virtio_blk: set the zone write granularitygrafted

virtblk_read_zoned_limits() reads the write granularity that the device
reports in virtio_blk_zoned_characteristics and assigns it to the
physical block size and to io_min, but never to the limit that is named
after it. queue_limits.zone_write_granularity is left at zero, so
blk_validate_zoned_limits() raises it to the logical block size:

	if (lim->zone_write_granularity < lim->logical_block_size)
		lim->zone_write_granularity = lim->logical_block_size;

A device that reports a granularity coarser than its logical block size,
which is what the field exists to express, therefore has it silently
reduced. A 512e host managed disk passed through to a guest reports a
logical block size of 512 and a write granularity of 4096, and the guest
ends up with a zone write granularity of 512.

bio_split_alignment() returns lim->zone_write_granularity if it is non-zero
and bio_split_io_at() may split a bio with as per bio_split_alignment().
This can real to the write getting rejected by the host drive, as the write
is not aligned to the physical block size.

zonefs also takes its block size from bdev_zone_write_granularity(), so it
would incorrectly use 512 on a disk that requires 4096.

sd_zbc_read_zones() sets the limit from the physical block size for the
same reason. NVMe ZNS and null_blk leave it unset, but the fallback
gives the right answer for them, as their write granularity is the
logical block size. virtio carries a separate value that may exceed it.

Set the zone write granularity from the value that the device reports.

Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://patch.msgid.link/20260918140641.2031075-2-cassel@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 365 insertions, 0 deletions

