J7F4K - Hard Lockups

[srfeo]

I have recently bought a J7F4K Jetway fanless VIA Eden 1.2GHz board and have been experiencing constant hardware lockups, especially when using the USB port to connect to external hard drive enclosures...

My server is running Linux (Debian Etch - testing) and I'm using it right now as a file server.

Since the case that I'm using (Mini-box M300) is too small for a large drive, I am booting off of CF flash and then using an eSATA enclosure connected to the motherboard or a USB enclosure to store the data.

I have had the computer lock up (no I/O to monitor, no keyboard access) under the following circumstances:

1) Using rsync to backup between the eSATA to external USB
2) SCP copying from the eSATA/USB drives from another computer over the network
3) Doing general intensive disk activity

When I am doing an rsync backup over the Internet, at about 384kbps, the system seems quite stable. It seems to only lock with higher transfer rates.

I am still trying to investigate further by setting up a console over serial so I can try to find out what could be causing the crashes...

Has anyone seen such problems?

[benou]

I had the same issue with mine.

The easiest way to freeze the system is to run 'dd' or 'badblocks' on the disks. I always won when doing that.

I tried many things trying to correct this, here's a very very short summary :
- the RAM has been tested with memtest, no defects found
- Bios upgraded to the latest version available from Jetway
- disable every on-board devices, except IDE disk (I also tried with SATA only)
- test with several disks (which work fine on another computer)
- test with another IDE cable
- test with and without DMA, at various DMA speed
- test underclocking the proc to 400MHz
- test with various OS : windows 2000, windows xp (SP1 and SP2), Linux Debian testing, Linux Gentoo, each with various kernel versions, including latest vanilla
I've read a lot of mailing lists, and also viaarena.com, but I didn't found what could be the problem.
Under Windows, it just doesn't work : it was very hard to install (many freeze during installs) and completely unsusable then (It was impossible to watch a DivX or wathever).
Under Linux, even with the more conservative tunings (kernel compiled with full debugging support - including kexec/kdump -, IRQ debugging, etc.), it keeps freezing, and sometimes panic inside interrupts handlers (most of the time it just freeze).

It's a hardware bug. I sent it back to the vendor and just received a new one (same model, though).
I will test it and if it keeps doing those nasty things, I'll just go for AMD Geode or similar.

[srfeo]

Benou,

Thanks for the feedback. I thought that I was going crazy since I couldn't find any debugging information that would point me to what was causing the problems. Many times I would get a lockup without any output to the console. I am working on connecting a console via serial, but I doubt that will work either. I can reproduce the problem with heavy I/O activity, especially through USB. As I said, if I sync large files (3GB+) from Canada to the UK over rsync/ssh it seems to be quite stable. I'll do a little more investigation, then I'll send it back to the vendor too...

I did update the BIOS to v4, but it was buggy and caused the bootup to be impossible slow. I reported this to Jetway and they released a new BIOS which seems to work better. I don't know if this has fixed the stability issue as I've had the server off for a couple of days now.

[TuxCrafter]

Got this problem to on my en12000E i have been pulling my hair out to debug the problem, so far no success. using xubuntu feisty latest testing

[TuxCrafter]

last week i ordered a new J7F4K because the en12000E was driving me crazy i really hope i don't get the same problems!

[TuxCrafter]

I also tested a AMD Geode but the power consumption is much higher, also the fan was to high preventing my case form closing. Also the SIS741CX doesn't have 3D drivers under linux. I made a test report of the test.

http://www.ubuntuforums.org/showthread.php?t=394589

[benou]

srfeo: I will test the new J7F4K on Wednesday, I'll post a report here, stay tuned.

TuxCrafter: a working AMD Geode is better than a crappy VIA, despite the higher consumption [img]i/expressions/face-icon-small-smile.gif[/img] - but I haven't abandoned yet.

[srfeo]

I ran some more long rsync tests (from one external eSata to an external USB) last night starting at 11pm. It looked as though everything was fine, but sometime past midnight (I can tell from the --mark-- in the syslog file) it locked up. I don't find any information in the serial console I have hooked up or in the log files. The system becomes unresponsive to anything - the ps/2 keyboard (even the kernel ALT+PRINTSCREEN commands), pings, etc. It is as if the whole I/O system is down. The funny thing is that I can see the Ethernet port lights flashing sometime, so I know at some level the system is still alive.

I'm going to try again to see if I can capture any more information. My vendor said something about a version 2 vs. version 3 of the motherboard. If I can't sort out something by Friday, this board goes back... I'll try to get an Intel Pentium M or AMD and enable the CPU frequency stepping to save Watts.

[benou]

Ok, I ran some test last night.

1/ My replacement board has some differences with the previous one:
- the layout is a little bit different on the board
- the "factory" bios is from November 2006 instead of September 2006
- it has now a buzzer.
So definitely, it's another version of the board (v2 ? v3 ? I don't know).

2/ I restart dd'ing with the default settings and kernel, and it still locks up.
There's hope though: when I test without ACPI, IOAPIC and local APIC, it doesn't lock up anymore. Tests are quite long, so I have to conduct more tests to be sure, and to try to isolate the problem.
With the previous board, It always locks up.

[srfeo]

Thanks for the feedback. If you every find the problematic settings, kernel or otherwise, please let me know.

I'm going to compile a new kernel now and strip out everything that I don't need and see what happens...