2
0
mirror of https://gitlab.isc.org/isc-projects/bind9 synced 2025-08-27 04:28:33 +00:00
bind/lib/isc/app.c
Ondřej Surý b540722bc3 Refactor taskmgr to run on top of netmgr
This commit changes the taskmgr to run the individual tasks on the
netmgr internal workers.  While an effort has been put into keeping the
taskmgr interface intact, couple of changes have been made:

 * The taskmgr has no concept of universal privileged mode - rather the
   tasks are either privileged or unprivileged (normal).  The privileged
   tasks are run as a first thing when the netmgr is unpaused.  There
   are now four different queues in in the netmgr:

   1. priority queue - netievent on the priority queue are run even when
      the taskmgr enter exclusive mode and netmgr is paused.  This is
      needed to properly start listening on the interfaces, free
      resources and resume.

   2. privileged task queue - only privileged tasks are queued here and
      this is the first queue that gets processed when network manager
      is unpaused using isc_nm_resume().  All netmgr workers need to
      clean the privileged task queue before they all proceed normal
      operation.  Both task queues are processed when the workers are
      finished.

   3. task queue - only (traditional) task are scheduled here and this
      queue along with privileged task queues are process when the
      netmgr workers are finishing.  This is needed to process the task
      shutdown events.

   4. normal queue - this is the queue with netmgr events, e.g. reading,
      sending, callbacks and pretty much everything is processed here.

 * The isc_taskmgr_create() now requires initialized netmgr (isc_nm_t)
   object.

 * The isc_nm_destroy() function now waits for indefinite time, but it
   will print out the active objects when in tracing mode
   (-DNETMGR_TRACE=1 and -DNETMGR_TRACE_VERBOSE=1), the netmgr has been
   made a little bit more asynchronous and it might take longer time to
   shutdown all the active networking connections.

 * Previously, the isc_nm_stoplistening() was a synchronous operation.
   This has been changed and the isc_nm_stoplistening() just schedules
   the child sockets to stop listening and exits.  This was needed to
   prevent a deadlock as the the (traditional) tasks are now executed on
   the netmgr threads.

 * The socket selection logic in isc__nm_udp_send() was flawed, but
   fortunatelly, it was broken, so we never hit the problem where we
   created uvreq_t on a socket from nmhandle_t, but then a different
   socket could be picked up and then we were trying to run the send
   callback on a socket that had different threadid than currently
   running.
2021-04-20 23:22:28 +02:00

