Merge remote-tracking branch 'stefanha/block' into staging

# By Stefan Hajnoczi (14) and others
# Via Stefan Hajnoczi
* stefanha/block: (28 commits)
  blockdev: Fix up copyright and permission notice
  qemu-iotests: use -nographic in test case 007
  qemu-iotests: add tests for rebasing zero clusters
  dataplane: fix hang introduced by AioContext transition
  coroutine: use AioContext for CoQueue BH
  threadpool: drop global thread pool
  block: add bdrv_get_aio_context()
  aio: add a ThreadPool instance to AioContext
  threadpool: add thread_pool_new() and thread_pool_free()
  threadpool: move globals into struct ThreadPool
  main-loop: add qemu_get_aio_context()
  sheepdog: set io_flush handler in do_co_req
  sheepdog: use non-blocking fd in coroutine context
  qcow2: make is_allocated return true for zero clusters
  qcow2: drop unnecessary flush in qcow2_update_snapshot_refcount()
  qcow2: drop flush in update_cluster_refcount()
  qcow2: flush in qcow2_update_snapshot_refcount()
  qcow2: set L2 cache dependency in qcow2_alloc_bytes()
  qcow2: flush refcount cache correctly in qcow2_write_snapshots()
  qcow2: flush refcount cache correctly in alloc_refcount_block()
  ...

async.c

@@ -24,6 +24,7 @@
 
 #include "qemu-common.h"
 #include "block/aio.h"
+#include "block/thread-pool.h"
 #include "qemu/main-loop.h"
 
 /***********************************************************/
@@ -172,6 +173,7 @@ aio_ctx_finalize(GSource     *source)
 {
     AioContext *ctx = (AioContext *) source;
 
+    thread_pool_free(ctx->thread_pool);
     aio_set_event_notifier(ctx, &ctx->notifier, NULL, NULL);
     event_notifier_cleanup(&ctx->notifier);
     g_array_free(ctx->pollfds, TRUE);
@@ -190,6 +192,14 @@ GSource *aio_get_g_source(AioContext *ctx)
     return &ctx->source;
 }
 
+ThreadPool *aio_get_thread_pool(AioContext *ctx)
+{
+    if (!ctx->thread_pool) {
+        ctx->thread_pool = thread_pool_new(ctx);
+    }
+    return ctx->thread_pool;
+}
+
 void aio_notify(AioContext *ctx)
 {
     event_notifier_set(&ctx->notifier);
@@ -200,6 +210,7 @@ AioContext *aio_context_new(void)
     AioContext *ctx;
     ctx = (AioContext *) g_source_new(&aio_source_funcs, sizeof(AioContext));
     ctx->pollfds = g_array_new(FALSE, FALSE, sizeof(GPollFD));
+    ctx->thread_pool = NULL;
     event_notifier_init(&ctx->notifier, false);
     aio_set_event_notifier(ctx, &ctx->notifier, 
                            (EventNotifierHandler *)

block.c

@@ -665,15 +665,18 @@ static int bdrv_open_flags(BlockDriverState *bs, int flags)
 
 /*
  * Common part for opening disk images and files
+ *
+ * Removes all processed options from *options.
  */
 static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file,
-    const char *filename,
+    const char *filename, QDict *options,
     int flags, BlockDriver *drv)
 {
     int ret, open_flags;
 
     assert(drv != NULL);
     assert(bs->file == NULL);
+    assert(options == NULL || bs->options != options);
 
     trace_bdrv_open_common(bs, filename, flags, drv->format_name);
 
@@ -710,7 +713,7 @@ static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file,
     } else {
         assert(file != NULL);
         bs->file = file;
-        ret = drv->bdrv_open(bs, open_flags);
+        ret = drv->bdrv_open(bs, options, open_flags);
     }
 
     if (ret < 0) {
@@ -752,7 +755,7 @@ int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
     }
 
     bs = bdrv_new("");
-    ret = bdrv_open_common(bs, NULL, filename, flags, drv);
+    ret = bdrv_open_common(bs, NULL, filename, NULL, flags, drv);
     if (ret < 0) {
         bdrv_delete(bs);
         return ret;
@@ -788,7 +791,8 @@ int bdrv_open_backing_file(BlockDriverState *bs)
     /* backing files always opened read-only */
     back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT);
 
