Re: USB hang setting ttyflags

[Patrick Welche <prlw1%cam.ac.uk@localhost>]

On Thu, May 12, 2016 at 09:08:39AM +0100, Patrick Welche wrote:
> Randomly, my -current/amd64 laptop hangs at "Setting tty flags".
> It seems this is because I have (for a long time now)
> 
> ttyU0   "/usr/libexec/getty std.9600"   unknown off local
> ttyU3   "/usr/libexec/getty std.9600"   unknown off local
> 
> in /etc/ttys. They are there on the off-chance that I have a ulpcom
> plugged in at boot or thereafter. There is a built-in
> 
> umodem0 at uhub3 port 6 configuration 1 interface 1
> umodem0: TOSHIBA F5521gw, rev 2.00/0.00, addr 5, iclass 2/2
> umodem0: data interface 2, has CM over data, has break
> umodem0: status change notification available
> ucom0 at umodem0
> 
> It seems that ttyflags get stuck in usbxfer. It remains stuck there
> after a ctl-c. Thereafter things like dev_mkdb get stuck in tstile.
> The only way out seems to be to reboot and roll the dice again.

Happened again Jul 25 2016 kernel:
PID    LID S CPU     FLAGS       STRUCT LWP *               NAME WAIT
78       1 3   2   8020000   fffffe810f9042e0           ttyflags usbxfer


crash> bt/a fffffe810f9042e0
trace: pid 78 lid 1 at 0xfffffe810fa1cb70
sleepq_block() at sleepq_block+0x97
cv_wait() at cv_wait+0x9e
usbd_transfer() at usbd_transfer+0x18d
usbd_do_request_flags() at usbd_do_request_flags+0x9e
umodem_set_line_state.isra.1() at umodem_set_line_state.isra.1+0x56
umodem_set() at umodem_set+0x5a
ucom_cleanup() at ucom_cleanup+0x42
ucomclose() at ucomclose+0x12c
spec_close() at spec_close+0x11a
VOP_CLOSE() at VOP_CLOSE+0x33
vn_close() at vn_close+0x4d
closef() at closef+0x4a
fd_close() at fd_close+0x108
sys_close() at sys_close+0x1f
syscall() at syscall+0x1a4