Re: [PATCH 5/10] mm/fs: route MADV_REMOVE to FALLOC_FL_PUNCH_HOLE

[Hugh Dickins]

On Mon, 21 May 2012, john stultz wrote:
> On Sat, May 12, 2012 at 5:13 AM, Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> > Now tmpfs supports hole-punching via fallocate(), switch madvise_remove()
> > to use do_fallocate() instead of vmtruncate_range(): which extends
> > madvise(,,MADV_REMOVE) support from tmpfs to ext4, ocfs2 and xfs.
> >
> > There is one more user of vmtruncate_range() in our tree, staging/android's
> > ashmem_shrink(): convert it to use do_fallocate() too (but if its unpinned
> > areas are already unmapped - I don't know - then it would do better to use
> > shmem_truncate_range() directly).
> 
> I suspect shmem_truncate_range directly would be the right approach,
> but am not totally sure.
> Arve: Any thoughts?
> 
> Hugh: Do you have a git tree with this set available somewhere?  I was
> working on my own tmpfs support for FALLOC_FL_PUNCH_HOLE, along with
> my volatile range work, so I'd like to rebase on top of your work
> here.

I don't, no, just the patch series posted.

I had hoped by now to say that it's in linux-next (though it would be
at the daily rebased end, which probably doesn't help you), but not yet.

If shmem_truncate_range() is all you need, then that doesn't depend on
these patches at all - but I expect you are aiming to be more general.

Hugh

> 
> thanks
> -john
> 
> 
> >
> > Based-on-patch-by: Cong Wang <amwang@xxxxxxxxxx>
> > Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
> > ---
> >  drivers/staging/android/ashmem.c |    8 +++++---
> >  mm/madvise.c                     |   15 +++++++--------
> >  2 files changed, 12 insertions(+), 11 deletions(-)
> >
> > --- 3045N.orig/drivers/staging/android/ashmem.c 2012-05-05 10:42:33.564056626 -0700
> > +++ 3045N/drivers/staging/android/ashmem.c      2012-05-05 10:46:25.692062478 -0700
> > @@ -19,6 +19,7 @@
> >  #include <linux/module.h>
> >  #include <linux/file.h>
> >  #include <linux/fs.h>
> > +#include <linux/falloc.h>
> >  #include <linux/miscdevice.h>
> >  #include <linux/security.h>
> >  #include <linux/mm.h>
> > @@ -363,11 +364,12 @@ static int ashmem_shrink(struct shrinker
> >
> >        mutex_lock(&ashmem_mutex);
> >        list_for_each_entry_safe(range, next, &ashmem_lru_list, lru) {
> > -               struct inode *inode = range->asma->file->f_dentry->d_inode;
> >                loff_t start = range->pgstart * PAGE_SIZE;
> > -               loff_t end = (range->pgend + 1) * PAGE_SIZE - 1;
> > +               loff_t end = (range->pgend + 1) * PAGE_SIZE;
> >
> > -               vmtruncate_range(inode, start, end);
> > +               do_fallocate(range->asma->file,
> > +                               FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
> > +                               start, end - start);
> >                range->purged = ASHMEM_WAS_PURGED;
> >                lru_del(range);
> >
> > --- 3045N.orig/mm/madvise.c     2012-05-05 10:42:33.572056784 -0700
> > +++ 3045N/mm/madvise.c  2012-05-05 10:46:25.692062478 -0700
> > @@ -11,8 +11,10 @@
> >  #include <linux/mempolicy.h>
> >  #include <linux/page-isolation.h>
> >  #include <linux/hugetlb.h>
> > +#include <linux/falloc.h>
> >  #include <linux/sched.h>
> >  #include <linux/ksm.h>
> > +#include <linux/fs.h>
> >
> >  /*
> >  * Any behaviour which results in changes to the vma->vm_flags needs to
> > @@ -200,8 +202,7 @@ static long madvise_remove(struct vm_are
> >                                struct vm_area_struct **prev,
> >                                unsigned long start, unsigned long end)
> >  {
> > -       struct address_space *mapping;
> > -       loff_t offset, endoff;
> > +       loff_t offset;
> >        int error;
> >
> >        *prev = NULL;   /* tell sys_madvise we drop mmap_sem */
> > @@ -217,16 +218,14 @@ static long madvise_remove(struct vm_are
> >        if ((vma->vm_flags & (VM_SHARED|VM_WRITE)) != (VM_SHARED|VM_WRITE))
> >                return -EACCES;
> >
> > -       mapping = vma->vm_file->f_mapping;
> > -
> >        offset = (loff_t)(start - vma->vm_start)
> >                        + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
> > -       endoff = (loff_t)(end - vma->vm_start - 1)
> > -                       + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
> >
> > -       /* vmtruncate_range needs to take i_mutex */
> > +       /* filesystem's fallocate may need to take i_mutex */
> >        up_read(&current->mm->mmap_sem);
> > -       error = vmtruncate_range(mapping->host, offset, endoff);
> > +       error = do_fallocate(vma->vm_file,
> > +                               FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
> > +                               offset, end - start);
> >        down_read(&current->mm->mmap_sem);
> >        return error;
> >  }
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at  http://www.tux.org/lkml/
>