Paolo Bonzini <pbonz...@redhat.com> wrote:
> This uses system calls directly for Unix file descriptors, so that the
> efficient writev_buffer can be used. Pay attention to the possibility
> of partial writes in writev.
>
> Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
Reviewed-by: Juan Quintela <quint...@redhat.com>
> +static ssize_t unix_writev_buffer(void *opaque, struct iovec *iov, int
> iovcnt)
> +{
> + QEMUFileSocket *s = opaque;
> + ssize_t len, offset;
> + ssize_t size = iov_size(iov, iovcnt);
> + ssize_t total = 0;
> +
> + assert(iovcnt > 0);
> + offset = 0;
> + while (size > 0) {
> + /* Find the next start position; skip all full-sized vector elements
> */
> + while (offset >= iov[0].iov_len) {
> + offset -= iov[0].iov_len;
> + iov++, iovcnt--;
> + }
> +
> + /* skip `offset' bytes from the (now) first element, undo it on exit
> */
> + assert(iovcnt > 0);
> + iov[0].iov_base += offset;
> + iov[0].iov_len -= offset;
> +
> + do {
> + len = writev(s->fd, iov, iovcnt);
> + } while (len == -1 && errno == EINTR);
> + if (len == -1) {
> + return -errno;
> + }
> +
> + /* Undo the changes above */
> + iov[0].iov_base -= offset;
> + iov[0].iov_len += offset;
> +
> + /* Prepare for the next iteration */
> + offset += len;
> + total += len;
> + size -= len;
> + }
> +
> + return total;
> +}
This code is very similar to the one in the iov_send_recv(), but I
can't think on a trivial way to share it :p
Later, Juan.
