Skip to contentSkip to conversation

f2fs: introduce metadata cache

PATCH In review Chao Yu 3 participants 30 messages Aug 20 – Aug 28

Subsystems (2)
lore
Share & tools
lore

Chao Yu

author maintainer
#0 lore raw

cover f2fs: introduce metadata cache

This patchset introduces a self-managed metadata block cache in f2fs, decoupling meta blocks, node blocks, and compressed data blocks from the Linux VFS page cache and fake internal inodes.

=== 1. Background & Motivation ===

Currently, F2FS uses fake VFS inodes (meta_inode, node_inode, and compress_inode) to manage internal block caching through the VFS page cache. Because of this implementation, the f2fs block size was historically coupled to the kernel page size.

We now want to unbind block size from page size to support configurations where block size <= PAGE_SIZE (e.g., mounting a 4KB-block F2FS image on a 16KB or 64KB page system).

One possible approach is to continue using the VFS page cache to store metadata blocks. However, doing so introduces three major architectural issues (illustrated by a 4KB block on a 16KB page system):

1. Memory Overhead:
Metadata access patterns are typically random and sparse. Caching a single 4KB metadata block inside a page cache folio forces the allocation of an entire 16KB folio, resulting in 4x memory waste.

2. Folio and Sub-block Conversion Complexity:
Using larger folios requires tracking individual sub-block dirty/uptodate states within each folio and performing index-to-offset conversions across function boundaries. Because core metadata structures (e.g., f2fs_checkpoint, f2fs_sit_block, f2fs_nat_block, f2fs_summary_block, f2fs_node) are accessed extensively throughout the filesystem, this sub-block management and offset calculation complexity would spread across the entire F2FS codebase.

3. Lock Contention:
Multiple independent node blocks (e.g., dnode blocks belonging to different files) can reside within the same folio. Concurrent fsync() calls on unrelated files would contend on the same folio_lock(), serializing metadata updates and degrading multi-threaded performance.

Decoupling metadata caching from PAGE_SIZE by allocating exact block-sized cache entries is the critical first step toward supporting 4KB-block F2FS images on 16KB/64KB page systems.

=== 2. Metadata Cache Architecture & Design ===

This patchset introduces a dedicated, block-size-aligned caching

3. Lock Contention:
Multiple independent node blocks (e.g., dnode blocks belonging to different files) can reside within the same folio. Concurrent fsync() calls on unrelated files would contend on the same folio_lock(), serializing metadata updates and degrading multi-threaded performance.

3. Lock Contention:
Multiple independent node blocks (e.g., dnode blocks belonging to different files) can reside within the same folio. Concurrent fsync() calls on unrelated files would contend on the same folio_lock(), serializing metadata updates and degrading multi-threaded performance.

Decoupling metadata caching from PAGE_SIZE by allocating exact block-sized cache entries is the critical first step toward supporting 4KB-block F2FS images on 16KB/64KB page systems.

=== 2. Metadata Cache Architecture & Design ===

This patchset introduces a dedicated, block-size-aligned caching infrastructure with the following key components:

- Block-Size Aligned Allocation:
Allocates memory buffers matching exactly the filesystem block size (4KB or 16KB) via kzalloc(), fully independent of the host architecture's PAGE_SIZE.

- Radix Tree Indexing with Fast Tag Scanning:
Each cache instance (META_CACHE, NODE_CACHE, COMPRESS_CACHE) indexes cached blocks via a radix tree (keyed by Physical Block Address for meta/ compress cache, and Node ID for node cache). Radix tree tags (F2FS_CACHE_TAG_DIRTY, F2FS_CACHE_TAG_WRITEBACK) provide O(1) batch gang lookups for flushing and writeback without dual-list shuffling.

- Lightweight Bit-Locking:
Individual entries use atomic bit locks (F2FS_BLOCK_LOCKED via wait_on_bit_lock() / clear_and_wake_up_bit()) rather than heavyweight embedded mutexes/semaphores, minimizing memory footprint per entry.

- Direct BIO Read/Write & BIO Merging:
Decouples metadata/node I/O from VFS address spaces by submitting direct BIOs (f2fs_submit_cache_read / f2fs_submit_cache_write) with chained adjacent vector merging (entry->next_entry) and dedicated completion handlers.

- Memory Reclamation Shrinker:
Integrates with the kernel shrinker subsystem via a 3-phase isolation algorithm (isolate unreferenced clean entries -> truncate from radix tree under lock -> splice un-reclaimed entries back to LRU) to safely reclaim clean cached blocks under system memory pressure.

- Background Writeback Kthread & Checkpoint Integration:
Provides a dedicated background kthread (f2fs_writeback-X:Y) for periodic dirty cache flushing, combined with synchronous flushing during checkpoint commit.

- Fault Injection, Tracepoints & Debugfs Observability:
Integrates FAULT_KALLOC fault injection, tracepoints for cache state transitions and batch writeback, and per-cache memory breakdowns in debugfs.

=== 3. Patchset Organization ===

- Patch 01: Implement the core metadata cache infrastructure & direct BIO I/O.
- Patch 02: Initialize and teardown META_CACHE in sb_info.
- Patch 03: Integrate metadata cache into the memory shrinker subsystem.
- Patch 04: Introduce the background writeback kernel thread.
- Patch 05: Migrate metadata block caching (SIT, NAT, SSA, CP, recovery, GC) from meta_inode to META_CACHE.
- Patch 06: Initialize and teardown NODE_CACHE in sb_info.
- Patch 07: Migrate node and inode block caching from node_inode to NODE_CACHE.
- Patch 08: Initialize and teardown COMPRESS_CACHE in sb_info.
- Patch 09: Migrate compressed cluster caching from compress_inode to COMPRESS_CACHE.
- Patch 10: Add fault injection support for cache allocation paths.
- Patch 11: Introduce ftrace tracepoints for cache dirty and writeback events.
- Patch 12: Expose per-cache memory usage in debugfs.

Chao Yu (12):
f2fs: cache: implement metadata cache
f2fs: cache: initialize meta cache
f2fs: cache: introduce shrinker
f2fs: cache: introduce writeback thread
f2fs: cache: use meta cache
f2fs: cache: initialize node cache
f2fs: cache: use node cache
f2fs: cache: initialize compress cache
f2fs: cache: use compress cache
f2fs: cache: support fault injection
f2fs: cache: introduce tracepoints
f2fs: cache: show per-cache usage in debugfs

 fs/f2fs/Makefile            |   2 +-
 fs/f2fs/acl.c               |  26 +-
 fs/f2fs/acl.h               |   8 +-
 fs/f2fs/cache.c             | 690 +++++++++++++++++++++++++
 fs/f2fs/cache.h             | 224 ++++++++
 fs/f2fs/checkpoint.c        | 404 +++++++--------
 fs/f2fs/compress.c          | 171 +++----
 fs/f2fs/data.c              | 527 +++++++++++++------
 fs/f2fs/debug.c             |  70 ++-
 fs/f2fs/dir.c               | 168 +++---
 fs/f2fs/extent_cache.c      |  14 +-
 fs/f2fs/f2fs.h              | 333 ++++++------
 fs/f2fs/file.c              |  78 ++-
 fs/f2fs/gc.c                | 178 ++++---
 fs/f2fs/inline.c            | 284 ++++++-----
 fs/f2fs/inode.c             | 205 +++-----
 fs/f2fs/iostat.h            |  11 +
 fs/f2fs/namei.c             | 118 ++---
 fs/f2fs/node.c              | 994 +++++++++++++++++-------------------
 fs/f2fs/node.h              | 109 ++--
 fs/f2fs/recovery.c          | 253 ++++-----
 fs/f2fs/segment.c           | 263 +++++-----
 fs/f2fs/segment.h           |  39 +-
 fs/f2fs/shrinker.c          |  13 +
 fs/f2fs/super.c             | 143 +++---
 fs/f2fs/xattr.c             | 123 +++--
 fs/f2fs/xattr.h             |  12 +-
 include/linux/f2fs_fs.h     |   3 -
 include/trace/events/f2fs.h |  71 +++
 29 files changed, 3331 insertions(+), 2203 deletions(-)
 create mode 100644 fs/f2fs/cache.c
 create mode 100644 fs/f2fs/cache.h
Signature · 1 line
2.49.0
author maintainer
#1 lore raw

PATCH 1/12 f2fs: cache: implement metadata cache

This patch introduces the core metadata block caching infrastructure to manage f2fs metadata independently of the page cache.

It implements:
- core cache APIs: get, create, put, drop, backed by a radix tree and a single global LRU list.
- support multiple status of cached block: LOCKED, UPTODATE, DIRTY, WRITEBACK, INLINE.
- internal bio based read/write helpers with adjacent block vector merging.

---
 fs/f2fs/Makefile |   2 +-
 fs/f2fs/cache.c  | 531 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h  | 179 ++++++++++++++++
 fs/f2fs/data.c   | 283 ++++++++++++++++++++++---
 fs/f2fs/f2fs.h   |  24 +++
 fs/f2fs/iostat.h |  11 +
 6 files changed, 1002 insertions(+), 28 deletions(-)
 create mode 100644 fs/f2fs/cache.c
 create mode 100644 fs/f2fs/cache.h
fs/f2fs/Makefile +1 -1
index 8a7322d229e4..fbf49c30b066 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_F2FS_FS) += f2fs.o
 
 f2fs-y		:= dir.o file.o inode.o namei.o hash.o super.o inline.o
 f2fs-y		+= checkpoint.o gc.o data.o node.o segment.o recovery.o
-f2fs-y		+= shrinker.o extent_cache.o sysfs.o
+f2fs-y		+= shrinker.o extent_cache.o sysfs.o cache.o
 f2fs-$(CONFIG_F2FS_STAT_FS) += debug.o
 f2fs-$(CONFIG_F2FS_FS_XATTR) += xattr.o
 f2fs-$(CONFIG_F2FS_FS_POSIX_ACL) += acl.o
fs/f2fs/cache.c new file +531 -0

531 lines not on this page · raw message

fs/f2fs/cache.h new file +179 -0

179 lines not on this page · raw message

fs/f2fs/data.c +256 -27

354 lines not on this page · raw message

fs/f2fs/f2fs.h +24 -0
index 1b96d8718c5c..8413983ea9d5 100644
@@ -221,6 +221,8 @@ struct f2fs_rwsem {
 #endif
 };
 
+#include "cache.h"
+
 struct f2fs_mount_info {
 	unsigned long long opt;
 	block_t root_reserved_blocks;	/* root reserved blocks */
@@ -1370,8 +1372,10 @@ struct f2fs_io_info {
 	unsigned int is_por:1;		/* indicate IO is from recovery or not */
 	unsigned int encrypted:1;	/* indicate file is encrypted */
 	unsigned int meta_gc:1;		/* require meta inode GC */
+	unsigned int is_cache:1;	/* indicate IO is from internal cache */
 	enum iostat_type io_type;	/* io type */
 	struct writeback_control *io_wbc; /* writeback control */
+	struct f2fs_cached_block *cache_entry;
 	struct bio **bio;		/* bio for ipu */
 	sector_t *last_block;		/* last block number in bio */
 };
@@ -1781,6 +1785,12 @@ struct f2fs_gc_kthread {
 	unsigned int boost_gc_greedy;
 };
 
+struct f2fs_bio {
+	struct work_struct work;
+	struct f2fs_cached_block *entry;
+	struct bio bio;
+};
+
 struct f2fs_sb_info {
 	struct super_block *sb;			/* pointer to VFS super block */
 	struct proc_dir_entry *s_proc;		/* proc entry */
@@ -2323,6 +2333,16 @@ static inline bool is_node_folio(struct folio *folio)
 	return folio->mapping == NODE_MAPPING(F2FS_F_SB(folio));
 }
 
+static inline struct f2fs_bio *F2FS_BIO(struct bio *bio)
+{
+	return container_of(bio, struct f2fs_bio, bio);
+}
+
+static inline bool f2fs_is_cache_bio(struct bio *bio)
+{
+	return F2FS_BIO(bio)->entry != NULL;
+}
+
 static inline bool is_sbi_flag_set(struct f2fs_sb_info *sbi, unsigned int type)
 {
 	return test_bit(type, &sbi->s_flag);
@@ -4237,6 +4257,8 @@ void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
 				nid_t ino, enum page_type type);
 void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
 				struct folio *folio, enum page_type type);
+bool f2fs_submit_merged_write_cache(struct f2fs_cached_block *entry,
+				enum page_type type);
 void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
 					struct bio **bio, struct folio *folio);
 void f2fs_submit_all_merged_ipu_writes(struct f2fs_sb_info *sbi);
@@ -4244,6 +4266,8 @@ void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi);
 int f2fs_submit_page_bio(struct f2fs_io_info *fio);
 int f2fs_merge_page_bio(struct f2fs_io_info *fio);
 void f2fs_submit_page_write(struct f2fs_io_info *fio);
+int f2fs_submit_cache_read(struct f2fs_io_info *fio);
+void f2fs_submit_cache_write(struct f2fs_io_info *fio);
 struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
 		block_t blk_addr, sector_t *sector);
 int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr);
fs/f2fs/iostat.h +11 -0
index 2025225b5bed..61c6bc8e3119 100644
@@ -60,6 +60,13 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 	return iostat_ctx->post_read_ctx;
 }
 
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
+
+	iostat_ctx->post_read_ctx = ctx;
+}
+
 extern void iostat_update_and_unbind_ctx(struct bio *bio);
 extern void iostat_alloc_and_bind_ctx(struct f2fs_sb_info *sbi,
 		struct bio *bio, struct bio_post_read_ctx *ctx);
@@ -81,6 +88,10 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 {
 	return bio->bi_private;
 }
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	bio->bi_private = ctx;
+}
 static inline int f2fs_init_iostat_processing(void) { return 0; }
 static inline void f2fs_destroy_iostat_processing(void) {}
 static inline int f2fs_init_iostat(struct f2fs_sb_info *sbi) { return 0; }
Signature · 1 line
2.49.0
maintainer
#2 lore raw

Re: [f2fs-dev] [PATCH v1 01/12] f2fs: cache: implement metadata cache

Quoted 41 lines · On 08/20, Chao Yu via Linux-f2fs-devel wrote:

This patch introduces the core metadata block caching infrastructure to manage f2fs metadata independently of the page cache.

It implements:
- core cache APIs: get, create, put, drop, backed by a radix tree and a single global LRU list.
- support multiple status of cached block: LOCKED, UPTODATE, DIRTY, WRITEBACK, INLINE.
- internal bio based read/write helpers with adjacent block vector merging.

---
 fs/f2fs/Makefile |   2 +-
 fs/f2fs/cache.c  | 531 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h  | 179 ++++++++++++++++
 fs/f2fs/data.c   | 283 ++++++++++++++++++++++---
 fs/f2fs/f2fs.h   |  24 +++
 fs/f2fs/iostat.h |  11 +
 6 files changed, 1002 insertions(+), 28 deletions(-)
 create mode 100644 fs/f2fs/cache.c
 create mode 100644 fs/f2fs/cache.h
fs/f2fs/Makefile +1 -1
index 8a7322d229e4..fbf49c30b066 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_F2FS_FS) += f2fs.o
 
 f2fs-y		:= dir.o file.o inode.o namei.o hash.o super.o inline.o
 f2fs-y		+= checkpoint.o gc.o data.o node.o segment.o recovery.o
-f2fs-y		+= shrinker.o extent_cache.o sysfs.o
+f2fs-y		+= shrinker.o extent_cache.o sysfs.o cache.o
 f2fs-$(CONFIG_F2FS_STAT_FS) += debug.o
 f2fs-$(CONFIG_F2FS_FS_XATTR) += xattr.o
 f2fs-$(CONFIG_F2FS_FS_POSIX_ACL) += acl.o
fs/f2fs/cache.c new file +0 -0
new file mode 100644
index 000000000000..08bc658166f7

fs/f2fs/cache.c line 6

Quoted 6 lines
fs/f2fs/cache.c new file +6 -0
@@ -0,0 +1,531 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2026 Google LLC
+ * Author: Chao Yu <[email protected]>
+ */
+#include "linux/spinlock.h"

#include <linux/spinlock.h>?
May need to Move below?

Quoted 33 lines
fs/f2fs/cache.c new file +33 -0
+#include <linux/fs.h>
+#include <linux/f2fs_fs.h>
+#include <linux/radix-tree.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/pagemap.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/delay.h>
+#include "f2fs.h"
+#include "cache.h"
+#include "node.h"
+#include "segment.h"
+
+void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
+					enum page_type type)
+{
+	/* in case the entry was truncated or on-going shrink */
+	if (!entry->cache)
+		return;
+
+	if (!f2fs_cache_test_writeback(entry))
+		return;
+
+	/* submit cached bio */
+	f2fs_submit_merged_write_cache(entry, type);
+
+	wait_on_bit_io(&entry->state, F2FS_BLOCK_WRITEBACK,
+					TASK_UNINTERRUPTIBLE);
+}
+
+void f2fs_cache_wait_writeback(struct f2fs_cached_block *entry)
+{

fs/f2fs/cache.c line 49

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+	/* in case the entry was truncated or on-going shrink */
+	if (!entry->cache)
+		return;
+
+	f2fs_cache_wait_writeback_cond(entry,
+			IS_META_CACHE(entry->cache) ? META : NODE);
+}
+
+static void f2fs_cache_update_tag(struct f2fs_cached_block *entry, unsigned int src,
+					unsigned int dst)

f2fs_cache_update_tag(entry, clear_from, set_to);

fs/f2fs/cache.c line 55

Quoted 6 lines
fs/f2fs/cache.c new file +6 -0
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	if (src)
	if (clear_fom != F2FS_CACHE_TAG_NONE)

fs/f2fs/cache.c line 57

Quoted 2 lines
fs/f2fs/cache.c new file +2 -0
+		radix_tree_tag_clear(&cache->root, entry->index, src);
+	if (dst)
	if (set_to != F2FS_CACHE_TAG_NONE)

fs/f2fs/cache.c line 67

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+		radix_tree_tag_set(&cache->root, entry->index, dst);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+}
+
+bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry)
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+
+	if (!f2fs_cache_test_uptodate(entry))
+		f2fs_cache_set_uptodate(entry);
	f2fs_cache_set_uptodate(entry);
Quoted 4 lines
fs/f2fs/cache.c new file +4 -0
+
+#ifdef CONFIG_F2FS_CHECK_FS
+	if (cache->type == F2FS_NODE_CACHE && IS_INODE(cache_folio(entry)))
+		f2fs_inode_chksum_set(cache->sbi, cache_folio(entry));

fs/f2fs/cache.c line 81

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+#endif
+
+	if (f2fs_cache_test_dirty(entry))
+		return false;
+
+	if (!f2fs_cache_test_and_set_dirty(entry)) {
+		enum count_type type = IS_META_CACHE(cache) ?
+				F2FS_DIRTY_META : F2FS_DIRTY_NODES;
+
+		f2fs_cache_update_tag(entry, 0, F2FS_CACHE_TAG_DIRTY);
		f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_NONE, F2FS_CACHE_TAG_DIRTY);
Quoted 7 lines
fs/f2fs/cache.c new file +7 -0
+		inc_page_count(cache->sbi, type);
+		return true;
+	}
+
+	return false;
+}
+

fs/f2fs/cache.c line 98

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+bool f2fs_clear_cache_dirty(struct f2fs_cached_block *entry)
+{
+	if (!f2fs_cache_test_dirty(entry))
+		return false;
+
+	f2fs_cache_clear_dirty(entry);
+	return true;
+}
+
+static void __f2fs_drop_cache_dirty(struct f2fs_cached_block *entry, bool force)

static void __drop_cache_dirty(struct f2fs_cached_block *entry, bool force)

Quoted 80 lines
fs/f2fs/cache.c new file +80 -0
+{
+
+	struct f2fs_cached_block_list *cache = entry->cache;
+	enum count_type type = IS_META_CACHE(cache) ?
+				F2FS_DIRTY_META : F2FS_DIRTY_NODES;
+
+	f2fs_cache_clear_uptodate(entry);
+
+	if (!force && !f2fs_clear_cache_dirty(entry))
+		return;
+
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_DIRTY, 0);
+	dec_page_count(cache->sbi, type);
+}
+
+void f2fs_drop_cache_dirty(struct f2fs_cached_block *entry)
+{
+	__f2fs_drop_cache_dirty(entry, false);
+}
+
+void f2fs_force_clear_cache_dirty(struct f2fs_cached_block *entry)
+{
+	__f2fs_drop_cache_dirty(entry, true);
+}
+
+void f2fs_start_cache_writeback(struct f2fs_cached_block *entry)
+{
+	f2fs_cache_set_writeback(entry);
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_DIRTY,
+				F2FS_CACHE_TAG_WRITEBACK);
+}
+
+void f2fs_end_cache_writeback(struct f2fs_cached_block *entry)
+{
+	/*
+	 * should call f2fs_cache_update_tag() before clearing writeback bit,
+	 * in case f2fs_truncate_cache() set entry->cache to NULL.
+	 */
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_WRITEBACK, 0);
+	clear_and_wake_up_bit(F2FS_BLOCK_WRITEBACK, &entry->state);
+}
+
+static int f2fs_cache_refcount(struct f2fs_cached_block *entry)
+{
+	return atomic_read(&entry->refcount);
+}
+
+static void __f2fs_free_cache(struct f2fs_cached_block *entry)
+{
+	kfree(entry->data);
+	kfree(entry);
+}
+
+static void f2fs_free_cache(struct f2fs_cached_block *entry)
+{
+	WARN_ON_ONCE(!list_empty(&entry->list));
+	WARN_ON_ONCE(f2fs_cache_refcount(entry));
+	__f2fs_free_cache(entry);
+}
+
+void f2fs_cache_get(struct f2fs_cached_block *entry)
+{
+	atomic_inc(&entry->refcount);
+}
+
+static bool f2fs_cache_put(struct f2fs_cached_block *entry)
+{
+	WARN_ON_ONCE(!f2fs_cache_refcount(entry));
+	if (atomic_dec_and_test(&entry->refcount)) {
+		f2fs_free_cache(entry);
+		return true;
+	}
+	return false;
+}
+
+static struct f2fs_cached_block *f2fs_create_cache(
+		struct f2fs_cached_block_list *cache,
+		unsigned long index, bool nofail)
+{
+	struct f2fs_cached_block *entry;

fs/f2fs/cache.c line 188

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+	unsigned int flags = GFP_NOFS;
+
+	if (nofail)
+		flags |= __GFP_NOFAIL;
+
+	entry = kzalloc_obj(*entry, flags);
+	if (!entry)
+		return ERR_PTR(-ENOMEM);
+
+	entry->data = kzalloc(cache->sbi->blocksize, flags);

We don't need kzalloc() since we have an uptodate flag.

Quoted 1,007 lines
fs/f2fs/cache.c new file +343 -0
+	if (!entry->data) {
+		kfree(entry);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	entry->index = index;
+
+	atomic_set(&entry->refcount, 0);
+	entry->next_entry = NULL;
+	INIT_LIST_HEAD(&entry->list);
+
+	entry->cache = cache;
+
+	return entry;
+}
+
+static struct f2fs_cached_block *f2fs_insert_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index,
+			struct f2fs_cached_block *new)
+{
+	struct f2fs_cached_block *e;
+	int ret;
+	unsigned long flags;
+
+	ret = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
+	f2fs_bug_on(cache->sbi, ret);
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	e = radix_tree_lookup(&cache->root, index);
+	if (!e) {
+		e = new;
+		f2fs_bug_on(cache->sbi, f2fs_cache_refcount(e));
+
+		ret = radix_tree_insert(&cache->root, index, e);
+		f2fs_bug_on(cache->sbi, ret);
+
+		/* radix tree referenced cache entry */
+		f2fs_cache_get(e);
+		f2fs_bug_on(cache->sbi, !list_empty(&e->list));
+		list_add_tail(&e->list, &cache->lru_list);
+		cache->num_entries++;
+	}
+	f2fs_cache_get(e);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+	radix_tree_preload_end();
+
+	if (new != e) {
+		f2fs_bug_on(cache->sbi, f2fs_cache_refcount(new));
+		__f2fs_free_cache(new);
+	}
+
+	return e;
+}
+
+struct f2fs_cached_block *f2fs_find_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index)
+{
+	struct f2fs_cached_block *entry;
+	unsigned long flags;
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	entry = radix_tree_lookup(&cache->root, index);
+	if (entry) {
+		f2fs_bug_on(cache->sbi, !f2fs_cache_refcount(entry));
+		f2fs_cache_get(entry);
+		if (!list_empty(&entry->list))
+			list_move_tail(&entry->list, &cache->lru_list);
+
+	} else {
+		entry = ERR_PTR(-ENOENT);
+	}
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+
+	return entry;
+}
+
+struct f2fs_cached_block *f2fs_grab_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index, int flags)
+
+{
+	struct f2fs_cached_block *entry, *new;
+	bool create = flags & F2FS_CACHE_CREATE;
+	bool nofail = flags & F2FS_CACHE_NOFAIL;
+	bool lock = flags & F2FS_CACHE_LOCK;
+
+repeat:
+	entry = f2fs_find_cache(cache, index);
+	if (!IS_ERR(entry))
+		goto found;
+
+	if (!create)
+		return ERR_PTR(-ENOENT);
+
+	new = f2fs_create_cache(cache, index, nofail);
+	if (IS_ERR(new))
+		return new;
+
+	entry = f2fs_insert_cache(cache, index, new);
+found:
+	if (lock) {
+		f2fs_lock_cache(entry);
+		/* has been truncated */
+		if (entry->cache != cache) {
+			f2fs_put_cache(entry, true);
+			goto repeat;
+		}
+	}
+	return entry;
+}
+
+bool f2fs_trylock_cache(struct f2fs_cached_block *entry)
+{
+	return !test_and_set_bit(F2FS_BLOCK_LOCKED, &entry->state);
+}
+
+void f2fs_lock_cache(struct f2fs_cached_block *entry)
+{
+	wait_on_bit_lock(&entry->state, F2FS_BLOCK_LOCKED,
+					TASK_UNINTERRUPTIBLE);
+}
+
+void f2fs_unlock_cache(struct f2fs_cached_block *entry)
+{
+	clear_and_wake_up_bit(F2FS_BLOCK_LOCKED, &entry->state);
+}
+
+bool f2fs_put_cache(struct f2fs_cached_block *entry, bool unlock)
+{
+	if (IS_ERR_OR_NULL(entry))
+		return false;
+	if (unlock)
+		f2fs_unlock_cache(entry);
+	return f2fs_cache_put(entry);
+}
+
+unsigned int f2fs_cache_gang_lookup_tag(struct f2fs_cached_block_list *cache,
+				struct f2fs_cached_block **entries,
+				pgoff_t *first_index, unsigned int max_nr,
+				int tag)
+{
+	unsigned long flags;
+	int nr, i;
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	nr = radix_tree_gang_lookup_tag(&cache->root, (void **)entries,
+				*first_index, max_nr, tag);
+	if (!nr)
+		goto out;
+
+	for (i = 0; i < nr; i++)
+		f2fs_cache_get(entries[i]);
+	*first_index = entries[nr - 1]->index + 1;
+out:
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	return nr;
+}
+
+void f2fs_cache_gang_release(struct f2fs_cached_block **entries,
+				unsigned int nr_entries)
+{
+	int i;
+
+	for (i = 0; i < nr_entries; i++)
+		f2fs_put_cache(entries[i], false);
+}
+
+void f2fs_cache_wait_on_all_writeback(struct f2fs_cached_block_list *cache)
+{
+	unsigned long index = 0;
+	struct f2fs_cached_block *entries[F2FS_ONSTACK_CACHES];
+	int nr, i;
+
+next:
+	nr = f2fs_cache_gang_lookup_tag(cache, entries, &index,
+			F2FS_ONSTACK_CACHES, F2FS_CACHE_TAG_WRITEBACK);
+	if (!nr)
+		return;
+
+	for (i = 0; i < nr; i++)
+		f2fs_cache_wait_writeback(entries[i]);
+	f2fs_cache_gang_release(entries, nr);
+	goto next;
+}
+
+static void f2fs_do_truncate_cache(struct f2fs_cached_block *entry,
+						bool drop_dirty)
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+	unsigned long flags;
+
+	if (!drop_dirty &&
+		(f2fs_cache_test_dirty(entry) ||
+		f2fs_cache_test_writeback(entry)))
+		return;
+
+	f2fs_cache_wait_writeback(entry);
+	f2fs_drop_cache_dirty(entry);
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+
+	f2fs_bug_on(cache->sbi, !entry->cache);
+	if (!radix_tree_delete(&cache->root, entry->index))
+		f2fs_bug_on(cache->sbi, !entry->cache);
+
+	entry->cache = NULL;
+	cache->num_entries--;
+
+	atomic_dec(&entry->refcount);
+	f2fs_bug_on(cache->sbi, !f2fs_cache_refcount(entry));
+
+	f2fs_bug_on(cache->sbi, list_empty(&entry->list));
+	list_del_init(&entry->list);
+
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+}
+
+static void f2fs_truncate_cache(struct f2fs_cached_block *entry,
+					bool drop_dirty)
+{
+	f2fs_lock_cache(entry);
+	if (entry->cache)
+		f2fs_do_truncate_cache(entry, drop_dirty);
+	f2fs_unlock_cache(entry);
+}
+
+static void f2fs_drop_cache(struct f2fs_cached_block_list *cache,
+			block_t blkaddr, bool drop_dirty)
+{
+	struct f2fs_cached_block *entry;
+
+	entry = f2fs_find_cache(cache, blkaddr);
+	if (IS_ERR(entry))
+		return;
+
+	f2fs_truncate_cache(entry, drop_dirty);
+	f2fs_put_cache(entry, false);
+}
+
+void f2fs_drop_cache_range(struct f2fs_cached_block_list *cache,
+		unsigned long start, unsigned long len, bool drop_dirty)
+{
+	unsigned long index = start;
+	unsigned long end = (ULONG_MAX - start < len) ?
+				ULONG_MAX : (start + len);
+	struct f2fs_cached_block *entries[F2FS_ONSTACK_CACHES];
+	unsigned long flags;
+	int nr, i;
+
+	if (len == 1)
+		return f2fs_drop_cache(cache, index, drop_dirty);
+
+next:
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	nr = radix_tree_gang_lookup(&cache->root, (void **)entries, index,
+		min((unsigned long)F2FS_ONSTACK_CACHES, end - index));
+	if (!nr)
+		goto out_unlock;
+
+	for (i = 0; i < nr; i++) {
+		struct f2fs_cached_block *entry = entries[i];
+
+		if (entry->index >= end) {
+			nr = i;
+			break;
+		}
+		f2fs_cache_get(entry);
+	}
+out_unlock:
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	if (!nr)
+		return;
+
+	for (i = 0; i < nr; i++) {
+		struct f2fs_cached_block *entry = entries[i];
+
+		index = entry->index + 1;
+
+		f2fs_truncate_cache(entry, drop_dirty);
+	}
+	f2fs_cache_gang_release(entries, nr);
+
+	if (index < end)
+		goto next;
+}
+
+int f2fs_init_cache(struct f2fs_sb_info *sbi,
+		struct f2fs_cached_block_list *cache,
+		enum f2fs_cache_type type)
+{
+	cache->sbi = sbi;
+	cache->type = type;
+	INIT_RADIX_TREE(&cache->root, GFP_ATOMIC);
+	spin_lock_init(&cache->tree_lock);
+	spin_lock_init(&cache->list_lock);
+	INIT_LIST_HEAD(&cache->lru_list);
+	cache->num_entries = 0;
+
+	return 0;
+}
+
+void f2fs_destroy_cache(struct f2fs_cached_block_list *cache)
+{
+	struct list_head *head = &cache->lru_list;
+	struct f2fs_cached_block *entry;
+	unsigned long flags;
+
+	f2fs_cache_wait_on_all_writeback(cache);
+next:
+	spin_lock(&cache->list_lock);
+	if (list_empty(head)) {
+		spin_unlock(&cache->list_lock);
+		return;
+	}
+	entry = list_first_entry(head, struct f2fs_cached_block, list);
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	radix_tree_delete(&cache->root, entry->index);
+	cache->num_entries--;
+	list_del_init(&entry->list);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+
+	spin_unlock(&cache->list_lock);
+
+	/* wait on read cache IO */
+	f2fs_lock_cache(entry);
+	/* wait on write cache IO */
+	f2fs_cache_wait_writeback(entry);
+	f2fs_bug_on(cache->sbi, f2fs_cache_test_dirty(entry));
+	f2fs_bug_on(cache->sbi, f2fs_cache_test_writeback(entry));
+	f2fs_bug_on(cache->sbi, !list_empty(&entry->list));
+	f2fs_bug_on(cache->sbi, f2fs_cache_refcount(entry) != 1);
+	f2fs_put_cache(entry, true);
+	goto next;
+}
fs/f2fs/cache.h new file +179 -0
new file mode 100644
index 000000000000..686a974008ae
@@ -0,0 +1,179 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2026 Google LLC
+ * Author: Chao Yu <[email protected]>
+ */
+#ifndef _LINUX_F2FS_CACHE_H
+#define _LINUX_F2FS_CACHE_H
+
+#include <linux/pagemap.h>
+#include <linux/mm.h>
+#include <linux/list.h>
+#include <linux/radix-tree.h>
+#include <linux/spinlock.h>
+#include <linux/wait.h>
+#include <linux/types.h>
+
+struct f2fs_rwsem;
+struct f2fs_io_info;
+enum page_type;
+
+/* Represents a single cached block (meta, node or compress) */
+struct f2fs_cached_block {
+	struct list_head list;		/* LRU list head */
+	struct f2fs_cached_block_list *cache; /* parent cache list */
+	union {
+		/* chain for merged BIO */
+		struct f2fs_cached_block *next_entry;
+		nid_t ino;		/* inode number for compress cache */
+	};
+	unsigned long index;		/* key in radix tree, (meta/compress: pba, node: nid) */
+	unsigned long state;		/* cache entry state (e.g., Dirty, UpToDate) */
+	void *data;			/* blocksize-aligned memory (4KB or 16KB) */
+	atomic_t refcount;		/* reference count */
+};
+
+struct f2fs_sb_info;
+
+enum f2fs_cache_type {
+	F2FS_META_CACHE,
+	F2FS_NODE_CACHE,
+};
+
+/* Main cache control structure (per sb_info) */
+struct f2fs_cached_block_list {
+	struct f2fs_sb_info *sbi;	/* Pointer to f2fs_sb_info */
+	struct radix_tree_root root;	/* Radix tree for cache lookup */
+	spinlock_t tree_lock;		/* Lock for radix tree */
+	struct list_head lru_list;	/* Single global LRU list */
+	spinlock_t list_lock;		/* Lock for LRU list */
+	enum f2fs_cache_type type;	/* Cache type (Node or Meta) */
+	unsigned long num_entries;	/* Current number of entries */
+};
+
+#define IS_META_CACHE(cache) (cache->type == F2FS_META_CACHE)
+
+/* Flags for f2fs_cached_block state */
+enum f2fs_cached_state {
+	F2FS_BLOCK_LOCKED,		/* cache entry is locked */
+	F2FS_BLOCK_UPTODATE,		/* cache data is valid */
+	F2FS_BLOCK_DIRTY,		/* cache data is dirty, need to writeback the data */
+	F2FS_BLOCK_WRITEBACK,		/* cache data is writeback state */
+	F2FS_BLOCK_INLINE_DATA,		/* indicate inline data */
+};
+
+enum {
+	__F2FS_CACHE_CREATE,		/* create the cache if there is no cache entry */
+	__F2FS_CACHE_LOCK,		/* get and lock the cache entry */
+	__F2FS_CACHE_NOFAIL,		/* do not allow failure */
+};
+
+enum f2fs_cache_request_flag {
+	F2FS_CACHE_CREATE	= 1 << __F2FS_CACHE_CREATE,
+	F2FS_CACHE_LOCK		= 1 << __F2FS_CACHE_LOCK,
+	F2FS_CACHE_NOFAIL	= 1 << __F2FS_CACHE_NOFAIL,
+};
+
+#define F2FS_CACHE_LOCK_CREATE	(F2FS_CACHE_LOCK | F2FS_CACHE_CREATE)
+
+#define F2FS_ONSTACK_CACHES		(32)
+
+#define F2FS_CACHE_FLAG_TEST_FUNC(name, flagname)			\
+static inline bool f2fs_cache_test_##name(				\
+			const struct f2fs_cached_block *entry)		\
+{									\
+	return test_bit(F2FS_BLOCK_##flagname, &entry->state);		\
+}									\
+
+#define F2FS_CACHE_FLAG_SET_FUNC(name, flagname)			\
+static inline void f2fs_cache_set_##name(				\
+			struct f2fs_cached_block *entry)		\
+{									\
+	set_bit(F2FS_BLOCK_##flagname, &entry->state);			\
+}									\
+
+#define F2FS_CACHE_FLAG_CLEAR_FUNC(name, flagname)			\
+static inline void f2fs_cache_clear_##name(				\
+			struct f2fs_cached_block *entry)		\
+{									\
+	clear_bit(F2FS_BLOCK_##flagname, &entry->state);		\
+}									\
+
+#define F2FS_CACHE_FLAG_TEST_AND_SET_FUNC(name, flagname)		\
+static inline bool f2fs_cache_test_and_set_##name(			\
+			struct f2fs_cached_block *entry)		\
+{									\
+	return test_and_set_bit(F2FS_BLOCK_##flagname, &entry->state);	\
+}									\
+
+F2FS_CACHE_FLAG_TEST_FUNC(locked, LOCKED);
+F2FS_CACHE_FLAG_SET_FUNC(locked, LOCKED);
+F2FS_CACHE_FLAG_CLEAR_FUNC(locked, LOCKED);
+
+F2FS_CACHE_FLAG_TEST_FUNC(uptodate, UPTODATE);
+F2FS_CACHE_FLAG_SET_FUNC(uptodate, UPTODATE);
+F2FS_CACHE_FLAG_CLEAR_FUNC(uptodate, UPTODATE);
+
+F2FS_CACHE_FLAG_TEST_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_SET_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_CLEAR_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_TEST_AND_SET_FUNC(dirty, DIRTY);
+
+F2FS_CACHE_FLAG_TEST_FUNC(writeback, WRITEBACK);
+F2FS_CACHE_FLAG_SET_FUNC(writeback, WRITEBACK);
+F2FS_CACHE_FLAG_CLEAR_FUNC(writeback, WRITEBACK);
+
+F2FS_CACHE_FLAG_TEST_FUNC(inline, INLINE_DATA);
+F2FS_CACHE_FLAG_SET_FUNC(inline, INLINE_DATA);
+F2FS_CACHE_FLAG_CLEAR_FUNC(inline, INLINE_DATA);
+
+static inline void *cache_address(const struct f2fs_cached_block *entry)
+{
+	return entry->data;
+}
+
+#define CACHED_NODE(entry)	((struct f2fs_node *)(cache_address(entry)))
+
+static inline struct folio *cache_folio(const struct f2fs_cached_block *entry)
+{
+	return virt_to_folio(entry->data);
+}
+
+int f2fs_init_cache(struct f2fs_sb_info *sbi,
+				struct f2fs_cached_block_list *cache,
+				enum f2fs_cache_type type);
+void f2fs_destroy_cache(struct f2fs_cached_block_list *cache);
+void f2fs_cache_get(struct f2fs_cached_block *entry);
+struct f2fs_cached_block *f2fs_find_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index);
+#define F2FS_CACHE_TAG_NONE		0
+#define F2FS_CACHE_TAG_DIRTY		1
+#define F2FS_CACHE_TAG_WRITEBACK	2
+
+bool f2fs_trylock_cache(struct f2fs_cached_block *entry);
+void f2fs_lock_cache(struct f2fs_cached_block *entry);
+void f2fs_unlock_cache(struct f2fs_cached_block *entry);
+bool f2fs_put_cache(struct f2fs_cached_block *entry, bool unlock);
+bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry);
+bool f2fs_clear_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_drop_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_force_clear_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_start_cache_writeback(struct f2fs_cached_block *entry);
+void f2fs_end_cache_writeback(struct f2fs_cached_block *entry);
+unsigned int f2fs_cache_gang_lookup_tag(struct f2fs_cached_block_list *cache,
+		struct f2fs_cached_block **results, pgoff_t *first_index,
+		unsigned int max_items, int tag);
+void f2fs_cache_gang_release(struct f2fs_cached_block **entries,
+				unsigned int nr_entries);
+int f2fs_writeback_cache(struct f2fs_cached_block_list *cache, bool sync);
+void f2fs_cache_wait_on_all_writeback(struct f2fs_cached_block_list *cache);
+void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
+					enum page_type type);
+void f2fs_cache_wait_writeback(struct f2fs_cached_block *entry);
+struct f2fs_cached_block *f2fs_grab_cache(struct f2fs_cached_block_list *cache,
+				unsigned long index, int flags);
+void f2fs_drop_cache_range(struct f2fs_cached_block_list *cache,
+		unsigned long start, unsigned long len, bool drop_dirty);
+
+#endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/data.c +256 -27
index 6ae0eb37d20f..09474790035b 100644
@@ -41,11 +41,6 @@ struct f2fs_folio_state {
 	unsigned int		read_pages_pending;
 };
 
