* [Devel] [PATCH hci-8.1 v2 2/3] ext4: discard the balloon on FITRIM
2026-11-10 22:27 [Devel] [PATCH hci-8.1 v2 0/3] ext4, xfs: discard the balloon on FITRIM Denis V. Lunev
2026-11-10 22:27 ` [Devel] [PATCH hci-8.1 v2 1/3] xfs: release the balloon inode when xfs_balloon_get() refuses it Denis V. Lunev
@ 2026-11-10 22:27 ` Denis V. Lunev
2026-11-10 22:35 ` no-reply
2026-11-10 22:27 ` [Devel] [PATCH hci-8.1 v2 3/3] xfs: " Denis V. Lunev
2 siblings, 1 reply; 7+ messages in thread
From: Denis V. Lunev @ 2026-11-10 22:27 UTC (permalink / raw)
To: khorenko; +Cc: devel
From: Denis V. Lunev <den@openvz.org>
The balloon holds file system space that never carries data, but its
blocks are allocated, so ext4_trim_all_free() never discards them, and
inflating the balloon is a plain fallocate() that does not discard
either. Once the balloon is inflated over blocks that held data, e.g.
on an online shrink after files were removed, the device below keeps
them allocated: a ploop or qcow2 image keeps host disk space the
container can never use, and compaction, which is FITRIM, does not give
it back.
Walk the balloon after the free space in ext4_trim_fs() and discard its
unwritten extents that fall into the requested range. The walk goes to
the end of the extent tree, as fallocate() with FALLOC_FL_KEEP_SIZE can
leave extents past i_size. Physically adjacent extents are merged into
one discard, otherwise every cut at the unwritten extent length limit
leaves a partly covered cluster of the image below. A merged run is
discarded in chunks, with a check for fatal signals and freezing
between them.
The walk holds i_rwsem and the invalidate lock exclusively, after
waiting for direct I/O and flushing the page cache of the balloon, as
punch hole does. No block of the balloon can be freed and reused under
the discard, and nothing is written to it meanwhile, so an extent that
is still unwritten reads as zeroes before and after the discard. The
balloon is hidden from lookup and opened only through
EXT4_IOC_OPEN_BALLOON, which takes CAP_SYS_ADMIN, so the lock only
waits for ploop inflating or shrinking it.
Every FITRIM discards the balloon again, which costs little: dm-ploop
completes the discard of an unmapped cluster at once. The balloon is
not counted in the trimmed length reported to userspace, and a discard
error fails FITRIM.
A balloon inflated before this change is released by the next FITRIM.
https://virtuozzo.atlassian.net/browse/VSTOR-153590
Feature: fs/ext4: fast online shrink support
Signed-off-by: Denis V. Lunev <den@openvz.org>
---
fs/ext4/mballoc.c | 126 ++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 126 insertions(+)
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index aaf2c24c514c..5fb87753975d 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -7009,6 +7009,126 @@ ext4_trim_all_free(struct super_block *sb, ext4_group_t group,
return ret;
}
+#define EXT4_BALLOON_DISCARD_CHUNK (128U << 20)
+
+static int ext4_trim_balloon_run(struct super_block *sb, ext4_fsblk_t pblk,
+ ext4_fsblk_t len, ext4_fsblk_t start,
+ ext4_fsblk_t end, ext4_fsblk_t minblocks)
+{
+ unsigned int gran = max_t(unsigned int, 1,
+ bdev_discard_granularity(sb->s_bdev) >> sb->s_blocksize_bits);
+ unsigned int chunk = roundup(EXT4_BALLOON_DISCARD_CHUNK >>
+ sb->s_blocksize_bits, gran);
+ ext4_fsblk_t last = pblk + len - 1, next;
+ int ret;
+
+ if (!len)
+ return 0;
+ if (pblk < start)
+ pblk = start;
+ if (last > end)
+ last = end;
+ if (last < pblk || last - pblk + 1 < minblocks)
+ return 0;
+
+ /*
+ * Chunks keep a long run interruptible. They end on multiples of the
+ * discard granularity, so no cluster below is split between two.
