diff --git a/STYLE_GUIDE.md b/STYLE_GUIDE.md new file mode 100644 index 00000000..8029b00b --- /dev/null +++ b/STYLE_GUIDE.md @@ -0,0 +1,76 @@ +# Kronosnet Style Guide + +This document outlines the coding style and conventions to be followed when contributing to the kronosnet project. + +## General Principles + +* Clarity: Code should be easy to read and understand. +* Consistency: Follow the established style throughout the codebase. +* Simplicity: Prefer simple solutions over complex ones. + +## C Language Style + +### Minimum Supported C Dialect +All C code should conform to the C99 standard. Do not use compiler-specific extensions unless absolutely necessary and properly guarded with preprocessor directives. + +### Indentation +* Use Tabs for indentation. Do not use spaces. + +### Curly Braces +* Place the opening curly brace on the same line as control flow statements (e.g., `if`, `for`, `while`). For function definitions, the opening curly brace should be on the next line. +* Always use curly braces for `if`, `for`, `while`, and `do-while` statements, even if the body is a single line. + +**Example:** +```c +if (condition) { + do_something(); +} + +for (int i = 0; i < 10; i++) { + process_item(i); +} +``` + +### Variable Declaration +* Declare variables at the beginning of their scope (e.g., at the start of a function or a block). + +**Example:** +```c +void my_function() +{ + int count; + char *name; + + // ... function logic ... +} +``` + +### Naming Conventions +* Use `snake_case` for variable names and function names. (e.g., `user_input`, `calculate_total_sum`). +* Public API elements (functions, structs, typedefs, etc.) must be prefixed according to the library they belong to. Use `knet_` for items in the Kronosnet core library and `nozzle_` for items in the Nozzle library. +* Public enums and defines (macros) must be prefixed with `KNET_` or `NOZZLE_` (all uppercase), followed by an uppercase `SNAKE_CASE` name that describes their purpose (e.g., `KNET_MAX_CLIENTS`, `NOZZLE_BUFFER_SIZE_DEFAULT`). +* Internal (non-public) functions, structs, enums, unions, and macros should have names that are descriptive, clearly indicating their purpose and the subsystem they belong to. Generally, no specific prefix is mandated for these internal elements. +* Functions that are designed to be shared and called from multiple different threads must be prefixed with a single underscore (`_`). For example: `_shared_resource_access()`. + +## Line Length +Preferred maximum line length is 120 characters. + +While this is a preference, it is understood that this limit may be exceeded in certain situations for better readability, such as with deeply nested structures or long string literals. + +## Comments +* Use `//` for single-line comments. +* Use `/* ... */` for single-line and multi-line comments. +* Write clear and concise comments to explain non-obvious code. + +## Best Practices +* API Changes Require Tests: Any modification to an internal or external API must be accompanied by new or updated tests in the project's test suite. These tests must validate the behavior of the changed API. + +--- +## Copyright + +Copyright (C) 2025 Red Hat, Inc. All rights reserved. + +Author: Jules +AI Driver Seat: Fabio M. Di Nitto + +This software licensed under GPL-2.0+ diff --git a/libknet/Makefile.am b/libknet/Makefile.am index 7fad9694..3790c3a8 100644 --- a/libknet/Makefile.am +++ b/libknet/Makefile.am @@ -1,170 +1,170 @@ # # Copyright (C) 2010-2025 Red Hat, Inc. All rights reserved. # # Authors: Fabio M. Di Nitto # Federico Simoncelli # # This software licensed under GPL-2.0+ # MAINTAINERCLEANFILES = Makefile.in include $(top_srcdir)/build-aux/check.mk SYMFILE = libknet_exported_syms EXTRA_DIST = $(SYMFILE) SUBDIRS = . tests # https://www.gnu.org/software/libtool/manual/html_node/Updating-version-info.html -libversion = 5:0:4 +libversion = 6:0:5 # override global LIBS that pulls in lots of craft we don't need here LIBS = sources = \ common.c \ compat.c \ compress.c \ crypto.c \ handle.c \ handle_api.c \ host.c \ lib_config.c \ links.c \ links_acl.c \ links_acl_ip.c \ links_acl_loopback.c \ logging.c \ netutils.c \ onwire.c \ threads_common.c \ threads_dsthandler.c \ threads_heartbeat.c \ threads_pmtud.c \ threads_rx.c \ threads_tx.c \ transports.c \ transport_common.c \ transport_loopback.c \ transport_udp.c \ transport_sctp.c include_HEADERS = libknet.h pkgconfigdir = $(libdir)/pkgconfig pkgconfig_DATA = libknet.pc noinst_HEADERS = \ common.h \ compat.h \ compress.h \ compress_model.h \ crypto.h \ crypto_model.h \ host.h \ internals.h \ links.h \ links_acl.h \ links_acl_ip.h \ links_acl_loopback.h \ logging.h \ netutils.h \ onwire.h \ threads_common.h \ threads_dsthandler.h \ threads_heartbeat.h \ threads_pmtud.h \ threads_rx.h \ threads_tx.h \ transports.h \ transport_common.h \ transport_loopback.h \ transport_udp.h \ transport_sctp.h lib_LTLIBRARIES = libknet.la libknet_la_SOURCES = $(sources) AM_CFLAGS += $(libqb_CFLAGS) libknet_la_CFLAGS = $(AM_CFLAGS) $(PTHREAD_CFLAGS) \ -DPLUGINPATH="\"$(pkglibdir)\"" EXTRA_libknet_la_DEPENDENCIES = $(SYMFILE) libknet_la_LDFLAGS = $(AM_LDFLAGS) \ -Wl,--version-script=$(srcdir)/$(SYMFILE) \ -version-info $(libversion) libknet_la_LIBADD = $(PTHREAD_LIBS) $(dl_LIBS) $(rt_LIBS) $(m_LIBS) check-local: check-annocheck-libs # Prepare empty value for appending pkglib_LTLIBRARIES = # MODULE_LDFLAGS would mean a target-specific variable for Automake MODULELDFLAGS = $(AM_LDFLAGS) -module -avoid-version -export-dynamic if BUILD_COMPRESS_ZSTD pkglib_LTLIBRARIES += compress_zstd.la compress_zstd_la_LDFLAGS = $(MODULELDFLAGS) compress_zstd_la_CFLAGS = $(AM_CFLAGS) $(libzstd_CFLAGS) compress_zstd_la_LIBADD = $(libzstd_LIBS) endif if BUILD_COMPRESS_ZLIB pkglib_LTLIBRARIES += compress_zlib.la compress_zlib_la_LDFLAGS = $(MODULELDFLAGS) compress_zlib_la_CFLAGS = $(AM_CFLAGS) $(zlib_CFLAGS) compress_zlib_la_LIBADD = $(zlib_LIBS) endif if BUILD_COMPRESS_LZ4 pkglib_LTLIBRARIES += compress_lz4.la compress_lz4hc.la compress_lz4_la_LDFLAGS = $(MODULELDFLAGS) compress_lz4_la_CFLAGS = $(AM_CFLAGS) $(liblz4_CFLAGS) compress_lz4_la_LIBADD = $(liblz4_LIBS) compress_lz4hc_la_LDFLAGS = $(MODULELDFLAGS) compress_lz4hc_la_CFLAGS = $(AM_CFLAGS) $(liblz4_CFLAGS) compress_lz4hc_la_LIBADD = $(liblz4_LIBS) endif if BUILD_COMPRESS_LZO2 pkglib_LTLIBRARIES += compress_lzo2.la compress_lzo2_la_LDFLAGS = $(MODULELDFLAGS) compress_lzo2_la_CFLAGS = $(AM_CFLAGS) $(lzo2_CFLAGS) compress_lzo2_la_LIBADD = $(lzo2_LIBS) endif if BUILD_COMPRESS_LZMA pkglib_LTLIBRARIES += compress_lzma.la compress_lzma_la_LDFLAGS = $(MODULELDFLAGS) compress_lzma_la_CFLAGS = $(AM_CFLAGS) $(liblzma_CFLAGS) compress_lzma_la_LIBADD = $(liblzma_LIBS) endif if BUILD_COMPRESS_BZIP2 pkglib_LTLIBRARIES += compress_bzip2.la compress_bzip2_la_LDFLAGS = $(MODULELDFLAGS) compress_bzip2_la_CFLAGS = $(AM_CFLAGS) $(bzip2_CFLAGS) compress_bzip2_la_LIBADD = $(bzip2_LIBS) endif if BUILD_CRYPTO_NSS pkglib_LTLIBRARIES += crypto_nss.la crypto_nss_la_LDFLAGS = $(MODULELDFLAGS) crypto_nss_la_CFLAGS = $(AM_CFLAGS) $(nss_CFLAGS) crypto_nss_la_LIBADD = $(nss_LIBS) endif if BUILD_CRYPTO_OPENSSL pkglib_LTLIBRARIES += crypto_openssl.la crypto_openssl_la_LDFLAGS = $(MODULELDFLAGS) crypto_openssl_la_CFLAGS = $(AM_CFLAGS) $(openssl_CFLAGS) crypto_openssl_la_LIBADD = $(openssl_LIBS) endif diff --git a/libknet/crypto_openssl.c b/libknet/crypto_openssl.c index 5eb078a5..8daea8b0 100644 --- a/libknet/crypto_openssl.c +++ b/libknet/crypto_openssl.c @@ -1,729 +1,752 @@ /* * Copyright (C) 2017-2025 Red Hat, Inc. All rights reserved. * * Author: Fabio M. Di Nitto * * This software licensed under LGPL-2.0+ */ #define KNET_MODULE #include "config.h" #include #include #include #include #include #include #if (OPENSSL_VERSION_NUMBER < 0x30000000L) #include #endif #include #include #include "logging.h" #include "crypto_model.h" /* * 1.0.2 requires at least 120 bytes * 1.1.0 requires at least 256 bytes */ #define SSLERR_BUF_SIZE 512 /* * crypto definitions and conversion tables */ #define SALT_SIZE 16 /* * required by OSSL_PARAM_construct_* * making them global and cost, saves 2 strncpy and some memory on each config */ #if (OPENSSL_VERSION_NUMBER >= 0x30000000L) static const char *hash = "digest"; #endif struct opensslcrypto_instance { void *private_key; int private_key_len; const EVP_CIPHER *crypto_cipher_type; const EVP_MD *crypto_hash_type; #if (OPENSSL_VERSION_NUMBER >= 0x30000000L) EVP_MAC *crypto_hash_mac; OSSL_PARAM params[3]; char hash_type[16]; /* Need to store a copy from knet_handle_crypto_cfg for OSSL_PARAM_construct_* */ #endif }; static int openssl_is_init = 0; /* * crypt/decrypt functions openssl1.0 */ #if (OPENSSL_VERSION_NUMBER < 0x10100000L) static int encrypt_openssl( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const struct iovec *iov, int iovcnt, unsigned char *buf_out, ssize_t *buf_out_len) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; EVP_CIPHER_CTX ctx; int tmplen = 0, offset = 0; unsigned char *salt = buf_out; unsigned char *data = buf_out + SALT_SIZE; int err = 0; int i; char sslerr[SSLERR_BUF_SIZE]; EVP_CIPHER_CTX_init(&ctx); if (!RAND_bytes(salt, SALT_SIZE)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to get random salt data: %s", sslerr); err = -1; goto out; } /* * add warning re keylength */ EVP_EncryptInit_ex(&ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt); for (i=0; imodel_instance; EVP_CIPHER_CTX ctx; int tmplen1 = 0, tmplen2 = 0; unsigned char *salt = (unsigned char *)buf_in; unsigned char *data = salt + SALT_SIZE; int datalen = buf_in_len - SALT_SIZE; int err = 0; char sslerr[SSLERR_BUF_SIZE]; EVP_CIPHER_CTX_init(&ctx); /* * add warning re keylength */ EVP_DecryptInit_ex(&ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt); if (!EVP_DecryptUpdate(&ctx, buf_out, &tmplen1, data, datalen)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to decrypt: %s", sslerr); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to decrypt: %s", sslerr); } err = -1; goto out; } if (!EVP_DecryptFinal_ex(&ctx, buf_out + tmplen1, &tmplen2)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to finalize decrypt: %s", sslerr); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to finalize decrypt: %s", sslerr); } err = -1; goto out; } *buf_out_len = tmplen1 + tmplen2; out: EVP_CIPHER_CTX_cleanup(&ctx); return err; } #else /* (OPENSSL_VERSION_NUMBER < 0x10100000L) */ static int encrypt_openssl( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const struct iovec *iov, int iovcnt, unsigned char *buf_out, ssize_t *buf_out_len) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; EVP_CIPHER_CTX *ctx; int tmplen = 0, offset = 0; unsigned char *salt = buf_out; unsigned char *data = buf_out + SALT_SIZE; int err = 0; int i; char sslerr[SSLERR_BUF_SIZE]; ctx = EVP_CIPHER_CTX_new(); + if (ctx == NULL) { + ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); + log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to allocate memory: %s", sslerr); + err = -1; + goto out; + } if (!RAND_bytes(salt, SALT_SIZE)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to get random salt data: %s", sslerr); err = -1; goto out; } /* * add warning re keylength */ - EVP_EncryptInit_ex(ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt); + if (EVP_EncryptInit_ex(ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt) <= 0) { + ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); + log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "EVP_EncryptInit_ex failed: %s", sslerr); + err = -1; + goto out; + } for (i=0; imodel_instance; EVP_CIPHER_CTX *ctx = NULL; int tmplen1 = 0, tmplen2 = 0; unsigned char *salt = (unsigned char *)buf_in; unsigned char *data = salt + SALT_SIZE; int datalen = buf_in_len - SALT_SIZE; int err = 0; char sslerr[SSLERR_BUF_SIZE]; if (datalen <= 0) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Packet is too short"); err = -1; goto out; } ctx = EVP_CIPHER_CTX_new(); + if (ctx == NULL) { + ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); + log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to allocate memory: %s", sslerr); + err = -1; + goto out; + } + /* * add warning re keylength */ - EVP_DecryptInit_ex(ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt); + if (EVP_DecryptInit_ex(ctx, instance->crypto_cipher_type, NULL, instance->private_key, salt) <= 0) { + ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); + log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "EVP_DecryptInit_ex failed: %s", sslerr); + err = -1; + goto out; + } if (!EVP_DecryptUpdate(ctx, buf_out, &tmplen1, data, datalen)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to decrypt: %s", sslerr); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to decrypt: %s", sslerr); } err = -1; goto out; } if (!EVP_DecryptFinal_ex(ctx, buf_out + tmplen1, &tmplen2)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to finalize decrypt: %s", sslerr); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to finalize decrypt: %s", sslerr); } err = -1; goto out; } *buf_out_len = tmplen1 + tmplen2; out: if (ctx) { EVP_CIPHER_CTX_free(ctx); } return err; } #endif /* * hash/hmac/digest functions */ #if (OPENSSL_VERSION_NUMBER < 0x30000000L) static int calculate_openssl_hash( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const unsigned char *buf, const size_t buf_len, unsigned char *hash, uint8_t log_level) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; unsigned int hash_len = 0; unsigned char *hash_out = NULL; char sslerr[SSLERR_BUF_SIZE]; hash_out = HMAC(instance->crypto_hash_type, instance->private_key, instance->private_key_len, buf, buf_len, hash, &hash_len); if ((!hash_out) || (hash_len != crypto_instance->sec_hash_size)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to calculate hash: %s", sslerr); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to calculate hash: %s", sslerr); } return -1; } return 0; } #else static int calculate_openssl_hash( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const unsigned char *buf, const size_t buf_len, unsigned char *hash, uint8_t log_level) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; EVP_MAC_CTX *ctx = NULL; char sslerr[SSLERR_BUF_SIZE]; int err = 0; size_t outlen = 0; ctx = EVP_MAC_CTX_new(instance->crypto_hash_mac); if (!ctx) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to allocate openssl context: %s", sslerr); err = -1; goto out_err; } if (!EVP_MAC_init(ctx, instance->private_key, instance->private_key_len, instance->params)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to set openssl context parameters: %s", sslerr); err = -1; goto out_err; } if (!EVP_MAC_update(ctx, buf, buf_len)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to update hash: %s", sslerr); err = -1; goto out_err; } if (!EVP_MAC_final(ctx, hash, &outlen, crypto_instance->sec_hash_size)) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to finalize hash: %s", sslerr); err = -1; goto out_err; } out_err: if (ctx) { EVP_MAC_CTX_free(ctx); } return err; } #endif /* * exported API */ static int opensslcrypto_encrypt_and_signv ( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const struct iovec *iov_in, int iovcnt_in, unsigned char *buf_out, ssize_t *buf_out_len) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; int i; if (instance->crypto_cipher_type) { if (encrypt_openssl(knet_h, crypto_instance, iov_in, iovcnt_in, buf_out, buf_out_len) < 0) { return -1; } } else { *buf_out_len = 0; for (i=0; icrypto_hash_type) { if (calculate_openssl_hash(knet_h, crypto_instance, buf_out, *buf_out_len, buf_out + *buf_out_len, KNET_LOG_ERR) < 0) { return -1; } *buf_out_len = *buf_out_len + crypto_instance->sec_hash_size; } return 0; } static int opensslcrypto_encrypt_and_sign ( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const unsigned char *buf_in, const ssize_t buf_in_len, unsigned char *buf_out, ssize_t *buf_out_len) { struct iovec iov_in; memset(&iov_in, 0, sizeof(iov_in)); iov_in.iov_base = (unsigned char *)buf_in; iov_in.iov_len = buf_in_len; return opensslcrypto_encrypt_and_signv(knet_h, crypto_instance, &iov_in, 1, buf_out, buf_out_len); } static int opensslcrypto_authenticate_and_decrypt ( knet_handle_t knet_h, struct crypto_instance *crypto_instance, const unsigned char *buf_in, const ssize_t buf_in_len, unsigned char *buf_out, ssize_t *buf_out_len, uint8_t log_level) { struct opensslcrypto_instance *instance = crypto_instance->model_instance; ssize_t temp_len = buf_in_len; if (instance->crypto_hash_type) { unsigned char tmp_hash[crypto_instance->sec_hash_size]; ssize_t temp_buf_len = buf_in_len - crypto_instance->sec_hash_size; memset(tmp_hash, 0, sizeof(tmp_hash)); if ((temp_buf_len <= 0) || (temp_buf_len > KNET_MAX_PACKET_SIZE)) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Received incorrect packet size: %zu for hash size: %zu", buf_in_len, crypto_instance->sec_hash_size); return -1; } if (calculate_openssl_hash(knet_h, crypto_instance, buf_in, temp_buf_len, tmp_hash, log_level) < 0) { return -1; } if (memcmp(tmp_hash, buf_in + temp_buf_len, crypto_instance->sec_hash_size) != 0) { if (log_level == KNET_LOG_DEBUG) { log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Digest does not match. Check crypto key and configuration."); } else { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Digest does not match. Check crypto key and configuration."); } return -1; } temp_len = temp_len - crypto_instance->sec_hash_size; *buf_out_len = temp_len; } if (instance->crypto_cipher_type) { if (decrypt_openssl(knet_h, crypto_instance, buf_in, temp_len, buf_out, buf_out_len, log_level) < 0) { return -1; } } else { memmove(buf_out, buf_in, temp_len); *buf_out_len = temp_len; } return 0; } #if (OPENSSL_VERSION_NUMBER < 0x10100000L) static pthread_mutex_t *openssl_internal_lock; static void openssl_internal_locking_callback(int mode, int type, char *file, int line) { if (mode & CRYPTO_LOCK) { (void)pthread_mutex_lock(&(openssl_internal_lock[type])); } else { pthread_mutex_unlock(&(openssl_internal_lock[type])); } } static pthread_t openssl_internal_thread_id(void) { return pthread_self(); } static void openssl_internal_lock_cleanup(void) { int i; CRYPTO_set_locking_callback(NULL); CRYPTO_set_id_callback(NULL); for (i = 0; i < CRYPTO_num_locks(); i++) { pthread_mutex_destroy(&(openssl_internal_lock[i])); } if (openssl_internal_lock) { free(openssl_internal_lock); } return; } static void openssl_atexit_handler(void) { openssl_internal_lock_cleanup(); } static int openssl_internal_lock_setup(void) { int savederrno = 0, err = 0; int i; openssl_internal_lock = malloc(CRYPTO_num_locks() * sizeof(pthread_mutex_t)); if (!openssl_internal_lock) { savederrno = errno; err = -1; goto out; } for (i = 0; i < CRYPTO_num_locks(); i++) { savederrno = pthread_mutex_init(&(openssl_internal_lock[i]), NULL); if (savederrno) { err = -1; goto out; } } CRYPTO_set_id_callback((void *)openssl_internal_thread_id); CRYPTO_set_locking_callback((void *)&openssl_internal_locking_callback); if (atexit(openssl_atexit_handler)) { err = -1; } out: if (err) { openssl_internal_lock_cleanup(); } errno = savederrno; return err; } #endif static void opensslcrypto_fini( knet_handle_t knet_h, struct crypto_instance *crypto_instance) { struct opensslcrypto_instance *opensslcrypto_instance = crypto_instance->model_instance; if (opensslcrypto_instance) { if (opensslcrypto_instance->private_key) { free(opensslcrypto_instance->private_key); opensslcrypto_instance->private_key = NULL; } #if (OPENSSL_VERSION_NUMBER >= 0x30000000L) if (opensslcrypto_instance->crypto_hash_mac) { EVP_MAC_free(opensslcrypto_instance->crypto_hash_mac); } #endif free(opensslcrypto_instance); crypto_instance->model_instance = NULL; } #if (OPENSSL_VERSION_NUMBER < 0x10100000L) ERR_free_strings(); #endif return; } static int opensslcrypto_init( knet_handle_t knet_h, struct crypto_instance *crypto_instance, struct knet_handle_crypto_cfg *knet_handle_crypto_cfg) { struct opensslcrypto_instance *opensslcrypto_instance = NULL; int savederrno; #if (OPENSSL_VERSION_NUMBER >= 0x30000000L) char sslerr[SSLERR_BUF_SIZE]; size_t params_n = 0; #endif log_debug(knet_h, KNET_SUB_OPENSSLCRYPTO, "Initializing openssl crypto module [%s/%s]", knet_handle_crypto_cfg->crypto_cipher_type, knet_handle_crypto_cfg->crypto_hash_type); if (!openssl_is_init) { #if (OPENSSL_VERSION_NUMBER < 0x10100000L) ERR_load_crypto_strings(); OPENSSL_add_all_algorithms_noconf(); if (openssl_internal_lock_setup() < 0) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to init openssl"); errno = EAGAIN; return -1; } #else if (!OPENSSL_init_crypto(OPENSSL_INIT_ADD_ALL_CIPHERS \ | OPENSSL_INIT_ADD_ALL_DIGESTS, NULL)) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to init openssl"); errno = EAGAIN; return -1; } #endif openssl_is_init = 1; } crypto_instance->model_instance = malloc(sizeof(struct opensslcrypto_instance)); if (!crypto_instance->model_instance) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to allocate memory for openssl model instance"); errno = ENOMEM; return -1; } opensslcrypto_instance = crypto_instance->model_instance; memset(opensslcrypto_instance, 0, sizeof(struct opensslcrypto_instance)); opensslcrypto_instance->private_key = malloc(knet_handle_crypto_cfg->private_key_len); if (!opensslcrypto_instance->private_key) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "Unable to allocate memory for openssl private key"); savederrno = ENOMEM; goto out_err; } memmove(opensslcrypto_instance->private_key, knet_handle_crypto_cfg->private_key, knet_handle_crypto_cfg->private_key_len); opensslcrypto_instance->private_key_len = knet_handle_crypto_cfg->private_key_len; if (strcmp(knet_handle_crypto_cfg->crypto_cipher_type, "none") == 0) { opensslcrypto_instance->crypto_cipher_type = NULL; } else { opensslcrypto_instance->crypto_cipher_type = EVP_get_cipherbyname(knet_handle_crypto_cfg->crypto_cipher_type); if (!opensslcrypto_instance->crypto_cipher_type) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "unknown crypto cipher type requested"); savederrno = ENXIO; goto out_err; } } if (strcmp(knet_handle_crypto_cfg->crypto_hash_type, "none") == 0) { opensslcrypto_instance->crypto_hash_type = NULL; } else { opensslcrypto_instance->crypto_hash_type = EVP_get_digestbyname(knet_handle_crypto_cfg->crypto_hash_type); if (!opensslcrypto_instance->crypto_hash_type) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "unknown crypto hash type requested"); savederrno = ENXIO; goto out_err; } #if (OPENSSL_VERSION_NUMBER >= 0x30000000L) opensslcrypto_instance->crypto_hash_mac = EVP_MAC_fetch(NULL, "HMAC", NULL); if (!opensslcrypto_instance->crypto_hash_mac) { ERR_error_string_n(ERR_get_error(), sslerr, sizeof(sslerr)); log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "unable to fetch HMAC: %s", sslerr); savederrno = ENXIO; goto out_err; } /* * OSSL_PARAM_construct_* store pointers to the data, it´s important that the referenced data are per-instance */ memmove(opensslcrypto_instance->hash_type, knet_handle_crypto_cfg->crypto_hash_type, sizeof(opensslcrypto_instance->hash_type)); opensslcrypto_instance->params[params_n++] = OSSL_PARAM_construct_utf8_string(hash, opensslcrypto_instance->hash_type, 0); opensslcrypto_instance->params[params_n] = OSSL_PARAM_construct_end(); #endif } if ((opensslcrypto_instance->crypto_cipher_type) && (!opensslcrypto_instance->crypto_hash_type)) { log_err(knet_h, KNET_SUB_OPENSSLCRYPTO, "crypto communication requires hash specified"); savederrno = EINVAL; goto out_err; } if (opensslcrypto_instance->crypto_hash_type) { crypto_instance->sec_hash_size = EVP_MD_size(opensslcrypto_instance->crypto_hash_type); } if (opensslcrypto_instance->crypto_cipher_type) { size_t block_size; block_size = EVP_CIPHER_block_size(opensslcrypto_instance->crypto_cipher_type); crypto_instance->sec_salt_size = SALT_SIZE; crypto_instance->sec_block_size = block_size; } return 0; out_err: opensslcrypto_fini(knet_h, crypto_instance); errno = savederrno; return -1; } crypto_ops_t crypto_model = { KNET_CRYPTO_MODEL_ABI, opensslcrypto_init, opensslcrypto_fini, opensslcrypto_encrypt_and_sign, opensslcrypto_encrypt_and_signv, opensslcrypto_authenticate_and_decrypt }; diff --git a/libknet/handle_api.c b/libknet/handle_api.c index 4ddcda1b..7624eb74 100644 --- a/libknet/handle_api.c +++ b/libknet/handle_api.c @@ -1,592 +1,621 @@ /* * Copyright (C) 2020-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * Federico Simoncelli * * This software licensed under LGPL-2.0+ */ #include "config.h" #include #include #include #include #include #include #include "internals.h" #include "crypto.h" #include "links.h" #include "common.h" #include "transport_common.h" #include "logging.h" int knet_handle_enable_sock_notify(knet_handle_t knet_h, void *sock_notify_fn_private_data, void (*sock_notify_fn) ( void *private_data, int datafd, int8_t channel, uint8_t tx_rx, int error, int errorno)) { int savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if (!sock_notify_fn) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } knet_h->sock_notify_fn_private_data = sock_notify_fn_private_data; knet_h->sock_notify_fn = sock_notify_fn; log_debug(knet_h, KNET_SUB_HANDLE, "sock_notify_fn enabled"); pthread_rwlock_unlock(&knet_h->global_rwlock); return 0; } int knet_handle_add_datafd(knet_handle_t knet_h, int *datafd, int8_t *channel) { int err = 0, savederrno = 0; int i; struct epoll_event ev; if (!