qemu/monitor/monitor.c
Kevin Wolf c81219a7dd monitor: Fix assertion failure on shutdown
Commit 357bda95 already tried to fix the order in monitor_cleanup() by
moving shutdown of the dispatcher coroutine further to the start.
However, it didn't go far enough:

iothread_stop() makes sure that all pending work (bottom halves) in the
AioContext of the monitor iothread is completed. iothread_destroy()
depends on this and fails an assertion if there is still a pending BH.

While the dispatcher coroutine is running, it will try to resume the
monitor after taking a request out of the queue, which involves a BH.
The dispatcher is run until it terminates in the AIO_WAIT_WHILE() loop.
However, adding new BHs between iothread_stop() and iothread_destroy()
is forbidden.

Fix this by stopping the dispatcher first before shutting down the other
parts of the monitor. This means we can now receive requests that aren't
handled any more when QEMU is shutting down, but this is unlikely to be
a problem for QMP clients.

Fixes: 357bda9590
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210212172028.288825-2-kwolf@redhat.com>
Tested-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2021-02-15 15:10:14 +01:00

775 lines
22 KiB
C

/*
* QEMU monitor
*
* Copyright (c) 2003-2004 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 "monitor-internal.h"
#include "qapi/error.h"
#include "qapi/opts-visitor.h"
#include "qapi/qapi-emit-events.h"
#include "qapi/qapi-visit-control.h"
#include "qapi/qmp/qdict.h"
#include "qemu/error-report.h"
#include "qemu/option.h"
#include "sysemu/qtest.h"
#include "sysemu/sysemu.h"
#include "trace.h"
/*
* To prevent flooding clients, events can be throttled. The
* throttling is calculated globally, rather than per-Monitor
* instance.
*/
typedef struct MonitorQAPIEventState {
QAPIEvent event; /* Throttling state for this event type and... */
QDict *data; /* ... data, see qapi_event_throttle_equal() */
QEMUTimer *timer; /* Timer for handling delayed events */
QDict *qdict; /* Delayed event (if any) */
} MonitorQAPIEventState;
typedef struct {
int64_t rate; /* Minimum time (in ns) between two events */
} MonitorQAPIEventConf;
/* Shared monitor I/O thread */
IOThread *mon_iothread;
/* Coroutine to dispatch the requests received from I/O thread */
Coroutine *qmp_dispatcher_co;
/* Set to true when the dispatcher coroutine should terminate */
bool qmp_dispatcher_co_shutdown;
/*
* qmp_dispatcher_co_busy is used for synchronisation between the
* monitor thread and the main thread to ensure that the dispatcher
* coroutine never gets scheduled a second time when it's already
* scheduled (scheduling the same coroutine twice is forbidden).
*
* It is true if the coroutine is active and processing requests.
* Additional requests may then be pushed onto mon->qmp_requests,
* and @qmp_dispatcher_co_shutdown may be set without further ado.
* @qmp_dispatcher_co_busy must not be woken up in this case.
*
* If false, you also have to set @qmp_dispatcher_co_busy to true and
* wake up @qmp_dispatcher_co after pushing the new requests.
*
* The coroutine will automatically change this variable back to false
* before it yields. Nobody else may set the variable to false.
*
* Access must be atomic for thread safety.
*/
bool qmp_dispatcher_co_busy;
/*
* Protects mon_list, monitor_qapi_event_state, coroutine_mon,
* monitor_destroyed.
*/
QemuMutex monitor_lock;
static GHashTable *monitor_qapi_event_state;
static GHashTable *coroutine_mon; /* Maps Coroutine* to Monitor* */
MonitorList mon_list;
int mon_refcount;
static bool monitor_destroyed;
Monitor *monitor_cur(void)
{
Monitor *mon;
qemu_mutex_lock(&monitor_lock);
mon = g_hash_table_lookup(coroutine_mon, qemu_coroutine_self());
qemu_mutex_unlock(&monitor_lock);
return mon;
}
/**
* Sets a new current monitor and returns the old one.
