[PATCH] drm/bridge: ti-sn65dsi86: Associate DSI device lifetime with auxiliary device

[Stephen Boyd]

Quoting Doug Anderson (2023-10-02 17:31:41)
> Hi,
>
> On Mon, Oct 2, 2023 at 4:54 PM Stephen Boyd <swboyd at chromium.org> wrote:
> >
> > The kernel produces a warning splat and the DSI device fails to register
> > in this driver if the i2c driver probes, populates child auxiliary
> > devices, and then somewhere in ti_sn_bridge_probe() a function call
> > returns -EPROBE_DEFER. When the auxiliary driver probe defers, the dsi
> > device created by devm_mipi_dsi_device_register_full() is left
> > registered because the devm managed device used to manage the lifetime
> > of the DSI device is the parent i2c device, not the auxiliary device
> > that is being probed.
> >
> > Associate the DSI device created and managed by this driver to the
> > lifetime of the auxiliary device, not the i2c device, so that the DSI
> > device is removed when the auxiliary driver unbinds. Similarly change
> > the device pointer used for dev_err_probe() so the deferred probe errors
> > are associated with the auxiliary device instead of the parent i2c
> > device so we can narrow down future problems faster.
> >
> > Cc: Douglas Anderson <dianders at chromium.org>
> > Cc: Maxime Ripard <maxime at cerno.tech>
> > Fixes: c3b75d4734cb ("drm/bridge: sn65dsi86: Register and attach our DSI device at probe")
>
> Even before that commit I think it was using the main "dev" instead of
> the auxiliary device's "dev" for some "devm" stuff. I guess the
> difference is that it wouldn't mess with probe deferral? Searching
> back, I think the first instance of a case that was using "devm_" with
> the wrong device was commit 4e5763f03e10 ("drm/bridge: ti-sn65dsi86:
> Wrap panel with panel-bridge")? Would it make sense to use that as a
> Fixes, you think?

The problem for me is that the dsi device is registered twice. That
happens because probe for the auxiliary device happens twice. I was
cautious about the fixes tag here because it didn't look like probe
deferral was happening before commit c3b75d4734cb.

>
> In any case, this looks reasonable to me:
>
> Reviewed-by: Douglas Anderson <dianders at chromium.org>
>
> I'll give it a week and then apply to "-fixes" if everything is quiet.

Thanks!