2 * QEMU aio implementation
4 * Copyright IBM Corp., 2008
5 * Copyright Red Hat Inc., 2012
8 * Anthony Liguori <aliguori@us.ibm.com>
9 * Paolo Bonzini <pbonzini@redhat.com>
11 * This work is licensed under the terms of the GNU GPL, version 2. See
12 * the COPYING file in the top-level directory.
14 * Contributions after 2012-01-13 are licensed under the terms of the
15 * GNU GPL, version 2 or (at your option) any later version.
18 #include "qemu/osdep.h"
19 #include "block/block.h"
20 #include "qemu/main-loop.h"
21 #include "qemu/queue.h"
22 #include "qemu/sockets.h"
23 #include "qapi/error.h"
24 #include "qemu/rcu_queue.h"
25 #include "qemu/error-report.h"
31 EventNotifierHandler
*io_notify
;
35 QLIST_ENTRY(AioHandler
) node
;
38 static void aio_remove_fd_handler(AioContext
*ctx
, AioHandler
*node
)
41 * If the GSource is in the process of being destroyed then
42 * g_source_remove_poll() causes an assertion failure. Skip
43 * removal in that case, because glib cleans up its state during
46 if (!g_source_is_destroyed(&ctx
->source
)) {
47 g_source_remove_poll(&ctx
->source
, &node
->pfd
);
50 /* If aio_poll is in progress, just mark the node as deleted */
51 if (qemu_lockcnt_count(&ctx
->list_lock
)) {
53 node
->pfd
.revents
= 0;
55 /* Otherwise, delete it for real. We can't just mark it as
56 * deleted because deleted nodes are only cleaned up after
57 * releasing the list_lock.
59 QLIST_REMOVE(node
, node
);
64 void aio_set_fd_handler(AioContext
*ctx
,
69 IOHandler
*io_poll_ready
,
73 AioHandler
*node
= NULL
;
76 if (!fd_is_socket(fd
)) {
77 error_report("fd=%d is not a socket, AIO implementation is missing", fd
);
81 s
= _get_osfhandle(fd
);
83 qemu_lockcnt_lock(&ctx
->list_lock
);
84 QLIST_FOREACH(old_node
, &ctx
->aio_handlers
, node
) {
85 if (old_node
->pfd
.fd
== s
&& !old_node
->deleted
) {
90 if (io_read
|| io_write
) {
94 /* Alloc and insert if it's not already there */
95 node
= g_new0(AioHandler
, 1);
100 node
->pfd
.events
|= G_IO_IN
;
102 if (node
->io_write
) {
103 node
->pfd
.events
|= G_IO_OUT
;
106 node
->e
= &ctx
->notifier
;
108 /* Update handler with latest information */
109 node
->opaque
= opaque
;
110 node
->io_read
= io_read
;
111 node
->io_write
= io_write
;
114 bitmask
|= FD_READ
| FD_ACCEPT
| FD_CLOSE
;
118 bitmask
|= FD_WRITE
| FD_CONNECT
;
121 QLIST_INSERT_HEAD_RCU(&ctx
->aio_handlers
, node
, node
);
122 event
= event_notifier_get_handle(&ctx
->notifier
);
123 qemu_socket_select(fd
, event
, bitmask
, NULL
);
126 aio_remove_fd_handler(ctx
, old_node
);
129 qemu_lockcnt_unlock(&ctx
->list_lock
);
133 void aio_set_event_notifier(AioContext
*ctx
,
135 EventNotifierHandler
*io_notify
,
137 EventNotifierHandler
*io_poll_ready
)
141 qemu_lockcnt_lock(&ctx
->list_lock
);
142 QLIST_FOREACH(node
, &ctx
->aio_handlers
, node
) {
143 if (node
->e
== e
&& !node
->deleted
) {
148 /* Are we deleting the fd handler? */
151 aio_remove_fd_handler(ctx
, node
);
155 /* Alloc and insert if it's not already there */
156 node
= g_new0(AioHandler
, 1);
158 node
->pfd
.fd
= (uintptr_t)event_notifier_get_handle(e
);
159 node
->pfd
.events
= G_IO_IN
;
160 QLIST_INSERT_HEAD_RCU(&ctx
->aio_handlers
, node
, node
);
162 g_source_add_poll(&ctx
->source
, &node
->pfd
);
164 /* Update handler with latest information */
165 node
->io_notify
= io_notify
;
168 qemu_lockcnt_unlock(&ctx
->list_lock
);
172 void aio_set_event_notifier_poll(AioContext
*ctx
,
173 EventNotifier
*notifier
,
174 EventNotifierHandler
*io_poll_begin
,
175 EventNotifierHandler
*io_poll_end
)
177 /* Not implemented */
180 bool aio_prepare(AioContext
*ctx
)
182 static struct timeval tv0
;
184 bool have_select_revents
= false;
188 * We have to walk very carefully in case aio_set_fd_handler is
189 * called while we're walking.
