On Wed, 5 Jul 2000, Jason R Thorpe wrote:

> On Wed, Jul 05, 2000 at 06:12:20PM -0700, Bill Studenmund wrote:

I think the only point of concern/disagreement is the change to the
interface to hand in a vnode rather than a dev_t

I guess one other thing I'd had in mind was that we'd make all of these
devices "real" devices. So in the kernel config we'd have:

pseudobus at root
ccd*	at pseudobus?
raid*	at pseudobus?
...

Then the config-a-new-one ioctl would trigger a set of events like what
happens when we find a usb sd. Then these devices behave just like
"real" devices. Among other things, the code:

        if (unit >= foo_cd.cd_ndevs ||
            (sc = foo_cd.cd_devs[unit]) == NULL)
                return (ENXIO);

would work, and it makes finding the softc quick.

> Oh, this is simple.  Let's just take the example of ccd:
> 
> int
> ccdopen(struct vnode *vp, int flags, int fmt, struct proc *p)
> {
> 	struct ccd_softc *cs;
> 	struct disklabel *lp;
> 	int error = 0, part, pmask;
> 
> 	cs = ccd_lookup(DISKUNIT(vp->v_rdev));

Uhm, that's not using the vnode, is it? :-)

I mean that for this code snippet to work, you've got to be saving that cs
you allocate below somwhere where ccd_lookup will find it. If that's so,
do we need the vnode?

> 	if (cs == NULL) {
> 		if (part == RAW_PART) {
> 			/*
> 			 * Device will be created and marked open
> 			 * in ccdioctl() later.
> 			 */
> 			return (0);
> 		}
> 		return (ENXIO);
> 	}
> 
> 	/* body of function */
> 
> 	vp->v_devcookie = cs;
> 	return (0);
> }
> 
> ...
> 
> int ccdioctl(struct vnode *vp, u_long cmd, caddr_t data, int flag,
>     struct proc *p)
> {
> 	...
> 
> 	case CCDIOCSET:
> 		if (cs != NULL) {
> 			error = EBUSY;
> 			goto out;
> 		}
> 
> 		/* Create the device. */
> 		cs = malloc(sizeof(*cs), M_DEVBUF, M_WAITOK);
> 		...
> 
> 		/* Creation was successful! */
> 		vp->v_devcookie = cs;
> 		return (0);
> 
> 	...
> }
> 
> ...etc :-)
> 
  [snip]

> Let's ignore things like ptys for now ... they're a different animal :-)
> 
> The advantage my scheme has for e.g. CCDs is that you need to issue an
> ioctl to configure them anyway, and there are flags to see if the device
> is initted in the CCD driver.  This merely uses the semi-cloning mechanism
> to get dynamic creation and to check for "initialized" :-)
> 
>  > Given that you have to be able to do ioctl's on ccd, raid, vnd, or vg
>  > devices when you might not have a softc defined, wouldn't it be simpler to
>  > leave things as they are, and just change these devices to live (in a
>  > degraded mode) with no softc around? They are the only ones which need
>  > this ability. :-)
>  > 
>  > i.e. let the open routine in these drivers always succeed, even for
>  > unconfigured devices. Then keep the checks to see if there is a softc
>  > around (like the nice one I cut out above) in all of the driver's
>  > routines. For everything other than ioctl or close, no softc == exit with
>  > error. close would succeed, and ioctl would only let you configure the
>  > device.
> 
> I guess we were sort of on the same wavelength here.

Yep. :-)

>  > Also, it makes us quite different from other BSD's. Porting drivers or
>  > trying to maintain one driver source tree for multiple OS's would become
>  > more complicated. While I know you and a number of us could deal with the
>  > complexity, I'd rather not add it if we don't have to. I'd like to get
>  > lots of folks into making drivers, and the easier it is to port them
>  > (within reason), the better.
> 
> We already have better technology inside than the other BSDs :-), so this
> isn't really much of an issue, as far as I'm concerned.

Well, how is having that cookie around in the vnode better? :-) You
mentioned that there was a second part - post it! :-) Because if we did
the pseudobus trick, finding the softc would be simply accessing an array.
:-)

Take care,

Bill