Paolo Bonzini <pbonz...@redhat.com> wrote:
> The recent patches to use vectored I/O for RAM migration caused a
> regression in savevm speed. To restore previous performance,
> add data to the buffer in qemu_put_buffer_async whenever writev_buffer
> is not available in the QEMUFile.
>
> Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
> ---
> savevm.c | 49 ++++++++++++++++++++++++++++++++++---------------
> 1 file changed, 34 insertions(+), 15 deletions(-)
>
> diff --git a/savevm.c b/savevm.c
> index c952c41..b32e0a7 100644
> --- a/savevm.c
> +++ b/savevm.c
> @@ -525,27 +525,24 @@ static void qemu_file_set_error(QEMUFile *f, int ret)
> static void qemu_fflush(QEMUFile *f)
> {
> ssize_t ret = 0;
> - int i = 0;
>
> if (!f->ops->writev_buffer && !f->ops->put_buffer) {
> return;
> }
>
> - if (f->is_write && f->iovcnt > 0) {
> + if (f->is_write) {
> if (f->ops->writev_buffer) {
> - ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt);
> - if (ret >= 0) {
> - f->pos += ret;
Code was there, and it is a case of of style, but adding 0 to one
ammount is a NOP.
> + if (f->iovcnt > 0) {
> + ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt);
> }
> } else {
> - for (i = 0; i < f->iovcnt && ret >= 0; i++) {
> - ret = f->ops->put_buffer(f->opaque, f->iov[i].iov_base,
> f->pos,
> - f->iov[i].iov_len);
> - if (ret >= 0) {
> - f->pos += ret;
> - }
> + if (f->buf_index > 0) {
> + ret = f->ops->put_buffer(f->opaque, f->buf, f->pos,
> f->buf_index);
> }
> }
> + if (ret >= 0) {
> + f->pos += ret;
> + }
> f->buf_index = 0;
> f->iovcnt = 0;
> }
> @@ -640,6 +637,11 @@ static void add_to_iovec(QEMUFile *f, const uint8_t
> *buf, int size)
>
> void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, int size)
> {
> + if (!f->ops->writev_buffer) {
> + qemu_put_buffer(f, buf, size);
> + return;
> + }
> +
This
> if (f->last_error) {
> return;
> }
> @@ -673,8 +675,17 @@ void qemu_put_buffer(QEMUFile *f, const uint8_t *buf,
> int size)
> if (l > size)
> l = size;
> memcpy(f->buf + f->buf_index, buf, l);
> - f->buf_index += l;
> - qemu_put_buffer_async(f, f->buf + (f->buf_index - l), l);
> + f->bytes_xfer += size;
> + if (f->ops->writev_buffer) {
> + add_to_iovec(f, f->buf + f->buf_index, l);
> + f->buf_index += l;
> + } else {
> + f->is_write = 1;
> + f->buf_index += l;
> + if (f->buf_index == IO_BUF_SIZE) {
> + qemu_fflush(f);
> + }
> + }
And this
> if (qemu_file_get_error(f)) {
> break;
> }
> @@ -697,8 +708,16 @@ void qemu_put_byte(QEMUFile *f, int v)
>
> f->buf[f->buf_index] = v;
> f->bytes_xfer++;
> - add_to_iovec(f, f->buf + f->buf_index, 1);
> - f->buf_index++;
> + if (f->ops->writev_buffer) {
> + add_to_iovec(f, f->buf + f->buf_index, 1);
> + f->buf_index++;
> + } else {
> + f->is_write = 1;
> + f->buf_index++;
> + if (f->buf_index == IO_BUF_SIZE) {
> + qemu_fflush(f);
> + }
> + }
And this
is not needed if we just put the code inside add_to_iovec() no?
Users of add_to_iovec() should not care if we are using vritev or
put_buffer() right.
and as a second note. Having qemu_put_buffer_async() calling
qemu_put_buffer() and qemu_put_buffer() calling qemu_put_buffer_async()
is ..... weird?
> }
>
> static void qemu_file_skip(QEMUFile *f, int size)
