1 files changed, 76 insertions, 54 deletions

diff --git a/block/io.c b/block/io.c
index 24205f5..95b1c56 100644
--- a/block/io.c
+++ b/block/io.c
@@ -754,55 +754,65 @@ static bool tracked_request_overlaps(BdrvTrackedRequest *req,
     return true;
 }
 
-static bool coroutine_fn
-bdrv_wait_serialising_requests_locked(BlockDriverState *bs,
-                                      BdrvTrackedRequest *self)
+/* Called with self->bs->reqs_lock held */
+static BdrvTrackedRequest *
+bdrv_find_conflicting_request(BdrvTrackedRequest *self)
 {
     BdrvTrackedRequest *req;
-    bool retry;
-    bool waited = false;
 
-    do {
-        retry = false;
-        QLIST_FOREACH(req, &bs->tracked_requests, list) {
-            if (req == self || (!req->serialising && !self->serialising)) {
-                continue;
-            }
-            if (tracked_request_overlaps(req, self->overlap_offset,
-                                         self->overlap_bytes))
-            {
-                /* Hitting this means there was a reentrant request, for
-                 * example, a block driver issuing nested requests.  This must
-                 * never happen since it means deadlock.
-                 */
-                assert(qemu_coroutine_self() != req->co);
-
-                /* If the request is already (indirectly) waiting for us, or
-                 * will wait for us as soon as it wakes up, then just go on
-                 * (instead of producing a deadlock in the former case). */
-                if (!req->waiting_for) {
-                    self->waiting_for = req;
-                    qemu_co_queue_wait(&req->wait_queue, &bs->reqs_lock);
-                    self->waiting_for = NULL;
-                    retry = true;
-                    waited = true;
-                    break;
-                }
+    QLIST_FOREACH(req, &self->bs->tracked_requests, list) {
+        if (req == self || (!req->serialising && !self->serialising)) {
+            continue;
+        }
+        if (tracked_request_overlaps(req, self->overlap_offset,
+                                     self->overlap_bytes))
+        {
+            /*
+             * Hitting this means there was a reentrant request, for
+             * example, a block driver issuing nested requests.  This must
+             * never happen since it means deadlock.
+             */
+            assert(qemu_coroutine_self() != req->co);
+
+            /*
+             * If the request is already (indirectly) waiting for us, or
+             * will wait for us as soon as it wakes up, then just go on
+             * (instead of producing a deadlock in the former case).
+             */
+            if (!req->waiting_for) {
+                return req;
             }
         }
-    } while (retry);
+    }
+
+    return NULL;
+}
+
+/* Called with self->bs->reqs_lock held */
+static bool coroutine_fn
+bdrv_wait_serialising_requests_locked(BdrvTrackedRequest *self)
+{
+    BdrvTrackedRequest *req;
+    bool waited = false;
+
+    while ((req = bdrv_find_conflicting_request(self))) {
+        self->waiting_for = req;
+        qemu_co_queue_wait(&req->wait_queue, &self->bs->reqs_lock);
+        self->waiting_for = NULL;
+        waited = true;
+    }
+
     return waited;
 }
 
-bool bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align)
+/* Called with req->bs->reqs_lock held */
+static void tracked_request_set_serialising(BdrvTrackedRequest *req,
+                                            uint64_t align)
 {
-    BlockDriverState *bs = req->bs;
     int64_t overlap_offset = req->offset & ~(align - 1);
     uint64_t overlap_bytes = ROUND_UP(req->offset + req->bytes, align)
                                - overlap_offset;
-    bool waited;
 
-    qemu_co_mutex_lock(&bs->reqs_lock);
     if (!req->serialising) {
         qatomic_inc(&req->bs->serialising_in_flight);
         req->serialising = true;
@@ -810,9 +820,6 @@ bool bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align)
 
     req->overlap_offset = MIN(req->overlap_offset, overlap_offset);
     req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes);
-    waited = bdrv_wait_serialising_requests_locked(bs, req);
-    qemu_co_mutex_unlock(&bs->reqs_lock);
-    return waited;
 }
 
