Re: [Qemu-devel] Re: [PATCH 1/10] Refactor QEMUFile for live migration

[Avi Kivity]

Glauber Costa wrote:
> > +static int fd_put_buffer(void *opaque, const uint8_t *buf,
> > +                         int64_t pos, int size)
> > +{
> > +    QEMUFileFD *s = opaque;
> > +    ssize_t len;
> > +
> > +    do {
> > +        len = write(s->fd, buf, size);
> > +    } while (len == -1 && errno == EINTR);
> >     
>
> What about the len == size case ?
>
>   

Should work, no?

> > +
> > +QEMUFile *qemu_fopen_fd(int fd)
> > +{
> > +    QEMUFileFD *s = qemu_mallocz(sizeof(QEMUFileFD));
> >     
>
> can't it fail?
>   

If it does, the entire world will disappear in a black hole.  Qemu isn't 
prepared to handle allocation failures.

> Isn't it possible to abstract the differences between bdrv and file so
> to have a common implementation
> between them? Do you think it's worthwhile ?
>
>   

They're very different.  bdrvs are sector-granularity, random access, 
concurrent, and want dma.  QEMUFiles are byte-granularity, sequential 
access, serial, and aren't too worried about dma (though it could be nice).

--
error compiling committee.c: too many arguments to function