Re: [PATCH v4] usb: gadget: f_fs: add "no_disconnect" mode

From: Felipe Balbi
Date: Fri Nov 21 2014 - 10:56:39 EST


On Thu, Nov 13, 2014 at 05:28:22PM +0100, Krzysztof Opasiak wrote:
> Hello,
>
> > -----Original Message-----
> > From: Felipe Balbi [mailto:balbi@xxxxxx]
> > Sent: Thursday, November 13, 2014 3:44 AM
> > To: Felipe Balbi
> > Cc: Robert Baldyga; David Cohen; gregkh@xxxxxxxxxxxxxxxxxxx; linux-
> > usb@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > mina86@xxxxxxxxxx; andrzej.p@xxxxxxxxxxx; m.szyprowski@xxxxxxxxxxx;
> > k.opasiak@xxxxxxxxxxx
> > Subject: Re: [PATCH v4] usb: gadget: f_fs: add "no_disconnect" mode
>
> (...)
>
> > > > David, you have been messing with ffs lately, can I get a
> > Tested-by
> > > > from you on this patch ?
> > >
> > > ping. David ? Any chance you can test this one out ?
> >
> > a reminder on this request, it's getting close to tree lockdown
> > (probably another 2 to 3 weeks).
>
> I have tested this patch today but I'm afraid that I can't give
> tested-by tag:
>
> When I kill the daemon and then reconnect my device I get on my host:
>
> [776092.877175] usb 1-1.1: can't set config #1, error -32
>
> While on device there is only:
>
> [ 2147.035340] configfs-gadget gadget: high-speed config #1: The only
> one
>
> Moreover, I'm not very enthusiastic about the convention in which we
> allow
> to reconnect the device without any changes.
>
> In my opinion when daemon has been killed we should allow user to finish
> communication with other functions and then simply unbind our gadget.
> This should happen not only on reconnection but also when host is trying
> to enable configuration which contains this function "zombie".
> This function is not fully operational so it is very good idea to
> allow user to finish his job with other functions but when he reconnects
> the
> device we should unbind gadget to notify userspace that something is
> wrong.
> The same when we switch configuration.
>
> What do you think about this Michal and Felipe?

let's wait another merge window to discuss this one better. If it's
causing confusion/regressions, we better delay it.

--
balbi

Attachment: signature.asc
Description: Digital signature