Merge tag 'xfs-zoned-allocator-2025-03-03' of git://git.infradead.org/users/hch/xfs into xfs-6.15-zoned_devices

xfs: add support for zoned devices

Add support for the new zoned space allocator and thus for zoned devices:

    https://zonedstorage.io/docs/introduction/zoned-storage

to XFS. This has been developed for and tested on both SMR hard drives,
which are the oldest and most common class of zoned devices:

   https://zonedstorage.io/docs/introduction/smr

and ZNS SSDs:

   https://zonedstorage.io/docs/introduction/zns

It has not been tested with zoned UFS devices, as their current capacity
points and performance characteristics aren't too interesting for XFS
use cases (but never say never).

Sequential write only zones are only supported for data using a new
allocator for the RT device, which maps each zone to a rtgroup which
is written sequentially.  All metadata and (for now) the log require
using randomly writable space. This means a realtime device is required
to support zoned storage, but for the common case of SMR hard drives
that contain random writable zones and sequential write required zones
on the same block device, the concept of an internal RT device is added
which means using XFS on a SMR HDD is as simple as:

$ mkfs.xfs /dev/sda
$ mount /dev/sda /mnt

When using NVMe ZNS SSDs that do not support conventional zones, the
traditional multi-device RT configuration is required.  E.g. for an
SSD with a conventional namespace 1 and a zoned namespace 2:

$ mkfs.xfs /dev/nvme0n1 -o rtdev=/dev/nvme0n2
$ mount -o rtdev=/dev/nvme0n2 /dev/nvme0n1 /mnt

The zoned allocator can also be used on conventional block devices, or
on conventional zones (e.g. when using an SMR HDD as the external RT
device).  For example using zoned XFS on normal SSDs shows very nice
performance advantages and write amplification reduction for intelligent
workloads like RocksDB.

Some work is still in progress or planned, but should not affect the
integration with the rest of XFS or the on-disk format:

 - support for quotas
 - support for reflinks

Note that the I/O path already supports reflink, but garbage collection
isn't refcount aware yet and would unshare shared blocks, thus rendering
the feature useless.

1 files changed, 107 insertions, 60 deletions

diff --git a/fs/xfs/libxfs/xfs_metafile.c b/fs/xfs/libxfs/xfs_metafile.c
index 2f5f554a36d4..225923e463c4 100644
--- a/fs/xfs/libxfs/xfs_metafile.c
+++ b/fs/xfs/libxfs/xfs_metafile.c
@@ -21,6 +21,9 @@
 #include "xfs_errortag.h"
 #include "xfs_error.h"
 #include "xfs_alloc.h"
+#include "xfs_rtgroup.h"
+#include "xfs_rtrmap_btree.h"
+#include "xfs_rtrefcount_btree.h"
 
 static const struct {
 	enum xfs_metafile_type	mtype;
@@ -74,12 +77,11 @@ xfs_metafile_clear_iflag(
 }
 
 /*
- * Is the amount of space that could be allocated towards a given metadata
- * file at or beneath a certain threshold?
+ * Is the metafile reservations at or beneath a certain threshold?
  */
 static inline bool
 xfs_metafile_resv_can_cover(
-	struct xfs_inode	*ip,
+	struct xfs_mount	*mp,
 	int64_t			rhs)
 {
 	/*
@@ -88,43 +90,38 @@ xfs_metafile_resv_can_cover(
 	 * global free block count.  Take care of the first case to avoid
 	 * touching the per-cpu counter.
 	 */
-	if (ip->i_delayed_blks >= rhs)
+	if (mp->m_metafile_resv_avail >= rhs)
 		return true;
 
 	/*
 	 * There aren't enough blocks left in the inode's reservation, but it
 	 * isn't critical unless there also isn't enough free space.
 	 */
-	return __percpu_counter_compare(&ip->i_mount->m_fdblocks,
-			rhs - ip->i_delayed_blks, 2048) >= 0;
+	return xfs_compare_freecounter(mp, XC_FREE_BLOCKS,
+			rhs - mp->m_metafile_resv_avail, 2048) >= 0;
 }
 
 /*
- * Is this metadata file critically low on blocks?  For now we'll define that
- * as the number of blocks we can get our hands on being less than 10% of what
- * we reserved or less than some arbitrary number (maximum btree height).
+ * Is the metafile reservation critically low on blocks?  For now we'll define
+ * that as the number of blocks we can get our hands on being less than 10% of
+ * what we reserved or less than some arbitrary number (maximum btree height).
  */
 bool
 xfs_metafile_resv_critical(
-	struct xfs_inode	*ip)
+	struct xfs_mount	*mp)
 {
-	uint64_t		asked_low_water;
+	ASSERT(xfs_has_metadir(mp));
 
