It's the receiver's FD that's inherited from the parent process, not
the listener's so the flag must move to the receiver so that appropriate
actions can be taken.
In order to split the receiver from the listener, we'll need to know that
a socket is already bound and ready to receive. We used to do that via
tha LI_O_ASSIGNED state but that's not sufficient anymore since the
receiver might not belong to a listener anymore. The new RX_F_BOUND flag
is used for this.
Some socket settings used to be retrieved via the listener and the
bind_conf. Now instead we use the receiver and its settings whenever
appropriate. This will simplify the removal of the dependency on the
listener.
The receiver is the one which depends on the protocol while the listener
relies on the receiver. Let's move the protocol there. Since there's also
a list element to get back to the listener from the proto list, this list
element (proto_list) was moved as well. For now when scanning protos, we
still see listeners which are linked by their rx.proto_list part.
The listening socket is represented by its file descriptor, which is
generic to all receivers and not just listeners, so it must move to
the rx struct.
It's worth noting that in order to extend receivers and listeners to
other protocols such as QUIC, we'll need other handles than file
descriptors here, and that either a union or a cast to uintptr_t
will have to be used. This was not done yet and the field was
preserved under the name "fd" to avoid adding confusion.
There currently is a large inconsistency in how binding parameters are
split between bind_conf and listeners. It happens that for historical
reasons some parameters are available at the listener level but cannot
be configured per-listener but only for a bind_conf, and thus, need to
be replicated. In addition, some of the bind_conf parameters are in fact
for the listening socket itself while others are for the instanciated
sockets.
A previous attempt at splitting listeners into receivers failed because
the boundary between all these settings is not well defined.
This patch introduces a level of listening socket settings in the
bind_conf, that will be detachable later. Such settings that are solely
for the listening socket are:
- unix socket permissions (used only during binding)
- interface (used for binding)
- network namespace (used for binding)
- process mask and thread mask (used during startup)
The rest seems to be used only to initialize the resulting sockets, or
to control the accept rate. For now, only the unix params (bind_conf->ux)
were moved there.
This is essentially a merge from tcp_find_compatible_fd() and
uxst_find_compatible_fd() that relies on a listener's address and
compare function and still checks for other variations. For AF_INET6
it compares a few of the listener's bind options. A minor change for
UNIX sockets is that transparent mode, interface and namespace used
to be ignored when trying to pick a previous socket while now if they
are changed, the socket will not be reused. This could be refined but
it's still better this way as there is no more risk of using a
differently bound socket by accident.
Eventually we should not pass a listener there but a set of binding
parameters (address, interface, namespace etc...) which ultimately will
be grouped into a receiver. For now this still doesn't exist so let's
stick to the listener to break dependencies in the rest of the code.
The new addrcmp() protocol member points to the function to be used to
compare two addresses of the same family.
When picking an FD from a previous process, we can now use the address
specific address comparison functions instead of having to rely on a
local implementation. This will help move that code to a more central
place.
The new file sock.c will contain generic code for standard sockets
relying on file descriptors. We currently have way too much duplication
between proto_uxst, proto_tcp, proto_sockpair and proto_udp.
For now only get_src, get_dst and sock_create_server_socket were moved,
and are used where appropriate.
Let's finish the cleanup and get rid of all bind and server keywords
parsers from proto_uxst.c. They're now moved to cfgparse-unix.c. Now
proto_uxst.c is clean and only contains code related to binding and
connecting.
This new flag will be used to mark FDs that must be passed to any future
process across the CLI's "_getsocks" command.
The scheme here is quite complex and full of special cases:
- FDs inherited from parent processes are *not* exported this way, as
they are supposed to instead be passed by the master process itself
across reloads. However such FDs ought never to be paused otherwise
this would disrupt the socket in the parent process as well;
- FDs resulting from a "bind" performed over a socket pair, which are
in fact one side of a socket pair passed inside another control socket
pair must not be passed either. Since all of them are used the same
way, for now it's enough never to put this "exported" flag to FDs
bound by the socketpair code.
- FDs belonging to temporary listeners (e.g. a passive FTP data port)
must not be passed either. Fortunately we don't have such FDs yet.
- the rest of the listeners for now are made of TCP, UNIX stream, ABNS
sockets and are exportable, so they get the flag.
