The following inlined functions are particularly large (and probably not
inlined at all by the compiler), and together represent roughly half of
the file, while they're used at most once per connection. They were moved
to connection.c.
conn_upgrade_mux_fe, conn_install_mux_fe, conn_install_mux_be,
conn_install_mux_chk, conn_delete_from_tree, conn_init, conn_new,
conn_free
We do not really need to have them inlined, and having xxhash.h included
by connection.h results in this 4700-lines file being processed 101 times
over the whole project, which accounts for 13.5% of the total size!
Additionally, half of the functions are only needed from connection.c.
Let's move the functions there and get rid of the painful include.
The build time is now down to 6.2s just due to this.
The hash type stored everywhere is XXH64_hash_t, which annoyingly forces
everyone to include the huge xxhash file. We know it's an uint64_t because
that's its purpose and the type is only made to abstract it on machines
where uint64_t is not availble. Let's switch the type to uint64_t
everywhere and avoid including xxhash from the type file.
This one doesn't use anything from an SSL context, it only checks the
type of the transport layer of a connection, thus it belongs to
connection.h. This is particularly visible due to all the ifdefs
around it in various call places.
We'll need to improve the API to pass other arguments in the future, so
let's start to adapt better to the current use cases. task_new() is used:
- 18 times as task_new(tid_bit)
- 18 times as task_new(MAX_THREADS_MASK)
- 2 times with a single bit (in a loop)
- 1 in the debug code that uses a mask
This patch provides 3 new functions to achieve this:
- task_new_here() to create a task on the calling thread
- task_new_anywhere() to create a task to be run anywhere
- task_new_on() to create a task to run on a specific thread
The change is trivial and will allow us to later concentrate the
required adaptations to these 3 functions only. It's still possible
to call task_new() if needed but a comment was added to encourage the
use of the new ones instead. The debug code was not changed and still
uses it.
Move the code to allocate/free the mux cleanup task outside of the polling
loop. A new thread_alloc/free handler is registered for this in
connection.c.
This has the benefit to clean up the polling loop code. And as another
benefit, if the task allocation fails, the handler can report an error
to exit the haproxy process. This prevents a potential null pointer
dereferencing.
This should fix the github issue #1389.
This must be backported up to 2.4.
The bc_conn_err and bc_conn_err_str sample fetches give the status of
the connection on the backend side. The error codes and error messages
are the same than the ones that can be raised by the fc_conn_err fetch.
The fc_conn_err and fc_conn_err_str sample fetches give information
about the problem that made the connection fail. This information would
previously only have been given by the error log messages meaning that
thanks to these fetches, the error log can now be included in a custom
log format. The log strings were all found in the conn_err_code_str
function.
This function appends to a buffer some information from a connection.
This will be used by traces and possibly some debugging as well. A
frontend/backend/server, transport/control layers, source/destination
ip:port, connection pointer and direction are reported depending on
the available information.
No idea why this was put inlined into connection.h, it's used only once
for haproxy -vv, and requires tools.h, causing an undesired dependency
from connection.h. Let's move it to connection.c instead where it ought
to have been.
Implement a safe mechanism to close front idling connection which
prevents the soft-stop to complete. Every h1/h2 front connection is
added in a new per-thread list instance. On shutdown, a new task is
waking up which calls wake mux operation on every connection still
present in the new list.
A new stopping_list attach point has been added in the connection
structure. As this member is only used for frontend connections, it
shared the same union as the session_list reserved for backend
connections.
bc_http_major sample fetch now works when it is called from a
tcp-check. When it happens, the session origin is a check. The backend
connection is retrieved from the conn-stream attached to the check.
If required, this path may easily be backported as far as 2.2.
fc_http_major and bc_http_major sample fetches return the major digit of the
HTTP version used, respectively, by the frontend and the backend
connections, based on the mux. However, in reality, "2" is returned if the
H2 mux is detected, otherwise "1" is inconditionally returned, regardless
the mux used. Thus, if called for a raw TCP connection, "1" is returned.
To fix this bug, we now get the multiplexer flags, if there is one, to be
sure MX_FL_HTX is set.
I guess it was made this way on purpose when the H2 multiplexer was
introduced in the 1.8 and with the legacy HTTP mode there is no other
solution at the connection level. Thus this patch should be backported as
far as 2.2. For the 2.0, it must be evaluated first because of the legacy
HTTP mode.
The default proxy was passed as a variable to all parsers instead of a
const, which is not without risk, especially when some timeout parsers used
to make some int pointers point to the default values for comparisons. We
want to be certain that none of these parsers will modify the defaults
sections by accident, so it's important to mark this proxy as const.
This patch touches all occurrences found (89).
There are multiple per-thread lists in the listeners, which isn't the
most efficient in terms of cache, and doesn't easily allow to store all
the per-thread stuff.
