diff options
Diffstat (limited to '')
41 files changed, 15913 insertions, 0 deletions
diff --git a/lib/isc/unix/Makefile.in b/lib/isc/unix/Makefile.in new file mode 100644 index 0000000..bdd7bfd --- /dev/null +++ b/lib/isc/unix/Makefile.in @@ -0,0 +1,44 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +CINCLUDES = -I${srcdir}/include \ + -I${srcdir}/../@ISC_THREAD_DIR@/include \ + -I../include \ + -I${srcdir}/../include \ + -I${srcdir}/.. @ISC_OPENSSL_INC@ + +CDEFINES = @CRYPTO@ +CWARNINGS = + +# Alphabetically +OBJS = @ISC_IPV6_O@ @ISC_PK11_API_O@ \ + app.@O@ dir.@O@ entropy.@O@ errno.@O@ errno2result.@O@ \ + file.@O@ fsaccess.@O@ interfaceiter.@O@ \ + keyboard.@O@ meminfo.@O@ \ + net.@O@ os.@O@ resource.@O@ socket.@O@ stdio.@O@ stdtime.@O@ \ + strerror.@O@ syslog.@O@ time.@O@ + +# Alphabetically +SRCS = @ISC_IPV6_C@ @ISC_PK11_API_C@ \ + app.c dir.c entropy.c errno.c errno2result.c \ + file.c fsaccess.c interfaceiter.c keyboard.c meminfo.c \ + net.c os.c resource.c socket.c stdio.c stdtime.c \ + strerror.c syslog.c time.c + +SUBDIRS = include +TARGETS = ${OBJS} + +@BIND9_MAKE_RULES@ + +interfaceiter.@O@: interfaceiter.c ifiter_ioctl.c ifiter_sysctl.c ifiter_getifaddrs.c + diff --git a/lib/isc/unix/app.c b/lib/isc/unix/app.c new file mode 100644 index 0000000..7e5a0ee --- /dev/null +++ b/lib/isc/unix/app.c @@ -0,0 +1,1038 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <config.h> + +#include <sys/param.h> /* Openserver 5.0.6A and FD_SETSIZE */ +#include <sys/types.h> + +#include <stddef.h> +#include <inttypes.h> +#include <stdbool.h> +#include <stdlib.h> +#include <errno.h> +#include <unistd.h> +#include <signal.h> +#include <sys/time.h> +#ifdef HAVE_EPOLL +#include <sys/epoll.h> +#endif + +#include <isc/app.h> +#include <isc/condition.h> +#include <isc/mem.h> +#include <isc/msgs.h> +#include <isc/mutex.h> +#include <isc/event.h> +#include <isc/platform.h> +#include <isc/strerror.h> +#include <isc/string.h> +#include <isc/task.h> +#include <isc/time.h> +#include <isc/util.h> + +#ifdef ISC_PLATFORM_USETHREADS +#include <pthread.h> +#endif + +/*% + * For BIND9 internal applications built with threads, we use a single app + * context and let multiple worker, I/O, timer threads do actual jobs. + * For other cases (including BIND9 built without threads) an app context acts + * as an event loop dispatching various events. + */ +#ifndef ISC_PLATFORM_USETHREADS +#include "../timer_p.h" +#include "../task_p.h" +#include "socket_p.h" +#endif /* ISC_PLATFORM_USETHREADS */ + +#ifdef ISC_PLATFORM_USETHREADS +static pthread_t blockedthread; +#endif /* ISC_PLATFORM_USETHREADS */ + +/*% + * The following are intended for internal use (indicated by "isc__" + * prefix) but are not declared as static, allowing direct access from + * unit tests etc. + */ +isc_result_t isc__app_start(void); +isc_result_t isc__app_ctxstart(isc_appctx_t *ctx); +isc_result_t isc__app_onrun(isc_mem_t *mctx, isc_task_t *task, + isc_taskaction_t action, void *arg); +isc_result_t isc__app_ctxrun(isc_appctx_t *ctx); +isc_result_t isc__app_run(void); +isc_result_t isc__app_ctxshutdown(isc_appctx_t *ctx); +isc_result_t isc__app_shutdown(void); +isc_result_t isc__app_reload(void); +isc_result_t isc__app_ctxsuspend(isc_appctx_t *ctx); +void isc__app_ctxfinish(isc_appctx_t *ctx); +void isc__app_finish(void); +void isc__app_block(void); +void isc__app_unblock(void); +isc_result_t isc__appctx_create(isc_mem_t *mctx, isc_appctx_t **ctxp); +void isc__appctx_destroy(isc_appctx_t **ctxp); +void isc__appctx_settaskmgr(isc_appctx_t *ctx, isc_taskmgr_t *taskmgr); +void isc__appctx_setsocketmgr(isc_appctx_t *ctx, isc_socketmgr_t *socketmgr); +void isc__appctx_settimermgr(isc_appctx_t *ctx, isc_timermgr_t *timermgr); +isc_result_t isc__app_ctxonrun(isc_appctx_t *ctx, isc_mem_t *mctx, + isc_task_t *task, isc_taskaction_t action, + void *arg); + +/* + * The application context of this module. This implementation actually + * doesn't use it. (This may change in the future). + */ +#define APPCTX_MAGIC ISC_MAGIC('A', 'p', 'c', 'x') +#define VALID_APPCTX(c) ISC_MAGIC_VALID(c, APPCTX_MAGIC) + +typedef struct isc__appctx { + isc_appctx_t common; + isc_mem_t *mctx; + isc_mutex_t lock; + isc_eventlist_t on_run; + bool shutdown_requested; + bool running; + + /*! + * We assume that 'want_shutdown' can be read and written atomically. + */ + bool want_shutdown; + /* + * We assume that 'want_reload' can be read and written atomically. + */ + bool want_reload; + + bool blocked; + + isc_taskmgr_t *taskmgr; + isc_socketmgr_t *socketmgr; + isc_timermgr_t *timermgr; +#ifdef ISC_PLATFORM_USETHREADS + isc_mutex_t readylock; + isc_condition_t ready; +#endif /* ISC_PLATFORM_USETHREADS */ +} isc__appctx_t; + +static isc__appctx_t isc_g_appctx; + +static struct { + isc_appmethods_t methods; + + /*% + * The following are defined just for avoiding unused static functions. + */ + void *run, *shutdown, *start, *reload, *finish, *block, *unblock; +} appmethods = { + { + isc__appctx_destroy, + isc__app_ctxstart, + isc__app_ctxrun, + isc__app_ctxsuspend, + isc__app_ctxshutdown, + isc__app_ctxfinish, + isc__appctx_settaskmgr, + isc__appctx_setsocketmgr, + isc__appctx_settimermgr, + isc__app_ctxonrun + }, + (void *)isc__app_run, + (void *)isc__app_shutdown, + (void *)isc__app_start, + (void *)isc__app_reload, + (void *)isc__app_finish, + (void *)isc__app_block, + (void *)isc__app_unblock +}; + +#ifdef HAVE_LINUXTHREADS +/*! + * Linux has sigwait(), but it appears to prevent signal handlers from + * running, even if they're not in the set being waited for. This makes + * it impossible to get the default actions for SIGILL, SIGSEGV, etc. + * Instead of messing with it, we just use sigsuspend() instead. + */ +#undef HAVE_SIGWAIT +/*! + * We need to remember which thread is the main thread... + */ +static pthread_t main_thread; +#endif + +#ifndef HAVE_SIGWAIT +static void +exit_action(int arg) { + UNUSED(arg); + isc_g_appctx.want_shutdown = true; +} + +static void +reload_action(int arg) { + UNUSED(arg); + isc_g_appctx.want_reload = true; +} +#endif + +static isc_result_t +handle_signal(int sig, void (*handler)(int)) { + struct sigaction sa; + char strbuf[ISC_STRERRORSIZE]; + + memset(&sa, 0, sizeof(sa)); + sa.sa_handler = handler; + + if (sigfillset(&sa.sa_mask) != 0 || + sigaction(sig, &sa, NULL) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_APP, + ISC_MSG_SIGNALSETUP, + "handle_signal() %d setup: %s"), + sig, strbuf); + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc__app_ctxstart(isc_appctx_t *ctx0) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + isc_result_t result; + int presult; + sigset_t sset; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(VALID_APPCTX(ctx)); + + /* + * Start an ISC library application. + */ + +#ifdef NEED_PTHREAD_INIT + /* + * BSDI 3.1 seg faults in pthread_sigmask() if we don't do this. + */ + presult = pthread_init(); + if (presult != 0) { + isc__strerror(presult, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_start() pthread_init: %s", strbuf); + return (ISC_R_UNEXPECTED); + } +#endif + +#ifdef ISC_PLATFORM_USETHREADS +#ifdef HAVE_LINUXTHREADS + main_thread = pthread_self(); +#endif /* HAVE_LINUXTHREADS */ + + result = isc_mutex_init(&ctx->readylock); + if (result != ISC_R_SUCCESS) + return (result); + + result = isc_condition_init(&ctx->ready); + if (result != ISC_R_SUCCESS) + goto cleanup_rlock; + + result = isc_mutex_init(&ctx->lock); + if (result != ISC_R_SUCCESS) + goto cleanup_rcond; +#else /* ISC_PLATFORM_USETHREADS */ + result = isc_mutex_init(&ctx->lock); + if (result != ISC_R_SUCCESS) + goto cleanup; +#endif /* ISC_PLATFORM_USETHREADS */ + + ISC_LIST_INIT(ctx->on_run); + + ctx->shutdown_requested = false; + ctx->running = false; + ctx->want_shutdown = false; + ctx->want_reload = false; + ctx->blocked = false; + +#ifndef HAVE_SIGWAIT + /* + * Install do-nothing handlers for SIGINT and SIGTERM. + * + * We install them now because BSDI 3.1 won't block + * the default actions, regardless of what we do with + * pthread_sigmask(). + */ + result = handle_signal(SIGINT, exit_action); + if (result != ISC_R_SUCCESS) + goto cleanup; + result = handle_signal(SIGTERM, exit_action); + if (result != ISC_R_SUCCESS) + goto cleanup; +#endif + + /* + * Always ignore SIGPIPE. + */ + result = handle_signal(SIGPIPE, SIG_IGN); + if (result != ISC_R_SUCCESS) + goto cleanup; + + /* + * On Solaris 2, delivery of a signal whose action is SIG_IGN + * will not cause sigwait() to return. We may have inherited + * unexpected actions for SIGHUP, SIGINT, and SIGTERM from our parent + * process (e.g, Solaris cron). Set an action of SIG_DFL to make + * sure sigwait() works as expected. Only do this for SIGTERM and + * SIGINT if we don't have sigwait(), since a different handler is + * installed above. + */ + result = handle_signal(SIGHUP, SIG_DFL); + if (result != ISC_R_SUCCESS) + goto cleanup; + +#ifdef HAVE_SIGWAIT + result = handle_signal(SIGTERM, SIG_DFL); + if (result != ISC_R_SUCCESS) + goto cleanup; + result = handle_signal(SIGINT, SIG_DFL); + if (result != ISC_R_SUCCESS) + goto cleanup; +#endif + +#ifdef ISC_PLATFORM_USETHREADS + /* + * Block SIGHUP, SIGINT, SIGTERM. + * + * If isc_app_start() is called from the main thread before any other + * threads have been created, then the pthread_sigmask() call below + * will result in all threads having SIGHUP, SIGINT and SIGTERM + * blocked by default, ensuring that only the thread that calls + * sigwait() for them will get those signals. + */ + if (sigemptyset(&sset) != 0 || + sigaddset(&sset, SIGHUP) != 0 || + sigaddset(&sset, SIGINT) != 0 || + sigaddset(&sset, SIGTERM) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_start() sigsetops: %s", strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup; + } + presult = pthread_sigmask(SIG_BLOCK, &sset, NULL); + if (presult != 0) { + isc__strerror(presult, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_start() pthread_sigmask: %s", + strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup; + } +#else /* ISC_PLATFORM_USETHREADS */ + /* + * Unblock SIGHUP, SIGINT, SIGTERM. + * + * If we're not using threads, we need to make sure that SIGHUP, + * SIGINT and SIGTERM are not inherited as blocked from the parent + * process. + */ + if (sigemptyset(&sset) != 0 || + sigaddset(&sset, SIGHUP) != 0 || + sigaddset(&sset, SIGINT) != 0 || + sigaddset(&sset, SIGTERM) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_start() sigsetops: %s", strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup; + } + presult = sigprocmask(SIG_UNBLOCK, &sset, NULL); + if (presult != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_start() sigprocmask: %s", strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup; + } +#endif /* ISC_PLATFORM_USETHREADS */ + + return (ISC_R_SUCCESS); + + cleanup: +#ifdef ISC_PLATFORM_USETHREADS + cleanup_rcond: + (void)isc_condition_destroy(&ctx->ready); + + cleanup_rlock: + (void)isc_mutex_destroy(&ctx->readylock); +#endif /* ISC_PLATFORM_USETHREADS */ + return (result); +} + +isc_result_t +isc__app_start(void) { + isc_g_appctx.common.impmagic = APPCTX_MAGIC; + isc_g_appctx.common.magic = ISCAPI_APPCTX_MAGIC; + isc_g_appctx.common.methods = &appmethods.methods; + isc_g_appctx.mctx = NULL; + /* The remaining members will be initialized in ctxstart() */ + + return (isc__app_ctxstart((isc_appctx_t *)&isc_g_appctx)); +} + +isc_result_t +isc__app_onrun(isc_mem_t *mctx, isc_task_t *task, isc_taskaction_t action, + void *arg) +{ + return (isc__app_ctxonrun((isc_appctx_t *)&isc_g_appctx, mctx, + task, action, arg)); +} + +isc_result_t +isc__app_ctxonrun(isc_appctx_t *ctx0, isc_mem_t *mctx, isc_task_t *task, + isc_taskaction_t action, void *arg) +{ + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + isc_event_t *event; + isc_task_t *cloned_task = NULL; + isc_result_t result; + + LOCK(&ctx->lock); + + if (ctx->running) { + result = ISC_R_ALREADYRUNNING; + goto unlock; + } + + /* + * Note that we store the task to which we're going to send the event + * in the event's "sender" field. + */ + isc_task_attach(task, &cloned_task); + event = isc_event_allocate(mctx, cloned_task, ISC_APPEVENT_SHUTDOWN, + action, arg, sizeof(*event)); + if (event == NULL) { + isc_task_detach(&cloned_task); + result = ISC_R_NOMEMORY; + goto unlock; + } + + ISC_LIST_APPEND(ctx->on_run, event, ev_link); + + result = ISC_R_SUCCESS; + + unlock: + UNLOCK(&ctx->lock); + + return (result); +} + +#ifndef ISC_PLATFORM_USETHREADS +/*! + * Event loop for nonthreaded programs. + */ +static isc_result_t +evloop(isc__appctx_t *ctx) { + isc_result_t result; + + while (!ctx->want_shutdown) { + int n; + isc_time_t when, now; + struct timeval tv, *tvp; + isc_socketwait_t *swait; + bool readytasks; + bool call_timer_dispatch = false; + + /* + * Check the reload (or suspend) case first for exiting the + * loop as fast as possible in case: + * - the direct call to isc__taskmgr_dispatch() in + * isc__app_ctxrun() completes all the tasks so far, + * - there is thus currently no active task, and + * - there is a timer event + */ + if (ctx->want_reload) { + ctx->want_reload = false; + return (ISC_R_RELOAD); + } + + readytasks = isc__taskmgr_ready(ctx->taskmgr); + if (readytasks) { + tv.tv_sec = 0; + tv.tv_usec = 0; + tvp = &tv; + call_timer_dispatch = true; + } else { + result = isc__timermgr_nextevent(ctx->timermgr, &when); + if (result != ISC_R_SUCCESS) + tvp = NULL; + else { + uint64_t us; + + TIME_NOW(&now); + us = isc_time_microdiff(&when, &now); + if (us == 0) + call_timer_dispatch = true; + tv.tv_sec = us / 1000000; + tv.tv_usec = us % 1000000; + tvp = &tv; + } + } + + swait = NULL; + n = isc__socketmgr_waitevents(ctx->socketmgr, tvp, &swait); + + if (n == 0 || call_timer_dispatch) { + /* + * We call isc__timermgr_dispatch() only when + * necessary, in order to reduce overhead. If the + * select() call indicates a timeout, we need the + * dispatch. Even if not, if we set the 0-timeout + * for the select() call, we need to check the timer + * events. In the 'readytasks' case, there may be no + * timeout event actually, but there is no other way + * to reduce the overhead. + * Note that we do not have to worry about the case + * where a new timer is inserted during the select() + * call, since this loop only runs in the non-thread + * mode. + */ + isc__timermgr_dispatch(ctx->timermgr); + } + if (n > 0) + (void)isc__socketmgr_dispatch(ctx->socketmgr, swait); + (void)isc__taskmgr_dispatch(ctx->taskmgr); + } + return (ISC_R_SUCCESS); +} + +/* + * This is a gross hack to support waiting for condition + * variables in nonthreaded programs in a limited way; + * see lib/isc/nothreads/include/isc/condition.h. + * We implement isc_condition_wait() by entering the + * event loop recursively until the want_shutdown flag + * is set by isc_condition_signal(). + */ + +/*! + * \brief True if we are currently executing in the recursive + * event loop. + */ +static bool in_recursive_evloop = false; + +/*! + * \brief True if we are exiting the event loop as the result of + * a call to isc_condition_signal() rather than a shutdown + * or reload. + */ +static bool signalled = false; + +isc_result_t +isc__nothread_wait_hack(isc_condition_t *cp, isc_mutex_t *mp) { + isc_result_t result; + + UNUSED(cp); + UNUSED(mp); + + INSIST(!in_recursive_evloop); + in_recursive_evloop = true; + + INSIST(*mp == 1); /* Mutex must be locked on entry. */ + --*mp; + + result = evloop(&isc_g_appctx); + if (result == ISC_R_RELOAD) + isc_g_appctx.want_reload = true; + if (signalled) { + isc_g_appctx.want_shutdown = false; + signalled = false; + } + + ++*mp; + in_recursive_evloop = false; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc__nothread_signal_hack(isc_condition_t *cp) { + + UNUSED(cp); + + INSIST(in_recursive_evloop); + + isc_g_appctx.want_shutdown = true; + signalled = true; + return (ISC_R_SUCCESS); +} +#endif /* ISC_PLATFORM_USETHREADS */ + +isc_result_t +isc__app_ctxrun(isc_appctx_t *ctx0) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + int result; + isc_event_t *event, *next_event; + isc_task_t *task; +#ifdef ISC_PLATFORM_USETHREADS + sigset_t sset; + char strbuf[ISC_STRERRORSIZE]; +#ifdef HAVE_SIGWAIT + int sig; +#endif /* HAVE_SIGWAIT */ +#endif /* ISC_PLATFORM_USETHREADS */ + + REQUIRE(VALID_APPCTX(ctx)); + +#ifdef HAVE_LINUXTHREADS + REQUIRE(main_thread == pthread_self()); +#endif + + LOCK(&ctx->lock); + + if (!ctx->running) { + ctx->running = true; + + /* + * Post any on-run events (in FIFO order). + */ + for (event = ISC_LIST_HEAD(ctx->on_run); + event != NULL; + event = next_event) { + next_event = ISC_LIST_NEXT(event, ev_link); + ISC_LIST_UNLINK(ctx->on_run, event, ev_link); + task = event->ev_sender; + event->ev_sender = NULL; + isc_task_sendanddetach(&task, &event); + } + + } + + UNLOCK(&ctx->lock); + +#ifndef ISC_PLATFORM_USETHREADS + if (isc_bind9 && ctx == &isc_g_appctx) { + result = handle_signal(SIGHUP, reload_action); + if (result != ISC_R_SUCCESS) + return (ISC_R_SUCCESS); + } + + (void) isc__taskmgr_dispatch(ctx->taskmgr); + result = evloop(ctx); + return (result); +#else /* ISC_PLATFORM_USETHREADS */ + /* + * BIND9 internal tools using multiple contexts do not + * rely on signal. + */ + if (isc_bind9 && ctx != &isc_g_appctx) + return (ISC_R_SUCCESS); + + /* + * There is no danger if isc_app_shutdown() is called before we + * wait for signals. Signals are blocked, so any such signal will + * simply be made pending and we will get it when we call + * sigwait(). + */ + while (!ctx->want_shutdown) { +#ifdef HAVE_SIGWAIT + if (isc_bind9) { + /* + * BIND9 internal; single context: + * Wait for SIGHUP, SIGINT, or SIGTERM. + */ + if (sigemptyset(&sset) != 0 || + sigaddset(&sset, SIGHUP) != 0 || + sigaddset(&sset, SIGINT) != 0 || + sigaddset(&sset, SIGTERM) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_run() sigsetops: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + +#ifndef HAVE_UNIXWARE_SIGWAIT + result = sigwait(&sset, &sig); + if (result == 0) { + if (sig == SIGINT || sig == SIGTERM) + ctx->want_shutdown = true; + else if (sig == SIGHUP) + ctx->want_reload = true; + } + +#else /* Using UnixWare sigwait semantics. */ + sig = sigwait(&sset); + if (sig >= 0) { + if (sig == SIGINT || sig == SIGTERM) + ctx->want_shutdown = true; + else if (sig == SIGHUP) + ctx->want_reload = true; + } +#endif /* HAVE_UNIXWARE_SIGWAIT */ + } else { + /* + * External, or BIND9 using multiple contexts: + * wait until woken up. + */ + LOCK(&ctx->readylock); + if (ctx->want_shutdown) { + /* shutdown() won the race. */ + UNLOCK(&ctx->readylock); + break; + } + if (!ctx->want_reload) + WAIT(&ctx->ready, &ctx->readylock); + UNLOCK(&ctx->readylock); + } +#else /* Don't have sigwait(). */ + if (isc_bind9) { + /* + * BIND9 internal; single context: + * Install a signal handler for SIGHUP, then wait for + * all signals. + */ + result = handle_signal(SIGHUP, reload_action); + if (result != ISC_R_SUCCESS) + return (ISC_R_SUCCESS); + + if (sigemptyset(&sset) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_run() sigsetops: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } +#ifdef HAVE_GPERFTOOLS_PROFILER + if (sigaddset(&sset, SIGALRM) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_run() sigsetops: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } +#endif + (void)sigsuspend(&sset); + } else { + /* + * External, or BIND9 using multiple contexts: + * wait until woken up. + */ + LOCK(&ctx->readylock); + if (ctx->want_shutdown) { + /* shutdown() won the race. */ + UNLOCK(&ctx->readylock); + break; + } + if (!ctx->want_reload) + WAIT(&ctx->ready, &ctx->readylock); + UNLOCK(&ctx->readylock); + } +#endif /* HAVE_SIGWAIT */ + + if (ctx->want_reload) { + ctx->want_reload = false; + return (ISC_R_RELOAD); + } + + if (ctx->want_shutdown && ctx->blocked) + exit(1); + } + + return (ISC_R_SUCCESS); +#endif /* ISC_PLATFORM_USETHREADS */ +} + +isc_result_t +isc__app_run(void) { + return (isc__app_ctxrun((isc_appctx_t *)&isc_g_appctx)); +} + +isc_result_t +isc__app_ctxshutdown(isc_appctx_t *ctx0) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + bool want_kill = true; +#ifdef ISC_PLATFORM_USETHREADS + char strbuf[ISC_STRERRORSIZE]; +#endif /* ISC_PLATFORM_USETHREADS */ + + REQUIRE(VALID_APPCTX(ctx)); + + LOCK(&ctx->lock); + + REQUIRE(ctx->running); + + if (ctx->shutdown_requested) + want_kill = false; + else + ctx->shutdown_requested = true; + + UNLOCK(&ctx->lock); + + if (want_kill) { + if (isc_bind9 && ctx != &isc_g_appctx) + /* BIND9 internal, but using multiple contexts */ + ctx->want_shutdown = true; + else { +#ifndef ISC_PLATFORM_USETHREADS + ctx->want_shutdown = true; +#else /* ISC_PLATFORM_USETHREADS */ +#ifdef HAVE_LINUXTHREADS + if (isc_bind9) { + /* BIND9 internal, single context */ + int result; + + result = pthread_kill(main_thread, SIGTERM); + if (result != 0) { + isc__strerror(result, + strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_shutdown() " + "pthread_kill: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + } +#else + if (isc_bind9) { + /* BIND9 internal, single context */ + if (kill(getpid(), SIGTERM) < 0) { + isc__strerror(errno, + strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_shutdown() " + "kill: %s", strbuf); + return (ISC_R_UNEXPECTED); + } + } +#endif /* HAVE_LINUXTHREADS */ + else { + /* External, multiple contexts */ + LOCK(&ctx->readylock); + ctx->want_shutdown = true; + UNLOCK(&ctx->readylock); + SIGNAL(&ctx->ready); + } +#endif /* ISC_PLATFORM_USETHREADS */ + } + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc__app_shutdown(void) { + return (isc__app_ctxshutdown((isc_appctx_t *)&isc_g_appctx)); +} + +isc_result_t +isc__app_ctxsuspend(isc_appctx_t *ctx0) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + bool want_kill = true; +#ifdef ISC_PLATFORM_USETHREADS + char strbuf[ISC_STRERRORSIZE]; +#endif + + REQUIRE(VALID_APPCTX(ctx)); + + LOCK(&ctx->lock); + + REQUIRE(ctx->running); + + /* + * Don't send the reload signal if we're shutting down. + */ + if (ctx->shutdown_requested) + want_kill = false; + + UNLOCK(&ctx->lock); + + if (want_kill) { + if (isc_bind9 && ctx != &isc_g_appctx) + /* BIND9 internal, but using multiple contexts */ + ctx->want_reload = true; + else { +#ifndef ISC_PLATFORM_USETHREADS + ctx->want_reload = true; +#else /* ISC_PLATFORM_USETHREADS */ +#ifdef HAVE_LINUXTHREADS + if (isc_bind9) { + /* BIND9 internal, single context */ + int result; + + result = pthread_kill(main_thread, SIGHUP); + if (result != 0) { + isc__strerror(result, + strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_reload() " + "pthread_kill: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + } +#else + if (isc_bind9) { + /* BIND9 internal, single context */ + if (kill(getpid(), SIGHUP) < 0) { + isc__strerror(errno, + strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_app_reload() " + "kill: %s", strbuf); + return (ISC_R_UNEXPECTED); + } + } +#endif /* HAVE_LINUXTHREADS */ + else { + /* External, multiple contexts */ + LOCK(&ctx->readylock); + ctx->want_reload = true; + UNLOCK(&ctx->readylock); + SIGNAL(&ctx->ready); + } +#endif /* ISC_PLATFORM_USETHREADS */ + } + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc__app_reload(void) { + return (isc__app_ctxsuspend((isc_appctx_t *)&isc_g_appctx)); +} + +void +isc__app_ctxfinish(isc_appctx_t *ctx0) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + + REQUIRE(VALID_APPCTX(ctx)); + + DESTROYLOCK(&ctx->lock); +} + +void +isc__app_finish(void) { + isc__app_ctxfinish((isc_appctx_t *)&isc_g_appctx); +} + +void +isc__app_block(void) { +#ifdef ISC_PLATFORM_USETHREADS + sigset_t sset; +#endif /* ISC_PLATFORM_USETHREADS */ + REQUIRE(isc_g_appctx.running); + REQUIRE(!isc_g_appctx.blocked); + + isc_g_appctx.blocked = true; +#ifdef ISC_PLATFORM_USETHREADS + blockedthread = pthread_self(); + RUNTIME_CHECK(sigemptyset(&sset) == 0 && + sigaddset(&sset, SIGINT) == 0 && + sigaddset(&sset, SIGTERM) == 0); + RUNTIME_CHECK(pthread_sigmask(SIG_UNBLOCK, &sset, NULL) == 0); +#endif /* ISC_PLATFORM_USETHREADS */ +} + +void +isc__app_unblock(void) { +#ifdef ISC_PLATFORM_USETHREADS + sigset_t sset; +#endif /* ISC_PLATFORM_USETHREADS */ + + REQUIRE(isc_g_appctx.running); + REQUIRE(isc_g_appctx.blocked); + + isc_g_appctx.blocked = false; + +#ifdef ISC_PLATFORM_USETHREADS + REQUIRE(blockedthread == pthread_self()); + + RUNTIME_CHECK(sigemptyset(&sset) == 0 && + sigaddset(&sset, SIGINT) == 0 && + sigaddset(&sset, SIGTERM) == 0); + RUNTIME_CHECK(pthread_sigmask(SIG_BLOCK, &sset, NULL) == 0); +#endif /* ISC_PLATFORM_USETHREADS */ +} + +isc_result_t +isc__appctx_create(isc_mem_t *mctx, isc_appctx_t **ctxp) { + isc__appctx_t *ctx; + + REQUIRE(mctx != NULL); + REQUIRE(ctxp != NULL && *ctxp == NULL); + + ctx = isc_mem_get(mctx, sizeof(*ctx)); + if (ctx == NULL) + return (ISC_R_NOMEMORY); + + ctx->common.impmagic = APPCTX_MAGIC; + ctx->common.magic = ISCAPI_APPCTX_MAGIC; + ctx->common.methods = &appmethods.methods; + + ctx->mctx = NULL; + isc_mem_attach(mctx, &ctx->mctx); + + ctx->taskmgr = NULL; + ctx->socketmgr = NULL; + ctx->timermgr = NULL; + + *ctxp = (isc_appctx_t *)ctx; + + return (ISC_R_SUCCESS); +} + +void +isc__appctx_destroy(isc_appctx_t **ctxp) { + isc__appctx_t *ctx; + + REQUIRE(ctxp != NULL); + ctx = (isc__appctx_t *)*ctxp; + REQUIRE(VALID_APPCTX(ctx)); + + isc_mem_putanddetach(&ctx->mctx, ctx, sizeof(*ctx)); + + *ctxp = NULL; +} + +void +isc__appctx_settaskmgr(isc_appctx_t *ctx0, isc_taskmgr_t *taskmgr) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + + REQUIRE(VALID_APPCTX(ctx)); + + ctx->taskmgr = taskmgr; +} + +void +isc__appctx_setsocketmgr(isc_appctx_t *ctx0, isc_socketmgr_t *socketmgr) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + + REQUIRE(VALID_APPCTX(ctx)); + + ctx->socketmgr = socketmgr; +} + +void +isc__appctx_settimermgr(isc_appctx_t *ctx0, isc_timermgr_t *timermgr) { + isc__appctx_t *ctx = (isc__appctx_t *)ctx0; + + REQUIRE(VALID_APPCTX(ctx)); + + ctx->timermgr = timermgr; +} + +isc_result_t +isc__app_register(void) { + return (isc_app_register(isc__appctx_create)); +} + +#include "../app_api.c" diff --git a/lib/isc/unix/dir.c b/lib/isc/unix/dir.c new file mode 100644 index 0000000..dae0e66 --- /dev/null +++ b/lib/isc/unix/dir.c @@ -0,0 +1,259 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <config.h> + +#include <sys/types.h> +#include <sys/stat.h> + +#include <ctype.h> +#include <errno.h> +#include <unistd.h> + +#include <isc/dir.h> +#include <isc/magic.h> +#include <isc/netdb.h> +#include <isc/print.h> +#include <isc/string.h> +#include <isc/util.h> + +#include "errno2result.h" + +#define ISC_DIR_MAGIC ISC_MAGIC('D', 'I', 'R', '*') +#define VALID_DIR(dir) ISC_MAGIC_VALID(dir, ISC_DIR_MAGIC) + +void +isc_dir_init(isc_dir_t *dir) { + REQUIRE(dir != NULL); + + dir->entry.name[0] = '\0'; + dir->entry.length = 0; + + dir->handle = NULL; + + dir->magic = ISC_DIR_MAGIC; +} + +/*! + * \brief Allocate workspace and open directory stream. If either one fails, + * NULL will be returned. + */ +isc_result_t +isc_dir_open(isc_dir_t *dir, const char *dirname) { + char *p; + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(VALID_DIR(dir)); + REQUIRE(dirname != NULL); + + /* + * Copy directory name. Need to have enough space for the name, + * a possible path separator, the wildcard, and the final NUL. + */ + if (strlen(dirname) + 3 > sizeof(dir->dirname)) { + /* XXXDCL ? */ + return (ISC_R_NOSPACE); + } + strlcpy(dir->dirname, dirname, sizeof(dir->dirname)); + + /* + * Append path separator, if needed, and "*". + */ + p = dir->dirname + strlen(dir->dirname); + if (dir->dirname < p && *(p - 1) != '/') + *p++ = '/'; + *p++ = '*'; + *p = '\0'; + + /* + * Open stream. + */ + dir->handle = opendir(dirname); + + if (dir->handle == NULL) { + return (isc__errno2result(errno)); + } + + return (result); +} + +/*! + * \brief Return previously retrieved file or get next one. + + * Unix's dirent has + * separate open and read functions, but the Win32 and DOS interfaces open + * the dir stream and reads the first file in one operation. + */ +isc_result_t +isc_dir_read(isc_dir_t *dir) { + struct dirent *entry; + + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + /* + * Fetch next file in directory. + */ + entry = readdir(dir->handle); + + if (entry == NULL) + return (ISC_R_NOMORE); + + /* + * Make sure that the space for the name is long enough. + */ + if (sizeof(dir->entry.name) <= strlen(entry->d_name)) + return (ISC_R_UNEXPECTED); + + strlcpy(dir->entry.name, entry->d_name, sizeof(dir->entry.name)); + + /* + * Some dirents have d_namlen, but it is not portable. + */ + dir->entry.length = strlen(entry->d_name); + + return (ISC_R_SUCCESS); +} + +/*! + * \brief Close directory stream. + */ +void +isc_dir_close(isc_dir_t *dir) { + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + (void)closedir(dir->handle); + dir->handle = NULL; +} + +/*! + * \brief Reposition directory stream at start. + */ +isc_result_t +isc_dir_reset(isc_dir_t *dir) { + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + rewinddir(dir->handle); + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_dir_chdir(const char *dirname) { + /*! + * \brief Change the current directory to 'dirname'. + */ + + REQUIRE(dirname != NULL); + + if (chdir(dirname) < 0) + return (isc__errno2result(errno)); + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_dir_chroot(const char *dirname) { +#ifdef HAVE_CHROOT + void *tmp; +#endif + + REQUIRE(dirname != NULL); + +#ifdef HAVE_CHROOT + /* + * Try to use getservbyname and getprotobyname before chroot. + * If WKS records are used in a zone under chroot, Name Service Switch + * may fail to load library in chroot. + * Do not report errors if it fails, we do not need any result now. + */ + tmp = getprotobyname("udp"); + if (tmp != NULL) + (void) getservbyname("domain", "udp"); + + if (chroot(dirname) < 0 || chdir("/") < 0) + return (isc__errno2result(errno)); + + return (ISC_R_SUCCESS); +#else + return (ISC_R_NOTIMPLEMENTED); +#endif +} + +isc_result_t +isc_dir_createunique(char *templet) { + isc_result_t result; + char *x; + char *p; + int i; + int pid; + + REQUIRE(templet != NULL); + + /*! + * \brief mkdtemp is not portable, so this emulates it. + */ + + pid = getpid(); + + /* + * Replace trailing Xs with the process-id, zero-filled. + */ + for (x = templet + strlen(templet) - 1; *x == 'X' && x >= templet; + x--, pid /= 10) + *x = pid % 10 + '0'; + + x++; /* Set x to start of ex-Xs. */ + + do { + i = mkdir(templet, 0700); + if (i == 0 || errno != EEXIST) + break; + + /* + * The BSD algorithm. + */ + p = x; + while (*p != '\0') { + if (isdigit(*p & 0xff)) + *p = 'a'; + else if (*p != 'z') + ++*p; + else { + /* + * Reset character and move to next. + */ + *p++ = 'a'; + continue; + } + + break; + } + + if (*p == '\0') { + /* + * Tried all combinations. errno should already + * be EEXIST, but ensure it is anyway for + * isc__errno2result(). + */ + errno = EEXIST; + break; + } + } while (1); + + if (i == -1) + result = isc__errno2result(errno); + else + result = ISC_R_SUCCESS; + + return (result); +} diff --git a/lib/isc/unix/entropy.c b/lib/isc/unix/entropy.c new file mode 100644 index 0000000..18eb938 --- /dev/null +++ b/lib/isc/unix/entropy.c @@ -0,0 +1,603 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/* \file unix/entropy.c + * \brief + * This is the system dependent part of the ISC entropy API. + */ + +#include <config.h> + +#include <stdbool.h> + +#include <sys/param.h> /* Openserver 5.0.6A and FD_SETSIZE */ +#include <sys/types.h> +#include <sys/time.h> +#include <sys/stat.h> +#include <sys/socket.h> +#include <sys/un.h> + +#ifdef HAVE_NANOSLEEP +#include <time.h> +#endif +#include <unistd.h> + +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/strerror.h> +#include <isc/string.h> + +#ifdef ISC_PLATFORM_NEEDSYSSELECTH +#include <sys/select.h> +#endif + +#include "errno2result.h" + +/*% + * There is only one variable in the entropy data structures that is not + * system independent, but pulling the structure that uses it into this file + * ultimately means pulling several other independent structures here also to + * resolve their interdependencies. Thus only the problem variable's type + * is defined here. + */ +#define FILESOURCE_HANDLE_TYPE int + +typedef struct { + int handle; + enum { + isc_usocketsource_disconnected, + isc_usocketsource_connecting, + isc_usocketsource_connected, + isc_usocketsource_ndesired, + isc_usocketsource_wrote, + isc_usocketsource_reading + } status; + size_t sz_to_recv; +} isc_entropyusocketsource_t; + +#include "../entropy.c" + +static unsigned int +get_from_filesource(isc_entropysource_t *source, uint32_t desired) { + isc_entropy_t *ent = source->ent; + unsigned char buf[128]; + int fd = source->sources.file.handle; + ssize_t n, ndesired; + unsigned int added; + + if (source->bad) + return (0); + + desired = desired / 8 + (((desired & 0x07) > 0) ? 1 : 0); + + added = 0; + while (desired > 0) { + ndesired = ISC_MIN(desired, sizeof(buf)); + n = read(fd, buf, ndesired); + if (n < 0) { + if (errno == EAGAIN || errno == EINTR) + goto out; + goto err; + } + if (n == 0) + goto err; + + entropypool_adddata(ent, buf, n, n * 8); + added += n * 8; + desired -= n; + } + goto out; + + err: + (void)close(fd); + source->sources.file.handle = -1; + source->bad = true; + + out: + return (added); +} + +static unsigned int +get_from_usocketsource(isc_entropysource_t *source, uint32_t desired) { + isc_entropy_t *ent = source->ent; + unsigned char buf[128]; + int fd = source->sources.usocket.handle; + ssize_t n = 0, ndesired; + unsigned int added; + size_t sz_to_recv = source->sources.usocket.sz_to_recv; + + if (source->bad) + return (0); + + desired = desired / 8 + (((desired & 0x07) > 0) ? 1 : 0); + + added = 0; + while (desired > 0) { + ndesired = ISC_MIN(desired, sizeof(buf)); + eagain_loop: + + switch ( source->sources.usocket.status ) { + case isc_usocketsource_ndesired: + buf[0] = ndesired; + if ((n = sendto(fd, buf, 1, 0, NULL, 0)) < 0) { + if (errno == EWOULDBLOCK || errno == EINTR || + errno == ECONNRESET) + goto out; + goto err; + } + INSIST(n == 1); + source->sources.usocket.status = + isc_usocketsource_wrote; + goto eagain_loop; + + case isc_usocketsource_connecting: + case isc_usocketsource_connected: + buf[0] = 1; + buf[1] = ndesired; + if ((n = sendto(fd, buf, 2, 0, NULL, 0)) < 0) { + if (errno == EWOULDBLOCK || errno == EINTR || + errno == ECONNRESET) + goto out; + goto err; + } + if (n == 1) { + source->sources.usocket.status = + isc_usocketsource_ndesired; + goto eagain_loop; + } + INSIST(n == 2); + source->sources.usocket.status = + isc_usocketsource_wrote; + /* FALLTHROUGH */ + + case isc_usocketsource_wrote: + if (recvfrom(fd, buf, 1, 0, NULL, NULL) != 1) { + if (errno == EAGAIN) { + /* + * The problem of EAGAIN (try again + * later) is a major issue on HP-UX. + * Solaris actually tries the recvfrom + * call again, while HP-UX just dies. + * This code is an attempt to let the + * entropy pool fill back up (at least + * that's what I think the problem is.) + * We go to eagain_loop because if we + * just "break", then the "desired" + * amount gets borked. + */ +#ifdef HAVE_NANOSLEEP + struct timespec ts; + + ts.tv_sec = 0; + ts.tv_nsec = 1000000; + nanosleep(&ts, NULL); +#else + usleep(1000); +#endif + goto eagain_loop; + } + if (errno == EWOULDBLOCK || errno == EINTR) + goto out; + goto err; + } + source->sources.usocket.status = + isc_usocketsource_reading; + sz_to_recv = buf[0]; + source->sources.usocket.sz_to_recv = sz_to_recv; + if (sz_to_recv > sizeof(buf)) + goto err; + /* FALLTHROUGH */ + + case isc_usocketsource_reading: + if (sz_to_recv != 0U) { + n = recv(fd, buf, sz_to_recv, 0); + if (n < 0) { + if (errno == EWOULDBLOCK || + errno == EINTR) + goto out; + goto err; + } + } else + n = 0; + break; + + default: + goto err; + } + + if ((size_t)n != sz_to_recv) + source->sources.usocket.sz_to_recv -= n; + else + source->sources.usocket.status = + isc_usocketsource_connected; + + if (n == 0) + goto out; + + entropypool_adddata(ent, buf, n, n * 8); + added += n * 8; + desired -= n; + } + goto out; + + err: + close(fd); + source->bad = true; + source->sources.usocket.status = isc_usocketsource_disconnected; + source->sources.usocket.handle = -1; + + out: + return (added); +} + +/* + * Poll each source, trying to get data from it to stuff into the entropy + * pool. + */ +static void +fillpool(isc_entropy_t *ent, unsigned int desired, bool blocking) { + unsigned int added; + unsigned int remaining; + unsigned int needed; + unsigned int nsource; + isc_entropysource_t *source; + + REQUIRE(VALID_ENTROPY(ent)); + + needed = desired; + + /* + * This logic is a little strange, so an explanation is in order. + * + * If needed is 0, it means we are being asked to "fill to whatever + * we think is best." This means that if we have at least a + * partially full pool (say, > 1/4th of the pool) we probably don't + * need to add anything. + * + * Also, we will check to see if the "pseudo" count is too high. + * If it is, try to mix in better data. Too high is currently + * defined as 1/4th of the pool. + * + * Next, if we are asked to add a specific bit of entropy, make + * certain that we will do so. Clamp how much we try to add to + * (DIGEST_SIZE * 8 < needed < POOLBITS - entropy). + * + * Note that if we are in a blocking mode, we will only try to + * get as much data as we need, not as much as we might want + * to build up. + */ + if (needed == 0) { + REQUIRE(!blocking); + + if ((ent->pool.entropy >= RND_POOLBITS / 4) + && (ent->pool.pseudo <= RND_POOLBITS / 4)) + return; + + needed = THRESHOLD_BITS * 4; + } else { + needed = ISC_MAX(needed, THRESHOLD_BITS); + needed = ISC_MIN(needed, RND_POOLBITS); + } + + /* + * In any case, clamp how much we need to how much we can add. + */ + needed = ISC_MIN(needed, RND_POOLBITS - ent->pool.entropy); + + /* + * But wait! If we're not yet initialized, we need at least + * THRESHOLD_BITS + * of randomness. + */ + if (ent->initialized < THRESHOLD_BITS) + needed = ISC_MAX(needed, THRESHOLD_BITS - ent->initialized); + + /* + * Poll each file source to see if we can read anything useful from + * it. XXXMLG When where are multiple sources, we should keep a + * record of which one we last used so we can start from it (or the + * next one) to avoid letting some sources build up entropy while + * others are always drained. + */ + + added = 0; + remaining = needed; + if (ent->nextsource == NULL) { + ent->nextsource = ISC_LIST_HEAD(ent->sources); + if (ent->nextsource == NULL) + return; + } + source = ent->nextsource; + again_file: + for (nsource = 0; nsource < ent->nsources; nsource++) { + unsigned int got; + + if (remaining == 0) + break; + + got = 0; + + switch ( source->type ) { + case ENTROPY_SOURCETYPE_FILE: + got = get_from_filesource(source, remaining); + break; + + case ENTROPY_SOURCETYPE_USOCKET: + got = get_from_usocketsource(source, remaining); + break; + } + + added += got; + + remaining -= ISC_MIN(remaining, got); + + source = ISC_LIST_NEXT(source, link); + if (source == NULL) + source = ISC_LIST_HEAD(ent->sources); + } + ent->nextsource = source; + + if (blocking && remaining != 0) { + int fds; + + fds = wait_for_sources(ent); + if (fds > 0) + goto again_file; + } + + /* + * Here, if there are bits remaining to be had and we can block, + * check to see if we have a callback source. If so, call them. + */ + source = ISC_LIST_HEAD(ent->sources); + while ((remaining != 0) && (source != NULL)) { + unsigned int got; + + got = 0; + + if (source->type == ENTROPY_SOURCETYPE_CALLBACK) + got = get_from_callback(source, remaining, blocking); + + added += got; + remaining -= ISC_MIN(remaining, got); + + if (added >= needed) + break; + + source = ISC_LIST_NEXT(source, link); + } + + /* + * Mark as initialized if we've added enough data. + */ + if (ent->initialized < THRESHOLD_BITS) + ent->initialized += added; +} + +static int +wait_for_sources(isc_entropy_t *ent) { + isc_entropysource_t *source; + int maxfd, fd; + int cc; + fd_set reads; + fd_set writes; + + maxfd = -1; + FD_ZERO(&reads); + FD_ZERO(&writes); + + source = ISC_LIST_HEAD(ent->sources); + while (source != NULL) { + if (source->type == ENTROPY_SOURCETYPE_FILE) { + fd = source->sources.file.handle; + if (fd >= 0) { + maxfd = ISC_MAX(maxfd, fd); + FD_SET(fd, &reads); + } + } + if (source->type == ENTROPY_SOURCETYPE_USOCKET) { + fd = source->sources.usocket.handle; + if (fd >= 0) { + switch (source->sources.usocket.status) { + case isc_usocketsource_disconnected: + break; + case isc_usocketsource_connecting: + case isc_usocketsource_connected: + case isc_usocketsource_ndesired: + maxfd = ISC_MAX(maxfd, fd); + FD_SET(fd, &writes); + break; + case isc_usocketsource_wrote: + case isc_usocketsource_reading: + maxfd = ISC_MAX(maxfd, fd); + FD_SET(fd, &reads); + break; + } + } + } + source = ISC_LIST_NEXT(source, link); + } + + if (maxfd < 0) + return (-1); + + cc = select(maxfd + 1, &reads, &writes, NULL, NULL); + if (cc < 0) + return (-1); + + return (cc); +} + +static void +destroyfilesource(isc_entropyfilesource_t *source) { + (void)close(source->handle); +} + +static void +destroyusocketsource(isc_entropyusocketsource_t *source) { + close(source->handle); +} + +/* + * Make a fd non-blocking + */ +static isc_result_t +make_nonblock(int fd) { + int ret; + char strbuf[ISC_STRERRORSIZE]; +#ifdef USE_FIONBIO_IOCTL + int on = 1; +#else + int flags; +#endif + +#ifdef USE_FIONBIO_IOCTL + ret = ioctl(fd, FIONBIO, (char *)&on); +#else + flags = fcntl(fd, F_GETFL, 0); + flags |= PORT_NONBLOCK; + ret = fcntl(fd, F_SETFL, flags); +#endif + + if (ret == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, +#ifdef USE_FIONBIO_IOCTL + "ioctl(%d, FIONBIO, &on): %s", fd, +#else + "fcntl(%d, F_SETFL, %d): %s", fd, flags, +#endif + strbuf); + + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_entropy_createfilesource(isc_entropy_t *ent, const char *fname) { + int fd; + struct stat _stat; + bool is_usocket = false; + bool is_connected = false; + isc_result_t ret; + isc_entropysource_t *source; + + REQUIRE(VALID_ENTROPY(ent)); + REQUIRE(fname != NULL); + + LOCK(&ent->lock); + + if (stat(fname, &_stat) < 0) { + ret = isc__errno2result(errno); + goto errout; + } + /* + * Solaris 2.5.1 does not have support for sockets (S_IFSOCK), + * but it does return type S_IFIFO (the OS believes that + * the socket is a fifo). This may be an issue if we tell + * the program to look at an actual FIFO as its source of + * entropy. + */ +#if defined(S_ISSOCK) + if (S_ISSOCK(_stat.st_mode)) + is_usocket = true; +#endif +#if defined(S_ISFIFO) && defined(sun) + if (S_ISFIFO(_stat.st_mode)) + is_usocket = true; +#endif + if (is_usocket) + fd = socket(PF_UNIX, SOCK_STREAM, 0); + else + fd = open(fname, O_RDONLY | PORT_NONBLOCK, 0); + + if (fd < 0) { + ret = isc__errno2result(errno); + goto errout; + } + + ret = make_nonblock(fd); + if (ret != ISC_R_SUCCESS) + goto closefd; + + if (is_usocket) { + struct sockaddr_un sname; + + memset(&sname, 0, sizeof(sname)); + sname.sun_family = AF_UNIX; + strlcpy(sname.sun_path, fname, sizeof(sname.sun_path)); +#ifdef ISC_PLATFORM_HAVESALEN +#if !defined(SUN_LEN) +#define SUN_LEN(su) \ + (sizeof(*(su)) - sizeof((su)->sun_path) + strlen((su)->sun_path)) +#endif + sname.sun_len = SUN_LEN(&sname); +#endif + + if (connect(fd, (struct sockaddr *) &sname, + sizeof(struct sockaddr_un)) < 0) { + if (errno != EINPROGRESS) { + ret = isc__errno2result(errno); + goto closefd; + } + } else + is_connected = true; + } + + source = isc_mem_get(ent->mctx, sizeof(isc_entropysource_t)); + if (source == NULL) { + ret = ISC_R_NOMEMORY; + goto closefd; + } + + /* + * From here down, no failures can occur. + */ + source->magic = SOURCE_MAGIC; + source->ent = ent; + source->total = 0; + source->bad = false; + memset(source->name, 0, sizeof(source->name)); + ISC_LINK_INIT(source, link); + if (is_usocket) { + source->sources.usocket.handle = fd; + if (is_connected) + source->sources.usocket.status = + isc_usocketsource_connected; + else + source->sources.usocket.status = + isc_usocketsource_connecting; + source->sources.usocket.sz_to_recv = 0; + source->type = ENTROPY_SOURCETYPE_USOCKET; + } else { + source->sources.file.handle = fd; + source->type = ENTROPY_SOURCETYPE_FILE; + } + + /* + * Hook it into the entropy system. + */ + ISC_LIST_APPEND(ent->sources, source, link); + ent->nsources++; + + UNLOCK(&ent->lock); + return (ISC_R_SUCCESS); + + closefd: + (void)close(fd); + + errout: + UNLOCK(&ent->lock); + + return (ret); +} diff --git a/lib/isc/unix/errno.c b/lib/isc/unix/errno.c new file mode 100644 index 0000000..a2445f3 --- /dev/null +++ b/lib/isc/unix/errno.c @@ -0,0 +1,24 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <config.h> + +#include <isc/errno.h> +#include <isc/util.h> + +#include "errno2result.h" + +isc_result_t +isc_errno_toresult(int err) { + return (isc___errno2result(err, false, 0, 0)); +} diff --git a/lib/isc/unix/errno2result.c b/lib/isc/unix/errno2result.c new file mode 100644 index 0000000..2f12bcc --- /dev/null +++ b/lib/isc/unix/errno2result.c @@ -0,0 +1,125 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <stdbool.h> + +#include <isc/result.h> +#include <isc/strerror.h> +#include <isc/util.h> + +#include "errno2result.h" + +/*% + * Convert a POSIX errno value into an isc_result_t. The + * list of supported errno values is not complete; new users + * of this function should add any expected errors that are + * not already there. + */ +isc_result_t +isc___errno2result(int posixerrno, bool dolog, + const char *file, unsigned int line) +{ + char strbuf[ISC_STRERRORSIZE]; + + switch (posixerrno) { + case ENOTDIR: + case ELOOP: + case EINVAL: /* XXX sometimes this is not for files */ + case ENAMETOOLONG: + case EBADF: + return (ISC_R_INVALIDFILE); + case ENOENT: + return (ISC_R_FILENOTFOUND); + case EACCES: + case EPERM: + return (ISC_R_NOPERM); + case EEXIST: + return (ISC_R_FILEEXISTS); + case EIO: + return (ISC_R_IOERROR); + case ENOMEM: + return (ISC_R_NOMEMORY); + case ENFILE: + case EMFILE: + return (ISC_R_TOOMANYOPENFILES); +#ifdef EOVERFLOW + case EOVERFLOW: + return (ISC_R_RANGE); +#endif + case EPIPE: +#ifdef ECONNRESET + case ECONNRESET: +#endif +#ifdef ECONNABORTED + case ECONNABORTED: +#endif + return (ISC_R_CONNECTIONRESET); +#ifdef ENOTCONN + case ENOTCONN: + return (ISC_R_NOTCONNECTED); +#endif +#ifdef ETIMEDOUT + case ETIMEDOUT: + return (ISC_R_TIMEDOUT); +#endif +#ifdef ENOBUFS + case ENOBUFS: + return (ISC_R_NORESOURCES); +#endif +#ifdef EAFNOSUPPORT + case EAFNOSUPPORT: + return (ISC_R_FAMILYNOSUPPORT); +#endif +#ifdef ENETDOWN + case ENETDOWN: + return (ISC_R_NETDOWN); +#endif +#ifdef EHOSTDOWN + case EHOSTDOWN: + return (ISC_R_HOSTDOWN); +#endif +#ifdef ENETUNREACH + case ENETUNREACH: + return (ISC_R_NETUNREACH); +#endif +#ifdef EHOSTUNREACH + case EHOSTUNREACH: + return (ISC_R_HOSTUNREACH); +#endif +#ifdef EADDRINUSE + case EADDRINUSE: + return (ISC_R_ADDRINUSE); +#endif + case EADDRNOTAVAIL: + return (ISC_R_ADDRNOTAVAIL); + case ECONNREFUSED: + return (ISC_R_CONNREFUSED); + default: + if (dolog) { + isc__strerror(posixerrno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(file, line, "unable to convert errno " + "to isc_result: %d: %s", + posixerrno, strbuf); + } + /* + * XXXDCL would be nice if perhaps this function could + * return the system's error string, so the caller + * might have something more descriptive than "unexpected + * error" to log with. + */ + return (ISC_R_UNEXPECTED); + } +} diff --git a/lib/isc/unix/errno2result.h b/lib/isc/unix/errno2result.h new file mode 100644 index 0000000..7733ab3 --- /dev/null +++ b/lib/isc/unix/errno2result.h @@ -0,0 +1,35 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef UNIX_ERRNO2RESULT_H +#define UNIX_ERRNO2RESULT_H 1 + +/*! \file */ + +/* XXXDCL this should be moved to lib/isc/include/isc/errno2result.h. */ + +#include <errno.h> /* Provides errno. */ +#include <stdbool.h> + +#include <isc/lang.h> +#include <isc/types.h> + +ISC_LANG_BEGINDECLS + +#define isc__errno2result(x) isc___errno2result(x, true, __FILE__, __LINE__) + +isc_result_t +isc___errno2result(int posixerrno, bool dolog, + const char *file, unsigned int line); + +ISC_LANG_ENDDECLS + +#endif /* UNIX_ERRNO2RESULT_H */ diff --git a/lib/isc/unix/file.c b/lib/isc/unix/file.c new file mode 100644 index 0000000..64d9596 --- /dev/null +++ b/lib/isc/unix/file.c @@ -0,0 +1,781 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/* + * Portions Copyright (c) 1987, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + + +/*! \file */ + +#include <config.h> + +#include <errno.h> +#include <fcntl.h> +#include <limits.h> +#include <inttypes.h> +#include <stdbool.h> +#include <stdlib.h> +#include <time.h> /* Required for utimes on some platforms. */ +#include <unistd.h> /* Required for mkstemp on NetBSD. */ + + +#include <sys/stat.h> +#include <sys/time.h> + +#ifdef HAVE_SYS_MMAN_H +#include <sys/mman.h> +#endif + +#include <isc/dir.h> +#include <isc/file.h> +#include <isc/log.h> +#include <isc/mem.h> +#include <isc/print.h> +#include <isc/random.h> +#include <isc/sha2.h> +#include <isc/string.h> +#include <isc/time.h> +#include <isc/util.h> + +#include "errno2result.h" + +/* + * XXXDCL As the API for accessing file statistics undoubtedly gets expanded, + * it might be good to provide a mechanism that allows for the results + * of a previous stat() to be used again without having to do another stat, + * such as perl's mechanism of using "_" in place of a file name to indicate + * that the results of the last stat should be used. But then you get into + * annoying MP issues. BTW, Win32 has stat(). + */ +static isc_result_t +file_stats(const char *file, struct stat *stats) { + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(file != NULL); + REQUIRE(stats != NULL); + + if (stat(file, stats) != 0) + result = isc__errno2result(errno); + + return (result); +} + +static isc_result_t +fd_stats(int fd, struct stat *stats) { + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(stats != NULL); + + if (fstat(fd, stats) != 0) + result = isc__errno2result(errno); + + return (result); +} + +isc_result_t +isc_file_getsizefd(int fd, off_t *size) { + isc_result_t result; + struct stat stats; + + REQUIRE(size != NULL); + + result = fd_stats(fd, &stats); + + if (result == ISC_R_SUCCESS) + *size = stats.st_size; + + return (result); +} + +isc_result_t +isc_file_mode(const char *file, mode_t *modep) { + isc_result_t result; + struct stat stats; + + REQUIRE(modep != NULL); + + result = file_stats(file, &stats); + if (result == ISC_R_SUCCESS) + *modep = (stats.st_mode & 07777); + + return (result); +} + +isc_result_t +isc_file_getmodtime(const char *file, isc_time_t *modtime) { + isc_result_t result; + struct stat stats; + + REQUIRE(file != NULL); + REQUIRE(modtime != NULL); + + result = file_stats(file, &stats); + + if (result == ISC_R_SUCCESS) +#ifdef ISC_PLATFORM_HAVESTATNSEC + isc_time_set(modtime, stats.st_mtime, stats.st_mtim.tv_nsec); +#else + isc_time_set(modtime, stats.st_mtime, 0); +#endif + + return (result); +} + +isc_result_t +isc_file_getsize(const char *file, off_t *size) { + isc_result_t result; + struct stat stats; + + REQUIRE(file != NULL); + REQUIRE(size != NULL); + + result = file_stats(file, &stats); + + if (result == ISC_R_SUCCESS) + *size = stats.st_size; + + return (result); +} + +isc_result_t +isc_file_settime(const char *file, isc_time_t *when) { + struct timeval times[2]; + + REQUIRE(file != NULL && when != NULL); + + /* + * tv_sec is at least a 32 bit quantity on all platforms we're + * dealing with, but it is signed on most (all?) of them, + * so we need to make sure the high bit isn't set. This unfortunately + * loses when either: + * * tv_sec becomes a signed 64 bit integer but long is 32 bits + * and isc_time_seconds > LONG_MAX, or + * * isc_time_seconds is changed to be > 32 bits but long is 32 bits + * and isc_time_seconds has at least 33 significant bits. + */ + times[0].tv_sec = times[1].tv_sec = (long)isc_time_seconds(when); + + /* + * Here is the real check for the high bit being set. + */ + if ((times[0].tv_sec & + (1ULL << (sizeof(times[0].tv_sec) * CHAR_BIT - 1))) != 0) + return (ISC_R_RANGE); + + /* + * isc_time_nanoseconds guarantees a value that divided by 1000 will + * fit into the minimum possible size tv_usec field. Unfortunately, + * we don't know what that type is so can't cast directly ... but + * we can at least cast to signed so the IRIX compiler shuts up. + */ + times[0].tv_usec = times[1].tv_usec = + (int32_t)(isc_time_nanoseconds(when) / 1000); + + if (utimes(file, times) < 0) + return (isc__errno2result(errno)); + + return (ISC_R_SUCCESS); +} + +#undef TEMPLATE +#define TEMPLATE "tmp-XXXXXXXXXX" /*%< 14 characters. */ + +isc_result_t +isc_file_mktemplate(const char *path, char *buf, size_t buflen) { + return (isc_file_template(path, TEMPLATE, buf, buflen)); +} + +isc_result_t +isc_file_template(const char *path, const char *templet, char *buf, + size_t buflen) +{ + const char *s; + + REQUIRE(templet != NULL); + REQUIRE(buf != NULL); + + if (path == NULL) + path = ""; + + s = strrchr(templet, '/'); + if (s != NULL) + templet = s + 1; + + s = strrchr(path, '/'); + + if (s != NULL) { + size_t prefixlen = s - path + 1; + if ((prefixlen + strlen(templet) + 1) > buflen) + return (ISC_R_NOSPACE); + + /* Copy 'prefixlen' bytes and NUL terminate. */ + strlcpy(buf, path, ISC_MIN(prefixlen + 1, buflen)); + strlcat(buf, templet, buflen); + } else { + if ((strlen(templet) + 1) > buflen) + return (ISC_R_NOSPACE); + + strlcpy(buf, templet, buflen); + } + + return (ISC_R_SUCCESS); +} + +static const char alphnum[] = + "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; + +isc_result_t +isc_file_renameunique(const char *file, char *templet) { + char *x; + char *cp; + + REQUIRE(file != NULL); + REQUIRE(templet != NULL); + + cp = templet; + while (*cp != '\0') + cp++; + if (cp == templet) + return (ISC_R_FAILURE); + + x = cp--; + while (cp >= templet && *cp == 'X') { + uint32_t which; + + isc_random_get(&which); + *cp = alphnum[which % (sizeof(alphnum) - 1)]; + x = cp--; + } + while (link(file, templet) == -1) { + if (errno != EEXIST) + return (isc__errno2result(errno)); + for (cp = x;;) { + const char *t; + if (*cp == '\0') + return (ISC_R_FAILURE); + t = strchr(alphnum, *cp); + if (t == NULL || *++t == '\0') + *cp++ = alphnum[0]; + else { + *cp = *t; + break; + } + } + } + if (unlink(file) < 0) + if (errno != ENOENT) + return (isc__errno2result(errno)); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_openunique(char *templet, FILE **fp) { + int mode = S_IWUSR|S_IRUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_openuniqueprivate(char *templet, FILE **fp) { + int mode = S_IWUSR|S_IRUSR; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_openuniquemode(char *templet, int mode, FILE **fp) { + int fd; + FILE *f; + isc_result_t result = ISC_R_SUCCESS; + char *x; + char *cp; + + REQUIRE(templet != NULL); + REQUIRE(fp != NULL && *fp == NULL); + + cp = templet; + while (*cp != '\0') + cp++; + if (cp == templet) + return (ISC_R_FAILURE); + + x = cp--; + while (cp >= templet && *cp == 'X') { + uint32_t which; + + isc_random_get(&which); + *cp = alphnum[which % (sizeof(alphnum) - 1)]; + x = cp--; + } + + + while ((fd = open(templet, O_RDWR|O_CREAT|O_EXCL, mode)) == -1) { + if (errno != EEXIST) + return (isc__errno2result(errno)); + for (cp = x;;) { + char *t; + if (*cp == '\0') + return (ISC_R_FAILURE); + t = strchr(alphnum, *cp); + if (t == NULL || *++t == '\0') + *cp++ = alphnum[0]; + else { + *cp = *t; + break; + } + } + } + f = fdopen(fd, "w+"); + if (f == NULL) { + result = isc__errno2result(errno); + if (remove(templet) < 0) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_FILE, ISC_LOG_ERROR, + "remove '%s': failed", templet); + } + (void)close(fd); + } else + *fp = f; + + return (result); +} + +isc_result_t +isc_file_bopenunique(char *templet, FILE **fp) { + int mode = S_IWUSR|S_IRUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_bopenuniqueprivate(char *templet, FILE **fp) { + int mode = S_IWUSR|S_IRUSR; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_bopenuniquemode(char *templet, int mode, FILE **fp) { + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_remove(const char *filename) { + int r; + + REQUIRE(filename != NULL); + + r = unlink(filename); + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + +isc_result_t +isc_file_rename(const char *oldname, const char *newname) { + int r; + + REQUIRE(oldname != NULL); + REQUIRE(newname != NULL); + + r = rename(oldname, newname); + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + +bool +isc_file_exists(const char *pathname) { + struct stat stats; + + REQUIRE(pathname != NULL); + + return (file_stats(pathname, &stats) == ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isplainfile(const char *filename) { + /* + * This function returns success if filename is a plain file. + */ + struct stat filestat; + memset(&filestat,0,sizeof(struct stat)); + + if ((stat(filename, &filestat)) == -1) + return(isc__errno2result(errno)); + + if(! S_ISREG(filestat.st_mode)) + return(ISC_R_INVALIDFILE); + + return(ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isplainfilefd(int fd) { + /* + * This function returns success if filename is a plain file. + */ + struct stat filestat; + memset(&filestat,0,sizeof(struct stat)); + + if ((fstat(fd, &filestat)) == -1) + return(isc__errno2result(errno)); + + if(! S_ISREG(filestat.st_mode)) + return(ISC_R_INVALIDFILE); + + return(ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isdirectory(const char *filename) { + /* + * This function returns success if filename exists and is a + * directory. + */ + struct stat filestat; + memset(&filestat,0,sizeof(struct stat)); + + if ((stat(filename, &filestat)) == -1) + return(isc__errno2result(errno)); + + if(! S_ISDIR(filestat.st_mode)) + return(ISC_R_INVALIDFILE); + + return(ISC_R_SUCCESS); +} + + +bool +isc_file_isabsolute(const char *filename) { + REQUIRE(filename != NULL); + return (filename[0] == '/'); +} + +bool +isc_file_iscurrentdir(const char *filename) { + REQUIRE(filename != NULL); + return (filename[0] == '.' && filename[1] == '\0'); +} + +bool +isc_file_ischdiridempotent(const char *filename) { + REQUIRE(filename != NULL); + if (isc_file_isabsolute(filename)) + return (true); + if (isc_file_iscurrentdir(filename)) + return (true); + return (false); +} + +const char * +isc_file_basename(const char *filename) { + const char *s; + + REQUIRE(filename != NULL); + + s = strrchr(filename, '/'); + if (s == NULL) + return (filename); + + return (s + 1); +} + +isc_result_t +isc_file_progname(const char *filename, char *buf, size_t buflen) { + const char *base; + size_t len; + + REQUIRE(filename != NULL); + REQUIRE(buf != NULL); + + base = isc_file_basename(filename); + len = strlen(base) + 1; + + if (len > buflen) + return (ISC_R_NOSPACE); + memmove(buf, base, len); + + return (ISC_R_SUCCESS); +} + +/* + * Put the absolute name of the current directory into 'dirname', which is + * a buffer of at least 'length' characters. End the string with the + * appropriate path separator, such that the final product could be + * concatenated with a relative pathname to make a valid pathname string. + */ +static isc_result_t +dir_current(char *dirname, size_t length) { + char *cwd; + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(dirname != NULL); + REQUIRE(length > 0U); + + cwd = getcwd(dirname, length); + + if (cwd == NULL) { + if (errno == ERANGE) { + result = ISC_R_NOSPACE; + } else { + result = isc__errno2result(errno); + } + } else { + if (strlen(dirname) + 1 == length) { + result = ISC_R_NOSPACE; + } else if (dirname[1] != '\0') { + strlcat(dirname, "/", length); + } + } + + return (result); +} + +isc_result_t +isc_file_absolutepath(const char *filename, char *path, size_t pathlen) { + isc_result_t result; + result = dir_current(path, pathlen); + if (result != ISC_R_SUCCESS) + return (result); + if (strlen(path) + strlen(filename) + 1 > pathlen) + return (ISC_R_NOSPACE); + strlcat(path, filename, pathlen); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_truncate(const char *filename, isc_offset_t size) { + isc_result_t result = ISC_R_SUCCESS; + + if (truncate(filename, size) < 0) + result = isc__errno2result(errno); + return (result); +} + +isc_result_t +isc_file_safecreate(const char *filename, FILE **fp) { + isc_result_t result; + int flags; + struct stat sb; + FILE *f; + int fd; + + REQUIRE(filename != NULL); + REQUIRE(fp != NULL && *fp == NULL); + + result = file_stats(filename, &sb); + if (result == ISC_R_SUCCESS) { + if ((sb.st_mode & S_IFREG) == 0) + return (ISC_R_INVALIDFILE); + flags = O_WRONLY | O_TRUNC; + } else if (result == ISC_R_FILENOTFOUND) { + flags = O_WRONLY | O_CREAT | O_EXCL; + } else + return (result); + + fd = open(filename, flags, S_IRUSR | S_IWUSR); + if (fd == -1) + return (isc__errno2result(errno)); + + f = fdopen(fd, "w"); + if (f == NULL) { + result = isc__errno2result(errno); + close(fd); + return (result); + } + + *fp = f; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_splitpath(isc_mem_t *mctx, const char *path, char **dirname, + char const **bname) +{ + char *dir; + const char *file, *slash; + + if (path == NULL) + return (ISC_R_INVALIDFILE); + + slash = strrchr(path, '/'); + + if (slash == path) { + file = ++slash; + dir = isc_mem_strdup(mctx, "/"); + } else if (slash != NULL) { + file = ++slash; + dir = isc_mem_allocate(mctx, slash - path); + if (dir != NULL) + strlcpy(dir, path, slash - path); + } else { + file = path; + dir = isc_mem_strdup(mctx, "."); + } + + if (dir == NULL) + return (ISC_R_NOMEMORY); + + if (*file == '\0') { + isc_mem_free(mctx, dir); + return (ISC_R_INVALIDFILE); + } + + *dirname = dir; + *bname = file; + + return (ISC_R_SUCCESS); +} + +void * +isc_file_mmap(void *addr, size_t len, int prot, + int flags, int fd, off_t offset) +{ +#ifdef HAVE_MMAP + return (mmap(addr, len, prot, flags, fd, offset)); +#else + void *buf; + ssize_t ret; + off_t end; + + UNUSED(addr); + UNUSED(prot); + UNUSED(flags); + + end = lseek(fd, 0, SEEK_END); + lseek(fd, offset, SEEK_SET); + if (end - offset < (off_t) len) + len = end - offset; + + buf = malloc(len); + if (buf == NULL) + return (NULL); + + ret = read(fd, buf, len); + if (ret != (ssize_t) len) { + free(buf); + buf = NULL; + } + + return (buf); +#endif +} + +int +isc_file_munmap(void *addr, size_t len) { +#ifdef HAVE_MMAP + return (munmap(addr, len)); +#else + UNUSED(len); + + free(addr); + return (0); +#endif +} + +#define DISALLOW "\\/ABCDEFGHIJKLMNOPQRSTUVWXYZ" +#ifndef PATH_MAX +#define PATH_MAX 1024 +#endif + +isc_result_t +isc_file_sanitize(const char *dir, const char *base, const char *ext, + char *path, size_t length) +{ + char buf[PATH_MAX], hash[ISC_SHA256_DIGESTSTRINGLENGTH]; + size_t l = 0; + + REQUIRE(base != NULL); + REQUIRE(path != NULL); + + l = strlen(base) + 1; + + /* + * allow room for a full sha256 hash (64 chars + * plus null terminator) + */ + if (l < 65U) + l = 65; + + if (dir != NULL) + l += strlen(dir) + 1; + if (ext != NULL) + l += strlen(ext) + 1; + + if (l > length || l > (unsigned)PATH_MAX) + return (ISC_R_NOSPACE); + + /* Check whether the full-length SHA256 hash filename exists */ + isc_sha256_data((const void *) base, strlen(base), hash); + snprintf(buf, sizeof(buf), "%s%s%s%s%s", + dir != NULL ? dir : "", dir != NULL ? "/" : "", + hash, ext != NULL ? "." : "", ext != NULL ? ext : ""); + if (isc_file_exists(buf)) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + /* Check for a truncated SHA256 hash filename */ + hash[16] = '\0'; + snprintf(buf, sizeof(buf), "%s%s%s%s%s", + dir != NULL ? dir : "", dir != NULL ? "/" : "", + hash, ext != NULL ? "." : "", ext != NULL ? ext : ""); + if (isc_file_exists(buf)) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + /* + * If neither hash filename already exists, then we'll use + * the original base name if it has no disallowed characters, + * or the truncated hash name if it does. + */ + if (strpbrk(base, DISALLOW) != NULL) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + snprintf(buf, sizeof(buf), "%s%s%s%s%s", + dir != NULL ? dir : "", dir != NULL ? "/" : "", + base, ext != NULL ? "." : "", ext != NULL ? ext : ""); + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); +} + +bool +isc_file_isdirwritable(const char *path) { + return (access(path, W_OK|X_OK) == 0); +} diff --git a/lib/isc/unix/fsaccess.c b/lib/isc/unix/fsaccess.c new file mode 100644 index 0000000..e3cd0bf --- /dev/null +++ b/lib/isc/unix/fsaccess.c @@ -0,0 +1,87 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <sys/types.h> +#include <sys/stat.h> + +#include <stdbool.h> +#include <errno.h> + +#include "errno2result.h" + +/*! \file + * \brief + * The OS-independent part of the API is in lib/isc. + */ +#include "../fsaccess.c" + +isc_result_t +isc_fsaccess_set(const char *path, isc_fsaccess_t access) { + struct stat statb; + mode_t mode; + bool is_dir = false; + isc_fsaccess_t bits; + isc_result_t result; + + if (stat(path, &statb) != 0) + return (isc__errno2result(errno)); + + if ((statb.st_mode & S_IFDIR) != 0) + is_dir = true; + else if ((statb.st_mode & S_IFREG) == 0) + return (ISC_R_INVALIDFILE); + + result = check_bad_bits(access, is_dir); + if (result != ISC_R_SUCCESS) + return (result); + + /* + * Done with checking bad bits. Set mode_t. + */ + mode = 0; + +#define SET_AND_CLEAR1(modebit) \ + if ((access & bits) != 0) { \ + mode |= modebit; \ + access &= ~bits; \ + } +#define SET_AND_CLEAR(user, group, other) \ + SET_AND_CLEAR1(user); \ + bits <<= STEP; \ + SET_AND_CLEAR1(group); \ + bits <<= STEP; \ + SET_AND_CLEAR1(other); + + bits = ISC_FSACCESS_READ | ISC_FSACCESS_LISTDIRECTORY; + + SET_AND_CLEAR(S_IRUSR, S_IRGRP, S_IROTH); + + bits = ISC_FSACCESS_WRITE | + ISC_FSACCESS_CREATECHILD | + ISC_FSACCESS_DELETECHILD; + + SET_AND_CLEAR(S_IWUSR, S_IWGRP, S_IWOTH); + + bits = ISC_FSACCESS_EXECUTE | + ISC_FSACCESS_ACCESSCHILD; + + SET_AND_CLEAR(S_IXUSR, S_IXGRP, S_IXOTH); + + INSIST(access == 0); + + if (chmod(path, mode) < 0) + return (isc__errno2result(errno)); + + return (ISC_R_SUCCESS); +} diff --git a/lib/isc/unix/ifiter_getifaddrs.c b/lib/isc/unix/ifiter_getifaddrs.c new file mode 100644 index 0000000..de6f060 --- /dev/null +++ b/lib/isc/unix/ifiter_getifaddrs.c @@ -0,0 +1,228 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file + * \brief + * Obtain the list of network interfaces using the getifaddrs(3) library. + */ + +#include <stdbool.h> +#include <ifaddrs.h> + +/*% Iterator Magic */ +#define IFITER_MAGIC ISC_MAGIC('I', 'F', 'I', 'G') +/*% Valid Iterator */ +#define VALID_IFITER(t) ISC_MAGIC_VALID(t, IFITER_MAGIC) + +#ifdef __linux +static bool seenv6 = false; +#endif + +/*% Iterator structure */ +struct isc_interfaceiter { + unsigned int magic; /*%< Magic number. */ + isc_mem_t *mctx; + void *buf; /*%< (unused) */ + unsigned int bufsize; /*%< (always 0) */ + struct ifaddrs *ifaddrs; /*%< List of ifaddrs */ + struct ifaddrs *pos; /*%< Ptr to current ifaddr */ + isc_interface_t current; /*%< Current interface data. */ + isc_result_t result; /*%< Last result code. */ +#ifdef __linux + FILE * proc; + char entry[ISC_IF_INET6_SZ]; + isc_result_t valid; +#endif +}; + +isc_result_t +isc_interfaceiter_create(isc_mem_t *mctx, isc_interfaceiter_t **iterp) { + isc_interfaceiter_t *iter; + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(mctx != NULL); + REQUIRE(iterp != NULL); + REQUIRE(*iterp == NULL); + + iter = isc_mem_get(mctx, sizeof(*iter)); + if (iter == NULL) + return (ISC_R_NOMEMORY); + + iter->mctx = mctx; + iter->buf = NULL; + iter->bufsize = 0; + iter->ifaddrs = NULL; +#ifdef __linux + /* + * Only open "/proc/net/if_inet6" if we have never seen a IPv6 + * address returned by getifaddrs(). + */ + if (!seenv6) + iter->proc = fopen("/proc/net/if_inet6", "r"); + else + iter->proc = NULL; + iter->valid = ISC_R_FAILURE; +#endif + + if (getifaddrs(&iter->ifaddrs) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERGETIFADDRS, + ISC_MSG_GETIFADDRS, + "getting interface " + "addresses: getifaddrs: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto failure; + } + + /* + * A newly created iterator has an undefined position + * until isc_interfaceiter_first() is called. + */ + iter->pos = NULL; + iter->result = ISC_R_FAILURE; + + iter->magic = IFITER_MAGIC; + *iterp = iter; + return (ISC_R_SUCCESS); + + failure: +#ifdef __linux + if (iter->proc != NULL) + fclose(iter->proc); +#endif + if (iter->ifaddrs != NULL) /* just in case */ + freeifaddrs(iter->ifaddrs); + isc_mem_put(mctx, iter, sizeof(*iter)); + return (result); +} + +/* + * Get information about the current interface to iter->current. + * If successful, return ISC_R_SUCCESS. + * If the interface has an unsupported address family, + * return ISC_R_IGNORE. + */ + +static isc_result_t +internal_current(isc_interfaceiter_t *iter) { + struct ifaddrs *ifa; + int family; + unsigned int namelen; + + REQUIRE(VALID_IFITER(iter)); + + ifa = iter->pos; + +#ifdef __linux + if (iter->pos == NULL) + return (linux_if_inet6_current(iter)); +#endif + + INSIST(ifa != NULL); + INSIST(ifa->ifa_name != NULL); + + if (ifa->ifa_addr == NULL) + return (ISC_R_IGNORE); + + family = ifa->ifa_addr->sa_family; + if (family != AF_INET && family != AF_INET6) + return (ISC_R_IGNORE); + +#ifdef __linux + if (family == AF_INET6) + seenv6 = true; +#endif + + memset(&iter->current, 0, sizeof(iter->current)); + + namelen = strlen(ifa->ifa_name); + if (namelen > sizeof(iter->current.name) - 1) + namelen = sizeof(iter->current.name) - 1; + + memset(iter->current.name, 0, sizeof(iter->current.name)); + memmove(iter->current.name, ifa->ifa_name, namelen); + + iter->current.flags = 0; + + if ((ifa->ifa_flags & IFF_UP) != 0) + iter->current.flags |= INTERFACE_F_UP; + + if ((ifa->ifa_flags & IFF_POINTOPOINT) != 0) + iter->current.flags |= INTERFACE_F_POINTTOPOINT; + + if ((ifa->ifa_flags & IFF_LOOPBACK) != 0) + iter->current.flags |= INTERFACE_F_LOOPBACK; + + iter->current.af = family; + + get_addr(family, &iter->current.address, ifa->ifa_addr, ifa->ifa_name); + + if (ifa->ifa_netmask != NULL) + get_addr(family, &iter->current.netmask, ifa->ifa_netmask, + ifa->ifa_name); + + if (ifa->ifa_dstaddr != NULL && + (iter->current.flags & INTERFACE_F_POINTTOPOINT) != 0) + get_addr(family, &iter->current.dstaddress, ifa->ifa_dstaddr, + ifa->ifa_name); + + return (ISC_R_SUCCESS); +} + +/* + * Step the iterator to the next interface. Unlike + * isc_interfaceiter_next(), this may leave the iterator + * positioned on an interface that will ultimately + * be ignored. Return ISC_R_NOMORE if there are no more + * interfaces, otherwise ISC_R_SUCCESS. + */ +static isc_result_t +internal_next(isc_interfaceiter_t *iter) { + + if (iter->pos != NULL) + iter->pos = iter->pos->ifa_next; + if (iter->pos == NULL) { +#ifdef __linux + if (!seenv6) + return (linux_if_inet6_next(iter)); +#endif + return (ISC_R_NOMORE); + } + + return (ISC_R_SUCCESS); +} + +static void +internal_destroy(isc_interfaceiter_t *iter) { + +#ifdef __linux + if (iter->proc != NULL) + fclose(iter->proc); + iter->proc = NULL; +#endif + if (iter->ifaddrs) + freeifaddrs(iter->ifaddrs); + iter->ifaddrs = NULL; +} + +static +void internal_first(isc_interfaceiter_t *iter) { + +#ifdef __linux + linux_if_inet6_first(iter); +#endif + iter->pos = iter->ifaddrs; +} diff --git a/lib/isc/unix/ifiter_ioctl.c b/lib/isc/unix/ifiter_ioctl.c new file mode 100644 index 0000000..9702fae --- /dev/null +++ b/lib/isc/unix/ifiter_ioctl.c @@ -0,0 +1,931 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <config.h> + +#include <stdbool.h> + +#include <isc/print.h> + +/*! \file + * \brief + * Obtain the list of network interfaces using the SIOCGLIFCONF ioctl. + * See netintro(4). + */ + +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) +#ifdef ISC_PLATFORM_HAVEIF_LADDRCONF +#define lifc_len iflc_len +#define lifc_buf iflc_buf +#define lifc_req iflc_req +#define LIFCONF if_laddrconf +#else +#define ISC_HAVE_LIFC_FAMILY 1 +#define ISC_HAVE_LIFC_FLAGS 1 +#define LIFCONF lifconf +#endif + +#ifdef ISC_PLATFORM_HAVEIF_LADDRREQ +#define lifr_addr iflr_addr +#define lifr_name iflr_name +#define lifr_dstaddr iflr_dstaddr +#define lifr_flags iflr_flags +#define ss_family sa_family +#define LIFREQ if_laddrreq +#else +#define LIFREQ lifreq +#endif +#endif + +#define IFITER_MAGIC ISC_MAGIC('I', 'F', 'I', 'T') +#define VALID_IFITER(t) ISC_MAGIC_VALID(t, IFITER_MAGIC) + +struct isc_interfaceiter { + unsigned int magic; /* Magic number. */ + isc_mem_t *mctx; + int mode; + int socket; + struct ifconf ifc; + void *buf; /* Buffer for sysctl data. */ + unsigned int bufsize; /* Bytes allocated. */ + unsigned int pos; /* Current offset in + SIOCGIFCONF data */ +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + int socket6; + struct LIFCONF lifc; + void *buf6; /* Buffer for sysctl data. */ + unsigned int bufsize6; /* Bytes allocated. */ + unsigned int pos6; /* Current offset in + SIOCGLIFCONF data */ + isc_result_t result6; /* Last result code. */ + bool first6; +#endif +#ifdef HAVE_TRUCLUSTER + int clua_context; /* Cluster alias context */ + bool clua_done; + struct sockaddr clua_sa; +#endif +#ifdef __linux + FILE * proc; + char entry[ISC_IF_INET6_SZ]; + isc_result_t valid; +#endif + isc_interface_t current; /* Current interface data. */ + isc_result_t result; /* Last result code. */ +}; + +#ifdef HAVE_TRUCLUSTER +#include <clua/clua.h> +#include <sys/socket.h> +#endif + + +/*% + * Size of buffer for SIOCGLIFCONF, in bytes. We assume no sane system + * will have more than a megabyte of interface configuration data. + */ +#define IFCONF_BUFSIZE_INITIAL 4096 +#define IFCONF_BUFSIZE_MAX 1048576 + +#ifdef __linux +#ifndef IF_NAMESIZE +# ifdef IFNAMSIZ +# define IF_NAMESIZE IFNAMSIZ +# else +# define IF_NAMESIZE 16 +# endif +#endif +#endif + +static isc_result_t +getbuf4(isc_interfaceiter_t *iter) { + char strbuf[ISC_STRERRORSIZE]; + + iter->bufsize = IFCONF_BUFSIZE_INITIAL; + + for (;;) { + iter->buf = isc_mem_get(iter->mctx, iter->bufsize); + if (iter->buf == NULL) + return (ISC_R_NOMEMORY); + + memset(&iter->ifc.ifc_len, 0, sizeof(iter->ifc.ifc_len)); + iter->ifc.ifc_len = iter->bufsize; + iter->ifc.ifc_buf = iter->buf; + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion". It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(iter->socket, SIOCGIFCONF, (char *)&iter->ifc) + == -1) { + if (errno != EINVAL) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETIFCONFIG, + "get interface " + "configuration: %s"), + strbuf); + goto unexpected; + } + /* + * EINVAL. Retry with a bigger buffer. + */ + } else { + /* + * The ioctl succeeded. + * Some OS's just return what will fit rather + * than set EINVAL if the buffer is too small + * to fit all the interfaces in. If + * ifc.lifc_len is too near to the end of the + * buffer we will grow it just in case and + * retry. + */ + if (iter->ifc.ifc_len + 2 * sizeof(struct ifreq) + < iter->bufsize) + break; + } + if (iter->bufsize >= IFCONF_BUFSIZE_MAX) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_BUFFERMAX, + "get interface " + "configuration: " + "maximum buffer " + "size exceeded")); + goto unexpected; + } + isc_mem_put(iter->mctx, iter->buf, iter->bufsize); + + iter->bufsize *= 2; + } + return (ISC_R_SUCCESS); + + unexpected: + isc_mem_put(iter->mctx, iter->buf, iter->bufsize); + iter->buf = NULL; + return (ISC_R_UNEXPECTED); +} + +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) +static isc_result_t +getbuf6(isc_interfaceiter_t *iter) { + char strbuf[ISC_STRERRORSIZE]; + isc_result_t result; + + iter->bufsize6 = IFCONF_BUFSIZE_INITIAL; + + for (;;) { + iter->buf6 = isc_mem_get(iter->mctx, iter->bufsize6); + if (iter->buf6 == NULL) + return (ISC_R_NOMEMORY); + + memset(&iter->lifc, 0, sizeof(iter->lifc)); +#ifdef ISC_HAVE_LIFC_FAMILY + iter->lifc.lifc_family = AF_INET6; +#endif +#ifdef ISC_HAVE_LIFC_FLAGS + iter->lifc.lifc_flags = 0; +#endif + iter->lifc.lifc_len = iter->bufsize6; + iter->lifc.lifc_buf = iter->buf6; + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion". It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(iter->socket6, SIOCGLIFCONF, (char *)&iter->lifc) + == -1) { +#ifdef __hpux + /* + * IPv6 interface scanning is not available on all + * kernels w/ IPv6 sockets. + */ + if (errno == ENOENT) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, + ISC_LOG_DEBUG(1), + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETIFCONFIG, + "get interface " + "configuration: %s"), + strbuf); + result = ISC_R_FAILURE; + goto cleanup; + } +#endif + if (errno != EINVAL) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETIFCONFIG, + "get interface " + "configuration: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup; + } + /* + * EINVAL. Retry with a bigger buffer. + */ + } else { + /* + * The ioctl succeeded. + * Some OS's just return what will fit rather + * than set EINVAL if the buffer is too small + * to fit all the interfaces in. If + * ifc.ifc_len is too near to the end of the + * buffer we will grow it just in case and + * retry. + */ + if (iter->lifc.lifc_len + 2 * sizeof(struct LIFREQ) + < iter->bufsize6) + break; + } + if (iter->bufsize6 >= IFCONF_BUFSIZE_MAX) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_BUFFERMAX, + "get interface " + "configuration: " + "maximum buffer " + "size exceeded")); + result = ISC_R_UNEXPECTED; + goto cleanup; + } + isc_mem_put(iter->mctx, iter->buf6, iter->bufsize6); + + iter->bufsize6 *= 2; + } + + if (iter->lifc.lifc_len != 0) + iter->mode = 6; + return (ISC_R_SUCCESS); + + cleanup: + isc_mem_put(iter->mctx, iter->buf6, iter->bufsize6); + iter->buf6 = NULL; + return (result); +} +#endif + +isc_result_t +isc_interfaceiter_create(isc_mem_t *mctx, isc_interfaceiter_t **iterp) { + isc_interfaceiter_t *iter; + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(mctx != NULL); + REQUIRE(iterp != NULL); + REQUIRE(*iterp == NULL); + + iter = isc_mem_get(mctx, sizeof(*iter)); + if (iter == NULL) + return (ISC_R_NOMEMORY); + + iter->mctx = mctx; + iter->mode = 4; + iter->buf = NULL; + iter->pos = (unsigned int) -1; +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + iter->buf6 = NULL; + iter->pos6 = (unsigned int) -1; + iter->result6 = ISC_R_NOMORE; + iter->socket6 = -1; + iter->first6 = false; +#endif + + /* + * Get the interface configuration, allocating more memory if + * necessary. + */ + +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + result = isc_net_probeipv6(); + if (result == ISC_R_SUCCESS) { + /* + * Create an unbound datagram socket to do the SIOCGLIFCONF + * ioctl on. HP/UX requires an AF_INET6 socket for + * SIOCGLIFCONF to get IPv6 addresses. + */ + if ((iter->socket6 = socket(AF_INET6, SOCK_DGRAM, 0)) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_MAKESCANSOCKET, + "making interface " + "scan socket: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto socket6_failure; + } + result = iter->result6 = getbuf6(iter); + if (result != ISC_R_NOTIMPLEMENTED && result != ISC_R_SUCCESS) + goto ioctl6_failure; + } +#endif + if ((iter->socket = socket(AF_INET, SOCK_DGRAM, 0)) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_MAKESCANSOCKET, + "making interface " + "scan socket: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto socket_failure; + } + result = getbuf4(iter); + if (result != ISC_R_SUCCESS) + goto ioctl_failure; + + /* + * A newly created iterator has an undefined position + * until isc_interfaceiter_first() is called. + */ +#ifdef HAVE_TRUCLUSTER + iter->clua_context = -1; + iter->clua_done = true; +#endif +#ifdef __linux + iter->proc = fopen("/proc/net/if_inet6", "r"); + iter->valid = ISC_R_FAILURE; +#endif + iter->result = ISC_R_FAILURE; + + iter->magic = IFITER_MAGIC; + *iterp = iter; + return (ISC_R_SUCCESS); + + ioctl_failure: + if (iter->buf != NULL) + isc_mem_put(mctx, iter->buf, iter->bufsize); + (void) close(iter->socket); + + socket_failure: +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + if (iter->buf6 != NULL) + isc_mem_put(mctx, iter->buf6, iter->bufsize6); + ioctl6_failure: + if (iter->socket6 != -1) + (void) close(iter->socket6); + socket6_failure: +#endif + + isc_mem_put(mctx, iter, sizeof(*iter)); + return (result); +} + +#ifdef HAVE_TRUCLUSTER +static void +get_inaddr(isc_netaddr_t *dst, struct in_addr *src) { + dst->family = AF_INET; + memmove(&dst->type.in, src, sizeof(struct in_addr)); +} + +static isc_result_t +internal_current_clusteralias(isc_interfaceiter_t *iter) { + struct clua_info ci; + if (clua_getaliasinfo(&iter->clua_sa, &ci) != CLUA_SUCCESS) + return (ISC_R_IGNORE); + memset(&iter->current, 0, sizeof(iter->current)); + iter->current.af = iter->clua_sa.sa_family; + memset(iter->current.name, 0, sizeof(iter->current.name)); + snprintf(iter->current.name, sizeof(iter->current.name), + "clua%d", ci.aliasid); + iter->current.flags = INTERFACE_F_UP; + get_inaddr(&iter->current.address, &ci.addr); + get_inaddr(&iter->current.netmask, &ci.netmask); + return (ISC_R_SUCCESS); +} +#endif + +/* + * Get information about the current interface to iter->current. + * If successful, return ISC_R_SUCCESS. + * If the interface has an unsupported address family, or if + * some operation on it fails, return ISC_R_IGNORE to make + * the higher-level iterator code ignore it. + */ + +static isc_result_t +internal_current4(isc_interfaceiter_t *iter) { + struct ifreq *ifrp; + struct ifreq ifreq; + int family; + char strbuf[ISC_STRERRORSIZE]; +#if !defined(ISC_PLATFORM_HAVEIF_LADDRREQ) && defined(SIOCGLIFADDR) + struct lifreq lifreq; +#else + char sabuf[256]; +#endif + int i, bits, prefixlen; + + REQUIRE(VALID_IFITER(iter)); + + if (iter->ifc.ifc_len == 0 || + iter->pos == (unsigned int)iter->ifc.ifc_len) { +#ifdef __linux + return (linux_if_inet6_current(iter)); +#else + return (ISC_R_NOMORE); +#endif + } + + INSIST( iter->pos < (unsigned int) iter->ifc.ifc_len); + + ifrp = (struct ifreq *)((char *) iter->ifc.ifc_req + iter->pos); + + memset(&ifreq, 0, sizeof(ifreq)); + memmove(&ifreq, ifrp, sizeof(ifreq)); + + family = ifreq.ifr_addr.sa_family; +#if defined(ISC_PLATFORM_HAVEIPV6) + if (family != AF_INET && family != AF_INET6) +#else + if (family != AF_INET) +#endif + return (ISC_R_IGNORE); + + memset(&iter->current, 0, sizeof(iter->current)); + iter->current.af = family; + + INSIST(sizeof(ifreq.ifr_name) <= sizeof(iter->current.name)); + memset(iter->current.name, 0, sizeof(iter->current.name)); + memmove(iter->current.name, ifreq.ifr_name, sizeof(ifreq.ifr_name)); + + get_addr(family, &iter->current.address, + (struct sockaddr *)&ifrp->ifr_addr, ifreq.ifr_name); + + /* + * If the interface does not have a address ignore it. + */ + switch (family) { + case AF_INET: + if (iter->current.address.type.in.s_addr == htonl(INADDR_ANY)) + return (ISC_R_IGNORE); + break; + case AF_INET6: + if (memcmp(&iter->current.address.type.in6, &in6addr_any, + sizeof(in6addr_any)) == 0) + return (ISC_R_IGNORE); + break; + } + + /* + * Get interface flags. + */ + + iter->current.flags = 0; + + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(iter->socket, SIOCGIFFLAGS, (char *) &ifreq) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "%s: getting interface flags: %s", + ifreq.ifr_name, strbuf); + return (ISC_R_IGNORE); + } + + if ((ifreq.ifr_flags & IFF_UP) != 0) + iter->current.flags |= INTERFACE_F_UP; + +#ifdef IFF_POINTOPOINT + if ((ifreq.ifr_flags & IFF_POINTOPOINT) != 0) + iter->current.flags |= INTERFACE_F_POINTTOPOINT; +#endif + + if ((ifreq.ifr_flags & IFF_LOOPBACK) != 0) + iter->current.flags |= INTERFACE_F_LOOPBACK; + + if (family == AF_INET) + goto inet; + +#if !defined(ISC_PLATFORM_HAVEIF_LADDRREQ) && defined(SIOCGLIFADDR) + memset(&lifreq, 0, sizeof(lifreq)); + memmove(lifreq.lifr_name, iter->current.name, sizeof(lifreq.lifr_name)); + memmove(&lifreq.lifr_addr, &iter->current.address.type.in6, + sizeof(iter->current.address.type.in6)); + + if (ioctl(iter->socket, SIOCGLIFADDR, &lifreq) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "%s: getting interface address: %s", + ifreq.ifr_name, strbuf); + return (ISC_R_IGNORE); + } + prefixlen = lifreq.lifr_addrlen; +#else + isc_netaddr_format(&iter->current.address, sabuf, sizeof(sabuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, + ISC_LOG_INFO, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETIFCONFIG, + "prefix length for %s is unknown " + "(assume 128)"), sabuf); + prefixlen = 128; +#endif + + /* + * Netmask already zeroed. + */ + iter->current.netmask.family = family; + for (i = 0; i < 16; i++) { + if (prefixlen > 8) { + bits = 0; + prefixlen -= 8; + } else { + bits = 8 - prefixlen; + prefixlen = 0; + } + iter->current.netmask.type.in6.s6_addr[i] = + (~0U << bits) & 0xff; + } + return (ISC_R_SUCCESS); + + inet: + if (family != AF_INET) + return (ISC_R_IGNORE); +#ifdef IFF_POINTOPOINT + /* + * If the interface is point-to-point, get the destination address. + */ + if ((iter->current.flags & INTERFACE_F_POINTTOPOINT) != 0) { + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(iter->socket, SIOCGIFDSTADDR, (char *)&ifreq) + < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETDESTADDR, + "%s: getting " + "destination address: %s"), + ifreq.ifr_name, strbuf); + return (ISC_R_IGNORE); + } + get_addr(family, &iter->current.dstaddress, + (struct sockaddr *)&ifreq.ifr_dstaddr, ifreq.ifr_name); + } +#endif + + /* + * Get the network mask. + */ + memset(&ifreq, 0, sizeof(ifreq)); + memmove(&ifreq, ifrp, sizeof(ifreq)); + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(iter->socket, SIOCGIFNETMASK, (char *)&ifreq) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETNETMASK, + "%s: getting netmask: %s"), + ifreq.ifr_name, strbuf); + return (ISC_R_IGNORE); + } + get_addr(family, &iter->current.netmask, + (struct sockaddr *)&ifreq.ifr_addr, ifreq.ifr_name); + return (ISC_R_SUCCESS); +} + +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) +static isc_result_t +internal_current6(isc_interfaceiter_t *iter) { + struct LIFREQ *ifrp; + struct LIFREQ lifreq; + int family; + char strbuf[ISC_STRERRORSIZE]; + int fd; + + REQUIRE(VALID_IFITER(iter)); + if (iter->result6 != ISC_R_SUCCESS) + return (iter->result6); + REQUIRE(iter->pos6 < (unsigned int) iter->lifc.lifc_len); + + ifrp = (struct LIFREQ *)((char *) iter->lifc.lifc_req + iter->pos6); + + memset(&lifreq, 0, sizeof(lifreq)); + memmove(&lifreq, ifrp, sizeof(lifreq)); + + family = lifreq.lifr_addr.ss_family; +#ifdef ISC_PLATFORM_HAVEIPV6 + if (family != AF_INET && family != AF_INET6) +#else + if (family != AF_INET) +#endif + return (ISC_R_IGNORE); + + memset(&iter->current, 0, sizeof(iter->current)); + iter->current.af = family; + + INSIST(sizeof(lifreq.lifr_name) <= sizeof(iter->current.name)); + memset(iter->current.name, 0, sizeof(iter->current.name)); + memmove(iter->current.name, lifreq.lifr_name, sizeof(lifreq.lifr_name)); + + get_addr(family, &iter->current.address, + (struct sockaddr *)&lifreq.lifr_addr, lifreq.lifr_name); + + /* + * If the interface does not have a address ignore it. + */ + switch (family) { + case AF_INET: + if (iter->current.address.type.in.s_addr == htonl(INADDR_ANY)) + return (ISC_R_IGNORE); + break; + case AF_INET6: + if (memcmp(&iter->current.address.type.in6, &in6addr_any, + sizeof(in6addr_any)) == 0) + return (ISC_R_IGNORE); + break; + } + + /* + * Get interface flags. + */ + + iter->current.flags = 0; + + if (family == AF_INET6) + fd = iter->socket6; + else + fd = iter->socket; + + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(fd, SIOCGLIFFLAGS, (char *) &lifreq) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "%s: getting interface flags: %s", + lifreq.lifr_name, strbuf); + return (ISC_R_IGNORE); + } + + if ((lifreq.lifr_flags & IFF_UP) != 0) + iter->current.flags |= INTERFACE_F_UP; + +#ifdef IFF_POINTOPOINT + if ((lifreq.lifr_flags & IFF_POINTOPOINT) != 0) + iter->current.flags |= INTERFACE_F_POINTTOPOINT; +#endif + + if ((lifreq.lifr_flags & IFF_LOOPBACK) != 0) + iter->current.flags |= INTERFACE_F_LOOPBACK; + +#ifdef IFF_POINTOPOINT + /* + * If the interface is point-to-point, get the destination address. + */ + if ((iter->current.flags & INTERFACE_F_POINTTOPOINT) != 0) { + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(fd, SIOCGLIFDSTADDR, (char *)&lifreq) + < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETDESTADDR, + "%s: getting " + "destination address: %s"), + lifreq.lifr_name, strbuf); + return (ISC_R_IGNORE); + } + get_addr(family, &iter->current.dstaddress, + (struct sockaddr *)&lifreq.lifr_dstaddr, + lifreq.lifr_name); + } +#endif + + /* + * Get the network mask. Netmask already zeroed. + */ + memset(&lifreq, 0, sizeof(lifreq)); + memmove(&lifreq, ifrp, sizeof(lifreq)); + +#ifdef lifr_addrlen + /* + * Special case: if the system provides lifr_addrlen member, the + * netmask of an IPv6 address can be derived from the length, since + * an IPv6 address always has a contiguous mask. + */ + if (family == AF_INET6) { + int i, bits; + + iter->current.netmask.family = family; + for (i = 0; i < lifreq.lifr_addrlen; i += 8) { + bits = lifreq.lifr_addrlen - i; + bits = (bits < 8) ? (8 - bits) : 0; + iter->current.netmask.type.in6.s6_addr[i / 8] = + (~0U << bits) & 0xff; + } + + return (ISC_R_SUCCESS); + } +#endif + + /* + * Ignore the HP/UX warning about "integer overflow during + * conversion. It comes from its own macro definition, + * and is really hard to shut up. + */ + if (ioctl(fd, SIOCGLIFNETMASK, (char *)&lifreq) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERIOCTL, + ISC_MSG_GETNETMASK, + "%s: getting netmask: %s"), + lifreq.lifr_name, strbuf); + return (ISC_R_IGNORE); + } + get_addr(family, &iter->current.netmask, + (struct sockaddr *)&lifreq.lifr_addr, lifreq.lifr_name); + + return (ISC_R_SUCCESS); +} +#endif + +static isc_result_t +internal_current(isc_interfaceiter_t *iter) { +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + if (iter->mode == 6) { + iter->result6 = internal_current6(iter); + if (iter->result6 != ISC_R_NOMORE) + return (iter->result6); + } +#endif +#ifdef HAVE_TRUCLUSTER + if (!iter->clua_done) + return(internal_current_clusteralias(iter)); +#endif + return (internal_current4(iter)); +} + +/* + * Step the iterator to the next interface. Unlike + * isc_interfaceiter_next(), this may leave the iterator + * positioned on an interface that will ultimately + * be ignored. Return ISC_R_NOMORE if there are no more + * interfaces, otherwise ISC_R_SUCCESS. + */ +static isc_result_t +internal_next4(isc_interfaceiter_t *iter) { +#ifdef ISC_PLATFORM_HAVESALEN + struct ifreq *ifrp; +#endif + + if (iter->pos < (unsigned int) iter->ifc.ifc_len) { +#ifdef ISC_PLATFORM_HAVESALEN + ifrp = (struct ifreq *)((char *) iter->ifc.ifc_req + iter->pos); + + if (ifrp->ifr_addr.sa_len > sizeof(struct sockaddr)) + iter->pos += sizeof(ifrp->ifr_name) + + ifrp->ifr_addr.sa_len; + else +#endif + iter->pos += sizeof(struct ifreq); + + } else { + INSIST(iter->pos == (unsigned int) iter->ifc.ifc_len); +#ifdef __linux + return (linux_if_inet6_next(iter)); +#else + return (ISC_R_NOMORE); +#endif + } + return (ISC_R_SUCCESS); +} + +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) +static isc_result_t +internal_next6(isc_interfaceiter_t *iter) { +#ifdef ISC_PLATFORM_HAVESALEN + struct LIFREQ *ifrp; +#endif + + if (iter->result6 != ISC_R_SUCCESS && iter->result6 != ISC_R_IGNORE) + return (iter->result6); + + REQUIRE(iter->pos6 < (unsigned int) iter->lifc.lifc_len); + +#ifdef ISC_PLATFORM_HAVESALEN + ifrp = (struct LIFREQ *)((char *) iter->lifc.lifc_req + iter->pos6); + + if (ifrp->lifr_addr.sa_len > sizeof(struct sockaddr)) + iter->pos6 += sizeof(ifrp->lifr_name) + ifrp->lifr_addr.sa_len; + else +#endif + iter->pos6 += sizeof(struct LIFREQ); + + if (iter->pos6 >= (unsigned int) iter->lifc.lifc_len) + return (ISC_R_NOMORE); + + return (ISC_R_SUCCESS); +} +#endif + +static isc_result_t +internal_next(isc_interfaceiter_t *iter) { +#ifdef HAVE_TRUCLUSTER + int clua_result; +#endif +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + if (iter->mode == 6) { + iter->result6 = internal_next6(iter); + if (iter->result6 != ISC_R_NOMORE) + return (iter->result6); + if (iter->first6) { + iter->first6 = false; + return (ISC_R_SUCCESS); + } + } +#endif +#ifdef HAVE_TRUCLUSTER + if (!iter->clua_done) { + clua_result = clua_getaliasaddress(&iter->clua_sa, + &iter->clua_context); + if (clua_result != CLUA_SUCCESS) + iter->clua_done = true; + return (ISC_R_SUCCESS); + } +#endif + return (internal_next4(iter)); +} + +static void +internal_destroy(isc_interfaceiter_t *iter) { + (void) close(iter->socket); +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + if (iter->socket6 != -1) + (void) close(iter->socket6); + if (iter->buf6 != NULL) { + isc_mem_put(iter->mctx, iter->buf6, iter->bufsize6); + } +#endif +#ifdef __linux + if (iter->proc != NULL) + fclose(iter->proc); +#endif +} + +static +void internal_first(isc_interfaceiter_t *iter) { +#ifdef HAVE_TRUCLUSTER + int clua_result; +#endif + iter->pos = 0; +#if defined(SIOCGLIFCONF) && defined(SIOCGLIFADDR) + iter->pos6 = 0; + if (iter->result6 == ISC_R_NOMORE) + iter->result6 = ISC_R_SUCCESS; + iter->first6 = true; +#endif +#ifdef HAVE_TRUCLUSTER + iter->clua_context = 0; + clua_result = clua_getaliasaddress(&iter->clua_sa, + &iter->clua_context); + iter->clua_done = (clua_result != CLUA_SUCCESS); +#endif +#ifdef __linux + linux_if_inet6_first(iter); +#endif +} diff --git a/lib/isc/unix/ifiter_sysctl.c b/lib/isc/unix/ifiter_sysctl.c new file mode 100644 index 0000000..ebbefe9 --- /dev/null +++ b/lib/isc/unix/ifiter_sysctl.c @@ -0,0 +1,300 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file + * \brief + * Obtain the list of network interfaces using sysctl. + * See TCP/IP Illustrated Volume 2, sections 19.8, 19.14, + * and 19.16. + */ + +#include <inttypes.h> + +#include <sys/param.h> +#include <sys/sysctl.h> + +#include <net/route.h> +#include <net/if_dl.h> + +#include <isc/print.h> + +/* XXX what about Alpha? */ +#ifdef sgi +#define ROUNDUP(a) ((a) > 0 ? \ + (1 + (((a) - 1) | (sizeof(__uint64_t) - 1))) : \ + sizeof(__uint64_t)) +#else +#define ROUNDUP(a) ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) \ + : sizeof(long)) +#endif + +#define IFITER_MAGIC ISC_MAGIC('I', 'F', 'I', 'S') +#define VALID_IFITER(t) ISC_MAGIC_VALID(t, IFITER_MAGIC) + +struct isc_interfaceiter { + unsigned int magic; /* Magic number. */ + isc_mem_t *mctx; + void *buf; /* Buffer for sysctl data. */ + unsigned int bufsize; /* Bytes allocated. */ + unsigned int bufused; /* Bytes used. */ + unsigned int pos; /* Current offset in + sysctl data. */ + isc_interface_t current; /* Current interface data. */ + isc_result_t result; /* Last result code. */ +}; + +static int mib[6] = { + CTL_NET, + PF_ROUTE, + 0, + 0, /* Any address family. */ + NET_RT_IFLIST, + 0 /* Flags. */ +}; + +isc_result_t +isc_interfaceiter_create(isc_mem_t *mctx, isc_interfaceiter_t **iterp) { + isc_interfaceiter_t *iter; + isc_result_t result; + size_t bufsize; + size_t bufused; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(mctx != NULL); + REQUIRE(iterp != NULL); + REQUIRE(*iterp == NULL); + + iter = isc_mem_get(mctx, sizeof(*iter)); + if (iter == NULL) + return (ISC_R_NOMEMORY); + + iter->mctx = mctx; + iter->buf = 0; + + /* + * Determine the amount of memory needed. + */ + bufsize = 0; + if (sysctl(mib, 6, NULL, &bufsize, NULL, (size_t) 0) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERSYSCTL, + ISC_MSG_GETIFLISTSIZE, + "getting interface " + "list size: sysctl: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto failure; + } + iter->bufsize = bufsize; + + iter->buf = isc_mem_get(iter->mctx, iter->bufsize); + if (iter->buf == NULL) { + result = ISC_R_NOMEMORY; + goto failure; + } + + bufused = bufsize; + if (sysctl(mib, 6, iter->buf, &bufused, NULL, (size_t) 0) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_IFITERSYSCTL, + ISC_MSG_GETIFLIST, + "getting interface list: " + "sysctl: %s"), + strbuf); + result = ISC_R_UNEXPECTED; + goto failure; + } + iter->bufused = bufused; + INSIST(iter->bufused <= iter->bufsize); + + /* + * A newly created iterator has an undefined position + * until isc_interfaceiter_first() is called. + */ + iter->pos = (unsigned int) -1; + iter->result = ISC_R_FAILURE; + + iter->magic = IFITER_MAGIC; + *iterp = iter; + return (ISC_R_SUCCESS); + + failure: + if (iter->buf != NULL) + isc_mem_put(mctx, iter->buf, iter->bufsize); + isc_mem_put(mctx, iter, sizeof(*iter)); + return (result); +} + +/* + * Get information about the current interface to iter->current. + * If successful, return ISC_R_SUCCESS. + * If the interface has an unsupported address family, + * return ISC_R_IGNORE. In case of other failure, + * return ISC_R_UNEXPECTED. + */ + +static isc_result_t +internal_current(isc_interfaceiter_t *iter) { + struct ifa_msghdr *ifam, *ifam_end; + + REQUIRE(VALID_IFITER(iter)); + REQUIRE (iter->pos < (unsigned int) iter->bufused); + + ifam = (struct ifa_msghdr *) ((char *) iter->buf + iter->pos); + ifam_end = (struct ifa_msghdr *) ((char *) iter->buf + iter->bufused); + + if (ifam->ifam_type == RTM_IFINFO) { + struct if_msghdr *ifm = (struct if_msghdr *) ifam; + struct sockaddr_dl *sdl = (struct sockaddr_dl *) (ifm + 1); + unsigned int namelen; + + memset(&iter->current, 0, sizeof(iter->current)); + + namelen = sdl->sdl_nlen; + if (namelen > sizeof(iter->current.name) - 1) + namelen = sizeof(iter->current.name) - 1; + + memset(iter->current.name, 0, sizeof(iter->current.name)); + memmove(iter->current.name, sdl->sdl_data, namelen); + + iter->current.flags = 0; + + if ((ifam->ifam_flags & IFF_UP) != 0) + iter->current.flags |= INTERFACE_F_UP; + + if ((ifam->ifam_flags & IFF_POINTOPOINT) != 0) + iter->current.flags |= INTERFACE_F_POINTTOPOINT; + + if ((ifam->ifam_flags & IFF_LOOPBACK) != 0) + iter->current.flags |= INTERFACE_F_LOOPBACK; + + /* + * This is not an interface address. + * Force another iteration. + */ + return (ISC_R_IGNORE); + } else if (ifam->ifam_type == RTM_NEWADDR) { + int i; + int family; + struct sockaddr *mask_sa = NULL; + struct sockaddr *addr_sa = NULL; + struct sockaddr *dst_sa = NULL; + + struct sockaddr *sa = (struct sockaddr *)(ifam + 1); + family = sa->sa_family; + + for (i = 0; i < RTAX_MAX; i++) + { + if ((ifam->ifam_addrs & (1 << i)) == 0) + continue; + + INSIST(sa < (struct sockaddr *) ifam_end); + + switch (i) { + case RTAX_NETMASK: /* Netmask */ + mask_sa = sa; + break; + case RTAX_IFA: /* Interface address */ + addr_sa = sa; + break; + case RTAX_BRD: /* Broadcast or destination address */ + dst_sa = sa; + break; + } +#ifdef ISC_PLATFORM_HAVESALEN + sa = (struct sockaddr *)((char*)(sa) + + ROUNDUP(sa->sa_len)); +#else +#ifdef sgi + /* + * Do as the contributed SGI code does. + */ + sa = (struct sockaddr *)((char*)(sa) + + ROUNDUP(_FAKE_SA_LEN_DST(sa))); +#else + /* XXX untested. */ + sa = (struct sockaddr *)((char*)(sa) + + ROUNDUP(sizeof(struct sockaddr))); +#endif +#endif + } + + if (addr_sa == NULL) + return (ISC_R_IGNORE); + + family = addr_sa->sa_family; + if (family != AF_INET && family != AF_INET6) + return (ISC_R_IGNORE); + + iter->current.af = family; + + get_addr(family, &iter->current.address, addr_sa, + iter->current.name); + + if (mask_sa != NULL) + get_addr(family, &iter->current.netmask, mask_sa, + iter->current.name); + + if (dst_sa != NULL && + (iter->current.flags & INTERFACE_F_POINTTOPOINT) != 0) + get_addr(family, &iter->current.dstaddress, dst_sa, + iter->current.name); + + return (ISC_R_SUCCESS); + } else { + printf("%s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_IFITERSYSCTL, + ISC_MSG_UNEXPECTEDTYPE, + "warning: unexpected interface list " + "message type\n")); + return (ISC_R_IGNORE); + } +} + +/* + * Step the iterator to the next interface. Unlike + * isc_interfaceiter_next(), this may leave the iterator + * positioned on an interface that will ultimately + * be ignored. Return ISC_R_NOMORE if there are no more + * interfaces, otherwise ISC_R_SUCCESS. + */ +static isc_result_t +internal_next(isc_interfaceiter_t *iter) { + struct ifa_msghdr *ifam; + REQUIRE (iter->pos < (unsigned int) iter->bufused); + + ifam = (struct ifa_msghdr *) ((char *) iter->buf + iter->pos); + + iter->pos += ifam->ifam_msglen; + + if (iter->pos >= iter->bufused) + return (ISC_R_NOMORE); + + return (ISC_R_SUCCESS); +} + +static void +internal_destroy(isc_interfaceiter_t *iter) { + UNUSED(iter); /* Unused. */ + /* + * Do nothing. + */ +} + +static +void internal_first(isc_interfaceiter_t *iter) { + iter->pos = 0; +} diff --git a/lib/isc/unix/include/Makefile.in b/lib/isc/unix/include/Makefile.in new file mode 100644 index 0000000..2cbf021 --- /dev/null +++ b/lib/isc/unix/include/Makefile.in @@ -0,0 +1,17 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +SUBDIRS = isc pkcs11 +TARGETS = + +@BIND9_MAKE_RULES@ diff --git a/lib/isc/unix/include/isc/Makefile.in b/lib/isc/unix/include/isc/Makefile.in new file mode 100644 index 0000000..977defe --- /dev/null +++ b/lib/isc/unix/include/isc/Makefile.in @@ -0,0 +1,35 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +VERSION=@BIND9_VERSION@ + +HEADERS = dir.h keyboard.h net.h netdb.h offset.h stat.h \ + stdtime.h strerror.h syslog.h time.h + +SUBDIRS = +TARGETS = + +@BIND9_MAKE_RULES@ + +installdirs: + $(SHELL) ${top_srcdir}/mkinstalldirs ${DESTDIR}${includedir}/isc + +install:: installdirs + for i in ${HEADERS}; do \ + ${INSTALL_DATA} $(srcdir)/$$i ${DESTDIR}${includedir}/isc || exit 1; \ + done + +uninstall:: + for i in ${HEADERS}; do \ + rm -f ${DESTDIR}${includedir}/isc/$$i || exit 1; \ + done diff --git a/lib/isc/unix/include/isc/dir.h b/lib/isc/unix/include/isc/dir.h new file mode 100644 index 0000000..36d0aad --- /dev/null +++ b/lib/isc/unix/include/isc/dir.h @@ -0,0 +1,84 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_DIR_H +#define ISC_DIR_H 1 + +/*! \file */ + +#include <sys/types.h> /* Required on some systems. */ +#include <dirent.h> + +#include <isc/lang.h> +#include <isc/result.h> + +#define ISC_DIR_NAMEMAX 256 +#define ISC_DIR_PATHMAX 1024 + +/*% Directory Entry */ +typedef struct isc_direntry { + /*! + * Ideally, this should be NAME_MAX, but AIX does not define it by + * default and dynamically allocating the space based on pathconf() + * complicates things undesirably, as does adding special conditionals + * just for AIX. So a comfortably sized buffer is chosen instead. + */ + char name[ISC_DIR_NAMEMAX]; + unsigned int length; +} isc_direntry_t; + +/*% Directory */ +typedef struct isc_dir { + unsigned int magic; + /*! + * As with isc_direntry_t->name, making this "right" for all systems + * is slightly problematic because AIX does not define PATH_MAX. + */ + char dirname[ISC_DIR_PATHMAX]; + isc_direntry_t entry; + DIR * handle; +} isc_dir_t; + +ISC_LANG_BEGINDECLS + +void +isc_dir_init(isc_dir_t *dir); + +isc_result_t +isc_dir_open(isc_dir_t *dir, const char *dirname); + +isc_result_t +isc_dir_read(isc_dir_t *dir); + +isc_result_t +isc_dir_reset(isc_dir_t *dir); + +void +isc_dir_close(isc_dir_t *dir); + +isc_result_t +isc_dir_chdir(const char *dirname); + +isc_result_t +isc_dir_chroot(const char *dirname); + +isc_result_t +isc_dir_createunique(char *templet); +/*!< + * Use a templet (such as from isc_file_mktemplate()) to create a uniquely + * named, empty directory. The templet string is modified in place. + * If result == ISC_R_SUCCESS, it is the name of the directory that was + * created. + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_DIR_H */ diff --git a/lib/isc/unix/include/isc/keyboard.h b/lib/isc/unix/include/isc/keyboard.h new file mode 100644 index 0000000..4dc8745 --- /dev/null +++ b/lib/isc/unix/include/isc/keyboard.h @@ -0,0 +1,46 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_KEYBOARD_H +#define ISC_KEYBOARD_H 1 + +/*! \file */ + +#include <stdbool.h> +#include <termios.h> + +#include <isc/lang.h> +#include <isc/result.h> + +ISC_LANG_BEGINDECLS + +typedef struct { + int fd; + struct termios saved_mode; + isc_result_t result; +} isc_keyboard_t; + +isc_result_t +isc_keyboard_open(isc_keyboard_t *keyboard); + +isc_result_t +isc_keyboard_close(isc_keyboard_t *keyboard, unsigned int sleepseconds); + +isc_result_t +isc_keyboard_getchar(isc_keyboard_t *keyboard, unsigned char *cp); + +bool +isc_keyboard_canceled(isc_keyboard_t *keyboard); + +ISC_LANG_ENDDECLS + +#endif /* ISC_KEYBOARD_H */ diff --git a/lib/isc/unix/include/isc/net.h b/lib/isc/unix/include/isc/net.h new file mode 100644 index 0000000..08be590 --- /dev/null +++ b/lib/isc/unix/include/isc/net.h @@ -0,0 +1,417 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_NET_H +#define ISC_NET_H 1 + +/***** + ***** Module Info + *****/ + +/*! \file + * \brief + * Basic Networking Types + * + * This module is responsible for defining the following basic networking + * types: + * + *\li struct in_addr + *\li struct in6_addr + *\li struct in6_pktinfo + *\li struct sockaddr + *\li struct sockaddr_in + *\li struct sockaddr_in6 + *\li struct sockaddr_storage + *\li in_port_t + * + * It ensures that the AF_ and PF_ macros are defined. + * + * It declares ntoh[sl]() and hton[sl](). + * + * It declares inet_aton(), inet_ntop(), and inet_pton(). + * + * It ensures that #INADDR_LOOPBACK, #INADDR_ANY, #IN6ADDR_ANY_INIT, + * IN6ADDR_V4MAPPED_INIT, in6addr_any, and in6addr_loopback are available. + * + * It ensures that IN_MULTICAST() is available to check for multicast + * addresses. + * + * MP: + *\li No impact. + * + * Reliability: + *\li No anticipated impact. + * + * Resources: + *\li N/A. + * + * Security: + *\li No anticipated impact. + * + * Standards: + *\li BSD Socket API + *\li RFC2553 + */ + +/*** + *** Imports. + ***/ +#include <isc/platform.h> + +#include <inttypes.h> + +#include <sys/types.h> +#include <sys/socket.h> /* Contractual promise. */ + +#include <net/if.h> + +#include <netinet/in.h> /* Contractual promise. */ +#include <arpa/inet.h> /* Contractual promise. */ +#ifdef ISC_PLATFORM_NEEDNETINETIN6H +#include <netinet/in6.h> /* Required on UnixWare. */ +#endif +#ifdef ISC_PLATFORM_NEEDNETINET6IN6H +#include <netinet6/in6.h> /* Required on BSD/OS for in6_pktinfo. */ +#endif + +#ifndef ISC_PLATFORM_HAVEIPV6 +#include <isc/ipv6.h> /* Contractual promise. */ +#endif + +#include <isc/lang.h> +#include <isc/types.h> + +#ifdef ISC_PLATFORM_HAVEINADDR6 +#define in6_addr in_addr6 /*%< Required for pre RFC2133 implementations. */ +#endif + +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifndef IN6ADDR_ANY_INIT +#ifdef s6_addr +/*% + * Required for some pre RFC2133 implementations. + * IN6ADDR_ANY_INIT and IN6ADDR_LOOPBACK_INIT were added in + * draft-ietf-ipngwg-bsd-api-04.txt or draft-ietf-ipngwg-bsd-api-05.txt. + * If 's6_addr' is defined then assume that there is a union and three + * levels otherwise assume two levels required. + */ +#define IN6ADDR_ANY_INIT { { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } } } +#else +#define IN6ADDR_ANY_INIT { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } } +#endif +#endif + +#ifndef IN6ADDR_LOOPBACK_INIT +#ifdef s6_addr +/*% IPv6 address loopback init */ +#define IN6ADDR_LOOPBACK_INIT { { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1 } } } +#else +#define IN6ADDR_LOOPBACK_INIT { { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1 } } +#endif +#endif + +#ifndef IN6ADDR_V4MAPPED_INIT +#ifdef s6_addr +/*% IPv6 v4mapped prefix init */ +#define IN6ADDR_V4MAPPED_INIT { { { 0,0,0,0,0,0,0,0,0,0,0xff,0xff,0,0,0,0 } } } +#else +#define IN6ADDR_V4MAPPED_INIT { { 0,0,0,0,0,0,0,0,0,0,0xff,0xff,0,0,0,0 } } +#endif +#endif + +#ifndef IN6_IS_ADDR_V4MAPPED +/*% Is IPv6 address V4 mapped? */ +#define IN6_IS_ADDR_V4MAPPED(x) \ + (memcmp((x)->s6_addr, in6addr_any.s6_addr, 10) == 0 && \ + (x)->s6_addr[10] == 0xff && (x)->s6_addr[11] == 0xff) +#endif + +#ifndef IN6_IS_ADDR_V4COMPAT +/*% Is IPv6 address V4 compatible? */ +#define IN6_IS_ADDR_V4COMPAT(x) \ + (memcmp((x)->s6_addr, in6addr_any.s6_addr, 12) == 0 && \ + ((x)->s6_addr[12] != 0 || (x)->s6_addr[13] != 0 || \ + (x)->s6_addr[14] != 0 || \ + ((x)->s6_addr[15] != 0 && (x)->s6_addr[15] != 1))) +#endif + +#ifndef IN6_IS_ADDR_MULTICAST +/*% Is IPv6 address multicast? */ +#define IN6_IS_ADDR_MULTICAST(a) ((a)->s6_addr[0] == 0xff) +#endif + +#ifndef IN6_IS_ADDR_LINKLOCAL +/*% Is IPv6 address linklocal? */ +#define IN6_IS_ADDR_LINKLOCAL(a) \ + (((a)->s6_addr[0] == 0xfe) && (((a)->s6_addr[1] & 0xc0) == 0x80)) +#endif + +#ifndef IN6_IS_ADDR_SITELOCAL +/*% is IPv6 address sitelocal? */ +#define IN6_IS_ADDR_SITELOCAL(a) \ + (((a)->s6_addr[0] == 0xfe) && (((a)->s6_addr[1] & 0xc0) == 0xc0)) +#endif + + +#ifndef IN6_IS_ADDR_LOOPBACK +/*% is IPv6 address loopback? */ +#define IN6_IS_ADDR_LOOPBACK(x) \ + (memcmp((x)->s6_addr, in6addr_loopback.s6_addr, 16) == 0) +#endif +#endif + +#ifndef AF_INET6 +/*% IPv6 */ +#define AF_INET6 99 +#endif + +#ifndef PF_INET6 +/*% IPv6 */ +#define PF_INET6 AF_INET6 +#endif + +#ifndef INADDR_ANY +/*% inaddr any */ +#define INADDR_ANY 0x00000000UL +#endif + +#ifndef INADDR_LOOPBACK +/*% inaddr loopback */ +#define INADDR_LOOPBACK 0x7f000001UL +#endif + +#ifndef ISC_PLATFORM_HAVEIN6PKTINFO +/*% IPv6 packet info */ +struct in6_pktinfo { + struct in6_addr ipi6_addr; /*%< src/dst IPv6 address */ + unsigned int ipi6_ifindex; /*%< send/recv interface index */ +}; +#endif + + +#ifndef ISC_PLATFORM_HAVESOCKADDRSTORAGE +#define _SS_MAXSIZE 128 +#define _SS_ALIGNSIZE (sizeof (uint64_t)) +#ifdef ISC_PLATFORM_HAVESALEN +#define _SS_PAD1SIZE (_SS_ALIGNSIZE - (2 * sizeof(uint8_t))) +#define _SS_PAD2SIZE (_SS_MAXSIZE - (_SS_ALIGNSIZE + _SS_PAD1SIZE \ + + 2 * sizeof(uint8_t))) +#else +#define _SS_PAD1SIZE (_SS_ALIGNSIZE - sizeof(uint16_t)) +#define _SS_PAD2SIZE (_SS_MAXSIZE - (_SS_ALIGNSIZE + _SS_PAD1SIZE \ + + sizeof(uint16_t))) +#endif + +struct sockaddr_storage { +#ifdef ISC_PLATFORM_HAVESALEN + uint8_t ss_len; + uint8_t ss_family; +#else + uint16_t ss_family; +#endif + char __ss_pad1[_SS_PAD1SIZE]; + uint64_t __ss_align; /* field to force desired structure */ + char __ss_pad2[_SS_PAD2SIZE]; +}; +#endif + +#if defined(ISC_PLATFORM_HAVEIPV6) && defined(ISC_PLATFORM_NEEDIN6ADDRANY) +extern const struct in6_addr isc_net_in6addrany; +/*% + * Cope with a missing in6addr_any and in6addr_loopback. + */ +#define in6addr_any isc_net_in6addrany +#endif + +#if defined(ISC_PLATFORM_HAVEIPV6) && defined(ISC_PLATFORM_NEEDIN6ADDRLOOPBACK) +extern const struct in6_addr isc_net_in6addrloop; +#define in6addr_loopback isc_net_in6addrloop +#endif + +#ifdef ISC_PLATFORM_FIXIN6ISADDR +#undef IN6_IS_ADDR_GEOGRAPHIC +/*! + * \brief + * Fix UnixWare 7.1.1's broken IN6_IS_ADDR_* definitions. + */ +#define IN6_IS_ADDR_GEOGRAPHIC(a) (((a)->S6_un.S6_l[0] & 0xE0) == 0x80) +#undef IN6_IS_ADDR_IPX +#define IN6_IS_ADDR_IPX(a) (((a)->S6_un.S6_l[0] & 0xFE) == 0x04) +#undef IN6_IS_ADDR_LINKLOCAL +#define IN6_IS_ADDR_LINKLOCAL(a) (((a)->S6_un.S6_l[0] & 0xC0FF) == 0x80FE) +#undef IN6_IS_ADDR_MULTICAST +#define IN6_IS_ADDR_MULTICAST(a) (((a)->S6_un.S6_l[0] & 0xFF) == 0xFF) +#undef IN6_IS_ADDR_NSAP +#define IN6_IS_ADDR_NSAP(a) (((a)->S6_un.S6_l[0] & 0xFE) == 0x02) +#undef IN6_IS_ADDR_PROVIDER +#define IN6_IS_ADDR_PROVIDER(a) (((a)->S6_un.S6_l[0] & 0xE0) == 0x40) +#undef IN6_IS_ADDR_SITELOCAL +#define IN6_IS_ADDR_SITELOCAL(a) (((a)->S6_un.S6_l[0] & 0xC0FF) == 0xC0FE) +#endif /* ISC_PLATFORM_FIXIN6ISADDR */ + +#ifdef ISC_PLATFORM_NEEDPORTT +/*% + * Ensure type in_port_t is defined. + */ +typedef uint16_t in_port_t; +#endif + +#ifndef MSG_TRUNC +/*% + * If this system does not have MSG_TRUNC (as returned from recvmsg()) + * ISC_PLATFORM_RECVOVERFLOW will be defined. This will enable the MSG_TRUNC + * faking code in socket.c. + */ +#define ISC_PLATFORM_RECVOVERFLOW +#endif + +/*% IP address. */ +#define ISC__IPADDR(x) ((uint32_t)htonl((uint32_t)(x))) + +/*% Is IP address multicast? */ +#define ISC_IPADDR_ISMULTICAST(i) \ + (((uint32_t)(i) & ISC__IPADDR(0xf0000000)) \ + == ISC__IPADDR(0xe0000000)) + +#define ISC_IPADDR_ISEXPERIMENTAL(i) \ + (((uint32_t)(i) & ISC__IPADDR(0xf0000000)) \ + == ISC__IPADDR(0xf0000000)) + +/*** + *** Functions. + ***/ + +ISC_LANG_BEGINDECLS + +isc_result_t +isc_net_probeipv4(void); +/*%< + * Check if the system's kernel supports IPv4. + * + * Returns: + * + *\li #ISC_R_SUCCESS IPv4 is supported. + *\li #ISC_R_NOTFOUND IPv4 is not supported. + *\li #ISC_R_DISABLED IPv4 is disabled. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probeipv6(void); +/*%< + * Check if the system's kernel supports IPv6. + * + * Returns: + * + *\li #ISC_R_SUCCESS IPv6 is supported. + *\li #ISC_R_NOTFOUND IPv6 is not supported. + *\li #ISC_R_DISABLED IPv6 is disabled. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probe_ipv6only(void); +/*%< + * Check if the system's kernel supports the IPV6_V6ONLY socket option. + * + * Returns: + * + *\li #ISC_R_SUCCESS the option is supported for both TCP and UDP. + *\li #ISC_R_NOTFOUND IPv6 itself or the option is not supported. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probe_ipv6pktinfo(void); +/* + * Check if the system's kernel supports the IPV6_(RECV)PKTINFO socket option + * for UDP sockets. + * + * Returns: + * + * \li #ISC_R_SUCCESS the option is supported. + * \li #ISC_R_NOTFOUND IPv6 itself or the option is not supported. + * \li #ISC_R_UNEXPECTED + */ + +void +isc_net_disableipv4(void); + +void +isc_net_disableipv6(void); + +void +isc_net_enableipv4(void); + +void +isc_net_enableipv6(void); + +isc_result_t +isc_net_probeunix(void); +/* + * Returns whether UNIX domain sockets are supported. + */ + +#define ISC_NET_DSCPRECVV4 0x01 /* Can receive sent DSCP value IPv4 */ +#define ISC_NET_DSCPRECVV6 0x02 /* Can receive sent DSCP value IPv6 */ +#define ISC_NET_DSCPSETV4 0x04 /* Can set DSCP on socket IPv4 */ +#define ISC_NET_DSCPSETV6 0x08 /* Can set DSCP on socket IPv6 */ +#define ISC_NET_DSCPPKTV4 0x10 /* Can set DSCP on per packet IPv4 */ +#define ISC_NET_DSCPPKTV6 0x20 /* Can set DSCP on per packet IPv6 */ +#define ISC_NET_DSCPALL 0x3f /* All valid flags */ + +unsigned int +isc_net_probedscp(void); +/*%< + * Probe the level of DSCP support. + */ + + +isc_result_t +isc_net_getudpportrange(int af, in_port_t *low, in_port_t *high); +/*%< + * Returns system's default range of ephemeral UDP ports, if defined. + * If the range is not available or unknown, ISC_NET_PORTRANGELOW and + * ISC_NET_PORTRANGEHIGH will be returned. + * + * Requires: + * + *\li 'low' and 'high' must be non NULL. + * + * Returns: + * + *\li *low and *high will be the ports specifying the low and high ends of + * the range. + */ + +#ifdef ISC_PLATFORM_NEEDNTOP +const char * +isc_net_ntop(int af, const void *src, char *dst, size_t size); +#undef inet_ntop +#define inet_ntop isc_net_ntop +#endif + +#ifdef ISC_PLATFORM_NEEDPTON +int +isc_net_pton(int af, const char *src, void *dst); +#undef inet_pton +#define inet_pton isc_net_pton +#endif + +int +isc_net_aton(const char *cp, struct in_addr *addr); +#undef inet_aton +#define inet_aton isc_net_aton + +ISC_LANG_ENDDECLS + +#endif /* ISC_NET_H */ diff --git a/lib/isc/unix/include/isc/netdb.h b/lib/isc/unix/include/isc/netdb.h new file mode 100644 index 0000000..5aaa9f6 --- /dev/null +++ b/lib/isc/unix/include/isc/netdb.h @@ -0,0 +1,50 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_NETDB_H +#define ISC_NETDB_H 1 + +/***** + ***** Module Info + *****/ + +/*! \file + * \brief + * Portable netdb.h support. + * + * This module is responsible for defining the get<x>by<y> APIs. + * + * MP: + *\li No impact. + * + * Reliability: + *\li No anticipated impact. + * + * Resources: + *\li N/A. + * + * Security: + *\li No anticipated impact. + * + * Standards: + *\li BSD API + */ + +/*** + *** Imports. + ***/ + +#include <isc/net.h> + +#include <netdb.h> + +#endif /* ISC_NETDB_H */ diff --git a/lib/isc/unix/include/isc/offset.h b/lib/isc/unix/include/isc/offset.h new file mode 100644 index 0000000..cf57292 --- /dev/null +++ b/lib/isc/unix/include/isc/offset.h @@ -0,0 +1,26 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_OFFSET_H +#define ISC_OFFSET_H 1 + +/*! \file + * \brief + * File offsets are operating-system dependent. + */ +#include <limits.h> /* Required for CHAR_BIT. */ +#include <sys/types.h> +#include <stddef.h> /* For Linux Standard Base. */ + +typedef off_t isc_offset_t; + +#endif /* ISC_OFFSET_H */ diff --git a/lib/isc/unix/include/isc/stat.h b/lib/isc/unix/include/isc/stat.h new file mode 100644 index 0000000..24c4ce5 --- /dev/null +++ b/lib/isc/unix/include/isc/stat.h @@ -0,0 +1,46 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_STAT_H +#define ISC_STAT_H 1 + +/***** + ***** Module Info + *****/ + +/* + * Portable <sys/stat.h> support. + * + * This module is responsible for defining S_IS??? macros. + * + * MP: + * No impact. + * + * Reliability: + * No anticipated impact. + * + * Resources: + * N/A. + * + * Security: + * No anticipated impact. + * + */ + +/*** + *** Imports. + ***/ + +#include <sys/types.h> +#include <sys/stat.h> + +#endif /* ISC_STAT_H */ diff --git a/lib/isc/unix/include/isc/stdtime.h b/lib/isc/unix/include/isc/stdtime.h new file mode 100644 index 0000000..1a3da82 --- /dev/null +++ b/lib/isc/unix/include/isc/stdtime.h @@ -0,0 +1,57 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_STDTIME_H +#define ISC_STDTIME_H 1 + +/*! \file */ + +#include <isc/lang.h> +#include <inttypes.h> + +/*% + * It's public information that 'isc_stdtime_t' is an unsigned integral type. + * Applications that want maximum portability should not assume anything + * about its size. + */ +typedef uint32_t isc_stdtime_t; + +/* but this flag helps... */ +#define STDTIME_ON_32BITS 1 + +/* + * isc_stdtime32_t is a 32-bit version of isc_stdtime_t. A variable of this + * type should only be used as an opaque integer (e.g.,) to compare two + * time values. + */ +typedef uint32_t isc_stdtime32_t; + +ISC_LANG_BEGINDECLS +/* */ +void +isc_stdtime_get(isc_stdtime_t *t); +/*%< + * Set 't' to the number of seconds since 00:00:00 UTC, January 1, 1970. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +#define isc_stdtime_convert32(t, t32p) (*(t32p) = t) +/* + * Convert the standard time to its 32-bit version. + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_STDTIME_H */ diff --git a/lib/isc/unix/include/isc/strerror.h b/lib/isc/unix/include/isc/strerror.h new file mode 100644 index 0000000..fa0e429 --- /dev/null +++ b/lib/isc/unix/include/isc/strerror.h @@ -0,0 +1,38 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_STRERROR_H +#define ISC_STRERROR_H + +/*! \file */ + +#include <sys/types.h> + +#include <isc/lang.h> + +ISC_LANG_BEGINDECLS + +/*% String Error Size */ +#define ISC_STRERRORSIZE 128 + +/*% + * Provide a thread safe wrapper to strerror(). + * + * Requires: + * 'buf' to be non NULL. + */ +void +isc__strerror(int num, char *buf, size_t bufsize); + +ISC_LANG_ENDDECLS + +#endif /* ISC_STRERROR_H */ diff --git a/lib/isc/unix/include/isc/syslog.h b/lib/isc/unix/include/isc/syslog.h new file mode 100644 index 0000000..e0b8e85 --- /dev/null +++ b/lib/isc/unix/include/isc/syslog.h @@ -0,0 +1,40 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_SYSLOG_H +#define ISC_SYSLOG_H 1 + +/*! \file */ + +#include <isc/lang.h> +#include <isc/types.h> + +ISC_LANG_BEGINDECLS + +isc_result_t +isc_syslog_facilityfromstring(const char *str, int *facilityp); +/*%< + * Convert 'str' to the appropriate syslog facility constant. + * + * Requires: + * + *\li 'str' is not NULL + *\li 'facilityp' is not NULL + * + * Returns: + * \li #ISC_R_SUCCESS + * \li #ISC_R_NOTFOUND + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_SYSLOG_H */ diff --git a/lib/isc/unix/include/isc/time.h b/lib/isc/unix/include/isc/time.h new file mode 100644 index 0000000..b864c29 --- /dev/null +++ b/lib/isc/unix/include/isc/time.h @@ -0,0 +1,361 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_TIME_H +#define ISC_TIME_H 1 + +/*! \file */ + +#include <inttypes.h> +#include <stdbool.h> + +#include <isc/lang.h> +#include <isc/types.h> + +/*** + *** Intervals + ***/ + +/*! + * \brief + * The contents of this structure are private, and MUST NOT be accessed + * directly by callers. + * + * The contents are exposed only to allow callers to avoid dynamic allocation. + */ +struct isc_interval { + unsigned int seconds; + unsigned int nanoseconds; +}; + +extern const isc_interval_t * const isc_interval_zero; + +/* + * ISC_FORMATHTTPTIMESTAMP_SIZE needs to be 30 in C locale and potentially + * more for other locales to handle longer national abbreviations when + * expanding strftime's %a and %b. + */ +#define ISC_FORMATHTTPTIMESTAMP_SIZE 50 + +ISC_LANG_BEGINDECLS + +void +isc_interval_set(isc_interval_t *i, + unsigned int seconds, unsigned int nanoseconds); +/*%< + * Set 'i' to a value representing an interval of 'seconds' seconds and + * 'nanoseconds' nanoseconds, suitable for use in isc_time_add() and + * isc_time_subtract(). + * + * Requires: + * + *\li 't' is a valid pointer. + *\li nanoseconds < 1000000000. + */ + +bool +isc_interval_iszero(const isc_interval_t *i); +/*%< + * Returns true iff. 'i' is the zero interval. + * + * Requires: + * + *\li 'i' is a valid pointer. + */ + +/*** + *** Absolute Times + ***/ + +/*% + * The contents of this structure are private, and MUST NOT be accessed + * directly by callers. + * + * The contents are exposed only to allow callers to avoid dynamic allocation. + */ + +struct isc_time { + unsigned int seconds; + unsigned int nanoseconds; +}; + +extern const isc_time_t * const isc_time_epoch; + +void +isc_time_set(isc_time_t *t, unsigned int seconds, unsigned int nanoseconds); +/*%< + * Set 't' to a value which represents the given number of seconds and + * nanoseconds since 00:00:00 January 1, 1970, UTC. + * + * Notes: + *\li The Unix version of this call is equivalent to: + *\code + * isc_time_settoepoch(t); + * isc_interval_set(i, seconds, nanoseconds); + * isc_time_add(t, i, t); + *\endcode + * + * Requires: + *\li 't' is a valid pointer. + *\li nanoseconds < 1000000000. + */ + +void +isc_time_settoepoch(isc_time_t *t); +/*%< + * Set 't' to the time of the epoch. + * + * Notes: + *\li The date of the epoch is platform-dependent. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +bool +isc_time_isepoch(const isc_time_t *t); +/*%< + * Returns true iff. 't' is the epoch ("time zero"). + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +isc_result_t +isc_time_now(isc_time_t *t); +/*%< + * Set 't' to the current absolute time. + * + * Requires: + * + *\li 't' is a valid pointer. + * + * Returns: + * + *\li Success + *\li Unexpected error + * Getting the time from the system failed. + *\li Out of range + * The time from the system is too large to be represented + * in the current definition of isc_time_t. + */ + +isc_result_t +isc_time_nowplusinterval(isc_time_t *t, const isc_interval_t *i); +/*%< + * Set *t to the current absolute time + i. + * + * Note: + *\li This call is equivalent to: + * + *\code + * isc_time_now(t); + * isc_time_add(t, i, t); + *\endcode + * + * Requires: + * + *\li 't' and 'i' are valid pointers. + * + * Returns: + * + *\li Success + *\li Unexpected error + * Getting the time from the system failed. + *\li Out of range + * The interval added to the time from the system is too large to + * be represented in the current definition of isc_time_t. + */ + +int +isc_time_compare(const isc_time_t *t1, const isc_time_t *t2); +/*%< + * Compare the times referenced by 't1' and 't2' + * + * Requires: + * + *\li 't1' and 't2' are valid pointers. + * + * Returns: + * + *\li -1 t1 < t2 (comparing times, not pointers) + *\li 0 t1 = t2 + *\li 1 t1 > t2 + */ + +isc_result_t +isc_time_add(const isc_time_t *t, const isc_interval_t *i, isc_time_t *result); +/*%< + * Add 'i' to 't', storing the result in 'result'. + * + * Requires: + * + *\li 't', 'i', and 'result' are valid pointers. + * + * Returns: + *\li Success + *\li Out of range + * The interval added to the time is too large to + * be represented in the current definition of isc_time_t. + */ + +isc_result_t +isc_time_subtract(const isc_time_t *t, const isc_interval_t *i, + isc_time_t *result); +/*%< + * Subtract 'i' from 't', storing the result in 'result'. + * + * Requires: + * + *\li 't', 'i', and 'result' are valid pointers. + * + * Returns: + *\li Success + *\li Out of range + * The interval is larger than the time since the epoch. + */ + +uint64_t +isc_time_microdiff(const isc_time_t *t1, const isc_time_t *t2); +/*%< + * Find the difference in microseconds between time t1 and time t2. + * t2 is the subtrahend of t1; ie, difference = t1 - t2. + * + * Requires: + * + *\li 't1' and 't2' are valid pointers. + * + * Returns: + *\li The difference of t1 - t2, or 0 if t1 <= t2. + */ + +uint32_t +isc_time_seconds(const isc_time_t *t); +/*%< + * Return the number of seconds since the epoch stored in a time structure. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +isc_result_t +isc_time_secondsastimet(const isc_time_t *t, time_t *secondsp); +/*%< + * Ensure the number of seconds in an isc_time_t is representable by a time_t. + * + * Notes: + *\li The number of seconds stored in an isc_time_t might be larger + * than the number of seconds a time_t is able to handle. Since + * time_t is mostly opaque according to the ANSI/ISO standard + * (essentially, all you can be sure of is that it is an arithmetic type, + * not even necessarily integral), it can be tricky to ensure that + * the isc_time_t is in the range a time_t can handle. Use this + * function in place of isc_time_seconds() any time you need to set a + * time_t from an isc_time_t. + * + * Requires: + *\li 't' is a valid pointer. + * + * Returns: + *\li Success + *\li Out of range + */ + +uint32_t +isc_time_nanoseconds(const isc_time_t *t); +/*%< + * Return the number of nanoseconds stored in a time structure. + * + * Notes: + *\li This is the number of nanoseconds in excess of the number + * of seconds since the epoch; it will always be less than one + * full second. + * + * Requires: + *\li 't' is a valid pointer. + * + * Ensures: + *\li The returned value is less than 1*10^9. + */ + +void +isc_time_formattimestamp(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using a format like "30-Aug-2000 04:06:47.997" and the local time zone. + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formathttptimestamp(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using a format like "Mon, 30 Aug 2000 04:06:47 GMT" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +isc_result_t +isc_time_parsehttptimestamp(char *input, isc_time_t *t); +/*%< + * Parse the time in 'input' into the isc_time_t pointed to by 't', + * expecting a format like "Mon, 30 Aug 2000 04:06:47 GMT" + * + * Requires: + *\li 'buf' and 't' are not NULL. + */ + +void +isc_time_formatISO8601(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ssZ" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601ms(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss.sssZ" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_TIME_H */ diff --git a/lib/isc/unix/include/pkcs11/Makefile.in b/lib/isc/unix/include/pkcs11/Makefile.in new file mode 100644 index 0000000..420fec1 --- /dev/null +++ b/lib/isc/unix/include/pkcs11/Makefile.in @@ -0,0 +1,31 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +HEADERS = cryptoki.h +SUBDIRS = +TARGETS = + +@BIND9_MAKE_RULES@ + +installdirs: + $(SHELL) ${top_srcdir}/mkinstalldirs ${DESTDIR}${includedir}/pkcs11 + +install:: installdirs + for i in ${HEADERS}; do \ + ${INSTALL_DATA} $(srcdir)/$$i ${DESTDIR}${includedir}/pkcs11 || exit 1; \ + done + +uninstall:: + for i in ${HEADERS}; do \ + rm -f ${DESTDIR}${includedir}/pkcs11/$$i || exit 1; \ + done diff --git a/lib/isc/unix/include/pkcs11/cryptoki.h b/lib/isc/unix/include/pkcs11/cryptoki.h new file mode 100644 index 0000000..7dc48b0 --- /dev/null +++ b/lib/isc/unix/include/pkcs11/cryptoki.h @@ -0,0 +1,66 @@ +/* cryptoki.h include file for PKCS #11. */ +/* + * Copyright (C) 2009 Internet Systems Consortium, Inc. ("ISC") + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND ISC AND NETWORK ASSOCIATES DISCLAIMS + * ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE + * FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR + * IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ +/* $Revision: 1.3 $ */ + +/* + * Portions Copyright RSA Security Inc. + * + * License to copy and use this software is granted provided that it is + * identified as "RSA Security Inc. PKCS #11 Cryptographic Token Interface + * (Cryptoki)" in all material mentioning or referencing this software. + + * License is also granted to make and use derivative works provided that + * such works are identified as "derived from the RSA Security Inc. PKCS #11 + * Cryptographic Token Interface (Cryptoki)" in all material mentioning or + * referencing the derived work. + + * RSA Security Inc. makes no representations concerning either the + * merchantability of this software or the suitability of this software for + * any particular purpose. It is provided "as is" without express or implied + * warranty of any kind. + */ + +/* This is a sample file containing the top level include directives + * for building Unix Cryptoki libraries and applications. + */ + +#ifndef ___CRYPTOKI_H_INC___ +#define ___CRYPTOKI_H_INC___ + +#define CK_PTR * + +#define CK_DEFINE_FUNCTION(returnType, name) \ + returnType name + +#define CK_DECLARE_FUNCTION(returnType, name) \ + returnType name + +#define CK_DECLARE_FUNCTION_POINTER(returnType, name) \ + returnType (* name) + +#define CK_CALLBACK_FUNCTION(returnType, name) \ + returnType (* name) + +/* NULL is in unistd.h */ +#include <unistd.h> +#define NULL_PTR NULL + +#undef CK_PKCS11_FUNCTION_INFO + +#include <pkcs11/pkcs11.h> + +#endif /* ___CRYPTOKI_H_INC___ */ diff --git a/lib/isc/unix/interfaceiter.c b/lib/isc/unix/interfaceiter.c new file mode 100644 index 0000000..4df4d34 --- /dev/null +++ b/lib/isc/unix/interfaceiter.c @@ -0,0 +1,306 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <sys/types.h> +#include <sys/ioctl.h> +#ifdef HAVE_SYS_SOCKIO_H +#include <sys/sockio.h> /* Required for ifiter_ioctl.c. */ +#endif + +#include <stdio.h> +#include <inttypes.h> +#include <stdlib.h> +#include <unistd.h> +#include <errno.h> + +#include <isc/interfaceiter.h> +#include <isc/log.h> +#include <isc/magic.h> +#include <isc/mem.h> +#include <isc/msgs.h> +#include <isc/net.h> +#include <isc/print.h> +#include <isc/result.h> +#include <isc/strerror.h> +#include <isc/string.h> +#include <isc/types.h> +#include <isc/util.h> + +/* Must follow <isc/net.h>. */ +#ifdef HAVE_NET_IF6_H +#include <net/if6.h> +#endif +#include <net/if.h> + +/* Common utility functions */ + +/*% + * Extract the network address part from a "struct sockaddr". + * \brief + * The address family is given explicitly + * instead of using src->sa_family, because the latter does not work + * for copying a network mask obtained by SIOCGIFNETMASK (it does + * not have a valid address family). + */ + +static void +get_addr(unsigned int family, isc_netaddr_t *dst, struct sockaddr *src, + char *ifname) +{ + struct sockaddr_in6 *sa6; + +#if !defined(ISC_PLATFORM_HAVEIFNAMETOINDEX) || \ + !defined(ISC_PLATFORM_HAVESCOPEID) + UNUSED(ifname); +#endif + + /* clear any remaining value for safety */ + memset(dst, 0, sizeof(*dst)); + + dst->family = family; + switch (family) { + case AF_INET: + memmove(&dst->type.in, + &((struct sockaddr_in *) src)->sin_addr, + sizeof(struct in_addr)); + break; + case AF_INET6: + sa6 = (struct sockaddr_in6 *)src; + memmove(&dst->type.in6, &sa6->sin6_addr, + sizeof(struct in6_addr)); +#ifdef ISC_PLATFORM_HAVESCOPEID + if (sa6->sin6_scope_id != 0) + isc_netaddr_setzone(dst, sa6->sin6_scope_id); + else { + /* + * BSD variants embed scope zone IDs in the 128bit + * address as a kernel internal form. Unfortunately, + * the embedded IDs are not hidden from applications + * when getting access to them by sysctl or ioctl. + * We convert the internal format to the pure address + * part and the zone ID part. + * Since multicast addresses should not appear here + * and they cannot be distinguished from netmasks, + * we only consider unicast link-local addresses. + */ + if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) { + uint16_t zone16; + + memmove(&zone16, &sa6->sin6_addr.s6_addr[2], + sizeof(zone16)); + zone16 = ntohs(zone16); + if (zone16 != 0) { + /* the zone ID is embedded */ + isc_netaddr_setzone(dst, + (uint32_t)zone16); + dst->type.in6.s6_addr[2] = 0; + dst->type.in6.s6_addr[3] = 0; +#ifdef ISC_PLATFORM_HAVEIFNAMETOINDEX + } else if (ifname != NULL) { + unsigned int zone; + + /* + * sin6_scope_id is still not provided, + * but the corresponding interface name + * is know. Use the interface ID as + * the link ID. + */ + zone = if_nametoindex(ifname); + if (zone != 0) { + isc_netaddr_setzone(dst, + (uint32_t)zone); + } +#endif + } + } + } +#endif + break; + default: + INSIST(0); + break; + } +} + +/* + * Include system-dependent code. + */ + +#ifdef __linux +#define ISC_IF_INET6_SZ \ + sizeof("00000000000000000000000000000001 01 80 10 80 XXXXXXloXXXXXXXX\n") +static isc_result_t linux_if_inet6_next(isc_interfaceiter_t *); +static isc_result_t linux_if_inet6_current(isc_interfaceiter_t *); +static void linux_if_inet6_first(isc_interfaceiter_t *iter); +#endif + +#if HAVE_GETIFADDRS +#include "ifiter_getifaddrs.c" +#elif HAVE_IFLIST_SYSCTL +#include "ifiter_sysctl.c" +#else +#include "ifiter_ioctl.c" +#endif + +#ifdef __linux +static void +linux_if_inet6_first(isc_interfaceiter_t *iter) { + if (iter->proc != NULL) { + rewind(iter->proc); + (void)linux_if_inet6_next(iter); + } else + iter->valid = ISC_R_NOMORE; +} + +static isc_result_t +linux_if_inet6_next(isc_interfaceiter_t *iter) { + if (iter->proc != NULL && + fgets(iter->entry, sizeof(iter->entry), iter->proc) != NULL) + iter->valid = ISC_R_SUCCESS; + else + iter->valid = ISC_R_NOMORE; + return (iter->valid); +} + +static isc_result_t +linux_if_inet6_current(isc_interfaceiter_t *iter) { + char address[33]; + char name[IF_NAMESIZE+1]; + struct in6_addr addr6; + unsigned int ifindex, prefix, flag3, flag4; + int res; + unsigned int i; + + if (iter->valid != ISC_R_SUCCESS) + return (iter->valid); + if (iter->proc == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:iter->proc == NULL"); + return (ISC_R_FAILURE); + } + + res = sscanf(iter->entry, "%32[a-f0-9] %x %x %x %x %16s\n", + address, &ifindex, &prefix, &flag3, &flag4, name); + if (res != 6) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:sscanf() -> %d (expected 6)", + res); + return (ISC_R_FAILURE); + } + if (strlen(address) != 32) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:strlen(%s) != 32", address); + return (ISC_R_FAILURE); + } + for (i = 0; i < 16; i++) { + unsigned char byte; + static const char hex[] = "0123456789abcdef"; + byte = ((strchr(hex, address[i * 2]) - hex) << 4) | + (strchr(hex, address[i * 2 + 1]) - hex); + addr6.s6_addr[i] = byte; + } + iter->current.af = AF_INET6; + iter->current.flags = INTERFACE_F_UP; + isc_netaddr_fromin6(&iter->current.address, &addr6); + if (isc_netaddr_islinklocal(&iter->current.address)) { + isc_netaddr_setzone(&iter->current.address, + (uint32_t)ifindex); + } + for (i = 0; i < 16; i++) { + if (prefix > 8) { + addr6.s6_addr[i] = 0xff; + prefix -= 8; + } else { + addr6.s6_addr[i] = (0xff << (8 - prefix)) & 0xff; + prefix = 0; + } + } + isc_netaddr_fromin6(&iter->current.netmask, &addr6); + strlcpy(iter->current.name, name, sizeof(iter->current.name)); + return (ISC_R_SUCCESS); +} +#endif + +/* + * The remaining code is common to the sysctl and ioctl case. + */ + +isc_result_t +isc_interfaceiter_current(isc_interfaceiter_t *iter, + isc_interface_t *ifdata) +{ + REQUIRE(iter->result == ISC_R_SUCCESS); + memmove(ifdata, &iter->current, sizeof(*ifdata)); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_interfaceiter_first(isc_interfaceiter_t *iter) { + isc_result_t result; + + REQUIRE(VALID_IFITER(iter)); + + internal_first(iter); + for (;;) { + result = internal_current(iter); + if (result != ISC_R_IGNORE) + break; + result = internal_next(iter); + if (result != ISC_R_SUCCESS) + break; + } + iter->result = result; + return (result); +} + +isc_result_t +isc_interfaceiter_next(isc_interfaceiter_t *iter) { + isc_result_t result; + + REQUIRE(VALID_IFITER(iter)); + REQUIRE(iter->result == ISC_R_SUCCESS); + + for (;;) { + result = internal_next(iter); + if (result != ISC_R_SUCCESS) + break; + result = internal_current(iter); + if (result != ISC_R_IGNORE) + break; + } + iter->result = result; + return (result); +} + +void +isc_interfaceiter_destroy(isc_interfaceiter_t **iterp) +{ + isc_interfaceiter_t *iter; + REQUIRE(iterp != NULL); + iter = *iterp; + REQUIRE(VALID_IFITER(iter)); + + internal_destroy(iter); + if (iter->buf != NULL) + isc_mem_put(iter->mctx, iter->buf, iter->bufsize); + + iter->magic = 0; + isc_mem_put(iter->mctx, iter, sizeof(*iter)); + *iterp = NULL; +} diff --git a/lib/isc/unix/ipv6.c b/lib/isc/unix/ipv6.c new file mode 100644 index 0000000..eb77377 --- /dev/null +++ b/lib/isc/unix/ipv6.c @@ -0,0 +1,20 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <isc/ipv6.h> + +const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT; +const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT; diff --git a/lib/isc/unix/keyboard.c b/lib/isc/unix/keyboard.c new file mode 100644 index 0000000..cb5f116 --- /dev/null +++ b/lib/isc/unix/keyboard.c @@ -0,0 +1,121 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <stdbool.h> + +#include <sys/param.h> +#include <sys/types.h> +#include <sys/time.h> +#include <sys/uio.h> + +#include <errno.h> +#include <stdlib.h> +#include <string.h> +#include <termios.h> +#include <unistd.h> +#include <fcntl.h> + +#include <isc/keyboard.h> +#include <isc/util.h> + +isc_result_t +isc_keyboard_open(isc_keyboard_t *keyboard) { + int fd; + isc_result_t ret; + struct termios current_mode; + + REQUIRE(keyboard != NULL); + + fd = open("/dev/tty", O_RDONLY, 0); + if (fd < 0) + return (ISC_R_IOERROR); + + keyboard->fd = fd; + + if (tcgetattr(fd, &keyboard->saved_mode) < 0) { + ret = ISC_R_IOERROR; + goto errout; + } + + current_mode = keyboard->saved_mode; + + current_mode.c_iflag &= + ~(IGNBRK|BRKINT|PARMRK|ISTRIP|INLCR|IGNCR|ICRNL|IXON); + current_mode.c_oflag &= ~OPOST; + current_mode.c_lflag &= ~(ECHO|ECHONL|ICANON|ISIG|IEXTEN); + current_mode.c_cflag &= ~(CSIZE|PARENB); + current_mode.c_cflag |= CS8; + + current_mode.c_cc[VMIN] = 1; + current_mode.c_cc[VTIME] = 0; + if (tcsetattr(fd, TCSAFLUSH, ¤t_mode) < 0) { + ret = ISC_R_IOERROR; + goto errout; + } + + keyboard->result = ISC_R_SUCCESS; + + return (ISC_R_SUCCESS); + + errout: + close (fd); + + return (ret); +} + +isc_result_t +isc_keyboard_close(isc_keyboard_t *keyboard, unsigned int sleeptime) { + REQUIRE(keyboard != NULL); + + if (sleeptime > 0 && keyboard->result != ISC_R_CANCELED) + (void)sleep(sleeptime); + + (void)tcsetattr(keyboard->fd, TCSAFLUSH, &keyboard->saved_mode); + (void)close(keyboard->fd); + + keyboard->fd = -1; + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_keyboard_getchar(isc_keyboard_t *keyboard, unsigned char *cp) { + ssize_t cc; + unsigned char c; + cc_t *controlchars; + + REQUIRE(keyboard != NULL); + REQUIRE(cp != NULL); + + cc = read(keyboard->fd, &c, 1); + if (cc < 0) { + keyboard->result = ISC_R_IOERROR; + return (keyboard->result); + } + + controlchars = keyboard->saved_mode.c_cc; + if (c == controlchars[VINTR] || c == controlchars[VQUIT]) { + keyboard->result = ISC_R_CANCELED; + return (keyboard->result); + } + + *cp = c; + + return (ISC_R_SUCCESS); +} + +bool +isc_keyboard_canceled(isc_keyboard_t *keyboard) { + return (keyboard->result == ISC_R_CANCELED); +} diff --git a/lib/isc/unix/meminfo.c b/lib/isc/unix/meminfo.c new file mode 100644 index 0000000..ad750a7 --- /dev/null +++ b/lib/isc/unix/meminfo.c @@ -0,0 +1,40 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <config.h> + +#include <isc/meminfo.h> +#include <inttypes.h> +#include <unistd.h> +#ifdef HAVE_SYS_SYSCTL_H +#include <sys/sysctl.h> +#endif + +uint64_t +isc_meminfo_totalphys(void) { +#if defined(CTL_HW) && (defined(HW_PHYSMEM64) || defined(HW_MEMSIZE)) + int mib[2]; + mib[0] = CTL_HW; +#if defined(HW_MEMSIZE) + mib[1] = HW_MEMSIZE; +#elif defined(HW_PHYSMEM64) + mib[1] = HW_PHYSMEM64; +#endif + uint64_t size = 0; + size_t len = sizeof(size); + if (sysctl(mib, 2, &size, &len, NULL, 0) == 0) + return (size); +#endif +#if defined(_SC_PHYS_PAGES) && defined(_SC_PAGESIZE) + return ((size_t) (sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE))); +#endif + return (0); +} diff --git a/lib/isc/unix/net.c b/lib/isc/unix/net.c new file mode 100644 index 0000000..efe17bd --- /dev/null +++ b/lib/isc/unix/net.c @@ -0,0 +1,913 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <sys/types.h> + +#if defined(HAVE_SYS_SYSCTL_H) +#if defined(HAVE_SYS_PARAM_H) +#include <sys/param.h> +#endif +#include <sys/sysctl.h> +#endif +#include <sys/uio.h> + +#include <errno.h> +#include <stdbool.h> +#include <unistd.h> +#include <fcntl.h> + +#include <isc/log.h> +#include <isc/msgs.h> +#include <isc/net.h> +#include <isc/netdb.h> +#include <isc/once.h> +#include <isc/strerror.h> +#include <isc/string.h> +#include <isc/util.h> + +#ifndef ISC_SOCKADDR_LEN_T +#define ISC_SOCKADDR_LEN_T unsigned int +#endif + +/*% + * Definitions about UDP port range specification. This is a total mess of + * portability variants: some use sysctl (but the sysctl names vary), some use + * system-specific interfaces, some have the same interface for IPv4 and IPv6, + * some separate them, etc... + */ + +/*% + * The last resort defaults: use all non well known port space + */ +#ifndef ISC_NET_PORTRANGELOW +#define ISC_NET_PORTRANGELOW 1024 +#endif /* ISC_NET_PORTRANGELOW */ +#ifndef ISC_NET_PORTRANGEHIGH +#define ISC_NET_PORTRANGEHIGH 65535 +#endif /* ISC_NET_PORTRANGEHIGH */ + +#ifdef HAVE_SYSCTLBYNAME + +/*% + * sysctl variants + */ +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__DragonFly__) +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW "net.inet.ip.portrange.hifirst" +#define SYSCTL_V4PORTRANGE_HIGH "net.inet.ip.portrange.hilast" +#define SYSCTL_V6PORTRANGE_LOW "net.inet.ip.portrange.hifirst" +#define SYSCTL_V6PORTRANGE_HIGH "net.inet.ip.portrange.hilast" +#endif + +#ifdef __NetBSD__ +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW "net.inet.ip.anonportmin" +#define SYSCTL_V4PORTRANGE_HIGH "net.inet.ip.anonportmax" +#define SYSCTL_V6PORTRANGE_LOW "net.inet6.ip6.anonportmin" +#define SYSCTL_V6PORTRANGE_HIGH "net.inet6.ip6.anonportmax" +#endif + +#else /* !HAVE_SYSCTLBYNAME */ + +#ifdef __OpenBSD__ +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW { CTL_NET, PF_INET, IPPROTO_IP, \ + IPCTL_IPPORT_HIFIRSTAUTO } +#define SYSCTL_V4PORTRANGE_HIGH { CTL_NET, PF_INET, IPPROTO_IP, \ + IPCTL_IPPORT_HILASTAUTO } +/* Same for IPv6 */ +#define SYSCTL_V6PORTRANGE_LOW SYSCTL_V4PORTRANGE_LOW +#define SYSCTL_V6PORTRANGE_HIGH SYSCTL_V4PORTRANGE_HIGH +#endif + +#endif /* HAVE_SYSCTLBYNAME */ + +#if defined(ISC_PLATFORM_HAVEIPV6) +# if defined(ISC_PLATFORM_NEEDIN6ADDRANY) +const struct in6_addr isc_net_in6addrany = IN6ADDR_ANY_INIT; +# endif + +# if defined(ISC_PLATFORM_NEEDIN6ADDRLOOPBACK) +const struct in6_addr isc_net_in6addrloop = IN6ADDR_LOOPBACK_INIT; +# endif + +# if defined(WANT_IPV6) +static isc_once_t once_ipv6only = ISC_ONCE_INIT; + +# if defined(ISC_PLATFORM_HAVEIN6PKTINFO) +static isc_once_t once_ipv6pktinfo = ISC_ONCE_INIT; +# endif +# endif /* WANT_IPV6 */ +#endif /* ISC_PLATFORM_HAVEIPV6 */ + +#ifndef ISC_CMSG_IP_TOS +#ifdef __APPLE__ +#define ISC_CMSG_IP_TOS 0 /* As of 10.8.2. */ +#else /* ! __APPLE__ */ +#define ISC_CMSG_IP_TOS 1 +#endif /* ! __APPLE__ */ +#endif /* ! ISC_CMSG_IP_TOS */ + +static isc_once_t once = ISC_ONCE_INIT; +static isc_once_t once_dscp = ISC_ONCE_INIT; + +static isc_result_t ipv4_result = ISC_R_NOTFOUND; +static isc_result_t ipv6_result = ISC_R_NOTFOUND; +static isc_result_t unix_result = ISC_R_NOTFOUND; +static isc_result_t ipv6only_result = ISC_R_NOTFOUND; +static isc_result_t ipv6pktinfo_result = ISC_R_NOTFOUND; +static unsigned int dscp_result = 0; + +static isc_result_t +try_proto(int domain) { + int s; + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + + s = socket(domain, SOCK_STREAM, 0); + if (s == -1) { + switch (errno) { +#ifdef EAFNOSUPPORT + case EAFNOSUPPORT: +#endif +#ifdef EPFNOSUPPORT + case EPFNOSUPPORT: +#endif +#ifdef EPROTONOSUPPORT + case EPROTONOSUPPORT: +#endif +#ifdef EINVAL + case EINVAL: +#endif + return (ISC_R_NOTFOUND); + default: + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "socket() %s: %s", + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + return (ISC_R_UNEXPECTED); + } + } + +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef WANT_IPV6 +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO + if (domain == PF_INET6) { + struct sockaddr_in6 sin6; + unsigned int len; + + /* + * Check to see if IPv6 is broken, as is common on Linux. + */ + len = sizeof(sin6); + if (getsockname(s, (struct sockaddr *)&sin6, (void *)&len) < 0) + { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "retrieving the address of an IPv6 " + "socket from the kernel failed."); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "IPv6 is not supported."); + result = ISC_R_NOTFOUND; + } else { + if (len == sizeof(struct sockaddr_in6)) + result = ISC_R_SUCCESS; + else { + isc_log_write(isc_lctx, + ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_ERROR, + "IPv6 structures in kernel and " + "user space do not match."); + isc_log_write(isc_lctx, + ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_ERROR, + "IPv6 is not supported."); + result = ISC_R_NOTFOUND; + } + } + } +#endif +#endif +#endif + + (void)close(s); + + return (result); +} + +static void +initialize_action(void) { + ipv4_result = try_proto(PF_INET); +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef WANT_IPV6 +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO + ipv6_result = try_proto(PF_INET6); +#endif +#endif +#endif +#ifdef ISC_PLATFORM_HAVESYSUNH + unix_result = try_proto(PF_UNIX); +#endif +} + +static void +initialize(void) { + RUNTIME_CHECK(isc_once_do(&once, initialize_action) == ISC_R_SUCCESS); +} + +isc_result_t +isc_net_probeipv4(void) { + initialize(); + return (ipv4_result); +} + +isc_result_t +isc_net_probeipv6(void) { + initialize(); + return (ipv6_result); +} + +isc_result_t +isc_net_probeunix(void) { + initialize(); + return (unix_result); +} + +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef WANT_IPV6 +static void +try_ipv6only(void) { +#ifdef IPV6_V6ONLY + int s, on; + char strbuf[ISC_STRERRORSIZE]; +#endif + isc_result_t result; + + result = isc_net_probeipv6(); + if (result != ISC_R_SUCCESS) { + ipv6only_result = result; + return; + } + +#ifndef IPV6_V6ONLY + ipv6only_result = ISC_R_NOTFOUND; + return; +#else + /* check for TCP sockets */ + s = socket(PF_INET6, SOCK_STREAM, 0); + if (s == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "socket() %s: %s", + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + ipv6only_result = ISC_R_UNEXPECTED; + return; + } + + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) < 0) { + ipv6only_result = ISC_R_NOTFOUND; + goto close; + } + + close(s); + + /* check for UDP sockets */ + s = socket(PF_INET6, SOCK_DGRAM, 0); + if (s == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "socket() %s: %s", + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + ipv6only_result = ISC_R_UNEXPECTED; + return; + } + + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) < 0) { + ipv6only_result = ISC_R_NOTFOUND; + goto close; + } + + ipv6only_result = ISC_R_SUCCESS; + +close: + close(s); + return; +#endif /* IPV6_V6ONLY */ +} + +static void +initialize_ipv6only(void) { + RUNTIME_CHECK(isc_once_do(&once_ipv6only, + try_ipv6only) == ISC_R_SUCCESS); +} +#endif /* WANT_IPV6 */ + +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO +#ifdef WANT_IPV6 +static void +try_ipv6pktinfo(void) { + int s, on; + char strbuf[ISC_STRERRORSIZE]; + isc_result_t result; + int optname; + + result = isc_net_probeipv6(); + if (result != ISC_R_SUCCESS) { + ipv6pktinfo_result = result; + return; + } + + /* we only use this for UDP sockets */ + s = socket(PF_INET6, SOCK_DGRAM, IPPROTO_UDP); + if (s == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "socket() %s: %s", + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + ipv6pktinfo_result = ISC_R_UNEXPECTED; + return; + } + +#ifdef IPV6_RECVPKTINFO + optname = IPV6_RECVPKTINFO; +#else + optname = IPV6_PKTINFO; +#endif + on = 1; + if (setsockopt(s, IPPROTO_IPV6, optname, &on, sizeof(on)) < 0) { + ipv6pktinfo_result = ISC_R_NOTFOUND; + goto close; + } + + ipv6pktinfo_result = ISC_R_SUCCESS; + +close: + close(s); + return; +} + +static void +initialize_ipv6pktinfo(void) { + RUNTIME_CHECK(isc_once_do(&once_ipv6pktinfo, + try_ipv6pktinfo) == ISC_R_SUCCESS); +} +#endif /* WANT_IPV6 */ +#endif /* ISC_PLATFORM_HAVEIN6PKTINFO */ +#endif /* ISC_PLATFORM_HAVEIPV6 */ + +isc_result_t +isc_net_probe_ipv6only(void) { +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef WANT_IPV6 + initialize_ipv6only(); +#else + ipv6only_result = ISC_R_NOTFOUND; +#endif +#endif + return (ipv6only_result); +} + +isc_result_t +isc_net_probe_ipv6pktinfo(void) { +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO +#ifdef WANT_IPV6 + initialize_ipv6pktinfo(); +#else + ipv6pktinfo_result = ISC_R_NOTFOUND; +#endif +#endif +#endif + return (ipv6pktinfo_result); +} + +#if ISC_CMSG_IP_TOS || \ + defined(ISC_NET_BSD44MSGHDR) && defined(IPV6_TCLASS) && defined(WANT_IPV6) + +static inline ISC_SOCKADDR_LEN_T +cmsg_len(ISC_SOCKADDR_LEN_T len) { +#ifdef CMSG_LEN + return (CMSG_LEN(len)); +#else + ISC_SOCKADDR_LEN_T hdrlen; + + /* + * Cast NULL so that any pointer arithmetic performed by CMSG_DATA + * is correct. + */ + hdrlen = (ISC_SOCKADDR_LEN_T)CMSG_DATA(((struct cmsghdr *)NULL)); + return (hdrlen + len); +#endif +} + +static inline ISC_SOCKADDR_LEN_T +cmsg_space(ISC_SOCKADDR_LEN_T len) { +#ifdef CMSG_SPACE + return (CMSG_SPACE(len)); +#else + struct msghdr msg; + struct cmsghdr *cmsgp; + /* + * XXX: The buffer length is an ad-hoc value, but should be enough + * in a practical sense. + */ + char dummybuf[sizeof(struct cmsghdr) + 1024]; + + memset(&msg, 0, sizeof(msg)); + msg.msg_control = dummybuf; + msg.msg_controllen = sizeof(dummybuf); + + cmsgp = (struct cmsghdr *)dummybuf; + cmsgp->cmsg_len = cmsg_len(len); + + cmsgp = CMSG_NXTHDR(&msg, cmsgp); + if (cmsgp != NULL) + return ((char *)cmsgp - (char *)msg.msg_control); + else + return (0); +#endif +} + +#ifdef ISC_NET_BSD44MSGHDR +/* + * Make a fd non-blocking. + */ +static isc_result_t +make_nonblock(int fd) { + int ret; + int flags; + char strbuf[ISC_STRERRORSIZE]; +#ifdef USE_FIONBIO_IOCTL + int on = 1; + + ret = ioctl(fd, FIONBIO, (char *)&on); +#else + flags = fcntl(fd, F_GETFL, 0); + flags |= PORT_NONBLOCK; + ret = fcntl(fd, F_SETFL, flags); +#endif + + if (ret == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, +#ifdef USE_FIONBIO_IOCTL + "ioctl(%d, FIONBIO, &on): %s", fd, +#else + "fcntl(%d, F_SETFL, %d): %s", fd, flags, +#endif + strbuf); + + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +static bool +cmsgsend(int s, int level, int type, struct addrinfo *res) { + char strbuf[ISC_STRERRORSIZE]; + struct sockaddr_storage ss; + ISC_SOCKADDR_LEN_T len = sizeof(ss); + struct msghdr msg; + union { + struct cmsghdr h; + unsigned char b[256]; + } control; + struct cmsghdr *cmsgp; + int dscp = (46 << 2); /* Expedited forwarding. */ + struct iovec iovec; + char buf[1] = { 0 }; + isc_result_t result; + + if (bind(s, res->ai_addr, res->ai_addrlen) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "bind: %s", strbuf); + return (false); + } + + if (getsockname(s, (struct sockaddr *)&ss, &len) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getsockname: %s", strbuf); + return (false); + } + + iovec.iov_base = buf; + iovec.iov_len = sizeof(buf); + + memset(&msg, 0, sizeof(msg)); + msg.msg_name = (struct sockaddr *)&ss; + msg.msg_namelen = len; + msg.msg_iov = &iovec; + msg.msg_iovlen = 1; + msg.msg_control = (void*)&control; + msg.msg_controllen = 0; + msg.msg_flags = 0; + + cmsgp = msg.msg_control; + + switch (type) { +#ifdef IP_TOS + case IP_TOS: + memset(cmsgp, 0, cmsg_space(sizeof(char))); + cmsgp->cmsg_level = level; + cmsgp->cmsg_type = type; + cmsgp->cmsg_len = cmsg_len(sizeof(char)); + *(unsigned char*)CMSG_DATA(cmsgp) = dscp; + msg.msg_controllen += cmsg_space(sizeof(char)); + break; +#endif +#ifdef IPV6_TCLASS + case IPV6_TCLASS: + memset(cmsgp, 0, cmsg_space(sizeof(dscp))); + cmsgp->cmsg_level = level; + cmsgp->cmsg_type = type; + cmsgp->cmsg_len = cmsg_len(sizeof(dscp)); + memmove(CMSG_DATA(cmsgp), &dscp, sizeof(dscp)); + msg.msg_controllen += cmsg_space(sizeof(dscp)); + break; +#endif + default: + INSIST(0); + } + + if (sendmsg(s, &msg, 0) < 0) { + int debug = ISC_LOG_DEBUG(10); + const char *typestr; + const char *msgstr; + switch (errno) { +#ifdef ENOPROTOOPT + case ENOPROTOOPT: +#endif +#ifdef EOPNOTSUPP + case EOPNOTSUPP: +#endif + case EINVAL: + case EPERM: + break; + default: + debug = ISC_LOG_NOTICE; + } + isc__strerror(errno, strbuf, sizeof(strbuf)); + if (debug != ISC_LOG_NOTICE) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "sendmsg: %s", strbuf); + } else { + typestr = (type == IP_TOS) ? "IP_TOS" : "IPV6_TCLASS"; + msgstr = isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"); + UNEXPECTED_ERROR(__FILE__, __LINE__, "probing " + "sendmsg() with %s=%02x %s: %s", + typestr, dscp, msgstr, strbuf); + } + return (false); + } + + /* + * Make sure the message actually got sent. + */ + result = make_nonblock(s); + RUNTIME_CHECK(result == ISC_R_SUCCESS); + + iovec.iov_base = buf; + iovec.iov_len = sizeof(buf); + + memset(&msg, 0, sizeof(msg)); + msg.msg_name = (struct sockaddr *)&ss; + msg.msg_namelen = sizeof(ss); + msg.msg_iov = &iovec; + msg.msg_iovlen = 1; + msg.msg_control = NULL; + msg.msg_controllen = 0; + msg.msg_flags = 0; + + if (recvmsg(s, &msg, 0) < 0) + return (false); + + return (true); +} +#endif +#endif + +static void +try_dscp_v4(void) { +#ifdef IP_TOS + char strbuf[ISC_STRERRORSIZE]; + struct addrinfo hints, *res0; + int s, dscp = 0, n; +#ifdef IP_RECVTOS + int on = 1; +#endif /* IP_RECVTOS */ + + memset(&hints, 0, sizeof(hints)); + hints.ai_family = AF_INET; + hints.ai_socktype = SOCK_DGRAM; + hints.ai_protocol = IPPROTO_UDP; +#ifdef AI_NUMERICHOST + hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST; +#else + hints.ai_flags = AI_PASSIVE; +#endif + + n = getaddrinfo("127.0.0.1", NULL, &hints, &res0); + if (n != 0 || res0 == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getaddrinfo(127.0.0.1): %s", gai_strerror(n)); + return; + } + + s = socket(res0->ai_family, res0->ai_socktype, res0->ai_protocol); + + if (s == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "socket: %s", strbuf); + freeaddrinfo(res0); + return; + } + + if (setsockopt(s, IPPROTO_IP, IP_TOS, &dscp, sizeof(dscp)) == 0) + dscp_result |= ISC_NET_DSCPSETV4; + +#ifdef IP_RECVTOS + on = 1; + if (setsockopt(s, IPPROTO_IP, IP_RECVTOS, &on, sizeof(on)) == 0) + dscp_result |= ISC_NET_DSCPRECVV4; +#endif /* IP_RECVTOS */ + +#ifdef ISC_NET_BSD44MSGHDR + +#if ISC_CMSG_IP_TOS + if (cmsgsend(s, IPPROTO_IP, IP_TOS, res0)) + dscp_result |= ISC_NET_DSCPPKTV4; +#endif /* ISC_CMSG_IP_TOS */ + +#endif /* ISC_NET_BSD44MSGHDR */ + + freeaddrinfo(res0); + close(s); + +#endif /* IP_TOS */ +} + +static void +try_dscp_v6(void) { +#ifdef ISC_PLATFORM_HAVEIPV6 +#ifdef WANT_IPV6 +#ifdef IPV6_TCLASS + char strbuf[ISC_STRERRORSIZE]; + struct addrinfo hints, *res0; + int s, dscp = 0, n; +#if defined(IPV6_RECVTCLASS) + int on = 1; +#endif /* IPV6_RECVTCLASS */ + + memset(&hints, 0, sizeof(hints)); + hints.ai_family = AF_INET6; + hints.ai_socktype = SOCK_DGRAM; + hints.ai_protocol = IPPROTO_UDP; +#ifdef AI_NUMERICHOST + hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST; +#else + hints.ai_flags = AI_PASSIVE; +#endif + + n = getaddrinfo("::1", NULL, &hints, &res0); + if (n != 0 || res0 == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getaddrinfo(::1): %s", gai_strerror(n)); + return; + } + + s = socket(res0->ai_family, res0->ai_socktype, res0->ai_protocol); + if (s == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "socket: %s", strbuf); + freeaddrinfo(res0); + return; + } + if (setsockopt(s, IPPROTO_IPV6, IPV6_TCLASS, &dscp, sizeof(dscp)) == 0) + dscp_result |= ISC_NET_DSCPSETV6; + +#ifdef IPV6_RECVTCLASS + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVTCLASS, &on, sizeof(on)) == 0) + dscp_result |= ISC_NET_DSCPRECVV6; +#endif /* IPV6_RECVTCLASS */ + +#ifdef ISC_NET_BSD44MSGHDR + if (cmsgsend(s, IPPROTO_IPV6, IPV6_TCLASS, res0)) + dscp_result |= ISC_NET_DSCPPKTV6; +#endif /* ISC_NET_BSD44MSGHDR */ + + freeaddrinfo(res0); + close(s); + +#endif /* IPV6_TCLASS */ +#endif /* WANT_IPV6 */ +#endif /* ISC_PLATFORM_HAVEIPV6 */ +} + +static void +try_dscp(void) { + try_dscp_v4(); + try_dscp_v6(); +} + +static void +initialize_dscp(void) { + RUNTIME_CHECK(isc_once_do(&once_dscp, try_dscp) == ISC_R_SUCCESS); +} + +unsigned int +isc_net_probedscp(void) { + initialize_dscp(); + return (dscp_result); +} + +#if defined(USE_SYSCTL_PORTRANGE) +#if defined(HAVE_SYSCTLBYNAME) +static isc_result_t +getudpportrange_sysctl(int af, in_port_t *low, in_port_t *high) { + int port_low, port_high; + size_t portlen; + const char *sysctlname_lowport, *sysctlname_hiport; + + if (af == AF_INET) { + sysctlname_lowport = SYSCTL_V4PORTRANGE_LOW; + sysctlname_hiport = SYSCTL_V4PORTRANGE_HIGH; + } else { + sysctlname_lowport = SYSCTL_V6PORTRANGE_LOW; + sysctlname_hiport = SYSCTL_V6PORTRANGE_HIGH; + } + portlen = sizeof(port_low); + if (sysctlbyname(sysctlname_lowport, &port_low, &portlen, + NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + portlen = sizeof(port_high); + if (sysctlbyname(sysctlname_hiport, &port_high, &portlen, + NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + if ((port_low & ~0xffff) != 0 || (port_high & ~0xffff) != 0) + return (ISC_R_RANGE); + + *low = (in_port_t)port_low; + *high = (in_port_t)port_high; + + return (ISC_R_SUCCESS); +} +#else /* !HAVE_SYSCTLBYNAME */ +static isc_result_t +getudpportrange_sysctl(int af, in_port_t *low, in_port_t *high) { + int mib_lo4[4] = SYSCTL_V4PORTRANGE_LOW; + int mib_hi4[4] = SYSCTL_V4PORTRANGE_HIGH; + int mib_lo6[4] = SYSCTL_V6PORTRANGE_LOW; + int mib_hi6[4] = SYSCTL_V6PORTRANGE_HIGH; + int *mib_lo, *mib_hi, miblen; + int port_low, port_high; + size_t portlen; + + if (af == AF_INET) { + mib_lo = mib_lo4; + mib_hi = mib_hi4; + miblen = sizeof(mib_lo4) / sizeof(mib_lo4[0]); + } else { + mib_lo = mib_lo6; + mib_hi = mib_hi6; + miblen = sizeof(mib_lo6) / sizeof(mib_lo6[0]); + } + + portlen = sizeof(port_low); + if (sysctl(mib_lo, miblen, &port_low, &portlen, NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + + portlen = sizeof(port_high); + if (sysctl(mib_hi, miblen, &port_high, &portlen, NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + + if ((port_low & ~0xffff) != 0 || (port_high & ~0xffff) != 0) + return (ISC_R_RANGE); + + *low = (in_port_t) port_low; + *high = (in_port_t) port_high; + + return (ISC_R_SUCCESS); +} +#endif /* HAVE_SYSCTLBYNAME */ +#endif /* USE_SYSCTL_PORTRANGE */ + +isc_result_t +isc_net_getudpportrange(int af, in_port_t *low, in_port_t *high) { + int result = ISC_R_FAILURE; +#if !defined(USE_SYSCTL_PORTRANGE) && defined(__linux) + FILE *fp; +#endif + + REQUIRE(low != NULL && high != NULL); + +#if defined(USE_SYSCTL_PORTRANGE) + result = getudpportrange_sysctl(af, low, high); +#elif defined(__linux) + + UNUSED(af); + + /* + * Linux local ports are address family agnostic. + */ + fp = fopen("/proc/sys/net/ipv4/ip_local_port_range", "r"); + if (fp != NULL) { + int n; + unsigned int l, h; + + n = fscanf(fp, "%u %u", &l, &h); + if (n == 2 && (l & ~0xffff) == 0 && (h & ~0xffff) == 0) { + *low = l; + *high = h; + result = ISC_R_SUCCESS; + } + fclose(fp); + } +#else + UNUSED(af); +#endif + + if (result != ISC_R_SUCCESS) { + *low = ISC_NET_PORTRANGELOW; + *high = ISC_NET_PORTRANGEHIGH; + } + + return (ISC_R_SUCCESS); /* we currently never fail in this function */ +} + +void +isc_net_disableipv4(void) { + initialize(); + if (ipv4_result == ISC_R_SUCCESS) + ipv4_result = ISC_R_DISABLED; +} + +void +isc_net_disableipv6(void) { + initialize(); + if (ipv6_result == ISC_R_SUCCESS) + ipv6_result = ISC_R_DISABLED; +} + +void +isc_net_enableipv4(void) { + initialize(); + if (ipv4_result == ISC_R_DISABLED) + ipv4_result = ISC_R_SUCCESS; +} + +void +isc_net_enableipv6(void) { + initialize(); + if (ipv6_result == ISC_R_DISABLED) + ipv6_result = ISC_R_SUCCESS; +} diff --git a/lib/isc/unix/os.c b/lib/isc/unix/os.c new file mode 100644 index 0000000..8cff6b6 --- /dev/null +++ b/lib/isc/unix/os.c @@ -0,0 +1,87 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <isc/os.h> + + +#ifdef HAVE_SYSCONF + +#include <unistd.h> + +#ifndef __hpux +static inline long +sysconf_ncpus(void) { +#if defined(_SC_NPROCESSORS_ONLN) + return sysconf((_SC_NPROCESSORS_ONLN)); +#elif defined(_SC_NPROC_ONLN) + return sysconf((_SC_NPROC_ONLN)); +#else + return (0); +#endif +} +#endif +#endif /* HAVE_SYSCONF */ + + +#ifdef __hpux + +#include <sys/pstat.h> + +static inline int +hpux_ncpus(void) { + struct pst_dynamic psd; + if (pstat_getdynamic(&psd, sizeof(psd), 1, 0) != -1) + return (psd.psd_proc_cnt); + else + return (0); +} + +#endif /* __hpux */ + +#if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) +#include <sys/types.h> /* for FreeBSD */ +#include <sys/param.h> /* for NetBSD */ +#include <sys/sysctl.h> + +static int +sysctl_ncpus(void) { + int ncpu, result; + size_t len; + + len = sizeof(ncpu); + result = sysctlbyname("hw.ncpu", &ncpu, &len , 0, 0); + if (result != -1) + return (ncpu); + return (0); +} +#endif + +unsigned int +isc_os_ncpus(void) { + long ncpus = 0; + +#ifdef __hpux + ncpus = hpux_ncpus(); +#elif defined(HAVE_SYSCONF) + ncpus = sysconf_ncpus(); +#endif +#if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) + if (ncpus <= 0) + ncpus = sysctl_ncpus(); +#endif + if (ncpus <= 0) + ncpus = 1; + + return ((unsigned int)ncpus); +} diff --git a/lib/isc/unix/pk11_api.c b/lib/isc/unix/pk11_api.c new file mode 100644 index 0000000..bbc55af --- /dev/null +++ b/lib/isc/unix/pk11_api.c @@ -0,0 +1,681 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <string.h> +#include <dlfcn.h> + +#include <isc/log.h> +#include <isc/mem.h> +#include <isc/once.h> +#include <isc/print.h> +#include <isc/stdio.h> +#include <isc/thread.h> +#include <isc/util.h> + +#include <pkcs11/cryptoki.h> +#include <pkcs11/pkcs11.h> + +#define KEEP_PKCS11_NAMES +#include <pk11/pk11.h> +#include <pk11/internal.h> + +static void *hPK11 = NULL; +static char loaderrmsg[1024]; + +CK_RV +pkcs_C_Initialize(CK_VOID_PTR pReserved) { + CK_C_Initialize sym; + + if (hPK11 != NULL) + return (CKR_LIBRARY_ALREADY_INITIALIZED); + + hPK11 = dlopen(pk11_get_lib_name(), RTLD_NOW); + + if (hPK11 == NULL) { + snprintf(loaderrmsg, sizeof(loaderrmsg), + "dlopen(\"%s\") failed: %s\n", + pk11_get_lib_name(), dlerror()); + return (CKR_LIBRARY_FAILED_TO_LOAD); + } + sym = (CK_C_Initialize)dlsym(hPK11, "C_Initialize"); + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(pReserved); +} + +char *pk11_get_load_error_message(void) { + return (loaderrmsg); +} + +CK_RV +pkcs_C_Finalize(CK_VOID_PTR pReserved) { + CK_C_Finalize sym; + CK_RV rv; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + sym = (CK_C_Finalize)dlsym(hPK11, "C_Finalize"); + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + rv = (*sym)(pReserved); + if ((rv == CKR_OK) && (dlclose(hPK11) != 0)) + return (CKR_LIBRARY_FAILED_TO_LOAD); + hPK11 = NULL; + return (rv); +} + +CK_RV +pkcs_C_GetSlotList(CK_BBOOL tokenPresent, CK_SLOT_ID_PTR pSlotList, + CK_ULONG_PTR pulCount) +{ + static CK_C_GetSlotList sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetSlotList)dlsym(hPK11, "C_GetSlotList"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(tokenPresent, pSlotList, pulCount); +} + +CK_RV +pkcs_C_GetTokenInfo(CK_SLOT_ID slotID, CK_TOKEN_INFO_PTR pInfo) { + static CK_C_GetTokenInfo sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetTokenInfo)dlsym(hPK11, "C_GetTokenInfo"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(slotID, pInfo); +} + +CK_RV +pkcs_C_GetMechanismInfo(CK_SLOT_ID slotID, CK_MECHANISM_TYPE type, + CK_MECHANISM_INFO_PTR pInfo) +{ + static CK_C_GetMechanismInfo sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetMechanismInfo)dlsym(hPK11, + "C_GetMechanismInfo"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(slotID, type, pInfo); +} + +CK_RV +pkcs_C_OpenSession(CK_SLOT_ID slotID, CK_FLAGS flags, + CK_VOID_PTR pApplication, + CK_RV (*Notify) (CK_SESSION_HANDLE hSession, + CK_NOTIFICATION event, + CK_VOID_PTR pApplication), + CK_SESSION_HANDLE_PTR phSession) +{ + static CK_C_OpenSession sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + hPK11 = dlopen(pk11_get_lib_name(), RTLD_NOW); + if (hPK11 == NULL) { + snprintf(loaderrmsg, sizeof(loaderrmsg), + "dlopen(\"%s\") failed: %s\n", + pk11_get_lib_name(), dlerror()); + return (CKR_LIBRARY_FAILED_TO_LOAD); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_OpenSession)dlsym(hPK11, "C_OpenSession"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(slotID, flags, pApplication, Notify, phSession); +} + +CK_RV +pkcs_C_CloseSession(CK_SESSION_HANDLE hSession) { + static CK_C_CloseSession sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_CloseSession)dlsym(hPK11, "C_CloseSession"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession); +} + +CK_RV +pkcs_C_Login(CK_SESSION_HANDLE hSession, CK_USER_TYPE userType, + CK_CHAR_PTR pPin, CK_ULONG usPinLen) +{ + static CK_C_Login sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Login)dlsym(hPK11, "C_Login"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, userType, pPin, usPinLen); +} + +CK_RV +pkcs_C_Logout(CK_SESSION_HANDLE hSession) { + static CK_C_Logout sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Logout)dlsym(hPK11, "C_Logout"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession); +} + +CK_RV +pkcs_C_CreateObject(CK_SESSION_HANDLE hSession, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG usCount, CK_OBJECT_HANDLE_PTR phObject) +{ + static CK_C_CreateObject sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_CreateObject)dlsym(hPK11, "C_CreateObject"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pTemplate, usCount, phObject); +} + +CK_RV +pkcs_C_DestroyObject(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject) { + static CK_C_DestroyObject sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DestroyObject)dlsym(hPK11, "C_DestroyObject"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, hObject); +} + +CK_RV +pkcs_C_GetAttributeValue(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG usCount) +{ + static CK_C_GetAttributeValue sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetAttributeValue)dlsym(hPK11, + "C_GetAttributeValue"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, hObject, pTemplate, usCount); +} + +CK_RV +pkcs_C_SetAttributeValue(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG usCount) +{ + static CK_C_SetAttributeValue sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SetAttributeValue)dlsym(hPK11, + "C_SetAttributeValue"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, hObject, pTemplate, usCount); +} + +CK_RV +pkcs_C_FindObjectsInit(CK_SESSION_HANDLE hSession, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG usCount) +{ + static CK_C_FindObjectsInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjectsInit)dlsym(hPK11, "C_FindObjectsInit"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pTemplate, usCount); +} + +CK_RV +pkcs_C_FindObjects(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE_PTR phObject, + CK_ULONG usMaxObjectCount, CK_ULONG_PTR pusObjectCount) +{ + static CK_C_FindObjects sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjects)dlsym(hPK11, "C_FindObjects"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, phObject, usMaxObjectCount, pusObjectCount); +} + +CK_RV +pkcs_C_FindObjectsFinal(CK_SESSION_HANDLE hSession) +{ + static CK_C_FindObjectsFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjectsFinal)dlsym(hPK11, + "C_FindObjectsFinal"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession); +} + +CK_RV +pkcs_C_EncryptInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) +{ + static CK_C_EncryptInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_EncryptInit)dlsym(hPK11, "C_EncryptInit"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pMechanism, hKey); +} + +CK_RV +pkcs_C_Encrypt(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, + CK_ULONG ulDataLen, CK_BYTE_PTR pEncryptedData, + CK_ULONG_PTR pulEncryptedDataLen) +{ + static CK_C_Encrypt sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Encrypt)dlsym(hPK11, "C_Encrypt"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pData, ulDataLen, + pEncryptedData, pulEncryptedDataLen); +} + +CK_RV +pkcs_C_DigestInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism) { + static CK_C_DigestInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestInit)dlsym(hPK11, "C_DigestInit"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pMechanism); +} + +CK_RV +pkcs_C_DigestUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) +{ + static CK_C_DigestUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestUpdate)dlsym(hPK11, "C_DigestUpdate"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pPart, ulPartLen); +} + +CK_RV +pkcs_C_DigestFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pDigest, + CK_ULONG_PTR pulDigestLen) +{ + static CK_C_DigestFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestFinal)dlsym(hPK11, "C_DigestFinal"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pDigest, pulDigestLen); +} + +CK_RV +pkcs_C_SignInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) +{ + static CK_C_SignInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignInit)dlsym(hPK11, "C_SignInit"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pMechanism, hKey); +} + +CK_RV +pkcs_C_Sign(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, + CK_ULONG ulDataLen, CK_BYTE_PTR pSignature, + CK_ULONG_PTR pulSignatureLen) +{ + static CK_C_Sign sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Sign)dlsym(hPK11, "C_Sign"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pData, ulDataLen, pSignature, pulSignatureLen); +} + +CK_RV +pkcs_C_SignUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) +{ + static CK_C_SignUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignUpdate)dlsym(hPK11, "C_SignUpdate"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pPart, ulPartLen); +} + +CK_RV +pkcs_C_SignFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSignature, + CK_ULONG_PTR pulSignatureLen) +{ + static CK_C_SignFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignFinal)dlsym(hPK11, "C_SignFinal"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pSignature, pulSignatureLen); +} + +CK_RV +pkcs_C_VerifyInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) +{ + static CK_C_VerifyInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyInit)dlsym(hPK11, "C_VerifyInit"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pMechanism, hKey); +} + +CK_RV +pkcs_C_Verify(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, + CK_ULONG ulDataLen, CK_BYTE_PTR pSignature, + CK_ULONG ulSignatureLen) +{ + static CK_C_Verify sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Verify)dlsym(hPK11, "C_Verify"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pData, ulDataLen, pSignature, ulSignatureLen); +} + +CK_RV +pkcs_C_VerifyUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) +{ + static CK_C_VerifyUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyUpdate)dlsym(hPK11, "C_VerifyUpdate"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pPart, ulPartLen); +} + +CK_RV +pkcs_C_VerifyFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSignature, + CK_ULONG ulSignatureLen) +{ + static CK_C_VerifyFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyFinal)dlsym(hPK11, "C_VerifyFinal"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pSignature, ulSignatureLen); +} + +CK_RV +pkcs_C_GenerateKey(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG ulCount, + CK_OBJECT_HANDLE_PTR phKey) +{ + static CK_C_GenerateKey sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateKey)dlsym(hPK11, "C_GenerateKey"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pMechanism, pTemplate, ulCount, phKey); +} + +CK_RV +pkcs_C_GenerateKeyPair(CK_SESSION_HANDLE hSession, + CK_MECHANISM_PTR pMechanism, + CK_ATTRIBUTE_PTR pPublicKeyTemplate, + CK_ULONG usPublicKeyAttributeCount, + CK_ATTRIBUTE_PTR pPrivateKeyTemplate, + CK_ULONG usPrivateKeyAttributeCount, + CK_OBJECT_HANDLE_PTR phPrivateKey, + CK_OBJECT_HANDLE_PTR phPublicKey) +{ + static CK_C_GenerateKeyPair sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateKeyPair)dlsym(hPK11, "C_GenerateKeyPair"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, + pMechanism, + pPublicKeyTemplate, + usPublicKeyAttributeCount, + pPrivateKeyTemplate, + usPrivateKeyAttributeCount, + phPrivateKey, + phPublicKey); +} + +CK_RV +pkcs_C_DeriveKey(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hBaseKey, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG ulAttributeCount, CK_OBJECT_HANDLE_PTR phKey) +{ + static CK_C_DeriveKey sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DeriveKey)dlsym(hPK11, "C_DeriveKey"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, + pMechanism, + hBaseKey, + pTemplate, + ulAttributeCount, + phKey); +} + +CK_RV +pkcs_C_SeedRandom(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSeed, + CK_ULONG ulSeedLen) +{ + static CK_C_SeedRandom sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SeedRandom)dlsym(hPK11, "C_SeedRandom"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, pSeed, ulSeedLen); +} + +CK_RV +pkcs_C_GenerateRandom(CK_SESSION_HANDLE hSession, CK_BYTE_PTR RandomData, + CK_ULONG ulRandomLen) +{ + static CK_C_GenerateRandom sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) + return (CKR_LIBRARY_FAILED_TO_LOAD); + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateRandom)dlsym(hPK11, "C_GenerateRandom"); + } + if (sym == NULL) + return (CKR_SYMBOL_RESOLUTION_FAILED); + return (*sym)(hSession, RandomData, ulRandomLen); +} diff --git a/lib/isc/unix/resource.c b/lib/isc/unix/resource.c new file mode 100644 index 0000000..6d5c5aa --- /dev/null +++ b/lib/isc/unix/resource.c @@ -0,0 +1,227 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <inttypes.h> +#include <stdbool.h> + +#include <sys/types.h> +#include <sys/time.h> /* Required on some systems for <sys/resource.h>. */ +#include <sys/resource.h> + +#include <isc/platform.h> +#include <isc/resource.h> +#include <isc/result.h> +#include <isc/util.h> + +#ifdef __linux__ +#include <linux/fs.h> /* To get the large NR_OPEN. */ +#endif + +#if defined(__hpux) && defined(HAVE_SYS_DYNTUNE_H) +#include <sys/dyntune.h> +#endif + +#include "errno2result.h" + +static isc_result_t +resource2rlim(isc_resource_t resource, int *rlim_resource) { + isc_result_t result = ISC_R_SUCCESS; + + switch (resource) { + case isc_resource_coresize: + *rlim_resource = RLIMIT_CORE; + break; + case isc_resource_cputime: + *rlim_resource = RLIMIT_CPU; + break; + case isc_resource_datasize: + *rlim_resource = RLIMIT_DATA; + break; + case isc_resource_filesize: + *rlim_resource = RLIMIT_FSIZE; + break; + case isc_resource_lockedmemory: +#ifdef RLIMIT_MEMLOCK + *rlim_resource = RLIMIT_MEMLOCK; +#else + result = ISC_R_NOTIMPLEMENTED; +#endif + break; + case isc_resource_openfiles: +#ifdef RLIMIT_NOFILE + *rlim_resource = RLIMIT_NOFILE; +#else + result = ISC_R_NOTIMPLEMENTED; +#endif + break; + case isc_resource_processes: +#ifdef RLIMIT_NPROC + *rlim_resource = RLIMIT_NPROC; +#else + result = ISC_R_NOTIMPLEMENTED; +#endif + break; + case isc_resource_residentsize: +#ifdef RLIMIT_RSS + *rlim_resource = RLIMIT_RSS; +#else + result = ISC_R_NOTIMPLEMENTED; +#endif + break; + case isc_resource_stacksize: + *rlim_resource = RLIMIT_STACK; + break; + default: + /* + * This test is not very robust if isc_resource_t + * changes, but generates a clear assertion message. + */ + REQUIRE(resource >= isc_resource_coresize && + resource <= isc_resource_stacksize); + + result = ISC_R_RANGE; + break; + } + + return (result); +} + +isc_result_t +isc_resource_setlimit(isc_resource_t resource, isc_resourcevalue_t value) { + struct rlimit rl; + ISC_PLATFORM_RLIMITTYPE rlim_value; + int unixresult; + int unixresource; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result != ISC_R_SUCCESS) + return (result); + + if (value == ISC_RESOURCE_UNLIMITED) + rlim_value = RLIM_INFINITY; + + else { + /* + * isc_resourcevalue_t was chosen as an unsigned 64 bit + * integer so that it could contain the maximum range of + * reasonable values. Unfortunately, this exceeds the typical + * range on Unix systems. Ensure the range of + * ISC_PLATFORM_RLIMITTYPE is not overflowed. + */ + isc_resourcevalue_t rlim_max; + bool rlim_t_is_signed = + (((double)(ISC_PLATFORM_RLIMITTYPE)-1) < 0); + + if (rlim_t_is_signed) + rlim_max = ~((ISC_PLATFORM_RLIMITTYPE)1 << + (sizeof(ISC_PLATFORM_RLIMITTYPE) * 8 - 1)); + else + rlim_max = (ISC_PLATFORM_RLIMITTYPE)-1; + + if (value > rlim_max) + value = rlim_max; + + rlim_value = value; + } + + rl.rlim_cur = rl.rlim_max = rlim_value; + unixresult = setrlimit(unixresource, &rl); + + if (unixresult == 0) + return (ISC_R_SUCCESS); + +#if defined(OPEN_MAX) && defined(__APPLE__) + /* + * The Darwin kernel doesn't accept RLIM_INFINITY for rlim_cur; the + * maximum possible value is OPEN_MAX. BIND8 used to use + * sysconf(_SC_OPEN_MAX) for such a case, but this value is much + * smaller than OPEN_MAX and is not really effective. + */ + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + rl.rlim_cur = OPEN_MAX; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) + return (ISC_R_SUCCESS); + } +#elif defined(__linux__) +#ifndef NR_OPEN +#define NR_OPEN (1024*1024) +#endif + + /* + * Some Linux kernels don't accept RLIM_INFINIT; the maximum + * possible value is the NR_OPEN defined in linux/fs.h. + */ + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + rl.rlim_cur = rl.rlim_max = NR_OPEN; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) + return (ISC_R_SUCCESS); + } +#elif defined(__hpux) && defined(HAVE_SYS_DYNTUNE_H) + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + uint64_t maxfiles; + if (gettune("maxfiles_lim", &maxfiles) == 0) { + rl.rlim_cur = rl.rlim_max = maxfiles; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) + return (ISC_R_SUCCESS); + } + } +#endif + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + if (getrlimit(unixresource, &rl) == 0) { + rl.rlim_cur = rl.rlim_max; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) + return (ISC_R_SUCCESS); + } + } + return (isc__errno2result(errno)); +} + +isc_result_t +isc_resource_getlimit(isc_resource_t resource, isc_resourcevalue_t *value) { + int unixresult; + int unixresource; + struct rlimit rl; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result == ISC_R_SUCCESS) { + unixresult = getrlimit(unixresource, &rl); + INSIST(unixresult == 0); + *value = rl.rlim_max; + } + + return (result); +} + +isc_result_t +isc_resource_getcurlimit(isc_resource_t resource, isc_resourcevalue_t *value) { + int unixresult; + int unixresource; + struct rlimit rl; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result == ISC_R_SUCCESS) { + unixresult = getrlimit(unixresource, &rl); + INSIST(unixresult == 0); + *value = rl.rlim_cur; + } + + return (result); +} diff --git a/lib/isc/unix/socket.c b/lib/isc/unix/socket.c new file mode 100644 index 0000000..5a40768 --- /dev/null +++ b/lib/isc/unix/socket.c @@ -0,0 +1,6848 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <config.h> + +#include <inttypes.h> +#include <stdbool.h> + +#include <sys/param.h> +#include <sys/types.h> +#include <sys/socket.h> +#include <sys/stat.h> +#ifdef HAVE_SYS_SYSCTL_H +#include <sys/sysctl.h> +#endif +#include <sys/time.h> +#include <sys/uio.h> + +#if defined(HAVE_LINUX_NETLINK_H) && defined(HAVE_LINUX_RTNETLINK_H) +#include <linux/netlink.h> +#include <linux/rtnetlink.h> +#endif + +#include <errno.h> +#include <fcntl.h> +#include <stddef.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> + +#include <isc/buffer.h> +#include <isc/bufferlist.h> +#include <isc/condition.h> +#include <isc/formatcheck.h> +#include <isc/json.h> +#include <isc/list.h> +#include <isc/log.h> +#include <isc/mem.h> +#include <isc/msgs.h> +#include <isc/mutex.h> +#include <isc/net.h> +#include <isc/once.h> +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/region.h> +#include <isc/resource.h> +#include <isc/socket.h> +#include <isc/stats.h> +#include <isc/strerror.h> +#include <isc/string.h> +#include <isc/task.h> +#include <isc/thread.h> +#include <isc/util.h> +#include <isc/xml.h> + +#ifdef ISC_PLATFORM_HAVESYSUNH +#include <sys/un.h> +#endif +#ifdef ISC_PLATFORM_HAVEKQUEUE +#include <sys/event.h> +#endif +#ifdef ISC_PLATFORM_HAVEEPOLL +#include <sys/epoll.h> +#endif +#ifdef ISC_PLATFORM_HAVEDEVPOLL +#if defined(HAVE_SYS_DEVPOLL_H) +#include <sys/devpoll.h> +#elif defined(HAVE_DEVPOLL_H) +#include <devpoll.h> +#endif +#endif + +#include <netinet/tcp.h> + +#include "errno2result.h" + +/* See task.c about the following definition: */ +#ifdef ISC_PLATFORM_USETHREADS +#define USE_WATCHER_THREAD +#else +#define USE_SHARED_MANAGER +#endif /* ISC_PLATFORM_USETHREADS */ + +#ifndef USE_WATCHER_THREAD +#include "socket_p.h" +#include "../task_p.h" +#endif /* USE_WATCHER_THREAD */ + +#if defined(SO_BSDCOMPAT) && defined(__linux__) +#include <sys/utsname.h> +#endif + +#ifdef ISC_PLATFORM_HAVETFO +#include <netinet/tcp.h> +#endif + +/*% + * Choose the most preferable multiplex method. + */ +#ifdef ISC_PLATFORM_HAVEKQUEUE +#define USE_KQUEUE +#elif defined (ISC_PLATFORM_HAVEEPOLL) +#define USE_EPOLL +#elif defined (ISC_PLATFORM_HAVEDEVPOLL) +#define USE_DEVPOLL +typedef struct { + unsigned int want_read : 1, + want_write : 1; +} pollinfo_t; +#else +#define USE_SELECT +#endif /* ISC_PLATFORM_HAVEKQUEUE */ + +#ifndef USE_WATCHER_THREAD +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) +struct isc_socketwait { + int nevents; +}; +#elif defined (USE_SELECT) +struct isc_socketwait { + fd_set *readset; + fd_set *writeset; + int nfds; + int maxfd; +}; +#endif /* USE_KQUEUE */ +#endif /* !USE_WATCHER_THREAD */ + +/* + * Set by the -T dscp option on the command line. If set to a value + * other than -1, we check to make sure DSCP values match it, and + * assert if not. + */ +int isc_dscp_check_value = -1; + +/*% + * Maximum number of allowable open sockets. This is also the maximum + * allowable socket file descriptor. + * + * Care should be taken before modifying this value for select(): + * The API standard doesn't ensure select() accept more than (the system default + * of) FD_SETSIZE descriptors, and the default size should in fact be fine in + * the vast majority of cases. This constant should therefore be increased only + * when absolutely necessary and possible, i.e., the server is exhausting all + * available file descriptors (up to FD_SETSIZE) and the select() function + * and FD_xxx macros support larger values than FD_SETSIZE (which may not + * always by true, but we keep using some of them to ensure as much + * portability as possible). Note also that overall server performance + * may be rather worsened with a larger value of this constant due to + * inherent scalability problems of select(). + * + * As a special note, this value shouldn't have to be touched if + * this is a build for an authoritative only DNS server. + */ +#ifndef ISC_SOCKET_MAXSOCKETS +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) +#ifdef TUNE_LARGE +#define ISC_SOCKET_MAXSOCKETS 21000 +#else +#define ISC_SOCKET_MAXSOCKETS 4096 +#endif /* TUNE_LARGE */ +#elif defined(USE_SELECT) +#define ISC_SOCKET_MAXSOCKETS FD_SETSIZE +#endif /* USE_KQUEUE... */ +#endif /* ISC_SOCKET_MAXSOCKETS */ + +#ifdef USE_SELECT +/*% + * Mac OS X needs a special definition to support larger values in select(). + * We always define this because a larger value can be specified run-time. + */ +#ifdef __APPLE__ +#define _DARWIN_UNLIMITED_SELECT +#endif /* __APPLE__ */ +#endif /* USE_SELECT */ + +#ifdef ISC_SOCKET_USE_POLLWATCH +/*% + * If this macro is defined, enable workaround for a Solaris /dev/poll kernel + * bug: DP_POLL ioctl could keep sleeping even if socket I/O is possible for + * some of the specified FD. The idea is based on the observation that it's + * likely for a busy server to keep receiving packets. It specifically works + * as follows: the socket watcher is first initialized with the state of + * "poll_idle". While it's in the idle state it keeps sleeping until a socket + * event occurs. When it wakes up for a socket I/O event, it moves to the + * poll_active state, and sets the poll timeout to a short period + * (ISC_SOCKET_POLLWATCH_TIMEOUT msec). If timeout occurs in this state, the + * watcher goes to the poll_checking state with the same timeout period. + * In this state, the watcher tries to detect whether this is a break + * during intermittent events or the kernel bug is triggered. If the next + * polling reports an event within the short period, the previous timeout is + * likely to be a kernel bug, and so the watcher goes back to the active state. + * Otherwise, it moves to the idle state again. + * + * It's not clear whether this is a thread-related bug, but since we've only + * seen this with threads, this workaround is used only when enabling threads. + */ + +typedef enum { poll_idle, poll_active, poll_checking } pollstate_t; + +#ifndef ISC_SOCKET_POLLWATCH_TIMEOUT +#define ISC_SOCKET_POLLWATCH_TIMEOUT 10 +#endif /* ISC_SOCKET_POLLWATCH_TIMEOUT */ +#endif /* ISC_SOCKET_USE_POLLWATCH */ + +/*% + * Size of per-FD lock buckets. + */ +#ifdef ISC_PLATFORM_USETHREADS +#define FDLOCK_COUNT 1024 +#define FDLOCK_ID(fd) ((fd) % FDLOCK_COUNT) +#else +#define FDLOCK_COUNT 1 +#define FDLOCK_ID(fd) 0 +#endif /* ISC_PLATFORM_USETHREADS */ + +/*% + * Maximum number of events communicated with the kernel. There should normally + * be no need for having a large number. + */ +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) +#ifndef ISC_SOCKET_MAXEVENTS +#ifdef TUNE_LARGE +#define ISC_SOCKET_MAXEVENTS 2048 +#else +#define ISC_SOCKET_MAXEVENTS 64 +#endif /* TUNE_LARGE */ +#endif +#endif + +/*% + * Some systems define the socket length argument as an int, some as size_t, + * some as socklen_t. This is here so it can be easily changed if needed. + */ +#ifndef ISC_SOCKADDR_LEN_T +#define ISC_SOCKADDR_LEN_T unsigned int +#endif + +/*% + * Define what the possible "soft" errors can be. These are non-fatal returns + * of various network related functions, like recv() and so on. + * + * For some reason, BSDI (and perhaps others) will sometimes return <0 + * from recv() but will have errno==0. This is broken, but we have to + * work around it here. + */ +#define SOFT_ERROR(e) ((e) == EAGAIN || \ + (e) == EWOULDBLOCK || \ + (e) == ENOBUFS || \ + (e) == EINTR || \ + (e) == 0) + +#define DLVL(x) ISC_LOGCATEGORY_GENERAL, ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(x) + +/*!