Fix truncation, add write-back cache support

debian/changelog

@@ -1,3 +1,10 @@
+pve-qemu-kvm (7.1.0-4+vitastor5) bullseye; urgency=medium
+
+  * Fix truncation
+  * Add write-back cache support
+
+ -- Vitaliy Filippov <vitalif@yourcmc.ru>  Fri, 27 Oct 2023 21:04:05 +0300
+
 pve-qemu-kvm (7.1.0-4+vitastor4) bullseye; urgency=medium
 
   * Improve performance by adding io_uring support

debian/patches/pve-qemu-7.1-vitastor.patch

@@ -171,7 +171,7 @@ Index: a/block/vitastor.c
 ===================================================================
 --- /dev/null
 +++ a/block/vitastor.c
-@@ -0,0 +1,1017 @@
+@@ -0,0 +1,1075 @@
 +// Copyright (c) Vitaliy Filippov, 2019+
 +// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
 +
@@ -371,7 +371,11 @@ Index: a/block/vitastor.c
 +            !strcmp(name, "rdma-mtu"))
 +        {
 +            unsigned long long num_val;
++#if QEMU_VERSION_MAJOR < 8 || QEMU_VERSION_MAJOR == 8 && QEMU_VERSION_MINOR < 1
 +            if (parse_uint_full(value, &num_val, 0))
++#else
++            if (parse_uint_full(value, 0, &num_val))
++#endif
 +            {
 +                error_setg(errp, "Illegal %s: %s", name, value);
 +                goto out;
@@ -414,10 +418,7 @@ Index: a/block/vitastor.c
 +    VitastorClient *client = (VitastorClient*)opaque;
 +    qemu_mutex_lock(&client->mutex);
 +    client->bh_uring_scheduled = 0;
-+    do
-+    {
-+        vitastor_c_uring_handle_events(client->proxy);
-+    } while (vitastor_c_uring_has_work(client->proxy));
++    vitastor_c_uring_handle_events(client->proxy);
 +    qemu_mutex_unlock(&client->mutex);
 +}
 +
@@ -497,7 +498,7 @@ Index: a/block/vitastor.c
 +static void universal_aio_set_fd_handler(AioContext *ctx, int fd, IOHandler *fd_read, IOHandler *fd_write, void *opaque)
 +{
 +    aio_set_fd_handler(ctx, fd,
-+#if QEMU_VERSION_MAJOR == 2 && QEMU_VERSION_MINOR >= 5 || QEMU_VERSION_MAJOR >= 3
++#if QEMU_VERSION_MAJOR == 2 && QEMU_VERSION_MINOR >= 5 || QEMU_VERSION_MAJOR >= 3 && (QEMU_VERSION_MAJOR < 8 || QEMU_VERSION_MAJOR == 8 && QEMU_VERSION_MINOR < 1)
 +        0 /*is_external*/,
 +#endif
 +        fd_read,
@@ -561,6 +562,45 @@ Index: a/block/vitastor.c
 +    );
 +}
 +
++#if defined VITASTOR_C_API_VERSION && VITASTOR_C_API_VERSION >= 2
++typedef struct str_array
++{
++    const char **items;
++    int len, alloc;
++} str_array;
++
++static void strarray_push(str_array *a, const char *str)
++{
++    if (a->len >= a->alloc)
++    {
++        a->alloc = !a->alloc ? 4 : 2*a->alloc;
++        a->items = (const char**)realloc(a->items, a->alloc*sizeof(char*));
++        if (!a->items)
++        {
++            fprintf(stderr, "bad alloc\n");
++            abort();
++        }
++    }
++    a->items[a->len++] = str;
++}
++
++static void strarray_push_kv(str_array *a, const char *key, const char *value)
++{
++    if (key && value)
++    {
++        strarray_push(a, key);
++        strarray_push(a, value);
++    }
++}
++
++static void strarray_free(str_array *a)
++{
++    free(a->items);
++    a->items = NULL;
++    a->len = a->alloc = 0;
++}
++#endif
++
 +static int vitastor_file_open(BlockDriverState *bs, QDict *options, int flags, Error **errp)
 +{
 +    VitastorRPC task;
@@ -580,22 +620,19 @@ Index: a/block/vitastor.c
 +    client->rdma_mtu = qdict_get_try_int(options, "rdma-mtu", 0);
 +    client->ctx = bdrv_get_aio_context(bs);
 +#if defined VITASTOR_C_API_VERSION && VITASTOR_C_API_VERSION >= 2
-+    client->proxy = vitastor_c_create_qemu_uring(
-+        vitastor_aio_set_fd_handler, client, client->config_path, client->etcd_host, client->etcd_prefix,
