Skip to contentSkip to conversation

f2fs: call __add_ino_entry out of the eviction path

PATCH In review Jaegeuk Kim 3 participants 6 messages Aug 4 – Aug 18 F2FS file system

lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/1 f2fs: call __add_ino_entry out of the eviction path

The f2fs_evict_inode() can be called during the direct reclaim path, but __add_ino_entry requires allocating some memory. Since we don't need to do that in that context, let's migrate it in other workqueue context.

---
 fs/f2fs/checkpoint.c | 15 ++++++++
 fs/f2fs/data.c       | 13 ++++++-
 fs/f2fs/f2fs.h       |  4 ++
 fs/f2fs/inode.c      | 88 +++++++++++++++++++++++++++++++++++++++++---
 fs/f2fs/super.c      |  9 ++++-
 5 files changed, 122 insertions(+), 7 deletions(-)
fs/f2fs/checkpoint.c +15 -0
index 064f5b537423..2d00b718974f 100644
@@ -766,6 +766,15 @@ static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 	spin_unlock(&im->ino_lock);
 }
 
+static void f2fs_wait_for_inode_record(struct f2fs_sb_info *sbi, int mode)
+{
+	if (mode != APPEND_INO && mode != UPDATE_INO)
+		return;
+
+	/* Let's wait for some pending updates for APPEND_INO and UPDATE_INO. */
+	flush_workqueue(sbi->evict_wq);
+}
+
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 {
 	/* add new dirty ino entry into list */
@@ -774,6 +783,8 @@ 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)
 {
+	f2fs_wait_for_inode_record(sbi, type);
+
 	/* remove dirty ino entry from list */
 	__remove_ino_entry(sbi, ino, type);
 }
@@ -784,6 +795,8 @@ bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
 	struct inode_management *im = &sbi->im[mode];
 	struct ino_entry *e;
 
+	f2fs_wait_for_inode_record(sbi, mode);
+
 	spin_lock(&im->ino_lock);
 	e = radix_tree_lookup(&im->ino_root, ino);
 	spin_unlock(&im->ino_lock);
@@ -798,6 +811,8 @@ void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all)
 	for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
 		struct inode_management *im = &sbi->im[i];
 
+		f2fs_wait_for_inode_record(sbi, i);
+
 		spin_lock(&im->ino_lock);
 		list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
 			list_del(&e->list);
fs/f2fs/data.c +12 -1
index c219ea76a3a7..6ae0eb37d20f 100644
@@ -4558,13 +4558,24 @@ int f2fs_init_wq(struct f2fs_sb_info *sbi)
 {
 	sbi->wq = alloc_workqueue("f2fs_wq", WQ_UNBOUND | WQ_HIGHPRI,
 				  num_online_cpus());
-	return sbi->wq ? 0 : -ENOMEM;
+	if (!sbi->wq)
+		return -ENOMEM;
+
+	sbi->evict_wq = alloc_workqueue("f2fs_evict_wq",
+			WQ_UNBOUND | WQ_HIGHPRI, num_online_cpus());
+	if (!sbi->evict_wq) {
+		destroy_workqueue(sbi->wq);
+		return -ENOMEM;
+	}
+	return 0;
 }
 
 void f2fs_destroy_wq(struct f2fs_sb_info *sbi)
 {
 	if (sbi->wq)
 		destroy_workqueue(sbi->wq);
+	if (sbi->evict_wq)
+		destroy_workqueue(sbi->evict_wq);
 }
 
 int __init f2fs_init_bio_entry_cache(void)
fs/f2fs/f2fs.h +4 -0
index c44908258dc1..54f9d3856b5c 100644
@@ -2016,6 +2016,8 @@ struct f2fs_sb_info {
 
 	struct workqueue_struct *wq;		/* bio completion workqueue */
 
+	struct workqueue_struct *evict_wq;	/* inode eviction workqueue */
+
 	/*
 	 * If we are in irq context, let's update error information into
 	 * on-disk superblock in the work.
@@ -3875,6 +3877,8 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc);
 void f2fs_remove_donate_inode(struct inode *inode);
 void f2fs_evict_inode(struct inode *inode);
 void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc);
+int f2fs_init_evict_inode_work(void);
+void f2fs_destroy_evict_inode_work(void);
 
 /*
  * namei.c
fs/f2fs/inode.c +83 -5

116 lines not on this page · raw message

fs/f2fs/super.c +8 -1
index 67abbf7ab477..6b55c30f0daa 100644
@@ -5763,10 +5763,16 @@ static int __init init_f2fs_fs(void)
 	err = f2fs_init_xattr_cache();
 	if (err)
 		goto free_casefold_cache;
-	err = register_filesystem(&f2fs_fs_type);
+	err = f2fs_init_evict_inode_work();
 	if (err)
 		goto free_xattr_cache;
+	err = register_filesystem(&f2fs_fs_type);
+	if (err)
+		goto free_evict_inode_cache;
 	return 0;
+
+free_evict_inode_cache:
+	f2fs_destroy_evict_inode_work();
 free_xattr_cache:
 	f2fs_destroy_xattr_cache();
 free_casefold_cache:
@@ -5809,6 +5815,7 @@ static int __init init_f2fs_fs(void)
 static void __exit exit_f2fs_fs(void)
 {
 	unregister_filesystem(&f2fs_fs_type);
+	f2fs_destroy_evict_inode_work();
 	f2fs_destroy_xattr_cache();
 	f2fs_destroy_casefold_cache();
 	f2fs_destroy_compress_cache();
Signature · 1 line
2.55.0.571.g244d577d93-goog
author maintainer
#1 lore raw

The f2fs_evict_inode() can be called during the direct reclaim path, but __add_ino_entry requires allocating some memory. Since we don't need to do that in that context, let's migrate it in other workqueue context.

