Am 31.10.2018 um 22:56 hat Kevin Wolf geschrieben: > aio_worker() for reads and writes isn't boring enough yet. It still does > some postprocessing for handling short reads and turning the result into > the right return value. > > However, there is no reason why handle_aiocb_rw() couldn't do the same, > and even without duplicating code between the read and write path. So > move the code there. > > Signed-off-by: Kevin Wolf <kw...@redhat.com>
> @@ -1354,7 +1356,21 @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) > } > qemu_vfree(buf); > > - return nbytes; > +out: > + if (nbytes == aiocb->aio_nbytes) { > + return 0; > + } else if (nbytes >= 0 && nbytes < aiocb->aio_nbytes) { > + if (aiocb->aio_type & QEMU_AIO_WRITE) { > + return -EINVAL; > + } else { > + iov_memset(aiocb->io.iov, aiocb->io.niov, nbytes, > + 0, aiocb->aio_nbytes - nbytes); > + return aiocb->aio_nbytes; While reviewing Eric's patches, I noticed that this should be return 0. Fixing it in my queue. Kevin