- UDP listeners were wrongly created as listeners and are not suitable
here. Their FDs should be passed but for now they are not since the
client doesn't even distinguish the SO_TYPE of the retrieved sockets.
In addition, it's important to keep in mind that:
- inherited FDs may never be closed in master process but may be closed
in worker processes if the service is shut down (useless since still
bound, but technically possible) ;
- inherited FDs may not be disabled ;
- exported FDs may be disabled because the caller will perform the
subsequent listen() on them. However that might not work for all OSes
- exported FDs may be closed, it just means the service was shut down
from the worker, and will be rebound in the new process. This implies
that we have to disable exported on close().
=> as such, contrary to an apparently obvious equivalence, the "exported"
status doesn't imply anything regarding the ability to close a
listener's FD or not.
When a connect() doesn't immediately succeed (i.e. most of the times),
fd_cant_send() is called to enable polling. But given that we don't
mark that we cannot receive either, we end up performing a failed
recvfrom() immediately when the connect() is finally confirmed, as
indicated in issue #253.
This patch simply adds fd_cant_recv() as well so that we're only
notified once the recv path is ready. The reason it was not there
is purely historic, as in the past when there was the fd cache,
doing it would have caused a pending recv request to be placed into
the fd cache, hence a useless recvfrom() upon success (i.e. what
happens now).
Without this patch, forwarding 100k connections does this:
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
17.51 0.704229 7 100000 100000 connect
16.75 0.673875 3 200000 sendto
16.24 0.653222 3 200036 close
10.82 0.435082 1 300000 100000 recvfrom
10.37 0.417266 1 300012 setsockopt
7.12 0.286511 1 199954 epoll_ctl
6.80 0.273447 2 100000 shutdown
5.34 0.214942 2 100005 socket
4.65 0.187137 1 105002 5002 accept4
3.35 0.134757 1 100004 fcntl
0.61 0.024585 4 5858 epoll_wait
With the patch:
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
18.04 0.697365 6 100000 100000 connect
17.40 0.672471 3 200000 sendto
17.03 0.658134 3 200036 close
10.57 0.408459 1 300012 setsockopt
7.69 0.297270 1 200000 recvfrom
7.32 0.282934 1 199922 epoll_ctl
7.09 0.274027 2 100000 shutdown
5.59 0.216041 2 100005 socket
4.87 0.188352 1 104697 4697 accept4
3.35 0.129641 1 100004 fcntl
0.65 0.024959 4 5337 1 epoll_wait
Note the total disappearance of 1/3 of failed recvfrom() *without*
adding any extra syscall anywhere else.
The trace of an HTTP health check is now totally clean, with no useless
syscall at all anymore:
09:14:21.959255 connect(9, {sa_family=AF_INET, sin_port=htons(8000), sin_addr=inet_addr("127.0.0.1")}, 16) = -1 EINPROGRESS (Operation now in progress)
09:14:21.959292 epoll_ctl(4, EPOLL_CTL_ADD, 9, {EPOLLIN|EPOLLOUT|EPOLLRDHUP, {u32=9, u64=9}}) = 0
09:14:21.959315 epoll_wait(4, [{EPOLLOUT, {u32=9, u64=9}}], 200, 1000) = 1
09:14:21.959376 sendto(9, "OPTIONS / HTTP/1.0\r\ncontent-leng"..., 41, MSG_DONTWAIT|MSG_NOSIGNAL, NULL, 0) = 41
09:14:21.959436 epoll_wait(4, [{EPOLLOUT, {u32=9, u64=9}}], 200, 1000) = 1
09:14:21.959456 epoll_ctl(4, EPOLL_CTL_MOD, 9, {EPOLLIN|EPOLLRDHUP, {u32=9, u64=9}}) = 0
09:14:21.959512 epoll_wait(4, [{EPOLLIN|EPOLLRDHUP, {u32=9, u64=9}}], 200, 1000) = 1
09:14:21.959548 recvfrom(9, "HTTP/1.0 200\r\nContent-length: 0\r"..., 16320, 0, NULL, NULL) = 126
09:14:21.959570 close(9) = 0
With the edge-triggered poller, it gets even better:
09:29:15.776201 connect(9, {sa_family=AF_INET, sin_port=htons(8000), sin_addr=inet_addr("127.0.0.1")}, 16) = -1 EINPROGRESS (Operation now in progress)
09:29:15.776256 epoll_ctl(4, EPOLL_CTL_ADD, 9, {EPOLLIN|EPOLLOUT|EPOLLRDHUP|EPOLLET, {u32=9, u64=9}}) = 0
09:29:15.776287 epoll_wait(4, [{EPOLLOUT, {u32=9, u64=9}}], 200, 1000) = 1
09:29:15.776320 sendto(9, "OPTIONS / HTTP/1.0\r\ncontent-leng"..., 41, MSG_DONTWAIT|MSG_NOSIGNAL, NULL, 0) = 41
09:29:15.776374 epoll_wait(4, [{EPOLLIN|EPOLLOUT|EPOLLRDHUP, {u32=9, u64=9}}], 200, 1000) = 1
09:29:15.776406 recvfrom(9, "HTTP/1.0 200\r\nContent-length: 0\r"..., 16320, 0, NULL, NULL) = 126
09:29:15.776434 close(9) = 0
It could make sense to backport this patch to 2.2 and maybe 2.1 after
it has been sufficiently checked for absence of side effects in 2.3-dev,
as some people had reported an extra overhead like in issue #168.