+ */
+ while (pblk <= last) {
+ u64 rem = pblk;
+
+ next = pblk + chunk - do_div(rem, chunk);
+ if (next > last + 1)
+ next = last + 1;
+ trace_ext4_discard_blocks(sb, pblk, next - pblk);
+ ret = sb_issue_discard(sb, pblk, next - pblk, GFP_NOFS, 0);
+ if (ret)
+ return ret;
+ pblk = next;
+ if (ext4_trim_interrupted())
+ break;
+ cond_resched();
+ }
+
+ return 0;
+}
+
+/*
+ * Blocks of the balloon never hold data, but they are allocated, so
+ * ext4_trim_all_free() never discards them. Discard its unwritten
+ * extents, merging physically adjacent ones, so that the device below
+ * can release the space as it does for free blocks.
+ */
+static int ext4_trim_balloon(struct super_block *sb, ext4_fsblk_t start,
+ ext4_fsblk_t end, ext4_fsblk_t minblocks)
+{
+ struct ext4_sb_info *sbi = EXT4_SB(sb);
+ ext4_fsblk_t run_pblk = 0, run_len = 0;
+ struct ext4_map_blocks map;
+ struct inode *inode = NULL;
+ ext4_lblk_t lblk;
+ bool unwritten;
+ int ret, err;
+
+ /* ext4_load_balloon() replaces s_balloon_ino under s_umount */
+ down_read(&sb->s_umount);
+ if (sbi->s_balloon_ino)
+ inode = igrab(sbi->s_balloon_ino);
+ up_read(&sb->s_umount);
+ if (!inode)
+ return 0;
+
+ /*
+ * Nobody may write the balloon meanwhile: data written into an
+ * unwritten extent would be discarded with it.
+ */
+ inode_lock(inode);
+ inode_dio_wait(inode);
+ filemap_invalidate_lock(inode->i_mapping);
+ ret = filemap_write_and_wait(inode->i_mapping);
+ if (ret)
+ goto out;
+
+ for (lblk = 0; lblk < EXT_MAX_BLOCKS; lblk += map.m_len) {
+ if (ext4_trim_interrupted())
+ break;
+ cond_resched();
+
+ map.m_lblk = lblk;
+ map.m_len = EXT_MAX_BLOCKS - lblk;
+ err = ext4_map_blocks(NULL, inode, &map, 0);
+ if (err < 0) {
+ ret = err;
+ break;
+ }
+ if (!map.m_len)
+ break;
+
+ unwritten = err > 0 && (map.m_flags & EXT4_MAP_UNWRITTEN);
+ if (unwritten && run_len && run_pblk + run_len == map.m_pblk) {
+ run_len += map.m_len;
+ continue;
+ }
+
+ ret = ext4_trim_balloon_run(sb, run_pblk, run_len, start, end,
+ minblocks);
+ if (ret)
+ break;
+ run_pblk = map.m_pblk;
+ run_len = unwritten ? map.m_len : 0;
+ }
+ if (!ret && !ext4_trim_interrupted())
+ ret = ext4_trim_balloon_run(sb, run_pblk, run_len, start, end,
+ minblocks);
+out:
+ filemap_invalidate_unlock(inode->i_mapping);
+ inode_unlock(inode);
+ iput(inode);
+
+ return ret;
+}
+
/**
* ext4_trim_fs() -- trim ioctl handle function
* @sb: superblock for filesystem
@@ -7028,6 +7148,7 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
ext4_group_t group, first_group, last_group;
ext4_grpblk_t cnt = 0, first_cluster, last_cluster;
uint64_t start, end, minlen, trimmed = 0;
+ ext4_fsblk_t last_blk;
ext4_fsblk_t first_data_blk =
le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
ext4_fsblk_t max_blks = ext4_blocks_count(EXT4_SB(sb)->s_es);
@@ -7061,6 +7182,7 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
&first_group, &first_cluster);
ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) end,
&last_group, &last_cluster);
+ last_blk = end;
/* end now represents the last cluster to discard in this group */
end = EXT4_CLUSTERS_PER_GROUP(sb) - 1;
@@ -7103,6 +7225,10 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
first_cluster = 0;
}
+ if (!ret && !ext4_trim_interrupted())
+ ret = ext4_trim_balloon(sb, start, last_blk,
+ EXT4_C2B(EXT4_SB(sb), minlen));
+
if (!ret)
EXT4_SB(sb)->s_last_trim_minblks = minlen;
--
2.53.0
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
^ permalink raw reply [flat|nested] 7+ messages in thread* [Devel] [PATCH hci-8.1 v2 3/3] xfs: discard the balloon on FITRIM
2026-11-10 22:27 [Devel] [PATCH hci-8.1 v2 0/3] ext4, xfs: discard the balloon on FITRIM Denis V. Lunev
2026-11-10 22:27 ` [Devel] [PATCH hci-8.1 v2 1/3] xfs: release the balloon inode when xfs_balloon_get() refuses it Denis V. Lunev
2026-11-10 22:27 ` [Devel] [PATCH hci-8.1 v2 2/3] ext4: discard the balloon on FITRIM Denis V. Lunev
@ 2026-11-10 22:27 ` Denis V. Lunev
2026-11-10 22:36 ` no-reply
2 siblings, 1 reply; 7+ messages in thread
From: Denis V. Lunev @ 2026-11-10 22:27 UTC (permalink / raw)
To: khorenko; +Cc: devel
From: Denis V. Lunev <den@openvz.org>
The balloon holds file system space that never carries data, but its
blocks are allocated, so the free space walk of xfs_ioc_trim() never
discards them, and inflating the balloon is a plain fallocate() that
does not discard either. Once the balloon is inflated over blocks that
held data, e.g. on an online shrink after files were removed, the
device below keeps them allocated: a ploop or qcow2 image keeps host
disk space the container can never use, and compaction, which is
FITRIM, does not give it back.
