Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging

Block layer patches

- virtio-blk: use blk_io_plug_call() instead of notification BH
- mirror: allow switching from background to active mode
- qemu-img rebase: add compression support
- Fix locking in media change monitor commands
- Fix a few blockjob-related deadlocks when using iothread

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEE3D3rFZqa+V09dFb+fwmycsiPL9YFAmVBTkERHGt3b2xmQHJl
# ZGhhdC5jb20ACgkQfwmycsiPL9ZiqRAAqvsWbblmEGJ7TBKYQK3f8QshJ66RxzbC
# 4eSjKHrciWNTeeIeU8r8OvFcPPoTcPXxpcmasD2gsAxG5W5N8vkPbBkW+YT4YdDJ
# pWJXrbJ15nILC4DmnR1ARVtvxKgv9zy5LSm5bjss1K+OSYJl/nx+ILjmfVZnYDF7
# z1dP/G0JxKKm4JzAIdBE3uZS+6Q5kx/wGYlJv8EQmlH3DYfsJfy6Lthe9jfw8ijg
# lSqLoQ+D0lEd6Bk4XbkUqqBxFcYBWTfU6qPZoyIO94zCTwTG9yIjmoivxmmfwQZq
# cJUTGGZjcxpJYnvcC6P13WgcWBtcD9L2kYFVH0JyjpwcSg9cCGHMF66n9pSlyEGq
# DUikwVzbTwOotwzYQyM88v4ET+2+Qdcwn8pRbv9PllEczh0kAsUAEuxSgtz4NEcN
# bZrap/16xHFybNOKkMZcmpqxspT5NXKbDODUP0IvbSYMOYpWS983nBTxwMRpyHog
# 2TFDZu4DjNiPkI2BcYM5VOKk6diNowZFShcEKvoaOLX/n9EBhP0tjoH9VUn1800F
# myHrhF2jpIf9GhErMWB7N2W3/0aK0pqdQgbpVnd1ARDdIdYkr7G/S+50D9K80b6n
# 0q2E7br4S5bcsY0HQzBL9YARSayY+lVOssLoolCWEsYzijdBQmAvs5THajFKcism
# /idI6nlp2Vs=
# =RdxS
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 01 Nov 2023 03:58:09 JST
# gpg:                using RSA key DC3DEB159A9AF95D3D7456FE7F09B272C88F2FD6
# gpg:                issuer "kwolf@redhat.com"
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* tag 'for-upstream' of https://repo.or.cz/qemu/kevin: (27 commits)
  iotests: add test for changing mirror's copy_mode
  mirror: return mirror-specific information upon query
  blockjob: query driver-specific info via a new 'query' driver method
  qapi/block-core: turn BlockJobInfo into a union
  qapi/block-core: use JobType for BlockJobInfo's type
  mirror: implement mirror_change method
  block/mirror: determine copy_to_target only once
  block/mirror: move dirty bitmap to filter
  block/mirror: set actively_synced even after the job is ready
  blockjob: introduce block-job-change QMP command
  virtio-blk: remove batch notification BH
  virtio: use defer_call() in virtio_irqfd_notify()
  util/defer-call: move defer_call() to util/
  block: rename blk_io_plug_call() API to defer_call()
  blockdev: mirror: avoid potential deadlock when using iothread
  block: avoid potential deadlock during bdrv_graph_wrlock() in bdrv_close()
  blockjob: drop AioContext lock before calling bdrv_graph_wrlock()
  iotests: Test media change with iothreads
  block: Fix locking in media change monitor commands
  iotests: add tests for "qemu-img rebase" with compression
  ...

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

1 files changed, 156 insertions, 0 deletions

diff --git a/util/defer-call.c b/util/defer-call.c
new file mode 100644
index 0000000000..037dc0abf0
--- /dev/null
+++ b/util/defer-call.c
@@ -0,0 +1,156 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Deferred calls
+ *
+ * Copyright Red Hat.
+ *
+ * This API defers a function call within a defer_call_begin()/defer_call_end()
+ * section, allowing multiple calls to batch up. This is a performance
+ * optimization that is used in the block layer to submit several I/O requests
+ * at once instead of individually:
+ *
+ *   defer_call_begin(); <-- start of section
+ *   ...
+ *   defer_call(my_func, my_obj); <-- deferred my_func(my_obj) call
+ *   defer_call(my_func, my_obj); <-- another
+ *   defer_call(my_func, my_obj); <-- another
+ *   ...
+ *   defer_call_end(); <-- end of section, my_func(my_obj) is called once
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/coroutine-tls.h"
+#include "qemu/notify.h"
+#include "qemu/thread.h"
+#include "qemu/defer-call.h"
+
+/* A function call that has been deferred until defer_call_end() */
+typedef struct {
+    void (*fn)(void *);
+    void *opaque;
+} DeferredCall;
+
+/* Per-thread state */
+typedef struct {
+    unsigned nesting_level;
+    GArray *deferred_call_array;
+} DeferCallThreadState;
+
+/* Use get_ptr_defer_call_thread_state() to fetch this thread-local value */
+QEMU_DEFINE_STATIC_CO_TLS(DeferCallThreadState, defer_call_thread_state);
+
+/* Called at thread cleanup time */
+static void defer_call_atexit(Notifier *n, void *value)
+{
+    DeferCallThreadState *thread_state = get_ptr_defer_call_thread_state();
+    g_array_free(thread_state->deferred_call_array, TRUE);
+}
+
+/* This won't involve coroutines, so use __thread */
+static __thread Notifier defer_call_atexit_notifier;
+
+/**
+ * defer_call:
+ * @fn: a function pointer to be invoked
+ * @opaque: a user-defined argument to @fn()
+ *
+ * Call @fn(@opaque) immediately if not within a
+ * defer_call_begin()/defer_call_end() section.
+ *
+ * Otherwise defer the call until the end of the outermost
+ * defer_call_begin()/defer_call_end() section in this thread. If the same
+ * @fn/@opaque pair has already been deferred, it will only be called once upon
+ * defer_call_end() so that accumulated calls are batched into a single call.
+ *
+ * The caller must ensure that @opaque is not freed before @fn() is invoked.
+ */
+void defer_call(void (*fn)(void *), void *opaque)
+{
+    DeferCallThreadState *thread_state = get_ptr_defer_call_thread_state();
+
+    /* Call immediately if we're not deferring calls */
+    if (thread_state->nesting_level == 0) {
+        fn(opaque);
+        return;
+    }
+
+    GArray *array = thread_state->deferred_call_array;
+    if (!array) {
+        array = g_array_new(FALSE, FALSE, sizeof(DeferredCall));
+        thread_state->deferred_call_array = array;
+        defer_call_atexit_notifier.notify = defer_call_atexit;
+        qemu_thread_atexit_add(&defer_call_atexit_notifier);
+    }
+
+    DeferredCall *fns = (DeferredCall *)array->data;
+    DeferredCall new_fn = {
+        .fn = fn,
+        .opaque = opaque,
+    };
+
+    /*
+     * There won't be many, so do a linear search. If this becomes a bottleneck
+     * then a binary search (glib 2.62+) or different data structure could be
+     * used.
+     */
+    for (guint i = 0; i < array->len; i++) {
+        if (memcmp(&fns[i], &new_fn, sizeof(new_fn)) == 0) {
+            return; /* already exists */
+        }
+    }
+
+    g_array_append_val(array, new_fn);
+}
+
+/**
+ * defer_call_begin: Defer defer_call() functions until defer_call_end()
+ *
+ * defer_call_begin() and defer_call_end() are thread-local operations. The
+ * caller must ensure that each defer_call_begin() has a matching
+ * defer_call_end() in the same thread.
+ *
+ * Nesting is supported. defer_call() functions are only called at the
+ * outermost defer_call_end().
+ */
+void defer_call_begin(void)
+{
+    DeferCallThreadState *thread_state = get_ptr_defer_call_thread_state();
+
+    assert(thread_state->nesting_level < UINT32_MAX);
+
+    thread_state->nesting_level++;
+}
+
+/**
+ * defer_call_end: Run any pending defer_call() functions
+ *
+ * There must have been a matching defer_call_begin() call in the same thread
+ * prior to this defer_call_end() call.
+ */
+void defer_call_end(void)
+{
+    DeferCallThreadState *thread_state = get_ptr_defer_call_thread_state();
+
+    assert(thread_state->nesting_level > 0);
+
+    if (--thread_state->nesting_level > 0) {
+        return;
+    }
+
+    GArray *array = thread_state->deferred_call_array;
+    if (!array) {
+        return;
+    }
+
+    DeferredCall *fns = (DeferredCall *)array->data;
+
+    for (guint i = 0; i < array->len; i++) {
+        fns[i].fn(fns[i].opaque);
+    }
+
+    /*
+     * This resets the array without freeing memory so that appending is cheap
+     * in the future.
+     */
+    g_array_set_size(array, 0);
+}