API Reference

Preliminaries

All declarations are in lsquic.h, so it is enough to

#incluide <lsquic.h>

in each source file.

Library Version

LSQUIC follows the following versioning model. The version number has the form MAJOR.MINOR.PATCH, where

  • MAJOR changes when a large redesign occurs;
  • MINOR changes when an API change or another significant change occurs; and
  • PATCH changes when a bug is fixed or another small, API-compatible change occurs.

QUIC Versions

LSQUIC supports two types of QUIC protocol: Google QUIC and IETF QUIC. The former will at some point become obsolete, while the latter is still being developed by the IETF. Both types are included in a single enum:

enum lsquic_version
LSQVER_043

Google QUIC version Q043

LSQVER_046

Google QUIC version Q046

LSQVER_050

Google QUIC version Q050

LSQVER_ID27

IETF QUIC version ID (Internet-Draft) 27

LSQVER_ID28

IETF QUIC version ID 28

N_LSQVER

Special value indicating the number of versions in the enum. It may be used as argument to lsquic_engine_connect().

Several version lists (as bitmasks) are defined in lsquic.h:

LSQUIC_SUPPORTED_VERSIONS

List of all supported versions.

LSQUIC_FORCED_TCID0_VERSIONS

List of versions in which the server never includes CID in short packets.

LSQUIC_EXPERIMENTAL_VERSIONS

Experimental versions.

Deprecated versions.

LSQUIC_GQUIC_HEADER_VERSIONS

Versions that have Google QUIC-like headers. Only Q043 remains in this list.

LSQUIC_IETF_VERSIONS

IETF QUIC versions.

LSQUIC_IETF_DRAFT_VERSIONS

IETF QUIC draft versions. When IETF QUIC v1 is released, it will not be included in this list.

LSQUIC Types

LSQUIC declares several types used by many of its public functions. They are:

lsquic_engine_t

Instance of LSQUIC engine.

lsquic_conn_t

QUIC connection.

lsquic_stream_t

QUIC stream.

lsquic_stream_id_t

Stream ID.

lsquic_conn_ctx_t

Connection context. This is the return value of on_new_conn(). To LSQUIC, this is just an opaque pointer. User code is expected to use it for its own purposes.

lsquic_stream_ctx_t

Stream context. This is the return value of on_new_stream(). To LSQUIC, this is just an opaque pointer. User code is expected to use it for its own purposes.

lsquic_http_headers_t

HTTP headers

Library Initialization

Before using the library, internal structures must be initialized using the global initialization function:

if (0 == lsquic_global_init(LSQUIC_GLOBAL_CLIENT|LSQUIC_GLOBAL_SERVER))
    /* OK, do something useful */
    ;

This call only needs to be made once. Afterwards, any number of LSQUIC engines may be instantiated.

After a process is done using LSQUIC, it should clean up:

lsquic_global_cleanup();

Logging

struct lsquic_logger_if
int (*log_buf)(void *logger_ctx, const char *buf, size_t len)
void lsquic_logger_init(const struct lsquic_logger_if *logger_if, void *logger_ctx, enum lsquic_logger_timestamp_style)

Call this if you want to do something with LSQUIC log messages, as they are thrown out by default.

int lsquic_set_log_level(const char *log_level)

Set log level for all LSQUIC modules.

Parameters:
  • log_level – Acceptable values are debug, info, notice, warning, error, alert, emerg, crit (case-insensitive).
Returns:

0 on success or -1 on failure (invalid log level).

int lsquic_logger_lopt(const char *log_specs)

Set log level for a particular module or several modules.

Parameters:
  • log_specs – One or more “module=level” specifications serapated by comma. For example, “event=debug,engine=info”. See List of Log Modules

Engine Instantiation and Destruction

To use the library, an instance of the struct lsquic_engine needs to be created:

lsquic_engine_t *lsquic_engine_new(unsigned flags, const struct lsquic_engine_api *api)

Create a new engine.

Parameters:
  • flags – This is is a bitmask of LSENG_SERVER` and LSENG_HTTP.
  • api – Pointer to an initialized lsquic_engine_api.

The engine can be instantiated either in server mode (when LSENG_SERVER is set) or client mode. If you need both server and client in your program, create two engines (or as many as you’d like).

Specifying LSENG_HTTP flag enables the HTTP functionality: HTTP/2-like for Google QUIC connections and HTTP/3 functionality for IETF QUIC connections.

void lsquic_engine_cooldown(lsquic_engine_t *engine)

This function closes all mini connections and marks all full connections as going away. In server mode, this also causes the engine to stop creating new connections.

void lsquic_engine_destroy(lsquic_engine_t *engine)

Destroy engine and all its resources.

Engine Callbacks

struct lsquic_engine_api contains a few mandatory members and several optional members.

struct lsquic_engine_api
const struct lsquic_stream_if *ea_stream_if
void *ea_stream_if_ctx

ea_stream_if is mandatory. This structure contains pointers to callbacks that handle connections and stream events.

lsquic_packets_out_f ea_packets_out
void *ea_packets_out_ctx

ea_packets_out is used by the engine to send packets.

const struct lsquic_engine_settings *ea_settings

If ea_settings is set to NULL, the engine uses default settings (see lsquic_engine_init_settings())

lsquic_lookup_cert_f ea_lookup_cert
void *ea_cert_lu_ctx

Look up certificate. Mandatory in server mode.

struct ssl_ctx_st * (*ea_get_ssl_ctx)(void *peer_ctx)

Get SSL_CTX associated with a peer context. Mandatory in server mode. This is use for default values for SSL instantiation.

const struct lsquic_hset_if *ea_hsi_if
void *ea_hsi_ctx

Optional header set interface. If not specified, the incoming headers are converted to HTTP/1.x format and are read from stream and have to be parsed again.

const struct lsquic_shared_hash_if *ea_shi
void *ea_shi_ctx

Shared hash interface can be used to share state between several processes of a single QUIC server.

const struct lsquic_packout_mem_if *ea_pmi
void *ea_pmi_ctx

Optional set of functions to manage memory allocation for outgoing packets.

lsquic_cids_update_f ea_new_scids
lsquic_cids_update_f ea_live_scids
lsquic_cids_update_f ea_old_scids
void *ea_cids_update_ctx

In a multi-process setup, it may be useful to observe the CID lifecycle. This optional set of callbacks makes it possible.

