On Thu, Oct 3, 2024 at 12:00 AM Matthew Wilcox (Oracle) wrote:
>
> Use memcpy_to_folio() instead of open-coding it, and use offset_in_folio()
> in case anybody wants to use nilfs2 on a device with large blocks.
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> ---
> fs/nilfs2/recovery.c | 11 ++++-------
> 1 file changed, 4 insertions(+), 7 deletions(-)
>
> diff --git a/fs/nilfs2/recovery.c b/fs/nilfs2/recovery.c
> index 21d81097a89f..1c665a32f002 100644
> --- a/fs/nilfs2/recovery.c
> +++ b/fs/nilfs2/recovery.c
> @@ -481,19 +481,16 @@ static int nilfs_prepare_segment_for_recovery(struct the_nilfs *nilfs,
>
> static int nilfs_recovery_copy_block(struct the_nilfs *nilfs,
> struct nilfs_recovery_block *rb,
> - loff_t pos, struct page *page)
> + loff_t pos, struct folio *folio)
> {
> struct buffer_head *bh_org;
> - size_t from = pos & ~PAGE_MASK;
> - void *kaddr;
> + size_t from = offset_in_folio(folio, pos);
>
> bh_org = __bread(nilfs->ns_bdev, rb->blocknr, nilfs->ns_blocksize);
> if (unlikely(!bh_org))
> return -EIO;
>
> - kaddr = kmap_local_page(page);
> - memcpy(kaddr + from, bh_org->b_data, bh_org->b_size);
> - kunmap_local(kaddr);
> + memcpy_to_folio(folio, from, bh_org->b_data, bh_org->b_size);
> brelse(bh_org);
> return 0;
> }
> @@ -531,7 +528,7 @@ static int nilfs_recover_dsync_blocks(struct the_nilfs *nilfs,
> goto failed_inode;
> }
>
> - err = nilfs_recovery_copy_block(nilfs, rb, pos, &folio->page);
> + err = nilfs_recovery_copy_block(nilfs, rb, pos, folio);
> if (unlikely(err))
> goto failed_page;
>
> --
> 2.43.0
>
This patch looks good to me.
One small thing: with this conversion, there is no reference to the
page structure in nilfs_recover_dsync_blocks, so how about changing
the jump label "failed_page" to "failed_folio"?
This will reduce noise when searching for "page" with grep.
Thanks,
Ryusuke Konishi