-struct f2fs_bio {
-	struct work_struct work;
-	struct bio bio;
-};
-
 #define	F2FS_BIO_POOL_SIZE	NR_CURSEG_TYPE
 
 int __init f2fs_init_bioset(void)
@@ -69,7 +64,6 @@ bool f2fs_is_cp_guaranteed(const struct folio *folio)
 		return folio_test_f2fs_gcing(fscrypt_pagecache_folio(folio));
 
 	inode = mapping->host;
-	sbi = F2FS_I_SB(inode);
 
 	if (inode->i_ino == F2FS_META_INO(sbi) ||
 			inode->i_ino == F2FS_NODE_INO(sbi) ||
@@ -437,6 +431,65 @@ static void f2fs_write_end_io(struct bio *bio)
 	}
 }
 
+static void f2fs_cache_read_end_io(struct bio *bio)
+{
+	struct f2fs_cached_block *entry = F2FS_BIO(bio)->entry;
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	enum count_type io_type = IS_META_CACHE(entry->cache) ?
+					F2FS_RD_META : F2FS_RD_NODE;
+	struct f2fs_cached_block *next;
+
+	iostat_update_and_unbind_ctx(bio);
+
+	if (time_to_inject(sbi, FAULT_READ_IO))
+		bio->bi_status = BLK_STS_IOERR;
+
+	while (entry) {
+		next = entry->next_entry;
+		entry->next_entry = NULL;
+
+		if (bio->bi_status == BLK_STS_OK)
+			f2fs_cache_set_uptodate(entry);
+
+		dec_page_count(sbi, io_type);
+
+		f2fs_unlock_cache(entry);
+		entry = next;
+	}
+	bio_put(bio);
+}
+
+static void f2fs_cache_write_end_io(struct bio *bio)
+{
+	struct f2fs_cached_block *entry = F2FS_BIO(bio)->entry;
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	struct f2fs_cached_block *next;
+
+	iostat_update_and_unbind_ctx(bio);
+
+	if (time_to_inject(sbi, FAULT_WRITE_IO))
+		bio->bi_status = BLK_STS_IOERR;
+
+	if (bio->bi_status != BLK_STS_OK)
+		f2fs_stop_checkpoint(sbi, true,
+			STOP_CP_REASON_WRITE_FAIL);
+
+	while (entry) {
+		next = entry->next_entry;
+		entry->next_entry = NULL;
+
+		dec_page_count(sbi, F2FS_WB_CP_DATA);
+
+		if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
+				wq_has_sleeper(&sbi->cp_wait))
+			wake_up(&sbi->cp_wait);
+
+		f2fs_end_cache_writeback(entry);
+		entry = next;
+	}
+	bio_put(bio);
+}
+
 #ifdef CONFIG_BLK_DEV_ZONED
 static void f2fs_zone_write_end_io(struct bio *bio)
 {
@@ -444,7 +497,10 @@ static void f2fs_zone_write_end_io(struct bio *bio)
 
 	bio->bi_private = io->bi_private;
 	complete(&io->zone_wait);
-	f2fs_write_end_io(bio);
+	if (io->fio.is_cache)
+		f2fs_cache_write_end_io(bio);
+	else
+		f2fs_write_end_io(bio);
 }
 #endif
 
@@ -531,12 +587,21 @@ static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
 				fio->op | fio->op_flags | f2fs_io_flags(fio),
 				GFP_NOIO, &f2fs_bioset);
 	bio->bi_iter.bi_sector = sector;
+	F2FS_BIO(bio)->entry = NULL;
+	bio->bi_private = NULL;
 	if (is_read_io(fio->op)) {
-		bio->bi_end_io = f2fs_read_end_io;
-		bio->bi_private = NULL;
+		if (fio->is_cache)
+			bio->bi_end_io = f2fs_cache_read_end_io;
+		else
+			bio->bi_end_io = f2fs_read_end_io;
 	} else {
-		bio->bi_end_io = f2fs_write_end_io;
-		bio->bi_private = sbi;
+		if (fio->is_cache) {
+			bio->bi_end_io = f2fs_cache_write_end_io;
+		} else {
+			bio->bi_end_io = f2fs_write_end_io;
+			bio->bi_private = sbi;
+		}
+
 		bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
 						fio->type, fio->temp);
 		bio->bi_write_stream = f2fs_io_type_to_write_stream(bdev, fio->type,
@@ -559,7 +624,7 @@ static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
 	 * read/write raw data without encryption.
 	 */
-	if (!fio || !fio->encrypted_page)
+	if (!fio || (!fio->encrypted_page && !fio->is_cache))
 		fscrypt_set_bio_crypt_ctx(bio, inode,
 				(loff_t)first_idx << inode->i_blkbits,
 				gfp_mask);
@@ -769,6 +834,53 @@ void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
 	__submit_merged_write_cond(sbi, NULL, folio, 0, type, true);
 }
 
+static bool __has_merged_cache(struct bio *bio,
+		struct f2fs_cached_block *target)
+{
+	struct f2fs_cached_block *entry;
+
+	if (!bio)
+		return false;
+
+	entry = F2FS_BIO(bio)->entry;
+
+	while (entry) {
+		if (target && entry == target)
+			return true;
+		entry = entry->next_entry;
+	}
+	return false;
+}
+
+bool f2fs_submit_merged_write_cache(struct f2fs_cached_block *entry,
+				enum page_type type)
+{
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	enum temp_type temp;
+	bool ret = false;
+
+	for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
+		enum page_type btype = PAGE_TYPE_OF_BIO(type);
+		struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
+		struct f2fs_lock_context lc;
+		bool merged;
+
+		f2fs_down_read_trace(&io->io_rwsem, &lc);
+		merged = __has_merged_cache(io->bio, entry);
+		f2fs_up_read_trace(&io->io_rwsem, &lc);
+
+		if (merged) {
+			__f2fs_submit_merged_write(sbi, type, temp);
+			ret = true;
+		}
+
+		/* TODO: use HOT temp only for meta pages now. */
+		if (type >= META)
+			break;
+	}
+	return ret;
+}
+
 void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
 {
 	f2fs_submit_merged_write(sbi, DATA);
@@ -832,6 +944,8 @@ static bool io_type_is_mergeable(struct f2fs_bio_info *io,
 
 	if (io->fio.op != fio->op)
 		return false;
+	if (io->fio.is_cache != fio->is_cache)
+		return false;
 	return (io->fio.op_flags & mask) == (fio->op_flags & mask);
 }
 
@@ -1063,6 +1177,34 @@ static bool is_end_zone_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr)
 		f2fs_blkz_is_seq(sbi, devi, blkaddr) &&
 		(blkaddr % sbi->blocks_per_blkz == sbi->blocks_per_blkz - 1);
 }
+
+static void f2fs_wait_zone_io_completion(struct f2fs_sb_info *sbi,
+			struct f2fs_bio_info *io, enum page_type btype)
+{
+	if (f2fs_sb_has_blkzoned(sbi) && btype < META && io->zone_pending_bio) {
+		wait_for_completion_io(&io->zone_wait);
+		bio_put(io->zone_pending_bio);
+		io->zone_pending_bio = NULL;
+		io->bi_private = NULL;
+	}
+}
+
+static void f2fs_submit_zone_io(struct f2fs_sb_info *sbi,
+			struct f2fs_io_info *fio, struct f2fs_bio_info *io,
+			enum page_type btype)
+{
+	if (f2fs_sb_has_blkzoned(sbi) && btype < META &&
+			is_end_zone_blkaddr(sbi, fio->new_blkaddr)) {
+		bio_get(io->bio);
+		reinit_completion(&io->zone_wait);
+		io->bi_private = io->bio->bi_private;
+		io->bio->bi_private = io;
+		io->bio->bi_end_io = f2fs_zone_write_end_io;
+		io->zone_pending_bio = io->bio;
+		__submit_merged_bio(io);
+	}
+
+}
 #endif
 
 void f2fs_submit_page_write(struct f2fs_io_info *fio)
@@ -1079,14 +1221,8 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
 	f2fs_down_write_trace(&io->io_rwsem, &lc);
 next:
 #ifdef CONFIG_BLK_DEV_ZONED
-	if (f2fs_sb_has_blkzoned(sbi) && btype < META && io->zone_pending_bio) {
-		wait_for_completion_io(&io->zone_wait);
-		bio_put(io->zone_pending_bio);
-		io->zone_pending_bio = NULL;
-		io->bi_private = NULL;
-	}
+	f2fs_wait_zone_io_completion(sbi, io, btype);
 #endif
-
 	if (fio->in_list) {
 		spin_lock(&io->io_lock);
 		if (list_empty(&io->io_list)) {
@@ -1141,16 +1277,109 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
 
 	trace_f2fs_submit_folio_write(fio->folio, fio);
 #ifdef CONFIG_BLK_DEV_ZONED
-	if (f2fs_sb_has_blkzoned(sbi) && btype < META &&
-			is_end_zone_blkaddr(sbi, fio->new_blkaddr)) {
-		bio_get(io->bio);
-		reinit_completion(&io->zone_wait);
-		io->bi_private = io->bio->bi_private;
-		io->bio->bi_private = io;
-		io->bio->bi_end_io = f2fs_zone_write_end_io;
-		io->zone_pending_bio = io->bio;
+	f2fs_submit_zone_io(sbi, fio, io, btype);
+#endif
+
+	if (fio->in_list)
+		goto next;
+out:
+	if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
+				!f2fs_is_checkpoint_ready(sbi))
 		__submit_merged_bio(io);
+	f2fs_up_write_trace(&io->io_rwsem, &lc);
+}
+
+static void f2fs_bio_add_cache(struct f2fs_io_info *fio, struct bio *bio)
+{
+	struct f2fs_bio *fbio = F2FS_BIO(bio);
+	struct f2fs_cached_block *head = fbio->entry;
+	struct f2fs_cached_block *new = fio->cache_entry;
+
+	new->next_entry = head;
+	fbio->entry = new;
+}
+
+int f2fs_submit_cache_read(struct f2fs_io_info *fio)
+{
+	struct f2fs_sb_info *sbi = fio->sbi;
+	struct f2fs_cached_block *entry = fio->cache_entry;
+	struct bio *bio;
+	enum count_type io_type = IS_META_CACHE(entry->cache) ?
+				F2FS_RD_META : F2FS_RD_NODE;
+
+	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
+			fio->is_por ? META_POR : (__is_meta_io(fio) ?
+			META_GENERIC : DATA_GENERIC_ENHANCE)))
+		return -EFSCORRUPTED;
+
+	bio = __bio_alloc(fio, 1);
+
+	bio_add_virt_nofail(bio, cache_address(entry), sbi->blocksize);
+	f2fs_bio_add_cache(fio, bio);
+	inc_page_count(sbi, io_type);
+
+	f2fs_submit_read_bio(sbi, bio, fio->type);
+	return 0;
+}
+
+void f2fs_submit_cache_write(struct f2fs_io_info *fio)
+{
+	struct f2fs_sb_info *sbi = fio->sbi;
+	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
+	struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
+	struct f2fs_lock_context lc;
+	struct folio *folio;
+
+	f2fs_bug_on(sbi, is_read_io(fio->op));
+
+	f2fs_down_write_trace(&io->io_rwsem, &lc);
+next:
+#ifdef CONFIG_BLK_DEV_ZONED
+	f2fs_wait_zone_io_completion(sbi, io, btype);
+#endif
+	if (fio->in_list) {
+		spin_lock(&io->io_lock);
+		if (list_empty(&io->io_list)) {
+			spin_unlock(&io->io_lock);
+			goto out;
+		}
+		fio = list_first_entry(&io->io_list,
+						struct f2fs_io_info, list);
+		list_del(&fio->list);
+		spin_unlock(&io->io_lock);
 	}
+
+	verify_fio_blkaddr(fio);
+
+	fio->submitted = 1;
+	inc_page_count(sbi, F2FS_WB_CP_DATA);
+
+	if (io->bio &&
+	    (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
+			      fio->new_blkaddr)))
+		__submit_merged_bio(io);
+alloc_new:
+	if (io->bio == NULL) {
+		io->bio = __bio_alloc(fio, BIO_MAX_VECS);
+		io->fio = *fio;
+	}
+
+	folio = cache_folio(fio->cache_entry);
+
+	if (!bio_add_folio(io->bio, folio, sbi->blocksize,
+			offset_in_folio(folio, cache_address(fio->cache_entry)))) {
+		f2fs_bug_on(sbi, !F2FS_BIO(io->bio)->entry);
+
+		__submit_merged_bio(io);
+		goto alloc_new;
+	}
+
+	f2fs_bio_add_cache(fio, io->bio);
+
+	io->last_block_in_bio = fio->new_blkaddr;
+
+#ifdef CONFIG_BLK_DEV_ZONED
+	f2fs_submit_zone_io(sbi, fio, io, btype);
 #endif
 	if (fio->in_list)
 		goto next;
fs/f2fs/f2fs.h +24 -0
index 1b96d8718c5c..8413983ea9d5 100644
@@ -221,6 +221,8 @@ struct f2fs_rwsem {
 #endif
 };
 
+#include "cache.h"
+
 struct f2fs_mount_info {
 	unsigned long long opt;
 	block_t root_reserved_blocks;	/* root reserved blocks */
@@ -1370,8 +1372,10 @@ struct f2fs_io_info {
 	unsigned int is_por:1;		/* indicate IO is from recovery or not */
 	unsigned int encrypted:1;	/* indicate file is encrypted */
 	unsigned int meta_gc:1;		/* require meta inode GC */
+	unsigned int is_cache:1;	/* indicate IO is from internal cache */
 	enum iostat_type io_type;	/* io type */
 	struct writeback_control *io_wbc; /* writeback control */
+	struct f2fs_cached_block *cache_entry;
 	struct bio **bio;		/* bio for ipu */
 	sector_t *last_block;		/* last block number in bio */
 };
@@ -1781,6 +1785,12 @@ struct f2fs_gc_kthread {
 	unsigned int boost_gc_greedy;
 };
 
+struct f2fs_bio {
+	struct work_struct work;
+	struct f2fs_cached_block *entry;
+	struct bio bio;
+};
+
 struct f2fs_sb_info {
 	struct super_block *sb;			/* pointer to VFS super block */
 	struct proc_dir_entry *s_proc;		/* proc entry */
@@ -2323,6 +2333,16 @@ static inline bool is_node_folio(struct folio *folio)
 	return folio->mapping == NODE_MAPPING(F2FS_F_SB(folio));
 }
 
+static inline struct f2fs_bio *F2FS_BIO(struct bio *bio)
+{
+	return container_of(bio, struct f2fs_bio, bio);
+}
+
+static inline bool f2fs_is_cache_bio(struct bio *bio)
+{
+	return F2FS_BIO(bio)->entry != NULL;
+}
+
 static inline bool is_sbi_flag_set(struct f2fs_sb_info *sbi, unsigned int type)
 {
 	return test_bit(type, &sbi->s_flag);
@@ -4237,6 +4257,8 @@ void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
 				nid_t ino, enum page_type type);
 void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
 				struct folio *folio, enum page_type type);
+bool f2fs_submit_merged_write_cache(struct f2fs_cached_block *entry,
+				enum page_type type);
 void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
 					struct bio **bio, struct folio *folio);
 void f2fs_submit_all_merged_ipu_writes(struct f2fs_sb_info *sbi);
@@ -4244,6 +4266,8 @@ void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi);
 int f2fs_submit_page_bio(struct f2fs_io_info *fio);
 int f2fs_merge_page_bio(struct f2fs_io_info *fio);
 void f2fs_submit_page_write(struct f2fs_io_info *fio);
+int f2fs_submit_cache_read(struct f2fs_io_info *fio);
+void f2fs_submit_cache_write(struct f2fs_io_info *fio);
 struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
 		block_t blk_addr, sector_t *sector);
 int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr);
fs/f2fs/iostat.h +11 -0
index 2025225b5bed..61c6bc8e3119 100644
@@ -60,6 +60,13 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 	return iostat_ctx->post_read_ctx;
 }
 
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
+
+	iostat_ctx->post_read_ctx = ctx;
+}
+
 extern void iostat_update_and_unbind_ctx(struct bio *bio);
 extern void iostat_alloc_and_bind_ctx(struct f2fs_sb_info *sbi,
 		struct bio *bio, struct bio_post_read_ctx *ctx);
@@ -81,6 +88,10 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 {
 	return bio->bi_private;
 }
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	bio->bi_private = ctx;
+}
 static inline int f2fs_init_iostat_processing(void) { return 0; }
 static inline void f2fs_destroy_iostat_processing(void) {}
 static inline int f2fs_init_iostat(struct f2fs_sb_info *sbi) { return 0; }
Signature · 8 lines
2.49.0



_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
author maintainer
#3 lore raw

Re: [f2fs-dev] [PATCH v1 01/12] f2fs: cache: implement metadata cache

Quoted 43 lines · Jaegeuk Kim
Quoted 42 lines · On 08/20, Chao Yu via Linux-f2fs-devel wrote:

This patch introduces the core metadata block caching infrastructure to manage f2fs metadata independently of the page cache.

It implements:
- core cache APIs: get, create, put, drop, backed by a radix tree and a single global LRU list.
- support multiple status of cached block: LOCKED, UPTODATE, DIRTY, WRITEBACK, INLINE.
- internal bio based read/write helpers with adjacent block vector merging.

---
 fs/f2fs/Makefile |   2 +-
 fs/f2fs/cache.c  | 531 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h  | 179 ++++++++++++++++
 fs/f2fs/data.c   | 283 ++++++++++++++++++++++---
 fs/f2fs/f2fs.h   |  24 +++
 fs/f2fs/iostat.h |  11 +
 6 files changed, 1002 insertions(+), 28 deletions(-)
 create mode 100644 fs/f2fs/cache.c
 create mode 100644 fs/f2fs/cache.h
fs/f2fs/Makefile +1 -1
index 8a7322d229e4..fbf49c30b066 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_F2FS_FS) += f2fs.o
 
 f2fs-y		:= dir.o file.o inode.o namei.o hash.o super.o inline.o
 f2fs-y		+= checkpoint.o gc.o data.o node.o segment.o recovery.o
-f2fs-y		+= shrinker.o extent_cache.o sysfs.o
+f2fs-y		+= shrinker.o extent_cache.o sysfs.o cache.o
 f2fs-$(CONFIG_F2FS_STAT_FS) += debug.o
 f2fs-$(CONFIG_F2FS_FS_XATTR) += xattr.o
 f2fs-$(CONFIG_F2FS_FS_POSIX_ACL) += acl.o
fs/f2fs/cache.c new file +1 -0
new file mode 100644
index 000000000000..08bc658166f7
@@ -0,0 +1,531 @@
+// SPDX-License-Identifier: GPL-2.0
Quoted 8 lines

fs/f2fs/cache.c line 6

Quoted 5 lines
fs/f2fs/cache.c new file +5 -0
+/*
+ * Copyright (c) 2026 Google LLC
+ * Author: Chao Yu <[email protected]>
+ */
+#include "linux/spinlock.h"

#include <linux/spinlock.h>?
May need to Move below?

Oh, we don't need this line, let's remove.

And will fix below all according to your suggestion.

Thanks,

Quoted 1,212 lines
Quoted 34 lines
fs/f2fs/cache.c new file +33 -0
+#include <linux/fs.h>
+#include <linux/f2fs_fs.h>
+#include <linux/radix-tree.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/pagemap.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/delay.h>
+#include "f2fs.h"
+#include "cache.h"
+#include "node.h"
+#include "segment.h"
+
+void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
+					enum page_type type)
+{
+	/* in case the entry was truncated or on-going shrink */
+	if (!entry->cache)
+		return;
+
+	if (!f2fs_cache_test_writeback(entry))
+		return;
+
+	/* submit cached bio */
+	f2fs_submit_merged_write_cache(entry, type);
+
+	wait_on_bit_io(&entry->state, F2FS_BLOCK_WRITEBACK,
+					TASK_UNINTERRUPTIBLE);
+}
+
+void f2fs_cache_wait_writeback(struct f2fs_cached_block *entry)
+{

fs/f2fs/cache.c line 49

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+	/* in case the entry was truncated or on-going shrink */
+	if (!entry->cache)
+		return;
+
+	f2fs_cache_wait_writeback_cond(entry,
+			IS_META_CACHE(entry->cache) ? META : NODE);
+}
+
+static void f2fs_cache_update_tag(struct f2fs_cached_block *entry, unsigned int src,
+					unsigned int dst)

f2fs_cache_update_tag(entry, clear_from, set_to);

fs/f2fs/cache.c line 55

Quoted 6 lines
fs/f2fs/cache.c new file +6 -0
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	if (src)
	if (clear_fom != F2FS_CACHE_TAG_NONE)

fs/f2fs/cache.c line 57

Quoted 2 lines
fs/f2fs/cache.c new file +2 -0
+		radix_tree_tag_clear(&cache->root, entry->index, src);
+	if (dst)
	if (set_to != F2FS_CACHE_TAG_NONE)

fs/f2fs/cache.c line 67

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+		radix_tree_tag_set(&cache->root, entry->index, dst);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+}
+
+bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry)
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+
+	if (!f2fs_cache_test_uptodate(entry))
+		f2fs_cache_set_uptodate(entry);
	f2fs_cache_set_uptodate(entry);
Quoted 4 lines
fs/f2fs/cache.c new file +4 -0
+
+#ifdef CONFIG_F2FS_CHECK_FS
+	if (cache->type == F2FS_NODE_CACHE && IS_INODE(cache_folio(entry)))
+		f2fs_inode_chksum_set(cache->sbi, cache_folio(entry));

fs/f2fs/cache.c line 81

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+#endif
+
+	if (f2fs_cache_test_dirty(entry))
+		return false;
+
+	if (!f2fs_cache_test_and_set_dirty(entry)) {
+		enum count_type type = IS_META_CACHE(cache) ?
+				F2FS_DIRTY_META : F2FS_DIRTY_NODES;
+
+		f2fs_cache_update_tag(entry, 0, F2FS_CACHE_TAG_DIRTY);
		f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_NONE, F2FS_CACHE_TAG_DIRTY);
Quoted 7 lines
fs/f2fs/cache.c new file +7 -0
+		inc_page_count(cache->sbi, type);
+		return true;
+	}
+
+	return false;
+}
+

fs/f2fs/cache.c line 98

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+bool f2fs_clear_cache_dirty(struct f2fs_cached_block *entry)
+{
+	if (!f2fs_cache_test_dirty(entry))
+		return false;
+
+	f2fs_cache_clear_dirty(entry);
+	return true;
+}
+
+static void __f2fs_drop_cache_dirty(struct f2fs_cached_block *entry, bool force)

static void __drop_cache_dirty(struct f2fs_cached_block *entry, bool force)

Quoted 80 lines
fs/f2fs/cache.c new file +80 -0
+{
+
+	struct f2fs_cached_block_list *cache = entry->cache;
+	enum count_type type = IS_META_CACHE(cache) ?
+				F2FS_DIRTY_META : F2FS_DIRTY_NODES;
+
+	f2fs_cache_clear_uptodate(entry);
+
+	if (!force && !f2fs_clear_cache_dirty(entry))
+		return;
+
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_DIRTY, 0);
+	dec_page_count(cache->sbi, type);
+}
+
+void f2fs_drop_cache_dirty(struct f2fs_cached_block *entry)
+{
+	__f2fs_drop_cache_dirty(entry, false);
+}
+
+void f2fs_force_clear_cache_dirty(struct f2fs_cached_block *entry)
+{
+	__f2fs_drop_cache_dirty(entry, true);
+}
+
+void f2fs_start_cache_writeback(struct f2fs_cached_block *entry)
+{
+	f2fs_cache_set_writeback(entry);
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_DIRTY,
+				F2FS_CACHE_TAG_WRITEBACK);
+}
+
+void f2fs_end_cache_writeback(struct f2fs_cached_block *entry)
+{
+	/*
+	 * should call f2fs_cache_update_tag() before clearing writeback bit,
+	 * in case f2fs_truncate_cache() set entry->cache to NULL.
+	 */
+	f2fs_cache_update_tag(entry, F2FS_CACHE_TAG_WRITEBACK, 0);
+	clear_and_wake_up_bit(F2FS_BLOCK_WRITEBACK, &entry->state);
+}
+
+static int f2fs_cache_refcount(struct f2fs_cached_block *entry)
+{
+	return atomic_read(&entry->refcount);
+}
+
+static void __f2fs_free_cache(struct f2fs_cached_block *entry)
+{
+	kfree(entry->data);
+	kfree(entry);
+}
+
+static void f2fs_free_cache(struct f2fs_cached_block *entry)
+{
+	WARN_ON_ONCE(!list_empty(&entry->list));
+	WARN_ON_ONCE(f2fs_cache_refcount(entry));
+	__f2fs_free_cache(entry);
+}
+
+void f2fs_cache_get(struct f2fs_cached_block *entry)
+{
+	atomic_inc(&entry->refcount);
+}
+
+static bool f2fs_cache_put(struct f2fs_cached_block *entry)
+{
+	WARN_ON_ONCE(!f2fs_cache_refcount(entry));
+	if (atomic_dec_and_test(&entry->refcount)) {
+		f2fs_free_cache(entry);
+		return true;
+	}
+	return false;
+}
+
+static struct f2fs_cached_block *f2fs_create_cache(
+		struct f2fs_cached_block_list *cache,
+		unsigned long index, bool nofail)
+{
+	struct f2fs_cached_block *entry;

fs/f2fs/cache.c line 188

Quoted 10 lines
fs/f2fs/cache.c new file +10 -0
+	unsigned int flags = GFP_NOFS;
+
+	if (nofail)
+		flags |= __GFP_NOFAIL;
+
+	entry = kzalloc_obj(*entry, flags);
+	if (!entry)
+		return ERR_PTR(-ENOMEM);
+
+	entry->data = kzalloc(cache->sbi->blocksize, flags);

We don't need kzalloc() since we have an uptodate flag.

Quoted 1,007 lines
fs/f2fs/cache.c new file +343 -0
+	if (!entry->data) {
+		kfree(entry);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	entry->index = index;
+
+	atomic_set(&entry->refcount, 0);
+	entry->next_entry = NULL;
+	INIT_LIST_HEAD(&entry->list);
+
+	entry->cache = cache;
+
+	return entry;
+}
+
+static struct f2fs_cached_block *f2fs_insert_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index,
+			struct f2fs_cached_block *new)
+{
+	struct f2fs_cached_block *e;
+	int ret;
+	unsigned long flags;
+
+	ret = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
+	f2fs_bug_on(cache->sbi, ret);
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	e = radix_tree_lookup(&cache->root, index);
+	if (!e) {
+		e = new;
+		f2fs_bug_on(cache->sbi, f2fs_cache_refcount(e));
+
+		ret = radix_tree_insert(&cache->root, index, e);
+		f2fs_bug_on(cache->sbi, ret);
+
+		/* radix tree referenced cache entry */
+		f2fs_cache_get(e);
+		f2fs_bug_on(cache->sbi, !list_empty(&e->list));
+		list_add_tail(&e->list, &cache->lru_list);
+		cache->num_entries++;
+	}
+	f2fs_cache_get(e);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+	radix_tree_preload_end();
+
+	if (new != e) {
+		f2fs_bug_on(cache->sbi, f2fs_cache_refcount(new));
+		__f2fs_free_cache(new);
+	}
+
+	return e;
+}
+
+struct f2fs_cached_block *f2fs_find_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index)
+{
+	struct f2fs_cached_block *entry;
+	unsigned long flags;
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	entry = radix_tree_lookup(&cache->root, index);
+	if (entry) {
+		f2fs_bug_on(cache->sbi, !f2fs_cache_refcount(entry));
+		f2fs_cache_get(entry);
+		if (!list_empty(&entry->list))
+			list_move_tail(&entry->list, &cache->lru_list);
+
+	} else {
+		entry = ERR_PTR(-ENOENT);
+	}
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+
+	return entry;
+}
+
+struct f2fs_cached_block *f2fs_grab_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index, int flags)
+
+{
+	struct f2fs_cached_block *entry, *new;
+	bool create = flags & F2FS_CACHE_CREATE;
+	bool nofail = flags & F2FS_CACHE_NOFAIL;
+	bool lock = flags & F2FS_CACHE_LOCK;
+
+repeat:
+	entry = f2fs_find_cache(cache, index);
+	if (!IS_ERR(entry))
+		goto found;
+
+	if (!create)
+		return ERR_PTR(-ENOENT);
+
+	new = f2fs_create_cache(cache, index, nofail);
+	if (IS_ERR(new))
+		return new;
+
+	entry = f2fs_insert_cache(cache, index, new);
+found:
+	if (lock) {
+		f2fs_lock_cache(entry);
+		/* has been truncated */
+		if (entry->cache != cache) {
+			f2fs_put_cache(entry, true);
+			goto repeat;
+		}
+	}
+	return entry;
+}
+
+bool f2fs_trylock_cache(struct f2fs_cached_block *entry)
+{
+	return !test_and_set_bit(F2FS_BLOCK_LOCKED, &entry->state);
+}
+
+void f2fs_lock_cache(struct f2fs_cached_block *entry)
+{
+	wait_on_bit_lock(&entry->state, F2FS_BLOCK_LOCKED,
+					TASK_UNINTERRUPTIBLE);
+}
+
+void f2fs_unlock_cache(struct f2fs_cached_block *entry)
+{
+	clear_and_wake_up_bit(F2FS_BLOCK_LOCKED, &entry->state);
+}
+
+bool f2fs_put_cache(struct f2fs_cached_block *entry, bool unlock)
+{
+	if (IS_ERR_OR_NULL(entry))
+		return false;
+	if (unlock)
+		f2fs_unlock_cache(entry);
+	return f2fs_cache_put(entry);
+}
+
+unsigned int f2fs_cache_gang_lookup_tag(struct f2fs_cached_block_list *cache,
+				struct f2fs_cached_block **entries,
+				pgoff_t *first_index, unsigned int max_nr,
+				int tag)
+{
+	unsigned long flags;
+	int nr, i;
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	nr = radix_tree_gang_lookup_tag(&cache->root, (void **)entries,
+				*first_index, max_nr, tag);
+	if (!nr)
+		goto out;
+
+	for (i = 0; i < nr; i++)
+		f2fs_cache_get(entries[i]);
+	*first_index = entries[nr - 1]->index + 1;
+out:
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	return nr;
+}
+
+void f2fs_cache_gang_release(struct f2fs_cached_block **entries,
+				unsigned int nr_entries)
+{
+	int i;
+
+	for (i = 0; i < nr_entries; i++)
+		f2fs_put_cache(entries[i], false);
+}
+
+void f2fs_cache_wait_on_all_writeback(struct f2fs_cached_block_list *cache)
+{
+	unsigned long index = 0;
+	struct f2fs_cached_block *entries[F2FS_ONSTACK_CACHES];
+	int nr, i;
+
+next:
+	nr = f2fs_cache_gang_lookup_tag(cache, entries, &index,
+			F2FS_ONSTACK_CACHES, F2FS_CACHE_TAG_WRITEBACK);
+	if (!nr)
+		return;
+
+	for (i = 0; i < nr; i++)
+		f2fs_cache_wait_writeback(entries[i]);
+	f2fs_cache_gang_release(entries, nr);
+	goto next;
+}
+
+static void f2fs_do_truncate_cache(struct f2fs_cached_block *entry,
+						bool drop_dirty)
+{
+	struct f2fs_cached_block_list *cache = entry->cache;
+	unsigned long flags;
+
+	if (!drop_dirty &&
+		(f2fs_cache_test_dirty(entry) ||
+		f2fs_cache_test_writeback(entry)))
+		return;
+
+	f2fs_cache_wait_writeback(entry);
+	f2fs_drop_cache_dirty(entry);
+
+	spin_lock(&cache->list_lock);
+	spin_lock_irqsave(&cache->tree_lock, flags);
+
+	f2fs_bug_on(cache->sbi, !entry->cache);
+	if (!radix_tree_delete(&cache->root, entry->index))
+		f2fs_bug_on(cache->sbi, !entry->cache);
+
+	entry->cache = NULL;
+	cache->num_entries--;
+
+	atomic_dec(&entry->refcount);
+	f2fs_bug_on(cache->sbi, !f2fs_cache_refcount(entry));
+
+	f2fs_bug_on(cache->sbi, list_empty(&entry->list));
+	list_del_init(&entry->list);
+
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	spin_unlock(&cache->list_lock);
+}
+
+static void f2fs_truncate_cache(struct f2fs_cached_block *entry,
+					bool drop_dirty)
+{
+	f2fs_lock_cache(entry);
+	if (entry->cache)
+		f2fs_do_truncate_cache(entry, drop_dirty);
+	f2fs_unlock_cache(entry);
+}
+
+static void f2fs_drop_cache(struct f2fs_cached_block_list *cache,
+			block_t blkaddr, bool drop_dirty)
+{
+	struct f2fs_cached_block *entry;
+
+	entry = f2fs_find_cache(cache, blkaddr);
+	if (IS_ERR(entry))
+		return;
+
+	f2fs_truncate_cache(entry, drop_dirty);
+	f2fs_put_cache(entry, false);
+}
+
+void f2fs_drop_cache_range(struct f2fs_cached_block_list *cache,
+		unsigned long start, unsigned long len, bool drop_dirty)
+{
+	unsigned long index = start;
+	unsigned long end = (ULONG_MAX - start < len) ?
+				ULONG_MAX : (start + len);
+	struct f2fs_cached_block *entries[F2FS_ONSTACK_CACHES];
+	unsigned long flags;
+	int nr, i;
+
+	if (len == 1)
+		return f2fs_drop_cache(cache, index, drop_dirty);
+
+next:
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	nr = radix_tree_gang_lookup(&cache->root, (void **)entries, index,
+		min((unsigned long)F2FS_ONSTACK_CACHES, end - index));
+	if (!nr)
+		goto out_unlock;
+
+	for (i = 0; i < nr; i++) {
+		struct f2fs_cached_block *entry = entries[i];
+
+		if (entry->index >= end) {
+			nr = i;
+			break;
+		}
+		f2fs_cache_get(entry);
+	}
+out_unlock:
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+	if (!nr)
+		return;
+
+	for (i = 0; i < nr; i++) {
+		struct f2fs_cached_block *entry = entries[i];
+
+		index = entry->index + 1;
+
+		f2fs_truncate_cache(entry, drop_dirty);
+	}
+	f2fs_cache_gang_release(entries, nr);
+
+	if (index < end)
+		goto next;
+}
+
+int f2fs_init_cache(struct f2fs_sb_info *sbi,
+		struct f2fs_cached_block_list *cache,
+		enum f2fs_cache_type type)
+{
+	cache->sbi = sbi;
+	cache->type = type;
+	INIT_RADIX_TREE(&cache->root, GFP_ATOMIC);
+	spin_lock_init(&cache->tree_lock);
+	spin_lock_init(&cache->list_lock);
+	INIT_LIST_HEAD(&cache->lru_list);
+	cache->num_entries = 0;
+
+	return 0;
+}
+
+void f2fs_destroy_cache(struct f2fs_cached_block_list *cache)
+{
+	struct list_head *head = &cache->lru_list;
+	struct f2fs_cached_block *entry;
+	unsigned long flags;
+
+	f2fs_cache_wait_on_all_writeback(cache);
+next:
+	spin_lock(&cache->list_lock);
+	if (list_empty(head)) {
+		spin_unlock(&cache->list_lock);
+		return;
+	}
+	entry = list_first_entry(head, struct f2fs_cached_block, list);
+
+	spin_lock_irqsave(&cache->tree_lock, flags);
+	radix_tree_delete(&cache->root, entry->index);
+	cache->num_entries--;
+	list_del_init(&entry->list);
+	spin_unlock_irqrestore(&cache->tree_lock, flags);
+
+	spin_unlock(&cache->list_lock);
+
+	/* wait on read cache IO */
+	f2fs_lock_cache(entry);
+	/* wait on write cache IO */
+	f2fs_cache_wait_writeback(entry);
+	f2fs_bug_on(cache->sbi, f2fs_cache_test_dirty(entry));
+	f2fs_bug_on(cache->sbi, f2fs_cache_test_writeback(entry));
+	f2fs_bug_on(cache->sbi, !list_empty(&entry->list));
+	f2fs_bug_on(cache->sbi, f2fs_cache_refcount(entry) != 1);
+	f2fs_put_cache(entry, true);
+	goto next;
+}
fs/f2fs/cache.h new file +179 -0
new file mode 100644
index 000000000000..686a974008ae
@@ -0,0 +1,179 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2026 Google LLC
+ * Author: Chao Yu <[email protected]>
+ */
+#ifndef _LINUX_F2FS_CACHE_H
+#define _LINUX_F2FS_CACHE_H
+
+#include <linux/pagemap.h>
+#include <linux/mm.h>
+#include <linux/list.h>
+#include <linux/radix-tree.h>
+#include <linux/spinlock.h>
+#include <linux/wait.h>
+#include <linux/types.h>
+
+struct f2fs_rwsem;
+struct f2fs_io_info;
+enum page_type;
+
+/* Represents a single cached block (meta, node or compress) */
+struct f2fs_cached_block {
+	struct list_head list;		/* LRU list head */
+	struct f2fs_cached_block_list *cache; /* parent cache list */
+	union {
+		/* chain for merged BIO */
+		struct f2fs_cached_block *next_entry;
+		nid_t ino;		/* inode number for compress cache */
+	};
+	unsigned long index;		/* key in radix tree, (meta/compress: pba, node: nid) */
+	unsigned long state;		/* cache entry state (e.g., Dirty, UpToDate) */
+	void *data;			/* blocksize-aligned memory (4KB or 16KB) */
+	atomic_t refcount;		/* reference count */
+};
+
+struct f2fs_sb_info;
+
+enum f2fs_cache_type {
+	F2FS_META_CACHE,
+	F2FS_NODE_CACHE,
+};
+
+/* Main cache control structure (per sb_info) */
+struct f2fs_cached_block_list {
+	struct f2fs_sb_info *sbi;	/* Pointer to f2fs_sb_info */
+	struct radix_tree_root root;	/* Radix tree for cache lookup */
+	spinlock_t tree_lock;		/* Lock for radix tree */
+	struct list_head lru_list;	/* Single global LRU list */
+	spinlock_t list_lock;		/* Lock for LRU list */
+	enum f2fs_cache_type type;	/* Cache type (Node or Meta) */
+	unsigned long num_entries;	/* Current number of entries */
+};
+
+#define IS_META_CACHE(cache) (cache->type == F2FS_META_CACHE)
+
+/* Flags for f2fs_cached_block state */
+enum f2fs_cached_state {
+	F2FS_BLOCK_LOCKED,		/* cache entry is locked */
+	F2FS_BLOCK_UPTODATE,		/* cache data is valid */
+	F2FS_BLOCK_DIRTY,		/* cache data is dirty, need to writeback the data */
+	F2FS_BLOCK_WRITEBACK,		/* cache data is writeback state */
+	F2FS_BLOCK_INLINE_DATA,		/* indicate inline data */
+};
+
+enum {
+	__F2FS_CACHE_CREATE,		/* create the cache if there is no cache entry */
+	__F2FS_CACHE_LOCK,		/* get and lock the cache entry */
+	__F2FS_CACHE_NOFAIL,		/* do not allow failure */
+};
+
+enum f2fs_cache_request_flag {
+	F2FS_CACHE_CREATE	= 1 << __F2FS_CACHE_CREATE,
+	F2FS_CACHE_LOCK		= 1 << __F2FS_CACHE_LOCK,
+	F2FS_CACHE_NOFAIL	= 1 << __F2FS_CACHE_NOFAIL,
+};
+
+#define F2FS_CACHE_LOCK_CREATE	(F2FS_CACHE_LOCK | F2FS_CACHE_CREATE)
+
+#define F2FS_ONSTACK_CACHES		(32)
+
+#define F2FS_CACHE_FLAG_TEST_FUNC(name, flagname)			\
+static inline bool f2fs_cache_test_##name(				\
+			const struct f2fs_cached_block *entry)		\
+{									\
+	return test_bit(F2FS_BLOCK_##flagname, &entry->state);		\
+}									\
+
+#define F2FS_CACHE_FLAG_SET_FUNC(name, flagname)			\
+static inline void f2fs_cache_set_##name(				\
+			struct f2fs_cached_block *entry)		\
+{									\
+	set_bit(F2FS_BLOCK_##flagname, &entry->state);			\
+}									\
+
+#define F2FS_CACHE_FLAG_CLEAR_FUNC(name, flagname)			\
+static inline void f2fs_cache_clear_##name(				\
+			struct f2fs_cached_block *entry)		\
+{									\
+	clear_bit(F2FS_BLOCK_##flagname, &entry->state);		\
+}									\
+
+#define F2FS_CACHE_FLAG_TEST_AND_SET_FUNC(name, flagname)		\
+static inline bool f2fs_cache_test_and_set_##name(			\
+			struct f2fs_cached_block *entry)		\
+{									\
+	return test_and_set_bit(F2FS_BLOCK_##flagname, &entry->state);	\
+}									\
+
+F2FS_CACHE_FLAG_TEST_FUNC(locked, LOCKED);
+F2FS_CACHE_FLAG_SET_FUNC(locked, LOCKED);
+F2FS_CACHE_FLAG_CLEAR_FUNC(locked, LOCKED);
+
+F2FS_CACHE_FLAG_TEST_FUNC(uptodate, UPTODATE);
+F2FS_CACHE_FLAG_SET_FUNC(uptodate, UPTODATE);
+F2FS_CACHE_FLAG_CLEAR_FUNC(uptodate, UPTODATE);
+
+F2FS_CACHE_FLAG_TEST_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_SET_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_CLEAR_FUNC(dirty, DIRTY);
+F2FS_CACHE_FLAG_TEST_AND_SET_FUNC(dirty, DIRTY);
+
+F2FS_CACHE_FLAG_TEST_FUNC(writeback, WRITEBACK);
+F2FS_CACHE_FLAG_SET_FUNC(writeback, WRITEBACK);
+F2FS_CACHE_FLAG_CLEAR_FUNC(writeback, WRITEBACK);
+
+F2FS_CACHE_FLAG_TEST_FUNC(inline, INLINE_DATA);
+F2FS_CACHE_FLAG_SET_FUNC(inline, INLINE_DATA);
+F2FS_CACHE_FLAG_CLEAR_FUNC(inline, INLINE_DATA);
+
+static inline void *cache_address(const struct f2fs_cached_block *entry)
+{
+	return entry->data;
+}
+
+#define CACHED_NODE(entry)	((struct f2fs_node *)(cache_address(entry)))
+
+static inline struct folio *cache_folio(const struct f2fs_cached_block *entry)
+{
+	return virt_to_folio(entry->data);
+}
+
+int f2fs_init_cache(struct f2fs_sb_info *sbi,
+				struct f2fs_cached_block_list *cache,
+				enum f2fs_cache_type type);
+void f2fs_destroy_cache(struct f2fs_cached_block_list *cache);
+void f2fs_cache_get(struct f2fs_cached_block *entry);
+struct f2fs_cached_block *f2fs_find_cache(
+			struct f2fs_cached_block_list *cache,
+			unsigned long index);
+#define F2FS_CACHE_TAG_NONE		0
+#define F2FS_CACHE_TAG_DIRTY		1
+#define F2FS_CACHE_TAG_WRITEBACK	2
+
+bool f2fs_trylock_cache(struct f2fs_cached_block *entry);
+void f2fs_lock_cache(struct f2fs_cached_block *entry);
+void f2fs_unlock_cache(struct f2fs_cached_block *entry);
+bool f2fs_put_cache(struct f2fs_cached_block *entry, bool unlock);
+bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry);
+bool f2fs_clear_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_drop_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_force_clear_cache_dirty(struct f2fs_cached_block *entry);
+void f2fs_start_cache_writeback(struct f2fs_cached_block *entry);
+void f2fs_end_cache_writeback(struct f2fs_cached_block *entry);
+unsigned int f2fs_cache_gang_lookup_tag(struct f2fs_cached_block_list *cache,
+		struct f2fs_cached_block **results, pgoff_t *first_index,
+		unsigned int max_items, int tag);
+void f2fs_cache_gang_release(struct f2fs_cached_block **entries,
+				unsigned int nr_entries);
+int f2fs_writeback_cache(struct f2fs_cached_block_list *cache, bool sync);
+void f2fs_cache_wait_on_all_writeback(struct f2fs_cached_block_list *cache);
+void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
+					enum page_type type);
+void f2fs_cache_wait_writeback(struct f2fs_cached_block *entry);
+struct f2fs_cached_block *f2fs_grab_cache(struct f2fs_cached_block_list *cache,
+				unsigned long index, int flags);
+void f2fs_drop_cache_range(struct f2fs_cached_block_list *cache,
+		unsigned long start, unsigned long len, bool drop_dirty);
+
+#endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/data.c +256 -27
index 6ae0eb37d20f..09474790035b 100644
@@ -41,11 +41,6 @@ struct f2fs_folio_state {
 	unsigned int		read_pages_pending;
 };
 
