qemu/util/main-loop.c
Paolo Bonzini 5f50be9b58 async: the main AioContext is only "current" if under the BQL
If we want to wake up a coroutine from a worker thread, aio_co_wake()
currently does not work.  In that scenario, aio_co_wake() calls
aio_co_enter(), but there is no current AioContext and therefore
qemu_get_current_aio_context() returns the main thread.  aio_co_wake()
then attempts to call aio_context_acquire() instead of going through
aio_co_schedule().

The default case of qemu_get_current_aio_context() was added to cover
synchronous I/O started from the vCPU thread, but the main and vCPU
threads are quite different.  The main thread is an I/O thread itself,
only running a more complicated event loop; the vCPU thread instead
is essentially a worker thread that occasionally calls
qemu_mutex_lock_iothread().  It is only in those critical sections
that it acts as if it were the home thread of the main AioContext.

Therefore, this patch detaches qemu_get_current_aio_context() from
iothreads, which is a useless complication.  The AioContext pointer
is stored directly in the thread-local variable, including for the
main loop.  Worker threads (including vCPU threads) optionally behave
as temporary home threads if they have taken the big QEMU lock,
but if that is not the case they will always schedule coroutines
on remote threads via aio_co_schedule().

With this change, the stub qemu_mutex_iothread_locked() must be changed
from true to false.  The previous value of true was needed because the
main thread did not have an AioContext in the thread-local variable,
but now it does have one.

Reported-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210609122234.544153-1-pbonzini@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Tested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
[eblake: tweak commit message per Vladimir's review]
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-06-18 10:59:52 -05:00

595 lines
16 KiB
C

/*
* QEMU System Emulator
*
* Copyright (c) 2003-2008 Fabrice Bellard
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "qemu/osdep.h"
#include "qapi/error.h"
#include "qemu/cutils.h"
#include "qemu/timer.h"
#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
#include "qemu/main-loop.h"
#include "block/aio.h"
#include "qemu/error-report.h"
#include "qemu/queue.h"
#include "qemu/compiler.h"
#ifndef _WIN32
#include <sys/wait.h>
#endif
#ifndef _WIN32
/* If we have signalfd, we mask out the signals we want to handle and then
* use signalfd to listen for them. We rely on whatever the current signal
* handler is to dispatch the signals when we receive them.
*/
/*
* Disable CFI checks.
* We are going to call a signal hander directly. Such handler may or may not
* have been defined in our binary, so there's no guarantee that the pointer
* used to set the handler is a cfi-valid pointer. Since the handlers are
* stored in kernel memory, changing the handler to an attacker-defined
* function requires being able to call a sigaction() syscall,
* which is not as easy as overwriting a pointer in memory.
*/
QEMU_DISABLE_CFI
static void sigfd_handler(void *opaque)
{
int fd = (intptr_t)opaque;
struct qemu_signalfd_siginfo info;
struct sigaction action;
ssize_t len;
while (1) {
do {
len = read(fd, &info, sizeof(info));
} while (len == -1 && errno == EINTR);
if (len == -1 && errno == EAGAIN) {
break;
}
if (len != sizeof(info)) {
error_report("read from sigfd returned %zd: %s", len,
g_strerror(errno));
return;
}
sigaction(info.ssi_signo, NULL, &action);
if ((action.sa_flags & SA_SIGINFO) && action.sa_sigaction) {
sigaction_invoke(&action, &info);
} else if (action.sa_handler) {
action.sa_handler(info.ssi_signo);
}
}
}
static int qemu_signal_init(Error **errp)
{
int sigfd;
sigset_t set;
/*
* SIG_IPI must be blocked in the main thread and must not be caught
* by sigwait() in the signal thread. Otherwise, the cpu thread will
* not catch it reliably.
*/
sigemptyset(&set);
sigaddset(&set, SIG_IPI);
sigaddset(&set, SIGIO);
sigaddset(&set, SIGALRM);
sigaddset(&set, SIGBUS);
/* SIGINT cannot be handled via signalfd, so that ^C can be used
* to interrupt QEMU when it is being run under gdb. SIGHUP and
* SIGTERM are also handled asynchronously, even though it is not
* strictly necessary, because they use the same handler as SIGINT.
