On 02/04, Chao Yu wrote:
> On 2022/2/4 1:42, Jaegeuk Kim wrote:
> > On 02/03, Chao Yu wrote:
> > > On 2022/2/3 8:34, Jaegeuk Kim wrote:
> > > > This adds a sysfs entry to call checkpoint during fsync() in order to
> > > > avoid
> > > > long elapsed time to run roll-forward recovery when booting the device.
> > > > Default value doesn't enforce the limitation which is same as before.
> > > >
> > > > Signed-off-by: Jaegeuk Kim <[email protected]>
> > > > ---
> > > > v2 from v1:
> > > > - make the default w/o enforcement
> > > >
> > > > Documentation/ABI/testing/sysfs-fs-f2fs | 6 ++++++
> > > > fs/f2fs/checkpoint.c | 1 +
> > > > fs/f2fs/f2fs.h | 3 +++
> > > > fs/f2fs/node.c | 2 ++
> > > > fs/f2fs/node.h | 3 +++
> > > > fs/f2fs/recovery.c | 4 ++++
> > > > fs/f2fs/sysfs.c | 2 ++
> > > > 7 files changed, 21 insertions(+)
> > > >
> > > > diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs
> > > > b/Documentation/ABI/testing/sysfs-fs-f2fs
> > > > index 87d3884c90ea..ce8103f522cb 100644
> > > > --- a/Documentation/ABI/testing/sysfs-fs-f2fs
> > > > +++ b/Documentation/ABI/testing/sysfs-fs-f2fs
> > > > @@ -567,3 +567,9 @@ Contact: "Daeho Jeong" <[email protected]>
> > > > Description: You can set the trial count limit for GC urgent high
> > > > mode with this value.
> > > > If GC thread gets to the limit, the mode will turn back
> > > > to GC normal mode.
> > > > By default, the value is zero, which means there is no
> > > > limit like before.
> > > > +
> > > > +What: /sys/fs/f2fs/<disk>/max_roll_forward_node_blocks
> > > > +Date: January 2022
> > > > +Contact: "Jaegeuk Kim" <[email protected]>
> > > > +Description: Controls max # of node block writes to be used for roll
> > > > forward
> > > > + recovery. This can limit the roll forward recovery time.
> > > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
> > > > index deeda95688f0..57a2d9164bee 100644
> > > > --- a/fs/f2fs/checkpoint.c
> > > > +++ b/fs/f2fs/checkpoint.c
> > > > @@ -1543,6 +1543,7 @@ static int do_checkpoint(struct f2fs_sb_info
> > > > *sbi, struct cp_control *cpc)
> > > > /* update user_block_counts */
> > > > sbi->last_valid_block_count = sbi->total_valid_block_count;
> > > > percpu_counter_set(&sbi->alloc_valid_block_count, 0);
> > > > + percpu_counter_set(&sbi->rf_node_block_count, 0);
> > > > /* Here, we have one bio having CP pack except cp pack 2 page */
> > > > f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO);
> > > > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > > > index 63c90416364b..6ddb98ff0b7c 100644
> > > > --- a/fs/f2fs/f2fs.h
> > > > +++ b/fs/f2fs/f2fs.h
> > > > @@ -913,6 +913,7 @@ struct f2fs_nm_info {
> > > > nid_t max_nid; /* maximum possible node ids */
> > > > nid_t available_nids; /* # of available node ids */
> > > > nid_t next_scan_nid; /* the next nid to be scanned */
> > > > + nid_t max_rf_node_blocks; /* max # of nodes for recovery
> > > > */
> > > > unsigned int ram_thresh; /* control the memory footprint
> > > > */
> > > > unsigned int ra_nid_pages; /* # of nid pages to be
> > > > readaheaded */
> > > > unsigned int dirty_nats_ratio; /* control dirty nats ratio
> > > > threshold */
> > > > @@ -1684,6 +1685,8 @@ struct f2fs_sb_info {
> > > > atomic_t nr_pages[NR_COUNT_TYPE];
> > > > /* # of allocated blocks */
> > > > struct percpu_counter alloc_valid_block_count;
> > > > + /* # of node block writes as roll forward recovery */
> > > > + struct percpu_counter rf_node_block_count;
> > > > /* writeback control */
> > > > atomic_t wb_sync_req[META]; /* count # of WB_SYNC threads */
> > > > diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
> > > > index 93512f8859d5..0d9883457579 100644
> > > > --- a/fs/f2fs/node.c
> > > > +++ b/fs/f2fs/node.c
> > > > @@ -1782,6 +1782,7 @@ int f2fs_fsync_node_pages(struct f2fs_sb_info
> > > > *sbi, struct inode *inode,
> > > > if (!atomic || page == last_page) {
> > > > set_fsync_mark(page, 1);
> > > > +
> > > > percpu_counter_inc(&sbi->rf_node_block_count);
> > >
> > > if (NM_I(sbi)->max_rf_node_blocks)
> > > percpu_counter_inc(&sbi->rf_node_block_count);
> >
> > I think we can just count this and adjust right away once sysfs is changed.