*
* If a non-NULL monitor is set for a coroutine, another call
* resetting it to NULL is required before the coroutine terminates,
* otherwise a stale entry would remain in the hash table.
*/
Monitor *monitor_set_cur(Coroutine *co, Monitor *mon)
{
Monitor *old_monitor = monitor_cur();
qemu_mutex_lock(&monitor_lock);
if (mon) {
g_hash_table_replace(coroutine_mon, co, mon);
} else {
g_hash_table_remove(coroutine_mon, co);
}
qemu_mutex_unlock(&monitor_lock);
return old_monitor;
}
/**
* Is the current monitor, if any, a QMP monitor?
*/
bool monitor_cur_is_qmp(void)
{
Monitor *cur_mon = monitor_cur();
return cur_mon && monitor_is_qmp(cur_mon);
}
/**
* Is @mon is using readline?
* Note: not all HMP monitors use readline, e.g., gdbserver has a
* non-interactive HMP monitor, so readline is not used there.
*/
static inline bool monitor_uses_readline(const MonitorHMP *mon)
{
return mon->use_readline;
}
static inline bool monitor_is_hmp_non_interactive(const Monitor *mon)
{
if (monitor_is_qmp(mon)) {
return false;
}
return !monitor_uses_readline(container_of(mon, MonitorHMP, common));
}
static void monitor_flush_locked(Monitor *mon);
static gboolean monitor_unblocked(GIOChannel *chan, GIOCondition cond,
void *opaque)
{
Monitor *mon = opaque;
qemu_mutex_lock(&mon->mon_lock);
mon->out_watch = 0;
monitor_flush_locked(mon);
qemu_mutex_unlock(&mon->mon_lock);
return FALSE;
}
/* Caller must hold mon->mon_lock */
static void monitor_flush_locked(Monitor *mon)
{
int rc;
size_t len;
const char *buf;
if (mon->skip_flush) {
return;
}
buf = mon->outbuf->str;
len = mon->outbuf->len;
if (len && !mon->mux_out) {
rc = qemu_chr_fe_write(&mon->chr, (const uint8_t *) buf, len);
if ((rc < 0 && errno != EAGAIN) || (rc == len)) {
/* all flushed or error */
g_string_truncate(mon->outbuf, 0);
return;
}
if (rc > 0) {
/* partial write */
g_string_erase(mon->outbuf, 0, rc);
}
if (mon->out_watch == 0) {
mon->out_watch =
qemu_chr_fe_add_watch(&mon->chr, G_IO_OUT | G_IO_HUP,
monitor_unblocked, mon);
}
}
}
void monitor_flush(Monitor *mon)
{
qemu_mutex_lock(&mon->mon_lock);
monitor_flush_locked(mon);
qemu_mutex_unlock(&mon->mon_lock);
}
/* flush at every end of line */
int monitor_puts(Monitor *mon, const char *str)
{
int i;
char c;
qemu_mutex_lock(&mon->mon_lock);
for (i = 0; str[i]; i++) {
c = str[i];
if (c == '\n') {
g_string_append_c(mon->outbuf, '\r');
}
g_string_append_c(mon->outbuf, c);
if (c == '\n') {
monitor_flush_locked(mon);
}
}
qemu_mutex_unlock(&mon->mon_lock);
return i;
}
int monitor_vprintf(Monitor *mon, const char *fmt, va_list ap)
{
char *buf;
int n;
if (!mon) {
return -1;
}
if (monitor_is_qmp(mon)) {
return -1;
}
buf = g_strdup_vprintf(fmt, ap);
n = monitor_puts(mon, buf);
g_free(buf);
return n;
}
int monitor_printf(Monitor *mon, const char *fmt, ...)
{
int ret;
va_list ap;
va_start(ap, fmt);
ret = monitor_vprintf(mon, fmt, ap);
va_end(ap);
return ret;
}
/*
* Print to current monitor if we have one, else to stderr.