Walk the balloon after the free space of the data device and discard
its unwritten extents that fall into the requested range, merging
physically adjacent ones into one discard. The walk goes to the end of
the file offset space, as fallocate() with FALLOC_FL_KEEP_SIZE can
leave extents past i_size. A merged run is discarded in chunks, with a
check for fatal signals and freezing between them.
The walk holds the IOLOCK and the MMAPLOCK exclusively, after waiting
for direct I/O and flushing the page cache of the balloon. No block of
the balloon can be freed and reused under the discard, and nothing is
written to it meanwhile, so an extent that is still unwritten reads as
zeroes before and after the discard. The balloon is hidden from lookup
and opened only through XFS_IOC_OPEN_BALLOON, which takes
CAP_SYS_ADMIN, so the locks only wait for ploop inflating or shrinking
it.
Every FITRIM discards the balloon again, which costs little: dm-ploop
completes the discard of an unmapped cluster at once. A discard error
or a failure to read the balloon inode fails FITRIM; a balloon that is
no longer a regular mutable file is skipped.
A balloon inflated before this change is released by the next FITRIM.
https://virtuozzo.atlassian.net/browse/VSTOR-153590
Feature: fs/xfs: fast online shrink support
Signed-off-by: Denis V. Lunev <den@openvz.org>
---
fs/xfs/xfs_discard.c | 144 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 144 insertions(+)
diff --git a/fs/xfs/xfs_discard.c b/fs/xfs/xfs_discard.c
index 93943e698eea..b9b5a28b0d96 100644
--- a/fs/xfs/xfs_discard.c
+++ b/fs/xfs/xfs_discard.c
@@ -22,6 +22,10 @@
#include "xfs_health.h"
#include "xfs_rtbitmap.h"
#include "xfs_rtgroup.h"
+#include "xfs_bit.h"
+#include "xfs_inode.h"
+#include "xfs_bmap.h"
+#include "xfs_super.h"
/*
* Notes on an efficient, low latency fstrim algorithm
@@ -833,6 +837,140 @@ xfs_trim_rtdev_extents(
# define xfs_trim_rtdev_extents(...) (-EOPNOTSUPP)
#endif /* CONFIG_XFS_RT */
+#define XFS_BALLOON_DISCARD_CHUNK_BB ((128U << 20) >> BBSHIFT)
+
+static int
+xfs_trim_balloon_run(
+ struct xfs_mount *mp,
+ xfs_daddr_t daddr,
+ uint64_t len,
+ xfs_daddr_t start,
+ xfs_daddr_t end,
+ xfs_extlen_t minlen)
+{
+ struct block_device *bdev = mp->m_ddev_targp->bt_bdev;
+ unsigned int gran = max(1U,
+ bdev_discard_granularity(bdev) >> BBSHIFT);
+ unsigned int chunk = roundup(XFS_BALLOON_DISCARD_CHUNK_BB, gran);
+ xfs_daddr_t last = daddr + len - 1, next;
+ int error;
+
+ if (!len)
+ return 0;
+ if (daddr < start)
+ daddr = start;
+ if (last > end)
+ last = end;
+ if (last < daddr || last - daddr + 1 < XFS_FSB_TO_BB(mp, minlen))
+ return 0;
+
+ /*
+ * Chunks keep a long run interruptible. They end on multiples of the
+ * discard granularity, so no cluster below is split between two.
