Hi Petr,
On 2026-10-02T21:19:31, Petr Hodina via B4 Relay
<[email protected]> wrote:
> video: console: clear rows stride-aware in set_row
>
> console_set_row() and console_set_row_2() clear a row with a single
> contiguous fill assuming line_length == xsize * pbytes. A stride-padded
> framebuffer inherited from a prior boot stage (e.g. via
> simple-framebuffer) breaks that assumption leaving stale pixels after
> scroll.
>
> Step per scanline by line_length instead matching the glyph renderer.
>
> Signed-off-by: Petr Hodina <[email protected]>
>
> drivers/video/console_normal.c | 16 +++++++++-------
> drivers/video/console_rotate.c | 16 +++++++++-------
> 2 files changed, 18 insertions(+), 14 deletions(-)
> diff --git a/drivers/video/console_normal.c b/drivers/video/console_normal.c
> @@ -29,11 +28,14 @@ static int console_set_row(struct udevice *dev, uint row,
> int clr)
> + for (j = 0; j < fontdata->height; j++) {
> + dst = line;
> + for (i = 0; i < vid_priv->xsize; i++)
> + fill_pixel_and_goto_next(&dst, clr, pbytes, pbytes);
> + line += vid_priv->line_length;
> + }
The old loop writes height * xsize pixels contiguously, so with a
padded stride it runs into the padding and stops short of the last
scanlines of the row.
BTW video_fill_part() already does a stride-aware fill and calls
video_damage(), so this could become something like:
return video_fill_part(dev->parent, 0, row * fontdata->height,
vid_priv->xsize,
(row + 1) * fontdata->height, clr);
That avoids open-coding the same loop here and in console_set_row_2().
It returns -ENOSYS for an unsupported bpix, so the
check_bpix_support() call could go too. What do you think? I am OK
with the patch as it is, if you prefer.
> diff --git a/drivers/video/console_rotate.c b/drivers/video/console_rotate.c
> @@ -112,18 +112,20 @@ static int console_set_row_2(struct udevice *dev, uint
> row, int clr)
> + void *start, *line, *dst;
> + int i, j;
> int pbytes = VNBYTES(vid_priv->bpix);
>
> start = vid_priv->fb + vid_priv->ysize * vid_priv->line_length -
> (row + 1) * fontdata->height * vid_priv->line_length;
> line = start;
Since you are changing this, start is no longer needed. You could
assign to line directly, as console_set_row() does.
Reviewed-by: Simon Glass <[email protected]>
Regards,
Simon
