On Mon, Aug 31, 2015 at 05:28:10PM +0300, Mathias Nyman wrote: > The max exit latency should tell how long it can maximum take for the entire link between host > and device to wake up and be fully functional. Isoc transfer will send a PING TP max exit > latency before the first data transfer to make sure link is up and running in time. > This is only for SuperSpeed Isoc transfers > > So either we have miscalculated to Max exit transfer, and link really isn't yet ready, > or then there could be some other powermanagement issue with getting the link up and running. I noticed another thing: I rewrote my application to have the USB thread run with realtime priority, and also move all disk I/O off that thread. This made it run with no drops on CONFIG_PM=n... but also fixed the error with CONFIG_PM=y! (Ran for ten hours straight without a single drop.) So it would seem the ping timeout only happens when I have overrun problems anyway. Now, they do seem to make the overrun problems a lot _worse_ when they first happen, so I think it should be fixed anyway, but it might be useful to try to narrow down what's going on. Perhaps the device goes into power-saving mode when it detects overflow? > The "ERROR Transfer event TRB DMA ptr not part of curren..." errors are a bit worrying. > I was hoping these would no longer occur after the last off by one fix: > > commit 7895086afde2a05fa24a0e410d8e6b75ca7c8fdd > xhci: fix off by one error in TRB DMA address boundary check That's in 3.12? Then I definitely have them included. But I also haven't seen them since that single occurrence. /* Steinar */ -- Homepage: http://www.sesse.net/ -- 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