---
Change log from v1:
- fix a bug on the wait logic

 fs/f2fs/checkpoint.c | 11 ++++++
 fs/f2fs/data.c       | 13 ++++++-
 fs/f2fs/f2fs.h       |  4 ++
 fs/f2fs/inode.c      | 91 +++++++++++++++++++++++++++++++++++++++++---
 fs/f2fs/super.c      |  9 ++++-
 5 files changed, 121 insertions(+), 7 deletions(-)
fs/f2fs/checkpoint.c +11 -0
index 064f5b537423..4413eccb5ecb 100644
@@ -766,6 +766,15 @@ static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 	spin_unlock(&im->ino_lock);
 }
 
+static void f2fs_wait_for_inode_record(struct f2fs_sb_info *sbi, int mode)
+{
+	if (mode != APPEND_INO && mode != UPDATE_INO)
+		return;
+
+	/* Let's wait for some pending updates for APPEND_INO and UPDATE_INO. */
+	flush_workqueue(sbi->evict_wq);
+}
+
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 {
 	/* add new dirty ino entry into list */
@@ -798,6 +807,8 @@ void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all)
 	for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
 		struct inode_management *im = &sbi->im[i];
 
+		f2fs_wait_for_inode_record(sbi, i);
+
 		spin_lock(&im->ino_lock);
 		list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
 			list_del(&e->list);
fs/f2fs/data.c +12 -1
index c219ea76a3a7..6ae0eb37d20f 100644
@@ -4558,13 +4558,24 @@ int f2fs_init_wq(struct f2fs_sb_info *sbi)
 {
 	sbi->wq = alloc_workqueue("f2fs_wq", WQ_UNBOUND | WQ_HIGHPRI,
 				  num_online_cpus());
-	return sbi->wq ? 0 : -ENOMEM;
+	if (!sbi->wq)
+		return -ENOMEM;
+
+	sbi->evict_wq = alloc_workqueue("f2fs_evict_wq",
+			WQ_UNBOUND | WQ_HIGHPRI, num_online_cpus());
+	if (!sbi->evict_wq) {
+		destroy_workqueue(sbi->wq);
+		return -ENOMEM;
+	}
+	return 0;
 }
 
 void f2fs_destroy_wq(struct f2fs_sb_info *sbi)
 {
 	if (sbi->wq)
 		destroy_workqueue(sbi->wq);
+	if (sbi->evict_wq)
+		destroy_workqueue(sbi->evict_wq);
 }
 
 int __init f2fs_init_bio_entry_cache(void)
fs/f2fs/f2fs.h +4 -0
index c44908258dc1..54f9d3856b5c 100644
@@ -2016,6 +2016,8 @@ struct f2fs_sb_info {
 
 	struct workqueue_struct *wq;		/* bio completion workqueue */
 
+	struct workqueue_struct *evict_wq;	/* inode eviction workqueue */
+
 	/*
 	 * If we are in irq context, let's update error information into
 	 * on-disk superblock in the work.
@@ -3875,6 +3877,8 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc);
 void f2fs_remove_donate_inode(struct inode *inode);
 void f2fs_evict_inode(struct inode *inode);
 void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc);
+int f2fs_init_evict_inode_work(void);
+void f2fs_destroy_evict_inode_work(void);
 
 /*
  * namei.c
fs/f2fs/inode.c +86 -5

125 lines not on this page · raw message

fs/f2fs/super.c +8 -1
index 67abbf7ab477..6b55c30f0daa 100644
@@ -5763,10 +5763,16 @@ static int __init init_f2fs_fs(void)
 	err = f2fs_init_xattr_cache();
 	if (err)
 		goto free_casefold_cache;
-	err = register_filesystem(&f2fs_fs_type);
+	err = f2fs_init_evict_inode_work();
 	if (err)
 		goto free_xattr_cache;
+	err = register_filesystem(&f2fs_fs_type);
+	if (err)
+		goto free_evict_inode_cache;
 	return 0;
+
+free_evict_inode_cache:
+	f2fs_destroy_evict_inode_work();
 free_xattr_cache:
 	f2fs_destroy_xattr_cache();
 free_casefold_cache:
@@ -5809,6 +5815,7 @@ static int __init init_f2fs_fs(void)
 static void __exit exit_f2fs_fs(void)
 {
 	unregister_filesystem(&f2fs_fs_type);
+	f2fs_destroy_evict_inode_work();
 	f2fs_destroy_xattr_cache();
 	f2fs_destroy_casefold_cache();
 	f2fs_destroy_compress_cache();
Signature · 1 line
2.55.0.571.g244d577d93-goog
maintainer
#2 lore raw
Quoted 189 lines · On 8/5/26 09:28, Jaegeuk Kim via Linux-f2fs-devel wrote:

The f2fs_evict_inode() can be called during the direct reclaim path, but __add_ino_entry requires allocating some memory. Since we don't need to do that in that context, let's migrate it in other workqueue context.

---
Change log from v1:
- fix a bug on the wait logic

 fs/f2fs/checkpoint.c | 11 ++++++
 fs/f2fs/data.c       | 13 ++++++-
 fs/f2fs/f2fs.h       |  4 ++
 fs/f2fs/inode.c      | 91 +++++++++++++++++++++++++++++++++++++++++---
 fs/f2fs/super.c      |  9 ++++-
 5 files changed, 121 insertions(+), 7 deletions(-)
fs/f2fs/checkpoint.c +11 -0
index 064f5b537423..4413eccb5ecb 100644
@@ -766,6 +766,15 @@ static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 	spin_unlock(&im->ino_lock);
 }
 
