Re: [PATCH v3] f2fs: zone: don't block IO if there is remained open zone

From: Daeho Jeong
Date: Wed Apr 10 2024 - 14:29:33 EST

Next message: Haitao Huang: "Re: Re: [PATCH v10 12/14] x86/sgx: Turn on per-cgroup EPC reclamation"
Previous message: Haitao Huang: "[PATCH v11 13/14] Docs/x86/sgx: Add description for cgroup support"
Next in thread: Jaegeuk Kim: "Re: [PATCH v3] f2fs: zone: don't block IO if there is remained open zone"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Sun, Apr 7, 2024 at 6:58 AM Chao Yu <chao@xxxxxxxxxx> wrote:
>
> max open zone may be larger than log header number of f2fs, for
> such case, it doesn't need to wait last IO in previous zone, let's
> introduce available_open_zone semaphore, and reduce it once we
> submit first write IO in a zone, and increase it after completion
> of last IO in the zone.
>
> Cc: Daeho Jeong <daeho43@xxxxxxxxx>
> Signed-off-by: Chao Yu <chao@xxxxxxxxxx>
> ---
> v3:
> - avoid race condition in between __submit_merged_bio()
> and __allocate_new_segment().
> fs/f2fs/data.c | 105 ++++++++++++++++++++++++++++++----------------
> fs/f2fs/f2fs.h | 34 ++++++++++++---
> fs/f2fs/iostat.c | 7 ++++
> fs/f2fs/iostat.h | 2 +
> fs/f2fs/segment.c | 43 ++++++++++++++++---
> fs/f2fs/segment.h | 12 +++++-
> fs/f2fs/super.c | 2 +
> 7 files changed, 156 insertions(+), 49 deletions(-)
>
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index 0d88649c60a5..18a4ac0a06bc 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -373,11 +373,10 @@ static void f2fs_write_end_io(struct bio *bio)
> #ifdef CONFIG_BLK_DEV_ZONED
> static void f2fs_zone_write_end_io(struct bio *bio)
> {
> - struct f2fs_bio_info *io = (struct f2fs_bio_info *)bio->bi_private;
> + struct f2fs_sb_info *sbi = iostat_get_bio_private(bio);
>
> - bio->bi_private = io->bi_private;
> - complete(&io->zone_wait);
> f2fs_write_end_io(bio);
> + up(&sbi->available_open_zones);
> }
> #endif
>
> @@ -531,6 +530,24 @@ static void __submit_merged_bio(struct f2fs_bio_info *io)
> if (!io->bio)
> return;
>
> +#ifdef CONFIG_BLK_DEV_ZONED
> + if (io->open_zone) {
> + /*
> + * if there is no open zone, it will wait for last IO in
> + * previous zone before submitting new IO.
> + */
> + down(&fio->sbi->available_open_zones);
> + io->open_zone = false;
> + io->zone_openned = true;
> + }
> +
> + if (io->close_zone) {
> + io->bio->bi_end_io = f2fs_zone_write_end_io;
> + io->zone_openned = false;
> + io->close_zone = false;
> + }
> +#endif
> +
> if (is_read_io(fio->op)) {
> trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
> f2fs_submit_read_bio(io->sbi, io->bio, fio->type);
> @@ -601,9 +618,9 @@ int f2fs_init_write_merge_io(struct f2fs_sb_info *sbi)
> INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
> init_f2fs_rwsem(&sbi->write_io[i][j].bio_list_lock);
> #ifdef CONFIG_BLK_DEV_ZONED
> - init_completion(&sbi->write_io[i][j].zone_wait);
> - sbi->write_io[i][j].zone_pending_bio = NULL;
> - sbi->write_io[i][j].bi_private = NULL;
> + sbi->write_io[i][j].open_zone = false;
> + sbi->write_io[i][j].zone_openned = false;
> + sbi->write_io[i][j].close_zone = false;
> #endif
> }
> }
> @@ -634,6 +651,31 @@ static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
> f2fs_up_write(&io->io_rwsem);
> }
>
> +void f2fs_blkzoned_submit_merged_write(struct f2fs_sb_info *sbi, int type)
> +{
> +#ifdef CONFIG_BLK_DEV_ZONED
> + struct f2fs_bio_info *io;
> +
> + if (!f2fs_sb_has_blkzoned(sbi))
> + return;
> +
> + io = sbi->write_io[PAGE_TYPE(type)] + type_to_temp(type);
> +
> + f2fs_down_write(&io->io_rwsem);
> + if (io->zone_openned) {
> + if (io->bio) {
> + io->close_zone = true;
> + __submit_merged_bio(io);
> + } else if (io->zone_openned) {

We already checked if this value is true above.

