nilfs2: convert recovery logic to use kmap_local

Patch series "nilfs2: eliminate kmap and kmap_atomic calls".

This series converts remaining kmap and kmap_atomic calls to use
kmap_local, mainly in metadata files, and eliminates calls to these
deprecated kmap functions from nilfs2.

This series does not include converting metadata files to use folios, but
it is a step in that direction.

Most conversions are straightforward, but some are not: the checkpoint
file, the inode file, and the persistent object allocator.  These have
been adjusted or rewritten to avoid multiple kmap_local calls or nest them
if necessary, and to eliminate long waits like block I/O within the
highmem mapping sections.

This series has been tested in both 32-bit and 64-bit environments with
varying block sizes.


This patch (of 15):

In the recovery function when mounting, nilfs_recovery_copy_block() uses
the deprecated kmap_atomic(), so convert it to use kmap_local.

Link: https://lkml.kernel.org/r/20240122140202.6950-1-konishi.ryusuke@gmail.com
Link: https://lkml.kernel.org/r/20240122140202.6950-2-konishi.ryusuke@gmail.com
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fs/nilfs2/recovery.c

@@ -482,9 +482,9 @@ static int nilfs_recovery_copy_block(struct the_nilfs *nilfs,
 	if (unlikely(!bh_org))
 		return -EIO;
 
-	kaddr = kmap_atomic(page);
+	kaddr = kmap_local_page(page);
 	memcpy(kaddr + from, bh_org->b_data, bh_org->b_size);
-	kunmap_atomic(kaddr);
+	kunmap_local(kaddr);
 	brelse(bh_org);
 	return 0;
 }