*/
int error_vprintf(const char *fmt, va_list ap)
{
Monitor *cur_mon = monitor_cur();
if (cur_mon && !monitor_cur_is_qmp()) {
return monitor_vprintf(cur_mon, fmt, ap);
}
return vfprintf(stderr, fmt, ap);
}
int error_vprintf_unless_qmp(const char *fmt, va_list ap)
{
Monitor *cur_mon = monitor_cur();
if (!cur_mon) {
return vfprintf(stderr, fmt, ap);
}
if (!monitor_cur_is_qmp()) {
return monitor_vprintf(cur_mon, fmt, ap);
}
return -1;
}
static MonitorQAPIEventConf monitor_qapi_event_conf[QAPI_EVENT__MAX] = {
/* Limit guest-triggerable events to 1 per second */
[QAPI_EVENT_RTC_CHANGE] = { 1000 * SCALE_MS },
[QAPI_EVENT_WATCHDOG] = { 1000 * SCALE_MS },
[QAPI_EVENT_BALLOON_CHANGE] = { 1000 * SCALE_MS },
[QAPI_EVENT_QUORUM_REPORT_BAD] = { 1000 * SCALE_MS },
[QAPI_EVENT_QUORUM_FAILURE] = { 1000 * SCALE_MS },
[QAPI_EVENT_VSERPORT_CHANGE] = { 1000 * SCALE_MS },
[QAPI_EVENT_MEMORY_DEVICE_SIZE_CHANGE] = { 1000 * SCALE_MS },
};
/*
* Return the clock to use for recording an event's time.
* It's QEMU_CLOCK_REALTIME, except for qtests it's
* QEMU_CLOCK_VIRTUAL, to support testing rate limits.
* Beware: result is invalid before configure_accelerator().
*/
static inline QEMUClockType monitor_get_event_clock(void)
{
return qtest_enabled() ? QEMU_CLOCK_VIRTUAL : QEMU_CLOCK_REALTIME;
}
/*
* Broadcast an event to all monitors.
* @qdict is the event object. Its member "event" must match @event.
* Caller must hold monitor_lock.
*/
static void monitor_qapi_event_emit(QAPIEvent event, QDict *qdict)
{
Monitor *mon;
MonitorQMP *qmp_mon;
trace_monitor_protocol_event_emit(event, qdict);
QTAILQ_FOREACH(mon, &mon_list, entry) {
if (!monitor_is_qmp(mon)) {
continue;
}
qmp_mon = container_of(mon, MonitorQMP, common);
if (qmp_mon->commands != &qmp_cap_negotiation_commands) {
qmp_send_response(qmp_mon, qdict);
}
}
}
static void monitor_qapi_event_handler(void *opaque);
/*
* Queue a new event for emission to Monitor instances,
* applying any rate limiting if required.
*/
static void
monitor_qapi_event_queue_no_reenter(QAPIEvent event, QDict *qdict)
{
MonitorQAPIEventConf *evconf;
MonitorQAPIEventState *evstate;
assert(event < QAPI_EVENT__MAX);
evconf = &monitor_qapi_event_conf[event];
trace_monitor_protocol_event_queue(event, qdict, evconf->rate);
qemu_mutex_lock(&monitor_lock);
if (!evconf->rate) {
/* Unthrottled event */
monitor_qapi_event_emit(event, qdict);
} else {
QDict *data = qobject_to(QDict, qdict_get(qdict, "data"));
MonitorQAPIEventState key = { .event = event, .data = data };
evstate = g_hash_table_lookup(monitor_qapi_event_state, &key);
assert(!evstate || timer_pending(evstate->timer));
if (evstate) {
/*
* Timer is pending for (at least) evconf->rate ns after
* last send. Store event for sending when timer fires,
* replacing a prior stored event if any.
*/
qobject_unref(evstate->qdict);
evstate->qdict = qobject_ref(qdict);
} else {
/*
* Last send was (at least) evconf->rate ns ago.
* Send immediately, and arm the timer to call
* monitor_qapi_event_handler() in evconf->rate ns. Any
* events arriving before then will be delayed until then.
