+
+/* Our block (disk) device should be really simple: the Guest asks for a block
+ * number and we read or write that position in the file. Unfortunately, that
+ * was amazingly slow: the Guest waits until the read is finished before
+ * running anything else, even if it could have been doing useful work.
+ *
+ * We could use async I/O, except it's reputed to suck so hard that characters
+ * actually go missing from your code when you try to use it.
+ *
+ * So we farm the I/O out to thread, and communicate with it via a pipe. */
+
+/* This hangs off device->priv. */
+struct vblk_info
+{
+ /* The size of the file. */
+ off64_t len;
+
+ /* The file descriptor for the file. */
+ int fd;
+
+ /* IO thread listens on this file descriptor [0]. */
+ int workpipe[2];
+
+ /* IO thread writes to this file descriptor to mark it done, then
+ * Launcher triggers interrupt to Guest. */
+ int done_fd;
+};
+
+/*L:210
+ * The Disk
+ *
+ * Remember that the block device is handled by a separate I/O thread. We head
+ * straight into the core of that thread here:
+ */
+static bool service_io(struct device *dev)
+{
+ struct vblk_info *vblk = dev->priv;
+ unsigned int head, out_num, in_num, wlen;
+ int ret;
+ u8 *in;
+ struct virtio_blk_outhdr *out;
+ struct iovec iov[dev->vq->vring.num];
+ off64_t off;
+
+ /* See if there's a request waiting. If not, nothing to do. */
+ head = get_vq_desc(dev->vq, iov, &out_num, &in_num);
+ if (head == dev->vq->vring.num)
+ return false;
+
+ /* Every block request should contain at least one output buffer
+ * (detailing the location on disk and the type of request) and one
+ * input buffer (to hold the result). */
+ if (out_num == 0 || in_num == 0)
+ errx(1, "Bad virtblk cmd %u out=%u in=%u",
+ head, out_num, in_num);
+
+ out = convert(&iov[0], struct virtio_blk_outhdr);
+ in = convert(&iov[out_num+in_num-1], u8);
+ off = out->sector * 512;
+
+ /* The block device implements "barriers", where the Guest indicates
+ * that it wants all previous writes to occur before this write. We
+ * don't have a way of asking our kernel to do a barrier, so we just
+ * synchronize all the data in the file. Pretty poor, no? */
+ if (out->type & VIRTIO_BLK_T_BARRIER)
+ fdatasync(vblk->fd);
+
+ /* In general the virtio block driver is allowed to try SCSI commands.
+ * It'd be nice if we supported eject, for example, but we don't. */
+ if (out->type & VIRTIO_BLK_T_SCSI_CMD) {
+ fprintf(stderr, "Scsi commands unsupported\n");
+ *in = VIRTIO_BLK_S_UNSUPP;
+ wlen = sizeof(*in);
+ } else if (out->type & VIRTIO_BLK_T_OUT) {
+ /* Write */
+
+ /* Move to the right location in the block file. This can fail
+ * if they try to write past end. */
+ if (lseek64(vblk->fd, off, SEEK_SET) != off)
+ err(1, "Bad seek to sector %llu", out->sector);
+
+ ret = writev(vblk->fd, iov+1, out_num-1);
+ verbose("WRITE to sector %llu: %i\n", out->sector, ret);
+
+ /* Grr... Now we know how long the descriptor they sent was, we
+ * make sure they didn't try to write over the end of the block
+ * file (possibly extending it). */
+ if (ret > 0 && off + ret > vblk->len) {
+ /* Trim it back to the correct length */
+ ftruncate64(vblk->fd, vblk->len);
+ /* Die, bad Guest, die. */
+ errx(1, "Write past end %llu+%u", off, ret);
+ }
+ wlen = sizeof(*in);
+ *in = (ret >= 0 ? VIRTIO_BLK_S_OK : VIRTIO_BLK_S_IOERR);
+ } else {
+ /* Read */
+
+ /* Move to the right location in the block file. This can fail
+ * if they try to read past end. */
+ if (lseek64(vblk->fd, off, SEEK_SET) != off)
+ err(1, "Bad seek to sector %llu", out->sector);
+
+ ret = readv(vblk->fd, iov+1, in_num-1);
+ verbose("READ from sector %llu: %i\n", out->sector, ret);
+ if (ret >= 0) {
+ wlen = sizeof(*in) + ret;
+ *in = VIRTIO_BLK_S_OK;
+ } else {
+ wlen = sizeof(*in);
+ *in = VIRTIO_BLK_S_IOERR;
+ }
+ }
+
+ /* We can't trigger an IRQ, because we're not the Launcher. It does
+ * that when we tell it we're done. */
+ add_used(dev->vq, head, wlen);
+ return true;
+}
+
+/* This is the thread which actually services the I/O. */
+static int io_thread(void *_dev)
+{
+ struct device *dev = _dev;
+ struct vblk_info *vblk = dev->priv;
+ char c;
+
+ /* Close other side of workpipe so we get 0 read when main dies. */
+ close(vblk->workpipe[1]);
+ /* Close the other side of the done_fd pipe. */
+ close(dev->fd);
+
+ /* When this read fails, it means Launcher died, so we follow. */
+ while (read(vblk->workpipe[0], &c, 1) == 1) {
+ /* We acknowledge each request immediately to reduce latency,
+ * rather than waiting until we've done them all. I haven't
+ * measured to see if it makes any difference.