< + * DLVL(90) -- Function entry/exit and other tracing. + * DLVL(70) -- Socket "correctness" -- including returning of events, etc. + * DLVL(60) -- Socket data send/receive + * DLVL(50) -- Event tracing, including receiving/sending completion events. + * DLVL(20) -- Socket creation/destruction. + */ +#define TRACE_LEVEL 90 +#define CORRECTNESS_LEVEL 70 +#define IOEVENT_LEVEL 60 +#define EVENT_LEVEL 50 +#define CREATION_LEVEL 20 + +#define TRACE DLVL(TRACE_LEVEL) +#define CORRECTNESS DLVL(CORRECTNESS_LEVEL) +#define IOEVENT DLVL(IOEVENT_LEVEL) +#define EVENT DLVL(EVENT_LEVEL) +#define CREATION DLVL(CREATION_LEVEL) + +typedef isc_event_t intev_t; + +#define SOCKET_MAGIC ISC_MAGIC('I', 'O', 'i', 'o') +#define VALID_SOCKET(s) ISC_MAGIC_VALID(s, SOCKET_MAGIC) + +/*! + * IPv6 control information. If the socket is an IPv6 socket we want + * to collect the destination address and interface so the client can + * set them on outgoing packets. + */ +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO +#ifndef USE_CMSG +#define USE_CMSG 1 +#endif +#endif + +/*% + * NetBSD and FreeBSD can timestamp packets. XXXMLG Should we have + * a setsockopt() like interface to request timestamps, and if the OS + * doesn't do it for us, call gettimeofday() on every UDP receive? + */ +#ifdef SO_TIMESTAMP +#ifndef USE_CMSG +#define USE_CMSG 1 +#endif +#endif + +/*% + * The size to raise the receive buffer to (from BIND 8). + */ +#ifdef TUNE_LARGE +#ifdef sun +#define RCVBUFSIZE (1*1024*1024) +#else +#define RCVBUFSIZE (16*1024*1024) +#endif +#else +#define RCVBUFSIZE (32*1024) +#endif /* TUNE_LARGE */ + +/*% + * Instead of calculating the cmsgbuf lengths every time we take + * a rule of thumb approach - sizes are taken from x86_64 linux, + * multiplied by 2, everything should fit. Those sizes are not + * large enough to cause any concern. + */ +#if defined(USE_CMSG) && defined(ISC_PLATFORM_HAVEIN6PKTINFO) +#define CMSG_SP_IN6PKT 40 +#else +#define CMSG_SP_IN6PKT 0 +#endif + +#if defined(USE_CMSG) && defined(SO_TIMESTAMP) +#define CMSG_SP_TIMESTAMP 32 +#else +#define CMSG_SP_TIMESTAMP 0 +#endif + +#if defined(USE_CMSG) && (defined(IPV6_TCLASS) || defined(IP_TOS)) +#define CMSG_SP_TCTOS 24 +#else +#define CMSG_SP_TCTOS 0 +#endif + +#define CMSG_SP_INT 24 + +/* Align cmsg buffers to be safe on SPARC etc. */ +#define RECVCMSGBUFLEN ISC_ALIGN(2*(CMSG_SP_IN6PKT + CMSG_SP_TIMESTAMP + CMSG_SP_TCTOS)+1, sizeof(void*)) +#define SENDCMSGBUFLEN ISC_ALIGN(2*(CMSG_SP_IN6PKT + CMSG_SP_INT + CMSG_SP_TCTOS)+1, sizeof(void*)) + +/*% + * The number of times a send operation is repeated if the result is EINTR. + */ +#define NRETRIES 10 + +typedef struct isc__socket isc__socket_t; +typedef struct isc__socketmgr isc__socketmgr_t; + +#define NEWCONNSOCK(ev) ((isc__socket_t *)(ev)->newsocket) + +struct isc__socket { + /* Not locked. */ + isc_socket_t common; + isc__socketmgr_t *manager; + isc_mutex_t lock; + isc_sockettype_t type; + const isc_statscounter_t *statsindex; + + /* Locked by socket lock. */ + ISC_LINK(isc__socket_t) link; + unsigned int references; + int fd; + int pf; + char name[16]; + void * tag; + + ISC_LIST(isc_socketevent_t) send_list; + ISC_LIST(isc_socketevent_t) recv_list; + ISC_LIST(isc_socket_newconnev_t) accept_list; + ISC_LIST(isc_socket_connev_t) connect_list; + + /* + * Internal events. Posted when a descriptor is readable or + * writable. These are statically allocated and never freed. + * They will be set to non-purgable before use. + */ + intev_t readable_ev; + intev_t writable_ev; + + isc_sockaddr_t peer_address; /* remote address */ + + unsigned int pending_recv : 1, + pending_send : 1, + pending_accept : 1, + listener : 1, /* listener socket */ + connected : 1, + connecting : 1, /* connect pending */ + bound : 1, /* bound to local addr */ + dupped : 1, + active : 1, /* currently active */ + pktdscp : 1; /* per packet dscp */ + +#ifdef ISC_PLATFORM_RECVOVERFLOW + unsigned char overflow; /* used for MSG_TRUNC fake */ +#endif + + void *fdwatcharg; + isc_sockfdwatch_t fdwatchcb; + int fdwatchflags; + isc_task_t *fdwatchtask; + unsigned int dscp; +}; + +#define SOCKET_MANAGER_MAGIC ISC_MAGIC('I', 'O', 'm', 'g') +#define VALID_MANAGER(m) ISC_MAGIC_VALID(m, SOCKET_MANAGER_MAGIC) + +struct isc__socketmgr { + /* Not locked. */ + isc_socketmgr_t common; + isc_mem_t *mctx; + isc_mutex_t lock; + isc_mutex_t *fdlock; + isc_stats_t *stats; +#ifdef USE_KQUEUE + int kqueue_fd; + int nevents; + struct kevent *events; +#endif /* USE_KQUEUE */ +#ifdef USE_EPOLL + int epoll_fd; + int nevents; + struct epoll_event *events; +#endif /* USE_EPOLL */ +#ifdef USE_DEVPOLL + int devpoll_fd; + isc_resourcevalue_t open_max; + unsigned int calls; + int nevents; + struct pollfd *events; +#endif /* USE_DEVPOLL */ +#ifdef USE_SELECT + int fd_bufsize; +#endif /* USE_SELECT */ + unsigned int maxsocks; +#ifdef ISC_PLATFORM_USETHREADS + int pipe_fds[2]; +#endif + + /* Locked by fdlock. */ + isc__socket_t **fds; + int *fdstate; +#if defined(USE_EPOLL) + uint32_t *epoll_events; +#endif +#ifdef USE_DEVPOLL + pollinfo_t *fdpollinfo; +#endif + + /* Locked by manager lock. */ + ISC_LIST(isc__socket_t) socklist; +#ifdef USE_SELECT + fd_set *read_fds; + fd_set *read_fds_copy; + fd_set *write_fds; + fd_set *write_fds_copy; + int maxfd; +#endif /* USE_SELECT */ + int reserved; /* unlocked */ +#ifdef USE_WATCHER_THREAD + isc_thread_t watcher; + isc_condition_t shutdown_ok; +#else /* USE_WATCHER_THREAD */ + unsigned int refs; +#endif /* USE_WATCHER_THREAD */ + int maxudp; +}; + +#ifdef USE_SHARED_MANAGER +static isc__socketmgr_t *socketmgr = NULL; +#endif /* USE_SHARED_MANAGER */ + +#define CLOSED 0 /* this one must be zero */ +#define MANAGED 1 +#define CLOSE_PENDING 2 + +/* + * send() and recv() iovec counts + */ +#define MAXSCATTERGATHER_SEND (ISC_SOCKET_MAXSCATTERGATHER) +#ifdef ISC_PLATFORM_RECVOVERFLOW +# define MAXSCATTERGATHER_RECV (ISC_SOCKET_MAXSCATTERGATHER + 1) +#else +# define MAXSCATTERGATHER_RECV (ISC_SOCKET_MAXSCATTERGATHER) +#endif + +static isc_result_t socket_create(isc_socketmgr_t *manager0, int pf, + isc_sockettype_t type, + isc_socket_t **socketp, + isc_socket_t *dup_socket); +static void send_recvdone_event(isc__socket_t *, isc_socketevent_t **); +static void send_senddone_event(isc__socket_t *, isc_socketevent_t **); +static void send_connectdone_event(isc__socket_t *, isc_socket_connev_t **); +static void free_socket(isc__socket_t **); +static isc_result_t allocate_socket(isc__socketmgr_t *, isc_sockettype_t, + isc__socket_t **); +static void destroy(isc__socket_t **); +static void internal_accept(isc_task_t *, isc_event_t *); +static void internal_connect(isc_task_t *, isc_event_t *); +static void internal_recv(isc_task_t *, isc_event_t *); +static void internal_send(isc_task_t *, isc_event_t *); +static void internal_fdwatch_write(isc_task_t *, isc_event_t *); +static void internal_fdwatch_read(isc_task_t *, isc_event_t *); +static void process_cmsg(isc__socket_t *, struct msghdr *, isc_socketevent_t *); +static void build_msghdr_send(isc__socket_t *, char *, isc_socketevent_t *, + struct msghdr *, struct iovec *, size_t *); +static void build_msghdr_recv(isc__socket_t *, char *, isc_socketevent_t *, + struct msghdr *, struct iovec *, size_t *); +#ifdef USE_WATCHER_THREAD +static bool process_ctlfd(isc__socketmgr_t *manager); +#endif +static void setdscp(isc__socket_t *sock, isc_dscp_t dscp); + +/*% + * The following are intended for internal use (indicated by "isc__" + * prefix) but are not declared as static, allowing direct access from + * unit tests etc. + */ + +isc_result_t +isc__socket_open(isc_socket_t *sock0); +isc_result_t +isc__socket_close(isc_socket_t *sock0); +isc_result_t +isc__socket_create(isc_socketmgr_t *manager, int pf, isc_sockettype_t type, + isc_socket_t **socketp); +void +isc__socket_attach(isc_socket_t *sock, isc_socket_t **socketp); +void +isc__socket_detach(isc_socket_t **socketp); +isc_result_t +isc__socket_recvv(isc_socket_t *sock, isc_bufferlist_t *buflist, + unsigned int minimum, isc_task_t *task, + isc_taskaction_t action, void *arg); +isc_result_t +isc__socket_recv(isc_socket_t *sock, isc_region_t *region, + unsigned int minimum, isc_task_t *task, + isc_taskaction_t action, void *arg); +isc_result_t +isc__socket_recv2(isc_socket_t *sock, isc_region_t *region, + unsigned int minimum, isc_task_t *task, + isc_socketevent_t *event, unsigned int flags); +isc_result_t +isc__socket_send(isc_socket_t *sock, isc_region_t *region, + isc_task_t *task, isc_taskaction_t action, void *arg); +isc_result_t +isc__socket_sendto(isc_socket_t *sock, isc_region_t *region, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo); +isc_result_t +isc__socket_sendv(isc_socket_t *sock, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg); +isc_result_t +isc__socket_sendtov(isc_socket_t *sock, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo); +isc_result_t +isc__socket_sendtov2(isc_socket_t *sock, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + unsigned int flags); +isc_result_t +isc__socket_sendto2(isc_socket_t *sock, isc_region_t *region, + isc_task_t *task, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + isc_socketevent_t *event, unsigned int flags); +isc_socketevent_t * +isc_socket_socketevent(isc_mem_t *mctx, void *sender, + isc_eventtype_t eventtype, isc_taskaction_t action, + void *arg); + +void +isc__socket_cleanunix(isc_sockaddr_t *sockaddr, bool active); +isc_result_t +isc__socket_permunix(isc_sockaddr_t *sockaddr, uint32_t perm, + uint32_t owner, uint32_t group); +isc_result_t +isc__socket_bind(isc_socket_t *sock, isc_sockaddr_t *sockaddr, + unsigned int options); +isc_result_t +isc__socket_filter(isc_socket_t *sock, const char *filter); +isc_result_t +isc__socket_listen(isc_socket_t *sock, unsigned int backlog); +isc_result_t +isc__socket_accept(isc_socket_t *sock, + isc_task_t *task, isc_taskaction_t action, void *arg); +isc_result_t +isc__socket_connect(isc_socket_t *sock, isc_sockaddr_t *addr, + isc_task_t *task, isc_taskaction_t action, + void *arg); +isc_result_t +isc__socket_getpeername(isc_socket_t *sock, isc_sockaddr_t *addressp); +isc_result_t +isc__socket_getsockname(isc_socket_t *sock, isc_sockaddr_t *addressp); +void +isc__socket_cancel(isc_socket_t *sock, isc_task_t *task, unsigned int how); +isc_sockettype_t +isc__socket_gettype(isc_socket_t *sock); +bool +isc__socket_isbound(isc_socket_t *sock); +void +isc__socket_ipv6only(isc_socket_t *sock, bool yes); +void +isc__socket_dscp(isc_socket_t *sock, isc_dscp_t dscp); +isc_result_t +isc__socket_fdwatchcreate(isc_socketmgr_t *manager, int fd, int flags, + isc_sockfdwatch_t callback, void *cbarg, + isc_task_t *task, isc_socket_t **socketp); +isc_result_t +isc__socket_fdwatchpoke(isc_socket_t *sock, int flags); +isc_result_t +isc__socket_dup(isc_socket_t *sock, isc_socket_t **socketp); +int +isc__socket_getfd(isc_socket_t *sock); + +isc_result_t +isc__socketmgr_create(isc_mem_t *mctx, isc_socketmgr_t **managerp); +isc_result_t +isc__socketmgr_create2(isc_mem_t *mctx, isc_socketmgr_t **managerp, + unsigned int maxsocks); +isc_result_t +isc_socketmgr_getmaxsockets(isc_socketmgr_t *manager0, unsigned int *nsockp); +void +isc_socketmgr_setstats(isc_socketmgr_t *manager0, isc_stats_t *stats); +void +isc__socketmgr_destroy(isc_socketmgr_t **managerp); +void +isc__socket_setname(isc_socket_t *socket0, const char *name, void *tag); +const char * +isc__socket_getname(isc_socket_t *socket0); +void * +isc__socket_gettag(isc_socket_t *socket0); + +#ifdef HAVE_LIBXML2 +void +isc__socketmgr_renderxml(isc_socketmgr_t *mgr0, xmlTextWriterPtr writer); +#endif +#ifdef HAVE_JSON +isc_result_t +isc__socketmgr_renderjson(isc_socketmgr_t *mgr0, json_object *stats); +#endif + +static struct { + isc_socketmethods_t methods; + + /*% + * The following are defined just for avoiding unused static functions. + */ + void *recvv, *send, *sendv, *sendto2, *cleanunix, *permunix, *filter, + *listen, *accept, *getpeername, *isbound; +} socketmethods = { + { + isc__socket_attach, + isc__socket_detach, + isc__socket_bind, + isc__socket_sendto, + isc__socket_sendto2, + isc__socket_connect, + isc__socket_recv, + isc__socket_recv2, + isc__socket_cancel, + isc__socket_getsockname, + isc__socket_gettype, + isc__socket_ipv6only, + isc__socket_fdwatchpoke, + isc__socket_dup, + isc__socket_getfd, + isc__socket_dscp + }, + (void *)isc__socket_recvv, (void *)isc__socket_send, + (void *)isc__socket_sendv, (void *)isc__socket_sendto2, + (void *)isc__socket_cleanunix, (void *)isc__socket_permunix, + (void *)isc__socket_filter, (void *)isc__socket_listen, + (void *)isc__socket_accept, (void *)isc__socket_getpeername, + (void *)isc__socket_isbound +}; + +static isc_socketmgrmethods_t socketmgrmethods = { + isc__socketmgr_destroy, + isc__socket_create, + isc__socket_fdwatchcreate +}; + +#define SELECT_POKE_SHUTDOWN (-1) +#define SELECT_POKE_NOTHING (-2) +#define SELECT_POKE_READ (-3) +#define SELECT_POKE_ACCEPT (-3) /*%< Same as _READ */ +#define SELECT_POKE_WRITE (-4) +#define SELECT_POKE_CONNECT (-4) /*%< Same as _WRITE */ +#define SELECT_POKE_CLOSE (-5) + +#define SOCK_DEAD(s) ((s)->references == 0) + +/*% + * Shortcut index arrays to get access to statistics counters. + */ +enum { + STATID_OPEN = 0, + STATID_OPENFAIL = 1, + STATID_CLOSE = 2, + STATID_BINDFAIL = 3, + STATID_CONNECTFAIL = 4, + STATID_CONNECT = 5, + STATID_ACCEPTFAIL = 6, + STATID_ACCEPT = 7, + STATID_SENDFAIL = 8, + STATID_RECVFAIL = 9, + STATID_ACTIVE = 10 +}; +static const isc_statscounter_t udp4statsindex[] = { + isc_sockstatscounter_udp4open, + isc_sockstatscounter_udp4openfail, + isc_sockstatscounter_udp4close, + isc_sockstatscounter_udp4bindfail, + isc_sockstatscounter_udp4connectfail, + isc_sockstatscounter_udp4connect, + -1, + -1, + isc_sockstatscounter_udp4sendfail, + isc_sockstatscounter_udp4recvfail, + isc_sockstatscounter_udp4active +}; +static const isc_statscounter_t udp6statsindex[] = { + isc_sockstatscounter_udp6open, + isc_sockstatscounter_udp6openfail, + isc_sockstatscounter_udp6close, + isc_sockstatscounter_udp6bindfail, + isc_sockstatscounter_udp6connectfail, + isc_sockstatscounter_udp6connect, + -1, + -1, + isc_sockstatscounter_udp6sendfail, + isc_sockstatscounter_udp6recvfail, + isc_sockstatscounter_udp6active +}; +static const isc_statscounter_t tcp4statsindex[] = { + isc_sockstatscounter_tcp4open, + isc_sockstatscounter_tcp4openfail, + isc_sockstatscounter_tcp4close, + isc_sockstatscounter_tcp4bindfail, + isc_sockstatscounter_tcp4connectfail, + isc_sockstatscounter_tcp4connect, + isc_sockstatscounter_tcp4acceptfail, + isc_sockstatscounter_tcp4accept, + isc_sockstatscounter_tcp4sendfail, + isc_sockstatscounter_tcp4recvfail, + isc_sockstatscounter_tcp4active +}; +static const isc_statscounter_t tcp6statsindex[] = { + isc_sockstatscounter_tcp6open, + isc_sockstatscounter_tcp6openfail, + isc_sockstatscounter_tcp6close, + isc_sockstatscounter_tcp6bindfail, + isc_sockstatscounter_tcp6connectfail, + isc_sockstatscounter_tcp6connect, + isc_sockstatscounter_tcp6acceptfail, + isc_sockstatscounter_tcp6accept, + isc_sockstatscounter_tcp6sendfail, + isc_sockstatscounter_tcp6recvfail, + isc_sockstatscounter_tcp6active +}; +static const isc_statscounter_t unixstatsindex[] = { + isc_sockstatscounter_unixopen, + isc_sockstatscounter_unixopenfail, + isc_sockstatscounter_unixclose, + isc_sockstatscounter_unixbindfail, + isc_sockstatscounter_unixconnectfail, + isc_sockstatscounter_unixconnect, + isc_sockstatscounter_unixacceptfail, + isc_sockstatscounter_unixaccept, + isc_sockstatscounter_unixsendfail, + isc_sockstatscounter_unixrecvfail, + isc_sockstatscounter_unixactive +}; +static const isc_statscounter_t fdwatchstatsindex[] = { + -1, + -1, + isc_sockstatscounter_fdwatchclose, + isc_sockstatscounter_fdwatchbindfail, + isc_sockstatscounter_fdwatchconnectfail, + isc_sockstatscounter_fdwatchconnect, + -1, + -1, + isc_sockstatscounter_fdwatchsendfail, + isc_sockstatscounter_fdwatchrecvfail, + -1 +}; +static const isc_statscounter_t rawstatsindex[] = { + isc_sockstatscounter_rawopen, + isc_sockstatscounter_rawopenfail, + isc_sockstatscounter_rawclose, + -1, + -1, + -1, + -1, + -1, + -1, + isc_sockstatscounter_rawrecvfail, + isc_sockstatscounter_rawactive +}; + +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) || \ + defined(USE_WATCHER_THREAD) +static void +manager_log(isc__socketmgr_t *sockmgr, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + const char *fmt, ...) ISC_FORMAT_PRINTF(5, 6); +static void +manager_log(isc__socketmgr_t *sockmgr, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + const char *fmt, ...) +{ + char msgbuf[2048]; + va_list ap; + + if (! isc_log_wouldlog(isc_lctx, level)) + return; + + va_start(ap, fmt); + vsnprintf(msgbuf, sizeof(msgbuf), fmt, ap); + va_end(ap); + + isc_log_write(isc_lctx, category, module, level, + "sockmgr %p: %s", sockmgr, msgbuf); +} +#endif + +static void +socket_log(isc__socket_t *sock, isc_sockaddr_t *address, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + isc_msgcat_t *msgcat, int msgset, int message, + const char *fmt, ...) ISC_FORMAT_PRINTF(9, 10); +static void +socket_log(isc__socket_t *sock, isc_sockaddr_t *address, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + isc_msgcat_t *msgcat, int msgset, int message, + const char *fmt, ...) +{ + char msgbuf[2048]; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + va_list ap; + + if (! isc_log_wouldlog(isc_lctx, level)) + return; + + va_start(ap, fmt); + vsnprintf(msgbuf, sizeof(msgbuf), fmt, ap); + va_end(ap); + + if (address == NULL) { + isc_log_iwrite(isc_lctx, category, module, level, + msgcat, msgset, message, + "socket %p: %s", sock, msgbuf); + } else { + isc_sockaddr_format(address, peerbuf, sizeof(peerbuf)); + isc_log_iwrite(isc_lctx, category, module, level, + msgcat, msgset, message, + "socket %p %s: %s", sock, peerbuf, msgbuf); + } +} + +#if defined(_AIX) && defined(ISC_NET_BSD44MSGHDR) && \ + defined(USE_CMSG) && defined(IPV6_RECVPKTINFO) +/* + * AIX has a kernel bug where IPV6_RECVPKTINFO gets cleared by + * setting IPV6_V6ONLY. + */ +static void +FIX_IPV6_RECVPKTINFO(isc__socket_t *sock) +{ + char strbuf[ISC_STRERRORSIZE]; + int on = 1; + + if (sock->pf != AF_INET6 || sock->type != isc_sockettype_udp) + return; + + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, + (void *)&on, sizeof(on)) < 0) { + + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_RECVPKTINFO) " + "%s: %s", sock->fd, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } +} +#else +#define FIX_IPV6_RECVPKTINFO(sock) (void)0 +#endif + +/*% + * Increment socket-related statistics counters. + */ +static inline void +inc_stats(isc_stats_t *stats, isc_statscounter_t counterid) { + REQUIRE(counterid != -1); + + if (stats != NULL) + isc_stats_increment(stats, counterid); +} + +/*% + * Decrement socket-related statistics counters. + */ +static inline void +dec_stats(isc_stats_t *stats, isc_statscounter_t counterid) { + REQUIRE(counterid != -1); + + if (stats != NULL) + isc_stats_decrement(stats, counterid); +} + +static inline isc_result_t +watch_fd(isc__socketmgr_t *manager, int fd, int msg) { + isc_result_t result = ISC_R_SUCCESS; + +#ifdef USE_KQUEUE + struct kevent evchange; + + memset(&evchange, 0, sizeof(evchange)); + if (msg == SELECT_POKE_READ) + evchange.filter = EVFILT_READ; + else + evchange.filter = EVFILT_WRITE; + evchange.flags = EV_ADD; + evchange.ident = fd; + if (kevent(manager->kqueue_fd, &evchange, 1, NULL, 0, NULL) != 0) + result = isc__errno2result(errno); + + return (result); +#elif defined(USE_EPOLL) + struct epoll_event event; + uint32_t oldevents; + int ret; + int op; + + oldevents = manager->epoll_events[fd]; + if (msg == SELECT_POKE_READ) + manager->epoll_events[fd] |= EPOLLIN; + else + manager->epoll_events[fd] |= EPOLLOUT; + + event.events = manager->epoll_events[fd]; + memset(&event.data, 0, sizeof(event.data)); + event.data.fd = fd; + + op = (oldevents == 0U) ? EPOLL_CTL_ADD : EPOLL_CTL_MOD; + ret = epoll_ctl(manager->epoll_fd, op, fd, &event); + if (ret == -1) { + if (errno == EEXIST) + UNEXPECTED_ERROR(__FILE__, __LINE__, + "epoll_ctl(ADD/MOD) returned " + "EEXIST for fd %d", fd); + result = isc__errno2result(errno); + } + + return (result); +#elif defined(USE_DEVPOLL) + struct pollfd pfd; + int lockid = FDLOCK_ID(fd); + + memset(&pfd, 0, sizeof(pfd)); + if (msg == SELECT_POKE_READ) + pfd.events = POLLIN; + else + pfd.events = POLLOUT; + pfd.fd = fd; + pfd.revents = 0; + LOCK(&manager->fdlock[lockid]); + if (write(manager->devpoll_fd, &pfd, sizeof(pfd)) == -1) + result = isc__errno2result(errno); + else { + if (msg == SELECT_POKE_READ) + manager->fdpollinfo[fd].want_read = 1; + else + manager->fdpollinfo[fd].want_write = 1; + } + UNLOCK(&manager->fdlock[lockid]); + + return (result); +#elif defined(USE_SELECT) + LOCK(&manager->lock); + if (msg == SELECT_POKE_READ) + FD_SET(fd, manager->read_fds); + if (msg == SELECT_POKE_WRITE) + FD_SET(fd, manager->write_fds); + UNLOCK(&manager->lock); + + return (result); +#endif +} + +static inline isc_result_t +unwatch_fd(isc__socketmgr_t *manager, int fd, int msg) { + isc_result_t result = ISC_R_SUCCESS; + +#ifdef USE_KQUEUE + struct kevent evchange; + + memset(&evchange, 0, sizeof(evchange)); + if (msg == SELECT_POKE_READ) + evchange.filter = EVFILT_READ; + else + evchange.filter = EVFILT_WRITE; + evchange.flags = EV_DELETE; + evchange.ident = fd; + if (kevent(manager->kqueue_fd, &evchange, 1, NULL, 0, NULL) != 0) + result = isc__errno2result(errno); + + return (result); +#elif defined(USE_EPOLL) + struct epoll_event event; + int ret; + int op; + + if (msg == SELECT_POKE_READ) + manager->epoll_events[fd] &= ~(EPOLLIN); + else + manager->epoll_events[fd] &= ~(EPOLLOUT); + + event.events = manager->epoll_events[fd]; + memset(&event.data, 0, sizeof(event.data)); + event.data.fd = fd; + + op = (event.events == 0U) ? EPOLL_CTL_DEL : EPOLL_CTL_MOD; + ret = epoll_ctl(manager->epoll_fd, op, fd, &event); + if (ret == -1 && errno != ENOENT) { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "epoll_ctl(DEL), %d: %s", fd, strbuf); + result = ISC_R_UNEXPECTED; + } + return (result); +#elif defined(USE_DEVPOLL) + struct pollfd pfds[2]; + size_t writelen = sizeof(pfds[0]); + int lockid = FDLOCK_ID(fd); + + memset(pfds, 0, sizeof(pfds)); + pfds[0].events = POLLREMOVE; + pfds[0].fd = fd; + + /* + * Canceling read or write polling via /dev/poll is tricky. Since it + * only provides a way of canceling per FD, we may need to re-poll the + * socket for the other operation. + */ + LOCK(&manager->fdlock[lockid]); + if (msg == SELECT_POKE_READ && + manager->fdpollinfo[fd].want_write == 1) { + pfds[1].events = POLLOUT; + pfds[1].fd = fd; + writelen += sizeof(pfds[1]); + } + if (msg == SELECT_POKE_WRITE && + manager->fdpollinfo[fd].want_read == 1) { + pfds[1].events = POLLIN; + pfds[1].fd = fd; + writelen += sizeof(pfds[1]); + } + + if (write(manager->devpoll_fd, pfds, writelen) == -1) + result = isc__errno2result(errno); + else { + if (msg == SELECT_POKE_READ) + manager->fdpollinfo[fd].want_read = 0; + else + manager->fdpollinfo[fd].want_write = 0; + } + UNLOCK(&manager->fdlock[lockid]); + + return (result); +#elif defined(USE_SELECT) + LOCK(&manager->lock); + if (msg == SELECT_POKE_READ) + FD_CLR(fd, manager->read_fds); + else if (msg == SELECT_POKE_WRITE) + FD_CLR(fd, manager->write_fds); + UNLOCK(&manager->lock); + + return (result); +#endif +} + +static void +wakeup_socket(isc__socketmgr_t *manager, int fd, int msg) { + isc_result_t result; + int lockid = FDLOCK_ID(fd); + + /* + * This is a wakeup on a socket. If the socket is not in the + * process of being closed, start watching it for either reads + * or writes. + */ + + INSIST(fd >= 0 && fd < (int)manager->maxsocks); + + if (msg == SELECT_POKE_CLOSE) { + /* No one should be updating fdstate, so no need to lock it */ + INSIST(manager->fdstate[fd] == CLOSE_PENDING); + manager->fdstate[fd] = CLOSED; + (void)unwatch_fd(manager, fd, SELECT_POKE_READ); + (void)unwatch_fd(manager, fd, SELECT_POKE_WRITE); + (void)close(fd); + return; + } + + LOCK(&manager->fdlock[lockid]); + if (manager->fdstate[fd] == CLOSE_PENDING) { + UNLOCK(&manager->fdlock[lockid]); + + /* + * We accept (and ignore) any error from unwatch_fd() as we are + * closing the socket, hoping it doesn't leave dangling state in + * the kernel. + * Note that unwatch_fd() must be called after releasing the + * fdlock; otherwise it could cause deadlock due to a lock order + * reversal. + */ + (void)unwatch_fd(manager, fd, SELECT_POKE_READ); + (void)unwatch_fd(manager, fd, SELECT_POKE_WRITE); + return; + } + if (manager->fdstate[fd] != MANAGED) { + UNLOCK(&manager->fdlock[lockid]); + return; + } + UNLOCK(&manager->fdlock[lockid]); + + /* + * Set requested bit. + */ + result = watch_fd(manager, fd, msg); + if (result != ISC_R_SUCCESS) { + /* + * XXXJT: what should we do? Ignoring the failure of watching + * a socket will make the application dysfunctional, but there + * seems to be no reasonable recovery process. + */ + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "failed to start watching FD (%d): %s", + fd, isc_result_totext(result)); + } +} + +#ifdef USE_WATCHER_THREAD +/* + * Poke the select loop when there is something for us to do. + * The write is required (by POSIX) to complete. That is, we + * will not get partial writes. + */ +static void +select_poke(isc__socketmgr_t *mgr, int fd, int msg) { + int cc; + int buf[2]; + char strbuf[ISC_STRERRORSIZE]; + + buf[0] = fd; + buf[1] = msg; + + do { + cc = write(mgr->pipe_fds[1], buf, sizeof(buf)); +#ifdef ENOSR + /* + * Treat ENOSR as EAGAIN but loop slowly as it is + * unlikely to clear fast. + */ + if (cc < 0 && errno == ENOSR) { + sleep(1); + errno = EAGAIN; + } +#endif + } while (cc < 0 && SOFT_ERROR(errno)); + + if (cc < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_WRITEFAILED, + "write() failed " + "during watcher poke: %s"), + strbuf); + } + + INSIST(cc == sizeof(buf)); +} + +/* + * Read a message on the internal fd. + */ +static void +select_readmsg(isc__socketmgr_t *mgr, int *fd, int *msg) { + int buf[2]; + int cc; + char strbuf[ISC_STRERRORSIZE]; + + cc = read(mgr->pipe_fds[0], buf, sizeof(buf)); + if (cc < 0) { + *msg = SELECT_POKE_NOTHING; + *fd = -1; /* Silence compiler. */ + if (SOFT_ERROR(errno)) + return; + + isc__strerror(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_READFAILED, + "read() failed " + "during watcher poke: %s"), + strbuf); + } + INSIST(cc == sizeof(buf)); + + *fd = buf[0]; + *msg = buf[1]; +} +#else /* USE_WATCHER_THREAD */ +/* + * Update the state of the socketmgr when something changes. + */ +static void +select_poke(isc__socketmgr_t *manager, int fd, int msg) { + if (msg == SELECT_POKE_SHUTDOWN) + return; + else if (fd >= 0) + wakeup_socket(manager, fd, msg); + return; +} +#endif /* USE_WATCHER_THREAD */ + +/* + * Make a fd non-blocking. + */ +static isc_result_t +make_nonblock(int fd) { + int ret; + char strbuf[ISC_STRERRORSIZE]; +#ifdef USE_FIONBIO_IOCTL + int on = 1; +#else + int flags; +#endif + +#ifdef USE_FIONBIO_IOCTL + ret = ioctl(fd, FIONBIO, (char *)&on); +#else + flags = fcntl(fd, F_GETFL, 0); + flags |= PORT_NONBLOCK; + ret = fcntl(fd, F_SETFL, flags); +#endif + + if (ret == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, +#ifdef USE_FIONBIO_IOCTL + "ioctl(%d, FIONBIO, &on): %s", fd, +#else + "fcntl(%d, F_SETFL, %d): %s", fd, flags, +#endif + strbuf); + + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +#ifdef USE_CMSG +/* + * Not all OSes support advanced CMSG macros: CMSG_LEN and CMSG_SPACE. + * In order to ensure as much portability as possible, we provide wrapper + * functions of these macros. + * Note that cmsg_space() could run slow on OSes that do not have + * CMSG_SPACE. + */ +static inline ISC_SOCKADDR_LEN_T +cmsg_len(ISC_SOCKADDR_LEN_T len) { +#ifdef CMSG_LEN + return (CMSG_LEN(len)); +#else + ISC_SOCKADDR_LEN_T hdrlen; + + /* + * Cast NULL so that any pointer arithmetic performed by CMSG_DATA + * is correct. + */ + hdrlen = (ISC_SOCKADDR_LEN_T)CMSG_DATA(((struct cmsghdr *)NULL)); + return (hdrlen + len); +#endif +} + +static inline ISC_SOCKADDR_LEN_T +cmsg_space(ISC_SOCKADDR_LEN_T len) { +#ifdef CMSG_SPACE + return (CMSG_SPACE(len)); +#else + struct msghdr msg; + struct cmsghdr *cmsgp; + /* + * XXX: The buffer length is an ad-hoc value, but should be enough + * in a practical sense. + */ + char dummybuf[sizeof(struct cmsghdr) + 1024]; + + memset(&msg, 0, sizeof(msg)); + msg.msg_control = dummybuf; + msg.msg_controllen = sizeof(dummybuf); + + cmsgp = (struct cmsghdr *)dummybuf; + cmsgp->cmsg_len = cmsg_len(len); + + cmsgp = CMSG_NXTHDR(&msg, cmsgp); + if (cmsgp != NULL) + return ((char *)cmsgp - (char *)msg.msg_control); + else + return (0); +#endif +} +#endif /* USE_CMSG */ + +/* + * Process control messages received on a socket. + */ +static void +process_cmsg(isc__socket_t *sock, struct msghdr *msg, isc_socketevent_t *dev) { +#ifdef ISC_NET_BSD44MSGHDR +#ifdef USE_CMSG + struct cmsghdr *cmsgp; +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO + struct in6_pktinfo *pktinfop; +#endif +#ifdef SO_TIMESTAMP + void *timevalp; +#endif +#endif +#endif + + /* + * sock is used only when ISC_NET_BSD44MSGHDR and USE_CMSG are defined. + * msg and dev are used only when ISC_NET_BSD44MSGHDR is defined. + * They are all here, outside of the CPP tests, because it is + * more consistent with the usual ISC coding style. + */ + UNUSED(sock); + UNUSED(msg); + UNUSED(dev); + +#ifdef ISC_NET_BSD44MSGHDR + +#ifdef MSG_TRUNC + if ((msg->msg_flags & MSG_TRUNC) == MSG_TRUNC) + dev->attributes |= ISC_SOCKEVENTATTR_TRUNC; +#endif + +#ifdef MSG_CTRUNC + if ((msg->msg_flags & MSG_CTRUNC) == MSG_CTRUNC) + dev->attributes |= ISC_SOCKEVENTATTR_CTRUNC; +#endif + +#ifndef USE_CMSG + return; +#else + if (msg->msg_controllen == 0U || msg->msg_control == NULL) + return; + +#ifdef SO_TIMESTAMP + timevalp = NULL; +#endif +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO + pktinfop = NULL; +#endif + + cmsgp = CMSG_FIRSTHDR(msg); + while (cmsgp != NULL) { + socket_log(sock, NULL, TRACE, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_PROCESSCMSG, + "processing cmsg %p", cmsgp); + +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO + if (cmsgp->cmsg_level == IPPROTO_IPV6 + && cmsgp->cmsg_type == IPV6_PKTINFO) { + + pktinfop = (struct in6_pktinfo *)CMSG_DATA(cmsgp); + memmove(&dev->pktinfo, pktinfop, + sizeof(struct in6_pktinfo)); + dev->attributes |= ISC_SOCKEVENTATTR_PKTINFO; + socket_log(sock, NULL, TRACE, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_IFRECEIVED, + "interface received on ifindex %u", + dev->pktinfo.ipi6_ifindex); + if (IN6_IS_ADDR_MULTICAST(&pktinfop->ipi6_addr)) + dev->attributes |= ISC_SOCKEVENTATTR_MULTICAST; + goto next; + } +#endif + +#ifdef SO_TIMESTAMP + if (cmsgp->cmsg_level == SOL_SOCKET + && cmsgp->cmsg_type == SCM_TIMESTAMP) { + struct timeval tv; + timevalp = CMSG_DATA(cmsgp); + memmove(&tv, timevalp, sizeof(tv)); + dev->timestamp.seconds = tv.tv_sec; + dev->timestamp.nanoseconds = tv.tv_usec * 1000; + dev->attributes |= ISC_SOCKEVENTATTR_TIMESTAMP; + goto next; + } +#endif + +#ifdef IPV6_TCLASS + if (cmsgp->cmsg_level == IPPROTO_IPV6 + && cmsgp->cmsg_type == IPV6_TCLASS) { + dev->dscp = *(int *)CMSG_DATA(cmsgp); + dev->dscp >>= 2; + dev->attributes |= ISC_SOCKEVENTATTR_DSCP; + goto next; + } +#endif + +#ifdef IP_TOS + if (cmsgp->cmsg_level == IPPROTO_IP + && (cmsgp->cmsg_type == IP_TOS +#ifdef IP_RECVTOS + || cmsgp->cmsg_type == IP_RECVTOS +#endif + )) { + dev->dscp = (int) *(unsigned char *)CMSG_DATA(cmsgp); + dev->dscp >>= 2; + dev->attributes |= ISC_SOCKEVENTATTR_DSCP; + goto next; + } +#endif + next: + cmsgp = CMSG_NXTHDR(msg, cmsgp); + } +#endif /* USE_CMSG */ + +#endif /* ISC_NET_BSD44MSGHDR */ +} + +/* + * Construct an iov array and attach it to the msghdr passed in. This is + * the SEND constructor, which will use the used region of the buffer + * (if using a buffer list) or will use the internal region (if a single + * buffer I/O is requested). + * + * Nothing can be NULL, and the done event must list at least one buffer + * on the buffer linked list for this function to be meaningful. + * + * If write_countp != NULL, *write_countp will hold the number of bytes + * this transaction can send. + */ +static void +build_msghdr_send(isc__socket_t *sock, char* cmsgbuf, isc_socketevent_t *dev, + struct msghdr *msg, struct iovec *iov, size_t *write_countp) +{ + unsigned int iovcount; + isc_buffer_t *buffer; + isc_region_t used; + size_t write_count; + size_t skip_count; +#ifdef ISC_NET_BSD44MSGHDR + struct cmsghdr *cmsgp; +#endif + + memset(msg, 0, sizeof(*msg)); + + if (!sock->connected) { + msg->msg_name = (void *)&dev->address.type.sa; + msg->msg_namelen = dev->address.length; + } else { + msg->msg_name = NULL; + msg->msg_namelen = 0; + } + + buffer = ISC_LIST_HEAD(dev->bufferlist); + write_count = 0; + iovcount = 0; + + /* + * Single buffer I/O? Skip what we've done so far in this region. + */ + if (buffer == NULL) { + write_count = dev->region.length - dev->n; + iov[0].iov_base = (void *)(dev->region.base + dev->n); + iov[0].iov_len = write_count; + iovcount = 1; + + goto config; + } + + /* + * Multibuffer I/O. + * Skip the data in the buffer list that we have already written. + */ + skip_count = dev->n; + while (buffer != NULL) { + REQUIRE(ISC_BUFFER_VALID(buffer)); + if (skip_count < isc_buffer_usedlength(buffer)) + break; + skip_count -= isc_buffer_usedlength(buffer); + buffer = ISC_LIST_NEXT(buffer, link); + } + + while (buffer != NULL) { + INSIST(iovcount < MAXSCATTERGATHER_SEND); + + isc_buffer_usedregion(buffer, &used); + + if (used.length > 0) { + iov[iovcount].iov_base = (void *)(used.base + + skip_count); + iov[iovcount].iov_len = used.length - skip_count; + write_count += (used.length - skip_count); + skip_count = 0; + iovcount++; + } + buffer = ISC_LIST_NEXT(buffer, link); + } + + INSIST(skip_count == 0U); + + config: + msg->msg_iov = iov; + msg->msg_iovlen = iovcount; + +#ifdef ISC_NET_BSD44MSGHDR + msg->msg_control = NULL; + msg->msg_controllen = 0; + msg->msg_flags = 0; +#if defined(USE_CMSG) + +#if defined(ISC_PLATFORM_HAVEIN6PKTINFO) + if ((sock->type == isc_sockettype_udp) && + ((dev->attributes & ISC_SOCKEVENTATTR_PKTINFO) != 0)) + { + struct in6_pktinfo *pktinfop; + + socket_log(sock, NULL, TRACE, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_SENDTODATA, + "sendto pktinfo data, ifindex %u", + dev->pktinfo.ipi6_ifindex); + + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen = cmsg_space(sizeof(struct in6_pktinfo)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp = (struct cmsghdr *)cmsgbuf; + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_PKTINFO; + cmsgp->cmsg_len = cmsg_len(sizeof(struct in6_pktinfo)); + pktinfop = (struct in6_pktinfo *)CMSG_DATA(cmsgp); + memmove(pktinfop, &dev->pktinfo, sizeof(struct in6_pktinfo)); + } +#endif + +#if defined(IPV6_USE_MIN_MTU) + if ((sock->type == isc_sockettype_udp) && + ((dev->attributes & ISC_SOCKEVENTATTR_USEMINMTU) != 0)) + { + int use_min_mtu = 1; /* -1, 0, 1 */ + + cmsgp = (struct cmsghdr *)(cmsgbuf + + msg->msg_controllen); + + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(use_min_mtu)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_USE_MIN_MTU; + cmsgp->cmsg_len = cmsg_len(sizeof(use_min_mtu)); + memmove(CMSG_DATA(cmsgp), &use_min_mtu, sizeof(use_min_mtu)); + } +#endif + + if (isc_dscp_check_value > -1) { + if (sock->type == isc_sockettype_udp) + INSIST((int)dev->dscp == isc_dscp_check_value); + else if (sock->type == isc_sockettype_tcp) + INSIST((int)sock->dscp == isc_dscp_check_value); + } + +#if defined(IP_TOS) || (defined(IPPROTO_IPV6) && defined(IPV6_TCLASS)) + if ((sock->type == isc_sockettype_udp) && + ((dev->attributes & ISC_SOCKEVENTATTR_DSCP) != 0)) + { + int dscp = (dev->dscp << 2) & 0xff; + + INSIST(dev->dscp < 0x40); + +#ifdef IP_TOS + if (sock->pf == AF_INET && sock->pktdscp) { + cmsgp = (struct cmsghdr *)(cmsgbuf + + msg->msg_controllen); + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(dscp)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IP; + cmsgp->cmsg_type = IP_TOS; + cmsgp->cmsg_len = cmsg_len(sizeof(char)); + *(unsigned char*)CMSG_DATA(cmsgp) = dscp; + } else if (sock->pf == AF_INET && sock->dscp != dev->dscp) { + if (setsockopt(sock->fd, IPPROTO_IP, IP_TOS, + (void *)&dscp, sizeof(int)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_TOS, %.02x)" + " %s: %s", + sock->fd, dscp >> 2, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } else + sock->dscp = dscp; + } +#endif +#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS) + if (sock->pf == AF_INET6 && sock->pktdscp) { + cmsgp = (struct cmsghdr *)(cmsgbuf + + msg->msg_controllen); + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(dscp)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_TCLASS; + cmsgp->cmsg_len = cmsg_len(sizeof(dscp)); + memmove(CMSG_DATA(cmsgp), &dscp, sizeof(dscp)); + } else if (sock->pf == AF_INET6 && sock->dscp != dev->dscp) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_TCLASS, + (void *)&dscp, sizeof(int)) < 0) { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_TCLASS, " + "%.02x) %s: %s", + sock->fd, dscp >> 2, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } else + sock->dscp = dscp; + } +#endif + if (msg->msg_controllen != 0 && + msg->msg_controllen < SENDCMSGBUFLEN) + { + memset(cmsgbuf + msg->msg_controllen, 0, + SENDCMSGBUFLEN - msg->msg_controllen); + } + } +#endif +#endif /* USE_CMSG */ +#else /* ISC_NET_BSD44MSGHDR */ + msg->msg_accrights = NULL; + msg->msg_accrightslen = 0; +#endif /* ISC_NET_BSD44MSGHDR */ + + if (write_countp != NULL) + *write_countp = write_count; +} + +/* + * Construct an iov array and attach it to the msghdr passed in. This is + * the RECV constructor, which will use the available region of the buffer + * (if using a buffer list) or will use the internal region (if a single + * buffer I/O is requested). + * + * Nothing can be NULL, and the done event must list at least one buffer + * on the buffer linked list for this function to be meaningful. + * + * If read_countp != NULL, *read_countp will hold the number of bytes + * this transaction can receive. + */ +static void +build_msghdr_recv(isc__socket_t *sock, char *cmsgbuf, isc_socketevent_t *dev, + struct msghdr *msg, struct iovec *iov, size_t *read_countp) +{ + unsigned int iovcount; + isc_buffer_t *buffer; + isc_region_t available; + size_t read_count; + + memset(msg, 0, sizeof(struct msghdr)); + + if (sock->type == isc_sockettype_udp) { + memset(&dev->address, 0, sizeof(dev->address)); +#ifdef BROKEN_RECVMSG + if (sock->pf == AF_INET) { + msg->msg_name = (void *)&dev->address.type.sin; + msg->msg_namelen = sizeof(dev->address.type.sin6); + } else if (sock->pf == AF_INET6) { + msg->msg_name = (void *)&dev->address.type.sin6; + msg->msg_namelen = sizeof(dev->address.type.sin6); +#ifdef ISC_PLATFORM_HAVESYSUNH + } else if (sock->pf == AF_UNIX) { + msg->msg_name = (void *)&dev->address.type.sunix; + msg->msg_namelen = sizeof(dev->address.type.sunix); +#endif + } else { + msg->msg_name = (void *)&dev->address.type.sa; + msg->msg_namelen = sizeof(dev->address.type); + } +#else + msg->msg_name = (void *)&dev->address.type.sa; + msg->msg_namelen = sizeof(dev->address.type); +#endif + } else { /* TCP */ + msg->msg_name = NULL; + msg->msg_namelen = 0; + dev->address = sock->peer_address; + } + + buffer = ISC_LIST_HEAD(dev->bufferlist); + read_count = 0; + + /* + * Single buffer I/O? Skip what we've done so far in this region. + */ + if (buffer == NULL) { + read_count = dev->region.length - dev->n; + iov[0].iov_base = (void *)(dev->region.base + dev->n); + iov[0].iov_len = read_count; + iovcount = 1; + + goto config; + } + + /* + * Multibuffer I/O. + * Skip empty buffers. + */ + while (buffer != NULL) { + REQUIRE(ISC_BUFFER_VALID(buffer)); + if (isc_buffer_availablelength(buffer) != 0) + break; + buffer = ISC_LIST_NEXT(buffer, link); + } + + iovcount = 0; + while (buffer != NULL) { + INSIST(iovcount < MAXSCATTERGATHER_RECV); + + isc_buffer_availableregion(buffer, &available); + + if (available.length > 0) { + iov[iovcount].iov_base = (void *)(available.base); + iov[iovcount].iov_len = available.length; + read_count += available.length; + iovcount++; + } + buffer = ISC_LIST_NEXT(buffer, link); + } + + config: + + /* + * If needed, set up to receive that one extra byte. + */ +#ifdef ISC_PLATFORM_RECVOVERFLOW + if (sock->type == isc_sockettype_udp) { + INSIST(iovcount < MAXSCATTERGATHER_RECV); + iov[iovcount].iov_base = (void *)(&sock->overflow); + iov[iovcount].iov_len = 1; + iovcount++; + } +#endif + + msg->msg_iov = iov; + msg->msg_iovlen = iovcount; + +#ifdef ISC_NET_BSD44MSGHDR +#if defined(USE_CMSG) + msg->msg_control = cmsgbuf; + msg->msg_controllen = RECVCMSGBUFLEN; +#else + msg->msg_control = NULL; + msg->msg_controllen = 0; +#endif /* USE_CMSG */ + msg->msg_flags = 0; +#else /* ISC_NET_BSD44MSGHDR */ + msg->msg_accrights = NULL; + msg->msg_accrightslen = 0; +#endif /* ISC_NET_BSD44MSGHDR */ + + if (read_countp != NULL) + *read_countp = read_count; +} + +static void +set_dev_address(isc_sockaddr_t *address, isc__socket_t *sock, + isc_socketevent_t *dev) +{ + if (sock->type == isc_sockettype_udp) { + if (address != NULL) + dev->address = *address; + else + dev->address = sock->peer_address; + } else if (sock->type == isc_sockettype_tcp) { + INSIST(address == NULL); + dev->address = sock->peer_address; + } +} + +static void +destroy_socketevent(isc_event_t *event) { + isc_socketevent_t *ev = (isc_socketevent_t *)event; + + INSIST(ISC_LIST_EMPTY(ev->bufferlist)); + + (ev->destroy)(event); +} + +static isc_socketevent_t * +allocate_socketevent(isc_mem_t *mctx, void *sender, + isc_eventtype_t eventtype, isc_taskaction_t action, + void *arg) +{ + isc_socketevent_t *ev; + + ev = (isc_socketevent_t *)isc_event_allocate(mctx, sender, + eventtype, action, arg, + sizeof(*ev)); + + if (ev == NULL) + return (NULL); + + ev->result = ISC_R_UNSET; + ISC_LINK_INIT(ev, ev_link); + ISC_LIST_INIT(ev->bufferlist); + ev->region.base = NULL; + ev->n = 0; + ev->offset = 0; + ev->attributes = 0; + ev->destroy = ev->ev_destroy; + ev->ev_destroy = destroy_socketevent; + ev->dscp = 0; + + return (ev); +} + +#if defined(ISC_SOCKET_DEBUG) +static void +dump_msg(struct msghdr *msg) { + unsigned int i; + + printf("MSGHDR %p\n", msg); + printf("\tname %p, namelen %ld\n", msg->msg_name, + (long) msg->msg_namelen); + printf("\tiov %p, iovlen %ld\n", msg->msg_iov, + (long) msg->msg_iovlen); + for (i = 0; i < (unsigned int)msg->msg_iovlen; i++) + printf("\t\t%u\tbase %p, len %ld\n", i, + msg->msg_iov[i].iov_base, + (long) msg->msg_iov[i].iov_len); +#ifdef ISC_NET_BSD44MSGHDR + printf("\tcontrol %p, controllen %ld\n", msg->msg_control, + (long) msg->msg_controllen); +#endif +} +#endif + +#define DOIO_SUCCESS 0 /* i/o ok, event sent */ +#define DOIO_SOFT 1 /* i/o