_is_valid_handle(knet_h)) { return -1; } if (datafd == NULL) { errno = EINVAL; return -1; } if (channel == NULL) { errno = EINVAL; return -1; } if (*channel >= KNET_DATAFD_MAX) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } if (!knet_h->sock_notify_fn) { log_err(knet_h, KNET_SUB_HANDLE, "Adding datafd requires sock notify callback enabled!"); savederrno = EINVAL; err = -1; goto out_unlock; } if (*datafd > 0) { for (i = 0; i < KNET_DATAFD_MAX; i++) { if ((knet_h->sockfd[i].in_use) && (knet_h->sockfd[i].sockfd[0] == *datafd)) { log_err(knet_h, KNET_SUB_HANDLE, "requested datafd: %d already exist in index: %d", *datafd, i); savederrno = EEXIST; err = -1; goto out_unlock; } } } /* * auto allocate a channel */ if (*channel < 0) { for (i = 0; i < KNET_DATAFD_MAX; i++) { if (!knet_h->sockfd[i].in_use) { *channel = i; break; } } if (*channel < 0) { savederrno = EBUSY; err = -1; goto out_unlock; } } else { if (knet_h->sockfd[*channel].in_use) { savederrno = EBUSY; err = -1; goto out_unlock; } } knet_h->sockfd[*channel].is_created = 0; knet_h->sockfd[*channel].is_socket = 0; knet_h->sockfd[*channel].has_error = 0; if (*datafd > 0) { int sockopt; socklen_t sockoptlen = sizeof(sockopt); if (_fdset_cloexec(*datafd)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_HANDLE, "Unable to set CLOEXEC on datafd: %s", strerror(savederrno)); goto out_unlock; } if (_fdset_nonblock(*datafd)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_HANDLE, "Unable to set NONBLOCK on datafd: %s", strerror(savederrno)); goto out_unlock; } knet_h->sockfd[*channel].sockfd[0] = *datafd; knet_h->sockfd[*channel].sockfd[1] = 0; if (!getsockopt(knet_h->sockfd[*channel].sockfd[0], SOL_SOCKET, SO_TYPE, &sockopt, &sockoptlen)) { knet_h->sockfd[*channel].is_socket = 1; } } else { if (_init_socketpair(knet_h, knet_h->sockfd[*channel].sockfd)) { savederrno = errno; err = -1; goto out_unlock; } knet_h->sockfd[*channel].is_created = 1; knet_h->sockfd[*channel].is_socket = 1; *datafd = knet_h->sockfd[*channel].sockfd[0]; } memset(&ev, 0, sizeof(struct epoll_event)); ev.events = EPOLLIN; ev.data.fd = knet_h->sockfd[*channel].sockfd[knet_h->sockfd[*channel].is_created]; if (epoll_ctl(knet_h->send_to_links_epollfd, EPOLL_CTL_ADD, knet_h->sockfd[*channel].sockfd[knet_h->sockfd[*channel].is_created], &ev)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_HANDLE, "Unable to add datafd %d to linkfd epoll pool: %s", knet_h->sockfd[*channel].sockfd[knet_h->sockfd[*channel].is_created], strerror(savederrno)); if (knet_h->sockfd[*channel].is_created) { _close_socketpair(knet_h, knet_h->sockfd[*channel].sockfd); } goto out_unlock; } knet_h->sockfd[*channel].in_use = 1; out_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); errno = err ? savederrno : 0; return err; } int knet_handle_remove_datafd(knet_handle_t knet_h, int datafd) { int err = 0, savederrno = 0; int8_t channel = -1; int i; struct epoll_event ev; if (!_is_valid_handle(knet_h)) { return -1; } if (datafd <= 0) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } for (i = 0; i < KNET_DATAFD_MAX; i++) { if ((knet_h->sockfd[i].in_use) && (knet_h->sockfd[i].sockfd[0] == datafd)) { channel = i; break; } } if (channel < 0) { savederrno = EINVAL; err = -1; goto out_unlock; } if (!knet_h->sockfd[channel].has_error) { memset(&ev, 0, sizeof(struct epoll_event)); if (epoll_ctl(knet_h->send_to_links_epollfd, EPOLL_CTL_DEL, knet_h->sockfd[channel].sockfd[knet_h->sockfd[channel].is_created], &ev)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_HANDLE, "Unable to del datafd %d from linkfd epoll pool: %s", knet_h->sockfd[channel].sockfd[0], strerror(savederrno)); goto out_unlock; } } if (knet_h->sockfd[channel].is_created) { _close_socketpair(knet_h, knet_h->sockfd[channel].sockfd); } memset(&knet_h->sockfd[channel], 0, sizeof(struct knet_sock)); out_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); errno = err ? savederrno : 0; return err; } int knet_handle_get_datafd(knet_handle_t knet_h, const int8_t channel, int *datafd) { int err = 0, savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if ((channel < 0) || (channel >= KNET_DATAFD_MAX)) { errno = EINVAL; return -1; } if (datafd == NULL) { errno = EINVAL; return -1; } savederrno = pthread_rwlock_rdlock(&knet_h->global_rwlock); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get read lock: %s", strerror(savederrno)); errno = savederrno; return -1; } if (!knet_h->sockfd[channel].in_use) { savederrno = EINVAL; err = -1; goto out_unlock; } *datafd = knet_h->sockfd[channel].sockfd[0]; out_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); errno = err ? savederrno : 0; return err; } int knet_handle_get_channel(knet_handle_t knet_h, const int datafd, int8_t *channel) { int err = 0, savederrno = 0; int i; if (!_is_valid_handle(knet_h)) { return -1; } if (datafd <= 0) { errno = EINVAL; return -1; } if (channel == NULL) { errno = EINVAL; return -1; } savederrno = pthread_rwlock_rdlock(&knet_h->global_rwlock); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get read lock: %s", strerror(savederrno)); errno = savederrno; return -1; } *channel = -1; for (i = 0; i < KNET_DATAFD_MAX; i++) { if ((knet_h->sockfd[i].in_use) && (knet_h->sockfd[i].sockfd[0] == datafd)) { *channel = i; break; } } if (*channel < 0) { savederrno = EINVAL; err = -1; goto out_unlock; } out_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); errno = err ? savederrno : 0; return err; } int knet_handle_enable_filter(knet_handle_t knet_h, void *dst_host_filter_fn_private_data, int (*dst_host_filter_fn) ( void *private_data, const unsigned char *outdata, ssize_t outdata_len, uint8_t tx_rx, knet_node_id_t this_host_id, knet_node_id_t src_node_id, int8_t *channel, knet_node_id_t *dst_host_ids, size_t *dst_host_ids_entries)) { int savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } knet_h->dst_host_filter_fn_private_data = dst_host_filter_fn_private_data; knet_h->dst_host_filter_fn = dst_host_filter_fn; if (knet_h->dst_host_filter_fn) { log_debug(knet_h, KNET_SUB_HANDLE, "dst_host_filter_fn enabled"); } else { log_debug(knet_h, KNET_SUB_HANDLE, "dst_host_filter_fn disabled"); } pthread_rwlock_unlock(&knet_h->global_rwlock); errno = 0; return 0; } int knet_handle_setfwd(knet_handle_t knet_h, unsigned int enabled) { int savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if (enabled > 1) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } if (enabled) { knet_h->enabled = enabled; log_debug(knet_h, KNET_SUB_HANDLE, "Data forwarding is enabled"); } else { /* * notify TX and RX threads to flush the queues */ if (set_thread_flush_queue(knet_h, KNET_THREAD_TX, KNET_THREAD_QUEUE_FLUSH) < 0) { log_debug(knet_h, KNET_SUB_HANDLE, "Unable to request queue flushing for TX thread"); } if (set_thread_flush_queue(knet_h, KNET_THREAD_RX, KNET_THREAD_QUEUE_FLUSH) < 0) { log_debug(knet_h, KNET_SUB_HANDLE, "Unable to request queue flushing for RX thread"); } } pthread_rwlock_unlock(&knet_h->global_rwlock); /* * when disabling data forward, we need to give time to TX and RX * to flush the queues. * * the TX thread is the main leader here. When there is no more * data in the TX queue, we will also close traffic for RX. */ if (!enabled) { /* * this usleep might be unnecessary, but wait_all_threads_flush_queue * adds extra locking delay. * * allow all threads to run free without extra locking interference * and then we switch to a more active wait in case the scheduler * has decided to delay one thread or another */ usleep(KNET_THREADS_TIMERES * 2); wait_all_threads_flush_queue(knet_h); /* * all threads have done flushing the queue, we can stop data forwarding */ savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } knet_h->enabled = enabled; log_debug(knet_h, KNET_SUB_HANDLE, "Data forwarding is disabled"); pthread_rwlock_unlock(&knet_h->global_rwlock); } errno = 0; return 0; } +int knet_handle_setprio_dscp(knet_handle_t knet_h, uint8_t dscp) +{ + int savederrno = 0; + + if (!_is_valid_handle(knet_h)) { + return -1; + } + + if (dscp > 0x3f) { + errno = EINVAL; + return -1; + } + + savederrno = get_global_wrlock(knet_h); + if (savederrno) { + log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", + strerror(savederrno)); + errno = savederrno; + return -1; + } + + knet_h->prio_dscp = dscp; + + pthread_rwlock_unlock(&knet_h->global_rwlock); + + errno = 0; + return 0; +} + int knet_handle_get_stats(knet_handle_t knet_h, struct knet_handle_stats *stats, size_t struct_size) { int err = 0, savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if (!stats) { errno = EINVAL; return -1; } savederrno = pthread_rwlock_rdlock(&knet_h->global_rwlock); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get read lock: %s", strerror(savederrno)); errno = savederrno; return -1; } savederrno = pthread_mutex_lock(&knet_h->handle_stats_mutex); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get mutex lock: %s", strerror(savederrno)); err = -1; goto out_unlock; } if (struct_size > sizeof(struct knet_handle_stats)) { struct_size = sizeof(struct knet_handle_stats); } memmove(stats, &knet_h->stats, struct_size); /* * TX crypt stats only count the data packets sent, so add in the ping/pong/pmtud figures * RX is OK as it counts them before they are sorted. */ stats->tx_crypt_packets += knet_h->stats_extra.tx_crypt_ping_packets + knet_h->stats_extra.tx_crypt_pong_packets + knet_h->stats_extra.tx_crypt_pmtu_packets + knet_h->stats_extra.tx_crypt_pmtu_reply_packets; /* Tell the caller our full size in case they have an old version */ stats->size = sizeof(struct knet_handle_stats); out_unlock: pthread_mutex_unlock(&knet_h->handle_stats_mutex); pthread_rwlock_unlock(&knet_h->global_rwlock); return err; } int knet_handle_clear_stats(knet_handle_t knet_h, int clear_option) { int savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if (clear_option != KNET_CLEARSTATS_HANDLE_ONLY && clear_option != KNET_CLEARSTATS_HANDLE_AND_LINK) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } memset(&knet_h->stats, 0, sizeof(struct knet_handle_stats)); memset(&knet_h->stats_extra, 0, sizeof(struct knet_handle_stats_extra)); if (clear_option == KNET_CLEARSTATS_HANDLE_AND_LINK) { _link_clear_stats(knet_h); } pthread_rwlock_unlock(&knet_h->global_rwlock); return 0; } int knet_handle_enable_access_lists(knet_handle_t knet_h, unsigned int enabled) { int savederrno = 0; if (!_is_valid_handle(knet_h)) { return -1; } if (enabled > 1) { errno = EINVAL; return -1; } savederrno = get_global_wrlock(knet_h); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get write lock: %s", strerror(savederrno)); errno = savederrno; return -1; } knet_h->use_access_lists = enabled; if (enabled) { log_debug(knet_h, KNET_SUB_HANDLE, "Links access lists are enabled"); } else { log_debug(knet_h, KNET_SUB_HANDLE, "Links access lists are disabled"); } pthread_rwlock_unlock(&knet_h->global_rwlock); errno = 0; return 0; } diff --git a/libknet/internals.h b/libknet/internals.h index 0cc5ce67..b4313362 100644 --- a/libknet/internals.h +++ b/libknet/internals.h @@ -1,433 +1,434 @@ /* * Copyright (C) 2010-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * Federico Simoncelli * * This software licensed under LGPL-2.0+ */ #ifndef __KNET_INTERNALS_H__ #define __KNET_INTERNALS_H__ /* * NOTE: you shouldn't need to include this header normally */ #include #include #include #include "libknet.h" #include "onwire.h" #include "compat.h" #include "threads_common.h" #define KNET_DATABUFSIZE KNET_MAX_PACKET_SIZE + KNET_HEADER_ALL_SIZE #define KNET_DATABUFSIZE_CRYPT_PAD 1024 #define KNET_DATABUFSIZE_CRYPT KNET_DATABUFSIZE + KNET_DATABUFSIZE_CRYPT_PAD #define KNET_DATABUFSIZE_COMPRESS_PAD 1024 #define KNET_DATABUFSIZE_COMPRESS KNET_DATABUFSIZE + KNET_DATABUFSIZE_COMPRESS_PAD #define KNET_RING_RCVBUFF 8388608 #define PCKT_FRAG_MAX UINT8_MAX #define PCKT_RX_BUFS 512 #define KNET_EPOLL_MAX_EVENTS KNET_DATAFD_MAX + 1 /* * Size of threads stack. Value is choosen by experimenting, how much is needed * to sucesfully finish test suite, and at the time of writing patch it was * ~300KiB. To have some room for future enhancement it is increased * by factor of 3 and rounded. */ #define KNET_THREAD_STACK_SIZE (1024 * 1024) typedef void *knet_transport_link_t; /* per link transport handle */ typedef void *knet_transport_t; /* per knet_h transport handle */ struct knet_transport_ops; /* Forward because of circular dependancy */ struct knet_mmsghdr { struct msghdr msg_hdr; /* Message header */ unsigned int msg_len; /* Number of bytes transmitted */ }; struct knet_link { /* required */ struct sockaddr_storage src_addr; struct sockaddr_storage dst_addr; /* configurable */ unsigned int dynamic; /* see KNET_LINK_DYN_ define above */ uint8_t priority; /* higher priority == preferred for A/P */ unsigned long long ping_interval; /* interval */ unsigned long long pong_timeout; /* timeout */ unsigned long long pong_timeout_adj; /* timeout adjusted for latency */ uint8_t pong_timeout_backoff; /* see link.h for definition */ unsigned int latency_max_samples; /* precision */ unsigned int latency_cur_samples; uint8_t pong_count; /* how many ping/pong to send/receive before link is up */ uint64_t flags; void *access_list_match_entry_head; /* pointer to access list match_entry list head */ /* status */ struct knet_link_status status; /* internals */ pthread_mutex_t link_stats_mutex; /* used to update link stats */ uint8_t link_id; uint8_t transport; /* #defined constant from API */ knet_transport_link_t transport_link; /* link_info_t from transport */ int outsock; unsigned int configured:1; /* set to 1 if src/dst have been configured transport initialized on this link*/ unsigned int transport_connected:1; /* set to 1 if lower level transport is connected */ uint8_t received_pong; struct timespec ping_last; /* used by PMTUD thread as temp per-link variables and should always contain the onwire_len value! */ uint32_t proto_overhead; /* IP + UDP/SCTP overhead. NOT to be confused with stats.proto_overhead that includes also knet headers and crypto headers */ struct timespec pmtud_last; uint32_t last_ping_size; uint32_t last_good_mtu; uint32_t last_bad_mtu; uint32_t last_sent_mtu; uint32_t last_recv_mtu; uint32_t pmtud_crypto_timeout_multiplier;/* used by PMTUd to adjust timeouts on high loads */ uint8_t has_valid_mtu; }; #define KNET_DEFRAG_BUFFERS 32 #define KNET_CBUFFER_SIZE 4096 struct knet_host_defrag_buf { char buf[KNET_DATABUFSIZE]; uint8_t in_use; /* 0 buffer is free, 1 is in use */ seq_num_t pckt_seq; /* identify the pckt we are receiving */ uint8_t frag_recv; /* how many frags did we receive */ uint8_t frag_map[PCKT_FRAG_MAX];/* bitmap of what we received? */ uint8_t last_first; /* special case if we receive the last fragment first */ ssize_t frag_size; /* normal frag size (not the last one) */ ssize_t last_frag_size; /* the last fragment might not be aligned with MTU size */ struct timespec last_update; /* keep time of the last pckt */ }; struct knet_host { /* required */ knet_node_id_t host_id; /* configurable */ uint8_t link_handler_policy; char name[KNET_MAX_HOST_LEN]; /* status */ struct knet_host_status status; /* internals */ char circular_buffer[KNET_CBUFFER_SIZE]; seq_num_t rx_seq_num; seq_num_t untimed_rx_seq_num; seq_num_t timed_rx_seq_num; uint8_t got_data; /* defrag/reassembly buffers */ struct knet_host_defrag_buf defrag_buf[KNET_DEFRAG_BUFFERS]; char circular_buffer_defrag[KNET_CBUFFER_SIZE]; /* link stuff */ struct knet_link link[KNET_MAX_LINK]; uint8_t active_link_entries; uint8_t active_links[KNET_MAX_LINK]; struct knet_host *next; }; struct knet_sock { int sockfd[2]; /* sockfd[0] will always be application facing * and sockfd[1] internal if sockpair has been created by knet */ int is_socket; /* check if it's a socket for recvmmsg usage */ int is_created; /* knet created this socket and has to clean up on exit/del */ int in_use; /* set to 1 if it's use, 0 if free */ int has_error; /* set to 1 if there were errors reading from the sock * and socket has been removed from epoll */ }; struct knet_fd_trackers { uint8_t transport; /* transport type (UDP/SCTP...) */ uint8_t data_type; /* internal use for transport to define what data are associated * with this fd */ socklen_t sockaddr_len; /* Size of sockaddr_in[6] structure for this socket */ void *data; /* pointer to the data */ int ifindex; /* interface index for this bound address */ }; #define KNET_MAX_FDS KNET_MAX_HOST * KNET_MAX_LINK * 4 #define KNET_MAX_COMPRESS_METHODS UINT8_MAX #define KNET_MAX_CRYPTO_INSTANCES 2 struct knet_handle_stats_extra { uint64_t tx_crypt_pmtu_packets; uint64_t tx_crypt_pmtu_reply_packets; uint64_t tx_crypt_ping_packets; uint64_t tx_crypt_pong_packets; }; struct knet_handle { knet_node_id_t host_id; unsigned int enabled:1; struct knet_sock sockfd[KNET_DATAFD_MAX + 1]; int logfd; uint8_t log_levels[KNET_MAX_SUBSYSTEMS]; int dstsockfd[2]; int send_to_links_epollfd; int recv_from_links_epollfd; int dst_link_handler_epollfd; uint8_t use_access_lists; /* set to 0 for disable, 1 for enable */ unsigned int pmtud_interval; unsigned int manual_mtu; unsigned int data_mtu; /* contains the max data size that we can send onwire * without frags */ + uint8_t prio_dscp; /* use this dscp value for KNET_LINK_FLAG_TRAFFICHIPRIO */ struct knet_host *host_head; struct knet_host *host_index[KNET_MAX_HOST]; knet_transport_t transports[KNET_MAX_TRANSPORTS+1]; struct knet_fd_trackers knet_transport_fd_tracker[KNET_MAX_FDS]; /* track status for each fd handled by transports */ struct knet_handle_stats stats; struct knet_handle_stats_extra stats_extra; pthread_mutex_t handle_stats_mutex; /* used to protect handle stats */ uint32_t reconnect_int; knet_node_id_t host_ids[KNET_MAX_HOST]; size_t host_ids_entries; struct knet_header *recv_from_sock_buf; struct knet_header *send_to_links_buf[PCKT_FRAG_MAX]; struct knet_header *recv_from_links_buf[PCKT_RX_BUFS]; struct knet_header *pingbuf; struct knet_header *pmtudbuf; uint8_t threads_status[KNET_THREAD_MAX]; uint8_t threads_flush_queue[KNET_THREAD_MAX]; pthread_mutex_t threads_status_mutex; pthread_t send_to_links_thread; pthread_t recv_from_links_thread; pthread_t heartbt_thread; pthread_t dst_link_handler_thread; pthread_t pmtud_link_handler_thread; pthread_rwlock_t global_rwlock; /* global config lock */ pthread_mutex_t pmtud_mutex; /* pmtud mutex to handle conditional send/recv + timeout */ pthread_cond_t pmtud_cond; /* conditional for above */ pthread_mutex_t tx_mutex; /* used to protect knet_send_sync and TX thread */ pthread_mutex_t hb_mutex; /* used to protect heartbeat thread and seq_num broadcasting */ pthread_mutex_t backoff_mutex; /* used to protect dst_link->pong_timeout_adj */ pthread_mutex_t kmtu_mutex; /* used to protect kernel_mtu */ uint32_t kernel_mtu; /* contains the MTU detected by the kernel on a given link */ int pmtud_waiting; int pmtud_running; int pmtud_forcerun; int pmtud_abort; struct crypto_instance *crypto_instance[KNET_MAX_CRYPTO_INSTANCES + 1]; /* store an extra pointer to allow 0|1|2 values without too much magic in the code */ uint8_t crypto_in_use_config; /* crypto config to use for TX */ uint8_t crypto_only; /* allow only crypto (1) or also clear (0) traffic */ size_t sec_block_size; size_t sec_hash_size; size_t sec_salt_size; unsigned char *send_to_links_buf_crypt[PCKT_FRAG_MAX]; unsigned char *recv_from_links_buf_crypt; unsigned char *recv_from_links_buf_decrypt; unsigned char *pingbuf_crypt; unsigned char *pmtudbuf_crypt; int compress_model; int compress_level; size_t compress_threshold; void *compress_int_data[KNET_MAX_COMPRESS_METHODS]; /* for compress method private data */ unsigned char *recv_from_links_buf_decompress; unsigned char *send_to_links_buf_compress; seq_num_t tx_seq_num; pthread_mutex_t tx_seq_num_mutex; uint8_t has_loop_link; uint8_t loop_link; void *dst_host_filter_fn_private_data; int (*dst_host_filter_fn) ( void *private_data, const unsigned char *outdata, ssize_t outdata_len, uint8_t tx_rx, knet_node_id_t this_host_id, knet_node_id_t src_node_id, int8_t *channel, knet_node_id_t *dst_host_ids, size_t *dst_host_ids_entries); void *pmtud_notify_fn_private_data; void (*pmtud_notify_fn) ( void *private_data, unsigned int data_mtu); void *host_status_change_notify_fn_private_data; void (*host_status_change_notify_fn) ( void *private_data, knet_node_id_t host_id, uint8_t reachable, uint8_t remote, uint8_t external); void *sock_notify_fn_private_data; void (*sock_notify_fn) ( void *private_data, int datafd, int8_t channel, uint8_t tx_rx, int error, int errorno); int fini_in_progress; uint64_t flags; struct qb_list_head list; const char *plugin_path; }; struct handle_tracker { struct qb_list_head head; }; /* * lib_config stuff shared across everything */ extern pthread_rwlock_t shlib_rwlock; /* global shared lib load lock */ extern pthread_mutex_t handle_config_mutex; extern struct handle_tracker handle_list; extern uint8_t handle_list_init; int _is_valid_handle(knet_handle_t knet_h); int _init_shlib_tracker(knet_handle_t knet_h); void _fini_shlib_tracker(void); /* * NOTE: every single operation must be implementend * for every protocol. */ /* * for now knet supports only IP protocols (udp/sctp) * in future there might be others like ARP * or TIPC. * keep this around as transport information * to use for access lists and other operations */ #define TRANSPORT_PROTO_LOOPBACK 0 #define TRANSPORT_PROTO_IP_PROTO 1 /* * some transports like SCTP can filter incoming * connections before knet has to process * any packets. * GENERIC_ACL -> packet has to be read and filterted * PROTO_ACL -> transport provides filtering at lower levels * and packet does not need to be processed */ typedef enum { USE_NO_ACL, USE_GENERIC_ACL, USE_PROTO_ACL } transport_acl; /* * make it easier to map values in transports.c */ #define TRANSPORT_PROTO_NOT_CONNECTION_ORIENTED 0 #define TRANSPORT_PROTO_IS_CONNECTION_ORIENTED 1 typedef struct knet_transport_ops { /* * transport generic information */ const char *transport_name; const uint8_t transport_id; const uint8_t built_in; uint8_t transport_protocol; transport_acl transport_acl_type; /* * connection oriented protocols like SCTP * don´t need dst_addr in sendto calls and * on some OSes are considered EINVAL. */ uint8_t transport_is_connection_oriented; uint32_t transport_mtu_overhead; /* * transport init must allocate the new transport * and perform all internal initializations * (threads, lists, etc). */ int (*transport_init)(knet_handle_t knet_h); /* * transport free must releases _all_ resources * allocated by tranport_init */ int (*transport_free)(knet_handle_t knet_h); /* * link operations should take care of all the * sockets and epoll management for a given link/transport set * transport_link_disable should return err = -1 and errno = EBUSY * if listener is still in use, and any other errno in case * the link cannot be disabled. * * set_config/clear_config are invoked in global write lock context */ int (*transport_link_set_config)(knet_handle_t knet_h, struct knet_link *link); int (*transport_link_clear_config)(knet_handle_t knet_h, struct knet_link *link); /* * transport callback for incoming dynamic connections * this is called in global read lock context */ int (*transport_link_dyn_connect)(knet_handle_t knet_h, int sockfd, struct knet_link *link); /* * per transport error handling of recvmmsg * (see _handle_recv_from_links comments for details) */ /* * transport_rx_sock_error is invoked when recvmmsg returns <= 0 * * transport_rx_sock_error is invoked with both global_rdlock */ int (*transport_rx_sock_error)(knet_handle_t knet_h, int sockfd, int recv_err, int recv_errno); /* * transport_tx_sock_error is invoked with global_rwlock and * it's invoked when sendto or sendmmsg returns =< 0 * * it should return: * -1 on internal error * 0 ignore error and continue * 1 retry * any sleep or wait action should happen inside the transport code */ int (*transport_tx_sock_error)(knet_handle_t knet_h, int sockfd, int subsys, int recv_err, int recv_errno); /* * this function is called on _every_ received packet * to verify if the packet is data or internal protocol error handling * * it should return: * -1 on error * 0 packet is not data and we should continue the packet process loop * 1 packet is not data and we should STOP the packet process loop * 2 packet is data and should be parsed as such * * transport_rx_is_data is invoked with both global_rwlock * and fd_tracker read lock (from RX thread) */ int (*transport_rx_is_data)(knet_handle_t knet_h, int sockfd, struct knet_mmsghdr *msg); /* * this function is called by links.c when a link down event is recorded * to notify the transport that packets are not going through, and give * transport the opportunity to take actions. */ int (*transport_link_is_down)(knet_handle_t knet_h, struct knet_link *link); } knet_transport_ops_t; struct pretty_names { const char *name; uint8_t val; }; #endif diff --git a/libknet/libknet.h