/*************************************************************************** * _ _ ____ _ * Project ___| | | | _ \| | * / __| | | | |_) | | * | (__| |_| | _ <| |___ * \___|\___/|_| \_\_____| * * Copyright (C) Daniel Stenberg, , et al. * * This software is licensed as described in the file COPYING, which * you should have received as part of this distribution. The terms * are also available at https://curl.se/docs/copyright.html. * * You may opt to use, copy, modify, merge, publish, distribute and/or sell * copies of the Software, and permit persons to whom the Software is * furnished to do so, under the terms of the COPYING file. * * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY * KIND, either express or implied. * * SPDX-License-Identifier: curl * ***************************************************************************/ #include "curl_setup.h" #if defined(USE_NGTCP2) && defined(USE_NGHTTP3) #include #include #ifdef USE_OPENSSL #include #if defined(OPENSSL_IS_BORINGSSL) || defined(OPENSSL_IS_AWSLC) #include #else #include #endif #include "vtls/openssl.h" #elif defined(USE_GNUTLS) #include #include "vtls/gtls.h" #elif defined(USE_WOLFSSL) #include #include "vtls/wolfssl.h" #endif #include "urldata.h" #include "sendf.h" #include "strdup.h" #include "rand.h" #include "multiif.h" #include "strcase.h" #include "cfilters.h" #include "cf-socket.h" #include "connect.h" #include "progress.h" #include "strerror.h" #include "dynbuf.h" #include "http1.h" #include "select.h" #include "inet_pton.h" #include "vquic.h" #include "vquic_int.h" #include "vtls/keylog.h" #include "vtls/vtls.h" #include "curl_ngtcp2.h" #include "warnless.h" /* The last 3 #include files should be in this order */ #include "curl_printf.h" #include "curl_memory.h" #include "memdebug.h" #define H3_ALPN_H3_29 "\x5h3-29" #define H3_ALPN_H3 "\x2h3" #define QUIC_MAX_STREAMS (256*1024) #define QUIC_MAX_DATA (1*1024*1024) #define QUIC_HANDSHAKE_TIMEOUT (10*NGTCP2_SECONDS) /* A stream window is the maximum amount we need to buffer for * each active transfer. We use HTTP/3 flow control and only ACK * when we take things out of the buffer. * Chunk size is large enough to take a full DATA frame */ #define H3_STREAM_WINDOW_SIZE (128 * 1024) #define H3_STREAM_CHUNK_SIZE (16 * 1024) /* The pool keeps spares around and half of a full stream windows * seems good. More does not seem to improve performance. * The benefit of the pool is that stream buffer to not keep * spares. So memory consumption goes down when streams run empty, * have a large upload done, etc. */ #define H3_STREAM_POOL_SPARES \ (H3_STREAM_WINDOW_SIZE / H3_STREAM_CHUNK_SIZE ) / 2 /* Receive and Send max number of chunks just follows from the * chunk size and window size */ #define H3_STREAM_RECV_CHUNKS \ (H3_STREAM_WINDOW_SIZE / H3_STREAM_CHUNK_SIZE) #define H3_STREAM_SEND_CHUNKS \ (H3_STREAM_WINDOW_SIZE / H3_STREAM_CHUNK_SIZE) #ifdef USE_OPENSSL #define QUIC_CIPHERS \ "TLS_AES_128_GCM_SHA256:TLS_AES_256_GCM_SHA384:TLS_CHACHA20_" \ "POLY1305_SHA256:TLS_AES_128_CCM_SHA256" #define QUIC_GROUPS "P-256:X25519:P-384:P-521" #elif defined(USE_GNUTLS) #define QUIC_PRIORITY \ "NORMAL:-VERS-ALL:+VERS-TLS1.3:-CIPHER-ALL:+AES-128-GCM:+AES-256-GCM:" \ "+CHACHA20-POLY1305:+AES-128-CCM:-GROUP-ALL:+GROUP-SECP256R1:" \ "+GROUP-X25519:+GROUP-SECP384R1:+GROUP-SECP521R1:" \ "%DISABLE_TLS13_COMPAT_MODE" #elif defined(USE_WOLFSSL) #define QUIC_CIPHERS \ "TLS_AES_128_GCM_SHA256:TLS_AES_256_GCM_SHA384:TLS_CHACHA20_" \ "POLY1305_SHA256:TLS_AES_128_CCM_SHA256" #define QUIC_GROUPS "P-256:P-384:P-521" #endif /* * Store ngtcp2 version info in this buffer. */ void Curl_ngtcp2_ver(char *p, size_t len) { const ngtcp2_info *ng2 = ngtcp2_version(0); const nghttp3_info *ht3 = nghttp3_version(0); (void)msnprintf(p, len, "ngtcp2/%s nghttp3/%s", ng2->version_str, ht3->version_str); } struct cf_ngtcp2_ctx { struct cf_quic_ctx q; struct ssl_peer peer; ngtcp2_path connected_path; ngtcp2_conn *qconn; ngtcp2_cid dcid; ngtcp2_cid scid; uint32_t version; ngtcp2_settings settings; ngtcp2_transport_params transport_params; ngtcp2_ccerr last_error; ngtcp2_crypto_conn_ref conn_ref; #ifdef USE_OPENSSL SSL_CTX *sslctx; SSL *ssl; #elif defined(USE_GNUTLS) struct gtls_instance *gtls; #elif defined(USE_WOLFSSL) WOLFSSL_CTX *sslctx; WOLFSSL *ssl; #endif struct cf_call_data call_data; nghttp3_conn *h3conn; nghttp3_settings h3settings; struct curltime started_at; /* time the current attempt started */ struct curltime handshake_at; /* time connect handshake finished */ struct curltime first_byte_at; /* when first byte was recvd */ struct curltime reconnect_at; /* time the next attempt should start */ struct bufc_pool stream_bufcp; /* chunk pool for streams */ size_t max_stream_window; /* max flow window for one stream */ uint64_t max_idle_ms; /* max idle time for QUIC connection */ int qlogfd; BIT(got_first_byte); /* if first byte was received */ #ifdef USE_OPENSSL BIT(x509_store_setup); /* if x509 store has been set up */ #endif }; /* How to access `call_data` from a cf_ngtcp2 filter */ #undef CF_CTX_CALL_DATA #define CF_CTX_CALL_DATA(cf) \ ((struct cf_ngtcp2_ctx *)(cf)->ctx)->call_data /** * All about the H3 internals of a stream */ struct h3_stream_ctx { int64_t id; /* HTTP/3 protocol identifier */ struct bufq sendbuf; /* h3 request body */ struct bufq recvbuf; /* h3 response body */ struct h1_req_parser h1; /* h1 request parsing */ size_t sendbuf_len_in_flight; /* sendbuf amount "in flight" */ size_t upload_blocked_len; /* the amount written last and EGAINed */ size_t recv_buf_nonflow; /* buffered bytes, not counting for flow control */ uint64_t error3; /* HTTP/3 stream error code */ curl_off_t upload_left; /* number of request bytes left to upload */ int status_code; /* HTTP status code */ bool resp_hds_complete; /* we have a complete, final response */ bool closed; /* TRUE on stream close */ bool reset; /* TRUE on stream reset */ bool send_closed; /* stream is local closed */ BIT(quic_flow_blocked); /* stream is blocked by QUIC flow control */ }; #define H3_STREAM_CTX(d) ((struct h3_stream_ctx *)(((d) && (d)->req.p.http)? \ ((struct HTTP *)(d)->req.p.http)->h3_ctx \ : NULL)) #define H3_STREAM_LCTX(d) ((struct HTTP *)(d)->req.p.http)->h3_ctx #define H3_STREAM_ID(d) (H3_STREAM_CTX(d)? \ H3_STREAM_CTX(d)->id : -2) static CURLcode h3_data_setup(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); if(!data || !data->req.p.http) { failf(data, "initialization failure, transfer not http initialized"); return CURLE_FAILED_INIT; } if(stream) return CURLE_OK; stream = calloc(1, sizeof(*stream)); if(!stream) return CURLE_OUT_OF_MEMORY; stream->id = -1; /* on send, we control how much we put into the buffer */ Curl_bufq_initp(&stream->sendbuf, &ctx->stream_bufcp, H3_STREAM_SEND_CHUNKS, BUFQ_OPT_NONE); stream->sendbuf_len_in_flight = 0; /* on recv, we need a flexible buffer limit since we also write * headers to it that are not counted against the nghttp3 flow limits. */ Curl_bufq_initp(&stream->recvbuf, &ctx->stream_bufcp, H3_STREAM_RECV_CHUNKS, BUFQ_OPT_SOFT_LIMIT); stream->recv_buf_nonflow = 0; Curl_h1_req_parse_init(&stream->h1, H1_PARSE_DEFAULT_MAX_LINE_LEN); H3_STREAM_LCTX(data) = stream; return CURLE_OK; } static void h3_data_done(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); (void)cf; if(stream) { CURL_TRC_CF(data, cf, "[%"PRId64"] easy handle is done", stream->id); if(ctx->h3conn && !stream->closed) { nghttp3_conn_shutdown_stream_read(ctx->h3conn, stream->id); nghttp3_conn_close_stream(ctx->h3conn, stream->id, NGHTTP3_H3_REQUEST_CANCELLED); nghttp3_conn_set_stream_user_data(ctx->h3conn, stream->id, NULL); ngtcp2_conn_set_stream_user_data(ctx->qconn, stream->id, NULL); stream->closed = TRUE; } Curl_bufq_free(&stream->sendbuf); Curl_bufq_free(&stream->recvbuf); Curl_h1_req_parse_free(&stream->h1); free(stream); H3_STREAM_LCTX(data) = NULL; } } static struct Curl_easy *get_stream_easy(struct Curl_cfilter *cf, struct Curl_easy *data, int64_t stream_id) { struct Curl_easy *sdata; (void)cf; if(H3_STREAM_ID(data) == stream_id) { return data; } else { DEBUGASSERT(data->multi); for(sdata = data->multi->easyp; sdata; sdata = sdata->next) { if((sdata->conn == data->conn) && H3_STREAM_ID(sdata) == stream_id) { return sdata; } } } return NULL; } static void h3_drain_stream(struct Curl_cfilter *cf, struct Curl_easy *data) { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); unsigned char bits; (void)cf; bits = CURL_CSELECT_IN; if(stream && stream->upload_left && !stream->send_closed) bits |= CURL_CSELECT_OUT; if(data->state.dselect_bits != bits) { data->state.dselect_bits = bits; Curl_expire(data, 0, EXPIRE_RUN_NOW); } } /* ngtcp2 default congestion controller does not perform pacing. Limit the maximum packet burst to MAX_PKT_BURST packets. */ #define MAX_PKT_BURST 10 struct pkt_io_ctx { struct Curl_cfilter *cf; struct Curl_easy *data; ngtcp2_tstamp ts; size_t pkt_count; ngtcp2_path_storage ps; }; static void pktx_update_time(struct pkt_io_ctx *pktx, struct Curl_cfilter *cf) { struct cf_ngtcp2_ctx *ctx = cf->ctx; vquic_ctx_update_time(&ctx->q); pktx->ts = ctx->q.last_op.tv_sec * NGTCP2_SECONDS + ctx->q.last_op.tv_usec * NGTCP2_MICROSECONDS; } static void pktx_init(struct pkt_io_ctx *pktx, struct Curl_cfilter *cf, struct Curl_easy *data) { pktx->cf = cf; pktx->data = data; pktx->pkt_count = 0; ngtcp2_path_storage_zero(&pktx->ps); pktx_update_time(pktx, cf); } static CURLcode cf_progress_ingress(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx); static CURLcode cf_progress_egress(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx); static int cb_h3_acked_req_body(nghttp3_conn *conn, int64_t stream_id, uint64_t datalen, void *user_data, void *stream_user_data); static ngtcp2_conn *get_conn(ngtcp2_crypto_conn_ref *conn_ref) { struct Curl_cfilter *cf = conn_ref->user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; return ctx->qconn; } #ifdef DEBUG_NGTCP2 static void quic_printf(void *user_data, const char *fmt, ...) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; (void)ctx; /* TODO: need an easy handle to infof() message */ va_list ap; va_start(ap, fmt); vfprintf(stderr, fmt, ap); va_end(ap); fprintf(stderr, "\n"); } #endif static void qlog_callback(void *user_data, uint32_t flags, const void *data, size_t datalen) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; (void)flags; if(ctx->qlogfd != -1) { ssize_t rc = write(ctx->qlogfd, data, datalen); if(rc == -1) { /* on write error, stop further write attempts */ close(ctx->qlogfd); ctx->qlogfd = -1; } } } static void quic_settings(struct cf_ngtcp2_ctx *ctx, struct Curl_easy *data, struct pkt_io_ctx *pktx) { ngtcp2_settings *s = &ctx->settings; ngtcp2_transport_params *t = &ctx->transport_params; ngtcp2_settings_default(s); ngtcp2_transport_params_default(t); #ifdef DEBUG_NGTCP2 s->log_printf = quic_printf; #else s->log_printf = NULL; #endif (void)data; s->initial_ts = pktx->ts; s->handshake_timeout = QUIC_HANDSHAKE_TIMEOUT; s->max_window = 100 * ctx->max_stream_window; s->max_stream_window = ctx->max_stream_window; t->initial_max_data = 10 * ctx->max_stream_window; t->initial_max_stream_data_bidi_local = ctx->max_stream_window; t->initial_max_stream_data_bidi_remote = ctx->max_stream_window; t->initial_max_stream_data_uni = ctx->max_stream_window; t->initial_max_streams_bidi = QUIC_MAX_STREAMS; t->initial_max_streams_uni = QUIC_MAX_STREAMS; t->max_idle_timeout = (ctx->max_idle_ms * NGTCP2_MILLISECONDS); if(ctx->qlogfd != -1) { s->qlog_write = qlog_callback; } } #ifdef USE_OPENSSL static void keylog_callback(const SSL *ssl, const char *line) { (void)ssl; Curl_tls_keylog_write_line(line); } #elif defined(USE_GNUTLS) static int keylog_callback(gnutls_session_t session, const char *label, const gnutls_datum_t *secret) { gnutls_datum_t crandom; gnutls_datum_t srandom; gnutls_session_get_random(session, &crandom, &srandom); if(crandom.size != 32) { return -1; } Curl_tls_keylog_write(label, crandom.data, secret->data, secret->size); return 0; } #elif defined(USE_WOLFSSL) #if defined(HAVE_SECRET_CALLBACK) static void keylog_callback(const WOLFSSL *ssl, const char *line) { (void)ssl; Curl_tls_keylog_write_line(line); } #endif #endif static int init_ngh3_conn(struct Curl_cfilter *cf); #ifdef USE_OPENSSL static CURLcode quic_ssl_ctx(SSL_CTX **pssl_ctx, struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct ssl_primary_config *conn_config; CURLcode result = CURLE_FAILED_INIT; SSL_CTX *ssl_ctx = SSL_CTX_new(TLS_method()); if(!ssl_ctx) { result = CURLE_OUT_OF_MEMORY; goto out; } conn_config = Curl_ssl_cf_get_primary_config(cf); if(!conn_config) { result = CURLE_FAILED_INIT; goto out; } #if defined(OPENSSL_IS_BORINGSSL) || defined(OPENSSL_IS_AWSLC) if(ngtcp2_crypto_boringssl_configure_client_context(ssl_ctx) != 0) { failf(data, "ngtcp2_crypto_boringssl_configure_client_context failed"); goto out; } #else if(ngtcp2_crypto_quictls_configure_client_context(ssl_ctx) != 0) { failf(data, "ngtcp2_crypto_quictls_configure_client_context failed"); goto out; } #endif SSL_CTX_set_default_verify_paths(ssl_ctx); { const char *curves = conn_config->curves ? conn_config->curves : QUIC_GROUPS; if(!SSL_CTX_set1_curves_list(ssl_ctx, curves)) { failf(data, "failed setting curves list for QUIC: '%s'", curves); return CURLE_SSL_CIPHER; } } #ifndef OPENSSL_IS_BORINGSSL { const char *ciphers13 = conn_config->cipher_list13 ? conn_config->cipher_list13 : QUIC_CIPHERS; if(SSL_CTX_set_ciphersuites(ssl_ctx, ciphers13) != 1) { failf(data, "failed setting QUIC cipher suite: %s", ciphers13); return CURLE_SSL_CIPHER; } infof(data, "QUIC cipher selection: %s", ciphers13); } #endif /* Open the file if a TLS or QUIC backend has not done this before. */ Curl_tls_keylog_open(); if(Curl_tls_keylog_enabled()) { SSL_CTX_set_keylog_callback(ssl_ctx, keylog_callback); } /* OpenSSL always tries to verify the peer, this only says whether it should * fail to connect if the verification fails, or if it should continue * anyway. In the latter case the result of the verification is checked with * SSL_get_verify_result() below. */ SSL_CTX_set_verify(ssl_ctx, conn_config->verifypeer ? SSL_VERIFY_PEER : SSL_VERIFY_NONE, NULL); /* give application a chance to interfere with SSL set up. */ if(data->set.ssl.fsslctx) { /* When a user callback is installed to modify the SSL_CTX, * we need to do the full initialization before calling it. * See: #11800 */ if(!ctx->x509_store_setup) { result = Curl_ssl_setup_x509_store(cf, data, ssl_ctx); if(result) goto out; ctx->x509_store_setup = TRUE; } Curl_set_in_callback(data, true); result = (*data->set.ssl.fsslctx)(data, ssl_ctx, data->set.ssl.fsslctxp); Curl_set_in_callback(data, false); if(result) { failf(data, "error signaled by ssl ctx callback"); goto out; } } result = CURLE_OK; out: *pssl_ctx = result? NULL : ssl_ctx; if(result && ssl_ctx) SSL_CTX_free(ssl_ctx); return result; } static CURLcode quic_set_client_cert(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; SSL_CTX *ssl_ctx = ctx->sslctx; const struct ssl_config_data *ssl_config; ssl_config = Curl_ssl_cf_get_config(cf, data); DEBUGASSERT(ssl_config); if(ssl_config->primary.clientcert || ssl_config->primary.cert_blob || ssl_config->cert_type) { return Curl_ossl_set_client_cert( data, ssl_ctx, ssl_config->primary.clientcert, ssl_config->primary.cert_blob, ssl_config->cert_type, ssl_config->key, ssl_config->key_blob, ssl_config->key_type, ssl_config->key_passwd); } return CURLE_OK; } /** SSL callbacks ***/ static CURLcode quic_init_ssl(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; const uint8_t *alpn = NULL; size_t alpnlen = 0; DEBUGASSERT(!ctx->ssl); ctx->ssl = SSL_new(ctx->sslctx); SSL_set_app_data(ctx->ssl, &ctx->conn_ref); SSL_set_connect_state(ctx->ssl); SSL_set_quic_use_legacy_codepoint(ctx->ssl, 0); alpn = (const uint8_t *)H3_ALPN_H3_29 H3_ALPN_H3; alpnlen = sizeof(H3_ALPN_H3_29) - 1 + sizeof(H3_ALPN_H3) - 1; if(alpn) SSL_set_alpn_protos(ctx->ssl, alpn, (int)alpnlen); /* set SNI */ if(ctx->peer.sni) { if(!SSL_set_tlsext_host_name(ctx->ssl, ctx->peer.sni)) { failf(data, "Failed set SNI"); SSL_free(ctx->ssl); ctx->ssl = NULL; return CURLE_QUIC_CONNECT_ERROR; } } return CURLE_OK; } #elif defined(USE_GNUTLS) static CURLcode quic_init_ssl(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct ssl_primary_config *conn_config; CURLcode result; gnutls_datum_t alpn[2]; /* this will need some attention when HTTPS proxy over QUIC get fixed */ long * const pverifyresult = &data->set.ssl.certverifyresult; int rc; conn_config = Curl_ssl_cf_get_primary_config(cf); if(!conn_config) return CURLE_FAILED_INIT; DEBUGASSERT(ctx->gtls == NULL); ctx->gtls = calloc(1, sizeof(*(ctx->gtls))); if(!ctx->gtls) return CURLE_OUT_OF_MEMORY; result = gtls_client_init(data, conn_config, &data->set.ssl, &ctx->peer, ctx->gtls, pverifyresult); if(result) return result; gnutls_session_set_ptr(ctx->gtls->session, &ctx->conn_ref); if(ngtcp2_crypto_gnutls_configure_client_session(ctx->gtls->session) != 0) { CURL_TRC_CF(data, cf, "ngtcp2_crypto_gnutls_configure_client_session failed\n"); return CURLE_QUIC_CONNECT_ERROR; } rc = gnutls_priority_set_direct(ctx->gtls->session, QUIC_PRIORITY, NULL); if(rc < 0) { CURL_TRC_CF(data, cf, "gnutls_priority_set_direct failed: %s\n", gnutls_strerror(rc)); return CURLE_QUIC_CONNECT_ERROR; } /* Open the file if a TLS or QUIC backend has not done this before. */ Curl_tls_keylog_open(); if(Curl_tls_keylog_enabled()) { gnutls_session_set_keylog_function(ctx->gtls->session, keylog_callback); } /* strip the first byte (the length) from NGHTTP3_ALPN_H3 */ alpn[0].data = (unsigned char *)H3_ALPN_H3_29 + 1; alpn[0].size = sizeof(H3_ALPN_H3_29) - 2; alpn[1].data = (unsigned char *)H3_ALPN_H3 + 1; alpn[1].size = sizeof(H3_ALPN_H3) - 2; gnutls_alpn_set_protocols(ctx->gtls->session, alpn, 2, GNUTLS_ALPN_MANDATORY); return CURLE_OK; } #elif defined(USE_WOLFSSL) static CURLcode quic_ssl_ctx(WOLFSSL_CTX **pssl_ctx, struct Curl_cfilter *cf, struct Curl_easy *data) { CURLcode result = CURLE_FAILED_INIT; struct ssl_primary_config *conn_config; WOLFSSL_CTX *ssl_ctx = NULL; conn_config = Curl_ssl_cf_get_primary_config(cf); if(!conn_config) { result = CURLE_FAILED_INIT; goto out; } ssl_ctx = wolfSSL_CTX_new(wolfTLSv1_3_client_method()); if(!ssl_ctx) { result = CURLE_OUT_OF_MEMORY; goto out; } if(ngtcp2_crypto_wolfssl_configure_client_context(ssl_ctx) != 0) { failf(data, "ngtcp2_crypto_wolfssl_configure_client_context failed"); result = CURLE_FAILED_INIT; goto out; } wolfSSL_CTX_set_default_verify_paths(ssl_ctx); if(wolfSSL_CTX_set_cipher_list(ssl_ctx, conn_config->cipher_list13 ? conn_config->cipher_list13 : QUIC_CIPHERS) != 1) { char error_buffer[256]; ERR_error_string_n(ERR_get_error(), error_buffer, sizeof(error_buffer)); failf(data, "wolfSSL failed to set ciphers: %s", error_buffer); goto out; } if(wolfSSL_CTX_set1_groups_list(ssl_ctx, conn_config->curves ? conn_config->curves : (char *)QUIC_GROUPS) != 1) { failf(data, "wolfSSL failed to set curves"); goto out; } /* Open the file if a TLS or QUIC backend has not done this before. */ Curl_tls_keylog_open(); if(Curl_tls_keylog_enabled()) { #if defined(HAVE_SECRET_CALLBACK) wolfSSL_CTX_set_keylog_callback(ssl_ctx, keylog_callback); #else failf(data, "wolfSSL was built without keylog callback"); goto out; #endif } if(conn_config->verifypeer) { const char * const ssl_cafile = conn_config->CAfile; const char * const ssl_capath = conn_config->CApath; wolfSSL_CTX_set_verify(ssl_ctx, SSL_VERIFY_PEER, NULL); if(ssl_cafile || ssl_capath) { /* tell wolfSSL where to find CA certificates that are used to verify the server's certificate. */ int rc = wolfSSL_CTX_load_verify_locations_ex(ssl_ctx, ssl_cafile, ssl_capath, WOLFSSL_LOAD_FLAG_IGNORE_ERR); if(SSL_SUCCESS != rc) { /* Fail if we insist on successfully verifying the server. */ failf(data, "error setting certificate verify locations:" " CAfile: %s CApath: %s", ssl_cafile ? ssl_cafile : "none", ssl_capath ? ssl_capath : "none"); goto out; } infof(data, " CAfile: %s", ssl_cafile ? ssl_cafile : "none"); infof(data, " CApath: %s", ssl_capath ? ssl_capath : "none"); } #ifdef CURL_CA_FALLBACK else { /* verifying the peer without any CA certificates won't work so use wolfssl's built-in default as fallback */ wolfSSL_CTX_set_default_verify_paths(ssl_ctx); } #endif } else { wolfSSL_CTX_set_verify(ssl_ctx, SSL_VERIFY_NONE, NULL); } /* give application a chance to interfere with SSL set up. */ if(data->set.ssl.fsslctx) { Curl_set_in_callback(data, true); result = (*data->set.ssl.fsslctx)(data, ssl_ctx, data->set.ssl.fsslctxp); Curl_set_in_callback(data, false); if(result) { failf(data, "error signaled by ssl ctx callback"); goto out; } } result = CURLE_OK; out: *pssl_ctx = result? NULL : ssl_ctx; if(result && ssl_ctx) SSL_CTX_free(ssl_ctx); return result; } /** SSL callbacks ***/ static CURLcode quic_init_ssl(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; const uint8_t *alpn = NULL; size_t alpnlen = 0; /* this will need some attention when HTTPS proxy over QUIC get fixed */ const char * const hostname = cf->conn->host.name; (void)data; DEBUGASSERT(!ctx->ssl); ctx->ssl = wolfSSL_new(ctx->sslctx); wolfSSL_set_app_data(ctx->ssl, &ctx->conn_ref); wolfSSL_set_connect_state(ctx->ssl); wolfSSL_set_quic_use_legacy_codepoint(ctx->ssl, 0); alpn = (const uint8_t *)H3_ALPN_H3_29 H3_ALPN_H3; alpnlen = sizeof(H3_ALPN_H3_29) - 1 + sizeof(H3_ALPN_H3) - 1; if(alpn) wolfSSL_set_alpn_protos(ctx->ssl, alpn, (int)alpnlen); /* set SNI */ wolfSSL_UseSNI(ctx->ssl, WOLFSSL_SNI_HOST_NAME, hostname, (unsigned short)strlen(hostname)); return CURLE_OK; } #endif /* defined(USE_WOLFSSL) */ static int cb_handshake_completed(ngtcp2_conn *tconn, void *user_data) { (void)user_data; (void)tconn; return 0; } static void report_consumed_data(struct Curl_cfilter *cf, struct Curl_easy *data, size_t consumed) { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); struct cf_ngtcp2_ctx *ctx = cf->ctx; if(!stream) return; /* the HTTP/1.1 response headers are written to the buffer, but * consuming those does not count against flow control. */ if(stream->recv_buf_nonflow) { if(consumed >= stream->recv_buf_nonflow) { consumed -= stream->recv_buf_nonflow; stream->recv_buf_nonflow = 0; } else { stream->recv_buf_nonflow -= consumed; consumed = 0; } } if(consumed > 0) { CURL_TRC_CF(data, cf, "[%" PRId64 "] ACK %zu bytes of DATA", stream->id, consumed); ngtcp2_conn_extend_max_stream_offset(ctx->qconn, stream->id, consumed); ngtcp2_conn_extend_max_offset(ctx->qconn, consumed); } } static int cb_recv_stream_data(ngtcp2_conn *tconn, uint32_t flags, int64_t stream_id, uint64_t offset, const uint8_t *buf, size_t buflen, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; nghttp3_ssize nconsumed; int fin = (flags & NGTCP2_STREAM_DATA_FLAG_FIN) ? 1 : 0; struct Curl_easy *data = stream_user_data; (void)offset; (void)data; nconsumed = nghttp3_conn_read_stream(ctx->h3conn, stream_id, buf, buflen, fin); CURL_TRC_CF(data, cf, "[%" PRId64 "] read_stream(len=%zu) -> %zd", stream_id, buflen, nconsumed); if(nconsumed < 0) { if(!data) { struct Curl_easy *cdata = CF_DATA_CURRENT(cf); CURL_TRC_CF(cdata, cf, "[%" PRId64 "] nghttp3 error on stream not " "used by us, ignored", stream_id); return 0; } ngtcp2_ccerr_set_application_error( &ctx->last_error, nghttp3_err_infer_quic_app_error_code((int)nconsumed), NULL, 0); return NGTCP2_ERR_CALLBACK_FAILURE; } /* number of bytes inside buflen which consists of framing overhead * including QPACK HEADERS. In other words, it does not consume payload of * DATA frame. */ ngtcp2_conn_extend_max_stream_offset(tconn, stream_id, nconsumed); ngtcp2_conn_extend_max_offset(tconn, nconsumed); return 0; } static int cb_acked_stream_data_offset(ngtcp2_conn *tconn, int64_t stream_id, uint64_t offset, uint64_t datalen, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; int rv; (void)stream_id; (void)tconn; (void)offset; (void)datalen; (void)stream_user_data; rv = nghttp3_conn_add_ack_offset(ctx->h3conn, stream_id, datalen); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static int cb_stream_close(ngtcp2_conn *tconn, uint32_t flags, int64_t stream3_id, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; int rv; (void)tconn; (void)data; /* stream is closed... */ if(!(flags & NGTCP2_STREAM_CLOSE_FLAG_APP_ERROR_CODE_SET)) { app_error_code = NGHTTP3_H3_NO_ERROR; } rv = nghttp3_conn_close_stream(ctx->h3conn, stream3_id, app_error_code); CURL_TRC_CF(data, cf, "[%" PRId64 "] quic close(err=%" PRIu64 ") -> %d", stream3_id, app_error_code, rv); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { ngtcp2_ccerr_set_application_error( &ctx->last_error, nghttp3_err_infer_quic_app_error_code(rv), NULL, 0); return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static int cb_stream_reset(ngtcp2_conn *tconn, int64_t stream_id, uint64_t final_size, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; struct Curl_easy *data = stream_user_data; int rv; (void)tconn; (void)final_size; (void)app_error_code; (void)data; rv = nghttp3_conn_shutdown_stream_read(ctx->h3conn, stream_id); CURL_TRC_CF(data, cf, "[%" PRId64 "] reset -> %d", stream_id, rv); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static int cb_stream_stop_sending(ngtcp2_conn *tconn, int64_t stream_id, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; int rv; (void)tconn; (void)app_error_code; (void)stream_user_data; rv = nghttp3_conn_shutdown_stream_read(ctx->h3conn, stream_id); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static int cb_extend_max_local_streams_bidi(ngtcp2_conn *tconn, uint64_t max_streams, void *user_data) { (void)tconn; (void)max_streams; (void)user_data; return 0; } static int cb_extend_max_stream_data(ngtcp2_conn *tconn, int64_t stream_id, uint64_t max_data, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; struct Curl_easy *data = CF_DATA_CURRENT(cf); struct Curl_easy *s_data; struct h3_stream_ctx *stream; int rv; (void)tconn; (void)max_data; (void)stream_user_data; rv = nghttp3_conn_unblock_stream(ctx->h3conn, stream_id); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } s_data = get_stream_easy(cf, data, stream_id); stream = H3_STREAM_CTX(s_data); if(stream && stream->quic_flow_blocked) { CURL_TRC_CF(data, cf, "[%" PRId64 "] unblock quic flow", stream_id); stream->quic_flow_blocked = FALSE; h3_drain_stream(cf, data); } return 0; } static void cb_rand(uint8_t *dest, size_t destlen, const ngtcp2_rand_ctx *rand_ctx) { CURLcode result; (void)rand_ctx; result = Curl_rand(NULL, dest, destlen); if(result) { /* cb_rand is only used for non-cryptographic context. If Curl_rand failed, just fill 0 and call it *random*. */ memset(dest, 0, destlen); } } static int cb_get_new_connection_id(ngtcp2_conn *tconn, ngtcp2_cid *cid, uint8_t *token, size_t cidlen, void *user_data) { CURLcode result; (void)tconn; (void)user_data; result = Curl_rand(NULL, cid->data, cidlen); if(result) return NGTCP2_ERR_CALLBACK_FAILURE; cid->datalen = cidlen; result = Curl_rand(NULL, token, NGTCP2_STATELESS_RESET_TOKENLEN); if(result) return NGTCP2_ERR_CALLBACK_FAILURE; return 0; } static int cb_recv_rx_key(ngtcp2_conn *tconn, ngtcp2_encryption_level level, void *user_data) { struct Curl_cfilter *cf = user_data; (void)tconn; if(level != NGTCP2_ENCRYPTION_LEVEL_1RTT) { return 0; } if(init_ngh3_conn(cf) != CURLE_OK) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static ngtcp2_callbacks ng_callbacks = { ngtcp2_crypto_client_initial_cb, NULL, /* recv_client_initial */ ngtcp2_crypto_recv_crypto_data_cb, cb_handshake_completed, NULL, /* recv_version_negotiation */ ngtcp2_crypto_encrypt_cb, ngtcp2_crypto_decrypt_cb, ngtcp2_crypto_hp_mask_cb, cb_recv_stream_data, cb_acked_stream_data_offset, NULL, /* stream_open */ cb_stream_close, NULL, /* recv_stateless_reset */ ngtcp2_crypto_recv_retry_cb, cb_extend_max_local_streams_bidi, NULL, /* extend_max_local_streams_uni */ cb_rand, cb_get_new_connection_id, NULL, /* remove_connection_id */ ngtcp2_crypto_update_key_cb, /* update_key */ NULL, /* path_validation */ NULL, /* select_preferred_addr */ cb_stream_reset, NULL, /* extend_max_remote_streams_bidi */ NULL, /* extend_max_remote_streams_uni */ cb_extend_max_stream_data, NULL, /* dcid_status */ NULL, /* handshake_confirmed */ NULL, /* recv_new_token */ ngtcp2_crypto_delete_crypto_aead_ctx_cb, ngtcp2_crypto_delete_crypto_cipher_ctx_cb, NULL, /* recv_datagram */ NULL, /* ack_datagram */ NULL, /* lost_datagram */ ngtcp2_crypto_get_path_challenge_data_cb, cb_stream_stop_sending, NULL, /* version_negotiation */ cb_recv_rx_key, NULL, /* recv_tx_key */ NULL, /* early_data_rejected */ }; /** * Connection maintenance like timeouts on packet ACKs etc. are done by us, not * the OS like for TCP. POLL events on the socket therefore are not * sufficient. * ngtcp2 tells us when it wants to be invoked again. We handle that via * the `Curl_expire()` mechanisms. */ static CURLcode check_and_set_expiry(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct pkt_io_ctx local_pktx; ngtcp2_tstamp expiry; if(!pktx) { pktx_init(&local_pktx, cf, data); pktx = &local_pktx; } else { pktx_update_time(pktx, cf); } expiry = ngtcp2_conn_get_expiry(ctx->qconn); if(expiry != UINT64_MAX) { if(expiry <= pktx->ts) { CURLcode result; int rv = ngtcp2_conn_handle_expiry(ctx->qconn, pktx->ts); if(rv) { failf(data, "ngtcp2_conn_handle_expiry returned error: %s", ngtcp2_strerror(rv)); ngtcp2_ccerr_set_liberr(&ctx->last_error, rv, NULL, 0); return CURLE_SEND_ERROR; } result = cf_progress_ingress(cf, data, pktx); if(result) return result; result = cf_progress_egress(cf, data, pktx); if(result) return result; /* ask again, things might have changed */ expiry = ngtcp2_conn_get_expiry(ctx->qconn); } if(expiry > pktx->ts) { ngtcp2_duration timeout = expiry - pktx->ts; if(timeout % NGTCP2_MILLISECONDS) { timeout += NGTCP2_MILLISECONDS; } Curl_expire(data, timeout / NGTCP2_MILLISECONDS, EXPIRE_QUIC); } } return CURLE_OK; } static void cf_ngtcp2_adjust_pollset(struct Curl_cfilter *cf, struct Curl_easy *data, struct easy_pollset *ps) { struct cf_ngtcp2_ctx *ctx = cf->ctx; bool want_recv = CURL_WANT_RECV(data); bool want_send = CURL_WANT_SEND(data); if(ctx->qconn && (want_recv || want_send)) { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); struct cf_call_data save; bool c_exhaust, s_exhaust; CF_DATA_SAVE(save, cf, data); c_exhaust = !ngtcp2_conn_get_cwnd_left(ctx->qconn) || !ngtcp2_conn_get_max_data_left(ctx->qconn); s_exhaust = stream && stream->id >= 0 && stream->quic_flow_blocked; want_recv = (want_recv || c_exhaust || s_exhaust); want_send = (!s_exhaust && want_send) || !Curl_bufq_is_empty(&ctx->q.sendbuf); Curl_pollset_set(data, ps, ctx->q.sockfd, want_recv, want_send); CF_DATA_RESTORE(cf, save); } } static int cb_h3_stream_close(nghttp3_conn *conn, int64_t stream_id, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); (void)conn; (void)stream_id; /* we might be called by nghttp3 after we already cleaned up */ if(!stream) return 0; stream->closed = TRUE; stream->error3 = app_error_code; if(stream->error3 != NGHTTP3_H3_NO_ERROR) { stream->reset = TRUE; stream->send_closed = TRUE; CURL_TRC_CF(data, cf, "[%" PRId64 "] RESET: error %" PRId64, stream->id, stream->error3); } else { CURL_TRC_CF(data, cf, "[%" PRId64 "] CLOSED", stream->id); } h3_drain_stream(cf, data); return 0; } /* * write_resp_raw() copies response data in raw format to the `data`'s * receive buffer. If not enough space is available, it appends to the * `data`'s overflow buffer. */ static CURLcode write_resp_raw(struct Curl_cfilter *cf, struct Curl_easy *data, const void *mem, size_t memlen, bool flow) { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); CURLcode result = CURLE_OK; ssize_t nwritten; (void)cf; if(!stream) { return CURLE_RECV_ERROR; } nwritten = Curl_bufq_write(&stream->recvbuf, mem, memlen, &result); if(nwritten < 0) { return result; } if(!flow) stream->recv_buf_nonflow += (size_t)nwritten; if((size_t)nwritten < memlen) { /* This MUST not happen. Our recbuf is dimensioned to hold the * full max_stream_window and then some for this very reason. */ DEBUGASSERT(0); return CURLE_RECV_ERROR; } return result; } static int cb_h3_recv_data(nghttp3_conn *conn, int64_t stream3_id, const uint8_t *buf, size_t buflen, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); CURLcode result; (void)conn; (void)stream3_id; if(!stream) return NGHTTP3_ERR_CALLBACK_FAILURE; result = write_resp_raw(cf, data, buf, buflen, TRUE); if(result) { CURL_TRC_CF(data, cf, "[%" PRId64 "] DATA len=%zu, ERROR receiving %d", stream->id, buflen, result); return NGHTTP3_ERR_CALLBACK_FAILURE; } CURL_TRC_CF(data, cf, "[%" PRId64 "] DATA len=%zu", stream->id, buflen); h3_drain_stream(cf, data); return 0; } static int cb_h3_deferred_consume(nghttp3_conn *conn, int64_t stream3_id, size_t consumed, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; (void)conn; (void)stream_user_data; /* nghttp3 has consumed bytes on the QUIC stream and we need to * tell the QUIC connection to increase its flow control */ ngtcp2_conn_extend_max_stream_offset(ctx->qconn, stream3_id, consumed); ngtcp2_conn_extend_max_offset(ctx->qconn, consumed); return 0; } static int cb_h3_end_headers(nghttp3_conn *conn, int64_t stream_id, int fin, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); CURLcode result = CURLE_OK; (void)conn; (void)stream_id; (void)fin; (void)cf; if(!stream) return 0; /* add a CRLF only if we've received some headers */ result = write_resp_raw(cf, data, "\r\n", 2, FALSE); if(result) { return -1; } CURL_TRC_CF(data, cf, "[%" PRId64 "] end_headers, status=%d", stream_id, stream->status_code); if(stream->status_code / 100 != 1) { stream->resp_hds_complete = TRUE; } h3_drain_stream(cf, data); return 0; } static int cb_h3_recv_header(nghttp3_conn *conn, int64_t stream_id, int32_t token, nghttp3_rcbuf *name, nghttp3_rcbuf *value, uint8_t flags, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; nghttp3_vec h3name = nghttp3_rcbuf_get_buf(name); nghttp3_vec h3val = nghttp3_rcbuf_get_buf(value); struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); CURLcode result = CURLE_OK; (void)conn; (void)stream_id; (void)token; (void)flags; (void)cf; /* we might have cleaned up this transfer already */ if(!stream) return 0; if(token == NGHTTP3_QPACK_TOKEN__STATUS) { char line[14]; /* status line is always 13 characters long */ size_t ncopy; result = Curl_http_decode_status(&stream->status_code, (const char *)h3val.base, h3val.len); if(result) return -1; ncopy = msnprintf(line, sizeof(line), "HTTP/3 %03d \r\n", stream->status_code); CURL_TRC_CF(data, cf, "[%" PRId64 "] status: %s", stream_id, line); result = write_resp_raw(cf, data, line, ncopy, FALSE); if(result) { return -1; } } else { /* store as an HTTP1-style header */ CURL_TRC_CF(data, cf, "[%" PRId64 "] header: %.*s: %.