-    ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
+    ret = bdrv_open(bs->backing_hd, backing_filename, NULL,
+                    back_flags, back_drv);
     if (ret < 0) {
         bdrv_delete(bs->backing_hd);
         bs->backing_hd = NULL;
@@ -800,15 +804,29 @@ int bdrv_open_backing_file(BlockDriverState *bs)
 
 /*
  * Opens a disk image (raw, qcow2, vmdk, ...)
+ *
+ * options is a QDict of options to pass to the block drivers, or NULL for an
+ * empty set of options. The reference to the QDict belongs to the block layer
+ * after the call (even on failure), so if the caller intends to reuse the
+ * dictionary, it needs to use QINCREF() before calling bdrv_open.
  */
-int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
-              BlockDriver *drv)
+int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options,
+              int flags, BlockDriver *drv)
 {
     int ret;
     /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
     char tmp_filename[PATH_MAX + 1];
     BlockDriverState *file = NULL;
 
+    /* NULL means an empty set of options */
+    if (options == NULL) {
+        options = qdict_new();
+    }
+
+    bs->options = options;
+    options = qdict_clone_shallow(options);
+
+    /* For snapshot=on, create a temporary qcow2 overlay */
     if (flags & BDRV_O_SNAPSHOT) {
         BlockDriverState *bs1;
         int64_t total_size;
@@ -822,10 +840,10 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
 
         /* if there is a backing file, use it */
         bs1 = bdrv_new("");
-        ret = bdrv_open(bs1, filename, 0, drv);
+        ret = bdrv_open(bs1, filename, NULL, 0, drv);
         if (ret < 0) {
             bdrv_delete(bs1);
-            return ret;
+            goto fail;
         }
         total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
 
@@ -836,15 +854,17 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
 
         ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename));
         if (ret < 0) {
-            return ret;
+            goto fail;
         }
 
         /* Real path is meaningless for protocols */
-        if (is_protocol)
+        if (is_protocol) {
             snprintf(backing_filename, sizeof(backing_filename),
                      "%s", filename);
-        else if (!realpath(filename, backing_filename))
-            return -errno;
+        } else if (!realpath(filename, backing_filename)) {
+            ret = -errno;
+            goto fail;
+        }
 
         bdrv_qcow2 = bdrv_find_format("qcow2");
         options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
@@ -859,7 +879,7 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
         ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
         free_option_parameters(options);
         if (ret < 0) {
-            return ret;
+            goto fail;
         }
 
         filename = tmp_filename;
@@ -874,7 +894,7 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
 
     ret = bdrv_file_open(&file, filename, bdrv_open_flags(bs, flags));
     if (ret < 0) {
-        return ret;
+        goto fail;
     }
 
     /* Find the right image format driver */
@@ -887,7 +907,7 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
     }
 
     /* Open the image */
-    ret = bdrv_open_common(bs, file, filename, flags, drv);
+    ret = bdrv_open_common(bs, file, filename, options, flags, drv);
     if (ret < 0) {
         goto unlink_and_fail;
     }
@@ -901,11 +921,22 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
     if ((flags & BDRV_O_NO_BACKING) == 0) {
         ret = bdrv_open_backing_file(bs);
         if (ret < 0) {
-            bdrv_close(bs);
-            return ret;
+            goto close_and_fail;
         }
     }
 
+    /* Check if any unknown options were used */
+    if (qdict_size(options) != 0) {
+        const QDictEntry *entry = qdict_first(options);
+        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by "
+            "device '%s' doesn't support the option '%s'",
+            drv->format_name, bs->device_name, entry->key);
+
+        ret = -EINVAL;
+        goto close_and_fail;
+    }
+    QDECREF(options);
+
     if (!bdrv_key_required(bs)) {
         bdrv_dev_change_media_cb(bs, true);
     }
@@ -924,6 +955,15 @@ unlink_and_fail:
     if (bs->is_temporary) {
         unlink(filename);
     }
+fail:
+    QDECREF(bs->options);
+    QDECREF(options);
+    bs->options = NULL;
+    return ret;
+
+close_and_fail:
+    bdrv_close(bs);
+    QDECREF(options);
     return ret;
 }
 
