MINOR: listener: define callback for accept queue push
Extend API for connection thread rebind API by replacing single callback
set_affinity by three different ones. Each one of them is used at a
different stage of the operation :
* set_affinity1 is used similarly to previous set_affinity
* set_affinity2 is called directly from accept_queue_push_mp() when an
entry has been found in accept ring. This operation cannot fail.
* reset_affinity is called after set_affinity1 in case of failure from
accept_queue_push_mp() due to no space left in accept ring. This is
necessary for protocols which must reconfigure resources before
fallback on the current tid.
This patch does not have any functional changes. However, it will be
required to fix crashes for QUIC connections when accept queue ring is
full. As such, it must be backported with it.
Previous commit removed access/manipulation to QUIC CID global tree
outside of quic_cid module. This ensures that proper locking is always
performed.
This commit finalizes this cleanup by marking CID global tree as static
only to quic_cid source file. Initialization of this tree is removed
from proto_quic and now performed using dedicated initcalls
quic_alloc_global_cid_tree().
As a side change, complete CID global tree documentation, in particular
to explain CID global tree artificial splitting and ODCID handling.
Overall, the code is now clearer and safer.
This fixes the fd leak, introduced in the commit d3fc982cd7
("MEDIUM: proto: make common fd checks in sock_create_server_socket").
Initially sock_create_server_socket() was designed to return only created
socket FD or -1. Its callers from upper protocol layers were required to test
the returned errno and were required then to apply different configuration
related checks to obtained positive sock_fd. A lot of this code was duplicated
among protocols implementations.
The new refactored version of sock_create_server_socket() gathers in one place
all duplicated checks, but in order to be complient with upper protocol
layers, it needs the 3rd parameter: 'stream_err', in which it sets the
Stream Error Flag for upper levels, if the obtained sock_fd has passed all
additional checks.
No backport needed since this was introduced in 3.0-dev10.
quic_connect_server(), tcp_connect_server(), uxst_connect_server() duplicate
same code to check different ERRNOs, that socket() and setns() may return.
They also duplicate some runtime condition checks, applied to the obtained
server socket fd.
So, in order to remove these duplications and to improve code readability,
let's encapsulate socket() and setns() ERRNOs handling in
sock_handle_system_err(). It must be called just before fd's runtime condition
checks, which we also move in sock_create_server_socket by the same reason.
tune.rcvbuf.client and tune.rcvbuf.server are not suitable for shared
dgram sockets because they're per connection so their units are not the
same. However, QUIC's listener and log servers are not connected and
take per-thread or per-process traffic where a socket log buffer might
be too small, causing undesirable packet losses and retransmits in the
case of QUIC. This essentially manifests in listener mode with new
connections taking a lot of time to set up under heavy traffic due to
the small queues causing delays. Let's add a few new settings allowing
to set these shared socket sizes on the frontend and backend side (which
reminds that these are per-front/back and not per client/server hence
not per connection).
Improve EACCES permission errors encounterd when using QUIC connection
socket at runtime :
* First occurence of the error on the process will generate a log
warning. This should prevent users from using a privileged port
without mandatory access rights.
* Socket mode will automatically fallback to listener socket for the
receiver instance. This requires to duplicate the settings from the
bind_conf to the receiver instance to support configurations with
multiple addresses on the same bind line.
Some protocol support SO_REUSEPORT and others not. Some have such a
limitation in the kernel, and others in haproxy itself (e.g. sock_unix
cannot support multiple bindings since each one will unbind the previous
one). Also it's really protocol-dependent and not just family-dependent
because on Linux for some time it was supported for TCP and not UDP.
Let's move the definition to the protocols instead. Now it's preset in
tcp/udp/quic when SO_REUSEPORT is defined, and is otherwise left unset.
The enabled() config condition test validates IPv4 (generally sufficient),
and -dR / noreuseport all protocols at once.
When migrating a quic_conn to another thread, we may need to also
switch the listener if the thread belongs to another group. When
this happens, the freshly created connection will already have the
target listener, so let's just pick it from the connection and use
it in qc_set_tid_affinity(). Note that it will be the caller's
responsibility to guarantee this.
Before this patch, QUIC protocol used a custom add_listener callback.
This was because a quic_conn instance was allocated before accept. Its
thread affinity was fixed and could not be changed after. The thread was
derived itself from the CID selected by the client which prevent an even
repartition of QUIC connections on multiple threads.
A series of patches was introduced with a lot of changes. The most
important ones :
* removal of affinity between an encoded CID and a thread
* possibility to rebind a quic_conn on a new thread
Thanks to this, it's possible to suppress the custom add_listener
callback. Accept is conducted for QUIC protocol as with the others. A
less loaded thread is selected on listener_accept() and the connection
stack is bind on it. This operation implies that quic_conn instance is
moved to the new thread using the set_affinity QUIC protocol callback.
