plug fails after powercut; cannot get into uboot; connect-debounce errors & more

[kyle]

! note - as per instructions:
 * (reading from back of plug) sheeva plug model:003-sp1001
 * can't give uboot environment as can't get to marvell prompt

-----TL/DR version-------
current status is that if I plug in the sheeva it keeps printing 'connect-debounce failed, port 6 disabled' on host pc until the blue light comes on. then I can connect via screen, but get a blank screen with a flashing cursor which accepts no keystrokes
-----------------------------

Hello folks, I've got a problem with my sheevaplug which happened after a brief powercut. Below is a quick run down of the things I've tried to fix it - I apologise for it being a little chaotic, but a lot of the time I couldn't get it to repeat errors and so some of the stuff is from memory

first a summary of what I did, then I'll detail the error messages.

-------

* first thing I noticed was I couldn't ssh in. I went up and connected the usb to the laptop and tried via screen. I did manage to get in, but I did't get a full prompt, I got "PTY allocation request failed on channel 0". I could run basic commands, but nothing which involves complex output (htop, tmux etc) "I got error opening terminal: unknown". I did then manage to mount the sheeva via fuse at this point and back everything up.

* Hence, I figured it wasn't a serious problem, but after experimenting and rebooting I found it was temperamental - sometimes kicking out masses of error messages during the marvell reboot, sometimes just giving me a dead flashing cursor.

* since it was all backed up, I tried to do a clean install. It all seemed to go ok, but I had the same problems connecting and the same temperamental behaviour. I checked the sd card and it's a clean arch install

* after playing around a bit more, I decided to follow the instructions for debricking (at http://www.newit.co.uk/forum/index.php/topic,2835.0.html). it all completed fine, no errors, except that the final stage I could not see the sheeva rebooting in the first terminal - just nothing, not accepting input either, but screen is up and happy that ttyusb0 is there.

* I subsequently realised that the thing didn't seem to be rebooting properly when I took the power out. If I hit the reset button or only kill the power for a short time, the blue light comes back on almost immediately. If I leave the power out for half an hour, the green light comes on first for a minute or two, flashing, before the blue light comes up, as if it's doing a full reboot

* so, I attempted unbricking again, taking care to kill the power fully, but this time I got the errors detailed below

-------
ERROR MESSAGES
-------

* at first, if I took the power out and put it back in with the sd card (fully formatted with arch arm) I got the following messages (or variants thereof) flashing up on the host terminal: 
 * unable to enumerate usb device
 * device not accepting address 82
 * device descriptor read/all error -71
* it then won't allow me to connect via screen at all

* if I take the power out and leave it out for a while, then plug it back in with no sd card, I get no errors from host machine and I can connect via screen - BUT - I get nothing: just a flashing cursor, cannot input anything. dmesg gives me the following:

[ 6437.762820] hub 1-0:1.0: connect-debounce failed, port 6 disabled
[ 6438.489515] usb 4-2: new full-speed USB device number 44 using uhci_hcd
[ 6438.697397] usb 4-2: Ignoring serial port reserved for JTAG
[ 6438.702409] ftdi_sio 4-2:1.1: FTDI USB Serial Device converter detected
[ 6438.702495] usb 4-2: Detected FT2232C
[ 6438.702501] usb 4-2: Number of endpoints 2
[ 6438.702506] usb 4-2: Endpoint 1 MaxPacketSize 64
[ 6438.702511] usb 4-2: Endpoint 2 MaxPacketSize 64
[ 6438.702515] usb 4-2: Setting MaxPacketSize 64
[ 6438.704360] usb 4-2: FTDI USB Serial Device converter now attached to ttyUSB0


* connect-debounce is the recurring theme throughout all my attempts

* often after rebooting it will spit out the connect-debounce error a dozen times or so until the blue light comes on, then it stops. at this point I can connect with screen via usb and get no errors, but I get no prompt and it doesn't respond - just a blinking cursor. if I now hit the physical reset button, screen will disconnect, BUT, the blue light comes back on almost immediately

* if I do a full reboot (power out for long time) and start up with a working sd card in i get the same errors (connect-debounce) until the blue light comes on, then it starts spitting out "unable to enumerate usb device" and "unable to read config" and "can't set config #1, error -71"

* Below is what I get when follwing the unbricking instructions:

Open On-Chip Debugger 0.5.0 (2011-12-17-08:00)
Licensed under GNU GPL v2
For bug reports, read
 http://openocd.berlios.de/doc/doxygen/bugs.html
Info : only one transport option; autoselect 'jtag'
2000 kHz
trst_and_srst separate srst_gates_jtag trst_push_pull srst_open_drain
adapter_nsrst_delay: 200
jtag_ntrst_delay: 200
dcc downloads are enabled
Warn : use 'feroceon.cpu' as target identifier, not '0'
sheevaplug_load_uboot
Info : clock speed 2000 kHz
Info : JTAG tap: feroceon.cpu tap/device found: 0x20a023d3 (mfg: 0x1e9, part: 0x0a02, ver: 0x2)
Info : Embedded ICE version 0
Info : feroceon.cpu: hardware has 1 breakpoint/watchpoint unit
Info : accepting 'telnet' connection from 4444
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
Error: TRST/SRST error
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
error: -104

in procedure 'reset'
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
Error: TRST/SRST error
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
Error: TRST/SRST error
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
Error: TRST/SRST error
Error: ftdi_write_data: usb bulk write failed
Error: couldn't write MPSSE commands to FT2232
error: -104


If anyone can offer any pointers they'd be most appreciated as I rely on my sheeva for everything. I'm posting here rather than on plugcomputing forums as I bought the device from New IT and they gave me excellent advice before. If needed, I can dig out the invoice somewhere, but obviously everything's a bit chaotic as all documents and emails were stored on the sheeva.