-struct f2fs_bio {
-	struct work_struct work;
-	struct bio bio;
-};
-
 #define	F2FS_BIO_POOL_SIZE	NR_CURSEG_TYPE
 
 int __init f2fs_init_bioset(void)
@@ -69,7 +64,6 @@ bool f2fs_is_cp_guaranteed(const struct folio *folio)
 		return folio_test_f2fs_gcing(fscrypt_pagecache_folio(folio));
 
 	inode = mapping->host;
-	sbi = F2FS_I_SB(inode);
 
 	if (inode->i_ino == F2FS_META_INO(sbi) ||
 			inode->i_ino == F2FS_NODE_INO(sbi) ||
@@ -437,6 +431,65 @@ static void f2fs_write_end_io(struct bio *bio)
 	}
 }
 
+static void f2fs_cache_read_end_io(struct bio *bio)
+{
+	struct f2fs_cached_block *entry = F2FS_BIO(bio)->entry;
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	enum count_type io_type = IS_META_CACHE(entry->cache) ?
+					F2FS_RD_META : F2FS_RD_NODE;
+	struct f2fs_cached_block *next;
+
+	iostat_update_and_unbind_ctx(bio);
+
+	if (time_to_inject(sbi, FAULT_READ_IO))
+		bio->bi_status = BLK_STS_IOERR;
+
+	while (entry) {
+		next = entry->next_entry;
+		entry->next_entry = NULL;
+
+		if (bio->bi_status == BLK_STS_OK)
+			f2fs_cache_set_uptodate(entry);
+
+		dec_page_count(sbi, io_type);
+
+		f2fs_unlock_cache(entry);
+		entry = next;
+	}
+	bio_put(bio);
+}
+
+static void f2fs_cache_write_end_io(struct bio *bio)
+{
+	struct f2fs_cached_block *entry = F2FS_BIO(bio)->entry;
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	struct f2fs_cached_block *next;
+
+	iostat_update_and_unbind_ctx(bio);
+
+	if (time_to_inject(sbi, FAULT_WRITE_IO))
+		bio->bi_status = BLK_STS_IOERR;
+
+	if (bio->bi_status != BLK_STS_OK)
+		f2fs_stop_checkpoint(sbi, true,
+			STOP_CP_REASON_WRITE_FAIL);
+
+	while (entry) {
+		next = entry->next_entry;
+		entry->next_entry = NULL;
+
+		dec_page_count(sbi, F2FS_WB_CP_DATA);
+
+		if (!get_pages(sbi, F2FS_WB_CP_DATA) &&
+				wq_has_sleeper(&sbi->cp_wait))
+			wake_up(&sbi->cp_wait);
+
+		f2fs_end_cache_writeback(entry);
+		entry = next;
+	}
+	bio_put(bio);
+}
+
 #ifdef CONFIG_BLK_DEV_ZONED
 static void f2fs_zone_write_end_io(struct bio *bio)
 {
@@ -444,7 +497,10 @@ static void f2fs_zone_write_end_io(struct bio *bio)
 
 	bio->bi_private = io->bi_private;
 	complete(&io->zone_wait);
-	f2fs_write_end_io(bio);
+	if (io->fio.is_cache)
+		f2fs_cache_write_end_io(bio);
+	else
+		f2fs_write_end_io(bio);
 }
 #endif
 
@@ -531,12 +587,21 @@ static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
 				fio->op | fio->op_flags | f2fs_io_flags(fio),
 				GFP_NOIO, &f2fs_bioset);
 	bio->bi_iter.bi_sector = sector;
+	F2FS_BIO(bio)->entry = NULL;
+	bio->bi_private = NULL;
 	if (is_read_io(fio->op)) {
-		bio->bi_end_io = f2fs_read_end_io;
-		bio->bi_private = NULL;
+		if (fio->is_cache)
+			bio->bi_end_io = f2fs_cache_read_end_io;
+		else
+			bio->bi_end_io = f2fs_read_end_io;
 	} else {
-		bio->bi_end_io = f2fs_write_end_io;
-		bio->bi_private = sbi;
+		if (fio->is_cache) {
+			bio->bi_end_io = f2fs_cache_write_end_io;
+		} else {
+			bio->bi_end_io = f2fs_write_end_io;
+			bio->bi_private = sbi;
+		}
+
 		bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
 						fio->type, fio->temp);
 		bio->bi_write_stream = f2fs_io_type_to_write_stream(bdev, fio->type,
@@ -559,7 +624,7 @@ static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
 	 * The f2fs garbage collector sets ->encrypted_page when it wants to
 	 * read/write raw data without encryption.
 	 */
-	if (!fio || !fio->encrypted_page)
+	if (!fio || (!fio->encrypted_page && !fio->is_cache))
 		fscrypt_set_bio_crypt_ctx(bio, inode,
 				(loff_t)first_idx << inode->i_blkbits,
 				gfp_mask);
@@ -769,6 +834,53 @@ void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
 	__submit_merged_write_cond(sbi, NULL, folio, 0, type, true);
 }
 
+static bool __has_merged_cache(struct bio *bio,
+		struct f2fs_cached_block *target)
+{
+	struct f2fs_cached_block *entry;
+
+	if (!bio)
+		return false;
+
+	entry = F2FS_BIO(bio)->entry;
+
+	while (entry) {
+		if (target && entry == target)
+			return true;
+		entry = entry->next_entry;
+	}
+	return false;
+}
+
+bool f2fs_submit_merged_write_cache(struct f2fs_cached_block *entry,
+				enum page_type type)
+{
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
+	enum temp_type temp;
+	bool ret = false;
+
+	for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
+		enum page_type btype = PAGE_TYPE_OF_BIO(type);
+		struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
+		struct f2fs_lock_context lc;
+		bool merged;
+
+		f2fs_down_read_trace(&io->io_rwsem, &lc);
+		merged = __has_merged_cache(io->bio, entry);
+		f2fs_up_read_trace(&io->io_rwsem, &lc);
+
+		if (merged) {
+			__f2fs_submit_merged_write(sbi, type, temp);
+			ret = true;
+		}
+
+		/* TODO: use HOT temp only for meta pages now. */
+		if (type >= META)
+			break;
+	}
+	return ret;
+}
+
 void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
 {
 	f2fs_submit_merged_write(sbi, DATA);
@@ -832,6 +944,8 @@ static bool io_type_is_mergeable(struct f2fs_bio_info *io,
 
 	if (io->fio.op != fio->op)
 		return false;
+	if (io->fio.is_cache != fio->is_cache)
+		return false;
 	return (io->fio.op_flags & mask) == (fio->op_flags & mask);
 }
 
@@ -1063,6 +1177,34 @@ static bool is_end_zone_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr)
 		f2fs_blkz_is_seq(sbi, devi, blkaddr) &&
 		(blkaddr % sbi->blocks_per_blkz == sbi->blocks_per_blkz - 1);
 }
+
+static void f2fs_wait_zone_io_completion(struct f2fs_sb_info *sbi,
+			struct f2fs_bio_info *io, enum page_type btype)
+{
+	if (f2fs_sb_has_blkzoned(sbi) && btype < META && io->zone_pending_bio) {
+		wait_for_completion_io(&io->zone_wait);
+		bio_put(io->zone_pending_bio);
+		io->zone_pending_bio = NULL;
+		io->bi_private = NULL;
+	}
+}
+
+static void f2fs_submit_zone_io(struct f2fs_sb_info *sbi,
+			struct f2fs_io_info *fio, struct f2fs_bio_info *io,
+			enum page_type btype)
+{
+	if (f2fs_sb_has_blkzoned(sbi) && btype < META &&
+			is_end_zone_blkaddr(sbi, fio->new_blkaddr)) {
+		bio_get(io->bio);
+		reinit_completion(&io->zone_wait);
+		io->bi_private = io->bio->bi_private;
+		io->bio->bi_private = io;
+		io->bio->bi_end_io = f2fs_zone_write_end_io;
+		io->zone_pending_bio = io->bio;
+		__submit_merged_bio(io);
+	}
+
+}
 #endif
 
 void f2fs_submit_page_write(struct f2fs_io_info *fio)
@@ -1079,14 +1221,8 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
 	f2fs_down_write_trace(&io->io_rwsem, &lc);
 next:
 #ifdef CONFIG_BLK_DEV_ZONED
-	if (f2fs_sb_has_blkzoned(sbi) && btype < META && io->zone_pending_bio) {
-		wait_for_completion_io(&io->zone_wait);
-		bio_put(io->zone_pending_bio);
-		io->zone_pending_bio = NULL;
-		io->bi_private = NULL;
-	}
+	f2fs_wait_zone_io_completion(sbi, io, btype);
 #endif
-
 	if (fio->in_list) {
 		spin_lock(&io->io_lock);
 		if (list_empty(&io->io_list)) {
@@ -1141,16 +1277,109 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
 
 	trace_f2fs_submit_folio_write(fio->folio, fio);
 #ifdef CONFIG_BLK_DEV_ZONED
-	if (f2fs_sb_has_blkzoned(sbi) && btype < META &&
-			is_end_zone_blkaddr(sbi, fio->new_blkaddr)) {
-		bio_get(io->bio);
-		reinit_completion(&io->zone_wait);
-		io->bi_private = io->bio->bi_private;
-		io->bio->bi_private = io;
-		io->bio->bi_end_io = f2fs_zone_write_end_io;
-		io->zone_pending_bio = io->bio;
+	f2fs_submit_zone_io(sbi, fio, io, btype);
+#endif
+
+	if (fio->in_list)
+		goto next;
+out:
+	if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
+				!f2fs_is_checkpoint_ready(sbi))
 		__submit_merged_bio(io);
+	f2fs_up_write_trace(&io->io_rwsem, &lc);
+}
+
+static void f2fs_bio_add_cache(struct f2fs_io_info *fio, struct bio *bio)
+{
+	struct f2fs_bio *fbio = F2FS_BIO(bio);
+	struct f2fs_cached_block *head = fbio->entry;
+	struct f2fs_cached_block *new = fio->cache_entry;
+
+	new->next_entry = head;
+	fbio->entry = new;
+}
+
+int f2fs_submit_cache_read(struct f2fs_io_info *fio)
+{
+	struct f2fs_sb_info *sbi = fio->sbi;
+	struct f2fs_cached_block *entry = fio->cache_entry;
+	struct bio *bio;
+	enum count_type io_type = IS_META_CACHE(entry->cache) ?
+				F2FS_RD_META : F2FS_RD_NODE;
+
+	if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
+			fio->is_por ? META_POR : (__is_meta_io(fio) ?
+			META_GENERIC : DATA_GENERIC_ENHANCE)))
+		return -EFSCORRUPTED;
+
+	bio = __bio_alloc(fio, 1);
+
+	bio_add_virt_nofail(bio, cache_address(entry), sbi->blocksize);
+	f2fs_bio_add_cache(fio, bio);
+	inc_page_count(sbi, io_type);
+
+	f2fs_submit_read_bio(sbi, bio, fio->type);
+	return 0;
+}
+
+void f2fs_submit_cache_write(struct f2fs_io_info *fio)
+{
+	struct f2fs_sb_info *sbi = fio->sbi;
+	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
+	struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
+	struct f2fs_lock_context lc;
+	struct folio *folio;
+
+	f2fs_bug_on(sbi, is_read_io(fio->op));
+
+	f2fs_down_write_trace(&io->io_rwsem, &lc);
+next:
+#ifdef CONFIG_BLK_DEV_ZONED
+	f2fs_wait_zone_io_completion(sbi, io, btype);
+#endif
+	if (fio->in_list) {
+		spin_lock(&io->io_lock);
+		if (list_empty(&io->io_list)) {
+			spin_unlock(&io->io_lock);
+			goto out;
+		}
+		fio = list_first_entry(&io->io_list,
+						struct f2fs_io_info, list);
+		list_del(&fio->list);
+		spin_unlock(&io->io_lock);
 	}
+
+	verify_fio_blkaddr(fio);
+
+	fio->submitted = 1;
+	inc_page_count(sbi, F2FS_WB_CP_DATA);
+
+	if (io->bio &&
+	    (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
+			      fio->new_blkaddr)))
+		__submit_merged_bio(io);
+alloc_new:
+	if (io->bio == NULL) {
+		io->bio = __bio_alloc(fio, BIO_MAX_VECS);
+		io->fio = *fio;
+	}
+
+	folio = cache_folio(fio->cache_entry);
+
+	if (!bio_add_folio(io->bio, folio, sbi->blocksize,
+			offset_in_folio(folio, cache_address(fio->cache_entry)))) {
+		f2fs_bug_on(sbi, !F2FS_BIO(io->bio)->entry);
+
+		__submit_merged_bio(io);
+		goto alloc_new;
+	}
+
+	f2fs_bio_add_cache(fio, io->bio);
+
+	io->last_block_in_bio = fio->new_blkaddr;
+
+#ifdef CONFIG_BLK_DEV_ZONED
+	f2fs_submit_zone_io(sbi, fio, io, btype);
 #endif
 	if (fio->in_list)
 		goto next;
fs/f2fs/f2fs.h +24 -0
index 1b96d8718c5c..8413983ea9d5 100644
@@ -221,6 +221,8 @@ struct f2fs_rwsem {
 #endif
 };
 
+#include "cache.h"
+
 struct f2fs_mount_info {
 	unsigned long long opt;
 	block_t root_reserved_blocks;	/* root reserved blocks */
@@ -1370,8 +1372,10 @@ struct f2fs_io_info {
 	unsigned int is_por:1;		/* indicate IO is from recovery or not */
 	unsigned int encrypted:1;	/* indicate file is encrypted */
 	unsigned int meta_gc:1;		/* require meta inode GC */
+	unsigned int is_cache:1;	/* indicate IO is from internal cache */
 	enum iostat_type io_type;	/* io type */
 	struct writeback_control *io_wbc; /* writeback control */
+	struct f2fs_cached_block *cache_entry;
 	struct bio **bio;		/* bio for ipu */
 	sector_t *last_block;		/* last block number in bio */
 };
@@ -1781,6 +1785,12 @@ struct f2fs_gc_kthread {
 	unsigned int boost_gc_greedy;
 };
 
+struct f2fs_bio {
+	struct work_struct work;
+	struct f2fs_cached_block *entry;
+	struct bio bio;
+};
+
 struct f2fs_sb_info {
 	struct super_block *sb;			/* pointer to VFS super block */
 	struct proc_dir_entry *s_proc;		/* proc entry */
@@ -2323,6 +2333,16 @@ static inline bool is_node_folio(struct folio *folio)
 	return folio->mapping == NODE_MAPPING(F2FS_F_SB(folio));
 }
 
+static inline struct f2fs_bio *F2FS_BIO(struct bio *bio)
+{
+	return container_of(bio, struct f2fs_bio, bio);
+}
+
+static inline bool f2fs_is_cache_bio(struct bio *bio)
+{
+	return F2FS_BIO(bio)->entry != NULL;
+}
+
 static inline bool is_sbi_flag_set(struct f2fs_sb_info *sbi, unsigned int type)
 {
 	return test_bit(type, &sbi->s_flag);
@@ -4237,6 +4257,8 @@ void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
 				nid_t ino, enum page_type type);
 void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
 				struct folio *folio, enum page_type type);
+bool f2fs_submit_merged_write_cache(struct f2fs_cached_block *entry,
+				enum page_type type);
 void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
 					struct bio **bio, struct folio *folio);
 void f2fs_submit_all_merged_ipu_writes(struct f2fs_sb_info *sbi);
@@ -4244,6 +4266,8 @@ void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi);
 int f2fs_submit_page_bio(struct f2fs_io_info *fio);
 int f2fs_merge_page_bio(struct f2fs_io_info *fio);
 void f2fs_submit_page_write(struct f2fs_io_info *fio);
+int f2fs_submit_cache_read(struct f2fs_io_info *fio);
+void f2fs_submit_cache_write(struct f2fs_io_info *fio);
 struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
 		block_t blk_addr, sector_t *sector);
 int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr);
fs/f2fs/iostat.h +11 -0
index 2025225b5bed..61c6bc8e3119 100644
@@ -60,6 +60,13 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 	return iostat_ctx->post_read_ctx;
 }
 
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
+
+	iostat_ctx->post_read_ctx = ctx;
+}
+
 extern void iostat_update_and_unbind_ctx(struct bio *bio);
 extern void iostat_alloc_and_bind_ctx(struct f2fs_sb_info *sbi,
 		struct bio *bio, struct bio_post_read_ctx *ctx);
@@ -81,6 +88,10 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
 {
 	return bio->bi_private;
 }
+static inline void iostat_set_post_read_ctx(struct bio *bio, void *ctx)
+{
+	bio->bi_private = ctx;
+}
 static inline int f2fs_init_iostat_processing(void) { return 0; }
 static inline void f2fs_destroy_iostat_processing(void) {}
 static inline int f2fs_init_iostat(struct f2fs_sb_info *sbi) { return 0; }
Signature · 8 lines
2.49.0



_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
#4 lore raw

Re: [PATCH v1 01/12] f2fs: cache: implement metadata cache

In reply to Chao Yu

Hi Chao,

Quoted 9 lines

static void f2fs_zone_write_end_io(struct bio *bio)

{
	bio->bi_private = io->bi_private;
	complete(&io->zone_wait);
	if (io->fio.is_cache)
		f2fs_cache_write_end_io(bio);
	else
		f2fs_write_end_io(bio);
}

io->fio is updated when the next BIO is prepared. Once complete(&io->zone_wait) wakes the waiting writer, that preparation can run before this callback reads io->fio.is_cache.

Is there another serialization condition that keeps io->fio.is_cache unchanged between complete() and this read?

author maintainer
#5 lore raw

Re: [PATCH v1 01/12] f2fs: cache: implement metadata cache

Quoted 12 lines · Wenjie Qi

Hi Chao,

Quoted 9 lines

static void f2fs_zone_write_end_io(struct bio *bio)

{
	bio->bi_private = io->bi_private;
	complete(&io->zone_wait);
	if (io->fio.is_cache)
		f2fs_cache_write_end_io(bio);
	else
		f2fs_write_end_io(bio);
}
Quoted 6 lines

io->fio is updated when the next BIO is prepared. Once complete(&io->zone_wait) wakes the waiting writer, that preparation can run before this callback reads io->fio.is_cache.

Is there another serialization condition that keeps io->fio.is_cache unchanged between complete() and this read?

Wenjie,

Thanks for the review, will fix.

Thanks,

author maintainer
#6 lore raw

PATCH 2/12 f2fs: cache: initialize meta cache

In reply to Chao Yu

This patch introduces meta_blocks in f2fs_sb_info structure, initializes and destroys the meta cache during filesystem mount and unmount.

It also introduces helper wrappers for meta cache operation.

---
 fs/f2fs/cache.h | 12 ++++++++++++
 fs/f2fs/data.c  |  3 +++
 fs/f2fs/f2fs.h  |  3 +++
 fs/f2fs/super.c | 10 +++++++++-
 4 files changed, 27 insertions(+), 1 deletion(-)
fs/f2fs/cache.h +12 -0
index 686a974008ae..3cd2abafe364 100644
@@ -176,4 +176,16 @@ struct f2fs_cached_block *f2fs_grab_cache(struct f2fs_cached_block_list *cache,
 void f2fs_drop_cache_range(struct f2fs_cached_block_list *cache,
 		unsigned long start, unsigned long len, bool drop_dirty);
 
+int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi);
+void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
+
+#define META_CACHE(sbi)		(&(sbi)->meta_blocks)
+
+#define f2fs_find_meta_cache(sbi, blkaddr)		\
+	f2fs_find_cache(META_CACHE(sbi), blkaddr)
+#define f2fs_invalidate_meta_caches(sbi, start, len)	\
+	f2fs_drop_cache_range(META_CACHE(sbi), start, len, false)
+#define f2fs_truncate_meta_caches(sbi, start, len)	\
+	f2fs_drop_cache_range(META_CACHE(sbi), start, len, true)
+
 #endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/data.c +3 -0
index 09474790035b..110282bb8dcd 100644
@@ -695,6 +695,9 @@ static bool __has_merged_page(struct bio *bio, struct inode *inode,
 	if (!inode && !folio && !ino)
 		return true;
 
+	if (f2fs_is_cache_bio(bio))
+		return false;
+
 	bio_for_each_folio_all(fi, bio) {
 		struct folio *target = fi.folio;
 
fs/f2fs/f2fs.h +3 -0
index 8413983ea9d5..9b7c60bbb137 100644
@@ -2105,6 +2105,9 @@ struct f2fs_sb_info {
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 	struct lock_class_key cp_global_sem_key;
 #endif
+
+	/* f2fs internal cache */
+	struct f2fs_cached_block_list meta_blocks;
 };
 
 /* Definitions to access f2fs_sb_info */
fs/f2fs/super.c +9 -1
index 3bdb0f891c35..bcfb1f97850e 100644
@@ -2053,6 +2053,8 @@ static void f2fs_put_super(struct super_block *sb)
 	iput(sbi->meta_inode);
 	sbi->meta_inode = NULL;
 
+	f2fs_destroy_cache(META_CACHE(sbi));
+
 	/* Should check the page counts after dropping all node/meta pages */
 	for (i = 0; i < NR_COUNT_TYPE; i++) {
 		if (!get_pages(sbi, i))
@@ -5197,12 +5199,16 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (err)
 		goto free_percpu;
 
+	err = f2fs_init_cache(sbi, META_CACHE(sbi), F2FS_META_CACHE);
+	if (err)
+		goto free_page_array_cache;
+
 	/* get an inode for meta space */
 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 	if (IS_ERR(sbi->meta_inode)) {
 		f2fs_err(sbi, "Failed to read F2FS meta data inode");
 		err = PTR_ERR(sbi->meta_inode);
-		goto free_page_array_cache;
+		goto free_meta_cache;
 	}
 
 	err = f2fs_get_valid_checkpoint(sbi);
@@ -5528,6 +5534,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	make_bad_inode(sbi->meta_inode);
 	iput(sbi->meta_inode);
 	sbi->meta_inode = NULL;
+free_meta_cache:
+	f2fs_destroy_cache(META_CACHE(sbi));
 free_page_array_cache:
 	f2fs_destroy_page_array_cache(sbi);
 free_percpu:
Signature · 1 line
2.49.0
author maintainer
#7 lore raw

PATCH 3/12 f2fs: cache: introduce shrinker

In reply to Chao Yu

This patch integrates the metadata cache into the F2FS memory shrinker subsystem to reclaim clean, unreferenced cached blocks under memory pressure.

It implements f2fs_shrink_cache() using a 3-phase cache reclamin method:
1. isolate clean entries from lru list
2. truncate from radix tree under lock
3. splice un-reclaimed entries back

And hooks the new interface into f2fs_shrink_count() and f2fs_shrink_scan().

---
 fs/f2fs/cache.c    | 82 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h    |  3 ++
 fs/f2fs/shrinker.c | 12 +++++++
 3 files changed, 97 insertions(+)
fs/f2fs/cache.c +82 -0
index 08bc658166f7..c071364822c4 100644
@@ -529,3 +529,85 @@ void f2fs_destroy_cache(struct f2fs_cached_block_list *cache)
 	f2fs_put_cache(entry, true);
 	goto next;
 }
+
+static unsigned long f2fs_do_shrink_cache(struct f2fs_cached_block_list *cache,
+						unsigned long nr_to_scan)
+{
+	struct f2fs_cached_block *entry, *next;
+	LIST_HEAD(dispose_list);
+	LIST_HEAD(keep_list);
+	unsigned long freed = 0;
+	unsigned long isolated = 0;
+
+	/* Phase 1: Isolate candidate entries from LRU list into dispose_list */
+	spin_lock(&cache->list_lock);
+	list_for_each_entry_safe(entry, next, &cache->lru_list, list) {
+		if (isolated >= nr_to_scan)
+			break;
+
+		if (f2fs_cache_test_dirty(entry) ||
+		    f2fs_cache_test_writeback(entry) ||
+		    f2fs_cache_test_locked(entry))
+			continue;
+
+		if (f2fs_cache_refcount(entry) != 1)
+			continue;
+
+		list_move_tail(&entry->list, &dispose_list);
+		isolated++;
+	}
+	spin_unlock(&cache->list_lock);
+
+	/* Phase 2: Process isolated candidates one by one */
+	while (1) {
+		spin_lock(&cache->list_lock);
+		entry = list_first_entry_or_null(&dispose_list,
+						struct f2fs_cached_block, list);
+		if (!entry) {
+			spin_unlock(&cache->list_lock);
+			break;
+		}
+		f2fs_cache_get(entry);
+		list_move_tail(&entry->list, &keep_list);
+		spin_unlock(&cache->list_lock);
+
+		if (!f2fs_trylock_cache(entry)) {
+			f2fs_put_cache(entry, false);
+			continue;
+		}
+
+		/* the entry has been truncated */
+		if (!entry->cache) {
+			f2fs_put_cache(entry, true);
+			continue;
+		}
+		/*
+		 * at least there are shrinker, radix tree and another user
+		 * has referenced the entry.
+		 */
+		if (f2fs_cache_refcount(entry) >= 3) {
+			f2fs_put_cache(entry, true);
+			continue;
+		}
+
+		f2fs_do_truncate_cache(entry, false);
+
+		if (f2fs_put_cache(entry, true))
+			freed++;
+	}
+
+	/* Phase 3: Splice un-reclaimed entries back onto cache->lru_list */
+	if (!list_empty(&keep_list)) {
+		spin_lock(&cache->list_lock);
+		list_splice_tail(&keep_list, &cache->lru_list);
+		spin_unlock(&cache->list_lock);
+	}
+
+	return freed;
+}
+
+unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
+					unsigned long nr_to_scan)
+{
+	return f2fs_do_shrink_cache(META_CACHE(sbi), nr_to_scan);
+}
fs/f2fs/cache.h +3 -0
index 3cd2abafe364..d92b6f3ed585 100644
@@ -188,4 +188,7 @@ void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 #define f2fs_truncate_meta_caches(sbi, start, len)	\
 	f2fs_drop_cache_range(META_CACHE(sbi), start, len, true)
 
+unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
+				unsigned long nr_to_scan);
+
 #endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/shrinker.c +12 -0
index 4f6bf5926de4..1755c85849e4 100644
@@ -37,6 +37,11 @@ static unsigned long __count_extent_cache(struct f2fs_sb_info *sbi,
 				atomic_read(&eti->total_ext_node);
 }
 
+static unsigned long __count_cache(struct f2fs_sb_info *sbi)
+{
+	return sbi->meta_blocks.num_entries;
+}
+
 unsigned long f2fs_shrink_count(struct shrinker *shrink,
 				struct shrink_control *sc)
 {
@@ -68,6 +73,9 @@ unsigned long f2fs_shrink_count(struct shrinker *shrink,
 		/* count free nids cache entries */
 		count += __count_free_nids(sbi);
 
+		/* count generic cache entries */
+		count += __count_cache(sbi);
+
 		spin_lock(&f2fs_list_lock);
 		p = p->next;
 		mutex_unlock(&sbi->umount_mutex);
@@ -120,6 +128,10 @@ unsigned long f2fs_shrink_scan(struct shrinker *shrink,
 		if (freed < nr)
 			freed += f2fs_try_to_free_nids(sbi, nr - freed);
 
+		/* shrink generic cache entries */
+		if (freed < nr)
+			freed += f2fs_shrink_cache(sbi, nr - freed);
+
 		spin_lock(&f2fs_list_lock);
 		p = p->next;
 		list_move_tail(&sbi->s_list, &f2fs_list);
Signature · 1 line
2.49.0
author maintainer
#8 lore raw

PATCH 4/12 f2fs: cache: introduce writeback thread

In reply to Chao Yu

This patch introduces a background writeback kthread (f2fs_writeback-x:y) to periodically flush dirty metadata cache entries with a default interval of 5 seconds.

It manages thread lifecycle across mount, unmount, and remount (rw/ro) transitions, and hooks synchronous flushing into checkpoint commits.

---
 fs/f2fs/cache.c      | 54 ++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h      | 13 +++++++++++
 fs/f2fs/checkpoint.c |  1 +
 fs/f2fs/f2fs.h       |  3 +++
 fs/f2fs/super.c      | 31 ++++++++++++++++++++++++-
 5 files changed, 101 insertions(+), 1 deletion(-)
fs/f2fs/cache.c +54 -0
index c071364822c4..cb5b26046162 100644
@@ -611,3 +611,57 @@ unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 {
 	return f2fs_do_shrink_cache(META_CACHE(sbi), nr_to_scan);
 }
+
+static int f2fs_cache_writeback_kthread(void *data)
+{
+	struct f2fs_sb_info *sbi = data;
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+	wait_queue_head_t *wq = &cache_thread->cache_wb_wq;
+	unsigned int interval = DEF_DIRTY_CACHE_TIMEOUT;
+
+	set_freezable();
+
+	while (!kthread_should_stop()) {
+		wait_event_freezable_timeout(*wq,
+				kthread_should_stop() ||
+				cache_thread->cache_wb_task == NULL,
+				msecs_to_jiffies(interval));
+
+		if (kthread_should_stop())
+			break;
+		if (f2fs_cp_error(sbi))
+			continue;
+	}
+	return 0;
+}
+
+int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+	dev_t dev = sbi->sb->s_dev;
+	char name[36];
+
+	if (cache_thread->cache_wb_task)
+		return 0;
+
+	init_waitqueue_head(&cache_thread->cache_wb_wq);
+	snprintf(name, sizeof(name), "f2fs_writeback-%u:%u",
+			MAJOR(dev), MINOR(dev));
+
+	cache_thread->cache_wb_task = kthread_run(f2fs_cache_writeback_kthread,
+						sbi, "%s", name);
+	if (IS_ERR(cache_thread->cache_wb_task))
+		return PTR_ERR(cache_thread->cache_wb_task);
+	return 0;
+}
+
+void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+
+	if (!cache_thread->cache_wb_task)
+		return;
+
+	kthread_stop(cache_thread->cache_wb_task);
+	cache_thread->cache_wb_task = NULL;
+}
fs/f2fs/cache.h +13 -0
index d92b6f3ed585..5e9fe8c0b15c 100644
@@ -191,4 +191,17 @@ void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 				unsigned long nr_to_scan);
 
+#define DEF_DIRTY_CACHE_TIMEOUT 5000
+
+struct f2fs_cache_kthread {
+	struct task_struct *cache_wb_task;
+	wait_queue_head_t cache_wb_wq;
+	atomic_t cache_wb_trigger;
+	unsigned int cache_wb_interval_ms;
+};
+
+int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi);
+void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
+void f2fs_sync_cache_wb(struct f2fs_sb_info *sbi);
+
 #endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/checkpoint.c +1 -0
index f81ba8cc861a..729d19680caf 100644
@@ -1596,6 +1596,7 @@ static int block_operations(struct f2fs_sb_info *sbi)
 	 * sbi->node_change is used only for AIO write_begin path which produces
 	 * dirty node blocks and some checkpoint values by block allocation.
 	 */
+
 	__prepare_cp_block(sbi);
 	f2fs_up_write(&sbi->node_change);
 	return err;
fs/f2fs/f2fs.h +3 -0
index 9b7c60bbb137..6e20b3586f26 100644
@@ -2108,6 +2108,9 @@ struct f2fs_sb_info {
 
 	/* f2fs internal cache */
 	struct f2fs_cached_block_list meta_blocks;
+
+	/* internal cache flush thread */
+	struct f2fs_cache_kthread cache_thread;
 };
 
 /* Definitions to access f2fs_sb_info */
fs/f2fs/super.c +30 -1
index bcfb1f97850e..89affe72f4fc 100644
@@ -1998,6 +1998,7 @@ static void f2fs_put_super(struct super_block *sb)
 	 * flush all issued checkpoints and stop checkpoint issue thread.
 	 * after then, all checkpoints should be done by each process context.
 	 */
+	f2fs_stop_cache_wb_thread(sbi);
 	f2fs_stop_ckpt_thread(sbi);
 
 	/*
@@ -2799,6 +2800,7 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 	unsigned int flags = fc->sb_flags;
 	int err;
 	bool need_restart_gc = false, need_stop_gc = false;
+	bool need_restart_wb = false, need_stop_wb = false;
 	bool need_restart_flush = false, need_stop_flush = false;
 	bool need_restart_discard = false, need_stop_discard = false;
 	bool need_enable_checkpoint = false, need_disable_checkpoint = false;
@@ -2957,6 +2959,18 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 		need_stop_gc = true;
 	}
 
+	if (flags & SB_RDONLY) {
+		if (sbi->cache_thread.cache_wb_task) {
+			f2fs_stop_cache_wb_thread(sbi);
+			need_restart_wb = true;
+		}
+	} else if (!sbi->cache_thread.cache_wb_task) {
+		err = f2fs_start_cache_wb_thread(sbi);
+		if (err)
+			goto restore_gc;
+		need_stop_wb = true;
+	}
+
 	if (flags & SB_RDONLY) {
 		sync_inodes_sb(sb);
 
@@ -2979,7 +2993,7 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 	} else {
 		err = f2fs_create_flush_cmd_control(sbi);
 		if (err)
-			goto restore_gc;
+			goto restore_wb;
 		need_stop_flush = true;
 	}
 
@@ -3075,6 +3089,13 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 		clear_opt(sbi, FLUSH_MERGE);
 		f2fs_destroy_flush_cmd_control(sbi, false);
 	}
+restore_wb:
+	if (need_restart_wb) {
+		if (f2fs_start_cache_wb_thread(sbi))
+			f2fs_warn(sbi, "background cache writeback thread has stopped");
+	} else if (need_stop_wb) {
+		f2fs_stop_cache_wb_thread(sbi);
+	}
 restore_gc:
 	if (need_restart_gc) {
 		if (f2fs_start_gc_thread(sbi))
@@ -5461,6 +5482,12 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 			goto sync_free_meta;
 	}
 
+	if (!f2fs_readonly(sb)) {
+		err = f2fs_start_cache_wb_thread(sbi);
+		if (err)
+			goto stop_gc_thread;
+	}
+
 	/* recover broken superblock */
 	if (recovery) {
 		err = f2fs_commit_super(sbi, true);
@@ -5483,6 +5510,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	sbi->umount_lock_holder = NULL;
 	return 0;
 
+stop_gc_thread:
+	f2fs_stop_gc_thread(sbi);
 sync_free_meta:
 	/* safe to flush all the data */
 	sync_filesystem(sbi->sb);
Signature · 1 line
2.49.0
maintainer
#9 lore raw

Re: [f2fs-dev] [PATCH v1 04/12] f2fs: cache: introduce writeback thread

Quoted 105 lines · On 08/20, Chao Yu via Linux-f2fs-devel wrote:

This patch introduces a background writeback kthread (f2fs_writeback-x:y) to periodically flush dirty metadata cache entries with a default interval of 5 seconds.

It manages thread lifecycle across mount, unmount, and remount (rw/ro) transitions, and hooks synchronous flushing into checkpoint commits.

