On Sun, 31 Mar 2019, Mikulas Patocka wrote:
>
>
> On Sat, 30 Mar 2019, Mikulas Patocka wrote:
>
> > Hi
> >
> > The patch a6dbe442755999960ca54a9b8ecfd9606be0ea75 ("vt: perform safe
> > console erase in the right order") introduces a bug.
> >
> > ---
> > drivers/tty/vt/vt.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > Index: linux-5.0.5/drivers/tty/vt/vt.c
> > ===================================================================
> > --- linux-5.0.5.orig/drivers/tty/vt/vt.c 2019-03-30 19:29:26.000000000 +0100
> > +++ linux-5.0.5/drivers/tty/vt/vt.c 2019-03-30 19:30:50.000000000 +0100
> > @@ -1518,7 +1518,7 @@ static void csi_J(struct vc_data *vc, in
> > return;
> > }
> > scr_memsetw(start, vc->vc_video_erase_char, 2 * count);
> > - update_region(vc, (unsigned long) start, count);
> > + do_update_region(vc, (unsigned long) start, count);
> ^^^^ this is wrong too - it will clear the screen if \e[2J is printed on
> inactive console. We need to use con_should_update(vc), just like it was
> before:
>
>
> The patch a6dbe442755999960ca54a9b8ecfd9606be0ea75 ("vt: perform safe
> console erase in the right order") introduces a bug.
>
> In order to reproduce the bug
> - use framebuffer console with the AMDGPU driver
> - type "links" to start the console www browser
> - press 'q' and space to exit links
>
> --- now, the cursor line will be permanently visible in the center of the
> screen. It will stay there until something overwrites it.
>
> Before the patch, there was a call to do_update_region, the patch changes
> it to update_region - and this seems to cause the bug with the cursor.
>
> The bug goes away if we change update_region back to do_update_region.
>
> Signed-off-by: Mikulas Patocka <mpatocka@xxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> Fixes: a6dbe4427559 ("vt: perform safe console erase in the right order")
The code in update_region(() does call do_update_region() conditionally
on con_should_update(vc).
But it does a set_cursor(vc) too which is the problem. So for the patch
below:
Acked-by: Nicolas Pitre <nico@xxxxxxxxxxx>
However there is another unconditional call to do_update_region() in
do_con_trol() (see case EShash) which might require a separate fix.
> ---
> drivers/tty/vt/vt.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> Index: linux-5.0.5/drivers/tty/vt/vt.c
> ===================================================================
> --- linux-5.0.5.orig/drivers/tty/vt/vt.c 2019-03-30 19:29:26.000000000 +0100
> +++ linux-5.0.5/drivers/tty/vt/vt.c 2019-03-31 13:06:24.000000000 +0200
> @@ -1518,7 +1518,8 @@ static void csi_J(struct vc_data *vc, in
> return;
> }
> scr_memsetw(start, vc->vc_video_erase_char, 2 * count);
> - update_region(vc, (unsigned long) start, count);
> + if (con_should_update(vc))
> + do_update_region(vc, (unsigned long) start, count);
> vc->vc_need_wrap = 0;
> }
>
>