Thanks

Kyle

[Confusticated]

An unstable power supply could possibly cause such problems
http://www.newit.co.uk/forum/index.php/topic,353.0.html

[kyle]

yeah, that was my worry, as the New IT guy told me before that my plug was one of the originals with the dodgy power supply. Do NewIT people monitor the forums or should I fire off an email? That thread said there was a 2 year warantee.

[NewIT_James]

Hi Kyle,

From your description I would guess the powercut and spike that probably went with it has damaged the jtag in your plug.

Can you still SSH into the plug when it boots?

[kyle]

Hi James, no I can't now, as I reinstalled and can't even get access via usb. But after it happened I did have ssh access, but not a proper prompt. I was getting the "PTY allocation request failed on channel 0" and then "terminal unknown" messages. I should probably have stopped there and posted here, but to be honest I thought it was a software/config error that would be solved by a reinstall.

As I said, it doesn't really boot. I'm guessing the behaviour I'm seeing with the lights is not normal? That is to say, it doesn't seem to reboot fully unless power is diconnected for a long time

Thanks

[NewIT_James]

You say you are getting the green followed by the blue light when booting which suggests uboot is working.

Is it configured to multiboot and if so have you tried booting to an SD image?

 

[kyle]

sorry I don't really know what you mean by multiboot, but it was running from an sd card before. The only thing I've ever really done at the marvell uboot prompt is install arch linux. Is the behaviour with the lights normal? That it doesn't reset itself unless all power is cut for a long time?

I can't find any info about the connect-debounce errors or any of the others I'm encountering.

[Confusticated]

The dodgy '5v into the USB port' trick might resolve if it is the power supply or not...or measure it to see what is coming out.

[NewIT_James]

but it was running from an sd card befor

That's a multiboot.

So without the SD installed and booted to the Nand can you access it via SSH?

Have you tried copying a new image to the SD card?

http://www.downloadsnewit.co.uk/SD-images/Sheevaplug/

Try an 'all' image if that doesn't work try a 'white-plug' image.

[kyle]

sorry, I know my original post was long and I should have edited down, but as I said in that, I did a reinstall and it all went fine. I've checked the sd card and it installed a clean copy of arch - all looks fine, so that's not the problem. And no, it won't let me ssh into it without (or with) the ssd card. As I said, it won't boot to anything - at least, nothing I can access via usb. I get the "connect-debounce" error message on the host pc when I boot up without the sd card.

As for 5v in the usb - I always found it odd that the lights stayed on even when the power was disconnected and I would have to remove the usb to get it to reboot properly. What should I do to test it?

[NewIT_James]

Have you tried a different cable?

Have you tried a different Host PC or at least a different usb port on the host you are using?

[kyle]

yep, both :-)
two different cables (both of which I know work as I use every day) and two different laptops. It's been two days, I've tried everything, my brain is close to imploding
--- edit ---
to clarify, two different laptops, random usb ports on both, plus the different cables

[kyle]

ok, interesting developments...
so, this morning I was setting up my laptop as a temporary server in place of the sheevaplug, and when I plugged in my external hdds I was getting a load of "unable to enumerate usb device on port X" errors, which was one of the same errors I got with the sheeva. So after a bit of searching I found that "modprobe -r ehci_hcd" stopped that problem, and it was probably related to a kernel bug/timing issue.

Next step was obviously to try the sheeva again. After a few attempts, I did manage to get a marvell prompt, but stupidly before doing any diagnostics I thought I'd solved the problem and moved it over to the other laptop, and that just gave me the dead screen. Since then I've had nothing but dead screen (meaning gnu screen connects, but just has a dead blinking cursor) or not recognized at all. There doesn't seem to be any pattern to any of this.

Any ideas? It's possible it was a kernel update that broke everything. If no-one else has any obvious suggestions I'll try downgrading the kernel on one of my machines, but i'll have to read up about it first as i've never done it

!!! update, I can now reliably get into a working marvell prompt if I reboot the laptop with the usb connected! however, if I disconnect it and reconnect, I'm back to a dead screen. At last, something repeatable! - note (to self), this is true without killing the ehci_hcd module

[kyle]

quick note - I just booted successfully into the fresh archarm install on the sd card. All looking good, going to keep experimenting

[kyle]

ok, I've figured out a few more things. Most important is what was going on when screen refused to connect at all - it was successfully at the arch login prompt, and for some reason screen just quits and dies at this point. I can't attach via usb once arch has started loading, but I CAN now ssh into it, so, progress.

Anyway, I've attached a file containing a printenv and a reset at the marvell prompt. You'll see there's a lot of errors come up on the reset. I don't know if these are normal or not. I'll keep experimenting and see where I can get to, but it's all looking good so far