and
parsed. To do so, stream_upgrade_from_cs() must be used. This function set
the SF_HTX flags to switch the stream to HTX mode, it removes the SF_IGNORE
flags and eventually it fills the request channel with some input data.
This patch is required to fix the TCP to H1 upgrades and is intimately
linked with the next commits.
An alive H1 connection may be in one of these 3 states :
* ST_IDLE : not active and is waiting to be reused (no h1s and no cs)
* ST_EMBRYONIC : active with a h1s but without any cs
* ST_ATTACHED : active with a h1s and a cs
ST_IDLE and ST_ATTACHED are possible for frontend and backend
connection. ST_EMBRYONIC is only possible on the client side, when we are
waiting for the request headers. The last one is the expected state for an
active connection processing data. These states are mutually exclusives.
Now, there is a new state, ST_READY. It may only be set if ST_ATTACHED is
also set and when the CS is considered as fully active. For now, ST_READY is
set in the same time of ST_ATTACHED. But it will be used to fix TCP to H1
upgrades. Idea is to have an H1 connection in ST_ATTACHED state but not
ST_READY yet and have more or less the same behavior than an H1 connection
in ST_EMBRYONIC state. And when the upgrade is fully achieved, the ST_READY
state may be set and the data layer may be notified accordingly.
So for now, this patch should not change anything. TCP to H1 upgrades are
still buggy. But it is mandatory to make it work properly.
When a H1 connection is created, we now wakeup the H1C tasklet if there are
some data in the input buffer. If not we only subscribe for reads.
This patch is required to fix the TCP to H1 upgrades.
Instead of calling the data layer wake callback function, we now first try
to wake it up. If the data layer is subscribed for receives or for sends,
its tasklet is woken up. The wake callback function is only called as the
last chance to notify the data layer.
Because si_cs_process() is also the SI wake callback function, it may be
called from the mux layer. Thus, in such cases, it is performed outside any
I/O event and si_cs_recv() is not called. If a read0 is reported by the mux,
via the CS_FL_EOS flag, the event is not handled, because only si_cs_recv()
take care of this flag for now.
It is not a bug, because this does not happens for now. All muxes set this
flag when the data layer retrieve data (via mux->rcv_buf()). But it is safer
to be prepared to handle it from the wake callback. And in fact, it will be
useful to fix the HTTP upgrades of TCP connections (especially TCP>H1>H2
upgrades).
To be sure to not handle the same event twice, it is only handled if the
shutr is not already set on the input channel.
The reuse of idle connections should only happen for a proxy with the
http mode. In case of a backend with the tcp mode, the reuse selection
and insertion in session list are skipped.
This behavior is present since commit :
MEDIUM: connection: Add private connections synchronously in session server list
It could also be further exagerated by :
MEDIUM: backend: add reused conn to sess if mux marked as HOL blocking
It can be backported up to 2.3.
A bug was introduced by the early insertion of idle connections at the
end of connect_server. It is possible to reuse a connection not yet
ready waiting for an handshake (for example with proxy protocol or ssl).
A wrong duplicate xprt_handshake_io_cb tasklet is thus registered as a
side-effect.
This triggers the BUG_ON statement of xprt_handshake_subscribe :
BUG_ON(ctx->subs && ctx->subs != es);
To counter this, a check is now present in session_get_conn to only
return a connection without the flag CO_FL_WAIT_XPRT. This might cause
sometimes the creation of dedicated server connections when in theory
reuse could have been used, but probably only occurs rarely in real
condition.
This behavior is present since commit :
MEDIUM: connection: Add private connections synchronously in session server list
It could also be further exagerated by :
MEDIUM: backend: add reused conn to sess if mux marked as HOL blocking
It can be backported up to 2.3.
NOTE : This bug seems to be only reproducible with mode tcp, for an
unknown reason. However, reuse should never happen when not in http
mode. This improper behavior will be the subject of a dedicated patch.
This bug can easily be reproducible with the following config (a
webserver is required to accept proxy protocol on port 31080) :
global
defaults
mode tcp
timeout connect 1s
timeout server 1s
timeout client 1s
listen li
bind 0.0.0.0:4444
server bla1 127.0.0.1:31080 check send-proxy-v2
with the inject client :
$ inject -u 10000 -d 10 -G 127.0.0.1:4444
This should fix the github issue #1058.
Use chunk_inistr() for a chunk initialisation in
ssl_sock_load_sctl_from_file() instead of a manual initialisation which
was not initialising head.
Fix issue #1073.
Must be backported as far as 2.2
The new ckch_inst_new_load_srv_store() function which mimics the
ckch_inst_new_load_store() function includes some dead code which was
used only in the former function.