+static void f2fs_wait_for_inode_record(struct f2fs_sb_info *sbi, int mode)
+{
+	if (mode != APPEND_INO && mode != UPDATE_INO)
+		return;
+
+	/* Let's wait for some pending updates for APPEND_INO and UPDATE_INO. */
+	flush_workqueue(sbi->evict_wq);
+}
+
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 {
 	/* add new dirty ino entry into list */
@@ -798,6 +807,8 @@ void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all)
 	for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
 		struct inode_management *im = &sbi->im[i];
 
+		f2fs_wait_for_inode_record(sbi, i);
+
 		spin_lock(&im->ino_lock);
 		list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
 			list_del(&e->list);
fs/f2fs/data.c +12 -1
index c219ea76a3a7..6ae0eb37d20f 100644
@@ -4558,13 +4558,24 @@ int f2fs_init_wq(struct f2fs_sb_info *sbi)
 {
 	sbi->wq = alloc_workqueue("f2fs_wq", WQ_UNBOUND | WQ_HIGHPRI,
 				  num_online_cpus());
-	return sbi->wq ? 0 : -ENOMEM;
+	if (!sbi->wq)
+		return -ENOMEM;
+
+	sbi->evict_wq = alloc_workqueue("f2fs_evict_wq",
+			WQ_UNBOUND | WQ_HIGHPRI, num_online_cpus());
+	if (!sbi->evict_wq) {
+		destroy_workqueue(sbi->wq);
+		return -ENOMEM;
+	}
+	return 0;
 }
 
 void f2fs_destroy_wq(struct f2fs_sb_info *sbi)
 {
 	if (sbi->wq)
 		destroy_workqueue(sbi->wq);
+	if (sbi->evict_wq)
+		destroy_workqueue(sbi->evict_wq);
 }
 
 int __init f2fs_init_bio_entry_cache(void)
fs/f2fs/f2fs.h +4 -0
index c44908258dc1..54f9d3856b5c 100644
@@ -2016,6 +2016,8 @@ struct f2fs_sb_info {
 
 	struct workqueue_struct *wq;		/* bio completion workqueue */
 
+	struct workqueue_struct *evict_wq;	/* inode eviction workqueue */
+
 	/*
 	 * If we are in irq context, let's update error information into
 	 * on-disk superblock in the work.
@@ -3875,6 +3877,8 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc);
 void f2fs_remove_donate_inode(struct inode *inode);
 void f2fs_evict_inode(struct inode *inode);
 void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc);
+int f2fs_init_evict_inode_work(void);
+void f2fs_destroy_evict_inode_work(void);
 
 /*
  * namei.c
fs/f2fs/inode.c +51 -5
index c95e0b126da4..0d30489cf083 100644
@@ -24,6 +24,18 @@
 extern const struct address_space_operations f2fs_compress_aops;
 #endif
 
+#define NUM_PREALLOC_EVICT_INODE_WORK 8
+
+static struct kmem_cache *evict_inode_work_cache;
+static mempool_t *evict_inode_work_pool;
+
+struct evict_inode_work {
+	struct work_struct work;
+	struct f2fs_sb_info *sbi;
+	nid_t ino;
+	unsigned int add_ino_entry_bits;
+};
+
 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
 {
 	if (is_inode_flag_set(inode, FI_NEW_INODE))
@@ -637,6 +649,9 @@ struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 		inode->i_fop = &f2fs_dir_operations;
 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
+
+		/* Let's prepare APPEND/UPDATE_INO before future access. */
+		flush_workqueue(sbi->evict_wq);
 	} else if (S_ISLNK(inode->i_mode)) {
 		if (file_is_encrypt(inode))
 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
@@ -854,6 +869,25 @@ void f2fs_remove_donate_inode(struct inode *inode)
 	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
 }
 
+static void f2fs_record_inode_state(struct f2fs_sb_info *sbi, nid_t ino,
+				    unsigned int bits)
+{
+	if (bits & BIT(APPEND_INO))
+		f2fs_add_ino_entry(sbi, ino, APPEND_INO);
+	if (bits & BIT(UPDATE_INO))
+		f2fs_add_ino_entry(sbi, ino, UPDATE_INO);
+}
+
+static void f2fs_evict_inode_work(struct work_struct *work)
+{
+	struct evict_inode_work *ew =
+		container_of(work, struct evict_inode_work, work);
+
+	f2fs_record_inode_state(ew->sbi, ew->ino, ew->add_ino_entry_bits);
+
+	mempool_free(ew, evict_inode_work_pool);
+}
+
 /*
  * Called at the last iput() if i_nlink is zero
  */
@@ -864,6 +898,7 @@ void f2fs_evict_inode(struct inode *inode)
 	nid_t xnid = fi->i_xattr_nid;
 	int err = 0;
 	bool freeze_protected = false;
+	unsigned int record_bits = 0;
 
 	f2fs_abort_atomic_write(inode, true);
 
@@ -1003,12 +1038,32 @@ void f2fs_evict_inode(struct inode *inode)
 							inode->i_ino);
 	if (xnid)
 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
-	if (inode->i_nlink) {
-		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
-			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
-		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
-			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
+
+	if (!inode->i_nlink)
+		goto skip_record;
+
+	if (is_inode_flag_set(inode, FI_APPEND_WRITE))
+		record_bits = BIT(APPEND_INO);
+	if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
+		record_bits = BIT(UPDATE_INO);
+
+	if (!record_bits)
+		goto skip_record;
+
+	/* Let's do this in workqueue out of the direct reclaim path. */
+	if (current_is_kswapd()) {
+		f2fs_record_inode_state(sbi, inode->i_ino, record_bits);
+	} else {

fs/f2fs/inode.c void f2fs_evict_inode(struct inode *inode) · line 1066

Quoted 10 lines
fs/f2fs/inode.c +9 -0
+		struct evict_inode_work *ew =
+			mempool_alloc(evict_inode_work_pool, GFP_NOFS);
+
+		ew->sbi = sbi;
+		ew->ino = inode->i_ino;
+		ew->add_ino_entry_bits = record_bits;
+		INIT_WORK(&ew->work, f2fs_evict_inode_work);
+		queue_work(sbi->evict_wq, &ew->work);
 	}
+skip_record:

What do you think of wrapping above codes into a static function for cleanup?