*s", stream_id, (int)h3name.len, h3name.base, (int)h3val.len, h3val.base); result = write_resp_raw(cf, data, h3name.base, h3name.len, FALSE); if(result) { return -1; } result = write_resp_raw(cf, data, ": ", 2, FALSE); if(result) { return -1; } result = write_resp_raw(cf, data, h3val.base, h3val.len, FALSE); if(result) { return -1; } result = write_resp_raw(cf, data, "\r\n", 2, FALSE); if(result) { return -1; } } return 0; } static int cb_h3_stop_sending(nghttp3_conn *conn, int64_t stream_id, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; int rv; (void)conn; (void)stream_user_data; rv = ngtcp2_conn_shutdown_stream_read(ctx->qconn, 0, stream_id, app_error_code); if(rv && rv != NGTCP2_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static int cb_h3_reset_stream(nghttp3_conn *conn, int64_t stream_id, uint64_t app_error_code, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct cf_ngtcp2_ctx *ctx = cf->ctx; struct Curl_easy *data = stream_user_data; int rv; (void)conn; (void)data; rv = ngtcp2_conn_shutdown_stream_write(ctx->qconn, 0, stream_id, app_error_code); CURL_TRC_CF(data, cf, "[%" PRId64 "] reset -> %d", stream_id, rv); if(rv && rv != NGTCP2_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } return 0; } static nghttp3_callbacks ngh3_callbacks = { cb_h3_acked_req_body, /* acked_stream_data */ cb_h3_stream_close, cb_h3_recv_data, cb_h3_deferred_consume, NULL, /* begin_headers */ cb_h3_recv_header, cb_h3_end_headers, NULL, /* begin_trailers */ cb_h3_recv_header, NULL, /* end_trailers */ cb_h3_stop_sending, NULL, /* end_stream */ cb_h3_reset_stream, NULL, /* shutdown */ NULL /* recv_settings */ }; static int init_ngh3_conn(struct Curl_cfilter *cf) { struct cf_ngtcp2_ctx *ctx = cf->ctx; CURLcode result; int rc; int64_t ctrl_stream_id, qpack_enc_stream_id, qpack_dec_stream_id; if(ngtcp2_conn_get_streams_uni_left(ctx->qconn) < 3) { return CURLE_QUIC_CONNECT_ERROR; } nghttp3_settings_default(&ctx->h3settings); rc = nghttp3_conn_client_new(&ctx->h3conn, &ngh3_callbacks, &ctx->h3settings, nghttp3_mem_default(), cf); if(rc) { result = CURLE_OUT_OF_MEMORY; goto fail; } rc = ngtcp2_conn_open_uni_stream(ctx->qconn, &ctrl_stream_id, NULL); if(rc) { result = CURLE_QUIC_CONNECT_ERROR; goto fail; } rc = nghttp3_conn_bind_control_stream(ctx->h3conn, ctrl_stream_id); if(rc) { result = CURLE_QUIC_CONNECT_ERROR; goto fail; } rc = ngtcp2_conn_open_uni_stream(ctx->qconn, &qpack_enc_stream_id, NULL); if(rc) { result = CURLE_QUIC_CONNECT_ERROR; goto fail; } rc = ngtcp2_conn_open_uni_stream(ctx->qconn, &qpack_dec_stream_id, NULL); if(rc) { result = CURLE_QUIC_CONNECT_ERROR; goto fail; } rc = nghttp3_conn_bind_qpack_streams(ctx->h3conn, qpack_enc_stream_id, qpack_dec_stream_id); if(rc) { result = CURLE_QUIC_CONNECT_ERROR; goto fail; } return CURLE_OK; fail: return result; } static ssize_t recv_closed_stream(struct Curl_cfilter *cf, struct Curl_easy *data, struct h3_stream_ctx *stream, CURLcode *err) { ssize_t nread = -1; (void)cf; if(stream->reset) { failf(data, "HTTP/3 stream %" PRId64 " reset by server", stream->id); *err = stream->resp_hds_complete? CURLE_PARTIAL_FILE : CURLE_HTTP3; goto out; } else if(!stream->resp_hds_complete) { failf(data, "HTTP/3 stream %" PRId64 " was closed cleanly, but before getting" " all response header fields, treated as error", stream->id); *err = CURLE_HTTP3; goto out; } *err = CURLE_OK; nread = 0; out: return nread; } /* incoming data frames on the h3 stream */ static ssize_t cf_ngtcp2_recv(struct Curl_cfilter *cf, struct Curl_easy *data, char *buf, size_t len, CURLcode *err) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); ssize_t nread = -1; struct cf_call_data save; struct pkt_io_ctx pktx; (void)ctx; CF_DATA_SAVE(save, cf, data); DEBUGASSERT(cf->connected); DEBUGASSERT(ctx); DEBUGASSERT(ctx->qconn); DEBUGASSERT(ctx->h3conn); *err = CURLE_OK; pktx_init(&pktx, cf, data); if(!stream) { *err = CURLE_RECV_ERROR; goto out; } if(!Curl_bufq_is_empty(&stream->recvbuf)) { nread = Curl_bufq_read(&stream->recvbuf, (unsigned char *)buf, len, err); if(nread < 0) { CURL_TRC_CF(data, cf, "[%" PRId64 "] read recvbuf(len=%zu) " "-> %zd, %d", stream->id, len, nread, *err); goto out; } report_consumed_data(cf, data, nread); } if(cf_progress_ingress(cf, data, &pktx)) { *err = CURLE_RECV_ERROR; nread = -1; goto out; } /* recvbuf had nothing before, maybe after progressing ingress? */ if(nread < 0 && !Curl_bufq_is_empty(&stream->recvbuf)) { nread = Curl_bufq_read(&stream->recvbuf, (unsigned char *)buf, len, err); if(nread < 0) { CURL_TRC_CF(data, cf, "[%" PRId64 "] read recvbuf(len=%zu) " "-> %zd, %d", stream->id, len, nread, *err); goto out; } report_consumed_data(cf, data, nread); } if(nread > 0) { h3_drain_stream(cf, data); } else { if(stream->closed) { nread = recv_closed_stream(cf, data, stream, err); goto out; } *err = CURLE_AGAIN; nread = -1; } out: if(cf_progress_egress(cf, data, &pktx)) { *err = CURLE_SEND_ERROR; nread = -1; } else { CURLcode result2 = check_and_set_expiry(cf, data, &pktx); if(result2) { *err = result2; nread = -1; } } CURL_TRC_CF(data, cf, "[%" PRId64 "] cf_recv(len=%zu) -> %zd, %d", stream? stream->id : -1, len, nread, *err); CF_DATA_RESTORE(cf, save); return nread; } static int cb_h3_acked_req_body(nghttp3_conn *conn, int64_t stream_id, uint64_t datalen, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); size_t skiplen; (void)cf; if(!stream) return 0; /* The server acknowledged `datalen` of bytes from our request body. * This is a delta. We have kept this data in `sendbuf` for * re-transmissions and can free it now. */ if(datalen >= (uint64_t)stream->sendbuf_len_in_flight) skiplen = stream->sendbuf_len_in_flight; else skiplen = (size_t)datalen; Curl_bufq_skip(&stream->sendbuf, skiplen); stream->sendbuf_len_in_flight -= skiplen; /* Everything ACKed, we resume upload processing */ if(!stream->sendbuf_len_in_flight) { int rv = nghttp3_conn_resume_stream(conn, stream_id); if(rv && rv != NGHTTP3_ERR_STREAM_NOT_FOUND) { return NGTCP2_ERR_CALLBACK_FAILURE; } } return 0; } static nghttp3_ssize cb_h3_read_req_body(nghttp3_conn *conn, int64_t stream_id, nghttp3_vec *vec, size_t veccnt, uint32_t *pflags, void *user_data, void *stream_user_data) { struct Curl_cfilter *cf = user_data; struct Curl_easy *data = stream_user_data; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); ssize_t nwritten = 0; size_t nvecs = 0; (void)cf; (void)conn; (void)stream_id; (void)user_data; (void)veccnt; if(!stream) return NGHTTP3_ERR_CALLBACK_FAILURE; /* nghttp3 keeps references to the sendbuf data until it is ACKed * by the server (see `cb_h3_acked_req_body()` for updates). * `sendbuf_len_in_flight` is the amount of bytes in `sendbuf` * that we have already passed to nghttp3, but which have not been * ACKed yet. * Any amount beyond `sendbuf_len_in_flight` we need still to pass * to nghttp3. Do that now, if we can. */ if(stream->sendbuf_len_in_flight < Curl_bufq_len(&stream->sendbuf)) { nvecs = 0; while(nvecs < veccnt && Curl_bufq_peek_at(&stream->sendbuf, stream->sendbuf_len_in_flight, (const unsigned char **)&vec[nvecs].base, &vec[nvecs].len)) { stream->sendbuf_len_in_flight += vec[nvecs].len; nwritten += vec[nvecs].len; ++nvecs; } DEBUGASSERT(nvecs > 0); /* we SHOULD have been be able to peek */ } if(nwritten > 0 && stream->upload_left != -1) stream->upload_left -= nwritten; /* When we stopped sending and everything in `sendbuf` is "in flight", * we are at the end of the request body. */ if(stream->upload_left == 0) { *pflags = NGHTTP3_DATA_FLAG_EOF; stream->send_closed = TRUE; } else if(!nwritten) { /* Not EOF, and nothing to give, we signal WOULDBLOCK. */ CURL_TRC_CF(data, cf, "[%" PRId64 "] read req body -> AGAIN", stream->id); return NGHTTP3_ERR_WOULDBLOCK; } CURL_TRC_CF(data, cf, "[%" PRId64 "] read req body -> " "%d vecs%s with %zu (buffered=%zu, left=%" CURL_FORMAT_CURL_OFF_T ")", stream->id, (int)nvecs, *pflags == NGHTTP3_DATA_FLAG_EOF?" EOF":"", nwritten, Curl_bufq_len(&stream->sendbuf), stream->upload_left); return (nghttp3_ssize)nvecs; } /* Index where :authority header field will appear in request header field list. */ #define AUTHORITY_DST_IDX 3 static ssize_t h3_stream_open(struct Curl_cfilter *cf, struct Curl_easy *data, const void *buf, size_t len, CURLcode *err) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct h3_stream_ctx *stream = NULL; struct dynhds h2_headers; size_t nheader; nghttp3_nv *nva = NULL; int rc = 0; unsigned int i; ssize_t nwritten = -1; nghttp3_data_reader reader; nghttp3_data_reader *preader = NULL; Curl_dynhds_init(&h2_headers, 0, DYN_HTTP_REQUEST); *err = h3_data_setup(cf, data); if(*err) goto out; stream = H3_STREAM_CTX(data); DEBUGASSERT(stream); if(!stream) { *err = CURLE_FAILED_INIT; goto out; } nwritten = Curl_h1_req_parse_read(&stream->h1, buf, len, NULL, 0, err); if(nwritten < 0) goto out; if(!stream->h1.done) { /* need more data */ goto out; } DEBUGASSERT(stream->h1.req); *err = Curl_http_req_to_h2(&h2_headers, stream->h1.req, data); if(*err) { nwritten = -1; goto out; } /* no longer needed */ Curl_h1_req_parse_free(&stream->h1); nheader = Curl_dynhds_count(&h2_headers); nva = malloc(sizeof(nghttp3_nv) * nheader); if(!nva) { *err = CURLE_OUT_OF_MEMORY; nwritten = -1; goto out; } for(i = 0; i < nheader; ++i) { struct dynhds_entry *e = Curl_dynhds_getn(&h2_headers, i); nva[i].name = (unsigned char *)e->name; nva[i].namelen = e->namelen; nva[i].value = (unsigned char *)e->value; nva[i].valuelen = e->valuelen; nva[i].flags = NGHTTP3_NV_FLAG_NONE; } rc = ngtcp2_conn_open_bidi_stream(ctx->qconn, &stream->id, data); if(rc) { failf(data, "can get bidi streams"); *err = CURLE_SEND_ERROR; goto out; } switch(data->state.httpreq) { case HTTPREQ_POST: case HTTPREQ_POST_FORM: case HTTPREQ_POST_MIME: case HTTPREQ_PUT: /* known request body size or -1 */ if(data->state.infilesize != -1) stream->upload_left = data->state.infilesize; else /* data sending without specifying the data amount up front */ stream->upload_left = -1; /* unknown */ break; default: /* there is not request body */ stream->upload_left = 0; /* no request body */ break; } stream->send_closed = (stream->upload_left == 0); if(!stream->send_closed) { reader.read_data = cb_h3_read_req_body; preader = &reader; } rc = nghttp3_conn_submit_request(ctx->h3conn, stream->id, nva, nheader, preader, data); if(rc) { switch(rc) { case NGHTTP3_ERR_CONN_CLOSING: CURL_TRC_CF(data, cf, "h3sid[%"PRId64"] failed to send, " "connection is closing", stream->id); break; default: CURL_TRC_CF(data, cf, "h3sid[%"PRId64"] failed to send -> %d (%s)", stream->id, rc, ngtcp2_strerror(rc)); break; } *err = CURLE_SEND_ERROR; nwritten = -1; goto out; } if(Curl_trc_is_verbose(data)) { infof(data, "[HTTP/3] [%" PRId64 "] OPENED stream for %s", stream->id, data->state.url); for(i = 0; i < nheader; ++i) { infof(data, "[HTTP/3] [%" PRId64 "] [%.*s: %.*s]", stream->id, (int)nva[i].namelen, nva[i].name, (int)nva[i].valuelen, nva[i].value); } } out: free(nva); Curl_dynhds_free(&h2_headers); return nwritten; } static ssize_t cf_ngtcp2_send(struct Curl_cfilter *cf, struct Curl_easy *data, const void *buf, size_t len, CURLcode *err) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct h3_stream_ctx *stream = H3_STREAM_CTX(data); ssize_t sent = 0; struct cf_call_data save; struct pkt_io_ctx pktx; CURLcode result; CF_DATA_SAVE(save, cf, data); DEBUGASSERT(cf->connected); DEBUGASSERT(ctx->qconn); DEBUGASSERT(ctx->h3conn); pktx_init(&pktx, cf, data); *err = CURLE_OK; result = cf_progress_ingress(cf, data, &pktx); if(result) { *err = result; sent = -1; } if(!stream || stream->id < 0) { sent = h3_stream_open(cf, data, buf, len, err); if(sent < 0) { CURL_TRC_CF(data, cf, "failed to open stream -> %d", *err); goto out; } stream = H3_STREAM_CTX(data); } else if(stream->upload_blocked_len) { /* the data in `buf` has already been submitted or added to the * buffers, but have been EAGAINed on the last invocation. */ DEBUGASSERT(len >= stream->upload_blocked_len); if(len < stream->upload_blocked_len) { /* Did we get called again with a smaller `len`? This should not * happen. We are not prepared to handle that. */ failf(data, "HTTP/3 send again with decreased length"); *err = CURLE_HTTP3; sent = -1; goto out; } sent = (ssize_t)stream->upload_blocked_len; stream->upload_blocked_len = 0; } else if(stream->closed) { if(stream->resp_hds_complete) { /* Server decided to close the stream after having sent us a final * response. This is valid if it is not interested in the request * body. This happens on 30x or 40x responses. * We silently discard the data sent, since this is not a transport * error situation. */ CURL_TRC_CF(data, cf, "[%" PRId64 "] discarding data" "on closed stream with response", stream->id); *err = CURLE_OK; sent = (ssize_t)len; goto out; } *err = CURLE_HTTP3; sent = -1; goto out; } else { sent = Curl_bufq_write(&stream->sendbuf, buf, len, err); CURL_TRC_CF(data, cf, "[%" PRId64 "] cf_send, add to " "sendbuf(len=%zu) -> %zd, %d", stream->id, len, sent, *err); if(sent < 0) { goto out; } (void)nghttp3_conn_resume_stream(ctx->h3conn, stream->id); } result = cf_progress_egress(cf, data, &pktx); if(result) { *err = result; sent = -1; } if(stream && sent > 0 && stream->sendbuf_len_in_flight) { /* We have unacknowledged DATA and cannot report success to our * caller. Instead we EAGAIN and remember how much we have already * "written" into our various internal connection buffers. */ stream->upload_blocked_len = sent; CURL_TRC_CF(data, cf, "[%" PRId64 "] cf_send(len=%zu), " "%zu bytes in flight -> EGAIN", stream->id, len, stream->sendbuf_len_in_flight); *err = CURLE_AGAIN; sent = -1; } out: result = check_and_set_expiry(cf, data, &pktx); if(result) { *err = result; sent = -1; } CURL_TRC_CF(data, cf, "[%" PRId64 "] cf_send(len=%zu) -> %zd, %d", stream? stream->id : -1, len, sent, *err); CF_DATA_RESTORE(cf, save); return sent; } static CURLcode qng_verify_peer(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct ssl_primary_config *conn_config; CURLcode result = CURLE_OK; conn_config = Curl_ssl_cf_get_primary_config(cf); if(!conn_config) return CURLE_FAILED_INIT; cf->conn->bits.multiplex = TRUE; /* at least potentially multiplexed */ cf->conn->httpversion = 30; cf->conn->bundle->multiuse = BUNDLE_MULTIPLEX; if(conn_config->verifyhost) { #ifdef USE_OPENSSL X509 *server_cert; server_cert = SSL_get1_peer_certificate(ctx->ssl); if(!server_cert) { return CURLE_PEER_FAILED_VERIFICATION; } result = Curl_ossl_verifyhost(data, cf->conn, &ctx->peer, server_cert); X509_free(server_cert); if(result) return result; #elif defined(USE_GNUTLS) result = Curl_gtls_verifyserver(data, ctx->gtls->session, conn_config, &data->set.ssl, &ctx->peer, data->set.str[STRING_SSL_PINNEDPUBLICKEY]); if(result) return result; #elif defined(USE_WOLFSSL) if(!ctx->peer.sni || wolfSSL_check_domain_name(ctx->ssl, ctx->peer.sni) == SSL_FAILURE) return CURLE_PEER_FAILED_VERIFICATION; #endif infof(data, "Verified certificate just fine"); } else infof(data, "Skipped certificate verification"); #ifdef USE_OPENSSL if(data->set.ssl.certinfo) /* asked to gather certificate info */ (void)Curl_ossl_certchain(data, ctx->ssl); #endif return result; } static CURLcode recv_pkt(const unsigned char *pkt, size_t pktlen, struct sockaddr_storage *remote_addr, socklen_t remote_addrlen, int ecn, void *userp) { struct pkt_io_ctx *pktx = userp; struct cf_ngtcp2_ctx *ctx = pktx->cf->ctx; ngtcp2_pkt_info pi; ngtcp2_path path; int rv; ++pktx->pkt_count; ngtcp2_addr_init(&path.local, (struct sockaddr *)&ctx->q.local_addr, ctx->q.local_addrlen); ngtcp2_addr_init(&path.remote, (struct sockaddr *)remote_addr, remote_addrlen); pi.ecn = (uint8_t)ecn; rv = ngtcp2_conn_read_pkt(ctx->qconn, &path, &pi, pkt, pktlen, pktx->ts); if(rv) { CURL_TRC_CF(pktx->data, pktx->cf, "ingress, read_pkt -> %s (%d)", ngtcp2_strerror(rv), rv); if(!ctx->last_error.error_code) { if(rv == NGTCP2_ERR_CRYPTO) { ngtcp2_ccerr_set_tls_alert(&ctx->last_error, ngtcp2_conn_get_tls_alert(ctx->qconn), NULL, 0); } else { ngtcp2_ccerr_set_liberr(&ctx->last_error, rv, NULL, 0); } } if(rv == NGTCP2_ERR_CRYPTO) /* this is a "TLS problem", but a failed certificate verification is a common reason for this */ return CURLE_PEER_FAILED_VERIFICATION; return CURLE_RECV_ERROR; } return CURLE_OK; } static CURLcode cf_progress_ingress(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct pkt_io_ctx local_pktx; size_t pkts_chunk = 128, i; size_t pkts_max = 10 * pkts_chunk; CURLcode result = CURLE_OK; if(!pktx) { pktx_init(&local_pktx, cf, data); pktx = &local_pktx; } else { pktx_update_time(pktx, cf); } #ifdef USE_OPENSSL if(!ctx->x509_store_setup) { result = Curl_ssl_setup_x509_store(cf, data, ctx->sslctx); if(result) return result; ctx->x509_store_setup = TRUE; } #endif for(i = 0; i < pkts_max; i += pkts_chunk) { pktx->pkt_count = 0; result = vquic_recv_packets(cf, data, &ctx->q, pkts_chunk, recv_pkt, pktx); if(result) /* error */ break; if(pktx->pkt_count < pkts_chunk) /* got less than we could */ break; /* give egress a chance before we receive more */ result = cf_progress_egress(cf, data, pktx); if(result) /* error */ break; } return result; } /** * Read a network packet to send from ngtcp2 into `buf`. * Return number of bytes written or -1 with *err set. */ static ssize_t read_pkt_to_send(void *userp, unsigned char *buf, size_t buflen, CURLcode *err) { struct pkt_io_ctx *x = userp; struct cf_ngtcp2_ctx *ctx = x->cf->ctx; nghttp3_vec vec[16]; nghttp3_ssize veccnt; ngtcp2_ssize ndatalen; uint32_t flags; int64_t stream_id; int fin; ssize_t nwritten, n; veccnt = 0; stream_id = -1; fin = 0; /* ngtcp2 may want to put several frames from different streams into * this packet. `NGTCP2_WRITE_STREAM_FLAG_MORE` tells it to do so. * When `NGTCP2_ERR_WRITE_MORE` is returned, we *need* to make * another iteration. * When ngtcp2 is happy (because it has no other frame that would fit * or it has nothing more to send), it returns the total length * of the assembled packet. This may be 0 if there was nothing to send. */ nwritten = 0; *err = CURLE_OK; for(;;) { if(ctx->h3conn && ngtcp2_conn_get_max_data_left(ctx->qconn)) { veccnt = nghttp3_conn_writev_stream(ctx->h3conn, &stream_id, &fin, vec, sizeof(vec) / sizeof(vec[0])); if(veccnt < 0) { failf(x->data, "nghttp3_conn_writev_stream returned error: %s", nghttp3_strerror((int)veccnt)); ngtcp2_ccerr_set_application_error( &ctx->last_error, nghttp3_err_infer_quic_app_error_code((int)veccnt), NULL, 0); *err = CURLE_SEND_ERROR; return -1; } } flags = NGTCP2_WRITE_STREAM_FLAG_MORE | (fin ? NGTCP2_WRITE_STREAM_FLAG_FIN : 0); n = ngtcp2_conn_writev_stream(ctx->qconn, &x->ps.path, NULL, buf, buflen, &ndatalen, flags, stream_id, (const ngtcp2_vec *)vec, veccnt, x->ts); if(n == 0) { /* nothing to send */ *err = CURLE_AGAIN; nwritten = -1; goto out; } else if(n < 0) { switch(n) { case NGTCP2_ERR_STREAM_DATA_BLOCKED: { struct h3_stream_ctx *stream = H3_STREAM_CTX(x->data); DEBUGASSERT(ndatalen == -1); nghttp3_conn_block_stream(ctx->h3conn, stream_id); CURL_TRC_CF(x->data, x->cf, "[%" PRId64 "] block quic flow", stream_id); DEBUGASSERT(stream); if(stream) stream->quic_flow_blocked = TRUE; n = 0; break; } case NGTCP2_ERR_STREAM_SHUT_WR: DEBUGASSERT(ndatalen == -1); nghttp3_conn_shutdown_stream_write(ctx->h3conn, stream_id); n = 0; break; case NGTCP2_ERR_WRITE_MORE: /* ngtcp2 wants to send more. update the flow of the stream whose data * is in the buffer and continue */ DEBUGASSERT(ndatalen >= 0); n = 0; break; default: DEBUGASSERT(ndatalen == -1); failf(x->data, "ngtcp2_conn_writev_stream returned error: %s", ngtcp2_strerror((int)n)); ngtcp2_ccerr_set_liberr(&ctx->last_error, (int)n, NULL, 0); *err = CURLE_SEND_ERROR; nwritten = -1; goto out; } } if(ndatalen >= 0) { /* we add the amount of data bytes to the flow windows */ int rv = nghttp3_conn_add_write_offset(ctx->h3conn, stream_id, ndatalen); if(rv) { failf(x->data, "nghttp3_conn_add_write_offset returned error: %s\n", nghttp3_strerror(rv)); return CURLE_SEND_ERROR; } } if(n > 0) { /* packet assembled, leave */ nwritten = n; goto out; } } out: return nwritten; } static CURLcode cf_progress_egress(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx) { struct cf_ngtcp2_ctx *ctx = cf->ctx; ssize_t nread; size_t max_payload_size, path_max_payload_size, max_pktcnt; size_t pktcnt = 0; size_t gsolen = 0; /* this disables gso until we have a clue */ CURLcode curlcode; struct pkt_io_ctx local_pktx; if(!pktx) { pktx_init(&local_pktx, cf, data); pktx = &local_pktx; } else { pktx_update_time(pktx, cf); ngtcp2_path_storage_zero(&pktx->ps); } curlcode = vquic_flush(cf, data, &ctx->q); if(curlcode) { if(curlcode == CURLE_AGAIN) { Curl_expire(data, 1, EXPIRE_QUIC); return CURLE_OK; } return curlcode; } /* In UDP, there is a maximum theoretical packet paload length and * a minimum payload length that is "guarantueed" to work. * To detect if this minimum payload can be increased, ngtcp2 sends * now and then a packet payload larger than the minimum. It that * is ACKed by the peer, both parties know that it works and * the subsequent packets can use a larger one. * This is called PMTUD (Path Maximum Transmission Unit Discovery). * Since a PMTUD might be rejected right on send, we do not want it * be followed by other packets of lesser size. Because those would * also fail then. So, if we detect a PMTUD while buffering, we flush. */ max_payload_size = ngtcp2_conn_get_max_tx_udp_payload_size(ctx->qconn); path_max_payload_size = ngtcp2_conn_get_path_max_tx_udp_payload_size(ctx->qconn); /* maximum number of packets buffered before we flush to the socket */ max_pktcnt = CURLMIN(MAX_PKT_BURST, ctx->q.sendbuf.chunk_size / max_payload_size); for(;;) { /* add the next packet to send, if any, to our buffer */ nread = Curl_bufq_sipn(&ctx->q.sendbuf, max_payload_size, read_pkt_to_send, pktx, &curlcode); if(nread < 0) { if(curlcode != CURLE_AGAIN) return curlcode; /* Nothing more to add, flush and leave */ curlcode = vquic_send(cf, data, &ctx->q, gsolen); if(curlcode) { if(curlcode == CURLE_AGAIN) { Curl_expire(data, 1, EXPIRE_QUIC); return CURLE_OK; } return curlcode; } goto out; } DEBUGASSERT(nread > 0); if(pktcnt == 0) { /* first packet in buffer. This is either of a known, "good" * payload size or it is a PMTUD. We'll see. */ gsolen = (size_t)nread; } else if((size_t)nread > gsolen || (gsolen > path_max_payload_size && (size_t)nread != gsolen)) { /* The just added packet is a PMTUD *or* the one(s) before the * just added were PMTUD and the last one is smaller. * Flush the buffer before the last add. */ curlcode = vquic_send_tail_split(cf, data, &ctx->q, gsolen, nread, nread); if(curlcode) { if(curlcode == CURLE_AGAIN) { Curl_expire(data, 1, EXPIRE_QUIC); return CURLE_OK; } return curlcode; } pktcnt = 0; continue; } if(++pktcnt >= max_pktcnt || (size_t)nread < gsolen) { /* Reached MAX_PKT_BURST *or* * the capacity of our buffer *or* * last add was shorter than the previous ones, flush */ curlcode = vquic_send(cf, data, &ctx->q, gsolen); if(curlcode) { if(curlcode == CURLE_AGAIN) { Curl_expire(data, 1, EXPIRE_QUIC); return CURLE_OK; } return curlcode; } /* pktbuf has been completely sent */ pktcnt = 0; } } out: return CURLE_OK; } /* * Called from transfer.c:data_pending to know if we should keep looping * to receive more data from the connection. */ static bool cf_ngtcp2_data_pending(struct Curl_cfilter *cf, const struct Curl_easy *data) { const struct h3_stream_ctx *stream = H3_STREAM_CTX(data); (void)cf; return stream && !Curl_bufq_is_empty(&stream->recvbuf); } static CURLcode h3_data_pause(struct Curl_cfilter *cf, struct Curl_easy *data, bool pause) { /* TODO: there seems right now no API in ngtcp2 to shrink/enlarge * the streams windows. As we do in HTTP/2. */ if(!pause) { h3_drain_stream(cf, data); Curl_expire(data, 0, EXPIRE_RUN_NOW); } return CURLE_OK; } static CURLcode cf_ngtcp2_data_event(struct Curl_cfilter *cf, struct Curl_easy *data, int event, int arg1, void *arg2) { struct cf_ngtcp2_ctx *ctx = cf->ctx; CURLcode result = CURLE_OK; struct cf_call_data save; CF_DATA_SAVE(save, cf, data); (void)arg1; (void)arg2; switch(event) { case CF_CTRL_DATA_SETUP: break; case CF_CTRL_DATA_PAUSE: result = h3_data_pause(cf, data, (arg1 != 0)); break; case CF_CTRL_DATA_DETACH: h3_data_done(cf, data); break; case CF_CTRL_DATA_DONE: h3_data_done(cf, data); break; case CF_CTRL_DATA_DONE_SEND: { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); if(stream && !stream->send_closed) { stream->send_closed = TRUE; stream->upload_left = Curl_bufq_len(&stream->sendbuf); (void)nghttp3_conn_resume_stream(ctx->h3conn, stream->id); } break; } case CF_CTRL_DATA_IDLE: { struct h3_stream_ctx *stream = H3_STREAM_CTX(data); CURL_TRC_CF(data, cf, "data idle"); if(stream && !stream->closed) { result = check_and_set_expiry(cf, data, NULL); if(result) CURL_TRC_CF(data, cf, "data idle, check_and_set_expiry -> %d", result); } break; } default: break; } CF_DATA_RESTORE(cf, save); return result; } static void cf_ngtcp2_ctx_clear(struct cf_ngtcp2_ctx *ctx) { struct cf_call_data save = ctx->call_data; if(ctx->qlogfd != -1) { close(ctx->qlogfd); } #ifdef USE_OPENSSL if(ctx->ssl) SSL_free(ctx->ssl); if(ctx->sslctx) SSL_CTX_free(ctx->sslctx); #elif defined(USE_GNUTLS) if(ctx->gtls) { if(ctx->gtls->cred) gnutls_certificate_free_credentials(ctx->gtls->cred); if(ctx->gtls->session) gnutls_deinit(ctx->gtls->session); free(ctx->gtls); } #elif defined(USE_WOLFSSL) if(ctx->ssl) wolfSSL_free(ctx->ssl); if(ctx->sslctx) wolfSSL_CTX_free(ctx->sslctx); #endif vquic_ctx_free(&ctx->q); if(ctx->h3conn) nghttp3_conn_del(ctx->h3conn); if(ctx->qconn) ngtcp2_conn_del(ctx->qconn); Curl_bufcp_free(&ctx->stream_bufcp); Curl_ssl_peer_cleanup(&ctx->peer); memset(ctx, 0, sizeof(*ctx)); ctx->qlogfd = -1; ctx->call_data = save; } static void cf_ngtcp2_close(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct cf_call_data save; CF_DATA_SAVE(save, cf, data); if(ctx && ctx->qconn) { char buffer[NGTCP2_MAX_UDP_PAYLOAD_SIZE]; struct pkt_io_ctx pktx; ngtcp2_ssize rc; CURL_TRC_CF(data, cf, "close"); pktx_init(&pktx, cf, data); rc = ngtcp2_conn_write_connection_close(ctx->qconn, NULL, /* path */ NULL, /* pkt_info */ (uint8_t *)buffer, sizeof(buffer), &ctx->last_error, pktx.ts); if(rc > 0) { while((send(ctx->q.sockfd, buffer, (SEND_TYPE_ARG3)rc, 0) == -1) && SOCKERRNO == EINTR); } cf_ngtcp2_ctx_clear(ctx); } cf->connected = FALSE; CF_DATA_RESTORE(cf, save); } static void cf_ngtcp2_destroy(struct Curl_cfilter *cf, struct Curl_easy *data) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct cf_call_data save; CF_DATA_SAVE(save, cf, data); CURL_TRC_CF(data, cf, "destroy"); if(ctx) { cf_ngtcp2_ctx_clear(ctx); free(ctx); } cf->ctx = NULL; /* No CF_DATA_RESTORE(cf, save) possible */ (void)save; } /* * Might be called twice for happy eyeballs. */ static CURLcode cf_connect_start(struct Curl_cfilter *cf, struct