Re: login(1) vs "dialout ttys"

[Alan Barrett <apb%cequrux.com@localhost>]

On Tue, 11 Nov 2014, Aymeric Vincent wrote:
> while looking at usr.bin/login/login.c, I noticed the following strange
> test:
>
> 	if (tty[sizeof("tty")-1] == 'd')
> 		syslog(LOG_INFO, "DIALUP %s, %s", tty, pwd->pw_name);

That's a little strange.  If we really want such log messages, 
then they should probably be conditional on bits in the device 
minor number, rather than conditional on the spelling of the name.

We could invent a new ioctl for "is this a dialout device?",  or 
add a new bit to the data returned by an existing ioctl (perhaps 
TIOCMGET).

Theoretically, each tty-like driver makes its own decision about 
how to interpret the minor bits.  In practice, they almost all use 
0x80000 for dialout:

sys/dev/ic/cd18xxvar.h:#define CD18XX_DIALOUT(x)   ((minor(x) & 0x80000) != 0)
sys/dev/ic/clmpcc.c:#define      CLMPCCDIALOUT(x)        (minor(x) & 0x80000)
sys/dev/ic/com.c:#define     COMDIALOUT_MASK 0x80000
sys/dev/ic/cy.c:#define     CYDIALOUT_MASK          0x80000
sys/dev/ic/z8530tty.c:#define     ZSDIALOUT(x)    (minor(x) & 0x80000)
sys/dev/marvell/gtmpsc.c:#define GTMPSCDIALOUT_MASK 0x80000
sys/dev/pci/cz.c:#define CZTTYDIALOUT_MASK   0x80000
sys/dev/usb/ucom.c:#define      UCOMDIALOUT_MASK        0x80000
sys/dev/usb/ucycom.c:#define UCYCOMDIALOUT_MASK   0x80000
sys/dev/usb/uhso.c:#define UHSO_DIALOUT_MASK   0x80000

However, these do not use 0x80000:

sys/dev/sbus/magmareg.h:#define MTTY_DIALOUT(x) (minor(x) & 0x10)
sys/dev/sbus/spifreg.h:#define STTY_DIALOUT(x) (minor(x) & 0x10)

--apb (Alan Barrett)