>
> Since this long recovery latency issue is a corner case, I guess we can avoid
> this
> to save cpu time...
I think we can show this in debugfs, as it won't give huge overhead.
>
> BTW, shouldn't we account all warn dnode blocks? as we will traverse all
> blocks there
> in warn node list.
I thought we don't need to track the whole bunch of chains, but would be enough
to record # of fsync calls roughly.
>
> Thanks,
>
> >
> > >
> > > Thanks,
> > >
> > > > if (IS_INODE(page)) {
> > > > if (is_inode_flag_set(inode,
> > > >
> > > > FI_DIRTY_INODE))
> > > > @@ -3218,6 +3219,7 @@ static int init_node_manager(struct f2fs_sb_info
> > > > *sbi)
> > > > nm_i->ram_thresh = DEF_RAM_THRESHOLD;
> > > > nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
> > > > nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
> > > > + nm_i->max_rf_node_blocks = DEF_RF_NODE_BLOCKS;
> > > > INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
> > > > INIT_LIST_HEAD(&nm_i->free_nid_list);
> > > > diff --git a/fs/f2fs/node.h b/fs/f2fs/node.h
> > > > index 18b98cf0465b..4c1d34bfea78 100644
> > > > --- a/fs/f2fs/node.h
> > > > +++ b/fs/f2fs/node.h
> > > > @@ -31,6 +31,9 @@
> > > > /* control total # of nats */
> > > > #define DEF_NAT_CACHE_THRESHOLD 100000
> > > > +/* control total # of node writes used for roll-fowrad recovery */
> > > > +#define DEF_RF_NODE_BLOCKS 0
> > > > +
> > > > /* vector size for gang look-up from nat cache that consists of
> > > > radix tree */
> > > > #define NATVEC_SIZE 64
> > > > #define SETVEC_SIZE 32
> > > > diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
> > > > index 10d152cfa58d..1c8041fd854e 100644
> > > > --- a/fs/f2fs/recovery.c
> > > > +++ b/fs/f2fs/recovery.c
> > > > @@ -53,9 +53,13 @@ extern struct kmem_cache *f2fs_cf_name_slab;
> > > > bool f2fs_space_for_roll_forward(struct f2fs_sb_info *sbi)
> > > > {
> > > > s64 nalloc =
> > > > percpu_counter_sum_positive(&sbi->alloc_valid_block_count);
> > > > + u32 rf_node =
> > > > percpu_counter_sum_positive(&sbi->rf_node_block_count);
> > > > if (sbi->last_valid_block_count + nalloc >
> > > > sbi->user_block_count)
> > > > return false;
> > > > + if (NM_I(sbi)->max_rf_node_blocks &&
> > > > + rf_node >= NM_I(sbi)->max_rf_node_blocks)
> > > > + return false;
> > > > return true;
> > > > }
> > > > diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
> > > > index 281bc0133ee6..47efcf233afd 100644
> > > > --- a/fs/f2fs/sysfs.c
> > > > +++ b/fs/f2fs/sysfs.c
> > > > @@ -732,6 +732,7 @@ F2FS_RW_ATTR(SM_INFO, f2fs_sm_info,
> > > > min_ssr_sections, min_ssr_sections);
> > > > F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
> > > > F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
> > > > F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio,
> > > > dirty_nats_ratio);
> > > > +F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, max_roll_forward_node_blocks,
> > > > max_rf_node_blocks);
> > > > F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search,
> > > > max_victim_search);
> > > > F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, migration_granularity,
> > > > migration_granularity);
> > > > F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
> > > > @@ -855,6 +856,7 @@ static struct attribute *f2fs_attrs[] = {
> > > > ATTR_LIST(ram_thresh),
> > > > ATTR_LIST(ra_nid_pages),
> > > > ATTR_LIST(dirty_nats_ratio),
> > > > + ATTR_LIST(max_roll_forward_node_blocks),
> > > > ATTR_LIST(cp_interval),
> > > > ATTR_LIST(idle_interval),
> > > > ATTR_LIST(discard_idle_interval),
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
