Hello all,

Eric B?nard wrote:

Valentin Longchamp a ?crit :

quoted

Alan Carvalho de Assis wrote:

quoted

On 12/1/09, Daniel Mack [off-list ref] wrote:

quoted

Hmm, I don't follow. You added some debug output and it started 
working,
is that what you say? Do we possibly need memory barriers somewhere?

Basically it, but I will do more tests to isolate the issue.

In fact there is a board where the issue remain, but using old kernel
(2.6.22) both work.

I am also investigating this issue. The same happens sometimes with my 
USB host with ULPI-ISP1504, but it never happens with the USB slave 
with the exact same hardware ULPI-ISP1504, that's why I have the 
feeling that it's a purely software issue.

we once met a similar problem on i.MX27 : in fact, during IOMUX 
configuration, some pins were togled on the ULPI and the ISP1504 was 
taking this toggles as a command and was taking control of the ULPI bus 
which means the IMX wasn't anymore able to talk to it through ULPI. Just 
reseting the ISP1504 allawed the boot to continue fine.
Maybe your issue is similar to this.

I have found a fix to this issue, and it looks a bit like what Eric just 
described. I will post a patch in the next minutes and it would be nice 
if (at least) Alan and Daniel could test it to see if it fixes the 
problem for them too.

With the patch, I have never noticed a problem during the whole 
afternoon (around 30-35 reboots with ULPI init).

Val