Simplifying devfs: minor == unit

[Ed Schouten]

* Kostik Belousov <kostikbel at gmail.com> wrote:
> On Tue, May 27, 2008 at 09:48:40PM +0200, Ed Schouten wrote:
> > * Kostik Belousov <kostikbel at gmail.com> wrote:
> > > In fact, I do not understand why you leave the UMINORMASK stuff. It makes
> > > no sense with the unit == minor, IMHO.
> > 
> > That's because the uminor() and umajor() routines operate on user device
> > numbers (dev_t's). We need to leave those intact, to make things like
> > the binary compat work.
> 
> What does it mean to take bits 8-15 from the the unit number to denote
> some fictitious major ? And, what binary compatibility you talk about
> ? You changed the layout of the encoding for the minor, but still take
> the major from it, while there is simply no space for major after your
> patch, and definitely not in [8:15].
> 
> Also, look at the sys/types.h:
> 
> #ifndef _KERNEL
> /*
> * minor() gives a cookie instead of an index since we don't want to
> * change the meanings of bits 0-15 or waste time and space shifting
> * bits 16-31 for devices that don't use them.
> */
> #define major(x)        ((int)(((u_int)(x) >> 8)&0xff)) /* major number */
> #define minor(x)        ((int)((x)&0xffff00ff))         /* minor number */
> #endif /* !_KERNEL */
> 
> #define makedev(x,y)    ((dev_t)(((x) << 8) | (y)))     /* create dev_t */
> 
> I think, if you want to do the change you proposed, the minors for cdevs
> shall be cleanly separated from the major/minor something, the later has no
> relations with the cdevs in the current FreeBSD.
> 
> I.e., in your patch, the uminor() can and shall be made identity
> function, to match minor2unit() and unit2minor(), and umajor() shall just
> return 0.
> 
> It would be nice to verify that the linuxolator emulation still works
> (look at the compat/linux/linux_stats.c).

Wait. uminor() and umajor() are used on the device number that's
returned by dev2udev(). They are only used to obtain the major and minor
number from a user device number, which can then be converted to a
native Linux/SVR4/etc dev_t. We should not turn this into a 1:1 mapping.
The values of dev2unit(cdev) and minor(cdev) should _never_ end up in
userspace.

It's quite confusing. MAXMINOR currently has a redundant meaning inside
the kernel. It is used for both the user major/minor numbers, but also
the kernel ones minor numbers (we don't store major numbers there),
which is why I renamed it to UMINORMASK while there.

-- 
 Ed Schouten <ed at 80386.nl>
 WWW: http://80386.nl/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-arch/attachments/20080527/4516fd78/attachment.pgp