*/
#include "qemu/osdep.h"
-#include "qemu-common.h"
#include "block/block.h"
+#include "qemu/rcu.h"
#include "qemu/rcu_queue.h"
#include "qemu/sockets.h"
#include "qemu/cutils.h"
#include "trace.h"
-#ifdef CONFIG_EPOLL_CREATE1
-#include <sys/epoll.h>
-#endif
+#include "aio-posix.h"
-struct AioHandler
-{
- GPollFD pfd;
- IOHandler *io_read;
- IOHandler *io_write;
- AioPollFn *io_poll;
- IOHandler *io_poll_begin;
- IOHandler *io_poll_end;
- int deleted;
- void *opaque;
- bool is_external;
- QLIST_ENTRY(AioHandler) node;
-};
-
-#ifdef CONFIG_EPOLL_CREATE1
-
-/* The fd number threashold to switch to epoll */
-#define EPOLL_ENABLE_THRESHOLD 64
-
-static void aio_epoll_disable(AioContext *ctx)
+/* Stop userspace polling on a handler if it isn't active for some time */
+#define POLL_IDLE_INTERVAL_NS (7 * NANOSECONDS_PER_SECOND)
+
+bool aio_poll_disabled(AioContext *ctx)
{
- ctx->epoll_available = false;
- if (!ctx->epoll_enabled) {
- return;
- }
- ctx->epoll_enabled = false;
- close(ctx->epollfd);
+ return atomic_read(&ctx->poll_disable_cnt);
}
-static inline int epoll_events_from_pfd(int pfd_events)
+void aio_add_ready_handler(AioHandlerList *ready_list,
+ AioHandler *node,
+ int revents)
{
- return (pfd_events & G_IO_IN ? EPOLLIN : 0) |
- (pfd_events & G_IO_OUT ? EPOLLOUT : 0) |
- (pfd_events & G_IO_HUP ? EPOLLHUP : 0) |
- (pfd_events & G_IO_ERR ? EPOLLERR : 0);
+ QLIST_SAFE_REMOVE(node, node_ready); /* remove from nested parent's list */
+ node->pfd.revents = revents;
+ QLIST_INSERT_HEAD(ready_list, node, node_ready);
}
-static bool aio_epoll_try_enable(AioContext *ctx)
+static AioHandler *find_aio_handler(AioContext *ctx, int fd)
{
AioHandler *node;
- struct epoll_event event;
- QLIST_FOREACH_RCU(node, &ctx->aio_handlers, node) {
- int r;
- if (node->deleted || !node->pfd.events) {
- continue;
- }
- event.events = epoll_events_from_pfd(node->pfd.events);
- event.data.ptr = node;
- r = epoll_ctl(ctx->epollfd, EPOLL_CTL_ADD, node->pfd.fd, &event);
- if (r) {
- return false;
+ QLIST_FOREACH(node, &ctx->aio_handlers, node) {
+ if (node->pfd.fd == fd) {
+ if (!QLIST_IS_INSERTED(node, node_deleted)) {
+ return node;
+ }
}
}
- ctx->epoll_enabled = true;
- return true;
-}
-static void aio_epoll_update(AioContext *ctx, AioHandler *node, bool is_new)
-{
- struct epoll_event event;
- int r;
- int ctl;
-
- if (!ctx->epoll_enabled) {
- return;
- }
- if (!node->pfd.events) {
- ctl = EPOLL_CTL_DEL;
- } else {
- event.data.ptr = node;
- event.events = epoll_events_from_pfd(node->pfd.events);
- ctl = is_new ? EPOLL_CTL_ADD : EPOLL_CTL_MOD;
- }
-
- r = epoll_ctl(ctx->epollfd, ctl, node->pfd.fd, &event);
- if (r) {
- aio_epoll_disable(ctx);
- }
+ return NULL;
}
-static int aio_epoll(AioContext *ctx, GPollFD *pfds,
- unsigned npfd, int64_t timeout)
+static bool aio_remove_fd_handler(AioContext *ctx, AioHandler *node)
{
- AioHandler *node;
- int i, ret = 0;
- struct epoll_event events[128];
-
- assert(npfd == 1);
- assert(pfds[0].fd == ctx->epollfd);
- if (timeout > 0) {
- ret = qemu_poll_ns(pfds, npfd, timeout);
- }
- if (timeout <= 0 || ret > 0) {
- ret = epoll_wait(ctx->epollfd, events,
- sizeof(events) / sizeof(events[0]),
- timeout);
- if (ret <= 0) {
- goto out;
- }
- for (i = 0; i < ret; i++) {
- int ev = events[i].events;
- node = events[i].data.ptr;
- node->pfd.revents = (ev & EPOLLIN ? G_IO_IN : 0) |
- (ev & EPOLLOUT ? G_IO_OUT : 0) |
- (ev & EPOLLHUP ? G_IO_HUP : 0) |
- (ev & EPOLLERR ? G_IO_ERR : 0);
- }
+ /* If the GSource is in the process of being destroyed then
+ * g_source_remove_poll() causes an assertion failure. Skip
+ * removal in that case, because glib cleans up its state during
+ * destruction anyway.