When building with gcc-8 -fsanitize=address, we get this warning once on
an strncpy() call in proto_uxst.c:
src/proto_uxst.c:262:3: warning: 'strncpy' output may be truncated copying 107 bytes from a string of length 4095 [-Wstringop-truncation]
strncpy(addr.sun_path, tempname, sizeof(addr.sun_path) - 1);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
It happens despite the test on snprintf() at the top (since gcc's string
handling is totally empiric), and requires the strlen() test to be placed
"very close" to the strncpy() call (with "very close" yet to be determined).
There's no other way to shut this one except disabling it. Given there's
only one instance of this warning and the cost of dealing with it in the
code is not huge, let's decorate the code to make gcc happily believe it
is smart since it seems to have a mind of itself.
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.
The current state of the logging is a real mess. The main problem is
that almost all files include log.h just in order to have access to
the alert/warning functions like ha_alert() etc, and don't care about
logs. But log.h also deals with real logging as well as log-format and
depends on stream.h and various other things. As such it forces a few
heavy files like stream.h to be loaded early and to hide missing
dependencies depending where it's loaded. Among the missing ones is
syslog.h which was often automatically included resulting in no less
than 3 users missing it.
Among 76 users, only 5 could be removed, and probably 70 don't need the
full set of dependencies.
A good approach would consist in splitting that file in 3 parts:
- one for error output ("errors" ?).
- one for log_format processing
- and one for actual logging.
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.
The TASK_IS_TASKLET() macro was moved to the proto file instead of the
type one. The proto part was a bit reordered to remove a number of ugly
forward declaration of static inline functions. About a tens of C and H
files had their dependency dropped since they were not using anything
from task.h.
global.h was one of the messiest files, it has accumulated tons of
implicit dependencies and declares many globals that make almost all
other file include it. It managed to silence a dependency loop between
server.h and proxy.h by being well placed to pre-define the required
structs, forcing struct proxy and struct server to be forward-declared
in a significant number of files.
It was split in to, one which is the global struct definition and the
few macros and flags, and the rest containing the functions prototypes.
The UNIX_MAX_PATH definition was moved to compat.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.
And also rename standard.c to tools.c. The original split between
tools.h and standard.h dates from version 1.3-dev and was mostly an
accident. This patch moves the files back to what they were expected
to be, and takes care of not changing anything else. However this
time tools.h was split between functions and types, because it contains
a small number of commonly used macros and structures (e.g. name_desc)
which in turn cause the massive list of includes of tools.h to conflict
with the callers.
They remain the ugliest files of the whole project and definitely need
to be cleaned and split apart. A few types are defined there only for
functions provided there, and some parts are even OS-specific and should
move somewhere else, such as the symbol resolution code.
The protocol.h files are pretty low in the dependency and (sadly) used
by some files from common/. Almost nothing was changed except lifting a
few comments.
This one is included almost everywhere and used to rely on a few other
.h that are not needed (unistd, stdlib, standard.h). It could possibly
make sense to split it into multiple parts to distinguish operations
performed on timers and the internal time accounting, but at this point
it does not appear much important.