To reactivate quic_conn instance after thread rebind,
qc_finalize_affinity_rebind() is called after accept on the new thread
by qc_xprt_start() through accept_queue_process() / session_accept_fd().
This should be backported up to 2.7 after a period of observation.
Previously, quic_connection_id were stored in a per-thread tree list.
Datagram were first dispatched to the correct thread using the encoded
TID before a tree lookup was done.
Remove these trees and replace it with a global trees list of 256
entries. A CID is using the list index corresponding to its first byte.
On datagram dispatch, CID is lookup on its tree and TID is retrieved
using new member quic_connection_id.tid. As such, a read-write lock
protects each list instances. With 256 entries, it is expected that
contention should be reduced.
A new structure quic_cid_tree served as a tree container associated with
its read-write lock. An API is implemented to ensure lock safety for
insert/lookup/delete operation.
This patch is a step forward to be able to break the affinity between a
CID and a TID encoded thread. This is required to be able to migrate a
quic_conn after accept to select thread based on their load.
This should be backported up to 2.7 after a period of observation.
First connection CID generation has been altered. It is now directly
derived from client ODCID since previous commit :
commit 162baaff7a
MINOR: quic: derive first DCID from client ODCID
This patch removes the ODCID tree which is now unneeded. On connection
lookup via CID, if a DCID is not found the hash derivation is performed
for an INITIAL/0-RTT packet only. In case a client has used multiple
times an ODCID, this will allow to retrieve our generated DCID in the
CID tree without storing the ODCID node.
The impact of this two combined patch is that it may improve slightly
haproxy memory footprint by removing a tree node from quic_conn
structure. The cpu calculation induced by hash derivation should only be
performed only a few times per connection as the client will start to
use our generated CID as soon as it received it.
This should be backported up to 2.7.
This flag is only used to tag a QUIC listener, which we now know by
its bind_conf's xprt as well. It's only used to decide whether or not
to perform an extra initialization step on the listener. Let's drop it
as well as the flags field.
With the various fields and options moved, the listener struct reduced
by 48 bytes total.
When QUIC thread binding was fixed by commit f5a0c8abf ("MEDIUM: quic:
respect the threads assigned to a bind line"), one point was overlooked
regarding rxbuf allocation. Indeed, there's one rxbuf per listener and
per bound thread. Originally the loop would iterate over all threads,
but this is not needed anymore and causes lots of memory to be allocated
in scenarios where shards are used, the worst one being "shards by-thread"
which allocates N^2 buffers for N threads. This gives us 2304 buffers
(or 576 MB of RAM) for 48 threads!
Let's only allocate one buffer per bound thread on each listener to fix
this.
This should be backported to 2.7 and generally wherever the commit
above is backported. It depends on the previous commit below:
"BUG/MEDIUM: quic: properly take shards into account on bind lines"
A startup check is done for first QUIC listener to detect if quic-conn
owned socket is supported by the system. This is done by creating a
dummy socket reusing the listener address. This socket must be closed as
soon as the check is done.
The socket condition is invalid as it excludes zero which is a valid
file-descriptor value. Fix this bug by adjusting this condition.
In theory, this bug could prevent the usage of quic-conn owned socket as
startup check would report a false error. Also, the file-descriptor
would leak as it is not closed. In practice, this cannot happen when
startup check is done after a 'quic4/quic6' listener is instantiated as
file-descriptor are allocated in ascending order by the system.
This should fix github issue #1954.
quic-conn owned socket implementation is scheduled for backport on 2.7.
This commit must be backported with it, more specifically to fix the
following patch :
75839a44e7
MINOR: quic: startup detect for quic-conn owned socket support
Try to use the quic-conn socket for reception if it is allocated. For
this, the socket is inserted in the fdtab. This will call the new
handler quic_conn_io_cb() which is responsible to process the recv()
system call. It will reuse datagram dispatch for simplicity. However,
this is guaranteed to be called on the quic-conn thread, so it will be
more efficient to use a dedicated buffer. This will be implemented in
another commit.
This patch should improve performance by reducing contention on the
receiver socket. However, more gain can be obtained when the datagram
dispatch operation will be skipped.
Older quic_sock_fd_iocb() is renamed to quic_lstnr_sock_fd_iocb() to
emphasize its usage for the receiver socket.
This change is part of quic-conn owned socket implementation.
It may be backported to 2.7 after a period of observation.
Extend the startup platform detection support test for quic-conn owned
socket. It is required to be able to retrieve destination address on a
recvfrom() system call so check if IP_PKTINFO or IP_RECVDSTADDR flags
are supported.
This change is part of quic-conn owned socket implementation.
It may be backported to 2.7 after a period of observation.