Thanks,

> + up(&sbi->available_open_zones);
> + io->zone_openned = false;
> + }
> + }
> + f2fs_up_write(&io->io_rwsem);
> +#endif
> +
> +}
> +
> static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
> struct inode *inode, struct page *page,
> nid_t ino, enum page_type type, bool force)
> @@ -918,22 +960,16 @@ int f2fs_merge_page_bio(struct f2fs_io_info *fio)
> }
>
> #ifdef CONFIG_BLK_DEV_ZONED
> -static bool is_end_zone_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr)
> +static bool is_blkaddr_zone_boundary(struct f2fs_sb_info *sbi,
> + block_t blkaddr, bool start)
> {
> - int devi = 0;
> + if (!f2fs_blkaddr_in_seqzone(sbi, blkaddr))
> + return false;
> +
> + if (start)
> + return (blkaddr % sbi->blocks_per_blkz) == 0;
> + return (blkaddr % sbi->blocks_per_blkz == sbi->blocks_per_blkz - 1);
>
> - if (f2fs_is_multi_device(sbi)) {
> - devi = f2fs_target_device_index(sbi, blkaddr);
> - if (blkaddr < FDEV(devi).start_blk ||
> - blkaddr > FDEV(devi).end_blk) {
> - f2fs_err(sbi, "Invalid block %x", blkaddr);
> - return false;
> - }
> - blkaddr -= FDEV(devi).start_blk;
> - }
> - return bdev_is_zoned(FDEV(devi).bdev) &&
> - f2fs_blkz_is_seq(sbi, devi, blkaddr) &&
> - (blkaddr % sbi->blocks_per_blkz == sbi->blocks_per_blkz - 1);
> }
> #endif
>
> @@ -944,20 +980,14 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
> struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
> struct page *bio_page;
> enum count_type type;
> +#ifdef CONFIG_BLK_DEV_ZONED
> + bool blkzoned = f2fs_sb_has_blkzoned(sbi) && btype < META;
> +#endif
>
> f2fs_bug_on(sbi, is_read_io(fio->op));
>
> f2fs_down_write(&io->io_rwsem);
> 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;
> - }
> -#endif
> -
> if (fio->in_list) {
> spin_lock(&io->io_lock);
> if (list_empty(&io->io_list)) {
> @@ -985,6 +1015,11 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
> type = WB_DATA_TYPE(bio_page, fio->compressed_page);
> inc_page_count(sbi, type);
>
> +#ifdef CONFIG_BLK_DEV_ZONED
> + if (blkzoned && is_blkaddr_zone_boundary(sbi, fio->new_blkaddr, true))
> + io->open_zone = true;
> +#endif
> +
> if (io->bio &&
> (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
> fio->new_blkaddr) ||
> @@ -1010,15 +1045,11 @@ void f2fs_submit_page_write(struct f2fs_io_info *fio)
> io->last_block_in_bio = fio->new_blkaddr;
>
> trace_f2fs_submit_page_write(fio->page, 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;
> + if (blkzoned &&
> + is_blkaddr_zone_boundary(sbi, fio->new_blkaddr, false)) {
> + io->close_zone = true;
> __submit_merged_bio(io);
> }
> #endif
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 694f8a52cb84..8a26530cf4fb 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -1234,16 +1234,16 @@ struct f2fs_bio_info {
> struct bio *bio; /* bios to merge */
> sector_t last_block_in_bio; /* last block number */
> struct f2fs_io_info fio; /* store buffered io info. */
> -#ifdef CONFIG_BLK_DEV_ZONED
> - struct completion zone_wait; /* condition value for the previous open zone to close */
> - struct bio *zone_pending_bio; /* pending bio for the previous zone */
> - void *bi_private; /* previous bi_private for pending bio */
> -#endif
> struct f2fs_rwsem io_rwsem; /* blocking op for bio */
> spinlock_t io_lock; /* serialize DATA/NODE IOs */
> struct list_head io_list; /* track fios */
> struct list_head bio_list; /* bio entry list head */
> struct f2fs_rwsem bio_list_lock; /* lock to protect bio entry list */