-	if (!ip)
-		return false;
-
-	ASSERT(xfs_is_metadir_inode(ip));
-	trace_xfs_metafile_resv_critical(ip, 0);
+	trace_xfs_metafile_resv_critical(mp, 0);
 
-	if (!xfs_metafile_resv_can_cover(ip, ip->i_mount->m_rtbtree_maxlevels))
+	if (!xfs_metafile_resv_can_cover(mp, mp->m_rtbtree_maxlevels))
 		return true;
 
-	asked_low_water = div_u64(ip->i_meta_resv_asked, 10);
-	if (!xfs_metafile_resv_can_cover(ip, asked_low_water))
+	if (!xfs_metafile_resv_can_cover(mp,
+			div_u64(mp->m_metafile_resv_target, 10)))
 		return true;
 
-	return XFS_TEST_ERROR(false, ip->i_mount,
-			XFS_ERRTAG_METAFILE_RESV_CRITICAL);
+	return XFS_TEST_ERROR(false, mp, XFS_ERRTAG_METAFILE_RESV_CRITICAL);
 }
 
 /* Allocate a block from the metadata file's reservation. */
@@ -133,22 +130,24 @@ xfs_metafile_resv_alloc_space(
 	struct xfs_inode	*ip,
 	struct xfs_alloc_arg	*args)
 {
+	struct xfs_mount	*mp = ip->i_mount;
 	int64_t			len = args->len;
 
 	ASSERT(xfs_is_metadir_inode(ip));
 	ASSERT(args->resv == XFS_AG_RESV_METAFILE);
 
-	trace_xfs_metafile_resv_alloc_space(ip, args->len);
+	trace_xfs_metafile_resv_alloc_space(mp, args->len);
 
 	/*
 	 * Allocate the blocks from the metadata inode's block reservation
 	 * and update the ondisk sb counter.
 	 */
-	if (ip->i_delayed_blks > 0) {
+	mutex_lock(&mp->m_metafile_resv_lock);
+	if (mp->m_metafile_resv_avail > 0) {
 		int64_t		from_resv;
 
-		from_resv = min_t(int64_t, len, ip->i_delayed_blks);
-		ip->i_delayed_blks -= from_resv;
+		from_resv = min_t(int64_t, len, mp->m_metafile_resv_avail);
+		mp->m_metafile_resv_avail -= from_resv;
 		xfs_mod_delalloc(ip, 0, -from_resv);
 		xfs_trans_mod_sb(args->tp, XFS_TRANS_SB_RES_FDBLOCKS,
 				-from_resv);
@@ -175,6 +174,9 @@ xfs_metafile_resv_alloc_space(
 		xfs_trans_mod_sb(args->tp, field, -len);
 	}
 
+	mp->m_metafile_resv_used += args->len;
+	mutex_unlock(&mp->m_metafile_resv_lock);
+
 	ip->i_nblocks += args->len;
 	xfs_trans_log_inode(args->tp, ip, XFS_ILOG_CORE);
 }
@@ -186,26 +188,33 @@ xfs_metafile_resv_free_space(
 	struct xfs_trans	*tp,
 	xfs_filblks_t		len)
 {
+	struct xfs_mount	*mp = ip->i_mount;
 	int64_t			to_resv;
 
 	ASSERT(xfs_is_metadir_inode(ip));
-	trace_xfs_metafile_resv_free_space(ip, len);
+
+	trace_xfs_metafile_resv_free_space(mp, len);
 
 	ip->i_nblocks -= len;
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
+	mutex_lock(&mp->m_metafile_resv_lock);
+	mp->m_metafile_resv_used -= len;
+
 	/*
 	 * Add the freed blocks back into the inode's delalloc reservation
 	 * until it reaches the maximum size.  Update the ondisk fdblocks only.
 	 */
-	to_resv = ip->i_meta_resv_asked - (ip->i_nblocks + ip->i_delayed_blks);
+	to_resv = mp->m_metafile_resv_target -
+		(mp->m_metafile_resv_used + mp->m_metafile_resv_avail);
 	if (to_resv > 0) {
 		to_resv = min_t(int64_t, to_resv, len);
-		ip->i_delayed_blks += to_resv;
+		mp->m_metafile_resv_avail += to_resv;
 		xfs_mod_delalloc(ip, 0, to_resv);
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_RES_FDBLOCKS, to_resv);
 		len -= to_resv;
 	}
+	mutex_unlock(&mp->m_metafile_resv_lock);
 
 	/*
 	 * Everything else goes back to the filesystem, so update the in-core
@@ -215,61 +224,99 @@ xfs_metafile_resv_free_space(
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, len);
 }
 
