Dave wrote:
> On 03/14/2013 10:50 AM, Kirill A. Shutemov wrote:
> > --- a/include/linux/huge_mm.h
> > +++ b/include/linux/huge_mm.h
> > @@ -113,11 +113,20 @@ extern void __split_huge_page_pmd(struct 
> > vm_area_struct *vma,
> >                     __split_huge_page_pmd(__vma, __address,         \
> >                                     ____pmd);                       \
> >     }  while (0)
> > -#define wait_split_huge_page(__anon_vma, __pmd)                            
> > \
> > +#define wait_split_huge_page(__vma, __pmd)                         \
> >     do {                                                            \
> >             pmd_t *____pmd = (__pmd);                               \
> > -           anon_vma_lock_write(__anon_vma);                        \
> > -           anon_vma_unlock_write(__anon_vma);                      \
> > +           struct address_space *__mapping =                       \
> > +                                   vma->vm_file->f_mapping;        \
> > +           struct anon_vma *__anon_vma = (__vma)->anon_vma;        \
> > +           if (__mapping)                                          \
> > +                   mutex_lock(&__mapping->i_mmap_mutex);           \
> > +           if (__anon_vma) {                                       \
> > +                   anon_vma_lock_write(__anon_vma);                \
> > +                   anon_vma_unlock_write(__anon_vma);              \
> > +           }                                                       \
> > +           if (__mapping)                                          \
> > +                   mutex_unlock(&__mapping->i_mmap_mutex);         \
> >             BUG_ON(pmd_trans_splitting(*____pmd) ||                 \
> >                    pmd_trans_huge(*____pmd));                       \
> >     } while (0)
> 
> That thing was pretty ugly _before_. :)  Any chance this can get turned
> in to a function?

Cyclic headers dependencies... :(

> What's the deal with the i_mmap_mutex operation getting split up?  I'm
> blanking on what kind of pages would have both anon_vmas and a valid
> mapping.

anon_vma_lock protects all anon pages on vma against splitting.
i_mmap_mutex protects shared pages. None pages can be on both sides, but
MAP_PRIVATE file vma can have both anon and shared pages.

As an option we can lookup for struct page with pmd_page(), check
PageAnon() and serialize only relevant lock, but...

Original macro, in fact, guarantees that *all* pages on the vma is not
splitting, not only the pages pmd is poinging on. PageAnon() check will
change semantics a bit. It shouldn't be a problem, but who knows.

Do you want me to add the check?

-- 
 Kirill A. Shutemov
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Reply via email to