Subject: Re: ptyfs fully working now...
To: None <>
From: Rui Paulo <>
List: tech-kern
Date: 11/11/2004 11:11:45
Content-Type: text/plain; charset=US-ASCII
Content-Disposition: inline
Content-Transfer-Encoding: 7bit

On Mon, 8 Nov 2004 15:04:52 -0500 (Christos Zoulas) wrote:

> This is work in progress for a ptyfs implementation.

I didn't tested it yet, but thanks for doing this.

> 3. I wanted the old bsd ptys to co-exist with /dev/pts. For that when ptyfs
>    starts up, it will copy the permissions from /dev/ptyXX and /dev/ttyXX
>    for its nodes. 

Indeed, a nice approach.

> 4. I did not like the way ttyname() worked with looking up the db file,
>    so I let the ioctl to find the pty name work on the slave too. Now
>    ttyname() will do the ioctl first to find the ttyname. If that fails,
>    then it will lookup in the db. This is a lot faster in the regular
>    case these days which is ptys. It is also silly to have entries for
>    all the ptys in /etc/ttys. You can now remove them, and the code
>    will take up the last + 1 slot for each pty (for either the old
>    or the new ptys).

Yes, the db file was huge, but how can I disable a specific tty under ptyfs ?

> 7. Kernels need COMPAT_BSDPTY to be able to allocated old style /dev/ttyXX
>    ptys using /dev/ptm. Regular pty allocation will work.

For the others:
As I saw in source-changes@, Christos Zoulas added COMPAT_BSDPTY to the GENERIC kernels.

 Rui Paulo                                

Content-Type: application/pgp-signature

Version: GnuPG v1.2.4 (NetBSD)