Now we introduce an srv_per_thread structure which the servers will have an
array of, and place the idle/safe/avail conns tree heads into. Overall this
was a fairly mechanical change, and the array is now always initialized for
all servers since we'll put more stuff there. It's worth noting that the Lua
code still has to deal with its own deinit by itself despite being in a
global list, because its server is not dynamically allocated.
If dispatch mode or transparent backend is used, the backend connection
target is a proxy instead of a server. In these cases, the reuse of
backend connections is not consistent.
With the default behavior, no reuse is done and every new request uses a
new connection. However, if http-reuse is set to never, the connection
are stored by the mux in the session and can be reused for future
requests in the same session.
As no server is used for these connections, no reuse can be made outside
of the session, similarly to http-reuse never mode. A different
http-reuse config value should not have an impact. To achieve this, mark
these connections as private to have a defined behavior.
For this feature to properly work, the connection hash has been slightly
adjusted. The server pointer as an input as been replaced by a generic
target pointer to refer to the server or proxy instance. The hash is
always calculated on connect_server even if the connection target is not
a server. This also requires to allocate the connection hash node for
every backend connections, not just the one with a server target.
Remove ebmb_node entry from struct connection and create a dedicated
struct conn_hash_node. struct connection contains now only a pointer to
a conn_hash_node, allocated only for connections where target is of type
OBJ_TYPE_SERVER. This will reduce memory footprints for every
connections that does not need http-reuse such as frontend connections.
Use the proxy protocol frame if proxy protocol is activated on the
server line. Do not add anymore these connections in the private list.
If some requests are made with the same proxy fields, they can reuse
the idle connection.
The reg-tests proxy_protocol_send_unique_id must be adapted has it
relied on the side effect behavior that every requests from a same
connection reused a private server connection. Now, a new connection is
created as expected if the proxy protocol fields differ.
The source address is used as an input to the the server connection hash. The
address and port are used as separate hash inputs. Do not add anymore these
connections in the private list.
This parameter is set only if used in the transparent-proxy mode.
The destination address is used as an input to the server connection hash. The
address and port are used as separated hash inputs. Note that they are not used
when statically specified on the server line. This is only useful for dynamic
destination address.
This is typically used when the server address is dynamically set via the
set-dst action. The address and port are separated hash parameters.
Most notably, it should fixed set-dst use case (cf github issue #947).
The sni parameter is an input to the server connection hash. Do not add
anymore connections with dynamic sni in the private list. Thus, it is
now possible to reuse a server connection if they use the same sni.
The pointer of the target server is used as a first parameter for the
server connection hash calcul. This prevents the hash to be null when no
specific parameters are present, and can serve as a simple defense
against an attacker trying to reuse a non-conform connection.
This is a preliminary work for the calcul of the backend connection
hash. A structure conn_hash_params is the input for the operation,
containing the various specific parameters of a connection.
The high bits of the hash will reflect the parameters present as input.
A set of macros is written to manipulate the connection hash and extract
the parameters/payload.
The server idle/safe/available connection lists are replaced with ebmb-
trees. This is used to store backend connections, with the new field
connection hash as the key. The hash is a 8-bytes size field, used to
reflect specific connection parameters.
This is a preliminary work to be able to reuse connection with SNI,
explicit src/dst address or PROXY protocol.
This changes the subscribe/unsubscribe functions to rely on the control
layer's check_events/ignore_events. At the moment only the socket version
of these functions is present so the code should basically be the same.
Now we don't touch the fd anymore there, instead we rely on the ->drain()
provided by the control layer. As such the function was renamed to
conn_ctrl_drain().
conn_fd_handler() is 100% specific to socket code. It's about time
it moves to sock.c which manipulates socket FDs. With it comes
conn_fd_check() which tests for the socket's readiness. The ugly
connection status check at the end of the iocb was moved to an inlined
function in connection.h so that if we need it for other socket layers
it's not too hard to reuse.
The code was really only moved and not changed at all.
The send() loop present in this function and the error handling is already
present in raw_sock_from_buf(). Let's rely on it instead and stop touching
the FD from this place. The send flag was changed to use a more agnostic
CO_SFL_*. The name was changed to "conn_ctrl_send()" to remind that it's
meant to be used to send at the lowest level.
This last call to conn_cond_update_polling() is now totally misleading as
the function only stops polling in case of unrecoverable connection error.
Let's open-code the test to make it more prominent and explain what we're
trying to do there. It's even almost certain this code is never executed
anymore, as the only remaining case should be a mux's wake function setting
CO_FL_ERROR without disabling the polling, but they need to be audited first
to make sure this is the case.
conn_stop_polling() in fact only calls fd_stop_both() after checking
that the ctrl layer is ready. It's the case in conn_fd_check() so
let's get rid of this next-to-last user of this function.
At a few places we used to rely on conn->owner to retrieve the session
while the session is already known. This is not correct because at some
of these points the reason the connection's owner was still the session
(instead of NULL) is a mistake. At one place a comparison is even made
between the session and conn->owner assuming it's valid without checking
if it's NULL. Let's clean this up to use the session all the time.