---
 fs/f2fs/cache.c      | 54 ++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/cache.h      | 13 +++++++++++
 fs/f2fs/checkpoint.c |  1 +
 fs/f2fs/f2fs.h       |  3 +++
 fs/f2fs/super.c      | 31 ++++++++++++++++++++++++-
 5 files changed, 101 insertions(+), 1 deletion(-)
fs/f2fs/cache.c +54 -0
index c071364822c4..cb5b26046162 100644
@@ -611,3 +611,57 @@ unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 {
 	return f2fs_do_shrink_cache(META_CACHE(sbi), nr_to_scan);
 }
+
+static int f2fs_cache_writeback_kthread(void *data)
+{
+	struct f2fs_sb_info *sbi = data;
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+	wait_queue_head_t *wq = &cache_thread->cache_wb_wq;
+	unsigned int interval = DEF_DIRTY_CACHE_TIMEOUT;
+
+	set_freezable();
+
+	while (!kthread_should_stop()) {
+		wait_event_freezable_timeout(*wq,
+				kthread_should_stop() ||
+				cache_thread->cache_wb_task == NULL,
+				msecs_to_jiffies(interval));
+
+		if (kthread_should_stop())
+			break;
+		if (f2fs_cp_error(sbi))
+			continue;
+	}
+	return 0;
+}
+
+int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+	dev_t dev = sbi->sb->s_dev;
+	char name[36];
+
+	if (cache_thread->cache_wb_task)
+		return 0;
+
+	init_waitqueue_head(&cache_thread->cache_wb_wq);
+	snprintf(name, sizeof(name), "f2fs_writeback-%u:%u",
+			MAJOR(dev), MINOR(dev));
+
+	cache_thread->cache_wb_task = kthread_run(f2fs_cache_writeback_kthread,
+						sbi, "%s", name);
+	if (IS_ERR(cache_thread->cache_wb_task))
+		return PTR_ERR(cache_thread->cache_wb_task);
+	return 0;
+}
+
+void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_cache_kthread *cache_thread = &sbi->cache_thread;
+
+	if (!cache_thread->cache_wb_task)
+		return;
+
+	kthread_stop(cache_thread->cache_wb_task);
+	cache_thread->cache_wb_task = NULL;
+}
fs/f2fs/cache.h +13 -0
index d92b6f3ed585..5e9fe8c0b15c 100644
@@ -191,4 +191,17 @@ void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 				unsigned long nr_to_scan);
 
+#define DEF_DIRTY_CACHE_TIMEOUT 5000
+
+struct f2fs_cache_kthread {
+	struct task_struct *cache_wb_task;
+	wait_queue_head_t cache_wb_wq;
+	atomic_t cache_wb_trigger;
+	unsigned int cache_wb_interval_ms;
+};
+
+int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi);
+void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
+void f2fs_sync_cache_wb(struct f2fs_sb_info *sbi);
+
 #endif /* _LINUX_F2FS_CACHE_H */
fs/f2fs/checkpoint.c +0 -0
index f81ba8cc861a..729d19680caf 100644

fs/f2fs/checkpoint.c static int block_operations(struct f2fs_sb_info *sbi) · line 1599

Quoted 4 lines
fs/f2fs/checkpoint.c +1 -0
@@ -1596,6 +1596,7 @@ static int block_operations(struct f2fs_sb_info *sbi)
 	 * sbi->node_change is used only for AIO write_begin path which produces
 	 * dirty node blocks and some checkpoint values by block allocation.
 	 */
+

Unnecessary line.

Quoted 110 lines
fs/f2fs/checkpoint.c +0 -0
 	__prepare_cp_block(sbi);
 	f2fs_up_write(&sbi->node_change);
 	return err;
fs/f2fs/f2fs.h +3 -0
index 9b7c60bbb137..6e20b3586f26 100644
@@ -2108,6 +2108,9 @@ struct f2fs_sb_info {
 
 	/* f2fs internal cache */
 	struct f2fs_cached_block_list meta_blocks;
+
+	/* internal cache flush thread */
+	struct f2fs_cache_kthread cache_thread;
 };
 
 /* Definitions to access f2fs_sb_info */
fs/f2fs/super.c +30 -1
index bcfb1f97850e..89affe72f4fc 100644
@@ -1998,6 +1998,7 @@ static void f2fs_put_super(struct super_block *sb)
 	 * flush all issued checkpoints and stop checkpoint issue thread.
 	 * after then, all checkpoints should be done by each process context.
 	 */
+	f2fs_stop_cache_wb_thread(sbi);
 	f2fs_stop_ckpt_thread(sbi);
 
 	/*
@@ -2799,6 +2800,7 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 	unsigned int flags = fc->sb_flags;
 	int err;
 	bool need_restart_gc = false, need_stop_gc = false;
+	bool need_restart_wb = false, need_stop_wb = false;
 	bool need_restart_flush = false, need_stop_flush = false;
 	bool need_restart_discard = false, need_stop_discard = false;
 	bool need_enable_checkpoint = false, need_disable_checkpoint = false;
@@ -2957,6 +2959,18 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 		need_stop_gc = true;
 	}
 
+	if (flags & SB_RDONLY) {
+		if (sbi->cache_thread.cache_wb_task) {
+			f2fs_stop_cache_wb_thread(sbi);
+			need_restart_wb = true;
+		}
+	} else if (!sbi->cache_thread.cache_wb_task) {
+		err = f2fs_start_cache_wb_thread(sbi);
+		if (err)
+			goto restore_gc;
+		need_stop_wb = true;
+	}
+
 	if (flags & SB_RDONLY) {
 		sync_inodes_sb(sb);
 
@@ -2979,7 +2993,7 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 	} else {
 		err = f2fs_create_flush_cmd_control(sbi);
 		if (err)
-			goto restore_gc;
+			goto restore_wb;
 		need_stop_flush = true;
 	}
 
@@ -3075,6 +3089,13 @@ static int __f2fs_remount(struct fs_context *fc, struct super_block *sb)
 		clear_opt(sbi, FLUSH_MERGE);
 		f2fs_destroy_flush_cmd_control(sbi, false);
 	}
+restore_wb:
+	if (need_restart_wb) {
+		if (f2fs_start_cache_wb_thread(sbi))
+			f2fs_warn(sbi, "background cache writeback thread has stopped");
+	} else if (need_stop_wb) {
+		f2fs_stop_cache_wb_thread(sbi);
+	}
 restore_gc:
 	if (need_restart_gc) {
 		if (f2fs_start_gc_thread(sbi))
@@ -5461,6 +5482,12 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 			goto sync_free_meta;
 	}
 
+	if (!f2fs_readonly(sb)) {
+		err = f2fs_start_cache_wb_thread(sbi);
+		if (err)
+			goto stop_gc_thread;
+	}
+
 	/* recover broken superblock */
 	if (recovery) {
 		err = f2fs_commit_super(sbi, true);
@@ -5483,6 +5510,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	sbi->umount_lock_holder = NULL;
 	return 0;
 
+stop_gc_thread:
+	f2fs_stop_gc_thread(sbi);
 sync_free_meta:
 	/* safe to flush all the data */
 	sync_filesystem(sbi->sb);
Signature · 8 lines
2.49.0



_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
#10 lore raw

Re: [PATCH v1 04/12] f2fs: cache: introduce writeback thread

In reply to Chao Yu

Hi Chao,

Quoted 4 lines

cache_thread->cache_wb_task = kthread_run(f2fs_cache_writeback_kthread,

						sbi, "%s", name);
if (IS_ERR(cache_thread->cache_wb_task))
	return PTR_ERR(cache_thread->cache_wb_task);

After kthread_run() returns an error pointer, cache_wb_task retains that value. The later startup and stop paths distinguish only NULL from non-NULL, and the stop path passes a non-NULL value to kthread_stop().

Is cache_wb_task expected to remain an error pointer after f2fs_start_cache_wb_thread() returns an error?

author maintainer
#11 lore raw

Re: [PATCH v1 04/12] f2fs: cache: introduce writeback thread

Quoted 7 lines · Wenjie Qi

Hi Chao,

Quoted 4 lines

cache_thread->cache_wb_task = kthread_run(f2fs_cache_writeback_kthread,

						sbi, "%s", name);
if (IS_ERR(cache_thread->cache_wb_task))
	return PTR_ERR(cache_thread->cache_wb_task);
Quoted 6 lines

After kthread_run() returns an error pointer, cache_wb_task retains that value. The later startup and stop paths distinguish only NULL from non-NULL, and the stop path passes a non-NULL value to kthread_stop().

Is cache_wb_task expected to remain an error pointer after f2fs_start_cache_wb_thread() returns an error?

Will fix.

Thanks,

author maintainer
#12 lore raw

PATCH 5/12 f2fs: cache: use meta cache

In reply to Chao Yu

This patch migrates F2FS meta block caching from the fake VFS inode page cache (sbi->meta_inode) to meta cache (sbi->meta_blocks).

It converts CP, SIT, NAT, SSA, recovery, and GC metadata I/O paths to operate on struct f2fs_cached_block instead of folio, and removes sbi->meta_inode.

---
 fs/f2fs/cache.c         |   2 +
 fs/f2fs/checkpoint.c    | 394 ++++++++++++++++++----------------------
 fs/f2fs/compress.c      |   4 +-
 fs/f2fs/data.c          |  29 ++-
 fs/f2fs/debug.c         |  12 +-
 fs/f2fs/f2fs.h          |  81 +++------
 fs/f2fs/file.c          |   4 +-
 fs/f2fs/gc.c            | 134 +++++++-------
 fs/f2fs/inline.c        |   9 +-
 fs/f2fs/inode.c         |   9 +-
 fs/f2fs/node.c          | 125 +++++++------
 fs/f2fs/node.h          |   4 +-
 fs/f2fs/recovery.c      | 198 ++++++++++----------
 fs/f2fs/segment.c       | 201 ++++++++++----------
 fs/f2fs/segment.h       |  31 +++-
 fs/f2fs/super.c         |  31 +---
 include/linux/f2fs_fs.h |   1 -
 17 files changed, 606 insertions(+), 663 deletions(-)
fs/f2fs/cache.c +2 -0
index cb5b26046162..afef808e485a 100644
@@ -631,6 +631,8 @@ static int f2fs_cache_writeback_kthread(void *data)
 			break;
 		if (f2fs_cp_error(sbi))
 			continue;
+
+		f2fs_write_meta_caches(sbi);
 	}
 	return 0;
 }
fs/f2fs/checkpoint.c +181 -213

770 lines not on this page · raw message

fs/f2fs/compress.c +2 -2
index 91855d91bbdd..bb749f6257a1 100644
@@ -1150,7 +1150,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 		f2fs_compress_ctx_add_page(cc, folio);
 
 		if (!folio_test_uptodate(folio)) {
-			f2fs_handle_page_eio(sbi, folio, DATA);
+			f2fs_handle_page_eio(sbi, folio->index, DATA);
 release_and_retry:
 			f2fs_put_rpages(cc);
 			f2fs_unlock_rpages(cc, i + 1);
@@ -1359,7 +1359,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 		fio.old_blkaddr = data_blkaddr(dn.inode, dn.node_folio,
 						dn.ofs_in_node + i + 1);
 
-		/* wait for GCed page writeback via META_MAPPING */
+		/* wait for GCed page writeback via generic cache */
 		f2fs_wait_on_block_writeback(inode, fio.old_blkaddr);
 
 		if (fio.encrypted) {
fs/f2fs/data.c +12 -17
index 110282bb8dcd..92c3293f0a1e 100644
@@ -65,8 +65,7 @@ bool f2fs_is_cp_guaranteed(const struct folio *folio)
 
 	inode = mapping->host;
 
-	if (inode->i_ino == F2FS_META_INO(sbi) ||
-			inode->i_ino == F2FS_NODE_INO(sbi) ||
+	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 			S_ISDIR(inode->i_mode))
 		return true;
 
@@ -84,9 +83,6 @@ static enum count_type __read_io_type(struct folio *folio)
 		struct inode *inode = mapping->host;
 		struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-		if (inode->i_ino == F2FS_META_INO(sbi))
-			return F2FS_RD_META;
-
 		if (inode->i_ino == F2FS_NODE_INO(sbi))
 			return F2FS_RD_NODE;
 	}
@@ -1452,7 +1448,7 @@ static void f2fs_submit_page_read(struct inode *inode, struct fsverity_info *vi,
 	bio = f2fs_grab_read_bio(inode, vi, blkaddr, 1, op_flags, folio->index,
 				 for_write);
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	f2fs_wait_on_block_writeback(inode, blkaddr);
 
 	if (!bio_add_folio(bio, folio, PAGE_SIZE, 0))
@@ -3110,7 +3106,7 @@ static void f2fs_readahead(struct readahead_control *rac)
 int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
 {
 	struct inode *inode = fio_inode(fio);
-	struct folio *mfolio;
+	struct f2fs_cached_block *entry;
 	struct page *page;
 
 	if (!f2fs_encrypted_file(inode))
@@ -3126,12 +3122,13 @@ int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
 	if (IS_ERR(fio->encrypted_page))
 		return PTR_ERR(fio->encrypted_page);
 
-	mfolio = filemap_lock_folio(META_MAPPING(fio->sbi), fio->old_blkaddr);
-	if (!IS_ERR(mfolio)) {
-		if (folio_test_uptodate(mfolio))
-			memcpy(folio_address(mfolio),
-				page_address(fio->encrypted_page), PAGE_SIZE);
-		f2fs_folio_put(mfolio, true);
+	entry = f2fs_find_cache(META_CACHE(fio->sbi), fio->old_blkaddr);
+	if (!IS_ERR(entry)) {
+		f2fs_lock_cache(entry);
+		if (f2fs_cache_test_uptodate(entry))
+			memcpy(cache_address(entry), page_address(fio->encrypted_page),
+						F2FS_BLKSIZE);
+		f2fs_put_cache(entry, true);
 	}
 	return 0;
 }
@@ -3297,7 +3294,7 @@ int f2fs_do_write_data_page(struct f2fs_io_info *fio)
 		goto out_writepage;
 	}
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	if (fio->meta_gc)
 		f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
 
@@ -4380,9 +4377,7 @@ void f2fs_invalidate_folio(struct folio *folio, size_t offset, size_t length)
 		return;
 
 	if (folio_test_dirty(folio)) {
-		if (inode->i_ino == F2FS_META_INO(sbi)) {
-			dec_page_count(sbi, F2FS_DIRTY_META);
-		} else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
+		if (inode->i_ino == F2FS_NODE_INO(sbi)) {
 			dec_page_count(sbi, F2FS_DIRTY_NODES);
 		} else {
 			inode_dec_dirty_pages(inode);
fs/f2fs/debug.c +4 -8
index ff379aff4472..14059a50739c 100644
@@ -224,8 +224,7 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->dirty_count = dirty_segments(sbi);
 	if (sbi->node_inode)
 		si->node_pages = NODE_MAPPING(sbi)->nrpages;
-	if (sbi->meta_inode)
-		si->meta_pages = META_MAPPING(sbi)->nrpages;
+	si->meta_caches = META_CACHE(sbi)->num_entries;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
 		si->compress_pages = COMPRESS_MAPPING(sbi)->nrpages;
@@ -388,11 +387,8 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 
 		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
 	}
-	if (sbi->meta_inode) {
-		unsigned long npages = META_MAPPING(sbi)->nrpages;
-
-		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
-	}
+	si->page_mem += (unsigned long long)META_CACHE(sbi)->num_entries << PAGE_SHIFT;
+	si->cache_mem += META_CACHE(sbi)->num_entries * sizeof(struct f2fs_cached_block);
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
 		unsigned long npages = COMPRESS_MAPPING(sbi)->nrpages;
@@ -708,7 +704,7 @@ static int stat_show(struct seq_file *s, void *v)
 		seq_printf(s, "  - quota data: %4d in quota files:%4d\n",
 			   si->ndirty_qdata, si->nquota_files);
 		seq_printf(s, "  - meta: %4d in %4d\n",
-			   si->ndirty_meta, si->meta_pages);
+			   si->ndirty_meta, si->meta_caches);
 		seq_printf(s, "  - imeta: %4d\n",
 			   si->ndirty_imeta);
 		seq_printf(s, "  - fsync mark: %4lld\n",
fs/f2fs/f2fs.h +23 -58

172 lines not on this page · raw message

fs/f2fs/file.c +2 -2
index a54b3ab52f1a..92daa41dd96d 100644
@@ -214,7 +214,7 @@ static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
 
 	f2fs_folio_wait_writeback(folio, DATA, false, true);
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 
 	/*
@@ -2563,7 +2563,7 @@ int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
 		break;
 	case F2FS_GOING_DOWN_METAFLUSH:
-		f2fs_sync_meta_pages(sbi, LONG_MAX, FS_META_IO);
+		f2fs_sync_meta_caches(sbi, LONG_MAX, true, FS_META_IO);
 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
 		break;
 	case F2FS_GOING_DOWN_NEED_FSCK:
fs/f2fs/gc.c +73 -61

291 lines not on this page · raw message

fs/f2fs/inline.c +5 -4
index aec06fb4fd76..2156fb1fc57d 100644
@@ -294,7 +294,7 @@ int f2fs_write_inline_data(struct inode *inode, struct folio *folio)
 	return 0;
 }
 
-int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
+int f2fs_recover_inline_data(struct inode *inode, struct f2fs_cached_block *entry)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct f2fs_inode *ri = NULL;
@@ -308,12 +308,13 @@ int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
 	 *    x       o  -> remove data blocks, and then recover inline_data
 	 *    x       x  -> recover data blocks
 	 */
-	if (IS_INODE(nfolio))
-		ri = F2FS_INODE(nfolio);
+	if (IS_INODE(cache_folio(entry)))
+		ri = &CACHED_NODE(entry)->i;
 
 	if (f2fs_has_inline_data(inode) &&
 			ri && (ri->i_inline & F2FS_INLINE_DATA)) {
 		struct folio *ifolio;
+
 process_inline:
 		ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
 		if (IS_ERR(ifolio))
@@ -321,7 +322,7 @@ int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
 
 		f2fs_folio_wait_writeback(ifolio, NODE, true, true);
 
-		src_addr = inline_data_addr(inode, nfolio);
+		src_addr = inline_data_addr(inode, cache_folio(entry));
 		dst_addr = inline_data_addr(inode, ifolio);
 		memcpy(dst_addr, src_addr, MAX_INLINE_DATA(inode));
 
fs/f2fs/inode.c +2 -7
index bac1e360d966..c533da4d4d70 100644
@@ -577,7 +577,7 @@ static int do_read_inode(struct inode *inode)
 
 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
 {
-	if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
+	if (ino == F2FS_NODE_INO(sbi))
 		return true;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (test_opt(sbi, COMPRESS_CACHE) && ino == F2FS_COMPRESS_INO(sbi))
@@ -624,9 +624,6 @@ struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 	if (ino == F2FS_NODE_INO(sbi)) {
 		inode->i_mapping->a_ops = &f2fs_node_aops;
 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
-	} else if (ino == F2FS_META_INO(sbi)) {
-		inode->i_mapping->a_ops = &f2fs_meta_aops;
-		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 		inode->i_mapping->a_ops = &f2fs_compress_aops;
@@ -824,8 +821,7 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi))
+	if (inode->i_ino == F2FS_NODE_INO(sbi))
 		return 0;
 
 	/*
@@ -919,7 +915,6 @@ static bool f2fs_pre_evict_inode(struct inode *inode)
 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
 
 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-	    inode->i_ino == F2FS_META_INO(sbi) ||
 	    inode->i_ino == F2FS_COMPRESS_INO(sbi))
 		return true;
 
fs/f2fs/node.c +62 -63

286 lines not on this page · raw message

fs/f2fs/node.h +2 -2
index 5e114f352099..de8dcd5d4392 100644
@@ -311,9 +311,9 @@ static inline void fill_node_footer_blkaddr(struct folio *folio, block_t blkaddr
 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
 }
 
-static inline bool is_recoverable_dnode(const struct folio *folio)
+static inline bool is_recoverable_dnode(struct f2fs_sb_info *sbi, const struct folio *folio)
 {
-	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_F_SB(folio));
+	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 	__u64 cp_ver = cur_cp_version(ckpt);
 
 	/* Don't care crc part, if fsck.f2fs sets it. */
fs/f2fs/recovery.c +100 -98

429 lines not on this page · raw message

fs/f2fs/segment.c +102 -99

440 lines not on this page · raw message

fs/f2fs/segment.h +27 -4
index db1079169a23..fff35c63a00f 100644
@@ -95,6 +95,8 @@ static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
 #define GET_SUM_BLKOFF(sbi, segno) (segno % (sbi)->sums_per_block)
 #define SUM_BLK_PAGE_ADDR(sbi, folio, segno)	\
 	(folio_address(folio) + GET_SUM_BLKOFF(sbi, segno) * (sbi)->sum_blocksize)
+#define SUM_BLK_ENTRY_ADDR(sbi, entry, segno)	\
+	(cache_address(entry) + GET_SUM_BLKOFF(sbi, segno) * (sbi)->sum_blocksize)
 
 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
@@ -417,8 +419,8 @@ static inline void __seg_info_to_raw_sit(struct seg_entry *se,
 	rs->mtime = cpu_to_le64(se->mtime);
 }
 
-static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi,
-				struct folio *folio, unsigned int start)
+static inline void seg_info_to_sit_block(struct f2fs_sb_info *sbi,
+			struct f2fs_cached_block *entry, unsigned int start)
 {
 	struct f2fs_sit_block *raw_sit;
 	struct seg_entry *se;
@@ -427,8 +429,8 @@ static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi,
 					(unsigned long)MAIN_SEGS(sbi));
 	int i;
 
-	raw_sit = folio_address(folio);
-	memset(raw_sit, 0, PAGE_SIZE);
+	raw_sit = cache_address(entry);
+	memset(raw_sit, 0, sbi->blocksize);
 	for (i = 0; i < end - start; i++) {
 		rs = &raw_sit->entries[i];
 		se = get_seg_entry(sbi, start + i);
@@ -991,6 +993,27 @@ static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
 		return 0;
 }
 
+/*
+ * When writing cache asynchronously, align nr_to_write to BIO_MAX_VECS.
+ */
+static inline long adjust_flush_cache_number(struct f2fs_sb_info *sbi, int type)
+{
+	long nr_to_write;
+
+	switch (type) {
+	case META:
+		nr_to_write = BIO_MAX_VECS;
+		break;
+	case NODE:
+		nr_to_write = BIO_MAX_VECS << 1;
+		break;
+	default:
+		f2fs_bug_on(sbi, 1);
+		return 0;
+	}
+	return nr_to_write;
+}
+
 /*
  * When writing pages, it'd better align nr_to_write for segment size.
  */
fs/f2fs/super.c +7 -24
index 89affe72f4fc..d3dee9b7d999 100644
@@ -1839,8 +1839,7 @@ static int f2fs_drop_inode(struct inode *inode)
 	 * drop useless meta/node dirty pages.
 	 */
 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
-		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi)) {
+		if (inode->i_ino == F2FS_NODE_INO(sbi)) {
 			trace_f2fs_drop_inode(inode, 1);
 			return 1;
 		}
@@ -1942,8 +1941,7 @@ static void f2fs_dirty_inode(struct inode *inode, int flags)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi))
+	if (inode->i_ino == F2FS_NODE_INO(sbi))
 		return;
 
 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
@@ -2039,9 +2037,10 @@ static void f2fs_put_super(struct super_block *sb)
 
 	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
 
-	if (err || f2fs_cp_error(sbi)) {
+	if (err || f2fs_cp_error(sbi) ||
+		unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 		truncate_inode_pages_final(NODE_MAPPING(sbi));
-		truncate_inode_pages_final(META_MAPPING(sbi));
+		f2fs_truncate_meta_caches(sbi, 0, ULONG_MAX);
 	}
 
 	f2fs_bug_on(sbi, sbi->fsync_node_num);
@@ -2051,9 +2050,6 @@ static void f2fs_put_super(struct super_block *sb)
 	iput(sbi->node_inode);
 	sbi->node_inode = NULL;
 
-	iput(sbi->meta_inode);
-	sbi->meta_inode = NULL;
-
 	f2fs_destroy_cache(META_CACHE(sbi));
 
 	/* Should check the page counts after dropping all node/meta pages */
@@ -4383,7 +4379,6 @@ static void init_sb_info(struct f2fs_sb_info *sbi)
 	sbi->allocate_section_policy = ALLOCATE_FORWARD_NOHINT;
 	F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
 	F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
-	F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
 	sbi->cur_victim_sec = NULL_SECNO;
 	sbi->gc_mode = GC_NORMAL;
 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
@@ -5224,18 +5219,10 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (err)
 		goto free_page_array_cache;
 
-	/* get an inode for meta space */
-	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
-	if (IS_ERR(sbi->meta_inode)) {
-		f2fs_err(sbi, "Failed to read F2FS meta data inode");
-		err = PTR_ERR(sbi->meta_inode);
-		goto free_meta_cache;
-	}
-
 	err = f2fs_get_valid_checkpoint(sbi);
 	if (err) {
 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
-		goto free_meta_inode;
+		goto free_meta_cache;
 	}
 
 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
@@ -5529,7 +5516,7 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
 	 * falls into an infinite loop in f2fs_sync_meta_pages().
 	 */
-	truncate_inode_pages_final(META_MAPPING(sbi));
+	f2fs_truncate_meta_caches(sbi, 0, ULONG_MAX);
 	/* evict some inodes being cached by GC */
 	evict_inodes(sb);
 	f2fs_unregister_sysfs(sbi);
@@ -5559,10 +5546,6 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 free_devices:
 	destroy_device_list(sbi);
 	kvfree(sbi->ckpt);
-free_meta_inode:
-	make_bad_inode(sbi->meta_inode);
-	iput(sbi->meta_inode);
-	sbi->meta_inode = NULL;
 free_meta_cache:
 	f2fs_destroy_cache(META_CACHE(sbi));
 free_page_array_cache:
include/linux/f2fs_fs.h +0 -1
index bb2b6cd5d507..0c027d00a1ea 100644
@@ -35,7 +35,6 @@
 
 #define F2FS_ROOT_INO(sbi)	((sbi)->root_ino_num)
 #define F2FS_NODE_INO(sbi)	((sbi)->node_ino_num)
-#define F2FS_META_INO(sbi)	((sbi)->meta_ino_num)
 #define F2FS_COMPRESS_INO(sbi)	(NM_I(sbi)->max_nid)
 
 #define F2FS_MAX_QUOTAS		3
Signature · 1 line
2.49.0
maintainer
#13 lore raw

Re: [f2fs-dev] [PATCH v1 05/12] f2fs: cache: use meta cache

Quoted 737 lines · On 08/20, Chao Yu via Linux-f2fs-devel wrote:

This patch migrates F2FS meta block caching from the fake VFS inode page cache (sbi->meta_inode) to meta cache (sbi->meta_blocks).

It converts CP, SIT, NAT, SSA, recovery, and GC metadata I/O paths to operate on struct f2fs_cached_block instead of folio, and removes sbi->meta_inode.

---
 fs/f2fs/cache.c         |   2 +
 fs/f2fs/checkpoint.c    | 394 ++++++++++++++++++----------------------
 fs/f2fs/compress.c      |   4 +-
 fs/f2fs/data.c          |  29 ++-
 fs/f2fs/debug.c         |  12 +-
 fs/f2fs/f2fs.h          |  81 +++------
 fs/f2fs/file.c          |   4 +-
 fs/f2fs/gc.c            | 134 +++++++-------
 fs/f2fs/inline.c        |   9 +-
 fs/f2fs/inode.c         |   9 +-
 fs/f2fs/node.c          | 125 +++++++------
 fs/f2fs/node.h          |   4 +-
 fs/f2fs/recovery.c      | 198 ++++++++++----------
 fs/f2fs/segment.c       | 201 ++++++++++----------
 fs/f2fs/segment.h       |  31 +++-
 fs/f2fs/super.c         |  31 +---
 include/linux/f2fs_fs.h |   1 -
 17 files changed, 606 insertions(+), 663 deletions(-)
fs/f2fs/cache.c +2 -0
index cb5b26046162..afef808e485a 100644
@@ -631,6 +631,8 @@ static int f2fs_cache_writeback_kthread(void *data)
 			break;
 		if (f2fs_cp_error(sbi))
 			continue;
+
+		f2fs_write_meta_caches(sbi);
 	}
 	return 0;
 }
fs/f2fs/checkpoint.c +157 -187
index 729d19680caf..1a7083540b82 100644
@@ -17,6 +17,7 @@
 #include <linux/delayacct.h>
 #include <linux/ioprio.h>
 #include <linux/math64.h>
+#include <linux/freezer.h>
 
 #include "f2fs.h"
 #include "node.h"
@@ -235,27 +236,27 @@ struct kmem_cache *f2fs_inode_entry_slab;
 /*
  * We guarantee no failure on the returned page.
  */
-struct folio *f2fs_grab_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index)
+struct f2fs_cached_block *f2fs_grab_meta_cache(struct f2fs_sb_info *sbi,
+							pgoff_t index)
 {
-	struct address_space *mapping = META_MAPPING(sbi);
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 repeat:
-	folio = f2fs_grab_cache_folio(mapping, index, false);
-	if (IS_ERR(folio)) {
+	entry = f2fs_grab_cache(META_CACHE(sbi), index,
+					F2FS_CACHE_LOCK_CREATE);
+	if (IS_ERR(entry)) {
 		cond_resched();
 		goto repeat;
 	}
-	f2fs_folio_wait_writeback(folio, META, true, true);
-	if (!folio_test_uptodate(folio))
-		folio_mark_uptodate(folio);
-	return folio;
+	f2fs_cache_wait_writeback(entry);
+	if (!f2fs_cache_test_uptodate(entry))
+		f2fs_cache_set_uptodate(entry);
+	return entry;
 }
 
-static struct folio *__get_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index,
-							bool is_meta)
+static struct f2fs_cached_block *__get_meta_cache(struct f2fs_sb_info *sbi,
+					pgoff_t index, bool is_meta)
 {
-	struct address_space *mapping = META_MAPPING(sbi);
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	struct f2fs_io_info fio = {
 		.sbi = sbi,
 		.type = META,
@@ -265,70 +266,75 @@ static struct folio *__get_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index,
 		.new_blkaddr = index,
 		.encrypted_page = NULL,
 		.is_por = !is_meta ? 1 : 0,
+		.is_cache = 1,
 	};
 	int err;
 
 	if (unlikely(!is_meta))
 		fio.op_flags &= ~REQ_META;
 repeat:
-	folio = f2fs_grab_cache_folio(mapping, index, false);
-	if (IS_ERR(folio)) {
+	entry = f2fs_grab_cache(META_CACHE(sbi), index,
+					F2FS_CACHE_LOCK_CREATE);
+	if (IS_ERR(entry)) {
 		cond_resched();
 		goto repeat;
 	}
-	if (folio_test_uptodate(folio))
+	if (f2fs_cache_test_uptodate(entry))
 		goto out;
 
-	fio.folio = folio;
+	fio.cache_entry = entry;
 
-	err = f2fs_submit_page_bio(&fio);
+	err = f2fs_submit_cache_read(&fio);
 	if (err) {
-		f2fs_folio_put(folio, true);
+		f2fs_put_cache(entry, true);
 		return ERR_PTR(err);
 	}
 
 	f2fs_update_iostat(sbi, NULL, FS_META_READ_IO, F2FS_BLKSIZE);
 
-	folio_lock(folio);
-	if (unlikely(!is_meta_folio(folio))) {
-		f2fs_folio_put(folio, true);
+	f2fs_lock_cache(entry);
+	if (unlikely(!f2fs_is_meta_cache(entry))) {
+		f2fs_put_cache(entry, true);
 		goto repeat;
 	}
 
-	if (unlikely(!folio_test_uptodate(folio))) {
-		f2fs_handle_page_eio(sbi, folio, META);
-		f2fs_folio_put(folio, true);
+	if (unlikely(!f2fs_cache_test_uptodate(entry))) {
+		f2fs_handle_page_eio(sbi, entry->index, META);
+		f2fs_put_cache(entry, true);
 		return ERR_PTR(-EIO);
 	}
 out:
-	return folio;
+	return entry;
 }
 
-struct folio *f2fs_get_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index)
+struct f2fs_cached_block *f2fs_get_meta_cache(struct f2fs_sb_info *sbi,
+							pgoff_t index)
 {
-	return __get_meta_folio(sbi, index, true);
+	return __get_meta_cache(sbi, index, true);
 }
 
-struct folio *f2fs_get_meta_folio_retry(struct f2fs_sb_info *sbi, pgoff_t index)
+struct f2fs_cached_block *f2fs_get_meta_cache_retry(struct f2fs_sb_info *sbi,
+							pgoff_t index)
 {
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	int count = 0;
 
 retry:
-	folio = __get_meta_folio(sbi, index, true);
-	if (IS_ERR(folio)) {
-		if (PTR_ERR(folio) == -EIO &&
+	entry = __get_meta_cache(sbi, index, true);
+	if (IS_ERR(entry)) {
+		if (PTR_ERR(entry) == -EIO &&
 				++count <= DEFAULT_RETRY_IO_COUNT)
 			goto retry;
 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_META_PAGE);
 	}
-	return folio;
+	return entry;
 }
 
 /* for POR only */
-struct folio *f2fs_get_tmp_folio(struct f2fs_sb_info *sbi, pgoff_t index)
+struct f2fs_cached_block *f2fs_get_tmp_cache(struct f2fs_sb_info *sbi,
+							pgoff_t index)
 {
-	return __get_meta_folio(sbi, index, false);
+	return __get_meta_cache(sbi, index, false);
 }
 
 static bool __is_bitmap_valid(struct f2fs_sb_info *sbi, block_t blkaddr,
@@ -446,9 +452,10 @@ bool f2fs_is_valid_blkaddr_raw(struct f2fs_sb_info *sbi,
 /*
  * Readahead CP/NAT/SIT/SSA/POR pages
  */
-int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
-							int type, bool sync)
+int f2fs_ra_meta_caches(struct f2fs_sb_info *sbi, block_t start,
+				int nrpages, int type, bool sync)
 {
+	struct f2fs_cached_block_list *cache = META_CACHE(sbi);
 	block_t blkno = start;
 	struct f2fs_io_info fio = {
 		.sbi = sbi,
@@ -458,6 +465,7 @@ int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 		.encrypted_page = NULL,
 		.in_list = 0,
 		.is_por = (type == META_POR) ? 1 : 0,
+		.is_cache = 1,
 	};
 	struct blk_plug plug;
 	int err;
@@ -467,7 +475,7 @@ int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 
 	blk_start_plug(&plug);
 	for (; nrpages-- > 0; blkno++) {
-		struct folio *folio;
+		struct f2fs_cached_block *entry;
 
 		if (!f2fs_is_valid_blkaddr(sbi, blkno, type))
 			goto out;
@@ -494,62 +502,58 @@ int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 			fio.new_blkaddr = blkno;
 			break;
 		default:
-			BUG();
+			f2fs_bug_on(sbi, 1);
 		}
 
-		folio = f2fs_grab_cache_folio(META_MAPPING(sbi),
-						fio.new_blkaddr, false);
-		if (IS_ERR(folio))
+		entry = f2fs_grab_cache(cache, fio.new_blkaddr,
+						F2FS_CACHE_LOCK_CREATE);
+		if (IS_ERR(entry))
 			continue;
-		if (folio_test_uptodate(folio)) {
-			f2fs_folio_put(folio, true);
+		if (f2fs_cache_test_uptodate(entry)) {
+			f2fs_put_cache(entry, true);
 			continue;
 		}
 
-		fio.folio = folio;
-		err = f2fs_submit_page_bio(&fio);
-		f2fs_folio_put(folio, err ? true : false);
+		fio.cache_entry = entry;
+		err = f2fs_submit_cache_read(&fio);
+		f2fs_put_cache(entry, err ? true : false);
 
 		if (!err)
 			f2fs_update_iostat(sbi, NULL, FS_META_READ_IO,
-							F2FS_BLKSIZE);
+							sbi->blocksize);
 	}
 out:
 	blk_finish_plug(&plug);
 	return blkno - start;
 }
 
-void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index,
-							unsigned int ra_blocks)
+void f2fs_ra_meta_caches_cond(struct f2fs_sb_info *sbi, pgoff_t index,
+						unsigned int ra_blocks)
 {
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	bool readahead = false;
 
 	if (ra_blocks == RECOVERY_MIN_RA_BLOCKS)
 		return;
 
-	folio = filemap_get_folio(META_MAPPING(sbi), index);
-	if (IS_ERR(folio) || !folio_test_uptodate(folio))
+	entry = f2fs_find_cache(META_CACHE(sbi), index);
+	if (IS_ERR(entry) || !f2fs_cache_test_uptodate(entry))
 		readahead = true;
-	f2fs_folio_put(folio, false);
+	f2fs_put_cache(entry, false);
 
 	if (readahead)
-		f2fs_ra_meta_pages(sbi, index, ra_blocks, META_POR, true);
+		f2fs_ra_meta_caches(sbi, index, ra_blocks, META_POR, true);
 }
 
-static bool __f2fs_write_meta_folio(struct folio *folio,
-				struct writeback_control *wbc,
+static bool __f2fs_write_meta_cache(struct f2fs_cached_block *entry,
 				enum iostat_type io_type)
 {
-	struct f2fs_sb_info *sbi = F2FS_F_SB(folio);
-
-	trace_f2fs_writepage(folio, META);
+	struct f2fs_sb_info *sbi = entry->cache->sbi;
 
 	if (unlikely(f2fs_cp_error(sbi))) {
 		if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
-			folio_clear_uptodate(folio);
-			dec_page_count(sbi, F2FS_DIRTY_META);
-			folio_unlock(folio);
+			f2fs_force_clear_cache_dirty(entry);
+			f2fs_unlock_cache(entry);
 			return true;
 		}
 		goto redirty_out;
@@ -557,10 +561,10 @@ static bool __f2fs_write_meta_folio(struct folio *folio,
 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
 		goto redirty_out;
 
-	f2fs_do_write_meta_page(sbi, folio, io_type);
+	f2fs_do_write_meta_cache(sbi, entry, io_type);
 	dec_page_count(sbi, F2FS_DIRTY_META);
 
-	folio_unlock(folio);
+	f2fs_unlock_cache(entry);
 
 	if (unlikely(f2fs_cp_error(sbi)))
 		f2fs_submit_merged_write(sbi, META);
@@ -568,101 +572,89 @@ static bool __f2fs_write_meta_folio(struct folio *folio,
 	return true;
 
 redirty_out:
-	folio_redirty_for_writepage(wbc, folio);
+	f2fs_cache_set_dirty(entry);
 	return false;
 }
 
-static int f2fs_write_meta_pages(struct address_space *mapping,
-				struct writeback_control *wbc)
+void f2fs_write_meta_caches(struct f2fs_sb_info *sbi)
 {
-	struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
 	struct f2fs_lock_context lc;
-	long diff, written;
+	long nr_to_write = LONG_MAX;
 
 	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
-		goto skip_write;
+		return;
 
-	/* collect a number of dirty meta pages and write together */
-	if (wbc->sync_mode != WB_SYNC_ALL &&
-			get_pages(sbi, F2FS_DIRTY_META) <
-					nr_pages_to_skip(sbi, META))
-		goto skip_write;
+	/* collect a number of dirty meta caches and write together */
+	if (get_pages(sbi, F2FS_DIRTY_META) <
+			nr_pages_to_skip(sbi, META))
+		return;
 
-	/* if locked failed, cp will flush dirty pages instead */
+	/* if locked failed, cp will flush dirty caches instead */
 	if (!f2fs_down_write_trylock_trace(&sbi->cp_global_sem, &lc))
-		goto skip_write;
+		return;
 
-	trace_f2fs_writepages(mapping->host, wbc, META);
-	diff = nr_pages_to_write(sbi, META, wbc);
-	written = f2fs_sync_meta_pages(sbi, wbc->nr_to_write, FS_META_IO);
+	nr_to_write = adjust_flush_cache_number(sbi, META);
+	f2fs_sync_meta_caches(sbi, nr_to_write, false, FS_META_IO);
 	f2fs_up_write_trace(&sbi->cp_global_sem, &lc);
-	wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff);
-	return 0;
-
-skip_write:
-	wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META);
-	trace_f2fs_writepages(mapping->host, wbc, META);
-	return 0;
 }
 
-long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, long nr_to_write,
-				enum iostat_type io_type)
+long f2fs_sync_meta_caches(struct f2fs_sb_info *sbi, long nr_to_write,
+				bool sync, enum iostat_type io_type)
 {
-	struct address_space *mapping = META_MAPPING(sbi);
 	pgoff_t index = 0, prev = ULONG_MAX;
-	struct folio_batch fbatch;
+	struct f2fs_cached_block *entries[F2FS_ONSTACK_CACHES];
 	long nwritten = 0;
-	int nr_folios;
-	struct writeback_control wbc = {};
+	int nr;
 	struct blk_plug plug;
-
-	folio_batch_init(&fbatch);
+	bool background = nr_to_write != LONG_MAX;
 
 	blk_start_plug(&plug);
 