 /**
@@ -892,12 +899,27 @@ static bool coroutine_fn bdrv_wait_serialising_requests(BdrvTrackedRequest *self
     }
 
     qemu_co_mutex_lock(&bs->reqs_lock);
-    waited = bdrv_wait_serialising_requests_locked(bs, self);
+    waited = bdrv_wait_serialising_requests_locked(self);
     qemu_co_mutex_unlock(&bs->reqs_lock);
 
     return waited;
 }
 
+bool coroutine_fn bdrv_make_request_serialising(BdrvTrackedRequest *req,
+                                                uint64_t align)
+{
+    bool waited;
+
+    qemu_co_mutex_lock(&req->bs->reqs_lock);
+
+    tracked_request_set_serialising(req, align);
+    waited = bdrv_wait_serialising_requests_locked(req);
+
+    qemu_co_mutex_unlock(&req->bs->reqs_lock);
+
+    return waited;
+}
+
 int bdrv_check_request(int64_t offset, int64_t bytes)
 {
     if (offset < 0 || bytes < 0) {
@@ -1423,7 +1445,7 @@ static int coroutine_fn bdrv_aligned_preadv(BdrvChild *child,
          * with each other for the same cluster.  For example, in copy-on-read
          * it ensures that the CoR read and write operations are atomic and
          * guest writes cannot interleave between them. */
-        bdrv_mark_request_serialising(req, bdrv_get_cluster_size(bs));
+        bdrv_make_request_serialising(req, bdrv_get_cluster_size(bs));
     } else {
         bdrv_wait_serialising_requests(req);
     }
@@ -1827,7 +1849,6 @@ bdrv_co_write_req_prepare(BdrvChild *child, int64_t offset, uint64_t bytes,
                           BdrvTrackedRequest *req, int flags)
 {
     BlockDriverState *bs = child->bs;
-    bool waited;
     int64_t end_sector = DIV_ROUND_UP(offset + bytes, BDRV_SECTOR_SIZE);
 
     if (bs->read_only) {
@@ -1837,17 +1858,18 @@ bdrv_co_write_req_prepare(BdrvChild *child, int64_t offset, uint64_t bytes,
     assert(!(bs->open_flags & BDRV_O_INACTIVE));
     assert((bs->open_flags & BDRV_O_NO_IO) == 0);
     assert(!(flags & ~BDRV_REQ_MASK));
+    assert(!((flags & BDRV_REQ_NO_WAIT) && !(flags & BDRV_REQ_SERIALISING)));
 
     if (flags & BDRV_REQ_SERIALISING) {
-        waited = bdrv_mark_request_serialising(req, bdrv_get_cluster_size(bs));
-        /*
-         * For a misaligned request we should have already waited earlier,
-         * because we come after bdrv_padding_rmw_read which must be called
-         * with the request already marked as serialising.
-         */
-        assert(!waited ||
-               (req->offset == req->overlap_offset &&
-                req->bytes == req->overlap_bytes));
+        QEMU_LOCK_GUARD(&bs->reqs_lock);
+
+        tracked_request_set_serialising(req, bdrv_get_cluster_size(bs));
+
+        if ((flags & BDRV_REQ_NO_WAIT) && bdrv_find_conflicting_request(req)) {
+            return -EBUSY;
+        }
+
+        bdrv_wait_serialising_requests_locked(req);
     } else {
         bdrv_wait_serialising_requests(req);
     }
@@ -2013,7 +2035,7 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child,
 
     padding = bdrv_init_padding(bs, offset, bytes, &pad);
     if (padding) {
-        bdrv_mark_request_serialising(req, align);
+        bdrv_make_request_serialising(req, align);
 
         bdrv_padding_rmw_read(child, req, &pad, true);
 
@@ -2127,7 +2149,7 @@ int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
     }
 
     if (bdrv_pad_request(bs, &qiov, &qiov_offset, &offset, &bytes, &pad)) {
-        bdrv_mark_request_serialising(&req, align);
+        bdrv_make_request_serialising(&req, align);
         bdrv_padding_rmw_read(child, &req, &pad, false);
     }
 
@@ -3248,7 +3270,7 @@ int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
      * new area, we need to make sure that no write requests are made to it
      * concurrently or they might be overwritten by preallocation. */
     if (new_bytes) {
-        bdrv_mark_request_serialising(&req, 1);
+        bdrv_make_request_serialising(&req, 1);
     }
     if (bs->read_only) {
         error_setg(errp, "Image is read-only");