On Mon, Sep 18, 2023 at 01:04:55PM +0200, Hannes Reinecke wrote:
> @@ -449,6 +450,22 @@ __bread(struct block_device *bdev, sector_t block, unsigned size)
>
> bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
>
> +static inline sector_t block_index_to_sector(pgoff_t index, unsigned int blkbits)
> +{
> + if (PAGE_SHIFT < blkbits)
> + return (sector_t)index >> (blkbits - PAGE_SHIFT);
> + else
> + return (sector_t)index << (PAGE_SHIFT - blkbits);
> +}
Is this actually more efficient than ...
loff_t pos = (loff_t)index * PAGE_SIZE;
return pos >> blkbits;
It feels like we're going to be doing this a lot, so we should find out
what's actually faster.