Engine Settings

Engine behavior can be controlled by several settings specified in the settings structure:

struct lsquic_engine_settings
unsigned es_versions

This is a bit mask wherein each bit corresponds to a value in lsquic_version. Client starts negotiating with the highest version and goes down. Server supports either of the versions specified here. This setting applies to both Google and IETF QUIC.

The default value is LSQUIC_DF_VERSIONS.

unsigned es_cfcw

Initial default connection flow control window.

In server mode, per-connection values may be set lower than this if resources are scarce.

Do not set es_cfcw and es_sfcw lower than LSQUIC_MIN_FCW.

unsigned es_sfcw

Initial default stream flow control window.

In server mode, per-connection values may be set lower than this if resources are scarce.

Do not set es_cfcw and es_sfcw lower than LSQUIC_MIN_FCW.

unsigned es_max_cfcw

This value is used to specify maximum allowed value connection flow control window is allowed to reach due to window auto-tuning. By default, this value is zero, which means that CFCW is not allowed to increase from its initial value.

unsigned es_max_sfcw

This value is used to specify maximum allowed value stream flow control window is allowed to reach due to window auto-tuning. By default, this value is zero, which means that CFCW is not allowed to increase from its initial value.

unsigned es_max_streams_in

Maximum incoming streams, a.k.a. MIDS.

Google QUIC only.

unsigned long es_handshake_to

Handshake timeout in microseconds.

For client, this can be set to an arbitrary value (zero turns the timeout off).

For server, this value is limited to about 16 seconds. Do not set it to zero.

Defaults to LSQUIC_DF_HANDSHAKE_TO.

unsigned long es_idle_conn_to

Idle connection timeout, a.k.a ICSL, in microseconds; GQUIC only.

Defaults to LSQUIC_DF_IDLE_CONN_TO

int es_silent_close

SCLS (silent close)

unsigned es_max_header_list_size

This corresponds to SETTINGS_MAX_HEADER_LIST_SIZE (RFC 7540#section-6.5.2). 0 means no limit. Defaults to LSQUIC_DF_MAX_HEADER_LIST_SIZE().

const char *es_ua

UAID – User-Agent ID. Defaults to LSQUIC_DF_UA.

Google QUIC only.

More parameters for server

unsigned es_max_inchoate

Maximum number of incoming connections in inchoate state. (In other words, maximum number of mini connections.)

This is only applicable in server mode.

Defaults to LSQUIC_DF_MAX_INCHOATE.

int es_support_push

Setting this value to 0 means that

For client:

  1. we send a SETTINGS frame to indicate that we do not support server push; and
  2. all incoming pushed streams get reset immediately.

(For maximum effect, set es_max_streams_in to 0.)

For server:

  1. lsquic_conn_push_stream() will return -1.
int es_support_tcid0

If set to true value, the server will not include connection ID in outgoing packets if client’s CHLO specifies TCID=0.

For client, this means including TCID=0 into CHLO message. Note that in this case, the engine tracks connections by the (source-addr, dest-addr) tuple, thereby making it necessary to create a socket for each connection.

This option has no effect in Q046, as the server never includes CIDs in the short packets.

The default is LSQUIC_DF_SUPPORT_TCID0().

int es_support_nstp

Q037 and higher support “No STOP_WAITING frame” mode. When set, the client will send NSTP option in its Client Hello message and will not sent STOP_WAITING frames, while ignoring incoming STOP_WAITING frames, if any. Note that if the version negotiation happens to downgrade the client below Q037, this mode will not be used.

This option does not affect the server, as it must support NSTP mode if it was specified by the client.

int es_honor_prst

If set to true value, the library will drop connections when it receives corresponding Public Reset packet. The default is to ignore these packets.

int es_send_prst

If set to true value, the library will send Public Reset packets in response to incoming packets with unknown Connection IDs.

The default is LSQUIC_DF_SEND_PRST.

unsigned es_progress_check

A non-zero value enables internal checks that identify suspected infinite loops in user on_read() and on_write() callbacks and break them. An infinite loop may occur if user code keeps on performing the same operation without checking status, e.g. reading from a closed stream etc.

The value of this parameter is as follows: should a callback return this number of times in a row without making progress (that is, reading, writing, or changing stream state), loop break will occur.

The defaut value is LSQUIC_DF_PROGRESS_CHECK.

int es_rw_once

A non-zero value make stream dispatch its read-write events once per call.

When zero, read and write events are dispatched until the stream is no longer readable or writeable, respectively, or until the user signals unwillingness to read or write using lsquic_stream_wantread() or lsquic_stream_wantwrite() or shuts down the stream.

The default value is LSQUIC_DF_RW_ONCE.

unsigned es_proc_time_thresh

If set, this value specifies that number of microseconds that lsquic_engine_process_conns() and lsquic_engine_send_unsent_packets() are allowed to spend before returning.

This is not an exact science and the connections must make progress, so the deadline is checked after all connections get a chance to tick (in the case of lsquic_engine_process_conns())() and at least one batch of packets is sent out.

When processing function runs out of its time slice, immediate calls to lsquic_engine_has_unsent_packets() return false.

The default value is LSQUIC_DF_PROC_TIME_THRESH().

int es_pace_packets

If set to true, packet pacing is implemented per connection.

The default value is LSQUIC_DF_PACE_PACKETS().

unsigned es_clock_granularity

Clock granularity information is used by the pacer. The value is in microseconds; default is LSQUIC_DF_CLOCK_GRANULARITY().

unsigned es_init_max_data

Initial max data.

This is a transport parameter.

Depending on the engine mode, the default value is either LSQUIC_DF_INIT_MAX_DATA_CLIENT or LSQUIC_DF_INIT_MAX_DATA_SERVER.

IETF QUIC only.

unsigned es_init_max_stream_data_bidi_remote

Initial max stream data.

This is a transport parameter.

Depending on the engine mode, the default value is either LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_REMOTE_CLIENT or LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_REMOTE_SERVER.

IETF QUIC only.

unsigned es_init_max_stream_data_bidi_local

Initial max stream data.

This is a transport parameter.

Depending on the engine mode, the default value is either LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_LOCAL_CLIENT or LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_LOCAL_SERVER.

IETF QUIC only.

unsigned es_init_max_stream_data_uni

Initial max stream data for unidirectional streams initiated by remote endpoint.

This is a transport parameter.

Depending on the engine mode, the default value is either LSQUIC_DF_INIT_MAX_STREAM_DATA_UNI_CLIENT or LSQUIC_DF_INIT_MAX_STREAM_DATA_UNI_SERVER.