+ *
+ * That would be an interesting test, wouldn't it? You could
+ * also try having more than one I/O thread. */
+ while (service_io(dev))
+ write(vblk->done_fd, &c, 1);
+ }
+ return 0;
+}
+
+/* Now we've seen the I/O thread, we return to the Launcher to see what happens
+ * when that thread tells us it's completed some I/O. */
+static bool handle_io_finish(int fd, struct device *dev)
+{
+ char c;
+
+ /* If the I/O thread died, presumably it printed the error, so we
+ * simply exit. */
+ if (read(dev->fd, &c, 1) != 1)
+ exit(1);
+
+ /* It did some work, so trigger the irq. */
+ trigger_irq(fd, dev->vq);
+ return true;
+}
+
+/* When the Guest submits some I/O, we just need to wake the I/O thread. */
+static void handle_virtblk_output(int fd, struct virtqueue *vq)
+{
+ struct vblk_info *vblk = vq->dev->priv;
+ char c = 0;
+
+ /* Wake up I/O thread and tell it to go to work! */
+ if (write(vblk->workpipe[1], &c, 1) != 1)
+ /* Presumably it indicated why it died. */
+ exit(1);
+}
+
+/*L:198 This actually sets up a virtual block device. */
+static void setup_block_file(const char *filename)
+{
+ int p[2];
+ struct device *dev;
+ struct vblk_info *vblk;
+ void *stack;
+ struct virtio_blk_config conf;
+
+ /* This is the pipe the I/O thread will use to tell us I/O is done. */
+ pipe(p);
+
+ /* The device responds to return from I/O thread. */
+ dev = new_device("block", VIRTIO_ID_BLOCK, p[0], handle_io_finish);
+
+ /* The device has one virtqueue, where the Guest places requests. */
+ add_virtqueue(dev, VIRTQUEUE_NUM, handle_virtblk_output);
+
+ /* Allocate the room for our own bookkeeping */
+ vblk = dev->priv = malloc(sizeof(*vblk));
+
+ /* First we open the file and store the length. */
+ vblk->fd = open_or_die(filename, O_RDWR|O_LARGEFILE);
+ vblk->len = lseek64(vblk->fd, 0, SEEK_END);
+
+ /* We support barriers. */
+ add_feature(dev, VIRTIO_BLK_F_BARRIER);
+
+ /* Tell Guest how many sectors this device has. */
+ conf.capacity = cpu_to_le64(vblk->len / 512);
+
+ /* Tell Guest not to put in too many descriptors at once: two are used
+ * for the in and out elements. */
+ add_feature(dev, VIRTIO_BLK_F_SEG_MAX);
+ conf.seg_max = cpu_to_le32(VIRTQUEUE_NUM - 2);
+
+ set_config(dev, sizeof(conf), &conf);
+
+ /* The I/O thread writes to this end of the pipe when done. */
+ vblk->done_fd = p[1];
+
+ /* This is the second pipe, which is how we tell the I/O thread about
+ * more work. */
+ pipe(vblk->workpipe);
+
+ /* Create stack for thread and run it. Since stack grows upwards, we
+ * point the stack pointer to the end of this region. */
+ stack = malloc(32768);
+ /* SIGCHLD - We dont "wait" for our cloned thread, so prevent it from
+ * becoming a zombie. */
+ if (clone(io_thread, stack + 32768, CLONE_VM | SIGCHLD, dev) == -1)
+ err(1, "Creating clone");
+
+ /* We don't need to keep the I/O thread's end of the pipes open. */
+ close(vblk->done_fd);
+ close(vblk->workpipe[0]);
+
+ verbose("device %u: virtblock %llu sectors\n",
+ devices.device_num, le64_to_cpu(conf.capacity));
+}