*/
pthread_sigmask(SIG_BLOCK, &set, NULL);
sigdelset(&set, SIG_IPI);
sigfd = qemu_signalfd(&set);
if (sigfd == -1) {
error_setg_errno(errp, errno, "failed to create signalfd");
return -errno;
}
fcntl_setfl(sigfd, O_NONBLOCK);
qemu_set_fd_handler(sigfd, sigfd_handler, NULL, (void *)(intptr_t)sigfd);
return 0;
}
#else /* _WIN32 */
static int qemu_signal_init(Error **errp)
{
return 0;
}
#endif
static AioContext *qemu_aio_context;
static QEMUBH *qemu_notify_bh;
static void notify_event_cb(void *opaque)
{
/* No need to do anything; this bottom half is only used to
* kick the kernel out of ppoll/poll/WaitForMultipleObjects.
*/
}
AioContext *qemu_get_aio_context(void)
{
return qemu_aio_context;
}
void qemu_notify_event(void)
{
if (!qemu_aio_context) {
return;
}
qemu_bh_schedule(qemu_notify_bh);
}
static GArray *gpollfds;
int qemu_init_main_loop(Error **errp)
{
int ret;
GSource *src;
init_clocks(qemu_timer_notify_cb);
ret = qemu_signal_init(errp);
if (ret) {
return ret;
}
qemu_aio_context = aio_context_new(errp);
if (!qemu_aio_context) {
return -EMFILE;
}
qemu_set_current_aio_context(qemu_aio_context);
qemu_notify_bh = qemu_bh_new(notify_event_cb, NULL);
gpollfds = g_array_new(FALSE, FALSE, sizeof(GPollFD));
src = aio_get_g_source(qemu_aio_context);
g_source_set_name(src, "aio-context");
g_source_attach(src, NULL);
g_source_unref(src);
src = iohandler_get_g_source();
g_source_set_name(src, "io-handler");
g_source_attach(src, NULL);
g_source_unref(src);
return 0;
}
static int max_priority;
#ifndef _WIN32
static int glib_pollfds_idx;
static int glib_n_poll_fds;
void qemu_fd_register(int fd)
{
}
static void glib_pollfds_fill(int64_t *cur_timeout)
{
GMainContext *context = g_main_context_default();
int timeout = 0;
int64_t timeout_ns;
int n;
g_main_context_prepare(context, &max_priority);
glib_pollfds_idx = gpollfds->len;
n = glib_n_poll_fds;
do {
GPollFD *pfds;
glib_n_poll_fds = n;
g_array_set_size(gpollfds, glib_pollfds_idx + glib_n_poll_fds);
pfds = &g_array_index(gpollfds, GPollFD, glib_pollfds_idx);
n = g_main_context_query(context, max_priority, &timeout, pfds,
glib_n_poll_fds);
} while (n != glib_n_poll_fds);
if (timeout < 0) {
timeout_ns = -1;
} else {
timeout_ns = (int64_t)timeout * (int64_t)SCALE_MS;
}
*cur_timeout = qemu_soonest_timeout(timeout_ns, *cur_timeout);
}
static void glib_pollfds_poll(void)
{
GMainContext *context = g_main_context_default();
GPollFD *pfds = &g_array_index(gpollfds, GPollFD, glib_pollfds_idx);
if (g_main_context_check(context, max_priority, pfds, glib_n_poll_fds)) {
g_main_context_dispatch(context);
}
}
#define MAX_MAIN_LOOP_SPIN (1000)
static int os_host_main_loop_wait(int64_t timeout)
{
GMainContext *context = g_main_context_default();
int ret;
g_main_context_acquire(context);
glib_pollfds_fill(&timeout);
qemu_mutex_unlock_iothread();
replay_mutex_unlock();
ret = qemu_poll_ns((GPollFD *)gpollfds->data, gpollfds->len, timeout);
replay_mutex_lock();
qemu_mutex_lock_iothread();
glib_pollfds_poll();
g_main_context_release(context);
return ret;
}
#else
/***********************************************************/
/* Polling handling */
typedef struct PollingEntry {
PollingFunc *func;
void *opaque;
struct PollingEntry *next;
} PollingEntry;
static PollingEntry *first_polling_entry;
int qemu_add_polling_cb(PollingFunc *func, void *opaque)
{
PollingEntry **ppe, *pe;
pe = g_malloc0(sizeof(PollingEntry));