-	while ((nr_folios = filemap_get_folios_tag(mapping, &index,
-					(pgoff_t)-1,
-					PAGECACHE_TAG_DIRTY, &fbatch))) {
+	while ((nr = f2fs_cache_gang_lookup_tag(META_CACHE(sbi), entries,
+			&index, F2FS_ONSTACK_CACHES, F2FS_CACHE_TAG_DIRTY))) {
 		int i;
 
-		for (i = 0; i < nr_folios; i++) {
-			struct folio *folio = fbatch.folios[i];
+		for (i = 0; i < nr; i++) {
+			struct f2fs_cached_block *entry = entries[i];
+
+			if (background && unlikely(freezing(current))) {
+				f2fs_cache_gang_release(entries, nr);
+				goto stop;
+			}
 
-			if (nr_to_write != LONG_MAX && i != 0 &&
-					folio->index != prev +
-					folio_nr_pages(fbatch.folios[i-1])) {
-				folio_batch_release(&fbatch);
+			if (background && i != 0 &&
+					entry->index != prev + 1) {
+				f2fs_cache_gang_release(entries, nr);
 				goto stop;
 			}
 
-			folio_lock(folio);
+			f2fs_lock_cache(entry);
 
-			if (unlikely(!is_meta_folio(folio))) {
+			if (unlikely(!f2fs_is_meta_cache(entry))) {
 continue_unlock:
-				folio_unlock(folio);
+				f2fs_unlock_cache(entry);
 				continue;
 			}
-			if (!folio_test_dirty(folio)) {
+			if (!f2fs_cache_test_dirty(entry)) {
 				/* someone wrote it for us */
 				goto continue_unlock;
 			}
 
-			f2fs_folio_wait_writeback(folio, META, true, true);
+			f2fs_cache_wait_writeback(entry);
 
-			if (!folio_clear_dirty_for_io(folio))
+			if (!f2fs_clear_cache_dirty(entry))
 				goto continue_unlock;
 
-			if (!__f2fs_write_meta_folio(folio, &wbc,
-						io_type)) {
-				folio_unlock(folio);
+			if (!__f2fs_write_meta_cache(entry, io_type)) {
+				f2fs_unlock_cache(entry);
 				break;
 			}
-			nwritten += folio_nr_pages(folio);
-			prev = folio->index;
+			nwritten += sbi->blocksize;
+			prev = entry->index;
 			if (unlikely(nwritten >= nr_to_write))
 				break;
 		}
-		folio_batch_release(&fbatch);
+		f2fs_cache_gang_release(entries, nr);
 		cond_resched();
 	}
 stop:
@@ -674,29 +666,6 @@ long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, long nr_to_write,
 	return nwritten;
 }
 
-static bool f2fs_dirty_meta_folio(struct address_space *mapping,
-		struct folio *folio)
-{
-	trace_f2fs_set_page_dirty(folio, META);
-
-	if (!folio_test_uptodate(folio))
-		folio_mark_uptodate(folio);
-	if (filemap_dirty_folio(mapping, folio)) {
-		inc_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_META);
-		folio_set_f2fs_reference(folio);
-		return true;
-	}
-	return false;
-}
-
-const struct address_space_operations f2fs_meta_aops = {
-	.writepages	= f2fs_write_meta_pages,
-	.dirty_folio	= f2fs_dirty_meta_folio,
-	.invalidate_folio = f2fs_invalidate_folio,
-	.release_folio	= f2fs_release_folio,
-	.migrate_folio	= filemap_migrate_folio,
-};
-
 static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino,
 						unsigned int devidx, int type)
 {
@@ -1036,20 +1005,20 @@ int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi)
 	start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi);
 	orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi);
 
-	f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true);
+	f2fs_ra_meta_caches(sbi, start_blk, orphan_blocks, META_CP, true);
 
 	for (i = 0; i < orphan_blocks; i++) {
-		struct folio *folio;
+		struct f2fs_cached_block *entry;
 		struct f2fs_orphan_block *orphan_blk;
 		unsigned int entry_count;
 
-		folio = f2fs_get_meta_folio(sbi, start_blk + i);
-		if (IS_ERR(folio)) {
-			err = PTR_ERR(folio);
+		entry = f2fs_get_meta_cache(sbi, start_blk + i);
+		if (IS_ERR(entry)) {
+			err = PTR_ERR(entry);
 			goto out;
 		}
 
-		orphan_blk = folio_address(folio);
+		orphan_blk = cache_address(entry);
 		entry_count = le32_to_cpu(orphan_blk->entry_count);
 		if (entry_count > F2FS_ORPHANS_PER_BLOCK) {
 			f2fs_err(sbi, "invalid orphan inode entry count %u",
@@ -1057,7 +1026,7 @@ int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi)
 			set_sbi_flag(sbi, SBI_NEED_FSCK);
 			f2fs_handle_error(sbi, ERROR_INCONSISTENT_ORPHAN);
 			err = -EFSCORRUPTED;
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 			goto out;
 		}
 
@@ -1066,11 +1035,11 @@ int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi)
 
 			err = recover_orphan_inode(sbi, ino);
 			if (err) {
-				f2fs_folio_put(folio, true);
+				f2fs_put_cache(entry, true);
 				goto out;
 			}
 		}
-		f2fs_folio_put(folio, true);
+		f2fs_put_cache(entry, true);
 	}
 	/* clear Orphan Flag */
 	clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG);
@@ -1087,9 +1056,9 @@ static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
 	unsigned int nentries = 0;
 	unsigned short index = 1;
 	unsigned short orphan_blocks;
-	struct folio *folio = NULL;
 	struct ino_entry *orphan = NULL;
 	struct inode_management *im = &sbi->im[ORPHAN_INO];
+	struct f2fs_cached_block *entry = NULL;
 
 	orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num);
 
@@ -1102,9 +1071,9 @@ static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
 
 	/* loop for each orphan inode entry and write them in journal block */
 	list_for_each_entry(orphan, head, list) {
-		if (!folio) {
-			folio = f2fs_grab_meta_folio(sbi, start_blk++);
-			orphan_blk = folio_address(folio);
+		if (!entry) {
+			entry = f2fs_grab_meta_cache(sbi, start_blk++);
+			orphan_blk = (struct f2fs_orphan_block *)cache_address(entry);
 			memset(orphan_blk, 0, sizeof(*orphan_blk));
 		}
 
@@ -1119,20 +1088,20 @@ static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
 			orphan_blk->blk_addr = cpu_to_le16(index);
 			orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
 			orphan_blk->entry_count = cpu_to_le32(nentries);
-			folio_mark_dirty(folio);
-			f2fs_folio_put(folio, true);
+			f2fs_mark_cache_dirty(entry);
+			f2fs_put_cache(entry, true);
 			index++;
 			nentries = 0;
-			folio = NULL;
+			entry = NULL;
 		}
 	}
 
-	if (folio) {
+	if (entry) {
 		orphan_blk->blk_addr = cpu_to_le16(index);
 		orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
 		orphan_blk->entry_count = cpu_to_le32(nentries);
-		folio_mark_dirty(folio);
-		f2fs_folio_put(folio, true);
+		f2fs_mark_cache_dirty(entry);
+		f2fs_put_cache(entry, true);
 	}
 }
 
@@ -1151,29 +1120,29 @@ static __u32 f2fs_checkpoint_chksum(struct f2fs_checkpoint *ckpt)
 }
 
 static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr,
-		struct f2fs_checkpoint **cp_block, struct folio **cp_folio,
+		struct f2fs_checkpoint **cp_block, struct f2fs_cached_block **cp_entry,
 		unsigned long long *version)
 {
 	size_t crc_offset = 0;
 	__u32 crc;
 
-	*cp_folio = f2fs_get_meta_folio(sbi, cp_addr);
-	if (IS_ERR(*cp_folio))
-		return PTR_ERR(*cp_folio);
+	*cp_entry = f2fs_get_meta_cache(sbi, cp_addr);
+	if (IS_ERR(*cp_entry))
+		return PTR_ERR(*cp_entry);
 
-	*cp_block = folio_address(*cp_folio);
+	*cp_block = (struct f2fs_checkpoint *)cache_address(*cp_entry);
 
 	crc_offset = le32_to_cpu((*cp_block)->checksum_offset);
 	if (crc_offset < CP_MIN_CHKSUM_OFFSET ||
 			crc_offset > CP_CHKSUM_OFFSET) {
-		f2fs_folio_put(*cp_folio, true);
+		f2fs_put_cache(*cp_entry, true);
 		f2fs_warn(sbi, "invalid crc_offset: %zu", crc_offset);
 		return -EINVAL;
 	}
 
 	crc = f2fs_checkpoint_chksum(*cp_block);
 	if (crc != cur_cp_crc(*cp_block)) {
-		f2fs_folio_put(*cp_folio, true);
+		f2fs_put_cache(*cp_entry, true);
 		f2fs_warn(sbi, "invalid crc value");
 		return -EINVAL;
 	}
@@ -1182,17 +1151,17 @@ static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr,
 	return 0;
 }
 
-static struct folio *validate_checkpoint(struct f2fs_sb_info *sbi,
+static struct f2fs_cached_block *validate_checkpoint(struct f2fs_sb_info *sbi,
 				block_t cp_addr, unsigned long long *version)
 {
-	struct folio *cp_folio_1 = NULL, *cp_folio_2 = NULL;
+	struct f2fs_cached_block *cp_entry_1 = NULL, *cp_entry_2 = NULL;
 	struct f2fs_checkpoint *cp_block = NULL;
 	unsigned long long cur_version = 0, pre_version = 0;
 	unsigned int cp_blocks;
 	int err;
 
 	err = get_checkpoint_version(sbi, cp_addr, &cp_block,
-					&cp_folio_1, version);
+					&cp_entry_1, version);
 	if (err)
 		return NULL;
 
@@ -1207,19 +1176,20 @@ static struct folio *validate_checkpoint(struct f2fs_sb_info *sbi,
 
 	cp_addr += cp_blocks - 1;
 	err = get_checkpoint_version(sbi, cp_addr, &cp_block,
-					&cp_folio_2, version);
+					&cp_entry_2, version);
 	if (err)
 		goto invalid_cp;
 	cur_version = *version;
 
 	if (cur_version == pre_version) {
 		*version = cur_version;
-		f2fs_folio_put(cp_folio_2, true);
-		return cp_folio_1;
+		f2fs_put_cache(cp_entry_2, true);
+		return cp_entry_1;
 	}
-	f2fs_folio_put(cp_folio_2, true);
+	f2fs_put_cache(cp_entry_2, true);
 invalid_cp:
-	f2fs_folio_put(cp_folio_1, true);
+	if (cp_entry_1)
+		f2fs_put_cache(cp_entry_1, true);
 	return NULL;
 }
 
@@ -1227,7 +1197,7 @@ int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi)
 {
 	struct f2fs_checkpoint *cp_block;
 	struct f2fs_super_block *fsb = sbi->raw_super;
-	struct folio *cp1, *cp2, *cur_folio;
+	struct f2fs_cached_block *cp1 = NULL, *cp2 = NULL, *cur_entry;
 	unsigned long blk_size = sbi->blocksize;
 	unsigned long long cp1_version = 0, cp2_version = 0;
 	unsigned long long cp_start_blk_no;
@@ -1254,22 +1224,22 @@ int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi)
 
 	if (cp1 && cp2) {
 		if (ver_after(cp2_version, cp1_version))
-			cur_folio = cp2;
+			cur_entry = cp2;
 		else
-			cur_folio = cp1;
+			cur_entry = cp1;
 	} else if (cp1) {
-		cur_folio = cp1;
+		cur_entry = cp1;
 	} else if (cp2) {
-		cur_folio = cp2;
+		cur_entry = cp2;
 	} else {
 		err = -EFSCORRUPTED;
 		goto fail_no_cp;
 	}
 
-	cp_block = folio_address(cur_folio);
+	cp_block = (struct f2fs_checkpoint *)cache_address(cur_entry);
 	memcpy(sbi->ckpt, cp_block, blk_size);
 
-	if (cur_folio == cp1)
+	if (cur_entry == cp1)
 		sbi->cur_cp_pack = 1;
 	else
 		sbi->cur_cp_pack = 2;
@@ -1284,30 +1254,35 @@ int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi)
 		goto done;
 
 	cp_blk_no = le32_to_cpu(fsb->cp_blkaddr);
-	if (cur_folio == cp2)
+	if (cur_entry == cp2)
 		cp_blk_no += BIT(le32_to_cpu(fsb->log_blocks_per_seg));
 
 	for (i = 1; i < cp_blks; i++) {
+		struct f2fs_cached_block *cur_entry_payload;
 		void *sit_bitmap_ptr;
 		unsigned char *ckpt = (unsigned char *)sbi->ckpt;
 
-		cur_folio = f2fs_get_meta_folio(sbi, cp_blk_no + i);
-		if (IS_ERR(cur_folio)) {
-			err = PTR_ERR(cur_folio);
+		cur_entry_payload = f2fs_get_meta_cache(sbi, cp_blk_no + i);
+		if (IS_ERR(cur_entry_payload)) {
+			err = PTR_ERR(cur_entry_payload);
 			goto free_fail_no_cp;
 		}
-		sit_bitmap_ptr = folio_address(cur_folio);

fs/f2fs/checkpoint.c int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi) · line 1276

Quoted 10 lines
fs/f2fs/checkpoint.c +4 -3
+		sit_bitmap_ptr = cache_address(cur_entry_payload);
 		memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size);
-		f2fs_folio_put(cur_folio, true);
+		f2fs_put_cache(cur_entry_payload, true);
 	}
 done:
-	f2fs_folio_put(cp1, true);
-	f2fs_folio_put(cp2, true);
+	if (cp1)
+		f2fs_put_cache(cp1, true);

In f2fs_put_cache(x, y),

	if (!x)
		return;
Quoted 2,265 lines
fs/f2fs/checkpoint.c +20 -23
+	if (cp2)
+		f2fs_put_cache(cp2, true);
 	return 0;
 
 free_fail_no_cp:
-	f2fs_folio_put(cp1, true);
-	f2fs_folio_put(cp2, true);
+	if (cp1)
+		f2fs_put_cache(cp1, true);
+	if (cp2)
+		f2fs_put_cache(cp2, true);
 fail_no_cp:
 	kvfree(sbi->ckpt);
 	return err;
@@ -1621,7 +1596,7 @@ void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type)
 			break;
 
 		if (type == F2FS_DIRTY_META)
-			f2fs_sync_meta_pages(sbi, LONG_MAX, FS_CP_META_IO);
+			f2fs_sync_meta_caches(sbi, LONG_MAX, true, FS_META_IO);
 		else if (type == F2FS_WB_CP_DATA)
 			f2fs_submit_merged_write(sbi, DATA);
 
@@ -1700,31 +1675,24 @@ static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 static void commit_checkpoint(struct f2fs_sb_info *sbi,
 	void *src, block_t blk_addr)
 {
-	struct writeback_control wbc = {};
-
-	/*
-	 * filemap_get_folios_tag and folio_lock again will take
-	 * some extra time. Therefore, f2fs_update_meta_pages and
-	 * f2fs_sync_meta_pages are combined in this function.
-	 */
-	struct folio *folio = f2fs_grab_meta_folio(sbi, blk_addr);
+	struct f2fs_cached_block *entry = f2fs_grab_meta_cache(sbi, blk_addr);
 
-	memcpy(folio_address(folio), src, PAGE_SIZE);
+	memcpy(cache_address(entry), src, F2FS_BLKSIZE);
 
-	folio_mark_dirty(folio);
-	if (unlikely(!folio_clear_dirty_for_io(folio)))
+	f2fs_mark_cache_dirty(entry);
+	if (unlikely(!f2fs_clear_cache_dirty(entry)))
 		f2fs_bug_on(sbi, 1);
 
 	/* writeout cp pack 2 page */
-	if (unlikely(!__f2fs_write_meta_folio(folio, &wbc, FS_CP_META_IO))) {
+	if (unlikely(!__f2fs_write_meta_cache(entry, FS_CP_META_IO))) {
 		if (f2fs_cp_error(sbi)) {
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 			return;
 		}
 		f2fs_bug_on(sbi, true);
 	}
 
-	f2fs_folio_put(folio, false);
+	f2fs_put_cache(entry, false);
 
 	/* submit checkpoint (with barrier if NOBARRIER is not set) */
 	f2fs_submit_merged_write(sbi, META_FLUSH);
@@ -1793,7 +1761,7 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	int err;
 
 	/* Flush all the NAT/SIT pages */
-	f2fs_sync_meta_pages(sbi, LONG_MAX, FS_CP_META_IO);
+	f2fs_sync_meta_caches(sbi, LONG_MAX, true, FS_CP_META_IO);
 
 	stat_cp_time(cpc, CP_TIME_SYNC_META);
 
@@ -1892,7 +1860,7 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	}
 
 	/* Here, we have one bio having CP pack except cp pack 2 page */
-	f2fs_sync_meta_pages(sbi, LONG_MAX, FS_CP_META_IO);
+	f2fs_sync_meta_caches(sbi, LONG_MAX, true, FS_CP_META_IO);
 	stat_cp_time(cpc, CP_TIME_SYNC_CP_META);
 
 	/* Wait for all dirty meta pages to be submitted for IO */
@@ -1919,10 +1887,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	 * used for migration of encrypted, verity or compressed inode's blocks.
 	 */
 	if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi) ||
-		f2fs_sb_has_compression(sbi))
-		f2fs_bug_on(sbi,
-			invalidate_inode_pages2_range(META_MAPPING(sbi),
-				MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1));
+		f2fs_sb_has_compression(sbi)) {
+		f2fs_truncate_meta_caches(sbi, MAIN_BLKADDR(sbi),
+				MAX_BLKADDR(sbi) - MAIN_BLKADDR(sbi));
+	}
 
 	f2fs_release_ino_entry(sbi, false);
 
fs/f2fs/compress.c +2 -2
index 91855d91bbdd..bb749f6257a1 100644
@@ -1150,7 +1150,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 		f2fs_compress_ctx_add_page(cc, folio);
 
 		if (!folio_test_uptodate(folio)) {
-			f2fs_handle_page_eio(sbi, folio, DATA);
+			f2fs_handle_page_eio(sbi, folio->index, DATA);
 release_and_retry:
 			f2fs_put_rpages(cc);
 			f2fs_unlock_rpages(cc, i + 1);
@@ -1359,7 +1359,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 		fio.old_blkaddr = data_blkaddr(dn.inode, dn.node_folio,
 						dn.ofs_in_node + i + 1);
 
-		/* wait for GCed page writeback via META_MAPPING */
+		/* wait for GCed page writeback via generic cache */
 		f2fs_wait_on_block_writeback(inode, fio.old_blkaddr);
 
 		if (fio.encrypted) {
fs/f2fs/data.c +12 -17
index 110282bb8dcd..92c3293f0a1e 100644
@@ -65,8 +65,7 @@ bool f2fs_is_cp_guaranteed(const struct folio *folio)
 
 	inode = mapping->host;
 
-	if (inode->i_ino == F2FS_META_INO(sbi) ||
-			inode->i_ino == F2FS_NODE_INO(sbi) ||
+	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 			S_ISDIR(inode->i_mode))
 		return true;
 
@@ -84,9 +83,6 @@ static enum count_type __read_io_type(struct folio *folio)
 		struct inode *inode = mapping->host;
 		struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-		if (inode->i_ino == F2FS_META_INO(sbi))
-			return F2FS_RD_META;
-
 		if (inode->i_ino == F2FS_NODE_INO(sbi))
 			return F2FS_RD_NODE;
 	}
@@ -1452,7 +1448,7 @@ static void f2fs_submit_page_read(struct inode *inode, struct fsverity_info *vi,
 	bio = f2fs_grab_read_bio(inode, vi, blkaddr, 1, op_flags, folio->index,
 				 for_write);
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	f2fs_wait_on_block_writeback(inode, blkaddr);
 
 	if (!bio_add_folio(bio, folio, PAGE_SIZE, 0))
@@ -3110,7 +3106,7 @@ static void f2fs_readahead(struct readahead_control *rac)
 int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
 {
 	struct inode *inode = fio_inode(fio);
-	struct folio *mfolio;
+	struct f2fs_cached_block *entry;
 	struct page *page;
 
 	if (!f2fs_encrypted_file(inode))
@@ -3126,12 +3122,13 @@ int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
 	if (IS_ERR(fio->encrypted_page))
 		return PTR_ERR(fio->encrypted_page);
 
-	mfolio = filemap_lock_folio(META_MAPPING(fio->sbi), fio->old_blkaddr);
-	if (!IS_ERR(mfolio)) {
-		if (folio_test_uptodate(mfolio))
-			memcpy(folio_address(mfolio),
-				page_address(fio->encrypted_page), PAGE_SIZE);
-		f2fs_folio_put(mfolio, true);
+	entry = f2fs_find_cache(META_CACHE(fio->sbi), fio->old_blkaddr);
+	if (!IS_ERR(entry)) {
+		f2fs_lock_cache(entry);
+		if (f2fs_cache_test_uptodate(entry))
+			memcpy(cache_address(entry), page_address(fio->encrypted_page),
+						F2FS_BLKSIZE);
+		f2fs_put_cache(entry, true);
 	}
 	return 0;
 }
@@ -3297,7 +3294,7 @@ int f2fs_do_write_data_page(struct f2fs_io_info *fio)
 		goto out_writepage;
 	}
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	if (fio->meta_gc)
 		f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
 
@@ -4380,9 +4377,7 @@ void f2fs_invalidate_folio(struct folio *folio, size_t offset, size_t length)
 		return;
 
 	if (folio_test_dirty(folio)) {
-		if (inode->i_ino == F2FS_META_INO(sbi)) {
-			dec_page_count(sbi, F2FS_DIRTY_META);
-		} else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
+		if (inode->i_ino == F2FS_NODE_INO(sbi)) {
 			dec_page_count(sbi, F2FS_DIRTY_NODES);
 		} else {
 			inode_dec_dirty_pages(inode);
fs/f2fs/debug.c +4 -8
index ff379aff4472..14059a50739c 100644
@@ -224,8 +224,7 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->dirty_count = dirty_segments(sbi);
 	if (sbi->node_inode)
 		si->node_pages = NODE_MAPPING(sbi)->nrpages;
-	if (sbi->meta_inode)
-		si->meta_pages = META_MAPPING(sbi)->nrpages;
+	si->meta_caches = META_CACHE(sbi)->num_entries;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
 		si->compress_pages = COMPRESS_MAPPING(sbi)->nrpages;
@@ -388,11 +387,8 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 
 		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
 	}
-	if (sbi->meta_inode) {
-		unsigned long npages = META_MAPPING(sbi)->nrpages;
-
-		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
-	}
+	si->page_mem += (unsigned long long)META_CACHE(sbi)->num_entries << PAGE_SHIFT;
+	si->cache_mem += META_CACHE(sbi)->num_entries * sizeof(struct f2fs_cached_block);
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
 		unsigned long npages = COMPRESS_MAPPING(sbi)->nrpages;
@@ -708,7 +704,7 @@ static int stat_show(struct seq_file *s, void *v)
 		seq_printf(s, "  - quota data: %4d in quota files:%4d\n",
 			   si->ndirty_qdata, si->nquota_files);
 		seq_printf(s, "  - meta: %4d in %4d\n",
-			   si->ndirty_meta, si->meta_pages);
+			   si->ndirty_meta, si->meta_caches);
 		seq_printf(s, "  - imeta: %4d\n",
 			   si->ndirty_imeta);
 		seq_printf(s, "  - fsync mark: %4lld\n",
fs/f2fs/f2fs.h +23 -58
index 6e20b3586f26..9a353dcd6658 100644
@@ -1828,7 +1828,6 @@ struct f2fs_sb_info {
 	struct f2fs_checkpoint *ckpt;		/* raw checkpoint pointer */
 	int cur_cp_pack;			/* remain current cp pack */
 	spinlock_t cp_lock;			/* for flag in ckpt */
-	struct inode *meta_inode;		/* cache meta blocks */
 	struct f2fs_rwsem cp_global_sem;	/* checkpoint procedure lock */
 	struct f2fs_rwsem cp_rwsem;		/* blocking FS operations */
 	struct f2fs_rwsem node_write;		/* locking node writes */
@@ -1873,7 +1872,6 @@ struct f2fs_sb_info {
 	unsigned int blocksize;			/* block size */
 	unsigned int root_ino_num;		/* root inode number*/
 	unsigned int node_ino_num;		/* node inode number*/
-	unsigned int meta_ino_num;		/* meta inode number*/
 	unsigned int log_blocks_per_seg;	/* log2 blocks per segment */
 	unsigned int blocks_per_seg;		/* blocks per segment */
 	unsigned int segs_per_sec;		/* segments per section */
@@ -2319,9 +2317,9 @@ static inline struct dirty_seglist_info *DIRTY_I(struct f2fs_sb_info *sbi)
 	return (struct dirty_seglist_info *)(SM_I(sbi)->dirty_info);
 }
 
-static inline struct address_space *META_MAPPING(struct f2fs_sb_info *sbi)
+static inline bool f2fs_is_meta_cache(struct f2fs_cached_block *entry)
 {
-	return sbi->meta_inode->i_mapping;
+	return entry->cache && entry->cache == META_CACHE(entry->cache->sbi);
 }
 
 static inline struct address_space *NODE_MAPPING(struct f2fs_sb_info *sbi)
@@ -2329,11 +2327,6 @@ static inline struct address_space *NODE_MAPPING(struct f2fs_sb_info *sbi)
 	return sbi->node_inode->i_mapping;
 }
 
-static inline bool is_meta_folio(struct folio *folio)
-{
-	return folio->mapping == META_MAPPING(F2FS_F_SB(folio));
-}
-
 static inline bool is_node_folio(struct folio *folio)
 {
 	return folio->mapping == NODE_MAPPING(F2FS_F_SB(folio));
@@ -4057,9 +4050,9 @@ bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid);
 void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid);
 void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid);
 int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink);
-int f2fs_recover_inline_xattr(struct inode *inode, struct folio *folio);
-int f2fs_recover_xattr_data(struct inode *inode, struct folio *folio);
-int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct folio *folio);
+int f2fs_recover_inline_xattr(struct inode *inode, struct f2fs_cached_block *entry);
+int f2fs_recover_xattr_data(struct inode *inode, struct f2fs_cached_block *entry);
+int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct f2fs_cached_block *entry);
 int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
 			unsigned int segno, struct f2fs_summary_block *sum);
 int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc);
@@ -4109,11 +4102,13 @@ int f2fs_allocate_new_segments(struct f2fs_sb_info *sbi);
 int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range);
 bool f2fs_exist_trim_candidates(struct f2fs_sb_info *sbi,
 					struct cp_control *cpc);
-struct folio *f2fs_get_sum_folio(struct f2fs_sb_info *sbi, unsigned int segno);
+struct f2fs_cached_block *f2fs_get_sum_cache(struct f2fs_sb_info *sbi,
+		unsigned int segno);
 void f2fs_update_meta_page(struct f2fs_sb_info *sbi, void *src,
 					block_t blk_addr);
-void f2fs_do_write_meta_page(struct f2fs_sb_info *sbi, struct folio *folio,
-						enum iostat_type io_type);
+void f2fs_do_write_meta_cache(struct f2fs_sb_info *sbi,
+			struct f2fs_cached_block *entry,
+			enum iostat_type io_type);
 void f2fs_do_write_node_page(unsigned int nid, struct f2fs_io_info *fio);
 void f2fs_outplace_write_data(struct dnode_of_data *dn,
 			struct f2fs_io_info *fio);
@@ -4136,8 +4131,6 @@ void f2fs_update_device_state(struct f2fs_sb_info *sbi, nid_t ino,
 					block_t blkaddr, unsigned int blkcnt);
 void f2fs_folio_wait_writeback(struct folio *folio, enum page_type type,
 		bool ordered, bool locked);
-#define f2fs_wait_on_page_writeback(page, type, ordered, locked)	\
-		f2fs_folio_wait_writeback(page_folio(page), type, ordered, locked)
 void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr);
 void f2fs_wait_on_block_writeback_range(struct inode *inode, block_t blkaddr,
 								block_t len);
@@ -4203,20 +4196,21 @@ void f2fs_unlock_op(struct f2fs_sb_info *sbi, struct f2fs_lock_context *lc);
 void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io,
 							unsigned char reason);
 void f2fs_flush_ckpt_thread(struct f2fs_sb_info *sbi);
-struct folio *f2fs_grab_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index);
-struct folio *f2fs_get_meta_folio(struct f2fs_sb_info *sbi, pgoff_t index);
-struct folio *f2fs_get_meta_folio_retry(struct f2fs_sb_info *sbi, pgoff_t index);
-struct folio *f2fs_get_tmp_folio(struct f2fs_sb_info *sbi, pgoff_t index);
+struct f2fs_cached_block *f2fs_grab_meta_cache(struct f2fs_sb_info *sbi, pgoff_t index);
+struct f2fs_cached_block *f2fs_get_meta_cache(struct f2fs_sb_info *sbi, pgoff_t index);
+struct f2fs_cached_block *f2fs_get_meta_cache_retry(struct f2fs_sb_info *sbi, pgoff_t index);
+struct f2fs_cached_block *f2fs_get_tmp_cache(struct f2fs_sb_info *sbi, pgoff_t index);
 bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi,
 					block_t blkaddr, int type);
 bool f2fs_is_valid_blkaddr_raw(struct f2fs_sb_info *sbi,
 					block_t blkaddr, int type);
-int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
+int f2fs_ra_meta_caches(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 			int type, bool sync);
-void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index,
+void f2fs_ra_meta_caches_cond(struct f2fs_sb_info *sbi, pgoff_t index,
 							unsigned int ra_blocks);
-long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, long nr_to_write,
-			enum iostat_type io_type);
+void f2fs_write_meta_caches(struct f2fs_sb_info *sbi);
+long f2fs_sync_meta_caches(struct f2fs_sb_info *sbi, long nr_to_write,
+				bool sync, enum iostat_type io_type);
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type);
 void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type);
 void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all);
@@ -4403,7 +4397,7 @@ struct f2fs_stat_info {
 	unsigned int bimodal, avg_vblocks;
 	int util_free, util_valid, util_invalid;
 	int rsvd_segs, overp_segs;
-	int dirty_count, node_pages, meta_pages, compress_pages;
+	int dirty_count, node_pages, meta_caches, compress_pages;
 	int compress_page_hit;
 	int prefree_count, free_segs, free_secs;
 	int cp_call_count[MAX_CALL_TYPE], cp_count;
@@ -4605,7 +4599,6 @@ extern const struct file_operations f2fs_file_operations;
 extern const struct inode_operations f2fs_file_inode_operations;
 extern const struct address_space_operations f2fs_dblock_aops;
 extern const struct address_space_operations f2fs_node_aops;
-extern const struct address_space_operations f2fs_meta_aops;
 extern const struct inode_operations f2fs_dir_inode_operations;
 extern const struct inode_operations f2fs_symlink_inode_operations;
 extern const struct inode_operations f2fs_encrypted_symlink_inode_operations;
@@ -4626,7 +4619,7 @@ int f2fs_convert_inline_folio(struct dnode_of_data *dn, struct folio *folio);
 int f2fs_convert_inline_inode(struct inode *inode);
 int f2fs_try_convert_inline_dir(struct inode *dir, struct dentry *dentry);
 int f2fs_write_inline_data(struct inode *inode, struct folio *folio);
-int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio);
+int f2fs_recover_inline_data(struct inode *inode, struct f2fs_cached_block *entry);
 struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
 		const struct f2fs_filename *fname, struct folio **res_folio,
 		bool use_hash);
@@ -5186,10 +5179,8 @@ static inline void f2fs_schedule_timeout_killable(long timeout, bool io)
 }
 
 static inline void f2fs_handle_page_eio(struct f2fs_sb_info *sbi,
-				struct folio *folio, enum page_type type)
+				pgoff_t ofs, enum page_type type)
 {
-	pgoff_t ofs = folio->index;
-
 	if (unlikely(f2fs_cp_error(sbi)))
 		return;
 
@@ -5224,36 +5215,10 @@ static inline bool f2fs_is_readonly(struct f2fs_sb_info *sbi)
 	return f2fs_sb_has_readonly(sbi) || f2fs_readonly(sbi->sb);
 }
 
-static inline void f2fs_truncate_meta_inode_pages(struct f2fs_sb_info *sbi,
-					block_t blkaddr, unsigned int cnt)
-{
-	bool need_submit = false;
-	int i = 0;
-
-	do {
-		struct folio *folio;
-
-		folio = filemap_get_folio(META_MAPPING(sbi), blkaddr + i);
-		if (!IS_ERR(folio)) {
-			if (folio_test_writeback(folio))
-				need_submit = true;
-			f2fs_folio_put(folio, false);
-		}
-	} while (++i < cnt && !need_submit);
-
-	if (need_submit)
-		f2fs_submit_merged_write_cond(sbi, sbi->meta_inode,
-							NULL, 0, DATA);
-
-	truncate_inode_pages_range(META_MAPPING(sbi),
-			F2FS_BLK_TO_BYTES((loff_t)blkaddr),
-			F2FS_BLK_END_BYTES((loff_t)(blkaddr + cnt - 1)));
-}
-
 static inline void f2fs_invalidate_internal_cache(struct f2fs_sb_info *sbi,
 						block_t blkaddr, unsigned int len)
 {
-	f2fs_truncate_meta_inode_pages(sbi, blkaddr, len);
+	f2fs_truncate_meta_caches(sbi, blkaddr, len);
 	f2fs_invalidate_compress_pages_range(sbi, blkaddr, len);
 }
 
fs/f2fs/file.c +2 -2
index a54b3ab52f1a..92daa41dd96d 100644
@@ -214,7 +214,7 @@ static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
 
 	f2fs_folio_wait_writeback(folio, DATA, false, true);
 
-	/* wait for GCed page writeback via META_MAPPING */
+	/* wait for GCed page writeback via generic cache */
 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 
 	/*
@@ -2563,7 +2563,7 @@ int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag,
 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
 		break;
 	case F2FS_GOING_DOWN_METAFLUSH:
-		f2fs_sync_meta_pages(sbi, LONG_MAX, FS_META_IO);
+		f2fs_sync_meta_caches(sbi, LONG_MAX, true, FS_META_IO);
 		f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN);
 		break;
 	case F2FS_GOING_DOWN_NEED_FSCK:
fs/f2fs/gc.c +73 -61
index c4da2f31805b..54327cb2e27e 100644
@@ -1066,7 +1066,7 @@ static int gc_node_segment(struct f2fs_sb_info *sbi,
 			continue;
 
 		if (phase == 0) {
-			f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), 1,
+			f2fs_ra_meta_caches(sbi, NAT_BLOCK_OFFSET(nid), 1,
 							META_NAT, true);
 			continue;
 		}
@@ -1217,7 +1217,8 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
 	struct address_space *mapping = inode->i_mapping;
 	struct inode *atomic_inode = NULL;
 	struct dnode_of_data dn;
-	struct folio *folio, *efolio;
+	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	struct f2fs_io_info fio = {
 		.sbi = sbi,
 		.ino = inode->i_ino,
@@ -1227,6 +1228,7 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
 		.op_flags = 0,
 		.encrypted_page = NULL,
 		.in_list = 0,
+		.is_cache = 1,
 	};
 	int err = 0;
 
@@ -1285,22 +1287,21 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
 
 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
 
-	efolio = f2fs_filemap_get_folio(META_MAPPING(sbi), dn.data_blkaddr,
-					FGP_LOCK | FGP_CREAT, GFP_NOFS);
-	if (IS_ERR(efolio)) {
-		err = PTR_ERR(efolio);
+	entry = f2fs_grab_cache(META_CACHE(sbi), dn.data_blkaddr,
+						F2FS_CACHE_LOCK_CREATE);
+	if (IS_ERR(entry)) {
+		err = PTR_ERR(entry);
 		goto put_folio;
 	}
 
-	fio.encrypted_page = &efolio->page;
-
-	if (folio_test_uptodate(efolio))
-		goto put_encrypted_page;
+	if (f2fs_cache_test_uptodate(entry))
+		goto put_cache;
 
-	err = f2fs_submit_page_bio(&fio);
+	fio.cache_entry = entry;
+	err = f2fs_submit_cache_read(&fio);
 	if (err)
-		goto put_encrypted_page;
-	f2fs_put_page(fio.encrypted_page, false);
+		goto put_cache;
+	f2fs_put_cache(entry, false);
 	f2fs_folio_put(folio, true);
 
 	f2fs_update_iostat(sbi, inode, FS_DATA_READ_IO, F2FS_BLKSIZE);
@@ -1309,8 +1310,8 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
 	if (atomic_inode)
 		iput(atomic_inode);
 	return 0;
-put_encrypted_page:
-	f2fs_put_page(fio.encrypted_page, true);
+put_cache:
+	f2fs_put_cache(entry, true);
 put_folio:
 	f2fs_folio_put(folio, true);
 out_iput:
@@ -1320,7 +1321,7 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
 }
 
 /*
- * Move data block via META_MAPPING while keeping locked data page.
+ * Move data block via meta cache while keeping locked data page.
  * This can be used to move blocks, aka LBAs, directly on disk.
  */
 static int move_data_block(struct inode *inode, block_t bidx,
@@ -1337,11 +1338,13 @@ static int move_data_block(struct inode *inode, block_t bidx,
 		.op_flags = 0,
 		.encrypted_page = NULL,
 		.in_list = 0,
+		.is_cache = 1,
 	};
 	struct dnode_of_data dn;
 	struct f2fs_summary sum;
 	struct node_info ni;
-	struct folio *folio, *mfolio, *efolio;
+	struct folio *folio;
+	struct f2fs_cached_block *sentry, *tentry;
 	block_t newaddr;
 	int err = 0;
 	bool lfs_mode = f2fs_lfs_mode(fio.sbi);
@@ -1406,20 +1409,20 @@ static int move_data_block(struct inode *inode, block_t bidx,
 	if (lfs_mode)
 		f2fs_down_write(&fio.sbi->io_order_lock);
 
-	mfolio = f2fs_grab_cache_folio(META_MAPPING(fio.sbi),
-					fio.old_blkaddr, false);
-	if (IS_ERR(mfolio)) {
-		err = PTR_ERR(mfolio);
+	sentry = f2fs_grab_cache(META_CACHE(fio.sbi), fio.old_blkaddr,
+						F2FS_CACHE_LOCK_CREATE);
+	if (IS_ERR(sentry)) {
+		err = PTR_ERR(sentry);
 		goto up_out;
 	}
 
-	fio.encrypted_page = folio_file_page(mfolio, fio.old_blkaddr);
+	fio.cache_entry = sentry;
 
-	/* read source block in mfolio */
-	if (!folio_test_uptodate(mfolio)) {
-		err = f2fs_submit_page_bio(&fio);
+	/* read source block in cache */
+	if (!f2fs_cache_test_uptodate(sentry)) {
+		err = f2fs_submit_cache_read(&fio);
 		if (err) {
-			f2fs_folio_put(mfolio, true);
+			f2fs_put_cache(sentry, true);
 			goto up_out;
 		}
 
@@ -1428,11 +1431,11 @@ static int move_data_block(struct inode *inode, block_t bidx,
 		f2fs_update_iostat(fio.sbi, NULL, FS_GDATA_READ_IO,
 							F2FS_BLKSIZE);
 
-		folio_lock(mfolio);
-		if (unlikely(!is_meta_folio(mfolio) ||
-			     !folio_test_uptodate(mfolio))) {
+		f2fs_lock_cache(sentry);
+		if (unlikely(!f2fs_is_meta_cache(sentry) ||
+				!f2fs_cache_test_uptodate(sentry))) {
 			err = -EIO;
-			f2fs_folio_put(mfolio, true);
+			f2fs_put_cache(sentry, true);
 			goto up_out;
 		}
 	}
@@ -1443,46 +1446,45 @@ static int move_data_block(struct inode *inode, block_t bidx,
 	err = f2fs_allocate_data_block(fio.sbi, NULL, fio.old_blkaddr, &newaddr,
 				&sum, type, NULL);
 	if (err) {
-		f2fs_folio_put(mfolio, true);
+		f2fs_put_cache(sentry, true);
 		/* filesystem should shutdown, no need to recovery block */
 		goto up_out;
 	}
 
