[PATCH] Fix KDFONTOP 32bit compatibility layer

[PATCH] Fix KDFONTOP 32bit compatibility layer

[Samuel Thibault]

KDFONTOP(GET) currently fails with EIO when being run in a 32bit
userland with a 64bit kernel if the font width is not 8. This is because
the compatibility layer introduced by e9216651 forces the addition of
the KD_FONT_FLAG_OLD flag, which makes con_font_get return EIO in such
case.  This flag should *not* be set for KDFONTOP, since it's actually
the whole point of this flag (see comment in con_font_set for instance).

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

diff --git a/drivers/tty/vt/vt_ioctl.c b/drivers/tty/vt/vt_ioctl.c
index 5e096f4..65447c5 100644
--- a/drivers/tty/vt/vt_ioctl.c
+++ b/drivers/tty/vt/vt_ioctl.c
@@ -1463,7 +1463,6 @@ compat_kdfontop_ioctl(struct compat_console_font_op __user *fontop,
 	if (!perm && op->op != KD_FONT_OP_GET)
 		return -EPERM;
 	op->data = compat_ptr(((struct compat_console_font_op *)op)->data);
-	op->flags |= KD_FONT_FLAG_OLD;
 	i = con_font_op(vc, op);
 	if (i)
 		return i;

Re: [PATCH] Fix KDFONTOP 32bit compatibility layer

[Arnd Bergmann]

On Saturday 28 January 2012, Samuel Thibault wrote:
> KDFONTOP(GET) currently fails with EIO when being run in a 32bit
> userland with a 64bit kernel if the font width is not 8. This is because
> the compatibility layer introduced by e9216651 forces the addition of
> the KD_FONT_FLAG_OLD flag, which makes con_font_get return EIO in such
> case.  This flag should not be set for KDFONTOP, since it's actually
> the whole point of this flag (see comment in con_font_set for instance).
> 
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

Hmm, this flag was not introduced in the patch you cite, it used to
live in fs/compat_ioctl.c before that, since before the start of the
git history. It seems to date back on the original sparc64 implementation
of do_kdfontop_ioctl that was written by Eddie Dost in 1998.

> diff --git a/drivers/tty/vt/vt_ioctl.c b/drivers/tty/vt/vt_ioctl.c
> index 5e096f4..65447c5 100644
> --- a/drivers/tty/vt/vt_ioctl.c
> +++ b/drivers/tty/vt/vt_ioctl.c
> @@ -1463,7 +1463,6 @@ compat_kdfontop_ioctl(struct compat_console_font_op __user *fontop,
>         if (!perm && op->op != KD_FONT_OP_GET)
>                 return -EPERM;
>         op->data = compat_ptr(((struct compat_console_font_op *)op)->data);
> -       op->flags |= KD_FONT_FLAG_OLD;
>         i = con_font_op(vc, op);
>         if (i)
>                 return i;

>From all I can tell, the patch looks correct, but please update the
description so you don't blame my innocent patch ;-)

Reviewed-by: Arnd Bergmann <arnd@arndb.de>

Re: [PATCH] Fix KDFONTOP 32bit compatibility layer

[Andrew Morton]

On Tue, 31 Jan 2012 16:35:36 +0000
Arnd Bergmann <arnd@arndb.de> wrote:

> On Saturday 28 January 2012, Samuel Thibault wrote:
> > KDFONTOP(GET) currently fails with EIO when being run in a 32bit
> > userland with a 64bit kernel if the font width is not 8. This is because
> > the compatibility layer introduced by e9216651 forces the addition of
> > the KD_FONT_FLAG_OLD flag, which makes con_font_get return EIO in such
> > case.  This flag should not be set for KDFONTOP, since it's actually
> > the whole point of this flag (see comment in con_font_set for instance).
> > 
> > Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> 
> Hmm, this flag was not introduced in the patch you cite, it used to
> live in fs/compat_ioctl.c before that, since before the start of the
> git history. It seems to date back on the original sparc64 implementation
> of do_kdfontop_ioctl that was written by Eddie Dost in 1998.
> 
> > diff --git a/drivers/tty/vt/vt_ioctl.c b/drivers/tty/vt/vt_ioctl.c
> > index 5e096f4..65447c5 100644
> > --- a/drivers/tty/vt/vt_ioctl.c
> > +++ b/drivers/tty/vt/vt_ioctl.c
> > @@ -1463,7 +1463,6 @@ compat_kdfontop_ioctl(struct compat_console_font_op __user *fontop,
> >         if (!perm && op->op != KD_FONT_OP_GET)
> >                 return -EPERM;
> >         op->data = compat_ptr(((struct compat_console_font_op *)op)->data);
> > -       op->flags |= KD_FONT_FLAG_OLD;
> >         i = con_font_op(vc, op);
> >         if (i)
> >                 return i;
> 
> From all I can tell, the patch looks correct, but please update the
> description so you don't blame my innocent patch ;-)
> 
> Reviewed-by: Arnd Bergmann <arnd@arndb.de>

Yes, we should note the fact that at least one kernel bug isn't your
fault ;)

I changed the changelog to

: KDFONTOP(GET) currently fails with EIO when being run in a 32bit userland
: with a 64bit kernel if the font width is not 8.
: 
: This is because of the setting of the KD_FONT_FLAG_OLD flag, which makes
: con_font_get return EIO in such case.  
: 
: This flag should *not* be set for KDFONTOP, since it's actually the whole
: point of this flag (see comment in con_font_set for instance).
: 
: Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
: Reviewed-by: Arnd Bergmann <arnd@arndb.de>
: Cc: Greg Kroah-Hartman <gregkh@suse.de>
: Cc: Arthur Taylor <art@ified.ca>
: Cc: Jiri Slaby <jslaby@suse.cz>
: Cc: Jiri Olsa <jolsa@redhat.com>
: Cc: <stable@vger.kernel.org>