On Thu, 14 Jul 2011, Guennadi Liakhovetski wrote: > Hi all > > I ran across an issue, which I cannot understand. In short: kernel threads > continue running also after freezing of the tasks has completed, late into > the device suspend. My scenario was suspending the system, while running > dmatest. At some point also the DMA driver is suspended, but the dmatest > threads continue running, so they issue further requests, which then > confuse the driver. In case this is a known design "decision," you do not > have to read on. Below I'll just provide some excerpts from kernel logs, > supporting my statement. Any explanation and fix suggestions would be > appreciated! I don't know how dmatest was designed, but it seems clear that the kernel threads it uses ought to be marked freezable. Probably nobody imagined running dmatest while suspending the system. Or maybe for some reason it's not possible to make those threads freezable. > Well, sure, I could hack the driver to reject any submits after suspend or > queue them without executing, but... Is this behaviour correct? The driver shouldn't have to worry about rejecting submissions after it is suspended. Rather, no thread should be running to make those submissions in the first place. Alan Stern _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