*/
int64_t now = qemu_clock_get_ns(monitor_get_event_clock());
monitor_qapi_event_emit(event, qdict);
evstate = g_new(MonitorQAPIEventState, 1);
evstate->event = event;
evstate->data = qobject_ref(data);
evstate->qdict = NULL;
evstate->timer = timer_new_ns(monitor_get_event_clock(),
monitor_qapi_event_handler,
evstate);
g_hash_table_add(monitor_qapi_event_state, evstate);
timer_mod_ns(evstate->timer, now + evconf->rate);
}
}
qemu_mutex_unlock(&monitor_lock);
}
void qapi_event_emit(QAPIEvent event, QDict *qdict)
{
/*
* monitor_qapi_event_queue_no_reenter() is not reentrant: it
* would deadlock on monitor_lock. Work around by queueing
* events in thread-local storage.
* TODO: remove this, make it re-enter safe.
*/
typedef struct MonitorQapiEvent {
QAPIEvent event;
QDict *qdict;
QSIMPLEQ_ENTRY(MonitorQapiEvent) entry;
} MonitorQapiEvent;
static __thread QSIMPLEQ_HEAD(, MonitorQapiEvent) event_queue;
static __thread bool reentered;
MonitorQapiEvent *ev;
if (!reentered) {
QSIMPLEQ_INIT(&event_queue);
}
ev = g_new(MonitorQapiEvent, 1);
ev->qdict = qobject_ref(qdict);
ev->event = event;
QSIMPLEQ_INSERT_TAIL(&event_queue, ev, entry);
if (reentered) {
return;
}
reentered = true;
while ((ev = QSIMPLEQ_FIRST(&event_queue)) != NULL) {
QSIMPLEQ_REMOVE_HEAD(&event_queue, entry);
monitor_qapi_event_queue_no_reenter(ev->event, ev->qdict);
qobject_unref(ev->qdict);
g_free(ev);
}
reentered = false;
}
/*
* This function runs evconf->rate ns after sending a throttled
* event.
* If another event has since been stored, send it.
*/
static void monitor_qapi_event_handler(void *opaque)
{
MonitorQAPIEventState *evstate = opaque;
MonitorQAPIEventConf *evconf = &monitor_qapi_event_conf[evstate->event];
trace_monitor_protocol_event_handler(evstate->event, evstate->qdict);
qemu_mutex_lock(&monitor_lock);
if (evstate->qdict) {
int64_t now = qemu_clock_get_ns(monitor_get_event_clock());
monitor_qapi_event_emit(evstate->event, evstate->qdict);
qobject_unref(evstate->qdict);
evstate->qdict = NULL;
timer_mod_ns(evstate->timer, now + evconf->rate);
} else {
g_hash_table_remove(monitor_qapi_event_state, evstate);
qobject_unref(evstate->data);
timer_free(evstate->timer);
g_free(evstate);
}
qemu_mutex_unlock(&monitor_lock);
}
static unsigned int qapi_event_throttle_hash(const void *key)
{
const MonitorQAPIEventState *evstate = key;
unsigned int hash = evstate->event * 255;
if (evstate->event == QAPI_EVENT_VSERPORT_CHANGE) {
hash += g_str_hash(qdict_get_str(evstate->data, "id"));
}
if (evstate->event == QAPI_EVENT_QUORUM_REPORT_BAD) {
hash += g_str_hash(qdict_get_str(evstate->data, "node-name"));
}
return hash;
}
static gboolean qapi_event_throttle_equal(const void *a, const void *b)
{
const MonitorQAPIEventState *eva = a;
const MonitorQAPIEventState *evb = b;
if (eva->event != evb->event) {
return FALSE;
}
if (eva->event == QAPI_EVENT_VSERPORT_CHANGE) {
return !strcmp(qdict_get_str(eva->data, "id"),
qdict_get_str(evb->data, "id"));
}
if (eva->event == QAPI_EVENT_QUORUM_REPORT_BAD) {
return !strcmp(qdict_get_str(eva->data, "node-name"),
qdict_get_str(evb->data, "node-name"));
}
return TRUE;
}
int monitor_suspend(Monitor *mon)
{
if (monitor_is_hmp_non_interactive(mon)) {
return -ENOTTY;
}
qatomic_inc(&mon->suspend_cnt);
if (mon->use_io_thread) {
/*
* Kick I/O thread to make sure this takes effect. It'll be
* evaluated again in prepare() of the watch object.
