Hi, On Thu, Oct 18, 2012 at 08:06:40PM +0300, Felipe Balbi wrote: > Hi, > > On Thu, Oct 18, 2012 at 10:07:31AM -0700, Kevin Hilman wrote: > > Felipe Balbi <balbi@xxxxxx> writes: > > > > > current omap_device PM implementation defines > > > omap-specific *_noirq methods but uses the > > > generic versions for all other PM methods. > > > > > > As it turns out, if a device decides to implement > > > non-runtime PM callbacks, we might fall into a > > > situation where the hwmod is still idled which > > > will generate an abort exception when we try > > > to access device's address space while clocks > > > are still gated. > > > > Please explain in more detail how this could happen. > > just follow the patchset. If I implement system suspend and I'm not > focefully calling runtime_* methods, who will be there to call > omap_device_enable() ? let me rephrase this. If we exit _noirq right in the beginning with the extra check I've added, who will call omap_device_enable() ? -- balbi
Attachment:
signature.asc
Description: Digital signature