@@ -1193,6 +1233,8 @@ void bdrv_close(BlockDriverState *bs)
         bs->valid_key = 0;
         bs->sg = 0;
         bs->growable = 0;
+        QDECREF(bs->options);
+        bs->options = NULL;
 
         if (bs->file != NULL) {
             bdrv_delete(bs->file);
@@ -3190,7 +3232,7 @@ int bdrv_snapshot_goto(BlockDriverState *bs,
     if (bs->file) {
         drv->bdrv_close(bs);
         ret = bdrv_snapshot_goto(bs->file, snapshot_id);
-        open_ret = drv->bdrv_open(bs, bs->open_flags);
+        open_ret = drv->bdrv_open(bs, NULL, bs->open_flags);
         if (open_ret < 0) {
             bdrv_delete(bs->file);
             bs->drv = NULL;
@@ -4594,7 +4636,8 @@ void bdrv_img_create(const char *filename, const char *fmt,
 
             bs = bdrv_new("");
 
-            ret = bdrv_open(bs, backing_file->value.s, back_flags, backing_drv);
+            ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags,
+                            backing_drv);
             if (ret < 0) {
                 error_setg_errno(errp, -ret, "Could not open '%s'",
                                  backing_file->value.s);
@@ -4638,3 +4681,9 @@ out:
         bdrv_delete(bs);
     }
 }
+
+AioContext *bdrv_get_aio_context(BlockDriverState *bs)
+{
+    /* Currently BlockDriverState always uses the main loop AioContext */
+    return qemu_get_aio_context();
+}

block/blkverify.c

@@ -98,7 +98,7 @@ static int blkverify_open(BlockDriverState *bs, const char *filename, int flags)
 
     /* Open the test file */
     s->test_file = bdrv_new("");
-    ret = bdrv_open(s->test_file, filename, flags, NULL);
+    ret = bdrv_open(s->test_file, filename, NULL, flags, NULL);
     if (ret < 0) {
         bdrv_delete(s->test_file);
         s->test_file = NULL;

block/bochs.c

@@ -108,7 +108,7 @@ static int bochs_probe(const uint8_t *buf, int buf_size, const char *filename)
     return 0;
 }
 
-static int bochs_open(BlockDriverState *bs, int flags)
+static int bochs_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVBochsState *s = bs->opaque;
     int i;

block/cloop.c

@@ -53,7 +53,7 @@ static int cloop_probe(const uint8_t *buf, int buf_size, const char *filename)
     return 0;
 }
 
-static int cloop_open(BlockDriverState *bs, int flags)
+static int cloop_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVCloopState *s = bs->opaque;
     uint32_t offsets_size, max_compressed_block_size = 1, i;

block/cow.c

@@ -58,7 +58,7 @@ static int cow_probe(const uint8_t *buf, int buf_size, const char *filename)
         return 0;
 }
 
-static int cow_open(BlockDriverState *bs, int flags)
+static int cow_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVCowState *s = bs->opaque;
     struct cow_header_v2 cow_header;

block/dmg.c

@@ -85,7 +85,7 @@ static int read_uint32(BlockDriverState *bs, int64_t offset, uint32_t *result)
     return 0;
 }
 
-static int dmg_open(BlockDriverState *bs, int flags)
+static int dmg_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVDMGState *s = bs->opaque;
     uint64_t info_begin,info_end,last_in_offset,last_out_offset;

block/parallels.c

@@ -68,7 +68,7 @@ static int parallels_probe(const uint8_t *buf, int buf_size, const char *filenam
     return 0;
 }
 
-static int parallels_open(BlockDriverState *bs, int flags)
+static int parallels_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVParallelsState *s = bs->opaque;
     int i;

block/qcow.c

@@ -92,7 +92,7 @@ static int qcow_probe(const uint8_t *buf, int buf_size, const char *filename)
         return 0;
 }
 
-static int qcow_open(BlockDriverState *bs, int flags)
+static int qcow_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVQcowState *s = bs->opaque;
     int len, i, shift, ret;

block/qcow2-cluster.c

@@ -454,6 +454,9 @@ int qcow2_get_cluster_offset(BlockDriverState *bs, uint64_t offset,
         *cluster_offset &= L2E_COMPRESSED_OFFSET_SIZE_MASK;
         break;
     case QCOW2_CLUSTER_ZERO:
+        if (s->qcow_version < 3) {
+            return -EIO;
+        }
         c = count_contiguous_clusters(nb_clusters, s->cluster_size,
                 &l2_table[l2_index], 0,
                 QCOW_OFLAG_COMPRESSED | QCOW_OFLAG_ZERO);
@@ -668,7 +671,7 @@ int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
     }
 
