Re: [Qemu-devel] [PATCH 01/14] Introduce qemu_write_full()

[Andreas Färber]

Am 31.12.2009 um 09:50 schrieb Kirill A. Shutemov:

> 2009/12/31 malc <address@hidden>:
> > On Thu, 31 Dec 2009, Kirill A. Shutemov wrote:
> >
> > > On Thu, Dec 31, 2009 at 4:00 AM, malc <address@hidden> wrote:
> > > > On Thu, 31 Dec 2009, Kirill A. Shutemov wrote:
> > > >
> > > > > A variant of write(2) which handles partial write.
> > > > >
> > > > > Signed-off-by: Kirill A. Shutemov <address@hidden>
> > > > > ---
> > > > >  osdep.c       |   27 +++++++++++++++++++++++++++
> > > > >  qemu-common.h |    1 +
> > > > >  2 files changed, 28 insertions(+), 0 deletions(-)
> > > > >
> > > > > diff --git a/osdep.c b/osdep.c
> > > > > index e4836e7..d2406f2 100644
> > > > > --- a/osdep.c
> > > > > +++ b/osdep.c
> > > > > @@ -243,6 +243,33 @@ int qemu_open(const char *name, int  
> > > > > flags, ...)
> > > > >      return ret;
> > > > >  }
> > > > >
> > > > > +/*
> > > > > + * A variant of write(2) which handles partial write.
> > > > > + *
> > > > > + * Return the number of bytes transferred.
> > > > > + * Set errno if fewer than `count' bytes are written.
> > > > > + */
> > > > > +ssize_t qemu_write_full(int fd, const void *buf, size_t count)
> > > > > +{
> > > > > +    ssize_t ret = 0;
> > > > > +    ssize_t total = 0;
> > > > > +
> > > > > +    while (count) {
> > > > > +        ret = write(fd, buf, count);
> > > > > +        if (ret < 0) {
> > > > > +            if (errno == EINTR)
> > > > > +                continue;
> > > > > +            break;
> > > > > +        }
> > > > > +
> > > > > +        count -= ret;
> > > > > +        buf += ret;
> > > > > +     total += ret;
> > > > > +    }
> > > > > +
> > > > > +    return total;
> > > > > +}
> > > >
> > > > This hides write errors.
> > >
> > > Why do you think so? Return value < count indicates error.
> >
> > It does not indicate _which_ error it was.
>
> ???
>
> You can check errno.

So all callers need to compare count to the return value, to know when  
to read errno.
Nitpick: count is size_t while total is ssize_t, so the required  
comparison seems a little ugly (without having tested it).

Andreas