Thanks,

Quoted 63 lines
fs/f2fs/inode.c +26 -0
 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
 		clear_inode_flag(inode, FI_FREE_NID);
@@ -1079,3 +1134,29 @@ void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc)
 	/* iput will drop the inode object */
 	iput(inode);
 }
+
+int __init f2fs_init_evict_inode_work(void)
+{
+	evict_inode_work_cache =
+		kmem_cache_create("f2fs_evict_inode_work",
+				  sizeof(struct evict_inode_work), 0, 0, NULL);
+	if (!evict_inode_work_cache)
+		goto fail;
+	evict_inode_work_pool =
+		mempool_create_slab_pool(NUM_PREALLOC_EVICT_INODE_WORK,
+					 evict_inode_work_cache);
+	if (!evict_inode_work_pool)
+		goto fail_free_cache;
+	return 0;
+
+fail_free_cache:
+	kmem_cache_destroy(evict_inode_work_cache);
+fail:
+	return -ENOMEM;
+}
+
+void f2fs_destroy_evict_inode_work(void)
+{
+	mempool_destroy(evict_inode_work_pool);
+	kmem_cache_destroy(evict_inode_work_cache);
+}
fs/f2fs/super.c +8 -1
index 67abbf7ab477..6b55c30f0daa 100644
@@ -5763,10 +5763,16 @@ static int __init init_f2fs_fs(void)
 	err = f2fs_init_xattr_cache();
 	if (err)
 		goto free_casefold_cache;
-	err = register_filesystem(&f2fs_fs_type);
+	err = f2fs_init_evict_inode_work();
 	if (err)
 		goto free_xattr_cache;
+	err = register_filesystem(&f2fs_fs_type);
+	if (err)
+		goto free_evict_inode_cache;
 	return 0;
+
+free_evict_inode_cache:
+	f2fs_destroy_evict_inode_work();
 free_xattr_cache:
 	f2fs_destroy_xattr_cache();
 free_casefold_cache:
@@ -5809,6 +5815,7 @@ static int __init init_f2fs_fs(void)
 static void __exit exit_f2fs_fs(void)
 {
 	unregister_filesystem(&f2fs_fs_type);
+	f2fs_destroy_evict_inode_work();
 	f2fs_destroy_xattr_cache();
 	f2fs_destroy_casefold_cache();
 	f2fs_destroy_compress_cache();
author maintainer
#3 lore raw
Quoted 194 lines · On 08/06, Chao Yu via Linux-f2fs-devel wrote:
Quoted 193 lines · On 8/5/26 09:28, Jaegeuk Kim via Linux-f2fs-devel wrote:

The f2fs_evict_inode() can be called during the direct reclaim path, but __add_ino_entry requires allocating some memory. Since we don't need to do that in that context, let's migrate it in other workqueue context.

---
Change log from v1:
- fix a bug on the wait logic

 fs/f2fs/checkpoint.c | 11 ++++++
 fs/f2fs/data.c       | 13 ++++++-
 fs/f2fs/f2fs.h       |  4 ++
 fs/f2fs/inode.c      | 91 +++++++++++++++++++++++++++++++++++++++++---
 fs/f2fs/super.c      |  9 ++++-
 5 files changed, 121 insertions(+), 7 deletions(-)
fs/f2fs/checkpoint.c +11 -0
index 064f5b537423..4413eccb5ecb 100644
@@ -766,6 +766,15 @@ static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 	spin_unlock(&im->ino_lock);
 }
 
+static void f2fs_wait_for_inode_record(struct f2fs_sb_info *sbi, int mode)
+{
+	if (mode != APPEND_INO && mode != UPDATE_INO)
+		return;
+
+	/* Let's wait for some pending updates for APPEND_INO and UPDATE_INO. */
+	flush_workqueue(sbi->evict_wq);
+}
+
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
 {
 	/* add new dirty ino entry into list */
@@ -798,6 +807,8 @@ void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all)
 	for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
 		struct inode_management *im = &sbi->im[i];
 
+		f2fs_wait_for_inode_record(sbi, i);
+
 		spin_lock(&im->ino_lock);
 		list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
 			list_del(&e->list);
fs/f2fs/data.c +12 -1
index c219ea76a3a7..6ae0eb37d20f 100644
@@ -4558,13 +4558,24 @@ int f2fs_init_wq(struct f2fs_sb_info *sbi)
 {
 	sbi->wq = alloc_workqueue("f2fs_wq", WQ_UNBOUND | WQ_HIGHPRI,
 				  num_online_cpus());
-	return sbi->wq ? 0 : -ENOMEM;
+	if (!sbi->wq)
+		return -ENOMEM;
+
+	sbi->evict_wq = alloc_workqueue("f2fs_evict_wq",
+			WQ_UNBOUND | WQ_HIGHPRI, num_online_cpus());
+	if (!sbi->evict_wq) {
+		destroy_workqueue(sbi->wq);
+		return -ENOMEM;
+	}
+	return 0;
 }
 
 void f2fs_destroy_wq(struct f2fs_sb_info *sbi)
 {
 	if (sbi->wq)
 		destroy_workqueue(sbi->wq);
+	if (sbi->evict_wq)
+		destroy_workqueue(sbi->evict_wq);
 }
 
 int __init f2fs_init_bio_entry_cache(void)
fs/f2fs/f2fs.h +4 -0
index c44908258dc1..54f9d3856b5c 100644
@@ -2016,6 +2016,8 @@ struct f2fs_sb_info {
 
 	struct workqueue_struct *wq;		/* bio completion workqueue */
 
+	struct workqueue_struct *evict_wq;	/* inode eviction workqueue */
+
 	/*
 	 * If we are in irq context, let's update error information into
 	 * on-disk superblock in the work.
@@ -3875,6 +3877,8 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc);
 void f2fs_remove_donate_inode(struct inode *inode);
 void f2fs_evict_inode(struct inode *inode);
 void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc);
+int f2fs_init_evict_inode_work(void);
+void f2fs_destroy_evict_inode_work(void);
 
