Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 27 Oct 2023 09:54:53 -0700
From:      bob prohaska <fbsd@www.zefox.net>
To:        freebsd-arm@freebsd.org
Subject:   Releng/14.0 dropping ssh connections to USB-serial adapter
Message-ID:  <ZTvrXXpzPW89DxFP@www.zefox.net>

next in thread | raw e-mail | index | archive | help
It appears that releng/14.0 is still dropping ssh connections to a
USB-serial adapter, though the problem seems absent now on -current.
Usbconfig reports

ugen1.4: <FTDI USB - Serial> at usbus1, cfg=0 md=HOST spd=FULL (12Mbps) pwr=ON (90mA)

The USB-end host is a Pi2 v1.1 running armv7, the usb adapter is 
an FTDI FT232 and the serial end is the serial console of a Pi3.
The adapter is plugged directly into the Pi, no USB hub is used.
The root filesystem is on a USB mechanical hard disk connected
via a powered hub and booting is vi microSD "bootcode.bin" mode.

With USB debugging turned on quite a lot of output is generated,
the only part I recognize is a "ucom_shutdown" on the USB end's
console. That makes termination of the serial connection look
intentional. Why the ssh connection to the host exits entirely
isn't evident to me. 

When the connection has dropped, restoring it is straightforward:

bob@generic:~ % su
Password:
# tip ucom
Stale lock on cuaU0 PID=1201... overriding.
connected
authentica [I didn't type this, it somehow got into the data stream]
Password: [I simply hit Enter]
Login incorrect
login: 

It looks as if part of the serial console login prompt has somehow 
leaked into the _inbound_ data stream. This is seen on other Pi hosts
in my collection, too, but seems unrelated to connections dropping. 
 
Meanwhile, the console of the USB-end host reports:

login: Oct 27 08:21:23 generic su[3278]: bob to root on /dev/pts/0
ucom_open: tp = 0xd6c3f400
ucom_cfg_open: 
ucom_dtr: onoff = 1
ucom_line_state: on=0x01, off=0x00
ucom_rts: onoff = 1
ucom_line_state: on=0x02, off=0x00
ucom_ring: onoff = 0
ucom_line_state: on=0x00, off=0x08
ucom_break: onoff = 0
ucom_line_state: on=0x00, off=0x04
ucom_status_change: 
ucom_param: sc = 0xd6c3f858
ucom_dtr: onoff = 1
ucom_line_state: on=0x01, off=0x00
ucom_rts: onoff = 1
ucom_line_state: on=0x02, off=0x00
ucom_ioctl: cmd = 0x402c7413
ucom_ioctl: cmd = 0x802c7416
ucom_outwakeup: sc = 0xd6c3f858
ucom_outwakeup: sc = 0xd6c3f858
ucom_get_data: cnt=13
ucom_inwakeup: tp=0xd6c3f400
ucom_inwakeup: tp=0xd6c3f400
ucom_get_data: cnt=0
ucom_ioctl: cmd = 0x2000740d
ucom_ioctl: cmd = 0x402c7413
ucom_ioctl: cmd = 0x802c7416
ucom_inwakeup: tp=0xd6c3f400
ucom_inwakeup: tp=0xd6c3f400
ucom_param: sc = 0xd6c3f858
ucom_ioctl: cmd = 0x8004667e
ucom_ioctl: cmd = 0x8004667d
ucom_get_data: cnt=0
ucom_inwakeup: tp=0xd6c3f400
ucom_inwakeup: tp=0xd6c3f400
ucom_inwakeup: tp=0xd6c3f400
ucom_outwakeup: sc = 0xd6c3f858
ucom_get_data: cnt=1
ucom_get_data: cnt=0
ucom_inwakeup: tp=0xd6c3f400

Hitting Enter restores the console login prompt.

The connections usually drop over the span of a few hours,
seemingly a bit faster when the host is active, as when
running buildworld, but I've not tested carefully.

This connection dropped after a bit over an hour. The
output generated is huge on both ends so I'll place it at 
http://www.zefox.net/~fbsd/rpi2/ssh_drops_usb_serial

Suggestions for betters tests are welcome!

Thanks for reading,

bob prohaska




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ZTvrXXpzPW89DxFP>