Re: [PATCH 3/3] f2fs: introduce io_list for serialize data/node IOs

From: Jaegeuk Kim
Date: Fri May 12 2017 - 13:49:54 EST

Next message: Mark Rutland: "[PATCHv4] arm64/cpufeature: don't use mutex in bringup path"
Previous message: Andreas Schwab: "Re: Is there an recommended way to refer to bitkeepr commits?"
In reply to: Chao Yu: "Re: [PATCH 3/3] f2fs: introduce io_list for serialize data/node IOs"
Next in thread: Jaegeuk Kim: "Re: [PATCH 1/3] f2fs: split wio_mutex"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 05/12, Chao Yu wrote:
> Hi Jaegeuk,
>
> On 2017/5/12 2:36, Jaegeuk Kim wrote:
> > Hi Chao,
> >
> > On 05/09, Chao Yu wrote:
> >> From: Chao Yu <yuchao0@xxxxxxxxxx>
> >>
> >> Serialize data/node IOs by using fifo list instead of mutex lock,
> >> it will help to enhance concurrency of f2fs, meanwhile keeping LFS
> >> IO semantics.
> >
> > I'm not against to give it a try, but not sure how much we can get a benefit
> > from this approach frankly. Have you got a trouble on any lock contention from
> > the below io_rwsem or mutex?
>
> Yes, because submitting IOs can be blocked in block layer since there may be:
> - limitation of some resources, e.g. request number.
> - IO throttle
> Holding a global mutex lock in the path is not good idea, as it may cause
> potential hungtask or concurrency performance regression.
>
> So I add this patch to relief impacting of global mutex.

Okay, could you send a modified patch? Then, let me evaluate it.