Note that this will be needed for a forthcoming fix and will have to be
backported.
Roughly half of the calls to sockadr_alloc() are made to copy an already
known address. Let's optionally pass it in argument so that the function
can handle the copy at the same time, this slightly simplifies its usage.
No protocol defines it anymore. The last user used to be the monitor-net
stuff that got partially broken already when the tcp_drain() function
moved to conn_sock_drain() with commit e215bba95 ("MINOR: connection:
make conn_sock_drain() work for all socket families") in 1.9-dev2.
A part of this will surely move back later when non-socket connections
arrive with QUIC but better keep the API clean and implement what's
needed in time instead.
When a connection is marked as private, it is now added in the session server
list. We don't wait a stream is detached from the mux to do so. When the
connection is created, this happens after the mux creation. Otherwise, it is
performed when the connection is marked as private.
To allow that, when a connection is created, the session is systematically set
as the connectin owner. Thus, a backend connection has always a owner during its
creation. And a private connection has always a owner until its death.
Note that outside the detach() callback, if the call to session_add_conn()
failed, the error is ignored. In this situation, we retry to add the connection
into the session server list in the detach() callback. If this fails at this
step, the multiplexer is destroyed and the connection is closed.
When the multiplexer creation is delayed after the handshakes phase, the
connection is added in the available connection list if http-reuse never is not
configured for the backend. But it is a wrong statement. At this step, the
connection is not safe because it is a new connection. So it must be added in
the available connection list only if http-reuse always is used.
No backport needed, this is 2.2-dev.
When a connection is created and the multiplexer is installed, if the connection
is marked as private, don't consider it as available, regardless the number of
available streams. This test is performed when the mux is installed when the
connection is created, in connect_server(), and when the mux is installed after
the handshakes stage.
No backport needed, this is 2.2-dev.
We used to have 3 thread-based arrays for toremove_lock, idle_cleanup,
and toremove_connections. The problem is that these items are small,
and that this creates false sharing between threads since it's possible
to pack up to 8-16 of these values into a single cache line. This can
cause real damage where there is contention on the lock.
This patch creates a new array of struct "idle_conns" that is aligned
on a cache line and which contains all three members above. This way
each thread has access to its variables without hindering the other
ones. Just doing this increased the HTTP/1 request rate by 5% on a
16-thread machine.
The definition was moved to connection.{c,h} since it appeared a more
natural evolution of the ongoing changes given that there was already
one of them declared in connection.h previously.
This effectively reverts the two following commits:
6f95f6e11 ("OPTIM: connection: disable receiving on disabled events when the run queue is too high")
065a02561 ("MEDIUM: connection: don't stop receiving events in the FD handler")
The problem as reported in issue #662 is that when the events signals
the readiness of input data that has to be forwarded over a congested
stream, the mux will read data and wake the stream up to forward them,
but the buffer full condition makes this impossible immediately, then
nobody in the chain will be able to disable the event after it was
first reported. And given we don't know at the connection level whether
an event was already reported or not, we can't decide anymore to
forcefully stop it if for any reason its processing gets delayed.
The problem is magnified in issue #662 by the fact that a shutdown is
reported with pending data occupying the buffer. The shutdown will
strike in loops and cause the upper layer stream to be notified until
it's handled, but with a buffer full it's not possible to call cs_recv()
hence to purge the event.
All this can only be handled optimally by implementing a lower layer,
direct mux-to-mux forwarding that will not require any scheduling. This
was no wake up will be needed and the event will be instantly handled
or paused for a long time.
For now let's simply revert these optimizations. Running a 1 MB transfer
test over H2 using 8 connections having each 32 streams with a limited
link of 320 Mbps shows the following profile before this fix:
calls syscall (100% CPU)
------ -------
259878 epoll_wait
519759 clock_gettime
17277 sendto
17129 recvfrom
672 epoll_ctl
And the following one after the fix:
calls syscall (2-3% CPU)
------ -------
17201 sendto
17357 recvfrom
2304 epoll_wait
4609 clock_gettime
1200 epoll_ctl
Thus the behavior is much better.
No backport is needed as these patches were only in 2.2-dev.
Many thanks to William Dauchy for reporting a lot of details around this
difficult issue.
This patch fixes all the leftovers from the include cleanup campaign. There
were not that many (~400 entries in ~150 files) but it was definitely worth
doing it as it revealed a few duplicates.
Most of the files dealing with error reports have to include log.h in order
to access ha_alert(), ha_warning() etc. But while these functions don't
depend on anything, log.h depends on a lot of stuff because it deals with
log-formats and samples. As a result it's impossible not to embark long
dependencies when using ha_warning() or qfprintf().
This patch moves these low-level functions to errors.h, which already
defines the error codes used at the same places. About half of the users
of log.h could be adjusted, sometimes revealing other issues such as
missing tools.h. Interestingly the total preprocessed size shrunk by
4%.