*/
aio_notify(iothread_get_aio_context(mon_iothread));
}
trace_monitor_suspend(mon, 1);
return 0;
}
static void monitor_accept_input(void *opaque)
{
Monitor *mon = opaque;
qemu_chr_fe_accept_input(&mon->chr);
}
void monitor_resume(Monitor *mon)
{
if (monitor_is_hmp_non_interactive(mon)) {
return;
}
if (qatomic_dec_fetch(&mon->suspend_cnt) == 0) {
AioContext *ctx;
if (mon->use_io_thread) {
ctx = iothread_get_aio_context(mon_iothread);
} else {
ctx = qemu_get_aio_context();
}
if (!monitor_is_qmp(mon)) {
MonitorHMP *hmp_mon = container_of(mon, MonitorHMP, common);
assert(hmp_mon->rs);
readline_show_prompt(hmp_mon->rs);
}
aio_bh_schedule_oneshot(ctx, monitor_accept_input, mon);
}
trace_monitor_suspend(mon, -1);
}
int monitor_can_read(void *opaque)
{
Monitor *mon = opaque;
return !qatomic_mb_read(&mon->suspend_cnt);
}
void monitor_list_append(Monitor *mon)
{
qemu_mutex_lock(&monitor_lock);
/*
* This prevents inserting new monitors during monitor_cleanup().
* A cleaner solution would involve the main thread telling other
* threads to terminate, waiting for their termination.
*/
if (!monitor_destroyed) {
QTAILQ_INSERT_HEAD(&mon_list, mon, entry);
mon = NULL;
}
qemu_mutex_unlock(&monitor_lock);
if (mon) {
monitor_data_destroy(mon);
g_free(mon);
}
}
static void monitor_iothread_init(void)
{
mon_iothread = iothread_create("mon_iothread", &error_abort);
}
void monitor_data_init(Monitor *mon, bool is_qmp, bool skip_flush,
bool use_io_thread)
{
if (use_io_thread && !mon_iothread) {
monitor_iothread_init();
}
qemu_mutex_init(&mon->mon_lock);
mon->is_qmp = is_qmp;
mon->outbuf = g_string_new(NULL);
mon->skip_flush = skip_flush;
mon->use_io_thread = use_io_thread;
}
void monitor_data_destroy(Monitor *mon)
{
g_free(mon->mon_cpu_path);
qemu_chr_fe_deinit(&mon->chr, false);
if (monitor_is_qmp(mon)) {
monitor_data_destroy_qmp(container_of(mon, MonitorQMP, common));
} else {
readline_free(container_of(mon, MonitorHMP, common)->rs);
}
g_string_free(mon->outbuf, true);
qemu_mutex_destroy(&mon->mon_lock);
}
void monitor_cleanup(void)
{
/*
* The dispatcher needs to stop before destroying the monitor and
* the I/O thread.
*
* We need to poll both qemu_aio_context and iohandler_ctx to make
* sure that the dispatcher coroutine keeps making progress and
* eventually terminates. qemu_aio_context is automatically
* polled by calling AIO_WAIT_WHILE on it, but we must poll
* iohandler_ctx manually.
*
* Letting the iothread continue while shutting down the dispatcher
* means that new requests may still be coming in. This is okay,
* we'll just leave them in the queue without sending a response
* and monitor_data_destroy() will free them.