Half of the users of this include only need the type definitions and
not the manipulation macros nor the inline functions. Moves the various
types into mini-clist-t.h makes the files cleaner. The other one had all
its includes grouped at the top. A few files continued to reference it
without using it and were cleaned.
In addition it was about time that we'd rename that file, it's not
"mini" anymore and contains a bit more than just circular lists.
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.
Just like previous commit, we don't need to pass through the connection
layer anymore to enable polling during a connect(), we know the FD, so
let's simply call fd_want_send().
When intializing a listener, let's make sure the bind_thread mask is
always limited to all_threads_mask when inserting the FD. This will
avoid seeing listening FDs with bits corresponding to threads that are
not active (e.g. when using "bind ... process 1/even"). The side effect
is very limited, all that was identified is that atomic operations are
used in fd_update_events() when not necessary. It's more a matter of
long-term correctness in practice.
This fix might be backported as far as 1.8 (then proto_sockpair must
be dropped).
We do have some checks for the UNIX socket path length to validate the
full pathname of a unix socket but the pathname extension is only taken
into account when using a bind_prefix. The second check only matches
against MAXPATHLEN. So this means that path names between 98 and 108
might successfully parse but fail to bind. Let's adjust the check in
the address parser and refine the error checking at the bind() step.
This addresses bug #493.
Apparently gcc developers decided that strncpy() semantics are no longer
valid and now deserve a warning, especially if used exactly as designed.
This results in issue #304. Let's just remove one to the target size to
please her majesty gcc, the God of C Compilers, who tries hard to make
users completely eliminate any use of string.h and reimplement it by
themselves at much higher risks. Pfff....
This can be backported to stable version, the fix is harmless since it
ignores the last zero that is already set on next line.
In order to address the absurd polling sequence described in issue #253,
let's make sure we disable receiving on a connection until it's established.
Previously with bottom-top I/Os, we were almost certain that a connection
was ready when the first I/O was confirmed. Now we can enter various
functions, including process_stream(), which will attempt to read
something, will fail, and will then subscribe. But we don't want them
to try to receive if we know the connection didn't complete. The first
prerequisite for this is to mark the connection as not ready for receiving
until it's validated. But we don't want to mark it as not ready for sending
because we know that attempting I/Os later is extremely likely to work
without polling.
Once the connection is confirmed we re-enable recv readiness. In order
for this event to be taken into account, the call to tcp_connect_probe()
was moved earlier, between the attempt to send() and the attempt to recv().
This way if tcp_connect_probe() enables reading, we have a chance to
immediately fall back to this and read the possibly pending data.
Now the trace looks like the following. It's far from being perfect
but we've already saved one recvfrom() and one epollctl():
epoll_wait(3, [], 200, 0) = 0
socket(AF_INET, SOCK_STREAM, IPPROTO_TCP) = 7
fcntl(7, F_SETFL, O_RDONLY|O_NONBLOCK) = 0
setsockopt(7, SOL_TCP, TCP_NODELAY, [1], 4) = 0
connect(7, {sa_family=AF_INET, sin_port=htons(8000), sin_addr=inet_addr("127.0.0.1")}, 16) = -1 EINPROGRESS (Operation now in progress)
epoll_ctl(3, EPOLL_CTL_ADD, 7, {EPOLLIN|EPOLLOUT|EPOLLRDHUP, {u32=7, u64=7}}) = 0
epoll_wait(3, [{EPOLLOUT, {u32=7, u64=7}}], 200, 1000) = 1
connect(7, {sa_family=AF_INET, sin_port=htons(8000), sin_addr=inet_addr("127.0.0.1")}, 16) = 0
getsockopt(7, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
sendto(7, "OPTIONS / HTTP/1.0\r\n\r\n", 22, MSG_DONTWAIT|MSG_NOSIGNAL, NULL, 0) = 22
epoll_ctl(3, EPOLL_CTL_MOD, 7, {EPOLLIN|EPOLLRDHUP, {u32=7, u64=7}}) = 0
epoll_wait(3, [{EPOLLIN|EPOLLRDHUP, {u32=7, u64=7}}], 200, 1000) = 1
getsockopt(7, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
getsockopt(7, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
recvfrom(7, "HTTP/1.0 200\r\nContent-length: 0\r\nX-req: size=22, time=0 ms\r\nX-rsp: id=dummy, code=200, cache=1, size=0, time=0 ms (0 real)\r\n\r\n", 16384, 0, NULL, NULL) = 126
close(7) = 0
Dragan Dosen found that the listeners lock is not sufficient to protect
the listeners list when proxies are stopping because the listeners are
also unlinked from the protocol list, and under certain situations like
bombing with soft-stop signals or shutting down many frontends in parallel
from multiple CLI connections, it could be possible to provoke multiple
instances of delete_listener() to be called in parallel for different
listeners, thus corrupting the protocol lists.