b/libknet/libknet.h index 5d7e6aa4..643a5aea 100644 --- a/libknet/libknet.h +++ b/libknet/libknet.h @@ -1,2459 +1,2484 @@ /* * Copyright (C) 2010-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * Federico Simoncelli * * This software licensed under LGPL-2.0+ */ #ifndef __LIBKNET_H__ #define __LIBKNET_H__ #include #include #include #include #include /** * @file libknet.h * @brief kronosnet API include file * @copyright Copyright (C) 2010-2025 Red Hat, Inc. All rights reserved. * * Kronosnet is an advanced VPN system for High Availability applications. */ #define KNET_API_VER 1 /* * libknet limits */ /** typedef for a knet node */ typedef uint16_t knet_node_id_t; /* * Maximum number of hosts */ #define KNET_MAX_HOST 65536 /* * Maximum number of links between 2 hosts */ #define KNET_MAX_LINK 8 /* * Maximum packet size that should be written to datafd * see knet_handle_new for details */ #define KNET_MAX_PACKET_SIZE 65536 /* * Buffers used for pretty logging * host is used to store both ip addresses and hostnames */ #define KNET_MAX_HOST_LEN 256 #define KNET_MAX_PORT_LEN 6 /* * Some notifications can be generated either on TX or RX */ #define KNET_NOTIFY_TX 0 #define KNET_NOTIFY_RX 1 /* * Link flags */ /* * Where possible, set traffic priority to high. * On Linux this sets the TOS to INTERACTIVE (6), * see tc-prio(8) for more infomation + * A dscp value may be configured, see knet_handle_setprio_dscp. */ #define KNET_LINK_FLAG_TRAFFICHIPRIO (1ULL << 0) /* * Handle flags */ /* * Use privileged operations during socket setup. */ #define KNET_HANDLE_FLAG_PRIVILEGED (1ULL << 0) /** * Opaque handle for this knet connection, created with knet_handle_new() and * freed with knet_handle_free() */ typedef struct knet_handle *knet_handle_t; /* * Handle structs/API calls */ /** * knet_handle_new_ex * * @brief create a new instance of a knet handle * * host_id - Each host in a knet is identified with a unique * ID. when creating a new handle local host_id * must be specified (0 to UINT16_MAX are all valid). * It is the user's responsibility to check that the value * is unique, or bad things might happen. * * log_fd - Write file descriptor. If set to a value > 0, it will be used * to write log packets from libknet to the application. * Setting to 0 will disable logging from libknet. * It is possible to enable logging at any given time (see logging API). * Make sure to either read from this filedescriptor properly and/or * mark it O_NONBLOCK, otherwise if the fd becomes full, libknet could * block. * It is strongly encouraged to use pipes (ex: pipe(2) or pipe2(2)) for * logging fds due to the atomic nature of writes between fds. * See also libknet test suite for reference and guidance. * The caller is responsible for management of the FD. eg. knet will not * close it when knet_handle_free(3) is called * * default_log_level - * If logfd is specified, it will initialize all subsystems to log * at default_log_level value. (see logging API) * * flags - bitwise OR of some of the following flags: * KNET_HANDLE_FLAG_PRIVILEGED: use privileged operations setting up the * communication sockets. If disabled, failure to acquire large * enough socket buffers is ignored but logged. Inadequate buffers * lead to poor performance. * * @return * on success, a new knet_handle_t is returned. * on failure, NULL is returned and errno is set. * knet-specific errno values: * ENAMETOOLONG - socket buffers couldn't be set big enough and KNET_HANDLE_FLAG_PRIVILEGED was specified * ERANGE - buffer size readback returned unexpected type */ knet_handle_t knet_handle_new_ex(knet_node_id_t host_id, int log_fd, uint8_t default_log_level, uint64_t flags); /** * knet_handle_new * * @brief knet_handle_new_ex with flags = KNET_HANDLE_FLAG_PRIVILEGED. */ knet_handle_t knet_handle_new(knet_node_id_t host_id, int log_fd, uint8_t default_log_level); /** * knet_handle_free * * @brief Destroy a knet handle, free all resources * * knet_h - pointer to knet_handle_t * * @return * knet_handle_free returns * 0 on success * -1 on error and errno is set. */ int knet_handle_free(knet_handle_t knet_h); /** * knet_handle_enable_sock_notify * * @brief Register a callback to receive socket events * * knet_h - pointer to knet_handle_t * * sock_notify_fn_private_data * void pointer to data that can be used to identify * the callback. * * sock_notify_fn * A callback function that is invoked every time * a socket in the datafd pool will report an error (-1) * or an end of read (0) (see socket.7). * This function MUST NEVER block or add substantial delays. * The callback is invoked in an internal unlocked area * to allow calls to knet_handle_add_datafd/knet_handle_remove_datafd * to swap/replace the bad fd. * if both err and errno are 0, it means that the socket * has received a 0 byte packet (EOF?). * The callback function must either remove the fd from knet * (by calling knet_handle_remove_fd()) or dup a new fd in its place. * Failure to do this can cause problems. * * @return * knet_handle_enable_sock_notify returns * 0 on success * -1 on error and errno is set. */ int knet_handle_enable_sock_notify(knet_handle_t knet_h, void *sock_notify_fn_private_data, void (*sock_notify_fn) ( void *private_data, int datafd, int8_t channel, uint8_t tx_rx, int error, int errorno)); /* sorry! can't call it errno ;) */ #define KNET_DATAFD_MAX 32 /** * knet_handle_add_datafd * * @brief Install a file descriptor for communication * * IMPORTANT: In order to add datafd to knet, knet_handle_enable_sock_notify * _MUST_ be set and be able to handle both errors (-1) and * 0 bytes read / write from the provided datafd. * On read error (< 0) from datafd, the socket is automatically * removed from polling to avoid spinning on dead sockets. * It is safe to call knet_handle_remove_datafd even on sockets * that have been removed. * * knet_h - pointer to knet_handle_t * * *datafd - read/write file descriptor. * knet will read data here to send to the other hosts * and will write data received from the network. * Each data packet can be of max size KNET_MAX_PACKET_SIZE! * Applications using knet_send/knet_recv will receive a * proper error if the packet size is not within boundaries. * Applications using their own functions to write to the * datafd should NOT write more than KNET_MAX_PACKET_SIZE. * * Please refer to handle.c on how to set up a socketpair. * * datafd can be 0, and knet_handle_add_datafd will create a properly * populated socket pair the same way as ping_test, or a value * higher than 0. A negative number will return an error. * On exit knet_handle_free will take care to cleanup the * socketpair only if they have been created by knet_handle_add_datafd. * * It is possible to pass either sockets or normal fds. * User provided datafd will be marked as non-blocking and close-on-exec. * * *channel - This value is analogous to the tag in VLAN tagging. * A negative value will auto-allocate a channel. * Setting a value between 0 and 31 will try to allocate that * specific channel (unless already in use). * * It is possible to add up to 32 datafds but be aware that each * one of them must have a receiving end on the other host. * * Example: * hostA channel 0 will be delivered to datafd on hostB channel 0 * hostA channel 1 to hostB channel 1. * * Each channel must have a unique file descriptor. * * If your application could have 2 channels on one host and one * channel on another host, then you can use dst_host_filter * to manipulate channel values on TX and RX. * * @return * knet_handle_add_datafd returns * @retval 0 on success, * *datafd will be populated with a socket if the original value was 0 * or if a specific fd was set, the value is untouched. * *channel will be populated with a channel number if the original value * was negative or the value is untouched if a specific channel * was requested. * * @retval -1 on error and errno is set. * *datafd and *channel are untouched or empty. */ int knet_handle_add_datafd(knet_handle_t knet_h, int *datafd, int8_t *channel); /** * knet_handle_remove_datafd * * @brief Remove a file descriptor from knet * * knet_h - pointer to knet_handle_t * * datafd - file descriptor to remove. * NOTE that if the socket/fd was created by knet_handle_add_datafd, * the socket will be closed by libknet. * * @return * knet_handle_remove_datafd returns * 0 on success * -1 on error and errno is set. */ int knet_handle_remove_datafd(knet_handle_t knet_h, int datafd); /** * knet_handle_get_channel * * @brief Get the channel associated with a file descriptor * * knet_h - pointer to knet_handle_t * * datafd - get the channel associated to this datafd * * *channel - will contain the result * * @return * knet_handle_get_channel returns * @retval 0 on success * and *channel will contain the result * @retval -1 on error and errno is set. * and *channel content is meaningless */ int knet_handle_get_channel(knet_handle_t knet_h, const int datafd, int8_t *channel); /** * knet_handle_get_datafd * * @brief Get the file descriptor associated with a channel * * knet_h - pointer to knet_handle_t * * channel - get the datafd associated to this channel * * *datafd - will contain the result * * @return * knet_handle_get_datafd returns * @retval 0 on success * and *datafd will contain the results * @retval -1 on error and errno is set. * and *datafd content is meaningless */ int knet_handle_get_datafd(knet_handle_t knet_h, const int8_t channel, int *datafd); /** * knet_recv * * @brief Receive data from knet nodes * * knet_h - pointer to knet_handle_t * * buff - pointer to buffer to store the received data * * buff_len - buffer length * * channel - channel number * * @return * knet_recv is a commodity function to wrap iovec operations * around a socket. It returns a call to readv(2). */ ssize_t knet_recv(knet_handle_t knet_h, char *buff, const size_t buff_len, const int8_t channel); /** * knet_send * * @brief Send data to knet nodes * * knet_h - pointer to knet_handle_t * * buff - pointer to the buffer of data to send * * buff_len - length of data to send * * channel - channel number * * @return * knet_send is a commodity function to wrap iovec operations * around a socket. It returns a call to writev(2). */ ssize_t knet_send(knet_handle_t knet_h, const char *buff, const size_t buff_len, const int8_t channel); /** * knet_send_sync * * @brief Synchronously send data to knet nodes * * knet_h - pointer to knet_handle_t * * buff - pointer to the buffer of data to send * * buff_len - length of data to send * * channel - data channel to use (see knet_handle_add_datafd(3)) * * All knet RX/TX operations are async for performance reasons. * There are applications that might need a sync version of data * transmission and receive errors in case of failure to deliver * to another host. * knet_send_sync bypasses the whole TX async layer and delivers * data directly to the link layer, and returns errors accordingly. * knet_send_sync sends only one packet to one host at a time. * It does NOT support multiple destinations or multicast packets. * Decision is still based on dst_host_filter_fn. * * @return * knet_send_sync returns 0 on success and -1 on error. * In addition to normal sendmmsg errors, knet_send_sync can fail * due to: * * @retval ECANCELED - data forward is disabled * @retval EFAULT - dst_host_filter fatal error * @retval EINVAL - dst_host_filter did not provide dst_host_ids_entries on unicast pckts * @retval E2BIG - dst_host_filter did return more than one dst_host_ids_entries on unicast pckts * @retval ENOMSG - received unknown message type * @retval EHOSTDOWN - unicast pckt cannot be delivered because dest host is not connected yet * @retval ECHILD - crypto failed * @retval EAGAIN - sendmmsg was unable to send all messages and there was no progress during retry * @retval ENETDOWN - a packet filter was not installed (necessary for knet_send_sync, but not knet_send) */ int knet_send_sync(knet_handle_t knet_h, const char *buff, const size_t buff_len, const int8_t channel); /** * knet_handle_enable_filter * * @brief install a filter to route packets * * knet_h - pointer to knet_handle_t * * dst_host_filter_fn_private_data * void pointer to data that can be used to identify * the callback. * * dst_host_filter_fn - * is a callback function that is invoked every time * a packet hits datafd (see knet_handle_new(3)). * the function allows users to tell libknet where the * packet has to be delivered. * * const unsigned char *outdata - is a pointer to the * current packet * ssize_t outdata_len - length of the above data * uint8_t tx_rx - filter is called on tx or rx * (KNET_NOTIFY_TX, KNET_NOTIFY_RX) * knet_node_id_t this_host_id - host_id processing the packet * knet_node_id_t src_host_id - host_id that generated the * packet * knet_node_id_t *dst_host_ids - array of KNET_MAX_HOST knet_node_id_t * where to store the destinations * (uninitialized by caller, callee should never * read it) * size_t *dst_host_ids_entries - number of hosts to send the message * * dst_host_filter_fn should return * -1 on error, packet is discarded. * 0 packet is unicast and should be sent to dst_host_ids and there are * dst_host_ids_entries in the buffer. * 1 packet is broadcast/multicast and is sent all hosts. * contents of dst_host_ids and dst_host_ids_entries are ignored. * * @return * knet_handle_enable_filter returns * 0 on success * -1 on error and errno is set. */ int knet_handle_enable_filter(knet_handle_t knet_h, void *dst_host_filter_fn_private_data, int (*dst_host_filter_fn) ( void *private_data, const unsigned char *outdata, ssize_t outdata_len, uint8_t tx_rx, knet_node_id_t this_host_id, knet_node_id_t src_host_id, int8_t *channel, knet_node_id_t *dst_host_ids, size_t *dst_host_ids_entries)); /** * knet_handle_setfwd * * @brief Start packet forwarding * * knet_h - pointer to knet_handle_t * * enable - set to 1 to allow data forwarding, 0 to disable data forwarding. * * @return * knet_handle_setfwd returns * 0 on success * -1 on error and errno is set. * * By default data forwarding is off and no traffic will pass through knet until * it is set on. */ int knet_handle_setfwd(knet_handle_t knet_h, unsigned int enabled); +/** + * knet_handle_setprio_dscp + * + * @brief Use dscp for IP_TOS on socket to implement KNET_LINK_FLAG_TRAFFICHIPRIO + * + * knet_h - pointer to knet_handle_t + * + * dscp - dscp value to set on all new sockets + * + * This function must be called prior to configure knet links. + * + * It disables the use of IPTOS_LOWDELAY and uses the given dscp value in the + * IP header's TOS field instead. + * + * Setting dscp to 0 reverts to using IPTOS_LOWDELAY. + * + * @return + * knet_handle_setprio_dscp returns + * 0 on success + * -1 on error and errno is set. + */ + +int knet_handle_setprio_dscp(knet_handle_t knet_h, uint8_t dscp); + /** * knet_handle_enable_access_lists * * @brief Enable or disable usage of access lists (default: off) * * knet_h - pointer to knet_handle_t * * enable - set to 1 to use access lists, 0 to disable access_lists. * * @return * knet_handle_enable_access_lists returns * 0 on success * -1 on error and errno is set. * * access lists are bound to links. There are 2 types of links: * 1) point to point, where both source and destinations are well known * at configuration time. * 2) open links, where only the source is known at configuration time. * * knet will automatically generate access lists for point to point links. * * For open links, knet provides 4 API calls to manipulate access lists: * knet_link_add_acl(3), knet_link_rm_acl(3), knet_link_insert_acl(3) * and knet_link_clear_acl(3). * Those API calls will work exclusively on open links as they * are of no use on point to point links. * * knet will not enforce any access list unless specifically enabled by * knet_handle_enable_access_lists(3). * * From a security / programming perspective we recommend: * - create the knet handle * - enable access lists * - configure hosts and links * - configure access lists for open links */ int knet_handle_enable_access_lists(knet_handle_t knet_h, unsigned int enabled); #define KNET_PMTUD_DEFAULT_INTERVAL 60 /** * knet_handle_pmtud_setfreq * * @brief Set the interval between PMTUd scans * * knet_h - pointer to knet_handle_t * * interval - define the interval in seconds between PMTUd scans * range from 1 to 86400 (24h) * * @return * knet_handle_pmtud_setfreq returns * 0 on success * -1 on error and errno is set. * * default interval is 60. */ int knet_handle_pmtud_setfreq(knet_handle_t knet_h, unsigned int interval); /** * knet_handle_pmtud_getfreq * * @brief Get the interval between PMTUd scans * * knet_h - pointer to knet_handle_t * * interval - pointer where to store the current interval value * * @return * knet_handle_pmtud_setfreq returns * 0 on success * -1 on error and errno is set. */ int knet_handle_pmtud_getfreq(knet_handle_t knet_h, unsigned int *interval); /** * knet_handle_enable_pmtud_notify * * @brief install a callback to receive PMTUd changes * * knet_h - pointer to knet_handle_t * * pmtud_notify_fn_private_data * void pointer to data that can be used to identify * the callback. * * pmtud_notify_fn * is a callback function that is invoked every time * a path MTU size change is detected. * The function allows libknet to notify the user * of data MTU, that's the max value that can be send * onwire without fragmentation. The data MTU will always * be lower than real link MTU because it accounts for * protocol overhead, knet packet header and (if configured) * crypto overhead, * This function MUST NEVER block or add substantial delays. * * @return * knet_handle_enable_pmtud_notify returns * 0 on success * -1 on error and errno is set. */ int knet_handle_enable_pmtud_notify(knet_handle_t knet_h, void *pmtud_notify_fn_private_data, void (*pmtud_notify_fn) ( void *private_data, unsigned int data_mtu)); /** * knet_handle_pmtud_set * * @brief Set the current interface MTU * * knet_h - pointer to knet_handle_t * * iface_mtu - current interface MTU, value 0 to 65535. 0 will * re-enable automatic MTU discovery. * In a setup with multiple interfaces, please specify * the lowest MTU between the selected intefaces. * knet will automatically adjust this value for * all headers overhead and set the correct data_mtu. * data_mtu can be retrivied with knet_handle_pmtud_get(3) * or applications will receive a pmtud_nofity event * if enabled via knet_handle_enable_pmtud_notify(3). * * @return * knet_handle_pmtud_set returns * 0 on success * -1 on error and errno is set. */ int knet_handle_pmtud_set(knet_handle_t knet_h, unsigned int iface_mtu); /** * knet_handle_pmtud_get * * @brief Get the current data MTU * * knet_h - pointer to knet_handle_t * * data_mtu - pointer where to store data_mtu * * @return * knet_handle_pmtud_get returns * 0 on success * -1 on error and errno is set. */ int knet_handle_pmtud_get(knet_handle_t knet_h, unsigned int *data_mtu); #define KNET_MIN_KEY_LEN 128 #define KNET_MAX_KEY_LEN 4096 /** * Structure passed into knet_handle_set_crypto_config() to determine * the crypto options to use for the current communications handle */ struct knet_handle_crypto_cfg { /** Model to use. nss, openssl, etc */ char crypto_model[16]; /** Cipher type name for encryption. aes 256 etc */ char crypto_cipher_type[16]; /** Hash type for digest. sha512 etc */ char crypto_hash_type[16]; /** Private key */ unsigned char private_key[KNET_MAX_KEY_LEN]; /** Length of private key */ unsigned int private_key_len; }; /** * knet_handle_crypto_set_config * * @brief set up packet cryptographic signing & encryption * * knet_h - pointer to knet_handle_t * * knet_handle_crypto_cfg - * pointer to a knet_handle_crypto_cfg structure * * crypto_model should contain the model name. * Currently only "openssl" and "nss" are supported. * Setting to "none" will disable crypto. * * crypto_cipher_type * should contain the cipher algo name. * It can be set to "none" to disable * encryption. * Currently supported by "nss" model: * "aes128", "aes192" and "aes256". * "openssl" model supports more modes and it strictly * depends on the openssl build. See: EVP_get_cipherbyname * openssl API call for details. * * crypto_hash_type * should contain the hashing algo name. * It can be set to "none" to disable * hashing. * Currently supported by "nss" model: * "md5", "sha1", "sha256", "sha384" and "sha512". * "openssl" model supports more modes and it strictly * depends on the openssl build. See: EVP_get_digestbyname * openssl API call for details. * * private_key will contain the private shared key. * It has to be at least KNET_MIN_KEY_LEN long. * * private_key_len * length of the provided private_key. * * config_num - knet supports 2 concurrent sets of crypto configurations, * to allow runtime change of crypto config and keys. * On RX both configurations will be used sequentially * in an attempt to decrypt/validate a packet (when 2 are available). * Note that this might slow down performance during a reconfiguration. * See also knet_handle_crypto_rx_clear_traffic(3) to enable / disable * processing of clear (unencrypted) traffic. * For TX, the user needs to specify which configuration to use via * knet_handle_crypto_use_config(3). * config_num accepts 0, 1 or 2 as the value. 0 should be used when * all crypto is being disabled. * Calling knet_handle_crypto_set_config(3) twice with * the same config_num will REPLACE the configuration and * NOT activate the second key. If the configuration is currently in use * EBUSY will be returned. See also knet_handle_crypto_use_config(3). * The correct sequence to perform a runtime rekey / reconfiguration * is: * - knet_handle_crypto_set_config(..., 1). -> first time config, will use config1 * - knet_handle_crypto_use_config(..., 1). -> switch TX to config 1 * - knet_handle_crypto_set_config(..., 2). -> install config2 and use it only for RX * - knet_handle_crypto_use_config(..., 2). -> switch TX to config 2 * - knet_handle_crypto_set_config(..., 1). -> with a "none"/"none"/"none" configuration to * release the resources previously allocated * The application is responsible for synchronizing calls on the nodes * to make sure the new config is in place before switching the TX configuration. * Failure to do so will result in knet being unable to talk to some of the nodes. * * Implementation notes/current limitations: * - enabling crypto, will increase latency as packets have * to processed. * - enabling crypto might reduce the overall throughtput * due to crypto data overhead. * - private/public key encryption/hashing is not currently * planned. * - crypto key must be the same for all hosts in the same * knet instance / configX. * - it is safe to call knet_handle_crypto_set_config multiple times at runtime. * The last config will be used. * IMPORTANT: a call to knet_handle_crypto_set_config can fail due to: * 1) failure to obtain locking * 2) errors to initializing the crypto level. * This can happen even in subsequent calls to knet_handle_crypto_set_config(3). * A failure in crypto init will restore the previous crypto configuration if any. * * @return * knet_handle_crypto_set_config returns: * @retval 0 on success * @retval -1 on error and errno is set. * @retval -2 on crypto subsystem initialization error. No errno is provided at the moment (yet). */ int knet_handle_crypto_set_config(knet_handle_t knet_h, struct knet_handle_crypto_cfg *knet_handle_crypto_cfg, uint8_t config_num); #define KNET_CRYPTO_RX_ALLOW_CLEAR_TRAFFIC 0 #define KNET_CRYPTO_RX_DISALLOW_CLEAR_TRAFFIC 1 /** * knet_handle_crypto_rx_clear_traffic * * @brief enable or disable RX processing of clear (unencrypted) traffic * * knet_h - pointer to knet_handle_t * * value - KNET_CRYPTO_RX_ALLOW_CLEAR_TRAFFIC or KNET_CRYPTO_RX_DISALLOW_CLEAR_TRAFFIC * * @return * knet_handle_crypto_use_config returns: * @retval 0 on success * @retval -1 on error and errno is set. */ int knet_handle_crypto_rx_clear_traffic(knet_handle_t knet_h, uint8_t value); /** * knet_handle_crypto_use_config * * @brief specify crypto configuration to use for TX * * knet_h - pointer to knet_handle_t * * config_num - 1|2 use configuration 1 or 2, 0 for clear (unencrypted) traffic. * * @return * knet_handle_crypto_use_config returns: * @retval 0 on success * @retval -1 on error and errno is set. */ int knet_handle_crypto_use_config(knet_handle_t knet_h, uint8_t config_num); /** * knet_handle_crypto * * @brief set up packet cryptographic signing & encryption * * knet_h - pointer to knet_handle_t * * knet_handle_crypto_cfg - * pointer to a knet_handle_crypto_cfg structure * see knet_handle_crypto_set_config(3) for details. * * * Implementation notes: * * knet_handle_crypto(3) is now a wrapper for knet_handle_crypto_set_config(3) * and knet_handle_crypto_use_config(3) with config_num set to 1. * * @return * knet_handle_crypto returns: * @retval 0 on success * @retval -1 on error and errno is set. * @retval -2 on crypto subsystem initialization error. No errno is provided at the moment (yet). */ int knet_handle_crypto(knet_handle_t knet_h, struct knet_handle_crypto_cfg *knet_handle_crypto_cfg); #define KNET_COMPRESS_THRESHOLD 100 /** * Structure passed into knet_handle_compress() * to tell knet what type of compression to use * for this communiction */ struct knet_handle_compress_cfg { /** Compression library to use, bzip2 etc... */ char compress_model[16]; /** Threshold. Packets smaller than this will not be compressed */ uint32_t compress_threshold; /** Passed into the compression library as an indication of the level of compression to apply */ int compress_level; }; /** * knet_handle_compress * * @brief Set up packet compression * * knet_h - pointer to knet_handle_t * * knet_handle_compress_cfg - * pointer to a knet_handle_compress_cfg structure * * compress_model contains the model name. * See "compress_level" for the list of accepted values. * Setting the value to "none" disables compression. * * compress_threshold * tells the transmission thread to NOT compress * any packets that are smaller than the value * indicated. Default 100 bytes. * Set to 0 to reset to the default. * Set to 1 to compress everything. * Max accepted value is KNET_MAX_PACKET_SIZE. * * compress_level is the "level" parameter for most models: * zlib: 0 (no compression), 1 (minimal) .. 