+ */
+ if (!g_source_is_destroyed(&ctx->source)) {
+ g_source_remove_poll(&ctx->source, &node->pfd);
}
-out:
- return ret;
-}
-static bool aio_epoll_enabled(AioContext *ctx)
-{
- /* Fall back to ppoll when external clients are disabled. */
- return !aio_external_disabled(ctx) && ctx->epoll_enabled;
-}
+ node->pfd.revents = 0;
-static bool aio_epoll_check_poll(AioContext *ctx, GPollFD *pfds,
- unsigned npfd, int64_t timeout)
-{
- if (!ctx->epoll_available) {
+ /* If the fd monitor has already marked it deleted, leave it alone */
+ if (QLIST_IS_INSERTED(node, node_deleted)) {
return false;
}
- if (aio_epoll_enabled(ctx)) {
- return true;
- }
- if (npfd >= EPOLL_ENABLE_THRESHOLD) {
- if (aio_epoll_try_enable(ctx)) {
- return true;
- } else {
- aio_epoll_disable(ctx);
- }
- }
- return false;
-}
-
-#else
-
-static void aio_epoll_update(AioContext *ctx, AioHandler *node, bool is_new)
-{
-}
-
-static int aio_epoll(AioContext *ctx, GPollFD *pfds,
- unsigned npfd, int64_t timeout)
-{
- assert(false);
-}
-
-static bool aio_epoll_enabled(AioContext *ctx)
-{
- return false;
-}
-
-static bool aio_epoll_check_poll(AioContext *ctx, GPollFD *pfds,
- unsigned npfd, int64_t timeout)
-{
- return false;
-}
-#endif
-
-static AioHandler *find_aio_handler(AioContext *ctx, int fd)
-{
- AioHandler *node;
-
- QLIST_FOREACH(node, &ctx->aio_handlers, node) {
- if (node->pfd.fd == fd)
- if (!node->deleted)
- return node;
+ /* If a read is in progress, just mark the node as deleted */
+ if (qemu_lockcnt_count(&ctx->list_lock)) {
+ QLIST_INSERT_HEAD_RCU(&ctx->deleted_aio_handlers, node, node_deleted);
+ return false;
}
-
- return NULL;
+ /* Otherwise, delete it for real. We can't just mark it as
+ * deleted because deleted nodes are only cleaned up while
+ * no one is walking the handlers list.
+ */
+ QLIST_SAFE_REMOVE(node, node_poll);
+ QLIST_REMOVE(node, node);
+ return true;
}
void aio_set_fd_handler(AioContext *ctx,
void *opaque)
{
AioHandler *node;
+ AioHandler *new_node = NULL;
bool is_new = false;
bool deleted = false;
+ int poll_disable_change;
qemu_lockcnt_lock(&ctx->list_lock);
qemu_lockcnt_unlock(&ctx->list_lock);
return;
}
+ /* Clean events in order to unregister fd from the ctx epoll. */
+ node->pfd.events = 0;
- /* If the GSource is in the process of being destroyed then
- * g_source_remove_poll() causes an assertion failure. Skip
- * removal in that case, because glib cleans up its state during
- * destruction anyway.
- */
- if (!g_source_is_destroyed(&ctx->source)) {
- g_source_remove_poll(&ctx->source, &node->pfd);
- }
-
- /* If the lock is held, just mark the node as deleted */
- if (qemu_lockcnt_count(&ctx->list_lock)) {
- node->deleted = 1;
- node->pfd.revents = 0;
- } else {
- /* Otherwise, delete it for real. We can't just mark it as
- * deleted because deleted nodes are only cleaned up while
- * no one is walking the handlers list.