+ */
+ while (daddr <= last) {
+ u64 rem = daddr;
+
+ next = daddr + chunk - do_div(rem, chunk);
+ if (next > last + 1)
+ next = last + 1;
+ error = blkdev_issue_discard(bdev, daddr, next - daddr,
+ GFP_KERNEL);
+ if (error)
+ return error;
+ daddr = next;
+ if (xfs_trim_should_stop())
+ break;
+ cond_resched();
+ }
+
+ return 0;
+}
+
+/*
+ * Blocks of the balloon never hold data, but they are allocated, so the
+ * free space btree walk never discards them. Discard its unwritten
+ * extents, merging physically adjacent ones, so that the device below
+ * can release the space as it does for free blocks.
+ */
+static int
+xfs_trim_balloon(
+ struct xfs_mount *mp,
+ xfs_daddr_t start,
+ xfs_daddr_t end,
+ xfs_extlen_t minlen)
+{
+ u64 balloon_ino = READ_ONCE(mp->m_balloon_ino);
+ xfs_daddr_t run_daddr = 0;
+ uint64_t run_len = 0;
+ struct xfs_bmbt_irec map;
+ struct xfs_inode *ip;
+ xfs_fileoff_t off = 0;
+ unsigned int lock_mode;
+ bool unwritten;
+ int nmaps, error;
+
+ if (!balloon_ino)
+ return 0;
+ ip = xfs_balloon_get(mp, balloon_ino, 0);
+ if (IS_ERR(ip))
+ return PTR_ERR(ip) == -EINVAL ? 0 : PTR_ERR(ip);
+ error = 0;
+ if (XFS_IS_REALTIME_INODE(ip))
+ goto out_rele;
+
+ /*
+ * Nobody may write the balloon meanwhile: data written into an
+ * unwritten extent would be discarded with it.
+ */
+ xfs_ilock(ip, XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL);
+ inode_dio_wait(VFS_I(ip));
+ error = filemap_write_and_wait(VFS_I(ip)->i_mapping);
+ if (error)
+ goto out_unlock;
+
+ while (off < XFS_MAX_FILEOFF) {
+ if (xfs_trim_should_stop())
+ break;
+ cond_resched();
+
+ nmaps = 1;
+ lock_mode = xfs_ilock_data_map_shared(ip);
+ error = xfs_bmapi_read(ip, off, XFS_MAX_FILEOFF - off, &map,
+ &nmaps, 0);
+ xfs_iunlock(ip, lock_mode);
+ if (error || !nmaps || !map.br_blockcount)
+ break;
+ off = map.br_startoff + map.br_blockcount;
+
+ unwritten = xfs_bmap_is_real_extent(&map) &&
+ map.br_state == XFS_EXT_UNWRITTEN;
+ if (unwritten && run_len && run_daddr + run_len ==
+ XFS_FSB_TO_DADDR(mp, map.br_startblock)) {
+ run_len += XFS_FSB_TO_BB(mp, map.br_blockcount);
+ continue;
+ }
+
+ error = xfs_trim_balloon_run(mp, run_daddr, run_len, start,
+ end, minlen);
+ if (error)
+ break;
+ run_len = 0;
+ if (unwritten) {
+ run_daddr = XFS_FSB_TO_DADDR(mp, map.br_startblock);
+ run_len = XFS_FSB_TO_BB(mp, map.br_blockcount);
+ }
+ }
+ if (!error && !xfs_trim_should_stop())
+ error = xfs_trim_balloon_run(mp, run_daddr, run_len, start,
+ end, minlen);
+out_unlock:
+ xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL);
+out_rele:
+ xfs_irele(ip);
+ return error;
+}
+
/*
* trim a range of the filesystem.
*
@@ -905,6 +1043,12 @@ xfs_ioc_trim(
error = xfs_trim_datadev_extents(mp, start, end, minlen);
if (error)
last_error = error;
+
+ if (!xfs_trim_should_stop()) {
+ error = xfs_trim_balloon(mp, start, end, minlen);
+ if (error)
+ last_error = error;
+ }
}
if (rt_bdev && !xfs_trim_should_stop()) {
--
2.53.0
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
^ permalink raw reply [flat|nested] 7+ messages in thread