Re: [PATCH] drm/omap: dsi: Fix deferred probe warnings

[Tony Lindgren]

* H. Nikolaus Schaller <hns@xxxxxxxxxxxxx> [230916 12:50]:
> Hi Tomi and Tony,
> 
> > Am 13.09.2023 um 13:59 schrieb Tomi Valkeinen <tomi.valkeinen@xxxxxxxxxxxxxxxx>:
> > 
> > On 12/04/2023 10:39, Tony Lindgren wrote:
> >> We may not have dsi->dsidev initialized during probe, and that can
> >> lead into various dsi related warnings as omap_dsi_host_detach() gets
> >> called with dsi->dsidev set to NULL.
> >> The warnings can be "Fixed dependency cycle(s)" followed by a
> >> WARNING: CPU: 0 PID: 787 at drivers/gpu/drm/omapdrm/dss/dsi.c:4414.
> >> Let's fix the warnings by checking for a valid dsi->dsidev.
> >> Signed-off-by: Tony Lindgren <tony@xxxxxxxxxxx>
> >> ---
> >>  drivers/gpu/drm/omapdrm/dss/dsi.c | 2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >> diff --git a/drivers/gpu/drm/omapdrm/dss/dsi.c b/drivers/gpu/drm/omapdrm/dss/dsi.c
> >> --- a/drivers/gpu/drm/omapdrm/dss/dsi.c
> >> +++ b/drivers/gpu/drm/omapdrm/dss/dsi.c
> >> @@ -4411,7 +4411,7 @@ static int omap_dsi_host_detach(struct mipi_dsi_host *host,
> >>  {
> >>  	struct dsi_data *dsi = host_to_omap(host);
> >>  -	if (WARN_ON(dsi->dsidev != client))
> >> +	if (dsi->dsidev && WARN_ON(dsi->dsidev != client))
> >>  		return -EINVAL;
> >>    	cancel_delayed_work_sync(&dsi->dsi_disable_work);
> > 
> > Shouldn't this rather be
> > 
> > if (!dsi->dsidev)
> > 	return 0;
> > 
> > before the if (WARN_ON(dsi->dsidev != client)) line?
> 
> Yes you are right. We have a different variant in our Pyra kernel:
> 
> What we currently have in our Pyra tree is: https://git.goldelico.com/?p=letux-kernel.git;a=commitdiff;h=5bf7bd64eec1eb924e794e8d6600919f0dae8c5a;hp=27a0cd6263194d1465e9c53293d35f8c8c988f9d
> 
>         struct dsi_data *dsi = host_to_omap(host);
>  
> -       if (WARN_ON(dsi->dsidev != client))
> +printk("%s\n", __func__);
> +
> +       if (!dsi->dsidev || WARN_ON(dsi->dsidev != client))
>                 return -EINVAL;
>  
>         cancel_delayed_work_sync(&dsi->dsi_disable_work);
> 
> > 
> > If dsi->dsidev is NULL, then attach hasn't been called, and we shouldn't do anything in the detach callback either.
> > 
> > With your change we'll end up doing all the work in the detach callback, without ever doing their counterpart in the attach side.
> 
> If useful, I can post above mentioned patch (without printk).

Sounds good to me.

Thanks,

Tony