There's no point splitting the file in two since only cfgparse uses the
types defined there. A few call places were updated and cleaned up. All
of them were in C files which register keywords.
There is nothing left in common/ now so this directory must not be used
anymore.
The type file is becoming a mess, half of it is for the proxy protocol,
another good part describes conn_streams and mux ops, it would deserve
being split again. At least it was reordered so that elements are easier
to find, with the PP-stuff left at the end. The MAX_SEND_FD macro was moved
to compat.h as it's said to be the value for Linux.
This one is particularly tricky to move because everyone uses it
and it depends on a lot of other types. For example it cannot include
arg-t.h and must absolutely only rely on forward declarations to avoid
dependency loops between vars -> sample_data -> arg. In order to address
this one, it would be nice to split the sample_data part out of sample.h.
A few includes were missing in each file. A definition of
struct polled_mask was moved to fd-t.h. The MAX_POLLERS macro was
moved to defaults.h
Stdio used to be silently inherited from whatever path but it's needed
for list_pollers() which takes a FILE* and which can thus not be
forward-declared.
The type was moved out as it's used by standard.h for netns_entry.
Instead of just being a forward declaration when not used, it's an
empty struct, which makes gdb happier (the resulting stripped executable
is the same).
This one used to be stored into debug.h but the debug tools got larger
and require a lot of other includes, which can't use BUG_ON() anymore
because of this. It does not make sense and instead this macro should
be placed into the lower includes and given its omnipresence, the best
solution is to create a new bug.h with the few surrounding macros needed
to trigger bugs and place assertions anywhere.
Another benefit is that it won't be required to add include <debug.h>
anymore to use BUG_ON, it will automatically be covered by api.h. No
less than 32 occurrences were dropped.
The FSM_PRINTF macro was dropped since not used at all anymore (probably
since 1.6 or so).
All files that were including one of the following include files have
been updated to only include haproxy/api.h or haproxy/api-t.h once instead:
- common/config.h
- common/compat.h
- common/compiler.h
- common/defaults.h
- common/initcall.h
- common/tools.h
The choice is simple: if the file only requires type definitions, it includes
api-t.h, otherwise it includes the full api.h.
In addition, in these files, explicit includes for inttypes.h and limits.h
were dropped since these are now covered by api.h and api-t.h.
No other change was performed, given that this patch is large and
affects 201 files. At least one (tools.h) was already freestanding and
didn't get the new one added.
It is possible to send a unique ID when the PROXY protocol v2 is used. It relies
on the stream to do so. So we must be sure to have a stream. Locally initiated
connections may not be linked to a stream. For instance, outgoing connections
created by health checks have no stream. Moreover, the stream is not retrieved
for mux-less connections (this bug will be fixed in another commit).
Unfortunately, in make_proxy_line_v2() function, the stream is not tested before
generating the unique-id. This bug leads to a segfault when a health check is
performed for a server with the PROXY protocol v2 and the unique-id option
enabled. It also crashes for servers using SSL connections with alpn. The bug
was introduced by the commit cf6e0c8a8 ("MEDIUM: proxy_protocol: Support sending
unique IDs using PPv2")
This patch should fix the issue #640. It must be backported to the same versions
as the commit above.
A bug in the PROXY protocol v2 implementation was present in HAProxy up to
version 2.1, causing it to emit a PROXY command instead of a LOCAL command
for health checks. This is particularly minor but confuses some servers'
logs. Sadly, the bug was discovered very late and revealed that some servers
which possibly only tested their PROXY protocol implementation against
HAProxy fail to properly handle the LOCAL command, and permanently remain in
the "down" state when HAProxy checks them. When this happens, it is possible
to enable this global option to revert to the older (bogus) behavior for the
time it takes to contact the affected components' vendors and get them fixed.
This option is disabled by default and acts on all servers having the
"send-proxy-v2" statement.
Older versions were reverted to the old behavior and should not attempt to
be fixed by default again. However a variant of this patch could possibly
be implemented to ask to explicitly send LOCAL if needed by some servers.
More context here:
https://www.mail-archive.com/haproxy@formilux.org/msg36890.htmlhttps://www.mail-archive.com/haproxy@formilux.org/msg37218.html
When a tcp-check connect rule is evaluated, the mux protocol corresponding to
the health-check is chosen. So for TCP based health-checks, the mux-pt is
used. For HTTP based health-checks, the mux-h1 is used. The connection is marked
as private to be sure to not ruse regular HTTP connection for
health-checks. Connections reuse will be evaluated later.
The functions evaluating HTTP send rules and expect rules have been updated to
be HTX compliant. The main change for users is that HTTP health-checks are now
stricter on the HTTP message format. While before, the HTTP formatting and
parsing were minimalist, now messages should be well formatted.