Such operations are pretty rare, they are performed once per proxy upon
startup and once per proxy on shut down. Thus there is no point trying
to optimize anything and we can use a global lock to protect the protocol
lists during these manipulations.
This fix (or a variant) will have to be backported as far as 1.8.
Now that the various handshakes come with their own XPRT, there's no
need for the CONN_FL_SOCK* flags, and the conn_sock_want|stop functions,
so garbage-collect them.
The connect() method had 2 arguments, "data", that tells if there's pending
data to be sent, and "delack" that tells if we have to use a delayed ack
inconditionally, or if the backend is configured with tcp-smart-connect.
Turn that into one argument, "flags".
That way it'll be easier to provide more informations to connect() without
adding extra arguments.
In an attempt to try to provide automatic maxconn settings, we need to
decorrelate a listner's backlog and maxconn so that these values can be
independent. This introduces a listener_backlog() function which retrieves
the backlog value from the listener's backlog, the frontend's, the
listener's maxconn, the frontend's or falls back to 1024. This
corresponds to what was done in cfgparse.c to force a value there except
the last fallback which was not set since the frontend's maxconn is always
known.
Now that nbproc and nbthread are exclusive, we can still provide more
detailed explanations about what we've found in the config when a bind
line appears on multiple threads and processes at the same time, then
ignore the setting.
This patch reduces the listener's thread mask to a single mask instead
of an array of masks per process. Now we have only one thread mask and
one process mask per bind-conf. This removes ~504 bytes of RAM per
bind-conf and will simplify handling of thread masks.
If a "bind" line only refers to process numbers not found by its parent
frontend or not covered by the global nbproc directive, or to a thread
not covered by the global nbthread directive, a warning is emitted saying
what will be used instead.
If the master was reloaded and there was a established connection to a
server, the FD resulting from the accept was leaking.
There was no CLOEXEC flag set on the FD of the socketpair created during
a connect call. This is specific to the socketpair in the master process
but it should be applied to every protocol in case we use them in the
master at some point.
No backport needed.
This switches explicit calls to various trivial registration methods for
keywords, muxes or protocols from constructors to INITCALL1 at stage
STG_REGISTER. All these calls have in common to consume a single pointer
and return void. Doing this removes 26 constructors. The following calls
were addressed :
- acl_register_keywords
- bind_register_keywords
- cfg_register_keywords
- cli_register_kw
- flt_register_keywords
- http_req_keywords_register
- http_res_keywords_register
- protocol_register
- register_mux_proto
- sample_register_convs
- sample_register_fetches
- srv_register_keywords
- tcp_req_conn_keywords_register
- tcp_req_cont_keywords_register
- tcp_req_sess_keywords_register
- tcp_res_cont_keywords_register
- flt_register_keywords
In the various connect_server() functions, don't reset the connection flags,
as some may have been set before. The flags are initialized in conn_init(),
anyway.
This patch improves the previous fix by implementing the socket draining
code directly in conn_sock_drain() so that it always applies regardless
of the protocol's family. Thus it gets rid of tcp_drain().
Right now conn_sock_drain() calls the protocol's ->drain() function if
it exists, otherwise it simply tries to disable polling for receiving
on the connection. This doesn't work well anymore since we've implemented
the muxes in 1.8, and it has a side effect with keep-alive backend
connections established over unix sockets. What happens is that if
during the idle time after a request, a connection reports some data,
si_idle_conn_null_cb() is called, which will call conn_sock_drain().
This one sees there's no drain() on unix sockets and will simply disable
polling for data on the connection. But it doesn't do anything on the
conn_stream. Thus while leaving the conn_fd_handler, the mux's polling
is updated and recomputed based on the conn_stream's polling state,
which is still enabled, and nothing changes, so we see the process
use 100% CPU in this case because the FD remains active in the cache.