     /* Update L2 table. */
-    if (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS) {
+    if (s->use_lazy_refcounts) {
         qcow2_mark_dirty(bs);
     }
     if (qcow2_need_accurate_refcounts(s)) {

block/qcow2-refcount.c

@@ -201,7 +201,10 @@ static int alloc_refcount_block(BlockDriverState *bs,
     *refcount_block = NULL;
 
     /* We write to the refcount table, so we might depend on L2 tables */
-    qcow2_cache_flush(bs, s->l2_table_cache);
+    ret = qcow2_cache_flush(bs, s->l2_table_cache);
+    if (ret < 0) {
+        return ret;
+    }
 
     /* Allocate the refcount block itself and mark it as used */
     int64_t new_block = alloc_clusters_noref(bs, s->cluster_size);
@@ -237,7 +240,10 @@ static int alloc_refcount_block(BlockDriverState *bs,
             goto fail_block;
         }
 
-        bdrv_flush(bs->file);
+        ret = qcow2_cache_flush(bs, s->refcount_block_cache);
+        if (ret < 0) {
+            goto fail_block;
+        }
 
         /* Initialize the new refcount block only after updating its refcount,
          * update_refcount uses the refcount cache itself */
@@ -526,8 +532,6 @@ static int update_cluster_refcount(BlockDriverState *bs,
         return ret;
     }
 
-    bdrv_flush(bs->file);
-
     return get_refcount(bs, cluster_index);
 }
 
@@ -663,7 +667,11 @@ int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
         }
     }
 
-    bdrv_flush(bs->file);
+    /* The cluster refcount was incremented, either by qcow2_alloc_clusters()
+     * or explicitly by update_cluster_refcount().  Refcount blocks must be
+     * flushed before the caller's L2 table updates.
+     */
+    qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache);
     return offset;
 }
 
@@ -782,10 +790,6 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
                             if (ret < 0) {
                                 goto fail;
                             }
-
-                            /* TODO Flushing once for the whole function should
-                             * be enough */
-                            bdrv_flush(bs->file);
                         }
                         /* compressed clusters are never modified */
                         refcount = 2;
@@ -841,7 +845,7 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
         }
     }
 
-    ret = 0;
+    ret = bdrv_flush(bs);
 fail:
     if (l2_table) {
         qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);

block/qcow2-snapshot.c

@@ -180,11 +180,14 @@ static int qcow2_write_snapshots(BlockDriverState *bs)
 
     /* Allocate space for the new snapshot list */
     snapshots_offset = qcow2_alloc_clusters(bs, snapshots_size);
-    bdrv_flush(bs->file);
     offset = snapshots_offset;
     if (offset < 0) {
         return offset;
     }
+    ret = bdrv_flush(bs);
+    if (ret < 0) {
+        return ret;
+    }
 
     /* Write all snapshots to the new list */
     for(i = 0; i < s->nb_snapshots; i++) {
@@ -378,11 +381,6 @@ int qcow2_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
         goto fail;
     }
 
-    ret = bdrv_flush(bs);
-    if (ret < 0) {
-        goto fail;
-    }
-
     /* Append the new snapshot to the snapshot list */
     new_snapshot_list = g_malloc((s->nb_snapshots + 1) * sizeof(QCowSnapshot));
     if (s->snapshots) {

block/qcow2.c

@@ -285,11 +285,26 @@ static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result,
     return ret;
 }
 
-static int qcow2_open(BlockDriverState *bs, int flags)
+static QemuOptsList qcow2_runtime_opts = {
+    .name = "qcow2",
+    .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head),
+    .desc = {
+        {
+            .name = "lazy_refcounts",
+            .type = QEMU_OPT_BOOL,
+            .help = "Postpone refcount updates",
+        },
+        { /* end of list */ }
+    },
+};
+
+static int qcow2_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVQcowState *s = bs->opaque;
     int len, i, ret = 0;
     QCowHeader header;
+    QemuOpts *opts;
+    Error *local_err = NULL;
     uint64_t ext_end;
 
     ret = bdrv_pread(bs->file, 0, &header, sizeof(header));
@@ -495,6 +510,28 @@ static int qcow2_open(BlockDriverState *bs, int flags)
         }
     }
 
