Erase blocks

[DaMacFunkin]

How significant is 11 bad erase blocks being reported during flash restore? I've also got 9 on my solo2.

[marty]

I wouldn't worry, I got 8 on my DUO²

[Rob van der Does]

12 here on Solo2.

[Rexstar]

5 here on duo2

[Milamber44]

7 on my solo 2, as to the significance, no idea. Regards.

[tope]

8 here no Solo2


Enviado desde mi iPad Air con Tapatalk

[betacentauri]

Every erase block is 131072 bytes big. So if you have 10 bad blocks it's around 1.3 MB. With todays flash memory sizes this shouldn't be a problem.
Only if the number of bad blocks increases, you might have a problem. But I haven't seen that yet. I flashed my et9200 around 300 times and it still has 4 bad blocks like on the first day.

[Rob van der Does]

Thanks for your explanation.
Reading

http://openpli.org/forums/topic/21460-backupsuite/?view=findpost&p=419717
it seems there might be a way to (try to) recover bad blocks. Is that a viable option? Or do you say, given your experience with the number of bad blocks remaining the same, that it isn't worth the trouble?

[betacentauri]

You'll have to distinguish between the rootfs and the kernel partition. Rootfs uses ubifs and ubifs was build to be able to handle bad blocks. And rootfs is so big that some bad blocks are really no problem.
For the kernel partition it's different. It's a raw device without CRC, ECC or whatever to ensure consistency. A bad block in kernel partition is in 2 different ways bad. First you can get problem with the backup. Second (more worse) the oe alliance kernels are partly so big that they don't fit in kernel partition when there is only one bad block. So the user can't use alliance images when he has one bad block in kernel partition.
Gutemine talked about a tool to "unmark" bad blocks if they are not really bad. That would be nice but it only helps when the block is really not bad.

I don't know whether it's quite common that you have bad blocks in kernel partition or not. But I would propose to make the kernels smaller, so that they fit even with 2-3 bad blocks.

[betacentauri]

And another thing:
I don't know how bootloaders handle bad blocks in kernel partitions. From that openpli post I would guess that they ignore the status of the block. But only the bootloader dev or a bigger research could tell us.

[marty]

But only the bootloader dev.

We'll ask TM then :sofahide:

[Rob van der Does]

I don't know how bootloaders handle bad blocks in kernel partitions. From that openpli post I would guess that they ignore the status of the block. But only the bootloader dev or a bigger research could tell us.
Can a bootlog (I attached one as an example) tell us that?

[betacentauri]

Maybe. But at first you need a box with a bad block in kernel the partition. I don't have one.
Currently I don't know whether it's easily possible to mark a good block as bad, so that it is possible to test whether boot log shows something or not.

[DaMacFunkin]

So if a box was to develop a bad block in kernel partion that could mean game over for that box, right? Or would there be a way to resize the Kernel partion by stealing a bit from the rootfs?

[betacentauri]

Game over I wouldn't say, but it gets more difficult.
We're here in the duo2 section. The duo2 seems to have a 7340032 bytes big kernel partition. How big is the kernel? (My pad cannot unzip the image zip.)
Then you can quite easy calculate how many bad blocks the kernel partition might have so that the kernel fits into it. Assumed that the bootloader can properly handle bad blocks.

What can you do, if the kernel don't fit:
- Compile own kernel without unneeded drivers
- Try to tell devs that a smaller kernel would be nice
- Use other image (didn't check that, but OpenPli had in the past a quite small kernel)

I don't know how to change partition sizes. I guess it's somehow possible. I would really love to see bigger kernel partitions. I don't know why they are still so small. New boxes have plenty of flash memory...

[Rob van der Does]

At the moment the ViX kernel for the DUO2 is 3.647.206 bytes, so it uses less than half the kernel partition. The kernel of the actual PLi image is 3.681.118 bytes.

[DaMacFunkin]

Well lets hope none of us ever need more kernel space lol.

[DaMacFunkin]

Talk about foot in mouth disease...
betacentauri if you read this can you have a look here please.
http://www.world-of-satellite.com/showthread.php?35908-Vu-Duo2-stucked-and-i-see-quot-Vu-your-Smart-Linux-TV-player-quot&p=287969&viewfull=1#post287969