9 (max compression). * lz4: 1 (max compression)... 9 (fastest compression). * lz4hc: 1 (min compression) ... LZ4HC_MAX_CLEVEL (16) or LZ4HC_CLEVEL_MAX (12) * depending on the version of lz4hc libknet was built with. * lzma: 0 (minimal) .. 9 (max compression) * bzip2: 1 (minimal) .. 9 (max compression) * For lzo2 it selects the algorithm to use: * 1 : lzo1x_1_compress (default) * 11 : lzo1x_1_11_compress * 12 : lzo1x_1_12_compress * 15 : lzo1x_1_15_compress * 999: lzo1x_999_compress * Other values select the default algorithm. * Please refer to the documentation of the respective * compression library for guidance about setting this * value. * * Implementation notes: * - it is possible to enable/disable compression at any time. * - nodes can be using a different compression algorithm at any time. * - knet does NOT implement the compression algorithm directly. it relies * on external libraries for this functionality. Please read * the libraries man pages to figure out which algorithm/compression * level is best for the data you are planning to transmit. * * @return * knet_handle_compress returns * 0 on success * -1 on error and errno is set. EINVAL means that either the model or the * level are not supported. */ int knet_handle_compress(knet_handle_t knet_h, struct knet_handle_compress_cfg *knet_handle_compress_cfg); /** * Detailed stats for this knet handle as returned by knet_handle_get_stats() */ struct knet_handle_stats { /** Size of the structure. set this to sizeof(struct knet_handle_stats) before calling */ size_t size; /** Number of uncompressed packets sent */ uint64_t tx_uncompressed_packets; /** Number of compressed packets sent */ uint64_t tx_compressed_packets; /** Number of bytes sent (as if uncompressed, ie actual data bytes) */ uint64_t tx_compressed_original_bytes; /** Number of bytes sent on the wire after compression */ uint64_t tx_compressed_size_bytes; /** Average(mean) time take to compress transmitted packets */ uint64_t tx_compress_time_ave; /** Minimum time taken to compress transmitted packets */ uint64_t tx_compress_time_min; /** Maximum time taken to compress transmitted packets */ uint64_t tx_compress_time_max; /** Number of compressed packets received */ uint64_t rx_compressed_packets; /** Number of bytes received - after decompression */ uint64_t rx_compressed_original_bytes; /** Number of compressed bytes received before decompression */ uint64_t rx_compressed_size_bytes; /** Average(mean) time take to decompress received packets */ uint64_t rx_compress_time_ave; /** Minimum time take to decompress received packets */ uint64_t rx_compress_time_min; /** Maximum time take to decompress received packets */ uint64_t rx_compress_time_max; /** Number of encrypted packets sent */ uint64_t tx_crypt_packets; /** Cumulative byte overhead of encrypted traffic */ uint64_t tx_crypt_byte_overhead; /** Average(mean) time take to encrypt packets in usecs */ uint64_t tx_crypt_time_ave; /** Minimum time take to encrypto packets in usecs */ uint64_t tx_crypt_time_min; /** Maximum time take to encrypto packets in usecs */ uint64_t tx_crypt_time_max; /** Number of encrypted packets received */ uint64_t rx_crypt_packets; /** Average(mean) time take to decrypt received packets */ uint64_t rx_crypt_time_ave; /** Minimum time take to decrypt received packets in usecs */ uint64_t rx_crypt_time_min; /** Maximum time take to decrypt received packets in usecs */ uint64_t rx_crypt_time_max; }; /** * knet_handle_get_stats * * @brief Get statistics for compression & crypto * * knet_h - pointer to knet_handle_t * * knet_handle_stats * pointer to a knet_handle_stats structure * * struct_size * size of knet_handle_stats structure to allow * for backwards compatibility. libknet will only * copy this much data into the stats structure * so that older callers will not get overflowed if * new fields are added. * * @return * 0 on success * -1 on error and errno is set. * */ int knet_handle_get_stats(knet_handle_t knet_h, struct knet_handle_stats *stats, size_t struct_size); /* * Tell knet_handle_clear_stats whether to clear just the handle stats * or all of them. */ #define KNET_CLEARSTATS_HANDLE_ONLY 1 #define KNET_CLEARSTATS_HANDLE_AND_LINK 2 /** * knet_handle_clear_stats * * @brief Clear knet stats, link and/or handle * * knet_h - pointer to knet_handle_t * * clear_option - Which stats to clear, must be one of * * KNET_CLEARSTATS_HANDLE_ONLY or * KNET_CLEARSTATS_HANDLE_AND_LINK * * @return * 0 on success * -1 on error and errno is set. * */ int knet_handle_clear_stats(knet_handle_t knet_h, int clear_option); /** * Structure returned from get_crypto_list() containing * information about the installed cryptographic systems */ struct knet_crypto_info { /** Name of the crypto library/ openssl, nss,etc .. */ const char *name; /** Properties - currently unused */ uint8_t properties; /** Currently unused padding */ char pad[256]; }; /** * knet_get_crypto_list * * @brief Get a list of supported crypto libraries * * crypto_list - array of struct knet_crypto_info * * If NULL then only the number of structs is returned in crypto_list_entries * to allow the caller to allocate sufficient space. * libknet does not allow more than 256 crypto methods at the moment. * it is safe to allocate 256 structs to avoid calling * knet_get_crypto_list twice. * * crypto_list_entries - returns the number of structs in crypto_list * * @return * knet_get_crypto_list returns * 0 on success * -1 on error and errno is set. */ int knet_get_crypto_list(struct knet_crypto_info *crypto_list, size_t *crypto_list_entries); /** * Structure returned from get_compress_list() containing * information about the installed compression systems */ struct knet_compress_info { /** Name of the compression type bzip2, lz4, etc.. */ const char *name; /** Properties - currently unused */ uint8_t properties; /** Currently unused padding */ char pad[256]; }; /** * knet_get_compress_list * * @brief Get a list of support compression types * * compress_list - array of struct knet_compress_info * * If NULL then only the number of structs is returned in compress_list_entries * to allow the caller to allocate sufficient space. * libknet does not allow more than 256 compress methods at the moment. * it is safe to allocate 256 structs to avoid calling * knet_get_compress_list twice. * * compress_list_entries - returns the number of structs in compress_list * * @return * knet_get_compress_list returns * 0 on success * -1 on error and errno is set. */ int knet_get_compress_list(struct knet_compress_info *compress_list, size_t *compress_list_entries); /* * host structs/API calls */ /** * knet_host_add * * @brief Add a new host ID to knet * * knet_h - pointer to knet_handle_t * * host_id - each host in a knet is identified with a unique ID * (see also knet_handle_new(3)) * * @return * knet_host_add returns: * 0 on success * -1 on error and errno is set. */ int knet_host_add(knet_handle_t knet_h, knet_node_id_t host_id); /** * knet_host_remove * * @brief Remove a host ID from knet * * knet_h - pointer to knet_handle_t * * host_id - each host in a knet is identified with a unique ID * (see also knet_handle_new(3)) * * @return * knet_host_remove returns: * 0 on success * -1 on error and errno is set. */ int knet_host_remove(knet_handle_t knet_h, knet_node_id_t host_id); /** * knet_host_set_name * * @brief Set the name of a knet host * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * name - this name will be used for pretty logging and eventually * search for hosts (see also knet_handle_host_get_name(2) and knet_handle_host_get_id(3)). * Only up to KNET_MAX_HOST_LEN - 1 bytes will be accepted and * name has to be unique for each host. * * @return * knet_host_set_name returns: * 0 on success * -1 on error and errno is set. */ int knet_host_set_name(knet_handle_t knet_h, knet_node_id_t host_id, const char *name); /** * knet_host_get_name_by_host_id * * @brief Get the name of a host given its ID * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * name - pointer to a preallocated buffer of at least size KNET_MAX_HOST_LEN * where the current host name will be stored * (as set by knet_host_set_name or default by knet_host_add) * * @return * knet_host_get_name_by_host_id returns: * 0 on success * -1 on error and errno is set (name is left untouched) */ int knet_host_get_name_by_host_id(knet_handle_t knet_h, knet_node_id_t host_id, char *name); /** * knet_host_get_id_by_host_name * * @brief Get the ID of a host given its name * * knet_h - pointer to knet_handle_t * * name - name to lookup, max len KNET_MAX_HOST_LEN * * host_id - where to store the result * * @return * knet_host_get_id_by_host_name returns: * 0 on success * -1 on error and errno is set. */ int knet_host_get_id_by_host_name(knet_handle_t knet_h, const char *name, knet_node_id_t *host_id); /** * knet_host_get_host_list * * @brief Get a list of hosts known to knet * * knet_h - pointer to knet_handle_t * * host_ids - array of at lest KNET_MAX_HOST size * * host_ids_entries - * number of entries writted in host_ids * * @return * knet_host_get_host_list returns * 0 on success * -1 on error and errno is set. */ int knet_host_get_host_list(knet_handle_t knet_h, knet_node_id_t *host_ids, size_t *host_ids_entries); /* * define switching policies */ #define KNET_LINK_POLICY_PASSIVE 0 #define KNET_LINK_POLICY_ACTIVE 1 #define KNET_LINK_POLICY_RR 2 /** * knet_host_set_policy * * @brief Set the switching policy for a host's links * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * policy - there are currently 3 kind of simple switching policies * based on link configuration. * KNET_LINK_POLICY_PASSIVE - the active link with the highest * priority (highest number) will be used. * if one or more active links share * the same priority, the one with * lowest link_id will be used. * * KNET_LINK_POLICY_ACTIVE - all active links will be used * simultaneously to send traffic. * link priority is ignored. * * KNET_LINK_POLICY_RR - round-robin policy, every packet * will be send on a different active * link. * * @return * knet_host_set_policy returns * 0 on success * -1 on error and errno is set. */ int knet_host_set_policy(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t policy); /** * knet_host_get_policy * * @brief Get the switching policy for a host's links * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * policy - will contain the current configured switching policy. * Default is passive when creating a new host. * * @return * knet_host_get_policy returns * 0 on success * -1 on error and errno is set. */ int knet_host_get_policy(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t *policy); /** * knet_host_enable_status_change_notify * * @brief Install a callback to get host status change events * * knet_h - pointer to knet_handle_t * * host_status_change_notify_fn_private_data - * void pointer to data that can be used to identify * the callback * * host_status_change_notify_fn - * is a callback function that is invoked every time * there is a change in the host status. * host status is identified by: * - reachable, this host can send/receive data to/from host_id * - remote, 0 if the host_id is connected locally or 1 if * the there is one or more knet host(s) in between. * NOTE: re-switching is NOT currently implemented, * but this is ready for future and can avoid * an API/ABI breakage later on. * - external, 0 if the host_id is configured locally or 1 if * it has been added from remote nodes config. * NOTE: dynamic topology is NOT currently implemented, * but this is ready for future and can avoid * an API/ABI breakage later on. * This function MUST NEVER block or add substantial delays. * * @return * knet_host_status_change_notify returns * 0 on success * -1 on error and errno is set. */ int knet_host_enable_status_change_notify(knet_handle_t knet_h, void *host_status_change_notify_fn_private_data, void (*host_status_change_notify_fn) ( void *private_data, knet_node_id_t host_id, uint8_t reachable, uint8_t remote, uint8_t external)); /* * define host status structure for quick lookup * struct is in flux as more stats will be added soon * * reachable host_id can be seen either directly connected * or via another host_id * * remote 0 = node is connected locally, 1 is visible via * via another host_id * * external 0 = node is configured/known locally, * 1 host_id has been received via another host_id */ /** * status of a knet host, returned from knet_host_get_status() */ struct knet_host_status { /** Whether the host is currently reachable */ uint8_t reachable; /** Whether the host is a remote node (not currently implemented) */ uint8_t remote; /** Whether the host is external (not currently implemented) */ uint8_t external; /* add host statistics */ }; /** * knet_host_get_status * * @brief Get the status of a host * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * status - pointer to knet_host_status struct * * @return * knet_handle_pmtud_get returns * 0 on success * -1 on error and errno is set. */ int knet_host_get_status(knet_handle_t knet_h, knet_node_id_t host_id, struct knet_host_status *status); /* * link structs/API calls * * every host allocated/managed by knet_host_* has * KNET_MAX_LINK structures to define the network * paths that connect 2 hosts. * * Each link is identified by a link_id that has a * values between 0 and KNET_MAX_LINK - 1. * * KNOWN LIMITATIONS: * * - let's assume the scenario where two hosts are connected * with any number of links. link_id must match on both sides. * If host_id 0 link_id 0 is configured to connect IP1 to IP2 and * host_id 0 link_id 1 is configured to connect IP3 to IP4, * host_id 1 link_id 0 _must_ connect IP2 to IP1 and likewise * host_id 1 link_id 1 _must_ connect IP4 to IP3. * We might be able to lift this restriction in future, by using * other data to determine src/dst link_id, but for now, deal with it. */ /* * commodity functions to convert strings to sockaddr and viceversa */ /** * knet_strtoaddr * * @brief Convert a hostname string to an address * * host - IPaddr/hostname to convert * be aware only the first IP address will be returned * in case a hostname resolves to multiple IP * * port - port to connect to * * ss - sockaddr_storage where to store the converted data * * sslen - len of the sockaddr_storage * * @return * knet_strtoaddr returns same error codes as getaddrinfo * */ int knet_strtoaddr(const char *host, const char *port, struct sockaddr_storage *ss, socklen_t sslen); /** * knet_addrtostr * * @brief Convert an address to a host name * * ss - sockaddr_storage to convert * * sslen - len of the sockaddr_storage * * host - IPaddr/hostname where to store data * (recommended size: KNET_MAX_HOST_LEN) * * port - port buffer where to store data * (recommended size: KNET_MAX_PORT_LEN) * * @return * knet_strtoaddr returns same error codes as getnameinfo */ int knet_addrtostr(const struct sockaddr_storage *ss, socklen_t sslen, char *addr_buf, size_t addr_buf_size, char *port_buf, size_t port_buf_size); #define KNET_TRANSPORT_LOOPBACK 0 #define KNET_TRANSPORT_UDP 1 #define KNET_TRANSPORT_SCTP 2 #define KNET_MAX_TRANSPORTS UINT8_MAX /* * The Loopback transport is only valid for connections to localhost, the host * with the same node_id specified in knet_handle_new(). Only one link of this * type is allowed. Data sent down a LOOPBACK link will be copied directly from * the knet send datafd to the knet receive datafd so the application must be set * up to take data from that socket at least as often as it is sent or deadlocks * could occur. If used, a LOOPBACK link must be the only link configured to the * local host. */ /** * Transport information returned from knet_get_transport_list() */ struct knet_transport_info { /** Transport name. UDP, SCTP, etc... */ const char *name; /** value that can be used for knet_link_set_config() */ uint8_t id; /** currently unused */ uint8_t properties; /** currently unused */ char pad[256]; }; /** * knet_get_transport_list * * @brief Get a list of the transports support by this build of knet * * transport_list - an array of struct transport_info that must be * at least of size struct transport_info * KNET_MAX_TRANSPORTS * * transport_list_entries - pointer to a size_t where to store how many transports * are available in this build of libknet. * * @return * knet_get_transport_list returns * 0 on success * -1 on error and errno is set. */ int knet_get_transport_list(struct knet_transport_info *transport_list, size_t *transport_list_entries); /** * knet_get_transport_name_by_id * * @brief Get a transport name from its ID number * * transport - one of the KNET_TRANSPORT_xxx constants * * @return * knet_get_transport_name_by_id returns: * * @retval pointer to the name on success or * @retval NULL on error and errno is set. */ const char *knet_get_transport_name_by_id(uint8_t transport); /** * knet_get_transport_id_by_name * * @brief Get a transport ID from its name * * name - transport name (UDP/SCTP/etc) * * @return * knet_get_transport_name_by_id returns: * * @retval KNET_MAX_TRANSPORTS on error and errno is set accordingly * @retval KNET_TRANSPORT_xxx on success. */ uint8_t knet_get_transport_id_by_name(const char *name); #define KNET_TRANSPORT_DEFAULT_RECONNECT_INTERVAL 1000 /** * knet_handle_set_transport_reconnect_interval * * @brief Set the interval between transport attempts to reconnect a failed link * * knet_h - pointer to knet_handle_t * * msecs - milliseconds * * @return * knet_handle_set_transport_reconnect_interval returns * 0 on success * -1 on error and errno is set. */ int knet_handle_set_transport_reconnect_interval(knet_handle_t knet_h, uint32_t msecs); /** * knet_handle_get_transport_reconnect_interval * * @brief Get the interval between transport attempts to reconnect a failed link * * knet_h - pointer to knet_handle_t * * msecs - milliseconds * * @return * knet_handle_get_transport_reconnect_interval returns * 0 on success * -1 on error and errno is set. */ int knet_handle_get_transport_reconnect_interval(knet_handle_t knet_h, uint32_t *msecs); /** * knet_link_set_config * * @brief Configure the link to a host * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * transport - one of the KNET_TRANSPORT_xxx constants * * src_addr - sockaddr_storage that can be either IPv4 or IPv6 * * dst_addr - sockaddr_storage that can be either IPv4 or IPv6 * this can be null if we don't know the incoming * IP address/port and the link will remain quiet * till the node on the other end will initiate a * connection * * flags - KNET_LINK_FLAG_* * * @return * knet_link_set_config returns * 0 on success * -1 on error and errno is set. */ int knet_link_set_config(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t transport, struct sockaddr_storage *src_addr, struct sockaddr_storage *dst_addr, uint64_t flags); /** * knet_link_get_config * * @brief Get the link configutation information * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * transport - see knet_link_set_config(3) * * src_addr - sockaddr_storage that can be either IPv4 or IPv6 * * dst_addr - sockaddr_storage that can be either IPv4 or IPv6 * * dynamic - 0 if dst_addr is static or 1 if dst_addr is dynamic. * In case of 1, dst_addr can be NULL and it will be left * untouched. * * flags - KNET_LINK_FLAG_* * * @return * knet_link_get_config returns * 0 on success. * -1 on error and errno is set. */ int knet_link_get_config(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t *transport, struct sockaddr_storage *src_addr, struct sockaddr_storage *dst_addr, uint8_t *dynamic, uint64_t *flags); /** * knet_link_clear_config * * @brief Clear link information and disconnect the link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * @return * knet_link_clear_config returns * 0 on success. * -1 on error and errno is set. */ int knet_link_clear_config(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id); /* * Access lists management for open links * see also knet_handle_enable_access_lists(3) */ /** * check_type_t * @brief address type enum for knet access lists * * CHECK_TYPE_ADDRESS is the equivalent of a single entry / IP address. * for example: 10.1.9.3 * and the entry is stored in ss1. ss2 can be NULL. * * CHECK_TYPE_MASK is used to configure network/netmask. * for example: 192.168.0.0/24 * the network is stored in ss1 and the netmask in ss2. * * CHECK_TYPE_RANGE defines a value / range of ip addresses. * for example: 172.16.0.1-172.16.0.10 * the start is stored in ss1 and the end in ss2. * * Please be aware that the above examples refer only to IP based protocols. * Other protocols might use ss1 and ss2 in slightly different ways. * At the moment knet only supports IP based protocol, though that might change * in the future. */ typedef enum { CHECK_TYPE_ADDRESS, CHECK_TYPE_MASK, CHECK_TYPE_RANGE } check_type_t; /** * check_acceptreject_t * * @brief enum for accept/reject in knet access lists * * accept or reject incoming packets defined in the access list entry */ typedef enum { CHECK_ACCEPT, CHECK_REJECT } check_acceptreject_t; /** * knet_link_add_acl * * @brief Add access list entry to an open link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * ss1 / ss2 / type / acceptreject - see typedef definitions for details * * IMPORTANT: the order in which access lists are added is critical and it * is left to the user to add them in the right order. knet * will not attempt to logically sort them. * * For example: * 1 - accept from 10.0.0.0/8 * 2 - reject from 10.0.0.1/32 * * is not the same as: * * 1 - reject from 10.0.0.1/32 * 2 - accept from 10.0.0.0/8 * * In the first example, rule number 2 will never match because * packets from 10.0.0.1 will be accepted by rule number 1. * * @return * knet_link_add_acl returns * 0 on success. * -1 on error and errno is set. */ int knet_link_add_acl(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, struct sockaddr_storage *ss1, struct sockaddr_storage *ss2, check_type_t type, check_acceptreject_t acceptreject); /** * knet_link_insert_acl * * @brief Insert access list entry to an open link at given index * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * index - insert at position "index" where 0 is the first entry and -1 * appends to the current list. * * ss1 / ss2 / type / acceptreject - see typedef definitions for details * * @return * knet_link_insert_acl returns * 0 on success. * -1 on error and errno is set. */ int knet_link_insert_acl(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, int index, struct sockaddr_storage *ss1, struct sockaddr_storage *ss2, check_type_t type, check_acceptreject_t acceptreject); /** * knet_link_rm_acl * * @brief Remove access list entry from an open link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * ss1 / ss2 / type / acceptreject - see typedef definitions for details * * IMPORTANT: the data passed to this API call must match exactly that passed * to knet_link_add_acl(3). * * @return * knet_link_rm_acl returns * 0 on success. * -1 on error and errno is set. */ int knet_link_rm_acl(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, struct sockaddr_storage *ss1, struct sockaddr_storage *ss2, check_type_t type, check_acceptreject_t acceptreject); /** * knet_link_clear_acl * * @brief Remove all access list entries from an open link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * @return * knet_link_clear_acl returns * 0 on success. * -1 on error and errno is set. */ int knet_link_clear_acl(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id); /** * knet_link_set_enable * * @brief Enable traffic on a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * enabled - 0 disable the link, 1 enable the link * * @return * knet_link_set_enable returns * 0 on success * -1 on error and errno is set. */ int knet_link_set_enable(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, unsigned int enabled); /** * knet_link_get_enable * * @brief Find out whether a link is enabled or not * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * enabled - 0 disable the link, 1 enable the link * * @return * knet_link_get_enable returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_enable(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, unsigned int *enabled); #define KNET_LINK_DEFAULT_PING_INTERVAL 1000 /* 1 second */ #define KNET_LINK_DEFAULT_PING_TIMEOUT 2000 /* 2 seconds */ #define KNET_LINK_DEFAULT_PING_PRECISION 2048 /* samples */ /** * knet_link_set_ping_timers * * @brief Set the ping timers for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * interval - specify the ping interval in milliseconds. * * timeout - if no pong is received within this time, * the link is declared dead, in milliseconds. * NOTE: in future it will be possible to set timeout to 0 * for an autocalculated timeout based on interval, pong_count * and latency. The API already accept 0 as value and it will * return ENOSYS / -1. Once the automatic calculation feature * will be implemented, this call will only return EINVAL * for incorrect values. * * precision - how many values of latency are used to calculate * the average link latency (see also knet_link_get_status(3)) * * @return * knet_link_set_ping_timers returns * 0 on success * -1 on error and errno is set. */ int knet_link_set_ping_timers(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, time_t interval, time_t timeout, unsigned int precision); /** * knet_link_get_ping_timers * * @brief Get the ping timers for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * interval - ping interval * * timeout - if no pong is received within this time, * the link is declared dead * * precision - how many values of latency are used to calculate * the average link latency (see also knet_link_get_status(3)) * * @return * knet_link_get_ping_timers returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_ping_timers(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, time_t *interval, time_t *timeout, unsigned int *precision); #define KNET_LINK_DEFAULT_PONG_COUNT 5 /** * knet_link_set_pong_count * * @brief Set the pong count for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * pong_count - how many valid ping/pongs before a link is marked UP. * default: 5, value should be > 0 * * @return * knet_link_set_pong_count returns * 0 on success * -1 on error and errno is set. */ int knet_link_set_pong_count(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t pong_count); /** * knet_link_get_pong_count * * @brief Get the pong count for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * pong_count - how many valid ping/pongs before a link is marked UP. * default: 5, value should be > 0 * * @return * knet_link_get_pong_count returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_pong_count(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t *pong_count); /** * knet_link_set_priority * * @brief Set the priority for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * priority - specify the switching priority for this link * see also knet_host_set_policy * * @return * knet_link_set_priority returns * 0 on success * -1 on error and errno is set. */ int knet_link_set_priority(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t priority); /** * knet_link_get_priority * * @brief Get the priority for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * priority - gather the switching priority for this link * see also knet_host_set_policy * * @return * knet_link_get_priority returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_priority(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, uint8_t *priority); /** * knet_link_get_link_list * * @brief Get a list of links connecting a host * * knet_h - pointer to knet_handle_t * * link_ids - array of at lest KNET_MAX_LINK size * with the list of configured links for a certain host. * * link_ids_entries - * number of entries contained in link_ids * * @return * knet_link_get_link_list returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_link_list(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t *link_ids, size_t *link_ids_entries); /* * define link status structure for quick lookup * * src/dst_{ipaddr,port} strings are filled by * getnameinfo(3) when configuring the link. * if the link is dynamic (see knet_link_set_config(3)) * dst_ipaddr/port will contain ipaddr/port of the currently * connected peer or "Unknown" if it was not possible * to determine the ipaddr/port at runtime. * * enabled see also knet_link_set/get_enable. * * connected the link is connected to a peer and ping/pong traffic * is flowing. * * dynconnected the link has dynamic ip on the other end, and * we can see the other host is sending pings to us. * * latency average latency of this link * see also knet_link_set/get_timeout. * * pong_last if the link is down, this value tells us how long * ago this link was active. A value of 0 means that the link * has never been active. * * knet_link_stats structure that contains details statistics for the link */ #define MAX_LINK_EVENTS 16 /** * Stats for a knet link * returned from knet_link_get_status() as part of a knet_link_status structure * link stats are 'onwire', ie they indicate the number of actual bytes/packets * sent including overheads, not just data packets. */ struct knet_link_stats { /** Number of data packets sent */ uint64_t tx_data_packets; /** Number of data packets received */ uint64_t rx_data_packets; /** Number of data bytes sent */ uint64_t tx_data_bytes; /** Number of data bytes received */ uint64_t rx_data_bytes; /** Number of ping packets sent */ uint64_t rx_ping_packets; /** Number of ping packets received */ uint64_t tx_ping_packets; /** Number of ping bytes sent */ uint64_t rx_ping_bytes; /** Number of ping bytes received */ uint64_t tx_ping_bytes; /** Number of pong packets sent */ uint64_t rx_pong_packets; /** Number of pong packets received */ uint64_t tx_pong_packets; /** Number of pong bytes sent */ uint64_t rx_pong_bytes; /** Number of pong bytes received */ uint64_t tx_pong_bytes; /** Number of pMTU packets sent */ uint64_t rx_pmtu_packets; /** Number of pMTU packets received */ uint64_t tx_pmtu_packets; /** Number of pMTU bytes sent */ uint64_t rx_pmtu_bytes; /** Number of pMTU bytes received */ uint64_t tx_pmtu_bytes; /* These are only filled in when requested ie. they are not collected in realtime */ /** Total of all packets sent */ uint64_t tx_total_packets; /** Total of all packets received */ uint64_t rx_total_packets; /** Total number of bytes sent */ uint64_t tx_total_bytes; /** Total number of bytes received */ uint64_t rx_total_bytes; /** Total number of errors that occurred while sending */ uint64_t tx_total_errors; /** Total number of retries that occurred while sending */ uint64_t tx_total_retries; /** Total number of errors that occurred while sending pMTU packets */ uint32_t tx_pmtu_errors; /** Total number of retries that occurred while sending pMTU packets */ uint32_t tx_pmtu_retries; /** Total number of errors that occurred while sending ping packets */ uint32_t tx_ping_errors; /** Total number of retries that occurred while sending ping packets */ uint32_t tx_ping_retries; /** Total number of errors that occurred while sending pong packets */ uint32_t tx_pong_errors; /** Total number of retries that occurred while sending pong packets */ uint32_t tx_pong_retries; /** Total number of errors that occurred while sending data packets */ uint32_t tx_data_errors; /** Total number of retries that occurred while sending data packets */ uint32_t tx_data_retries; /** Minimum latency measured in usecs */ uint32_t latency_min; /** Maximum latency measured in usecs */ uint32_t latency_max; /** Average(mean) latency measured in usecs */ uint32_t latency_ave; /** Number of samples used to calculate latency */ uint32_t latency_samples; /** How many times the link has gone down */ uint32_t down_count; /** How many times the link has come up */ uint32_t up_count; /** * A circular buffer of time_t structs collecting the history * of up events on this link. * The index indicates current/last event. * it is safe to walk back the history by decreasing the index */ time_t last_up_times[MAX_LINK_EVENTS]; /** * A circular buffer of time_t structs collecting the history * of down events on this link. * The index indicates current/last event. * it is safe to walk back the history by decreasing the index */ time_t last_down_times[MAX_LINK_EVENTS]; /** Index of last element in the last_up_times[] array */ int8_t last_up_time_index; /** Index of last element in the last_down_times[] array */ int8_t last_down_time_index; /* Always add new stats at the end */ }; /** * Status of a knet link as returned from knet_link_get_status() */ struct knet_link_status { /** Size of the structure for ABI checking, set this to sizeof(knet_link_status) before calling knet_link_get_status() */ size_t size; /** Local IP address as a string*/ char src_ipaddr[KNET_MAX_HOST_LEN]; /** Local IP port as a string */ char src_port[KNET_MAX_PORT_LEN]; /** Remote IP address as a string */ char dst_ipaddr[KNET_MAX_HOST_LEN]; /** Remote IP port as a string*/ char dst_port[KNET_MAX_PORT_LEN]; /** Link is configured and admin enabled for traffic */ uint8_t enabled; /** Link is connected for data (local view) */ uint8_t connected; /** Link has been activated by remote dynip */ uint8_t dynconnected; /** average latency computed by fix/exp */ unsigned long long latency; /** Timestamp of the past pong received */ struct timespec pong_last; /** Currently detected MTU on this link */ unsigned int mtu; /** * Contains the size of the IP protocol, knet headers and * crypto headers (if configured). This value is filled in * ONLY after the first PMTUd run on that given link, * and can change if link configuration or crypto configuration * changes at runtime. * WARNING: in general mtu + proto_overhead might or might * not match the output of ifconfig mtu due to crypto * requirements to pad packets to some specific boundaries. */ unsigned int proto_overhead; /** Link statistics */ struct knet_link_stats stats; }; /** * knet_link_get_status * * @brief Get the status (and statistics) for a link * * knet_h - pointer to knet_handle_t * * host_id - see knet_host_add(3) * * link_id - see knet_link_set_config(3) * * status - pointer to knet_link_status struct * * struct_size - max size of knet_link_status - allows library to * add fields without ABI change. Returned structure * will be truncated to this length and .size member * indicates the full size. * * @return * knet_link_get_status returns * 0 on success * -1 on error and errno is set. */ int knet_link_get_status(knet_handle_t knet_h, knet_node_id_t host_id, uint8_t link_id, struct knet_link_status *status, size_t struct_size); /* * logging structs/API calls */ /* * libknet is composed of several subsystems. In order * to easily distinguish log messages coming from different * places, each subsystem has its own ID. * * 0-19 config/management * 20-39 internal threads * 40-59 transports * 60-69 crypto implementations */ #define KNET_SUB_COMMON 0 /* common.c */ #define KNET_SUB_HANDLE 1 /* handle.c alloc/dealloc config changes */ #define KNET_SUB_HOST 2 /* host add/del/modify */ #define KNET_SUB_LISTENER 3 /* listeners add/del/modify... */ #define KNET_SUB_LINK 4 /* link add/del/modify */ #define KNET_SUB_TRANSPORT 5 /* Transport common */ #define KNET_SUB_CRYPTO 6 /* crypto.c config generic layer */ #define KNET_SUB_COMPRESS 7 /* compress.c config generic layer */ #define KNET_SUB_FILTER 19 /* allocated for users to log from dst_filter */ #define KNET_SUB_DSTCACHE 20 /* switching thread (destination cache handling) */ #define KNET_SUB_HEARTBEAT 21 /* heartbeat thread */ #define KNET_SUB_PMTUD 22 /* Path MTU Discovery thread */ #define KNET_SUB_TX 23 /* send to link thread */ #define KNET_SUB_RX 24 /* recv from link thread */ #define KNET_SUB_TRANSP_BASE 40 /* Base log level for transports */ #define KNET_SUB_TRANSP_LOOPBACK (KNET_SUB_TRANSP_BASE + KNET_TRANSPORT_LOOPBACK) #define KNET_SUB_TRANSP_UDP (KNET_SUB_TRANSP_BASE + KNET_TRANSPORT_UDP) #define KNET_SUB_TRANSP_SCTP (KNET_SUB_TRANSP_BASE + KNET_TRANSPORT_SCTP) #define KNET_SUB_NSSCRYPTO 60 /* nsscrypto.c */ #define KNET_SUB_OPENSSLCRYPTO 61 /* opensslcrypto.c */ #define KNET_SUB_ZLIBCOMP 70 /* compress_zlib.c */ #define KNET_SUB_LZ4COMP 71 /* compress_lz4.c */ #define KNET_SUB_LZ4HCCOMP 72 /* compress_lz4.c */ #define KNET_SUB_LZO2COMP 73 /* compress_lzo.c */ #define KNET_SUB_LZMACOMP 74 /* compress_lzma.c */ #define KNET_SUB_BZIP2COMP 75 /* compress_bzip2.c */ #define KNET_SUB_ZSTDCOMP 76 /* compress_zstd.c */ #define KNET_SUB_UNKNOWN UINT8_MAX - 1 #define KNET_MAX_SUBSYSTEMS UINT8_MAX /* * Convert between subsystem IDs and names */ /** * knet_log_get_subsystem_name * * @brief Get a logging system name from its numeric ID * * @return * returns internal name of the subsystem or "common" */ const char *knet_log_get_subsystem_name(uint8_t subsystem); /** * knet_log_get_subsystem_id * * @brief Get a logging system ID from its name * * @return * returns internal ID of the subsystem or KNET_SUB_COMMON */ uint8_t knet_log_get_subsystem_id(const char *name); /* * 5 log levels are enough for everybody */ #define KNET_LOG_ERR 0 /* unrecoverable errors/conditions */ #define KNET_LOG_WARN 1 /* recoverable errors/conditions */ #define KNET_LOG_INFO 2 /* info, link up/down, config changes.. */ #define KNET_LOG_DEBUG 3 #define KNET_LOG_TRACE 4 /* * Convert between log level values and names */ /** * knet_log_get_loglevel_name * * @brief Get a logging level name from its numeric ID * * @return * returns internal name of the log level or "ERROR" for unknown values */ const char *knet_log_get_loglevel_name(uint8_t level); /** * knet_log_get_loglevel_id * * @brief Get a logging level ID from its name * * @return * returns internal log level ID or KNET_LOG_ERR for invalid names */ uint8_t knet_log_get_loglevel_id(const char *name); /* * every log message is composed by a text message * and message level/subsystem IDs. * In order to make debugging easier it is possible to send those packets * straight to stdout/stderr (see knet_bench.c stdout option). */ #define KNET_MAX_LOG_MSG_SIZE 254 #if KNET_MAX_LOG_MSG_SIZE > PIPE_BUF #error KNET_MAX_LOG_MSG_SIZE cannot be bigger than PIPE_BUF for guaranteed system atomic writes #endif /** * Structure of a log message sent to the logging fd */ struct knet_log_msg { /** Text of the log message */ char msg[KNET_MAX_LOG_MSG_SIZE]; /** Subsystem that sent this message. KNET_SUB_* */ uint8_t subsystem; /** Logging level of this message. KNET_LOG_* */ uint8_t msglevel; }; /** * knet_log_set_loglevel * * @brief Set the logging level for a subsystem * * knet_h - same as above * * subsystem - same as above * * level - same as above * * knet_log_set_loglevel allows fine control of log levels by subsystem. * See also knet_handle_new for defaults. * * @return * knet_log_set_loglevel returns * 0 on success * -1 on error and errno is set. */ int knet_log_set_loglevel(knet_handle_t knet_h, uint8_t subsystem, uint8_t level); /** * knet_log_get_loglevel * * @brief Get the logging level for a subsystem * * knet_h - same as above * * subsystem - same as above * * level - same as above * * @return * knet_log_get_loglevel returns * 0 on success * -1 on error and errno is set. */ int knet_log_get_loglevel(knet_handle_t knet_h, uint8_t subsystem, uint8_t *level); #endif diff --git a/libknet/tests/api-check.mk b/libknet/tests/api-check.mk index 96f7f2a7..bd26ad0c 100644 --- a/libknet/tests/api-check.mk +++ b/libknet/tests/api-check.mk @@ -1,290 +1,294 @@ # # Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. # # Authors: Fabio M. Di Nitto # # This software licensed under GPL-2.0+ # api_checks = \ api_knet_handle_new_test \ api_knet_handle_free_test \ api_knet_handle_compress_test \ api_knet_handle_crypto_test \ api_knet_handle_setfwd_test \ api_knet_handle_enable_access_lists_test \ api_knet_handle_enable_filter_test \ api_knet_handle_enable_sock_notify_test \ api_knet_handle_add_datafd_test \ api_knet_handle_remove_datafd_test \ api_knet_handle_get_channel_test \ api_knet_handle_get_datafd_test \ api_knet_handle_get_stats_test \ api_knet_get_crypto_list_test \ api_knet_get_compress_list_test \ api_knet_handle_clear_stats_test \ api_knet_get_transport_list_test \ api_knet_get_transport_name_by_id_test \ api_knet_get_transport_id_by_name_test \ api_knet_handle_set_transport_reconnect_interval_test \ api_knet_handle_get_transport_reconnect_interval_test \ api_knet_recv_test \ api_knet_send_test \ api_knet_send_crypto_test \ api_knet_send_compress_test \ api_knet_send_sync_test \ api_knet_send_loopback_test \ api_knet_handle_pmtud_setfreq_test \ api_knet_handle_pmtud_getfreq_test \ api_knet_handle_enable_pmtud_notify_test \ api_knet_handle_pmtud_get_test \ api_knet_handle_pmtud_set_test \ api_knet_host_add_test \ api_knet_host_remove_test \ api_knet_host_set_name_test \ api_knet_host_get_name_by_host_id_test \ api_knet_host_get_id_by_host_name_test \ api_knet_host_get_host_list_test \ api_knet_host_set_policy_test \ api_knet_host_get_policy_test \ api_knet_host_get_status_test \ api_knet_host_enable_status_change_notify_test \ api_knet_log_get_subsystem_name_test \ api_knet_log_get_subsystem_id_test \ api_knet_log_get_loglevel_name_test \ api_knet_log_get_loglevel_id_test \ api_knet_log_set_loglevel_test \ api_knet_log_get_loglevel_test \ api_knet_strtoaddr_test \ api_knet_addrtostr_test \ api_knet_link_set_config_test \ api_knet_link_clear_config_test \ api_knet_link_get_config_test \ api_knet_link_set_ping_timers_test \ api_knet_link_get_ping_timers_test \ api_knet_link_set_pong_count_test \ api_knet_link_get_pong_count_test \ api_knet_link_set_priority_test \ api_knet_link_get_priority_test \ api_knet_link_set_enable_test \ api_knet_link_get_enable_test \ api_knet_link_get_link_list_test \ api_knet_link_get_status_test \ api_knet_link_add_acl_test \ api_knet_link_insert_acl_test \ api_knet_link_rm_acl_test \ api_knet_link_clear_acl_test \ api_knet_handle_crypto_set_config_test \ api_knet_handle_crypto_use_config_test \ - api_knet_handle_crypto_rx_clear_traffic_test + api_knet_handle_crypto_rx_clear_traffic_test \ + api_knet_handle_setprio_dscp_test api_knet_handle_new_test_SOURCES = api_knet_handle_new.c \ test-common.c api_knet_handle_free_test_SOURCES = api_knet_handle_free.c \ test-common.c api_knet_handle_new_limit_test_SOURCES = api_knet_handle_new_limit.c \ test-common.c api_knet_handle_compress_test_SOURCES = api_knet_handle_compress.c \ test-common.c api_knet_handle_crypto_test_SOURCES = api_knet_handle_crypto.c \ test-common.c api_knet_handle_setfwd_test_SOURCES = api_knet_handle_setfwd.c \ test-common.c api_knet_handle_enable_access_lists_test_SOURCES = api_knet_handle_enable_access_lists.c \ test-common.c api_knet_handle_enable_filter_test_SOURCES = api_knet_handle_enable_filter.c \ test-common.c api_knet_handle_enable_sock_notify_test_SOURCES = api_knet_handle_enable_sock_notify.c \ test-common.c api_knet_handle_add_datafd_test_SOURCES = api_knet_handle_add_datafd.c \ test-common.c api_knet_handle_remove_datafd_test_SOURCES = api_knet_handle_remove_datafd.c \ test-common.c api_knet_handle_get_channel_test_SOURCES = api_knet_handle_get_channel.c \ test-common.c api_knet_handle_get_datafd_test_SOURCES = api_knet_handle_get_datafd.c \ test-common.c api_knet_handle_get_stats_test_SOURCES = api_knet_handle_get_stats.c \ test-common.c api_knet_get_crypto_list_test_SOURCES = api_knet_get_crypto_list.c \ test-common.c api_knet_get_compress_list_test_SOURCES = api_knet_get_compress_list.c \ test-common.c api_knet_handle_clear_stats_test_SOURCES = api_knet_handle_clear_stats.c \ test-common.c api_knet_get_transport_list_test_SOURCES = api_knet_get_transport_list.c \ test-common.c api_knet_get_transport_name_by_id_test_SOURCES = api_knet_get_transport_name_by_id.c \ test-common.c api_knet_get_transport_id_by_name_test_SOURCES = api_knet_get_transport_id_by_name.c \ test-common.c api_knet_handle_set_transport_reconnect_interval_test_SOURCES = api_knet_handle_set_transport_reconnect_interval.c \ test-common.c api_knet_handle_get_transport_reconnect_interval_test_SOURCES = api_knet_handle_get_transport_reconnect_interval.c \ test-common.c api_knet_recv_test_SOURCES = api_knet_recv.c \ test-common.c api_knet_send_test_SOURCES = api_knet_send.c \ test-common.c api_knet_send_compress_test_SOURCES = api_knet_send_compress.c \ test-common.c api_knet_send_crypto_test_SOURCES = api_knet_send_crypto.c \ test-common.c api_knet_send_loopback_test_SOURCES = api_knet_send_loopback.c \ test-common.c api_knet_send_sync_test_SOURCES = api_knet_send_sync.c \ test-common.c api_knet_handle_pmtud_setfreq_test_SOURCES = api_knet_handle_pmtud_setfreq.c \ test-common.c api_knet_handle_pmtud_getfreq_test_SOURCES = api_knet_handle_pmtud_getfreq.c \ test-common.c api_knet_handle_enable_pmtud_notify_test_SOURCES = api_knet_handle_enable_pmtud_notify.c \ test-common.c api_knet_handle_pmtud_get_test_SOURCES = api_knet_handle_pmtud_get.c \ test-common.c api_knet_handle_pmtud_set_test_SOURCES = api_knet_handle_pmtud_set.c \ test-common.c api_knet_host_add_test_SOURCES = api_knet_host_add.c \ test-common.c api_knet_host_remove_test_SOURCES = api_knet_host_remove.c \ test-common.c api_knet_host_set_name_test_SOURCES = api_knet_host_set_name.c \ test-common.c api_knet_host_get_name_by_host_id_test_SOURCES = api_knet_host_get_name_by_host_id.c \ test-common.c api_knet_host_get_id_by_host_name_test_SOURCES = api_knet_host_get_id_by_host_name.c \ test-common.c api_knet_host_get_host_list_test_SOURCES = api_knet_host_get_host_list.c \ test-common.c api_knet_host_set_policy_test_SOURCES = api_knet_host_set_policy.c \ test-common.c api_knet_host_get_policy_test_SOURCES = api_knet_host_get_policy.c \ test-common.c api_knet_host_get_status_test_SOURCES = api_knet_host_get_status.c \ test-common.c api_knet_host_enable_status_change_notify_test_SOURCES = api_knet_host_enable_status_change_notify.c \ test-common.c api_knet_log_get_subsystem_name_test_SOURCES = api_knet_log_get_subsystem_name.c \ test-common.c api_knet_log_get_subsystem_id_test_SOURCES = api_knet_log_get_subsystem_id.c \ test-common.c api_knet_log_get_loglevel_name_test_SOURCES = api_knet_log_get_loglevel_name.c \ test-common.c api_knet_log_get_loglevel_id_test_SOURCES = api_knet_log_get_loglevel_id.c \ test-common.c api_knet_log_set_loglevel_test_SOURCES = api_knet_log_set_loglevel.c \ test-common.c api_knet_log_get_loglevel_test_SOURCES = api_knet_log_get_loglevel.c \ test-common.c api_knet_strtoaddr_test_SOURCES = api_knet_strtoaddr.c api_knet_addrtostr_test_SOURCES = api_knet_addrtostr.c api_knet_link_set_config_test_SOURCES = api_knet_link_set_config.c \ test-common.c api_knet_link_clear_config_test_SOURCES = api_knet_link_clear_config.c \ test-common.c api_knet_link_get_config_test_SOURCES = api_knet_link_get_config.c \ test-common.c api_knet_link_set_ping_timers_test_SOURCES = api_knet_link_set_ping_timers.c \ test-common.c api_knet_link_get_ping_timers_test_SOURCES = api_knet_link_get_ping_timers.c \ test-common.c api_knet_link_set_pong_count_test_SOURCES = api_knet_link_set_pong_count.c \ test-common.c api_knet_link_get_pong_count_test_SOURCES = api_knet_link_get_pong_count.c \ test-common.c api_knet_link_set_priority_test_SOURCES = api_knet_link_set_priority.c \ test-common.c api_knet_link_get_priority_test_SOURCES = api_knet_link_get_priority.c \ test-common.c api_knet_link_set_enable_test_SOURCES = api_knet_link_set_enable.c \ test-common.c api_knet_link_get_enable_test_SOURCES = api_knet_link_get_enable.c \ test-common.c api_knet_link_get_link_list_test_SOURCES = api_knet_link_get_link_list.c \ test-common.c api_knet_link_get_status_test_SOURCES = api_knet_link_get_status.c \ test-common.c api_knet_link_add_acl_test_SOURCES = api_knet_link_add_acl.c \ test-common.c api_knet_link_insert_acl_test_SOURCES = api_knet_link_insert_acl.c \ test-common.c api_knet_link_rm_acl_test_SOURCES = api_knet_link_rm_acl.c \ test-common.c api_knet_link_clear_acl_test_SOURCES = api_knet_link_clear_acl.c \ test-common.c api_knet_handle_crypto_set_config_test_SOURCES = api_knet_handle_crypto_set_config.c \ test-common.c api_knet_handle_crypto_use_config_test_SOURCES = api_knet_handle_crypto_use_config.c \ test-common.c api_knet_handle_crypto_rx_clear_traffic_test_SOURCES = api_knet_handle_crypto_rx_clear_traffic.c \ test-common.c + +api_knet_handle_setprio_dscp_test_SOURCES = api_knet_handle_setprio_dscp.c \ + test-common.c diff --git a/libknet/tests/api_knet_handle_free.c b/libknet/tests/api_knet_handle_free.c index 1ceba443..83b46aa6 100644 --- a/libknet/tests/api_knet_handle_free.c +++ b/libknet/tests/api_knet_handle_free.c @@ -1,61 +1,62 @@ /* * Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * * This software licensed under GPL-2.0+ */ #include "config.h" #include #include #include #include #include #include "libknet.h" #include "internals.h" #include "test-common.h" #define TESTNODES 1 static void test(void) { knet_handle_t knet_h1, knet_h[2]; int logfds[2]; int res; setup_logpipes(logfds); printf("Test knet_handle_free with invalid knet_h (part 1)\n"); if ((!knet_handle_free(NULL)) || (errno != EINVAL)) { printf("knet_handle_free failed to detect