Light off on awakening PW3

[knc1]

Originally Posted by knc1

So reboot it.

Yes, it is an indication of a hardware problem that can grow worse.
But the start-up boot code includes tests and corrections for the problem.

Be sure you have a full charge on the battery and then, as mentioned above, reboot it.

Originally Posted by AlanHK

I know you were trying to be helpful.,
But I did not ask how to fix it (which I have done anyway, by open/close; changing brightness, or indeed rebooting) I asked what it meant. If it was a symptom of a serious problem.

Naively I thought Kindle OS was more stable than PCs and glitches indicated a hardware issue rather than in the Windows world where the OS is so flaky that it just gets unstable for no external reason and "reboot" is the answer to 90% of problems.

Better that you re-read my post that included the answer to "what it meant" then.
I have quoted it above.

Do you want the technical details, complete with terms of art?
I thought you just wanted a yes/no type of answer to "does it indicate a hardware issue" (ANS: YES).

= = = =

Windows is a poor choice for a baseline standard.
Google for "uptime records".

[salamanderjuice]

I've had that happen before several times. Never particularly bothered me. I think it has something to do with the way the Kindle turns on from having the case opened. I never saw the same thing happen when using power button outside the case.

[AlanHK]

Originally Posted by knc1

Better that you re-read my post that included the answer to "what it meant" then.
I have quoted it above.

I was not asking the question again, I restated it only because there was a question about what question I had asked.

Thus my comment on on recursion.

I hope we could let the question of the question go now.

Do you want the technical details, complete with terms of art?
I thought you just wanted a yes/no type of answer to "does it indicate a hardware issue" (ANS: YES).

Initially, I wanted to know if it is hardware related (unfixable) or software (potentially fixable).
If I could actually prove it was hardware I might be able to get it replaced though.

Windows is a poor choice for a baseline standard.
Google for "uptime records".

It's a good choice for illustrating random inexplicable errors "cured" by a reboot though. Since Kindle is Linux based, I did expect it to be more stable.

[knc1]

Originally Posted by AlanHK

- - -
Initially, I wanted to know if it is hardware related (unfixable) or software (potentially fixable).
If I could actually prove it was hardware I might be able to get it replaced though.
- - - -

Is it hardware related: YES
Is it unfixable: NO
Does the hardware need to be replaced: NO
Is it fixable: YES
If it is fixable, can I fix it: YES
How to fix it then: Reboot it, the system's startup code will detect and fix the hardware problem.

How can it do that?
1) You have a eMMC (the flash storage device) with a erase block with an uncorrectable error.
2) The eMMC device has spare erase blocks
3) The eMMC device has a record (list) of the spare erase blocks
4) The eMMC has hardware error reporting
5) The Linux kernel's startup code includes checking for (1) above.
6) The end result will be to execute commands that cause the eMMC to move the in-use erase block with errors to a list of bad blocks, that content copied (using ECC to fix the error) to one of the spare erase blocks removed from (3).
7) The file system will be checked, and any errors found corrected.
8) Normal startup will continue.

How is the above sequence triggered? Reboot the damn thing!

It isn't magic, it is written into the start up code.
If you where interested, you could get a copy from the dev forum posts, disassemble it, and read it for yourself.

Sure, it is the same directions that are often given to dumbshit customers by dumbershit "service representatives" - but in this case, it is the way engineered into the system to fix what is most likely wrong.

You can either take my word for it, that I have read it, and that is really the way the startup code is written, or you can read it for yourself.

[AlanHK]

Originally Posted by knc1

Is it hardware related: YES
Is it unfixable: NO.

Well, if you look at it like that, everything is hardware.
If you can fix a hardware issue by software commands (i.e., rebooting) it's a philosophical question which domain the problem is in.

Anyway, as I said, I have rebooted a few times. I said that early on. It's just a button press so why not?

You seem to assume that the error is in the flash memory. How do you know that?

[j.p.s]

Originally Posted by knc1

Windows is a poor choice for a baseline standard.
Google for "uptime records".

I remember when microsoft windows was guaranteed not to stay up for more than 49.7 days because of a counter rollover bug. A similar 497 day bug in linux was discovered and fixed before the windows bug was even discovered. Most windows computers would crash well before 49 days and those that experienced the freeze assumed it was just another random windows lockup.

[j.p.s]

Originally Posted by AlanHK

It's a good choice for illustrating random inexplicable errors "cured" by a reboot though. Since Kindle is Linux based, I did expect it to be more stable.

Like just about any OS, linux is no match for a poorly written GUI application.

My phone runs linux, and when it feezes after a few months, weeks, or days, I often can ssh in and reboot it instead of forcing a power off, so linux itself was still stably running and mostly in control.

[NullNix]

Originally Posted by AlanHK

You seem to assume that the error is in the flash memory. How do you know that?

Quite. I had a problem with near-identical symptoms on my PW1 for many years: sometimes the light would not go off when you turned it off, and would then never turn off until you rebooted: sometimes the inverse problem occurred, with the light not turning on.

This obviously (to my mind) had nothing whatsoever to do with bad eMMC -- among other things, it happened for the first time the week I bought it, and flash doesn't wear out that fast. It was fairly clearly a counting-semaphore problem: perhaps "screen on" is denoted by a semaphore >=1 in a signed integer range, and there was, somewhere in the code, a screen off/screen on imbalance so that it ended up oscillating around 1/2 or something. Rebooting obviously fixes dynamic state problems like that.

To me this looks like exactly the same class of bug: purely software. I see no reason to assume anything relating to hardware here.