- */
- QLIST_REMOVE(node, node);
- deleted = true;
- }
-
- if (!node->io_poll) {
- ctx->poll_disable_cnt--;
- }
+ poll_disable_change = -!node->io_poll;
} else {
+ poll_disable_change = !io_poll - (node && !node->io_poll);
if (node == NULL) {
- /* Alloc and insert if it's not already there */
- node = g_new0(AioHandler, 1);
- node->pfd.fd = fd;
- QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, node, node);
-
- g_source_add_poll(&ctx->source, &node->pfd);
is_new = true;
+ }
+ /* Alloc and insert if it's not already there */
+ new_node = g_new0(AioHandler, 1);
- ctx->poll_disable_cnt += !io_poll;
+ /* Update handler with latest information */
+ new_node->io_read = io_read;
+ new_node->io_write = io_write;
+ new_node->io_poll = io_poll;
+ new_node->opaque = opaque;
+ new_node->is_external = is_external;
+
+ if (is_new) {
+ new_node->pfd.fd = fd;
} else {
- ctx->poll_disable_cnt += !io_poll - !node->io_poll;
+ new_node->pfd = node->pfd;
}
+ g_source_add_poll(&ctx->source, &new_node->pfd);
- /* Update handler with latest information */
- node->io_read = io_read;
- node->io_write = io_write;
- node->io_poll = io_poll;
- node->opaque = opaque;
- node->is_external = is_external;
+ new_node->pfd.events = (io_read ? G_IO_IN | G_IO_HUP | G_IO_ERR : 0);
+ new_node->pfd.events |= (io_write ? G_IO_OUT | G_IO_ERR : 0);
- node->pfd.events = (io_read ? G_IO_IN | G_IO_HUP | G_IO_ERR : 0);
- node->pfd.events |= (io_write ? G_IO_OUT | G_IO_ERR : 0);
+ QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, new_node, node);
}
- aio_epoll_update(ctx, node, is_new);
+ /* No need to order poll_disable_cnt writes against other updates;
+ * the counter is only used to avoid wasting time and latency on
+ * iterated polling when the system call will be ultimately necessary.
+ * Changing handlers is a rare event, and a little wasted polling until
+ * the aio_notify below is not an issue.
+ */
+ atomic_set(&ctx->poll_disable_cnt,
+ atomic_read(&ctx->poll_disable_cnt) + poll_disable_change);
+
+ ctx->fdmon_ops->update(ctx, node, new_node);
+ if (node) {
+ deleted = aio_remove_fd_handler(ctx, node);
+ }
qemu_lockcnt_unlock(&ctx->list_lock);
aio_notify(ctx);
(IOHandler *)io_poll_end);
}
-static void poll_set_started(AioContext *ctx, bool started)
+static bool poll_set_started(AioContext *ctx, bool started)
{
AioHandler *node;
+ bool progress = false;
if (started == ctx->poll_started) {
- return;
+ return false;
}
ctx->poll_started = started;
qemu_lockcnt_inc(&ctx->list_lock);
- QLIST_FOREACH_RCU(node, &ctx->aio_handlers, node) {
+ QLIST_FOREACH(node, &ctx->poll_aio_handlers, node_poll) {
IOHandler *fn;
- if (node->deleted) {
+ if (QLIST_IS_INSERTED(node, node_deleted)) {
continue;
}
if (fn) {
fn(node->opaque);
}
+
+ /* Poll one last time in case ->io_poll_end() raced with the event */
+ if (!started) {
+ progress = node->io_poll(node->opaque) || progress;
+ }
}
qemu_lockcnt_dec(&ctx->list_lock);
+
+ return progress;
}
return result;
}
-static bool aio_dispatch_handlers(AioContext *ctx)
+static void aio_free_deleted_handlers(AioContext *ctx)
{
- AioHandler *node, *tmp;
- bool progress = false;
+ AioHandler *node;
- QLIST_FOREACH_SAFE_RCU(node, &ctx->aio_handlers, node, tmp) {
- int revents;
+ if (QLIST_EMPTY_RCU(&ctx->deleted_aio_handlers)) {
+ return;
+ }
+ if (!qemu_lockcnt_dec_if_lock(&ctx->list_lock)) {
+ return; /* we are nested, let the parent do the freeing */
+ }
- revents = node->pfd.revents & node->pfd.events;
- node->pfd.revents = 0;
+ while ((node = QLIST_FIRST_RCU(&ctx->deleted_aio_handlers))) {
+ QLIST_REMOVE(node, node);
+ QLIST_REMOVE(node, node_deleted);
+ QLIST_SAFE_REMOVE(node, node_poll);
+ g_free(node);
+ }
- if (!node->deleted &&
- (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) &&
- aio_node_check(ctx, node->is_external) &&
- node->io_read) {
- node->io_read(node->opaque);
+ qemu_lockcnt_inc_and_unlock(&ctx->list_lock);
+}
- /* aio_notify() does not count as progress */
- if (node->opaque != &ctx->notifier) {
- progress = true;
- }
+static bool aio_dispatch_handler(AioContext *ctx, AioHandler *node)
+{
+ bool progress = false;
+ int revents;
+
+ revents = node->pfd.revents & node->pfd.events;
+ node->pfd.revents = 0;
+
+ /*
+ * Start polling AioHandlers when they become ready because activity is
+ * likely to continue. Note that starvation is theoretically possible when
+ * fdmon_supports_polling(), but only until the fd fires for the first
+ * time.