IETF QUIC only.

unsigned es_init_max_streams_bidi

Maximum initial number of bidirectional stream.

This is a transport parameter.

Default value is LSQUIC_DF_INIT_MAX_STREAMS_BIDI.

IETF QUIC only.

unsigned es_init_max_streams_uni

Maximum initial number of unidirectional stream.

This is a transport parameter.

Default value is LSQUIC_DF_INIT_MAX_STREAMS_UNI_CLIENT or LSQUIC_DF_INIT_MAX_STREAM_DATA_UNI_SERVER.

IETF QUIC only.

unsigned es_idle_timeout

Idle connection timeout.

This is a transport parameter.

(Note: es_idle_conn_to() is not reused because it is in microseconds, which, I now realize, was not a good choice. Since it will be obsoleted some time after the switchover to IETF QUIC, we do not have to keep on using strange units.)

Default value is LSQUIC_DF_IDLE_TIMEOUT.

Maximum value is 600 seconds.

IETF QUIC only.

unsigned es_ping_period

Ping period. If set to non-zero value, the connection will generate and send PING frames in the absence of other activity.

By default, the server does not send PINGs and the period is set to zero. The client’s defaut value is LSQUIC_DF_PING_PERIOD.

IETF QUIC only.

unsigned es_scid_len

Source Connection ID length. Valid values are 0 through 20, inclusive.

Default value is LSQUIC_DF_SCID_LEN.

IETF QUIC only.

unsigned es_scid_iss_rate

Source Connection ID issuance rate. This field is measured in CIDs per minute. Using value 0 indicates that there is no rate limit for CID issuance.

Default value is LSQUIC_DF_SCID_ISS_RATE.

IETF QUIC only.

unsigned es_qpack_dec_max_size

Maximum size of the QPACK dynamic table that the QPACK decoder will use.

The default is LSQUIC_DF_QPACK_DEC_MAX_SIZE.

IETF QUIC only.

unsigned es_qpack_dec_max_blocked

Maximum number of blocked streams that the QPACK decoder is willing to tolerate.

The default is LSQUIC_DF_QPACK_DEC_MAX_BLOCKED.

IETF QUIC only.

unsigned es_qpack_enc_max_size

Maximum size of the dynamic table that the encoder is willing to use. The actual size of the dynamic table will not exceed the minimum of this value and the value advertized by peer.

The default is LSQUIC_DF_QPACK_ENC_MAX_SIZE.

IETF QUIC only.

unsigned es_qpack_enc_max_blocked

Maximum number of blocked streams that the QPACK encoder is willing to risk. The actual number of blocked streams will not exceed the minimum of this value and the value advertized by peer.

The default is LSQUIC_DF_QPACK_ENC_MAX_BLOCKED.

IETF QUIC only.

int es_ecn

Enable ECN support.

The default is LSQUIC_DF_ECN

IETF QUIC only.

int es_allow_migration

Allow peer to migrate connection.

The default is LSQUIC_DF_ALLOW_MIGRATION

IETF QUIC only.

unsigned es_cc_algo

Congestion control algorithm to use.

  • 0: Use default (LSQUIC_DF_CC_ALGO)
  • 1: Cubic
  • 2: BBR

IETF QUIC only.

int es_ql_bits

Use QL loss bits. Allowed values are:

  • 0: Do not use loss bits
  • 1: Allow loss bits
  • 2: Allow and send loss bits

Default value is LSQUIC_DF_QL_BITS

int es_spin

Enable spin bit. Allowed values are 0 and 1.

Default value is LSQUIC_DF_SPIN

int es_delayed_acks

Enable delayed ACKs extension. Allowed values are 0 and 1.

Warning: this is an experimental feature. Using it will most likely lead to degraded performance.

Default value is LSQUIC_DF_DELAYED_ACKS

int es_timestamps

Enable timestamps extension. Allowed values are 0 and 1.

Default value is @ref LSQUIC_DF_TIMESTAMPS

unsigned short es_max_udp_payload_size_rx

Maximum packet size we are willing to receive. This is sent to peer in transport parameters: the library does not enforce this limit for incoming packets.

If set to zero, limit is not set.

Default value is LSQUIC_DF_MAX_UDP_PAYLOAD_SIZE_RX

unsigned es_noprogress_timeout

No progress timeout.

If connection does not make progress for this number of seconds, the connection is dropped. Here, progress is defined as user streams being written to or read from.

If this value is zero, this timeout is disabled.

Default value is LSQUIC_DF_NOPROGRESS_TIMEOUT_SERVER in server mode and LSQUIC_DF_NOPROGRESS_TIMEOUT_CLIENT in client mode.

To initialize the settings structure to library defaults, use the following convenience function:

lsquic_engine_init_settings(struct lsquic_engine_settings *, unsigned flags)

flags is a bitmask of LSENG_SERVER and LSENG_HTTP

After doing this, change just the settings you’d like. To check whether the values are correct, another convenience function is provided:

lsquic_engine_check_settings(const struct lsquic_engine_settings *, unsigned flags, char *err_buf, size_t err_buf_sz)

Check settings for errors. Return 0 if settings are OK, -1 otherwise.

If err_buf() and err_buf_sz() are set, an error string is written to the buffers.

The following macros in lsquic.h specify default values:

Note that, despite our best efforts, documentation may accidentally get out of date. Please check your :file:`lsquic.h` for actual values.

LSQUIC_MIN_FCW

Minimum flow control window is set to 16 KB for both client and server. This means we can send up to this amount of data before handshake gets completed.

LSQUIC_DF_VERSIONS

By default, deprecated and experimental versions are not included.

LSQUIC_DF_CFCW_SERVER
LSQUIC_DF_CFCW_CLIENT
LSQUIC_DF_SFCW_SERVER
LSQUIC_DF_SFCW_CLIENT
LSQUIC_DF_MAX_STREAMS_IN
LSQUIC_DF_INIT_MAX_DATA_SERVER
LSQUIC_DF_INIT_MAX_DATA_CLIENT
LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_REMOTE_SERVER
LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_LOCAL_SERVER
LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_REMOTE_CLIENT
LSQUIC_DF_INIT_MAX_STREAM_DATA_BIDI_LOCAL_CLIENT
LSQUIC_DF_INIT_MAX_STREAMS_BIDI
LSQUIC_DF_INIT_MAX_STREAMS_UNI_CLIENT
LSQUIC_DF_INIT_MAX_STREAMS_UNI_SERVER
LSQUIC_DF_INIT_MAX_STREAM_DATA_UNI_CLIENT
LSQUIC_DF_INIT_MAX_STREAM_DATA_UNI_SERVER
LSQUIC_DF_IDLE_TIMEOUT

Default idle connection timeout is 30 seconds.

