On 11/20/16 1:29 , Jorge Schrauwen wrote: > > On 2016-11-19 17:11, Robert Mustacchi wrote: >> On 11/19/16 5:14 , Jorge Schrauwen wrote: >>> I just upgrade one of my compute nodes. >>> No crashes yet, but nothing plugged into the USB3 ports is working. >>> I tried a USB-keyboard and a USB3 Drive. >>> >>> [root@mantle ~]# mdb -ke '::prtusb' >>> INDEX DRIVER INST NODE GEN VID.PID PRODUCT >>> 1 xhci 0 pci15d9,813 3.0 0000.0000 No Product >>> String >>> 2 ehci 0 pci8086,7270 2.0 0000.0000 No Product >>> String >>> 3 hubd 0 hub 2.0 8087.07db No Product >>> String >>> 4 hubd 2 hub 2.0 0557.7000 No Product >>> String >>> 5 usb_mid 2 device 1.1 0557.2419 No Product >>> String >>> 6 hubd 1 hub 2.0 1a40.0101 USB 2.0 Hub >>> 7 scsa2usb 0 storage 2.1 0781.5583 Ultra Fit >>> 8 ugen 1 input 2.0 051d.0002 >>> Back-UPS RS 900G FW:879.L4 .I USB FW:L4 >>> 9 hubd 4 hub 1.1 058f.9254 Generic USB Hub >>> a usb_mid 3 device 2.0 046d.c52b USB Receiver >>> b usb_mid 4 device 1.1 10d5.55a2 2Port >>> KVMSwitcher >>> [root@mantle ~]# prtconf -dD | grep -i xhci >>> pci15d9,813 (pciex1912,14) [Renesas Technology Corp. >>> uPD720201 USB 3.0 Host Controller], instance #0 (driver name: xhci) >> >> Hmm, one of the Renesas controllers, not one we've seen yet. Also one of >> the ones more often reported to be troublesome. >> >> If you have the time, can you please plug things back into USB 3 ports, >> reboot, and then run the following: >> >> /usr/lib/xhci/xhci_portsc -v >> >> Along with the mdb -ke '::prtusb'. Could you also check if there's >> anything got logged about that? > > I added everything to a gist for easier reading: > https://gist.github.com/sjorge/47cb7f5e14b335c3c3c382dd3daf4e4f > > This is with the boot USB plugged into the first USB3 port (it always > works at boot, even before the xhci work) > The small USB2 hub with my keyboard (KVM) and UPS plugged in attached to > the second USB3 port.
Thanks for the additional information. It looks like one of your ports is stuck trying to transition in the device state machine. Unfortunately it's a bit hard to get a mapping between the virtual ports on the controller and the physical ports on the machine. I'm not entirely sure why it would be stuck in that state, from the spec it makes it sound like it's supposed to transition that way on its own without any intervention from the host. It does look like we detected one of the devices, but for some reason that fact didn't make it to the kernel per se. Would it be possible for you to run cfgadm -al in both cases? Sorry I don't have a more immediate answer for what's happening here. I'm not sure yet what's different about your setup from others -- though more likely than not there's some driver bug at the root of it all. Robert ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