Thanks,

>
> Thanks,
>
> >
> > Thanks,
> >
> >>
> >> Signed-off-by: Chao Yu <yuchao0@xxxxxxxxxx>
> >> ---
> >> fs/f2fs/checkpoint.c | 1 +
> >> fs/f2fs/data.c | 28 ++++++++++++++++++++++++----
> >> fs/f2fs/f2fs.h | 5 ++++-
> >> fs/f2fs/gc.c | 3 ++-
> >> fs/f2fs/segment.c | 20 ++++++++++++++------
> >> fs/f2fs/segment.h | 3 ++-
> >> 6 files changed, 47 insertions(+), 13 deletions(-)
> >>
> >> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
> >> index 2a475e83a092..7b3393474f6b 100644
> >> --- a/fs/f2fs/checkpoint.c
> >> +++ b/fs/f2fs/checkpoint.c
> >> @@ -162,6 +162,7 @@ int ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
> >> .op = REQ_OP_READ,
> >> .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD,
> >> .encrypted_page = NULL,
> >> + .in_list = false,
> >> };
> >> struct blk_plug plug;
> >>
> >> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> >> index 5f001b471252..89eaa8aaa97b 100644
> >> --- a/fs/f2fs/data.c
> >> +++ b/fs/f2fs/data.c
> >> @@ -393,10 +393,28 @@ int f2fs_submit_page_mbio(struct f2fs_io_info *fio)
> >> struct f2fs_bio_info *io;
> >> bool is_read = is_read_io(fio->op);
> >> struct page *bio_page;
> >> + struct curseg_info *curseg;
> >> int err = 0;
> >>
> >> + if (fio->in_list)
> >> + curseg = CURSEG_I(sbi, fio->seg_type);
> >> +
> >> io = __get_bio_info(sbi, fio->op, fio->type, fio->seg_type);
> >>
> >> + down_write(&io->io_rwsem);
> >> +next:
> >> + if (fio->in_list) {
> >> + spin_lock(&curseg->io_lock);
> >> + if (list_empty(&curseg->io_list)) {
> >> + spin_unlock(&curseg->io_lock);
> >> + goto out_fail;
> >> + }
> >> + fio = list_first_entry(&curseg->io_list,
> >> + struct f2fs_io_info, list);
> >> + list_del(&fio->list);
> >> + spin_unlock(&curseg->io_lock);
> >> + }
> >> +
> >> if (fio->old_blkaddr != NEW_ADDR)
> >> verify_block_addr(sbi, fio->old_blkaddr);
> >> verify_block_addr(sbi, fio->new_blkaddr);
> >> @@ -409,8 +427,6 @@ int f2fs_submit_page_mbio(struct f2fs_io_info *fio)
> >> if (!is_read)
> >> inc_page_count(sbi, WB_DATA_TYPE(bio_page));
> >>
> >> - down_write(&io->io_rwsem);
> >> -
> >> if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 ||
> >> (io->fio.op != fio->op || io->fio.op_flags != fio->op_flags) ||
> >> !__same_bdev(sbi, fio->new_blkaddr, io->bio)))
> >> @@ -437,9 +453,13 @@ int f2fs_submit_page_mbio(struct f2fs_io_info *fio)
> >>
> >> io->last_block_in_bio = fio->new_blkaddr;
> >> f2fs_trace_ios(fio, 0);
> >> +
> >> + trace_f2fs_submit_page_mbio(fio->page, fio);
> >> +
> >> + if (fio->in_list)
> >> + goto next;
> >> out_fail:
> >> up_write(&io->io_rwsem);
> >> - trace_f2fs_submit_page_mbio(fio->page, fio);
> >> return err;
> >> }
> >>
> >> @@ -752,7 +772,7 @@ static int __allocate_data_block(struct dnode_of_data *dn)
> >> set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
> >>
> >> allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr,
> >> - &sum, CURSEG_WARM_DATA);
> >> + &sum, CURSEG_WARM_DATA, NULL, false);
> >> set_data_blkaddr(dn);
> >>
> >> /* update i_size */
> >> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> >> index 9129a6229bc8..6b8e9f051aa2 100644
> >> --- a/fs/f2fs/f2fs.h
> >> +++ b/fs/f2fs/f2fs.h
> >> @@ -802,8 +802,10 @@ struct f2fs_io_info {
> >> block_t old_blkaddr; /* old block address before Cow */
> >> struct page *page; /* page to be written */
> >> struct page *encrypted_page; /* encrypted page */
> >> + struct list_head list; /* serialize IOs */
> >> bool submitted; /* indicate IO submission */
> >> bool need_lock; /* indicate we need to lock cp_rwsem */
> >> + bool in_list; /* indicate fio is in io_list */
> >> };
> >>
> >> #define is_read_io(rw) ((rw) == READ)
> >> @@ -2274,7 +2276,8 @@ void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
> >> bool recover_newaddr);
> >> void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
> >> block_t old_blkaddr, block_t *new_blkaddr,
> >> - struct f2fs_summary *sum, int type);
> >> + struct f2fs_summary *sum, int type,
> >> + struct f2fs_io_info *fio, bool add_list);
> >> void f2fs_wait_on_page_writeback(struct page *page,
> >> enum page_type type, bool ordered);
> >> void f2fs_wait_on_encrypted_page_writeback(struct f2fs_sb_info *sbi,
> >> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> >> index 8b267ca30926..ac2f74e40eea 100644
> >> --- a/fs/f2fs/gc.c
> >> +++ b/fs/f2fs/gc.c