diff --git a/fs/affs/bitmap.c b/fs/affs/bitmap.c
new file mode 100644
index 000000000..40bc4ce6a
--- /dev/null
+++ b/fs/affs/bitmap.c
@@ -0,0 +1,365 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/fs/affs/bitmap.c
+ *
+ *  (c) 1996 Hans-Joachim Widmaier
+ *
+ *  bitmap.c contains the code that handles all bitmap related stuff -
+ *  block allocation, deallocation, calculation of free space.
+ */
+
+#include <linux/slab.h>
+#include "affs.h"
+
+u32
+affs_count_free_blocks(struct super_block *sb)
+{
+	struct affs_bm_info *bm;
+	u32 free;
+	int i;
+
+	pr_debug("%s()\n", __func__);
+
+	if (sb_rdonly(sb))
+		return 0;
+
+	mutex_lock(&AFFS_SB(sb)->s_bmlock);
+
+	bm = AFFS_SB(sb)->s_bitmap;
+	free = 0;
+	for (i = AFFS_SB(sb)->s_bmap_count; i > 0; bm++, i--)
+		free += bm->bm_free;
+
+	mutex_unlock(&AFFS_SB(sb)->s_bmlock);
+
+	return free;
+}
+
+void
+affs_free_block(struct super_block *sb, u32 block)
+{
+	struct affs_sb_info *sbi = AFFS_SB(sb);
+	struct affs_bm_info *bm;
+	struct buffer_head *bh;
+	u32 blk, bmap, bit, mask, tmp;
+	__be32 *data;
+
+	pr_debug("%s(%u)\n", __func__, block);
+
+	if (block > sbi->s_partition_size)
+		goto err_range;
+
+	blk     = block - sbi->s_reserved;
+	bmap    = blk / sbi->s_bmap_bits;
+	bit     = blk % sbi->s_bmap_bits;
+	bm      = &sbi->s_bitmap[bmap];
+
+	mutex_lock(&sbi->s_bmlock);
+
+	bh = sbi->s_bmap_bh;
+	if (sbi->s_last_bmap != bmap) {
+		affs_brelse(bh);
+		bh = affs_bread(sb, bm->bm_key);
+		if (!bh)
+			goto err_bh_read;
+		sbi->s_bmap_bh = bh;
+		sbi->s_last_bmap = bmap;
+	}
+
+	mask = 1 << (bit & 31);
+	data = (__be32 *)bh->b_data + bit / 32 + 1;
+
+	/* mark block free */
+	tmp = be32_to_cpu(*data);
+	if (tmp & mask)
+		goto err_free;
+	*data = cpu_to_be32(tmp | mask);
+
+	/* fix checksum */
+	tmp = be32_to_cpu(*(__be32 *)bh->b_data);
+	*(__be32 *)bh->b_data = cpu_to_be32(tmp - mask);
+
+	mark_buffer_dirty(bh);
+	affs_mark_sb_dirty(sb);
+	bm->bm_free++;
+
+	mutex_unlock(&sbi->s_bmlock);
+	return;
+
+err_free:
+	affs_warning(sb,"affs_free_block","Trying to free block %u which is already free", block);
+	mutex_unlock(&sbi->s_bmlock);
+	return;
+
+err_bh_read:
+	affs_error(sb,"affs_free_block","Cannot read bitmap block %u", bm->bm_key);
+	sbi->s_bmap_bh = NULL;
+	sbi->s_last_bmap = ~0;
+	mutex_unlock(&sbi->s_bmlock);
+	return;
+
+err_range:
+	affs_error(sb, "affs_free_block","Block %u outside partition", block);
+}
+
+/*
+ * Allocate a block in the given allocation zone.
+ * Since we have to byte-swap the bitmap on little-endian
+ * machines, this is rather expensive. Therefore we will
+ * preallocate up to 16 blocks from the same word, if
+ * possible. We are not doing preallocations in the
+ * header zone, though.
+ */
+
+u32
+affs_alloc_block(struct inode *inode, u32 goal)
+{
+	struct super_block *sb;
+	struct affs_sb_info *sbi;
+	struct affs_bm_info *bm;
+	struct buffer_head *bh;
+	__be32 *data, *enddata;
+	u32 blk, bmap, bit, mask, mask2, tmp;
+	int i;
+
+	sb = inode->i_sb;
+	sbi = AFFS_SB(sb);
+
+	pr_debug("balloc(inode=%llu,goal=%u): ", inode->i_ino, goal);
+
+	if (AFFS_I(inode)->i_pa_cnt) {
+		pr_debug("%d\n", AFFS_I(inode)->i_lastalloc+1);
+		AFFS_I(inode)->i_pa_cnt--;
+		return ++AFFS_I(inode)->i_lastalloc;
+	}
+
+	if (!goal || goal > sbi->s_partition_size) {
+		if (goal)
+			affs_warning(sb, "affs_balloc", "invalid goal %d", goal);
+		//if (!AFFS_I(inode)->i_last_block)
+		//	affs_warning(sb, "affs_balloc", "no last alloc block");
+		goal = sbi->s_reserved;
+	}
+
+	blk = goal - sbi->s_reserved;
+	bmap = blk / sbi->s_bmap_bits;
+	bm = &sbi->s_bitmap[bmap];
+
+	mutex_lock(&sbi->s_bmlock);
+
+	if (bm->bm_free)
+		goto find_bmap_bit;
+
+find_bmap:
+	/* search for the next bmap buffer with free bits */
+	i = sbi->s_bmap_count;
+	do {
+		if (--i < 0)
+			goto err_full;
+		bmap++;
+		bm++;
+		if (bmap < sbi->s_bmap_count)
+			continue;
+		/* restart search at zero */
+		bmap = 0;
+		bm = sbi->s_bitmap;
+	} while (!bm->bm_free);
+	blk = bmap * sbi->s_bmap_bits;
+
+find_bmap_bit:
+
+	bh = sbi->s_bmap_bh;
+	if (sbi->s_last_bmap != bmap) {
+		affs_brelse(bh);
+		bh = affs_bread(sb, bm->bm_key);
+		if (!bh)
+			goto err_bh_read;
+		sbi->s_bmap_bh = bh;
+		sbi->s_last_bmap = bmap;
+	}
+
+	/* find an unused block in this bitmap block */
+	bit = blk % sbi->s_bmap_bits;
+	data = (__be32 *)bh->b_data + bit / 32 + 1;
+	enddata = (__be32 *)((u8 *)bh->b_data + sb->s_blocksize);
+	mask = ~0UL << (bit & 31);
+	blk &= ~31UL;
+
+	tmp = be32_to_cpu(*data);
+	if (tmp & mask)
+		goto find_bit;