ok, soft error, no event sent */ +#define DOIO_HARD 2 /* i/o error, event sent */ +#define DOIO_EOF 3 /* EOF, no event sent */ + +static int +doio_recv(isc__socket_t *sock, isc_socketevent_t *dev) { + int cc; + struct iovec iov[MAXSCATTERGATHER_RECV]; + size_t read_count; + size_t actual_count; + struct msghdr msghdr; + isc_buffer_t *buffer; + int recv_errno; + char strbuf[ISC_STRERRORSIZE]; + char cmsgbuf[RECVCMSGBUFLEN] = {0}; + + build_msghdr_recv(sock, cmsgbuf, dev, &msghdr, iov, &read_count); + +#if defined(ISC_SOCKET_DEBUG) + dump_msg(&msghdr); +#endif + + cc = recvmsg(sock->fd, &msghdr, 0); + recv_errno = errno; + +#if defined(ISC_SOCKET_DEBUG) + dump_msg(&msghdr); +#endif + + if (cc < 0) { + if (SOFT_ERROR(recv_errno)) + return (DOIO_SOFT); + + if (isc_log_wouldlog(isc_lctx, IOEVENT_LEVEL)) { + isc__strerror(recv_errno, strbuf, sizeof(strbuf)); + socket_log(sock, NULL, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_DOIORECV, + "doio_recv: recvmsg(%d) %d bytes, err %d/%s", + sock->fd, cc, recv_errno, strbuf); + } + +#define SOFT_OR_HARD(_system, _isc) \ + if (recv_errno == _system) { \ + if (sock->connected) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_RECVFAIL]); \ + return (DOIO_HARD); \ + } \ + return (DOIO_SOFT); \ + } +#define ALWAYS_HARD(_system, _isc) \ + if (recv_errno == _system) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_RECVFAIL]); \ + return (DOIO_HARD); \ + } + + SOFT_OR_HARD(ECONNREFUSED, ISC_R_CONNREFUSED); + SOFT_OR_HARD(ENETUNREACH, ISC_R_NETUNREACH); + SOFT_OR_HARD(EHOSTUNREACH, ISC_R_HOSTUNREACH); + SOFT_OR_HARD(EHOSTDOWN, ISC_R_HOSTDOWN); + /* HPUX 11.11 can return EADDRNOTAVAIL. */ + SOFT_OR_HARD(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + SOFT_OR_HARD(ENOBUFS, ISC_R_NORESOURCES); + /* Should never get this one but it was seen. */ +#ifdef ENOPROTOOPT + SOFT_OR_HARD(ENOPROTOOPT, ISC_R_HOSTUNREACH); +#endif + /* + * HPUX returns EPROTO and EINVAL on receiving some ICMP/ICMPv6 + * errors. + */ +#ifdef EPROTO + SOFT_OR_HARD(EPROTO, ISC_R_HOSTUNREACH); +#endif + SOFT_OR_HARD(EINVAL, ISC_R_HOSTUNREACH); + +#undef SOFT_OR_HARD +#undef ALWAYS_HARD + + dev->result = isc__errno2result(recv_errno); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_RECVFAIL]); + return (DOIO_HARD); + } + + /* + * On TCP and UNIX sockets, zero length reads indicate EOF, + * while on UDP sockets, zero length reads are perfectly valid, + * although strange. + */ + switch (sock->type) { + case isc_sockettype_tcp: + case isc_sockettype_unix: + if (cc == 0) + return (DOIO_EOF); + break; + case isc_sockettype_udp: + case isc_sockettype_raw: + break; + case isc_sockettype_fdwatch: + default: + INSIST(0); + } + + if (sock->type == isc_sockettype_udp) { + dev->address.length = msghdr.msg_namelen; + if (isc_sockaddr_getport(&dev->address) == 0) { + if (isc_log_wouldlog(isc_lctx, IOEVENT_LEVEL)) { + socket_log(sock, &dev->address, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_ZEROPORT, + "dropping source port zero packet"); + } + return (DOIO_SOFT); + } + /* + * Simulate a firewall blocking UDP responses bigger than + * 'maxudp' bytes. + */ + if (sock->manager->maxudp != 0 && cc > sock->manager->maxudp) + return (DOIO_SOFT); + } + + socket_log(sock, &dev->address, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_PKTRECV, + "packet received correctly"); + + /* + * Overflow bit detection. If we received MORE bytes than we should, + * this indicates an overflow situation. Set the flag in the + * dev entry and adjust how much we read by one. + */ +#ifdef ISC_PLATFORM_RECVOVERFLOW + if ((sock->type == isc_sockettype_udp) && ((size_t)cc > read_count)) { + dev->attributes |= ISC_SOCKEVENTATTR_TRUNC; + cc--; + } +#endif + + /* + * If there are control messages attached, run through them and pull + * out the interesting bits. + */ + process_cmsg(sock, &msghdr, dev); + + /* + * update the buffers (if any) and the i/o count + */ + dev->n += cc; + actual_count = cc; + buffer = ISC_LIST_HEAD(dev->bufferlist); + while (buffer != NULL && actual_count > 0U) { + REQUIRE(ISC_BUFFER_VALID(buffer)); + if (isc_buffer_availablelength(buffer) <= actual_count) { + actual_count -= isc_buffer_availablelength(buffer); + isc_buffer_add(buffer, + isc_buffer_availablelength(buffer)); + } else { + isc_buffer_add(buffer, actual_count); + actual_count = 0; + POST(actual_count); + break; + } + buffer = ISC_LIST_NEXT(buffer, link); + if (buffer == NULL) { + INSIST(actual_count == 0U); + } + } + + /* + * If we read less than we expected, update counters, + * and let the upper layer poke the descriptor. + */ + if (((size_t)cc != read_count) && (dev->n < dev->minimum)) + return (DOIO_SOFT); + + /* + * Full reads are posted, or partials if partials are ok. + */ + dev->result = ISC_R_SUCCESS; + return (DOIO_SUCCESS); +} + +/* + * Returns: + * DOIO_SUCCESS The operation succeeded. dev->result contains + * ISC_R_SUCCESS. + * + * DOIO_HARD A hard or unexpected I/O error was encountered. + * dev->result contains the appropriate error. + * + * DOIO_SOFT A soft I/O error was encountered. No senddone + * event was sent. The operation should be retried. + * + * No other return values are possible. + */ +static int +doio_send(isc__socket_t *sock, isc_socketevent_t *dev) { + int cc; + struct iovec iov[MAXSCATTERGATHER_SEND]; + size_t write_count; + struct msghdr msghdr; + char addrbuf[ISC_SOCKADDR_FORMATSIZE]; + int attempts = 0; + int send_errno; + char strbuf[ISC_STRERRORSIZE]; + char cmsgbuf[SENDCMSGBUFLEN] = {0}; + + build_msghdr_send(sock, cmsgbuf, dev, &msghdr, iov, &write_count); + + resend: + if (sock->type == isc_sockettype_udp && + sock->manager->maxudp != 0 && + write_count > (size_t)sock->manager->maxudp) + cc = write_count; + else + cc = sendmsg(sock->fd, &msghdr, 0); + send_errno = errno; + + /* + * Check for error or block condition. + */ + if (cc < 0) { + if (send_errno == EINTR && ++attempts < NRETRIES) + goto resend; + + if (SOFT_ERROR(send_errno)) { + if (errno == EWOULDBLOCK || errno == EAGAIN) + dev->result = ISC_R_WOULDBLOCK; + return (DOIO_SOFT); + } + +#define SOFT_OR_HARD(_system, _isc) \ + if (send_errno == _system) { \ + if (sock->connected) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_SENDFAIL]); \ + return (DOIO_HARD); \ + } \ + return (DOIO_SOFT); \ + } +#define ALWAYS_HARD(_system, _isc) \ + if (send_errno == _system) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_SENDFAIL]); \ + return (DOIO_HARD); \ + } + + SOFT_OR_HARD(ECONNREFUSED, ISC_R_CONNREFUSED); + ALWAYS_HARD(EACCES, ISC_R_NOPERM); + ALWAYS_HARD(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ALWAYS_HARD(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ALWAYS_HARD(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ALWAYS_HARD(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif + ALWAYS_HARD(ENETUNREACH, ISC_R_NETUNREACH); + SOFT_OR_HARD(ENOBUFS, ISC_R_NORESOURCES); + ALWAYS_HARD(EPERM, ISC_R_HOSTUNREACH); + ALWAYS_HARD(EPIPE, ISC_R_NOTCONNECTED); + ALWAYS_HARD(ECONNRESET, ISC_R_CONNECTIONRESET); + +#undef SOFT_OR_HARD +#undef ALWAYS_HARD + + /* + * The other error types depend on whether or not the + * socket is UDP or TCP. If it is UDP, some errors + * that we expect to be fatal under TCP are merely + * annoying, and are really soft errors. + * + * However, these soft errors are still returned as + * a status. + */ + isc_sockaddr_format(&dev->address, addrbuf, sizeof(addrbuf)); + isc__strerror(send_errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "internal_send: %s: %s", + addrbuf, strbuf); + dev->result = isc__errno2result(send_errno); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_SENDFAIL]); + return (DOIO_HARD); + } + + if (cc == 0) { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_SENDFAIL]); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "doio_send: send() %s 0", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_RETURNED, "returned")); + } + + /* + * If we write less than we expected, update counters, poke. + */ + dev->n += cc; + if ((size_t)cc != write_count) + return (DOIO_SOFT); + + /* + * Exactly what we wanted to write. We're done with this + * entry. Post its completion event. + */ + dev->result = ISC_R_SUCCESS; + return (DOIO_SUCCESS); +} + +/* + * Kill. + * + * Caller must ensure that the socket is not locked and no external + * references exist. + */ +static void +socketclose(isc__socketmgr_t *manager, isc__socket_t *sock, int fd) { + isc_sockettype_t type = sock->type; + int lockid = FDLOCK_ID(fd); + + /* + * No one has this socket open, so the watcher doesn't have to be + * poked, and the socket doesn't have to be locked. + */ + LOCK(&manager->fdlock[lockid]); + manager->fds[fd] = NULL; + if (type == isc_sockettype_fdwatch) + manager->fdstate[fd] = CLOSED; + else + manager->fdstate[fd] = CLOSE_PENDING; + UNLOCK(&manager->fdlock[lockid]); + if (type == isc_sockettype_fdwatch) { + /* + * The caller may close the socket once this function returns, + * and `fd' may be reassigned for a new socket. So we do + * unwatch_fd() here, rather than defer it via select_poke(). + * Note: this may complicate data protection among threads and + * may reduce performance due to additional locks. One way to + * solve this would be to dup() the watched descriptor, but we + * take a simpler approach at this moment. + */ + (void)unwatch_fd(manager, fd, SELECT_POKE_READ); + (void)unwatch_fd(manager, fd, SELECT_POKE_WRITE); + } else + select_poke(manager, fd, SELECT_POKE_CLOSE); + + inc_stats(manager->stats, sock->statsindex[STATID_CLOSE]); + if (sock->active == 1) { + dec_stats(manager->stats, sock->statsindex[STATID_ACTIVE]); + sock->active = 0; + } + + /* + * update manager->maxfd here (XXX: this should be implemented more + * efficiently) + */ +#ifdef USE_SELECT + LOCK(&manager->lock); + if (manager->maxfd == fd) { + int i; + + manager->maxfd = 0; + for (i = fd - 1; i >= 0; i--) { + lockid = FDLOCK_ID(i); + + LOCK(&manager->fdlock[lockid]); + if (manager->fdstate[i] == MANAGED) { + manager->maxfd = i; + UNLOCK(&manager->fdlock[lockid]); + break; + } + UNLOCK(&manager->fdlock[lockid]); + } +#ifdef ISC_PLATFORM_USETHREADS + if (manager->maxfd < manager->pipe_fds[0]) + manager->maxfd = manager->pipe_fds[0]; +#endif + } + + UNLOCK(&manager->lock); +#endif /* USE_SELECT */ +} + +static void +destroy(isc__socket_t **sockp) { + int fd; + isc__socket_t *sock = *sockp; + isc__socketmgr_t *manager = sock->manager; + + socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_DESTROYING, "destroying"); + + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(sock->fd >= -1 && sock->fd < (int)manager->maxsocks); + + if (sock->fd >= 0) { + fd = sock->fd; + sock->fd = -1; + socketclose(manager, sock, fd); + } + + LOCK(&manager->lock); + + ISC_LIST_UNLINK(manager->socklist, sock, link); + +#ifdef USE_WATCHER_THREAD + if (ISC_LIST_EMPTY(manager->socklist)) + SIGNAL(&manager->shutdown_ok); +#endif /* USE_WATCHER_THREAD */ + + /* can't unlock manager as its memory context is still used */ + free_socket(sockp); + + UNLOCK(&manager->lock); +} + +static isc_result_t +allocate_socket(isc__socketmgr_t *manager, isc_sockettype_t type, + isc__socket_t **socketp) +{ + isc__socket_t *sock; + isc_result_t result; + + sock = isc_mem_get(manager->mctx, sizeof(*sock)); + + if (sock == NULL) + return (ISC_R_NOMEMORY); + + sock->common.magic = 0; + sock->common.impmagic = 0; + sock->references = 0; + + sock->manager = manager; + sock->type = type; + sock->fd = -1; + sock->dscp = 0; /* TOS/TCLASS is zero until set. */ + sock->dupped = 0; + sock->statsindex = NULL; + sock->active = 0; + + ISC_LINK_INIT(sock, link); + + + memset(sock->name, 0, sizeof(sock->name)); + sock->tag = NULL; + + /* + * Set up list of readers and writers to be initially empty. + */ + ISC_LIST_INIT(sock->recv_list); + ISC_LIST_INIT(sock->send_list); + ISC_LIST_INIT(sock->accept_list); + ISC_LIST_INIT(sock->connect_list); + sock->pending_recv = 0; + sock->pending_send = 0; + sock->pending_accept = 0; + sock->listener = 0; + sock->connected = 0; + sock->connecting = 0; + sock->bound = 0; + sock->pktdscp = 0; + + /* + * Initialize the lock. + */ + result = isc_mutex_init(&sock->lock); + if (result != ISC_R_SUCCESS) { + sock->common.magic = 0; + sock->common.impmagic = 0; + goto error; + } + + /* + * Initialize readable and writable events. + */ + ISC_EVENT_INIT(&sock->readable_ev, sizeof(intev_t), + ISC_EVENTATTR_NOPURGE, NULL, ISC_SOCKEVENT_INTR, + NULL, sock, sock, NULL, NULL); + ISC_EVENT_INIT(&sock->writable_ev, sizeof(intev_t), + ISC_EVENTATTR_NOPURGE, NULL, ISC_SOCKEVENT_INTW, + NULL, sock, sock, NULL, NULL); + + sock->common.magic = ISCAPI_SOCKET_MAGIC; + sock->common.impmagic = SOCKET_MAGIC; + *socketp = sock; + + return (ISC_R_SUCCESS); + + error: + isc_mem_put(manager->mctx, sock, sizeof(*sock)); + + return (result); +} + +/* + * This event requires that the various lists be empty, that the reference + * count be 1, and that the magic number is valid. The other socket bits, + * like the lock, must be initialized as well. The fd associated must be + * marked as closed, by setting it to -1 on close, or this routine will + * also close the socket. + */ +static void +free_socket(isc__socket_t **socketp) { + isc__socket_t *sock = *socketp; + + INSIST(VALID_SOCKET(sock)); + INSIST(sock->references == 0); + INSIST(!sock->connecting); + INSIST(!sock->pending_recv); + INSIST(!sock->pending_send); + INSIST(!sock->pending_accept); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + INSIST(!ISC_LINK_LINKED(sock, link)); + + sock->common.magic = 0; + sock->common.impmagic = 0; + + DESTROYLOCK(&sock->lock); + + isc_mem_put(sock->manager->mctx, sock, sizeof(*sock)); + + *socketp = NULL; +} + +#ifdef SO_RCVBUF +static isc_once_t rcvbuf_once = ISC_ONCE_INIT; +static int rcvbuf = RCVBUFSIZE; + +static void +set_rcvbuf(void) { + int fd; + int max = rcvbuf, min; + ISC_SOCKADDR_LEN_T len; + + fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); +#if defined(ISC_PLATFORM_HAVEIPV6) + if (fd == -1) { + switch (errno) { + case EPROTONOSUPPORT: + case EPFNOSUPPORT: + case EAFNOSUPPORT: + /* + * Linux 2.2 (and maybe others) return EINVAL instead of + * EAFNOSUPPORT. + */ + case EINVAL: + fd = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP); + break; + } + } +#endif + if (fd == -1) + return; + + len = sizeof(min); + if (getsockopt(fd, SOL_SOCKET, SO_RCVBUF, (void *)&min, &len) == 0 && + min < rcvbuf) { + again: + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, (void *)&rcvbuf, + sizeof(rcvbuf)) == -1) { + if (errno == ENOBUFS && rcvbuf > min) { + max = rcvbuf - 1; + rcvbuf = (rcvbuf + min) / 2; + goto again; + } else { + rcvbuf = min; + goto cleanup; + } + } else + min = rcvbuf; + if (min != max) { + rcvbuf = max; + goto again; + } + } + cleanup: + close (fd); +} +#endif + +#ifdef SO_BSDCOMPAT +/* + * This really should not be necessary to do. Having to workout + * which kernel version we are on at run time so that we don't cause + * the kernel to issue a warning about us using a deprecated socket option. + * Such warnings should *never* be on by default in production kernels. + * + * We can't do this a build time because executables are moved between + * machines and hence kernels. + * + * We can't just not set SO_BSDCOMAT because some kernels require it. + */ + +static isc_once_t bsdcompat_once = ISC_ONCE_INIT; +bool bsdcompat = true; + +static void +clear_bsdcompat(void) { +#ifdef __linux__ + struct utsname buf; + char *endp; + long int major; + long int minor; + + uname(&buf); /* Can only fail if buf is bad in Linux. */ + + /* Paranoia in parsing can be increased, but we trust uname(). */ + major = strtol(buf.release, &endp, 10); + if (*endp == '.') { + minor = strtol(endp+1, &endp, 10); + if ((major > 2) || ((major == 2) && (minor >= 4))) { + bsdcompat = false; + } + } +#endif /* __linux __ */ +} +#endif + +static void +use_min_mtu(isc__socket_t *sock) { +#if !defined(IPV6_USE_MIN_MTU) && !defined(IPV6_MTU) + UNUSED(sock); +#endif +#ifdef IPV6_USE_MIN_MTU + /* use minimum MTU */ + if (sock->pf == AF_INET6) { + int on = 1; + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_USE_MIN_MTU, + (void *)&on, sizeof(on)); + } +#endif +#if defined(IPV6_MTU) + /* + * Use minimum MTU on IPv6 sockets. + */ + if (sock->pf == AF_INET6) { + int mtu = 1280; + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_MTU, + &mtu, sizeof(mtu)); + } +#endif +} + +static void +set_tcp_maxseg(isc__socket_t *sock, int size) { +#ifdef TCP_MAXSEG + if (sock->type == isc_sockettype_tcp) + (void)setsockopt(sock->fd, IPPROTO_TCP, TCP_MAXSEG, + (void *)&size, sizeof(size)); +#endif +} + +static isc_result_t +opensocket(isc__socketmgr_t *manager, isc__socket_t *sock, + isc__socket_t *dup_socket) +{ + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + const char *err = "socket"; + int tries = 0; +#if defined(USE_CMSG) || defined(SO_BSDCOMPAT) || defined(SO_NOSIGPIPE) + int on = 1; +#endif +#if defined(SO_RCVBUF) + ISC_SOCKADDR_LEN_T optlen; + int size = 0; +#endif + + again: + if (dup_socket == NULL) { + switch (sock->type) { + case isc_sockettype_udp: + sock->fd = socket(sock->pf, SOCK_DGRAM, IPPROTO_UDP); + break; + case isc_sockettype_tcp: + sock->fd = socket(sock->pf, SOCK_STREAM, IPPROTO_TCP); + break; + case isc_sockettype_unix: + sock->fd = socket(sock->pf, SOCK_STREAM, 0); + break; + case isc_sockettype_raw: + errno = EPFNOSUPPORT; + /* + * PF_ROUTE is a alias for PF_NETLINK on linux. + */ +#if defined(PF_ROUTE) + if (sock->fd == -1 && sock->pf == PF_ROUTE) { +#ifdef NETLINK_ROUTE + sock->fd = socket(sock->pf, SOCK_RAW, + NETLINK_ROUTE); +#else + sock->fd = socket(sock->pf, SOCK_RAW, 0); +#endif + if (sock->fd != -1) { +#ifdef NETLINK_ROUTE + struct sockaddr_nl sa; + int n; + + /* + * Do an implicit bind. + */ + memset(&sa, 0, sizeof(sa)); + sa.nl_family = AF_NETLINK; + sa.nl_groups = RTMGRP_IPV4_IFADDR | + RTMGRP_IPV6_IFADDR; + n = bind(sock->fd, + (struct sockaddr *) &sa, + sizeof(sa)); + if (n < 0) { + close(sock->fd); + sock->fd = -1; + } +#endif + sock->bound = 1; + } + } +#endif + break; + case isc_sockettype_fdwatch: + /* + * We should not be called for isc_sockettype_fdwatch + * sockets. + */ + INSIST(0); + break; + } + } else { + sock->fd = dup(dup_socket->fd); + sock->dupped = 1; + sock->bound = dup_socket->bound; + } + if (sock->fd == -1 && errno == EINTR && tries++ < 42) + goto again; + +#ifdef F_DUPFD + /* + * Leave a space for stdio and TCP to work in. + */ + if (manager->reserved != 0 && sock->type == isc_sockettype_udp && + sock->fd >= 0 && sock->fd < manager->reserved) { + int newfd, tmp; + newfd = fcntl(sock->fd, F_DUPFD, manager->reserved); + tmp = errno; + (void)close(sock->fd); + errno = tmp; + sock->fd = newfd; + err = "isc_socket_create: fcntl/reserved"; + } else if (sock->fd >= 0 && sock->fd < 20) { + int newfd, tmp; + newfd = fcntl(sock->fd, F_DUPFD, 20); + tmp = errno; + (void)close(sock->fd); + errno = tmp; + sock->fd = newfd; + err = "isc_socket_create: fcntl"; + } +#endif + + if (sock->fd >= (int)manager->maxsocks) { + (void)close(sock->fd); + isc_log_iwrite(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_TOOMANYFDS, + "socket: file descriptor exceeds limit (%d/%u)", + sock->fd, manager->maxsocks); + inc_stats(manager->stats, sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_NORESOURCES); + } + + if (sock->fd < 0) { + switch (errno) { + case EMFILE: + case ENFILE: + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_iwrite(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_TOOMANYFDS, + "%s: %s", err, strbuf); + /* fallthrough */ + case ENOBUFS: + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_NORESOURCES); + + case EPROTONOSUPPORT: + case EPFNOSUPPORT: + case EAFNOSUPPORT: + /* + * Linux 2.2 (and maybe others) return EINVAL instead of + * EAFNOSUPPORT. + */ + case EINVAL: + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_FAMILYNOSUPPORT); + + default: + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "%s() %s: %s", err, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_UNEXPECTED); + } + } + + if (dup_socket != NULL) + goto setup_done; + + result = make_nonblock(sock->fd); + if (result != ISC_R_SUCCESS) { + (void)close(sock->fd); + inc_stats(manager->stats, sock->statsindex[STATID_OPENFAIL]); + return (result); + } + +#ifdef SO_BSDCOMPAT + RUNTIME_CHECK(isc_once_do(&bsdcompat_once, + clear_bsdcompat) == ISC_R_SUCCESS); + if (sock->type != isc_sockettype_unix && bsdcompat && + setsockopt(sock->fd, SOL_SOCKET, SO_BSDCOMPAT, + (void *)&on, sizeof(on)) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_BSDCOMPAT) %s: %s", + sock->fd, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + /* Press on... */ + } +#endif + +#ifdef SO_NOSIGPIPE + if (setsockopt(sock->fd, SOL_SOCKET, SO_NOSIGPIPE, + (void *)&on, sizeof(on)) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_NOSIGPIPE) %s: %s", + sock->fd, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + /* Press on... */ + } +#endif + + /* + * Use minimum mtu if possible. + */ + if (sock->type == isc_sockettype_tcp && sock->pf == AF_INET6) { + use_min_mtu(sock); + set_tcp_maxseg(sock, 1280 - 20 - 40); /* 1280 - TCP - IPV6 */ + } + +#if defined(USE_CMSG) || defined(SO_RCVBUF) + if (sock->type == isc_sockettype_udp) { + +#if defined(USE_CMSG) +#if defined(SO_TIMESTAMP) + if (setsockopt(sock->fd, SOL_SOCKET, SO_TIMESTAMP, + (void *)&on, sizeof(on)) < 0 + && errno != ENOPROTOOPT) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_TIMESTAMP) %s: %s", + sock->fd, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + /* Press on... */ + } +#endif /* SO_TIMESTAMP */ + +#if defined(ISC_PLATFORM_HAVEIPV6) +#ifdef ISC_PLATFORM_HAVEIN6PKTINFO +#ifdef IPV6_RECVPKTINFO + /* RFC 3542 */ + if ((sock->pf == AF_INET6) + && (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, + (void *)&on, sizeof(on)) < 0)) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_RECVPKTINFO) " + "%s: %s", sock->fd, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } +#else + /* RFC 2292 */ + if ((sock->pf == AF_INET6) + && (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_PKTINFO, + (void *)&on, sizeof(on)) < 0)) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_PKTINFO) %s: %s", + sock->fd, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } +#endif /* IPV6_RECVPKTINFO */ +#endif /* ISC_PLATFORM_HAVEIN6PKTINFO */ +#if defined(IPV6_MTU_DISCOVER) && defined(IPV6_PMTUDISC_DONT) + /* + * Turn off Path MTU discovery on IPv6/UDP sockets. + */ + if (sock->pf == AF_INET6) { + int action = IPV6_PMTUDISC_DONT; + (void)setsockopt(sock->fd, IPPROTO_IPV6, + IPV6_MTU_DISCOVER, &action, + sizeof(action)); + } +#endif +#endif /* ISC_PLATFORM_HAVEIPV6 */ +#endif /* defined(USE_CMSG) */ + +#if defined(IP_MTU_DISCOVER) && defined(IP_PMTUDISC_DONT) + /* + * Turn off Path MTU discovery on IPv4/UDP sockets. + * Prefer IP_PMTUDISC_OMIT over IP_PMTUDISC_DONT + * if it available. + */ + if (sock->pf == AF_INET) { + int action; +#if defined(IP_PMTUDISC_OMIT) + action = IP_PMTUDISC_OMIT; + if (setsockopt(sock->fd, IPPROTO_IP, + IP_MTU_DISCOVER, &action, + sizeof(action)) < 0) { +#endif + action = IP_PMTUDISC_DONT; + (void)setsockopt(sock->fd, IPPROTO_IP, + IP_MTU_DISCOVER, + &action, sizeof(action)); +#if defined(IP_PMTUDISC_OMIT) + } +#endif + } +#endif +#if defined(IP_DONTFRAG) + /* + * Turn off Path MTU discovery on IPv4/UDP sockets. + */ + if (sock->pf == AF_INET) { + int off = 0; + (void)setsockopt(sock->fd, IPPROTO_IP, IP_DONTFRAG, + &off, sizeof(off)); + } +#endif + +#if defined(SO_RCVBUF) + optlen = sizeof(size); + if (getsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF, + (void *)&size, &optlen) == 0 && size < rcvbuf) { + RUNTIME_CHECK(isc_once_do(&rcvbuf_once, + set_rcvbuf) == ISC_R_SUCCESS); + if (setsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF, + (void *)&rcvbuf, sizeof(rcvbuf)) == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_RCVBUF, %d) %s: %s", + sock->fd, rcvbuf, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } + } +#endif + } +#ifdef IPV6_RECVTCLASS + if ((sock->pf == AF_INET6) + && (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVTCLASS, + (void *)&on, sizeof(on)) < 0)) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_RECVTCLASS) " + "%s: %s", sock->fd, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + } +#endif +#ifdef IP_RECVTOS + if ((sock->pf == AF_INET) + && (setsockopt(sock->fd, IPPROTO_IP, IP_RECVTOS, + (void *)&on, sizeof(on)) < 0)) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_RECVTOS) " + "%s: %s", sock->fd, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + } +#endif +#endif /* defined(USE_CMSG) || defined(SO_RCVBUF) */ + +setup_done: + inc_stats(manager->stats, sock->statsindex[STATID_OPEN]); + if (sock->active == 0) { + inc_stats(manager->stats, sock->statsindex[STATID_ACTIVE]); + sock->active = 1; + } + + return (ISC_R_SUCCESS); +} + +/* + * Create a 'type' socket or duplicate an existing socket, managed + * by 'manager'. Events will be posted to 'task' and when dispatched + * 'action' will be called with 'arg' as the arg value. The new + * socket is returned in 'socketp'. + */ +static isc_result_t +socket_create(isc_socketmgr_t *manager0, int pf, isc_sockettype_t type, + isc_socket_t **socketp, isc_socket_t *dup_socket) +{ + isc__socket_t *sock = NULL; + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + isc_result_t result; + int lockid; + + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(socketp != NULL && *socketp == NULL); + REQUIRE(type != isc_sockettype_fdwatch); + + result = allocate_socket(manager, type, &sock); + if (result != ISC_R_SUCCESS) + return (result); + + switch (sock->type) { + case isc_sockettype_udp: + sock->statsindex = + (pf == AF_INET) ? udp4statsindex : udp6statsindex; +#define DCSPPKT(pf) ((pf == AF_INET) ? ISC_NET_DSCPPKTV4 : ISC_NET_DSCPPKTV6) + sock->pktdscp = (isc_net_probedscp() & DCSPPKT(pf)) != 0; + break; + case isc_sockettype_tcp: + sock->statsindex = + (pf == AF_INET) ? tcp4statsindex : tcp6statsindex; + break; + case isc_sockettype_unix: + sock->statsindex = unixstatsindex; + break; + case isc_sockettype_raw: + sock->statsindex = rawstatsindex; + break; + default: + INSIST(0); + } + + sock->pf = pf; + + result = opensocket(manager, sock, (isc__socket_t *)dup_socket); + if (result != ISC_R_SUCCESS) { + free_socket(&sock); + return (result); + } + + sock->common.methods = (isc_socketmethods_t *)&socketmethods; + sock->references = 1; + *socketp = (isc_socket_t *)sock; + + /* + * Note we don't have to lock the socket like we normally would because + * there are no external references to it yet. + */ + + lockid = FDLOCK_ID(sock->fd); + LOCK(&manager->fdlock[lockid]); + manager->fds[sock->fd] = sock; + manager->fdstate[sock->fd] = MANAGED; +#if defined(USE_EPOLL) + manager->epoll_events[sock->fd] = 0; +#endif +#ifdef USE_DEVPOLL + INSIST(sock->manager->fdpollinfo[sock->fd].want_read == 0 && + sock->manager->fdpollinfo[sock->fd].want_write == 0); +#endif + UNLOCK(&manager->fdlock[lockid]); + + LOCK(&manager->lock); + ISC_LIST_APPEND(manager->socklist, sock, link); +#ifdef USE_SELECT + if (manager->maxfd < sock->fd) + manager->maxfd = sock->fd; +#endif + UNLOCK(&manager->lock); + + socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_CREATED, dup_socket != NULL ? "dupped" : "created"); + + return (ISC_R_SUCCESS); +} + +/*% + * Create a new 'type' socket managed by 'manager'. Events + * will be posted to 'task' and when dispatched 'action' will be + * called with 'arg' as the arg value. The new socket is returned + * in 'socketp'. + */ +isc_result_t +isc__socket_create(isc_socketmgr_t *manager0, int pf, isc_sockettype_t type, + isc_socket_t **socketp) +{ + return (socket_create(manager0, pf, type, socketp, NULL)); +} + +/*% + * Duplicate an existing socket. The new socket is returned + * in 'socketp'. + */ +isc_result_t +isc__socket_dup(isc_socket_t *sock0, isc_socket_t **socketp) { + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(socketp != NULL && *socketp == NULL); + + return (socket_create((isc_socketmgr_t *) sock->manager, + sock->pf, sock->type, socketp, + sock0)); +} + +isc_result_t +isc__socket_open(isc_socket_t *sock0) { + isc_result_t result; + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + REQUIRE(sock->references == 1); + REQUIRE(sock->type != isc_sockettype_fdwatch); + UNLOCK(&sock->lock); + /* + * We don't need to retain the lock hereafter, since no one else has + * this socket. + */ + REQUIRE(sock->fd == -1); + + result = opensocket(sock->manager, sock, NULL); + if (result != ISC_R_SUCCESS) + sock->fd = -1; + + if (result == ISC_R_SUCCESS) { + int lockid = FDLOCK_ID(sock->fd); + + LOCK(&sock->manager->fdlock[lockid]); + sock->manager->fds[sock->fd] = sock; + sock->manager->fdstate[sock->fd] = MANAGED; +#if defined(USE_EPOLL) + sock->manager->epoll_events[sock->fd] = 0; +#endif +#ifdef USE_DEVPOLL + INSIST(sock->manager->fdpollinfo[sock->fd].want_read == 0 && + sock->manager->fdpollinfo[sock->fd].want_write == 0); +#endif + UNLOCK(&sock->manager->fdlock[lockid]); + +#ifdef USE_SELECT + LOCK(&sock->manager->lock); + if (sock->manager->maxfd < sock->fd) + sock->manager->maxfd = sock->fd; + UNLOCK(&sock->manager->lock); +#endif + } + + return (result); +} + +/* + * Create a new 'type' socket managed by 'manager'. Events + * will be posted to 'task' and when dispatched 'action' will be + * called with 'arg' as the arg value. The new socket is returned + * in 'socketp'. + */ +isc_result_t +isc__socket_fdwatchcreate(isc_socketmgr_t *manager0, int fd, int flags, + isc_sockfdwatch_t callback, void *cbarg, + isc_task_t *task, isc_socket_t **socketp) +{ + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + isc__socket_t *sock = NULL; + isc_result_t result; + int lockid; + + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(socketp != NULL && *socketp == NULL); + + if (fd < 0 || (unsigned int)fd >= manager->maxsocks) + return (ISC_R_RANGE); + + result = allocate_socket(manager, isc_sockettype_fdwatch, &sock); + if (result != ISC_R_SUCCESS) + return (result); + + sock->fd = fd; + sock->fdwatcharg = cbarg; + sock->fdwatchcb = callback; + sock->fdwatchflags = flags; + sock->fdwatchtask = task; + sock->statsindex = fdwatchstatsindex; + + sock->common.methods = (isc_socketmethods_t *)&socketmethods; + sock->references = 1; + *socketp = (isc_socket_t *)sock; + + /* + * Note we don't have to lock the socket like we normally would because + * there are no external references to it yet. + */ + + lockid = FDLOCK_ID(sock->fd); + LOCK(&manager->fdlock[lockid]); + manager->fds[sock->fd] = sock; + manager->fdstate[sock->fd] = MANAGED; +#if defined(USE_EPOLL) + manager->epoll_events[sock->fd] = 0; +#endif + UNLOCK(&manager->fdlock[lockid]); + + LOCK(&manager->lock); + ISC_LIST_APPEND(manager->socklist, sock, link); +#ifdef USE_SELECT + if (manager->maxfd < sock->fd) + manager->maxfd = sock->fd; +#endif + UNLOCK(&manager->lock); + + if (flags & ISC_SOCKFDWATCH_READ) + select_poke(sock->manager, sock->fd, SELECT_POKE_READ); + if (flags & ISC_SOCKFDWATCH_WRITE) + select_poke(sock->manager, sock->fd, SELECT_POKE_WRITE); + + socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_CREATED, "fdwatch-created"); + + return (ISC_R_SUCCESS); +} + +/* + * Indicate to the manager that it should watch the socket again. + * This can be used to restart watching if the previous event handler + * didn't indicate there was more data to be processed. Primarily + * it is for writing but could be used for reading if desired + */ + +isc_result_t +isc__socket_fdwatchpoke(isc_socket_t *sock0, int flags) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + + /* + * We check both flags first to allow us to get the lock + * once but only if we need it. + */ + + if ((flags & (ISC_SOCKFDWATCH_READ | ISC_SOCKFDWATCH_WRITE)) != 0) { + LOCK(&sock->lock); + if (((flags & ISC_SOCKFDWATCH_READ) != 0) && + !sock->pending_recv) + select_poke(sock->manager, sock->fd, + SELECT_POKE_READ); + if (((flags & ISC_SOCKFDWATCH_WRITE) != 0) && + !sock->pending_send) + select_poke(sock->manager, sock->fd, + SELECT_POKE_WRITE); + UNLOCK(&sock->lock); + } + + socket_log(sock, NULL, TRACE, isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_POKED, "fdwatch-poked flags: %d", flags); + + return (ISC_R_SUCCESS); +} + +/* + * Attach to a socket. Caller must explicitly detach when it is done. + */ +void +isc__socket_attach(isc_socket_t *sock0, isc_socket_t **socketp) { + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(socketp != NULL && *socketp == NULL); + + LOCK(&sock->lock); + sock->references++; + UNLOCK(&sock->lock); + + *socketp = (isc_socket_t *)sock; +} + +/* + * Dereference a socket. If this is the last reference to it, clean things + * up by destroying the socket. + */ +void +isc__socket_detach(isc_socket_t **socketp) { + isc__socket_t *sock; + bool kill_socket = false; + + REQUIRE(socketp != NULL); + sock = (isc__socket_t *)*socketp; + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + REQUIRE(sock->references > 0); + sock->references--; + if (sock->references == 0) + kill_socket = true; + UNLOCK(&sock->lock); + + if (kill_socket) + destroy(&sock); + + *socketp = NULL; +} + +isc_result_t +isc__socket_close(isc_socket_t *sock0) { + isc__socket_t *sock = (isc__socket_t *)sock0; + int fd; + isc__socketmgr_t *manager; + + fflush(stdout); + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + REQUIRE(sock->references == 1); + REQUIRE(sock->type != isc_sockettype_fdwatch); + REQUIRE(sock->fd >= 0 && sock->fd < (int)sock->manager->maxsocks); + + INSIST(!sock->connecting); + INSIST(!sock->pending_recv); + INSIST(!sock->pending_send); + INSIST(!sock->pending_accept); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + + manager = sock->manager; + fd = sock->fd; + sock->fd = -1; + sock->dupped = 0; + memset(sock->name, 0, sizeof(sock->name)); + sock->tag = NULL; + sock->listener = 0; + sock->connected = 0; + sock->connecting = 0; + sock->bound = 0; + isc_sockaddr_any(&sock->peer_address); + + UNLOCK(&sock->lock); + + socketclose(manager, sock, fd); + + return (ISC_R_SUCCESS); +} + +/* + * I/O is possible on a given socket. Schedule an event to this task that + * will call an internal function to do the I/O. This will charge the + * task with the I/O operation and let our select loop handler get back + * to doing something real as fast as possible. + * + * The socket and manager must be locked before calling this function. + */ +static void +dispatch_recv(isc__socket_t *sock) { + intev_t *iev; + isc_socketevent_t *ev; + isc_task_t *sender; + + INSIST(!sock->pending_recv); + + if (sock->type != isc_sockettype_fdwatch) { + ev = ISC_LIST_HEAD(sock->recv_list); + if (ev == NULL) + return; + socket_log(sock, NULL, EVENT, NULL, 0, 0, + "dispatch_recv: event %p -> task %p", + ev, ev->ev_sender); + sender = ev->ev_sender; + } else { + sender = sock->fdwatchtask; + } + + sock->pending_recv = 1; + iev = &sock->readable_ev; + + sock->references++; + iev->ev_sender = sock; + if (sock->type == isc_sockettype_fdwatch) + iev->ev_action = internal_fdwatch_read; + else + iev->ev_action = internal_recv; + iev->ev_arg = sock; + + isc_task_send(sender, (isc_event_t **)&iev); +} + +static void +dispatch_send(isc__socket_t *sock) { + intev_t *iev; + isc_socketevent_t *ev; + isc_task_t *sender; + + INSIST(!sock->pending_send); + + if (sock->type != isc_sockettype_fdwatch) { + ev = ISC_LIST_HEAD(sock->send_list); + if (ev == NULL) + return; + socket_log(sock, NULL, EVENT, NULL, 0, 0, + "dispatch_send: event %p -> task %p", + ev, ev->ev_sender); + sender = ev->ev_sender; + } else { + sender = sock->fdwatchtask; + } + + sock->pending_send = 1; + iev = &sock->writable_ev; + + sock->references++; + iev->ev_sender = sock; + if (sock->type == isc_sockettype_fdwatch) + iev->ev_action = internal_fdwatch_write; + else + iev->ev_action = internal_send; + iev->ev_arg = sock; + + isc_task_send(sender, (isc_event_t **)&iev); +} + +/* + * Dispatch an internal accept event. + */ +static void +dispatch_accept(isc__socket_t *sock) { + intev_t *iev; + isc_socket_newconnev_t *ev; + + INSIST(!sock->pending_accept); + + /* + * Are there any done events left, or were they all canceled + * before the manager got the socket lock? + */ + ev = ISC_LIST_HEAD(sock->accept_list); + if (ev == NULL) + return; + + sock->pending_accept = 1; + iev = &sock->readable_ev; + + sock->references++; /* keep socket around for this internal event */ + iev->ev_sender = sock; + iev->ev_action = internal_accept; + iev->ev_arg = sock; + + isc_task_send(ev->ev_sender, (isc_event_t **)&iev); +} + +static void +dispatch_connect(isc__socket_t *sock) { + intev_t *iev; + isc_socket_connev_t *ev; + + iev = &sock->writable_ev; + + ev = ISC_LIST_HEAD(sock->connect_list); + INSIST(ev != NULL); /* XXX */ + + INSIST(sock->connecting); + + sock->references++; /* keep socket around for this internal event */ + iev->ev_sender = sock; + iev->ev_action = internal_connect; + iev->ev_arg = sock; + + isc_task_send(ev->ev_sender, (isc_event_t **)&iev); +} + +/* + * Dequeue an item off the given socket's read queue, set the result code + * in the done event to the one provided, and send it to the task it was + * destined for. + * + * If the event to be sent is on a list, remove it before sending. If + * asked to, send and detach from the socket as well. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_recvdone_event(isc__socket_t *sock, isc_socketevent_t **dev) { + isc_task_t *task; + + task = (*dev)->ev_sender; + + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) + ISC_LIST_DEQUEUE(sock->recv_list, *dev, ev_link); + + if (((*dev)->attributes & ISC_SOCKEVENTATTR_ATTACHED) + == ISC_SOCKEVENTATTR_ATTACHED) + isc_task_sendanddetach(&task, (isc_event_t **)dev); + else + isc_task_send(task, (isc_event_t **)dev); +} + +/* + * See comments for send_recvdone_event() above. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_senddone_event(isc__socket_t *sock, isc_socketevent_t **dev) { + isc_task_t *task; + + INSIST(dev != NULL && *dev != NULL); + + task = (*dev)->ev_sender; + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) + ISC_LIST_DEQUEUE(sock->send_list, *dev, ev_link); + + if (((*dev)->attributes & ISC_SOCKEVENTATTR_ATTACHED) + == ISC_SOCKEVENTATTR_ATTACHED) + isc_task_sendanddetach(&task, (isc_event_t **)dev); + else + isc_task_send(task, (isc_event_t **)dev); +} + +/* + * See comments for send_recvdone_event() above. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_connectdone_event(isc__socket_t *sock, isc_socket_connev_t **dev) { + isc_task_t *task; + + INSIST(dev != NULL && *dev != NULL); + + task = (*dev)->ev_sender; + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) + ISC_LIST_DEQUEUE(sock->connect_list, *dev, ev_link); + + isc_task_sendanddetach(&task, (isc_event_t **)dev); +} + +/* + * Call accept() on a socket, to get the new file descriptor. The listen + * socket is used as a prototype to create a new isc_socket_t. The new + * socket has one outstanding reference. The task receiving the event + * will be detached from just after the event is delivered. + * + * On entry to this function, the event delivered is the internal + * readable event, and the first item on the accept_list should be + * the done event we want to send. If the list is empty, this is a no-op, + * so just unlock and return. + */ +static void +internal_accept(isc_task_t *me, isc_event_t *ev) { + isc__socket_t *sock; + isc__socketmgr_t *manager; + isc_socket_newconnev_t *dev; + isc_task_t *task; + ISC_SOCKADDR_LEN_T addrlen; + int fd; + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + const char *err = "accept"; + + UNUSED(me); + + sock = ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + socket_log(sock, NULL, TRACE, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_ACCEPTLOCK, + "internal_accept called, locked socket"); + + manager = sock->manager; + INSIST(VALID_MANAGER(manager)); + + INSIST(sock->listener); + INSIST(sock->pending_accept == 1); + sock->pending_accept = 0; + + INSIST(sock->references > 0); + sock->references--; /* the internal event is done with this socket */ + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + /* + * Get the first item off the accept list. + * If it is empty, unlock the socket and return. + */ + dev = ISC_LIST_HEAD(sock->accept_list); + if (dev == NULL) { + UNLOCK(&sock->lock); + return; + } + + /* + * Try to accept the new connection. If the accept fails with + * EAGAIN or EINTR, simply poke the watcher to watch this socket + * again. Also ignore ECONNRESET, which has been reported to + * be spuriously returned on Linux 2.2.19 although it is not + * a documented error for accept(). ECONNABORTED has been + * reported for Solaris 8. The rest are thrown in not because + * we have seen them but because they are ignored by other + * daemons such as BIND 8 and Apache. + */ + + addrlen = sizeof(NEWCONNSOCK(dev)->peer_address.type); + memset(&NEWCONNSOCK(dev)->peer_address.type, 0, addrlen); + fd = accept(sock->fd, &NEWCONNSOCK(dev)->peer_address.type.sa, + (void *)&addrlen); + +#ifdef F_DUPFD + /* + * Leave a space for stdio to work in. + */ + if (fd >= 0 && fd < 20) { + int newfd, tmp; + newfd = fcntl(fd, F_DUPFD, 20); + tmp = errno; + (void)close(fd); + errno = tmp; + fd = newfd; + err = "accept/fcntl"; + } +#endif + + if (fd < 0) { + if (SOFT_ERROR(errno)) + goto soft_error; + switch (errno) { + case ENFILE: + case EMFILE: + isc_log_iwrite(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_TOOMANYFDS, + "%s: too many open file descriptors", + err); + goto soft_error; + + case ENOBUFS: + case ENOMEM: + case ECONNRESET: + case ECONNABORTED: + case EHOSTUNREACH: + case EHOSTDOWN: + case ENETUNREACH: + case ENETDOWN: + case ECONNREFUSED: +#ifdef EPROTO + case EPROTO: +#endif +#ifdef ENONET + case ENONET: +#endif + goto soft_error; + default: + break; + } + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_accept: %s() %s: %s", err, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + fd = -1; + result = ISC_R_UNEXPECTED; + } else { + if (addrlen == 0U) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_accept(): " + "accept() failed to return " + "remote address"); + + (void)close(fd); + goto soft_error; + } else if (NEWCONNSOCK(dev)->peer_address.type.sa.sa_family != + sock->pf) + { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_accept(): " + "accept() returned peer address " + "family %u (expected %u)", + NEWCONNSOCK(dev)->peer_address. + type.sa.sa_family, + sock->pf); + (void)close(fd); + goto soft_error; + } else if (fd >= (int)manager->maxsocks) { + isc_log_iwrite(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_TOOMANYFDS, + "accept: " + "file descriptor exceeds limit (%d/%u)", + fd, manager->maxsocks); + (void)close(fd); + goto soft_error; + } + } + + if (fd != -1) { + NEWCONNSOCK(dev)->peer_address.length = addrlen; + NEWCONNSOCK(dev)->pf = sock->pf; + } + + /* + * Pull off the done event. + */ + ISC_LIST_UNLINK(sock->accept_list, dev, ev_link); + + /* + * Poke watcher if there are more pending accepts. + */ + if (!ISC_LIST_EMPTY(sock->accept_list)) + select_poke(sock->manager, sock->fd, SELECT_POKE_ACCEPT); + + UNLOCK(&sock->lock); + + if (fd != -1) { + result = make_nonblock(fd); + if (result != ISC_R_SUCCESS) { + (void)close(fd); + fd = -1; + } + } + + /* + * -1 means the new socket didn't happen. + */ + if (fd != -1) { + int lockid = FDLOCK_ID(fd); + + NEWCONNSOCK(dev)->fd = fd; + NEWCONNSOCK(dev)->bound = 1; + NEWCONNSOCK(dev)->connected = 1; + + /* + * Use minimum mtu if possible. + */ + use_min_mtu(NEWCONNSOCK(dev)); + set_tcp_maxseg(NEWCONNSOCK(dev), 1280 - 20 - 40); + + /* + * Ensure DSCP settings are inherited across accept. + */ + setdscp(NEWCONNSOCK(dev), sock->dscp); + + /* + * Save away the remote address + */ + dev->address = NEWCONNSOCK(dev)->peer_address; + + if (NEWCONNSOCK(dev)->active == 0) { + inc_stats(manager->stats, + NEWCONNSOCK(dev)->statsindex[STATID_ACTIVE]); + NEWCONNSOCK(dev)->active = 1; + } + + LOCK(&manager->fdlock[lockid]); + manager->fds[fd] = NEWCONNSOCK(dev); + manager->fdstate[fd] = MANAGED; +#if defined(USE_EPOLL) + manager->epoll_events[fd] = 0; +#endif + UNLOCK(&manager->fdlock[lockid]); + + LOCK(&manager->lock); + +#ifdef USE_SELECT + if (manager->maxfd < fd) + manager->maxfd = fd; +#endif + + socket_log(sock, &NEWCONNSOCK(dev)->peer_address, CREATION, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_ACCEPTEDCXN, + "accepted connection, new socket %p", + dev->newsocket); + + ISC_LIST_APPEND(manager->socklist, NEWCONNSOCK(dev), link); + + UNLOCK(&manager->lock); + + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPT]); + } else { + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPTFAIL]); + NEWCONNSOCK(dev)->references--; + free_socket((isc__socket_t **)&dev->newsocket); + } + + /* + * Fill in the done event details and send it off. + */ + dev->result = result; + task = dev->ev_sender; + dev->ev_sender = sock; + + isc_task_sendanddetach(&task, ISC_EVENT_PTR(&dev)); + return; + + soft_error: + select_poke(sock->manager, sock->fd, SELECT_POKE_ACCEPT); + UNLOCK(&sock->lock); + + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPTFAIL]); + return; +} + +static void +internal_recv(isc_task_t *me, isc_event_t *ev) { + isc_socketevent_t *dev; + isc__socket_t *sock; + + INSIST(ev->ev_type == ISC_SOCKEVENT_INTR); + + sock = ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + socket_log(sock, NULL, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_INTERNALRECV, + "internal_recv: task %p got event %p", me, ev); + + INSIST(sock->pending_recv == 1); + sock->pending_recv = 0; + + INSIST(sock->references > 0); + sock->references--; /* the internal event is done with this socket */ + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + /* + * Try to do as much I/O as possible on this socket. There are no + * limits here, currently. + */ + dev = ISC_LIST_HEAD(sock->recv_list); + while (dev != NULL) { + switch (doio_recv(sock, dev)) { + case DOIO_SOFT: + goto poke; + + case DOIO_EOF: + /* + * read of 0 means the remote end was closed. + * Run