> +#ifdef CONFIG_BLK_DEV_ZONED
> + bool open_zone; /* open a zone */
> + bool zone_openned; /* zone has been openned */
> + bool close_zone; /* close a zone */
> +#endif
> };
>
> #define FDEV(i) (sbi->devs[i])
> @@ -1560,6 +1560,7 @@ struct f2fs_sb_info {
> #ifdef CONFIG_BLK_DEV_ZONED
> unsigned int blocks_per_blkz; /* F2FS blocks per zone */
> unsigned int max_open_zones; /* max open zone resources of the zoned device */
> + struct semaphore available_open_zones; /* available open zones */
> #endif
>
> /* for node-related operations */
> @@ -3822,6 +3823,7 @@ void f2fs_destroy_bio_entry_cache(void);
> void f2fs_submit_read_bio(struct f2fs_sb_info *sbi, struct bio *bio,
> enum page_type type);
> int f2fs_init_write_merge_io(struct f2fs_sb_info *sbi);
> +void f2fs_blkzoned_submit_merged_write(struct f2fs_sb_info *sbi, int type);
> void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type);
> void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
> struct inode *inode, struct page *page,
> @@ -4469,6 +4471,28 @@ static inline bool f2fs_blkz_is_seq(struct f2fs_sb_info *sbi, int devi,
>
> return test_bit(zno, FDEV(devi).blkz_seq);
> }
> +
> +static inline bool f2fs_blkaddr_in_seqzone(struct f2fs_sb_info *sbi,
> + block_t blkaddr)
> +{
> + int devi = 0;
> +
> + if (f2fs_is_multi_device(sbi)) {
> + devi = f2fs_target_device_index(sbi, blkaddr);
> + if (blkaddr < FDEV(devi).start_blk ||
> + blkaddr > FDEV(devi).end_blk) {
> + f2fs_err(sbi, "Invalid block %x", blkaddr);
> + return false;
> + }
> + blkaddr -= FDEV(devi).start_blk;
> + }
> +
> + if (!bdev_is_zoned(FDEV(devi).bdev) ||
> + !f2fs_blkz_is_seq(sbi, devi, blkaddr))
> + return false;
> +
> + return true;
> +}
> #endif
>
> static inline int f2fs_bdev_index(struct f2fs_sb_info *sbi,
> diff --git a/fs/f2fs/iostat.c b/fs/f2fs/iostat.c
> index f8703038e1d8..a8626e297876 100644
> --- a/fs/f2fs/iostat.c
> +++ b/fs/f2fs/iostat.c
> @@ -237,6 +237,13 @@ static inline void __update_iostat_latency(struct bio_iostat_ctx *iostat_ctx,
> spin_unlock_irqrestore(&sbi->iostat_lat_lock, flags);
> }
>
> +void *iostat_get_bio_private(struct bio *bio)
> +{
> + struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
> +
> + return iostat_ctx->sbi;
> +}
> +
> void iostat_update_and_unbind_ctx(struct bio *bio)
> {
> struct bio_iostat_ctx *iostat_ctx = bio->bi_private;
> diff --git a/fs/f2fs/iostat.h b/fs/f2fs/iostat.h
> index eb99d05cf272..9006c3d41590 100644
> --- a/fs/f2fs/iostat.h
> +++ b/fs/f2fs/iostat.h
> @@ -58,6 +58,7 @@ static inline struct bio_post_read_ctx *get_post_read_ctx(struct bio *bio)
> return iostat_ctx->post_read_ctx;
> }
>
> +extern void *iostat_get_bio_private(struct bio *bio);
> 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);
> @@ -68,6 +69,7 @@ extern void f2fs_destroy_iostat(struct f2fs_sb_info *sbi);
> #else
> static inline void f2fs_update_iostat(struct f2fs_sb_info *sbi, struct inode *inode,
> enum iostat_type type, unsigned long long io_bytes) {}
> +static inline void *iostat_get_bio_private(struct bio *bio) { return bio->bi_private; }
> static inline void iostat_update_and_unbind_ctx(struct bio *bio) {}
> static inline void iostat_alloc_and_bind_ctx(struct f2fs_sb_info *sbi,
> struct bio *bio, struct bio_post_read_ctx *ctx) {}
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 4fd76e867e0a..4a3cf2888faf 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -3140,6 +3140,9 @@ static int __allocate_new_segment(struct f2fs_sb_info *sbi, int type,
