On Tue, 10 Jul 2007, Dmitry Monakhov wrote: > On 18:27 Втр 10 Июл , Hugh Dickins wrote: > > Or would this simpler patch be the right one? I suspect the > > mspec driver only says VM_LOCKED because of a deep-seated but > > irrational fear that its pages might fall into reclaim. > No mspec is not the only one :( , in my case it was fglrx (ati driver).
Ah, then fglrx is wrong too. We don't want to check locked_vm, but we don't much want it to wrap negative later. Perhaps we should mask VM_LOCKED off, perhaps we should BUG, perhaps we should ask ATI/AMD to change it (a BUG would be persuasive ;) > BTW: where is comment about it in do_mmap_pgoff: Sorry, I don't understand. You'd like a comment saying which flags the driver may change and which it may not? Hmmm... not today. > /* Can addr have changed?? > * > * Answer: Yes, several device drivers can do > * it in their > * f_op->mmap method. -DaveM > */ > addr = vma->vm_start; > pgoff = vma->vm_pgoff; > vm_flags = vma->vm_flags; Hugh