pe->func = func;
pe->opaque = opaque;
for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next);
*ppe = pe;
return 0;
}
void qemu_del_polling_cb(PollingFunc *func, void *opaque)
{
PollingEntry **ppe, *pe;
for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next) {
pe = *ppe;
if (pe->func == func && pe->opaque == opaque) {
*ppe = pe->next;
g_free(pe);
break;
}
}
}
/***********************************************************/
/* Wait objects support */
typedef struct WaitObjects {
int num;
int revents[MAXIMUM_WAIT_OBJECTS + 1];
HANDLE events[MAXIMUM_WAIT_OBJECTS + 1];
WaitObjectFunc *func[MAXIMUM_WAIT_OBJECTS + 1];
void *opaque[MAXIMUM_WAIT_OBJECTS + 1];
} WaitObjects;
static WaitObjects wait_objects = {0};
int qemu_add_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
{
WaitObjects *w = &wait_objects;
if (w->num >= MAXIMUM_WAIT_OBJECTS) {
return -1;
}
w->events[w->num] = handle;
w->func[w->num] = func;
w->opaque[w->num] = opaque;
w->revents[w->num] = 0;
w->num++;
return 0;
}
void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
{
int i, found;
WaitObjects *w = &wait_objects;
found = 0;
for (i = 0; i < w->num; i++) {
if (w->events[i] == handle) {
found = 1;
}
if (found) {
w->events[i] = w->events[i + 1];
w->func[i] = w->func[i + 1];
w->opaque[i] = w->opaque[i + 1];
w->revents[i] = w->revents[i + 1];
}
}
if (found) {
w->num--;
}
}
void qemu_fd_register(int fd)
{
WSAEventSelect(fd, event_notifier_get_handle(&qemu_aio_context->notifier),
FD_READ | FD_ACCEPT | FD_CLOSE |
FD_CONNECT | FD_WRITE | FD_OOB);
}
static int pollfds_fill(GArray *pollfds, fd_set *rfds, fd_set *wfds,
fd_set *xfds)
{
int nfds = -1;
int i;
for (i = 0; i < pollfds->len; i++) {
GPollFD *pfd = &g_array_index(pollfds, GPollFD, i);
int fd = pfd->fd;
int events = pfd->events;
if (events & G_IO_IN) {
FD_SET(fd, rfds);
nfds = MAX(nfds, fd);
}
if (events & G_IO_OUT) {
FD_SET(fd, wfds);
nfds = MAX(nfds, fd);
}
if (events & G_IO_PRI) {
FD_SET(fd, xfds);
nfds = MAX(nfds, fd);
}
}
return nfds;
}
static void pollfds_poll(GArray *pollfds, int nfds, fd_set *rfds,
fd_set *wfds, fd_set *xfds)
{
int i;
for (i = 0; i < pollfds->len; i++) {
GPollFD *pfd = &g_array_index(pollfds, GPollFD, i);
int fd = pfd->fd;
int revents = 0;
if (FD_ISSET(fd, rfds)) {
revents |= G_IO_IN;
}
if (FD_ISSET(fd, wfds)) {
revents |= G_IO_OUT;
}
if (FD_ISSET(fd, xfds)) {
revents |= G_IO_PRI;
}
pfd->revents = revents & pfd->events;
}
}
static int os_host_main_loop_wait(int64_t timeout)
{
GMainContext *context = g_main_context_default();
GPollFD poll_fds[1024 * 2]; /* this is probably overkill */
int select_ret = 0;
int g_poll_ret, ret, i, n_poll_fds;
PollingEntry *pe;
WaitObjects *w = &wait_objects;
gint poll_timeout;
int64_t poll_timeout_ns;
static struct timeval tv0;
fd_set rfds, wfds, xfds;
int nfds;
g_main_context_acquire(context);
/* XXX: need to suppress polling by better using win32 events */
ret = 0;
for (pe = first_polling_entry; pe != NULL; pe = pe->next) {
ret |= pe->func(pe->opaque);
}
if (ret != 0) {
g_main_context_release(context);
return ret;
}
FD_ZERO(&rfds);
FD_ZERO(&wfds);
FD_ZERO(&xfds);
nfds = pollfds_fill(gpollfds, &rfds, &wfds, &xfds);
if (nfds >= 0) {
select_ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv0);
if (select_ret != 0) {
timeout = 0;
}
if (select_ret > 0) {
pollfds_poll(gpollfds, nfds, &rfds, &wfds, &xfds);
}
}