+    /* Enable lazy_refcounts according to image and command line options */
+    opts = qemu_opts_create_nofail(&qcow2_runtime_opts);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);
+        error_free(local_err);
+        ret = -EINVAL;
+        goto fail;
+    }
+
+    s->use_lazy_refcounts = qemu_opt_get_bool(opts, "lazy_refcounts",
+        (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS));
+
+    qemu_opts_del(opts);
+
+    if (s->use_lazy_refcounts && s->qcow_version < 3) {
+        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Lazy refcounts require "
+            "a qcow2 image with at least qemu 1.1 compatibility level");
+        ret = -EINVAL;
+        goto fail;
+    }
+
 #ifdef DEBUG_ALLOC
     {
         BdrvCheckResult result = {0};
@@ -584,7 +621,7 @@ static int coroutine_fn qcow2_co_is_allocated(BlockDriverState *bs,
         *pnum = 0;
     }
 
-    return (cluster_offset != 0);
+    return (cluster_offset != 0) || (ret == QCOW2_CLUSTER_ZERO);
 }
 
 /* handle reading after the end of the backing file */
@@ -665,10 +702,6 @@ static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num,
             break;
 
         case QCOW2_CLUSTER_ZERO:
-            if (s->qcow_version < 3) {
-                ret = -EIO;
-                goto fail;
-            }
             qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors);
             break;
 
@@ -912,7 +945,7 @@ static void qcow2_invalidate_cache(BlockDriverState *bs)
     qcow2_close(bs);
 
     memset(s, 0, sizeof(BDRVQcowState));
-    qcow2_open(bs, flags);
+    qcow2_open(bs, NULL, flags);
 
     if (crypt_method) {
         s->crypt_method = crypt_method;
@@ -1265,7 +1298,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
      */
     BlockDriver* drv = bdrv_find_format("qcow2");
     assert(drv != NULL);
-    ret = bdrv_open(bs, filename,
+    ret = bdrv_open(bs, filename, NULL,
         BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, drv);
     if (ret < 0) {
         goto out;

block/qcow2.h

@@ -173,6 +173,7 @@ typedef struct BDRVQcowState {
 
     int flags;
     int qcow_version;
+    bool use_lazy_refcounts;
 
     uint64_t incompatible_features;
     uint64_t compatible_features;

block/qed.c

@@ -373,7 +373,7 @@ static void bdrv_qed_rebind(BlockDriverState *bs)
     s->bs = bs;
 }
 
-static int bdrv_qed_open(BlockDriverState *bs, int flags)
+static int bdrv_qed_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVQEDState *s = bs->opaque;
     QEDHeader le_header;
@@ -1526,7 +1526,7 @@ static void bdrv_qed_invalidate_cache(BlockDriverState *bs)
 
     bdrv_qed_close(bs);
     memset(s, 0, sizeof(BDRVQEDState));
-    bdrv_qed_open(bs, bs->open_flags);
+    bdrv_qed_open(bs, NULL, bs->open_flags);
 }
 
 static int bdrv_qed_check(BlockDriverState *bs, BdrvCheckResult *result,

block/raw-posix.c

@@ -750,6 +750,7 @@ static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, int fd,
         BlockDriverCompletionFunc *cb, void *opaque, int type)
 {
     RawPosixAIOData *acb = g_slice_new(RawPosixAIOData);
+    ThreadPool *pool;
 
     acb->bs = bs;
     acb->aio_type = type;
@@ -763,7 +764,8 @@ static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, int fd,
     acb->aio_offset = sector_num * 512;
 
     trace_paio_submit(acb, opaque, sector_num, nb_sectors, type);
-    return thread_pool_submit_aio(aio_worker, acb, cb, opaque);
+    pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
+    return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque);
 }
 
 static BlockDriverAIOCB *raw_aio_submit(BlockDriverState *bs,
@@ -1413,6 +1415,7 @@ static BlockDriverAIOCB *hdev_aio_ioctl(BlockDriverState *bs,
 {
     BDRVRawState *s = bs->opaque;
     RawPosixAIOData *acb;
+    ThreadPool *pool;
 
     if (fd_open(bs) < 0)
         return NULL;
@@ -1424,7 +1427,8 @@ static BlockDriverAIOCB *hdev_aio_ioctl(BlockDriverState *bs,
     acb->aio_offset = 0;
     acb->aio_ioctl_buf = buf;
     acb->aio_ioctl_cmd = req;
-    return thread_pool_submit_aio(aio_worker, acb, cb, opaque);
+    pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
+    return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque);
 }
 
