SanDisk Sansa c200v2, m200v4, clipv1, clipv2, clip+, and fuzev2

[FlynDice]

Well I have my daughter's rockboxed fuzev2 today and I get the same uSD problems as with the clip+ so maybe it's the card.  We had problems with these same Transcend class 6 SDHC cards with the as3525v1 development but that involved the slow times while in the PRG state and not GPIOA.

[DonDan]

Hi to everyone!
I use a fuzev2 8GB with a 8GB Transcend cl6 uSDHC card.
On all recent rockbox builds (r25733-r25763) i was able to boot with an 8GB uSDHC inserted, but I can only browse the files, if I want to play a file from it, it goes to the WPS but it doesn't load the tag and it gets stuck on 0:00 seconds (not reading the file?). I am still able to change the volume and to exit the WPS but if I try to go again to the "Files"-Menu or the "Plugin > Games"-Menu it hangs (hold still works but nothing else, need to reboot).

From the internal memory I can browse and play music. Sometimes the music stops after 10-15 songs and after that I cannot load any files anymore (Music or Plugin).

With the r25753-r25763 builds all plugins work great. No black screen after quitting anymore. Volume control works. Doom works. Image viewer works!

Looking forward to read/write support for the internal/external memory. Everything else is working pretty well in the latest build!

Great job folks! Thumbs up!

[FlynDice]

Seems a good old dose of fsck has cleared up my booting problems.  Good thing I spent all that time tracking it down to INT_GPIOA.....

I'm still not able to play off of the uSD though.

Now, back to pio writes with a poor attitude!

[funman]

On my Fuze v2, I can startup with a usDHC inserted, but the 8gb card will not be recognized (the 2gb one will). Reverted back to 25733 for the meantime.

Can you replace "udelay(1000)" by "sleep(1)" in firmware/target/arm/as3525/sd-as3525v2.c and see if the card works?

[bilditup1]

I'll try to do this tomorrow night when I'm home; thanks for the suggestion.

-e

[kugel.]

It's possible that r25770 fixed the microsd problems for some of you.

[bilditup1]

Funman - the change you suggested appears to have worked perfectly, both if the card was already in at startup and on insertion. You can go ahead and patch it in. (This was tried with the latest build, 25791).

-e

[kugel.]

What about the latest build without funman's change? I committed a fix which potentially solves the problem. If it doesn't it would be helpful if you change the "udelay(1000)" so that it fits your cards (e.g. maybe higher it until your cards work).

[bilditup1]

If you're referring to changes you made in 25770, I tried it without success. Did you change anything since then in 25791?
What is the technical issue here? Why is it preferable to supply udelay with a higher or lower number than just sleep(1)?

[kugel.]

If you're referring to changes you made in 25770, I tried it without success.  Did you change anything since then in 25791?

Yes, I meant that. And no I haven't made further changes.

What is the technical issue here? Why is it preferable to supply udelay with a higher or lower number than just sleep(1)?

Sleep(1) is 10 times longer. Delays should be as low as possible since they block other tasks (although sleep() only blocks the current one), forcing the cpu to do nothing.
I'm also curious why udelay(1000) doesn't work, since I calculated the 1000 from the old delay which worked for everyone.

[bilditup1]

Interesting. So do you have any other suggestions then? Should I proceed by using random inputs for udelay less than 10000? I'd like to help but really don't quite know what's flying here

[kugel.]

Interesting. So do you have any other suggestions then? Should I proceed by using random inputs for udelay less than 10000? I'd like to help but really don't quite know what's flying here

Yes, you could bisect the needed delay. E.g. start with 5000, then try 2500 or 7500 depending on the success, then 1250 or 3750 (assuming 2500 worked). Basically always half the interval, and add or subtract it from the current number, until the interval is like 50.

[funman]

Write support committed in r25799, should work on all models including µSD.

Unless of course the µSD isn't detected in the first place

EDIT: Please run a battery bench and post the results (with model, revision, volume, and DSP effects enabled), and pray for it doesn't crash while benching

[xiang]

holy shit.. i've waited so long for this!!! thanks!! i'll donate some money to u guys. when are the new builds up?

[funman]

The current builds are at the usual address:
Clipv2
Clip+
Fuzev2