191 qemu_lockcnt_inc(&ctx
->list_lock
);
196 QLIST_FOREACH_RCU(node
, &ctx
->aio_handlers
, node
) {
198 FD_SET ((SOCKET
)node
->pfd
.fd
, &rfds
);
200 if (node
->io_write
) {
201 FD_SET ((SOCKET
)node
->pfd
.fd
, &wfds
);
205 if (select(0, &rfds
, &wfds
, NULL
, &tv0
) > 0) {
206 QLIST_FOREACH_RCU(node
, &ctx
->aio_handlers
, node
) {
207 node
->pfd
.revents
= 0;
208 if (FD_ISSET(node
->pfd
.fd
, &rfds
)) {
209 node
->pfd
.revents
|= G_IO_IN
;
210 have_select_revents
= true;
213 if (FD_ISSET(node
->pfd
.fd
, &wfds
)) {
214 node
->pfd
.revents
|= G_IO_OUT
;
215 have_select_revents
= true;
220 qemu_lockcnt_dec(&ctx
->list_lock
);
221 return have_select_revents
;
224 bool aio_pending(AioContext
*ctx
)
230 * We have to walk very carefully in case aio_set_fd_handler is
231 * called while we're walking.
233 qemu_lockcnt_inc(&ctx
->list_lock
);
234 QLIST_FOREACH_RCU(node
, &ctx
->aio_handlers
, node
) {
235 if (node
->pfd
.revents
&& node
->io_notify
) {
240 if ((node
->pfd
.revents
& G_IO_IN
) && node
->io_read
) {
244 if ((node
->pfd
.revents
& G_IO_OUT
) && node
->io_write
) {
250 qemu_lockcnt_dec(&ctx
->list_lock
);
254 static bool aio_dispatch_handlers(AioContext
*ctx
, HANDLE event
)
257 bool progress
= false;
261 * We have to walk very carefully in case aio_set_fd_handler is
262 * called while we're walking.
264 QLIST_FOREACH_SAFE_RCU(node
, &ctx
->aio_handlers
, node
, tmp
) {
265 int revents
= node
->pfd
.revents
;
267 if (!node
->deleted
&&
268 (revents
|| event_notifier_get_handle(node
->e
) == event
) &&
270 node
->pfd
.revents
= 0;
271 node
->io_notify(node
->e
);
273 /* aio_notify() does not count as progress */
274 if (node
->e
!= &ctx
->notifier
) {
279 if (!node
->deleted
&&
280 (node
->io_read
|| node
->io_write
)) {
281 node
->pfd
.revents
= 0;
282 if ((revents
& G_IO_IN
) && node
->io_read
) {
283 node
->io_read(node
->opaque
);
286 if ((revents
& G_IO_OUT
) && node
->io_write
) {
287 node
->io_write(node
->opaque
);
291 /* if the next select() will return an event, we have progressed */
292 if (event
== event_notifier_get_handle(&ctx
->notifier
)) {
294 WSAEnumNetworkEvents(node
->pfd
.fd
, event
, &ev
);
295 if (ev
.lNetworkEvents
) {
302 if (qemu_lockcnt_dec_if_lock(&ctx
->list_lock
)) {
303 QLIST_REMOVE(node
, node
);
305 qemu_lockcnt_inc_and_unlock(&ctx
->list_lock
);
313 void aio_dispatch(AioContext
*ctx
)
315 qemu_lockcnt_inc(&ctx
->list_lock
);
317 aio_dispatch_handlers(ctx
, INVALID_HANDLE_VALUE
);
318 qemu_lockcnt_dec(&ctx
->list_lock
);
319 timerlistgroup_run_timers(&ctx
->tlg
);
322 bool aio_poll(AioContext
*ctx
, bool blocking
)
325 HANDLE events
[MAXIMUM_WAIT_OBJECTS
];
326 bool progress
, have_select_revents
, first
;
331 * There cannot be two concurrent aio_poll calls for the same AioContext (or
332 * an aio_poll concurrent with a GSource prepare/check/dispatch callback).
