Issues with Headless Server - Never boots.

[baordog]

Hello! I am new to the parallella community. I recently bought one of the "server" models off Amazon. I installed the "headless" build included in the quick start notes on the SD card. I am familiar with this process from working with the Rasberry Pi. I do not intend to ever use a HDMI monitor. Ever. I'm currently pretty unhappy as the thing has never booted to the best of my reckoning.

Upon attempting booting up the machine, I get blinking lights, seeming to indicate the machine has booted. Viewing my network directly from my gateway, I see that the parallella never appears as a connected device. I do see blinking lights on the parallella's ethernet jack. However, I did a test and removed the SD card and noticed that the same blinking lights appear even without an attached SD card. That's very irritating, as there are absolutely zero external signs whether the parallella has booted successfully.

I want to emphasize that I have not been able to successfully boot this machine at all, ever. I'm beginning to think I bought a brick here.

Also, I have looked, but I have found no checksums to determine the image you provided me for the SD card is valid. This makes me very sad.

Does SSH listen on a nonstandard port?
How can I trouble shoot this on a "headless" build? I do not want to use HDMI! Is there a serial interface I can somehow use to see where in the boot process it is dying? How is it connected? What cable is used? baud rate? How do I set this up?
How can I ensure that that the image written to the SD card is valid? Do you have checksums? How do I know dd has written correctly to the SD card even if the original image is valid?

I have also tried:
-rewriting the SD card image several times.
-connecting directly via my laptop, with the suggested network settings.

~/parabuntu $ ssh 10.11.12.13
ssh: connect to host 10.11.12.13 port 22: No route to host
$ ping 10.11.12.13
PING 10.11.12.13 (10.11.12.13) 56(84) bytes of data.
From 10.11.12.1 icmp_seq=1 Destination Host Unreachable
From 10.11.12.1 icmp_seq=2 Destination Host Unreachable
From 10.11.12.1 icmp_seq=3 Destination Host Unreachable
cFrom 10.11.12.1 icmp_seq=4 Destination Host Unreachable
From 10.11.12.1 icmp_seq=5 Destination Host Unreachable
From 10.11.12.1 icmp_seq=6 Destination Host Unreachable
From 10.11.12.1 icmp_seq=7 Destination Host Unreachable
From 10.11.12.1 icmp_seq=8 Destination Host Unreachable
From 10.11.12.1 icmp_seq=9 Destination Host Unreachable
From 10.11.12.1 icmp_seq=10 Destination Host Unreachable
From 10.11.12.1 icmp_seq=11 Destination Host Unreachable
From 10.11.12.1 icmp_seq=12 Destination Host Unreachable
From 10.11.12.1 icmp_seq=13 Destination Host Unreachable
From 10.11.12.1 icmp_seq=14 Destination Host Unreachable
From 10.11.12.1 icmp_seq=15 Destination Host Unreachable
From 10.11.12.1 icmp_seq=16 Destination Host Unreachable
From 10.11.12.1 icmp_seq=17 Destination Host Unreachable
From 10.11.12.1 icmp_seq=18 Destination Host Unreachable
^C
--- 10.11.12.13 ping statistics ---
21 packets transmitted, 0 received, +18 errors, 100% packet loss, time 19999ms
pipe 4


Image used:
parabuntu-2016.11.1-headless-z7010.img

I believe the model number here is the P1600-DK03, at least that's what Amazon says. The box you sent me has no outward signs of a model number. Box says "Parallella 16 Micro Server, New" - I can provide more detailed information on the individual board over PM.

Lastly so we can save some time here are my reactions to the troubleshooting information on github:

Your board won't come up on the network?

The most common error is that the wrong image is used (see install instructions above to determine which version is the correct one)

According to your published information this is the correct image.

Log in to your router and see if you can find the board in the DHCP client list.

I can't.