LSQUIC_DF_PING_PERIOD

Default ping period is 15 seconds.

LSQUIC_DF_HANDSHAKE_TO

Default handshake timeout is 10,000,000 microseconds (10 seconds).

LSQUIC_DF_IDLE_CONN_TO

Default idle connection timeout is 30,000,000 microseconds.

LSQUIC_DF_SILENT_CLOSE

By default, connections are closed silenty when they time out (no CONNECTION_CLOSE frame is sent).

LSQUIC_DF_MAX_HEADER_LIST_SIZE

Default value of maximum header list size. If set to non-zero value, SETTINGS_MAX_HEADER_LIST_SIZE will be sent to peer after handshake is completed (assuming the peer supports this setting frame type).

LSQUIC_DF_UA

Default value of UAID (user-agent ID).

LSQUIC_DF_MAX_INCHOATE

Default is 1,000,000.

LSQUIC_DF_SUPPORT_NSTP

NSTP is not used by default.

LSQUIC_DF_SUPPORT_PUSH

Push promises are supported by default.

LSQUIC_DF_SUPPORT_TCID0

Support for TCID=0 is enabled by default.

LSQUIC_DF_HONOR_PRST

By default, LSQUIC ignores Public Reset packets.

LSQUIC_DF_SEND_PRST

By default, LSQUIC will not send Public Reset packets in response to packets that specify unknown connections.

LSQUIC_DF_PROGRESS_CHECK

By default, infinite loop checks are turned on.

LSQUIC_DF_RW_ONCE

By default, read/write events are dispatched in a loop.

LSQUIC_DF_PROC_TIME_THRESH

By default, the threshold is not enabled.

LSQUIC_DF_PACE_PACKETS

By default, packets are paced

LSQUIC_DF_CLOCK_GRANULARITY

Default clock granularity is 1000 microseconds.

LSQUIC_DF_SCID_LEN 8

The default value is 8 for simplicity and speed.

LSQUIC_DF_SCID_ISS_RATE

The default value is 60 CIDs per minute.

LSQUIC_DF_QPACK_DEC_MAX_BLOCKED

Default value is 100.

LSQUIC_DF_QPACK_DEC_MAX_SIZE

Default value is 4,096 bytes.

LSQUIC_DF_QPACK_ENC_MAX_BLOCKED

Default value is 100.

LSQUIC_DF_QPACK_ENC_MAX_SIZE

Default value is 4,096 bytes.

LSQUIC_DF_ECN

ECN is disabled by default.

LSQUIC_DF_ALLOW_MIGRATION

Allow migration by default.

LSQUIC_DF_QL_BITS

Use QL loss bits by default.

LSQUIC_DF_SPIN

Turn spin bit on by default.

LSQUIC_DF_CC_ALGO

Use Cubic by default.

LSQUIC_DF_DELAYED_ACKS

Delayed ACKs are off by default.

LSQUIC_DF_MAX_UDP_PAYLOAD_SIZE_RX

By default, incoming packet size is not limited.

LSQUIC_DF_NOPROGRESS_TIMEOUT_SERVER

By default, drop no-progress connections after 60 seconds on the server.

LSQUIC_DF_NOPROGRESS_TIMEOUT_CLIENT

By default, do not use no-progress timeout on the client.

Receiving Packets

Incoming packets are supplied to the engine using lsquic_engine_packet_in(). It is up to the engine to decide what do to with the packet. It can find an existing connection and dispatch the packet there, create a new connection (in server mode), or schedule a version negotiation or stateless reset packet.

int lsquic_engine_packet_in(lsquic_engine_t *engine, const unsigned char *data, size_t size, const struct sockaddr *local, const struct sockaddr *peer, void *peer_ctx, int ecn)

Pass incoming packet to the QUIC engine. This function can be called more than once in a row. After you add one or more packets, call lsquic_engine_process_conns() to schedule outgoing packets, if any.

Parameters:
  • engine – Engine instance.
  • data – Pointer to UDP datagram payload.
  • size – Size of UDP datagram.
  • local – Local address.
  • peer – Peer address.
  • peer_ctx – Peer context.
  • ecn – ECN marking associated with this UDP datagram.
Returns:

  • 0: Packet was processed by a real connection.
  • 1: Packet was handled successfully, but not by a connection. This may happen with version negotiation and public reset packets as well as some packets that may be ignored.
  • -1: Some error occurred. Possible reasons are invalid packet size or failure to allocate memory.

int lsquic_engine_earliest_adv_tick(lsquic_engine_t *engine, int *diff)

Returns true if there are connections to be processed, false otherwise.

Parameters:
  • engine – Engine instance.
  • diff – If the function returns a true value, the pointed to integer is set to the difference between the earliest advisory tick time and now. If the former is in the past, this difference is negative.
Returns:

True if there are connections to be processed, false otherwise.

Sending Packets

User specifies a callback lsquic_packets_out_f in lsquic_engine_api that the library uses to send packets.

struct lsquic_out_spec

This structure describes an outgoing packet.

struct iovec *iov

A vector with payload.

size_t iovlen

Vector length.

const struct sockaddr *local_sa

Local address.

const struct sockaddr *dest_sa

Destination address.

void *peer_ctx

Peer context associated with the local address.

int ecn

ECN: Valid values are 0 - 3. See RFC 3168.

ECN may be set by IETF QUIC connections if es_ecn is set.

void lsquic_engine_process_conns(lsquic_engine_t *engine)

Process tickable connections. This function must be called often enough so that packets and connections do not expire. The preferred method of doing so is by using lsquic_engine_earliest_adv_tick().

int lsquic_engine_has_unsent_packets(lsquic_engine_t *engine)

Returns true if engine has some unsent packets. This happens if ea_packets_out() could not send everything out.

void lsquic_engine_send_unsent_packets(lsquic_engine_t *engine)

Send out as many unsent packets as possibe: until we are out of unsent packets or until ea_packets_out() fails.

If ea_packets_out() cannot send all packets, this function must be called to signify that sending of packets is possible again.

