Hi Laurent, On Sat, Feb 22, 2020 at 12:43:48PM +0200, Laurent Pinchart wrote: > Hi Sakari, > > On Fri, Feb 21, 2020 at 09:46:41PM +0200, Sakari Ailus wrote: > > Hi Hans, Laurent, > > > > On Fri, Feb 21, 2020 at 04:31:01PM +0200, Laurent Pinchart wrote: > > > Hi Hans, > > > > > > CC'ing Sakari for the omap3isp part. > > > > Thanks. > > > > The omap3isp is behind an IOMMU, so the USERPTR memory does not need to be > > physically contiguous. I don't see a reason to drop userptr support from > > the driver. > > Apart from the fact that this API should be discouraged :-) I wonder if > it's used, that's my real question. As we can't rule it out, I'd be > cautious about dropping it. Indeed. If we'd drop USERPTR, it should apply to the uAPI, the framework and all the drivers, not an individual driver. But again that would be certain to break a lot of applications... I guess we could document USERPTR as deprecated, and not recommended for new applications? -- Regards, Sakari Ailus
