Re: SlugOS 4.0 HEAD status update

Great work!
I'm all for faster bootups.
I've been using apex for a while now to speed slugos-3.10 on my robot
and it could use some more speedups. 

Take care,

--- In nslu2-developers-hHKSG33TihhbjbujkaE4pw@xxxxxxxxxxxxxxxx, Michael-Luke 
Jones
<mlj28@...> wrote:
>
> Hi there,
> 
> This afternoon I have pushed some large changes to org.openembedded.dev.
> I have checked out a clean version of the org.openembedded.dev tree  
> and am testing a full build right now.
> 
> SUMMARY OF CHANGES
> The kernel has been upgraded to 2.6.18. I've put in place a new build  
> system that grabs patches straight from the kernel SVN repository,  
> allowing NSLU2-Linux to have a single site for kernel development.
> 
> USB support, and some other drivers are now compiled as modules. This  
> is to reduce the kernel image size to allow a second-stage bootloader  
> to be prepended.
> 
> The open source ixp_npe driver is now enabled by default, and thanks  
> to this the Intel Access Library has been jettisoned. It autoloads  
> correctly on boot, allowing SSH access to the NSLU2. It has not been  
> heavily tested.
> 
> For now, it is required to install the Intel NPE Microcode directly  
> into the root filesystem. The ixp4xx-npe package handles this within  
> the OE build environment. (Eventually, the microcode will be  
> installed onto a specific location on the NSLU2 flash, so that it is  
> not overwritten by kernel/rootfs changes.)
> 
> udev has been replaced by mdev, provided by busybox 1.2.1. hotplug-ng  
> has been removed. This is the most controversial change made,  
> partially because it will probably break turnup (not tested).
> 
> The reasons to make this change are as follows:
> 1) udev seems to be permanently undergoing rapid development and its  
> behaviour is not consistent between versions, increasing time  
> required to maintain SlugOS
> 2) A functioning firmware upload implementation is absolutely  
> required for ethernet access
> 3) No distribution, however large and well resourced, seems to have a  
> functioning and up-to-date udev (>94) shipping (with firmware upload  
> functioning) without the use of crufty outdated hotplug/coldplug  
> scripts (thinking of gentoo here). This does not bode well for our  
> small and resource-limited distribution.
> 4) For the life of me, I can't get udev to upload microcode to the  
> NPE. With mdev, a one-line config change was required, which took  
> less than a minute (see the ucode_dl line in slugos/mdev.conf in the  
> busybox commit).
> 5) mdev is fast, so booting is speeded up.
> 6) While mdev does not load modules automatically, this does mean  
> that its behaviour is predictable, and modules can easily be loaded  
> in boot scripts at appropriate times.
> 7) This is a matter of opinion, but I think that less hacky behaviour  
> is required in my implementation of mdev - compare the mdev init  
> script to the udev init script to see
> 
> The downsides of the change are:
> 1) People with custom udev rules will not be able to use them (Debian/ 
> NSLU2 should include a full and functioning udev implementation with  
> the final etch release)
> 2) Some script 'logic' will be required to load the right modules for  
> different embedded IXP boards with different peripherals
> 3) USB hard disks are not currently automounted when plugged in
> 4) Turnup and associated boot scripts will need possibly more  
> rewriting than was necessary to deal with modularisation alone
> 
> mdev documentation can be found here: http://busybox.net/downloads/ 
> BusyBox.html#item_mdev
> Scripts can be run automatically after a device node has been added  
> or removed in order to provide functionality similar to udev
> 
> Summaries of the changes can be be seen at the links below:
> 
> busybox
> http://tinyurl.com/emg2y
> 
> slugos-init
> http://tinyurl.com/zt2oe
> 
> ixp4xx-npe
> http://tinyurl.com/jstwb
> 
> slugos image & conf files
> http://tinyurl.com/e5kd3
> 
> ixp4xx-kernel
> http://tinyurl.com/pnpxe
> http://tinyurl.com/zc72g
> 
> WHAT'S BROKEN
> HEAD only boots correctly on the NSLU2 currently (and theoretically  
> on an IXDP425).
> turnup (see above and below)
> Kernel does not take into account 66MHz crystal, so timer is inaccurate
> 
> TODO
> Test and fix turnup (disk first, then nfs)
> Add logic in slugos-init to allow different module loading for  
> specific ixp4xx boards
> Kernel updates (DSM-G600, loft support, new clocksource timer patches)
> Specifically: Add mac definition for NAS-100D as in http:// 
> tinyurl.com/ro6r3 & http://www.hohnstaedt.de/ixp_npe/readme
> Copy ixp_npe branch in kernel SVN repository to trunk, and archive  
> current trunk
> 
> Again, my thanks go to Christian Hohnstaedt who wrote this driver and  
> has provided assistance throughout the process of integration.
> http://www.hohnstaedt.de/ixp_npe/
> 
> Please reply with comments and complaints,
> 
> Michael-Luke (blaster8)
> (Apparently) NSLU2-Linux.org Core Team
>






 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/nslu2-developers/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/nslu2-developers/join
    (Yahoo! ID required)

<*> To change settings via email:
    mailto:nslu2-developers-digest-hHKSG33TihhbjbujkaE4pw@xxxxxxxxxxxxxxxx 
    mailto:nslu2-developers-fullfeatured-hHKSG33TihhbjbujkaE4pw@xxxxxxxxxxxxxxxx

<*> To unsubscribe from this group, send an email to:
    nslu2-developers-unsubscribe-hHKSG33TihhbjbujkaE4pw@xxxxxxxxxxxxxxxx

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/