Re: [PATCH v4 55/66] mm/mempolicy: Use maple tree iterators instead of vma linked list

[Liam Howlett <liam.howlett@oracle.com>]

* Vlastimil Babka <vbabka@suse.cz> [220126 04:23]:
> On 1/26/22 03:48, Liam Howlett wrote:
> > * Vlastimil Babka <vbabka@suse.cz> [220120 06:58]:
> >> On 12/1/21 15:30, Liam Howlett wrote:
> >> > From: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
> >> > 
> >> > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> >> > Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
> >> > ---
> >> >  mm/mempolicy.c | 53 ++++++++++++++++++++++++++------------------------
> >> >  1 file changed, 28 insertions(+), 25 deletions(-)
> >> > 
> >> > diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> >> > index 10e9c87260ed..0e2d918f4f30 100644
> >> > --- a/mm/mempolicy.c
> >> > +++ b/mm/mempolicy.c
> >> > @@ -377,9 +377,10 @@ void mpol_rebind_task(struct task_struct *tsk, const nodemask_t *new)
> >> >  void mpol_rebind_mm(struct mm_struct *mm, nodemask_t *new)
> >> >  {
> >> >  	struct vm_area_struct *vma;
> >> > +	MA_STATE(mas, &mm->mm_mt, 0, 0);
> >> 
> >> VMA_ITERATOR?
> > 
> > Yes, I will make this change.  Thanks.
> > 
> >> 
> >> >  
> >> >  	mmap_write_lock(mm);
> >> > -	for (vma = mm->mmap; vma; vma = vma->vm_next)
> >> > +	mas_for_each(&mas, vma, ULONG_MAX)
> >> >  		mpol_rebind_policy(vma->vm_policy, new);
> >> >  	mmap_write_unlock(mm);
> >> >  }
> >> > @@ -652,7 +653,7 @@ static unsigned long change_prot_numa(struct vm_area_struct *vma,
> >> >  static int queue_pages_test_walk(unsigned long start, unsigned long end,
> >> >  				struct mm_walk *walk)
> >> >  {
> >> > -	struct vm_area_struct *vma = walk->vma;
> >> > +	struct vm_area_struct *next, *vma = walk->vma;
> >> >  	struct queue_pages *qp = walk->private;
> >> >  	unsigned long endvma = vma->vm_end;
> >> >  	unsigned long flags = qp->flags;
> >> > @@ -667,9 +668,10 @@ static int queue_pages_test_walk(unsigned long start, unsigned long end,
> >> >  			/* hole at head side of range */
> >> >  			return -EFAULT;
> >> >  	}
> >> > +	next = find_vma(vma->vm_mm, vma->vm_end);
> >> >  	if (!(flags & MPOL_MF_DISCONTIG_OK) &&
> >> >  		((vma->vm_end < qp->end) &&
> >> > -		(!vma->vm_next || vma->vm_end < vma->vm_next->vm_start)))
> >> > +		(!next || vma->vm_end < next->vm_start)))
> >> >  		/* hole at middle or tail of range */
> >> >  		return -EFAULT;
> >> >  
> >> > @@ -783,28 +785,24 @@ static int vma_replace_policy(struct vm_area_struct *vma,
> >> >  static int mbind_range(struct mm_struct *mm, unsigned long start,
> >> >  		       unsigned long end, struct mempolicy *new_pol)
> >> >  {
> >> > -	struct vm_area_struct *next;
> >> > +	MA_STATE(mas, &mm->mm_mt, start - 1, start - 1);
> >> >  	struct vm_area_struct *prev;
> >> >  	struct vm_area_struct *vma;
> >> >  	int err = 0;
> >> >  	pgoff_t pgoff;
> >> > -	unsigned long vmstart;
> >> > -	unsigned long vmend;
> >> > -
> >> > -	vma = find_vma(mm, start);
> >> > -	VM_BUG_ON(!vma);
> >> >  
> >> > -	prev = vma->vm_prev;
> >> > -	if (start > vma->vm_start)
> >> > -		prev = vma;
> >> > +	prev = mas_find_rev(&mas, 0);
> >> > +	if (prev && (start < prev->vm_end))
> >> > +		vma = prev;
> >> > +	else
> >> > +		vma = mas_next(&mas, end - 1);
> >> >  
> >> > -	for (; vma && vma->vm_start < end; prev = vma, vma = next) {
> >> > -		next = vma->vm_next;
> >> > -		vmstart = max(start, vma->vm_start);
> >> > -		vmend   = min(end, vma->vm_end);
> >> > +	do {
> >> > +		unsigned long vmstart = max(start, vma->vm_start);
> >> > +		unsigned long vmend = min(end, vma->vm_end);
> >> 
> >> What if vma is null? Shouldn't this rather be a "for (; vma; vma =
> >> mas_next(...)"
> > 
> > We have already found the vma above.
> 
> AFAICS if the range intersects no vmas, we might have found a 'prev', but
> 'vma' might be NULL after the "vma = mas_next(&mas, end - 1);"?

Yes, I agree.  I was going to restore VM_BUG_ON(!vma) after mas_next(),
but that's not the same as the existing code.  The VM_BUG_ON(!vma)
only catches if we search for 'start' above any VMA.  It seems
mbind_range() would have returned without error if there were no VMAs
within the range but would error if the 'start' was sufficiently large.

I think it is better to write it as you suggested to restore the
functionality of not failing on an empty list.  I don't see a decent
way of checking if we searched for an address above any VMA to restore
the VM_BUG_ON() which existed - although I see little value in it to
begin with.  I will mention this in the commit message.