+ */
+ if (!QLIST_IS_INSERTED(node, node_deleted) &&
+ !QLIST_IS_INSERTED(node, node_poll) &&
+ node->io_poll) {
+ trace_poll_add(ctx, node, node->pfd.fd, revents);
+ if (ctx->poll_started && node->io_poll_begin) {
+ node->io_poll_begin(node->opaque);
}
- if (!node->deleted &&
- (revents & (G_IO_OUT | G_IO_ERR)) &&
- aio_node_check(ctx, node->is_external) &&
- node->io_write) {
- node->io_write(node->opaque);
+ QLIST_INSERT_HEAD(&ctx->poll_aio_handlers, node, node_poll);
+ }
+
+ if (!QLIST_IS_INSERTED(node, node_deleted) &&
+ (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) &&
+ aio_node_check(ctx, node->is_external) &&
+ node->io_read) {
+ node->io_read(node->opaque);
+
+ /* aio_notify() does not count as progress */
+ if (node->opaque != &ctx->notifier) {
progress = true;
}
+ }
+ if (!QLIST_IS_INSERTED(node, node_deleted) &&
+ (revents & (G_IO_OUT | G_IO_ERR)) &&
+ aio_node_check(ctx, node->is_external) &&
+ node->io_write) {
+ node->io_write(node->opaque);
+ progress = true;
+ }
- if (node->deleted) {
- if (qemu_lockcnt_dec_if_lock(&ctx->list_lock)) {
- QLIST_REMOVE(node, node);
- g_free(node);
- qemu_lockcnt_inc_and_unlock(&ctx->list_lock);
- }
- }
+ return progress;
+}
+
+/*
+ * If we have a list of ready handlers then this is more efficient than
+ * scanning all handlers with aio_dispatch_handlers().
+ */
+static bool aio_dispatch_ready_handlers(AioContext *ctx,
+ AioHandlerList *ready_list)
+{
+ bool progress = false;
+ AioHandler *node;
+
+ while ((node = QLIST_FIRST(ready_list))) {
+ QLIST_REMOVE(node, node_ready);
+ progress = aio_dispatch_handler(ctx, node) || progress;
+ }
+
+ return progress;
+}
+
+/* Slower than aio_dispatch_ready_handlers() but only used via glib */
+static bool aio_dispatch_handlers(AioContext *ctx)
+{
+ AioHandler *node, *tmp;
+ bool progress = false;
+
+ QLIST_FOREACH_SAFE_RCU(node, &ctx->aio_handlers, node, tmp) {
+ progress = aio_dispatch_handler(ctx, node) || progress;
}
return progress;
qemu_lockcnt_inc(&ctx->list_lock);
aio_bh_poll(ctx);
aio_dispatch_handlers(ctx);
+ aio_free_deleted_handlers(ctx);
qemu_lockcnt_dec(&ctx->list_lock);
timerlistgroup_run_timers(&ctx->tlg);
}
-/* These thread-local variables are used only in a small part of aio_poll
- * around the call to the poll() system call. In particular they are not
- * used while aio_poll is performing callbacks, which makes it much easier
- * to think about reentrancy!