Fix issue #1081.
The previous patch was pushed too quickly (399bf72f6 "BUG/MINOR: stats:
Remove a break preventing ST_F_QCUR to be set for servers"). It was not an
extra break but a misplaced break statement. Thus, now a break statement
must be added after filling the ST_F_MODE field in stats_fill_sv_stats().
No backport needed except if the above commit is backported.
There is an extra break statement wrongly placed in stats_fill_sv_stats()
function, just before filling the ST_F_QCUR field. It prevents this field to
be set to the right value for servers.
No backport needed except if commit 3a9a4992 ("MEDIUM: stats: allow to
select one field in `stats_fill_sv_stats`") is backported.
Building with `"DEBUG=-DDEBUG_STRICT=1 -DDEBUG_USE_ABORT=1"` previously emitted the warning:
In file included from include/haproxy/api.h:35:0,
from src/mux_pt.c:13:
include/haproxy/buf.h: In function ‘br_init’:
include/haproxy/bug.h:42:90: warning: implicit declaration of function ‘abort’ [-Wimplicit-function-declaration]
#define ABORT_NOW() do { extern void ha_backtrace_to_stderr(); ha_backtrace_to_stderr(); abort(); } while (0)
^
include/haproxy/bug.h:56:21: note: in expansion of macro ‘ABORT_NOW’
#define CRASH_NOW() ABORT_NOW()
^
include/haproxy/bug.h:68:4: note: in expansion of macro ‘CRASH_NOW’
CRASH_NOW(); \
^
include/haproxy/bug.h:62:35: note: in expansion of macro ‘__BUG_ON’
#define _BUG_ON(cond, file, line) __BUG_ON(cond, file, line)
^
include/haproxy/bug.h:61:22: note: in expansion of macro ‘_BUG_ON’
#define BUG_ON(cond) _BUG_ON(cond, __FILE__, __LINE__)
^
include/haproxy/buf.h:875:2: note: in expansion of macro ‘BUG_ON’
BUG_ON(size < 2);
^
This patch fixes that issue. The `DEBUG_USE_ABORT` option exists for use with
static analysis tools. No backport needed.
This patch makes things more consistent between the bind_conf functions
and the server ones:
- ssl_sock_load_srv_ckchs() loads the SSL_CTX in the server
(ssl_sock_load_ckchs() load the SNIs in the bind_conf)
- add the server parameter to ssl_sock_load_srv_ckchs()
- changes made to the ckch_inst are done in
ckch_inst_new_load_srv_store()
Since the server SSL_CTX is now stored in the ckch_inst, it is not
needed anymore to pass an SSL_CTX to ckch_inst_new_load_srv_store() and
ssl_sock_load_srv_ckchs().
The new feature allowing the change of server side certificates
introduced duplicated free code. Rework the code in
cli_io_handler_commit_cert() to be more consistent.
The client_crt member is not used anymore since the server's ssl context
initialization now behaves the same way as the bind lines one (using
ckch stores and instances).
When trying to update a backend certificate, we should find a
server-side ckch instance thanks to which we can rebuild a new ssl
context and a new ckch instance that replace the previous ones in the
server structure. This way any new ssl session will be built out of the
new ssl context and the newly updated certificate.
This resolves a subpart of GitHub issue #427 (the certificate part)
In order for the backend server's certificate to be hot-updatable, it
needs to fit into the implementation used for the "bind" certificates.
This patch follows the architecture implemented for the frontend
implementation and reuses its structures and general function calls
(adapted for the server side).
The ckch store logic is kept and a dedicated ckch instance is used (one
per server). The whole sni_ctx logic was not kept though because it is
not needed.
All the new functions added in this patch are basically server-side
copies of functions that already exist on the frontend side with all the
sni and bind_cond references removed.
The ckch_inst structure has a new 'is_server_instance' flag which is
used to distinguish regular instances from the server-side ones, and a
new pointer to the server's structure in case of backend instance.
Since the new server ckch instances are linked to a standard ckch_store,
a lookup in the ckch store table will succeed so the cli code used to
update bind certificates needs to be covered to manage those new server
side ckch instances.
Split the server's ssl context initialization into the general ssl
related initializations and the actual initialization of a single
SSL_CTX structure. This way the context's initialization will be
usable by itself from elsewhere.
Reorganize the conditions for the reuse of idle/safe connections :
- reduce code by using variable to store reuse mode and idle/safe conns
counts
- consider that idle/safe/avail lists are properly allocated if
max_idle_conns not null. An allocation failure prevents haproxy
startup.