-	efolio = f2fs_filemap_get_folio(META_MAPPING(fio.sbi), newaddr,
-					FGP_LOCK | FGP_CREAT, GFP_NOFS);
-	if (IS_ERR(efolio)) {
-		err = PTR_ERR(efolio);
-		f2fs_folio_put(mfolio, true);
+	tentry = f2fs_grab_cache(META_CACHE(fio.sbi), newaddr,
+					F2FS_CACHE_LOCK_CREATE);
+	if (IS_ERR(tentry)) {
+		err = PTR_ERR(tentry);
+		f2fs_put_cache(sentry, true);
 		goto recover_block;
 	}
 
-	fio.encrypted_page = &efolio->page;
+	fio.cache_entry = tentry;
 
 	/* write target block */
-	f2fs_wait_on_page_writeback(fio.encrypted_page, DATA, true, true);
-	memcpy(page_address(fio.encrypted_page),
-				folio_address(mfolio), PAGE_SIZE);
-	f2fs_folio_put(mfolio, true);
+	f2fs_cache_wait_writeback_cond(tentry, DATA);
+	memcpy(cache_address(tentry), cache_address(sentry), PAGE_SIZE);
+	f2fs_put_cache(sentry, true);
 
 	f2fs_invalidate_internal_cache(fio.sbi, fio.old_blkaddr, 1);
 
-	set_page_dirty(fio.encrypted_page);
-	if (clear_page_dirty_for_io(fio.encrypted_page))
+	f2fs_mark_cache_dirty(tentry);
+	if (f2fs_clear_cache_dirty(tentry))
 		dec_page_count(fio.sbi, F2FS_DIRTY_META);
 
-	set_page_writeback(fio.encrypted_page);
+	f2fs_start_cache_writeback(tentry);
 
 	fio.op = REQ_OP_WRITE;
 	fio.op_flags = REQ_SYNC;
 	fio.new_blkaddr = newaddr;
-	f2fs_submit_page_write(&fio);
+	f2fs_submit_cache_write(&fio);
 
 	f2fs_update_iostat(fio.sbi, NULL, FS_GC_DATA_IO, F2FS_BLKSIZE);
 
 	f2fs_update_data_blkaddr(&dn, newaddr);
 	set_inode_flag(inode, FI_APPEND_WRITE);
 
-	f2fs_put_page(fio.encrypted_page, true);
+	f2fs_put_cache(tentry, true);
 recover_block:
 	if (err)
 		f2fs_do_replace_block(fio.sbi, &sum, newaddr, fio.old_blkaddr,
@@ -1614,7 +1616,7 @@ static int gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
 			continue;
 
 		if (phase == 0) {
-			f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), 1,
+			f2fs_ra_meta_caches(sbi, NAT_BLOCK_OFFSET(nid), 1,
 							META_NAT, true);
 			continue;
 		}
@@ -1818,28 +1820,31 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
 	sum_blk_cnt = DIV_ROUND_UP(end_segno - segno, sbi->sums_per_block);
 	/* readahead multi ssa blocks those have contiguous address */
 	if (__is_large_section(sbi))
-		f2fs_ra_meta_pages(sbi, GET_SUM_BLOCK(sbi, segno),
+		f2fs_ra_meta_caches(sbi, GET_SUM_BLOCK(sbi, segno),
 					sum_blk_cnt, META_SSA, true);
 
 	/* reference all summary page */
 	while (segno < end_segno) {
-		struct folio *sum_folio = f2fs_get_sum_folio(sbi, segno);
+		struct f2fs_cached_block *sum_entry =
+				f2fs_get_sum_cache(sbi, segno);
 
 		segno += sbi->sums_per_block;
-		if (IS_ERR(sum_folio)) {
-			int err = PTR_ERR(sum_folio);
+		if (IS_ERR(sum_entry)) {
+			int err = PTR_ERR(sum_entry);
 
 			end_segno = segno - sbi->sums_per_block;
 			segno = rounddown(start_segno, sbi->sums_per_block);
 			while (segno < end_segno) {
-				sum_folio = filemap_get_folio(META_MAPPING(sbi),
+				sum_entry = f2fs_find_meta_cache(sbi,
 						GET_SUM_BLOCK(sbi, segno));
-				folio_put_refs(sum_folio, 2);
+				f2fs_put_cache(sum_entry, false);
+				f2fs_put_cache(sum_entry, false);
 				segno += sbi->sums_per_block;
 			}
 			return err;
 		}
-		folio_unlock(sum_folio);
+		f2fs_unlock_cache(sum_entry);
+
 	}
 
 	blk_start_plug(&plug);
@@ -1847,11 +1852,16 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
 	segno = start_segno;
 	while (segno < end_segno) {
 		unsigned int cur_segno;
+		unsigned int block_end_segno;
 
 		/* find segment summary of victim */
-		struct folio *sum_folio = filemap_get_folio(META_MAPPING(sbi),
+		struct f2fs_cached_block *sum_entry =
+				f2fs_find_meta_cache(sbi,
 					GET_SUM_BLOCK(sbi, segno));
-		unsigned int block_end_segno = rounddown(segno, sbi->sums_per_block)
+
+		f2fs_bug_on(sbi, IS_ERR(sum_entry));
+
+		block_end_segno = rounddown(segno, sbi->sums_per_block)
 					+ sbi->sums_per_block;
 
 		if (block_end_segno > end_segno)
@@ -1864,8 +1874,8 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
 			goto next_block;
 		}
 
-		if (!folio_test_uptodate(sum_folio) ||
-		    unlikely(f2fs_cp_error(sbi)))
+		if (!f2fs_cache_test_uptodate(sum_entry) ||
+			unlikely(f2fs_cp_error(sbi)))
 			goto next_block;
 
 		for (cur_segno = segno; cur_segno < block_end_segno;
@@ -1884,7 +1894,7 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
 				data_type = (type == SUM_TYPE_DATA) ? DATA : NODE;
 			}
 
-			sum = SUM_BLK_PAGE_ADDR(sbi, sum_folio, cur_segno);
+			sum = SUM_BLK_ENTRY_ADDR(sbi, sum_entry, cur_segno);
 			if (type != GET_SUM_TYPE(sum_footer(sbi, sum))) {
 				f2fs_err(sbi, "Inconsistent segment (%u) type "
 						"[%d, %d] in SIT and SSA",
@@ -1926,12 +1936,14 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
 					cur_segno + 1 : NULL_SEGNO;
 
 			if (unlikely(freezing(current))) {
-				folio_put_refs(sum_folio, 2);
+				f2fs_put_cache(sum_entry, false);
+				f2fs_put_cache(sum_entry, false);
 				goto stop;
 			}
 		}
 next_block:
-		folio_put_refs(sum_folio, 2);
+		f2fs_put_cache(sum_entry, false);
+		f2fs_put_cache(sum_entry, false);
 		segno = block_end_segno;
 	}
 
fs/f2fs/inline.c +5 -4
index aec06fb4fd76..2156fb1fc57d 100644
@@ -294,7 +294,7 @@ int f2fs_write_inline_data(struct inode *inode, struct folio *folio)
 	return 0;
 }
 
-int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
+int f2fs_recover_inline_data(struct inode *inode, struct f2fs_cached_block *entry)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct f2fs_inode *ri = NULL;
@@ -308,12 +308,13 @@ int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
 	 *    x       o  -> remove data blocks, and then recover inline_data
 	 *    x       x  -> recover data blocks
 	 */
-	if (IS_INODE(nfolio))
-		ri = F2FS_INODE(nfolio);
+	if (IS_INODE(cache_folio(entry)))
+		ri = &CACHED_NODE(entry)->i;
 
 	if (f2fs_has_inline_data(inode) &&
 			ri && (ri->i_inline & F2FS_INLINE_DATA)) {
 		struct folio *ifolio;
+
 process_inline:
 		ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
 		if (IS_ERR(ifolio))
@@ -321,7 +322,7 @@ int f2fs_recover_inline_data(struct inode *inode, struct folio *nfolio)
 
 		f2fs_folio_wait_writeback(ifolio, NODE, true, true);
 
-		src_addr = inline_data_addr(inode, nfolio);
+		src_addr = inline_data_addr(inode, cache_folio(entry));
 		dst_addr = inline_data_addr(inode, ifolio);
 		memcpy(dst_addr, src_addr, MAX_INLINE_DATA(inode));
 
fs/f2fs/inode.c +2 -7
index bac1e360d966..c533da4d4d70 100644
@@ -577,7 +577,7 @@ static int do_read_inode(struct inode *inode)
 
 static bool is_meta_ino(struct f2fs_sb_info *sbi, unsigned int ino)
 {
-	if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
+	if (ino == F2FS_NODE_INO(sbi))
 		return true;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (test_opt(sbi, COMPRESS_CACHE) && ino == F2FS_COMPRESS_INO(sbi))
@@ -624,9 +624,6 @@ struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 	if (ino == F2FS_NODE_INO(sbi)) {
 		inode->i_mapping->a_ops = &f2fs_node_aops;
 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
-	} else if (ino == F2FS_META_INO(sbi)) {
-		inode->i_mapping->a_ops = &f2fs_meta_aops;
-		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
 	} else if (ino == F2FS_COMPRESS_INO(sbi)) {
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 		inode->i_mapping->a_ops = &f2fs_compress_aops;
@@ -824,8 +821,7 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi))
+	if (inode->i_ino == F2FS_NODE_INO(sbi))
 		return 0;
 
 	/*
@@ -919,7 +915,6 @@ static bool f2fs_pre_evict_inode(struct inode *inode)
 		f2fs_invalidate_compress_pages(sbi, inode->i_ino);
 
 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-	    inode->i_ino == F2FS_META_INO(sbi) ||
 	    inode->i_ino == F2FS_COMPRESS_INO(sbi))
 		return true;
 
fs/f2fs/node.c +62 -63
index 46bea52e35c3..398c58fc6cad 100644
@@ -140,38 +140,36 @@ static void clear_node_folio_dirty(struct folio *folio)
 	folio_clear_uptodate(folio);
 }
 
-static struct folio *get_current_nat_folio(struct f2fs_sb_info *sbi, nid_t nid)
+static struct f2fs_cached_block *get_current_nat_cache(struct f2fs_sb_info *sbi,
+		nid_t nid)
 {
-	return f2fs_get_meta_folio_retry(sbi, current_nat_addr(sbi, nid));
+	return f2fs_get_meta_cache_retry(sbi, current_nat_addr(sbi, nid));
 }
 
-static struct folio *get_next_nat_folio(struct f2fs_sb_info *sbi, nid_t nid)
+static struct f2fs_cached_block *get_next_nat_cache(struct f2fs_sb_info *sbi,
+		nid_t nid)
 {
-	struct folio *src_folio;
-	struct folio *dst_folio;
+	struct f2fs_cached_block *src_entry;
+	struct f2fs_cached_block *dst_entry;
 	pgoff_t dst_off;
-	void *src_addr;
-	void *dst_addr;
 	struct f2fs_nm_info *nm_i = NM_I(sbi);
 
 	dst_off = next_nat_addr(sbi, current_nat_addr(sbi, nid));
 
 	/* get current nat block page with lock */
-	src_folio = get_current_nat_folio(sbi, nid);
-	if (IS_ERR(src_folio))
-		return src_folio;
-	dst_folio = f2fs_grab_meta_folio(sbi, dst_off);
-	f2fs_bug_on(sbi, folio_test_dirty(src_folio));
-
-	src_addr = folio_address(src_folio);
-	dst_addr = folio_address(dst_folio);
-	memcpy(dst_addr, src_addr, PAGE_SIZE);
-	folio_mark_dirty(dst_folio);
-	f2fs_folio_put(src_folio, true);
+	src_entry = get_current_nat_cache(sbi, nid);
+	if (IS_ERR(src_entry))
+		return src_entry;
+	dst_entry = f2fs_grab_meta_cache(sbi, dst_off);
+	f2fs_bug_on(sbi, f2fs_cache_test_dirty(src_entry));
+
+	memcpy(cache_address(dst_entry), cache_address(src_entry), sbi->blocksize);
+	f2fs_mark_cache_dirty(dst_entry);
+	f2fs_put_cache(src_entry, true);
 
 	set_to_next_nat(nm_i, nid);
 
-	return dst_folio;
+	return dst_entry;
 }
 
 static struct nat_entry *__alloc_nat_entry(struct f2fs_sb_info *sbi,
@@ -575,7 +573,7 @@ int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
 	struct f2fs_journal *journal = curseg->journal;
 	nid_t start_nid = START_NID(nid);
 	struct f2fs_nat_block *nat_blk;
-	struct folio *folio = NULL;
+	struct f2fs_cached_block *entry = NULL;
 	struct f2fs_nat_entry ne;
 	struct nat_entry *e;
 	pgoff_t index;
@@ -629,14 +627,14 @@ int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
 	index = current_nat_addr(sbi, nid);
 	f2fs_up_read(&nm_i->nat_tree_lock);
 
-	folio = f2fs_get_meta_folio(sbi, index);
-	if (IS_ERR(folio))
-		return PTR_ERR(folio);
+	entry = f2fs_get_meta_cache(sbi, index);
+	if (IS_ERR(entry))
+		return PTR_ERR(entry);
 
-	nat_blk = folio_address(folio);
+	nat_blk = cache_address(entry);
 	ne = nat_blk->entries[nid - start_nid];
 	node_info_from_raw_nat(ni, &ne);
-	f2fs_folio_put(folio, true);
+	f2fs_put_cache(entry, true);
 sanity_check:
 	if (__is_valid_data_blkaddr(ni->blk_addr) &&
 		!f2fs_is_valid_blkaddr(sbi, ni->blk_addr,
@@ -1622,7 +1620,7 @@ static struct folio *__get_node_folio(struct f2fs_sb_info *sbi, pgoff_t nid,
 out_put_err:
 	/* ENOENT comes from read_node_folio which is not an error. */
 	if (err != -ENOENT)
-		f2fs_handle_page_eio(sbi, folio, NODE);
+		f2fs_handle_page_eio(sbi, folio->index, NODE);
 	f2fs_folio_put(folio, true);
 	return ERR_PTR(err);
 }
@@ -2598,6 +2596,7 @@ static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
 						bool sync, bool mount)
 {
 	struct f2fs_nm_info *nm_i = NM_I(sbi);
+	struct f2fs_cached_block *entry = NULL;
 	int i = 0, ret;
 	nid_t nid = nm_i->next_scan_nid;
 
@@ -2623,7 +2622,7 @@ static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
 	}
 
 	/* readahead nat pages to be scanned */
-	f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
+	f2fs_ra_meta_caches(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
 							META_NAT, true);
 
 	f2fs_down_read(&nm_i->nat_tree_lock);
@@ -2631,14 +2630,14 @@ static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
 	while (1) {
 		if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
 						nm_i->nat_block_bitmap)) {
-			struct folio *folio = get_current_nat_folio(sbi, nid);
+			entry = get_current_nat_cache(sbi, nid);
 
-			if (IS_ERR(folio)) {
-				ret = PTR_ERR(folio);
+			if (IS_ERR(entry)) {
+				ret = PTR_ERR(entry);
 			} else {
-				ret = scan_nat_page(sbi, folio_address(folio),
+				ret = scan_nat_page(sbi, cache_address(entry),
 						nid);
-				f2fs_folio_put(folio, true);
+				f2fs_put_cache(entry, true);
 			}
 
 			if (ret) {
@@ -2671,7 +2670,7 @@ static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
 
 	f2fs_up_read(&nm_i->nat_tree_lock);
 
-	f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
+	f2fs_ra_meta_caches(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
 					nm_i->ra_nid_pages, META_NAT, false);
 
 	return 0;
@@ -2826,7 +2825,7 @@ int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
 	return nr - nr_shrink;
 }
 
-int f2fs_recover_inline_xattr(struct inode *inode, struct folio *folio)
+int f2fs_recover_inline_xattr(struct inode *inode, struct f2fs_cached_block *entry)
 {
 	void *src_addr, *dst_addr;
 	size_t inline_size;
@@ -2837,7 +2836,7 @@ int f2fs_recover_inline_xattr(struct inode *inode, struct folio *folio)
 	if (IS_ERR(ifolio))
 		return PTR_ERR(ifolio);
 
-	ri = F2FS_INODE(folio);
+	ri = &CACHED_NODE(entry)->i;
 	if (ri->i_inline & F2FS_INLINE_XATTR) {
 		if (!f2fs_has_inline_xattr(inode)) {
 			set_inode_flag(inode, FI_INLINE_XATTR);
@@ -2852,7 +2851,7 @@ int f2fs_recover_inline_xattr(struct inode *inode, struct folio *folio)
 	}
 
 	dst_addr = inline_xattr_addr(inode, ifolio);
-	src_addr = inline_xattr_addr(inode, folio);
+	src_addr = inline_xattr_addr(inode, cache_folio(entry));
 	inline_size = inline_xattr_size(inode);
 
 	f2fs_folio_wait_writeback(ifolio, NODE, true, true);
@@ -2863,7 +2862,7 @@ int f2fs_recover_inline_xattr(struct inode *inode, struct folio *folio)
 	return 0;
 }
 
-int f2fs_recover_xattr_data(struct inode *inode, struct folio *folio)
+int f2fs_recover_xattr_data(struct inode *inode, struct f2fs_cached_block *entry)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
@@ -2901,8 +2900,8 @@ int f2fs_recover_xattr_data(struct inode *inode, struct folio *folio)
 	f2fs_update_inode_page(inode);
 
 	/* 3: update and set xattr node page dirty */
-	if (folio) {
-		memcpy(F2FS_NODE(xfolio), F2FS_NODE(folio),
+	if (entry) {
+		memcpy(F2FS_NODE(xfolio), CACHED_NODE(entry),
 				VALID_XATTR_BLOCK_SIZE);
 		folio_mark_dirty(xfolio);
 	}
@@ -2911,10 +2910,10 @@ int f2fs_recover_xattr_data(struct inode *inode, struct folio *folio)
 	return 0;
 }
 
-int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct folio *folio)
+int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct f2fs_cached_block *entry)
 {
 	struct f2fs_inode *src, *dst;
-	nid_t ino = ino_of_node(folio);
+	nid_t ino = ino_of_node(cache_folio(entry));
 	struct node_info old_ni, new_ni;
 	struct folio *ifolio;
 	int err;
@@ -2940,7 +2939,7 @@ int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct folio *folio)
 	fill_node_footer(ifolio, ino, ino, 0, true);
 	set_cold_node(ifolio, false);
 
-	src = F2FS_INODE(folio);
+	src = &CACHED_NODE(entry)->i;
 	dst = F2FS_INODE(ifolio);
 
 	memcpy(dst, src, offsetof(struct f2fs_inode, i_ext));
@@ -2999,24 +2998,24 @@ int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
 		nrpages = bio_max_segs(last_offset - i);
 
 		/* readahead node pages */
-		f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true);
+		f2fs_ra_meta_caches(sbi, addr, nrpages, META_POR, true);
 
 		for (idx = addr; idx < addr + nrpages; idx++) {
-			struct folio *folio = f2fs_get_tmp_folio(sbi, idx);
+			struct f2fs_cached_block *entry =
+				f2fs_get_tmp_cache(sbi, idx);
 
-			if (IS_ERR(folio))
-				return PTR_ERR(folio);
+			if (IS_ERR(entry))
+				return PTR_ERR(entry);
 
-			rn = F2FS_NODE(folio);
+			rn = CACHED_NODE(entry);
 			sum_entry->nid = rn->footer.nid;
 			sum_entry->version = 0;
 			sum_entry->ofs_in_node = 0;
 			sum_entry++;
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 		}
 
-		invalidate_mapping_pages(META_MAPPING(sbi), addr,
-							addr + nrpages);
+		f2fs_truncate_meta_caches(sbi, addr, nrpages);
 	}
 	return 0;
 }
@@ -3126,7 +3125,7 @@ static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
 	bool to_journal = true;
 	struct f2fs_nat_block *nat_blk;
 	struct nat_entry *ne, *cur;
-	struct folio *folio = NULL;
+	struct f2fs_cached_block *entry = NULL;
 
 	/*
 	 * there are two steps to flush nat entries:
@@ -3140,11 +3139,11 @@ static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
 	if (to_journal) {
 		down_write(&curseg->journal_rwsem);
 	} else {
-		folio = get_next_nat_folio(sbi, start_nid);
-		if (IS_ERR(folio))
-			return PTR_ERR(folio);
+		entry = get_next_nat_cache(sbi, start_nid);
+		if (IS_ERR(entry))
+			return PTR_ERR(entry);
 
-		nat_blk = folio_address(folio);
+		nat_blk = cache_address(entry);
 		f2fs_bug_on(sbi, !nat_blk);
 	}
 
@@ -3181,7 +3180,7 @@ static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
 		up_write(&curseg->journal_rwsem);
 	} else {
 		__update_nat_bits(sbi, start_nid, nat_blk);
-		f2fs_folio_put(folio, true);
+		f2fs_put_cache(entry, true);
 	}
 
 	/* Allow dirty nats by node block allocation in write_begin */
@@ -3252,7 +3251,7 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 			__has_cursum_space(sbi, journal,
 					entry_count, NAT_JOURNAL))
 			continue;
-		f2fs_ra_meta_pages(sbi, set->set, 1, META_NAT, true);
+		f2fs_ra_meta_caches(sbi, set->set, 1, META_NAT, true);
 	}
 	/* flush dirty nats in nat entry set */
 	list_for_each_entry_safe(set, tmp, &sets, set_list) {
@@ -3288,15 +3287,15 @@ static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
 	nat_bits_addr = __start_cp_addr(sbi) + BLKS_PER_SEG(sbi) -
 						nm_i->nat_bits_blocks;
 	for (i = 0; i < nm_i->nat_bits_blocks; i++) {
-		struct folio *folio;
+		struct f2fs_cached_block *entry;
 
-		folio = f2fs_get_meta_folio(sbi, nat_bits_addr++);
-		if (IS_ERR(folio))
-			return PTR_ERR(folio);
+		entry = f2fs_get_meta_cache(sbi, nat_bits_addr++);
+		if (IS_ERR(entry))
+			return PTR_ERR(entry);
 
 		memcpy(nm_i->nat_bits + F2FS_BLK_TO_BYTES(i),
-					folio_address(folio), F2FS_BLKSIZE);
-		f2fs_folio_put(folio, true);
+					cache_address(entry), F2FS_BLKSIZE);
+		f2fs_put_cache(entry, true);
 	}
 
 	cp_ver |= (cur_cp_crc(ckpt) << 32);
fs/f2fs/node.h +2 -2
index 5e114f352099..de8dcd5d4392 100644
@@ -311,9 +311,9 @@ static inline void fill_node_footer_blkaddr(struct folio *folio, block_t blkaddr
 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
 }
 
-static inline bool is_recoverable_dnode(const struct folio *folio)
+static inline bool is_recoverable_dnode(struct f2fs_sb_info *sbi, const struct folio *folio)
 {
-	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_F_SB(folio));
+	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 	__u64 cp_ver = cur_cp_version(ckpt);
 
 	/* Don't care crc part, if fsck.f2fs sets it. */
fs/f2fs/recovery.c +100 -98
index aaa5227739c8..5cb56b4c1879 100644
@@ -182,33 +182,33 @@ static const char *recover_printable_name(struct inode *inode,
 	return raw->i_name;
 }
 
-static int recover_dentry(struct inode *inode, struct folio *ifolio,
+static int recover_dentry(struct inode *inode, struct f2fs_cached_block *entry,
 						struct list_head *dir_list)
 {
-	struct f2fs_inode *raw_inode = F2FS_INODE(ifolio);
+	struct f2fs_inode *raw_inode = &CACHED_NODE(entry)->i;
 	nid_t pino = le32_to_cpu(raw_inode->i_pino);
 	struct f2fs_dir_entry *de;
 	struct f2fs_filename fname;
 	struct qstr usr_fname;
 	struct folio *folio;
 	struct inode *dir, *einode;
-	struct fsync_inode_entry *entry;
+	struct fsync_inode_entry *fsync_entry;
 	int err = 0;
 	const char *name;
 	int name_len;
 
-	entry = get_fsync_inode(dir_list, pino);
-	if (!entry) {
-		entry = add_fsync_inode(F2FS_I_SB(inode), dir_list,
+	fsync_entry = get_fsync_inode(dir_list, pino);
+	if (!fsync_entry) {
+		fsync_entry = add_fsync_inode(F2FS_I_SB(inode), dir_list,
 							pino, false);
-		if (IS_ERR(entry)) {
-			dir = ERR_CAST(entry);
-			err = PTR_ERR(entry);
+		if (IS_ERR(fsync_entry)) {
+			dir = ERR_CAST(fsync_entry);
+			err = PTR_ERR(fsync_entry);
 			goto out;
 		}
 	}
 
-	dir = entry->inode;
+	dir = fsync_entry->inode;
 	err = init_recovered_filename(dir, inode, raw_inode, &fname, &usr_fname);
 	if (err)
 		goto out;
@@ -256,14 +256,14 @@ static int recover_dentry(struct inode *inode, struct folio *ifolio,
 out:
 	name = recover_printable_name(inode, raw_inode, &name_len);
 	f2fs_notice(F2FS_I_SB(inode), "%s: ino = %x, name = %.*s, dir = %llu, err = %d",
-		    __func__, ino_of_node(ifolio), name_len, name,
+		    __func__, ino_of_node(cache_folio(entry)), name_len, name,
 		    IS_ERR(dir) ? 0 : dir->i_ino, err);
 	return err;
 }
 
-static int recover_quota_data(struct inode *inode, struct folio *folio)
+static int recover_quota_data(struct inode *inode, struct f2fs_cached_block *entry)
 {
-	struct f2fs_inode *raw = F2FS_INODE(folio);
+	struct f2fs_inode *raw = &CACHED_NODE(entry)->i;
 	struct iattr attr;
 	uid_t i_uid = le32_to_cpu(raw->i_uid);
 	gid_t i_gid = le32_to_cpu(raw->i_gid);
@@ -300,9 +300,9 @@ static void recover_inline_flags(struct inode *inode, struct f2fs_inode *ri)
 		clear_inode_flag(inode, FI_DATA_EXIST);
 }
 
-static int recover_inode(struct inode *inode, struct folio *folio)
+static int recover_inode(struct inode *inode, struct f2fs_cached_block *entry)
 {
-	struct f2fs_inode *raw = F2FS_INODE(folio);
+	struct f2fs_inode *raw = &CACHED_NODE(entry)->i;
 	struct f2fs_inode_info *fi = F2FS_I(inode);
 	const char *name;
 	int name_len;
@@ -310,7 +310,7 @@ static int recover_inode(struct inode *inode, struct folio *folio)
 
 	inode->i_mode = le16_to_cpu(raw->i_mode);
 
-	err = recover_quota_data(inode, folio);
+	err = recover_quota_data(inode, entry);
 	if (err)
 		return err;
 
@@ -357,7 +357,7 @@ static int recover_inode(struct inode *inode, struct folio *folio)
 	name = recover_printable_name(inode, raw, &name_len);
 
 	f2fs_notice(F2FS_I_SB(inode), "%s: ino = %x, name = %.*s, inline = %x",
-		    __func__, ino_of_node(folio), name_len, name,
+		    __func__, ino_of_node(cache_folio(entry)), name_len, name,
 		    raw->i_inline);
 	return 0;
 }
@@ -386,30 +386,30 @@ static int sanity_check_node_chain(struct f2fs_sb_info *sbi, block_t blkaddr,
 		return 0;
 
 	for (i = 0; i < 2; i++) {
-		struct folio *folio;
+		struct f2fs_cached_block *entry;
 
 		if (!f2fs_is_valid_blkaddr(sbi, *blkaddr_fast, META_POR)) {
 			*is_detecting = false;
 			return 0;
 		}
 
-		folio = f2fs_get_tmp_folio(sbi, *blkaddr_fast);
-		if (IS_ERR(folio))
-			return PTR_ERR(folio);
+		entry = f2fs_get_tmp_cache(sbi, *blkaddr_fast);
+		if (IS_ERR(entry))
+			return PTR_ERR(entry);
 
-		if (!is_recoverable_dnode(folio)) {
-			f2fs_folio_put(folio, true);
+		if (!is_recoverable_dnode(sbi, cache_folio(entry))) {
+			f2fs_put_cache(entry, true);
 			*is_detecting = false;
 			return 0;
 		}
 
 		ra_blocks = adjust_por_ra_blocks(sbi, ra_blocks, *blkaddr_fast,
-					next_blkaddr_of_node(folio));
+					next_blkaddr_of_node(cache_folio(entry)));
 
-		*blkaddr_fast = next_blkaddr_of_node(folio);
-		f2fs_folio_put(folio, true);
+		*blkaddr_fast = next_blkaddr_of_node(cache_folio(entry));
+		f2fs_put_cache(entry, true);
 
-		f2fs_ra_meta_pages_cond(sbi, *blkaddr_fast, ra_blocks);
+		f2fs_ra_meta_caches_cond(sbi, *blkaddr_fast, ra_blocks);
 	}
 
 	if (*blkaddr_fast == blkaddr) {
@@ -434,45 +434,47 @@ static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
 	blkaddr_fast = blkaddr;
 
 	while (1) {
-		struct fsync_inode_entry *entry;
-		struct folio *folio;
+		struct fsync_inode_entry *fsync_entry;
+		struct f2fs_cached_block *entry;
+		struct f2fs_node *rn;
 
 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
 			return 0;
 
-		folio = f2fs_get_tmp_folio(sbi, blkaddr);
-		if (IS_ERR(folio)) {
-			err = PTR_ERR(folio);
+		entry = f2fs_get_tmp_cache(sbi, blkaddr);
+		if (IS_ERR(entry)) {
+			err = PTR_ERR(entry);
 			break;
 		}
+		rn = CACHED_NODE(entry);
 
-		if (!is_recoverable_dnode(folio)) {
-			f2fs_folio_put(folio, true);
+		if (!is_recoverable_dnode(sbi, cache_folio(entry))) {
+			f2fs_put_cache(entry, true);
 			break;
 		}
 
-		if (!is_fsync_dnode(folio))
+		if (!is_fsync_dnode(cache_folio(entry)))
 			goto next;
 
-		entry = get_fsync_inode(head, ino_of_node(folio));
-		if (!entry) {
+		fsync_entry = get_fsync_inode(head, ino_of_node(cache_folio(entry)));
+		if (!fsync_entry) {
 			bool quota_inode = false;
 
 			if (!check_only &&
-					IS_INODE(folio) &&
-					is_dent_dnode(folio)) {
-				err = f2fs_recover_inode_page(sbi, folio);
+					IS_INODE(cache_folio(entry)) &&
+					is_dent_dnode(cache_folio(entry))) {
+				err = f2fs_recover_inode_page(sbi, entry);
 				if (err) {
-					f2fs_folio_put(folio, true);
+					f2fs_put_cache(entry, true);
 					break;
 				}
 				quota_inode = true;
 			}
 
-			entry = add_fsync_inode(sbi, head, ino_of_node(folio),
+			fsync_entry = add_fsync_inode(sbi, head, ino_of_node(cache_folio(entry)),
 								quota_inode);
-			if (IS_ERR(entry)) {
-				err = PTR_ERR(entry);
+			if (IS_ERR(fsync_entry)) {
+				err = PTR_ERR(fsync_entry);
 				/*
 				 * CP | dnode(F) | inode(DF)
 				 * For this case, we should not give up now.
@@ -482,18 +484,18 @@ static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
 						*new_inode = true;
 					goto next;
 				}
-				f2fs_folio_put(folio, true);
+				f2fs_put_cache(entry, true);
 				break;
 			}
 		}
-		entry->blkaddr = blkaddr;
+		fsync_entry->blkaddr = blkaddr;
 
-		if (IS_INODE(folio) && is_dent_dnode(folio))
-			entry->last_dentry = blkaddr;
+		if (IS_INODE(cache_folio(entry)) && is_dent_dnode(cache_folio(entry)))
+			fsync_entry->last_dentry = blkaddr;
 next:
 		/* check next segment */
-		blkaddr = next_blkaddr_of_node(folio);
-		f2fs_folio_put(folio, true);
+		blkaddr = next_blkaddr_of_node(cache_folio(entry));
+		f2fs_put_cache(entry, true);
 
 		err = sanity_check_node_chain(sbi, blkaddr, &blkaddr_fast,
 				&is_detecting);
@@ -519,7 +521,8 @@ static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi,
 	unsigned short blkoff = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
 	struct f2fs_summary_block *sum_node;
 	struct f2fs_summary sum;
-	struct folio *sum_folio, *node_folio;
+	struct f2fs_cached_block *entry = NULL;
+	struct folio *node_folio;
 	struct dnode_of_data tdn = *dn;
 	nid_t ino, nid;
 	struct inode *inode;
@@ -541,12 +544,12 @@ static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi,
 		}
 	}
 
-	sum_folio = f2fs_get_sum_folio(sbi, segno);
-	if (IS_ERR(sum_folio))
-		return PTR_ERR(sum_folio);
-	sum_node = SUM_BLK_PAGE_ADDR(sbi, sum_folio, segno);
+	entry = f2fs_get_sum_cache(sbi, segno);
+	if (IS_ERR(entry))
+		return PTR_ERR(entry);
+	sum_node = SUM_BLK_ENTRY_ADDR(sbi, entry, segno);
 	sum = sum_entries(sum_node)[blkoff];
-	f2fs_folio_put(sum_folio, true);
+	f2fs_put_cache(entry, true);
 got_it:
 	/* Use the locked dnode page and inode */
 	nid = le32_to_cpu(sum.nid);
@@ -645,7 +648,7 @@ static int f2fs_reserve_new_block_retry(struct dnode_of_data *dn)
 }
 
 static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
-					struct folio *folio)
+					struct f2fs_cached_block *entry)
 {
 	struct dnode_of_data dn;
 	struct node_info ni;
@@ -653,19 +656,19 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 	int err = 0, recovered = 0;
 
 	/* step 1: recover xattr */
-	if (IS_INODE(folio)) {
-		err = f2fs_recover_inline_xattr(inode, folio);
+	if (IS_INODE(cache_folio(entry))) {
+		err = f2fs_recover_inline_xattr(inode, entry);
 		if (err)
 			goto out;
-	} else if (f2fs_has_xattr_block(ofs_of_node(folio))) {
-		err = f2fs_recover_xattr_data(inode, folio);
+	} else if (f2fs_has_xattr_block(ofs_of_node(cache_folio(entry)))) {
+		err = f2fs_recover_xattr_data(inode, entry);
 		if (!err)
 			recovered++;
 		goto out;
 	}
 
 	/* step 2: recover inline data */
-	err = f2fs_recover_inline_data(inode, folio);
+	err = f2fs_recover_inline_data(inode, entry);
 	if (err) {
 		if (err == 1)
 			err = 0;
@@ -673,8 +676,8 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 	}
 
 	/* step 3: recover data indices */
-	start = f2fs_start_bidx_of_node(ofs_of_node(folio), inode);
-	end = start + ADDRS_PER_PAGE(folio, inode);
+	start = f2fs_start_bidx_of_node(ofs_of_node(cache_folio(entry)), inode);
+	end = start + addrs_per_page(inode, IS_INODE(cache_folio(entry)));
 
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
 retry_dn:
@@ -693,12 +696,12 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 	if (err)
 		goto err;
 
-	f2fs_bug_on(sbi, ni.ino != ino_of_node(folio));
+	f2fs_bug_on(sbi, ni.ino != ino_of_node(cache_folio(entry)));
 
-	if (ofs_of_node(dn.node_folio) != ofs_of_node(folio)) {
+	if (ofs_of_node(dn.node_folio) != ofs_of_node(cache_folio(entry))) {
 		f2fs_warn(sbi, "Inconsistent ofs_of_node, ino:%llu, ofs:%u, %u",
 			  inode->i_ino, ofs_of_node(dn.node_folio),
-			  ofs_of_node(folio));
+			  ofs_of_node(cache_folio(entry)));
 		err = -EFSCORRUPTED;
 		f2fs_handle_error(sbi, ERROR_INCONSISTENT_FOOTER);
 		fserror_report_file_metadata(dn.inode, err, GFP_NOFS);
@@ -709,7 +712,7 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 		block_t src, dest;
 
 		src = f2fs_data_blkaddr(&dn);
-		dest = data_blkaddr(dn.inode, folio, dn.ofs_in_node);
+		dest = data_blkaddr(dn.inode, cache_folio(entry), dn.ofs_in_node);
 
 		if (__is_valid_data_blkaddr(src) &&
 			!f2fs_is_valid_blkaddr(sbi, src, META_POR)) {
@@ -784,16 +787,16 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 		}
 	}
 
-	copy_node_footer(dn.node_folio, folio);
+	copy_node_footer(dn.node_folio, cache_folio(entry));
 	fill_node_footer(dn.node_folio, dn.nid, ni.ino,
-					ofs_of_node(folio), false);
+					ofs_of_node(cache_folio(entry)), false);
 	folio_mark_dirty(dn.node_folio);
 err:
 	f2fs_put_dnode(&dn);
 out:
 	f2fs_notice(sbi, "recover_data: ino = %llx, nid = %x (i_size: %s), "
 		    "range (%u, %u), recovered = %d, err = %d",
-		    inode->i_ino, nid_of_node(folio),
+		    inode->i_ino, nid_of_node(cache_folio(entry)),
 		    file_keep_isize(inode) ? "keep" : "recover",
 		    start, end, recovered, err);
 	return err;
@@ -820,26 +823,26 @@ static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list,
 	blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
 
 	while (1) {
-		struct fsync_inode_entry *entry;
-		struct folio *folio;
+		struct fsync_inode_entry *fsync_entry;
+		struct f2fs_cached_block *entry;
 
 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
 			break;
 
-		folio = f2fs_get_tmp_folio(sbi, blkaddr);
-		if (IS_ERR(folio)) {
-			err = PTR_ERR(folio);
+		entry = f2fs_get_tmp_cache(sbi, blkaddr);
+		if (IS_ERR(entry)) {
+			err = PTR_ERR(entry);
 			break;
 		}
 
-		if (!is_recoverable_dnode(folio)) {
-			f2fs_folio_put(folio, true);
+		if (!is_recoverable_dnode(sbi, cache_folio(entry))) {
+			f2fs_put_cache(entry, true);
 			break;
 		}
 		recoverable_dnode++;
 
-		entry = get_fsync_inode(inode_list, ino_of_node(folio));
-		if (!entry)
+		fsync_entry = get_fsync_inode(inode_list, ino_of_node(cache_folio(entry)));
+		if (!fsync_entry)
 			goto next;
 		fsynced_dnode++;
 		/*
@@ -847,40 +850,40 @@ static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list,
 		 * In this case, we can lose the latest inode(x).
 		 * So, call recover_inode for the inode update.
 		 */
-		if (IS_INODE(folio)) {
-			err = recover_inode(entry->inode, folio);
+		if (IS_INODE(cache_folio(entry))) {
+			err = recover_inode(fsync_entry->inode, entry);
 			if (err) {
-				f2fs_folio_put(folio, true);
+				f2fs_put_cache(entry, true);
 				break;
 			}
 			recovered_inode++;
 		}
-		if (entry->last_dentry == blkaddr) {
-			err = recover_dentry(entry->inode, folio, dir_list);
+		if (fsync_entry->last_dentry == blkaddr) {
+			err = recover_dentry(fsync_entry->inode, entry, dir_list);
 			if (err) {
-				f2fs_folio_put(folio, true);
+				f2fs_put_cache(entry, true);
 				break;
 			}
 			recovered_dentry++;
 		}
-		err = do_recover_data(sbi, entry->inode, folio);
+		err = do_recover_data(sbi, fsync_entry->inode, entry);
 		if (err) {
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 			break;
 		}
 		recovered_dnode++;
 
-		if (entry->blkaddr == blkaddr)
-			list_move_tail(&entry->list, tmp_inode_list);
+		if (fsync_entry->blkaddr == blkaddr)
+			list_move_tail(&fsync_entry->list, tmp_inode_list);
 next:
 		ra_blocks = adjust_por_ra_blocks(sbi, ra_blocks, blkaddr,
-					next_blkaddr_of_node(folio));
+					next_blkaddr_of_node(cache_folio(entry)));
 
 		/* check next segment */
-		blkaddr = next_blkaddr_of_node(folio);
-		f2fs_folio_put(folio, true);
+		blkaddr = next_blkaddr_of_node(cache_folio(entry));
+		f2fs_put_cache(entry, true);
 
