/dev/vcs or /dev/tty local colormap retrieving

/dev/vcs or /dev/tty local colormap retrieving

[Cedric Roux]

Hi linux kernel hackers,

I have a problem.
I want to retrieve the current colormap of a given tty.

/dev/vcsX (or /dev/vcsaX) don't dump the colormap
of the given tty through their read() interface.

The GIO_CMAP ioctl call only returns the current *default*
colormap, which is not what I want, since someone may
send terminal commands to modify the colormap of a
specific tty.

(I am working with EGA/VGA text virtual terminals,
source code in drivers/char/vt.c)

I will do my own private hack to dump the colormap,
but maybe it might be useful for other people as
well, so I send you an email to know if you
are interested.

If yes, what is the prefered way (according to
your internal design policy) to do that?

I think about several methods:
1 - dump the 48 bytes of the colormap at the
    end of read(/dev/vcsaX), so after the screen
    content. Obviously, userland programs that
    don't take into account the dimensions of
    the console may behave strangely after this
    modification, but it's the simplest IMHO,
2 - create a new /dev/vcspX device. Some minor
    number after 191 are still available for
    major 7 c mode. It would pollute /dev,
    but why not,
3 - have a new ioctl, preferably on /dev/vcsaX
    or /dev/vcsX, but might also be available
    for /dev/ttyX (/dev/vcsaX seems more
    natural, because this device already exports
    attributes of characters).

We might also want a write access, why not?
And maybe also get the cursor shape (I didn't
check for that because I don't need it, so it
might already be done somewhere).

If you think this feature is useful, I will
send a patch in a few days.

So, well, that's it.
If it appears that I didn't read the sources
carefully enough and if the feature I want
is already there, I apologize. In which case,
please direct me to the correct source file.

Take care and happy hacking,
Cédric.

Re: /dev/vcs or /dev/tty local colormap retrieving

[Alan Cox]

> 3 - have a new ioctl, preferably on /dev/vcsaX
>     or /dev/vcsX, but might also be available
>     for /dev/ttyX (/dev/vcsaX seems more
>     natural, because this device already exports
>     attributes of characters).

This is probably the right approach.