On Wed, 2020-07-08 at 09:49 +0200, Thomas Zimmermann wrote:
> Posting the GPU requires the correct DRAM type to be stored in
> struct ast_private. Therefore first initialize the DRAM info and
> then post the GPU. This restores the original order of instructions
> in this function.
I no longer have access to the relevant test POWER systems,
however the patch looks good to me.
> Signed-off-by: Thomas Zimmermann <tzimmermann@xxxxxxx>
> Fixes: bad09da6deab ("drm/ast: Fixed vram size incorrect issue on
> POWER")
> Cc: Joel Stanley <joel@xxxxxxxxx>
> Cc: Y.C. Chen <yc_chen@xxxxxxxxxxxxxx>
Acked-by: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx>
> Cc: Dave Airlie <airlied@xxxxxxxxxx>
> Cc: Thomas Zimmermann <tzimmermann@xxxxxxx>
> Cc: Gerd Hoffmann <kraxel@xxxxxxxxxx>
> Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
> Cc: Sam Ravnborg <sam@xxxxxxxxxxxx>
> Cc: Emil Velikov <emil.l.velikov@xxxxxxxxx>
> Cc: "Y.C. Chen" <yc_chen@xxxxxxxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> # v4.11+
> ---
> drivers/gpu/drm/ast/ast_main.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/ast/ast_main.c
> b/drivers/gpu/drm/ast/ast_main.c
> index b162cc82204d..87e5baded2a7 100644
> --- a/drivers/gpu/drm/ast/ast_main.c
> +++ b/drivers/gpu/drm/ast/ast_main.c
> @@ -418,15 +418,15 @@ int ast_driver_load(struct drm_device *dev,
> unsigned long flags)
>
> ast_detect_chip(dev, &need_post);
>
> - if (need_post)
> - ast_post_gpu(dev);
> -
> ret = ast_get_dram_info(dev);
> if (ret)
> goto out_free;
> drm_info(dev, "dram MCLK=%u Mhz type=%d bus_width=%d\n",
> ast->mclk, ast->dram_type, ast->dram_bus_width);
>
> + if (need_post)
> + ast_post_gpu(dev);
> +
> ret = ast_mm_init(ast);
> if (ret)
> goto out_free;
