Kindle Hung after downgrade

[knc1]

Originally Posted by geekmaster

Again, plug it into the charge until we can deal with this. I think perhaps we need the full device info (settings menu) recorded before downgrading, so we can find a pattern in this. Do not publish it -- just a PM to knc1 and NiLuJe (who needs to find a pattern in mysterious new serial numbers for kindletool) and me to look at too when I have time (very limited now). Need sleep... zzzzz.

P.S. We are considering taking the PW3 image offline even though it has about a 98% success rate, just to be safe, until we figure this out. Howeve, *I* think we just post a "2% chance of bricking your PW3" warning and let the user decide it he can afford that risk. What do you think?

@GM:

When I left, at 1:30am, the directions DID HAVE the "2% warning" in it -
pdurrant "improved" our directions at 2:09am - I haven't a clue what was changed -
You made the above post at 2:52am.

Now I am clueless also.

When I left, I was bringing up another distributed cloud distribution network (cdn77.com) - -
Now, I'll back away from that and put the directions into a public repository where interested people can post 'pull requests' rather than directly edit them.

Edit:
Downgrade thread updated for the activity during my 4 hours of down (sleep) time.
Please, everyone understand that I may not be writing from the standpoint of a tolerant mood this morning.

[geekmaster]

Originally Posted by knc1

... When I left, at 1:30am, the directions DID HAVE the "2% warning" in it -
pdurrant "improved" our directions at 2:09am - I haven't a clue what was changed -
... When I left, I was bringing up another distributed cloud distribution network (cdn77.com) - -
Now, I'll back away from that and put the directions into a public repository where interested people can post 'pull requests' rather than directly edit them.

You can see exactly what changed by clicking the "editted" text at the bottom of your posts, and compare all the edits.

An editor at one of my jobs got fired after she repeatedly editted one of my highly technical manuals to make it more "UX friendly" and less "technically accurate". I reverted all her edits every time she submitted it to me for my approval step. Then she made major edits without my approval and sent it to the print factory without my final approval. Millions of dollars of manuals had to be discarded after the paper manual was sent to me to double-check before sending to customer, and she was terminated on the spot. Gotta love it when non-techs edit critically accurate technical content, eh? Sad that it resulted in you terminating your work on this here and moving it where no "non-nerds" cannot damage it again. Very sorry to hear that...

I suggest reverting those edits. But beware, one of our active tech members almost got suspended for repairing damage to code formatting done in a post that had been editted by another member who has edit privileges, so beware of attracting such mistreatment in your direction.

Moving our content offsite (as suggested by other active members too) may be our only hope to protect our nerdly technically accurate content.

EDIT: Maybe he just fixed some spelling errors? I hope not more than that, for the sake of technical accuracy. Technical instructions are not at all like story books, and require a different set of editting skills, even for grammtical corrections,. That is a fact.

[knc1]

Not a problem,
pdurrant only fixed some of my write-up errors (for instance, BD -> BranchDelay -> Branch Delay).

Thanks for the tip on where to click - I knew the forum posts used the wiki engine, I just didn't know where to access the difference feature.

[pdurrant]

Originally Posted by knc1

pdurrant "improved" our directions at 2:09am - I haven't a clue what was changed .

The full extent of my changes was to add Branch Delay's full user name to the credits instead of BD.

Only the member, forum mods or MobileRead Editors can change posts.

[knc1]

Originally Posted by pdurrant

The full extent of my changes was to add Branch Delay's full user name to the credits instead of BD.

Thanks.

Since the original version of my post, I have learned how to find out what changed -
Then I went back and edited my own posts on this subject, also giving you credit for helpful changes.

Blame this all on lack of morning coffee please.

Edit:
PS: You missed one in step 5, but I just fixed it also, thanks for spotting the problem.

[geekmaster]

Originally Posted by pdurrant

The full extent of my changes was to add Branch Delay's full user name to the credits instead of BD.

Only the member, forum mods or MobileRead Editors can change posts.

Okay, that is useful. I would not mind if you changed a few hundred GM references to geekmaster too.

[knc1]

Originally Posted by jprince

- - - -
What was showing on the screen when it 'hung'? It cycled to boot, however, the "active" graphics don't fill the whole screen, rather the upper right. Perhaps as if it doesn't know its dimensions?