 /*
  * namei.c
fs/f2fs/inode.c +55 -5
index c95e0b126da4..0d30489cf083 100644
@@ -24,6 +24,18 @@
 extern const struct address_space_operations f2fs_compress_aops;
 #endif
 
+#define NUM_PREALLOC_EVICT_INODE_WORK 8
+
+static struct kmem_cache *evict_inode_work_cache;
+static mempool_t *evict_inode_work_pool;
+
+struct evict_inode_work {
+	struct work_struct work;
+	struct f2fs_sb_info *sbi;
+	nid_t ino;
+	unsigned int add_ino_entry_bits;
+};
+
 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
 {
 	if (is_inode_flag_set(inode, FI_NEW_INODE))
@@ -637,6 +649,9 @@ struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 		inode->i_fop = &f2fs_dir_operations;
 		inode->i_mapping->a_ops = &f2fs_dblock_aops;
 		mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
+
+		/* Let's prepare APPEND/UPDATE_INO before future access. */
+		flush_workqueue(sbi->evict_wq);
 	} else if (S_ISLNK(inode->i_mode)) {
 		if (file_is_encrypt(inode))
 			inode->i_op = &f2fs_encrypted_symlink_inode_operations;
@@ -854,6 +869,25 @@ void f2fs_remove_donate_inode(struct inode *inode)
 	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
 }
 
+static void f2fs_record_inode_state(struct f2fs_sb_info *sbi, nid_t ino,
+				    unsigned int bits)
+{
+	if (bits & BIT(APPEND_INO))
+		f2fs_add_ino_entry(sbi, ino, APPEND_INO);
+	if (bits & BIT(UPDATE_INO))
+		f2fs_add_ino_entry(sbi, ino, UPDATE_INO);
+}
+
+static void f2fs_evict_inode_work(struct work_struct *work)
+{
+	struct evict_inode_work *ew =
+		container_of(work, struct evict_inode_work, work);
+
+	f2fs_record_inode_state(ew->sbi, ew->ino, ew->add_ino_entry_bits);
+
+	mempool_free(ew, evict_inode_work_pool);
+}
+
 /*
  * Called at the last iput() if i_nlink is zero
  */
@@ -864,6 +898,7 @@ void f2fs_evict_inode(struct inode *inode)
 	nid_t xnid = fi->i_xattr_nid;
 	int err = 0;
 	bool freeze_protected = false;
+	unsigned int record_bits = 0;
 
 	f2fs_abort_atomic_write(inode, true);
 
@@ -1003,12 +1038,32 @@ void f2fs_evict_inode(struct inode *inode)
 							inode->i_ino);
 	if (xnid)
 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
-	if (inode->i_nlink) {
-		if (is_inode_flag_set(inode, FI_APPEND_WRITE))
-			f2fs_add_ino_entry(sbi, inode->i_ino, APPEND_INO);
-		if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
-			f2fs_add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
+
+	if (!inode->i_nlink)
+		goto skip_record;
+
+	if (is_inode_flag_set(inode, FI_APPEND_WRITE))
+		record_bits = BIT(APPEND_INO);
+	if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
+		record_bits = BIT(UPDATE_INO);
+
+	if (!record_bits)
+		goto skip_record;
+
+	/* Let's do this in workqueue out of the direct reclaim path. */
+	if (current_is_kswapd()) {
+		f2fs_record_inode_state(sbi, inode->i_ino, record_bits);
+	} else {
+		struct evict_inode_work *ew =
+			mempool_alloc(evict_inode_work_pool, GFP_NOFS);
+
+		ew->sbi = sbi;
Quoted 8 lines

fs/f2fs/inode.c void f2fs_evict_inode(struct inode *inode) · line 1066

Quoted 6 lines
fs/f2fs/inode.c +5 -0
+		ew->ino = inode->i_ino;
+		ew->add_ino_entry_bits = record_bits;
+		INIT_WORK(&ew->work, f2fs_evict_inode_work);
+		queue_work(sbi->evict_wq, &ew->work);
 	}
+skip_record:

What do you think of wrapping above codes into a static function for cleanup?

Sumbitted another patch series having the refactoring patch. Thanks,

Quoted 73 lines

Thanks,

Quoted 56 lines
fs/f2fs/inode.c +26 -0
 	if (is_inode_flag_set(inode, FI_FREE_NID)) {
 		f2fs_alloc_nid_failed(sbi, inode->i_ino);
 		clear_inode_flag(inode, FI_FREE_NID);
@@ -1079,3 +1134,29 @@ void f2fs_handle_failed_inode(struct inode *inode, struct f2fs_lock_context *lc)
 	/* iput will drop the inode object */
 	iput(inode);
 }
+
+int __init f2fs_init_evict_inode_work(void)
+{
+	evict_inode_work_cache =
+		kmem_cache_create("f2fs_evict_inode_work",
+				  sizeof(struct evict_inode_work), 0, 0, NULL);
+	if (!evict_inode_work_cache)
+		goto fail;
+	evict_inode_work_pool =
+		mempool_create_slab_pool(NUM_PREALLOC_EVICT_INODE_WORK,
+					 evict_inode_work_cache);
+	if (!evict_inode_work_pool)
+		goto fail_free_cache;
+	return 0;
+
+fail_free_cache:
+	kmem_cache_destroy(evict_inode_work_cache);
+fail:
+	return -ENOMEM;
+}
+
+void f2fs_destroy_evict_inode_work(void)
+{
+	mempool_destroy(evict_inode_work_pool);
+	kmem_cache_destroy(evict_inode_work_cache);
+}
fs/f2fs/super.c +7 -1
index 67abbf7ab477..6b55c30f0daa 100644
@@ -5763,10 +5763,16 @@ static int __init init_f2fs_fs(void)
 	err = f2fs_init_xattr_cache();
 	if (err)
 		goto free_casefold_cache;
-	err = register_filesystem(&f2fs_fs_type);
+	err = f2fs_init_evict_inode_work();
 	if (err)
 		goto free_xattr_cache;
+	err = register_filesystem(&f2fs_fs_type);
+	if (err)
+		goto free_evict_inode_cache;
 	return 0;
+
+free_evict_inode_cache:
+	f2fs_destroy_evict_inode_work();
 free_xattr_cache:
 	f2fs_destroy_xattr_cache();
 free_casefold_cache:

