f2fs: Run f2fs_write_end_io() asynchronously

The bio_for_each_segment_all() loop can take more than 10 ms for a large
bio on an ARM little core. This is too much for interrupt context. Hence
perform the write bio completion work asynchronously if a bio is large and
if f2fs_write_end_io() is called from atomic context. This patch reduces
the time spent in f2fs_write_end_io() from about 10 ms to about 150
microseconds on an Arm Cortex-A520 core if the max_atc_write_bio_size
parameter is changed to 16384.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

Documentation/ABI/testing/sysfs-fs-f2fs

@@ -1002,3 +1002,13 @@ Description:	It can be used to tune priority of f2fs critical task, e.g. f2fs_ck
 		threads, limitation as below:
 		- it requires user has CAP_SYS_NICE capability.
 		- the range is [100, 139], by default the value is 120.
+
+What:		/sys/fs/f2fs/<disk>/max_atc_write_bio_size
+Date:		June 2026
+Contact:	Bart Van Assche <bvanassche@acm.org>
+Description:	Every time a write operation completes f2fs_write_end_io() is
+		called. This function may be called from an atomic context,
+		e.g. from inside an interrupt handler. This attribute controls
+		the maximum size of a write bio that is completed in atomic
+		(atc) context. The default value for this attribute is UINT_MAX
+		which means that this functionality is disabled by default.

fs/f2fs/data.c

@@ -411,13 +411,32 @@ static void f2fs_write_end_bio(struct bio *bio)
 	bio_put(bio);
 }
 
-static void f2fs_write_end_io(struct bio *bio)
+static void f2fs_write_end_io_work(struct work_struct *work)
 {
-	iostat_update_and_unbind_ctx(bio);
+	struct bio *bio = &container_of(work, struct f2fs_bio, work)->bio;
 
 	f2fs_write_end_bio(bio);
 }
 
+static void f2fs_write_end_io(struct bio *bio)
+{
+	struct f2fs_sb_info *sbi;
+
+	iostat_update_and_unbind_ctx(bio);
+
+	sbi = bio->bi_private;
+
+	if (in_atomic() && bio->bi_iter.bi_size > sbi->max_atc_write_bio_size) {
+		struct work_struct *w;
+
+		w = &container_of(bio, struct f2fs_bio, bio)->work;
+		INIT_WORK(w, f2fs_write_end_io_work);
+		queue_work(sbi->wq, w);
+	} else {
+		f2fs_write_end_bio(bio);
+	}
+}
+
 #ifdef CONFIG_BLK_DEV_ZONED
 static void f2fs_zone_write_end_io(struct bio *bio)
 {

fs/f2fs/f2fs.h

@@ -1800,6 +1800,8 @@ struct f2fs_sb_info {
 	struct f2fs_sm_info *sm_info;		/* segment manager */
 
 	/* for bio operations */
+	/* Largest write bio size completed in atomic context (atc). */
+	u32 max_atc_write_bio_size;
 	struct f2fs_bio_info *write_io[NR_PAGE_TYPE];	/* for write bios */
 	/* keep migration IO order for LFS mode */
 	struct f2fs_rwsem io_order_lock;

fs/f2fs/super.c

@@ -5068,6 +5068,7 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc)
 
 	sb->s_fs_info = sbi;
 	sbi->raw_super = raw_super;
+	sbi->max_atc_write_bio_size = UINT_MAX;
 
 	INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
 	memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);

fs/f2fs/sysfs.c

@@ -1266,6 +1266,7 @@ F2FS_SBI_RW_ATTR(gc_idle_interval, interval_time[GC_TIME]);
 F2FS_SBI_RW_ATTR(umount_discard_timeout, interval_time[UMOUNT_DISCARD_TIMEOUT]);
 F2FS_SBI_RW_ATTR(gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_SBI_RW_ATTR(gc_reclaimed_segments, gc_reclaimed_segs);
+F2FS_SBI_RW_ATTR(max_atc_write_bio_size, max_atc_write_bio_size);
 F2FS_SBI_GENERAL_RW_ATTR(max_victim_search);
 F2FS_SBI_GENERAL_RW_ATTR(migration_granularity);
 F2FS_SBI_GENERAL_RW_ATTR(migration_window_granularity);
@@ -1509,6 +1510,7 @@ static struct attribute *f2fs_attrs[] = {
 	ATTR_LIST(seq_file_ra_mul),
 	ATTR_LIST(gc_segment_mode),
 	ATTR_LIST(gc_reclaimed_segments),
+	ATTR_LIST(max_atc_write_bio_size),
 	ATTR_LIST(max_fragment_chunk),
 	ATTR_LIST(max_fragment_hole),
 	ATTR_LIST(current_atomic_write),