f2fs: keep POSIX_FADV_NOREUSE ranges

[ Upstream commit ef0c333 ]

This patch records POSIX_FADV_NOREUSE ranges for users to reclaim the caches
instantly off from LRU.

Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Stable-dep-of: e462fc4 ("f2fs: maintain one time GC mode is enabled during whole zoned GC cycle")
Signed-off-by: Sasha Levin <sashal@kernel.org>

Author: Jaegeuk Kim

fs/f2fs/debug.c

@@ -100,6 +100,7 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->ndirty_imeta = get_pages(sbi, F2FS_DIRTY_IMETA);
 	si->ndirty_dirs = sbi->ndirty_inode[DIR_INODE];
 	si->ndirty_files = sbi->ndirty_inode[FILE_INODE];
+	si->ndonate_files = sbi->donate_files;
 	si->nquota_files = sbi->nquota_files;
 	si->ndirty_all = sbi->ndirty_inode[DIRTY_META];
 	si->aw_cnt = atomic_read(&sbi->atomic_files);
@@ -435,6 +436,8 @@ static int stat_show(struct seq_file *s, void *v)
 			   si->compr_inode, si->compr_blocks);
 		seq_printf(s, "  - Swapfile Inode: %u\n",
 			   si->swapfile_inode);
+		seq_printf(s, "  - Donate Inode: %u\n",
+			   si->ndonate_files);
 		seq_printf(s, "  - Orphan/Append/Update Inode: %u, %u, %u\n",
 			   si->orphans, si->append, si->update);
 		seq_printf(s, "\nMain area: %d segs, %d secs %d zones\n",

fs/f2fs/f2fs.h

@@ -849,6 +849,11 @@ struct f2fs_inode_info {
 #endif
 	struct list_head dirty_list;	/* dirty list for dirs and files */
 	struct list_head gdirty_list;	/* linked in global dirty list */
+
+	/* linked in global inode list for cache donation */
+	struct list_head gdonate_list;
+	pgoff_t donate_start, donate_end; /* inclusive */
+
 	struct task_struct *atomic_write_task;	/* store atomic write task */
 	struct extent_tree *extent_tree[NR_EXTENT_CACHES];
 					/* cached extent_tree entry */
@@ -1274,6 +1279,7 @@ enum inode_type {
 	DIR_INODE,			/* for dirty dir inode */
 	FILE_INODE,			/* for dirty regular/symlink inode */
 	DIRTY_META,			/* for all dirtied inode metadata */
+	DONATE_INODE,			/* for all inode to donate pages */
 	NR_INODE_TYPE,
 };
 
@@ -1629,6 +1635,9 @@ struct f2fs_sb_info {
 	unsigned int warm_data_age_threshold;
 	unsigned int last_age_weight;
 
+	/* control donate caches */
+	unsigned int donate_files;
+
 	/* basic filesystem units */
 	unsigned int log_sectors_per_block;	/* log2 sectors per block */
 	unsigned int log_blocksize;		/* log2 block size */
@@ -3997,7 +4006,8 @@ struct f2fs_stat_info {
 	unsigned long long allocated_data_blocks;
 	int ndirty_node, ndirty_dent, ndirty_meta, ndirty_imeta;
 	int ndirty_data, ndirty_qdata;
-	unsigned int ndirty_dirs, ndirty_files, nquota_files, ndirty_all;
+	unsigned int ndirty_dirs, ndirty_files, ndirty_all;
+	unsigned int nquota_files, ndonate_files;
 	int nats, dirty_nats, sits, dirty_sits;
 	int free_nids, avail_nids, alloc_nids;
 	int total_count, utilization;

fs/f2fs/file.c

@@ -2452,6 +2452,52 @@ static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
 	return ret;
 }
 
+static void f2fs_keep_noreuse_range(struct inode *inode,
+				loff_t offset, loff_t len)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	u64 max_bytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
+	u64 start, end;
+
+	if (!S_ISREG(inode->i_mode))
+		return;
+
+	if (offset >= max_bytes || len > max_bytes ||
+	    (offset + len) > max_bytes)
+		return;
+
+	start = offset >> PAGE_SHIFT;
+	end = DIV_ROUND_UP(offset + len, PAGE_SIZE);
+
+	inode_lock(inode);
+	if (f2fs_is_atomic_file(inode)) {
+		inode_unlock(inode);
+		return;
+	}
+
+	spin_lock(&sbi->inode_lock[DONATE_INODE]);
+	/* let's remove the range, if len = 0 */
+	if (!len) {
+		if (!list_empty(&F2FS_I(inode)->gdonate_list)) {
+			list_del_init(&F2FS_I(inode)->gdonate_list);
+			sbi->donate_files--;
+		}
+	} else {
+		if (list_empty(&F2FS_I(inode)->gdonate_list)) {
+			list_add_tail(&F2FS_I(inode)->gdonate_list,
+					&sbi->inode_list[DONATE_INODE]);
+			sbi->donate_files++;
+		} else {
+			list_move_tail(&F2FS_I(inode)->gdonate_list,
+					&sbi->inode_list[DONATE_INODE]);
+		}
+		F2FS_I(inode)->donate_start = start;
+		F2FS_I(inode)->donate_end = end - 1;
+	}
+	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
+	inode_unlock(inode);
+}
+
 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
 {
 	struct inode *inode = file_inode(filp);
@@ -5144,12 +5190,16 @@ static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len,
 	}
 
 	err = generic_fadvise(filp, offset, len, advice);
-	if (!err && advice == POSIX_FADV_DONTNEED &&
-		test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
-		f2fs_compressed_file(inode))
-		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
+	if (err)
+		return err;
 
-	return err;
+	if (advice == POSIX_FADV_DONTNEED &&
+	    (test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) &&
+	     f2fs_compressed_file(inode)))
+		f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino);
+	else if (advice == POSIX_FADV_NOREUSE)
+		f2fs_keep_noreuse_range(inode, offset, len);
+	return 0;
 }
 
 #ifdef CONFIG_COMPAT

fs/f2fs/inode.c

@@ -807,6 +807,19 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 	return 0;
 }
 
+static void f2fs_remove_donate_inode(struct inode *inode)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+
+	if (list_empty(&F2FS_I(inode)->gdonate_list))
+		return;
+
+	spin_lock(&sbi->inode_lock[DONATE_INODE]);
+	list_del_init(&F2FS_I(inode)->gdonate_list);
+	sbi->donate_files--;
+	spin_unlock(&sbi->inode_lock[DONATE_INODE]);
+}
+
 /*
  * Called at the last iput() if i_nlink is zero
  */
@@ -841,6 +854,7 @@ void f2fs_evict_inode(struct inode *inode)
 
 	f2fs_bug_on(sbi, get_dirty_pages(inode));
 	f2fs_remove_dirty_inode(inode);
+	f2fs_remove_donate_inode(inode);
 
 	f2fs_destroy_extent_tree(inode);
 

fs/f2fs/super.c

@@ -1429,6 +1429,7 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
 	spin_lock_init(&fi->i_size_lock);
 	INIT_LIST_HEAD(&fi->dirty_list);
 	INIT_LIST_HEAD(&fi->gdirty_list);
+	INIT_LIST_HEAD(&fi->gdonate_list);
 	init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
 	init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
 	init_f2fs_rwsem(&fi->i_xattr_sem);