Stream Callback Interface

The stream callback interface structure lists the callbacks used by the engine to communicate with the user code:

struct lsquic_stream_if
lsquic_conn_ctx_t *(*on_new_conn)(void *stream_if_ctx,
lsquic_conn_t *);

Called when a new connection has been created. In server mode, this means that the handshake has been successful. In client mode, on the other hand, this callback is called as soon as connection object is created inside the engine, but before the handshake is done.

The return value is the connection context associated with this connection. Use lsquic_conn_get_ctx() to get back this context. It is OK for this function to return NULL.

This callback is mandatory.

void (*on_conn_closed)(lsquic_conn_t *)

Connection is closed.

This callback is mandatory.

lsquic_stream_ctx_t * (*on_new_stream)(void *stream_if_ctx, lsquic_stream_t *)

If you need to initiate a connection, call lsquic_conn_make_stream(). This will cause on_new_stream() callback to be called when appropriate (this operation is delayed when maximum number of outgoing streams is reached).

If connection is going away, this callback may be called with the second parameter set to NULL.

The return value is the stream context associated with the stream. A pointer to it is passed to on_read(), on_write(), and on_close() callbacks. It is OK for this function to return NULL.

This callback is mandatory.

void (*on_read)(lsquic_stream_t *s, lsquic_stream_ctx_t *h)

Stream is readable: either there are bytes to be read or an error is ready to be collected.

This callback is mandatory.

void (*on_write)(lsquic_stream_t *s, lsquic_stream_ctx_t *h)

Stream is writeable.

This callback is mandatory.

void (*on_close)(lsquic_stream_t *s, lsquic_stream_ctx_t *h)

After this callback returns, the stream is no longer accessible. This is a good time to clean up the stream context.

This callback is mandatory.

void (*on_hsk_done)(lsquic_conn_t *c, enum lsquic_hsk_status s)

When handshake is completed, this callback is called.

This callback is optional.

void (*on_goaway_received)(lsquic_conn_t *)

This is called when our side received GOAWAY frame. After this, new streams should not be created.

This callback is optional.

void (*on_new_token)(lsquic_conn_t *c, const unsigned char *token, size_t token_size)

When client receives a token in NEW_TOKEN frame, this callback is called.

This callback is optional.

void (*on_zero_rtt_info)(lsquic_conn_t *c, const unsigned char *, size_t)

This callback lets client record information needed to perform a zero-RTT handshake next time around.

This callback is optional.

Creating Connections

In server mode, the connections are created by the library based on incoming packets. After handshake is completed, the library calls on_new_conn() callback.

In client mode, a new connection is created by

lsquic_conn_t * lsquic_engine_connect(lsquic_engine_t *engine, enum lsquic_version version, const struct sockaddr *local_sa, const struct sockaddr *peer_sa, void *peer_ctx, lsquic_conn_ctx_t *conn_ctx, const char *sni, unsigned short max_udp_payload_size, const unsigned char *zero_rtt, size_t zero_rtt_len, const unsigned char *token, size_t token_sz)
Parameters:
  • engine – Engine to use.
  • version – To let the engine specify QUIC version, use N_LSQVER. If zero-rtt info is supplied, version is picked from there instead.
  • local_sa – Local address.
  • peer_sa – Address of the server.
  • peer_ctx – Context associated with the connection. This is what gets passed to TODO.
  • conn_ctx – Connection context can be set early using this parameter. Useful if you need the connection context to be available in on_conn_new(). Note that that callback’s return value replaces the connection context set here.
  • sni – The SNI is required for Google QUIC connections; it is optional for IETF QUIC and may be set to NULL.
  • max_udp_payload_size – Maximum packet size. If set to zero, it is inferred based on peer_sa() and version().
  • zero_rtt – Pointer to previously saved zero-RTT data needed for TLS resumption. May be NULL.
  • zero_rtt_len – Size of zero-RTT data.
  • token

    Pointer to previously received token to include in the Initial packet. Tokens are used by IETF QUIC to pre-validate client connections, potentially avoiding a retry.

    See on_new_token callback in lsquic_stream_if:

    May be NULL.

  • token_sz – Size of data pointed to by token.

Closing Connections

void lsquic_conn_going_away(lsquic_conn_t *conn)

Mark connection as going away: send GOAWAY frame and do not accept any more incoming streams, nor generate streams of our own.

Only applicable to HTTP/3 and GQUIC connections. Otherwise a no-op.

void lsquic_conn_close(lsquic_conn_t *conn)

This closes the connection. on_conn_closed() and on_close() callbacks will be called.

Creating Streams

Similar to connections, streams are created by the library in server mode; they correspond to requests. In client mode, a new stream is created by

void lsquic_conn_make_stream(lsquic_conn_t *)

Create a new request stream. This causes on_new_stream() callback to be called. If creating more requests is not permitted at the moment (due to number of concurrent streams limit), stream creation is registered as “pending” and the stream is created later when number of streams dips under the limit again. Any number of pending streams can be created. Use lsquic_conn_n_pending_streams() and lsquic_conn_cancel_pending_streams() to manage pending streams.

If connection is going away, on_new_stream() is called with the stream parameter set to NULL.

Stream Events

To register or unregister an interest in a read or write event, use the following functions:

int lsquic_stream_wantread(lsquic_stream_t *stream, int want)
Parameters:
  • stream – Stream to read from.
  • want – Boolean value indicating whether the caller wants to read from stream.
Returns:

Previous value of want or -1 if the stream has already been closed for reading.

A stream becomes readable if there is was an error: for example, the peer may have reset the stream. In this case, reading from the stream will return an error.

int lsquic_stream_wantwrite(lsquic_stream_t *stream, int want)
Parameters:
  • stream – Stream to write to.
  • want – Boolean value indicating whether the caller wants to write to stream.
Returns:

Previous value of want or -1 if the stream has already been closed for writing.

Reading From Streams

ssize_t lsquic_stream_read(lsquic_stream_t *stream, unsigned char *buf, size_t sz)
Parameters:
  • stream – Stream to read from.
  • buf – Buffer to copy data to.
  • sz – Size of the buffer.
Returns:

Number of bytes read, zero if EOS has been reached, or -1 on error.

Read up to sz bytes from stream into buffer buf.

-1 is returned on error, in which case errno is set:

  • EBADF: The stream is closed.
  • ECONNRESET: The stream has been reset.
  • EWOULDBLOCK: There is no data to be read.
ssize_t lsquic_stream_readv(lsquic_stream_t *stream, const struct iovec *vec, int iovcnt)
Parameters:
  • stream – Stream to read from.
  • vec – Array of iovec structures.
  • iovcnt – Number of elements in vec.
Returns:

Number of bytes read, zero if EOS has been reached, or -1 on error.

