> > ASPM must be configured on both ends of the link, so for completeness, > > can you also collect the "PCI" output for the bridge leading to the > > 7260 device? Based on the Linux lspci output, this should be > > 0000:00:1c.1. > > > > And I assume the device works well with the Windows driver? > > > Here are the "PCI" and "PCI Index" data for 0000:00:1c:1. > > And yes the NIC works nice in Windows. > Small update from the bugzilla (https://bugzilla.kernel.org/show_bug.cgi?id=64541). The bug is solved... There seem to be a hardware bug with the L1 OFF exit timer. To solve this bug we need *not* to rely on the internal clock and need to keep using the external clock. The internal clock isn't reliable enough and can lead to loss of synchronization between the bridge and the device upon L1 OFF transition. This issue has been seen in simulation and not on real hardware... until now... The windows driver has a workaround for this hardware bug, this is why the issue wasn't seen on Windows. I am porting the work around to the Linux driver. Thank you wzyboy for your patience... Then end. ��.n��������+%������w��{.n�����{���"�)��jg��������ݢj����G�������j:+v���w�m������w�������h�����٥