Zedboard booting up issue

Hi,

Are you using a Sandisk SD card? A variety of problems have been seen when working with other brands.

Regards,
   Eli

I am using a Sandisk 8GB HDSC card. I used quite a lot of other cards before and they were really slow, Scandisk is relatively fast but the problem described is prevailing.  

Hi,

After some deliberation I can say that the Xillinux boots up without any problem and shuts down without errors.

If I start the x-windows using startx the board hangs up randomly and will not shut down properly and displays a message which states "Failure to communicate with kernel-device-mapper" and the board will not reboot and displays the decompression error as stated originally on doing a manual re-boot. 

Does this mean that there is a problem due to bandwidth mismatch and SDHC card is unable to feed data to VGA display or it could be a problem with the xserver installed.   

Hello,

Problems with booting and/or startx have been reported repeatedly, and have so far turned out to be a data integrity issue: SD cards other that Sandisk corrupted the data, which lead to random failures. The only special thing with the X server is that it consumes a lot of data from the disk as it goes up, which is why it's likely to be the first opportunity to notice that the SD card is corrupted.

So I insist on Sandisk not for the speed, but for the reliability. The X server is installed fine.

Now, since you're saying that you are already using Sandisk, the next step is to suspect data corruption of the DDR memories or an issue with the MMC interface's driver in Linux.

So I suggest a shot in the dark -- there are two versions of Xillinux out there, 1.1 and 1.3. Whatever version you tried now, please switch to the other, and see if that matters. It shouldn't, but if it does -- it would be interesting to know about it.

See http://xillybus.com/doc/old-xillinux-zedboard for more information about Xillinux' versions.

Regards,
   Eli

Hi Eli,
As you suggested I was using 1.3 Version of xillinux previously and replaced it with 1.1 Version. Surprisingly the zedboard is really stable and X server is working perfectly fine. I kept the display up running for more than 5 hrs and the performance did not really degrade. Also rebooted the board a few times and it is booting up perfectly.

So changing from V 1.3 to V 1.1 solved the stability problem for me.  

Hi,

Even though it's good to know that Xillinux-1.1 solves your issue, it's not such good news for Xillinux-1.3, which was released about a month ago.

The thing is that the root filesystems are almost the same. The differences are mainly in the kernel version and Uboot, which have both been upgraded (the latter based upon another initialization routine, which is supplied by a newer version of XPS).

So question #1 is: Which revision of Zedboard do you have? Maybe it's an issue with some specific revision.

The v1.3 bundle for Zedboard has been downloaded >200 times by now, so it's safe to assume that the vast majority of users haven't seen any problem (or there would have been more complaints).

As I mentioned earlier, the immediate suspects are poor signal integrity on the wires going to the DDR memories and SD card, which can be a result of inadequate initialization of their registers on the processor.

If you'd like to help me investigate this, could you please try running the C program that is listed on this post on Xillinux-1.3?

http://billauer.co.il/blog/?p=3932

Just copy it somehow into the SD card's file system (say, as memtest.c) and then go (on the board's shell propmt):

gcc -O3 memtest.c -o memtest

and then run it as shown on the post. At least for a few hours.

Don't bother starting an X session.

Thanks in advance,
   Eli

As you suggested I did the test for over 6 hours and it did not return any errors. I am e-mailing you the snap-shot of the result.

Please suggest me if I have to re-run the test again. It ran for over 6 hours but did not return the time of run. I paced the program in the background and brought it back again.

Harsha

Hi,

Thanks for your checkup. It would be nice trying out another thing. On the Zedboard, running Xillinux-1.3, please type at shell prompt:

dd if=/dev/urandom of=rndfile bs=1k count=512k

This will take 7 minutes or so, and create a 512 MB file. It has to be that big to make sure that the disk cache is flushed.

Then type:
while [ 1 ] ; do cat rndfile > rndfile-copy; cmp rndfile rndfile-copy || break ; echo OK $((i++)) $(date); done

This bash oneliner runs indefinitely or until there's a difference between a file and its copy. It should, of course run forever. Or until you stop it, preferably no sooner than after a few hours.

Each round takes about two and a half minutes. I believe that if there's a problem with the MMC interface, something should happen on this test.

I should mention that this causes a certain wear on the SD card, but I suppose a few hours shouldn't make any difference.

And the SD card must have at least 1 GB of free space. The Getting Started guide explains how to resize the partition, if you haven't done that yet.

Thanks again,
   Eli