Re: [Qemu-devel] [PATCH v1 5/7] io: get rid of bounce buffering in websock write path

[Eric Blake]

On 10/10/2017 10:43 AM, Daniel P. Berrange wrote:
> Currently most outbound I/O on the websock channel gets copied into the
> rawoutput buffer, and then immediately copied again into the encoutput
> buffer, with a header prepended. Now that qio_channel_websock_encode
> accepts a struct iovec, we can trivially remove this bounce buffering
> and write directly to encoutput.
> 
> In doing so, we also now correctly validate the encoutput size against
> the QIO_CHANNEL_WEBSOCK_MAX_BUFFER limit.
> 
> Signed-off-by: Daniel P. Berrange <address@hidden>
> ---
>  include/io/channel-websock.h |  1 -
>  io/channel-websock.c         | 64 
> +++++++++++++++++++-------------------------
>  2 files changed, 28 insertions(+), 37 deletions(-)
> 

> -    iov.iov_base = ioc->rawoutput.buffer;
> -    iov.iov_len = ioc->rawoutput.offset;
>      qio_channel_websock_encode(ioc, QIO_CHANNEL_WEBSOCK_OPCODE_CLOSE,
> -                               &iov, 1, iov.iov_len);
> -    buffer_reset(&ioc->rawoutput);
> +                               iov, niov, size);

If 4/7 changes this to compute size from the iov/niov, then this has
knock-on impact.

> @@ -1115,32 +1118,21 @@ static ssize_t qio_channel_websock_writev(QIOChannel 
> *ioc,
>          return -1;

> +    avail = wioc->encoutput.offset >= QIO_CHANNEL_WEBSOCK_MAX_BUFFER ?
> +        0 : (QIO_CHANNEL_WEBSOCK_MAX_BUFFER - wioc->encoutput.offset);
> +    if (want > avail) {
> +        want = avail;
>      }
>  
> - done:
> -    if (wioc->rawoutput.offset) {
> -        struct iovec iov = { .iov_base = wioc->rawoutput.buffer,
> -                             .iov_len = wioc->rawoutput.offset };
> +    if (want) {
>          qio_channel_websock_encode(wioc,
>                                     QIO_CHANNEL_WEBSOCK_OPCODE_BINARY_FRAME,
> -                                   &iov, 1, iov.iov_len);
> -        buffer_reset(&wioc->rawoutput);
> +                                   iov, niov, want);

then again, I see that you DO support want < iov_size(iov, niov).

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org

signature.asc
Description: OpenPGP digital signature