Wolfgang Spraul wrote: > So normally, association is remembered through a suspend/resume cycle?
If we don't take down the WLAN module, then it should keep its state, yes. > And right now we would loose it? Is there a chance you can fix this or > is it an inherent limitation of the firmware we have or the hardware > module/how it's wired up? It's basically that s3cmci asks the stack to remove the driver on suspend, and to re-insert it on resume. This causes the stack to try to identify the "new" card. Unfortunately, without a prior reset, this will fail, because the commands used for this are only allowed after reset. So the solution will be to avoid removing the driver on suspend. This probably has to be done in s3cmci, e.g., indicated by a flag in the platform data. s3cmci would then just flush the request queue and save the register state across suspend/resume. (That's basically what the Atheros SDIO stack does as well.) So that should be quite fixable. I just haven't done it yet :-) That change will partially obsolete the patch set I just sent in the sense that the HIF will never be reached on resume, but all the rest of the power handling is still needed for modules. - Werner
