Merge remote-tracking branch 'kwolf/for-anthony' into staging

* kwolf/for-anthony:
  dataplane: handle misaligned virtio-blk requests
  dataplane: extract virtio-blk read/write processing into do_rdwr_cmd()
  block: make qiov_is_aligned() public
  raw-posix: fix bdrv_aio_ioctl
  sheepdog: implement direct write semantics
  block: do not probe zero-sized disks

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

5 files changed, 120 insertions, 68 deletions

diff --git a/block.c b/block.c
index 60873eafea..b5e64ecf11 100644
--- a/block.c
+++ b/block.c
@@ -527,7 +527,7 @@ static int find_image_format(BlockDriverState *bs, const char *filename,
     int ret = 0;
 
     /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
-    if (bs->sg || !bdrv_is_inserted(bs)) {
+    if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) {
         drv = bdrv_find_format("raw");
         if (!drv) {
             ret = -ENOENT;
@@ -4313,6 +4313,22 @@ void *qemu_blockalign(BlockDriverState *bs, size_t size)
     return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
 }
 
+/*
+ * Check if all memory in this vector is sector aligned.
+ */
+bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov)
+{
+    int i;
+
+    for (i = 0; i < qiov->niov; i++) {
+        if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) {
+            return false;
+        }
+    }
+
+    return true;
+}
+
 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
 {
     int64_t bitmap_size;
diff --git a/block/raw-posix.c b/block/raw-posix.c
index 87d888ed01..c3d7fda7b7 100644
--- a/block/raw-posix.c
+++ b/block/raw-posix.c
@@ -430,22 +430,6 @@ static void raw_reopen_abort(BDRVReopenState *state)
 #endif
 */
 
-/*
- * Check if all memory in this vector is sector aligned.
- */
-static int qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov)
-{
-    int i;
-
-    for (i = 0; i < qiov->niov; i++) {
-        if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) {
-            return 0;
-        }
-    }
-
-    return 1;
-}
-
 static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb)
 {
     int ret;
@@ -455,15 +439,7 @@ static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb)
         return -errno;
     }
 
-    /*
-     * This looks weird, but the aio code only considers a request
-     * successful if it has written the full number of bytes.
-     *
-     * Now we overload aio_nbytes as aio_ioctl_cmd for the ioctl command,
-     * so in fact we return the ioctl command here to make posix_aio_read()
-     * happy..
-     */
-    return aiocb->aio_nbytes;
+    return 0;
 }
 
 static ssize_t handle_aiocb_flush(RawPosixAIOData *aiocb)
@@ -722,7 +698,7 @@ static BlockDriverAIOCB *raw_aio_submit(BlockDriverState *bs,
      * driver that it needs to copy the buffer.
      */
     if ((bs->open_flags & BDRV_O_NOCACHE)) {
-        if (!qiov_is_aligned(bs, qiov)) {
+        if (!bdrv_qiov_is_aligned(bs, qiov)) {
             type |= QEMU_AIO_MISALIGNED;
 #ifdef CONFIG_LINUX_AIO
         } else if (s->use_aio) {
diff --git a/block/sheepdog.c b/block/sheepdog.c
index e821746116..462c4b2d5d 100644
--- a/block/sheepdog.c
+++ b/block/sheepdog.c
@@ -36,7 +36,8 @@
 
 #define SD_FLAG_CMD_WRITE    0x01
 #define SD_FLAG_CMD_COW      0x02
-#define SD_FLAG_CMD_CACHE    0x04
+#define SD_FLAG_CMD_CACHE    0x04 /* Writeback mode for cache */
+#define SD_FLAG_CMD_DIRECT   0x08 /* Don't use cache */
 
 #define SD_RES_SUCCESS       0x00 /* Success */
 #define SD_RES_UNKNOWN       0x01 /* Unknown error */
@@ -293,7 +294,7 @@ typedef struct BDRVSheepdogState {
 
     char name[SD_MAX_VDI_LEN];
     bool is_snapshot;
-    bool cache_enabled;
+    uint32_t cache_flags;
 
     char *addr;
     char *port;
@@ -977,8 +978,8 @@ static int coroutine_fn add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,
         hdr.flags = SD_FLAG_CMD_WRITE | flags;
     }
 