> return err;
> stat_inc_seg_type(sbi, curseg);
> locate_dirty_segment(sbi, old_segno);
> +
> + f2fs_blkzoned_submit_merged_write(sbi, type);
> +
> return 0;
> }
>
> @@ -3461,12 +3464,7 @@ static int __get_segment_type(struct f2fs_io_info *fio)
> f2fs_bug_on(fio->sbi, true);
> }
>
> - if (IS_HOT(type))
> - fio->temp = HOT;
> - else if (IS_WARM(type))
> - fio->temp = WARM;
> - else
> - fio->temp = COLD;
> + fio->temp = type_to_temp(type);
> return type;
> }
>
> @@ -4132,6 +4130,27 @@ static int restore_curseg_summaries(struct f2fs_sb_info *sbi)
> return -EINVAL;
> }
>
> +#ifdef CONFIG_BLK_DEV_ZONED
> + if (f2fs_sb_has_blkzoned(sbi)) {
> + for (type = 0; type < NR_PERSISTENT_LOG; type++) {
> + struct curseg_info *curseg = CURSEG_I(sbi, type);
> + enum page_type ptype;
> + enum temp_type temp;
> +
> + if (!(curseg->next_blkoff % sbi->blocks_per_blkz))
> + continue;
> +
> + if (!f2fs_blkaddr_in_seqzone(sbi,
> + START_BLOCK(sbi, curseg->segno)))
> + continue;
> +
> + ptype = PAGE_TYPE(type);
> + temp = type_to_temp(type);
> + down(&sbi->available_open_zones);
> + sbi->write_io[ptype][temp].zone_openned = true;
> + }
> + }
> +#endif
> return 0;
> }
>
> @@ -5451,6 +5470,18 @@ static void destroy_curseg(struct f2fs_sb_info *sbi)
> for (i = 0; i < NR_CURSEG_TYPE; i++) {
> kfree(array[i].sum_blk);
> kfree(array[i].journal);
> +
> +#ifdef CONFIG_BLK_DEV_ZONED
> + if (f2fs_sb_has_blkzoned(sbi)) {
> + enum page_type ptype = PAGE_TYPE(i);
> + enum temp_type temp = type_to_temp(i);
> +
> + if (sbi->write_io[ptype][temp].zone_openned) {
> + up(&sbi->available_open_zones);
> + sbi->write_io[ptype][temp].zone_openned = false;
> + }
> + }
> +#endif
> }
> kfree(array);
> }
> diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
> index e1c0f418aa11..855978ca869f 100644
> --- a/fs/f2fs/segment.h
> +++ b/fs/f2fs/segment.h
> @@ -24,7 +24,8 @@
>
> #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
> #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
> -#define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA))
> +#define PAGE_TYPE(t) (IS_NODESEG(t) ? NODE : DATA)
> +#define SE_PAGETYPE(se) (PAGE_TYPE((se)->type))
>
> static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
> unsigned short seg_type)
> @@ -965,3 +966,12 @@ static inline unsigned int first_zoned_segno(struct f2fs_sb_info *sbi)
> return GET_SEGNO(sbi, FDEV(devi).start_blk);
> return 0;
> }
> +
> +static inline enum temp_type type_to_temp(int type)
> +{
> + if (IS_HOT(type))
> + return HOT;
> + else if (IS_WARM(type))
> + return WARM;
> + return COLD;
> +}
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index fdf358c7f808..954baa6c100d 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -3893,6 +3893,8 @@ static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
> sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
> return -EINVAL;
> }
> +
> + sema_init(&sbi->available_open_zones, sbi->max_open_zones);
> }
>
> zone_sectors = bdev_zone_sectors(bdev);
> --
> 2.40.1
>

Next message: Haitao Huang: "Re: Re: [PATCH v10 12/14] x86/sgx: Turn on per-cgroup EPC reclamation"
Previous message: Haitao Huang: "[PATCH v11 13/14] Docs/x86/sgx: Add description for cgroup support"
Next in thread: Jaegeuk Kim: "Re: [PATCH v3] f2fs: zone: don't block IO if there is remained open zone"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]