Commit 7f26391bc5 ("BUG/MINOR: connection: make sure to correctly tag
local PROXY connections") revealed that some implementations do not
properly ignore addresses in LOCAL connections (at least Dovecot was
spotted). More context information in the thread below:
https://www.mail-archive.com/haproxy@formilux.org/msg36890.html
The patch above was using LOCAL on top of local addresses in order to
minimize the risk of breakage but revealed worse than a clean fix. So
let's partially revert it and send pure LOCAL connections instead now.
After a bit of observation, this patch should be progressively backported
to stable branches. However if it reveals new breakage, the backport of
the patch above will have to be reverted from stable branches while other
products work on fixing their code based on the master branch.
Make the "list" element a struct mt_list, and explicitely use
list_from_mt_list to get a struct list * where it is used as such, so that
mt_list_for_each_entry will be usable with it.
Revamp the server connection lists. We know have 3 lists :
- idle_conns, which contains idling connections
- safe_conns, which contains idling connections that are safe to use even
for the first request
- available_conns, which contains connections that are not idling, but can
still accept new streams (those are HTTP/2 or fastcgi, and are always
considered safe).
This adds the missing blank lines in `make_proxy_line_v2` and
`conn_recv_proxy`. It also adjusts the type of the temporary variable
used for the return value of `recv` to be `ssize_t` instead of `int`.
This patch adds the `unique-id` option to `proxy-v2-options`. If this
option is set a unique ID will be generated based on the `unique-id-format`
while sending the proxy protocol v2 header and stored as the unique id for
the first stream of the connection.
This feature is meant to be used in `tcp` mode. It works on HTTP mode, but
might result in inconsistent unique IDs for the first request on a keep-alive
connection, because the unique ID for the first stream is generated earlier
than the others.
Now that we can send unique IDs in `tcp` mode the `%ID` log variable is made
available in TCP mode.
This patch fixes PROXYv2 parsing when the payload of the TCP connection is
fused with the PROXYv2 header within a single recv() call.
Previously HAProxy ignored the PROXYv2 header length when attempting to
parse the TLV, possibly interpreting the first byte of the payload as a
TLV type.
This patch adds proper validation. It ensures that:
1. TLV parsing stops when the end of the PROXYv2 header is reached.
2. TLV lengths cannot exceed the PROXYv2 header length.
3. The PROXYv2 header ends together with the last TLV, not allowing for
"stray bytes" to be ignored.
A reg-test was added to ensure proper behavior.
This patch tries to find the sweat spot between a small and easily
backportable one, and a cleaner one that's more easily adaptable to
older versions, hence why it merges the "if" and "while" blocks which
causes a reindent of the whole block. It should be used as-is for
versions 1.9 to 2.1, the block about PP2_TYPE_AUTHORITY should be
dropped for 2.0 and the block about CRC32C should be dropped for 1.8.
This bug was introduced when TLV parsing was added. This happened in commit
b3e54fe387. This commit was first released
with HAProxy 1.6-dev1.
A similar issue was fixed in commit 7209c204bd.
This patch must be backported to HAProxy 1.6+.
In order to save a lot on syscalls, we currently don't disable receiving
on a file descriptor anymore if its handler was already woken up. But if
the run queue is huge and the poller collects a lot of events, this causes
excess wakeups which take CPU time which is not used to flush these tasklets.
This patch simply considers the run queue size to decide whether or not to
stop receiving. Tests show that by stopping receiving when the run queue
reaches ~16 times its configured size, we can still hold maximal performance
in extreme situations like maxpollevents=20k for runqueue_depth=2, and still
totally avoid calling epoll_event under moderate load using default settings
on keep-alive connections.
We used to call ->wake() for any I/O event for which there was no
subscriber. But this is a problem because this causes massive wake()
storms since we disabled fd_stop_recv() to save syscalls.
The only reason for the io_available condition is to detect that an
asynchronous connect() just finished and will not be handled by any
registered event handler. Since we now properly handle synchronous
connects, we can detect this situation by the fact that we had a
success on conn_fd_check() and no requested I/O took over.
With commit 065a025610 ("MEDIUM: connection: don't stop receiving events
in the FD handler") we disabled a number of fd_stop_* in conn_fd_handler(),
in order to wait for their respective handlers to deal with them. But it
is not correct to do that for the send direction, as we may very well
have nothing to send. This is visible when connecting in TCP mode to
a server with no data to send, there's nobody anymore to disable the
polling for the send direction.
And it is logical, on the recv() path we know the system has data to
deliver and that some code will be in charge of it. On the send
direction we simply don't know if it was the result of a successful
connect() or if there is still something to send. In any case we
almost never fill the network buffer on a single send() after being
woken up by the system, so disabling the FD immediately or much later
will not change the number of operations.
No backport is needed, this is 2.2-dev.
The remaining epoll_ctl() calls are exclusively caused by the disagreement
between conn_fd_handler() and the mux receiving the data: the fd handler
wants to stop after having woken up the tasklet, then the mux after
receiving data wants to receive again. Given that they don't happen in
the same poll loop when there are many FDs, this causes a lot of state
changes.