-		f2fs_ra_meta_pages_cond(sbi, blkaddr, ra_blocks);
+		f2fs_ra_meta_caches_cond(sbi, blkaddr, ra_blocks);
 		total_dnode++;
 	}
 	if (!err)
@@ -937,12 +940,11 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
 	destroy_fsync_dnodes(&tmp_inode_list, err);
 
 	/* truncate meta pages to be used by the recovery */
-	truncate_inode_pages_range(META_MAPPING(sbi),
-			(loff_t)MAIN_BLKADDR(sbi) << PAGE_SHIFT, -1);
-
+	f2fs_truncate_meta_caches(sbi, MAIN_BLKADDR(sbi),
+				MAX_BLKADDR(sbi) - MAIN_BLKADDR(sbi));
 	if (err) {
 		truncate_inode_pages_final(NODE_MAPPING(sbi));
-		truncate_inode_pages_final(META_MAPPING(sbi));
+		f2fs_truncate_meta_caches(sbi, 0, ULONG_MAX);
 	}
 
 	/*
fs/f2fs/segment.c +102 -99
index 56decf9c691c..cbb3a8c9f4ab 100644
@@ -2774,63 +2774,62 @@ int f2fs_npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra)
 	return 3;
 }
 
-/*
- * Caller should put this summary folio
- */
-struct folio *f2fs_get_sum_folio(struct f2fs_sb_info *sbi, unsigned int segno)
+
+struct f2fs_cached_block *f2fs_get_sum_cache(struct f2fs_sb_info *sbi,
+		unsigned int segno)
 {
 	if (unlikely(f2fs_cp_error(sbi)))
 		return ERR_PTR(-EIO);
-	return f2fs_get_meta_folio_retry(sbi, GET_SUM_BLOCK(sbi, segno));
+	return f2fs_get_meta_cache_retry(sbi, GET_SUM_BLOCK(sbi, segno));
 }
 
 void f2fs_update_meta_page(struct f2fs_sb_info *sbi,
 					void *src, block_t blk_addr)
 {
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 
 	if (!f2fs_sb_has_packed_ssa(sbi))
-		folio = f2fs_grab_meta_folio(sbi, blk_addr);
+		entry = f2fs_grab_meta_cache(sbi, blk_addr);
 	else
-		folio = f2fs_get_meta_folio_retry(sbi, blk_addr);
+		entry = f2fs_get_meta_cache_retry(sbi, blk_addr);
 
-	if (IS_ERR(folio))
+	if (IS_ERR(entry))
 		return;
 
-	memcpy(folio_address(folio), src, PAGE_SIZE);
-	folio_mark_dirty(folio);
-	f2fs_folio_put(folio, true);
+	memcpy(cache_address(entry), src, sbi->blocksize);
+	f2fs_mark_cache_dirty(entry);
+	f2fs_put_cache(entry, true);
 }
 
 static void write_sum_page(struct f2fs_sb_info *sbi,
 		struct f2fs_summary_block *sum_blk, unsigned int segno)
 {
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 
 	if (!f2fs_sb_has_packed_ssa(sbi))
 		return f2fs_update_meta_page(sbi, (void *)sum_blk,
 				GET_SUM_BLOCK(sbi, segno));
 
-	folio = f2fs_get_sum_folio(sbi, segno);
-	if (IS_ERR(folio))
+	entry = f2fs_get_sum_cache(sbi, GET_SUM_BLOCK(sbi, segno));
+	if (IS_ERR(entry))
 		return;
 
-	memcpy(SUM_BLK_PAGE_ADDR(sbi, folio, segno), sum_blk,
+	memcpy(SUM_BLK_ENTRY_ADDR(sbi, entry, segno), sum_blk,
 			sbi->sum_blocksize);
-	folio_mark_dirty(folio);
-	f2fs_folio_put(folio, true);
+	f2fs_mark_cache_dirty(entry);
+	f2fs_put_cache(entry, true);
 }
 
 static void write_current_sum_page(struct f2fs_sb_info *sbi,
 						int type, block_t blk_addr)
 {
 	struct curseg_info *curseg = CURSEG_I(sbi, type);
-	struct folio *folio = f2fs_grab_meta_folio(sbi, blk_addr);
+	struct f2fs_cached_block *entry = f2fs_grab_meta_cache(sbi, blk_addr);
 	struct f2fs_summary_block *src = curseg->sum_blk;
 	struct f2fs_summary_block *dst;
 
-	dst = folio_address(folio);
-	memset(dst, 0, PAGE_SIZE);
+	dst = cache_address(entry);
+	memset(dst, 0, sbi->blocksize);
 
 	mutex_lock(&curseg->curseg_mutex);
 
@@ -2843,8 +2842,8 @@ static void write_current_sum_page(struct f2fs_sb_info *sbi,
 
 	mutex_unlock(&curseg->curseg_mutex);
 
-	folio_mark_dirty(folio);
-	f2fs_folio_put(folio, true);
+	f2fs_mark_cache_dirty(entry);
+	f2fs_put_cache(entry, true);
 }
 
 static int is_next_segment_free(struct f2fs_sb_info *sbi,
@@ -3160,7 +3159,7 @@ static int change_curseg(struct f2fs_sb_info *sbi, int type)
 	struct curseg_info *curseg = CURSEG_I(sbi, type);
 	unsigned int new_segno = curseg->next_segno;
 	struct f2fs_summary_block *sum_node;
-	struct folio *sum_folio;
+	struct f2fs_cached_block *entry = NULL;
 
 	if (curseg->inited)
 		write_sum_page(sbi, curseg->sum_blk, curseg->segno);
@@ -3176,15 +3175,15 @@ static int change_curseg(struct f2fs_sb_info *sbi, int type)
 	curseg->alloc_type = SSR;
 	curseg->next_blkoff = __next_free_blkoff(sbi, curseg->segno, 0);
 
-	sum_folio = f2fs_get_sum_folio(sbi, new_segno);
-	if (IS_ERR(sum_folio)) {
+	entry = f2fs_get_sum_cache(sbi, new_segno);
+	if (IS_ERR(entry)) {
 		/* GC won't be able to use stale summary pages by cp_error */
 		memset(curseg->sum_blk, 0, sbi->sum_entry_size);
-		return PTR_ERR(sum_folio);
+		return PTR_ERR(entry);
 	}
-	sum_node = SUM_BLK_PAGE_ADDR(sbi, sum_folio, new_segno);
+	sum_node = SUM_BLK_ENTRY_ADDR(sbi, entry, new_segno);
 	memcpy(curseg->sum_blk, sum_node, sbi->sum_entry_size);
-	f2fs_folio_put(sum_folio, true);
+	f2fs_put_cache(entry, true);
 	return 0;
 }
 
@@ -4127,8 +4126,9 @@ static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
 		f2fs_up_read(&fio->sbi->io_order_lock);
 }
 
-void f2fs_do_write_meta_page(struct f2fs_sb_info *sbi, struct folio *folio,
-					enum iostat_type io_type)
+void f2fs_do_write_meta_cache(struct f2fs_sb_info *sbi,
+			struct f2fs_cached_block *entry,
+			enum iostat_type io_type)
 {
 	struct f2fs_io_info fio = {
 		.sbi = sbi,
@@ -4136,20 +4136,21 @@ void f2fs_do_write_meta_page(struct f2fs_sb_info *sbi, struct folio *folio,
 		.temp = HOT,
 		.op = REQ_OP_WRITE,
 		.op_flags = REQ_SYNC | REQ_META | REQ_PRIO,
-		.old_blkaddr = folio->index,
-		.new_blkaddr = folio->index,
-		.folio = folio,
+		.old_blkaddr = entry->index,
+		.new_blkaddr = entry->index,
 		.encrypted_page = NULL,
 		.in_list = 0,
+		.cache_entry = entry,
+		.is_cache = 1,
 	};
 
-	if (unlikely(folio->index >= MAIN_BLKADDR(sbi)))
+	if (unlikely(entry->index >= MAIN_BLKADDR(sbi)))
 		fio.op_flags &= ~REQ_META;
 
-	folio_start_writeback(folio);
-	f2fs_submit_page_write(&fio);
+	f2fs_start_cache_writeback(entry);
+	f2fs_submit_cache_write(&fio);
 
-	stat_inc_meta_count(sbi, folio->index);
+	stat_inc_meta_count(sbi, entry->index);
 	f2fs_update_iostat(sbi, NULL, io_type, F2FS_BLKSIZE);
 }
 
@@ -4206,7 +4207,7 @@ int f2fs_inplace_write_data(struct f2fs_io_info *fio)
 	}
 
 	if (fio->meta_gc)
-		f2fs_truncate_meta_inode_pages(sbi, fio->new_blkaddr, 1);
+		f2fs_truncate_meta_caches(sbi, fio->new_blkaddr, 1);
 
 	stat_inc_inplace_blocks(fio->sbi);
 
@@ -4372,7 +4373,7 @@ void f2fs_folio_wait_writeback(struct folio *folio, enum page_type type,
 void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
-	struct folio *cfolio;
+	struct f2fs_cached_block *entry;
 
 	if (!f2fs_meta_inode_gc_required(inode))
 		return;
@@ -4380,11 +4381,12 @@ void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr)
 	if (!__is_valid_data_blkaddr(blkaddr))
 		return;
 
-	cfolio = filemap_lock_folio(META_MAPPING(sbi), blkaddr);
-	if (!IS_ERR(cfolio)) {
-		f2fs_folio_wait_writeback(cfolio, DATA, true, true);
-		f2fs_folio_put(cfolio, true);
-	}
+	entry = f2fs_find_cache(META_CACHE(sbi), blkaddr);
+	if (IS_ERR(entry))
+		return;
+	f2fs_lock_cache(entry);
+	f2fs_cache_wait_writeback_cond(entry, DATA);
+	f2fs_put_cache(entry, true);
 }
 
 void f2fs_wait_on_block_writeback_range(struct inode *inode, block_t blkaddr,
@@ -4399,7 +4401,7 @@ void f2fs_wait_on_block_writeback_range(struct inode *inode, block_t blkaddr,
 	for (i = 0; i < len; i++)
 		f2fs_wait_on_block_writeback(inode, blkaddr + i);
 
-	f2fs_truncate_meta_inode_pages(sbi, blkaddr, len);
+	f2fs_truncate_meta_caches(sbi, blkaddr, len);
 }
 
 static int read_compacted_summaries(struct f2fs_sb_info *sbi)
@@ -4407,16 +4409,16 @@ static int read_compacted_summaries(struct f2fs_sb_info *sbi)
 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 	struct curseg_info *seg_i;
 	unsigned char *kaddr;
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	block_t start;
 	int i, j, offset;
 
 	start = start_sum_block(sbi);
 
-	folio = f2fs_get_meta_folio(sbi, start++);
-	if (IS_ERR(folio))
-		return PTR_ERR(folio);
-	kaddr = folio_address(folio);
+	entry = f2fs_get_meta_cache(sbi, start++);
+	if (IS_ERR(entry))
+		return PTR_ERR(entry);
+	kaddr = cache_address(entry);
 
 	/* Step 1: restore nat cache */
 	seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
@@ -4453,16 +4455,16 @@ static int read_compacted_summaries(struct f2fs_sb_info *sbi)
 						SUM_FOOTER_SIZE)
 				continue;
 
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 
-			folio = f2fs_get_meta_folio(sbi, start++);
-			if (IS_ERR(folio))
-				return PTR_ERR(folio);
-			kaddr = folio_address(folio);
+			entry = f2fs_get_meta_cache(sbi, start++);
+			if (IS_ERR(entry))
+				return PTR_ERR(entry);
+			kaddr = cache_address(entry);
 			offset = 0;
 		}
 	}
-	f2fs_folio_put(folio, true);
+	f2fs_put_cache(entry, true);
 	return 0;
 }
 
@@ -4471,7 +4473,7 @@ static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 	struct f2fs_summary_block *sum;
 	struct curseg_info *curseg;
-	struct folio *new;
+	struct f2fs_cached_block *entry;
 	unsigned short blk_off;
 	unsigned int segno = 0;
 	block_t blk_addr = 0;
@@ -4498,10 +4500,10 @@ static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
 			blk_addr = GET_SUM_BLOCK(sbi, segno);
 	}
 
-	new = f2fs_get_meta_folio(sbi, blk_addr);
-	if (IS_ERR(new))
-		return PTR_ERR(new);
-	sum = folio_address(new);
+	entry = f2fs_get_meta_cache(sbi, blk_addr);
+	if (IS_ERR(entry))
+		return PTR_ERR(entry);
+	sum = cache_address(entry);
 
 	if (IS_NODESEG(type)) {
 		if (__exist_node_summaries(sbi)) {
@@ -4538,7 +4540,7 @@ static int read_normal_summaries(struct f2fs_sb_info *sbi, int type)
 	curseg->next_blkoff = blk_off;
 	mutex_unlock(&curseg->curseg_mutex);
 out:
-	f2fs_folio_put(new, true);
+	f2fs_put_cache(entry, true);
 	return err;
 }
 
@@ -4553,8 +4555,8 @@ static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
 		int npages = f2fs_npages_for_summary_flush(sbi, true);
 
 		if (npages >= 2)
-			f2fs_ra_meta_pages(sbi, start_sum_block(sbi), npages,
-							META_CP, true);
+			f2fs_ra_meta_caches(sbi, start_sum_block(sbi),
+					npages, META_CP, true);
 
 		/* restore for compacted data summary */
 		err = read_compacted_summaries(sbi);
@@ -4564,7 +4566,7 @@ static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
 	}
 
 	if (__exist_node_summaries(sbi))
-		f2fs_ra_meta_pages(sbi,
+		f2fs_ra_meta_caches(sbi,
 				sum_blk_addr(sbi, NR_CURSEG_PERSIST_TYPE, type),
 				NR_CURSEG_PERSIST_TYPE - type, META_CP, true);
 
@@ -4587,16 +4589,16 @@ static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
 
 static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
 {
-	struct folio *folio;
+	struct f2fs_cached_block *entry = NULL;
 	unsigned char *kaddr;
 	struct f2fs_summary *summary;
 	struct curseg_info *seg_i;
 	int written_size = 0;
 	int i, j;
 
-	folio = f2fs_grab_meta_folio(sbi, blkaddr++);
-	kaddr = folio_address(folio);
-	memset(kaddr, 0, PAGE_SIZE);
+	entry = f2fs_grab_meta_cache(sbi, blkaddr++);
+	kaddr = cache_address(entry);
+	memset(kaddr, 0, sbi->blocksize);
 
 	/* Step 1: write nat cache */
 	seg_i = CURSEG_I(sbi, CURSEG_HOT_DATA);
@@ -4612,10 +4614,10 @@ static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
 	for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
 		seg_i = CURSEG_I(sbi, i);
 		for (j = 0; j < f2fs_curseg_valid_blocks(sbi, i); j++) {
-			if (!folio) {
-				folio = f2fs_grab_meta_folio(sbi, blkaddr++);
-				kaddr = folio_address(folio);
-				memset(kaddr, 0, PAGE_SIZE);
+			if (!entry) {
+				entry = f2fs_grab_meta_cache(sbi, blkaddr++);
+				kaddr = cache_address(entry);
+				memset(kaddr, 0, sbi->blocksize);
 				written_size = 0;
 			}
 			summary = (struct f2fs_summary *)(kaddr + written_size);
@@ -4626,14 +4628,14 @@ static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
 							SUM_FOOTER_SIZE)
 				continue;
 
-			folio_mark_dirty(folio);
-			f2fs_folio_put(folio, true);
-			folio = NULL;
+			f2fs_mark_cache_dirty(entry);
+			f2fs_put_cache(entry, true);
+			entry = NULL;
 		}
 	}
-	if (folio) {
-		folio_mark_dirty(folio);
-		f2fs_folio_put(folio, true);
+	if (entry) {
+		f2fs_mark_cache_dirty(entry);
+		f2fs_put_cache(entry, true);
 	}
 }
 
@@ -4687,29 +4689,29 @@ int f2fs_lookup_journal_in_cursum(struct f2fs_sb_info *sbi,
 	return -1;
 }
 
-static struct folio *get_current_sit_folio(struct f2fs_sb_info *sbi,
+static struct f2fs_cached_block *get_current_sit_cache(struct f2fs_sb_info *sbi,
 					unsigned int segno)
 {
-	return f2fs_get_meta_folio(sbi, current_sit_addr(sbi, segno));
+	return f2fs_get_meta_cache(sbi, current_sit_addr(sbi, segno));
 }
 
-static struct folio *get_next_sit_folio(struct f2fs_sb_info *sbi,
+static struct f2fs_cached_block *get_next_sit_cache(struct f2fs_sb_info *sbi,
 					unsigned int start)
 {
 	struct sit_info *sit_i = SIT_I(sbi);
-	struct folio *folio;
+	struct f2fs_cached_block *entry;
 	pgoff_t src_off, dst_off;
 
 	src_off = current_sit_addr(sbi, start);
 	dst_off = next_sit_addr(sbi, src_off);
 
-	folio = f2fs_grab_meta_folio(sbi, dst_off);
-	seg_info_to_sit_folio(sbi, folio, start);
+	entry = f2fs_grab_meta_cache(sbi, dst_off);
+	seg_info_to_sit_block(sbi, entry, start);
 
-	folio_mark_dirty(folio);
+	f2fs_mark_cache_dirty(entry);
 	set_to_next_sit(sit_i, start);
 
-	return folio;
+	return entry;
 }
 
 static struct sit_entry_set *grab_sit_entry_set(void)
@@ -4839,8 +4841,9 @@ void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	 * #2, flush sit entries to sit page.
 	 */
 	list_for_each_entry_safe(ses, tmp, head, set_list) {
-		struct folio *folio = NULL;
+		struct f2fs_cached_block *entry = NULL;
 		struct f2fs_sit_block *raw_sit = NULL;
+
 		unsigned int start_segno = ses->start_segno;
 		unsigned int end = min(start_segno + SIT_ENTRY_PER_BLOCK,
 						(unsigned long)MAIN_SEGS(sbi));
@@ -4854,8 +4857,8 @@ void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 		if (to_journal) {
 			down_write(&curseg->journal_rwsem);
 		} else {
-			folio = get_next_sit_folio(sbi, start_segno);
-			raw_sit = folio_address(folio);
+			entry = get_next_sit_cache(sbi, start_segno);
+			raw_sit = cache_address(entry);
 		}
 
 		/* flush dirty sit entries in region of current sit set */
@@ -4900,7 +4903,7 @@ void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 		if (to_journal)
 			up_write(&curseg->journal_rwsem);
 		else
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 
 		f2fs_bug_on(sbi, ses->entry_cnt);
 		release_sit_entry_set(ses);
@@ -5091,7 +5094,7 @@ static int build_sit_entries(struct f2fs_sb_info *sbi)
 	block_t sit_valid_blocks[2] = {0, 0};
 
 	do {
-		readed = f2fs_ra_meta_pages(sbi, start_blk, BIO_MAX_VECS,
+		readed = f2fs_ra_meta_caches(sbi, start_blk, BIO_MAX_VECS,
 							META_SIT, true);
 
 		start = start_blk * sit_i->sents_per_block;
@@ -5099,15 +5102,15 @@ static int build_sit_entries(struct f2fs_sb_info *sbi)
 
 		for (; start < end && start < MAIN_SEGS(sbi); start++) {
 			struct f2fs_sit_block *sit_blk;
-			struct folio *folio;
+			struct f2fs_cached_block *entry;
 
 			se = &sit_i->sentries[start];
-			folio = get_current_sit_folio(sbi, start);
-			if (IS_ERR(folio))
-				return PTR_ERR(folio);
-			sit_blk = folio_address(folio);
+			entry = get_current_sit_cache(sbi, start);
+			if (IS_ERR(entry))
+				return PTR_ERR(entry);
+			sit_blk = cache_address(entry);
 			sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
-			f2fs_folio_put(folio, true);
+			f2fs_put_cache(entry, true);
 
 			err = check_block_count(sbi, start, &sit);
 			if (err)
fs/f2fs/segment.h +27 -4
index db1079169a23..fff35c63a00f 100644
@@ -95,6 +95,8 @@ static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
 #define GET_SUM_BLKOFF(sbi, segno) (segno % (sbi)->sums_per_block)
 #define SUM_BLK_PAGE_ADDR(sbi, folio, segno)	\
 	(folio_address(folio) + GET_SUM_BLKOFF(sbi, segno) * (sbi)->sum_blocksize)
+#define SUM_BLK_ENTRY_ADDR(sbi, entry, segno)	\
+	(cache_address(entry) + GET_SUM_BLKOFF(sbi, segno) * (sbi)->sum_blocksize)
 
 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
@@ -417,8 +419,8 @@ static inline void __seg_info_to_raw_sit(struct seg_entry *se,
 	rs->mtime = cpu_to_le64(se->mtime);
 }
 
-static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi,
-				struct folio *folio, unsigned int start)
+static inline void seg_info_to_sit_block(struct f2fs_sb_info *sbi,
+			struct f2fs_cached_block *entry, unsigned int start)
 {
 	struct f2fs_sit_block *raw_sit;
 	struct seg_entry *se;
@@ -427,8 +429,8 @@ static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi,
 					(unsigned long)MAIN_SEGS(sbi));
 	int i;
 
-	raw_sit = folio_address(folio);
-	memset(raw_sit, 0, PAGE_SIZE);
+	raw_sit = cache_address(entry);
+	memset(raw_sit, 0, sbi->blocksize);
 	for (i = 0; i < end - start; i++) {
 		rs = &raw_sit->entries[i];
 		se = get_seg_entry(sbi, start + i);
@@ -991,6 +993,27 @@ static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
 		return 0;
 }
 
+/*
+ * When writing cache asynchronously, align nr_to_write to BIO_MAX_VECS.
+ */
+static inline long adjust_flush_cache_number(struct f2fs_sb_info *sbi, int type)
+{
+	long nr_to_write;
+
+	switch (type) {
+	case META:
+		nr_to_write = BIO_MAX_VECS;
+		break;
+	case NODE:
+		nr_to_write = BIO_MAX_VECS << 1;
+		break;
+	default:
+		f2fs_bug_on(sbi, 1);
+		return 0;
+	}
+	return nr_to_write;
+}
+
 /*
  * When writing pages, it'd better align nr_to_write for segment size.
  */
fs/f2fs/super.c +7 -24
index 89affe72f4fc..d3dee9b7d999 100644
@@ -1839,8 +1839,7 @@ static int f2fs_drop_inode(struct inode *inode)
 	 * drop useless meta/node dirty pages.
 	 */
 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
-		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi)) {
+		if (inode->i_ino == F2FS_NODE_INO(sbi)) {
 			trace_f2fs_drop_inode(inode, 1);
 			return 1;
 		}
@@ -1942,8 +1941,7 @@ static void f2fs_dirty_inode(struct inode *inode, int flags)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
-			inode->i_ino == F2FS_META_INO(sbi))
+	if (inode->i_ino == F2FS_NODE_INO(sbi))
 		return;
 
 	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
@@ -2039,9 +2037,10 @@ static void f2fs_put_super(struct super_block *sb)
 
 	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
 
-	if (err || f2fs_cp_error(sbi)) {
+	if (err || f2fs_cp_error(sbi) ||
+		unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 		truncate_inode_pages_final(NODE_MAPPING(sbi));
-		truncate_inode_pages_final(META_MAPPING(sbi));
+		f2fs_truncate_meta_caches(sbi, 0, ULONG_MAX);
 	}
 
 	f2fs_bug_on(sbi, sbi->fsync_node_num);
@@ -2051,9 +2050,6 @@ static void f2fs_put_super(struct super_block *sb)
 	iput(sbi->node_inode);
 	sbi->node_inode = NULL;
 
-	iput(sbi->meta_inode);
-	sbi->meta_inode = NULL;
-
 	f2fs_destroy_cache(META_CACHE(sbi));
 
 	/* Should check the page counts after dropping all node/meta pages */
@@ -4383,7 +4379,6 @@ static void init_sb_info(struct f2fs_sb_info *sbi)
 	sbi->allocate_section_policy = ALLOCATE_FORWARD_NOHINT;
 	F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
 	F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
-	F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
 	sbi->cur_victim_sec = NULL_SECNO;
 	sbi->gc_mode = GC_NORMAL;
 	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
@@ -5224,18 +5219,10 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (err)
 		goto free_page_array_cache;
 
-	/* get an inode for meta space */
-	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
-	if (IS_ERR(sbi->meta_inode)) {
-		f2fs_err(sbi, "Failed to read F2FS meta data inode");
-		err = PTR_ERR(sbi->meta_inode);
-		goto free_meta_cache;
-	}
-
 	err = f2fs_get_valid_checkpoint(sbi);
 	if (err) {
 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
-		goto free_meta_inode;
+		goto free_meta_cache;
 	}
 
 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
@@ -5529,7 +5516,7 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
 	 * falls into an infinite loop in f2fs_sync_meta_pages().
 	 */
-	truncate_inode_pages_final(META_MAPPING(sbi));
+	f2fs_truncate_meta_caches(sbi, 0, ULONG_MAX);
 	/* evict some inodes being cached by GC */
 	evict_inodes(sb);
 	f2fs_unregister_sysfs(sbi);
@@ -5559,10 +5546,6 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 free_devices:
 	destroy_device_list(sbi);
 	kvfree(sbi->ckpt);
-free_meta_inode:
-	make_bad_inode(sbi->meta_inode);
-	iput(sbi->meta_inode);
-	sbi->meta_inode = NULL;
 free_meta_cache:
 	f2fs_destroy_cache(META_CACHE(sbi));
 free_page_array_cache:
include/linux/f2fs_fs.h +0 -1
index bb2b6cd5d507..0c027d00a1ea 100644
@@ -35,7 +35,6 @@
 
 #define F2FS_ROOT_INO(sbi)	((sbi)->root_ino_num)
 #define F2FS_NODE_INO(sbi)	((sbi)->node_ino_num)
-#define F2FS_META_INO(sbi)	((sbi)->meta_ino_num)
 #define F2FS_COMPRESS_INO(sbi)	(NM_I(sbi)->max_nid)
 
 #define F2FS_MAX_QUOTAS		3
Signature · 8 lines
2.49.0



_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
#14 lore raw

Re: [PATCH v1 05/12] f2fs: cache: use meta cache

In reply to Chao Yu

Hi Chao,

Quoted 9 lines

struct f2fs_cached_block *f2fs_get_sum_cache(struct f2fs_sb_info *sbi,
unsigned int segno)

{
	if (unlikely(f2fs_cp_error(sbi)))
		return ERR_PTR(-EIO);
	return f2fs_get_meta_cache_retry(sbi, GET_SUM_BLOCK(sbi, segno));
}

entry = f2fs_get_sum_cache(sbi, GET_SUM_BLOCK(sbi, segno));

f2fs_get_sum_cache() applies GET_SUM_BLOCK() to its argument. Its other callers pass a logical segment number, while this packed-SSA path passes an already translated value.

Is the argument contract of f2fs_get_sum_cache() a logical segment number or an already translated block address?

author maintainer
#15 lore raw

Re: [PATCH v1 05/12] f2fs: cache: use meta cache

Quoted 10 lines · Wenjie Qi

Hi Chao,

Quoted 8 lines

struct f2fs_cached_block *f2fs_get_sum_cache(struct f2fs_sb_info *sbi,
unsigned int segno)

{
	if (unlikely(f2fs_cp_error(sbi)))
		return ERR_PTR(-EIO);
	return f2fs_get_meta_cache_retry(sbi, GET_SUM_BLOCK(sbi, segno));
}
Quoted 8 lines
Quoted 1 line

entry = f2fs_get_sum_cache(sbi, GET_SUM_BLOCK(sbi, segno));

f2fs_get_sum_cache() applies GET_SUM_BLOCK() to its argument. Its other callers pass a logical segment number, while this packed-SSA path passes an already translated value.

Is the argument contract of f2fs_get_sum_cache() a logical segment number or an already translated block address?

Oops, should use f2fs_get_sum_cache(sbi, segno) here.

Thanks,

author maintainer
#16 lore raw

PATCH 6/12 f2fs: cache: initialize node cache

In reply to Chao Yu

This patch introduces node_blocks in f2fs_sb_info structure, initializes and destroys the node cache during filesystem mount and unmount.

It also introduces helper wrappers for node cache operation, and registers node cache into the memory shrinker and writeback kthread.

---
 fs/f2fs/cache.c |  9 ++++++++-
 fs/f2fs/cache.h | 11 +++++++++++
 fs/f2fs/f2fs.h  |  1 +
 fs/f2fs/super.c |  9 ++++++++-
 4 files changed, 28 insertions(+), 2 deletions(-)
fs/f2fs/cache.c +8 -1
index afef808e485a..6c9b7a6d2313 100644
@@ -609,7 +609,14 @@ static unsigned long f2fs_do_shrink_cache(struct f2fs_cached_block_list *cache,
 unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 					unsigned long nr_to_scan)
 {
-	return f2fs_do_shrink_cache(META_CACHE(sbi), nr_to_scan);
+	unsigned long freed;
+
+	freed = f2fs_do_shrink_cache(META_CACHE(sbi), nr_to_scan);
+	if (freed >= nr_to_scan)
+		return freed;
+
+	freed += f2fs_do_shrink_cache(NODE_CACHE(sbi), nr_to_scan - freed);
+	return freed;
 }
 
 static int f2fs_cache_writeback_kthread(void *data)
fs/f2fs/cache.h +11 -0
index 5e9fe8c0b15c..d843b9caba47 100644
@@ -180,6 +180,7 @@ int f2fs_start_cache_wb_thread(struct f2fs_sb_info *sbi);
 void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 
 #define META_CACHE(sbi)		(&(sbi)->meta_blocks)
+#define NODE_CACHE(sbi)		(&(sbi)->node_blocks)
 
 #define f2fs_find_meta_cache(sbi, blkaddr)		\
 	f2fs_find_cache(META_CACHE(sbi), blkaddr)
@@ -188,6 +189,16 @@ void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 #define f2fs_truncate_meta_caches(sbi, start, len)	\
 	f2fs_drop_cache_range(META_CACHE(sbi), start, len, true)
 
+#define f2fs_grab_node_cache(sbi, blkaddr)		\
+	f2fs_grab_cache(NODE_CACHE(sbi), blkaddr,	\
+	F2FS_CACHE_LOCK_CREATE)
+#define f2fs_find_node_cache(sbi, blkaddr)		\
+	f2fs_find_cache(NODE_CACHE(sbi), blkaddr)
+#define f2fs_invalidate_node_cache(sbi, blkaddr)	\
+	f2fs_drop_cache_range(NODE_CACHE(sbi), blkaddr, 1, false)
+#define f2fs_truncate_node_caches(sbi, start, len)	\
+	f2fs_drop_cache_range(NODE_CACHE(sbi), start, len, true)
+
 unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 				unsigned long nr_to_scan);
 
fs/f2fs/f2fs.h +1 -0
index 9a353dcd6658..4a811e9d325b 100644
@@ -2106,6 +2106,7 @@ struct f2fs_sb_info {
 
 	/* f2fs internal cache */
 	struct f2fs_cached_block_list meta_blocks;
+	struct f2fs_cached_block_list node_blocks;
 
 	/* internal cache flush thread */
 	struct f2fs_cache_kthread cache_thread;
fs/f2fs/super.c +8 -1
index d3dee9b7d999..83496c46c89f 100644
@@ -2051,6 +2051,7 @@ static void f2fs_put_super(struct super_block *sb)
 	sbi->node_inode = NULL;
 
 	f2fs_destroy_cache(META_CACHE(sbi));
+	f2fs_destroy_cache(NODE_CACHE(sbi));
 
 	/* Should check the page counts after dropping all node/meta pages */
 	for (i = 0; i < NR_COUNT_TYPE; i++) {
@@ -5219,10 +5220,14 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (err)
 		goto free_page_array_cache;
 
+	err = f2fs_init_cache(sbi, NODE_CACHE(sbi), F2FS_NODE_CACHE);
+	if (err)
+		goto free_meta_cache;
+
 	err = f2fs_get_valid_checkpoint(sbi);
 	if (err) {
 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
-		goto free_meta_cache;
+		goto free_node_cache;
 	}
 
 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
@@ -5546,6 +5551,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 free_devices:
 	destroy_device_list(sbi);
 	kvfree(sbi->ckpt);
+free_node_cache:
+	f2fs_destroy_cache(NODE_CACHE(sbi));
 free_meta_cache:
 	f2fs_destroy_cache(META_CACHE(sbi));
 free_page_array_cache:
Signature · 1 line
2.49.0
author maintainer
#17 lore raw

PATCH 7/12 f2fs: cache: use node cache

In reply to Chao Yu

This patch migrates F2FS node block caching from the fake VFS inode page cache (sbi->node_inode) to node cache (sbi->node_blocks).

It updates node related helpers to use node cache APIs and reference to struct f2fs_cached_block, and removes sbi->node_inode, especially, unifies inline and regular dentry block handling via struct f2fs_dentry_block_ref.

---
 fs/f2fs/acl.c           |  26 +-
 fs/f2fs/acl.h           |   8 +-
 fs/f2fs/cache.c         |   7 +-
 fs/f2fs/cache.h         |   1 +
 fs/f2fs/checkpoint.c    |   7 +-
 fs/f2fs/compress.c      |  18 +-
 fs/f2fs/data.c          | 236 +++++------
 fs/f2fs/debug.c         |  12 +-
 fs/f2fs/dir.c           | 168 ++++----
 fs/f2fs/extent_cache.c  |  14 +-
 fs/f2fs/f2fs.h          | 223 +++++-----
 fs/f2fs/file.c          |  74 ++--
 fs/f2fs/gc.c            |  44 +-
 fs/f2fs/inline.c        | 279 ++++++-------
 fs/f2fs/inode.c         | 155 ++++---
 fs/f2fs/namei.c         | 118 +++---
 fs/f2fs/node.c          | 875 +++++++++++++++++++---------------------
 fs/f2fs/node.h          | 107 +++--
 fs/f2fs/recovery.c      | 115 +++---
 fs/f2fs/segment.c       |  59 +--
 fs/f2fs/segment.h       |  20 -
 fs/f2fs/shrinker.c      |   3 +-
 fs/f2fs/super.c         |  56 +--
 fs/f2fs/xattr.c         | 123 +++---
 fs/f2fs/xattr.h         |  12 +-
 include/linux/f2fs_fs.h |   1 -
 26 files changed, 1329 insertions(+), 1432 deletions(-)
fs/f2fs/acl.c +13 -13
index d3253549173e..34c9aa279040 100644
@@ -181,7 +181,7 @@ static void *f2fs_acl_to_disk(struct f2fs_sb_info *sbi,
 }
 
 static struct posix_acl *__f2fs_get_acl(struct inode *inode, int type,
-						struct folio *dfolio)
+						struct f2fs_cached_block *entry)
 {
 	int name_index = F2FS_XATTR_INDEX_POSIX_ACL_DEFAULT;
 	void *value = NULL;
@@ -191,13 +191,13 @@ static struct posix_acl *__f2fs_get_acl(struct inode *inode, int type,
 	if (type == ACL_TYPE_ACCESS)
 		name_index = F2FS_XATTR_INDEX_POSIX_ACL_ACCESS;
 
-	retval = f2fs_getxattr(inode, name_index, "", NULL, 0, dfolio);
+	retval = f2fs_getxattr(inode, name_index, "", NULL, 0, entry);
 	if (retval > 0) {
 		value = f2fs_kmalloc(F2FS_I_SB(inode), retval, GFP_F2FS_ZERO);
 		if (!value)
 			return ERR_PTR(-ENOMEM);
 		retval = f2fs_getxattr(inode, name_index, "", value,
-							retval, dfolio);
+							retval, entry);
 	}
 
 	if (retval > 0)
@@ -242,7 +242,7 @@ static int f2fs_acl_update_mode(struct mnt_idmap *idmap,
 
 static int __f2fs_set_acl(struct mnt_idmap *idmap,
 			struct inode *inode, int type,
-			struct posix_acl *acl, struct folio *ifolio)
+			struct posix_acl *acl, struct f2fs_cached_block *ientry)
 {
 	int name_index;
 	void *value = NULL;
@@ -253,7 +253,7 @@ static int __f2fs_set_acl(struct mnt_idmap *idmap,
 	switch (type) {
 	case ACL_TYPE_ACCESS:
 		name_index = F2FS_XATTR_INDEX_POSIX_ACL_ACCESS;
-		if (acl && !ifolio) {
+		if (acl && !ientry) {
 			error = f2fs_acl_update_mode(idmap, inode, &mode, &acl);
 			if (error)
 				return error;
@@ -279,7 +279,7 @@ static int __f2fs_set_acl(struct mnt_idmap *idmap,
 		}
 	}
 
-	error = f2fs_setxattr(inode, name_index, "", value, size, ifolio, 0);
+	error = f2fs_setxattr(inode, name_index, "", value, size, ientry, 0);
 
 	kfree(value);
 	if (!error)
@@ -374,7 +374,7 @@ static int f2fs_acl_create_masq(struct posix_acl *acl, umode_t *mode_p)
 
 static int f2fs_acl_create(struct inode *dir, umode_t *mode,
 		struct posix_acl **default_acl, struct posix_acl **acl,
-		struct folio *dfolio)
+		struct f2fs_cached_block *entry)
 {
 	struct posix_acl *p;
 	struct posix_acl *clone;
@@ -386,7 +386,7 @@ static int f2fs_acl_create(struct inode *dir, umode_t *mode,
 	if (S_ISLNK(*mode) || !IS_POSIXACL(dir))
 		return 0;
 
-	p = __f2fs_get_acl(dir, ACL_TYPE_DEFAULT, dfolio);
+	p = __f2fs_get_acl(dir, ACL_TYPE_DEFAULT, entry);
 	if (!p || p == ERR_PTR(-EOPNOTSUPP)) {
 		*mode &= ~current_umask();
 		return 0;
@@ -423,13 +423,13 @@ static int f2fs_acl_create(struct inode *dir, umode_t *mode,
 	return ret;
 }
 
-int f2fs_init_acl(struct inode *inode, struct inode *dir, struct folio *ifolio,
-							struct folio *dfolio)
+int f2fs_init_acl(struct inode *inode, struct inode *dir, struct f2fs_cached_block *ientry,
+							struct f2fs_cached_block *dentry)
 {
 	struct posix_acl *default_acl = NULL, *acl = NULL;
 	int error;
 
-	error = f2fs_acl_create(dir, &inode->i_mode, &default_acl, &acl, dfolio);
+	error = f2fs_acl_create(dir, &inode->i_mode, &default_acl, &acl, dentry);
 	if (error)
 		return error;
 
@@ -437,7 +437,7 @@ int f2fs_init_acl(struct inode *inode, struct inode *dir, struct folio *ifolio,
 
 	if (default_acl) {
 		error = __f2fs_set_acl(NULL, inode, ACL_TYPE_DEFAULT,
-				default_acl, ifolio);
+				default_acl, ientry);
 		posix_acl_release(default_acl);
 	} else {
 		inode->i_default_acl = NULL;
@@ -445,7 +445,7 @@ int f2fs_init_acl(struct inode *inode, struct inode *dir, struct folio *ifolio,
 	if (acl) {
 		if (!error)
 			error = __f2fs_set_acl(NULL, inode, ACL_TYPE_ACCESS,
-					acl, ifolio);
+					acl, ientry);
 		posix_acl_release(acl);
 	} else {
 		inode->i_acl = NULL;
fs/f2fs/acl.h +5 -3
index 20e87e63c089..0f639367a0ab 100644
@@ -36,14 +36,16 @@ struct f2fs_acl_header {
 struct posix_acl *f2fs_get_acl(struct inode *, int, bool);
 int f2fs_set_acl(struct mnt_idmap *, struct dentry *,
 			struct posix_acl *, int);
-int f2fs_init_acl(struct inode *, struct inode *, struct folio *ifolio,
-		struct folio *dfolio);
+int f2fs_init_acl(struct inode *inode, struct inode *dir,
+			struct f2fs_cached_block *ientry,
+			struct f2fs_cached_block *dentry);
 #else
 #define f2fs_get_acl	NULL
 #define f2fs_set_acl	NULL
 
 static inline int f2fs_init_acl(struct inode *inode, struct inode *dir,
-				struct folio *ifolio, struct folio *dfolio)
+				struct f2fs_cached_block *ientry,
+				struct f2fs_cached_block *dentry)
 {
 	return 0;
 }
fs/f2fs/cache.c +4 -3
index 6c9b7a6d2313..3176d62ce25b 100644
@@ -29,7 +29,7 @@ void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
 		return;
 
