Re: Re[8]: cdrom ioctls

"Also sprach Esmont Alexander:"
> I'm sorry. My mistake. My standart template at work contains text in
> russian.

It's the "charset=Windows-XXXX" that causes the problem for me, no
matter what the XXXX is!  (1251, I think I recall).

> >> Äīįšūé äåķü Peter,

> 1) I change "enbd_ioctl_table.h" for read/write option

Well, OK - I wouldn't have done that ...  I'd have added another ioctl,
but you can go about it that way as a temporary hack.

> #ifdef CDROM_DRIVE_STATUS
>    { CDROM_DRIVE_STATUS, _NEW_IOWR(CDROM_DRIVE_STATUS, int), },
> #endif /* CDROM_DRIVE_STATUS */

You want IOR, don't you?  You are not writing, I think? Or is the
argument of the ioctl call used for something?  Anyway ...

> 2) I change in user code for getting data in the variable with
> parammeter for write (code in python)
> 
>             st = CDSL_CURRENT
>             s = ioctl(fd, CDROM_DRIVE_STATUS, st)

It should be:
                st = CDSL_CURRENT;
                errval = ioctl(fd, CDROM_DRIVE_STATUS, &st);
                s = st;

And you may wish to shorten to

                s = CDSL_CURRENT;
                errval = ioctl(fd, CDROM_DRIVE_STATUS, &s);

(and you need to check the return errval code).

But that isn't enough. YES, you need to do that, but you also need to
modify the enbd-server to send the _original_ ioctl to disk instead!


>             # s = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT)
>             _debug_('from cdrom %s ' % media.devicename)
>             _debug_('return %s' % st)
> 
> but I still not getting the right result. Value of st variable not

You wouldn't expect to!

In the nbd/fileserver.c code, in ioctlfile(), you will want to intercept
your hacked ioctl call and suobstitute the real/right/corrected call:

      # warning Will Robinson! Gag me with a tablespoon. CDROM ioctl hack.
      case CDROM_DRIVE_STATUS:
          intval = *(unsigned long*)arg;
          err = file->ioctl(file, CDROM_DRIVE_STATUS, intval);
          *(unsigned long*)arg = err;
          return (err < 0) ? err : 0;
              
There is a built in generic interface for ioctls that need ultraspecial
treatment, but I've never used it after "designing" it, so I don't
recall how to do it! It looks as though this case of IOW ioctls which
also return a meaningful value is rare, but sufficient to require
looking at and thinking about ... I notice that there is a call to

 ioctl = enbd_ioctl_revert(ioctl);


at the start of the ioctlfile() routine, so presumably that is where
prior transformations should really take place before executing the
ioctl, and I see that there is a space after the ioctl has been called
to allow changes there too ...

       // PTB in certain cases fixup answer too
       switch(ioctl) {
       ...

You might want to decompose my suggested stanza into two parts,
"before" and "after" to better fit in with the existing code.

Peter