fs/f2fs/super.c static int __init init_f2fs_fs(void) · line 5821

Quoted 7 lines
fs/f2fs/super.c +1 -0
@@ -5809,6 +5815,7 @@ static int __init init_f2fs_fs(void)
 static void __exit exit_f2fs_fs(void)
 {
 	unregister_filesystem(&f2fs_fs_type);
+	f2fs_destroy_evict_inode_work();
 	f2fs_destroy_xattr_cache();
 	f2fs_destroy_casefold_cache();
 	f2fs_destroy_compress_cache();

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

#4 lore raw
In reply to Jaegeuk Kim
kernel test robot noticed the following build errors:

Hi Jaegeuk,

kernel test robot noticed the following build errors:

[auto build test ERROR on v7.2-rc7]
[also build test ERROR on linus/master]
[cannot apply to jaegeuk-f2fs/dev-test jaegeuk-f2fs/dev next-20260814] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Jaegeuk-Kim/f2fs-call-__add_ino_entry-out-of-the-eviction-path/20260815-151820
base:   v7.2-rc7
patch link:    https://lore.kernel.org/r/20260804002027.2196300-1-jaegeuk%40kernel.org
patch subject: [PATCH] f2fs: call __add_ino_entry out of the eviction path
config: x86_64-kexec (https://download.01.org/0day-ci/archive/20260816/[email protected]/config)
compiler: clang version 22.1.8 (https://github.com/llvm/llvm-project ca7933e47d3a3451d81e72ac174dcb5aa28b59d1)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260816/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

>> fs/f2fs/inode.c:1051:6: error: call to undeclared function 'current_is_kswapd'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
    1051 |         if (current_is_kswapd()) {
         |             ^

1 error generated.

vim +/current_is_kswapd +1051 fs/f2fs/inode.c

887
888 /*
889 * Called at the last iput() if i_nlink is zero
890 */
891 void f2fs_evict_inode(struct inode *inode)
892 {
893 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
894 struct f2fs_inode_info *fi = F2FS_I(inode);
895 nid_t xnid = fi->i_xattr_nid;
896 int err = 0;
897 bool freeze_protected = false;
898 unsigned int record_bits = 0;
899
900 f2fs_abort_atomic_write(inode, true);
901
902 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
903 struct inode *cow_inode = fi->cow_inode;
904
905 f2fs_down_write(&F2FS_I(cow_inode)->i_sem);
906 clear_inode_flag(cow_inode, FI_COW_FILE);
907 F2FS_I(cow_inode)->atomic_inode = NULL;
908 fi->cow_inode = NULL;
909 f2fs_up_write(&F2FS_I(cow_inode)->i_sem);
910
911 iput(cow_inode);
912 }
913
914 trace_f2fs_evict_inode(inode);
915 truncate_inode_pages_final(&inode->i_data);
916
917 if ((inode->i_nlink || is_bad_inode(inode)) &&
918 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
919 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
920
921 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
922 inode->i_ino == F2FS_META_INO(sbi) ||
923 inode->i_ino == F2FS_COMPRESS_INO(sbi))
924 goto out_clear;
925
926 f2fs_bug_on(sbi, get_dirty_pages(inode));
927 f2fs_remove_dirty_inode(inode);
928 f2fs_remove_donate_inode(inode);
929
930 if (!IS_DEVICE_ALIASING(inode))
931 f2fs_destroy_extent_tree(inode);
932
933 if (inode->i_nlink || is_bad_inode(inode))
934 goto no_delete;
935
936 err = f2fs_dquot_initialize(inode);
937 if (err) {
938 err = 0;
939 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
940 }
941
942 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
943 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
944 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
945
946 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
947 sb_start_intwrite(inode->i_sb);
948 freeze_protected = true;
949 }
950 set_inode_flag(inode, FI_NO_ALLOC);
951 i_size_write(inode, 0);
952 retry:
953 if (F2FS_HAS_BLOCKS(inode))
954 err = f2fs_truncate(inode);
955
956 if (time_to_inject(sbi, FAULT_EVICT_INODE))
957 err = -EIO;
958
959 if (!err) {
960 struct f2fs_lock_context lc;
961
962 f2fs_lock_op(sbi, &lc);
963 err = f2fs_remove_inode_page(inode);
964 f2fs_unlock_op(sbi, &lc);
965 if (err == -ENOENT) {
966 err = 0;
967
968 /*
969 * in fuzzed image, another node may has the same
970 * block address as inode's, if it was truncated
971 * previously, truncation of inode node will fail.
972 */
973 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
974 f2fs_warn(F2FS_I_SB(inode),
975 "f2fs_evict_inode: inconsistent node id, ino:%llu",
976 inode->i_ino);
977 f2fs_inode_synced(inode);
978 set_sbi_flag(sbi, SBI_NEED_FSCK);
979 }
980 }
981 }
982
983 /* give more chances, if ENOMEM case */
984 if (err == -ENOMEM) {
985 err = 0;
986 goto retry;
987 }
988
989 if (IS_DEVICE_ALIASING(inode))
990 f2fs_destroy_extent_tree(inode);
991
992 if (err) {
993 f2fs_update_inode_page(inode);
994 if (dquot_initialize_needed(inode))
995 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
996
997 /*
998 * If both f2fs_truncate() and f2fs_update_inode_page() failed
999 * due to fuzzed corrupted inode, call f2fs_inode_synced() to
1000 * avoid triggering later f2fs_bug_on().
1001 */
1002 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1003 f2fs_warn(sbi,
1004 "f2fs_evict_inode: inode is dirty, ino:%llu",
1005 inode->i_ino);
1006 f2fs_inode_synced(inode);
1007 set_sbi_flag(sbi, SBI_NEED_FSCK);
1008 }
1009 }
1010 if (freeze_protected)
1011 sb_end_intwrite(inode->i_sb);
1012 no_delete:
1013 dquot_drop(inode);
1014
1015 stat_dec_inline_xattr(inode);
1016 stat_dec_inline_dir(inode);
1017 stat_dec_inline_inode(inode);
1018 stat_dec_compr_inode(inode);
1019 stat_sub_compr_blocks(inode,
1020 atomic_read(&fi->i_compr_blocks));
1021
1022 if (likely(!f2fs_cp_error(sbi) &&
1023 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1024 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
1025
1026 /*
1027 * anyway, it needs to remove the inode from sbi->inode_list[DIRTY_META]
1028 * list to avoid UAF in f2fs_sync_inode_meta() during checkpoint.
1029 */
1030 f2fs_inode_synced(inode);
1031
1032 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
1033 if (inode->i_ino)
1034 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
1035 inode->i_ino);
1036 if (xnid)
1037 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
1038
1039 if (!inode->i_nlink)
1040 goto skip_record;
1041
1042 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
1043 record_bits = BIT(APPEND_INO);
1044 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
1045 record_bits = BIT(UPDATE_INO);
1046
1047 if (!record_bits)
1048 goto skip_record;
1049
1050 /* Let's do this in workqueue out of the direct reclaim path. */