 #elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__)

block/raw-win32.c

@@ -144,6 +144,7 @@ static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, HANDLE hfile,
         BlockDriverCompletionFunc *cb, void *opaque, int type)
 {
     RawWin32AIOData *acb = g_slice_new(RawWin32AIOData);
+    ThreadPool *pool;
 
     acb->bs = bs;
     acb->hfile = hfile;
@@ -157,7 +158,8 @@ static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, HANDLE hfile,
     acb->aio_offset = sector_num * 512;
 
     trace_paio_submit(acb, opaque, sector_num, nb_sectors, type);
-    return thread_pool_submit_aio(aio_worker, acb, cb, opaque);
+    pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
+    return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque);
 }
 
 int qemu_ftruncate64(int fd, int64_t length)

block/raw.c

@@ -3,7 +3,7 @@
 #include "block/block_int.h"
 #include "qemu/module.h"
 
-static int raw_open(BlockDriverState *bs, int flags)
+static int raw_open(BlockDriverState *bs, QDict *options, int flags)
 {
     bs->sg = bs->file->sg;
     return 0;

block/sheepdog.c

@@ -468,6 +468,8 @@ static int connect_to_sdog(BDRVSheepdogState *s)
     if (err != NULL) {
         qerror_report_err(err);
         error_free(err);
+    } else {
+        socket_set_nonblock(fd);
     }
 
     return fd;
@@ -499,6 +501,13 @@ static void restart_co_req(void *opaque)
     qemu_coroutine_enter(co, NULL);
 }
 
+static int have_co_req(void *opaque)
+{
+    /* this handler is set only when there is a pending request, so
+     * always returns 1. */
+    return 1;
+}
+
 typedef struct SheepdogReqCo {
     int sockfd;
     SheepdogReq *hdr;
@@ -521,15 +530,14 @@ static coroutine_fn void do_co_req(void *opaque)
     unsigned int *rlen = srco->rlen;
 
     co = qemu_coroutine_self();
-    qemu_aio_set_fd_handler(sockfd, NULL, restart_co_req, NULL, co);
+    qemu_aio_set_fd_handler(sockfd, NULL, restart_co_req, have_co_req, co);
 
-    socket_set_block(sockfd);
     ret = send_co_req(sockfd, hdr, data, wlen);
     if (ret < 0) {
         goto out;
     }
 
-    qemu_aio_set_fd_handler(sockfd, restart_co_req, NULL, NULL, co);
+    qemu_aio_set_fd_handler(sockfd, restart_co_req, NULL, have_co_req, co);
 
     ret = qemu_co_recv(sockfd, hdr, sizeof(*hdr));
     if (ret < sizeof(*hdr)) {
@@ -552,8 +560,9 @@ static coroutine_fn void do_co_req(void *opaque)
     }
     ret = 0;
 out:
+    /* there is at most one request for this sockfd, so it is safe to
+     * set each handler to NULL. */
     qemu_aio_set_fd_handler(sockfd, NULL, NULL, NULL, NULL);
-    socket_set_nonblock(sockfd);
 
     srco->ret = ret;
     srco->finished = true;
@@ -776,8 +785,6 @@ static int get_sheep_fd(BDRVSheepdogState *s)
         return fd;
     }
 
-    socket_set_nonblock(fd);
-
     qemu_aio_set_fd_handler(fd, co_read_response, NULL, aio_flush_request, s);
     return fd;
 }

block/vdi.c

@@ -364,7 +364,7 @@ static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename)
     return result;
 }
 
-static int vdi_open(BlockDriverState *bs, int flags)
+static int vdi_open(BlockDriverState *bs, QDict *options, int flags)
 {
     BDRVVdiState *s = bs->opaque;
     VdiHeader header;