Re: [Qemu-devel] [PATCH v2] raw-posix: Detect CDROM via ioctl

[Jamie Lokier <jamie@shareable.org>]

Christoph Hellwig wrote:
> On Tue, Jan 12, 2010 at 10:29:11AM -0500, Cole Robinson wrote:
> >  static int cdrom_probe_device(const char *filename)
> >  {
> > +    int fd, ret, prio;
> > +
> >      if (strstart(filename, "/dev/cd", NULL))
> > -        return 100;
> > -    return 0;
> > +        prio = 50;
> > +
> > +    fd = open(filename, O_RDONLY | O_NONBLOCK);
> > +    if (fd < 0) {
> > +        goto out;
> > +    }
> > +
> > +    /* Attempt to detect via a CDROM specific ioctl */
> > +    ret = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
> > +    if (!(ret < 0 && errno == EINVAL))
> > +        prio = 100;
> > +
> > +    close(fd);
> > +out:
> > +    return prio;
> >  }
> 
> Looks good.  We'll now get an open each from the cdrom and floppy
> drivers for each image we're trying to open, but I guess that should be
> fine.

A theoretical problem, that applies to all devices detected in this way:

A few ioctl values (the integers which the names expand to) overlap
between different devices.  It's not policy but it has happened.  This
shows up in strace sometimes, for example on Linux the isatty() C
library function uses TCGETS, but in strace it shows as:

    SNDCTL_TMR_TIMEBASE or TCGETS

I tried isatty() on Linux sound devices and it seemed to be reliable anyway :-)

But, in theory, on some host or other, the CD-ROM and floppy checks
could do accidentally do something else on the wrong sort of device.
I don't know how assiduous hosts are about keeping the number spaces separate.

-- Jamie