-    if (s->cache_enabled) {
-        hdr.flags |= SD_FLAG_CMD_CACHE;
+    if (s->cache_flags) {
+        hdr.flags |= s->cache_flags;
     }
 
     hdr.oid = oid;
@@ -1023,7 +1024,7 @@ static int coroutine_fn add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,
 
 static int read_write_object(int fd, char *buf, uint64_t oid, int copies,
                              unsigned int datalen, uint64_t offset,
-                             bool write, bool create, bool cache)
+                             bool write, bool create, uint32_t cache_flags)
 {
     SheepdogObjReq hdr;
     SheepdogObjRsp *rsp = (SheepdogObjRsp *)&hdr;
@@ -1047,9 +1048,7 @@ static int read_write_object(int fd, char *buf, uint64_t oid, int copies,
         hdr.opcode = SD_OP_READ_OBJ;
     }
 
-    if (cache) {
-        hdr.flags |= SD_FLAG_CMD_CACHE;
-    }
+    hdr.flags |= cache_flags;
 
     hdr.oid = oid;
     hdr.data_length = datalen;
@@ -1072,18 +1071,19 @@ static int read_write_object(int fd, char *buf, uint64_t oid, int copies,
 }
 
 static int read_object(int fd, char *buf, uint64_t oid, int copies,
-                       unsigned int datalen, uint64_t offset, bool cache)
+                       unsigned int datalen, uint64_t offset,
+                       uint32_t cache_flags)
 {
     return read_write_object(fd, buf, oid, copies, datalen, offset, false,
-                             false, cache);
+                             false, cache_flags);
 }
 
 static int write_object(int fd, char *buf, uint64_t oid, int copies,
                         unsigned int datalen, uint64_t offset, bool create,
-                        bool cache)
+                        uint32_t cache_flags)
 {
     return read_write_object(fd, buf, oid, copies, datalen, offset, true,
-                             create, cache);
+                             create, cache_flags);
 }
 
 static int sd_open(BlockDriverState *bs, const char *filename, int flags)
@@ -1118,12 +1118,22 @@ static int sd_open(BlockDriverState *bs, const char *filename, int flags)
         goto out;
     }
 
-    s->cache_enabled = true;
-    s->flush_fd = connect_to_sdog(s->addr, s->port);
-    if (s->flush_fd < 0) {
-        error_report("failed to connect");
-        ret = s->flush_fd;
-        goto out;
+    /*
+     * QEMU block layer emulates writethrough cache as 'writeback + flush', so
+     * we always set SD_FLAG_CMD_CACHE (writeback cache) as default.
+     */
+    s->cache_flags = SD_FLAG_CMD_CACHE;
+    if (flags & BDRV_O_NOCACHE) {
+        s->cache_flags = SD_FLAG_CMD_DIRECT;
+    }
+
+    if (s->cache_flags == SD_FLAG_CMD_CACHE) {
+        s->flush_fd = connect_to_sdog(s->addr, s->port);
+        if (s->flush_fd < 0) {
+            error_report("failed to connect");
+            ret = s->flush_fd;
+            goto out;
+        }
     }
 
     if (snapid || tag[0] != '\0') {
@@ -1140,7 +1150,7 @@ static int sd_open(BlockDriverState *bs, const char *filename, int flags)
 
     buf = g_malloc(SD_INODE_SIZE);
     ret = read_object(fd, buf, vid_to_vdi_oid(vid), 0, SD_INODE_SIZE, 0,
-                      s->cache_enabled);
+                      s->cache_flags);
 
     closesocket(fd);
 
@@ -1387,7 +1397,7 @@ static void sd_close(BlockDriverState *bs)
 
     qemu_aio_set_fd_handler(s->fd, NULL, NULL, NULL, NULL);
     closesocket(s->fd);
-    if (s->cache_enabled) {
+    if (s->cache_flags) {
         closesocket(s->flush_fd);
     }
     g_free(s->addr);
@@ -1423,7 +1433,7 @@ static int sd_truncate(BlockDriverState *bs, int64_t offset)
     datalen = SD_INODE_SIZE - sizeof(s->inode.data_vdi_id);
     s->inode.vdi_size = offset;
     ret = write_object(fd, (char *)&s->inode, vid_to_vdi_oid(s->inode.vdi_id),
-                       s->inode.nr_copies, datalen, 0, false, s->cache_enabled);
+                       s->inode.nr_copies, datalen, 0, false, s->cache_flags);
     close(fd);
 
     if (ret < 0) {
@@ -1506,7 +1516,7 @@ static int sd_create_branch(BDRVSheepdogState *s)
     }
 
     ret = read_object(fd, buf, vid_to_vdi_oid(vid), s->inode.nr_copies,
-                      SD_INODE_SIZE, 0, s->cache_enabled);
+                      SD_INODE_SIZE, 0, s->cache_flags);
 
     closesocket(fd);
 