There are several issues that need to be addressed here. The first and
most important is that we cannot expect some protocols to simply stop
reading data when asked to drain pending data. So this patch make the
unix sockets rely on tcp_drain() since the functions are the same. This
solution is appropriate for backporting, but a better one is desired for
the long term. The second issue is that si_idle_conn_null_cb() shouldn't
drain the connection but the conn_stream.
At the moment we don't have any way to drain a conn_stream, though a flag
on rcv_buf() will do it well. Until we support muxes on the server side
it is not a problem so this part can be addressed later.
This fix must be backported to 1.8.
When checking if a socket we got from the parent is suitable for a listener,
we just checked that the path matched sockname.tmp, however this is
unsuitable for abns sockets, where we don't have to create a temporary
file and rename it later.
To detect that, check that the first character of the sun_path is 0 for
both, and if so, that &sun_path[1] is the same too.
This should be backported to 1.8.
When a listener is temporarily disabled, we start by locking it and then we call
.pause callback of the underlying protocol (tcp/unix). For TCP listeners, this
is not a problem. But listeners bound on an unix socket are in fact closed
instead. So .pause callback relies on unbind_listener function to do its job.
Unfortunatly, unbind_listener hold the listener's lock and then call an internal
function to unbind it. So, there is a deadlock here. This happens during a
reload. To fix the problemn, the function do_unbind_listener, which is lockless,
is now exported and is called when a listener bound on an unix socket is
temporarily disabled.
This patch must be backported in 1.8.
When removing the socket from the xfer_sock_list, we want to set
next->prev to prev, not to next->prev, which is useless.
This should be backported to 1.8.
fd_insert() is currently called just after setting the owner and iocb,
but proceeding like this prevents the operation from being atomic and
requires a lock to protect the maxfd computation in another thread from
meeting an incompletely initialized FD and computing a wrong maxfd.
Fortunately for now all fdtab[].owner are set before calling fd_insert(),
and the first lock in fd_insert() enforces a memory barrier so the code
is safe.
This patch moves the initialization of the owner and iocb to fd_insert()
so that the function will be able to properly arrange its operations and
remain safe even when modified to become lockless. There's no other change
beyond the internal API.
The listeners and connectors may complain that process-wide or
system-wide FD limits have been reached and will in this case report
maxfd as the limit. This is wrong in fact since there's no reason for
the whole FD space to be contiguous when the total # of FD is reached.
A better approach would consist in reporting the accurate number of
opened FDs, but this is pointless as what matters here is to give a
hint about what might be wrong. So let's simply report the configured
maxsock, which will generally explain why the process' limits were
reached, which is the most common reason. This removes another
dependency on maxfd.
These flags are not exactly for the data layer, they instead indicate
what is expected from the transport layer. Since we're going to split
the connection between the transport and the data layers to insert a
mux layer, it's important to have a clear idea of what each layer does.
All function conn_data_* used to manipulate these flags were renamed to
conn_xprt_*.
A config containing "stats socket /path/to/socket mode admin" used to
silently start and be unusable (mode 0, level user) because the "mode"
parser doesn't take care of non-digits. Now it properly reports :
[ALERT] 276/144303 (7019) : parsing [ext-check.cfg:4] : 'stats socket' : ''mode' : missing or invalid mode 'admin' (octal integer expected)'
This can probably be backported to 1.7, 1.6 and 1.5, though reporting
parsing errors in very old versions probably isn't a good idea if the
feature was left unused for years.
Since everything is self contained in proto_uxst.c there's no need to
export anything. The same should be done for proto_tcp.c but the file
contains other stuff that's not related to the TCP protocol itself
and which should first be moved somewhere else.
cfgparse has no business directly calling each individual protocol's 'add'
function to create a listener. Now that they're all registered, better
perform a protocol lookup on the family and have a standard ->add method
for all of them.
It's a shame that cfgparse() has to make special cases of each protocol
just to cast the port to the target address family. Let's pass the port
in argument to the function. The unix listener simply ignores it.
Till now connections used to rely exclusively on file descriptors. It
was planned in the past that alternative solutions would be implemented,
leading to member "union t" presenting sock.fd only for now.
With QUIC, the connection will need to continue to exist but will not
rely on a file descriptor but a connection ID.