Similar to lsquic_stream_read(), but reads data into a vector.

ssize_t lsquic_stream_readf(lsquic_stream_t *stream, size_t (*readf)(void *ctx, const unsigned char *buf, size_t len, int fin), void *ctx)
Parameters:
  • stream – Stream to read from.
  • readf

    The callback takes four parameters:

    • Pointer to user-supplied context;
    • Pointer to the data;
    • Data size (can be zero); and
    • Indicator whether the FIN follows the data.

    The callback returns number of bytes processed. If this number is zero or is smaller than len, reading from stream stops.

  • ctx – Context pointer passed to readf.

This function allows user-supplied callback to read the stream contents. It is meant to be used for zero-copy stream processing.

Return value and errors are same as in lsquic_stream_read().

Writing To Streams

ssize_t lsquic_stream_write(lsquic_stream_t *stream, const void *buf, size_t len)
Parameters:
  • stream – Stream to write to.
  • buf – Buffer to copy data from.
  • len – Number of bytes to copy.
Returns:

Number of bytes written – which may be smaller than len – or a negative value when an error occurs.

Write len bytes to the stream. Returns number of bytes written, which may be smaller that len.

A negative return value indicates a serious error (the library is likely to have aborted the connection because of it).

ssize_t lsquic_stream_writev(lsquic_stream_t *s, const struct iovec *vec, int count)

Like lsquic_stream_write(), but read data from a vector.

struct lsquic_reader

Used as argument to lsquic_stream_writef().

size_t (*lsqr_read)(void *lsqr_ctx, void *buf, size_t count)
Parameters:
  • lsqr_ctx – Pointer to user-specified context.
  • buf – Memory location to write to.
  • count – Size of available memory pointed to by buf.
Returns:

Number of bytes written. This is not a ssize_t because the read function is not supposed to return an error. If an error occurs in the read function (for example, when reading from a file fails), it is supposed to deal with the error itself.

size_t (*lsqr_size)(void *lsqr_ctx)

Return number of bytes remaining in the reader.

void *lsqr_ctx

Context pointer passed both to lsqr_read() and to lsqr_size().

ssize_t lsquic_stream_writef(lsquic_stream_t *stream, struct lsquic_reader *reader)
Parameters:
  • stream – Stream to write to.
  • reader – Reader to read from.
Returns:

Number of bytes written or -1 on error.

Write to stream using lsquic_reader. This is the most generic of the write functions – lsquic_stream_write() and lsquic_stream_writev() utilize the same mechanism.

int lsquic_stream_flush(lsquic_stream_t *stream)
Parameters:
  • stream – Stream to flush.
Returns:

0 on success and -1 on failure.

Flush any buffered data. This triggers packetizing even a single byte into a separate frame. Flushing a closed stream is an error.

Closing Streams

Streams can be closed for reading, writing, or both. on_close() callback is called at some point after a stream is closed for both reading and writing,

int lsquic_stream_shutdown(lsquic_stream_t *stream, int how)
Parameters:
  • stream – Stream to shut down.
  • how

    This parameter specifies what do to. Allowed values are:

    • 0: Stop reading.
    • 1: Stop writing.
    • 2: Stop both reading and writing.
Returns:

0 on success or -1 on failure.

int lsquic_stream_close(lsquic_stream_t *stream)
Parameters:
  • stream – Stream to close.
Returns:

0 on success or -1 on failure.

Sending HTTP Headers

struct lsxpack_header

This type is defined in _lsxpack_header.h_. See that header file for more information.

char *buf

the buffer for headers

const char *name_ptr

the name pointer can be optionally set for encoding

uint32_t name_hash

hash value for name

uint32_t nameval_hash

hash value for name + value

lsxpack_strlen_t name_offset

the offset for name in the buffer

lsxpack_strlen_t name_len

the length of name

lsxpack_strlen_t val_offset

the offset for value in the buffer

lsxpack_strlen_t val_len

the length of value

uint16_t chain_next_idx

mainly for cookie value chain

uint8_t hpack_index

HPACK static table index

uint8_t qpack_index

QPACK static table index

uint8_t app_index

APP header index

enum lsxpack_flag flags:8

combination of lsxpack_flag

uint8_t indexed_type

control to disable index or not

uint8_t dec_overhead

num of extra bytes written to decoded buffer

lsquic_http_headers_t
int count

Number of headers in headers.

struct lsxpack_header *headers

Pointer to an array of HTTP headers.

HTTP header list structure. Contains a list of HTTP headers.

int lsquic_stream_send_headers(lsquic_stream_t *stream, const lsquic_http_headers_t *headers, int eos)
Parameters:
  • stream – Stream to send headers on.
  • headers – Headers to send.
  • eos – Boolean value to indicate whether these headers constitute the whole HTTP message.
Returns:

0 on success or -1 on error.

Receiving HTTP Headers

If ea_hsi_if is not set in lsquic_engine_api, the library will translate HPACK- and QPACK-encoded headers into HTTP/1.x-like headers and prepend them to the stream. To the stream-reading function, it will look as if a standard HTTP/1.x message.

Alternatively, you can specify header-processing set of functions and manage header fields yourself. In that case, the header set must be “read” from the stream via lsquic_stream_get_hset().

struct lsquic_hset_if
void * (*hsi_create_header_set)(void *hsi_ctx, lsquic_stream_t *stream, int is_push_promise)
Parameters:
  • hsi_ctx – User context. This is the pointer specifed in ea_hsi_ctx.
  • stream – Stream with which the header set is associated. May be set to NULL in server mode.
  • is_push_promise – Boolean value indicating whether this header set is for a push promise.
Returns:

Pointer to user-defined header set object.

Create a new header set. This object is (and must be) fetched from a stream by calling lsquic_stream_get_hset() before the stream can be read.

struct lsxpack_header * (*hsi_prepare_decode)(void *hdr_set, struct lsxpack_header *hdr, size_t space)

Return a header set prepared for decoding. If hdr is NULL, this means return a new structure with at least space bytes available in the decoder buffer. On success, a newly prepared header is returned.

If hdr is not NULL, it means there was not enough decoder buffer and it must be increased to at least space bytes. buf, val_len, and name_offset member of the hdr structure may change. On success, the return value is the same as hdr.

