More VIA Woes on 1.2.3



  • I picked up a used Wyse 941GXL thin client on ebay. It's a 1Ghz VIA C3 Nehemiah processor with padlock. It has 512MB RAM and a 512MB Flash DOM (Disk on module). I Added an Intel Dual 10/100 WAN PCI card to the expansion slot. I proceeded to boot it up in knoppix via USB CD Rom and dd'd the 1.2.3-RELEASE nanobsd image from a thumb drive onto the internal flash drive. Rebooted and it came up without issue.

    LAN: fxp0
    WAN: fxp1
    OPT1: vr0 (down)

    My IPSec tunnel came up fine and I didn't see the pseudo-random number generator message leading me to believe the hardware RNG is being utilized.

    After 12 hours of boot the system rebooted unexpectedly. This is the same exact behavior I was getting with a JetWay motherboard using a VIA C7 proc with padlock on 1.2.3-RELEASE.

    I'm currently running the 033109 build of 2.0 Beta to see how long it stays up. Something is definitely not right with the padlock/VIA code it seems and it's causing kernel panic. This time there are no Realtek interfaces to blame either.

    I guess if 2.0 is crashing too I will have to download one of the older RC candidates of 1.2.3 which didn't have padlock and test that.



  • It might be too late now, but I do not recommend VIA for…. well, anything.  When VIA bought Cyrix, they were third string to AMD and Intel not just because of speed, but they weren't fully x86 compatible.  I'm not sure they ever got their act together.



  • Yes, I agree, VIA was absolute crap in the mid-to-late 90's. However, these thin clients are used by hundreds of banks and merchants as point-of-sale terminals, and I doubt a company as big as WYSE would sell hardware that is not reliable. I think there are enough VIA CPU based systems out there in the field to at least raise a flag when there is a glaring bug causing kernel panics on this hardware.

    I'm not the only one bringing this issue up. There are at least 3 other threads about 1.2.3-RELEASE crashing on VIA hardware.

    Anyway, I do have some good news. 2.0-BETA1 (built march 31st 2010) is so far stable on this VIA box. it's ticking up 24 hours of uptime as I write this :) I stressed it out with 5000 states using bit torrent and so far it's running fine with low load.

    This setup has significantly more horsepower than my alix box (I was able to route 97 mbps thru it), and it only draws 25 watts on my kill-o-watt. It's also completely fanless so all you hear is the pfSense startup chime :)



  • This morning the router was hung. I guess I will have to test a pre-padlock build of 1.2.3. Any other suggestions?



  • I just got a Jetway J74K1G5S-LF with dual Realtek Lan ports running 1.2.3 and had trouble with the firewall becoming non responsive at random intervals, the symptoms being no internet access and could not log on to the firewall via the web GUI. I resolved the issue by disabling Hardware Checksum Unloading on the Advanced functions web GUI page near the bottom of the page. This has resolved the issue and now everything works fine



  • @bmwkeith:

    I just got a Jetway J74K1G5S-LF with dual Realtek Lan ports running 1.2.3 and had trouble with the firewall becoming non responsive at random intervals, the symptoms being no internet access and could not log on to the firewall via the web GUI. I resolved the issue by disabling Hardware Checksum Unloading on the Advanced functions web GUI page near the bottom of the page. This has resolved the issue and now everything works fine

    But I'm assuming your JetWay board has RealTek nics on it, this one has one VIA Rhyne and 2 Intel interfaces. I've had to disable TSO even on Intel CPU Realtek nic mobos so I think that is more related to the realtek interfaces. I suppose I should test that combination as well anyway. Thanks for the suggestion.



  • Very curious about how you make out on this – the VIA Nano chipset supposedly blows away the Atom in benchmarks, plus has Padlock/AES acceleration AND an "Intel compatible" Virtual Technology (VT) instruction set.  Sounds too good to be true...



  • My pfSense host runs on Via and it has only gone down when the power has gone out for extended periods (ie, longer than the UPS can handle). I have 2 other VIA boxes, one of which has an uptime in excess of 2 years.

    Your claims of VIA hardware being third rate isn't supported by any evidence I've ever seen.



  • @Cry:

    My pfSense host runs on Via and it has only gone down when the power has gone out for extended periods (ie, longer than the UPS can handle). I have 2 other VIA boxes, one of which has an uptime in excess of 2 years.

    Your claims of VIA hardware being third rate isn't supported by any evidence I've ever seen.

    Load a different x86 compatible OS.  Its quite possible pfSense isn't hitting any of the inherent bugs on that platform.  So in your case, that's good.



  • I've run FreeBSD from 4.4 onwards and Linux from kernel 2.2 onwards on these boxes (and XP).  Doing everything from file serving, firewalls, desktops, mail servers.  I've never had any problems that weren't caused by something outside of the processor and core chipset (a few cases of bad memory, a couple of failed CPU fans and one case of an idiot who couldn't tell the difference between Ethernet and ISDN).

    I think if there were "inherent bugs" I'd have found at least one of them by now ;)



  • ive got mine running on a via 533mhz epia, runs just fine



  • Yes, I also have a VIA C3 600 Mhz running for the last 3-4 years without a problem.



  • Count me (and many of our customers) among those who have VIA hardware in production with 0 issues. Not sure what it is that a small number of people appear to hit on that hardware, but I have some very busy deployments (including the firewall in front of the web servers that host this forum and our other sites) running on VIA hardware. The padlock is in heavy use as well, offsite backups go over IPsec.



  • @networkninja:

    However, these thin clients are used by hundreds of banks and merchants as point-of-sale terminals, and I doubt a company as big as WYSE would sell hardware that is not reliable. I think there are enough VIA CPU based systems out there in the field to at least raise a flag when there is a glaring bug causing kernel panics on this hardware.

    Are you handy with a screwdriver? Maybe worth a look inside for thermal clues, or at least running memtest86 on it for a few days continuously just in case.

    Reading this a month later just I'm wondering how many of those utilize the hardware crypto side of things to run WindowsCE or whatever they run in the MS-world
    I appreciate that a fair proportion of them run unix-y OSs as terminals in enterprise use but not all. I'm curious - does anyone know (especially from a MS perspective) if the hardware crypto elements are heavily used on these sort of systems in practice?.

    I've run early VIA C3 systems (at least 5 years ago, in light use desktop/terminal situations) with no CPU-related problems, though in ALL cases the motherboard hardware failed prematurely from badly-specified capacitors (mainly Jetway ISTR). Swapping the CPUs to identical new motherboards cured the varied symptoms immediately. None of these cores had the Padlock feature, I mention it to illustrate that not all the early blame these systems took might have been due to the CPU architecture.

    I'd hate to think you were chasing a hardware/design/fabrication feature masquerading as a coding bug :(

    That said, I know of a few relatively recent VIA pfsense installs in current service with no problems (at least mentioned to me), so I'm minded to agree some of the later above posts.
    G



  • I'm running a via VB8001 board in 3 different applications, Pfsense being one of them. My current uptime is about 400 days, or the last time the power was out for an extended duration. Is it a quad core processor running multiple apps or windows pulling 400 watts? No, but it draws 14 watts and handles our connectivity just fine. As someone else posted earlier, check your thermals or power supply.


Locked