*/
qmp_dispatcher_co_shutdown = true;
if (!qatomic_xchg(&qmp_dispatcher_co_busy, true)) {
aio_co_wake(qmp_dispatcher_co);
}
AIO_WAIT_WHILE(qemu_get_aio_context(),
(aio_poll(iohandler_get_aio_context(), false),
qatomic_mb_read(&qmp_dispatcher_co_busy)));
/*
* We need to explicitly stop the I/O thread (but not destroy it),
* clean up the monitor resources, then destroy the I/O thread since
* we need to unregister from chardev below in
* monitor_data_destroy(), and chardev is not thread-safe yet
*/
if (mon_iothread) {
iothread_stop(mon_iothread);
}
/* Flush output buffers and destroy monitors */
qemu_mutex_lock(&monitor_lock);
monitor_destroyed = true;
while (!QTAILQ_EMPTY(&mon_list)) {
Monitor *mon = QTAILQ_FIRST(&mon_list);
QTAILQ_REMOVE(&mon_list, mon, entry);
/* Permit QAPI event emission from character frontend release */
qemu_mutex_unlock(&monitor_lock);
monitor_flush(mon);
monitor_data_destroy(mon);
qemu_mutex_lock(&monitor_lock);
g_free(mon);
}
qemu_mutex_unlock(&monitor_lock);
if (mon_iothread) {
iothread_destroy(mon_iothread);
mon_iothread = NULL;
}
}
static void monitor_qapi_event_init(void)
{
monitor_qapi_event_state = g_hash_table_new(qapi_event_throttle_hash,
qapi_event_throttle_equal);
}
void monitor_init_globals_core(void)
{
monitor_qapi_event_init();
qemu_mutex_init(&monitor_lock);
coroutine_mon = g_hash_table_new(NULL, NULL);
/*
* The dispatcher BH must run in the main loop thread, since we
* have commands assuming that context. It would be nice to get
* rid of those assumptions.
*/
qmp_dispatcher_co = qemu_coroutine_create(monitor_qmp_dispatcher_co, NULL);
qatomic_mb_set(&qmp_dispatcher_co_busy, true);
aio_co_schedule(iohandler_get_aio_context(), qmp_dispatcher_co);
}
int monitor_init(MonitorOptions *opts, bool allow_hmp, Error **errp)
{
Chardev *chr;
Error *local_err = NULL;
chr = qemu_chr_find(opts->chardev);
if (chr == NULL) {
error_setg(errp, "chardev \"%s\" not found", opts->chardev);
return -1;
}
if (!opts->has_mode) {
opts->mode = allow_hmp ? MONITOR_MODE_READLINE : MONITOR_MODE_CONTROL;
}
switch (opts->mode) {
case MONITOR_MODE_CONTROL:
monitor_init_qmp(chr, opts->pretty, &local_err);
break;
case MONITOR_MODE_READLINE:
if (!allow_hmp) {
error_setg(errp, "Only QMP is supported");
return -1;
}
if (opts->pretty) {
warn_report("'pretty' is deprecated for HMP monitors, it has no "
"effect and will be removed in future versions");
}
monitor_init_hmp(chr, true, &local_err);
break;
default:
g_assert_not_reached();
}
if (local_err) {
error_propagate(errp, local_err);
return -1;
}
return 0;
}
int monitor_init_opts(QemuOpts *opts, Error **errp)
{
Visitor *v;
MonitorOptions *options;
int ret;
v = opts_visitor_new(opts);
visit_type_MonitorOptions(v, NULL, &options, errp);
visit_free(v);
if (!options) {
return -1;
}
ret = monitor_init(options, true, errp);
qapi_free_MonitorOptions(options);
return ret;
}
QemuOptsList qemu_mon_opts = {
.name = "mon",
.implied_opt_name = "chardev",
.head = QTAILQ_HEAD_INITIALIZER(qemu_mon_opts.head),
.desc = {
{
.name = "mode",
.type = QEMU_OPT_STRING,
},{
.name = "chardev",
.type = QEMU_OPT_STRING,
},{
.name = "pretty",
.type = QEMU_OPT_BOOL,
},
{ /* end of list */ }
},
};