Curl_easy *data, struct pkt_io_ctx *pktx) { struct cf_ngtcp2_ctx *ctx = cf->ctx; int rc; int rv; CURLcode result; const struct Curl_sockaddr_ex *sockaddr = NULL; int qfd; ctx->version = NGTCP2_PROTO_VER_MAX; ctx->max_stream_window = H3_STREAM_WINDOW_SIZE; ctx->max_idle_ms = CURL_QUIC_MAX_IDLE_MS; Curl_bufcp_init(&ctx->stream_bufcp, H3_STREAM_CHUNK_SIZE, H3_STREAM_POOL_SPARES); result = Curl_ssl_peer_init(&ctx->peer, cf); if(result) return result; #ifdef USE_OPENSSL result = quic_ssl_ctx(&ctx->sslctx, cf, data); if(result) return result; result = quic_set_client_cert(cf, data); if(result) return result; #elif defined(USE_WOLFSSL) result = quic_ssl_ctx(&ctx->sslctx, cf, data); if(result) return result; #endif result = quic_init_ssl(cf, data); if(result) return result; ctx->dcid.datalen = NGTCP2_MAX_CIDLEN; result = Curl_rand(data, ctx->dcid.data, NGTCP2_MAX_CIDLEN); if(result) return result; ctx->scid.datalen = NGTCP2_MAX_CIDLEN; result = Curl_rand(data, ctx->scid.data, NGTCP2_MAX_CIDLEN); if(result) return result; (void)Curl_qlogdir(data, ctx->scid.data, NGTCP2_MAX_CIDLEN, &qfd); ctx->qlogfd = qfd; /* -1 if failure above */ quic_settings(ctx, data, pktx); result = vquic_ctx_init(&ctx->q); if(result) return result; Curl_cf_socket_peek(cf->next, data, &ctx->q.sockfd, &sockaddr, NULL, NULL, NULL, NULL); if(!sockaddr) return CURLE_QUIC_CONNECT_ERROR; ctx->q.local_addrlen = sizeof(ctx->q.local_addr); rv = getsockname(ctx->q.sockfd, (struct sockaddr *)&ctx->q.local_addr, &ctx->q.local_addrlen); if(rv == -1) return CURLE_QUIC_CONNECT_ERROR; ngtcp2_addr_init(&ctx->connected_path.local, (struct sockaddr *)&ctx->q.local_addr, ctx->q.local_addrlen); ngtcp2_addr_init(&ctx->connected_path.remote, &sockaddr->sa_addr, sockaddr->addrlen); rc = ngtcp2_conn_client_new(&ctx->qconn, &ctx->dcid, &ctx->scid, &ctx->connected_path, NGTCP2_PROTO_VER_V1, &ng_callbacks, &ctx->settings, &ctx->transport_params, NULL, cf); if(rc) return CURLE_QUIC_CONNECT_ERROR; #ifdef USE_GNUTLS ngtcp2_conn_set_tls_native_handle(ctx->qconn, ctx->gtls->session); #else ngtcp2_conn_set_tls_native_handle(ctx->qconn, ctx->ssl); #endif ngtcp2_ccerr_default(&ctx->last_error); ctx->conn_ref.get_conn = get_conn; ctx->conn_ref.user_data = cf; return CURLE_OK; } static CURLcode cf_ngtcp2_connect(struct Curl_cfilter *cf, struct Curl_easy *data, bool blocking, bool *done) { struct cf_ngtcp2_ctx *ctx = cf->ctx; CURLcode result = CURLE_OK; struct cf_call_data save; struct curltime now; struct pkt_io_ctx pktx; if(cf->connected) { *done = TRUE; return CURLE_OK; } /* Connect the UDP filter first */ if(!cf->next->connected) { result = Curl_conn_cf_connect(cf->next, data, blocking, done); if(result || !*done) return result; } *done = FALSE; now = Curl_now(); pktx_init(&pktx, cf, data); CF_DATA_SAVE(save, cf, data); if(ctx->reconnect_at.tv_sec && Curl_timediff(now, ctx->reconnect_at) < 0) { /* Not time yet to attempt the next connect */ CURL_TRC_CF(data, cf, "waiting for reconnect time"); goto out; } if(!ctx->qconn) { ctx->started_at = now; result = cf_connect_start(cf, data, &pktx); if(result) goto out; result = cf_progress_egress(cf, data, &pktx); /* we do not expect to be able to recv anything yet */ goto out; } result = cf_progress_ingress(cf, data, &pktx); if(result) goto out; result = cf_progress_egress(cf, data, &pktx); if(result) goto out; if(ngtcp2_conn_get_handshake_completed(ctx->qconn)) { ctx->handshake_at = now; CURL_TRC_CF(data, cf, "handshake complete after %dms", (int)Curl_timediff(now, ctx->started_at)); result = qng_verify_peer(cf, data); if(!result) { CURL_TRC_CF(data, cf, "peer verified"); cf->connected = TRUE; cf->conn->alpn = CURL_HTTP_VERSION_3; *done = TRUE; connkeep(cf->conn, "HTTP/3 default"); } } out: if(result == CURLE_RECV_ERROR && ctx->qconn && ngtcp2_conn_in_draining_period(ctx->qconn)) { /* When a QUIC server instance is shutting down, it may send us a * CONNECTION_CLOSE right away. Our connection then enters the DRAINING * state. The CONNECT may work in the near future again. Indicate * that as a "weird" reply. */ result = CURLE_WEIRD_SERVER_REPLY; } #ifndef CURL_DISABLE_VERBOSE_STRINGS if(result) { const char *r_ip = NULL; int r_port = 0; Curl_cf_socket_peek(cf->next, data, NULL, NULL, &r_ip, &r_port, NULL, NULL); infof(data, "QUIC connect to %s port %u failed: %s", r_ip, r_port, curl_easy_strerror(result)); } #endif if(!result && ctx->qconn) { result = check_and_set_expiry(cf, data, &pktx); } if(result || *done) CURL_TRC_CF(data, cf, "connect -> %d, done=%d", result, *done); CF_DATA_RESTORE(cf, save); return result; } static CURLcode cf_ngtcp2_query(struct Curl_cfilter *cf, struct Curl_easy *data, int query, int *pres1, void *pres2) { struct cf_ngtcp2_ctx *ctx = cf->ctx; struct cf_call_data save; switch(query) { case CF_QUERY_MAX_CONCURRENT: { const ngtcp2_transport_params *rp; DEBUGASSERT(pres1); CF_DATA_SAVE(save, cf, data); rp = ngtcp2_conn_get_remote_transport_params(ctx->qconn); if(rp) *pres1 = (rp->initial_max_streams_bidi > INT_MAX)? INT_MAX : (int)rp->initial_max_streams_bidi; else /* not arrived yet? */ *pres1 = Curl_multi_max_concurrent_streams(data->multi); CURL_TRC_CF(data, cf, "query max_conncurrent -> %d", *pres1); CF_DATA_RESTORE(cf, save); return CURLE_OK; } case CF_QUERY_CONNECT_REPLY_MS: if(ctx->got_first_byte) { timediff_t ms = Curl_timediff(ctx->first_byte_at, ctx->started_at); *pres1 = (ms < INT_MAX)? (int)ms : INT_MAX; } else *pres1 = -1; return CURLE_OK; case CF_QUERY_TIMER_CONNECT: { struct curltime *when = pres2; if(ctx->got_first_byte) *when = ctx->first_byte_at; return CURLE_OK; } case CF_QUERY_TIMER_APPCONNECT: { struct curltime *when = pres2; if(cf->connected) *when = ctx->handshake_at; return CURLE_OK; } default: break; } return cf->next? cf->next->cft->query(cf->next, data, query, pres1, pres2) : CURLE_UNKNOWN_OPTION; } static bool cf_ngtcp2_conn_is_alive(struct Curl_cfilter *cf, struct Curl_easy *data, bool *input_pending) { struct cf_ngtcp2_ctx *ctx = cf->ctx; bool alive = FALSE; const ngtcp2_transport_params *rp; struct cf_call_data save; CF_DATA_SAVE(save, cf, data); *input_pending = FALSE; if(!ctx->qconn) goto out; /* Both sides of the QUIC connection announce they max idle times in * the transport parameters. Look at the minimum of both and if * we exceed this, regard the connection as dead. The other side * may have completely purged it and will no longer respond * to any packets from us. */ rp = ngtcp2_conn_get_remote_transport_params(ctx->qconn); if(rp) { timediff_t idletime; uint64_t idle_ms = ctx->max_idle_ms; if(rp->max_idle_timeout && (rp->max_idle_timeout / NGTCP2_MILLISECONDS) < idle_ms) idle_ms = (rp->max_idle_timeout / NGTCP2_MILLISECONDS); idletime = Curl_timediff(Curl_now(), ctx->q.last_io); if(idletime > 0 && (uint64_t)idletime > idle_ms) goto out; } if(!cf->next || !cf->next->cft->is_alive(cf->next, data, input_pending)) goto out; alive = TRUE; if(*input_pending) { CURLcode result; /* This happens before we've sent off a request and the connection is not in use by any other transfer, there shouldn't be any data here, only "protocol frames" */ *input_pending = FALSE; result = cf_progress_ingress(cf, data, NULL); CURL_TRC_CF(data, cf, "is_alive, progress ingress -> %d", result); alive = result? FALSE : TRUE; } out: CF_DATA_RESTORE(cf, save); return alive; } struct Curl_cftype Curl_cft_http3 = { "HTTP/3", CF_TYPE_IP_CONNECT | CF_TYPE_SSL | CF_TYPE_MULTIPLEX, 0, cf_ngtcp2_destroy, cf_ngtcp2_connect, cf_ngtcp2_close, Curl_cf_def_get_host, cf_ngtcp2_adjust_pollset, cf_ngtcp2_data_pending, cf_ngtcp2_send, cf_ngtcp2_recv, cf_ngtcp2_data_event, cf_ngtcp2_conn_is_alive, Curl_cf_def_conn_keep_alive, cf_ngtcp2_query, }; CURLcode Curl_cf_ngtcp2_create(struct Curl_cfilter **pcf, struct Curl_easy *data, struct connectdata *conn, const struct Curl_addrinfo *ai) { struct cf_ngtcp2_ctx *ctx = NULL; struct Curl_cfilter *cf = NULL, *udp_cf = NULL; CURLcode result; (void)data; ctx = calloc(1, sizeof(*ctx)); if(!ctx) { result = CURLE_OUT_OF_MEMORY; goto out; } ctx->qlogfd = -1; cf_ngtcp2_ctx_clear(ctx); result = Curl_cf_create(&cf, &Curl_cft_http3, ctx); if(result) goto out; result = Curl_cf_udp_create(&udp_cf, data, conn, ai, TRNSPRT_QUIC); if(result) goto out; cf->conn = conn; udp_cf->conn = cf->conn; udp_cf->sockindex = cf->sockindex; cf->next = udp_cf; out: *pcf = (!result)? cf : NULL; if(result) { if(udp_cf) Curl_conn_cf_discard_sub(cf, udp_cf, data, TRUE); Curl_safefree(cf); Curl_safefree(ctx); } return result; } bool Curl_conn_is_ngtcp2(const struct Curl_easy *data, const struct connectdata *conn, int sockindex) { struct Curl_cfilter *cf = conn? conn->cfilter[sockindex] : NULL; (void)data; for(; cf; cf = cf->next) { if(cf->cft == &Curl_cft_http3) return TRUE; if(cf->cft->flags & CF_TYPE_IP_CONNECT) return FALSE; } return FALSE; } #endif