Numonyx NOR and chip->mutex bug?

Michael Cashwell mboards at
Thu Feb 24 09:28:15 EST 2011

Anders Grafström <grfstrm at> wrote on 2011/02/10 14:21:36:

> It produced this output at one occasion:
> 6000 loops
> status 00800080
> ffff0000
> ffff0000
> 4783 suspends
> ---------------------
> See the status 00800080 ?
> Looking at the code this should not be possible, right?

It's similar but not identical to my issue. I was seeing a buffered programming operation (done while an erase was suspended) appear to fail because the status read looking for program completion would return garbage. Your case is only erase, not programming.

But yours looks similar to mine in that your resumed erase completes (00800080) and then the next several status read returns garbage. Interestingly, the low chip returns correct status (WSM busy, nothing suspended, no errors, etc.) but the high chip returns FFFF. That was often what mine did during buffer program.

I also sometimes would see other values, not always FFFF. But with bits set in the upper byte whatever was being read was clearly NOT status. Since FFFF is correct for erased array data I assumed the part had dropped out of status-read mode but I've not been able to determine why.

Also, on my parts there is a loose restriction on the minimum time between starting or resuming an erase and the next suspend of it. It's 500µs for my part and they say that "repeated violations" of that time can cause erase failures. The hard part is that the 500µs is only a typical value and no hard number is given for how many violations are needed to cause a failure. I suspect it varies from part to part.

Your code suspends immediately after starting the erase, so that would be a violation on my part. And depending on the value of loops you might be violating it often.

On Feb 24, 2011, at 6:36 AM, Joakim Tjernlund wrote:

> Michael, have you made any progress?

Sadly, management was breathing down my neck to get it working and I've been pulled completely off onto a different project. I do still want to find the real answer but currently the people who control the purse strings have other ideas. There is another board in the pipeline here that might have this part on it. When that arrives on my desk I'll see if I can pick this up again.

> Can you run Anders program too?

Not as written. My board only has one flash chip where Anders' clearly has two chips in parallel. It wouldn't be too hard to change it to 1-wide but I'm overloaded on my other project at present.


More information about the linux-mtd mailing list