 	/* submit cached bio */
-	f2fs_submit_merged_write_cache(entry, type);
+	f2fs_submit_merged_write_cache(entry->cache->sbi, entry, 0, type);
 
 	wait_on_bit_io(&entry->state, F2FS_BLOCK_WRITEBACK,
 					TASK_UNINTERRUPTIBLE);
@@ -67,8 +67,8 @@ bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry)
 		f2fs_cache_set_uptodate(entry);
 
 #ifdef CONFIG_F2FS_CHECK_FS
-	if (cache->type == F2FS_NODE_CACHE && IS_INODE(cache_folio(entry)))
-		f2fs_inode_chksum_set(cache->sbi, cache_folio(entry));
+	if (f2fs_is_node_cache(entry) && IS_INODE(entry))
+		f2fs_inode_chksum_set(cache->sbi, entry);
 #endif
 
 	if (f2fs_cache_test_dirty(entry))
@@ -640,6 +640,7 @@ static int f2fs_cache_writeback_kthread(void *data)
 			continue;
 
 		f2fs_write_meta_caches(sbi);
+		f2fs_write_node_caches(sbi);
 	}
 	return 0;
 }
fs/f2fs/cache.h +1 -0
index d843b9caba47..397019cfb861 100644
@@ -52,6 +52,7 @@ struct f2fs_cached_block_list {
 };
 
 #define IS_META_CACHE(cache) (cache->type == F2FS_META_CACHE)
+#define IS_NODE_CACHE(cache) (cache->type == F2FS_NODE_CACHE)
 
 /* Flags for f2fs_cached_block state */
 enum f2fs_cached_state {
fs/f2fs/checkpoint.c +2 -5
index 1a7083540b82..508132652693 100644
@@ -1489,10 +1489,6 @@ static bool __need_flush_quota(struct f2fs_sb_info *sbi)
  */
 static int block_operations(struct f2fs_sb_info *sbi)
 {
-	struct writeback_control wbc = {
-		.sync_mode = WB_SYNC_ALL,
-		.nr_to_write = LONG_MAX,
-	};
 	int err = 0, cnt = 0;
 
 	/*
@@ -1556,7 +1552,8 @@ static int block_operations(struct f2fs_sb_info *sbi)
 	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
 		f2fs_up_write(&sbi->node_write);
 		atomic_inc(&sbi->wb_sync_req[NODE]);
-		err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
+		err = f2fs_writeback_node_caches(sbi, LONG_MAX,
+			true, false, FS_CP_NODE_IO);
 		atomic_dec(&sbi->wb_sync_req[NODE]);
 		if (err) {
 			f2fs_up_write(&sbi->node_change);
fs/f2fs/compress.c +9 -9
index bb749f6257a1..676a5559357f 100644
@@ -909,7 +909,7 @@ bool f2fs_sanity_check_cluster(struct dnode_of_data *dn)
 	}
 
 	for (i = 1, count = 1; i < cluster_size; i++, count++) {
-		block_t blkaddr = data_blkaddr(dn->inode, dn->node_folio,
+		block_t blkaddr = data_blkaddr(dn->inode, dn->node_entry,
 							dn->ofs_in_node + i);
 
 		/* [COMPR_ADDR, ..., COMPR_ADDR] */
@@ -950,7 +950,7 @@ static int __f2fs_get_cluster_blocks(struct inode *inode,
 	int count, i;
 
 	for (i = 0, count = 0; i < cluster_size; i++) {
-		block_t blkaddr = data_blkaddr(dn->inode, dn->node_folio,
+		block_t blkaddr = data_blkaddr(dn->inode, dn->node_entry,
 							dn->ofs_in_node + i);
 
 		if (__is_valid_data_blkaddr(blkaddr))
@@ -1146,7 +1146,7 @@ static int prepare_compress_overwrite(struct compress_ctx *cc,
 			goto release_and_retry;
 		}
 
-		f2fs_folio_wait_writeback(folio, DATA, true, true);
+		f2fs_folio_wait_writeback(folio, true, true);
 		f2fs_compress_ctx_add_page(cc, folio);
 
 		if (!folio_test_uptodate(folio)) {
@@ -1324,7 +1324,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 		goto out_unlock_op;
 
 	for (i = 0; i < cc->cluster_size; i++) {
-		if (data_blkaddr(dn.inode, dn.node_folio,
+		if (data_blkaddr(dn.inode, dn.node_entry,
 					dn.ofs_in_node + i) == NULL_ADDR)
 			goto out_put_dnode;
 	}
@@ -1356,7 +1356,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
 				page_folio(cc->rpages[i + 1])->index, cic);
 		fio.compressed_page = cc->cpages[i];
 
-		fio.old_blkaddr = data_blkaddr(dn.inode, dn.node_folio,
+		fio.old_blkaddr = data_blkaddr(dn.inode, dn.node_entry,
 						dn.ofs_in_node + i + 1);
 
 		/* wait for GCed page writeback via generic cache */
@@ -1567,7 +1567,7 @@ static int f2fs_write_raw_pages(struct compress_ctx *cc,
 		if (folio_test_writeback(folio)) {
 			if (wbc->sync_mode == WB_SYNC_NONE)
 				goto continue_unlock;
-			f2fs_folio_wait_writeback(folio, DATA, true, true);
+			f2fs_folio_wait_writeback(folio, true, true);
 		}
 
 		if (!folio_clear_dirty_for_io(folio))
@@ -1899,14 +1899,14 @@ void f2fs_put_folio_dic(struct folio *folio, bool in_task)
 unsigned int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn,
 						unsigned int ofs_in_node)
 {
-	bool compressed = data_blkaddr(dn->inode, dn->node_folio,
+	bool compressed = data_blkaddr(dn->inode, dn->node_entry,
 					ofs_in_node) == COMPRESS_ADDR;
 	int i = compressed ? 1 : 0;
-	block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_folio,
+	block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_entry,
 							ofs_in_node + i);
 
 	for (i += 1; i < F2FS_I(dn->inode)->i_cluster_size; i++) {
-		block_t blkaddr = data_blkaddr(dn->inode, dn->node_folio,
+		block_t blkaddr = data_blkaddr(dn->inode, dn->node_entry,
 							ofs_in_node + i);
 
 		if (!__is_valid_data_blkaddr(blkaddr))
fs/f2fs/data.c +108 -128

557 lines not on this page · raw message

fs/f2fs/debug.c +4 -8
index 14059a50739c..bedaade92677 100644
@@ -222,8 +222,7 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->free_secs = free_sections(sbi);
 	si->prefree_count = prefree_segments(sbi);
 	si->dirty_count = dirty_segments(sbi);
-	if (sbi->node_inode)
-		si->node_pages = NODE_MAPPING(sbi)->nrpages;
+	si->node_caches = NODE_CACHE(sbi)->num_entries;
 	si->meta_caches = META_CACHE(sbi)->num_entries;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
@@ -382,13 +381,10 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 	}
 
 	si->page_mem = 0;
-	if (sbi->node_inode) {
-		unsigned long npages = NODE_MAPPING(sbi)->nrpages;
-
-		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
-	}
 	si->page_mem += (unsigned long long)META_CACHE(sbi)->num_entries << PAGE_SHIFT;
 	si->cache_mem += META_CACHE(sbi)->num_entries * sizeof(struct f2fs_cached_block);
+	si->page_mem += (unsigned long long)NODE_CACHE(sbi)->num_entries << PAGE_SHIFT;
+	si->cache_mem += NODE_CACHE(sbi)->num_entries * sizeof(struct f2fs_cached_block);
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	if (sbi->compress_inode) {
 		unsigned long npages = COMPRESS_MAPPING(sbi)->nrpages;
@@ -696,7 +692,7 @@ static int stat_show(struct seq_file *s, void *v)
 			   si->aw_cnt, si->max_aw_cnt);
 		seq_printf(s, "  - compress: %4d, hit:%8d\n", si->compress_pages, si->compress_page_hit);
 		seq_printf(s, "  - nodes: %4d in %4d\n",
-			   si->ndirty_node, si->node_pages);
+			   si->ndirty_node, si->node_caches);
 		seq_printf(s, "  - dents: %4d in dirs:%4d (%4d)\n",
 			   si->ndirty_dent, si->ndirty_dirs, si->ndirty_all);
 		seq_printf(s, "  - data: %4d in files:%4d\n",
fs/f2fs/dir.c +92 -76

407 lines not on this page · raw message

fs/f2fs/extent_cache.c +7 -7
index 37cf9fa8d537..e52527c68c02 100644
@@ -20,10 +20,10 @@
 #include "segment.h"
 #include <trace/events/f2fs.h>
 
-bool sanity_check_extent_cache(struct inode *inode, struct folio *ifolio)
+bool sanity_check_extent_cache(struct inode *inode, struct f2fs_cached_block *ientry)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
-	struct f2fs_extent *i_ext = &F2FS_INODE(ifolio)->i_ext;
+	struct f2fs_extent *i_ext = &F2FS_INODE(ientry)->i_ext;
 	struct extent_info ei;
 	int devi;
 
@@ -416,11 +416,11 @@ static void __drop_largest_extent(struct extent_tree *et,
 	}
 }
 
-void f2fs_init_read_extent_tree(struct inode *inode, struct folio *ifolio)
+void f2fs_init_read_extent_tree(struct inode *inode, struct f2fs_cached_block *ientry)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct extent_tree_info *eti = &sbi->extent_tree[EX_READ];
-	struct f2fs_extent *i_ext = &F2FS_INODE(ifolio)->i_ext;
+	struct f2fs_extent *i_ext = &F2FS_INODE(ientry)->i_ext;
 	struct extent_tree *et;
 	struct extent_node *en;
 	struct extent_info ei = {0};
@@ -428,9 +428,9 @@ void f2fs_init_read_extent_tree(struct inode *inode, struct folio *ifolio)
 	if (!__may_extent_tree(inode, EX_READ)) {
 		/* drop largest read extent */
 		if (i_ext->len) {
-			f2fs_folio_wait_writeback(ifolio, NODE, true, true);
+			f2fs_cache_wait_writeback(ientry);
 			i_ext->len = 0;
-			folio_mark_dirty(ifolio);
+			f2fs_mark_cache_dirty(ientry);
 		}
 		set_inode_flag(inode, FI_NO_EXTENT);
 		return;
@@ -956,7 +956,7 @@ static void __update_extent_cache(struct dnode_of_data *dn, enum extent_type typ
 	if (!__may_extent_tree(dn->inode, type))
 		return;
 
-	ei.fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_folio), dn->inode) +
+	ei.fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_entry), dn->inode) +
 								dn->ofs_in_node;
 	ei.len = 1;
 
fs/f2fs/f2fs.h +125 -98

464 lines not on this page · raw message

fs/f2fs/file.c +35 -39

258 lines not on this page · raw message

fs/f2fs/gc.c +22 -22

128 lines not on this page · raw message

fs/f2fs/inline.c +141 -138

670 lines not on this page · raw message

fs/f2fs/inode.c +72 -83

210 lines not on this page · raw message

fs/f2fs/namei.c +61 -57

118 lines not on this page · raw message

fs/f2fs/node.c +404 -471

875 lines not on this page · raw message

fs/f2fs/node.h +50 -57

107 lines not on this page · raw message

fs/f2fs/recovery.c +58 -57

115 lines not on this page · raw message

fs/f2fs/segment.c +35 -24

59 lines not on this page · raw message

fs/f2fs/segment.h +0 -20
index fff35c63a00f..341939478fd4 100644

Lines of this file are not shown. View the raw message.

fs/f2fs/shrinker.c +2 -1
index 1755c85849e4..20b3fe1f8c07 100644

Lines of this file are not shown. View the raw message.

fs/f2fs/super.c +12 -44

56 lines not on this page · raw message

fs/f2fs/xattr.c +61 -62

123 lines not on this page · raw message

fs/f2fs/xattr.h +6 -6
index bce3d93e4755..f50ed518c5cb 100644

Lines of this file are not shown. View the raw message.

include/linux/f2fs_fs.h +0 -1
index 0c027d00a1ea..105cfeedea74 100644

Lines of this file are not shown. View the raw message.

The rest of this diff is not shown. View the raw message.

Signature · 1 line
2.49.0

Part of the diff is not shown here. View the raw message.

#18 lore raw

Re: [PATCH v1 07/12] f2fs: cache: use node cache

Hi Chao,

Quoted 5 lines

entry = fn->entry;
f2fs_cache_get(entry); spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);

f2fs_cache_wait_writeback(entry);

The entry reference keeps the entry allocated, but the waiter does not hold the entry lock. After WRITEBACK is cleared, invalidation can detach the entry and set entry->cache to NULL. f2fs_cache_wait_writeback() checks and uses entry->cache through separate accesses.

Does the reference taken here also guarantee that the parent cache remains attached across the wait helper's check and use?

author maintainer
#19 lore raw

Re: [PATCH v1 07/12] f2fs: cache: use node cache

Quoted 8 lines · Wenjie Qi

Hi Chao,

Quoted 5 lines

entry = fn->entry;
f2fs_cache_get(entry); spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);

f2fs_cache_wait_writeback(entry);

Quoted 7 lines

The entry reference keeps the entry allocated, but the waiter does not hold the entry lock. After WRITEBACK is cleared, invalidation can detach the entry and set entry->cache to NULL. f2fs_cache_wait_writeback() checks and uses entry->cache through separate accesses.

Does the reference taken here also guarantee that the parent cache remains attached across the wait helper's check and use?

Do you mean race like this?

- f2fs_wait_on_node_caches_writeback
- f2fs_cache_get
- f2fs_cache_wait_writeback
- checked entry->cache, it's valid
- f2fs_truncate_cache
- f2fs_do_truncate_cache
- entry->cache = NULL
- f2fs_cache_wait_writeback_cond accesses

    invalid entry->cache

Thanks,

#20 lore raw

Re: [PATCH v1 07/12] f2fs: cache: use node cache

Yes, that is the race I meant.

Quoted 31 lines · Chao Yu
Quoted 8 lines · Wenjie Qi

Hi Chao,

Quoted 5 lines

entry = fn->entry;
f2fs_cache_get(entry); spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);

f2fs_cache_wait_writeback(entry);

Quoted 7 lines

The entry reference keeps the entry allocated, but the waiter does not hold the entry lock. After WRITEBACK is cleared, invalidation can detach the entry and set entry->cache to NULL. f2fs_cache_wait_writeback() checks and uses entry->cache through separate accesses.

Does the reference taken here also guarantee that the parent cache remains attached across the wait helper's check and use?

Do you mean race like this?

- f2fs_wait_on_node_caches_writeback
- f2fs_cache_get
- f2fs_cache_wait_writeback
- checked entry->cache, it's valid
- f2fs_truncate_cache
- f2fs_do_truncate_cache
- entry->cache = NULL
- f2fs_cache_wait_writeback_cond accesses

     invalid entry->cache

Thanks,

#21 lore raw

Re: [PATCH v1 07/12] f2fs: cache: use node cache

In reply to Chao Yu

Hi Chao,

Quoted 2 lines

out_err:

	f2fs_clear_cache_dirty(entry);

Both the inode-checksum and footer-validation failure paths reach this label. f2fs_clear_cache_dirty() clears the DIRTY state bit, but it does not clear UPTODATE or synchronize the DIRTY radix tag and dirty-node counter.

What invariant keeps these states consistent after node validation fails?

author maintainer
#22 lore raw

Re: [PATCH v1 07/12] f2fs: cache: use node cache

Quoted 10 lines · Wenjie Qi

Hi Chao,

Quoted 2 lines

out_err:

	f2fs_clear_cache_dirty(entry);

Both the inode-checksum and footer-validation failure paths reach this label. f2fs_clear_cache_dirty() clears the DIRTY state bit, but it does not clear UPTODATE or synchronize the DIRTY radix tag and dirty-node counter.

What invariant keeps these states consistent after node validation fails?

Will fix.

Thanks,

author maintainer
#23 lore raw

PATCH 8/12 f2fs: cache: initialize compress cache

In reply to Chao Yu

This patch introduces compress_blocks in f2fs_sb_info structure, initializes and destroys the compress cache during filesystem mount and unmount.

It adds an ino union field in struct f2fs_cached_block (sharing space with writeback linkage for zero memory overhead) to track per-inode cached blocks, and registers compress cache into the memory shrinker.

---
 fs/f2fs/cache.c | 9 ++++++++-
 fs/f2fs/cache.h | 5 ++++-
 fs/f2fs/f2fs.h  | 6 ++++++
 fs/f2fs/super.c | 9 ++++++++-
 4 files changed, 26 insertions(+), 3 deletions(-)
fs/f2fs/cache.c +8 -1
index 3176d62ce25b..2cecf6287207 100644
@@ -194,7 +194,10 @@ static struct f2fs_cached_block *f2fs_create_cache(
 	entry->index = index;
 
 	atomic_set(&entry->refcount, 0);
-	entry->next_entry = NULL;
+	if (!IS_COMPRESS_CACHE(cache))
+		entry->next_entry = NULL;
+	else
+		entry->ino = 0;
 	INIT_LIST_HEAD(&entry->list);
 
 	entry->cache = cache;
@@ -616,6 +619,10 @@ unsigned long f2fs_shrink_cache(struct f2fs_sb_info *sbi,
 		return freed;
 
 	freed += f2fs_do_shrink_cache(NODE_CACHE(sbi), nr_to_scan - freed);
+	if (freed >= nr_to_scan)
+		return freed;
+
+	freed += f2fs_do_shrink_cache(COMPRESS_CACHE(sbi), nr_to_scan - freed);
 	return freed;
 }
 
fs/f2fs/cache.h +4 -1
index 397019cfb861..23583a52b8f8 100644
@@ -38,6 +38,7 @@ struct f2fs_sb_info;
 enum f2fs_cache_type {
 	F2FS_META_CACHE,
 	F2FS_NODE_CACHE,
+	F2FS_COMPRESS_CACHE,
 };
 
 /* Main cache control structure (per sb_info) */
@@ -47,12 +48,13 @@ struct f2fs_cached_block_list {
 	spinlock_t tree_lock;		/* Lock for radix tree */
 	struct list_head lru_list;	/* Single global LRU list */
 	spinlock_t list_lock;		/* Lock for LRU list */
-	enum f2fs_cache_type type;	/* Cache type (Node or Meta) */
+	enum f2fs_cache_type type;	/* Cache type (Node, Meta, Compress) */
 	unsigned long num_entries;	/* Current number of entries */
 };
 
 #define IS_META_CACHE(cache) (cache->type == F2FS_META_CACHE)
 #define IS_NODE_CACHE(cache) (cache->type == F2FS_NODE_CACHE)
+#define IS_COMPRESS_CACHE(cache) (cache->type == F2FS_COMPRESS_CACHE)
 
 /* Flags for f2fs_cached_block state */
 enum f2fs_cached_state {
@@ -182,6 +184,7 @@ void f2fs_stop_cache_wb_thread(struct f2fs_sb_info *sbi);
 
 #define META_CACHE(sbi)		(&(sbi)->meta_blocks)
 #define NODE_CACHE(sbi)		(&(sbi)->node_blocks)
+#define COMPRESS_CACHE(sbi)	(&(sbi)->compress_blocks)
 
 #define f2fs_find_meta_cache(sbi, blkaddr)		\
 	f2fs_find_cache(META_CACHE(sbi), blkaddr)
fs/f2fs/f2fs.h +6 -0
index 4dd165ac05c9..36a6afbf3ba1 100644
@@ -2104,6 +2104,7 @@ struct f2fs_sb_info {
 	/* f2fs internal cache */
 	struct f2fs_cached_block_list meta_blocks;
 	struct f2fs_cached_block_list node_blocks;
+	struct f2fs_cached_block_list compress_blocks;
 
 	/* internal cache flush thread */
 	struct f2fs_cache_kthread cache_thread;
@@ -2325,6 +2326,11 @@ static inline bool f2fs_is_node_cache(struct f2fs_cached_block *entry)
 	return entry->cache && entry->cache == NODE_CACHE(entry->cache->sbi);
 }
 
+static inline bool f2fs_is_compress_cache(struct f2fs_cached_block *entry)
+{
+	return entry->cache && entry->cache == COMPRESS_CACHE(entry->cache->sbi);
+}
+
 static inline struct f2fs_bio *F2FS_BIO(struct bio *bio)
 {
 	return container_of(bio, struct f2fs_bio, bio);
fs/f2fs/super.c +8 -1
index bc1cff6cad0d..ac3a94736a91 100644
@@ -2030,6 +2030,7 @@ static void f2fs_put_super(struct super_block *sb)
 
 	f2fs_destroy_compress_inode(sbi);
 
+	f2fs_destroy_cache(COMPRESS_CACHE(sbi));
 	f2fs_destroy_cache(NODE_CACHE(sbi));
 	f2fs_destroy_cache(META_CACHE(sbi));
 
@@ -5203,10 +5204,14 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	if (err)
 		goto free_meta_cache;
 
+	err = f2fs_init_cache(sbi, COMPRESS_CACHE(sbi), F2FS_COMPRESS_CACHE);
+	if (err)
+		goto free_node_cache;
+
 	err = f2fs_get_valid_checkpoint(sbi);
 	if (err) {
 		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
-		goto free_node_cache;
+		goto free_compress_cache;
 	}
 
 	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
@@ -5519,6 +5524,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 free_devices:
 	destroy_device_list(sbi);
 	kvfree(sbi->ckpt);
+free_compress_cache:
+	f2fs_destroy_cache(COMPRESS_CACHE(sbi));
 free_node_cache:
 	f2fs_destroy_cache(NODE_CACHE(sbi));
 free_meta_cache:
Signature · 1 line
2.49.0
author maintainer
#24 lore raw

PATCH 9/12 f2fs: cache: use compress cache

In reply to Chao Yu

This patch migrates compressed cluster caching from the fake VFS inode page cache (sbi->compress_inode) to compress cache (sbi->compress_blocks).

It converts compression caching and decompression paths to use compress cache APIs, uses entry->ino for per-inode invalidation, and removes sbi->compress_inode.

---
 fs/f2fs/cache.c         |   2 +-
 fs/f2fs/cache.h         |   1 +
 fs/f2fs/compress.c      | 149 ++++++++++++++++------------------------
 fs/f2fs/debug.c         |   9 +--
 fs/f2fs/f2fs.h          |   8 +--
 fs/f2fs/inode.c         |  49 ++-----------
 fs/f2fs/node.c          |   2 +-
 fs/f2fs/super.c         |  15 ++--
 include/linux/f2fs_fs.h |   1 -
 9 files changed, 79 insertions(+), 157 deletions(-)
fs/f2fs/cache.c +1 -1
index 2cecf6287207..173bbf3aa94c 100644
@@ -414,7 +414,7 @@ static void f2fs_do_truncate_cache(struct f2fs_cached_block *entry,
 	spin_unlock(&cache->list_lock);
 }
 
-static void f2fs_truncate_cache(struct f2fs_cached_block *entry,
+void f2fs_truncate_cache(struct f2fs_cached_block *entry,
 					bool drop_dirty)
 {
 	f2fs_lock_cache(entry);
fs/f2fs/cache.h +1 -0
index 23583a52b8f8..12c061d28d23 100644
@@ -169,6 +169,7 @@ unsigned int f2fs_cache_gang_lookup_tag(struct f2fs_cached_block_list *cache,
 		unsigned int max_items, int tag);
 void f2fs_cache_gang_release(struct f2fs_cached_block **entries,
 				unsigned int nr_entries);
+void f2fs_truncate_cache(struct f2fs_cached_block *entry, bool drop_dirty);
 int f2fs_writeback_cache(struct f2fs_cached_block_list *cache, bool sync);
 void f2fs_cache_wait_on_all_writeback(struct f2fs_cached_block_list *cache);
 void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
fs/f2fs/compress.c +60 -89

210 lines not on this page · raw message

fs/f2fs/debug.c +5 -4
index bedaade92677..8cd06f7ba9e7 100644
@@ -225,8 +225,8 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->node_caches = NODE_CACHE(sbi)->num_entries;
 	si->meta_caches = META_CACHE(sbi)->num_entries;
 #ifdef CONFIG_F2FS_FS_COMPRESSION
-	if (sbi->compress_inode) {
-		si->compress_pages = COMPRESS_MAPPING(sbi)->nrpages;
+	if (test_opt(sbi, COMPRESS_CACHE)) {
+		si->compress_pages = COMPRESS_CACHE(sbi)->num_entries;
 		si->compress_page_hit = atomic_read(&sbi->compress_page_hit);
 	}
 #endif
@@ -386,10 +386,11 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 	si->page_mem += (unsigned long long)NODE_CACHE(sbi)->num_entries << PAGE_SHIFT;
 	si->cache_mem += NODE_CACHE(sbi)->num_entries * sizeof(struct f2fs_cached_block);
 #ifdef CONFIG_F2FS_FS_COMPRESSION
-	if (sbi->compress_inode) {
-		unsigned long npages = COMPRESS_MAPPING(sbi)->nrpages;
+	if (test_opt(sbi, COMPRESS_CACHE)) {
+		unsigned long npages = COMPRESS_CACHE(sbi)->num_entries;
 
 		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
+		si->cache_mem += npages * sizeof(struct f2fs_cached_block);
 	}
 #endif
 }
fs/f2fs/f2fs.h +2 -6
index 36a6afbf3ba1..0a749a59f386 100644
@@ -2075,7 +2075,6 @@ struct f2fs_sb_info {
 	u32 compr_new_inode;
 
 	/* For compressed block cache */
-	struct inode *compress_inode;		/* cache compressed blocks */
 	unsigned int compress_percent;		/* cache page percentage */
 	unsigned int compress_watermark;	/* cache page watermark */
 	atomic_t compress_page_hit;		/* cache hit count */
@@ -4817,13 +4816,11 @@ unsigned int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn,
 int f2fs_init_compress_ctx(struct compress_ctx *cc);
 void f2fs_destroy_compress_ctx(struct compress_ctx *cc, bool reuse);
 void f2fs_init_compress_info(struct f2fs_sb_info *sbi);
-int f2fs_init_compress_inode(struct f2fs_sb_info *sbi);
-void f2fs_destroy_compress_inode(struct f2fs_sb_info *sbi);
+void f2fs_init_compress_cache_context(struct f2fs_sb_info *sbi);
 int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi);
 void f2fs_destroy_page_array_cache(struct f2fs_sb_info *sbi);
 int __init f2fs_init_compress_cache(void);
 void f2fs_destroy_compress_cache(void);
-struct address_space *COMPRESS_MAPPING(struct f2fs_sb_info *sbi);
 void f2fs_invalidate_compress_pages_range(struct f2fs_sb_info *sbi,
 					block_t blkaddr, unsigned int len);
 bool f2fs_load_compressed_folio(struct f2fs_sb_info *sbi, struct folio *folio,
@@ -4872,8 +4869,7 @@ static inline void f2fs_put_folio_dic(struct folio *folio, bool in_task)
 static inline unsigned int f2fs_cluster_blocks_are_contiguous(
 			struct dnode_of_data *dn, unsigned int ofs_in_node) { return 0; }
 static inline bool f2fs_sanity_check_cluster(struct dnode_of_data *dn) { return false; }
-static inline int f2fs_init_compress_inode(struct f2fs_sb_info *sbi) { return 0; }
-static inline void f2fs_destroy_compress_inode(struct f2fs_sb_info *sbi) { }
+static inline void f2fs_init_compress_cache_context(struct f2fs_sb_info *sbi) { }
 static inline int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi) { return 0; }
 static inline void f2fs_destroy_page_array_cache(struct f2fs_sb_info *sbi) { }
 static inline int __init f2fs_init_compress_cache(void) { return 0; }
fs/f2fs/inode.c +4 -45

100 lines not on this page · raw message

fs/f2fs/node.c +1 -1
index 9482c86b5506..4b611ed3d8ef 100644
@@ -119,7 +119,7 @@ bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type)
 		 * exceed threshold, deny caching compress page.
 		 */
 		res = (free_ram > avail_ram * sbi->compress_watermark / 100) &&
-			(COMPRESS_MAPPING(sbi)->nrpages <
+			(COMPRESS_CACHE(sbi)->num_entries <
 			 free_ram * sbi->compress_percent / 100);
 #else
 		res = false;
fs/f2fs/super.c +5 -10
index ac3a94736a91..64c39286486d 100644
@@ -2028,8 +2028,6 @@ static void f2fs_put_super(struct super_block *sb)
 
 	f2fs_bug_on(sbi, sbi->fsync_node_num);
 
-	f2fs_destroy_compress_inode(sbi);
-
 	f2fs_destroy_cache(COMPRESS_CACHE(sbi));
 	f2fs_destroy_cache(NODE_CACHE(sbi));
 	f2fs_destroy_cache(META_CACHE(sbi));
@@ -5257,6 +5255,8 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 
 	f2fs_init_fsync_node_info(sbi);
 
+	f2fs_init_compress_cache_context(sbi);
+
 	/* setup checkpoint request control and start checkpoint issue thread */
 	f2fs_init_ckpt_req_control(sbi);
 	if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
@@ -5327,13 +5327,9 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 		goto free_ino_entry;
 	}
 
-	err = f2fs_init_compress_inode(sbi);
-	if (err)
-		goto free_root_inode;
-
 	err = f2fs_register_sysfs(sbi);
 	if (err)
-		goto free_compress_inode;
+		goto free_root_inode;
 
 	sbi->umount_lock_holder = current;
 #ifdef CONFIG_QUOTA
@@ -5501,8 +5497,6 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 	/* evict some inodes being cached by GC */
 	evict_inodes(sb);
 	f2fs_unregister_sysfs(sbi);
-free_compress_inode:
-	f2fs_destroy_compress_inode(sbi);
 free_root_inode:
 	dput(sb->s_root);
 	sb->s_root = NULL;
@@ -5619,7 +5613,8 @@ static void kill_f2fs_super(struct super_block *sb)
 		 * compress inode cache.
 		 */
 		if (test_opt(sbi, COMPRESS_CACHE))
-			truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
+			f2fs_invalidate_compress_pages_range(sbi,
+					0, UINT_MAX);
 #endif
 
 		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
include/linux/f2fs_fs.h +0 -1
index 105cfeedea74..53344e1f2b44 100644
@@ -34,7 +34,6 @@
 #define F2FS_RESERVED_NODE_NUM		3
 
 #define F2FS_ROOT_INO(sbi)	((sbi)->root_ino_num)
-#define F2FS_COMPRESS_INO(sbi)	(NM_I(sbi)->max_nid)
 
 #define F2FS_MAX_QUOTAS		3
 
Signature · 1 line
2.49.0
author maintainer
#25 lore raw

PATCH 10/12 f2fs: cache: support fault injection

In reply to Chao Yu

This patch adds fault injection support for metadata cache allocations to improve error-path test coverage.
- it integrates entry allocations with FAULT_KALLOC

---
 fs/f2fs/cache.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
fs/f2fs/cache.c +14 -10
index 173bbf3aa94c..c1e5c945391f 100644
@@ -175,20 +175,24 @@ static struct f2fs_cached_block *f2fs_create_cache(
 		struct f2fs_cached_block_list *cache,
 		unsigned long index, bool nofail)
 {
+	struct f2fs_sb_info *sbi = cache->sbi;
 	struct f2fs_cached_block *entry;
 	unsigned int flags = GFP_NOFS;
 
-	if (nofail)
+	if (nofail) {
 		flags |= __GFP_NOFAIL;
-
-	entry = kzalloc_obj(*entry, flags);
-	if (!entry)
-		return ERR_PTR(-ENOMEM);
-
-	entry->data = kzalloc(cache->sbi->blocksize, flags);
-	if (!entry->data) {
-		kfree(entry);
-		return ERR_PTR(-ENOMEM);
+		entry = kzalloc_obj(*entry, flags);
+		entry->data = kzalloc(sbi->blocksize, flags);
+	} else {
+		entry = f2fs_kzalloc(sbi, sizeof(*entry), flags);
+		if (!entry)
+			return ERR_PTR(-ENOMEM);
+
+		entry->data = f2fs_kzalloc(sbi, sbi->blocksize, flags);
+		if (!entry->data) {
+			kfree(entry);
+			return ERR_PTR(-ENOMEM);
+		}
 	}
 
 	entry->index = index;
Signature · 1 line
2.49.0
author maintainer
#26 lore raw

PATCH 11/12 f2fs: cache: introduce tracepoints

In reply to Chao Yu

This patch introduces ftrace tracepoints to observe and profile metadata cache operations:
- trace_f2fs_cache_set_dirty to trace marking a cached block dirty
- trace_f2fs_write_cache to trace single block writeback submission
- trace_f2fs_write_caches to tracesbatch writeback and sync sessions

---
 fs/f2fs/cache.c             |  2 ++
 fs/f2fs/checkpoint.c        |  6 ++++
 fs/f2fs/segment.c           |  3 ++
 include/trace/events/f2fs.h | 71 +++++++++++++++++++++++++++++++++++++
 4 files changed, 82 insertions(+)
fs/f2fs/cache.c +2 -0
index c1e5c945391f..d57adf52c3a5 100644
@@ -16,6 +16,7 @@
 #include "f2fs.h"
 #include "cache.h"
 #include "node.h"
+#include <trace/events/f2fs.h>
 #include "segment.h"
 
 void f2fs_cache_wait_writeback_cond(struct f2fs_cached_block *entry,
@@ -78,6 +79,7 @@ bool f2fs_mark_cache_dirty(struct f2fs_cached_block *entry)
 		enum count_type type = IS_META_CACHE(cache) ?
 				F2FS_DIRTY_META : F2FS_DIRTY_NODES;
 
+		trace_f2fs_cache_set_dirty(entry, IS_META_CACHE(cache) ? META : NODE);
 		f2fs_cache_update_tag(entry, 0, F2FS_CACHE_TAG_DIRTY);
 		inc_page_count(cache->sbi, type);
 		return true;
fs/f2fs/checkpoint.c +6 -0
index 508132652693..8bb3b701ce89 100644
@@ -550,6 +550,8 @@ static bool __f2fs_write_meta_cache(struct f2fs_cached_block *entry,
 {
 	struct f2fs_sb_info *sbi = entry->cache->sbi;
 
+	trace_f2fs_write_cache(entry, META);
+
 	if (unlikely(f2fs_cp_error(sbi))) {
 		if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
 			f2fs_force_clear_cache_dirty(entry);
@@ -608,6 +610,8 @@ long f2fs_sync_meta_caches(struct f2fs_sb_info *sbi, long nr_to_write,
 	struct blk_plug plug;
 	bool background = nr_to_write != LONG_MAX;
 
+	trace_f2fs_write_caches(sbi, nr_to_write, 0, META);
+
 	blk_start_plug(&plug);
 
 	while ((nr = f2fs_cache_gang_lookup_tag(META_CACHE(sbi), entries,
@@ -663,6 +667,8 @@ long f2fs_sync_meta_caches(struct f2fs_sb_info *sbi, long nr_to_write,
 
 	blk_finish_plug(&plug);
 
+	trace_f2fs_write_caches(sbi, nr_to_write, nwritten, META);
+
 	return nwritten;
 }
 
fs/f2fs/segment.c +3 -0
index 3a585fc7c197..e37d57052b49 100644
@@ -4170,6 +4170,9 @@ void f2fs_do_write_node_page(unsigned int nid, struct f2fs_io_info *fio)
 {
 	struct f2fs_summary sum;
 
+	if (fio->is_cache)
+		trace_f2fs_write_cache(fio->cache_entry, NODE);
+
 	set_summary(&sum, nid, 0, 0);
 	do_write_page(&sum, fio);
 
include/trace/events/f2fs.h +71 -0

83 lines not on this page · raw message

Signature · 1 line
2.49.0
author maintainer
#27 lore raw

PATCH 12/12 f2fs: cache: show per-cache usage in debugfs

In reply to Chao Yu

This patch exposes per-cache memory usage (entry struct memory vs. cached block buffer memory) for meta, node, and compress caches under the memory section in debugfs:
- meta entry: xxx KB, meta cache: xxx KB
- node entry: xxx KB, node cache: xxx KB
- compress entry: xxx KB, compress cache: xxx KB

---
 fs/f2fs/cache.h |  1 +
 fs/f2fs/debug.c | 49 ++++++++++++++++++++++++++++++++++++++++---------
 fs/f2fs/f2fs.h  |  2 ++
 3 files changed, 43 insertions(+), 9 deletions(-)
fs/f2fs/cache.h +1 -0
index 12c061d28d23..dccfc90fbab1 100644
@@ -39,6 +39,7 @@ enum f2fs_cache_type {
 	F2FS_META_CACHE,
 	F2FS_NODE_CACHE,
 	F2FS_COMPRESS_CACHE,
+	NR_CACHE_TYPES,
 };
 
 /* Main cache control structure (per sb_info) */
fs/f2fs/debug.c +40 -9

63 lines not on this page · raw message

fs/f2fs/f2fs.h +2 -0
index 0a749a59f386..42952dde79e0 100644
@@ -4453,6 +4453,8 @@ struct f2fs_stat_info {
 	unsigned int block_count[2];
 	unsigned int inplace_count;
 	unsigned long long base_mem, cache_mem, page_mem;
+	unsigned long long cache_entry_mem[NR_CACHE_TYPES];
+	unsigned long long cache_data_mem[NR_CACHE_TYPES];
 	struct f2fs_dev_stats *dev_stats;
 };
 
Signature · 1 line
2.49.0
#28 lore raw
In reply to Chao Yu

Hi Chao,

At Xiaomi, we have an out-of-tree F2FS node-cache prototype for mobile workloads that retains clean node contents in compressed memory after page-cache reclaim. This is different from this series:
F2FS_NODE_CACHE is the primary cache and can hold dirty authoritative data, while our compressed contents are clean and disposable.

In our mobile workloads, clean F2FS node pages can represent roughly 100 MB of uncompressed working-set memory. We have observed an LZ4 payload-to-original ratio of about 0.1 for these pages, excluding zsmalloc fragmentation, entry metadata, and workspaces. Node reads are also on latency-sensitive paths, while memory pressure can repeatedly reclaim and later reread the same node pages from storage.

These are observations from our out-of-tree mobile workloads, not benchmark results from this series or current upstream Linux.

If the primary metadata-cache model moves forward, the direction we would like to explore is one F2FS_NODE_CACHE slot per NID with mutually exclusive clean representations:

    UNCOMPRESSED_CLEAN <-> COMPRESSED_CLEAN

Active, dirty, and writeback entries would stay uncompressed, and a slot would not retain stable copies in both forms. Compression would be prepared outside the shrinker and direct-reclaim paths. Shrinker reclaim would continue to discard clean entries without compression, allocation, or I/O.

If an optional compressed representation for clean, idle F2FS_NODE_CACHE entries fits your intended metadata-cache roadmap, would it make sense to discuss the clean-entry state and lifetime contract first, and prototype it on top of v2 once the base cache is ready?

author maintainer
#29 lore raw

Wenjie,

Thanks for sharing the info, I guess we can discuss after we upstream this serial. :)

Quoted 34 lines · Wenjie Qi

Hi Chao,

At Xiaomi, we have an out-of-tree F2FS node-cache prototype for mobile workloads that retains clean node contents in compressed memory after page-cache reclaim. This is different from this series:
F2FS_NODE_CACHE is the primary cache and can hold dirty authoritative data, while our compressed contents are clean and disposable.

In our mobile workloads, clean F2FS node pages can represent roughly 100 MB of uncompressed working-set memory. We have observed an LZ4 payload-to-original ratio of about 0.1 for these pages, excluding zsmalloc fragmentation, entry metadata, and workspaces. Node reads are also on latency-sensitive paths, while memory pressure can repeatedly reclaim and later reread the same node pages from storage.

These are observations from our out-of-tree mobile workloads, not benchmark results from this series or current upstream Linux.

If the primary metadata-cache model moves forward, the direction we would like to explore is one F2FS_NODE_CACHE slot per NID with mutually exclusive clean representations:

     UNCOMPRESSED_CLEAN <-> COMPRESSED_CLEAN

Active, dirty, and writeback entries would stay uncompressed, and a slot would not retain stable copies in both forms. Compression would be prepared outside the shrinker and direct-reclaim paths. Shrinker reclaim would continue to discard clean entries without compression, allocation, or I/O.

If an optional compressed representation for clean, idle F2FS_NODE_CACHE entries fits your intended metadata-cache roadmap, would it make sense to discuss the clean-entry state and lifetime contract first, and prototype it on top of v2 once the base cache is ready?