X-Git-Url: https://repo.jachan.dev/qemu.git/blobdiff_plain/b5d54bd42158b90b239bb6ce9c13072eb3a53fd2..453f842bc4cab49f10c267cff9ad3cf657265d49:/aio-posix.c

diff --git a/aio-posix.c b/aio-posix.c
index bd06f33c78..0467f23a63 100644
--- a/aio-posix.c
+++ b/aio-posix.c
@@ -24,8 +24,8 @@ struct AioHandler
     IOHandler *io_read;
     IOHandler *io_write;
     int deleted;
-    int pollfds_idx;
     void *opaque;
+    bool is_external;
     QLIST_ENTRY(AioHandler) node;
 };
 
@@ -44,6 +44,7 @@ static AioHandler *find_aio_handler(AioContext *ctx, int fd)
 
 void aio_set_fd_handler(AioContext *ctx,
                         int fd,
+                        bool is_external,
                         IOHandler *io_read,
                         IOHandler *io_write,
                         void *opaque)
@@ -73,7 +74,7 @@ void aio_set_fd_handler(AioContext *ctx,
     } else {
         if (node == NULL) {
             /* Alloc and insert if it's not already there */
-            node = g_malloc0(sizeof(AioHandler));
+            node = g_new0(AioHandler, 1);
             node->pfd.fd = fd;
             QLIST_INSERT_HEAD(&ctx->aio_handlers, node, node);
 
@@ -83,7 +84,7 @@ void aio_set_fd_handler(AioContext *ctx,
         node->io_read = io_read;
         node->io_write = io_write;
         node->opaque = opaque;
-        node->pollfds_idx = -1;
+        node->is_external = is_external;
 
         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);
@@ -94,10 +95,16 @@ void aio_set_fd_handler(AioContext *ctx,
 
 void aio_set_event_notifier(AioContext *ctx,
                             EventNotifier *notifier,
+                            bool is_external,
                             EventNotifierHandler *io_read)
 {
     aio_set_fd_handler(ctx, event_notifier_get_fd(notifier),
-                       (IOHandler *)io_read, NULL, notifier);
+                       is_external, (IOHandler *)io_read, NULL, notifier);
+}
+
+bool aio_prepare(AioContext *ctx)
+{
+    return false;
 }
 
 bool aio_pending(AioContext *ctx)
@@ -119,13 +126,22 @@ bool aio_pending(AioContext *ctx)
     return false;
 }
 
-static bool aio_dispatch(AioContext *ctx)
+bool aio_dispatch(AioContext *ctx)
 {
     AioHandler *node;
     bool progress = false;
 
     /*
-     * We have to walk very carefully in case qemu_aio_set_fd_handler is
+     * If there are callbacks left that have been queued, we need to call them.
+     * Do not call select in this case, because it is possible that the caller
+     * does not need a complete flush (as is the case for aio_poll loops).
+     */
+    if (aio_bh_poll(ctx)) {
+        progress = true;
+    }
+
+    /*
+     * We have to walk very carefully in case aio_set_fd_handler is
      * called while we're walking.
      */
     node = QLIST_FIRST(&ctx->aio_handlers);
@@ -172,76 +188,117 @@ static bool aio_dispatch(AioContext *ctx)
     return progress;
 }
 
+/* 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)
+{
+    g_assert(npfd == 0);
+    g_free(pollfds);
+    g_free(nodes);
+    nalloc = 0;
+}
+
+static void add_pollfd(AioHandler *node)
+{
+    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++;
+}
+
 bool aio_poll(AioContext *ctx, bool blocking)
 {
     AioHandler *node;
-    int ret;
+    int i, ret;
     bool progress;
+    int64_t timeout;
 
+    aio_context_acquire(ctx);
     progress = false;
 
-    /*
-     * If there are callbacks left that have been queued, we need to call them.
-     * Do not call select in this case, because it is possible that the caller
-     * does not need a complete flush (as is the case for qemu_aio_wait loops).
+    /* 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
+     * already true when aio_poll is called with blocking == false;
+     * if blocking == true, it is only true after poll() returns,
+     * so disable the optimization now.
      */
-    if (aio_bh_poll(ctx)) {
-        blocking = false;
-        progress = true;
-    }
-
-    if (aio_dispatch(ctx)) {
-        progress = true;
-    }
-
-    if (progress && !blocking) {
-        return true;
+    if (blocking) {
+        atomic_add(&ctx->notify_me, 2);
     }
 
     ctx->walking_handlers++;
 
-    g_array_set_size(ctx->pollfds, 0);
+    assert(npfd == 0);
 
     /* fill pollfds */
     QLIST_FOREACH(node, &ctx->aio_handlers, node) {
-        node->pollfds_idx = -1;
-        if (!node->deleted && node->pfd.events) {
-            GPollFD pfd = {
-                .fd = node->pfd.fd,
-                .events = node->pfd.events,
-            };
-            node->pollfds_idx = ctx->pollfds->len;
-            g_array_append_val(ctx->pollfds, pfd);
+        if (!node->deleted && node->pfd.events
+            && aio_node_check(ctx, node->is_external)) {
+            add_pollfd(node);
         }
     }
 
-    ctx->walking_handlers--;
+    timeout = blocking ? aio_compute_timeout(ctx) : 0;
 
-    /* early return if we only have the aio_notify() fd */
-    if (ctx->pollfds->len == 1) {
-        return progress;
+    /* wait until next event */
+    if (timeout) {
+        aio_context_release(ctx);
+    }
+    ret = qemu_poll_ns((GPollFD *)pollfds, npfd, timeout);
+    if (blocking) {
+        atomic_sub(&ctx->notify_me, 2);
+    }
+    if (timeout) {
+        aio_context_acquire(ctx);
     }
 
-    /* wait until next event */
-    ret = qemu_poll_ns((GPollFD *)ctx->pollfds->data,
-                         ctx->pollfds->len,
-                         blocking ? timerlistgroup_deadline_ns(&ctx->tlg) : 0);
+    aio_notify_accept(ctx);
 
     /* if we have any readable fds, dispatch event */
     if (ret > 0) {
-        QLIST_FOREACH(node, &ctx->aio_handlers, node) {
-            if (node->pollfds_idx != -1) {
-                GPollFD *pfd = &g_array_index(ctx->pollfds, GPollFD,
-                                              node->pollfds_idx);
-                node->pfd.revents = pfd->revents;
-            }
+        for (i = 0; i < npfd; i++) {
+            nodes[i]->pfd.revents = pollfds[i].revents;
         }
     }
 
+    npfd = 0;
+    ctx->walking_handlers--;
+
     /* Run dispatch even if there were no readable fds to run timers */
     if (aio_dispatch(ctx)) {
         progress = true;
     }
 
+    aio_context_release(ctx);
+
     return progress;
 }