Wolfgang Denk a écrit :
Dear Stefan Roese,
I have to admit, that I'm not totally sure about this problem. Perhaps it would be best, if you would try to "fix" to the CFI code. To verify, if your theory is correct. And then post the resulting fix to the list, so that we can discuss it here.
Agreed. Eventually the READ ID could / should be combined with the 8/16 bit test, at least for this specific device ID.
That is 'should', really, because the only way to tell apart a real pure word mode QRY from this faulty miwed-mode QRY is to try the READ ID, because i) it *is* correclty mixed-mode and ii) the Mfg and device ID are needed for verifying that we indeed had a faulty part.
Um... what does the Linux kernel code do in this case?
Not entierly sure yet. I'll look.
Amicalement,