To be able to use individual sockets for QUIC connections, we rely on
the OS network stack which must support UDP sockets binding on the same
local address.
Add a detection code for this feature executed on startup. When the
first QUIC listener socket is binded, a test socket is created and
binded on the same address. If the bind call fails, we consider that
it's impossible to use individual socket for QUIC connections.
A new global option GTUNE_QUIC_SOCK_PER_CONN is defined. If startup
detect fails, this value is resetted from global options. For the
moment, there is no code to activate the option : this will be in a
follow-up patch with the introduction of a new configuration option.
This change is part of quic-conn owned socket implementation.
It may be backported to 2.7 after a period of observation.
QUIC datagrams are read from a random thread. They are then redispatch
to the connection thread according to the first packet DCID. These
operations are implemented through a special buffer designed to avoid
locking.
Refactor this code with the following changes :
* <rxbuf> type is renamed <quic_receiver_buf>. Its list element is also
renamed to highligh its attach point to a receiver.
* <quic_dgram> and <quic_receiver_buf> definition are moved to
quic_sock-t.h. This helps to reduce the size of quic_conn-t.h.
* <quic_dgram> list elements are renamed to highlight their attach point
into a <quic_receiver_buf> and a <quic_dghdlr>.
This should be backported up to 2.6.
rxbuf is the structure used to store QUIC datagrams and redispatch them
to the connection thread.
Each receiver manages a list of rxbuf. This was stored both as an array
and a mt_list. Currently, only mt_list is needed so removed <rxbufs>
member from receiver structure.
This should be backported up to 2.6.
Move code which activates IP_PKTINFO socket option (or affiliated
options) from sock_inet_bind_receiver() to quic_bind_listener()
function. This change is useful for two reasons :
* first, and the most important one : this activates IP_PKTINFO only for
QUIC receivers. The previous version impacted all datagram receivers,
used for example by log-forwarder. This should reduce memory usage for
these datagram sockets which do not need this option.
* second, USE_QUIC preprocessor statements are removed from
src/sock_inet.c which clean up the code.
IP_PKTINFO was introduced recently by the following patch :
97ecc7a8ea (quic-dev/qns)
MEDIUM: quic: retrieve frontend destination address
For the moment, this does not impact any stable release. However, as
previous patch is scheduled for 2.6 backporting, the current change must
also be backported to the same versions.
The tx_qrings[] and tx_qring_list in the receiver are not used
anymore since commit f2476053f ("MINOR: quic: replace custom buf on Tx
by default struct buffer"), the only place where they're referenced
was in quic_alloc_tx_rings_listener(), which by the way implies that
these were not even freed on exit.
Let's just remove them. This should be backported to 2.6 since the
commit above also was.
xprt_quic module was too large and did not reflect the true architecture
by contrast to the other protocols in haproxy.
Extract code related to XPRT layer and keep it under xprt_quic module.
This code should only contains a simple API to communicate between QUIC
lower layer and connection/MUX.
The vast majority of the code has been moved into a new module named
quic_conn. This module is responsible to the implementation of QUIC
lower layer. Conceptually, it overlaps with TCP kernel implementation
when comparing QUIC and HTTP1/2 stacks of haproxy.
This should be backported up to 2.6.
Clean up quic sources by adjusting headers list included depending
on the actual dependency of each source file.
On some occasion, xprt_quic.h was removed from included list. This is
useful to help reducing the dependency on this single file and cleaning
up QUIC haproxy architecture.
This should be backported up to 2.6.
A curious practise seems to have started long ago and contaminated various
code areas, consisting in appending "_pool" at the end of the name of a
given pool. That makes no sense as the name is only used to name the pool
in diags such as "show pools", and since names are truncated there, this
adds some confusion when analysing the dump outputs. Let's just clean all
of them at once. there were essentially in SSL and QUIC.
There's been some great confusion between proto_type, ctrl_type and
sock_type. It turns out that ctrl_type was improperly chosen because
it's not the control layer that is of this or that type, but the
transport layer, and it turns out that the transport layer doesn't
(normally) denaturate the underlying control layer, except for QUIC
which turns dgrams to streams. The fact that the SOCK_{DGRAM|STREAM}
set of values was used added to the confusion.
Let's replace it with xprt_type which reuses the later introduced
PROTO_TYPE_* values, and update the comments to explain which one
works at what level.
Just like for the conn_stream, now that these addresses are dynamically
allocated, there is no single case where the pointer is set without the
corresponding flag, and the flag is used as a permission to dereference
the pointer. Let's just replace the test of the flag with a test of the
pointer and remove all flag assignment. This makes the code clearer
(especially in "if" conditions) and saves the need for future code to
think about properly setting the flag after setting the pointer.