@@ -1707,7 +1717,7 @@ static int coroutine_fn sd_co_flush_to_disk(BlockDriverState *bs)
     int ret;
     unsigned int wlen = 0, rlen = 0;
 
-    if (!s->cache_enabled) {
+    if (s->cache_flags != SD_FLAG_CMD_CACHE) {
         return 0;
     }
 
@@ -1723,7 +1733,7 @@ static int coroutine_fn sd_co_flush_to_disk(BlockDriverState *bs)
     if (rsp->result == SD_RES_INVALID_PARMS) {
         dprintf("disable write cache since the server doesn't support it\n");
 
-        s->cache_enabled = false;
+        s->cache_flags = SD_FLAG_CMD_DIRECT;
         closesocket(s->flush_fd);
         return 0;
     }
@@ -1774,7 +1784,7 @@ static int sd_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
     }
 
     ret = write_object(fd, (char *)&s->inode, vid_to_vdi_oid(s->inode.vdi_id),
-                       s->inode.nr_copies, datalen, 0, false, s->cache_enabled);
+                       s->inode.nr_copies, datalen, 0, false, s->cache_flags);
     if (ret < 0) {
         error_report("failed to write snapshot's inode.");
         goto cleanup;
@@ -1791,7 +1801,7 @@ static int sd_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
     inode = (SheepdogInode *)g_malloc(datalen);
 
     ret = read_object(fd, (char *)inode, vid_to_vdi_oid(new_vid),
-                      s->inode.nr_copies, datalen, 0, s->cache_enabled);
+                      s->inode.nr_copies, datalen, 0, s->cache_flags);
 
     if (ret < 0) {
         error_report("failed to read new inode info. %s", strerror(errno));
@@ -1845,7 +1855,7 @@ static int sd_snapshot_goto(BlockDriverState *bs, const char *snapshot_id)
 
     buf = g_malloc(SD_INODE_SIZE);
     ret = read_object(fd, buf, vid_to_vdi_oid(vid), s->inode.nr_copies,
-                      SD_INODE_SIZE, 0, s->cache_enabled);
+                      SD_INODE_SIZE, 0, s->cache_flags);
 
     closesocket(fd);
 
@@ -1942,7 +1952,7 @@ static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
         /* we don't need to read entire object */
         ret = read_object(fd, (char *)&inode, vid_to_vdi_oid(vid),
                           0, SD_INODE_SIZE - sizeof(inode.data_vdi_id), 0,
-                          s->cache_enabled);
+                          s->cache_flags);
 
         if (ret) {
             continue;
@@ -2003,11 +2013,11 @@ static int do_load_save_vmstate(BDRVSheepdogState *s, uint8_t *data,
         if (load) {
             ret = read_object(fd, (char *)data, vmstate_oid,
                               s->inode.nr_copies, data_len, offset,
-                              s->cache_enabled);
+                              s->cache_flags);
         } else {
             ret = write_object(fd, (char *)data, vmstate_oid,
                                s->inode.nr_copies, data_len, offset, create,
-                               s->cache_enabled);
+                               s->cache_flags);
         }
 
         if (ret < 0) {
diff --git a/hw/dataplane/virtio-blk.c b/hw/dataplane/virtio-blk.c
index 4c4ad8422a..1f7346ea19 100644
--- a/hw/dataplane/virtio-blk.c
+++ b/hw/dataplane/virtio-blk.c
@@ -34,6 +34,8 @@ typedef struct {
     struct iocb iocb;               /* Linux AIO control block */
     QEMUIOVector *inhdr;            /* iovecs for virtio_blk_inhdr */
     unsigned int head;              /* vring descriptor index */
+    struct iovec *bounce_iov;       /* used if guest buffers are unaligned */
+    QEMUIOVector *read_qiov;        /* for read completion /w bounce buffer */
 } VirtIOBlockRequest;
 
 struct VirtIOBlockDataPlane {
@@ -89,6 +91,18 @@ static void complete_request(struct iocb *iocb, ssize_t ret, void *opaque)
 
     trace_virtio_blk_data_plane_complete_request(s, req->head, ret);
 
+    if (req->read_qiov) {
+        assert(req->bounce_iov);
+        qemu_iovec_from_buf(req->read_qiov, 0, req->bounce_iov->iov_base, len);
+        qemu_iovec_destroy(req->read_qiov);
+        g_slice_free(QEMUIOVector, req->read_qiov);
+    }
+
+    if (req->bounce_iov) {
+        qemu_vfree(req->bounce_iov->iov_base);
+        g_slice_free(struct iovec, req->bounce_iov);
+    }
+
     qemu_iovec_from_buf(req->inhdr, 0, &hdr, sizeof(hdr));
     qemu_iovec_destroy(req->inhdr);
     g_slice_free(QEMUIOVector, req->inhdr);
@@ -130,6 +144,48 @@ static void do_get_id_cmd(VirtIOBlockDataPlane *s,
     complete_request_early(s, head, inhdr, VIRTIO_BLK_S_OK);
 }
 
