On Sat, Dec 1, 2012 at 12:25 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote: > On Sat, 1 Dec 2012 00:06:43 +0100, Daniel Vetter <daniel at ffwll.ch> wrote: >> Quick review: >> - I don't like the internal byte offset we add/substract for relocations. >> mmap works on page boundaries, imo gpu mmap should do the same. > > However, the necessity is that I work on unaligned boundaries since I > have no control over how the clients allocate the memory to be mapped, > or which bytes within that memory are associated with a particular > pixmap. The XShm API is such that not even the ShmSegment is passed down > to the driver. That vies with the lack of explicit synchronisation > points as being the worst elements of its design. Well, I just want to move the page_offset computation from the kernel to userspace. I know, this requires that you store that offset in userspace somewhere, which requires a new field right next to the existing u32 handle. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
