Hi, Oliver Neukum <oneukum@xxxxxxxx> writes: > Am Mittwoch, den 11.01.2017, 21:06 +0200 schrieb Felipe Balbi: > >> Hi, >> >> Todd Brandt <todd.e.brandt@xxxxxxxxxxxxxxx> writes: >> > >> > On Mon, 2017-01-09 at 09:44 +0100, Oliver Neukum wrote: >> > Well my main purpose in getting this out there is to get the seed >> > planted for further optimization. The focus right now is just >> > advanced >> > users who want to optimize their personal machines. >> > >> > But you're right, it wouldn't be responsible to enable this feature >> > by >> > default in a distro like android or chromium. >> > >> > > >> > > For example it would be much safer to enable this for hubs of the >> > > first level and internal devices. >> >> I disagree. I think it's only safe to do so for devices known to be > > If you want absolute safety, yes. > Internal devices however are manufactured approximately at the same > time as the host, so we can hope they are improving with time. I'm thinking you have a romanticized view of internal devices. They are nothing more than common, off-the-shelf USB silicon that, instead of being packaged in a separate PCB with a USB connector, is layed out on the same PCB without the USB connector. They are just as likely to be non-compliant as any other device. > While as for external devices what is out there needs to be supported. > > In addition, they are a much smaller group for which we can hope > to get a reasonable testing coverage. A generic white list will always > be a fraction of all possibilities. not true. The full list of certified devices exists from usb.org. With idVendor:idProduct pairs and all. It's just a matter of exposing a "compliant" flag on sysfs and having a udev rule that parses the list from usb.org and sets "compliant" flag. > And the tests are harder as external devices are likelier to be > depowered during S3. And why does that matter? If the device is depowered, once it's powered again it'll be reenumerated, not resumed. It looks exactly as if you have just physically plugged the device. -- balbi
Attachment:
signature.asc
Description: PGP signature
