DSM-G600, DNS-3xx and NSA-220 Hack Forum

Yes, I'm looking at pivot_root, unfortunately it is not part of the supplied busybox and there is no room in the ramdisk image, so I'll need to do something non-standard. It is not possible to add any functionailty to u-boot as it already fills the flash partition allocated to it.

I think I know what I need to do to use pivot_root to use a usbkey as a root device.  I still need to come up with a scheme that will reliably identify such a key, probably a file name that would be hard to create by chance.  Something like ".BootThisFlashDrive" and only set the rootdrive if that file exists and is excutable.  I'm not sure if the file should be hidden (start with a ".") or not.

I rebult busybox and if it was bigger than the one currently there is was not by much, I think it was smaller, but for some reason, when gzipped the ramdisk was bigger, and too big to fit in flash.  I'll try it again and verify that -Os is on, but the real issue is not the size of busybox, but the size of the compressed ramdisk image. Hmmmmm, I wonder what would happen if I gziped the running ramdisk...

BTW if anybody wants to play with ramdisk images, the gziped image needs to be <= (0x2e0000 - 64)

Well I tried -Os and yes the busybox was smaller, but the gziped image was still too large, To check for a problem in gzip I gziped the ungziped original and it gives the the same size within a few bytes, but when I gzip the image with the new buzybox it is too large (3133606 bytes).

Last edited by beattie (2006-12-13 05:43:29)

My goal is to leave the existing functionality in place.

Though I'm starting to think about completely redoing the flash layout, which means a new u-boot.

probably cut the ramdisk to 2MB an up the kernel flash to 1 full MB.  It will be a while before I get to that.

Last edited by beattie (2006-12-13 08:52:45)

beattie wrote:

I think I know why the modified ramdisk is bigger, dirty blocks,

The blocks freed by deleting the old busybox are put back on the free list but no contain "random" data and thus are now much less compressible.  I tried to make a new filesystem but that was even larger so I need to figure out what parameters were used to build the filesystem, I suspect it has a smaler than standard inode list.  possibly other things as well.

After you have done modifying your ramdisk, run

cat /dev/zero > data

This should write one big file with zeros and it should fill all free space that you have left on ramdisk. And if it's done then delete data file and unmount ramdisk
Alternatively you may also use dd to fill all free space with dummy file.
I haven't tested neither one but I guess they should work.

Some progress to report, by zeroing out the ramdisk image before doing mkfs I was able to buld a ramdisk image that fits with pivot_root and vi plus a few other functions in busybox.  I put together a fwbuild that builds a firmware file give a ubootable kernel and ramdisk.

Thus I built and uploaded a firmware image, the ramdisk has some problems (telnetd does not come up), but it a lot faster to upload fw that way.

Source for the utilities (still prety rough) http://www.beattie-home.net/beattie/DSM … re-0.2.tgz

Looking at the firmware a bit, it is appears to be failing to mount the harddrive or usb drives for some reason.

Last edited by beattie (2006-12-14 17:42:25)

qn1234 wrote:

That's good news.  Are you planning on creating a firmware image for those who want to flash using the existing web interface?

--
Quang

Long term I have some plans, but testing is going to be an issue, since I don't want to be blamed for bricking other peoples hardware.  There is not way to unbrick a dsm without a serial port if the web interface is hosed.

On a different question, does anybody here know how to get busybox to give a shell prompt on the console?

BertrandB wrote:

What's exacly the problem you don't have shell on console ? or the shel don't print prompt ?
For the fist case i look in the /etc/inittab it seems me ok
on second case did you never get prompt ?

I have never gotten a prompt, inittab does start /bin/ash on the console and busybox prints out it's banner but no prompt.  Characters are echoed and <ctl>c cause the shell to restart, but nothing else seems to have an effect.  This is what I see:

BusyBox v1.00-pre1 (2005.10.27-08:04+0000) Built-in shell (ash)
Enter 'help' for a list of built-in commands.

beattie wrote:

I have never gotten a prompt, inittab does start /bin/ash on the console and busybox prints out it's banner but no prompt.  Characters are echoed and <ctl>c cause the shell to restart, but nothing else seems to have an effect.  This is what I see:

Code:

BusyBox v1.00-pre1 (2005.10.27-08:04+0000) Built-in shell (ash)
Enter 'help' for a list of built-in commands.

Here's a tested patch to fix this.

sala wrote:

Here's a tested patch to fix this.

Great, thanks!  I now have built a firmware image that, brings up a shell and now mounts the hard-drive, which was the problem I wanted the shell to debug.

Of course I still have the problem that my unit seems to lock up after sitting idle  for a day or so.  I haven't been able to categorize it better yet, but if this sounds familar...

beattie wrote:

I hope none of the developers from D-Link read this,

I hope some d-link developper read this forum

May be then they would make better firmware ...  ;-)

beattie wrote:

In any case once I get pivot root working maybe somebody will tell us how to install debian on a DSM-G600

Sort of webinstaller would be nice just kidding
Meanwhile here is basic topic about debian.
Also here is good guide to bootstrap, just replace arm with powerpc.

qn1234 wrote:

Better yet, leverage some of the features and design from DD-WRT (www.dd-wrt.com)'s web user interface.

Web design is not my forte,  I'm sort of thinking about an Access Point, with local contant and Mesh networking.  The Mesh networking will probably require a different radio, maybe even a different (2.6) kernel.  If anybody wants to do some web design I'd be happy to work with them.

sala wrote:

[
Web user interface is not the priority, but eventually it is good idea to throw webs away and replace it with cgi compatible web server (eg. busybox httpd, fnord) and haserl scripts

Yeah, I sort of agree about getting rid of goahead (webs) I'd probably used the httpd in busybox, but my first priority after getting pivot_root working, is figureing out how to read the switches and blink the lights.  I've tried strings on the executables that seem to do that, and looked at the kernel diffs, with no clues, I suppose the next step is to compile strace and see what system calls it's making.