542 lines
12 KiB
C

/*
* 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 https://mozilla.org/MPL/2.0/.
*
* See the COPYRIGHT file distributed with this work for additional
* information regarding copyright ownership.
*/
/*! \file */
#include <errno.h>
#include <stdbool.h>
#include <stddef.h>
#include <stdlib.h>
#include <sys/types.h>
#include <unistd.h>
#ifndef WIN32
#include <inttypes.h>
#include <signal.h>
#include <sys/time.h>
#endif /* WIN32 */
#include <isc/app.h>
#include <isc/atomic.h>
#include <isc/condition.h>
#include <isc/event.h>
#include <isc/mem.h>
#include <isc/mutex.h>
#include <isc/platform.h>
#include <isc/strerr.h>
#include <isc/string.h>
#include <isc/task.h>
#include <isc/thread.h>
#include <isc/time.h>
#include <isc/util.h>
#ifdef WIN32
#include <process.h>
#else /* WIN32 */
#include <pthread.h>
#endif /* WIN32 */
/*%
* For BIND9 internal applications built with threads, we use a single app
* context and let multiple worker, I/O, timer threads do actual jobs.
*/
static isc_thread_t blockedthread;
static atomic_bool is_running;
#ifdef WIN32
/*
* We need to remember which thread is the main thread...
*/
static isc_thread_t main_thread;
#endif /* ifdef WIN32 */
/*
* The application context of this module.
*/
#define APPCTX_MAGIC ISC_MAGIC('A', 'p', 'c', 'x')
#define VALID_APPCTX(c) ISC_MAGIC_VALID(c, APPCTX_MAGIC)
#ifdef WIN32
#define NUM_EVENTS 2
enum { RELOAD_EVENT, SHUTDOWN_EVENT };
#endif /* WIN32 */
struct isc_appctx {
unsigned int magic;
isc_mem_t *mctx;
isc_mutex_t lock;
isc_eventlist_t on_run;
atomic_bool shutdown_requested;
atomic_bool running;
atomic_bool want_shutdown;
atomic_bool want_reload;
atomic_bool blocked;
#ifdef WIN32
HANDLE hEvents[NUM_EVENTS];
#else /* WIN32 */
isc_mutex_t readylock;
isc_condition_t ready;
#endif /* WIN32 */
};
static isc_appctx_t isc_g_appctx;
#ifndef WIN32
static void
handle_signal(int sig, void (*handler)(int)) {
struct sigaction sa;
memset(&sa, 0, sizeof(sa));
sa.sa_handler = handler;
if (sigfillset(&sa.sa_mask) != 0 || sigaction(sig, &sa, NULL) < 0) {
char strbuf[ISC_STRERRORSIZE];
strerror_r(errno, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"handle_signal() %d setup: %s", sig, strbuf);
}
}
#endif /* ifndef WIN32 */
isc_result_t
isc_app_ctxstart(isc_appctx_t *ctx) {
REQUIRE(VALID_APPCTX(ctx));
/*
* Start an ISC library application.
*/
isc_mutex_init(&ctx->lock);
#ifndef WIN32
isc_mutex_init(&ctx->readylock);
isc_condition_init(&ctx->ready);
#endif /* WIN32 */
ISC_LIST_INIT(ctx->on_run);
atomic_init(&ctx->shutdown_requested, false);
atomic_init(&ctx->running, false);
atomic_init(&ctx->want_shutdown, false);
atomic_init(&ctx->want_reload, false);
atomic_init(&ctx->blocked, false);
#ifdef WIN32
main_thread = GetCurrentThread();
/* Create the reload event in a non-signaled state */
ctx->hEvents[RELOAD_EVENT] = CreateEvent(NULL, FALSE, FALSE, NULL);
/* Create the shutdown event in a non-signaled state */
ctx->hEvents[SHUTDOWN_EVENT] = CreateEvent(NULL, FALSE, FALSE, NULL);
#else /* WIN32 */
int presult;
sigset_t sset;
char strbuf[ISC_STRERRORSIZE];
/*
* Always ignore SIGPIPE.
*/
handle_signal(SIGPIPE, SIG_IGN);
handle_signal(SIGHUP, SIG_DFL);
handle_signal(SIGTERM, SIG_DFL);
handle_signal(SIGINT, SIG_DFL);
/*
* 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)
{
strerror_r(errno, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"isc_app_start() sigsetops: %s", strbuf);
