Re: [PATCH v2 3/4] mm: Introduce page_needs_cow_for_dma() for deciding whether cow

On Thu, Feb 04, 2021 at 09:54:37AM -0800, Linus Torvalds wrote:

On Thu, Feb 4, 2021 at 6:50 AM Peter Xu [off-list ref] wrote:

quoted

--- a/include/linux/mm.h
+++ b/include/linux/mm.h

@@ -1291,6 +1291,27 @@ static inline bool page_maybe_dma_pinned(struct page *page)
                GUP_PIN_COUNTING_BIAS;
 }

+static inline bool is_cow_mapping(vm_flags_t flags)
+{
+       return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
+}

Oh, and I just realized: moving this to <linux/mm.h> means that this
patch could/should also get rid of

 - manual copy of this in mm/hugetlb.c:

        cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;

 - private #define for the same thing in fs/proc/task_mmu.c

    #define is_cow_mapping(flags) (((flags) & (VM_SHARED |
VM_MAYWRITE)) == VM_MAYWRITE)

 - manual copy in drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c:

        bool is_cow_mapping =
                (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;

I guess it could be a later cleanup patch too, but maybe best done in
this series just to not forget about it.

Will do in v3 if there is.  Or I'll post a standalone patch to not disturb Gal
just in case he's kicked a test.  Thanks,

-- 
Peter Xu