It is only a problem on the response path because the request payload length
it always known. But when a filter is registered to analyze the response
payload, the filtering may hang if the server closes just after the headers.
The root cause of the bug comes from an attempt to allow the filters to not
immediately forward the headers if necessary. A filter may choose to hold
the headers by not forwarding any bytes of the payload. For a message with
no payload but a known payload length, there is always a EOM block to
forward. Thus holding the EOM block for bodyless messages is a good way to
also hold the headers. However, messages with an unknown payload length,
there is no EOM block finishing the message, but only a SHUTR flag on the
channel to mark the end of the stream. If there is no payload when it
happens, there is no payload at all to forward. In the filters API, it is
wrongly detected as a condition to not forward the headers.
Because it is not the most used feature and not the obvious one, this patch
introduces another way to hold the message headers at the begining of the
forwarding. A filter flag is added to explicitly says the headers should be
hold. A filter may choose to set the STRM_FLT_FL_HOLD_HTTP_HDRS flag and not
forwad anything to hold the headers. This flag is removed at each call, thus
it must always be explicitly set by filters. This flag is only evaluated if
no byte has ever been forwarded because the headers are forwarded with the
first byte of the payload.
reg-tests/filters/random-forwarding.vtc reg-test is updated to also test
responses with unknown payload length (with and without payload).
This patch must be backported as far as 2.0.
use `stats_fill_sv_stats` when possible to avoid duplicating code.
the following metrics have a change of behaviour:
haproxy_server_limit_sessions
haproxy_server_queue_limit
haproxy_server_check_failures_total
haproxy_server_check_up_down_total
haproxy_server_downtime_seconds_total
haproxy_server_current_throttle
haproxy_server_idle_connections_limit
depending on cases, if the limit was not configured or enabled, NaN is
returned instead. It should not be an issue for users, even better than
before as it provides more precise info.
Signed-off-by: William Dauchy <wdauchy@gmail.com>
prometheus approach requires to output all values for a given metric
name; meaning we iterate through all metrics, and then iterate in the
inner loop on all objects for this metric.
In order to allow more code reuse, adapt the stats API to be able to
select one field or fill them all otherwise.
This patch follows what has already been done on frontend and backend
side.
From this patch it should be possible to remove most of the duplicate
code on prometheuse side for the server.
A few things to note though:
- state require prior calculation, so I moved that to a sort of helper
`stats_fill_be_stats_computestate`.
- all ST_F*TIME fields requires some minor compute, so I moved it at te
beginning of the function under a condition.
Signed-off-by: William Dauchy <wdauchy@gmail.com>
use `stats_fill_be_stats` when possible to avoid duplicating code; make
use of field selector to get the needed field only.
the only difference is on `haproxy_backend_downtime_seconds_total` as
stats.c is testing `px->srv`. This behaviour is present since commit
7344f47893 ("MINOR: stats: only report
backend's down time if it has servers"). The end result is a NaN instead
of a zero when no server are present.
Signed-off-by: William Dauchy <wdauchy@gmail.com>
prometheus approach requires to output all values for a given metric
name; meaning we iterate through all metrics, and then iterate in the
inner loop on all objects for this metric.
In order to allow more code reuse, adapt the stats API to be able to
select one field or fill them all otherwise.
This patch follows what has already been done on frontend side.
From this patch it should be possible to remove most of the duplicate
code on prometheuse side for the backend
A few things to note though:
- status and uweight field requires prior compute, so I moved that to a
sort of helper `stats_fill_be_stats_computesrv`.
- all ST_F*TIME fields requires some minor compute, so I moved it at te
beginning of the function under a condition.
Signed-off-by: William Dauchy <wdauchy@gmail.com>
GitHub issue #796 notes that many administrators miss the fact that the `hdr()`
fetch (without the `f`) splits the header value at commas. This is only
mentioned at the end of a long paragraph.
This patch attempts to improve the documentation by:
- Explaning the "comma issue" as early as possible.
- Adding newlines to split the explanation into distinct sections.
- Reducing duplication by making the `res` siblings refer to their `req`
counterparts.
This patch may be backported as long as it applies cleanly. During the
refactoring I needed to adjust several explanations for consistency and not all
of them might be available in older branches.
while working on backend/servers I realised I could have written that in
a better way and avoid one extra break. This is slightly improving
readiness.
also while being here, fix function declaration which was not 100%
accurate.
this patch does not change the behaviour of the code.
Signed-off-by: William Dauchy <wdauchy@gmail.com>
Not necessarily mandatory but I saw a few prometheus client parsing only
`NaN`. Also most librarries do output `NaN`
Signed-off-by: William Dauchy <wdauchy@gmail.com>
