Re: [Qemu-devel] [PATCH v2 3/7] block/qapi: Move 'aio' option to file driver

[Kevin Wolf]

Am 26.09.2016 um 18:49 hat Max Reitz geschrieben:
> On 23.09.2016 16:32, Kevin Wolf wrote:
> > The option whether or not to use a native AIO interface really isn't a
> > generic option for all drivers, but only applies to the native file
> > protocols. This patch moves the option in blockdev-add to the
> > appropriate places (raw-posix and raw-win32).
> > 
> > We still have to keep the flag BDRV_O_NATIVE_AIO for compatibility
> > because so far the AIO option was usually specified on the wrong layer
> > (the top-level format driver, which didn't even look at it) and then
> > inherited by the protocol driver (where it was actually used). We can't
> > forbid this use except in new interfaces.
> > 
> > Signed-off-by: Kevin Wolf <address@hidden>
> > ---
> >  block/raw-posix.c      | 44 ++++++++++++++++++++++++---------------
> >  block/raw-win32.c      | 56 
> > +++++++++++++++++++++++++++++++++++++++++++++-----
> >  qapi/block-core.json   |  6 +++---
> >  tests/qemu-iotests/087 |  4 ++--
> >  4 files changed, 83 insertions(+), 27 deletions(-)
> 
> [...]
> 
> > diff --git a/block/raw-win32.c b/block/raw-win32.c
> > index 56f45fe..734bb10 100644
> > --- a/block/raw-win32.c
> > +++ b/block/raw-win32.c
> 
> [...]
> 
> > +static bool get_aio_option(QemuOpts *opts, int flags, Error **errp)
> > +{
> > +    BlockdevAioOptions aio, aio_default;
> > +
> > +    aio_default = (flags & BDRV_O_NATIVE_AIO) ? BLOCKDEV_AIO_OPTIONS_NATIVE
> > +                                              : 
> > BLOCKDEV_AIO_OPTIONS_THREADS;
> > +    aio = qapi_enum_parse(BlockdevAioOptions_lookup, qemu_opt_get(opts, 
> > "aio"),
> > +                          BLOCKDEV_AIO_OPTIONS__MAX, aio_default, errp);
> > +
> > +    switch (aio) {
> > +    case BLOCKDEV_AIO_OPTIONS_NATIVE:
> > +        return true;
> > +    case BLOCKDEV_AIO_OPTIONS_THREADS:
> > +        return false;
> > +    default:
> > +        error_setg(errp, "Invalid AIO option");
> 
> Any reason for catching this case here but not in raw-posix?
> 
> (Not that it really matters, though.)

Nobody will forget raw-posix when adding a new AIO mode to win32, so I
didn't feel like the additional code was worth it there. But if we add a
new AIO mode to raw-posix, I'm pretty sure we will forget win32.

Kevin

pgp1gEz40VxkG.pgp
Description: PGP signature