invalid parameter\n"); exit(FAIL); } knet_h1 = knet_handle_start(logfds, KNET_LOG_DEBUG, knet_h); printf("Test knet_handle_free with one host configured\n"); FAIL_ON_ERR(knet_host_add(knet_h1, 1)); if ((!knet_handle_free(knet_h1)) || (errno != EBUSY)) { CLEAN_EXIT(FAIL); } FAIL_ON_ERR(knet_host_remove(knet_h1, 1)); printf("Test knet_handle_free with invalid knet_h (part 2)\n"); + // coverity[BAD_FREE:SUPPRESS] - deliberate bad handle FAIL_ON_SUCCESS(knet_handle_free(knet_h1 + 1), EINVAL); FAIL_ON_ERR(knet_handle_free(knet_h1)); CLEAN_EXIT(CONTINUE); } int main(int argc, char *argv[]) { test(); return PASS; } diff --git a/libknet/tests/api_knet_handle_pmtud_setfreq.c b/libknet/tests/api_knet_handle_pmtud_setfreq.c index 5645eb9f..eadba5e4 100644 --- a/libknet/tests/api_knet_handle_pmtud_setfreq.c +++ b/libknet/tests/api_knet_handle_pmtud_setfreq.c @@ -1,62 +1,62 @@ /* * Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * * This software licensed under GPL-2.0+ */ #include "config.h" #include #include #include #include #include #include "libknet.h" #include "internals.h" #include "test-common.h" static void test(void) { knet_handle_t knet_h1, knet_h[2]; int res; int logfds[2]; printf("Test knet_handle_pmtud_setfreq incorrect knet_h\n"); if ((!knet_handle_pmtud_setfreq(NULL, 1)) || (errno != EINVAL)) { printf("knet_handle_pmtud_setfreq accepted invalid knet_h or returned incorrect error: %s\n", strerror(errno)); exit(FAIL); } setup_logpipes(logfds); knet_h1 = knet_handle_start(logfds, KNET_LOG_DEBUG, knet_h); flush_logs(logfds[0], stdout); printf("Test knet_handle_pmtud_setfreq with 0 (incorrect)\n"); - FAIL_ON_SUCCESS(knet_handle_pmtud_setfreq(NULL, 0), EINVAL); + FAIL_ON_SUCCESS(knet_handle_pmtud_setfreq(knet_h1, 0), EINVAL); printf("Test knet_handle_pmtud_setfreq with 86401 (incorrect)\n"); - FAIL_ON_SUCCESS(knet_handle_pmtud_setfreq(NULL, 86401), EINVAL); + FAIL_ON_SUCCESS(knet_handle_pmtud_setfreq(knet_h1, 86401), EINVAL); printf("Test knet_handle_pmtud_setfreq with 1 (correct)\n"); FAIL_ON_ERR(knet_handle_pmtud_setfreq(knet_h1, 1)); if (knet_h1->pmtud_interval != 1) { printf("knet_handle_pmtud_setfreq failed to set the value\n"); CLEAN_EXIT(FAIL); } CLEAN_EXIT(CONTINUE); } int main(int argc, char *argv[]) { test(); return PASS; } diff --git a/libknet/tests/api_knet_handle_set_transport_reconnect_interval.c b/libknet/tests/api_knet_handle_set_transport_reconnect_interval.c index f0c6b2f3..1d2cdc96 100644 --- a/libknet/tests/api_knet_handle_set_transport_reconnect_interval.c +++ b/libknet/tests/api_knet_handle_set_transport_reconnect_interval.c @@ -1,58 +1,59 @@ /* * Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * * This software licensed under GPL-2.0+ */ #include "config.h" #include #include #include #include #include #include "libknet.h" #include "internals.h" #include "test-common.h" static void test(void) { knet_handle_t knet_h1, knet_h[2]; int res; int logfds[2]; printf("Test knet_handle_set_transport_reconnect_interval with incorrect knet_h\n"); if ((!knet_handle_set_transport_reconnect_interval(NULL, 1000)) || (errno != EINVAL)) { printf("knet_handle_set_transport_reconnect_interval accepted invalid knet_h or returned incorrect error: %s\n", strerror(errno)); exit(FAIL); } setup_logpipes(logfds); knet_h1 = knet_handle_start(logfds, KNET_LOG_DEBUG, knet_h); printf("Test knet_handle_set_transport_reconnect_interval with incorrect msecs\n"); FAIL_ON_SUCCESS(knet_handle_set_transport_reconnect_interval(knet_h1, 0), EINVAL); printf("Test knet_handle_set_transport_reconnect_interval with correct values\n"); FAIL_ON_ERR(knet_handle_set_transport_reconnect_interval(knet_h1, 2000)); + // coverity[MISSING_LOCK:SUPPRESS] use out of the main library is 'OK' here. ish if (knet_h1->reconnect_int != 2000) { printf("knet_handle_set_transport_reconnect_interval failed to set correct value\n"); CLEAN_EXIT(FAIL); } CLEAN_EXIT(CONTINUE); } int main(int argc, char *argv[]) { test(); return PASS; } diff --git a/libknet/tests/api_knet_handle_setprio_dscp.c b/libknet/tests/api_knet_handle_setprio_dscp.c new file mode 100644 index 00000000..8c809070 --- /dev/null +++ b/libknet/tests/api_knet_handle_setprio_dscp.c @@ -0,0 +1,59 @@ +/* + * Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. + * + * Authors: David Hanisch + * + * This software licensed under GPL-2.0+ + */ + +#include "config.h" + +#include +#include +#include +#include +#include + +#include "libknet.h" + +#include "internals.h" +#include "test-common.h" + +static void test(void) +{ + knet_handle_t knet_h1, knet_h[2]; + int res; + int logfds[2]; + + printf("Test knet_handle_setprio_dscp incorrect knet_h\n"); + + if ((!knet_handle_setprio_dscp(NULL, 1)) || (errno != EINVAL)) { + printf("knet_handle_setprio_dscp accepted invalid knet_h or returned incorrect error: %s\n", strerror(errno)); + exit(FAIL); + } + + setup_logpipes(logfds); + + knet_h1 = knet_handle_start(logfds, KNET_LOG_DEBUG, knet_h); + + flush_logs(logfds[0], stdout); + + printf("Test knet_handle_setprio_dscp with 100 (incorrect)\n"); + FAIL_ON_SUCCESS(knet_handle_setprio_dscp(knet_h1, 100), EINVAL); + + printf("Test knet_handle_setprio_dscp with 40 (correct)\n"); + FAIL_ON_ERR(knet_handle_setprio_dscp(knet_h1, 40)); + + if (knet_h1->prio_dscp != 40) { + printf("knet_handle_setprio_dscp failed to set the value\n"); + CLEAN_EXIT(FAIL); + } + CLEAN_EXIT(CONTINUE); +} + +int main(int argc, char *argv[]) +{ + test(); + + return PASS; +} diff --git a/libknet/threads_rx.c b/libknet/threads_rx.c index cc79374f..e2a3aa9c 100644 --- a/libknet/threads_rx.c +++ b/libknet/threads_rx.c @@ -1,1090 +1,1090 @@ /* * Copyright (C) 2012-2025 Red Hat, Inc. All rights reserved. * * Authors: Fabio M. Di Nitto * Federico Simoncelli * * This software licensed under LGPL-2.0+ */ #include "config.h" #include #include #include #include #include #include "compat.h" #include "compress.h" #include "crypto.h" #include "host.h" #include "links.h" #include "links_acl.h" #include "logging.h" #include "transports.h" #include "transport_common.h" #include "threads_common.h" #include "threads_heartbeat.h" #include "threads_rx.h" #include "netutils.h" /* * RECV */ /* * return 1 if a > b * return -1 if b > a * return 0 if they are equal */ static inline int timecmp(struct timespec a, struct timespec b) { if (a.tv_sec != b.tv_sec) { if (a.tv_sec > b.tv_sec) { return 1; } else { return -1; } } else { if (a.tv_nsec > b.tv_nsec) { return 1; } else if (a.tv_nsec < b.tv_nsec) { return -1; } else { return 0; } } } /* * this functions needs to return an index (0 to 7) * to a knet_host_defrag_buf. (-1 on errors) */ static int find_pckt_defrag_buf(knet_handle_t knet_h, struct knet_header *inbuf) { struct knet_host *src_host = knet_h->host_index[inbuf->kh_node]; int i, oldest; /* * check if there is a buffer already in use handling the same seq_num */ for (i = 0; i < KNET_DEFRAG_BUFFERS; i++) { if (src_host->defrag_buf[i].in_use) { if (src_host->defrag_buf[i].pckt_seq == inbuf->khp_data_seq_num) { return i; } } } /* * If there is no buffer that's handling the current seq_num * either it's new or it's been reclaimed already. * check if it's been reclaimed/seen before using the defrag circular * buffer. If the pckt has been seen before, the buffer expired (ETIME) * and there is no point to try to defrag it again. */ if (!_seq_num_lookup(src_host, inbuf->khp_data_seq_num, 1, 0)) { errno = ETIME; return -1; } /* * register the pckt as seen */ _seq_num_set(src_host, inbuf->khp_data_seq_num, 1); /* * see if there is a free buffer */ for (i = 0; i < KNET_DEFRAG_BUFFERS; i++) { if (!src_host->defrag_buf[i].in_use) { return i; } } /* * at this point, there are no free buffers, the pckt is new * and we need to reclaim a buffer, and we will take the one * with the oldest timestamp. It's as good as any. */ oldest = 0; for (i = 0; i < KNET_DEFRAG_BUFFERS; i++) { if (timecmp(src_host->defrag_buf[i].last_update, src_host->defrag_buf[oldest].last_update) < 0) { oldest = i; } } src_host->defrag_buf[oldest].in_use = 0; return oldest; } static int pckt_defrag(knet_handle_t knet_h, struct knet_header *inbuf, ssize_t *len) { struct knet_host_defrag_buf *defrag_buf; int defrag_buf_idx; defrag_buf_idx = find_pckt_defrag_buf(knet_h, inbuf); if (defrag_buf_idx < 0) { return 1; } defrag_buf = &knet_h->host_index[inbuf->kh_node]->defrag_buf[defrag_buf_idx]; /* * if the buf is not is use, then make sure it's clean */ if (!defrag_buf->in_use) { memset(defrag_buf, 0, sizeof(struct knet_host_defrag_buf)); defrag_buf->in_use = 1; defrag_buf->pckt_seq = inbuf->khp_data_seq_num; } /* * update timestamp on the buffer */ clock_gettime(CLOCK_MONOTONIC, &defrag_buf->last_update); /* * check if we already received this fragment */ if (defrag_buf->frag_map[inbuf->khp_data_frag_seq]) { /* * if we have received this fragment and we didn't clear the buffer * it means that we don't have all fragments yet */ return 1; } /* * we need to handle the last packet with gloves due to its different size */ if (inbuf->khp_data_frag_seq == inbuf->khp_data_frag_num) { defrag_buf->last_frag_size = *len; /* * in the event when the last packet arrives first, * we still don't know the offset vs the other fragments (based on MTU), * so we store the fragment at the end of the buffer where it's safe * and take a copy of the len so that we can restore its offset later. * remember we can't use the local MTU for this calculation because pMTU * can be asymettric between the same hosts. */ if (!defrag_buf->frag_size) { defrag_buf->last_first = 1; memmove(defrag_buf->buf + (KNET_MAX_PACKET_SIZE - *len), inbuf->khp_data_userdata, *len); } } else { defrag_buf->frag_size = *len; } if (defrag_buf->frag_size) { memmove(defrag_buf->buf + ((inbuf->khp_data_frag_seq - 1) * defrag_buf->frag_size), inbuf->khp_data_userdata, *len); } defrag_buf->frag_recv++; defrag_buf->frag_map[inbuf->khp_data_frag_seq] = 1; /* * check if we received all the fragments */ if (defrag_buf->frag_recv == inbuf->khp_data_frag_num) { /* * special case the last pckt */ if (defrag_buf->last_first) { memmove(defrag_buf->buf + ((inbuf->khp_data_frag_num - 1) * defrag_buf->frag_size), defrag_buf->buf + (KNET_MAX_PACKET_SIZE - defrag_buf->last_frag_size), defrag_buf->last_frag_size); } /* * recalculate packet lenght */ *len = ((inbuf->khp_data_frag_num - 1) * defrag_buf->frag_size) + defrag_buf->last_frag_size; /* * copy the pckt back in the user data */ memmove(inbuf->khp_data_userdata, defrag_buf->buf, *len); /* * free this buffer */ defrag_buf->in_use = 0; return 0; } return 1; } /* * processing incoming packets vs access lists */ static int _check_rx_acl(knet_handle_t knet_h, struct knet_link *src_link, const struct knet_mmsghdr *msg) { if (knet_h->use_access_lists) { if (!check_validate(knet_h, src_link, msg->msg_hdr.msg_name)) { char src_ipaddr[KNET_MAX_HOST_LEN]; char src_port[KNET_MAX_PORT_LEN]; memset(src_ipaddr, 0, KNET_MAX_HOST_LEN); memset(src_port, 0, KNET_MAX_PORT_LEN); if (knet_addrtostr(msg->msg_hdr.msg_name, sockaddr_len(msg->msg_hdr.msg_name), src_ipaddr, KNET_MAX_HOST_LEN, src_port, KNET_MAX_PORT_LEN) < 0) { log_warn(knet_h, KNET_SUB_RX, "Packet rejected: unable to resolve host/port"); } else { log_warn(knet_h, KNET_SUB_RX, "Packet rejected from %s:%s", src_ipaddr, src_port); } return 0; } } return 1; } static int _fast_data_up(knet_handle_t knet_h, struct knet_host *src_host, struct knet_link *src_link) { if (src_link->received_pong) { log_debug(knet_h, KNET_SUB_RX, "host: %u link: %u received data during valid ping/pong activity. Force link up.", src_host->host_id, src_link->link_id); _link_updown(knet_h, src_host->host_id, src_link->link_id, src_link->status.enabled, 1, 0); return 1; } // host is not eligible for fast data up return 0; } static void _parse_recv_from_links(knet_handle_t knet_h, int sockfd, const struct knet_mmsghdr *msg) { int err = 0, savederrno = 0, stats_err = 0; ssize_t outlen; struct knet_host *src_host; struct knet_link *src_link; unsigned long long latency_last; knet_node_id_t dst_host_ids[KNET_MAX_HOST]; size_t dst_host_ids_entries = 0; int bcast = 1; uint64_t decrypt_time = 0; struct timespec recvtime; struct knet_header *inbuf = msg->msg_hdr.msg_iov->iov_base; unsigned char *outbuf = (unsigned char *)msg->msg_hdr.msg_iov->iov_base; ssize_t len = msg->msg_len; struct iovec iov_out[1]; int8_t channel; seq_num_t recv_seq_num; int wipe_bufs = 0; int try_decrypt = 0, decrypted = 0, i, found_link = 0; for (i = 1; i <= KNET_MAX_CRYPTO_INSTANCES; i++) { if (knet_h->crypto_instance[i]) { try_decrypt = 1; break; } } if ((!try_decrypt) && (knet_h->crypto_only == KNET_CRYPTO_RX_DISALLOW_CLEAR_TRAFFIC)) { log_debug(knet_h, KNET_SUB_RX, "RX thread configured to accept only crypto packets, but no crypto configs are configured!"); return; } if (try_decrypt) { struct timespec start_time; struct timespec end_time; clock_gettime(CLOCK_MONOTONIC, &start_time); if (crypto_authenticate_and_decrypt(knet_h, (unsigned char *)inbuf, len, knet_h->recv_from_links_buf_decrypt, &outlen) < 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to decrypt/auth packet"); if (knet_h->crypto_only == KNET_CRYPTO_RX_DISALLOW_CLEAR_TRAFFIC) { char src_ipaddr[KNET_MAX_HOST_LEN]; char src_port[KNET_MAX_PORT_LEN]; memset(src_ipaddr, 0, KNET_MAX_HOST_LEN); memset(src_port, 0, KNET_MAX_PORT_LEN); if (knet_addrtostr(msg->msg_hdr.msg_name, sockaddr_len(msg->msg_hdr.msg_name), src_ipaddr, KNET_MAX_HOST_LEN, src_port, KNET_MAX_PORT_LEN) < 0) { log_err(knet_h, KNET_SUB_RX, "Unable to decrypt packet from unknown host/port (size %zu)!", len); } else { log_err(knet_h, KNET_SUB_RX, "Unable to decrypt packet from %s:%s (size %zu)!", src_ipaddr, src_port, len); } return; } log_debug(knet_h, KNET_SUB_RX, "Attempting to process packet as clear data"); } else { clock_gettime(CLOCK_MONOTONIC, &end_time); timespec_diff(start_time, end_time, &decrypt_time); len = outlen; inbuf = (struct knet_header *)knet_h->recv_from_links_buf_decrypt; decrypted = 1; } } if (len < (ssize_t)(KNET_HEADER_SIZE + 1)) { log_debug(knet_h, KNET_SUB_RX, "Packet is too short: %ld", (long)len); return; } if (inbuf->kh_version != KNET_HEADER_VERSION) { log_debug(knet_h, KNET_SUB_RX, "Packet version does not match"); return; } inbuf->kh_node = ntohs(inbuf->kh_node); src_host = knet_h->host_index[inbuf->kh_node]; if (src_host == NULL) { /* host not found */ log_debug(knet_h, KNET_SUB_RX, "Unable to find source host for this packet"); return; } if ((inbuf->kh_type & KNET_HEADER_TYPE_PMSK) != 0) { /* be aware this works only for PING / PONG and PMTUd packets! */ src_link = src_host->link + (inbuf->khp_ping_link % KNET_MAX_LINK); if (!_check_rx_acl(knet_h, src_link, msg)) { return; } if (src_link->dynamic == KNET_LINK_DYNIP) { if (cmpaddr(&src_link->dst_addr, msg->msg_hdr.msg_name) != 0) { log_debug(knet_h, KNET_SUB_RX, "host: %u link: %u appears to have changed ip address", src_host->host_id, src_link->link_id); memmove(&src_link->dst_addr, msg->msg_hdr.msg_name, sizeof(struct sockaddr_storage)); if (knet_addrtostr(&src_link->dst_addr, sockaddr_len(&src_link->dst_addr), src_link->status.dst_ipaddr, KNET_MAX_HOST_LEN, src_link->status.dst_port, KNET_MAX_PORT_LEN) != 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to resolve ???"); snprintf(src_link->status.dst_ipaddr, KNET_MAX_HOST_LEN - 1, "Unknown!!!"); snprintf(src_link->status.dst_port, KNET_MAX_PORT_LEN - 1, "??"); } else { log_info(knet_h, KNET_SUB_RX, "host: %u link: %u new connection established from: %s:%s", src_host->host_id, src_link->link_id, src_link->status.dst_ipaddr, src_link->status.dst_port); } } /* * transport has already accepted the connection here * otherwise we would not be receiving packets */ transport_link_dyn_connect(knet_h, sockfd, src_link); } } else { /* data packet */ for (i = 0; i < KNET_MAX_LINK; i++) { src_link = &src_host->link[i]; if (cmpaddr(&src_link->dst_addr, msg->msg_hdr.msg_name) == 0) { found_link = 1; break; } } if (found_link) { /* * this check is currently redundant.. Keep it here for now */ if (!_check_rx_acl(knet_h, src_link, msg)) { return; } } else { log_debug(knet_h, KNET_SUB_RX, "Unable to determine source link for data packet. Discarding packet."); return; } } stats_err = pthread_mutex_lock(&src_link->link_stats_mutex); if (stats_err) { log_err(knet_h, KNET_SUB_RX, "Unable to get stats mutex lock for host %u link %u: %s", src_host->host_id, src_link->link_id, strerror(savederrno)); return; } switch (inbuf->kh_type) { case KNET_HEADER_TYPE_DATA: /* data stats at the top for consistency with TX */ src_link->status.stats.rx_data_packets++; src_link->status.stats.rx_data_bytes += len; if (decrypted) { stats_err = pthread_mutex_lock(&knet_h->handle_stats_mutex); if (stats_err < 0) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_err(knet_h, KNET_SUB_RX, "Unable to get mutex lock: %s", strerror(stats_err)); return; } /* Only update the crypto overhead for data packets. Mainly to be consistent with TX */ if (decrypt_time < knet_h->stats.rx_crypt_time_min) { knet_h->stats.rx_crypt_time_min = decrypt_time; } if (decrypt_time > knet_h->stats.rx_crypt_time_max) { knet_h->stats.rx_crypt_time_max = decrypt_time; } knet_h->stats.rx_crypt_time_ave = (knet_h->stats.rx_crypt_time_ave * knet_h->stats.rx_crypt_packets + decrypt_time) / (knet_h->stats.rx_crypt_packets+1); knet_h->stats.rx_crypt_packets++; pthread_mutex_unlock(&knet_h->handle_stats_mutex); } if (!src_host->status.reachable) { if (!_fast_data_up(knet_h, src_host, src_link)) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "Source host %u not reachable yet. Discarding packet.", src_host->host_id); return; } } inbuf->khp_data_seq_num = ntohs(inbuf->khp_data_seq_num); channel = inbuf->khp_data_channel; src_host->got_data = 1; if (!_seq_num_lookup(src_host, inbuf->khp_data_seq_num, 0, 0)) { pthread_mutex_unlock(&src_link->link_stats_mutex); if (src_host->link_handler_policy != KNET_LINK_POLICY_ACTIVE) { log_debug(knet_h, KNET_SUB_RX, "Packet has already been delivered"); } return; } if (inbuf->khp_data_frag_num > 1) { /* * len as received from the socket also includes extra stuff * that the defrag code doesn't care about. So strip it * here and readd only for repadding once we are done * defragging */ len = len - KNET_HEADER_DATA_SIZE; if (pckt_defrag(knet_h, inbuf, &len)) { pthread_mutex_unlock(&src_link->link_stats_mutex); return; } len = len + KNET_HEADER_DATA_SIZE; } if (inbuf->khp_data_compress) { ssize_t decmp_outlen = KNET_DATABUFSIZE_COMPRESS; struct timespec start_time; struct timespec end_time; uint64_t compress_time; clock_gettime(CLOCK_MONOTONIC, &start_time); err = decompress(knet_h, inbuf->khp_data_compress, (const unsigned char *)inbuf->khp_data_userdata, len - KNET_HEADER_DATA_SIZE, knet_h->recv_from_links_buf_decompress, &decmp_outlen); stats_err = pthread_mutex_lock(&knet_h->handle_stats_mutex); if (stats_err < 0) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_err(knet_h, KNET_SUB_RX, "Unable to get mutex lock: %s", strerror(stats_err)); return; } clock_gettime(CLOCK_MONOTONIC, &end_time); timespec_diff(start_time, end_time, &compress_time); if (!err) { /* Collect stats */ if (compress_time < knet_h->stats.rx_compress_time_min) { knet_h->stats.rx_compress_time_min = compress_time; } if (compress_time > knet_h->stats.rx_compress_time_max) { knet_h->stats.rx_compress_time_max = compress_time; } knet_h->stats.rx_compress_time_ave = (knet_h->stats.rx_compress_time_ave * knet_h->stats.rx_compressed_packets + compress_time) / (knet_h->stats.rx_compressed_packets+1); knet_h->stats.rx_compressed_packets++; knet_h->stats.rx_compressed_original_bytes += decmp_outlen; knet_h->stats.rx_compressed_size_bytes += len - KNET_HEADER_SIZE; memmove(inbuf->khp_data_userdata, knet_h->recv_from_links_buf_decompress, decmp_outlen); len = decmp_outlen + KNET_HEADER_DATA_SIZE; } else { pthread_mutex_unlock(&knet_h->handle_stats_mutex); pthread_mutex_unlock(&src_link->link_stats_mutex); log_err(knet_h, KNET_SUB_COMPRESS, "Unable to decompress packet (%d): %s", err, strerror(errno)); return; } pthread_mutex_unlock(&knet_h->handle_stats_mutex); } if (knet_h->enabled != 1) /* data forward is disabled */ break; if (knet_h->dst_host_filter_fn) { size_t host_idx; int found = 0; bcast = knet_h->dst_host_filter_fn( knet_h->dst_host_filter_fn_private_data, (const unsigned char *)inbuf->khp_data_userdata, len - KNET_HEADER_DATA_SIZE, KNET_NOTIFY_RX, knet_h->host_id, inbuf->kh_node, &channel, dst_host_ids, &dst_host_ids_entries); if (bcast < 0) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "Error from dst_host_filter_fn: %d", bcast); return; } if ((!bcast) && (!dst_host_ids_entries)) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "Message is unicast but no dst_host_ids_entries"); return; } /* check if we are dst for this packet */ if (!bcast) { if (dst_host_ids_entries > KNET_MAX_HOST) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "dst_host_filter_fn returned too many destinations"); return; } for (host_idx = 0; host_idx < dst_host_ids_entries; host_idx++) { if (dst_host_ids[host_idx] == knet_h->host_id) { found = 1; break; } } if (!found) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "Packet is not for us"); return; } } } if (!knet_h->sockfd[channel].in_use) { pthread_mutex_unlock(&src_link->link_stats_mutex); log_debug(knet_h, KNET_SUB_RX, "received packet for channel %d but there is no local sock connected", channel); return; } outlen = 0; memset(iov_out, 0, sizeof(iov_out)); retry: iov_out[0].iov_base = (void *) inbuf->khp_data_userdata + outlen; iov_out[0].iov_len = len - (outlen + KNET_HEADER_DATA_SIZE); outlen = writev(knet_h->sockfd[channel].sockfd[knet_h->sockfd[channel].is_created], iov_out, 1); if ((outlen > 0) && (outlen < (ssize_t)iov_out[0].iov_len)) { log_debug(knet_h, KNET_SUB_RX, "Unable to send all data to the application in one go. Expected: %zu Sent: %zd\n", iov_out[0].iov_len, outlen); goto retry; } if (outlen <= 0) { knet_h->sock_notify_fn(knet_h->sock_notify_fn_private_data, knet_h->sockfd[channel].sockfd[0], channel, KNET_NOTIFY_RX, outlen, errno); pthread_mutex_unlock(&src_link->link_stats_mutex); return; } if ((size_t)outlen == iov_out[0].iov_len) { _seq_num_set(src_host, inbuf->khp_data_seq_num, 0); } break; case KNET_HEADER_TYPE_PING: outlen = KNET_HEADER_PING_SIZE; inbuf->kh_type = KNET_HEADER_TYPE_PONG; inbuf->kh_node = htons(knet_h->host_id); recv_seq_num = ntohs(inbuf->khp_ping_seq_num); src_link->status.stats.rx_ping_packets++; src_link->status.stats.rx_ping_bytes += len; wipe_bufs = 0; if (!inbuf->khp_ping_timed) { /* * we might be receiving this message from all links, but we want * to process it only the first time */ if (recv_seq_num != src_host->untimed_rx_seq_num) { /* * cache the untimed seq num */ src_host->untimed_rx_seq_num = recv_seq_num; /* * if the host has received data in between * untimed ping, then we don't need to wipe the bufs */ if (src_host->got_data) { src_host->got_data = 0; wipe_bufs = 0; } else { wipe_bufs = 1; } } _seq_num_lookup(src_host, recv_seq_num, 0, wipe_bufs); } else { /* * pings always arrives in bursts over all the link * catch the first of them to cache the seq num and * avoid duplicate processing */ if (recv_seq_num != src_host->timed_rx_seq_num) { src_host->timed_rx_seq_num = recv_seq_num; if (recv_seq_num == 0) { _seq_num_lookup(src_host, recv_seq_num, 0, 1); } } } if (knet_h->crypto_in_use_config) { if (crypto_encrypt_and_sign(knet_h, (const unsigned char *)inbuf, outlen, knet_h->recv_from_links_buf_crypt, &outlen) < 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to encrypt pong packet"); break; } outbuf = knet_h->recv_from_links_buf_crypt; stats_err = pthread_mutex_lock(&knet_h->handle_stats_mutex); if (stats_err < 0) { log_err(knet_h, KNET_SUB_RX, "Unable to get mutex lock: %s", strerror(stats_err)); break; } knet_h->stats_extra.tx_crypt_pong_packets++; pthread_mutex_unlock(&knet_h->handle_stats_mutex); } retry_pong: if (src_link->transport_connected) { if (transport_get_connection_oriented(knet_h, src_link->transport) == TRANSPORT_PROTO_NOT_CONNECTION_ORIENTED) { len = sendto(src_link->outsock, outbuf, outlen, MSG_DONTWAIT | MSG_NOSIGNAL, (struct sockaddr *) &src_link->dst_addr, knet_h->knet_transport_fd_tracker[src_link->outsock].sockaddr_len); } else { len = sendto(src_link->outsock, outbuf, outlen, MSG_DONTWAIT | MSG_NOSIGNAL, NULL, 0); } savederrno = errno; if (len != outlen) { err = transport_tx_sock_error(knet_h, src_link->transport, src_link->outsock, KNET_SUB_RX, len, savederrno); switch(err) { case -1: /* unrecoverable error */ log_debug(knet_h, KNET_SUB_RX, "Unable to send pong reply (sock: %d) packet (sendto): %d %s. recorded src ip: %s src port: %s dst ip: %s dst port: %s", src_link->outsock, errno, strerror(errno), src_link->status.src_ipaddr, src_link->status.src_port, src_link->status.dst_ipaddr, src_link->status.dst_port); src_link->status.stats.tx_pong_errors++; break; case 0: /* ignore error and continue */ break; case 1: /* retry to send those same data */ src_link->status.stats.tx_pong_retries++; goto retry_pong; break; } } src_link->status.stats.tx_pong_packets++; src_link->status.stats.tx_pong_bytes += outlen; } break; case KNET_HEADER_TYPE_PONG: src_link->status.stats.rx_pong_packets++; src_link->status.stats.rx_pong_bytes += len; clock_gettime(CLOCK_MONOTONIC, &src_link->status.pong_last); memmove(&recvtime, &inbuf->khp_ping_time[0], sizeof(struct timespec)); timespec_diff(recvtime, src_link->status.pong_last, &latency_last); if ((latency_last / 1000llu) > src_link->pong_timeout) { log_debug(knet_h, KNET_SUB_RX, "Incoming pong packet from host: %u link: %u has higher latency than pong_timeout. Discarding", src_host->host_id, src_link->link_id); } else { /* * in words : ('previous mean' * '(count -1)') + 'new value') / 'count' */ src_link->latency_cur_samples++; /* * limit to max_samples (precision) */ if (src_link->latency_cur_samples >= src_link->latency_max_samples) { src_link->latency_cur_samples = src_link->latency_max_samples; } src_link->status.latency = (((src_link->status.latency * (src_link->latency_cur_samples - 1)) + (latency_last / 1000llu)) / src_link->latency_cur_samples); if (src_link->status.latency < src_link->pong_timeout_adj) { if (!src_link->status.connected) { if (src_link->received_pong >= src_link->pong_count) { log_info(knet_h, KNET_SUB_RX, "host: %u link: %u is up", src_host->host_id, src_link->link_id); _link_updown(knet_h, src_host->host_id, src_link->link_id, src_link->status.enabled, 1, 0); } else { src_link->received_pong++; log_debug(knet_h, KNET_SUB_RX, "host: %u link: %u received pong: %u", src_host->host_id, src_link->link_id, src_link->received_pong); } } } /* Calculate latency stats */ if (src_link->status.latency > src_link->status.stats.latency_max) { src_link->status.stats.latency_max = src_link->status.latency; } if (src_link->status.latency < src_link->status.stats.latency_min) { src_link->status.stats.latency_min = src_link->status.latency; } /* * those 2 lines below make all latency average calculations consistent and capped to * link precision. In future we will kill the one above to keep only this one in * the stats structure, but for now we leave it around to avoid API/ABI * breakage as we backport the fixes to stable */ src_link->status.stats.latency_ave = src_link->status.latency; src_link->status.stats.latency_samples = src_link->latency_cur_samples; } break; case KNET_HEADER_TYPE_PMTUD: src_link->status.stats.rx_pmtu_packets++; src_link->status.stats.rx_pmtu_bytes += len; outlen = KNET_HEADER_PMTUD_SIZE; inbuf->kh_type = KNET_HEADER_TYPE_PMTUD_REPLY; inbuf->kh_node = htons(knet_h->host_id); if (knet_h->crypto_in_use_config) { if (crypto_encrypt_and_sign(knet_h, (const unsigned char *)inbuf, outlen, knet_h->recv_from_links_buf_crypt, &outlen) < 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to encrypt PMTUd reply packet"); break; } outbuf = knet_h->recv_from_links_buf_crypt; stats_err = pthread_mutex_lock(&knet_h->handle_stats_mutex); if (stats_err < 0) { log_err(knet_h, KNET_SUB_RX, "Unable to get mutex lock: %s", strerror(stats_err)); break; } knet_h->stats_extra.tx_crypt_pmtu_reply_packets++; pthread_mutex_unlock(&knet_h->handle_stats_mutex); } /* Unlock so we don't deadlock with tx_mutex */ pthread_mutex_unlock(&src_link->link_stats_mutex); savederrno = pthread_mutex_lock(&knet_h->tx_mutex); if (savederrno) { log_err(knet_h, KNET_SUB_RX, "Unable to get TX mutex lock: %s", strerror(savederrno)); goto out_pmtud; } retry_pmtud: if (src_link->transport_connected) { if (transport_get_connection_oriented(knet_h, src_link->transport) == TRANSPORT_PROTO_NOT_CONNECTION_ORIENTED) { len = sendto(src_link->outsock, outbuf, outlen, MSG_DONTWAIT | MSG_NOSIGNAL, (struct sockaddr *) &src_link->dst_addr, knet_h->knet_transport_fd_tracker[src_link->outsock].sockaddr_len); } else { len = sendto(src_link->outsock, outbuf, outlen, MSG_DONTWAIT | MSG_NOSIGNAL, NULL, 0); } savederrno = errno; if (len != outlen) { err = transport_tx_sock_error(knet_h, src_link->transport, src_link->outsock, KNET_SUB_RX, len, savederrno); stats_err = pthread_mutex_lock(&src_link->link_stats_mutex); if (stats_err < 0) { log_err(knet_h, KNET_SUB_RX, "Unable to get mutex lock: %s", strerror(stats_err)); break; } switch(err) { case -1: /* unrecoverable error */ log_debug(knet_h, KNET_SUB_RX, "Unable to send PMTUd reply (sock: %d) packet (sendto): %d %s. recorded src ip: %s src port: %s dst ip: %s dst port: %s", src_link->outsock, errno, strerror(errno), src_link->status.src_ipaddr, src_link->status.src_port, src_link->status.dst_ipaddr, src_link->status.dst_port); src_link->status.stats.tx_pmtu_errors++; break; case 0: /* ignore error and continue */ src_link->status.stats.tx_pmtu_errors++; break; case 1: /* retry to send those same data */ src_link->status.stats.tx_pmtu_retries++; pthread_mutex_unlock(&src_link->link_stats_mutex); goto retry_pmtud; break; } pthread_mutex_unlock(&src_link->link_stats_mutex); } } pthread_mutex_unlock(&knet_h->tx_mutex); out_pmtud: return; /* Don't need to unlock link_stats_mutex */ case KNET_HEADER_TYPE_PMTUD_REPLY: src_link->status.stats.rx_pmtu_packets++; src_link->status.stats.rx_pmtu_bytes += len; /* pmtud_mutex can't be acquired while we hold a link_stats_mutex (ordering) */ pthread_mutex_unlock(&src_link->link_stats_mutex); if (pthread_mutex_lock(&knet_h->pmtud_mutex) != 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to get mutex lock"); break; } src_link->last_recv_mtu = inbuf->khp_pmtud_size; pthread_cond_signal(&knet_h->pmtud_cond); pthread_mutex_unlock(&knet_h->pmtud_mutex); return; default: pthread_mutex_unlock(&src_link->link_stats_mutex); return; } pthread_mutex_unlock(&src_link->link_stats_mutex); } static void _handle_recv_from_links(knet_handle_t knet_h, int sockfd, struct knet_mmsghdr *msg) { int err, savederrno; int i, msg_recv, transport; if (pthread_rwlock_rdlock(&knet_h->global_rwlock) != 0) { log_debug(knet_h, KNET_SUB_RX, "Unable to get global read lock"); return; } if (_is_valid_fd(knet_h, sockfd) < 1) { /* * this is normal if a fd got an event and before we grab the read lock * and the link is removed by another thread */ goto exit_unlock; } transport = knet_h->knet_transport_fd_tracker[sockfd].transport; /* * reset msg_namelen to buffer size because after recvmmsg * each msg_namelen will contain sizeof sockaddr_in or sockaddr_in6 */ for (i = 0; i < PCKT_RX_BUFS; i++) { msg[i].msg_hdr.msg_namelen = knet_h->knet_transport_fd_tracker[sockfd].sockaddr_len; } msg_recv = _recvmmsg(sockfd, &msg[0], PCKT_RX_BUFS, MSG_DONTWAIT | MSG_NOSIGNAL); savederrno = errno; /* * WARNING: man page for recvmmsg is wrong. Kernel implementation here: * recvmmsg can return: * -1 on error * 0 if the previous run of recvmmsg recorded an error on the socket * N number of messages (see exception below). * * If there is an error from recvmsg after receiving a frame or more, the recvmmsg * loop is interrupted, error recorded in the socket (getsockopt(SO_ERROR) and * it will be visibile in the next run. * * Need to be careful how we handle errors at this stage. * * error messages need to be handled on a per transport/protocol base * at this point we have different layers of error handling * - msg_recv < 0 -> error from this run * msg_recv = 0 -> error from previous run and error on socket needs to be cleared * - per-transport message data * example: msg[i].msg_hdr.msg_flags & MSG_NOTIFICATION or msg_len for SCTP == EOF, * but for UDP it is perfectly legal to receive a 0 bytes message.. go figure * - NOTE: on SCTP MSG_NOTIFICATION we get msg_recv == PCKT_FRAG_MAX messages and no * errno set. That means the error api needs to be able to abort the loop below. */ if (msg_recv <= 0) { transport_rx_sock_error(knet_h, transport, sockfd, msg_recv, savederrno); goto exit_unlock; } for (i = 0; i < msg_recv; i++) { err = transport_rx_is_data(knet_h, transport, sockfd, &msg[i]); /* * TODO: make this section silent once we are confident * all protocols packet handlers are good */ switch(err) { case KNET_TRANSPORT_RX_ERROR: /* on error */ log_debug(knet_h, KNET_SUB_RX, "Transport reported error parsing packet"); goto exit_unlock; break; case KNET_TRANSPORT_RX_NOT_DATA_CONTINUE: /* packet is not data and we should continue the packet process loop */ log_debug(knet_h, KNET_SUB_RX, "Transport reported no data, continue"); break; case KNET_TRANSPORT_RX_NOT_DATA_STOP: /* packet is not data and we should STOP the packet process loop */ log_debug(knet_h, KNET_SUB_RX, "Transport reported no data, stop"); goto exit_unlock; break; case KNET_TRANSPORT_RX_IS_DATA: /* packet is data and should be parsed as such */ _parse_recv_from_links(knet_h, sockfd, &msg[i]); break; case KNET_TRANSPORT_RX_OOB_DATA_CONTINUE: log_debug(knet_h, KNET_SUB_RX, "Transport is processing sock OOB data, continue"); break; case KNET_TRANSPORT_RX_OOB_DATA_STOP: log_debug(knet_h, KNET_SUB_RX, "Transport has completed processing sock OOB data, stop"); goto exit_unlock; break; } } exit_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); } void *_handle_recv_from_links_thread(void *data) { int i, nev; knet_handle_t knet_h = (knet_handle_t) data; struct epoll_event events[KNET_EPOLL_MAX_EVENTS]; struct sockaddr_storage address[PCKT_RX_BUFS]; struct knet_mmsghdr msg[PCKT_RX_BUFS]; struct iovec iov_in[PCKT_RX_BUFS]; #if defined(IP_PKTINFO) || defined(IPV6_PKTINFO) - unsigned char control_in[CMSG_SPACE(sizeof(struct in6_pktinfo))][PCKT_RX_BUFS]; + unsigned char control_in[PCKT_RX_BUFS][CMSG_SPACE(sizeof(struct in6_pktinfo))]; #endif set_thread_status(knet_h, KNET_THREAD_RX, KNET_THREAD_STARTED); memset(&msg, 0, sizeof(msg)); memset(&events, 0, sizeof(events)); for (i = 0; i < PCKT_RX_BUFS; i++) { iov_in[i].iov_base = (void *)knet_h->recv_from_links_buf[i]; iov_in[i].iov_len = KNET_DATABUFSIZE; memset(&msg[i].msg_hdr, 0, sizeof(struct msghdr)); msg[i].msg_hdr.msg_name = &address[i]; msg[i].msg_hdr.msg_namelen = sizeof(struct sockaddr_storage); /* Real value filled in before actual use */ msg[i].msg_hdr.msg_iov = &iov_in[i]; msg[i].msg_hdr.msg_iovlen = 1; #if defined(IP_PKTINFO) || defined(IPV6_PKTINFO) - msg[i].msg_hdr.msg_control = &control_in[0][i]; + msg[i].msg_hdr.msg_control = &control_in[i][0]; msg[i].msg_hdr.msg_controllen = CMSG_SPACE(sizeof(struct in6_pktinfo)); /* Largest of the two pktinfo structs */ #endif } while (!shutdown_in_progress(knet_h)) { nev = epoll_wait(knet_h->recv_from_links_epollfd, events, KNET_EPOLL_MAX_EVENTS, KNET_THREADS_TIMERES / 1000); /* * the RX threads only need to notify that there has been at least * one successful run after queue flush has been requested. * See setfwd in handle.c */ if (get_thread_flush_queue(knet_h, KNET_THREAD_RX) == KNET_THREAD_QUEUE_FLUSH) { set_thread_flush_queue(knet_h, KNET_THREAD_RX, KNET_THREAD_QUEUE_FLUSHED); } /* * we use timeout to detect if thread is shutting down */ if (nev == 0) { continue; } for (i = 0; i < nev; i++) { _handle_recv_from_links(knet_h, events[i].data.fd, msg); } } set_thread_status(knet_h, KNET_THREAD_RX, KNET_THREAD_STOPPED); return NULL; } ssize_t knet_recv(knet_handle_t knet_h, char *buff, const size_t buff_len, const int8_t channel) { int savederrno = 0; ssize_t err = 0; struct iovec iov_in; if (!_is_valid_handle(knet_h)) { return -1; } if (buff == NULL) { errno = EINVAL; return -1; } if (buff_len <= 0) { errno = EINVAL; return -1; } if (buff_len > KNET_MAX_PACKET_SIZE) { errno = EINVAL; return -1; } if (channel < 0) { errno = EINVAL; return -1; } if (channel >= KNET_DATAFD_MAX) { errno = EINVAL; return -1; } savederrno = pthread_rwlock_rdlock(&knet_h->global_rwlock); if (savederrno) { log_err(knet_h, KNET_SUB_HANDLE, "Unable to get read lock: %s", strerror(savederrno)); errno = savederrno; return -1; } if (!knet_h->sockfd[channel].in_use) { savederrno = EINVAL; err = -1; goto out_unlock; } memset(&iov_in, 0, sizeof(iov_in)); iov_in.iov_base = (void *)buff; iov_in.iov_len = buff_len; err = readv(knet_h->sockfd[channel].sockfd[0], &iov_in, 1); savederrno = errno; out_unlock: pthread_rwlock_unlock(&knet_h->global_rwlock); errno = err ? savederrno : 0; return err; } diff --git a/libknet/transport_common.c b/libknet/transport_common.c index abd86403..afe938ca 100644 --- a/libknet/transport_common.c +++ b/libknet/transport_common.c @@ -1,448 +1,465 @@ /* * Copyright (C) 2016-2025 Red Hat, Inc. All rights reserved. * * Author: Fabio M. Di Nitto * * This software licensed under LGPL-2.0+ */ #include "config.h" #include #include #include #include #include #include #include #include #include "libknet.h" #include "compat.h" #include "host.h" #include "link.h" #include "logging.h" #include "common.h" #include "transport_common.h" /* * reuse Jan Friesse's compat layer as wrapper to drop usage of sendmmsg * * TODO: kill those wrappers once we work on packet delivery guarantees */ int _recvmmsg(int sockfd, struct knet_mmsghdr *msgvec, unsigned int vlen, unsigned int flags) { int savederrno = 0, err = 0; unsigned int i; for (i = 0; i < vlen; i++) { err = recvmsg(sockfd, &msgvec[i].msg_hdr, flags); savederrno = errno; if (err >= 0) { msgvec[i].msg_len = err; if (err == 0) { /* No point in reading anything more until we know this has been dealt with or we'll just get a vector full of them. Several in fact */ i++; break; } } else { if ((i > 0) && ((errno == EAGAIN) || (errno == EWOULDBLOCK))) { savederrno = 0; } break; } } errno = savederrno; return ((i > 0) ? (int)i : err); } int _sendmmsg(int sockfd, int connection_oriented, struct knet_mmsghdr *msgvec, unsigned int vlen, unsigned int flags) { int savederrno = 0, err = 0; unsigned int i; struct msghdr temp_msg; struct msghdr *use_msghdr; for (i = 0; i < vlen; i++) { if (connection_oriented == TRANSPORT_PROTO_IS_CONNECTION_ORIENTED) { memcpy(&temp_msg, &msgvec[i].msg_hdr, sizeof(struct msghdr)); temp_msg.msg_name = NULL; temp_msg.msg_namelen = 0; use_msghdr = &temp_msg; } else { use_msghdr = &msgvec[i].msg_hdr; } err = sendmsg(sockfd, use_msghdr, flags); savederrno = errno; if (err < 0) { break; } } errno = savederrno; return ((i > 0) ? (int)i : err); } /* Assume neither of these constants can ever be zero */ #ifndef SO_RCVBUFFORCE #define SO_RCVBUFFORCE 0 #endif #ifndef SO_SNDBUFFORCE #define SO_SNDBUFFORCE 0 #endif static int _configure_sockbuf(knet_handle_t knet_h, int sock, int option, int force, int target) { int savederrno = 0; int new_value; socklen_t value_len = sizeof new_value; if (setsockopt(sock, SOL_SOCKET, option, &target, sizeof target) != 0) { savederrno = errno; log_err(knet_h, KNET_SUB_TRANSPORT, "Error setting socket buffer via option %d to value %d: %s\n", option, target, strerror(savederrno)); errno = savederrno; return -1; } if (getsockopt(sock, SOL_SOCKET, option, &new_value, &value_len) != 0) { savederrno = errno; log_err(knet_h, KNET_SUB_TRANSPORT, "Error getting socket buffer via option %d: %s\n", option, strerror(savederrno)); errno = savederrno; return -1; } if (value_len != sizeof new_value) { log_err(knet_h, KNET_SUB_TRANSPORT, "Socket option %d returned unexpected size %u\n", option, value_len); errno = ERANGE; return -1; } if (target <= new_value) { return 0; } if (!force || !(knet_h->flags & KNET_HANDLE_FLAG_PRIVILEGED)) { log_err(knet_h, KNET_SUB_TRANSPORT, "Failed to set socket buffer via option %d to value %d: capped at %d", option, target, new_value); if (!(knet_h->flags & KNET_HANDLE_FLAG_PRIVILEGED)) { log_err(knet_h, KNET_SUB_TRANSPORT, "Continuing regardless, as the handle is not privileged." " Expect poor performance!"); return 0; } else { errno = ENAMETOOLONG; return -1; } } if (setsockopt(sock, SOL_SOCKET, force, &target, sizeof target) < 0) { savederrno = errno; log_err(knet_h, KNET_SUB_TRANSPORT, "Failed to set socket buffer via force option %d: %s", force, strerror(savederrno)); if (savederrno == EPERM) { errno = ENAMETOOLONG; } else { errno = savederrno; } return -1; } return 0; } int _configure_common_socket(knet_handle_t knet_h, int sock, uint64_t flags, const char *type) { int err = 0, savederrno = 0; - int value; if (_fdset_cloexec(sock)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s CLOEXEC socket opts: %s", type, strerror(savederrno)); goto exit_error; } if (_fdset_nonblock(sock)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s NONBLOCK socket opts: %s", type, strerror(savederrno)); goto exit_error; } if (_configure_sockbuf(knet_h, sock, SO_RCVBUF, SO_RCVBUFFORCE, KNET_RING_RCVBUFF)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s receive buffer: %s", type, strerror(savederrno)); goto exit_error; } if (_configure_sockbuf(knet_h, sock, SO_SNDBUF, SO_SNDBUFFORCE, KNET_RING_RCVBUFF)) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s send buffer: %s", type, strerror(savederrno)); goto exit_error; } if (flags & KNET_LINK_FLAG_TRAFFICHIPRIO) { #ifdef KNET_LINUX #ifdef SO_PRIORITY - value = 6; /* TC_PRIO_INTERACTIVE */ + int value = 6; /* TC_PRIO_INTERACTIVE */ + if (setsockopt(sock, SOL_SOCKET, SO_PRIORITY, &value, sizeof(value)) < 0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s priority: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "TC_PRIO_INTERACTIVE enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "TC_PRIO_INTERACTIVE not available in this build/platform"); #endif #endif -#if defined(IP_TOS) && defined(IPTOS_LOWDELAY) - value = IPTOS_LOWDELAY; - if (setsockopt(sock, IPPROTO_IP, IP_TOS, &value, sizeof(value)) < 0) { - savederrno = errno; - err = -1; - log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s priority: %s", - type, strerror(savederrno)); - goto exit_error; - } - log_debug(knet_h, KNET_SUB_TRANSPORT, "IPTOS_LOWDELAY enabled on socket: %i", sock); +#if defined(IP_TOS) + if (knet_h->prio_dscp) { + /* dscp is the 6 highest bits of TOS IP header field, RFC 2474 */ + int value = (knet_h->prio_dscp & 0x3f) << 2; + + if (setsockopt(sock, IPPROTO_IP, IP_TOS, &value, sizeof(value)) < 0) { + savederrno = errno; + err = -1; + log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s priority: %s", + type, strerror(savederrno)); + goto exit_error; + } + log_debug(knet_h, KNET_SUB_TRANSPORT, "dscp %d set on socket: %i", knet_h->prio_dscp, sock); + } else { +#if defined(IPTOS_LOWDELAY) + int value = IPTOS_LOWDELAY; + + if (setsockopt(sock, IPPROTO_IP, IP_TOS, &value, sizeof(value)) < 0) { + savederrno = errno; + err = -1; + log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s priority: %s", + type, strerror(savederrno)); + goto exit_error; + } + log_debug(knet_h, KNET_SUB_TRANSPORT, "IPTOS_LOWDELAY enabled on socket: %i", sock); #else - log_debug(knet_h, KNET_SUB_TRANSPORT, "IPTOS_LOWDELAY not available in this build/platform"); + log_debug(knet_h, KNET_SUB_TRANSPORT, "IPTOS_LOWDELAY not available in this build/platform"); +#endif + } #endif } exit_error: errno = savederrno; return err; } int _configure_transport_socket(knet_handle_t knet_h, int sock, struct sockaddr_storage *address, uint64_t flags, const char *type) { int err = 0, savederrno = 0; int value; if (_configure_common_socket(knet_h, sock, flags, type) < 0) { savederrno = errno; err = -1; goto exit_error; } #ifdef KNET_LINUX #ifdef IP_FREEBIND value = 1; if (setsockopt(sock, SOL_IP, IP_FREEBIND, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set FREEBIND on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "FREEBIND enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "FREEBIND not available in this build/platform"); #endif #endif #ifdef KNET_BSD #ifdef IP_BINDANY /* BSD */ value = 1; if (setsockopt(sock, IPPROTO_IP, IP_BINDANY, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set BINDANY on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "BINDANY enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "BINDANY not available in this build/platform"); #endif #endif if (address->ss_family == AF_INET6) { value = 1; if (setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY, &value, sizeof(value)) < 0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set %s IPv6 only: %s", type, strerror(savederrno)); goto exit_error; } #ifdef KNET_LINUX #ifdef IPV6_MTU_DISCOVER value = IPV6_PMTUDISC_PROBE; if (setsockopt(sock, SOL_IPV6, IPV6_MTU_DISCOVER, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set PMTUDISC on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "IPV6_MTU_DISCOVER enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "IPV6_MTU_DISCOVER not available in this build/platform"); #endif #endif #ifdef IPV6_DONTFRAG value = 1; if (setsockopt(sock, IPPROTO_IPV6, IPV6_DONTFRAG, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set DONTFRAG on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "IPV6_DONTFRAG enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "IPV6_DONTFRAG not available in this build/platform"); #endif } else { #ifdef KNET_LINUX #ifdef IP_MTU_DISCOVER value = IP_PMTUDISC_PROBE; if (setsockopt(sock, SOL_IP, IP_MTU_DISCOVER, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set PMTUDISC on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "PMTUDISC enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "PMTUDISC not available in this build/platform"); #endif #endif #ifdef KNET_BSD #ifdef IP_DONTFRAG value = 1; if (setsockopt(sock, IPPROTO_IP, IP_DONTFRAG, &value, sizeof(value)) <0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_TRANSPORT, "Unable to set DONTFRAG on %s socket: %s", type, strerror(savederrno)); goto exit_error; } log_debug(knet_h, KNET_SUB_TRANSPORT, "DONTFRAG enabled on socket: %i", sock); #else log_debug(knet_h, KNET_SUB_TRANSPORT, "DONTFRAG not available in this build/platform"); #endif #endif } exit_error: errno = savederrno; return err; } int _init_socketpair(knet_handle_t knet_h, int *sock) { int err = 0, savederrno = 0; int i; if (socketpair(AF_UNIX, SOCK_SEQPACKET, 0, sock) != 0) { savederrno = errno; err = -1; log_err(knet_h, KNET_SUB_HANDLE, "Unable to initialize socketpair: %s", strerror(savederrno)); goto exit_fail; } for (i = 0; i < 2; i++) { if (_configure_common_socket(knet_h, sock[i], 0, "local socketpair") < 0) { savederrno = errno; err = -1; goto exit_fail; } } exit_fail: errno = savederrno; return err; } void _close_socketpair(knet_handle_t knet_h, int *sock) { int i; for (i = 0; i < 2; i++) { if (sock[i]) { close(sock[i]); sock[i] = 0; } } } /* * must be called with global read lock * * return -1 on error * return 0 if fd is invalid * return 1 if fd is valid */ int _is_valid_fd(knet_handle_t knet_h, int sockfd) { int ret = 0; if (sockfd < 0) { errno = EINVAL; return -1; } if (sockfd >= KNET_MAX_FDS) { errno = EINVAL; return -1; } if (knet_h->knet_transport_fd_tracker[sockfd].transport >= KNET_MAX_TRANSPORTS) { ret = 0; } else { ret = 1; } return ret; } /* * must be called with global write lock */ int _set_fd_tracker(knet_handle_t knet_h, int sockfd, uint8_t transport, uint8_t data_type, socklen_t socklen, void *data, int ifindex) { if (sockfd < 0) { errno = EINVAL; return -1; } if (sockfd >= KNET_MAX_FDS) { errno = EINVAL; return -1; } knet_h->knet_transport_fd_tracker[sockfd].transport = transport; knet_h->knet_transport_fd_tracker[sockfd].data_type = data_type; knet_h->knet_transport_fd_tracker[sockfd].sockaddr_len = socklen; knet_h->knet_transport_fd_tracker[sockfd].data = data; knet_h->knet_transport_fd_tracker[sockfd].ifindex = ifindex; return 0; } diff --git a/libnozzle/libnozzle.c b/libnozzle/libnozzle.c index d84b8f2a..01878bd9 100644 --- a/libnozzle/libnozzle.c +++ b/libnozzle/libnozzle.c @@ -1,1243 +1,1231 @@ /* * Copyright (C) 2010-2025 Red Hat, Inc. All rights reserved. * * Author: Fabio M. Di Nitto * * This software licensed under LGPL-2.0+ */ #include "config.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef KNET_LINUX #include /* * libnl3 < 3.3 includes kernel headers directly * causing conflicts with net/if.h included above */ #ifdef LIBNL3_WORKAROUND #define _LINUX_IF_H 1 #endif #include #include #include #include #endif #ifdef KNET_BSD #include #include #endif #include "libnozzle.h" #include "internals.h" /* * internal functions are all _unlocked_ * locking should be handled at external API functions */ static int lib_init = 0; static struct nozzle_lib_config lib_cfg; static pthread_mutex_t config_mutex = PTHREAD_MUTEX_INITIALIZER; /* * internal helpers */ static void lib_fini(void) { if (lib_cfg.head == NULL) { #ifdef KNET_LINUX nl_close(lib_cfg.nlsock); nl_socket_free(lib_cfg.nlsock); #endif close(lib_cfg.ioctlfd); lib_init = 0; } } static int is_valid_nozzle(const nozzle_t nozzle) { nozzle_t temp; if (!nozzle) { return 0; } if (!lib_init) { return 0; } temp = lib_cfg.head; while (temp != NULL) { if (nozzle == temp) return 1; temp = temp->next; } return 0; } static void destroy_iface(nozzle_t nozzle) { #ifdef KNET_BSD struct ifreq ifr; #endif if (!nozzle) return; if (nozzle->fd >= 0) close(nozzle->fd); #ifdef KNET_BSD memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); ioctl(lib_cfg.ioctlfd, SIOCIFDESTROY, &ifr); + ioctl(lib_cfg.ioctlfd, SIOCGIFFLAGS, &ifr); #endif free(nozzle); lib_fini(); return; } static int get_iface_mtu(const nozzle_t nozzle) { int err = 0, savederrno = 0; struct ifreq ifr; memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); err = ioctl(lib_cfg.ioctlfd, SIOCGIFMTU, &ifr); if (err) { savederrno = errno; goto out_clean; } err = ifr.ifr_mtu; out_clean: errno = savederrno; return err; } static int get_iface_mac(const nozzle_t nozzle, char **ether_addr) { int err = 0, savederrno = 0; struct ifreq ifr; char mac[MACADDR_CHAR_MAX]; #ifdef KNET_BSD struct ifaddrs *ifap = NULL; struct ifaddrs *ifa; int found = 0; #endif memset(&mac, 0, MACADDR_CHAR_MAX); memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); #ifdef KNET_LINUX err = ioctl(lib_cfg.ioctlfd, SIOCGIFHWADDR, &ifr); if (err) { savederrno = errno; goto out_clean; } ether_ntoa_r((struct ether_addr *)ifr.ifr_hwaddr.sa_data, mac); #endif #ifdef KNET_BSD /* * there is no ioctl to get the ether address of an interface on FreeBSD * (not to be confused with hwaddr). Use workaround described here: * https://lists.freebsd.org/pipermail/freebsd-hackers/2004-June/007394.html */ err = getifaddrs(&ifap); if (err < 0) { savederrno = errno; goto out_clean; } ifa = ifap; while (ifa) { if (!strncmp(nozzle->name, ifa->ifa_name, IFNAMSIZ)) { found = 1; break; } ifa=ifa->ifa_next; } if (found) { ether_ntoa_r((struct ether_addr *)LLADDR((struct sockaddr_dl *)ifa->ifa_addr), mac); } else { errno = EINVAL; err = -1; } freeifaddrs(ifap); if (err) { goto out_clean; } #endif *ether_addr = strdup(mac); if (!