Best Kernel for Atom D510 with no video output? (and optimise for CF/no video)



  • Hi, I have a Lanner 7535 which is basically a small (cheap) Atom D510 board (64bit + dual core/quad thread).  For various reasons I will be installing onto a 4GB SLC compact flash card

    Initially I have installed pfSense 2.0 from the liveusb install and chosen the "embedded" kernel, however, this would appear to support only a single CPU?  I think I need to switch to a multi-core kernel - but a) how to do that switch (without re-installing)? and b) I dont have video output (other than during install), so how to switch to only console output on that kernel?

    Also, although my calcs suggest that an SLC CF card should last long enough for my requirements, it would also be prudent to minimise writes.  Any tips? It would appear I should set syslog to async writes? How can I see file io by process, eg is there an iotop binary (or equiv)?

    Finally, I formatted my CF card as approximately 2x 2GB partitions with the idea that I can use the "slices" concept to have a backup installation in case of a failed upgrade.  However, seems that the web interface to copy across to multiple slices only exists in the "embedded" install?  How to backup partition 1 to partition 2 (such that it's subsequently bootable if need be?)

    Thanks for any pointers?



  • Is there any reason not to use the NanoBSD builds?
    They will support packages just like the full installs will and would support the dual-slice out of the box.

    If you wish to follow through with a full install for some reason or another but need Console access via serial, then you should not select 'Embedded Kernel'.
    What you need to do is to select 'SMP Kernel' and after you get into the WebGUI, go to:
    System -> Advanced -> Admin Access -> Serial Communications -> Serial Terminal

    Check the box that says "This will enable the first serial port with 9600/8/N/1".



  • If you can find how to extract and install the SMP kernel from the upgrade tgz files, doing that on the NanoBSD images might be the best option.



  • @dreamslacker:

    Is there any reason not to use the NanoBSD builds?
    They will support packages just like the full installs will and would support the dual-slice out of the box.

    I don't see an AMD64 NanoBSD build?  OK, lacking 64bit builds is a minor inconvenience

    However, what else is missing on a nanoBSD build?  Why isn't everyone using such a build if it's not more limited?  My understanding is that the nanobsd won't save graphs between reboots and packages which need to save state won't work? Actually, assuming an SLC CF card is available, just wondering why not to use a full install?

    However, back to the original question please - how do I enable dual slices on a normal build please? How to copy images from one device to the other? (I'm a linux user, not really familiar with BSD..?)

    Thanks


  • Rebel Alliance Developer Netgate

    There are no amd64 NanoBSD builds (yet, probably not for 2.0). The NanoBSD builds have their own set of features and limitations. They aren't right for everyone.

    The major differences are:

    • NanoBSD only outputs to serial console (though there may be a VGA variant before 2.0 release)
    • NanoBSD keeps the media read only except for discrete write operations. (This makes it unusable for many packages)
    • NanoBSD has two main OS slices, and one separate configuration slice. Upgrade operations overwrite the currently unused "alternate" slice, and then boot from that.
    • NanoBSD has no installer, it can only be imaged to a disk and run directly.
    • NanoBSD only comes in 512MB, 1GB, 2GB, and 4GB sizes, so using it on HDDs would waste quite a lot of space. Using variable sizes is a non-trivial task.

    The OS itself is basically the same between the two variants, though there are some kernel differences (a few drivers are left out, like certain SCSI drivers and RAID controllers).


Locked