As suggested by Olivier, if the task is already scheduled for running,
we don't need to disable the event because it's in the run queue, poll()
cannot stop, and reporting it again will be harmless. What *might*
happen however is that a sampling-based poller like epoll() would report
many times the same event and has trouble getting others behind. But if
it would happen, it would still indicate the run queue has plenty of
pending operations, so it would in fact only displace the problem from
the poller to the run queue, which doesn't seem to be worse (and in
fact we do support priorities while the poller does not).
By doing this change, the keep-alive test with 1k conns and 100k reqs
completely gets rid of the per-request epoll_ctl changes, while still
not causing extra recvfrom() :
$ ./h1load -n 100000 -t 4 -c 1000 -T 20 -F 127.0.0.1:8001/?s=1k/t=20
200000 sendto 1
200000 recvfrom 1
10762 epoll_wait 1
3664 epoll_ctl 1
1999 recvfrom -1
In close mode, it didn't change anything, we're still in the optimal
case (2 epoll per connection) :
$ ./h1load -n 100000 -r 1 -t 4 -c 1000 -T 20 -F 127.0.0.1:8001/?s=1k/t=20
203764 epoll_ctl 1
200000 sendto 1
200000 recvfrom 1
6091 epoll_wait 1
2994 recvfrom -1
There's currently an internal API limitation at the connection layer
regarding conn_subscribe(). We must not subscribe if we haven't yet
met EAGAIN or such a condition, so we sometimes force ourselves to read
in order to meet this condition and being allowed to call subscribe.
But reading cannot always be done (e.g. at the end of a loop where we
cannot afford to retrieve new data and start again) so we instead
perform a tasklet_wakeup() of the requester's io_cb. This is what is
done in mux_h1 for example. The problem with this is that it forces
a new receive when we're not necessarily certain we need one. And if
the FD is not ready and was already being polled, it's a useless
wakeup.
The current patch improves the connection-level subscribe() so that
it really manipulates the polling if the FD is marked not-ready, but
instead schedules the argument tasklet for a wakeup if the FD is
ready. This guarantees we'll wake this tasklet up in any case once the
FD is ready, either immediately or after polling.
By doing so, a test on pure close mode shows we cut in half the number
of epoll_ctl() calls and almost eliminate failed recvfrom():
$ ./h1load -n 100000 -r 1 -t 4 -c 1000 -T 20 -F 127.0.0.1:8001/?s=1k/t=20
before:
399464 epoll_ctl 1
200007 recvfrom 1
200000 sendto 1
100000 recvfrom -1
7508 epoll_wait 1
after:
205739 epoll_ctl 1
200000 sendto 1
200000 recvfrom 1
6084 epoll_wait 1
2651 recvfrom -1
On keep-alive there is no change however.
As reported in issue #511, when sending an outgoing local connection
(e.g. health check) we must set the "local" tag and not a "proxy" tag.
The issue comes from historic support on v1 which required to steal the
address on the outgoing connection for such ones, creating confusion in
the v2 code which believes it sees the incoming connection.
In order not to risk to break existing setups which might rely on seeing
the LB's address in the connection's source field, let's just change the
connection type from proxy to local and keep the addresses. The protocol
spec states that for local, the addresses must be ignored anyway.
This problem has always existed, this can be backported as far as 1.5,
though it's probably not a good idea to change such setups, thus maybe
2.0 would be more reasonable.
This marks the end of the transition from the connection polling states
introduced in 1.5-dev12 and the subscriptions in that arrived in 1.9.
The socket layer can now safely use its FD while all upper layers rely
exclusively on subscriptions. These old functions were removed. Some may
deserve some renaming to improved clarty though. The single call to
conn_xprt_stop_both() was dropped in favor of conn_cond_update_polling()
which already does the same.
The last few calls to conn_xprt_{want,stop}_{recv,send} in the central
connection code were replaced with their strictly exact equivalent fd_*,
adding the call to conn_ctrl_ready() when it was missing.
Historically we used to require that the connections held the desired
polling states for the data layer and the socket layer. Then with muxes
these were more or less merged into the transport layer, and now it
happens that with all transport layers having their own state, the
"transport layer state" as we have it in the connection (XPRT_RD_ENA,
XPRT_WR_ENA) is only an exact copy of the undelying file descriptor
state, but with a delay. All of this is causing some difficulties at
many places in the code because there are still some locations which
use the conn_want_* API to remain clean and only rely on connection,
and count on a later collection call to conn_cond_update_polling(),
while others need an immediate action and directly use the FD updates.
Since our updates are now much cheaper, most of them being only an
atomic test-and-set operation, and since our I/O callbacks are deferred,
there's no benefit anymore in trying to "cache" the transient state
change in the connection flags hoping to cancel them before they
become an FD event. Better make such calls transparent indirections
to the FD layer instead and get rid of the deferred operations which
needlessly complicate the logic inside.