When the USB cable was connected, Kindle to PC, could you see the storage on the Kindle? yes, originally, however, after trying DO_FACTORY_RESTORE (and I mean, I did a touch -m /Volumes/Kindle/DO_FACTORY_RESTORE to create the file, it no longer seems to mount. However, the first time after that, it proceeded to ask what language I wanted to use. The graphics were/are still small, but the onscreen hit targets matched as if it were full size. Regardless, hitting Next lead back into boot cycling
- - - -

That is clear enough for me to make an educated guess as to what is happening (without a clue yet on how to fix it):

The "Recovery" style packages contain a kernel image with an embedded initramfs (which includes the 'at boot time' screen drivers).

Somehow, the process is running the wrong combination kernel/initramfs image file.

There probably needs to be a re-boot (maybe to the second kernel/initramfs image) in the process somewhere.

If the above is so - then the looping is probably caused by the update code trying to replace the kernel/initramfs while it is still running from eMMC storage (a programming Oops).

No clue at the moment if the above guess-work is true.

- - - -

And the initramfs code is in violation of the GPLv2 license of the Linux kernel -
It is embedded in the kernel image file (making it GPLv2 also) but Amazon has never released that source code as required by the license.

Note:
There is a way for u-boot to load a kernel and a non-embedded initramfs, added to the kernel code just for the purpose of keeping initramfs code proprietary, only Amazon has never use it.
I even know the person who did that, and why.

[geekmaster]

Originally Posted by jprince

The graphics I get pretty much match theirs—filling the upper right of the screen, as if it were using graphics intended for the Kindle PW version with less DPI.

The dimensions are hardcoded as determined by the device serial number and/or PCB ID. These device identifiers are "idme vars" stored in a hidden area you may access in diags or with the "idme" command. It is possible that they got damaged, but more likely that the firmware is looking for them in the wrong location. Those screen artifacts in the photos look scarily like a broken screen, but no need to panic in this case. We can fix this (with serial port, or even simpler).

[Cinisajoy]

I don't know if it means anything but I noticed all three reports state that the PW3 have
special offers.

[geekmaster]

Originally Posted by Cinisajoy

I don't know if it means anything but I noticed all three reports state that the PW3 have
special offers.

Good catch. It may be important to know to help prevent this problem in the future.

[knc1]

Originally Posted by Cinisajoy

I don't know if it means anything but I noticed all three reports state that the PW3 have
special offers.

We have at least one report where the owner used the Kindle's account to remove special offers before starting the downgrade.

But that report was not clear if the Kindle was given enough on-line time afterwards to sync its local files with the status change.
Which is the part of special offers removal that counts in this situation.

[Cinisajoy]

Originally Posted by knc1

We have at least one report where the owner used the Kindle's account to remove special offers before starting the downgrade.

But that report was not clear if the Kindle was given enough on-line time afterwards to sync its local files with the status change.
Which is the part of special offers removal that counts in this situation.

Ok, so the 4th one originally had special offers. That leads me to ask: does Amazon delete the special offers file or just add an override file.

[geekmaster]

Originally Posted by Cinisajoy

Ok, so the 4th one originally had special offers. That leads me to ask: does Amazon delete the special offers file or just add an override file.

Pretty sure they just delete special offers. Unless recent firmwares do things differently than in the past. However, great observation and recommendation.

[knc1]

Originally Posted by geekmaster

Pretty sure they just delete special offers. Unless recent firmwares do things differently than in the past. However, great observation and recommendation.

And would probably explain why the problem is hard to find.
Developers rarely use Special Offers enabled devices for development and testing.

Also, I suspect geekmaster's speculation that the internal control variables moved or changed format between PW-3 firmware versions (maybe in preparation for the White PW-3 release).

Hmm...
We probably need a "White PW-3" factory initial update image (I already have a place in the "credits" for that person).

[geekmaster]

Moving the idme vars might involve a change in PCB ID or other hardware-related identifying information stored there too. I know u-boot looks at multiple idme vars to figure out how to init SD-RAM (or it did when I made the custom u-boot images for MfgTool -- I wish I still had that source code).