On Sat, 2016-07-09 at 20:31 +0200, Bjørn Mork wrote:
> [removed the stable CC since this discussion isn't relevant to stable
> anymore]
>
> Oliver Neukum <oneukum@xxxxxxxx> writes:
[..]
> How about splitting the behaviour locally in cdc-wdm, keeping the
> current behaviour for CDC WDM devices and changing to "drain on open"
> for the netdev drivers? Something like this plus the necessary logic
> dealing with the "drain_on_open" ("next_desc" is a label in wdm_probe):
Even better solution.
[..]
> Yes. We could alternatively filter the EPIPE from read(), since it
> isn't supposed to happen anyway. But it's not going to look less ugly :(
I'd rather not. Dropping errors really is evil.
> Execpt for the extreme ugliness, I don't think it will hurt to apply
> this unconditionally for all the network drivers using cdc-wdm, as long
> as it is limited to open only.
It is very hard to see how would can avoid it in the long run in resume.
> I certainly want to avoid any blacklist. Device IDs are cheap in this
> market. The MBIM modems typically run Android and the device ID is
> configured in NVRAM for whatever OEM laptop vendor it is sold to.
> Having a catch-all class driver for MBIM is an absolute requirement.
> Making it depend on lists of devices is not an option, IMHO.
Obviously. I am just not optimistic that we can do without specific
exceptions in the long run.
Regards
Oliver
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
