View Single Post
Old 09-28-2012, 12:16 AM   #70
knc1
Embedded Cheerleader
knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.knc1 ought to be getting tired of karma fortunes by now.
 
knc1's Avatar
 
Posts: 5,774
Karma: 5722276
Join Date: Feb 2012
Device: Too many.
Quote:
Originally Posted by NiLuJe View Post
Take my double kernel comment with a grain of salt, it's been a good while since I wrote that, and it may not be entirely accurate ^^.

The only thing I'm sure of is that there are two kernels .
The "flash update" utilities are implemented as shell scripts.
I have not studied them enough to know in what order they are called.
But they do share a common organization structure:
They update U-Boot (if available)
They update the uImage (if available)
They update the system image (if available)

And there is scripting for each "set" of U-Boot / uImage
(They could have done with fewer scripts if they had duplicated the thing I labeled: "machine specific data" - but they didn't.

Since all of this is in the mtdblock device, none of which has a file system that could be mounted during normal operation - it should be "safe" to re-write the contents at any time.

In the case of this thread - all that is lacking is the storage addresses as seen from U-Boot and this problem is solved.

There are two DX(G) machines involved - one with a working serial port, reported as booting to u-boot (although I suspect it is actually booting the kernel and the kernel is doing a panic halt) ;

The other machine - without a serial port, that works normally.
Separated by at least 6 time zones.

It isn't exactly a "real-time interactive" troubleshooting session.

It may take a bit of "peeking" into /dev/mem on the working machine for "known values" of the various images.
And then some confirmation by "peeking" from U-Boot (which should be the same - /dev/mem translates from virtual to physical addresses).

At this point -
getting closer to that noob's DX cure.
knc1 is offline   Reply With Quote