-/* Release a metadata file's space reservation. */
+static void
+__xfs_metafile_resv_free(
+	struct xfs_mount	*mp)
+{
+	if (mp->m_metafile_resv_avail) {
+		xfs_mod_sb_delalloc(mp, -(int64_t)mp->m_metafile_resv_avail);
+		xfs_add_fdblocks(mp, mp->m_metafile_resv_avail);
+	}
+	mp->m_metafile_resv_avail = 0;
+	mp->m_metafile_resv_used = 0;
+	mp->m_metafile_resv_target = 0;
+}
+
+/* Release unused metafile space reservation. */
 void
 xfs_metafile_resv_free(
-	struct xfs_inode	*ip)
+	struct xfs_mount	*mp)
 {
-	/* Non-btree metadata inodes don't need space reservations. */
-	if (!ip || !ip->i_meta_resv_asked)
+	if (!xfs_has_metadir(mp))
 		return;
 
-	ASSERT(xfs_is_metadir_inode(ip));
-	trace_xfs_metafile_resv_free(ip, 0);
+	trace_xfs_metafile_resv_free(mp, 0);
 
-	if (ip->i_delayed_blks) {
-		xfs_mod_delalloc(ip, 0, -ip->i_delayed_blks);
-		xfs_add_fdblocks(ip->i_mount, ip->i_delayed_blks);
-		ip->i_delayed_blks = 0;
-	}
-	ip->i_meta_resv_asked = 0;
+	mutex_lock(&mp->m_metafile_resv_lock);
+	__xfs_metafile_resv_free(mp);
+	mutex_unlock(&mp->m_metafile_resv_lock);
 }
 
-/* Set up a metadata file's space reservation. */
+/* Set up a metafile space reservation. */
 int
 xfs_metafile_resv_init(
-	struct xfs_inode	*ip,
-	xfs_filblks_t		ask)
+	struct xfs_mount	*mp)
 {
+	struct xfs_rtgroup	*rtg = NULL;
+	xfs_filblks_t		used = 0, target = 0;
 	xfs_filblks_t		hidden_space;
-	xfs_filblks_t		used;
-	int			error;
+	xfs_rfsblock_t		dblocks_avail = mp->m_sb.sb_dblocks / 4;
+	int			error = 0;
 
-	if (!ip || ip->i_meta_resv_asked > 0)
+	if (!xfs_has_metadir(mp))
 		return 0;
 
-	ASSERT(xfs_is_metadir_inode(ip));
+	/*
+	 * Free any previous reservation to have a clean slate.
+	 */
+	mutex_lock(&mp->m_metafile_resv_lock);
+	__xfs_metafile_resv_free(mp);
+
+	/*
+	 * Currently the only btree metafiles that require reservations are the
+	 * rtrmap and the rtrefcount.  Anything new will have to be added here
+	 * as well.
+	 */
+	while ((rtg = xfs_rtgroup_next(mp, rtg))) {
+		if (xfs_has_rtrmapbt(mp)) {
+			used += rtg_rmap(rtg)->i_nblocks;
+			target += xfs_rtrmapbt_calc_reserves(mp);
+		}
+		if (xfs_has_rtreflink(mp)) {
+			used += rtg_refcount(rtg)->i_nblocks;
+			target += xfs_rtrefcountbt_calc_reserves(mp);
+		}
+	}
+
+	if (!target)
+		goto out_unlock;
 
 	/*
-	 * Space taken by all other metadata btrees are accounted on-disk as
+	 * Space taken by the per-AG metadata btrees are accounted on-disk as
 	 * used space.  We therefore only hide the space that is reserved but
 	 * not used by the trees.
 	 */
-	used = ip->i_nblocks;
-	if (used > ask)
-		ask = used;
-	hidden_space = ask - used;
+	if (used > target)
+		target = used;
+	else if (target > dblocks_avail)
+		target = dblocks_avail;
+	hidden_space = target - used;
 
-	error = xfs_dec_fdblocks(ip->i_mount, hidden_space, true);
+	error = xfs_dec_fdblocks(mp, hidden_space, true);
 	if (error) {
-		trace_xfs_metafile_resv_init_error(ip, error, _RET_IP_);
-		return error;
+		trace_xfs_metafile_resv_init_error(mp, 0);
+		goto out_unlock;
 	}
 
-	xfs_mod_delalloc(ip, 0, hidden_space);
-	ip->i_delayed_blks = hidden_space;
-	ip->i_meta_resv_asked = ask;
+	xfs_mod_sb_delalloc(mp, hidden_space);
+
+	mp->m_metafile_resv_target = target;
+	mp->m_metafile_resv_used = used;
+	mp->m_metafile_resv_avail = hidden_space;
+
+	trace_xfs_metafile_resv_init(mp, target);
 
-	trace_xfs_metafile_resv_init(ip, ask);
-	return 0;
+out_unlock:
+	mutex_unlock(&mp->m_metafile_resv_lock);
+	return error;
 }