On Tue, Oct 09, 2012 at 03:36:59PM -0400, Alan Stern wrote: > On Tue, 9 Oct 2012, Sarah Sharp wrote: > > > Hmm, so I've tested the VIA hubs under a different Intel chipset (Lynx > > Point), and they exhibit the same issues as when a bus analyzer is > > connected on Panther Point. Only the issues show up without the > > analyzer in between. > > Come to think of it, you might want to report the Panther Point results > to the company that made the bus analyzer. It's the sort of thing they > should be interested in fixing. Yeah, I should send them an email. They also display the Set SEL command wrong, so I don't think the LPM portion of their firmware/software is very well tested. > > I'll do some more triaging, but it looks like some devices may work with > > LPM enabled on some hosts and not others. > > Have you spoken to the hardware people at Intel about this? They > should be interested in hearing about failures in their host > controllers. It could be that the device is marginal electrically, and putting the analyzer in between just pushes it over the edge. Maybe the new host controller isn't as tolerant of the signaling as the old host controller. I'm not a hardware expert, so yes, I'm talking to Intel's chipset team. > > What do you think about > > dynamically adding device VID:PIDs to a blacklist? > > How safe would it be to populate this list? That is, how sure can you > be that once a device fails with LPM enabled, you can get it to work > again simply by turning off LPM? When LPM fails, the device often goes into the Inactive state, which also marks the port as not connected. The USB core will issue a warm reset to the device and then do a logical disconnect. VIA hubs come back fine after the warm reset, but that does mean the user will see a disconnect and reconnect of devices beneath the tree. > Also, how annoying will it be for users if their device fails and has > to be reset (or replugged) every time they boot? Yep, you're right that it would be pretty annoying. I'm leaning towards the static blacklist in the kernel. One of the things I've found out is that VIA hubs keep their firmware revision in the bcdDevice field of their device descriptor. If I find one version of the VIA firmware works with LPM, can the quirks table create a rule that would effectively be something like "bcdDevice < 0x9a90"? Sarah Sharp -- 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