-+        client->use_rdma, client->rdma_device, client->rdma_port_num, client->rdma_gid_index, client->rdma_mtu, 0
-+    );
-+    if (!client->proxy)
-+    {
-+        fprintf(stderr, "vitastor: failed to create io_uring: %s - I/O will be slower\n", strerror(errno));
-+        client->uring_eventfd = -1;
-+#endif
-+        client->proxy = vitastor_c_create_qemu(
-+            vitastor_aio_set_fd_handler, client, client->config_path, client->etcd_host, client->etcd_prefix,
-+            client->use_rdma, client->rdma_device, client->rdma_port_num, client->rdma_gid_index, client->rdma_mtu, 0
-+        );
-+#if defined VITASTOR_C_API_VERSION && VITASTOR_C_API_VERSION >= 2
-+    }
-+    else
++    str_array opt = {};
++    strarray_push_kv(&opt, "config_path", qdict_get_try_str(options, "config-path"));
++    strarray_push_kv(&opt, "etcd_address", qdict_get_try_str(options, "etcd-host"));
++    strarray_push_kv(&opt, "etcd_prefix", qdict_get_try_str(options, "etcd-prefix"));
++    strarray_push_kv(&opt, "use_rdma", qdict_get_try_str(options, "use-rdma"));
++    strarray_push_kv(&opt, "rdma_device", qdict_get_try_str(options, "rdma-device"));
++    strarray_push_kv(&opt, "rdma_port_num", qdict_get_try_str(options, "rdma-port-num"));
++    strarray_push_kv(&opt, "rdma_gid_index", qdict_get_try_str(options, "rdma-gid-index"));
++    strarray_push_kv(&opt, "rdma_mtu", qdict_get_try_str(options, "rdma-mtu"));
++    strarray_push_kv(&opt, "client_writeback_allowed", (flags & BDRV_O_NOCACHE) ? "0" : "1");
++    client->proxy = vitastor_c_create_uring_json(opt.items, opt.len);
++    strarray_free(&opt);
++    if (client->proxy)
 +    {
 +        client->uring_eventfd = vitastor_c_uring_register_eventfd(client->proxy);
 +        if (client->uring_eventfd < 0)
@@ -607,6 +644,19 @@ Index: a/block/vitastor.c
 +        }
 +        universal_aio_set_fd_handler(client->ctx, client->uring_eventfd, vitastor_uring_handler, NULL, client);
 +    }
++    else
++    {
++        // Writeback cache is unusable without io_uring because the client can't correctly flush on exit
++        fprintf(stderr, "vitastor: failed to create io_uring: %s - I/O will be slower%s\n",
++            strerror(errno), (flags & BDRV_O_NOCACHE ? "" : " and writeback cache will be disabled"));
++#endif
++        client->uring_eventfd = -1;
++        client->proxy = vitastor_c_create_qemu(
++            vitastor_aio_set_fd_handler, client, client->config_path, client->etcd_host, client->etcd_prefix,
++            client->use_rdma, client->rdma_device, client->rdma_port_num, client->rdma_gid_index, client->rdma_mtu, 0
++        );
++#if defined VITASTOR_C_API_VERSION && VITASTOR_C_API_VERSION >= 2
++    }
 +#endif
 +    image = client->image = g_strdup(qdict_get_try_str(options, "image"));
 +    client->readonly = (flags & BDRV_O_RDWR) ? 1 : 0;
@@ -666,6 +716,10 @@ Index: a/block/vitastor.c
 +        return -1;
 +    }
 +    bs->total_sectors = client->size / BDRV_SECTOR_SIZE;
++#if QEMU_VERSION_MAJOR > 5 || QEMU_VERSION_MAJOR == 5 && QEMU_VERSION_MINOR >= 1
++    /* When extending regular files, we get zeros from the OS */
++    bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
++#endif
 +    //client->aio_context = bdrv_get_aio_context(bs);
 +    qdict_del(options, "use-rdma");
 +    qdict_del(options, "rdma-mtu");
@@ -762,7 +816,11 @@ Index: a/block/vitastor.c
 +    }
 +
 +    // TODO: Resize inode to <offset> bytes
-+    client->size = offset / BDRV_SECTOR_SIZE;
++#if QEMU_VERSION_MAJOR >= 4
++    client->size = exact || client->size < offset ? offset : client->size;
++#else
++    client->size = offset;
++#endif
 +
 +    return 0;
 +}