> 1051		if (current_is_kswapd()) {

1052 f2fs_record_inode_state(sbi, inode->i_ino, record_bits);
1053 } else {
1054 struct evict_inode_work *ew =
1055 mempool_alloc(evict_inode_work_pool, GFP_NOFS);
1056
1057 ew->sbi = sbi;
1058 ew->ino = inode->i_ino;
1059 ew->add_ino_entry_bits = record_bits;
1060 INIT_WORK(&ew->work, f2fs_evict_inode_work);
1061 queue_work(sbi->evict_wq, &ew->work);
1062 }
1063 skip_record:
1064 if (is_inode_flag_set(inode, FI_FREE_NID)) {
1065 f2fs_alloc_nid_failed(sbi, inode->i_ino);
1066 clear_inode_flag(inode, FI_FREE_NID);
1067 } else {
1068 /*
1069 * If xattr nid is corrupted, we can reach out error condition,
1070 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
1071 * In that case, f2fs_check_nid_range() is enough to give a clue.
1072 */
1073 }
1074 out_clear:
1075 fscrypt_put_encryption_info(inode);
1076 clear_inode(inode);
1077 }
1078

--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

#5 lore raw
In reply to Jaegeuk Kim
kernel test robot noticed the following build errors:

Hi Jaegeuk,

kernel test robot noticed the following build errors:

[auto build test ERROR on v7.2-rc7]
[also build test ERROR on linus/master]
[cannot apply to jaegeuk-f2fs/dev-test jaegeuk-f2fs/dev next-20260817] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Jaegeuk-Kim/f2fs-call-__add_ino_entry-out-of-the-eviction-path/20260815-151820
base:   v7.2-rc7
patch link:    https://lore.kernel.org/r/20260804002027.2196300-1-jaegeuk%40kernel.org
patch subject: [PATCH] f2fs: call __add_ino_entry out of the eviction path
config: x86_64-kexec (https://download.01.org/0day-ci/archive/20260818/[email protected]/config)
compiler: clang version 22.1.3 (https://github.com/llvm/llvm-project e9846648fd6183ee6d8cbdb4502213fcf902a211)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260818/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

>> fs/f2fs/inode.c:1051:6: error: call to undeclared function 'current_is_kswapd'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
    1051 |         if (current_is_kswapd()) {
         |             ^

1 error generated.

