iriver H120 after CF mod: *PANIC* error dc_writeback_callback ()

[entne]

Hi again,

I just completed a CF mod on my H120 and all seemed fine (see this thread of mine: http://forums.rockbox.org/index.php/topic,52424.0.html).

But after beginning to test the upgraded device, it quickly turned out that there is something wrong... it is the following error:

*PANIC*
dc_writeback_callback ()
- Could not write sector
  15747 (error -5)

(The 5-digit number varies a little, sometimes it's  15746)

It occurs sometimes when I just click "artist" in database. The device gets stuck for a while with "Searching... 0 found" and then the error appears. I have to press REC + Play then to reboot. Sometimes I can get through to an artist and a song and the same appears when I start the track... "Searching..." and then the error after a while. When I browse via Files I can play a track, but after the track has finished the device gets stuck and then after a while comes the error again.

I also witnessed that USB connection takes a while when I plug the cable in.

Any help is highly appreciated!

[Bilgus]

I added retry code to the dc_writeback function try this to see if it works if this fixes the issue we can go from there to fix it right (write?)

http://www.mediafire.com/file/nbu3kmq59gszqt6/IriverH120140-rockbox-dcwritecb._10-18-18.zip/file

[entne]

Thank you very much for your helpfulness and effort!

I just replaced the whole .rockbox folder with your modified version, is that correct? After that, it said that booting had changed and asked for rebooting which I confirmed. I think I have to test it now for a little while if it is stable now. I'll give you a feedback the next day. Thanks again! :-)

[Bilgus]

No problem, yes just copy whole folder, generally you should rename or delete the old one first but it shouldn't cause any issues not to unless its a particular setting causing the issue or an older version

[entne]

Ok, so I don't have to do any further actions with the new folder/files like flashing or something?

[entne]

After a quick try I discovered the following (I deleted the .rockbox directory and replaced it with your modified version):

I can't play any tracks anymore. Database loaded sufficiently and I could walk through menu and database quite well so far in the first minutes, but when I played a track, f. e. the first of an album, the tracks just rushed through without playing like "1 of 9... 2 of 9... 3 of 9" and so on as if the files were corrupt. I experienced no error or hanging but I didn't try very long because the described phenomenon appeared for all tracks I tried.

[Bilgus]

well that was based on the current dev version so now I'm unsure if that was an issue or if it was the retry causing the issue
that is not the behavior I'd expect with a retry unless its now succeeding and there is junk data which is possible

I assume you were using 3.14 prior to this could you try the latest dev version?
http://build.rockbox.org/

[entne]

I have to correct myself, I think the device reacts pretty unforeseeable... as I turned it on today it got stuck when I wanted to access an album via database and came out with a similar PANIC error like before after a while:

*PANIC*
dc_writeback_callback ()
- Could not write sector
  17541698 (error -5)

Then I deleted the .rockbox folder again and copied the dev build over. I could access a song which was played. I could skip to the next which played too. But when I wanted to go back to albums again it got stuck with the above error... 

I would be very grateful about further ideas. 

And to be sure: do I have to do anything else to update to dev build after I copied the folder? Flash to RAM/ROM again or something?

[Bilgus]

No updating needed of the bootloader in flash, after booting it is no longer used.

It still fails after a while because it is a genuine failure to write so that means the issue is either in the card or the adapter

What you need to do from here is take the card out and do a thorough test of the card

on linux see here:
https://ccollins.wordpress.com/2016/01/18/testing-sd-cards-with-linux/

on windows maybe something from here (no clue about which might be best):
https://www.geckoandfly.com/22803/detect-fake-usb-flash-drives-sd-cards-ssd-disk/

Be aware this will destroy the data on the sdcard so back up important files first (if any)

Next if you have a CF reader do the same with the adapter and the sdcard together

Finally if test(s) pass reformat the card to fat32 and try again

You should probably completely wipe sdcard and partitions..

I can verify HDD LLF Low Level Format Tool works on windows:
http://forums.rockbox.org/index.php/topic,51941.msg240386.html#msg240386

[entne]

Thanks for your quick answers.

Currently I am in huge trouble even connecting the CF card (via card reader) with the PC for testing. It was buggy from the beginning... I have to figure out if it's the fault of the reader or the card (which would make sense concerning the H120 behavior and your hypothesis, but odd because it's freshly bought).

When I have checked that out, I'll test with Check Flash 1.17.0. I can only test the CF card via card reader and have no option to test the adapter. But if the card turns out to be ok, maybe it's a good idea to go for a new adapter...

Besides that I was asking myself if cutting down the jumper of the adapter for reasons of space could cause any problems? Or the rubber/gum part I layed between the adapter and the board?

[entne]

Now I took the CF card out of the player. I borrowed another card reader and it could be connected to the pc without any trouble. Then I did a Full Pattern Write And Read Test with CheckFlash (first app from your link). I used the set value on access type "use temporary file" because I wasn't sure about the other options for NT-based systems only ("as logical drive" or "as physical device"). It passed without errors. Afterwards, I built the CF card into the player again, connected it to the pc (which took some time just as before) and repeated the test in CheckFlash (for testing via built-in CF adapter). It passed without errors.

I took the CF card out again and connected it via card reader to pc. Then I used HDD LLF Low Level Format Tool to fully format it. After that the app noted "remember to create new partitions". I just accessed the drive in windows and windows asked to format it - which I did with "cluster size" (I am using german Win 10 so I am not sure about the translation) set to standard size (initially at first use of the card I chose 16 which was the set value). Then the drive was accessible again and I copied over the stable rockbox build and some music files. I turned on the player and went to the database. Player asked to initialize because db was not ready. I confirmed and unfortunately it got stuck again with "searching... 0 found" and presented the error after a while, though it was slightly different than before:

*PANIC*
dc_writeback_callback ()
- Could not write sector
  32898 (error -2)

Next step trying new CF card and/or adapter (latter is much cheaper so that would be my idea at the moment)? Or any other idea I can try?

[Bilgus]

Sorry to hear it, I find it odd that the sector failing has moved but maybe it is genuinely bad or maybe it is Rockbox,
maybe you could try older firmware this user says 3.9.1 works for his cf modded device

http://forums.rockbox.org/index.php/topic,52528.msg243012.html#msg243012

[entne]

How could it be that after replacing the current build with 3.9.1 the booting screen still says it's running 3.14? I deleted the current build folder before I copied  the older release...

[Bilgus]

probably just the bootloader version

what does System>Rockbox Info>Version show?

[entne]

3.14!

I thought bootloader has a different version naming ("7pre4 or something").