}
presult = pthread_sigmask(SIG_BLOCK, &sset, NULL);
if (presult != 0) {
strerror_r(presult, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"isc_app_start() pthread_sigmask: %s", strbuf);
}
#endif /* WIN32 */
return (ISC_R_SUCCESS);
}
isc_result_t
isc_app_start(void) {
isc_g_appctx.magic = APPCTX_MAGIC;
isc_g_appctx.mctx = NULL;
/* The remaining members will be initialized in ctxstart() */
return (isc_app_ctxstart(&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_g_appctx, mctx, task, action, arg));
}
isc_result_t
isc_app_ctxonrun(isc_appctx_t *ctx, isc_mem_t *mctx, isc_task_t *task,
isc_taskaction_t action, void *arg) {
isc_event_t *event;
isc_task_t *cloned_task = NULL;
if (atomic_load_acquire(&ctx->running)) {
return (ISC_R_ALREADYRUNNING);
}
/*
* 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));
LOCK(&ctx->lock);
ISC_LINK_INIT(event, ev_link);
ISC_LIST_APPEND(ctx->on_run, event, ev_link);
UNLOCK(&ctx->lock);
return (ISC_R_SUCCESS);
}
isc_result_t
isc_app_ctxrun(isc_appctx_t *ctx) {
isc_event_t *event, *next_event;
isc_task_t *task;
REQUIRE(VALID_APPCTX(ctx));
#ifdef WIN32
REQUIRE(main_thread == GetCurrentThread());
#endif /* ifdef WIN32 */
if (atomic_compare_exchange_strong_acq_rel(&ctx->running,
&(bool){ false }, true))
{
/*
* Post any on-run events (in FIFO order).
*/
LOCK(&ctx->lock);
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 WIN32
/*
* BIND9 internal tools using multiple contexts do not
* rely on signal. */
if (isc_bind9 && ctx != &isc_g_appctx) {
return (ISC_R_SUCCESS);
}
#endif /* WIN32 */
/*
* 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 (!atomic_load_acquire(&ctx->want_shutdown)) {
#ifdef WIN32
DWORD dwWaitResult = WaitForMultipleObjects(
NUM_EVENTS, ctx->hEvents, FALSE, INFINITE);
/* See why we returned */
if (WaitSucceeded(dwWaitResult, NUM_EVENTS)) {
/*
* The return was due to one of the events
* being signaled
*/
switch (WaitSucceededIndex(dwWaitResult)) {
case RELOAD_EVENT:
atomic_store_release(&ctx->want_reload, true);
break;
case SHUTDOWN_EVENT:
atomic_store_release(&ctx->want_shutdown, true);
break;
}
}
#else /* WIN32 */
if (isc_bind9) {
sigset_t sset;
int sig;
/*
* 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)
{
char strbuf[ISC_STRERRORSIZE];
strerror_r(errno, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"isc_app_run() sigsetops: %s",
strbuf);
}
if (sigwait(&sset, &sig) == 0) {
switch (sig) {
case SIGINT:
case SIGTERM:
atomic_store_release(
&ctx->want_shutdown, true);
break;
case SIGHUP:
atomic_store_release(&ctx->want_reload,
true);
break;
default:
INSIST(0);
ISC_UNREACHABLE();
}
}
} else {
/*
* External, or BIND9 using multiple contexts:
* wait until woken up.
*/
if (atomic_load_acquire(&ctx->want_shutdown)) {
break;
}
if (!atomic_load_acquire(&ctx->want_reload)) {
LOCK(&ctx->readylock);
WAIT(&ctx->ready, &ctx->readylock);
UNLOCK(&ctx->readylock);
}
}
#endif /* WIN32 */
if (atomic_compare_exchange_strong_acq_rel(
&ctx->want_reload, &(bool){ true }, false))
{
return (ISC_R_RELOAD);
}
if (atomic_load_acquire(&ctx->want_shutdown) &&
atomic_load_acquire(&ctx->blocked))
{
exit(1);
}
}
return (ISC_R_SUCCESS);
}
isc_result_t
isc_app_run(void) {
isc_result_t result;
REQUIRE(atomic_compare_exchange_strong_acq_rel(&is_running,
&(bool){ false }, true));
result = isc_app_ctxrun(&isc_g_appctx);
atomic_store_release(&is_running, false);
return (result);
}
bool
isc_app_isrunning(void) {