333 * We rely on this below to avoid slow locked accesses to ctx->notify_me.
335 * aio_poll() may only be called in the AioContext's thread. iohandler_ctx
336 * is special in that it runs in the main thread, but that thread's context
337 * is qemu_aio_context.
339 assert(in_aio_context_home_thread(ctx
== iohandler_get_aio_context() ?
340 qemu_get_aio_context() : ctx
));
343 /* aio_notify can avoid the expensive event_notifier_set if
344 * everything (file descriptors, bottom halves, timers) will
345 * be re-evaluated before the next blocking poll(). This is
346 * already true when aio_poll is called with blocking == false;
347 * if blocking == true, it is only true after poll() returns,
348 * so disable the optimization now.
351 qatomic_set(&ctx
->notify_me
, qatomic_read(&ctx
->notify_me
) + 2);
353 * Write ctx->notify_me before computing the timeout
354 * (reading bottom half flags, etc.). Pairs with
355 * smp_mb in aio_notify().
360 qemu_lockcnt_inc(&ctx
->list_lock
);
361 have_select_revents
= aio_prepare(ctx
);
365 QLIST_FOREACH_RCU(node
, &ctx
->aio_handlers
, node
) {
366 if (!node
->deleted
&& node
->io_notify
) {
367 assert(count
< MAXIMUM_WAIT_OBJECTS
);
368 events
[count
++] = event_notifier_get_handle(node
->e
);
374 /* ctx->notifier is always registered. */
377 /* Multiple iterations, all of them non-blocking except the first,
378 * may be necessary to process all pending events. After the first
379 * WaitForMultipleObjects call ctx->notify_me will be decremented.
385 timeout
= blocking
&& !have_select_revents
386 ? qemu_timeout_ns_to_ms(aio_compute_timeout(ctx
)) : 0;
387 ret
= WaitForMultipleObjects(count
, events
, FALSE
, timeout
);
390 qatomic_store_release(&ctx
->notify_me
,
391 qatomic_read(&ctx
->notify_me
) - 2);
392 aio_notify_accept(ctx
);
396 progress
|= aio_bh_poll(ctx
);
400 /* if we have any signaled events, dispatch event */
402 if ((DWORD
) (ret
- WAIT_OBJECT_0
) < count
) {
403 event
= events
[ret
- WAIT_OBJECT_0
];
404 events
[ret
- WAIT_OBJECT_0
] = events
[--count
];
405 } else if (!have_select_revents
) {
409 have_select_revents
= false;
412 progress
|= aio_dispatch_handlers(ctx
, event
);
415 qemu_lockcnt_dec(&ctx
->list_lock
);
417 progress
|= timerlistgroup_run_timers(&ctx
->tlg
);
421 void aio_context_setup(AioContext
*ctx
)
425 void aio_context_destroy(AioContext
*ctx
)
429 void aio_context_use_g_source(AioContext
*ctx
)
433 void aio_context_set_poll_params(AioContext
*ctx
, int64_t max_ns
,
434 int64_t grow
, int64_t shrink
, Error
**errp
)
437 error_setg(errp
, "AioContext polling is not implemented on Windows");
441 void aio_context_set_aio_params(AioContext
*ctx
, int64_t max_batch
)