This removes flags CO_FL_XPRT_{RD,WR}_ENA and CO_FL_WILL_UPDATE.
A number of functions related to polling updates were either greatly
simplified or removed.
Two places were using CO_FL_XPRT_WR_ENA as a hint to know if more data
were expected to be sent after a PROXY protocol or SOCKSv4 header. These
ones were simply replaced with a check on the subscription which is
where we ought to get the autoritative information from.
Now the __conn_xprt_want_* and their conn_xprt_want_* counterparts
are the same. conn_stop_polling() and conn_xprt_stop_both() are the
same as well. conn_cond_update_polling() only causes errors to stop
polling. It also becomes way more obvious that muxes should not at
all employ conn_xprt_{want|stop}_{recv,send}(), and that the call
to __conn_xprt_stop_recv() in case a mux failed to allocate a buffer
is inappropriate, it ought to unsubscribe from reads instead. All of
this definitely requires a serious cleanup.
Issue #490 reports that there are a few bogus constructs of the famous
"do { if (cond) continue; } while (0)" in the connection code, that are
used to retry on I/O failures caused by receipt of a signal. Let's turn
them into the more correct "while (1) { if (cond) continue; break }"
instead. This may or may not be backported, it shouldn't have any
visible effect.
triggered by coverity; src_port is set earlier.
this should fix github issue #467
Fixes: 7fec021537 ("MEDIUM: proxy_protocol: Convert IPs to v6 when
protocols are mixed")
This should be backported to 1.8.
Signed-off-by: William Dauchy <w.dauchy@criteo.com>
Reviewed-by: Tim Duesterhus <tim@bastelstu.be>
Just like with CO_FL_SOCK_RD_SH, we don't need to check for this flag too
early because conn_sock_send() already does it. No error was lost so it
was harmless, it was only useless code.
The handshake functions dedicated to proxy proto, netscaler and
socks4 all check for this flag before proceeding. This is wrong,
they must not do and instead perform the call to recv() then
report the close. The reason for this is that the current
construct managed to lose the CO_ER_CIP_EMPTY error code in case
the connection was already shut, thus causing a race condition
with some errors being reported correctly or as unknown depending
on the timing.
As mentioned in commit c192b0ab95 ("MEDIUM: connection: remove
CO_FL_CONNECTED and only rely on CO_FL_WAIT_*"), there is a lack of
consistency on which flags are checked among L4/L6/HANDSHAKE depending
on the code areas. A number of sample fetch functions only check for
L4L6 to report MAY_CHANGE, some places only check for HANDSHAKE and
many check both L4L6 and HANDSHAKE.
This patch starts to make all of this more consistent by introducing a
new mask CO_FL_WAIT_XPRT which is the union of L4/L6/HANDSHAKE and
reports whether the transport layer is ready or not.
All inconsistent call places were updated to rely on this one each time
the goal was to check for the readiness of the transport layer.
Commit 477902bd2e ("MEDIUM: connections: Get ride of the xprt_done
callback.") broke the master CLI for a very obscure reason. It happens
that short requests immediately terminated by a shutdown are properly
received, CS_FL_EOS is correctly set, but in si_cs_recv(), we refrain
from setting CF_SHUTR on the channel because CO_FL_CONNECTED was not
yet set on the connection since we've not passed again through
conn_fd_handler() and it was not done in conn_complete_session(). While
commit a8a415d31a ("BUG/MEDIUM: connections: Set CO_FL_CONNECTED in
conn_complete_session()") fixed the issue, such accident may happen
again as the root cause is deeper and actually comes down to the fact
that CO_FL_CONNECTED is lazily set at various check points in the code
but not every time we drop one wait bit. It is not the first time we
face this situation.
Originally this flag was used to detect the transition between WAIT_*
and CONNECTED in order to call ->wake() from the FD handler. But since
at least 1.8-dev1 with commit 7bf3fa3c23 ("BUG/MAJOR: connection: update
CO_FL_CONNECTED before calling the data layer"), CO_FL_CONNECTED is
always synchronized against the two others before being checked. Moreover,
with the I/Os moved to tasklets, the decision to call the ->wake() function
is performed after the I/Os in si_cs_process() and equivalent, which don't
care about this transition either.
So in essence, checking for CO_FL_CONNECTED has become a lazy wait to
check for (CO_FL_WAIT_L4_CONN | CO_FL_WAIT_L6_CONN), but that always
relies on someone else having synchronized it.
This patch addresses it once for all by killing this flag and only checking
the two others (for which a composite mask CO_FL_WAIT_L4L6 was added). This
revealed a number of inconsistencies that were purposely not addressed here
for the sake of bisectability:
- while most places do check both L4+L6 and HANDSHAKE at the same time,
some places like assign_server() or back_handle_st_con() and a few
sample fetches looking for proxy protocol do check for L4+L6 but
don't care about HANDSHAKE ; these ones will probably fail on TCP
request session rules if the handshake is not complete.
