Re: [RFC v2 02/14] filemap: align the index to mapping_min_order in the page cache

[Dave Chinner <david@fromorbit.com>]

On Tue, Feb 13, 2024 at 10:37:01AM +0100, Pankaj Raghav (Samsung) wrote:
> From: Luis Chamberlain <mcgrof@kernel.org>
> 
> Supporting mapping_min_order implies that we guarantee each folio in the
> page cache has at least an order of mapping_min_order. So when adding new
> folios to the page cache we must ensure the index used is aligned to the
> mapping_min_order as the page cache requires the index to be aligned to
> the order of the folio.
> 
> A higher order folio than min_order by definition is a multiple of the
> min_order. If an index is aligned to an order higher than a min_order, it
> will also be aligned to the min order.
> 
> This effectively introduces no new functional changes when min order is
> not set other than a few rounding computations that should result in the
> same value.
> 
> Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
> Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
> ---
>  mm/filemap.c | 34 ++++++++++++++++++++++++++--------
>  1 file changed, 26 insertions(+), 8 deletions(-)
> 
> diff --git a/mm/filemap.c b/mm/filemap.c
> index 750e779c23db..323a8e169581 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -2479,14 +2479,16 @@ static int filemap_get_pages(struct kiocb *iocb, size_t count,
>  {
>  	struct file *filp = iocb->ki_filp;
>  	struct address_space *mapping = filp->f_mapping;
> +	unsigned int min_nrpages = mapping_min_folio_nrpages(mapping);
>  	struct file_ra_state *ra = &filp->f_ra;
> -	pgoff_t index = iocb->ki_pos >> PAGE_SHIFT;
> +	pgoff_t index = round_down(iocb->ki_pos >> PAGE_SHIFT, min_nrpages);

This is pretty magic - this patch adds a bunch of what appear to be
random rounding operations for some undocumented reason.

>  	pgoff_t last_index;
>  	struct folio *folio;
>  	int err = 0;
>  
>  	/* "last_index" is the index of the page beyond the end of the read */
>  	last_index = DIV_ROUND_UP(iocb->ki_pos + count, PAGE_SIZE);
> +	last_index = round_up(last_index, min_nrpages);

Same here - this is pretty nasty - we round up twice, but no obvious
reason as to why the second round up exists or why it can't be done
by the DIV_ROUND_UP() call. Just looking at the code it's impossible
to reason why this is being done, let alone determine if it has been
implemented correctly.

>  retry:
>  	if (fatal_signal_pending(current))
>  		return -EINTR;
> @@ -2502,8 +2504,7 @@ static int filemap_get_pages(struct kiocb *iocb, size_t count,
>  	if (!folio_batch_count(fbatch)) {
>  		if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_WAITQ))
>  			return -EAGAIN;
> -		err = filemap_create_folio(filp, mapping,
> -				iocb->ki_pos >> PAGE_SHIFT, fbatch);
> +		err = filemap_create_folio(filp, mapping, index, fbatch);
>  		if (err == AOP_TRUNCATED_PAGE)
>  			goto retry;
>  		return err;
> @@ -3095,7 +3096,10 @@ static struct file *do_sync_mmap_readahead(struct vm_fault *vmf)
>  	struct file *file = vmf->vma->vm_file;
>  	struct file_ra_state *ra = &file->f_ra;
>  	struct address_space *mapping = file->f_mapping;
> -	DEFINE_READAHEAD(ractl, file, ra, mapping, vmf->pgoff);
> +	unsigned int min_order = mapping_min_folio_order(mapping);
> +	unsigned int min_nrpages = mapping_min_folio_nrpages(file->f_mapping);

Why use file->f_mapping here and not mapping? And why not just

	unsigned int min_nrpages = 1U < min_order;

So it's obvious how the index alignment is related to the folio
order?

> +	pgoff_t index = round_down(vmf->pgoff, min_nrpages);
> +	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
>  	struct file *fpin = NULL;
>  	unsigned long vm_flags = vmf->vma->vm_flags;
>  	unsigned int mmap_miss;
> @@ -3147,10 +3151,11 @@ static struct file *do_sync_mmap_readahead(struct vm_fault *vmf)
>  	 */
>  	fpin = maybe_unlock_mmap_for_io(vmf, fpin);
>  	ra->start = max_t(long, 0, vmf->pgoff - ra->ra_pages / 2);
> +	ra->start = round_down(ra->start, min_nrpages);

Again, another random rounding operation....