If NULL is returned, the space cannot be allocated.

int (*hsi_process_header)(void *hdr_set, struct lsxpack_header *hdr)

Process new header.

Parameters:
  • hdr_set – Header set to add the new header field to. This is the object returned by hsi_create_header_set().
  • hdr – The header returned by @ref hsi_prepare_decode().
Returns:

Return 0 on success, a positive value if a header error occured, or a negative value on any other error. A positive return value will result in cancellation of associated stream. A negative return value will result in connection being aborted.

void (*hsi_discard_header_set)(void *hdr_set)
Parameters:
  • hdr_set – Header set to discard.

Discard header set. This is called for unclaimed header sets and header sets that had an error.

enum lsquic_hsi_flag hsi_flags

These flags specify properties of decoded headers passed to hsi_process_header(). This is only applicable to QPACK headers; HPACK library header properties are based on compilation, not run-time, options.

void * lsquic_stream_get_hset(lsquic_stream_t *stream)
Parameters:
  • stream – Stream to fetch header set from.
Returns:

Header set associated with the stream.

Get header set associated with the stream. The header set is created by hsi_create_header_set() callback. After this call, the ownership of the header set is transferred to the caller.

This call must precede calls to lsquic_stream_read(), lsquic_stream_readv(), and lsquic_stream_readf().

If the optional header set interface is not specified, this function returns NULL.

Push Promises

int lsquic_conn_push_stream(lsquic_conn_t *conn, void *hdr_set, lsquic_stream_t *stream, const lsquic_http_headers_t *headers)
Returns:
  • 0: Stream pushed successfully.
  • 1: Stream push failed because it is disabled or because we hit
    stream limit or connection is going away.
  • -1: Stream push failed because of an internal error.

A server may push a stream. This call creates a new stream in reference to stream stream. It will behave as if the client made a request: it will trigger on_new_stream() event and it can be used as a regular client-initiated stream.

hdr_set must be set. It is passed as-is to lsquic_stream_get_hset().

int lsquic_conn_is_push_enabled(lsquic_conn_t *conn)
Returns:Boolean value indicating whether push promises are enabled.

Only makes sense in server mode: the client cannot push a stream and this function always returns false in client mode.

int lsquic_stream_refuse_push(lsquic_stream_t *stream)

Refuse pushed stream. Call it from on_new_stream(). No need to call lsquic_stream_close() after this. on_close() will be called.

int lsquic_stream_push_info(const lsquic_stream_t *stream, lsquic_stream_id_t *ref_stream_id, void **hdr_set)

Get information associated with pushed stream

Parameters:
  • ref_stream_id – Stream ID in response to which push promise was sent.
  • hdr_set – Header set. This object was passed to or generated by lsquic_conn_push_stream().
Returns:

0 on success and -1 if this is not a pushed stream.

Stream Priorities

unsigned lsquic_stream_priority(const lsquic_stream_t *stream)

Return current priority of the stream.

int lsquic_stream_set_priority(lsquic_stream_t *stream, unsigned priority)

Set stream priority. Valid priority values are 1 through 256, inclusive.

Returns:0 on success of -1 on failure (this happens if priority value is invalid).

Miscellaneous Engine Functions

unsigned lsquic_engine_quic_versions(const lsquic_engine_t *engine)

Return the list of QUIC versions (as bitmask) this engine instance supports.

unsigned lsquic_engine_count_attq(lsquic_engine_t *engine, int from_now)

Return number of connections whose advisory tick time is before current time plus from_now microseconds from now. from_now can be negative.

Miscellaneous Connection Functions

enum lsquic_version lsquic_conn_quic_version(const lsquic_conn_t *conn)

Get QUIC version used by the connection.

If version has not yet been negotiated (can happen in client mode), -1 is returned.

const lsquic_cid_t * lsquic_conn_id(const lsquic_conn_t *conn)

Get connection ID.

lsquic_engine_t * lsquic_conn_get_engine(lsquic_conn_t *conn)

Get pointer to the engine.

int lsquic_conn_get_sockaddr(lsquic_conn_t *conn, const struct sockaddr **local, const struct sockaddr **peer)

Get current (last used) addresses associated with the current path used by the connection.

struct stack_st_X509 * lsquic_conn_get_server_cert_chain(lsquic_conn_t *conn)

Get certificate chain returned by the server. This can be used for server certificate verification.

The caller releases the stack using sk_X509_free().

lsquic_conn_ctx_t * lsquic_conn_get_ctx(const lsquic_conn_t *conn)

Get user-supplied context associated with the connection.

void lsquic_conn_set_ctx(lsquic_conn_t *conn, lsquic_conn_ctx_t *ctx)

Set user-supplied context associated with the connection.

void * lsquic_conn_get_peer_ctx(lsquic_conn_t *conn, const struct sockaddr *local_sa)

Get peer context associated with the connection and local address.

enum LSQUIC_CONN_STATUS lsquic_conn_status(lsquic_conn_t *conn, char *errbuf, size_t bufsz)

Get connection status.

Miscellaneous Stream Functions

unsigned lsquic_conn_n_avail_streams(const lsquic_conn_t *conn)

Return max allowed outbound streams less current outbound streams.

unsigned lsquic_conn_n_pending_streams(const lsquic_conn_t *conn)

Return number of delayed streams currently pending.

unsigned lsquic_conn_cancel_pending_streams(lsquic_conn_t *, unsigned n)

Cancel n pending streams. Returns new number of pending streams.

lsquic_conn_t * lsquic_stream_conn(const lsquic_stream_t *stream)

Get a pointer to the connection object. Use it with connection functions.

int lsquic_stream_is_rejected(const lsquic_stream_t *stream)

Returns true if this stream was rejected, false otherwise. Use this as an aid to distinguish between errors.

Other Functions

enum lsquic_version lsquic_str2ver(const char *str, size_t len)

Translate string QUIC version to LSQUIC QUIC version representation.

enum lsquic_version lsquic_alpn2ver(const char *alpn, size_t len)

Translate ALPN (e.g. “h3”, “h3-23”, “h3-Q046”) to LSQUIC enum.

Miscellaneous Types

struct lsquic_shared_hash_if

The shared hash interface is used to share data between multiple LSQUIC instances.

int (*shi_insert)(void *shi_ctx, void *key, unsigned key_sz, void *data, unsigned data_sz, time_t expiry)
Parameters:
  • shi_ctx – Shared memory context pointer
  • key – Key data.
  • key_sz – Key size.
  • data – Pointer to the data to store.
  • data_sz – Data size.
  • expiry – When this item expires. If you want your item to never expire, set this to zero.
Returns:

0 on success, -1 on failure.

