[REVIEW] move tty lock/initial up in the stack
phk at phk.freebsd.dk
Sat Jun 19 18:22:11 GMT 2004
In message <20040619193901.W770 at gamplex.bde.org>, Bruce Evans writes:
>On Sat, 19 Jun 2004, Poul-Henning Kamp wrote:
>> This patch moves the "lock/initial" facility known from sio(4) up
>> to the generic tty layer.
>Just moving them is OK. Unfortunately, the patch does less than move
>... starting here. Control devices in general must be separate so that
>they can have different ownership and permissions, and can be opened
>without side effects. For sio devices, the non-control devices can't
>even be opened if the complementary non-control device is open.
I'm not keen on adding four control devices for all ttys in the system,
and therefore I decided to collapse the states for both directions and
use inband signalling for the changes. I could be persuaded to have
a single control-device per tty.
>This is missing moving the lock state handling, which comprises about half
>of the code that can be moved.
This is only the first step of the journey.
>The new cases don't belong in this case statement anyway, since the new
>ioctls don't involve modification of the active part of the tty struct.
I'm not sure I agree, but I don't care much either way.
I want to get rid of this duplication of the switch anyway, and use
a function call to do the background check and add that functioncall
where each ioctl is implemented.
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
More information about the freebsd-current