If you connect your Parallella directly to your computer's ethernet port, enable internet sharing, or make sure that your computer and Parallella is on the same network by setting your computer's IP to 10.11.12.x (e.g. IP 10.11.12.1 netmask 255.255.255.0).

I did this. Same result.

Connect a serial cable, anything weird in the boot log?

What is a "serial cable" in this case? - do you have any detailed instructions for how to speak to this over serial? Must a solder a connection? Baud rates?

Please give me some indication of how I can debug all this. I'm beginning to be very frustrated.

[mchampigny]

I had a similar problem with my board and it did turn out to be a brick. However, I would not have been able to conclude that without the serial cable.

Go to Amazon and order:

https://www.amazon.com/gp/product/B008A ... UTF8&psc=1

There is a serial header on the Parallella near the ethernet jack (3 pins). I used gtkterm to connect on Linux, but any debug terminal will work.
The box includes a sticker showing which wires connect where. You'll need to match those up with the pins on the Parallella. There are several
posts on this forum describing the mapping of pins on the board. Just search for "serial cable" on this forum and something will come up.

Don't connect the red wire! That much I remember.

You can then paste the output from the terminal to this forum and someone will be able to help.

[baordog]

Hooked up the UART cable and lo and behold:



U-Boot 2012.10-00003-g792c31c (Jan 03 2014 - 12:24:08)

I2C: ready
DRAM: 992 MiB
WARNING: Caches not enabled
MMC: SDHCI: 0
SF: Detected N25Q128 with page size 64 KiB, total 16 MiB
In: serial
Out: serial
Err: serial
Net: zynq_gem
Hit any key to stop autoboot: 0
Configuring PL and Booting Linux...
Device: SDHCI
Manufacturer ID: 3
OEM: 5344
Name: SL32G
Tran Speed: 50000000
Rd Block Len: 512
SD version 2.0
High Capacity: Yes
Capacity: 29.7 GiB
Bus Width: 4-bit

** Unable to use mmc 0:1 for fatload **
Error: Timeout waiting for FPGA to config.
fpga - loadable FPGA image support

Usage:
fpga [operation type] [device number] [image address] [image size]
fpga operations:
dump [dev] Load device to memory buffer
info [dev] list known device information
load [dev] [address] [size] Load device from memory buffer
loadb [dev] [address] [size] Load device from bitstream buffer (Xilinx only)
loadmk [dev] [address] Load device generated with mkimage
For loadmk operating on FIT format uImage address must include
subimage unit name in the form of addr:<subimg_uname>

** Unable to use mmc 0:1 for fatload **

** Unable to use mmc 0:1 for fatload **
Wrong Image Format for bootm command
ERROR: can't get kernel image!
zynq-uboot> <INTERRUPT>
zynq-uboot>


Now in the past I tried formatting the damn SD card with no formatting at all (just fully 0'ed and dd'ed with the image) and with ext4. It shouldn't matter with DD, but I'm thinking maybe I need this to be fat32 or something?

Any ideas on how I screwed this up?

[mchampigny]

That doesn't look like the failure pattern I saw unfortunately. There are some similarities (FPGA timeout) but the logs and error messages are not the same.

Check this board (same topic) for "Bad MMC device" thread and you'll see the log messages I was getting to compare. It's just a few posts down from yours.
I concluded my board was dead. I sent for another one and the next one did work using the same SD image that was failing.

If you are burning on Linux, make sure you sync after using dd to make sure disk buffers are flushed. It should take a long time to burn. If it doesn't then you're doing something wrong in that step.

[sebraa]

Specifically, how did you write the image? There were some reports that using Windows 10 to write the image produces non-working SD cards. Try to use an older Windows XP or 7 machine instead, or use Linux to write the image.

There should be two partitions, one with the kernel, devicetree and bitstream, and one with the Linux file system. The latter one won't be visible on Windows.

Also, try a different SD cards, they vary a lot in quality.

[baordog]

[sebraa]

[ioiomi]

I met the same problem with my two cards and two machines, and I would continue to pay attention to the problem.