- *
- * Stack-allocated arrays would be perfect but they have size limitations;
- * heap allocation is expensive enough that we want to reuse arrays across
- * calls to aio_poll(). And because poll() has to be called without holding
- * any lock, the arrays cannot be stored in AioContext. Thread-local data
- * has none of the disadvantages of these three options.
- */
-static __thread GPollFD *pollfds;
-static __thread AioHandler **nodes;
-static __thread unsigned npfd, nalloc;
-static __thread Notifier pollfds_cleanup_notifier;
-
-static void pollfds_cleanup(Notifier *n, void *unused)
+static bool run_poll_handlers_once(AioContext *ctx,
+ int64_t now,
+ int64_t *timeout)
{
- g_assert(npfd == 0);
- g_free(pollfds);
- g_free(nodes);
- nalloc = 0;
+ bool progress = false;
+ AioHandler *node;
+ AioHandler *tmp;
+
+ QLIST_FOREACH_SAFE(node, &ctx->poll_aio_handlers, node_poll, tmp) {
+ if (aio_node_check(ctx, node->is_external) &&
+ node->io_poll(node->opaque)) {
+ node->poll_idle_timeout = now + POLL_IDLE_INTERVAL_NS;
+
+ /*
+ * Polling was successful, exit try_poll_mode immediately
+ * to adjust the next polling time.
+ */
+ *timeout = 0;
+ if (node->opaque != &ctx->notifier) {
+ progress = true;
+ }
+ }
+
+ /* Caller handles freeing deleted nodes. Don't do it here. */
+ }
+
+ return progress;
}
-static void add_pollfd(AioHandler *node)
+static bool fdmon_supports_polling(AioContext *ctx)
{
- if (npfd == nalloc) {
- if (nalloc == 0) {
- pollfds_cleanup_notifier.notify = pollfds_cleanup;
- qemu_thread_atexit_add(&pollfds_cleanup_notifier);
- nalloc = 8;
- } else {
- g_assert(nalloc <= INT_MAX);
- nalloc *= 2;
- }
- pollfds = g_renew(GPollFD, pollfds, nalloc);
- nodes = g_renew(AioHandler *, nodes, nalloc);
- }
- nodes[npfd] = node;
- pollfds[npfd] = (GPollFD) {
- .fd = node->pfd.fd,
- .events = node->pfd.events,
- };
- npfd++;
+ return ctx->fdmon_ops->need_wait != aio_poll_disabled;
}
-static bool run_poll_handlers_once(AioContext *ctx)
+static bool remove_idle_poll_handlers(AioContext *ctx, int64_t now)
{
- bool progress = false;
AioHandler *node;
+ AioHandler *tmp;
+ bool progress = false;
- QLIST_FOREACH_RCU(node, &ctx->aio_handlers, node) {
- if (!node->deleted && node->io_poll &&
- aio_node_check(ctx, node->is_external) &&
- node->io_poll(node->opaque)) {
- progress = true;
- }
+ /*
+ * File descriptor monitoring implementations without userspace polling
+ * support suffer from starvation when a subset of handlers is polled
+ * because fds will not be processed in a timely fashion. Don't remove
+ * idle poll handlers.
+ */
+ if (!fdmon_supports_polling(ctx)) {
+ return false;
+ }
- /* Caller handles freeing deleted nodes. Don't do it here. */
+ QLIST_FOREACH_SAFE(node, &ctx->poll_aio_handlers, node_poll, tmp) {
+ if (node->poll_idle_timeout == 0LL) {
+ node->poll_idle_timeout = now + POLL_IDLE_INTERVAL_NS;
+ } else if (now >= node->poll_idle_timeout) {
+ trace_poll_remove(ctx, node, node->pfd.fd);
+ node->poll_idle_timeout = 0LL;
+ QLIST_SAFE_REMOVE(node, node_poll);
+ if (ctx->poll_started && node->io_poll_end) {
+ node->io_poll_end(node->opaque);
+
+ /*
+ * Final poll in case ->io_poll_end() races with an event.
+ * Nevermind about re-adding the handler in the rare case where
+ * this causes progress.
