Re: [PATCH 3/9] maple_tree: Modify the allocation method of mtree_alloc_range/rrange()

[Liam R. Howlett]

* Peng Zhang <perlyzhang@xxxxxxxxx> [230426 08:34]:
> 
> 
> 在 2023/4/26 00:08, Liam R. Howlett 写道:
> > * Peng Zhang <zhangpeng.00@xxxxxxxxxxxxx> [230425 07:05]:
> > > Let mtree_alloc_range() and mtree_alloc_rrange() use mas_empty_area()
> > > and mas_empty_area_rev() respectively for allocation to reduce code
> > > redundancy. And after doing this, we don't need to maintain two logically
> > > identical codes to improve maintainability.
> > > 
> > > In fact, mtree_alloc_range/rrange() has some bugs. For example, when
> > > dealing with min equals to max (mas_empty_area/area_rev() has been fixed),
> > > the allocation will fail.
> > > There are still some other bugs in it, I saw it with my naked eyes, but
> > > I didn't test it, for example:
> > > When mtree_alloc_range()->mas_alloc()->mas_awalk(), we set mas.index = min,
> > > mas.last = max - size. However, mas_awalk() requires mas.index = min,
> > > mas.last = max, which may lead to allocation failures.
> > 
> > Please don't re-state code in your commit messages.
> > 
> > Try to focus on what you did, and not why.
> > 
> > ie: Aligned mtree_alloc_range() to use the same internal function as
> > mas_empty_area().
> > 
> > > 
> > > Right now no users are using these two functions so the bug won't trigger,
> > > but this might trigger in the future.
> > > 
> > > Also use mas_store_gfp() instead of mas_fill_gap() as I don't see any
> > > difference between them.
> > 
> > Yeah, evolution of the code converged on the same design.  Thanks for
> > seeing this.
> > 
> > > 
> > > After doing this, we no longer need the three functions
> > > mas_fill_gap(), mas_alloc(), and mas_rev_alloc().
> > 
> > Let's just drop mtree_alloc_range() and mtree_alloc_rrange() and
> > whatever else you found here.  They were planned to simplify the mmap
> > code allocations, but since there would need to be arch involvement
> > (coloring, etc) and alignment, etc; it is better to leave this job to
> > the mm code itself.
> Ok, I will remove some useless functions here.
> But mtree_alloc_range() and mtree_alloc_rrange() really don't need to be
> reserved? Because I don't know if there will be users using it in other
> scenarios in the future.

As you showed, a lot of the code is now the same elsewhere, so it
wouldn't take much to make a version of this outside of the tree if
someone needs the functionality.

> 
> Thank you for all your suggestions on this patch set, I will update them.
> > 
> > > 
> > > Fixes: 54a611b60590 ("Maple Tree: add new data structure")
> > > Signed-off-by: Peng Zhang <zhangpeng.00@xxxxxxxxxxxxx>
> > > ---
> > >   lib/maple_tree.c | 45 ++++++++++++---------------------------------
> > >   1 file changed, 12 insertions(+), 33 deletions(-)
> > > 
> > > diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> > > index aa55c914818a0..294d4c8668323 100644
> > > --- a/lib/maple_tree.c
> > > +++ b/lib/maple_tree.c
> > > @@ -6362,32 +6362,20 @@ int mtree_alloc_range(struct maple_tree *mt, unsigned long *startp,
> > >   {
> > >   	int ret = 0;
> > > -	MA_STATE(mas, mt, min, max - size);
> > > +	MA_STATE(mas, mt, 0, 0);
> > >   	if (!mt_is_alloc(mt))
> > >   		return -EINVAL;
> > >   	if (WARN_ON_ONCE(mt_is_reserved(entry)))
> > >   		return -EINVAL;
> > > -	if (min > max)
> > > -		return -EINVAL;
> > > -
> > > -	if (max < size)
> > > -		return -EINVAL;
> > > -
> > > -	if (!size)
> > > -		return -EINVAL;
> > > -
> > >   	mtree_lock(mt);
> > > -retry:
> > > -	mas.offset = 0;
> > > -	mas.index = min;
> > > -	mas.last = max - size;
> > > -	ret = mas_alloc(&mas, entry, size, startp);
> > > -	if (mas_nomem(&mas, gfp))
> > > -		goto retry;
> > > -
> > > +	ret = mas_empty_area(&mas, min, max, size);
> > > +	if (!ret)
> > > +		ret = mas_store_gfp(&mas, entry, gfp);
> > >   	mtree_unlock(mt);
> > > +	if (!ret)
> > > +		*startp = mas.index;
> > >   	return ret;
> > >   }
> > >   EXPORT_SYMBOL(mtree_alloc_range);
> > > @@ -6398,29 +6386,20 @@ int mtree_alloc_rrange(struct maple_tree *mt, unsigned long *startp,
> > >   {
> > >   	int ret = 0;
> > > -	MA_STATE(mas, mt, min, max - size);
> > > +	MA_STATE(mas, mt, 0, 0);
> > >   	if (!mt_is_alloc(mt))
> > >   		return -EINVAL;
> > >   	if (WARN_ON_ONCE(mt_is_reserved(entry)))
> > >   		return -EINVAL;
> > > -	if (min >= max)
> > > -		return -EINVAL;
> > > -
> > > -	if (max < size - 1)
> > > -		return -EINVAL;
> > > -
> > > -	if (!size)
> > > -		return -EINVAL;
> > > -
> > >   	mtree_lock(mt);
> > > -retry:
> > > -	ret = mas_rev_alloc(&mas, min, max, entry, size, startp);
> > > -	if (mas_nomem(&mas, gfp))
> > > -		goto retry;
> > > -
> > > +	ret = mas_empty_area_rev(&mas, min, max, size);
> > > +	if (!ret)
> > > +		ret = mas_store_gfp(&mas, entry, gfp);
> > >   	mtree_unlock(mt);
> > > +	if (!ret)
> > > +		*startp = mas.index;
> > >   	return ret;
> > >   }
> > >   EXPORT_SYMBOL(mtree_alloc_rrange);
> > > -- 
> > > 2.20.1
> > >