Well, in that case it would be even simpler. We don't even need the "unplugged" flag check in udl_detect because all connectors are already unplugged in udl_usb_disconnect (in drm_connector_unplug_all()).

It is not necessary to check the flag in udl_fb_open() either, if the intention is to keep the device alive.

As for code paths that uses udl->udev, I only see the following 3 places:

(1) udl_parse_vendor_description and udl_alloc_urb_list: both are only called at udl_driver_load(), so not a problem

(2) udl_usb_probe: won't happen after USB device is unplugged

(3) udl_get_edid: only called from udl_get_modes, won't be an issue because connectors are already unplugged

So, it seems that I *really* just need to drop the "unplugged" flag and update the commit message.

On Thu, Feb 11, 2016 at 2:30 AM, David Herrmann <dh.herrmann@gmail.com> wrote:

Hi

On Thu, Feb 11, 2016 at 12:18 AM, Haixia Shi <hshi@chromium.org> wrote:
> When USB cable is disconnected, we mark udl device as unplugged so that
> udl_detect reports connector status as disconnected, but still keep
> the drm device alive until user-space closes it.
>
> Signed-off-by: Haixia Shi <hshi@chromium.org>
> Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>

I assume this is based on the discussion I had with Stephane on IRC.
I'm fine with going this way and keeping the device fully alive, and
just treat the device removal as a connector-unplug. However, you
really must document all this in your commit message.

Anyway, if you want to keep the device alive, then please change the
code to entirely drop the "unplugged" flag and all related code. Then
make sure you somehow reset "udl->udev" to NULL and make sure no
code-path uses the usb-device after it was unplugged. I guess there
should be appropriate locks already.

This way, you end up with a fully functional UDL device, which just
happens to have to connector plugged. But you *really* have to be
careful that no-one touches the usb device, as the interface might be
re-used by some other device any time (or in case the usb-core
provides protection against this, please document it).

Thanks
David