+ */
+ progress = node->io_poll(node->opaque) || progress;
+ }
+ }
}
return progress;
*
* Returns: true if progress was made, false otherwise
*/
-static bool run_poll_handlers(AioContext *ctx, int64_t max_ns)
+static bool run_poll_handlers(AioContext *ctx, int64_t max_ns, int64_t *timeout)
{
bool progress;
- int64_t end_time;
+ int64_t start_time, elapsed_time;
assert(ctx->notify_me);
assert(qemu_lockcnt_count(&ctx->list_lock) > 0);
- assert(ctx->poll_disable_cnt == 0);
- trace_run_poll_handlers_begin(ctx, max_ns);
+ trace_run_poll_handlers_begin(ctx, max_ns, *timeout);
- end_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + max_ns;
+ /*
+ * Optimization: ->io_poll() handlers often contain RCU read critical
+ * sections and we therefore see many rcu_read_lock() -> rcu_read_unlock()
+ * -> rcu_read_lock() -> ... sequences with expensive memory
+ * synchronization primitives. Make the entire polling loop an RCU
+ * critical section because nested rcu_read_lock()/rcu_read_unlock() calls
+ * are cheap.
+ */
+ RCU_READ_LOCK_GUARD();
+ start_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
do {
- progress = run_poll_handlers_once(ctx);
- } while (!progress && qemu_clock_get_ns(QEMU_CLOCK_REALTIME) < end_time);
+ progress = run_poll_handlers_once(ctx, start_time, timeout);
+ elapsed_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - start_time;
+ max_ns = qemu_soonest_timeout(*timeout, max_ns);
+ assert(!(max_ns && progress));
+ } while (elapsed_time < max_ns && !ctx->fdmon_ops->need_wait(ctx));
+
+ if (remove_idle_poll_handlers(ctx, start_time + elapsed_time)) {
+ *timeout = 0;
+ progress = true;
+ }
- trace_run_poll_handlers_end(ctx, progress);
+ /* If time has passed with no successful polling, adjust *timeout to
+ * keep the same ending time.
+ */
+ if (*timeout != -1) {
+ *timeout -= MIN(*timeout, elapsed_time);
+ }
+ trace_run_poll_handlers_end(ctx, progress, *timeout);
return progress;
}
/* try_poll_mode:
* @ctx: the AioContext
- * @blocking: busy polling is only attempted when blocking is true
+ * @timeout: timeout for blocking wait, computed by the caller and updated if
+ * polling succeeds.
*
* ctx->notify_me must be non-zero so this function can detect aio_notify().
*
*
* Returns: true if progress was made, false otherwise
*/
-static bool try_poll_mode(AioContext *ctx, bool blocking)
+static bool try_poll_mode(AioContext *ctx, int64_t *timeout)
{
- if (blocking && ctx->poll_max_ns && ctx->poll_disable_cnt == 0) {
- /* See qemu_soonest_timeout() uint64_t hack */
- int64_t max_ns = MIN((uint64_t)aio_compute_timeout(ctx),
- (uint64_t)ctx->poll_ns);
+ int64_t max_ns;
- if (max_ns) {
- poll_set_started(ctx, true);
+ if (QLIST_EMPTY_RCU(&ctx->poll_aio_handlers)) {
+ return false;
+ }
- if (run_poll_handlers(ctx, max_ns)) {
- return true;
- }
+ max_ns = qemu_soonest_timeout(*timeout, ctx->poll_ns);
+ if (max_ns && !ctx->fdmon_ops->need_wait(ctx)) {
+ poll_set_started(ctx, true);
+
+ if (run_poll_handlers(ctx, max_ns, timeout)) {
+ return true;
}
}
- poll_set_started(ctx, false);
+ if (poll_set_started(ctx, false)) {
+ *timeout = 0;
+ return true;
+ }
- /* Even if we don't run busy polling, try polling once in case it can make
- * progress and the caller will be able to avoid ppoll(2)/epoll_wait(2).
- */
- return run_poll_handlers_once(ctx);
+ return false;
}
bool aio_poll(AioContext *ctx, bool blocking)
{
- AioHandler *node;
- int i;
+ AioHandlerList ready_list = QLIST_HEAD_INITIALIZER(ready_list);
int ret = 0;
bool progress;
int64_t timeout;
int64_t start = 0;
+ /*
+ * There cannot be two concurrent aio_poll calls for the same AioContext (or
+ * an aio_poll concurrent with a GSource prepare/check/dispatch callback).
+ * We rely on this below to avoid slow locked accesses to ctx->notify_me.
+ */
+ assert(in_aio_context_home_thread(ctx));
+
/* aio_notify can avoid the expensive event_notifier_set if
* everything (file descriptors, bottom halves, timers) will
* be re-evaluated before the next blocking poll(). This is
* so disable the optimization now.