Some of the protocol-level ->connect() functions currently dereference
the connection's destination address while others test it and return an
error. There's normally no more non-bogus code path that calls such
functions without a valid destination address on the connection, so
let's unify these functions and just place a BUG_ON() there, and drop
the useless test that's supposed to return an internal error.
This gets rid of most open-coded fcntl() calls, some of which were passed
through DISGUISE() to avoid a useless test. The FD_CLOEXEC was most often
set without preserving previous flags, which could become a problem once
new flags are created. Now this will not happen anymore.
Some older systems may routinely return EWOULDBLOCK for some syscalls
while we tend to check only for EAGAIN nowadays. Modern systems define
EWOULDBLOCK as EAGAIN so that solves it, but on a few older ones (AIX,
VMS etc) both are different, and for portability we'd need to test for
both or we never know if we risk to confuse some status codes with
plain errors.
There were few entries, the most annoying ones are the switch/case
because they require to only add the entry when it differs, but the
other ones are really trivial.
These functions will allow the connection layer to retrieve a quic_conn's
source or destination when possible. The quic_conn holds the peer's address
but not the local one, and the sockets API doesn't always makes that easy
for datagrams. Thus for frontend connection what we're doing here is to
retrieve the listener's address when the destination address is desired.
Now it finally becomes possible to fetch the source and destination using
"src" and "dst", and to pass an incoming connection's endpoints via the
proxy protocol.
As reported by Tim in issue #1428, our sources are clean, there are
just a few files with a few rare non-ASCII chars for the paragraph
symbol, a few typos, or in Fred's name. Given that Fred already uses
the non-accentuated form at other places like on the public list,
let's uniformize all this and make sure the code displays equally
everywhere.
This is the same fixe as for this commit:
"BUILD: tree-wide: avoid warnings caused by redundant checks of obj_types"
Should fix CID 1469649 for GH #1546
Move the QUIC datagram handlers oustide of the receivers. Use a global
handler per-thread which is allocated on post-config. Implement a free
function on process deinit to avoid a memory leak.
The CID trees are no more attached to the listener receiver but to the
underlying datagram handlers (one by thread) which run always on the same thread.
So, any operation on these trees do not require any locking.
As mentionned in the comment, the tx_qrings and rxbufs members of
receiver struct must be pointers to pointers!
Modify the functions responsible of their allocations consequently.
Note that this code could work because sizeof rxbuf and sizeof tx_qrings
are greater than the size of pointer!
quic_dgram_read() parses all the QUIC packets from a UDP datagram. It is the best
candidate to be converted into a task, because is processing data unit is the UDP
datagram received by the QUIC sock i/o handler. If correct, this datagram is
added to the context of a task, quic_lstnr_dghdlr(), a conversion of quic_dgram_read()
into a task. This task pop a datagram from an mt_list and passes it among to
the packet handler (quic_lstnr_pkt_rcv()).
Modify the quic_dgram struct to play the role of the old quic_dgram_ctx struct when
passed to quic_lstnr_pkt_rcv().
Modify the datagram handlers allocation to set their tasks to quic_lstnr_dghdlr().
Add quic_dghdlr new struct do define datagram handler tasks, one by thread.
Allocate them and attach them to the listener receiver part calling
quic_alloc_dghdlrs_listener() newly implemented function.
Do not proceed to direct accept when creating a new quic_conn. Wait for
the QUIC handshake to succeeds to insert the quic_conn in the accept
queue. A tasklet is then woken up to call listener_accept to accept the
quic_conn.
The most important effect is that the connection/mux layers are not
instantiated at the same time as the quic_conn. This forces to delay
some process to be sure that the mux is allocated :
* initialization of mux transport parameters
* installation of the app-ops
Also, the mux instance is not checked now to wake up the quic_conn
tasklet. This is safe because the xprt-quic code is now ready to handle
the absence of the connection/mux layers.
Note that this commit has a deep impact as it changes significantly the
lower QUIC architecture. Most notably, it breaks the 0-RTT feature.
Create a new type quic_accept_queue to handle QUIC connections accept.
A queue will be allocated for each thread. It contains a list of
listeners which contains at least one quic_conn ready to be accepted and
the tasklet to run listener_accept for these listeners.
Mark QUIC listeners with the flag LI_F_QUIC_LISTENER. It is set by the
proto-quic layer on the add listener callback. This allows to override
more clearly the accept callback on quic_session_accept.
QUIC connections are distributed accross threads by xprt-quic according
to their CIDs. As such disable the thread selection in listener_accept
for QUIC listeners.
This prevents connection from migrating to another threads after its
allocation which can results in unexpected side-effects.
Define a special accept cb for QUIC listeners to quic_session_accept().
This operation is conducted during the proto.add callback when creating
listeners.
A special care is now taken care when setting the standard callback
session_accept_fd() to not overwrite if already defined by the proto
layer.