+
+	/* scan the rest of the buffer */
+	do {
+		blk += 32;
+		if (++data >= enddata)
+			/* didn't find something, can only happen
+			 * if scan didn't start at 0, try next bmap
+			 */
+			goto find_bmap;
+	} while (!*data);
+	tmp = be32_to_cpu(*data);
+	mask = ~0;
+
+find_bit:
+	/* finally look for a free bit in the word */
+	bit = ffs(tmp & mask) - 1;
+	blk += bit + sbi->s_reserved;
+	mask2 = mask = 1 << (bit & 31);
+	AFFS_I(inode)->i_lastalloc = blk;
+
+	/* prealloc as much as possible within this word */
+	while ((mask2 <<= 1)) {
+		if (!(tmp & mask2))
+			break;
+		AFFS_I(inode)->i_pa_cnt++;
+		mask |= mask2;
+	}
+	bm->bm_free -= AFFS_I(inode)->i_pa_cnt + 1;
+
+	*data = cpu_to_be32(tmp & ~mask);
+
+	/* fix checksum */
+	tmp = be32_to_cpu(*(__be32 *)bh->b_data);
+	*(__be32 *)bh->b_data = cpu_to_be32(tmp + mask);
+
+	mark_buffer_dirty(bh);
+	affs_mark_sb_dirty(sb);
+
+	mutex_unlock(&sbi->s_bmlock);
+
+	pr_debug("%d\n", blk);
+	return blk;
+
+err_bh_read:
+	affs_error(sb,"affs_read_block","Cannot read bitmap block %u", bm->bm_key);
+	sbi->s_bmap_bh = NULL;
+	sbi->s_last_bmap = ~0;
+err_full:
+	mutex_unlock(&sbi->s_bmlock);
+	pr_debug("failed\n");
+	return 0;
+}
+
+int affs_init_bitmap(struct super_block *sb, int *flags)
+{
+	struct affs_bm_info *bm;
+	struct buffer_head *bmap_bh = NULL, *bh = NULL;
+	__be32 *bmap_blk;
+	u32 size, blk, end, offset, mask;
+	int i, res = 0;
+	struct affs_sb_info *sbi = AFFS_SB(sb);
+
+	if (*flags & SB_RDONLY)
+		return 0;
+
+	if (!AFFS_ROOT_TAIL(sb, sbi->s_root_bh)->bm_flag) {
+		pr_notice("Bitmap invalid - mounting %s read only\n", sb->s_id);
+		*flags |= SB_RDONLY;
+		return 0;
+	}
+
+	sbi->s_last_bmap = ~0;
+	sbi->s_bmap_bh = NULL;
+	sbi->s_bmap_bits = sb->s_blocksize * 8 - 32;
+	sbi->s_bmap_count = (sbi->s_partition_size - sbi->s_reserved +
+				 sbi->s_bmap_bits - 1) / sbi->s_bmap_bits;
+	size = sbi->s_bmap_count * sizeof(*bm);
+	bm = sbi->s_bitmap = kzalloc(size, GFP_KERNEL);
+	if (!sbi->s_bitmap) {
+		pr_err("Bitmap allocation failed\n");
+		return -ENOMEM;
+	}
+
+	bmap_blk = (__be32 *)sbi->s_root_bh->b_data;
+	blk = sb->s_blocksize / 4 - 49;
+	end = blk + 25;
+
+	for (i = sbi->s_bmap_count; i > 0; bm++, i--) {
+		affs_brelse(bh);
+
+		bm->bm_key = be32_to_cpu(bmap_blk[blk]);
+		bh = affs_bread(sb, bm->bm_key);
+		if (!bh) {
+			pr_err("Cannot read bitmap\n");
+			res = -EIO;
+			goto out;
+		}
+		if (affs_checksum_block(sb, bh)) {
+			pr_warn("Bitmap %u invalid - mounting %s read only.\n",
+				bm->bm_key, sb->s_id);
+			*flags |= SB_RDONLY;
+			goto out;
+		}
+		pr_debug("read bitmap block %d: %d\n", blk, bm->bm_key);
+		bm->bm_free = memweight(bh->b_data + 4, sb->s_blocksize - 4);
+
+		/* Don't try read the extension if this is the last block,
+		 * but we also need the right bm pointer below
+		 */
+		if (++blk < end || i == 1)
+			continue;
+		if (bmap_bh)
+			affs_brelse(bmap_bh);
+		bmap_bh = affs_bread(sb, be32_to_cpu(bmap_blk[blk]));
+		if (!bmap_bh) {
+			pr_err("Cannot read bitmap extension\n");
+			res = -EIO;
+			goto out;
+		}
+		bmap_blk = (__be32 *)bmap_bh->b_data;
+		blk = 0;
+		end = sb->s_blocksize / 4 - 1;
+	}
+
+	offset = (sbi->s_partition_size - sbi->s_reserved) % sbi->s_bmap_bits;
+	mask = ~(0xFFFFFFFFU << (offset & 31));
+	pr_debug("last word: %d %d %d\n", offset, offset / 32 + 1, mask);
+	offset = offset / 32 + 1;
+
+	if (mask) {
+		u32 old, new;
+
+		/* Mark unused bits in the last word as allocated */
+		old = be32_to_cpu(((__be32 *)bh->b_data)[offset]);
+		new = old & mask;
+		//if (old != new) {
+			((__be32 *)bh->b_data)[offset] = cpu_to_be32(new);
+			/* fix checksum */
+			//new -= old;
+			//old = be32_to_cpu(*(__be32 *)bh->b_data);
+			//*(__be32 *)bh->b_data = cpu_to_be32(old - new);
+			//mark_buffer_dirty(bh);
+		//}
+		/* correct offset for the bitmap count below */
+		//offset++;
+	}
+	while (++offset < sb->s_blocksize / 4)
+		((__be32 *)bh->b_data)[offset] = 0;
+	((__be32 *)bh->b_data)[0] = 0;
+	((__be32 *)bh->b_data)[0] = cpu_to_be32(-affs_checksum_block(sb, bh));
+	mark_buffer_dirty(bh);
+
+	/* recalculate bitmap count for last block */
+	bm--;
+	bm->bm_free = memweight(bh->b_data + 4, sb->s_blocksize - 4);
+
+out:
+	affs_brelse(bh);
+	affs_brelse(bmap_bh);
+	return res;
+}
+
+void affs_free_bitmap(struct super_block *sb)
+{
+	struct affs_sb_info *sbi = AFFS_SB(sb);
+
+	if (!sbi->s_bitmap)
+		return;
+
+	affs_brelse(sbi->s_bmap_bh);
+	sbi->s_bmap_bh = NULL;
+	sbi->s_last_bmap = ~0;
+	kfree(sbi->s_bitmap);
+	sbi->s_bitmap = NULL;
+}