through the event queue and dispatch all + * the events with an EOF result code. + */ + do { + dev->result = ISC_R_EOF; + send_recvdone_event(sock, &dev); + dev = ISC_LIST_HEAD(sock->recv_list); + } while (dev != NULL); + goto poke; + + case DOIO_SUCCESS: + case DOIO_HARD: + send_recvdone_event(sock, &dev); + break; + } + + dev = ISC_LIST_HEAD(sock->recv_list); + } + + poke: + if (!ISC_LIST_EMPTY(sock->recv_list)) + select_poke(sock->manager, sock->fd, SELECT_POKE_READ); + + UNLOCK(&sock->lock); +} + +static void +internal_send(isc_task_t *me, isc_event_t *ev) { + isc_socketevent_t *dev; + isc__socket_t *sock; + + INSIST(ev->ev_type == ISC_SOCKEVENT_INTW); + + /* + * Find out what socket this is and lock it. + */ + sock = (isc__socket_t *)ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + socket_log(sock, NULL, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_INTERNALSEND, + "internal_send: task %p got event %p", me, ev); + + INSIST(sock->pending_send == 1); + sock->pending_send = 0; + + INSIST(sock->references > 0); + sock->references--; /* the internal event is done with this socket */ + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + /* + * Try to do as much I/O as possible on this socket. There are no + * limits here, currently. + */ + dev = ISC_LIST_HEAD(sock->send_list); + while (dev != NULL) { + switch (doio_send(sock, dev)) { + case DOIO_SOFT: + goto poke; + + case DOIO_HARD: + case DOIO_SUCCESS: + send_senddone_event(sock, &dev); + break; + } + + dev = ISC_LIST_HEAD(sock->send_list); + } + + poke: + if (!ISC_LIST_EMPTY(sock->send_list)) + select_poke(sock->manager, sock->fd, SELECT_POKE_WRITE); + + UNLOCK(&sock->lock); +} + +static void +internal_fdwatch_write(isc_task_t *me, isc_event_t *ev) { + isc__socket_t *sock; + int more_data; + + INSIST(ev->ev_type == ISC_SOCKEVENT_INTW); + + /* + * Find out what socket this is and lock it. + */ + sock = (isc__socket_t *)ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + socket_log(sock, NULL, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_INTERNALSEND, + "internal_fdwatch_write: task %p got event %p", me, ev); + + INSIST(sock->pending_send == 1); + + UNLOCK(&sock->lock); + more_data = (sock->fdwatchcb)(me, (isc_socket_t *)sock, + sock->fdwatcharg, ISC_SOCKFDWATCH_WRITE); + LOCK(&sock->lock); + + sock->pending_send = 0; + + INSIST(sock->references > 0); + sock->references--; /* the internal event is done with this socket */ + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + if (more_data) + select_poke(sock->manager, sock->fd, SELECT_POKE_WRITE); + + UNLOCK(&sock->lock); +} + +static void +internal_fdwatch_read(isc_task_t *me, isc_event_t *ev) { + isc__socket_t *sock; + int more_data; + + INSIST(ev->ev_type == ISC_SOCKEVENT_INTR); + + /* + * Find out what socket this is and lock it. + */ + sock = (isc__socket_t *)ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + socket_log(sock, NULL, IOEVENT, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_INTERNALRECV, + "internal_fdwatch_read: task %p got event %p", me, ev); + + INSIST(sock->pending_recv == 1); + + UNLOCK(&sock->lock); + more_data = (sock->fdwatchcb)(me, (isc_socket_t *)sock, + sock->fdwatcharg, ISC_SOCKFDWATCH_READ); + LOCK(&sock->lock); + + sock->pending_recv = 0; + + INSIST(sock->references > 0); + sock->references--; /* the internal event is done with this socket */ + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + if (more_data) + select_poke(sock->manager, sock->fd, SELECT_POKE_READ); + + UNLOCK(&sock->lock); +} + +/* + * Process read/writes on each fd here. Avoid locking + * and unlocking twice if both reads and writes are possible. + */ +static void +process_fd(isc__socketmgr_t *manager, int fd, bool readable, + bool writeable) +{ + isc__socket_t *sock; + bool unlock_sock; + bool unwatch_read = false, unwatch_write = false; + int lockid = FDLOCK_ID(fd); + + /* + * If the socket is going to be closed, don't do more I/O. + */ + LOCK(&manager->fdlock[lockid]); + if (manager->fdstate[fd] == CLOSE_PENDING) { + UNLOCK(&manager->fdlock[lockid]); + + (void)unwatch_fd(manager, fd, SELECT_POKE_READ); + (void)unwatch_fd(manager, fd, SELECT_POKE_WRITE); + return; + } + + sock = manager->fds[fd]; + unlock_sock = false; + if (readable) { + if (sock == NULL) { + unwatch_read = true; + goto check_write; + } + unlock_sock = true; + LOCK(&sock->lock); + if (!SOCK_DEAD(sock)) { + if (sock->listener) + dispatch_accept(sock); + else + dispatch_recv(sock); + } + unwatch_read = true; + } +check_write: + if (writeable) { + if (sock == NULL) { + unwatch_write = true; + goto unlock_fd; + } + if (!unlock_sock) { + unlock_sock = true; + LOCK(&sock->lock); + } + if (!SOCK_DEAD(sock)) { + if (sock->connecting) + dispatch_connect(sock); + else + dispatch_send(sock); + } + unwatch_write = true; + } + if (unlock_sock) + UNLOCK(&sock->lock); + + unlock_fd: + UNLOCK(&manager->fdlock[lockid]); + if (unwatch_read) + (void)unwatch_fd(manager, fd, SELECT_POKE_READ); + if (unwatch_write) + (void)unwatch_fd(manager, fd, SELECT_POKE_WRITE); + +} + +#ifdef USE_KQUEUE +static bool +process_fds(isc__socketmgr_t *manager, struct kevent *events, int nevents) { + int i; + bool readable, writable; + bool done = false; +#ifdef USE_WATCHER_THREAD + bool have_ctlevent = false; +#endif + + if (nevents == manager->nevents) { + /* + * This is not an error, but something unexpected. If this + * happens, it may indicate the need for increasing + * ISC_SOCKET_MAXEVENTS. + */ + manager_log(manager, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].ident < manager->maxsocks); +#ifdef USE_WATCHER_THREAD + if (events[i].ident == (uintptr_t)manager->pipe_fds[0]) { + have_ctlevent = true; + continue; + } +#endif + readable = (events[i].filter == EVFILT_READ); + writable = (events[i].filter == EVFILT_WRITE); + process_fd(manager, events[i].ident, readable, writable); + } + +#ifdef USE_WATCHER_THREAD + if (have_ctlevent) + done = process_ctlfd(manager); +#endif + + return (done); +} +#elif defined(USE_EPOLL) +static bool +process_fds(isc__socketmgr_t *manager, struct epoll_event *events, int nevents) +{ + int i; + bool done = false; +#ifdef USE_WATCHER_THREAD + bool have_ctlevent = false; +#endif + + if (nevents == manager->nevents) { + manager_log(manager, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].data.fd < (int)manager->maxsocks); +#ifdef USE_WATCHER_THREAD + if (events[i].data.fd == manager->pipe_fds[0]) { + have_ctlevent = true; + continue; + } +#endif + if ((events[i].events & EPOLLERR) != 0 || + (events[i].events & EPOLLHUP) != 0) { + /* + * epoll does not set IN/OUT bits on an erroneous + * condition, so we need to try both anyway. This is a + * bit inefficient, but should be okay for such rare + * events. Note also that the read or write attempt + * won't block because we use non-blocking sockets. + */ + int fd = events[i].data.fd; + events[i].events |= manager->epoll_events[fd]; + } + process_fd(manager, events[i].data.fd, + (events[i].events & EPOLLIN) != 0, + (events[i].events & EPOLLOUT) != 0); + } + +#ifdef USE_WATCHER_THREAD + if (have_ctlevent) + done = process_ctlfd(manager); +#endif + + return (done); +} +#elif defined(USE_DEVPOLL) +static bool +process_fds(isc__socketmgr_t *manager, struct pollfd *events, int nevents) { + int i; + bool done = false; +#ifdef USE_WATCHER_THREAD + bool have_ctlevent = false; +#endif + + if (nevents == manager->nevents) { + manager_log(manager, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].fd < (int)manager->maxsocks); +#ifdef USE_WATCHER_THREAD + if (events[i].fd == manager->pipe_fds[0]) { + have_ctlevent = true; + continue; + } +#endif + process_fd(manager, events[i].fd, + (events[i].events & POLLIN) != 0, + (events[i].events & POLLOUT) != 0); + } + +#ifdef USE_WATCHER_THREAD + if (have_ctlevent) + done = process_ctlfd(manager); +#endif + + return (done); +} +#elif defined(USE_SELECT) +static void +process_fds(isc__socketmgr_t *manager, int maxfd, fd_set *readfds, + fd_set *writefds) +{ + int i; + + REQUIRE(maxfd <= (int)manager->maxsocks); + + for (i = 0; i < maxfd; i++) { +#ifdef USE_WATCHER_THREAD + if (i == manager->pipe_fds[0] || i == manager->pipe_fds[1]) + continue; +#endif /* USE_WATCHER_THREAD */ + process_fd(manager, i, FD_ISSET(i, readfds), + FD_ISSET(i, writefds)); + } +} +#endif + +#ifdef USE_WATCHER_THREAD +static bool +process_ctlfd(isc__socketmgr_t *manager) { + int msg, fd; + + for (;;) { + select_readmsg(manager, &fd, &msg); + + manager_log(manager, IOEVENT, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_WATCHERMSG, + "watcher got message %d " + "for socket %d"), msg, fd); + + /* + * Nothing to read? + */ + if (msg == SELECT_POKE_NOTHING) + break; + + /* + * Handle shutdown message. We really should + * jump out of this loop right away, but + * it doesn't matter if we have to do a little + * more work first. + */ + if (msg == SELECT_POKE_SHUTDOWN) + return (true); + + /* + * This is a wakeup on a socket. Look + * at the event queue for both read and write, + * and decide if we need to watch on it now + * or not. + */ + wakeup_socket(manager, fd, msg); + } + + return (false); +} + +/* + * This is the thread that will loop forever, always in a select or poll + * call. + * + * When select returns something to do, track down what thread gets to do + * this I/O and post the event to it. + */ +static isc_threadresult_t +watcher(void *uap) { + isc__socketmgr_t *manager = uap; + bool done; + int cc; +#ifdef USE_KQUEUE + const char *fnname = "kevent()"; +#elif defined (USE_EPOLL) + const char *fnname = "epoll_wait()"; +#elif defined(USE_DEVPOLL) + isc_result_t result; + const char *fnname = "ioctl(DP_POLL)"; + struct dvpoll dvp; + int pass; +#if defined(ISC_SOCKET_USE_POLLWATCH) + pollstate_t pollstate = poll_idle; +#endif +#elif defined (USE_SELECT) + const char *fnname = "select()"; + int maxfd; + int ctlfd; +#endif + char strbuf[ISC_STRERRORSIZE]; + +#if defined (USE_SELECT) + /* + * Get the control fd here. This will never change. + */ + ctlfd = manager->pipe_fds[0]; +#endif + done = false; + while (!done) { + do { +#ifdef USE_KQUEUE + cc = kevent(manager->kqueue_fd, NULL, 0, + manager->events, manager->nevents, NULL); +#elif defined(USE_EPOLL) + cc = epoll_wait(manager->epoll_fd, manager->events, + manager->nevents, -1); +#elif defined(USE_DEVPOLL) + /* + * Re-probe every thousand calls. + */ + if (manager->calls++ > 1000U) { + result = isc_resource_getcurlimit( + isc_resource_openfiles, + &manager->open_max); + if (result != ISC_R_SUCCESS) + manager->open_max = 64; + manager->calls = 0; + } + for (pass = 0; pass < 2; pass++) { + dvp.dp_fds = manager->events; + dvp.dp_nfds = manager->nevents; + if (dvp.dp_nfds >= manager->open_max) + dvp.dp_nfds = manager->open_max - 1; +#ifndef ISC_SOCKET_USE_POLLWATCH + dvp.dp_timeout = -1; +#else + if (pollstate == poll_idle) + dvp.dp_timeout = -1; + else + dvp.dp_timeout = + ISC_SOCKET_POLLWATCH_TIMEOUT; +#endif /* ISC_SOCKET_USE_POLLWATCH */ + cc = ioctl(manager->devpoll_fd, DP_POLL, &dvp); + if (cc == -1 && errno == EINVAL) { + /* + * {OPEN_MAX} may have dropped. Look + * up the current value and try again. + */ + result = isc_resource_getcurlimit( + isc_resource_openfiles, + &manager->open_max); + if (result != ISC_R_SUCCESS) + manager->open_max = 64; + } else + break; + } +#elif defined(USE_SELECT) + LOCK(&manager->lock); + memmove(manager->read_fds_copy, manager->read_fds, + manager->fd_bufsize); + memmove(manager->write_fds_copy, manager->write_fds, + manager->fd_bufsize); + maxfd = manager->maxfd + 1; + UNLOCK(&manager->lock); + + cc = select(maxfd, manager->read_fds_copy, + manager->write_fds_copy, NULL, NULL); +#endif /* USE_KQUEUE */ + + if (cc < 0 && !SOFT_ERROR(errno)) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, + "%s %s: %s", fnname, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), strbuf); + } + +#if defined(USE_DEVPOLL) && defined(ISC_SOCKET_USE_POLLWATCH) + if (cc == 0) { + if (pollstate == poll_active) + pollstate = poll_checking; + else if (pollstate == poll_checking) + pollstate = poll_idle; + } else if (cc > 0) { + if (pollstate == poll_checking) { + /* + * XXX: We'd like to use a more + * verbose log level as it's actually an + * unexpected event, but the kernel bug + * reportedly happens pretty frequently + * (and it can also be a false positive) + * so it would be just too noisy. + */ + manager_log(manager, + ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_DEBUG(1), + "unexpected POLL timeout"); + } + pollstate = poll_active; + } +#endif + } while (cc < 0); + +#if defined(USE_KQUEUE) || defined (USE_EPOLL) || defined (USE_DEVPOLL) + done = process_fds(manager, manager->events, cc); +#elif defined(USE_SELECT) + process_fds(manager, maxfd, manager->read_fds_copy, + manager->write_fds_copy); + + /* + * Process reads on internal, control fd. + */ + if (FD_ISSET(ctlfd, manager->read_fds_copy)) + done = process_ctlfd(manager); +#endif + } + + manager_log(manager, TRACE, "%s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_EXITING, "watcher exiting")); + + return ((isc_threadresult_t)0); +} +#endif /* USE_WATCHER_THREAD */ + +void +isc__socketmgr_setreserved(isc_socketmgr_t *manager0, uint32_t reserved) { + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + + REQUIRE(VALID_MANAGER(manager)); + + manager->reserved = reserved; +} + +void +isc__socketmgr_maxudp(isc_socketmgr_t *manager0, int maxudp) { + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + + REQUIRE(VALID_MANAGER(manager)); + + manager->maxudp = maxudp; +} + +/* + * Create a new socket manager. + */ + +static isc_result_t +setup_watcher(isc_mem_t *mctx, isc__socketmgr_t *manager) { + isc_result_t result; +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) + char strbuf[ISC_STRERRORSIZE]; +#endif + +#ifdef USE_KQUEUE + manager->nevents = ISC_SOCKET_MAXEVENTS; + manager->events = isc_mem_get(mctx, sizeof(struct kevent) * + manager->nevents); + if (manager->events == NULL) + return (ISC_R_NOMEMORY); + manager->kqueue_fd = kqueue(); + if (manager->kqueue_fd == -1) { + result = isc__errno2result(errno); + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "kqueue %s: %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + isc_mem_put(mctx, manager->events, + sizeof(struct kevent) * manager->nevents); + return (result); + } + +#ifdef USE_WATCHER_THREAD + result = watch_fd(manager, manager->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + close(manager->kqueue_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct kevent) * manager->nevents); + return (result); + } +#endif /* USE_WATCHER_THREAD */ +#elif defined(USE_EPOLL) + manager->nevents = ISC_SOCKET_MAXEVENTS; + manager->events = isc_mem_get(mctx, sizeof(struct epoll_event) * + manager->nevents); + if (manager->events == NULL) + return (ISC_R_NOMEMORY); + manager->epoll_fd = epoll_create(manager->nevents); + if (manager->epoll_fd == -1) { + result = isc__errno2result(errno); + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "epoll_create %s: %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + isc_mem_put(mctx, manager->events, + sizeof(struct epoll_event) * manager->nevents); + return (result); + } +#ifdef USE_WATCHER_THREAD + result = watch_fd(manager, manager->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + close(manager->epoll_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct epoll_event) * manager->nevents); + return (result); + } +#endif /* USE_WATCHER_THREAD */ +#elif defined(USE_DEVPOLL) + manager->nevents = ISC_SOCKET_MAXEVENTS; + result = isc_resource_getcurlimit(isc_resource_openfiles, + &manager->open_max); + if (result != ISC_R_SUCCESS) + manager->open_max = 64; + manager->calls = 0; + manager->events = isc_mem_get(mctx, sizeof(struct pollfd) * + manager->nevents); + if (manager->events == NULL) + return (ISC_R_NOMEMORY); + /* + * Note: fdpollinfo should be able to support all possible FDs, so + * it must have maxsocks entries (not nevents). + */ + manager->fdpollinfo = isc_mem_get(mctx, sizeof(pollinfo_t) * + manager->maxsocks); + if (manager->fdpollinfo == NULL) { + isc_mem_put(mctx, manager->events, + sizeof(struct pollfd) * manager->nevents); + return (ISC_R_NOMEMORY); + } + memset(manager->fdpollinfo, 0, sizeof(pollinfo_t) * manager->maxsocks); + manager->devpoll_fd = open("/dev/poll", O_RDWR); + if (manager->devpoll_fd == -1) { + result = isc__errno2result(errno); + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "open(/dev/poll) %s: %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + isc_mem_put(mctx, manager->events, + sizeof(struct pollfd) * manager->nevents); + isc_mem_put(mctx, manager->fdpollinfo, + sizeof(pollinfo_t) * manager->maxsocks); + return (result); + } +#ifdef USE_WATCHER_THREAD + result = watch_fd(manager, manager->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + close(manager->devpoll_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct pollfd) * manager->nevents); + isc_mem_put(mctx, manager->fdpollinfo, + sizeof(pollinfo_t) * manager->maxsocks); + return (result); + } +#endif /* USE_WATCHER_THREAD */ +#elif defined(USE_SELECT) + UNUSED(result); + +#if ISC_SOCKET_MAXSOCKETS > FD_SETSIZE + /* + * Note: this code should also cover the case of MAXSOCKETS <= + * FD_SETSIZE, but we separate the cases to avoid possible portability + * issues regarding howmany() and the actual representation of fd_set. + */ + manager->fd_bufsize = howmany(manager->maxsocks, NFDBITS) * + sizeof(fd_mask); +#else + manager->fd_bufsize = sizeof(fd_set); +#endif + + manager->read_fds = NULL; + manager->read_fds_copy = NULL; + manager->write_fds = NULL; + manager->write_fds_copy = NULL; + + manager->read_fds = isc_mem_get(mctx, manager->fd_bufsize); + if (manager->read_fds != NULL) + manager->read_fds_copy = isc_mem_get(mctx, manager->fd_bufsize); + if (manager->read_fds_copy != NULL) + manager->write_fds = isc_mem_get(mctx, manager->fd_bufsize); + if (manager->write_fds != NULL) { + manager->write_fds_copy = isc_mem_get(mctx, + manager->fd_bufsize); + } + if (manager->write_fds_copy == NULL) { + if (manager->write_fds != NULL) { + isc_mem_put(mctx, manager->write_fds, + manager->fd_bufsize); + } + if (manager->read_fds_copy != NULL) { + isc_mem_put(mctx, manager->read_fds_copy, + manager->fd_bufsize); + } + if (manager->read_fds != NULL) { + isc_mem_put(mctx, manager->read_fds, + manager->fd_bufsize); + } + return (ISC_R_NOMEMORY); + } + memset(manager->read_fds, 0, manager->fd_bufsize); + memset(manager->write_fds, 0, manager->fd_bufsize); + +#ifdef USE_WATCHER_THREAD + (void)watch_fd(manager, manager->pipe_fds[0], SELECT_POKE_READ); + manager->maxfd = manager->pipe_fds[0]; +#else /* USE_WATCHER_THREAD */ + manager->maxfd = 0; +#endif /* USE_WATCHER_THREAD */ +#endif /* USE_KQUEUE */ + + return (ISC_R_SUCCESS); +} + +static void +cleanup_watcher(isc_mem_t *mctx, isc__socketmgr_t *manager) { +#ifdef USE_WATCHER_THREAD + isc_result_t result; + + result = unwatch_fd(manager, manager->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "epoll_ctl(DEL) %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed")); + } +#endif /* USE_WATCHER_THREAD */ + +#ifdef USE_KQUEUE + close(manager->kqueue_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct kevent) * manager->nevents); +#elif defined(USE_EPOLL) + close(manager->epoll_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct epoll_event) * manager->nevents); +#elif defined(USE_DEVPOLL) + close(manager->devpoll_fd); + isc_mem_put(mctx, manager->events, + sizeof(struct pollfd) * manager->nevents); + isc_mem_put(mctx, manager->fdpollinfo, + sizeof(pollinfo_t) * manager->maxsocks); +#elif defined(USE_SELECT) + if (manager->read_fds != NULL) + isc_mem_put(mctx, manager->read_fds, manager->fd_bufsize); + if (manager->read_fds_copy != NULL) + isc_mem_put(mctx, manager->read_fds_copy, manager->fd_bufsize); + if (manager->write_fds != NULL) + isc_mem_put(mctx, manager->write_fds, manager->fd_bufsize); + if (manager->write_fds_copy != NULL) + isc_mem_put(mctx, manager->write_fds_copy, manager->fd_bufsize); +#endif /* USE_KQUEUE */ +} + +isc_result_t +isc__socketmgr_create(isc_mem_t *mctx, isc_socketmgr_t **managerp) { + return (isc__socketmgr_create2(mctx, managerp, 0)); +} + +isc_result_t +isc__socketmgr_create2(isc_mem_t *mctx, isc_socketmgr_t **managerp, + unsigned int maxsocks) +{ + int i; + isc__socketmgr_t *manager; +#ifdef USE_WATCHER_THREAD + char strbuf[ISC_STRERRORSIZE]; +#endif + isc_result_t result; + + REQUIRE(managerp != NULL && *managerp == NULL); + +#ifdef USE_SHARED_MANAGER + if (socketmgr != NULL) { + /* Don't allow maxsocks to be updated */ + if (maxsocks > 0 && socketmgr->maxsocks != maxsocks) + return (ISC_R_EXISTS); + + socketmgr->refs++; + *managerp = (isc_socketmgr_t *)socketmgr; + return (ISC_R_SUCCESS); + } +#endif /* USE_SHARED_MANAGER */ + + if (maxsocks == 0) + maxsocks = ISC_SOCKET_MAXSOCKETS; + + manager = isc_mem_get(mctx, sizeof(*manager)); + if (manager == NULL) + return (ISC_R_NOMEMORY); + + /* zero-clear so that necessary cleanup on failure will be easy */ + memset(manager, 0, sizeof(*manager)); + manager->maxsocks = maxsocks; + manager->reserved = 0; + manager->maxudp = 0; + manager->fds = isc_mem_get(mctx, + manager->maxsocks * sizeof(isc__socket_t *)); + if (manager->fds == NULL) { + result = ISC_R_NOMEMORY; + goto free_manager; + } + manager->fdstate = isc_mem_get(mctx, manager->maxsocks * sizeof(int)); + if (manager->fdstate == NULL) { + result = ISC_R_NOMEMORY; + goto free_manager; + } +#if defined(USE_EPOLL) + manager->epoll_events = isc_mem_get(mctx, (manager->maxsocks * + sizeof(uint32_t))); + if (manager->epoll_events == NULL) { + result = ISC_R_NOMEMORY; + goto free_manager; + } + memset(manager->epoll_events, 0, manager->maxsocks * sizeof(uint32_t)); +#endif + manager->stats = NULL; + + manager->common.methods = &socketmgrmethods; + manager->common.magic = ISCAPI_SOCKETMGR_MAGIC; + manager->common.impmagic = SOCKET_MANAGER_MAGIC; + manager->mctx = NULL; + memset(manager->fds, 0, manager->maxsocks * sizeof(isc_socket_t *)); + ISC_LIST_INIT(manager->socklist); + result = isc_mutex_init(&manager->lock); + if (result != ISC_R_SUCCESS) + goto free_manager; + manager->fdlock = isc_mem_get(mctx, FDLOCK_COUNT * sizeof(isc_mutex_t)); + if (manager->fdlock == NULL) { + result = ISC_R_NOMEMORY; + goto cleanup_lock; + } + for (i = 0; i < FDLOCK_COUNT; i++) { + result = isc_mutex_init(&manager->fdlock[i]); + if (result != ISC_R_SUCCESS) { + while (--i >= 0) + DESTROYLOCK(&manager->fdlock[i]); + isc_mem_put(mctx, manager->fdlock, + FDLOCK_COUNT * sizeof(isc_mutex_t)); + manager->fdlock = NULL; + goto cleanup_lock; + } + } + +#ifdef USE_WATCHER_THREAD + if (isc_condition_init(&manager->shutdown_ok) != ISC_R_SUCCESS) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_condition_init() %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed")); + result = ISC_R_UNEXPECTED; + goto cleanup_lock; + } + + /* + * Create the special fds that will be used to wake up the + * select/poll loop when something internal needs to be done. + */ + if (pipe(manager->pipe_fds) != 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "pipe() %s: %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed"), + strbuf); + result = ISC_R_UNEXPECTED; + goto cleanup_condition; + } + + RUNTIME_CHECK(make_nonblock(manager->pipe_fds[0]) == ISC_R_SUCCESS); +#if 0 + RUNTIME_CHECK(make_nonblock(manager->pipe_fds[1]) == ISC_R_SUCCESS); +#endif +#endif /* USE_WATCHER_THREAD */ + +#ifdef USE_SHARED_MANAGER + manager->refs = 1; +#endif /* USE_SHARED_MANAGER */ + + /* + * Set up initial state for the select loop + */ + result = setup_watcher(mctx, manager); + if (result != ISC_R_SUCCESS) + goto cleanup; + + memset(manager->fdstate, 0, manager->maxsocks * sizeof(int)); + +#ifdef USE_WATCHER_THREAD + /* + * Start up the select/poll thread. + */ + if (isc_thread_create(watcher, manager, &manager->watcher) != + ISC_R_SUCCESS) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_thread_create() %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed")); + cleanup_watcher(mctx, manager); + result = ISC_R_UNEXPECTED; + goto cleanup; + } + isc_thread_setname(manager->watcher, "isc-socket"); +#endif /* USE_WATCHER_THREAD */ + isc_mem_attach(mctx, &manager->mctx); + +#ifdef USE_SHARED_MANAGER + socketmgr = manager; +#endif /* USE_SHARED_MANAGER */ + *managerp = (isc_socketmgr_t *)manager; + + return (ISC_R_SUCCESS); + +cleanup: +#ifdef USE_WATCHER_THREAD + (void)close(manager->pipe_fds[0]); + (void)close(manager->pipe_fds[1]); +#endif /* USE_WATCHER_THREAD */ + +#ifdef USE_WATCHER_THREAD +cleanup_condition: + (void)isc_condition_destroy(&manager->shutdown_ok); +#endif /* USE_WATCHER_THREAD */ + + +cleanup_lock: + if (manager->fdlock != NULL) { + for (i = 0; i < FDLOCK_COUNT; i++) + DESTROYLOCK(&manager->fdlock[i]); + } + DESTROYLOCK(&manager->lock); + +free_manager: + if (manager->fdlock != NULL) { + isc_mem_put(mctx, manager->fdlock, + FDLOCK_COUNT * sizeof(isc_mutex_t)); + } +#if defined(USE_EPOLL) + if (manager->epoll_events != NULL) { + isc_mem_put(mctx, manager->epoll_events, + manager->maxsocks * sizeof(uint32_t)); + } +#endif + if (manager->fdstate != NULL) { + isc_mem_put(mctx, manager->fdstate, + manager->maxsocks * sizeof(int)); + } + if (manager->fds != NULL) { + isc_mem_put(mctx, manager->fds, + manager->maxsocks * sizeof(isc_socket_t *)); + } + isc_mem_put(mctx, manager, sizeof(*manager)); + + return (result); +} + +isc_result_t +isc_socketmgr_getmaxsockets(isc_socketmgr_t *manager0, unsigned int *nsockp) { + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(nsockp != NULL); + + *nsockp = manager->maxsocks; + + return (ISC_R_SUCCESS); +} + +void +isc_socketmgr_setstats(isc_socketmgr_t *manager0, isc_stats_t *stats) { + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(ISC_LIST_EMPTY(manager->socklist)); + REQUIRE(manager->stats == NULL); + REQUIRE(isc_stats_ncounters(stats) == isc_sockstatscounter_max); + + isc_stats_attach(stats, &manager->stats); +} + +void +isc__socketmgr_destroy(isc_socketmgr_t **managerp) { + isc__socketmgr_t *manager; + int i; + isc_mem_t *mctx; + + /* + * Destroy a socket manager. + */ + + REQUIRE(managerp != NULL); + manager = (isc__socketmgr_t *)*managerp; + REQUIRE(VALID_MANAGER(manager)); + +#ifdef USE_SHARED_MANAGER + manager->refs--; + if (manager->refs > 0) { + *managerp = NULL; + return; + } + socketmgr = NULL; +#endif /* USE_SHARED_MANAGER */ + + LOCK(&manager->lock); + + /* + * Wait for all sockets to be destroyed. + */ + while (!ISC_LIST_EMPTY(manager->socklist)) { +#ifdef USE_WATCHER_THREAD + manager_log(manager, CREATION, "%s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_SOCKETSREMAIN, + "sockets exist")); + WAIT(&manager->shutdown_ok, &manager->lock); +#else /* USE_WATCHER_THREAD */ + UNLOCK(&manager->lock); + isc__taskmgr_dispatch(NULL); + LOCK(&manager->lock); +#endif /* USE_WATCHER_THREAD */ + } + + UNLOCK(&manager->lock); + + /* + * Here, poke our select/poll thread. Do this by closing the write + * half of the pipe, which will send EOF to the read half. + * This is currently a no-op in the non-threaded case. + */ + select_poke(manager, 0, SELECT_POKE_SHUTDOWN); + +#ifdef USE_WATCHER_THREAD + /* + * Wait for thread to exit. + */ + if (isc_thread_join(manager->watcher, NULL) != ISC_R_SUCCESS) + UNEXPECTED_ERROR(__FILE__, __LINE__, + "isc_thread_join() %s", + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed")); +#endif /* USE_WATCHER_THREAD */ + + /* + * Clean up. + */ + cleanup_watcher(manager->mctx, manager); + +#ifdef USE_WATCHER_THREAD + (void)close(manager->pipe_fds[0]); + (void)close(manager->pipe_fds[1]); + (void)isc_condition_destroy(&manager->shutdown_ok); +#endif /* USE_WATCHER_THREAD */ + + for (i = 0; i < (int)manager->maxsocks; i++) + if (manager->fdstate[i] == CLOSE_PENDING) /* no need to lock */ + (void)close(i); + +#if defined(USE_EPOLL) + isc_mem_put(manager->mctx, manager->epoll_events, + manager->maxsocks * sizeof(uint32_t)); +#endif + isc_mem_put(manager->mctx, manager->fds, + manager->maxsocks * sizeof(isc__socket_t *)); + isc_mem_put(manager->mctx, manager->fdstate, + manager->maxsocks * sizeof(int)); + + if (manager->stats != NULL) + isc_stats_detach(&manager->stats); + + if (manager->fdlock != NULL) { + for (i = 0; i < FDLOCK_COUNT; i++) + DESTROYLOCK(&manager->fdlock[i]); + isc_mem_put(manager->mctx, manager->fdlock, + FDLOCK_COUNT * sizeof(isc_mutex_t)); + } + DESTROYLOCK(&manager->lock); + manager->common.magic = 0; + manager->common.impmagic = 0; + mctx= manager->mctx; + isc_mem_put(mctx, manager, sizeof(*manager)); + + isc_mem_detach(&mctx); + + *managerp = NULL; + +#ifdef USE_SHARED_MANAGER + socketmgr = NULL; +#endif +} + +static isc_result_t +socket_recv(isc__socket_t *sock, isc_socketevent_t *dev, isc_task_t *task, + unsigned int flags) +{ + int io_state; + bool have_lock = false; + isc_task_t *ntask = NULL; + isc_result_t result = ISC_R_SUCCESS; + + dev->ev_sender = task; + + if (sock->type == isc_sockettype_udp) { + io_state = doio_recv(sock, dev); + } else { + LOCK(&sock->lock); + have_lock = true; + + if (ISC_LIST_EMPTY(sock->recv_list)) + io_state = doio_recv(sock, dev); + else + io_state = DOIO_SOFT; + } + + switch (io_state) { + case DOIO_SOFT: + /* + * We couldn't read all or part of the request right now, so + * queue it. + * + * Attach to socket and to task + */ + isc_task_attach(task, &ntask); + dev->attributes |= ISC_SOCKEVENTATTR_ATTACHED; + + if (!have_lock) { + LOCK(&sock->lock); + have_lock = true; + } + + /* + * Enqueue the request. If the socket was previously not being + * watched, poke the watcher to start paying attention to it. + */ + if (ISC_LIST_EMPTY(sock->recv_list) && !sock->pending_recv) + select_poke(sock->manager, sock->fd, SELECT_POKE_READ); + ISC_LIST_ENQUEUE(sock->recv_list, dev, ev_link); + + socket_log(sock, NULL, EVENT, NULL, 0, 0, + "socket_recv: event %p -> task %p", + dev, ntask); + + if ((flags & ISC_SOCKFLAG_IMMEDIATE) != 0) + result = ISC_R_INPROGRESS; + break; + + case DOIO_EOF: + dev->result = ISC_R_EOF; + /* fallthrough */ + + case DOIO_HARD: + case DOIO_SUCCESS: + if ((flags & ISC_SOCKFLAG_IMMEDIATE) == 0) + send_recvdone_event(sock, &dev); + break; + } + + if (have_lock) + UNLOCK(&sock->lock); + + return (result); +} + +isc_result_t +isc__socket_recvv(isc_socket_t *sock0, isc_bufferlist_t *buflist, + unsigned int minimum, isc_task_t *task, + isc_taskaction_t action, void *arg) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socketevent_t *dev; + isc__socketmgr_t *manager; + unsigned int iocount; + isc_buffer_t *buffer; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(buflist != NULL); + REQUIRE(!ISC_LIST_EMPTY(*buflist)); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + iocount = isc_bufferlist_availablecount(buflist); + REQUIRE(iocount > 0); + + INSIST(sock->bound); + + dev = allocate_socketevent(manager->mctx, sock, + ISC_SOCKEVENT_RECVDONE, action, arg); + if (dev == NULL) + return (ISC_R_NOMEMORY); + + /* + * UDP sockets are always partial read + */ + if (sock->type == isc_sockettype_udp) + dev->minimum = 1; + else { + if (minimum == 0) + dev->minimum = iocount; + else + dev->minimum = minimum; + } + + /* + * Move each buffer from the passed in list to our internal one. + */ + buffer = ISC_LIST_HEAD(*buflist); + while (buffer != NULL) { + ISC_LIST_DEQUEUE(*buflist, buffer, link); + ISC_LIST_ENQUEUE(dev->bufferlist, buffer, link); + buffer = ISC_LIST_HEAD(*buflist); + } + + return (socket_recv(sock, dev, task, 0)); +} + +isc_result_t +isc__socket_recv(isc_socket_t *sock0, isc_region_t *region, + unsigned int minimum, isc_task_t *task, + isc_taskaction_t action, void *arg) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socketevent_t *dev; + isc__socketmgr_t *manager; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + INSIST(sock->bound); + + dev = allocate_socketevent(manager->mctx, sock, + ISC_SOCKEVENT_RECVDONE, action, arg); + if (dev == NULL) + return (ISC_R_NOMEMORY); + + return (isc__socket_recv2(sock0, region, minimum, task, dev, 0)); +} + +isc_result_t +isc__socket_recv2(isc_socket_t *sock0, isc_region_t *region, + unsigned int minimum, isc_task_t *task, + isc_socketevent_t *event, unsigned int flags) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + + event->ev_sender = sock; + event->result = ISC_R_UNSET; + ISC_LIST_INIT(event->bufferlist); + event->region = *region; + event->n = 0; + event->offset = 0; + event->attributes = 0; + + /* + * UDP sockets are always partial read. + */ + if (sock->type == isc_sockettype_udp) + event->minimum = 1; + else { + if (minimum == 0) + event->minimum = region->length; + else + event->minimum = minimum; + } + + return (socket_recv(sock, event, task, flags)); +} + +static isc_result_t +socket_send(isc__socket_t *sock, isc_socketevent_t *dev, isc_task_t *task, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + unsigned int flags) +{ + int io_state; + bool have_lock = false; + isc_task_t *ntask = NULL; + isc_result_t result = ISC_R_SUCCESS; + + dev->ev_sender = task; + + set_dev_address(address, sock, dev); + if (pktinfo != NULL) { + dev->attributes |= ISC_SOCKEVENTATTR_PKTINFO; + dev->pktinfo = *pktinfo; + + if (!isc_sockaddr_issitelocal(&dev->address) && + !isc_sockaddr_islinklocal(&dev->address)) { + socket_log(sock, NULL, TRACE, isc_msgcat, + ISC_MSGSET_SOCKET, ISC_MSG_PKTINFOPROVIDED, + "pktinfo structure provided, ifindex %u " + "(set to 0)", pktinfo->ipi6_ifindex); + + /* + * Set the pktinfo index to 0 here, to let the + * kernel decide what interface it should send on. + */ + dev->pktinfo.ipi6_ifindex = 0; + } + } + + if (sock->type == isc_sockettype_udp) + io_state = doio_send(sock, dev); + else { + LOCK(&sock->lock); + have_lock = true; + + if (ISC_LIST_EMPTY(sock->send_list)) + io_state = doio_send(sock, dev); + else + io_state = DOIO_SOFT; + } + + switch (io_state) { + case DOIO_SOFT: + /* + * We couldn't send all or part of the request right now, so + * queue it unless ISC_SOCKFLAG_NORETRY is set. + */ + if ((flags & ISC_SOCKFLAG_NORETRY) == 0) { + isc_task_attach(task, &ntask); + dev->attributes |= ISC_SOCKEVENTATTR_ATTACHED; + + if (!have_lock) { + LOCK(&sock->lock); + have_lock = true; + } + + /* + * Enqueue the request. If the socket was previously + * not being watched, poke the watcher to start + * paying attention to it. + */ + if (ISC_LIST_EMPTY(sock->send_list) && + !sock->pending_send) + select_poke(sock->manager, sock->fd, + SELECT_POKE_WRITE); + ISC_LIST_ENQUEUE(sock->send_list, dev, ev_link); + + socket_log(sock, NULL, EVENT, NULL, 0, 0, + "socket_send: event %p -> task %p", + dev, ntask); + + if ((flags & ISC_SOCKFLAG_IMMEDIATE) != 0) + result = ISC_R_INPROGRESS; + break; + } + + /* FALLTHROUGH */ + + case DOIO_HARD: + case DOIO_SUCCESS: + if ((flags & ISC_SOCKFLAG_IMMEDIATE) == 0) + send_senddone_event(sock, &dev); + break; + } + + if (have_lock) + UNLOCK(&sock->lock); + + return (result); +} + +isc_result_t +isc__socket_send(isc_socket_t *sock, isc_region_t *region, + isc_task_t *task, isc_taskaction_t action, void *arg) +{ + /* + * REQUIRE() checking is performed in isc_socket_sendto(). + */ + return (isc__socket_sendto(sock, region, task, action, arg, NULL, + NULL)); +} + +isc_result_t +isc__socket_sendto(isc_socket_t *sock0, isc_region_t *region, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socketevent_t *dev; + isc__socketmgr_t *manager; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(region != NULL); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + INSIST(sock->bound); + + dev = allocate_socketevent(manager->mctx, sock, + ISC_SOCKEVENT_SENDDONE, action, arg); + if (dev == NULL) + return (ISC_R_NOMEMORY); + + dev->region = *region; + + return (socket_send(sock, dev, task, address, pktinfo, 0)); +} + +isc_result_t +isc__socket_sendv(isc_socket_t *sock, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg) +{ + return (isc__socket_sendtov2(sock, buflist, task, action, arg, NULL, + NULL, 0)); +} + +isc_result_t +isc__socket_sendtov(isc_socket_t *sock, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo) +{ + return (isc__socket_sendtov2(sock, buflist, task, action, arg, address, + pktinfo, 0)); +} + +isc_result_t +isc__socket_sendtov2(isc_socket_t *sock0, isc_bufferlist_t *buflist, + isc_task_t *task, isc_taskaction_t action, void *arg, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + unsigned int flags) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socketevent_t *dev; + isc__socketmgr_t *manager; + unsigned int iocount; + isc_buffer_t *buffer; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(buflist != NULL); + REQUIRE(!ISC_LIST_EMPTY(*buflist)); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + iocount = isc_bufferlist_usedcount(buflist); + REQUIRE(iocount > 0); + + dev = allocate_socketevent(manager->mctx, sock, + ISC_SOCKEVENT_SENDDONE, action, arg); + if (dev == NULL) + return (ISC_R_NOMEMORY); + + /* + * Move each buffer from the passed in list to our internal one. + */ + buffer = ISC_LIST_HEAD(*buflist); + while (buffer != NULL) { + ISC_LIST_DEQUEUE(*buflist, buffer, link); + ISC_LIST_ENQUEUE(dev->bufferlist, buffer, link); + buffer = ISC_LIST_HEAD(*buflist); + } + + return (socket_send(sock, dev, task, address, pktinfo, flags)); +} + +isc_result_t +isc__socket_sendto2(isc_socket_t *sock0, isc_region_t *region, + isc_task_t *task, + isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + isc_socketevent_t *event, unsigned int flags) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE((flags & ~(ISC_SOCKFLAG_IMMEDIATE|ISC_SOCKFLAG_NORETRY)) == 0); + if ((flags & ISC_SOCKFLAG_NORETRY) != 0) + REQUIRE(sock->type == isc_sockettype_udp); + event->ev_sender = sock; + event->result = ISC_R_UNSET; + ISC_LIST_INIT(event->bufferlist); + event->region = *region; + event->n = 0; + event->offset = 0; + event->attributes &= ~ISC_SOCKEVENTATTR_ATTACHED; + + return (socket_send(sock, event, task, address, pktinfo, flags)); +} + +void +isc__socket_cleanunix(isc_sockaddr_t *sockaddr, bool active) { +#ifdef ISC_PLATFORM_HAVESYSUNH + int s; + struct stat sb; + char strbuf[ISC_STRERRORSIZE]; + + if (sockaddr->type.sa.sa_family != AF_UNIX) + return; + +#ifndef S_ISSOCK +#if defined(S_IFMT) && defined(S_IFSOCK) +#define S_ISSOCK(mode) ((mode & S_IFMT)==S_IFSOCK) +#elif defined(_S_IFMT) && defined(S_IFSOCK) +#define S_ISSOCK(mode) ((mode & _S_IFMT)==S_IFSOCK) +#endif +#endif + +#ifndef S_ISFIFO +#if defined(S_IFMT) && defined(S_IFIFO) +#define S_ISFIFO(mode) ((mode & S_IFMT)==S_IFIFO) +#elif defined(_S_IFMT) && defined(S_IFIFO) +#define S_ISFIFO(mode) ((mode & _S_IFMT)==S_IFIFO) +#endif +#endif + +#if !defined(S_ISFIFO) && !defined(S_ISSOCK) +#error You need to define S_ISFIFO and S_ISSOCK as appropriate for your platform. See <sys/stat.h>. +#endif + +#ifndef S_ISFIFO +#define S_ISFIFO(mode) 0 +#endif + +#ifndef S_ISSOCK +#define S_ISSOCK(mode) 0 +#endif + + if (active) { + if (stat(sockaddr->type.sunix.sun_path, &sb) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_cleanunix: stat(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + return; + } + if (!(S_ISSOCK(sb.st_mode) || S_ISFIFO(sb.st_mode))) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_cleanunix: %s: not a socket", + sockaddr->type.sunix.sun_path); + return; + } + if (unlink(sockaddr->type.sunix.sun_path) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_cleanunix: unlink(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + } + return; + } + + s = socket(AF_UNIX, SOCK_STREAM, 0); + if (s < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: socket(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + return; + } + + if (stat(sockaddr->type.sunix.sun_path, &sb) < 0) { + switch (errno) { + case ENOENT: /* We exited cleanly last time */ + break; + default: + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: stat(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + break; + } + goto cleanup; + } + + if (!