+static int do_rdwr_cmd(VirtIOBlockDataPlane *s, bool read,
+                       struct iovec *iov, unsigned int iov_cnt,
+                       long long offset, unsigned int head,
+                       QEMUIOVector *inhdr)
+{
+    struct iocb *iocb;
+    QEMUIOVector qiov;
+    struct iovec *bounce_iov = NULL;
+    QEMUIOVector *read_qiov = NULL;
+
+    qemu_iovec_init_external(&qiov, iov, iov_cnt);
+    if (!bdrv_qiov_is_aligned(s->blk->conf.bs, &qiov)) {
+        void *bounce_buffer = qemu_blockalign(s->blk->conf.bs, qiov.size);
+
+        if (read) {
+            /* Need to copy back from bounce buffer on completion */
+            read_qiov = g_slice_new(QEMUIOVector);
+            qemu_iovec_init(read_qiov, iov_cnt);
+            qemu_iovec_concat_iov(read_qiov, iov, iov_cnt, 0, qiov.size);
+        } else {
+            qemu_iovec_to_buf(&qiov, 0, bounce_buffer, qiov.size);
+        }
+
+        /* Redirect I/O to aligned bounce buffer */
+        bounce_iov = g_slice_new(struct iovec);
+        bounce_iov->iov_base = bounce_buffer;
+        bounce_iov->iov_len = qiov.size;
+        iov = bounce_iov;
+        iov_cnt = 1;
+    }
+
+    iocb = ioq_rdwr(&s->ioqueue, read, iov, iov_cnt, offset);
+
+    /* Fill in virtio block metadata needed for completion */
+    VirtIOBlockRequest *req = container_of(iocb, VirtIOBlockRequest, iocb);
+    req->head = head;
+    req->inhdr = inhdr;
+    req->bounce_iov = bounce_iov;
+    req->read_qiov = read_qiov;
+    return 0;
+}
+
 static int process_request(IOQueue *ioq, struct iovec iov[],
                            unsigned int out_num, unsigned int in_num,
                            unsigned int head)
@@ -139,7 +195,6 @@ static int process_request(IOQueue *ioq, struct iovec iov[],
     struct virtio_blk_outhdr outhdr;
     QEMUIOVector *inhdr;
     size_t in_size;
-    struct iocb *iocb;
 
     /* Copy in outhdr */
     if (unlikely(iov_to_buf(iov, out_num, 0, &outhdr,
@@ -167,12 +222,12 @@ static int process_request(IOQueue *ioq, struct iovec iov[],
 
     switch (outhdr.type) {
     case VIRTIO_BLK_T_IN:
-        iocb = ioq_rdwr(ioq, true, in_iov, in_num, outhdr.sector * 512);
-        break;
+        do_rdwr_cmd(s, true, in_iov, in_num, outhdr.sector * 512, head, inhdr);
+        return 0;
 
     case VIRTIO_BLK_T_OUT:
-        iocb = ioq_rdwr(ioq, false, iov, out_num, outhdr.sector * 512);
-        break;
+        do_rdwr_cmd(s, false, iov, out_num, outhdr.sector * 512, head, inhdr);
+        return 0;
 
     case VIRTIO_BLK_T_SCSI_CMD:
         /* TODO support SCSI commands */
@@ -198,12 +253,6 @@ static int process_request(IOQueue *ioq, struct iovec iov[],
         g_slice_free(QEMUIOVector, inhdr);
         return -EFAULT;
     }
-
-    /* Fill in virtio block metadata needed for completion */
-    VirtIOBlockRequest *req = container_of(iocb, VirtIOBlockRequest, iocb);
-    req->head = head;
-    req->inhdr = inhdr;
-    return 0;
 }
 
 static void handle_notify(EventHandler *handler)
diff --git a/include/block/block.h b/include/block/block.h
index 0719339231..ffd193637d 100644
--- a/include/block/block.h
+++ b/include/block/block.h
@@ -349,6 +349,7 @@ void bdrv_img_create(const char *filename, const char *fmt,
 
 void bdrv_set_buffer_alignment(BlockDriverState *bs, int align);
 void *qemu_blockalign(BlockDriverState *bs, size_t size);
+bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
 
 #define BDRV_SECTORS_PER_DIRTY_CHUNK 2048