2.0.1 -> 2.1 fails on 4GB CF in Alix



  • First off: I'm not sure whether this is related to the 2.0.1 -> 2.1 beta update or if my Alix-board is to blame. But I'm not sure how to tell which is…

    So I have an Alix 2c2-board, updated the BIOS to 0.99h (was just 0.99 before), CHS selected in the BIOS (according to http://doc.pfsense.org/index.php/ALIX_BIOS_Update_Procedure). I also have a 4GB CF-card (well, slightly under 4GB as the 2.0.1 4G-nano-image didn't fit) with the 2GB-image flashed on it using physdiskwrite and a CF-adapter in my PC.
    Plugged the 4GB-card in the Alix, booted, did some initial configuration (selecting NIC, assigning IP-adresses, selecting the 2.1 development-branch to check for updates) all without problems. Then used the Auto Update, let it download the new image (about 80MB?), and then "the system will reboot once the upgrade is completed". But using a serial connection, I can see it has failed:

    Installing root/latest.tgz
    
    GEOM: ad0s2: media size does not match label.
    Unknown: TIMEOUT - WRITE retrying (1 retry left) LBA=338519
    ata0: timeout waiting to issue command
    ata0: error issuing WRITE command
    g_vfs_done(): ufs/pfsense0[WRITE(offset=173281280, length=4096)]error = 5
    Device cf went missing before all of the data could be written to it; expect data loss
    Device cf went missing before all of the data could be written to it; expect data loss
    

    And the first LED on the Alix-board blinks SOS in Morse code. Nice touch.

    After hard rebooting the device, it gives lots of errors, but in the end, it can't start a working pfSense. So I plugged the 4GB-card back in my windows machine, did a chkdisk which didn't find any problems, flashed the 2.0.1 2GB-image again, back in the Alix, initial setup, download the update… and it failed with the same message.

    It's the first time I use a 2.1-snapshot, but I guess it's supposed to work? Can anyone derive the source of this issue from the information above? Is it some update-procedure or the build of 2.1? Or is it the Alix-board? Or the CF-card? I think I have another 4GB CF-card here, so I'll try it again with the card (probably tomorrow) to potentially rule that last one out.

    Moderators: If it's Alix-related, please feel free to move this thread to the appropriate sub-forum.

    Thanks in advance!



  • That's your CF failing to be written at first, then completely disappearing. Almost certainly a hardware issue, most likely with the CF itself.



  • I upgraded a 2GB Alix nanobsd system from 1.2.3 to 2.0.1, made sure it was working, then upgraded from 2.0.1 to 2.1-BETA0 a few weeks ago. Everything went smoothly. Your issues do look like a hardware problem!



  • @phil.davis:

    I upgraded a 2GB Alix nanobsd system from 1.2.3 to 2.0.1, made sure it was working, then upgraded from 2.0.1 to 2.1-BETA0 a few weeks ago. Everything went smoothly. Your issues do look like a hardware problem!

    Same HW, properly updated from 2.0.1 to 2.1-BETA0.
    Already seen that kind of errors with a fault (i.e. not original) SanDisk CF.



  • Indeed, it seems to be the card: another card works just fine.
    Too bad the card didn't fail while using physdiskwrite or doing a chkdisk… But luckily I noticed before using the Alix in "production".



  • I have had CF cards that spat out IO errors on the console when used for real (and caused system hangs, crashes, whatever). I took them out, put them in my CF card reader, and used phydiskwrite to write a complete new image with no errors reported. Put the card back in the system, booted OK, reloaded the config from backup. Not long after, IO errors are happening again.
    So physdiskwrite must have done a lot of good stuff (the card got a bootable FreeBSD/pfSense slice, the files that pfSense actually uses must have been written OK, the config restored OK…), but I have trouble believing that it really wrote to the whole CF card successfully. Maybe physdiskwrite doesn't check the IO return status of its writes?
    Obviously, I replaced that card with a new one!


Log in to reply