(S_ISSOCK(sb.st_mode) || S_ISFIFO(sb.st_mode))) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: %s: not a socket", + sockaddr->type.sunix.sun_path); + goto cleanup; + } + + if (connect(s, (struct sockaddr *)&sockaddr->type.sunix, + sizeof(sockaddr->type.sunix)) < 0) { + switch (errno) { + case ECONNREFUSED: + case ECONNRESET: + if (unlink(sockaddr->type.sunix.sun_path) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_WARNING, + "isc_socket_cleanunix: " + "unlink(%s): %s", + sockaddr->type.sunix.sun_path, + strbuf); + } + break; + default: + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: connect(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + break; + } + } + cleanup: + close(s); +#else + UNUSED(sockaddr); + UNUSED(active); +#endif +} + +isc_result_t +isc__socket_permunix(isc_sockaddr_t *sockaddr, uint32_t perm, + uint32_t owner, uint32_t group) +{ +#ifdef ISC_PLATFORM_HAVESYSUNH + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + char path[sizeof(sockaddr->type.sunix.sun_path)]; +#ifdef NEED_SECURE_DIRECTORY + char *slash; +#endif + + REQUIRE(sockaddr->type.sa.sa_family == AF_UNIX); + INSIST(strlen(sockaddr->type.sunix.sun_path) < sizeof(path)); + strlcpy(path, sockaddr->type.sunix.sun_path, sizeof(path)); + +#ifdef NEED_SECURE_DIRECTORY + slash = strrchr(path, '/'); + if (slash != NULL) { + if (slash != path) { + *slash = '\0'; + } else { + strlcpy(path, "/", sizeof(path)); + } + } else { + strlcpy(path, ".", sizeof(path)); + } +#endif + + if (chmod(path, perm) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_permunix: chmod(%s, %d): %s", + path, perm, strbuf); + result = ISC_R_FAILURE; + } + if (chown(path, owner, group) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_permunix: chown(%s, %d, %d): %s", + path, owner, group, + strbuf); + result = ISC_R_FAILURE; + } + return (result); +#else + UNUSED(sockaddr); + UNUSED(perm); + UNUSED(owner); + UNUSED(group); + return (ISC_R_NOTIMPLEMENTED); +#endif +} + +isc_result_t +isc__socket_bind(isc_socket_t *sock0, isc_sockaddr_t *sockaddr, + unsigned int options) { + isc__socket_t *sock = (isc__socket_t *)sock0; + char strbuf[ISC_STRERRORSIZE]; + int on = 1; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + INSIST(!sock->bound); + INSIST(!sock->dupped); + + if (sock->pf != sockaddr->type.sa.sa_family) { + UNLOCK(&sock->lock); + return (ISC_R_FAMILYMISMATCH); + } + + /* + * Only set SO_REUSEADDR when we want a specific port. + */ +#ifdef AF_UNIX + if (sock->pf == AF_UNIX) + goto bind_socket; +#endif + if ((options & ISC_SOCKET_REUSEADDRESS) != 0 && + isc_sockaddr_getport(sockaddr) != (in_port_t)0 && + setsockopt(sock->fd, SOL_SOCKET, SO_REUSEADDR, (void *)&on, + sizeof(on)) < 0) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d) %s", sock->fd, + isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, "failed")); + /* Press on... */ + } +#ifdef AF_UNIX + bind_socket: +#endif + if (bind(sock->fd, &sockaddr->type.sa, sockaddr->length) < 0) { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_BINDFAIL]); + + UNLOCK(&sock->lock); + switch (errno) { + case EACCES: + return (ISC_R_NOPERM); + case EADDRNOTAVAIL: + return (ISC_R_ADDRNOTAVAIL); + case EADDRINUSE: + return (ISC_R_ADDRINUSE); + case EINVAL: + return (ISC_R_BOUND); + default: + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "bind: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + } + + socket_log(sock, sockaddr, TRACE, + isc_msgcat, ISC_MSGSET_SOCKET, ISC_MSG_BOUND, "bound"); + sock->bound = 1; + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * Enable this only for specific OS versions, and only when they have repaired + * their problems with it. Until then, this is is broken and needs to be + * diabled by default. See RT22589 for details. + */ +#undef ENABLE_ACCEPTFILTER + +isc_result_t +isc__socket_filter(isc_socket_t *sock0, const char *filter) { + isc__socket_t *sock = (isc__socket_t *)sock0; +#if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) + char strbuf[ISC_STRERRORSIZE]; + struct accept_filter_arg afa; +#else + UNUSED(sock); + UNUSED(filter); +#endif + + REQUIRE(VALID_SOCKET(sock)); + +#if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) + bzero(&afa, sizeof(afa)); + strlcpy(afa.af_name, filter, sizeof(afa.af_name)); + if (setsockopt(sock->fd, SOL_SOCKET, SO_ACCEPTFILTER, + &afa, sizeof(afa)) == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET, + ISC_MSG_FILTER, "setsockopt(SO_ACCEPTFILTER): %s", + strbuf); + return (ISC_R_FAILURE); + } + return (ISC_R_SUCCESS); +#else + return (ISC_R_NOTIMPLEMENTED); +#endif +} + +/* + * Try enabling TCP Fast Open for a given socket if the OS supports it. + */ +static void +set_tcp_fastopen(isc__socket_t *sock, unsigned int backlog) { +#if defined(ISC_PLATFORM_HAVETFO) && defined(TCP_FASTOPEN) + char strbuf[ISC_STRERRORSIZE]; + +/* + * FreeBSD, as of versions 10.3 and 11.0, defines TCP_FASTOPEN while also + * shipping a default kernel without TFO support, so we special-case it by + * performing an additional runtime check for TFO support using sysctl to + * prevent setsockopt() errors from being logged. + */ +#if defined(__FreeBSD__) && defined(HAVE_SYSCTLBYNAME) +#define SYSCTL_TFO "net.inet.tcp.fastopen.enabled" + unsigned int enabled; + size_t enabledlen = sizeof(enabled); + static bool tfo_notice_logged = false; + + if (sysctlbyname(SYSCTL_TFO, &enabled, &enabledlen, NULL, 0) < 0) { + /* + * This kernel does not support TCP Fast Open. There is + * nothing more we can do. + */ + return; + } else if (enabled == 0) { + /* + * This kernel does support TCP Fast Open, but it is disabled + * by sysctl. Notify the user, but do not nag. + */ + if (!tfo_notice_logged) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_NOTICE, + "TCP_FASTOPEN support is disabled by " + "sysctl (" SYSCTL_TFO " = 0)"); + tfo_notice_logged = true; + } + return; + } +#endif + +#ifdef __APPLE__ + backlog = 1; +#else + backlog = backlog / 2; + if (backlog == 0) + backlog = 1; +#endif + if (setsockopt(sock->fd, IPPROTO_TCP, TCP_FASTOPEN, + (void *)&backlog, sizeof(backlog)) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, TCP_FASTOPEN) failed with %s", + sock->fd, strbuf); + /* TCP_FASTOPEN is experimental so ignore failures */ + } +#else + UNUSED(sock); + UNUSED(backlog); +#endif +} + +/* + * Set up to listen on a given socket. We do this by creating an internal + * event that will be dispatched when the socket has read activity. The + * watcher will send the internal event to the task when there is a new + * connection. + * + * Unlike in read, we don't preallocate a done event here. Every time there + * is a new connection we'll have to allocate a new one anyway, so we might + * as well keep things simple rather than having to track them. + */ +isc_result_t +isc__socket_listen(isc_socket_t *sock0, unsigned int backlog) { + isc__socket_t *sock = (isc__socket_t *)sock0; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + REQUIRE(!sock->listener); + REQUIRE(sock->bound); + REQUIRE(sock->type == isc_sockettype_tcp || + sock->type == isc_sockettype_unix); + + if (backlog == 0) + backlog = SOMAXCONN; + + if (listen(sock->fd, (int)backlog) < 0) { + UNLOCK(&sock->lock); + isc__strerror(errno, strbuf, sizeof(strbuf)); + + UNEXPECTED_ERROR(__FILE__, __LINE__, "listen: %s", strbuf); + + return (ISC_R_UNEXPECTED); + } + + set_tcp_fastopen(sock, backlog); + + sock->listener = 1; + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * This should try to do aggressive accept() XXXMLG + */ +isc_result_t +isc__socket_accept(isc_socket_t *sock0, + isc_task_t *task, isc_taskaction_t action, void *arg) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socket_newconnev_t *dev; + isc__socketmgr_t *manager; + isc_task_t *ntask = NULL; + isc__socket_t *nsock; + isc_result_t result; + bool do_poke = false; + + REQUIRE(VALID_SOCKET(sock)); + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + LOCK(&sock->lock); + + REQUIRE(sock->listener); + + /* + * Sender field is overloaded here with the task we will be sending + * this event to. Just before the actual event is delivered the + * actual ev_sender will be touched up to be the socket. + */ + dev = (isc_socket_newconnev_t *) + isc_event_allocate(manager->mctx, task, ISC_SOCKEVENT_NEWCONN, + action, arg, sizeof(*dev)); + if (dev == NULL) { + UNLOCK(&sock->lock); + return (ISC_R_NOMEMORY); + } + ISC_LINK_INIT(dev, ev_link); + + result = allocate_socket(manager, sock->type, &nsock); + if (result != ISC_R_SUCCESS) { + isc_event_free(ISC_EVENT_PTR(&dev)); + UNLOCK(&sock->lock); + return (result); + } + + /* + * Attach to socket and to task. + */ + isc_task_attach(task, &ntask); + if (isc_task_exiting(ntask)) { + free_socket(&nsock); + isc_task_detach(&ntask); + isc_event_free(ISC_EVENT_PTR(&dev)); + UNLOCK(&sock->lock); + return (ISC_R_SHUTTINGDOWN); + } + nsock->references++; + nsock->statsindex = sock->statsindex; + + dev->ev_sender = ntask; + dev->newsocket = (isc_socket_t *)nsock; + + /* + * Poke watcher here. We still have the socket locked, so there + * is no race condition. We will keep the lock for such a short + * bit of time waking it up now or later won't matter all that much. + */ + if (ISC_LIST_EMPTY(sock->accept_list)) + do_poke = true; + + ISC_LIST_ENQUEUE(sock->accept_list, dev, ev_link); + + if (do_poke) + select_poke(manager, sock->fd, SELECT_POKE_ACCEPT); + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc__socket_connect(isc_socket_t *sock0, isc_sockaddr_t *addr, + isc_task_t *task, isc_taskaction_t action, void *arg) +{ + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_socket_connev_t *dev; + isc_task_t *ntask = NULL; + isc__socketmgr_t *manager; + int cc; + char strbuf[ISC_STRERRORSIZE]; + char addrbuf[ISC_SOCKADDR_FORMATSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addr != NULL); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(addr != NULL); + + if (isc_sockaddr_ismulticast(addr)) + return (ISC_R_MULTICAST); + + LOCK(&sock->lock); + + dev = (isc_socket_connev_t *)isc_event_allocate(manager->mctx, sock, + ISC_SOCKEVENT_CONNECT, + action, arg, + sizeof(*dev)); + if (dev == NULL) { + UNLOCK(&sock->lock); + return (ISC_R_NOMEMORY); + } + ISC_LINK_INIT(dev, ev_link); + + if (sock->connecting) { + INSIST(isc_sockaddr_equal(&sock->peer_address, addr)); + goto queue; + } + + if (sock->connected) { + INSIST(isc_sockaddr_equal(&sock->peer_address, addr)); + dev->result = ISC_R_SUCCESS; + isc_task_send(task, ISC_EVENT_PTR(&dev)); + + UNLOCK(&sock->lock); + + return (ISC_R_SUCCESS); + } + + /* + * Try to do the connect right away, as there can be only one + * outstanding, and it might happen to complete. + */ + sock->peer_address = *addr; + cc = connect(sock->fd, &addr->type.sa, addr->length); + if (cc < 0) { + /* + * HP-UX "fails" to connect a UDP socket and sets errno to + * EINPROGRESS if it's non-blocking. We'd rather regard this as + * a success and let the user detect it if it's really an error + * at the time of sending a packet on the socket. + */ + if (sock->type == isc_sockettype_udp && errno == EINPROGRESS) { + cc = 0; + goto success; + } + if (SOFT_ERROR(errno) || errno == EINPROGRESS) + goto queue; + + switch (errno) { +#define ERROR_MATCH(a, b) case a: dev->result = b; goto err_exit; + ERROR_MATCH(EACCES, ISC_R_NOPERM); + ERROR_MATCH(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(ECONNREFUSED, ISC_R_CONNREFUSED); + ERROR_MATCH(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ERROR_MATCH(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif + ERROR_MATCH(ENETUNREACH, ISC_R_NETUNREACH); + ERROR_MATCH(ENOBUFS, ISC_R_NORESOURCES); + ERROR_MATCH(EPERM, ISC_R_HOSTUNREACH); + ERROR_MATCH(EPIPE, ISC_R_NOTCONNECTED); + ERROR_MATCH(ECONNRESET, ISC_R_CONNECTIONRESET); +#undef ERROR_MATCH + } + + sock->connected = 0; + + isc__strerror(errno, strbuf, sizeof(strbuf)); + isc_sockaddr_format(addr, addrbuf, sizeof(addrbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "connect(%s) %d/%s", + addrbuf, errno, strbuf); + + UNLOCK(&sock->lock); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + isc_event_free(ISC_EVENT_PTR(&dev)); + return (ISC_R_UNEXPECTED); + + err_exit: + sock->connected = 0; + isc_task_send(task, ISC_EVENT_PTR(&dev)); + + UNLOCK(&sock->lock); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + return (ISC_R_SUCCESS); + } + + /* + * If connect completed, fire off the done event. + */ + success: + if (cc == 0) { + sock->connected = 1; + sock->bound = 1; + dev->result = ISC_R_SUCCESS; + isc_task_send(task, ISC_EVENT_PTR(&dev)); + + UNLOCK(&sock->lock); + + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECT]); + + return (ISC_R_SUCCESS); + } + + queue: + + /* + * Attach to task. + */ + isc_task_attach(task, &ntask); + + dev->ev_sender = ntask; + + /* + * Poke watcher here. We still have the socket locked, so there + * is no race condition. We will keep the lock for such a short + * bit of time waking it up now or later won't matter all that much. + */ + if (ISC_LIST_EMPTY(sock->connect_list) && !sock->connecting) + select_poke(manager, sock->fd, SELECT_POKE_CONNECT); + + sock->connecting = 1; + + ISC_LIST_ENQUEUE(sock->connect_list, dev, ev_link); + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * Called when a socket with a pending connect() finishes. + */ +static void +internal_connect(isc_task_t *me, isc_event_t *ev) { + isc__socket_t *sock; + isc_socket_connev_t *dev; + int cc; + isc_result_t result; + ISC_SOCKADDR_LEN_T optlen; + char strbuf[ISC_STRERRORSIZE]; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + + UNUSED(me); + INSIST(ev->ev_type == ISC_SOCKEVENT_INTW); + + sock = ev->ev_sender; + INSIST(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + /* + * When the internal event was sent the reference count was bumped + * to keep the socket around for us. Decrement the count here. + */ + INSIST(sock->references > 0); + sock->references--; + if (sock->references == 0) { + UNLOCK(&sock->lock); + destroy(&sock); + return; + } + + /* + * Get the first item off the connect list. + * If it is empty, unlock the socket and return. + */ + dev = ISC_LIST_HEAD(sock->connect_list); + if (dev == NULL) { + INSIST(!sock->connecting); + UNLOCK(&sock->lock); + return; + } + + INSIST(sock->connecting); + sock->connecting = 0; + + /* + * Get any possible error status here. + */ + optlen = sizeof(cc); + if (getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, + (void *)&cc, (void *)&optlen) != 0) + cc = errno; + else + errno = cc; + + if (errno != 0) { + /* + * If the error is EAGAIN, just re-select on this + * fd and pretend nothing strange happened. + */ + if (SOFT_ERROR(errno) || errno == EINPROGRESS) { + sock->connecting = 1; + select_poke(sock->manager, sock->fd, + SELECT_POKE_CONNECT); + UNLOCK(&sock->lock); + + return; + } + + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + + /* + * Translate other errors into ISC_R_* flavors. + */ + switch (errno) { +#define ERROR_MATCH(a, b) case a: result = b; break; + ERROR_MATCH(EACCES, ISC_R_NOPERM); + ERROR_MATCH(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(ECONNREFUSED, ISC_R_CONNREFUSED); + ERROR_MATCH(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ERROR_MATCH(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif + ERROR_MATCH(ENETUNREACH, ISC_R_NETUNREACH); + ERROR_MATCH(ENOBUFS, ISC_R_NORESOURCES); + ERROR_MATCH(EPERM, ISC_R_HOSTUNREACH); + ERROR_MATCH(EPIPE, ISC_R_NOTCONNECTED); + ERROR_MATCH(ETIMEDOUT, ISC_R_TIMEDOUT); + ERROR_MATCH(ECONNRESET, ISC_R_CONNECTIONRESET); +#undef ERROR_MATCH + default: + result = ISC_R_UNEXPECTED; + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_connect: connect(%s) %s", + peerbuf, strbuf); + } + } else { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECT]); + result = ISC_R_SUCCESS; + sock->connected = 1; + sock->bound = 1; + } + + do { + dev->result = result; + send_connectdone_event(sock, &dev); + dev = ISC_LIST_HEAD(sock->connect_list); + } while (dev != NULL); + + UNLOCK(&sock->lock); +} + +isc_result_t +isc__socket_getpeername(isc_socket_t *sock0, isc_sockaddr_t *addressp) { + isc__socket_t *sock = (isc__socket_t *)sock0; + isc_result_t result; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addressp != NULL); + + LOCK(&sock->lock); + + if (sock->connected) { + *addressp = sock->peer_address; + result = ISC_R_SUCCESS; + } else { + result = ISC_R_NOTCONNECTED; + } + + UNLOCK(&sock->lock); + + return (result); +} + +isc_result_t +isc__socket_getsockname(isc_socket_t *sock0, isc_sockaddr_t *addressp) { + isc__socket_t *sock = (isc__socket_t *)sock0; + ISC_SOCKADDR_LEN_T len; + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addressp != NULL); + + LOCK(&sock->lock); + + if (!sock->bound) { + result = ISC_R_NOTBOUND; + goto out; + } + + result = ISC_R_SUCCESS; + + len = sizeof(addressp->type); + if (getsockname(sock->fd, &addressp->type.sa, (void *)&len) < 0) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "getsockname: %s", + strbuf); + result = ISC_R_UNEXPECTED; + goto out; + } + addressp->length = (unsigned int)len; + + out: + UNLOCK(&sock->lock); + + return (result); +} + +/* + * Run through the list of events on this socket, and cancel the ones + * queued for task "task" of type "how". "how" is a bitmask. + */ +void +isc__socket_cancel(isc_socket_t *sock0, isc_task_t *task, unsigned int how) { + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + + /* + * Quick exit if there is nothing to do. Don't even bother locking + * in this case. + */ + if (how == 0) + return; + + LOCK(&sock->lock); + + /* + * All of these do the same thing, more or less. + * Each will: + * o If the internal event is marked as "posted" try to + * remove it from the task's queue. If this fails, mark it + * as canceled instead, and let the task clean it up later. + * o For each I/O request for that task of that type, post + * its done event with status of "ISC_R_CANCELED". + * o Reset any state needed. + */ + if (((how & ISC_SOCKCANCEL_RECV) == ISC_SOCKCANCEL_RECV) + && !ISC_LIST_EMPTY(sock->recv_list)) { + isc_socketevent_t *dev; + isc_socketevent_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->recv_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_recvdone_event(sock, &dev); + } + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_SEND) == ISC_SOCKCANCEL_SEND) + && !ISC_LIST_EMPTY(sock->send_list)) { + isc_socketevent_t *dev; + isc_socketevent_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->send_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_senddone_event(sock, &dev); + } + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_ACCEPT) == ISC_SOCKCANCEL_ACCEPT) + && !ISC_LIST_EMPTY(sock->accept_list)) { + isc_socket_newconnev_t *dev; + isc_socket_newconnev_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->accept_list); + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + + ISC_LIST_UNLINK(sock->accept_list, dev, + ev_link); + + NEWCONNSOCK(dev)->references--; + free_socket((isc__socket_t **)&dev->newsocket); + + dev->result = ISC_R_CANCELED; + dev->ev_sender = sock; + isc_task_sendanddetach(¤t_task, + ISC_EVENT_PTR(&dev)); + } + + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_CONNECT) == ISC_SOCKCANCEL_CONNECT) + && !ISC_LIST_EMPTY(sock->connect_list)) { + isc_socket_connev_t *dev; + isc_socket_connev_t *next; + isc_task_t *current_task; + + INSIST(sock->connecting); + sock->connecting = 0; + + dev = ISC_LIST_HEAD(sock->connect_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_connectdone_event(sock, &dev); + } + dev = next; + } + } + + UNLOCK(&sock->lock); +} + +isc_sockettype_t +isc__socket_gettype(isc_socket_t *sock0) { + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + + return (sock->type); +} + +bool +isc__socket_isbound(isc_socket_t *sock0) { + isc__socket_t *sock = (isc__socket_t *)sock0; + bool val; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + val = ((sock->bound) ? true : false); + UNLOCK(&sock->lock); + + return (val); +} + +void +isc__socket_ipv6only(isc_socket_t *sock0, bool yes) { + isc__socket_t *sock = (isc__socket_t *)sock0; +#if defined(IPV6_V6ONLY) + int onoff = yes ? 1 : 0; +#else + UNUSED(yes); + UNUSED(sock); +#endif + + REQUIRE(VALID_SOCKET(sock)); + INSIST(!sock->dupped); + +#ifdef IPV6_V6ONLY + if (sock->pf == AF_INET6) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_V6ONLY, + (void *)&onoff, sizeof(int)) < 0) { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_V6ONLY) " + "%s: %s", sock->fd, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } + } + FIX_IPV6_RECVPKTINFO(sock); /* AIX */ +#endif +} + +static void +setdscp(isc__socket_t *sock, isc_dscp_t dscp) { +#if defined(IP_TOS) || defined(IPV6_TCLASS) + int value = dscp << 2; +#endif + + sock->dscp = dscp; + +#ifdef IP_TOS + if (sock->pf == AF_INET) { + if (setsockopt(sock->fd, IPPROTO_IP, IP_TOS, + (void *)&value, sizeof(value)) < 0) { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_TOS, %.02x) " + "%s: %s", sock->fd, value >> 2, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } + } +#endif +#ifdef IPV6_TCLASS + if (sock->pf == AF_INET6) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_TCLASS, + (void *)&value, sizeof(value)) < 0) { + char strbuf[ISC_STRERRORSIZE]; + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_TCLASS, %.02x) " + "%s: %s", sock->fd, dscp >> 2, + isc_msgcat_get(isc_msgcat, + ISC_MSGSET_GENERAL, + ISC_MSG_FAILED, + "failed"), + strbuf); + } + } +#endif +} + +void +isc__socket_dscp(isc_socket_t *sock0, isc_dscp_t dscp) { + isc__socket_t *sock = (isc__socket_t *)sock0; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(dscp < 0x40); + +#if !defined(IP_TOS) && !defined(IPV6_TCLASS) + UNUSED(dscp); +#else + if (dscp < 0) + return; + + /* The DSCP value must not be changed once it has been set. */ + if (isc_dscp_check_value != -1) + INSIST(dscp == isc_dscp_check_value); +#endif + + +#ifdef notyet + REQUIRE(!sock->dupped); +#endif + + setdscp(sock, dscp); +} + +isc_socketevent_t * +isc_socket_socketevent(isc_mem_t *mctx, void *sender, + isc_eventtype_t eventtype, isc_taskaction_t action, + void *arg) +{ + return (allocate_socketevent(mctx, sender, eventtype, action, arg)); +} + +#ifndef USE_WATCHER_THREAD +/* + * In our assumed scenario, we can simply use a single static object. + * XXX: this is not true if the application uses multiple threads with + * 'multi-context' mode. Fixing this is a future TODO item. + */ +static isc_socketwait_t swait_private; + +int +isc__socketmgr_waitevents(isc_socketmgr_t *manager0, struct timeval *tvp, + isc_socketwait_t **swaitp) +{ + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + int n; +#ifdef USE_KQUEUE + struct timespec ts, *tsp; +#endif +#ifdef USE_EPOLL + int timeout; +#endif +#ifdef USE_DEVPOLL + isc_result_t result; + int pass; + struct dvpoll dvp; +#endif + + REQUIRE(swaitp != NULL && *swaitp == NULL); + +#ifdef USE_SHARED_MANAGER + if (manager == NULL) + manager = socketmgr; +#endif + if (manager == NULL) + return (0); + +#ifdef USE_KQUEUE + if (tvp != NULL) { + ts.tv_sec = tvp->tv_sec; + ts.tv_nsec = tvp->tv_usec * 1000; + tsp = &ts; + } else + tsp = NULL; + swait_private.nevents = kevent(manager->kqueue_fd, NULL, 0, + manager->events, manager->nevents, + tsp); + n = swait_private.nevents; +#elif defined(USE_EPOLL) + if (tvp != NULL) + timeout = tvp->tv_sec * 1000 + (tvp->tv_usec + 999) / 1000; + else + timeout = -1; + swait_private.nevents = epoll_wait(manager->epoll_fd, + manager->events, + manager->nevents, timeout); + n = swait_private.nevents; +#elif defined(USE_DEVPOLL) + /* + * Re-probe every thousand calls. + */ + if (manager->calls++ > 1000U) { + result = isc_resource_getcurlimit(isc_resource_openfiles, + &manager->open_max); + if (result != ISC_R_SUCCESS) + manager->open_max = 64; + manager->calls = 0; + } + for (pass = 0; pass < 2; pass++) { + dvp.dp_fds = manager->events; + dvp.dp_nfds = manager->nevents; + if (dvp.dp_nfds >= manager->open_max) + dvp.dp_nfds = manager->open_max - 1; + if (tvp != NULL) { + dvp.dp_timeout = tvp->tv_sec * 1000 + + (tvp->tv_usec + 999) / 1000; + } else + dvp.dp_timeout = -1; + n = ioctl(manager->devpoll_fd, DP_POLL, &dvp); + if (n == -1 && errno == EINVAL) { + /* + * {OPEN_MAX} may have dropped. Look + * up the current value and try again. + */ + result = isc_resource_getcurlimit( + isc_resource_openfiles, + &manager->open_max); + if (result != ISC_R_SUCCESS) + manager->open_max = 64; + } else + break; + } + swait_private.nevents = n; +#elif defined(USE_SELECT) + memmove(manager->read_fds_copy, manager->read_fds, manager->fd_bufsize); + memmove(manager->write_fds_copy, manager->write_fds, + manager->fd_bufsize); + + swait_private.readset = manager->read_fds_copy; + swait_private.writeset = manager->write_fds_copy; + swait_private.maxfd = manager->maxfd + 1; + + n = select(swait_private.maxfd, swait_private.readset, + swait_private.writeset, NULL, tvp); +#endif + + *swaitp = &swait_private; + return (n); +} + +isc_result_t +isc__socketmgr_dispatch(isc_socketmgr_t *manager0, isc_socketwait_t *swait) { + isc__socketmgr_t *manager = (isc__socketmgr_t *)manager0; + + REQUIRE(swait == &swait_private); + +#ifdef USE_SHARED_MANAGER + if (manager == NULL) + manager = socketmgr; +#endif + if (manager == NULL) + return (ISC_R_NOTFOUND); + +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) + (void)process_fds(manager, manager->events, swait->nevents); + return (ISC_R_SUCCESS); +#elif defined(USE_SELECT) + process_fds(manager, swait->maxfd, swait->readset, swait->writeset); + return (ISC_R_SUCCESS); +#endif +} +#endif /* USE_WATCHER_THREAD */ + +void +isc__socket_setname(isc_socket_t *socket0, const char *name, void *tag) { + isc__socket_t *sock = (isc__socket_t *)socket0; + + /* + * Name 'sock'. + */ + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + strlcpy(sock->name, name, sizeof(sock->name)); + sock->tag = tag; + UNLOCK(&sock->lock); +} + +const char * +isc__socket_getname(isc_socket_t *socket0) { + isc__socket_t *sock = (isc__socket_t *)socket0; + + return (sock->name); +} + +void * +isc__socket_gettag(isc_socket_t *socket0) { + isc__socket_t *sock = (isc__socket_t *)socket0; + + return (sock->tag); +} + +isc_result_t +isc__socket_register(void) { + return (isc_socket_register(isc__socketmgr_create)); +} + +int +isc__socket_getfd(isc_socket_t *socket0) { + isc__socket_t *sock = (isc__socket_t *)socket0; + + return ((short) sock->fd); +} + +#if defined(HAVE_LIBXML2) || defined(HAVE_JSON) +static const char * +_socktype(isc_sockettype_t type) +{ + if (type == isc_sockettype_udp) + return ("udp"); + else if (type == isc_sockettype_tcp) + return ("tcp"); + else if (type == isc_sockettype_unix) + return ("unix"); + else if (type == isc_sockettype_fdwatch) + return ("fdwatch"); + else + return ("not-initialized"); +} +#endif + +#ifdef HAVE_LIBXML2 +#define TRY0(a) do { xmlrc = (a); if (xmlrc < 0) goto error; } while(0) +int +isc_socketmgr_renderxml(isc_socketmgr_t *mgr0, xmlTextWriterPtr writer) { + isc__socketmgr_t *mgr = (isc__socketmgr_t *)mgr0; + isc__socket_t *sock = NULL; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + isc_sockaddr_t addr; + ISC_SOCKADDR_LEN_T len; + int xmlrc; + + LOCK(&mgr->lock); + +#ifdef USE_SHARED_MANAGER + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "references")); + TRY0(xmlTextWriterWriteFormatString(writer, "%d", mgr->refs)); + TRY0(xmlTextWriterEndElement(writer)); +#endif /* USE_SHARED_MANAGER */ + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "sockets")); + sock = ISC_LIST_HEAD(mgr->socklist); + while (sock != NULL) { + LOCK(&sock->lock); + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "socket")); + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "id")); + TRY0(xmlTextWriterWriteFormatString(writer, "%p", sock)); + TRY0(xmlTextWriterEndElement(writer)); + + if (sock->name[0] != 0) { + TRY0(xmlTextWriterStartElement(writer, + ISC_XMLCHAR "name")); + TRY0(xmlTextWriterWriteFormatString(writer, "%s", + sock->name)); + TRY0(xmlTextWriterEndElement(writer)); /* name */ + } + + TRY0(xmlTextWriterStartElement(writer, + ISC_XMLCHAR "references")); + TRY0(xmlTextWriterWriteFormatString(writer, "%d", + sock->references)); + TRY0(xmlTextWriterEndElement(writer)); + + TRY0(xmlTextWriterWriteElement(writer, ISC_XMLCHAR "type", + ISC_XMLCHAR _socktype(sock->type))); + + if (sock->connected) { + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "peer-address", + ISC_XMLCHAR peerbuf)); + } + + len = sizeof(addr); + if (getsockname(sock->fd, &addr.type.sa, (void *)&len) == 0) { + isc_sockaddr_format(&addr, peerbuf, sizeof(peerbuf)); + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "local-address", + ISC_XMLCHAR peerbuf)); + } + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "states")); + if (sock->pending_recv) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "pending-receive")); + if (sock->pending_send) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "pending-send")); + if (sock->pending_accept) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "pending_accept")); + if (sock->listener) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "listener")); + if (sock->connected) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "connected")); + if (sock->connecting) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "connecting")); + if (sock->bound) + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "bound")); + + TRY0(xmlTextWriterEndElement(writer)); /* states */ + + TRY0(xmlTextWriterEndElement(writer)); /* socket */ + + UNLOCK(&sock->lock); + sock = ISC_LIST_NEXT(sock, link); + } + TRY0(xmlTextWriterEndElement(writer)); /* sockets */ + + error: + if (sock != NULL) + UNLOCK(&sock->lock); + + UNLOCK(&mgr->lock); + + return (xmlrc); +} +#endif /* HAVE_LIBXML2 */ + +#ifdef HAVE_JSON +#define CHECKMEM(m) do { \ + if (m == NULL) { \ + result = ISC_R_NOMEMORY;\ + goto error;\ + } \ +} while(0) + +isc_result_t +isc_socketmgr_renderjson(isc_socketmgr_t *mgr0, json_object *stats) { + isc_result_t result = ISC_R_SUCCESS; + isc__socketmgr_t *mgr = (isc__socketmgr_t *)mgr0; + isc__socket_t *sock = NULL; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + isc_sockaddr_t addr; + ISC_SOCKADDR_LEN_T len; + json_object *obj, *array = json_object_new_array(); + + CHECKMEM(array); + + LOCK(&mgr->lock); + +#ifdef USE_SHARED_MANAGER + obj = json_object_new_int(mgr->refs); + CHECKMEM(obj); + json_object_object_add(stats, "references", obj); +#endif /* USE_SHARED_MANAGER */ + + sock = ISC_LIST_HEAD(mgr->socklist); + while (sock != NULL) { + json_object *states, *entry = json_object_new_object(); + char buf[255]; + + CHECKMEM(entry); + json_object_array_add(array, entry); + + LOCK(&sock->lock); + + snprintf(buf, sizeof(buf), "%p", sock); + obj = json_object_new_string(buf); + CHECKMEM(obj); + json_object_object_add(entry, "id", obj); + + if (sock->name[0] != 0) { + obj = json_object_new_string(sock->name); + CHECKMEM(obj); + json_object_object_add(entry, "name", obj); + } + + obj = json_object_new_int(sock->references); + CHECKMEM(obj); + json_object_object_add(entry, "references", obj); + + obj = json_object_new_string(_socktype(sock->type)); + CHECKMEM(obj); + json_object_object_add(entry, "type", obj); + + if (sock->connected) { + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + obj = json_object_new_string(peerbuf); + CHECKMEM(obj); + json_object_object_add(entry, "peer-address", obj); + } + + len = sizeof(addr); + if (getsockname(sock->fd, &addr.type.sa, (void *)&len) == 0) { + isc_sockaddr_format(&addr, peerbuf, sizeof(peerbuf)); + obj = json_object_new_string(peerbuf); + CHECKMEM(obj); + json_object_object_add(entry, "local-address", obj); + } + + states = json_object_new_array(); + CHECKMEM(states); + json_object_object_add(entry, "states", states); + + if (sock->pending_recv) { + obj = json_object_new_string("pending-receive"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->pending_send) { + obj = json_object_new_string("pending-send"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->pending_accept) { + obj = json_object_new_string("pending-accept"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->listener) { + obj = json_object_new_string("listener"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->connected) { + obj = json_object_new_string("connected"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->connecting) { + obj = json_object_new_string("connecting"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->bound) { + obj = json_object_new_string("bound"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + UNLOCK(&sock->lock); + sock = ISC_LIST_NEXT(sock, link); + } + + json_object_object_add(stats, "sockets", array); + array = NULL; + result = ISC_R_SUCCESS; + + error: + if (array != NULL) + json_object_put(array); + + if (sock != NULL) + UNLOCK(&sock->lock); + + UNLOCK(&mgr->lock); + + return (result); +} +#endif /* HAVE_JSON */ + +#include "../socket_api.c" diff --git a/lib/isc/unix/socket_p.h b/lib/isc/unix/socket_p.h new file mode 100644 index 0000000..fb4fdb8 --- /dev/null +++ b/lib/isc/unix/socket_p.h @@ -0,0 +1,26 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#ifndef ISC_SOCKET_P_H +#define ISC_SOCKET_P_H + +/*! \file */ + +#ifdef ISC_PLATFORM_NEEDSYSSELECTH +#include <sys/select.h> +#endif + +typedef struct isc_socketwait isc_socketwait_t; +int isc__socketmgr_waitevents(isc_socketmgr_t *, struct timeval *, + isc_socketwait_t **); +isc_result_t isc__socketmgr_dispatch(isc_socketmgr_t *, isc_socketwait_t *); +#endif /* ISC_SOCKET_P_H */ diff --git a/lib/isc/unix/stdio.c b/lib/isc/unix/stdio.c new file mode 100644 index 0000000..e60fa65 --- /dev/null +++ b/lib/isc/unix/stdio.c @@ -0,0 +1,151 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +#include <config.h> + +#include <errno.h> +#include <unistd.h> + +#include <isc/stdio.h> +#include <isc/stat.h> +#include <isc/util.h> + +#include "errno2result.h" + +isc_result_t +isc_stdio_open(const char *filename, const char *mode, FILE **fp) { + FILE *f; + + f = fopen(filename, mode); + if (f == NULL) + return (isc__errno2result(errno)); + *fp = f; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_stdio_close(FILE *f) { + int r; + + r = fclose(f); + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + +isc_result_t +isc_stdio_seek(FILE *f, off_t offset, int whence) { + int r; + +#ifdef HAVE_FSEEKO + r = fseeko(f, offset, whence); +#else + r = fseek(f, offset, whence); +#endif + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + +isc_result_t +isc_stdio_tell(FILE *f, off_t *offsetp) { + off_t r; + + REQUIRE(offsetp != NULL); + +#ifdef HAVE_FTELLO + r = ftello(f); +#else + r = ftell(f); +#endif + if (r >= 0) { + *offsetp = r; + return (ISC_R_SUCCESS); + } else + return (isc__errno2result(errno)); +} + +isc_result_t +isc_stdio_read(void *ptr, size_t size, size_t nmemb, FILE *f, size_t *nret) { + isc_result_t result = ISC_R_SUCCESS; + size_t r; + + clearerr(f); + r = fread(ptr, size, nmemb, f); + if (r != nmemb) { + if (feof(f)) + result = ISC_R_EOF; + else + result = isc__errno2result(errno); + } + if (nret != NULL) + *nret = r; + return (result); +} + +isc_result_t +isc_stdio_write(const void *ptr, size_t size, size_t nmemb, FILE *f, + size_t *nret) +{ + isc_result_t result = ISC_R_SUCCESS; + size_t r; + + clearerr(f); + r = fwrite(ptr, size, nmemb, f); + if (r != nmemb) + result = isc__errno2result(errno); + if (nret != NULL) + *nret = r; + return (result); +} + +isc_result_t +isc_stdio_flush(FILE *f) { + int r; + + r = fflush(f); + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + +/* + * OpenBSD has deprecated ENOTSUP in favor of EOPNOTSUPP. + */ +#if defined(EOPNOTSUPP) && !defined(ENOTSUP) +#define ENOTSUP EOPNOTSUPP +#endif + +isc_result_t +isc_stdio_sync(FILE *f) { + struct stat buf; + int r; + + if (fstat(fileno(f), &buf) != 0) + return (isc__errno2result(errno)); + + /* + * Only call fsync() on regular files. + */ + if ((buf.st_mode & S_IFMT) != S_IFREG) + return (ISC_R_SUCCESS); + + r = fsync(fileno(f)); + if (r == 0) + return (ISC_R_SUCCESS); + else + return (isc__errno2result(errno)); +} + diff --git a/lib/isc/unix/stdtime.c b/lib/isc/unix/stdtime.c new file mode 100644 index 0000000..989bde4 --- /dev/null +++ b/lib/isc/unix/stdtime.c @@ -0,0 +1,80 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <stdbool.h> +#include <stddef.h> /* NULL */ +#include <stdlib.h> /* NULL */ +#include <syslog.h> + +#include <sys/time.h> + +#include <isc/stdtime.h> +#include <isc/util.h> + +#ifndef ISC_FIX_TV_USEC +#define ISC_FIX_TV_USEC 1 +#endif + +#define US_PER_S 1000000 + +#if ISC_FIX_TV_USEC +static inline void +fix_tv_usec(struct timeval *tv) { + bool fixed = false; + + if (tv->tv_usec < 0) { + fixed = true; + do { + tv->tv_sec -= 1; + tv->tv_usec += US_PER_S; + } while (tv->tv_usec < 0); + } else if (tv->tv_usec >= US_PER_S) { + fixed = true; + do { + tv->tv_sec += 1; + tv->tv_usec -= US_PER_S; + } while (tv->tv_usec >=US_PER_S); + } + /* + * Call syslog directly as we are called from the logging functions. + */ + if (fixed) + (void)syslog(LOG_ERR, "gettimeofday returned bad tv_usec: corrected"); +} +#endif + +void +isc_stdtime_get(isc_stdtime_t *t) { + struct timeval tv; + + /* + * Set 't' to the number of seconds since 00:00:00 UTC, January 1, + * 1970. + */ + + REQUIRE(t != NULL); + + RUNTIME_CHECK(gettimeofday(&tv, NULL) != -1); + +#if ISC_FIX_TV_USEC + fix_tv_usec(&tv); + INSIST(tv.tv_usec >= 0); +#else + INSIST(tv.tv_usec >= 0 && tv.tv_usec < US_PER_S); +#endif + + *t = (unsigned int)tv.tv_sec; +} diff --git a/lib/isc/unix/strerror.c b/lib/isc/unix/strerror.c new file mode 100644 index 0000000..e4fb793 --- /dev/null +++ b/lib/isc/unix/strerror.c @@ -0,0 +1,67 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <stdio.h> +#include <string.h> + +#include <isc/mutex.h> +#include <isc/once.h> +#include <isc/print.h> +#include <isc/strerror.h> +#include <isc/util.h> + +#ifdef HAVE_STRERROR +/*% + * We need to do this this way for profiled locks. + */ +static isc_mutex_t isc_strerror_lock; +static void init_lock(void) { + RUNTIME_CHECK(isc_mutex_init(&isc_strerror_lock) == ISC_R_SUCCESS); +} +#else +extern const char * const sys_errlist[]; +extern const int sys_nerr; +#endif + +void +isc__strerror(int num, char *buf, size_t size) { +#ifdef HAVE_STRERROR + char *msg; + unsigned int unum = (unsigned int)num; + static isc_once_t once = ISC_ONCE_INIT; + + REQUIRE(buf != NULL); + + RUNTIME_CHECK(isc_once_do(&once, init_lock) == ISC_R_SUCCESS); + + LOCK(&isc_strerror_lock); + msg = strerror(num); + if (msg != NULL) + snprintf(buf, size, "%s", msg); + else + snprintf(buf, size, "Unknown error: %u", unum); + UNLOCK(&isc_strerror_lock); +#else + unsigned int unum = (unsigned int)num; + + REQUIRE(buf != NULL); + + if (num >= 0 && num < sys_nerr) + snprintf(buf, size, "%s", sys_errlist[num]); + else + snprintf(buf, size, "Unknown error: %u", unum); +#endif +} diff --git a/lib/isc/unix/syslog.c b/lib/isc/unix/syslog.c new file mode 100644 index 0000000..4ce3272 --- /dev/null +++ b/lib/isc/unix/syslog.c @@ -0,0 +1,77 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <stdlib.h> +#include <syslog.h> + +#include <isc/result.h> +#include <isc/string.h> +#include <isc/syslog.h> +#include <isc/util.h> + +static struct dsn_c_pvt_sfnt { + int val; + const char *strval; +} facilities[] = { + { LOG_KERN, "kern" }, + { LOG_USER, "user" }, + { LOG_MAIL, "mail" }, + { LOG_DAEMON, "daemon" }, + { LOG_AUTH, "auth" }, + { LOG_SYSLOG, "syslog" }, + { LOG_LPR, "lpr" }, +#ifdef LOG_NEWS + { LOG_NEWS, "news" }, +#endif +#ifdef LOG_UUCP + { LOG_UUCP, "uucp" }, +#endif +#ifdef LOG_CRON + { LOG_CRON, "cron" }, +#endif +#ifdef LOG_AUTHPRIV + { LOG_AUTHPRIV, "authpriv" }, +#endif +#ifdef LOG_FTP + { LOG_FTP, "ftp" }, +#endif + { LOG_LOCAL0, "local0"}, + { LOG_LOCAL1, "local1"}, + { LOG_LOCAL2, "local2"}, + { LOG_LOCAL3, "local3"}, + { LOG_LOCAL4, "local4"}, + { LOG_LOCAL5, "local5"}, + { LOG_LOCAL6, "local6"}, + { LOG_LOCAL7, "local7"}, + { 0, NULL } +}; + +isc_result_t +isc_syslog_facilityfromstring(const char *str, int *facilityp) { + int i; + + REQUIRE(str != NULL); + REQUIRE(facilityp != NULL); + + for (i = 0; facilities[i].strval != NULL; i++) { + if (strcasecmp(facilities[i].strval, str) == 0) { + *facilityp = facilities[i].val; + return (ISC_R_SUCCESS); + } + } + return (ISC_R_NOTFOUND); + +} diff --git a/lib/isc/unix/time.c b/lib/isc/unix/time.c new file mode 100644 index 0000000..8edc9df --- /dev/null +++ b/lib/isc/unix/time.c @@ -0,0 +1,500 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + + +/*! \file */ + +#include <config.h> + +#include <errno.h> +#include <limits.h> +#include <inttypes.h> +#include <stdbool.h> +#include <stdlib.h> +#include <syslog.h> +#include <time.h> + +#include <sys/time.h> /* Required for struct timeval on some platforms. */ + +#include <isc/log.h> +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/strerror.h> +#include <isc/string.h> +#include <isc/time.h> +#include <isc/tm.h> +#include <isc/util.h> + +#define NS_PER_S 1000000000 /*%< Nanoseconds per second. */ +#define NS_PER_US 1000 /*%< Nanoseconds per microsecond. */ +#define NS_PER_MS 1000000 /*%< Nanoseconds per millisecond. */ +#define US_PER_S 1000000 /*%< Microseconds per second. */ + +/* + * All of the INSIST()s checks of nanoseconds < NS_PER_S are for + * consistency checking of the type. In lieu of magic numbers, it + * is the best we've got. The check is only performed on functions which + * need an initialized type. + */ + +#ifndef ISC_FIX_TV_USEC +#define ISC_FIX_TV_USEC 1 +#endif + +/*% + *** Intervals + ***/ + +static const isc_interval_t zero_interval = { 0, 0 }; +const isc_interval_t * const isc_interval_zero = &zero_interval; + +#if ISC_FIX_TV_USEC +static inline void +fix_tv_usec(struct timeval *tv) { + bool fixed = false; + + if (tv->tv_usec < 0) { + fixed = true; + do { + tv->tv_sec -= 1; + tv->tv_usec += US_PER_S; + } while (tv->tv_usec < 0); + } else if (tv->tv_usec >= US_PER_S) { + fixed = true; + do { + tv->tv_sec += 1; + tv->tv_usec -= US_PER_S; + } while (tv->tv_usec >=US_PER_S); + } + /* + * Call syslog directly as was are called from the logging functions. + */ + if (fixed) + (void)syslog(LOG_ERR, "gettimeofday returned bad tv_usec: corrected"); +} +#endif + +void +isc_interval_set(isc_interval_t *i, + unsigned int seconds, unsigned int nanoseconds) +{ + REQUIRE(i != NULL); + REQUIRE(nanoseconds < NS_PER_S); + + i->seconds = seconds; + i->nanoseconds = nanoseconds; +} + +bool +isc_interval_iszero(const isc_interval_t *i) { + REQUIRE(i != NULL); + INSIST(i->nanoseconds < NS_PER_S); + + if (i->seconds == 0 && i->nanoseconds == 0) + return (true); + + return (false); +} + + +/*** + *** Absolute Times + ***/ + +static const isc_time_t epoch = { 0, 0 }; +const isc_time_t * const isc_time_epoch = &epoch; + +void +isc_time_set(isc_time_t *t, unsigned int seconds, unsigned int nanoseconds) { + REQUIRE(t != NULL); + REQUIRE(nanoseconds < NS_PER_S); + + t->seconds = seconds; + t->nanoseconds = nanoseconds; +} + +void +isc_time_settoepoch(isc_time_t *t) { + REQUIRE(t != NULL); + + t->seconds = 0; + t->nanoseconds = 0; +} + +bool +isc_time_isepoch(const isc_time_t *t) { + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + if (t->seconds == 0 && t->nanoseconds == 0) + return (true); + + return (false); +} + + +isc_result_t +isc_time_now(isc_time_t *t) { + struct timeval tv; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(t != NULL); + + if (gettimeofday(&tv, NULL) == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "%s", strbuf); + return (ISC_R_UNEXPECTED); + } + + /* + * Does POSIX guarantee the signedness of tv_sec and tv_usec? If not, + * then this test will generate warnings for platforms on which it is + * unsigned. In any event, the chances of any of these problems + * happening are pretty much zero, but since the libisc library ensures + * certain things to be true ... + */ +#if ISC_FIX_TV_USEC + fix_tv_usec(&tv); + if (tv.tv_sec < 0) + return (ISC_R_UNEXPECTED); +#else + if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= US_PER_S) + return (ISC_R_UNEXPECTED); +#endif + + /* + * Ensure the tv_sec value fits in t->seconds. + */ + if (sizeof(tv.tv_sec) > sizeof(t->seconds) && + ((tv.tv_sec | (unsigned int)-1) ^ (unsigned int)-1) != 0U) + return (ISC_R_RANGE); + + t->seconds = tv.tv_sec; + t->nanoseconds = tv.tv_usec * NS_PER_US; + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_time_nowplusinterval(isc_time_t *t, const isc_interval_t *i) { + struct timeval tv; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(t != NULL); + REQUIRE(i != NULL); + INSIST(i->nanoseconds < NS_PER_S); + + if (gettimeofday(&tv, NULL) == -1) { + isc__strerror(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "%s", strbuf); + return (ISC_R_UNEXPECTED); + } + + /* + * Does POSIX guarantee the signedness of tv_sec and tv_usec? If not, + * then this test will generate warnings for platforms on which it is + * unsigned. In any event, the chances of any of these problems + * happening are pretty much zero, but since the libisc library ensures + * certain things to be true ... + */ +#if ISC_FIX_TV_USEC + fix_tv_usec(&tv); + if (tv.tv_sec < 0) + return (ISC_R_UNEXPECTED); +#else + if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= US_PER_S) + return (ISC_R_UNEXPECTED); +#endif + + /* + * Ensure the resulting seconds value fits in the size of an + * unsigned int. (It is written this way as a slight optimization; + * note that even if both values == INT_MAX, then when added + * and getting another 1 added below the result is UINT_MAX.) + */ + if ((tv.tv_sec > INT_MAX || i->seconds > INT_MAX) && + ((long long)tv.tv_sec + i->seconds > UINT_MAX)) + return (ISC_R_RANGE); + + t->seconds = tv.tv_sec + i->seconds; + t->nanoseconds = tv.tv_usec * NS_PER_US + i->nanoseconds; + if (t->nanoseconds >= NS_PER_S) { + t->seconds++; + t->nanoseconds -= NS_PER_S; + } + + return (ISC_R_SUCCESS); +} + +int +isc_time_compare(const isc_time_t *t1, const isc_time_t *t2) { + REQUIRE(t1 != NULL && t2 != NULL); + INSIST(t1->nanoseconds < NS_PER_S && t2->nanoseconds < NS_PER_S); + + if (t1->seconds < t2->seconds) + return (-1); + if (t1->seconds > t2->seconds) + return (1); + if (t1->nanoseconds < t2->nanoseconds) + return (-1); + if (t1->nanoseconds > t2->nanoseconds) + return (1); + return (0); +} + +isc_result_t +isc_time_add(const isc_time_t *t, const isc_interval_t *i, isc_time_t *result) +{ + REQUIRE(t != NULL && i != NULL && result != NULL); + INSIST(t->nanoseconds < NS_PER_S && i->nanoseconds < NS_PER_S); + + /* + * Ensure the resulting seconds value fits in the size of an + * unsigned int. (It is written this way as a slight optimization; + * note that even if both values == INT_MAX, then when added + * and getting another 1 added below the result is UINT_MAX.) + */ + if ((t->seconds > INT_MAX || i->seconds > INT_MAX) && + ((long long)t->seconds + i->seconds > UINT_MAX)) + return (ISC_R_RANGE); + + result->seconds = t->seconds + i->seconds; + result->nanoseconds = t->nanoseconds + i->nanoseconds; + if (result->nanoseconds >= NS_PER_S) { + result->seconds++; + result->nanoseconds -= NS_PER_S; + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_time_subtract(const isc_time_t *t, const isc_interval_t *i, + isc_time_t *result) +{ + REQUIRE(t != NULL && i != NULL && result != NULL); + INSIST(t->nanoseconds < NS_PER_S && i->nanoseconds < NS_PER_S); + + if ((unsigned int)t->seconds < i->seconds || + ((unsigned int)t->seconds == i->seconds && + t->nanoseconds < i->nanoseconds)) + return (ISC_R_RANGE); + + result->seconds = t->seconds - i->seconds; + if (t->nanoseconds >= i->nanoseconds) + result->nanoseconds = t->nanoseconds - i->nanoseconds; + else { + result->nanoseconds = NS_PER_S - i->nanoseconds + + t->nanoseconds; + result->seconds--; + } + + return (ISC_R_SUCCESS); +} + +uint64_t +isc_time_microdiff(const isc_time_t *t1, const isc_time_t *t2) { + uint64_t i1, i2, i3; + + REQUIRE(t1 != NULL && t2 != NULL); + INSIST(t1->nanoseconds < NS_PER_S && t2->nanoseconds < NS_PER_S); + + i1 = (uint64_t)t1->seconds * NS_PER_S + t1->nanoseconds; + i2 = (uint64_t)t2->seconds * NS_PER_S + t2->nanoseconds; + + if (i1 <= i2) + return (0); + + i3 = i1 - i2; + + /* + * Convert to microseconds. + */ + i3 /= NS_PER_US; + + return (i3); +} + +uint32_t +isc_time_seconds(const isc_time_t *t) { + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + return ((uint32_t)t->seconds); +} + +isc_result_t +isc_time_secondsastimet(const isc_time_t *t, time_t *secondsp) { + time_t seconds; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + /* + * Ensure that the number of seconds represented by t->seconds + * can be represented by a time_t. Since t->seconds is an unsigned + * int and since time_t is mostly opaque, this is trickier than + * it seems. (This standardized opaqueness of time_t is *very* + * frustrating; time_t is not even limited to being an integral + * type.) + * + * The mission, then, is to avoid generating any kind of warning + * about "signed versus unsigned" while trying to determine if the + * the unsigned int t->seconds is out range for tv_sec, which is + * pretty much only true if time_t is a signed integer of the same + * size as the return value of isc_time_seconds. + * + * If the paradox in the if clause below is true, t->seconds is out + * of range for time_t. + */ + seconds = (time_t)t->seconds; + + INSIST(sizeof(unsigned int) == sizeof(uint32_t)); + INSIST(sizeof(time_t) >= sizeof(uint32_t)); + + if (t->seconds > (~0U>>1) && seconds <= (time_t)(~0U>>1)) + return (ISC_R_RANGE); + + *secondsp = seconds; + + return (ISC_R_SUCCESS); +} + +uint32_t +isc_time_nanoseconds(const isc_time_t *t) { + REQUIRE(t != NULL); + + ENSURE(t->nanoseconds < NS_PER_S); + + return ((uint32_t)t->nanoseconds); +} + +void +isc_time_formattimestamp(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; +#ifdef ISC_PLATFORM_USETHREADS + struct tm tm; +#endif + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t) t->seconds; +#ifdef ISC_PLATFORM_USETHREADS + flen = strftime(buf, len, "%d-%b-%Y %X", localtime_r(&now, &tm)); +#else + flen = strftime(buf, len, "%d-%b-%Y %X", localtime(&now)); +#endif + INSIST(flen < len); + if (flen != 0) + snprintf(buf + flen, len - flen, + ".%03u", t->nanoseconds / NS_PER_MS); + else { + strlcpy(buf, "99-Bad-9999 99:99:99.999", len); + } +} + +void +isc_time_formathttptimestamp(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; +#ifdef ISC_PLATFORM_USETHREADS + struct tm tm; +#endif + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + /* + * 5 spaces, 1 comma, 3 GMT, 2 %d, 4 %Y, 8 %H:%M:%S, 3+ %a, 3+ %b (29+) + */ + now = (time_t)t->seconds; +#ifdef ISC_PLATFORM_USETHREADS + flen = strftime(buf, len, "%a, %d %b %Y %H:%M:%S GMT", + gmtime_r(&now, &tm)); +#else + flen = strftime(buf, len, "%a, %d %b %Y %H:%M:%S GMT", gmtime(&now)); +#endif + INSIST(flen < len); +} + +isc_result_t +isc_time_parsehttptimestamp(char *buf, isc_time_t *t) { + struct tm t_tm; + time_t when; + char *p; + + REQUIRE(buf != NULL); + REQUIRE(t != NULL); + + p = isc_tm_strptime(buf, "%a, %d %b %Y %H:%M:%S", &t_tm); + if (p == NULL) + return (ISC_R_UNEXPECTED); + when = isc_tm_timegm(&t_tm); + if (when == -1) + return (ISC_R_UNEXPECTED); + isc_time_set(t, when, 0); + return (ISC_R_SUCCESS); +} + +void +isc_time_formatISO8601(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; +#ifdef ISC_PLATFORM_USETHREADS + struct tm tm; +#endif + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; +#ifdef ISC_PLATFORM_USETHREADS + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime_r(&now, &tm)); +#else + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime(&now)); +#endif + INSIST(flen < len); +} + +void +isc_time_formatISO8601ms(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; +#ifdef ISC_PLATFORM_USETHREADS + struct tm tm; +#endif + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; +#ifdef ISC_PLATFORM_USETHREADS + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime_r(&now, &tm)); +#else + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime(&now)); +#endif + INSIST(flen < len); + if (flen > 0U && len - flen >= 5) { + flen -= 1; /* rewind one character (Z) */ + snprintf(buf + flen, len - flen, ".%03uZ", + t->nanoseconds / NS_PER_MS); + } +} |