*/
if (blocking) {
- atomic_add(&ctx->notify_me, 2);
+ atomic_set(&ctx->notify_me, atomic_read(&ctx->notify_me) + 2);
+ /*
+ * Write ctx->notify_me before computing the timeout
+ * (reading bottom half flags, etc.). Pairs with
+ * smp_mb in aio_notify().
+ */
+ smp_mb();
}
qemu_lockcnt_inc(&ctx->list_lock);
start = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
}
- progress = try_poll_mode(ctx, blocking);
- if (!progress) {
- assert(npfd == 0);
+ timeout = blocking ? aio_compute_timeout(ctx) : 0;
+ progress = try_poll_mode(ctx, &timeout);
+ assert(!(timeout && progress));
- /* fill pollfds */
-
- if (!aio_epoll_enabled(ctx)) {
- QLIST_FOREACH_RCU(node, &ctx->aio_handlers, node) {
- if (!node->deleted && node->pfd.events
- && aio_node_check(ctx, node->is_external)) {
- add_pollfd(node);
- }
- }
- }
-
- timeout = blocking ? aio_compute_timeout(ctx) : 0;
-
- /* wait until next event */
- if (aio_epoll_check_poll(ctx, pollfds, npfd, timeout)) {
- AioHandler epoll_handler;
-
- epoll_handler.pfd.fd = ctx->epollfd;
- epoll_handler.pfd.events = G_IO_IN | G_IO_OUT | G_IO_HUP | G_IO_ERR;
- npfd = 0;
- add_pollfd(&epoll_handler);
- ret = aio_epoll(ctx, pollfds, npfd, timeout);
- } else {
- ret = qemu_poll_ns(pollfds, npfd, timeout);
- }
+ /* If polling is allowed, non-blocking aio_poll does not need the
+ * system call---a single round of run_poll_handlers_once suffices.
+ */
+ if (timeout || ctx->fdmon_ops->need_wait(ctx)) {
+ ret = ctx->fdmon_ops->wait(ctx, &ready_list, timeout);
}
if (blocking) {
- atomic_sub(&ctx->notify_me, 2);
+ /* Finish the poll before clearing the flag. */
+ atomic_store_release(&ctx->notify_me, atomic_read(&ctx->notify_me) - 2);
+ aio_notify_accept(ctx);
}
/* Adjust polling time */
}
}
- aio_notify_accept(ctx);
-
- /* if we have any readable fds, dispatch event */
- if (ret > 0) {
- for (i = 0; i < npfd; i++) {
- nodes[i]->pfd.revents = pollfds[i].revents;
- }
- }
-
- npfd = 0;
-
progress |= aio_bh_poll(ctx);
if (ret > 0) {
- progress |= aio_dispatch_handlers(ctx);
+ progress |= aio_dispatch_ready_handlers(ctx, &ready_list);
}
+ aio_free_deleted_handlers(ctx);
+
qemu_lockcnt_dec(&ctx->list_lock);
progress |= timerlistgroup_run_timers(&ctx->tlg);
void aio_context_setup(AioContext *ctx)
{
- /* TODO remove this in final patch submission */
- if (getenv("QEMU_AIO_POLL_MAX_NS")) {
- fprintf(stderr, "The QEMU_AIO_POLL_MAX_NS environment variable has "
- "been replaced with -object iothread,poll-max-ns=NUM\n");
- exit(1);
- }
-
-#ifdef CONFIG_EPOLL_CREATE1
- assert(!ctx->epollfd);
- ctx->epollfd = epoll_create1(EPOLL_CLOEXEC);
- if (ctx->epollfd == -1) {
- fprintf(stderr, "Failed to create epoll instance: %s", strerror(errno));
- ctx->epoll_available = false;
- } else {
- ctx->epoll_available = true;
+ ctx->fdmon_ops = &fdmon_poll_ops;
+ ctx->epollfd = -1;
+
+ /* Use the fastest fd monitoring implementation if available */
+ if (fdmon_io_uring_setup(ctx)) {
+ return;
}
-#endif
+
+ fdmon_epoll_setup(ctx);
+}
+
+void aio_context_destroy(AioContext *ctx)
+{
+ fdmon_io_uring_destroy(ctx);
+ fdmon_epoll_disable(ctx);
}
void aio_context_set_poll_params(AioContext *ctx, int64_t max_ns,
projects / qemu.git / blobdiff