2008-10-31 05:59:20 BF527 / SPI / DMA
Eric BENARD (FRANCE)
I'm facing big problems to activate DMA in order to interface a 8686 Wifi Chipset to a BF527, it works fine without DMA but fails to load firmware once DMA is activated.
I've seen 2 threads (first, second) discussing of this kind of problem, and in both threads, Robine Getz & Sonic Zhang are saying that DMA & SPI are unusable and send extra unwanted bits.
May you have a link to a document describing the problem as our AD's Application Enginer says " there is no problem with SPI & DMA especially in the latests silicon release".
Could this be a problem in the silicon we have and in this case : are you aware of a fix concerning SPI & DMA in the latest silicon ?
Should we switch to the latest silicon release (we actually have a 0.1 on the evaluation board) or won't that change anything for SPI & DMA ?
Many thanks and best regards
2008-10-31 06:02:47 Re: BF527 / SPI / DMA
Mike Frysinger (UNITED STATES)
there is no documentation nor is there any fixed silicon
SPI DMA in the driver may be broken in the release regardless of those issues depending on the SPI device and how it responds to the CS toggling in the middle of a transfer. you can try the latest svn trunk.
2008-10-31 14:53:44 Re: BF527 / SPI / DMA
Robin Getz (UNITED STATES)
When we refer to things as "broken" it means undesirable behaviour from a standard software implmentation. It is unusuable in many (but not all) circumstances. Your FAE means "it works like described in the documentation". This is correct - the documentation describes the existing (what we call "undesirable", and when we are upset - "buggy") behaviour.
ADI has documented the issue. In the latest 548 HRM it states:
"The SPI continues receiving words until SPI DMA mode is disabled."
What is missing - is it is up to software to do so. By the time that happens - it is normally too late - and extra clocks have gone out, confusing/corrupting the SPI device.