g_main_context_prepare(context, &max_priority);
n_poll_fds = g_main_context_query(context, max_priority, &poll_timeout,
poll_fds, ARRAY_SIZE(poll_fds));
g_assert(n_poll_fds + w->num <= ARRAY_SIZE(poll_fds));
for (i = 0; i < w->num; i++) {
poll_fds[n_poll_fds + i].fd = (DWORD_PTR)w->events[i];
poll_fds[n_poll_fds + i].events = G_IO_IN;
}
if (poll_timeout < 0) {
poll_timeout_ns = -1;
} else {
poll_timeout_ns = (int64_t)poll_timeout * (int64_t)SCALE_MS;
}
poll_timeout_ns = qemu_soonest_timeout(poll_timeout_ns, timeout);
qemu_mutex_unlock_iothread();
replay_mutex_unlock();
g_poll_ret = qemu_poll_ns(poll_fds, n_poll_fds + w->num, poll_timeout_ns);
replay_mutex_lock();
qemu_mutex_lock_iothread();
if (g_poll_ret > 0) {
for (i = 0; i < w->num; i++) {
w->revents[i] = poll_fds[n_poll_fds + i].revents;
}
for (i = 0; i < w->num; i++) {
if (w->revents[i] && w->func[i]) {
w->func[i](w->opaque[i]);
}
}
}
if (g_main_context_check(context, max_priority, poll_fds, n_poll_fds)) {
g_main_context_dispatch(context);
}
g_main_context_release(context);
return select_ret || g_poll_ret;
}
#endif
static NotifierList main_loop_poll_notifiers =
NOTIFIER_LIST_INITIALIZER(main_loop_poll_notifiers);
void main_loop_poll_add_notifier(Notifier *notify)
{
notifier_list_add(&main_loop_poll_notifiers, notify);
}
void main_loop_poll_remove_notifier(Notifier *notify)
{
notifier_remove(notify);
}
void main_loop_wait(int nonblocking)
{
MainLoopPoll mlpoll = {
.state = MAIN_LOOP_POLL_FILL,
.timeout = UINT32_MAX,
.pollfds = gpollfds,
};
int ret;
int64_t timeout_ns;
if (nonblocking) {
mlpoll.timeout = 0;
}
/* poll any events */
g_array_set_size(gpollfds, 0); /* reset for new iteration */
/* XXX: separate device handlers from system ones */
notifier_list_notify(&main_loop_poll_notifiers, &mlpoll);
if (mlpoll.timeout == UINT32_MAX) {
timeout_ns = -1;
} else {
timeout_ns = (uint64_t)mlpoll.timeout * (int64_t)(SCALE_MS);
}
timeout_ns = qemu_soonest_timeout(timeout_ns,
timerlistgroup_deadline_ns(
&main_loop_tlg));
ret = os_host_main_loop_wait(timeout_ns);
mlpoll.state = ret < 0 ? MAIN_LOOP_POLL_ERR : MAIN_LOOP_POLL_OK;
notifier_list_notify(&main_loop_poll_notifiers, &mlpoll);
if (icount_enabled()) {
/*
* CPU thread can infinitely wait for event after
* missing the warp
*/
icount_start_warp_timer();
}
qemu_clock_run_all_timers();
}
/* Functions to operate on the main QEMU AioContext. */
QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
{
return aio_bh_new(qemu_aio_context, cb, opaque);
}
/*
* Functions to operate on the I/O handler AioContext.
* This context runs on top of main loop. We can't reuse qemu_aio_context
* because iohandlers mustn't be polled by aio_poll(qemu_aio_context).
*/
static AioContext *iohandler_ctx;
static void iohandler_init(void)
{
if (!iohandler_ctx) {
iohandler_ctx = aio_context_new(&error_abort);
}
}
AioContext *iohandler_get_aio_context(void)
{
iohandler_init();
return iohandler_ctx;
}
GSource *iohandler_get_g_source(void)
{
iohandler_init();
return aio_get_g_source(iohandler_ctx);
}
void qemu_set_fd_handler(int fd,
IOHandler *fd_read,
IOHandler *fd_write,
void *opaque)
{
iohandler_init();
aio_set_fd_handler(iohandler_ctx, fd, false,
fd_read, fd_write, NULL, opaque);
}
void event_notifier_set_handler(EventNotifier *e,
EventNotifierHandler *handler)
{
iohandler_init();
aio_set_event_notifier(iohandler_ctx, e, false,
handler, NULL);
}