return (atomic_load_acquire(&is_running));
}
void
isc_app_ctxshutdown(isc_appctx_t *ctx) {
REQUIRE(VALID_APPCTX(ctx));
REQUIRE(atomic_load_acquire(&ctx->running));
/* If ctx->shutdown_requested == true, we are already shutting
* down and we want to just bail out.
*/
if (atomic_compare_exchange_strong_acq_rel(&ctx->shutdown_requested,
&(bool){ false }, true))
{
#ifdef WIN32
SetEvent(ctx->hEvents[SHUTDOWN_EVENT]);
#else /* WIN32 */
if (isc_bind9 && ctx != &isc_g_appctx) {
/* BIND9 internal, but using multiple contexts */
atomic_store_release(&ctx->want_shutdown, true);
} else if (isc_bind9) {
/* BIND9 internal, single context */
if (kill(getpid(), SIGTERM) < 0) {
char strbuf[ISC_STRERRORSIZE];
strerror_r(errno, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"isc_app_shutdown() "
"kill: %s",
strbuf);
}
} else {
/* External, multiple contexts */
atomic_store_release(&ctx->want_shutdown, true);
SIGNAL(&ctx->ready);
}
#endif /* WIN32 */
}
}
void
isc_app_shutdown(void) {
isc_app_ctxshutdown(&isc_g_appctx);
}
void
isc_app_ctxsuspend(isc_appctx_t *ctx) {
REQUIRE(VALID_APPCTX(ctx));
REQUIRE(atomic_load(&ctx->running));
/*
* Don't send the reload signal if we're shutting down.
*/
if (!atomic_load_acquire(&ctx->shutdown_requested)) {
#ifdef WIN32
SetEvent(ctx->hEvents[RELOAD_EVENT]);
#else /* WIN32 */
if (isc_bind9 && ctx != &isc_g_appctx) {
/* BIND9 internal, but using multiple contexts */
atomic_store_release(&ctx->want_reload, true);
} else if (isc_bind9) {
/* BIND9 internal, single context */
if (kill(getpid(), SIGHUP) < 0) {
char strbuf[ISC_STRERRORSIZE];
strerror_r(errno, strbuf, sizeof(strbuf));
isc_error_fatal(__FILE__, __LINE__,
"isc_app_reload() "
"kill: %s",
strbuf);
}
} else {
/* External, multiple contexts */
atomic_store_release(&ctx->want_reload, true);
SIGNAL(&ctx->ready);
}
#endif /* WIN32 */
}
}
void
isc_app_reload(void) {
isc_app_ctxsuspend(&isc_g_appctx);
}
void
isc_app_ctxfinish(isc_appctx_t *ctx) {
REQUIRE(VALID_APPCTX(ctx));
isc_mutex_destroy(&ctx->lock);
#ifndef WIN32
isc_mutex_destroy(&ctx->readylock);
isc_condition_destroy(&ctx->ready);
#endif /* WIN32 */
}
void
isc_app_finish(void) {
isc_app_ctxfinish(&isc_g_appctx);
}
void
isc_app_block(void) {
REQUIRE(atomic_load_acquire(&isc_g_appctx.running));
REQUIRE(atomic_compare_exchange_strong_acq_rel(&isc_g_appctx.blocked,
&(bool){ false }, true));
#ifdef WIN32
blockedthread = GetCurrentThread();
#else /* WIN32 */
sigset_t sset;
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 /* WIN32 */
}
void
isc_app_unblock(void) {
REQUIRE(atomic_load_acquire(&isc_g_appctx.running));
REQUIRE(atomic_compare_exchange_strong_acq_rel(&isc_g_appctx.blocked,
&(bool){ true }, false));
#ifdef WIN32
REQUIRE(blockedthread == GetCurrentThread());
#else /* WIN32 */
REQUIRE(blockedthread == pthread_self());
sigset_t sset;
RUNTIME_CHECK(sigemptyset(&sset) == 0 &&
sigaddset(&sset, SIGINT) == 0 &&
sigaddset(&sset, SIGTERM) == 0);
RUNTIME_CHECK(pthread_sigmask(SIG_BLOCK, &sset, NULL) == 0);
#endif /* WIN32 */
}
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));
*ctx = (isc_appctx_t){ .magic = 0 };
isc_mem_attach(mctx, &ctx->mctx);
ctx->magic = APPCTX_MAGIC;
*ctxp = ctx;
return (ISC_R_SUCCESS);
}
void
isc_appctx_destroy(isc_appctx_t **ctxp) {
isc_appctx_t *ctx;
REQUIRE(ctxp != NULL);
ctx = *ctxp;
*ctxp = NULL;
REQUIRE(VALID_APPCTX(ctx));
ctx->magic = 0;
isc_mem_putanddetach(&ctx->mctx, ctx, sizeof(*ctx));
}