Date:	Tue, 17 Jul 2007 01:51:27 -0700
From:	"Shawn Rutledge" <shawn.t.rutledge@...il.com>
To:	linux-kernel@...r.kernel.org
Subject: attempting to open a VT from elsewhere and take raw keyboard input

I'm messing around with framebuffer graphics stuff and trying to write
a keyboard driver similar to the one in X, which will allow you to
start up the graphics system from any tty (not just a virtual
console), find an available VT, switch to that VT, and then take input
from the physically attached keyboard rather than necessarily from the
controlling tty (which might not be a VT).  I have looked at the X
implementation (hw/dmx/input/lnx-keyboard.c) which seems
straightforward but I am not getting the same results when I try to do
the same thing, so must be missing something.  I started with this
very simple program:

void main()
{
        int i;
        char c;
        int fd;

        fd = open("/dev/tty1", O_RDWR | O_ASYNC, 0);
        printf("opened /dev/tty1 as fd %d\n", fd);
        if(ioctl(fd, KDSKBMODE, K_RAW) < 0) perror(
                "Warning: failed to change the virtual console to Raw mode.");

        while((i = read(fd, &c, 1)) == 1)
        {
                if( (c &= 255) == 0xBE) /* Cancel */
                        break;
                printf( "%x\n\r", c);
        }
        printf("read returned %d\n", i);
        if(ioctl(fd, KDSKBMODE, K_XLATE) < 0) perror(
                "Warning: failed to change the virtual console to XLATE mode.");

}

If I have the test machine already on VT 1, run the program remotely
over ssh, and start banging on the test machine's keyboard, usually it
seems I do not see any scan codes at first.  If I start by holding
down a key other than the Enter key, there isn't much hope at all.  If
I hold down the Enter key, sometimes I see scan codes fairly soon, and
sometimes it requires several tries.

I tried turning the ICANON flag off via tcsetattr; it doesn't help.  I
tried copying the flags being set in kbdLinuxOn in the X code; it also
doesn't help.  If I use non-blocking I/O, then every read returns -1
regardless if a key has been pressed, regardless which key was being
held down and regardless how long.  If I use blocking I/O and select()
to check if a keystroke is available, then I get one keystroke
whenever there has been so much input that the line has wrapped, plus
an extra one in between; but most of the keystrokes never show up.

Oddly what did help was this scenario:  on the test machine I had an
open ssh session on VT 1, going to another machine.  From the other
machine I opened an ssh session to the test machine and ran the
program; so now the keyboard input which would have gone from the test
machine to the ssh client running on VT 1, was going to my program
instead (also on VT 1 because it opened /dev/tty1 directly).  I got
every keystroke instantly.  So I copied from openssh the tcsetattr
stuff (enter_raw_mode); no change in behavior.  I looked at kdrive
too... but regular X is the only program I know of that can use a VT
despite not having been started in a situation in which stdin is
itself a VT.  I want mine to work like that.

FWIW I'm testing mostly on a Zaurus but get similar results on my main
Linux PC so it's not just a Zaurus thing.

Sorry for bothering the list with this silly userspace stuff but I'm
running out of options, and there might even be a bug in there
somewhere.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists