On Thu, Apr 19, 2012 at 1:38 AM, Albert ARIBAUD albert.u.boot@aribaud.net wrote:
Hi Dave,
Not my main area of expertise here, but I am not sure how this plays on Marvell non-kirkwood platforms (e.g., orion5x).
ISTR it is not the first time we deal with the whole IDE number of bus / number of ports [ / useable ports ] issue, and we may be running in circles here, fixing one platform and breaking another.
I, too, remember remember reading some discussions about changes to that very line.
I'll try this on EDMiniV2 in the coming days, and let people know the results in this thread.
Amicalement,
Albert.
After further testing, I've found a possible (minor?) anomaly, but not on a supported (official) Kirkwood device. The device I tested it on is a Seagate GoFlex Net (dual SATA).
The behavior I see is this:
With Toshiba 160GB and Seagate 320GB hdd's in ports 0 and 1 (right & left) respectively, I get perfect results. I can access both drives from U-Boot, and boot from both.
If I switch the drives around (ie. swap them) so the Toshiba is is port 1 (left), I can no longer probe it. It could very well be a timing problem, or it could be simply a quirk that is simply attributable to this hdd (which seems to check out completely healthy).
If I place _just_ the Seagate drive in port 1 (left), it is still correctly recognized. This is true whether it is the lone hdd, or if another hdd is in port 0.
Just thought I'd let you know. Maybe this is a _specific_ behavior subset that you all can test on your respective boxes.
My feeling is that the patch is nevertheless still valid - for what I've seen on Kirkwood, it is an improvement over the previous situation. Just my 2 cents.
Dave