- some handshake handlers do validate that a connection is established
at L4 but didn't clear CO_FL_WAIT_L4_CONN
- the ->ctl method of mux_fcgi, mux_pt and mux_h1 only checks for L4+L6
before declaring the mux ready while the snd_buf function also checks
for the handshake's completion. Likely the former should validate the
handshake as well and we should get rid of these extra tests in snd_buf.
- raw_sock_from_buf() would directly set CO_FL_CONNECTED and would only
later clear CO_FL_WAIT_L4_CONN.
- xprt_handshake would set CO_FL_CONNECTED itself without actually
clearing CO_FL_WAIT_L4_CONN, which could apparently happen only if
waiting for a pure Rx handshake.
- most places in ssl_sock that were checking CO_FL_CONNECTED don't need
to include the L4 check as an L6 check is enough to decide whether to
wait for more info or not.
It also becomes obvious when reading the test in si_cs_recv() that caused
the failure mentioned above that once converted it doesn't make any sense
anymore: having CS_FL_EOS set while still waiting for L4 and L6 to complete
cannot happen since for CS_FL_EOS to be set, the other ones must have been
validated.
Some of these parts will still deserve further cleanup, and some of the
observations above may induce some backports of potential bug fixes once
totally analyzed in their context. The risk of breaking existing stuff
is too high to blindly backport everything.
We can't just assume conn_create_mux() will be called, and set CO_FL_CONNECTED,
conn_complete_session() might be call synchronously if we're not using SSL,
so ew haee no choice but to set CO_FL_CONNECTED in there. This should fix
the recent breakage of the mcli reg tests.
The xprt_done_cb callback was used to defer some connection initialization
until we're connected and the handshake are done. As it mostly consists of
creating the mux, instead of using the callback, introduce a conn_create_mux()
function, that will just call conn_complete_session() for frontend, and
create the mux for backend.
In h2_wake(), make sure we call the wake method of the stream_interface,
as we no longer wakeup the stream task.
In conn_recv_netscaler_cip(), don't forget to allocate conn->src and
conn->dst, as those are now dynamically allocated. Not doing so results in
getting a crash when using netscaler.
This should fix github issue #460.
This should be backported to 2.1.
The subscriber used to be passed as a "void *param" that was systematically
cast to a struct wait_event*. By now it appears clear that the subscribe()
call at every layer is well defined and always takes a pointer to an event
subscriber of type wait_event, so let's enforce this in the functions'
prototypes, remove the intermediary variables used to cast it and clean up
the comments to clarify what all these functions do in their context.
In practice all callers use the same wait_event notification for any I/O
so instead of keeping specific code to handle them separately, let's merge
them and it will allow us to create new events later.
These ones used to serve as a set of switches between CO_FL_SOCK_* and
CO_FL_XPRT_*, and now that the SOCK layer is gone, they're always a
copy of the last know CO_FL_XPRT_* ones that is resynchronized before
I/O events by calling conn_refresh_polling_flags(), and that are pushed
back to FDs when detecting changes with conn_xprt_polling_changes().
While these functions are not particularly heavy, what they do is
totally redundant by now because the fd_want_*/fd_stop_*() actions
already perform test-and-set operations to decide to create an entry
or not, so they do the exact same thing that is done by
conn_xprt_polling_changes(). As such it is pointless to call that
one, and given that the only reason to keep CO_FL_CURR_* is to detect
changes there, we can now remove them.
Even if this does only save very few cycles, this removes a significant
complexity that has been responsible for many bugs in the past, including
the last one affecting FreeBSD.
All tests look good, and no performance regressions were observed.
Both flags became equal in commit 82967bf9 ("MINOR: connection: adjust
CO_FL_NOTIFY_DATA after removal of flags"), which already predicted the
overlap between xprt_done_cb() and wake() after the removal of the DATA
specific flags in 1.8. Let's simply remove CO_FL_NOTIFY_DATA since the
"_DONE" version already covers everything and explains the intent well
enough.
The conn_fd_handler used to have one possible call to this function to
notify about end of handshakes, and another one to notify about connection
setup or error. But given that we're now only performing wakeup calls
after connection validation, we don't need to keep two places to run
this test since the conditions do not change in between.
This patch merges the two tests into a single one and moves the
CO_FL_CONNECTED test appropriately as well so that it's called even
on the error path if needed.
In conn_fd_handler() we used to first give a chance to the send()
callback to try to send data and validate the connection at the same
time. But since 1.9 we do not call this callback anymore inline, it's
scheduled. So let's validate the connection ealier so that all other
decisions can be taken based on this confirmation. This may notably
be useful to the xprt_done_cb() to know that the connection was
properly validated.
The function is not TCP-specific at all, it covers all FD-based sockets
so let's move this where other similar functions are, in connection.c,
and rename it conn_fd_check().