If inserted successfully, free() will be called on data and key pointer when the element is deleted, whether due to expiration or explicit deletion.

int (*shi_delete)(void *shi_ctx, const void *key, unsigned key_sz)

Delete item from shared hash

Returns:0 on success, -1 on failure.
int (*shi_lookup)(void *shi_ctx, const void *key, unsigned key_sz, void **data, unsigned *data_sz)
Parameters:
  • shi_ctx – Shared memory context pointer
  • key – Key data.
  • key_sz – Key size.
  • data – Pointer to set to the result.
  • data_sz – Pointer to the data size.
Returns:

  • 1: found.
  • 0: not found.
  • -1: error (perhaps not enough room in data if copy was attempted).

The implementation may choose to copy the object into buffer pointed to by data, so you should have it ready.

struct lsquic_packout_mem_if

The packet out memory interface is used by LSQUIC to get buffers to which outgoing packets will be written before they are passed to lsquic_engine_api.ea_packets_out callback.

If not specified, malloc() and free() are used.

void * (*pmi_allocate)(void *pmi_ctx, void *conn_ctx, unsigned short sz, char is_ipv6)

Allocate buffer for sending.

void (*pmi_release)(void *pmi_ctx, void *conn_ctx, void *buf, char is_ipv6)

This function is used to release the allocated buffer after it is sent via ea_packets_out().

void (*pmi_return)(void *pmi_ctx, void *conn_ctx, void *buf, char is_ipv6)

If allocated buffer is not going to be sent, return it to the caller using this function.

typedef void (*lsquic_cids_update_f)(void *ctx, void **peer_ctx, const lsquic_cid_t *cids, unsigned n_cids)
Parameters:
  • ctx – Context associated with the CID lifecycle callbacks (ea_cids_update_ctx).
  • peer_ctx – Array of peer context pointers.
  • cids – Array of connection IDs.
  • n_cids – Number of elements in the peer context pointer and connection ID arrays.
struct lsquic_keylog_if

SSL keylog interface.

void * (*kli_open)(void *keylog_ctx, lsquic_conn_t *conn)

Return keylog handle or NULL if no key logging is desired.

void (*kli_log_line)(void *handle, const char *line)

Log line. The first argument is the pointer returned by kli_open().

void (*kli_close)(void *handle)

Close handle.

enum lsquic_logger_timestamp_style

Enumerate timestamp styles supported by LSQUIC logger mechanism.

LLTS_NONE

No timestamp is generated.

LLTS_HHMMSSMS

The timestamp consists of 24 hours, minutes, seconds, and milliseconds. Example: 13:43:46.671

LLTS_YYYYMMDD_HHMMSSMS

Like above, plus date, e.g: 2017-03-21 13:43:46.671

LLTS_CHROMELIKE

This is Chrome-like timestamp used by proto-quic. The timestamp includes month, date, hours, minutes, seconds, and microseconds.

Example: 1223/104613.946956 (instead of 12/23 10:46:13.946956).

This is to facilitate reading two logs side-by-side.

LLTS_HHMMSSUS

The timestamp consists of 24 hours, minutes, seconds, and microseconds. Example: 13:43:46.671123

LLTS_YYYYMMDD_HHMMSSUS

Date and time using microsecond resolution, e.g: 2017-03-21 13:43:46.671123

enum LSQUIC_CONN_STATUS
LSCONN_ST_HSK_IN_PROGRESS
LSCONN_ST_CONNECTED
LSCONN_ST_HSK_FAILURE
LSCONN_ST_GOING_AWAY
LSCONN_ST_TIMED_OUT
LSCONN_ST_RESET

If es_honor_prst is not set, the connection will never get public reset packets and this flag will not be set.

LSCONN_ST_USER_ABORTED
LSCONN_ST_ERROR
LSCONN_ST_CLOSED
LSCONN_ST_PEER_GOING_AWAY
enum lsquic_hsi_flag

These flags are ORed together to specify properties of lsxpack_header passed to lsquic_hset_if.hsi_process_header.

LSQUIC_HSI_HTTP1X

Turn HTTP/1.x mode on or off. In this mode, decoded name and value pair are separated by ": " and "\r\n" is appended to the end of the string. By default, this mode is off.

LSQUIC_HSI_HASH_NAME

Include name hash into lsxpack_header.

LSQUIC_HSI_HASH_NAMEVAL

Include nameval hash into lsxpack_header.

Global Variables

const char *const lsquic_ver2str[N_LSQVER]

Convert LSQUIC version to human-readable string

List of Log Modules

The following log modules are defined:

  • alarmset: Alarm processing.
  • bbr: BBR congestion controller.
  • bw-sampler: Bandwidth sampler (used by BBR).
  • cfcw: Connection flow control window.
  • conn: Connection.
  • crypto: Low-level Google QUIC cryptography tracing.
  • cubic: Cubic congestion controller.
  • di: “Data In” handler (storing incoming data before it is read).
  • eng-hist: Engine history.
  • engine: Engine.
  • event: Cross-module significant events.
  • frame-reader: Reader of the HEADERS stream in Google QUIC.
  • frame-writer: Writer of the HEADERS stream in Google QUIC.
  • handshake: Handshake and packet encryption and decryption.
  • hcsi-reader: Reader of the HTTP/3 control stream.
  • hcso-writer: Writer of the HTTP/3 control stream.
  • headers: HEADERS stream (Google QUIC).
  • hsk-adapter:
  • http1x: Header conversion to HTTP/1.x.
  • logger: Logger.
  • mini-conn: Mini connection.
  • pacer: Pacer.
  • parse: Parsing.
  • prq: PRQ stands for Packet Request Queue. This logs scheduling and sending packets not associated with a connection: version negotiation and stateless resets.
  • purga: CID purgatory.
  • qdec-hdl: QPACK decoder stream handler.
  • qenc-hdl: QPACK encoder stream handler.
  • qlog: QLOG output. At the moment, it is out of date.
  • qpack-dec: QPACK decoder.
  • qpack-enc: QPACK encoder.
  • rechist: Receive history.
  • sendctl: Send controller.
  • sfcw: Stream flow control window.
  • spi: Stream priority iterator.
  • stream: Stream operation.
  • tokgen: Token generation and validation.
  • trapa: Transport parameter processing.