> On Sat, Nov 28, 2015 at 02:02:57PM +0100, Mark Kettenis wrote: > > If you find any regressions, please send me: > > > > * dmesg (with and without this diff) > > * pcidump -vxx (with and without this diff) > > First off, it's great to have the Apple Thunderbolt to Ethernet adapter > working! Thank you! I did this round of testing on the MacBookAir7,2 > (2015 Broadwell model). I have four different dmesgs and pcidump outputs > following. The regression I found is a panic if I unplug the Apple > Thunderbolt to Ethernet adapter while using the ppb.c patch if I booted > with the adapter plugged in. The panic that results is shown as: > > panic: config_detach: forced detach of bge0 failed (45)
Yes, it is known that the bge driver does not currently have detach code. There's a bit of a story to that. Detach code in our tree initially arrived for pcmcia, then later for usb, then cardbus... PCI hotplug only arrived when pci express connectors showed up on laptops, and Mark wrote code for that. The first device we could add support for msk(4) ethernet. Other devices were hard to find. Many devices are not really PCI, but use the usb-passthrough found on the same connector. Around that time, I found a fairly cheap pciexpress to pci32 cage. Maybe that was late 2009? I plugged a variety of cards into it, and tried to make a few of them hotplug. em, fxp, ... seemed kind of pointless, but a few changes made did help suspend/resume logic of those drivers (so it ended up making a difference later). I plugged a bge(4) into that chassis back in those days, but never got detach working correctly; late attach was also flaky. I think back in those days BIOS ROM code had to be run to initialize the devices correctly. Mark has told me he thinks this chip model is likely to come up clean, so let's see what he gets working!