vim +/current_is_kswapd +1051 fs/f2fs/inode.c

887
888 /*
889 * Called at the last iput() if i_nlink is zero
890 */
891 void f2fs_evict_inode(struct inode *inode)
892 {
893 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
894 struct f2fs_inode_info *fi = F2FS_I(inode);
895 nid_t xnid = fi->i_xattr_nid;
896 int err = 0;
897 bool freeze_protected = false;
898 unsigned int record_bits = 0;
899
900 f2fs_abort_atomic_write(inode, true);
901
902 if (fi->cow_inode && f2fs_is_cow_file(fi->cow_inode)) {
903 struct inode *cow_inode = fi->cow_inode;
904
905 f2fs_down_write(&F2FS_I(cow_inode)->i_sem);
906 clear_inode_flag(cow_inode, FI_COW_FILE);
907 F2FS_I(cow_inode)->atomic_inode = NULL;
908 fi->cow_inode = NULL;
909 f2fs_up_write(&F2FS_I(cow_inode)->i_sem);
910
911 iput(cow_inode);
912 }
913
914 trace_f2fs_evict_inode(inode);
915 truncate_inode_pages_final(&inode->i_data);
916
917 if ((inode->i_nlink || is_bad_inode(inode)) &&
918 test_opt(sbi, COMPRESS_CACHE) && f2fs_compressed_file(inode))
919 f2fs_invalidate_compress_pages(sbi, inode->i_ino);
920
921 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
922 inode->i_ino == F2FS_META_INO(sbi) ||
923 inode->i_ino == F2FS_COMPRESS_INO(sbi))
924 goto out_clear;
925
926 f2fs_bug_on(sbi, get_dirty_pages(inode));
927 f2fs_remove_dirty_inode(inode);
928 f2fs_remove_donate_inode(inode);
929
930 if (!IS_DEVICE_ALIASING(inode))
931 f2fs_destroy_extent_tree(inode);
932
933 if (inode->i_nlink || is_bad_inode(inode))
934 goto no_delete;
935
936 err = f2fs_dquot_initialize(inode);
937 if (err) {
938 err = 0;
939 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
940 }
941
942 f2fs_remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
943 f2fs_remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
944 f2fs_remove_ino_entry(sbi, inode->i_ino, FLUSH_INO);
945
946 if (!is_sbi_flag_set(sbi, SBI_IS_FREEZING)) {
947 sb_start_intwrite(inode->i_sb);
948 freeze_protected = true;
949 }
950 set_inode_flag(inode, FI_NO_ALLOC);
951 i_size_write(inode, 0);
952 retry:
953 if (F2FS_HAS_BLOCKS(inode))
954 err = f2fs_truncate(inode);
955
956 if (time_to_inject(sbi, FAULT_EVICT_INODE))
957 err = -EIO;
958
959 if (!err) {
960 struct f2fs_lock_context lc;
961
962 f2fs_lock_op(sbi, &lc);
963 err = f2fs_remove_inode_page(inode);
964 f2fs_unlock_op(sbi, &lc);
965 if (err == -ENOENT) {
966 err = 0;
967
968 /*
969 * in fuzzed image, another node may has the same
970 * block address as inode's, if it was truncated
971 * previously, truncation of inode node will fail.
972 */
973 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
974 f2fs_warn(F2FS_I_SB(inode),
975 "f2fs_evict_inode: inconsistent node id, ino:%llu",
976 inode->i_ino);
977 f2fs_inode_synced(inode);
978 set_sbi_flag(sbi, SBI_NEED_FSCK);
979 }
980 }
981 }
982
983 /* give more chances, if ENOMEM case */
984 if (err == -ENOMEM) {
985 err = 0;
986 goto retry;
987 }
988
989 if (IS_DEVICE_ALIASING(inode))
990 f2fs_destroy_extent_tree(inode);
991
992 if (err) {
993 f2fs_update_inode_page(inode);
994 if (dquot_initialize_needed(inode))
995 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
996
997 /*
998 * If both f2fs_truncate() and f2fs_update_inode_page() failed
999 * due to fuzzed corrupted inode, call f2fs_inode_synced() to
1000 * avoid triggering later f2fs_bug_on().
1001 */
1002 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1003 f2fs_warn(sbi,
1004 "f2fs_evict_inode: inode is dirty, ino:%llu",
1005 inode->i_ino);
1006 f2fs_inode_synced(inode);
1007 set_sbi_flag(sbi, SBI_NEED_FSCK);
1008 }
1009 }
1010 if (freeze_protected)
1011 sb_end_intwrite(inode->i_sb);
1012 no_delete:
1013 dquot_drop(inode);
1014
1015 stat_dec_inline_xattr(inode);
1016 stat_dec_inline_dir(inode);
1017 stat_dec_inline_inode(inode);
1018 stat_dec_compr_inode(inode);
1019 stat_sub_compr_blocks(inode,
1020 atomic_read(&fi->i_compr_blocks));
1021
1022 if (likely(!f2fs_cp_error(sbi) &&
1023 !is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1024 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
1025
1026 /*
1027 * anyway, it needs to remove the inode from sbi->inode_list[DIRTY_META]
1028 * list to avoid UAF in f2fs_sync_inode_meta() during checkpoint.
1029 */
1030 f2fs_inode_synced(inode);
1031
1032 /* for the case f2fs_new_inode() was failed, .i_ino is zero, skip it */
1033 if (inode->i_ino)
1034 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
1035 inode->i_ino);
1036 if (xnid)
1037 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
1038
1039 if (!inode->i_nlink)
1040 goto skip_record;
1041
1042 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
1043 record_bits = BIT(APPEND_INO);
1044 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
1045 record_bits = BIT(UPDATE_INO);
1046
1047 if (!record_bits)
1048 goto skip_record;
1049
1050 /* Let's do this in workqueue out of the direct reclaim path. */

> 1051		if (current_is_kswapd()) {

1052 f2fs_record_inode_state(sbi, inode->i_ino, record_bits);
1053 } else {
1054 struct evict_inode_work *ew =
1055 mempool_alloc(evict_inode_work_pool, GFP_NOFS);
1056
1057 ew->sbi = sbi;
1058 ew->ino = inode->i_ino;
1059 ew->add_ino_entry_bits = record_bits;
1060 INIT_WORK(&ew->work, f2fs_evict_inode_work);
1061 queue_work(sbi->evict_wq, &ew->work);
1062 }
1063 skip_record:
1064 if (is_inode_flag_set(inode, FI_FREE_NID)) {
1065 f2fs_alloc_nid_failed(sbi, inode->i_ino);
1066 clear_inode_flag(inode, FI_FREE_NID);
1067 } else {
1068 /*
1069 * If xattr nid is corrupted, we can reach out error condition,
1070 * err & !f2fs_exist_written_data(sbi, inode->i_ino, ORPHAN_INO)).
1071 * In that case, f2fs_check_nid_range() is enough to give a clue.
1072 */
1073 }
1074 out_clear:
1075 fscrypt_put_encryption_info(inode);
1076 clear_inode(inode);
1077 }
1078

--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki