From 1dcc336b02bff3d38f173feac55a2b6c25a5fb54 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 5 Feb 2015 17:57:31 +0800 Subject: f2fs: enable rb-tree extent cache This patch enables rb-tree based extent cache in f2fs. When we mount with "-o extent_cache", f2fs will try to add recently accessed page-block mappings into rb-tree based extent cache as much as possible, instead of original one extent info cache. By this way, f2fs can support more effective cache between dnode page cache and disk. It will supply high hit ratio in the cache with fewer memory when dnode page cache are reclaimed in environment of low memory. Storage: Sandisk sd card 64g 1.append write file (offset: 0, size: 128M); 2.override write file (offset: 2M, size: 1M); 3.override write file (offset: 4M, size: 1M); ... 4.override write file (offset: 48M, size: 1M); ... 5.override write file (offset: 112M, size: 1M); 6.sync 7.echo 3 > /proc/sys/vm/drop_caches 8.read file (size:128M, unit: 4k, count: 32768) (time dd if=/mnt/f2fs/128m bs=4k count=32768) Extent Hit Ratio: before patched Hit Ratio 121 / 1071 1071 / 1071 Performance: before patched real 0m37.051s 0m35.556s user 0m0.040s 0m0.026s sys 0m2.990s 0m2.251s Memory Cost: before patched Tree Count: 0 1 (size: 24 bytes) Node Count: 0 45 (size: 1440 bytes) v3: o retest and given more details of test result. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/segment.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs/f2fs/segment.c') diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index daee4ab913da..29c04edcb843 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -277,6 +277,9 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi) void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi) { + /* try to shrink extent cache when there is no enough memory */ + f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER); + /* check the # of cached NAT entries and prefree segments */ if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK) || excess_prefree_segs(sbi) || -- cgit From 2b11a74b21b040270d19fc3b61265c75413c0dc1 Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Fri, 27 Feb 2015 16:52:50 +0800 Subject: f2fs: don't need to collect dirty sit entries and flush journal when there's no dirty sit entries Don't need to collect dirty sit entries and flush sit journal to sit entries when there's no dirty sit entries. This patch check dirty_sentries earlier just like flush_nat_entries. Signed-off-by: Wanpeng Li Reviewed-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/segment.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/f2fs/segment.c') diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 29c04edcb843..1e909340842a 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -1733,6 +1733,9 @@ void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc) mutex_lock(&curseg->curseg_mutex); mutex_lock(&sit_i->sentry_lock); + if (!sit_i->dirty_sentries) + goto out; + /* * add and account sit entries of dirty bitmap in sit entry * set temporarily @@ -1747,9 +1750,6 @@ void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc) if (!__has_cursum_space(sum, sit_i->dirty_sentries, SIT_JOURNAL)) remove_sits_in_journal(sbi); - if (!sit_i->dirty_sentries) - goto out; - /* * there are two steps to flush sit entries: * #1, flush sit entries to journal in current cold data summary block. -- cgit From b28c3f94933dae663eeca24637bf9ff962bf02dc Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Sat, 28 Feb 2015 17:23:30 +0800 Subject: f2fs: fix to issue small discard in real-time mode discard Now in f2fs, we share functions and structures for batch mode and real-time mode discard. For real-time mode discard, in shared function add_discard_addrs, we will use uninitialized trim_minlen in struct cp_control to compare with length of contiguous free blocks to decide whether skipping discard fragmented freespace or not, this makes us ignore small discard sometimes. Fix it. Signed-off-by: Chao Yu Reviewed-by : Changman Lee Signed-off-by: Jaegeuk Kim --- fs/f2fs/segment.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/f2fs/segment.c') diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 1e909340842a..3e80bd69d4a0 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -552,7 +552,7 @@ static void add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc) end = __find_rev_next_zero_bit(dmap, max_blocks, start + 1); - if (end - start < cpc->trim_minlen) + if (force && end - start < cpc->trim_minlen) continue; __add_discard_entry(sbi, cpc, start, end); -- cgit From 21cb1d99bcc77252e6426010bcc6433f75b581bb Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 11 Mar 2015 13:42:48 -0400 Subject: f2fs: fix to cover sentry_lock for block allocation In the following call stack, f2fs changes the bitmap for dirty segments and # of dirty sentries without grabbing sit_i->sentry_lock. This can result in mismatch on bitmap and # of dirty sentries, since if there are some direct_io operations. In allocate_data_block, - __allocate_new_segments - mutex_lock(&curseg->curseg_mutex); - s_ops->allocate_segment - new_curseg/change_curseg - reset_curseg - __set_sit_entry_type - __mark_sit_entry_dirty - set_bit(dirty_sentries_bitmap) - dirty_sentries++; Signed-off-by: Jaegeuk Kim --- fs/f2fs/segment.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/f2fs/segment.c') diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 3e80bd69d4a0..eafaf728236b 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -1167,6 +1167,7 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page, curseg = CURSEG_I(sbi, type); mutex_lock(&curseg->curseg_mutex); + mutex_lock(&sit_i->sentry_lock); /* direct_io'ed data is aligned to the segment for better performance */ if (direct_io && curseg->next_blkoff) @@ -1181,7 +1182,6 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page, */ __add_sum_entry(sbi, type, sum); - mutex_lock(&sit_i->sentry_lock); __refresh_next_blkoff(sbi, curseg); stat_inc_block_count(sbi, curseg); -- cgit From 8ce67cb07dbf6ba35aea1e07e8ad1ea004ced27b Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Tue, 17 Mar 2015 17:58:08 -0700 Subject: f2fs: add some tracepoints to debug volatile and atomic writes Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 2 ++ fs/f2fs/f2fs.h | 4 ++++ fs/f2fs/segment.c | 4 ++++ include/trace/events/f2fs.h | 27 ++++++++++++++++++++++++++- 4 files changed, 36 insertions(+), 1 deletion(-) (limited to 'fs/f2fs/segment.c') diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 486113db97b1..241b98741df9 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -1287,9 +1287,11 @@ int do_write_data_page(struct page *page, struct f2fs_io_info *fio) need_inplace_update(inode))) { rewrite_data_page(page, fio); set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE); + trace_f2fs_do_write_data_page(page, IPU); } else { write_data_page(page, &dn, fio); f2fs_update_extent_cache(&dn); + trace_f2fs_do_write_data_page(page, OPU); set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); if (page->index == 0) set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 9e378179ff93..c9818193e15e 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -556,6 +556,10 @@ enum page_type { META, NR_PAGE_TYPE, META_FLUSH, + INMEM, /* the below types are used by tracepoints only. */ + INMEM_DROP, + IPU, + OPU, }; struct f2fs_io_info { diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index eafaf728236b..f939660941bb 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -205,6 +205,8 @@ retry: list_add_tail(&new->list, &fi->inmem_pages); inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES); mutex_unlock(&fi->inmem_lock); + + trace_f2fs_register_inmem_page(page, INMEM); } void commit_inmem_pages(struct inode *inode, bool abort) @@ -238,11 +240,13 @@ void commit_inmem_pages(struct inode *inode, bool abort) f2fs_wait_on_page_writeback(cur->page, DATA); if (clear_page_dirty_for_io(cur->page)) inode_dec_dirty_pages(inode); + trace_f2fs_commit_inmem_page(cur->page, INMEM); do_write_data_page(cur->page, &fio); submit_bio = true; } f2fs_put_page(cur->page, 1); } else { + trace_f2fs_commit_inmem_page(cur->page, INMEM_DROP); put_page(cur->page); } radix_tree_delete(&fi->inmem_root, cur->page->index); diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index 09a56dc75f8d..75724bd28356 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -14,7 +14,11 @@ { NODE, "NODE" }, \ { DATA, "DATA" }, \ { META, "META" }, \ - { META_FLUSH, "META_FLUSH" }) + { META_FLUSH, "META_FLUSH" }, \ + { INMEM, "INMEM" }, \ + { INMEM_DROP, "INMEM_DROP" }, \ + { IPU, "IN-PLACE" }, \ + { OPU, "OUT-OF-PLACE" }) #define F2FS_BIO_MASK(t) (t & (READA | WRITE_FLUSH_FUA)) #define F2FS_BIO_EXTRA_MASK(t) (t & (REQ_META | REQ_PRIO)) @@ -854,6 +858,13 @@ DEFINE_EVENT(f2fs__page, f2fs_writepage, TP_ARGS(page, type) ); +DEFINE_EVENT(f2fs__page, f2fs_do_write_data_page, + + TP_PROTO(struct page *page, int type), + + TP_ARGS(page, type) +); + DEFINE_EVENT(f2fs__page, f2fs_readpage, TP_PROTO(struct page *page, int type), @@ -875,6 +886,20 @@ DEFINE_EVENT(f2fs__page, f2fs_vm_page_mkwrite, TP_ARGS(page, type) ); +DEFINE_EVENT(f2fs__page, f2fs_register_inmem_page, + + TP_PROTO(struct page *page, int type), + + TP_ARGS(page, type) +); + +DEFINE_EVENT(f2fs__page, f2fs_commit_inmem_page, + + TP_PROTO(struct page *page, int type), + + TP_ARGS(page, type) +); + TRACE_EVENT(f2fs_writepages, TP_PROTO(struct inode *inode, struct writeback_control *wbc, int type), -- cgit