*ether_addr) { savederrno = errno; err = -1; } out_clean: errno = savederrno; return err; } #define IP_ADD 1 #define IP_DEL 2 static int _set_ip(nozzle_t nozzle, int command, const char *ipaddr, const char *prefix, int secondary) { int fam; char *broadcast = NULL; int err = 0; #ifdef KNET_LINUX struct rtnl_addr *addr = NULL; struct nl_addr *local_addr = NULL; struct nl_addr *bcast_addr = NULL; struct nl_cache *cache = NULL; int ifindex; #endif #ifdef KNET_BSD char cmdline[4096]; char proto[6]; char *error_string = NULL; #endif if (!strchr(ipaddr, ':')) { fam = AF_INET; broadcast = generate_v4_broadcast(ipaddr, prefix); if (!broadcast) { errno = EINVAL; return -1; } } else { fam = AF_INET6; } #ifdef KNET_LINUX addr = rtnl_addr_alloc(); if (!addr) { errno = ENOMEM; err = -1; goto out; } if (rtnl_link_alloc_cache(lib_cfg.nlsock, AF_UNSPEC, &cache) < 0) { errno = ENOMEM; err = -1; goto out; } ifindex = rtnl_link_name2i(cache, nozzle->name); if (ifindex == 0) { errno = ENOENT; err = -1; goto out; } rtnl_addr_set_ifindex(addr, ifindex); if (nl_addr_parse(ipaddr, fam, &local_addr) < 0) { errno = EINVAL; err = -1; goto out; } if (rtnl_addr_set_local(addr, local_addr) < 0) { errno = EINVAL; err = -1; goto out; } if (broadcast) { if (nl_addr_parse(broadcast, fam, &bcast_addr) < 0) { errno = EINVAL; err = -1; goto out; } if (rtnl_addr_set_broadcast(addr, bcast_addr) < 0) { errno = EINVAL; err = -1; goto out; } } rtnl_addr_set_prefixlen(addr, atoi(prefix)); if (command == IP_ADD) { if (rtnl_addr_add(lib_cfg.nlsock, addr, 0) < 0) { errno = EINVAL; err = -1; goto out; } } else { if (rtnl_addr_delete(lib_cfg.nlsock, addr, 0) < 0) { errno = EINVAL; err = -1; goto out; } } out: if (addr) { rtnl_addr_put(addr); } if (local_addr) { nl_addr_put(local_addr); } if (bcast_addr) { nl_addr_put(bcast_addr); } if (cache) { nl_cache_put(cache); } if (broadcast) { free(broadcast); } return err; #endif #ifdef KNET_BSD /* * TODO: port to use ioctl and such, drop shell forking here */ memset(cmdline, 0, sizeof(cmdline)); if (fam == AF_INET) { snprintf(proto, sizeof(proto), "inet"); } else { snprintf(proto, sizeof(proto), "inet6"); } if (command == IP_ADD) { snprintf(cmdline, sizeof(cmdline)-1, "ifconfig %s %s %s/%s", nozzle->name, proto, ipaddr, prefix); if (broadcast) { snprintf(cmdline + strlen(cmdline), sizeof(cmdline) - strlen(cmdline) -1, " broadcast %s", broadcast); } if ((secondary) && (fam == AF_INET)) { snprintf(cmdline + strlen(cmdline), sizeof(cmdline) - strlen(cmdline) -1, " alias"); } } else { snprintf(cmdline, sizeof(cmdline)-1, "ifconfig %s %s %s/%s delete", nozzle->name, proto, ipaddr, prefix); } if (broadcast) { free(broadcast); } /* * temporary workaround as we port libnozzle to BSD ioctl * for IP address management */ err = execute_bin_sh_command(cmdline, &error_string); if (error_string) { free(error_string); error_string = NULL; } return err; #endif } /* * Exported public API */ nozzle_t nozzle_open(char *devname, size_t devname_size, const char *updownpath) { int savederrno = 0; nozzle_t nozzle = NULL; char *temp_mac = NULL; #ifdef KNET_LINUX struct ifreq ifr; #endif #ifdef KNET_BSD uint16_t i; long int nozzlenum = 0; char curnozzle[IFNAMSIZ]; struct ifreq ifr; #endif if (devname == NULL) { errno = EINVAL; return NULL; } if (devname_size < IFNAMSIZ) { errno = EINVAL; return NULL; } /* Need to allow space for trailing NUL */ if (strlen(devname) >= IFNAMSIZ) { errno = E2BIG; return NULL; } #ifdef KNET_BSD /* * BSD does not support named devices like Linux * but it is possible to force a nozzleX device number * where X is 0 to 255. */ if (strlen(devname)) { if (strncmp(devname, "tap", 3)) { errno = EINVAL; return NULL; } errno = 0; nozzlenum = strtol(devname+3, NULL, 10); if (errno) { errno = EINVAL; return NULL; } if ((nozzlenum < 0) || (nozzlenum > 255)) { errno = EINVAL; return NULL; } } #endif if (updownpath) { /* only absolute paths */ if (updownpath[0] != '/') { errno = EINVAL; return NULL; } if (strlen(updownpath) >= UPDOWN_PATH_MAX) { errno = E2BIG; return NULL; } } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return NULL; } if (!lib_init) { lib_cfg.head = NULL; #ifdef KNET_LINUX lib_cfg.nlsock = nl_socket_alloc(); if (!lib_cfg.nlsock) { savederrno = errno; goto out_error; } if (nl_connect(lib_cfg.nlsock, NETLINK_ROUTE) < 0) { savederrno = EBUSY; goto out_error; } lib_cfg.ioctlfd = socket(AF_INET, SOCK_STREAM, 0); #endif #ifdef KNET_BSD lib_cfg.ioctlfd = socket(AF_LOCAL, SOCK_DGRAM, 0); #endif if (lib_cfg.ioctlfd < 0) { savederrno = errno; goto out_error; } lib_init = 1; } nozzle = malloc(sizeof(struct nozzle_iface)); if (!nozzle) { savederrno = ENOMEM; goto out_error; } memset(nozzle, 0, sizeof(struct nozzle_iface)); #ifdef KNET_BSD if (!strlen(devname)) { - /* - * FreeBSD 13 kernel has changed how the tap module - * works and tap0 cannot be removed from the system. - * This means that tap0 settings are never reset to default - * and nozzle cannot control the default state of the device - * when taking over. - * nozzle expects some parameters to be default when opening - * a tap device (such as random mac address, default MTU, no - * other attributes, etc.) - * - * For 13 and higher, simply skip tap0 as usable device. - */ -#if __FreeBSD__ >= 13 - for (i = 1; i < 256; i++) { -#else for (i = 0; i < 256; i++) { -#endif memset(&ifr, 0, sizeof(struct ifreq)); snprintf(curnozzle, sizeof(curnozzle) - 1, "tap%u", i); memmove(ifr.ifr_name, curnozzle, IFNAMSIZ); if (ioctl(lib_cfg.ioctlfd, SIOCIFCREATE2, &ifr) < 0) { continue; } snprintf(curnozzle, sizeof(curnozzle) - 1, "/dev/tap%u", i); nozzle->fd = open(curnozzle, O_RDWR); savederrno = errno; if (nozzle->fd > 0) { break; } /* For some reason we can't open that device, keep trying but don't leave debris */ (void)ioctl(lib_cfg.ioctlfd, SIOCIFDESTROY, &ifr); + (void)ioctl(lib_cfg.ioctlfd, SIOCGIFFLAGS, &ifr); } snprintf(curnozzle, sizeof(curnozzle) -1 , "tap%u", i); } else { memmove(ifr.ifr_name, devname, IFNAMSIZ); if (ioctl(lib_cfg.ioctlfd, SIOCIFCREATE2, &ifr) < 0) { goto out_error; } snprintf(curnozzle, sizeof(curnozzle) - 1, "/dev/%s", devname); nozzle->fd = open(curnozzle, O_RDWR); savederrno = errno; snprintf(curnozzle, sizeof(curnozzle) - 1, "%s", devname); } if (nozzle->fd < 0) { savederrno = EBUSY; goto out_error; } memmove(devname, curnozzle, IFNAMSIZ); memmove(nozzle->name, curnozzle, IFNAMSIZ); #endif #ifdef KNET_LINUX if ((nozzle->fd = open("/dev/net/tun", O_RDWR)) < 0) { savederrno = errno; goto out_error; } memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, devname, IFNAMSIZ); ifr.ifr_flags = IFF_TAP | IFF_NO_PI; if (ioctl(nozzle->fd, TUNSETIFF, &ifr) < 0) { savederrno = errno; goto out_error; } if ((strlen(devname) > 0) && (strcmp(devname, ifname) != 0)) { savederrno = EBUSY; goto out_error; } memmove(devname, ifname, IFNAMSIZ); memmove(nozzle->name, ifname, IFNAMSIZ); #endif nozzle->default_mtu = get_iface_mtu(nozzle); if (nozzle->default_mtu < 0) { savederrno = errno; goto out_error; } if (get_iface_mac(nozzle, &temp_mac) < 0) { savederrno = errno; goto out_error; } strncpy(nozzle->default_mac, temp_mac, 18); free(temp_mac); if (updownpath) { int len = strlen(updownpath); strcpy(nozzle->updownpath, updownpath); if (nozzle->updownpath[len-1] != '/') { nozzle->updownpath[len] = '/'; } nozzle->hasupdown = 1; } nozzle->next = lib_cfg.head; lib_cfg.head = nozzle; pthread_mutex_unlock(&config_mutex); errno = savederrno; return nozzle; out_error: destroy_iface(nozzle); pthread_mutex_unlock(&config_mutex); errno = savederrno; return NULL; } int nozzle_close(nozzle_t nozzle) { int err = 0, savederrno = 0; - nozzle_t temp = lib_cfg.head; - nozzle_t prev = lib_cfg.head; + nozzle_t temp; + nozzle_t prev; struct nozzle_ip *ip, *ip_next; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } + temp = lib_cfg.head; + prev = lib_cfg.head; while ((temp) && (temp != nozzle)) { prev = temp; temp = temp->next; } if (nozzle == prev) { lib_cfg.head = nozzle->next; } else { prev->next = nozzle->next; } ip = nozzle->ip; while (ip) { ip_next = ip->next; free(ip); ip = ip_next; } destroy_iface(nozzle); out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_run_updown(const nozzle_t nozzle, uint8_t action, char **exec_string) { int err = 0, savederrno = 0; char command[PATH_MAX]; const char *action_str = NULL; struct stat sb; if (action > NOZZLE_POSTDOWN) { errno = EINVAL; return -1; } if (!exec_string) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } if (!nozzle->hasupdown) { savederrno = EINVAL; err = -1; goto out_clean; } switch(action) { case NOZZLE_PREUP: action_str = "pre-up.d"; break; case NOZZLE_UP: action_str = "up.d"; break; case NOZZLE_DOWN: action_str = "down.d"; break; case NOZZLE_POSTDOWN: action_str = "post-down.d"; break; } memset(command, 0, PATH_MAX); snprintf(command, PATH_MAX, "%s/%s/%s", nozzle->updownpath, action_str, nozzle->name); err = stat(command, &sb); if (err) { savederrno = errno; goto out_clean; } /* * clear errno from previous calls as there is no errno * returned from execute_bin_sh_command */ savederrno = 0; err = execute_bin_sh_command(command, exec_string); if (err) { err = -2; } out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_set_up(nozzle_t nozzle) { int err = 0, savederrno = 0; struct ifreq ifr; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } if (nozzle->up) { goto out_clean; } memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); err = ioctl(lib_cfg.ioctlfd, SIOCGIFFLAGS, &ifr); if (err) { savederrno = errno; goto out_clean; } ifr.ifr_flags |= IFF_UP | IFF_RUNNING; err = ioctl(lib_cfg.ioctlfd, SIOCSIFFLAGS, &ifr); if (err) { savederrno = errno; goto out_clean; } nozzle->up = 1; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_set_down(nozzle_t nozzle) { int err = 0, savederrno = 0; struct ifreq ifr; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } if (!nozzle->up) { goto out_clean; } memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); err = ioctl(lib_cfg.ioctlfd, SIOCGIFFLAGS, &ifr); if (err) { savederrno = errno; goto out_clean; } ifr.ifr_flags &= ~IFF_UP; err = ioctl(lib_cfg.ioctlfd, SIOCSIFFLAGS, &ifr); if (err) { savederrno = errno; goto out_clean; } nozzle->up = 0; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_get_mtu(const nozzle_t nozzle) { int err = 0, savederrno = 0; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } err = get_iface_mtu(nozzle); savederrno = errno; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_get_mac(const nozzle_t nozzle, char **ether_addr) { int err = 0, savederrno = 0; if (!ether_addr) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } err = get_iface_mac(nozzle, ether_addr); out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_set_mac(nozzle_t nozzle, const char *ether_addr) { int err = 0, savederrno = 0; struct ifreq ifr; if (!ether_addr) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); #ifdef KNET_LINUX err = ioctl(lib_cfg.ioctlfd, SIOCGIFHWADDR, &ifr); if (err) { savederrno = errno; goto out_clean; } memmove(ifr.ifr_hwaddr.sa_data, ether_aton(ether_addr), ETH_ALEN); err = ioctl(lib_cfg.ioctlfd, SIOCSIFHWADDR, &ifr); savederrno = errno; #endif #ifdef KNET_BSD err = ioctl(lib_cfg.ioctlfd, SIOCGIFADDR, &ifr); if (err) { savederrno = errno; goto out_clean; } memmove(ifr.ifr_addr.sa_data, ether_aton(ether_addr), ETHER_ADDR_LEN); ifr.ifr_addr.sa_len = ETHER_ADDR_LEN; err = ioctl(lib_cfg.ioctlfd, SIOCSIFLLADDR, &ifr); savederrno = errno; #endif out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_reset_mac(nozzle_t nozzle) { return nozzle_set_mac(nozzle, nozzle->default_mac); } nozzle_t nozzle_get_handle_by_name(const char *devname) { int savederrno = 0; nozzle_t nozzle; if ((devname == NULL) || (strlen(devname) > IFNAMSIZ)) { errno = EINVAL; return NULL; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return NULL; } nozzle = lib_cfg.head; while (nozzle != NULL) { if (!strcmp(devname, nozzle->name)) break; nozzle = nozzle->next; } if (!nozzle) { savederrno = ENOENT; } pthread_mutex_unlock(&config_mutex); errno = savederrno; return nozzle; } const char *nozzle_get_name_by_handle(const nozzle_t nozzle) { int savederrno = 0; char *name = NULL; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return NULL; } if (!is_valid_nozzle(nozzle)) { savederrno = ENOENT; goto out_clean; } name = nozzle->name; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return name; } int nozzle_get_fd(const nozzle_t nozzle) { int fd = -1, savederrno = 0; savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = ENOENT; fd = -1; goto out_clean; } fd = nozzle->fd; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return fd; } int nozzle_set_mtu(nozzle_t nozzle, const int mtu) { int err = 0, savederrno = 0; struct nozzle_ip *tmp_ip; struct ifreq ifr; if (!mtu) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } err = nozzle->current_mtu = get_iface_mtu(nozzle); if (err < 0) { savederrno = errno; goto out_clean; } memset(&ifr, 0, sizeof(struct ifreq)); memmove(ifname, nozzle->name, IFNAMSIZ); ifr.ifr_mtu = mtu; err = ioctl(lib_cfg.ioctlfd, SIOCSIFMTU, &ifr); if (err) { savederrno = errno; goto out_clean; } if ((nozzle->current_mtu < 1280) && (mtu >= 1280)) { tmp_ip = nozzle->ip; while(tmp_ip) { if (tmp_ip->domain == AF_INET6) { err = _set_ip(nozzle, IP_ADD, tmp_ip->ipaddr, tmp_ip->prefix, 0); if (err) { savederrno = errno; err = -1; goto out_clean; } } tmp_ip = tmp_ip->next; } } out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_reset_mtu(nozzle_t nozzle) { return nozzle_set_mtu(nozzle, nozzle->default_mtu); } int nozzle_add_ip(nozzle_t nozzle, const char *ipaddr, const char *prefix) { int err = 0, savederrno = 0; int found = 0; struct nozzle_ip *ip = NULL, *ip_prev = NULL, *ip_last = NULL; int secondary = 0; if ((!ipaddr) || (!prefix)) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } found = find_ip(nozzle, ipaddr, prefix, &ip, &ip_prev); if (found) { goto out_clean; } ip = malloc(sizeof(struct nozzle_ip)); if (!ip) { savederrno = errno; err = -1 ; goto out_clean; } memset(ip, 0, sizeof(struct nozzle_ip)); strncpy(ip->ipaddr, ipaddr, IPADDR_CHAR_MAX); strncpy(ip->prefix, prefix, PREFIX_CHAR_MAX); if (!strchr(ip->ipaddr, ':')) { ip->domain = AF_INET; } else { ip->domain = AF_INET6; } /* * if user asks for an IPv6 address, but MTU < 1280 * store the IP and bring it up later if and when MTU > 1280 */ if ((ip->domain == AF_INET6) && (get_iface_mtu(nozzle) < 1280)) { err = 0; } else { if (nozzle->ip) { secondary = 1; } err = _set_ip(nozzle, IP_ADD, ipaddr, prefix, secondary); savederrno = errno; } if (err) { free(ip); goto out_clean; } if (nozzle->ip) { ip_last = nozzle->ip; while (ip_last->next != NULL) { ip_last = ip_last->next; } ip_last->next = ip; } else { nozzle->ip = ip; } out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_del_ip(nozzle_t nozzle, const char *ipaddr, const char *prefix) { int err = 0, savederrno = 0; int found = 0; struct nozzle_ip *ip = NULL, *ip_prev = NULL; if ((!ipaddr) || (!prefix)) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { savederrno = EINVAL; err = -1; goto out_clean; } found = find_ip(nozzle, ipaddr, prefix, &ip, &ip_prev); if (!found) { goto out_clean; } /* * if user asks for an IPv6 address, but MTU < 1280 * the IP might not be configured on the interface and we only need to * remove it from our internal database */ if ((ip->domain == AF_INET6) && (get_iface_mtu(nozzle) < 1280)) { err = 0; } else { err = _set_ip(nozzle, IP_DEL, ipaddr, prefix, 0); savederrno = errno; } if (!err) { if (ip == ip_prev) { nozzle->ip = ip->next; } else { ip_prev->next = ip->next; } free(ip); } out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } int nozzle_get_ips(const nozzle_t nozzle, struct nozzle_ip **nozzle_ip) { int err = 0, savederrno = 0; if (!nozzle_ip) { errno = EINVAL; return -1; } savederrno = pthread_mutex_lock(&config_mutex); if (savederrno) { errno = savederrno; return -1; } if (!is_valid_nozzle(nozzle)) { err = -1; savederrno = EINVAL; goto out_clean; } *nozzle_ip = nozzle->ip; out_clean: pthread_mutex_unlock(&config_mutex); errno = savederrno; return err; } diff --git a/libnozzle/tests/test-common.c b/libnozzle/tests/test-common.c index 70649bb0..00f26c88 100644 --- a/libnozzle/tests/test-common.c +++ b/libnozzle/tests/test-common.c @@ -1,201 +1,202 @@ /* * Copyright (C) 2018-2025 Red Hat, Inc. All rights reserved. * * Author: Fabio M. Di Nitto * * This software licensed under GPL-2.0+ */ #include "config.h" #include #include #include #include #include #include #include #include #include #include #ifdef KNET_BSD #include #include #endif #include "test-common.h" void need_root(void) { if (geteuid() != 0) { printf("This test requires root privileges\n"); exit(SKIP); } } void need_tun(void) { int fd; #ifdef KNET_LINUX const char *tundev = "/dev/net/tun"; #endif #ifdef KNET_BSD const char *tundev = "/dev/tap"; struct ifreq ifr; int ioctlfd = socket(AF_LOCAL, SOCK_DGRAM, 0); if (ioctlfd < 0) { printf("Unable to init ioctlfd (errno=%d)\n", errno); exit(FAIL); } #endif fd = open(tundev, O_RDWR); if (fd < 0) { printf("Failed to open %s (errno=%d); this test requires TUN support\n", tundev, errno); #ifdef KNET_BSD close(ioctlfd); #endif exit(SKIP); } #ifdef KNET_BSD memset(&ifr, 0, sizeof(struct ifreq)); ioctl(fd, TAPGIFNAME, &ifr); #endif close(fd); #ifdef KNET_BSD ioctl(ioctlfd, SIOCIFDESTROY, &ifr); + ioctl(ioctlfd, SIOCGIFFLAGS, &ifr); close(ioctlfd); #endif } int test_iface(char *name, size_t size, const char *updownpath) { nozzle_t nozzle; nozzle=nozzle_open(name, size, updownpath); if (!nozzle) { printf("Unable to open nozzle (errno=%d).\n", errno); return -1; } printf("Created interface: %s\n", name); if (is_if_in_system(name) > 0) { printf("Found interface %s on the system\n", name); } else { printf("Unable to find interface %s on the system\n", name); } if (!nozzle_get_handle_by_name(name)) { printf("Unable to find interface %s in nozzle db\n", name); } else { printf("Found interface %s in nozzle db\n", name); } nozzle_close(nozzle); if (is_if_in_system(name) == 0) printf("Successfully removed interface %s from the system\n", name); return 0; } int is_if_in_system(char *name) { struct ifaddrs *ifap = NULL; struct ifaddrs *ifa; int found = 0; if (getifaddrs(&ifap) < 0) { printf("Unable to get interface list.\n"); return -1; } ifa = ifap; while (ifa) { if (!strncmp(name, ifa->ifa_name, IFNAMSIZ)) { found = 1; break; } ifa=ifa->ifa_next; } freeifaddrs(ifap); return found; } int get_random_byte(void) { pid_t mypid; uint8_t *pid; uint8_t randombyte = 0; uint8_t i; if (sizeof(pid_t) < 4) { printf("pid_t is smaller than 4 bytes?\n"); exit(77); } mypid = getpid(); pid = (uint8_t *)&mypid; for (i = 0; i < sizeof(pid_t); i++) { if (pid[i] == 0) { pid[i] = 128; } } randombyte = pid[1]; return randombyte; } void make_local_ips(char *testipv4_1, char *testipv4_2, char *testipv6_1, char *testipv6_2) { pid_t mypid; uint8_t *pid; uint8_t i; memset(testipv4_1, 0, IPBUFSIZE); memset(testipv4_2, 0, IPBUFSIZE); memset(testipv6_1, 0, IPBUFSIZE); memset(testipv6_2, 0, IPBUFSIZE); mypid = getpid(); pid = (uint8_t *)&mypid; for (i = 0; i < sizeof(pid_t); i++) { if ((pid[i] == 0) || (pid[i] == 255)) { pid[i] = 128; } } snprintf(testipv4_1, IPBUFSIZE - 1, "127.%u.%u.%u", pid[1], pid[2], pid[0]); snprintf(testipv4_2, IPBUFSIZE - 1, "127.%u.%d.%u", pid[1], pid[2]+1, pid[0]); snprintf(testipv6_1, IPBUFSIZE - 1, "fd%x:%x%x::1", pid[1], pid[2], pid[0]); snprintf(testipv6_2, IPBUFSIZE - 1, "fd%x:%x%x:1::1", pid[1], pid[2], pid[0]); } diff --git a/man/Makefile.am b/man/Makefile.am index 2388b8c5..3de4da83 100644 --- a/man/Makefile.am +++ b/man/Makefile.am @@ -1,161 +1,162 @@ # # Copyright (C) 2017-2025 Red Hat, Inc. All rights reserved. # # Authors: Fabio M. Di Nitto # Federico Simoncelli # # This software licensed under GPL-2.0+ # MAINTAINERCLEANFILES = Makefile.in include $(top_srcdir)/build-aux/check.mk EXTRA_DIST = \ api-to-man-page-coverage # Avoid Automake warnings about overriding these user variables. # Programs in this directory are used during the build only. AUTOMAKE_OPTIONS = -Wno-gnu EXEEXT=$(BUILD_EXEEXT) CC=$(CC_FOR_BUILD) CFLAGS=$(CFLAGS_FOR_BUILD) CPPFLAGS=$(CPPFLAGS_FOR_BUILD) LDFLAGS=$(LDFLAGS_FOR_BUILD) if BUILD_MAN if BUILD_DOXYXML noinst_PROGRAMS = doxyxml noinst_HEADERS = cstring.h doxyxml_SOURCES = doxyxml.c cstring.c doxyxml_CFLAGS = $(AM_CFLAGS) $(libqb_BUILD_CFLAGS) $(libxml_BUILD_CFLAGS) doxyxml_LDADD = $(libqb_BUILD_LIBS) $(libxml_BUILD_LIBS) endif knet_man3_MANS = \ knet_addrtostr.3 \ knet_handle_add_datafd.3 \ knet_handle_clear_stats.3 \ knet_handle_compress.3 \ knet_handle_crypto.3 \ knet_handle_enable_filter.3 \ knet_handle_enable_pmtud_notify.3 \ knet_handle_enable_sock_notify.3 \ knet_handle_free.3 \ knet_handle_get_channel.3 \ knet_get_compress_list.3 \ knet_get_crypto_list.3 \ knet_handle_get_datafd.3 \ knet_handle_get_stats.3 \ knet_get_transport_id_by_name.3 \ knet_get_transport_list.3 \ knet_get_transport_name_by_id.3 \ knet_handle_get_transport_reconnect_interval.3 \ knet_handle_new.3 \ knet_handle_new_ex.3 \ knet_handle_pmtud_get.3 \ knet_handle_pmtud_set.3 \ knet_handle_pmtud_getfreq.3 \ knet_handle_pmtud_setfreq.3 \ knet_handle_remove_datafd.3 \ knet_handle_setfwd.3 \ knet_handle_set_transport_reconnect_interval.3 \ knet_host_add.3 \ knet_host_enable_status_change_notify.3 \ knet_host_get_host_list.3 \ knet_host_get_id_by_host_name.3 \ knet_host_get_name_by_host_id.3 \ knet_host_get_policy.3 \ knet_host_get_status.3 \ knet_host_remove.3 \ knet_host_set_name.3 \ knet_host_set_policy.3 \ knet_link_clear_config.3 \ knet_link_get_config.3 \ knet_link_get_enable.3 \ knet_link_get_link_list.3 \ knet_link_get_ping_timers.3 \ knet_link_get_pong_count.3 \ knet_link_get_priority.3 \ knet_link_get_status.3 \ knet_link_set_config.3 \ knet_link_set_enable.3 \ knet_link_set_ping_timers.3 \ knet_link_set_pong_count.3 \ knet_link_set_priority.3 \ knet_log_get_loglevel.3 \ knet_log_get_loglevel_id.3 \ knet_log_get_loglevel_name.3 \ knet_log_get_subsystem_id.3 \ knet_log_get_subsystem_name.3 \ knet_log_set_loglevel.3 \ knet_recv.3 \ knet_send.3 \ knet_send_sync.3 \ knet_strtoaddr.3 \ knet_handle_enable_access_lists.3 \ knet_link_add_acl.3 \ knet_link_insert_acl.3 \ knet_link_rm_acl.3 \ knet_link_clear_acl.3 \ knet_handle_crypto_set_config.3 \ knet_handle_crypto_use_config.3 \ - knet_handle_crypto_rx_clear_traffic.3 + knet_handle_crypto_rx_clear_traffic.3 \ + knet_handle_setprio_dscp.3 if BUILD_LIBNOZZLE nozzle_man3_MANS = \ nozzle_add_ip.3 \ nozzle_close.3 \ nozzle_del_ip.3 \ nozzle_get_fd.3 \ nozzle_get_handle_by_name.3 \ nozzle_get_ips.3 \ nozzle_get_mac.3 \ nozzle_get_mtu.3 \ nozzle_get_name_by_handle.3 \ nozzle_open.3 \ nozzle_reset_mac.3 \ nozzle_reset_mtu.3 \ nozzle_run_updown.3 \ nozzle_set_down.3 \ nozzle_set_mac.3 \ nozzle_set_mtu.3 \ nozzle_set_up.3 endif man3_MANS = $(knet_man3_MANS) $(nozzle_man3_MANS) $(MANS): doxyfile-knet.stamp doxyfile-nozzle.stamp doxyfile-knet.stamp: $(noinst_PROGRAMS) Doxyfile-knet $(top_srcdir)/libknet/libknet.h $(DOXYGEN) Doxyfile-knet 2>&1 | $(EGREP) -v 'warning.*macro definition' $(DOXYGEN2MAN) -m -P -o $(builddir) -s 3 -p @PACKAGE_NAME@ -H "Kronosnet Programmer's Manual" \ $$($(UTC_DATE_AT)$(SOURCE_EPOCH) +"-D %F -Y %Y") -d $(builddir)/xml-knet/ libknet_8h.xml touch doxyfile-knet.stamp doxyfile-nozzle.stamp: $(noinst_PROGRAMS) Doxyfile-nozzle $(top_srcdir)/libnozzle/libnozzle.h if BUILD_LIBNOZZLE $(DOXYGEN) Doxyfile-nozzle 2>&1 | $(EGREP) -v 'warning.*macro definition' $(DOXYGEN2MAN) -m -P -o $(builddir) -s 3 -p @PACKAGE_NAME@ -H "Kronosnet Programmer's Manual" \ $$($(UTC_DATE_AT)$(SOURCE_EPOCH) +"-D %F -Y %Y") -d $(builddir)/xml-nozzle/ libnozzle_8h.xml endif touch doxyfile-nozzle.stamp noinst_SCRIPTS = api-to-man-page-coverage check-local: check-api-to-man-page-coverage-libknet check-api-to-man-page-coverage-libnozzle check-api-to-man-page-coverage-libnozzle: if BUILD_LIBNOZZLE $(srcdir)/api-to-man-page-coverage $(top_srcdir) nozzle endif check-api-to-man-page-coverage-libknet: $(srcdir)/api-to-man-page-coverage $(top_srcdir) knet endif clean-local: rm -rf doxyxml doxyfile*.stamp xml* *.3