So this patch introduces a "connection handle" which is either a file
descriptor or a connection ID, to replace the existing "union t". We've
now removed the intermediate "struct sock" which was never used. There
is no functional change at all, though the struct connection was inflated
by 32 bits on 64-bit platforms due to alignment.
James Brown reported some cases where a race condition happens between
the old and the new processes resulting in the leaving process removing
a newly bound unix socket. Jeff gave all the details he observed here :
https://www.mail-archive.com/haproxy@formilux.org/msg25001.html
The unix socket removal was an attempt at an optimal cleanup, which
almost never works anyway since the process is supposed to be chrooted.
And in the rare cases where it works it occasionally creates trouble.
There was already a workaround in place to avoid removing this socket
when it's been inherited from a parent's file descriptor.
So let's finally kill this useless stuff now to definitely get rid of
this persistent problem.
This fix should be backported to all stable releases.
Add a new command that will send all the listening sockets, via the
stats socket, and their properties.
This is a first step to workaround the linux problem when reloading
haproxy.
There's a test after a successful synchronous connect() consisting
in waking the data layer up asap if there's no more handshake.
Unfortunately this test is run before setting the CO_FL_SEND_PROXY
flag and before the transport layer adds its own flags, so it can
indicate a willingness to send data while it's not the case and it
will have to be handled later.
This has no visible effect except a useless call to a function in
case of health checks making use of the proxy protocol for example.
Additionally a corner case where EALREADY was returned and considered
equivalent to EISCONN was fixed so that it's considered equivalent to
EINPROGRESS given that the connection is not complete yet. But this
code should never return on the first call anyway so it's mostly a
cleanup.
This fix should be backported to 1.7 and 1.6 at least to avoid
headaches during some debugging.
When a connect() to a unix socket returns EAGAIN we talk about
"no free ports" in the error/debug message, which only makes
sense when using TCP.
Explain connect() failure and suggest troubleshooting server
backlog size.
Abstract namespace sockets ignore the shutdown() call and do not make
it possible to temporarily stop listening. The issue it causes is that
during a soft reload, the new process cannot bind, complaining that the
address is already in use.
This change registers a new pause() function for unix sockets and
completely unbinds the abstract ones since it's possible to rebind
them later. It requires the two previous patches as well as preceeding
fixes.
This fix should be backported into 1.5 since the issue apperas there.
Jan Seda noticed that abstract sockets are incompatible with soft reload,
because the new process cannot bind and immediately fails. This patch marks
the binding as retryable and not fatal so that the new process can try to
bind again after sending a signal to the old process.
Note that this fix is not enough to completely solve the problem, but it
is necessary. This patch should be backported to 1.5.
When bind() fails (function uxst_bind_listener()), the fail path doesn't
consider the abstract namespace and tries to unlink paths held in
uninitiliazed memory (tempname and backname). See the strace excerpt;
the strings still hold the path from test1.
===============================================================================================
23722 bind(5, {sa_family=AF_FILE, path=@"test2"}, 110) = -1 EADDRINUSE (Address already in use)
23722 unlink("/tmp/test1.sock.23722.tmp") = -1 ENOENT (No such file or directory)
23722 close(5) = 0
23722 unlink("/tmp/test1.sock.23722.bak") = -1 ENOENT (No such file or directory)
===============================================================================================
This patch should be backported to 1.5.
Plain "tcp" health checks sent to a unix socket cause two connect()
calls to be made, one to connect, and a second one to verify that the
connection properly established. But with unix sockets, we get
immediate notification of success, so we can avoid this second
attempt. However we need to ensure that we'll visit the connection
handler even if there's no remaining handshake pending, so for this
we claim we have some data to send in order to enable polling for
writes if there's no more handshake.
These sockets are the same as Unix sockets except that there's no need
for any filesystem access. The address may be whatever string both sides
agree upon. This can be really convenient for inter-process communications
as well as for chaining backends to frontends.
These addresses are forced by prepending their address with "abns@" for
"abstract namespace".
We've had everything in place for this for a while now, we just missed
the connect function for UNIX sockets. Note that in order to connect to
a UNIX socket inside a chroot, the path will have to be relative to the
chroot.
UNIX sockets connect about twice as fast as TCP sockets (or consume
about half of the CPU at the same rate). This is interesting for
internal communications between SSL processes and HTTP processes
for example, or simply to avoid allocating source ports on the
loopback.
