Re: ioctl(2) numbers

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Wed, 01 Aug 2018 15:43:06 +0100
    From:        Robert Swindells <rjs%fdy2.co.uk@localhost>
    Message-ID:  <x7h8ke2lmt.fsf%ren.fdy2.co.uk@localhost>

  | I understand how ioctl works.

OK, apologies...

  | I asked because it looked as if there was a convention within that file
  | that numbers be unique, even if they were in different groups.

That looks to date back to original BSD definitions ...  just some desire
to be "nice" I would expect, and I doubt it makes any difference (that is,
I cannot see anyone complaining - sockio.h is something of a mess...)

I did do a little checking though, after the last message, and I see ...

./dev/isa/satlinkio.h:#define   SATIOSBUFSIZE   _IOW('s', 2, int) /* set buffer size */
./sys/sockio.h:#define  SIOCSLOWAT       _IOW('s',  2, int)             /* set low watermark */

which is not wonderful, though I doubt that there is much chance of
the wrong one of those ever being used in the unintended situation.

Slightly more problematic ...

./external/bsd/ipf/netinet/ip_fil.h:# define    SIOCADAFR       _IOW('r', 60, struct ipfobj)
./external/bsd/ipf/netinet/ip_nat.h:#define     SIOCADNAT       _IOW('r', 60, struct ipfobj)

('r', 60) turns out to be a popular choice...  (there are 5, only those two 
conflict however).

We should probably make a system wide global registry of such things.

kre