>  	ra->size = ra->ra_pages;
>  	ra->async_size = ra->ra_pages / 4;
>  	ractl._index = ra->start;
> -	page_cache_ra_order(&ractl, ra, 0);
> +	page_cache_ra_order(&ractl, ra, min_order);
>  	return fpin;
>  }
>  
> @@ -3164,7 +3169,9 @@ static struct file *do_async_mmap_readahead(struct vm_fault *vmf,
>  {
>  	struct file *file = vmf->vma->vm_file;
>  	struct file_ra_state *ra = &file->f_ra;
> -	DEFINE_READAHEAD(ractl, file, ra, file->f_mapping, vmf->pgoff);
> +	unsigned int min_nrpages = mapping_min_folio_nrpages(file->f_mapping);
> +	pgoff_t index = round_down(vmf->pgoff, min_nrpages);
> +	DEFINE_READAHEAD(ractl, file, ra, file->f_mapping, index);

Ok, this is begging for a new DEFINE_READAHEAD_ALIGNED() macro which
internally grabs the mapping_min_folio_nrpages() from the mapping
passed to the macro.

>  	struct file *fpin = NULL;
>  	unsigned int mmap_miss;
>  
> @@ -3212,13 +3219,17 @@ vm_fault_t filemap_fault(struct vm_fault *vmf)
>  	struct file *file = vmf->vma->vm_file;
>  	struct file *fpin = NULL;
>  	struct address_space *mapping = file->f_mapping;
> +	unsigned int min_order = mapping_min_folio_order(mapping);
> +	unsigned int nrpages = 1UL << min_order;

You didn't use mapping_min_folio_nrpages() for this.

At least initialise all the variables the same way in the same
patch!

>  	struct inode *inode = mapping->host;
> -	pgoff_t max_idx, index = vmf->pgoff;
> +	pgoff_t max_idx, index = round_down(vmf->pgoff, nrpages);

Yup, I can't help but think that with how many times this is being
repeated in this patchset that a helper or two is in order:

	index = mapping_align_start_index(mapping, vmf->pgoff);

And then most of the calls to mapping_min_folio_order() and
mapping_min_folio_nrpages() can be removed from this code, too.


>  	struct folio *folio;
>  	vm_fault_t ret = 0;
>  	bool mapping_locked = false;
>  
>  	max_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
> +	max_idx = round_up(max_idx, nrpages);

	max_index = mapping_align_end_index(mapping,
			DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));

>  	if (unlikely(index >= max_idx))
>  		return VM_FAULT_SIGBUS;
>  
> @@ -3317,13 +3328,17 @@ vm_fault_t filemap_fault(struct vm_fault *vmf)
>  	 * We must recheck i_size under page lock.
>  	 */
>  	max_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
> +	max_idx = round_up(max_idx, nrpages);

Same again:

	max_index = mapping_align_end_index(mapping,
			DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE));
> +
>  	if (unlikely(index >= max_idx)) {
>  		folio_unlock(folio);
>  		folio_put(folio);
>  		return VM_FAULT_SIGBUS;
>  	}
>  
> -	vmf->page = folio_file_page(folio, index);
> +	VM_BUG_ON_FOLIO(folio_order(folio) < min_order, folio);
> +
> +	vmf->page = folio_file_page(folio, vmf->pgoff);
>  	return ret | VM_FAULT_LOCKED;
>  
>  page_not_uptodate:
> @@ -3658,6 +3673,9 @@ static struct folio *do_read_cache_folio(struct address_space *mapping,
>  {
>  	struct folio *folio;
>  	int err;
> +	unsigned int min_nrpages = mapping_min_folio_nrpages(mapping);
> +
> +	index = round_down(index, min_nrpages);

And more magic rounding.

	index = mapping_align_start_index(mapping, index);

-Dave.
-- 
Dave Chinner
david@fromorbit.com