The tcp_connect_probe() function is still used to probe a dataless
connection, but it is compatible so that's not an issue for now.
Health checks are not yet fully supported since they require a port.
Using the address syntax "fd@<num>", a listener may inherit a file
descriptor that the caller process has already bound and passed as
this number. The fd's socket family is detected using getsockname(),
and the usual initialization is performed through the existing code
for that family, but the socket creation is skipped.
Whether the parent has performed the listen() call or not is not
important as this is detected.
For UNIX sockets, we immediately clear the path after preparing a
socket so that we never remove it in case an abort would happen due
to a late error during startup.
Unix permissions are per-bind configuration line and not per listener,
so let's concretize this in the way the config is stored. This avoids
some unneeded loops to set permissions on all listeners.
The access level is not part of the unix perms so it has been moved
away. Once we can use str2listener() to set all listener addresses,
we'll have a bind keyword parser for this one.
Navigating through listeners was very inconvenient and error-prone. Not to
mention that listeners were linked in reverse order and reverted afterwards.
In order to definitely get rid of these issues, we now do the following :
- frontends have a dual-linked list of bind_conf
- frontends have a dual-linked list of listeners
- bind_conf have a dual-linked list of listeners
- listeners have a pointer to their bind_conf
This way we can now navigate from anywhere to anywhere and always find the
proper bind_conf for a given listener, as well as find the list of listeners
for a current bind_conf.
The "mode", "uid", "gid", "user" and "group" bind options were moved to
proto_uxst as they are unix-specific.
Note that previous versions had a bug here, only the last listener was
updated with the specified settings. However, it almost never happens
that bind lines contain multiple UNIX socket paths so this is not that
much of a problem anyway.
The "raw_sock" prefix will be more convenient for naming functions as
it will be prefixed with the data layer and suffixed with the data
direction. So let's rename the files now to avoid any further confusion.
The #include directive was also removed from a number of files which do
not need it anymore.
fdtab[].state was only used to know whether a connection was in progress
or an error was encountered. Instead we now use connection->flags to store
a flag for both. This way, connection management will be able to update the
connection status on I/O.
The destination address is purely a connection thing and not an fd thing.
It's also likely that later the address will be stored into the connection
and linked to by the SI.
struct fdinfo only keeps the pointer to the port range and the local port
for now. All of this also needs to move to the connection but before this
the release of the port range must move from fd_delete() to a new function
dedicated to the connection.
These pointers were used to hold pointers to buffers in the past, but
since we introduced the stream interface, they're no longer used but
they were still sometimes set.
Removing them shrink the struct fdtab from 32 to 24 bytes on 32-bit machines,
and from 52 to 36 bytes on 64-bit machines, which is a significant saving. A
quick tests shows a steady 0.5% performance gain, probably due to the better
cache efficiency.
Commit e164e7a removed get_src/get_dst setting in the stream interfaces but
forgot to set it in proto_tcp. Get the feature back because we need it for
logging, transparent mode, ACLs etc... We now rely on the stream interface
direction to know what syscall to use.
One benefit of doing it this way is that we don't use getsockopt() anymore
on outgoing stream interfaces nor on UNIX sockets.
We'll soon have an SSL socket layer, and in order to ease the difference
between the two, we use the name "sock_raw" to designate the one which
directly talks to the sockets without any conversion.
The previous sockstream_accept() function uses nothing from sockstream, and
is totally irrelevant to stream interfaces. Move this to the protocols.c
file which handles listeners and protocols, and call it listener_accept().
It now makes much more sense that the code dealing with listen() also handles
accept() and passes it to upper layers.
Managing listeners state is difficult because they have their own state
and can at the same time have theirs dictated by their proxy. The pause
is not done properly, as the proxy code is fiddling with sockets. By
introducing new functions such as pause_listener()/resume_listener(), we
make it a bit more obvious how/when they're supposed to be used. The
listen_proxies() function was also renamed to resume_proxies() since
it's only used for pause/resume.
This patch is the first in a series aiming at getting rid of the maintain_proxies
mess. In the end, proxies should not call enable_listener()/disable_listener()
anymore.
Since unix sockets are supported for bind, the default backlog size was not
enough to accept the traffic. The size is now inherited from the listener
to behave like the tcp listeners.
This also affects the "stats socket" backlog, which is now determined by
"stats maxconn".