> >> @@ -590,6 +590,7 @@ static void move_encrypted_block(struct inode *inode, block_t bidx,
> >> .op = REQ_OP_READ,
> >> .op_flags = 0,
> >> .encrypted_page = NULL,
> >> + .in_list = false,
> >> };
> >> struct dnode_of_data dn;
> >> struct f2fs_summary sum;
> >> @@ -633,7 +634,7 @@ static void move_encrypted_block(struct inode *inode, block_t bidx,
> >> fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
> >>
> >> allocate_data_block(fio.sbi, NULL, fio.old_blkaddr, &newaddr,
> >> - &sum, fio.seg_type);
> >> + &sum, fio.seg_type, NULL, false);
> >>
> >> fio.encrypted_page = pagecache_get_page(META_MAPPING(fio.sbi), newaddr,
> >> FGP_LOCK | FGP_CREAT, GFP_NOFS);
> >> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> >> index c047b5d8b9d3..d4975b8f4620 100644
> >> --- a/fs/f2fs/segment.c
> >> +++ b/fs/f2fs/segment.c
> >> @@ -2100,7 +2100,8 @@ static int __get_segment_type(struct page *page, enum page_type p_type)
> >>
> >> void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
> >> block_t old_blkaddr, block_t *new_blkaddr,
> >> - struct f2fs_summary *sum, int type)
> >> + struct f2fs_summary *sum, int type,
> >> + struct f2fs_io_info *fio, bool add_list)
> >> {
> >> struct sit_info *sit_i = SIT_I(sbi);
> >> struct curseg_info *curseg = CURSEG_I(sbi, type);
> >> @@ -2136,6 +2137,14 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
> >> if (page && IS_NODESEG(type))
> >> fill_node_footer_blkaddr(page, NEXT_FREE_BLKADDR(sbi, curseg));
> >>
> >> + if (add_list) {
> >> + INIT_LIST_HEAD(&fio->list);
> >> + fio->in_list = true;
> >> + spin_lock(&curseg->io_lock);
> >> + list_add_tail(&fio->list, &curseg->io_list);
> >> + spin_unlock(&curseg->io_lock);
> >> + }
> >> +
> >> mutex_unlock(&curseg->curseg_mutex);
> >> }
> >>
> >> @@ -2145,10 +2154,9 @@ static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
> >>
> >> fio->seg_type = __get_segment_type(fio->page, fio->type);
> >>
> >> - mutex_lock(&CURSEG_I(fio->sbi, fio->seg_type)->wio_mutex);
> >> reallocate:
> >> allocate_data_block(fio->sbi, fio->page, fio->old_blkaddr,
> >> - &fio->new_blkaddr, sum, fio->seg_type);
> >> + &fio->new_blkaddr, sum, fio->seg_type, fio, true);
> >>
> >> /* writeout dirty page into bdev */
> >> err = f2fs_submit_page_mbio(fio);
> >> @@ -2156,8 +2164,6 @@ static void do_write_page(struct f2fs_summary *sum, struct f2fs_io_info *fio)
> >> fio->old_blkaddr = fio->new_blkaddr;
> >> goto reallocate;
> >> }
> >> -
> >> - mutex_unlock(&CURSEG_I(fio->sbi, fio->seg_type)->wio_mutex);
> >> }
> >>
> >> void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
> >> @@ -2171,6 +2177,7 @@ void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
> >> .new_blkaddr = page->index,
> >> .page = page,
> >> .encrypted_page = NULL,
> >> + .in_list = false,
> >> };
> >>
> >> if (unlikely(page->index >= MAIN_BLKADDR(sbi)))
> >> @@ -2978,7 +2985,8 @@ static int build_curseg(struct f2fs_sb_info *sbi)
> >> array[i].bio_info.sbi = sbi;
> >> array[i].bio_info.bio = NULL;
> >>
> >> - mutex_init(&array[i].wio_mutex);
> >> + spin_lock_init(&array[i].io_lock);
> >> + INIT_LIST_HEAD(&array[i].io_list);
> >> }
> >> return restore_curseg_summaries(sbi);
> >> }
> >> diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
> >> index 701944b462cd..b6f5dffeaa61 100644
> >> --- a/fs/f2fs/segment.h
> >> +++ b/fs/f2fs/segment.h
> >> @@ -283,7 +283,8 @@ struct curseg_info {
> >> struct rw_semaphore journal_rwsem; /* protect journal area */
> >> struct f2fs_journal *journal; /* cached journal info */
> >> struct f2fs_bio_info bio_info; /* for log bios */
> >> - struct mutex wio_mutex; /* serialize DATA/NODE IOs */
> >> + spinlock_t io_lock; /* serialize DATA/NODE IOs */
> >> + struct list_head io_list; /* tracking fios */
> >> unsigned char alloc_type; /* current allocation type */
> >> unsigned int segno; /* current segment number */
> >> unsigned short next_blkoff; /* next block offset to write */
> >> --
> >> 2.12.2.575.gb14f27f
> >
> > .
> >

Next message: Mark Rutland: "[PATCHv4] arm64/cpufeature: don't use mutex in bringup path"
Previous message: Andreas Schwab: "Re: Is there an recommended way to refer to bitkeepr commits?"
In reply to: Chao Yu: "Re: [PATCH 3/3] f2fs: introduce io_list for serialize data/node IOs"
Next in thread: Jaegeuk Kim: "Re: [PATCH 1/3] f2fs: split wio_mutex"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]