From 3843154598a00408f4214a68bd536fdf27b1df10 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Thu, 27 Feb 2014 18:20:00 +0900 Subject: f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.txt | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'Documentation/filesystems') diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt index b8d284975f0f..8eb06b0a7d2b 100644 --- a/Documentation/filesystems/f2fs.txt +++ b/Documentation/filesystems/f2fs.txt @@ -444,9 +444,11 @@ The number of blocks and buckets are determined by, # of blocks in level #n = | `- 4, Otherwise - ,- 2^n, if n < MAX_DIR_HASH_DEPTH / 2, + ,- 2^ (n + dir_level), + | if n < MAX_DIR_HASH_DEPTH / 2, # of buckets in level #n = | - `- 2^((MAX_DIR_HASH_DEPTH / 2) - 1), Otherwise + `- 2^((MAX_DIR_HASH_DEPTH / 2 + dir_level) - 1), + Otherwise When F2FS finds a file name in a directory, at first a hash value of the file name is calculated. Then, F2FS scans the hash table in level #0 to find the -- cgit v1.2.3 From ab9fa662e4867455f44f4de96d29a7f09cf292c6 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Thu, 27 Feb 2014 20:09:05 +0900 Subject: f2fs: add an sysfs entry to control the directory level This patch adds an sysfs entry to control dir_level used by the large directory. The description of this entry is: dir_level This parameter controls the directory level to support large directory. If a directory has a number of files, it can reduce the file lookup latency by increasing this dir_level value. Otherwise, it needs to decrease this value to reduce the space overhead. The default value is 0. Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.txt | 7 +++++++ fs/f2fs/f2fs.h | 3 +++ fs/f2fs/super.c | 7 +++++++ 3 files changed, 17 insertions(+) (limited to 'Documentation/filesystems') diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt index 8eb06b0a7d2b..803784e1e8ef 100644 --- a/Documentation/filesystems/f2fs.txt +++ b/Documentation/filesystems/f2fs.txt @@ -195,6 +195,13 @@ Files in /sys/fs/f2fs/ cleaning operations. The default value is 4096 which covers 8GB block address range. + dir_level This parameter controls the directory level to + support large directory. If a directory has a + number of files, it can reduce the file lookup + latency by increasing this dir_level value. + Otherwise, it needs to decrease this value to + reduce the space overhead. The default value is 0. + ================================================================================ USAGE ================================================================================ diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index a82691674918..6f88191f2a34 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -196,6 +196,8 @@ struct extent_info { #define FADVISE_COLD_BIT 0x01 #define FADVISE_LOST_PINO_BIT 0x02 +#define DEF_DIR_LEVEL 0 + struct f2fs_inode_info { struct inode vfs_inode; /* serve a vfs inode */ unsigned long i_flags; /* keep an inode flags for ioctl */ @@ -447,6 +449,7 @@ struct f2fs_sb_info { unsigned int total_valid_node_count; /* valid node block count */ unsigned int total_valid_inode_count; /* valid inode count */ int active_logs; /* # of active logs */ + int dir_level; /* directory level */ block_t user_block_count; /* # of user blocks */ block_t total_valid_block_count; /* # of valid blocks */ diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 475560e5ee71..1bd915362154 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -184,6 +184,7 @@ F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards); F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy); F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util); F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search); +F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); #define ATTR_LIST(name) (&f2fs_attr_##name.attr) static struct attribute *f2fs_attrs[] = { @@ -196,6 +197,7 @@ static struct attribute *f2fs_attrs[] = { ATTR_LIST(ipu_policy), ATTR_LIST(min_ipu_util), ATTR_LIST(max_victim_search), + ATTR_LIST(dir_level), NULL, }; @@ -359,6 +361,9 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb) if (test_opt(F2FS_SB(sb), INLINE_XATTR)) set_inode_flag(fi, FI_INLINE_XATTR); + /* Will be used by directory only */ + fi->i_dir_level = F2FS_SB(sb)->dir_level; + return &fi->vfs_inode; } @@ -785,6 +790,8 @@ static void init_sb_info(struct f2fs_sb_info *sbi) for (i = 0; i < NR_COUNT_TYPE; i++) atomic_set(&sbi->nr_pages[i], 0); + + sbi->dir_level = DEF_DIR_LEVEL; } /* -- cgit v1.2.3 From cdfc41c134d48c1923066bcfa6630b94588ad6bc Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 19 Mar 2014 13:31:37 +0900 Subject: f2fs: throttle the memory footprint with a sysfs entry This patch introduces ram_thresh, a sysfs entry, which controls the memory footprint used by the free nid list and the nat cache. Previously, the free nid list was controlled by MAX_FREE_NIDS, while the nat cache was managed by NM_WOUT_THRESHOLD. However, this approach cannot be applied dynamically according to the system. So, this patch adds ram_thresh that users can specify the threshold, which is in order of 1 / 1024. For example, if the total ram size is 4GB and the value is set to 10 by default, f2fs tries to control the number of free nids and nat caches not to consume over 10 * (4GB / 1024) = 10MB. Signed-off-by: Jaegeuk Kim --- Documentation/ABI/testing/sysfs-fs-f2fs | 6 ++++++ Documentation/filesystems/f2fs.txt | 4 ++++ fs/f2fs/f2fs.h | 1 + fs/f2fs/node.c | 23 ++++++++++++++++++++--- fs/f2fs/node.h | 11 ++++++++--- fs/f2fs/super.c | 5 +++++ 6 files changed, 44 insertions(+), 6 deletions(-) (limited to 'Documentation/filesystems') diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fs index 32b0809203dd..4ac2c25025ea 100644 --- a/Documentation/ABI/testing/sysfs-fs-f2fs +++ b/Documentation/ABI/testing/sysfs-fs-f2fs @@ -55,3 +55,9 @@ Date: January 2014 Contact: "Jaegeuk Kim" Description: Controls the number of trials to find a victim segment. + +What: /sys/fs/f2fs//ram_thresh +Date: March 2014 +Contact: "Jaegeuk Kim" +Description: + Controls the memory footprint used by f2fs. diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt index 803784e1e8ef..f415b9fc4cc8 100644 --- a/Documentation/filesystems/f2fs.txt +++ b/Documentation/filesystems/f2fs.txt @@ -202,6 +202,10 @@ Files in /sys/fs/f2fs/ Otherwise, it needs to decrease this value to reduce the space overhead. The default value is 0. + ram_thresh This parameter controls the memory footprint used + by free nids and cached nat entries. By default, + 10 is set, which indicates 10 MB / 1 GB RAM. + ================================================================================ USAGE ================================================================================ diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 59fac1a9f0c2..05c652493f04 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -242,6 +242,7 @@ struct f2fs_nm_info { block_t nat_blkaddr; /* base disk address of NAT */ nid_t max_nid; /* maximum possible node ids */ nid_t next_scan_nid; /* the next nid to be scanned */ + unsigned int ram_thresh; /* control the memory footprint */ /* NAT cache management */ struct radix_tree_root nat_root;/* root of the nat entry cache */ diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c index d27e65a1fb0b..fec4967fb8d2 100644 --- a/fs/f2fs/node.c +++ b/fs/f2fs/node.c @@ -26,6 +26,22 @@ static struct kmem_cache *nat_entry_slab; static struct kmem_cache *free_nid_slab; +static inline bool available_free_memory(struct f2fs_nm_info *nm_i, int type) +{ + struct sysinfo val; + unsigned long mem_size = 0; + + si_meminfo(&val); + if (type == FREE_NIDS) + mem_size = nm_i->fcnt * sizeof(struct free_nid); + else if (type == NAT_ENTRIES) + mem_size += nm_i->nat_cnt * sizeof(struct nat_entry); + mem_size >>= 12; + + /* give 50:50 memory for free nids and nat caches respectively */ + return (mem_size < ((val.totalram * nm_i->ram_thresh) >> 11)); +} + static void clear_node_page_dirty(struct page *page) { struct address_space *mapping = page->mapping; @@ -208,7 +224,7 @@ int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) { struct f2fs_nm_info *nm_i = NM_I(sbi); - if (nm_i->nat_cnt <= NM_WOUT_THRESHOLD || nr_shrink <= 0) + if (available_free_memory(nm_i, NAT_ENTRIES) || nr_shrink <= 0) return 0; write_lock(&nm_i->nat_tree_lock); @@ -1288,7 +1304,7 @@ static int add_free_nid(struct f2fs_nm_info *nm_i, nid_t nid, bool build) struct nat_entry *ne; bool allocated = false; - if (nm_i->fcnt > 2 * MAX_FREE_NIDS) + if (!available_free_memory(nm_i, FREE_NIDS)) return -1; /* 0 nid should not be used */ @@ -1473,7 +1489,7 @@ void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) spin_lock(&nm_i->free_nid_list_lock); i = __lookup_free_nid_list(nm_i, nid); f2fs_bug_on(!i || i->state != NID_ALLOC); - if (nm_i->fcnt > 2 * MAX_FREE_NIDS) { + if (!available_free_memory(nm_i, FREE_NIDS)) { __del_from_free_nid_list(nm_i, i); } else { i->state = NID_NEW; @@ -1836,6 +1852,7 @@ static int init_node_manager(struct f2fs_sb_info *sbi) nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks - 3; nm_i->fcnt = 0; nm_i->nat_cnt = 0; + nm_i->ram_thresh = DEF_RAM_THRESHOLD; 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 ee6d28684ee8..c97215432100 100644 --- a/fs/f2fs/node.h +++ b/fs/f2fs/node.h @@ -17,15 +17,15 @@ /* # of pages to perform readahead before building free nids */ #define FREE_NID_PAGES 4 -/* maximum # of free node ids to produce during build_free_nids */ -#define MAX_FREE_NIDS (NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES) - /* maximum readahead size for node during getting data blocks */ #define MAX_RA_NODE 128 /* maximum cached nat entries to manage memory footprint */ #define NM_WOUT_THRESHOLD (64 * NAT_ENTRY_PER_BLOCK) +/* control the memory footprint threshold (10MB per 1GB ram) */ +#define DEF_RAM_THRESHOLD 10 + /* vector size for gang look-up from nat cache that consists of radix tree */ #define NATVEC_SIZE 64 @@ -77,6 +77,11 @@ static inline void node_info_from_raw_nat(struct node_info *ni, ni->version = raw_ne->version; } +enum nid_type { + FREE_NIDS, /* indicates the free nid list */ + NAT_ENTRIES /* indicates the cached nat entry */ +}; + /* * For free nid mangement */ diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index dbe402b1a4b7..34c47b2010bc 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -74,6 +74,7 @@ static match_table_t f2fs_tokens = { enum { GC_THREAD, /* struct f2fs_gc_thread */ SM_INFO, /* struct f2fs_sm_info */ + NM_INFO, /* struct f2fs_nm_info */ F2FS_SBI, /* struct f2fs_sb_info */ }; @@ -92,6 +93,8 @@ static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type) return (unsigned char *)sbi->gc_thread; else if (struct_type == SM_INFO) return (unsigned char *)SM_I(sbi); + else if (struct_type == NM_INFO) + return (unsigned char *)NM_I(sbi); else if (struct_type == F2FS_SBI) return (unsigned char *)sbi; return NULL; @@ -183,6 +186,7 @@ F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments); F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards); F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy); F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util); +F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh); F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search); F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); @@ -198,6 +202,7 @@ static struct attribute *f2fs_attrs[] = { ATTR_LIST(min_ipu_util), ATTR_LIST(max_victim_search), ATTR_LIST(dir_level), + ATTR_LIST(ram_thresh), NULL, }; -- cgit v1.2.3 From 58c410351eba3d24f741c85a0eb9eaf15c94047d Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 19 Mar 2014 14:17:21 +0900 Subject: f2fs: change reclaim rate in percentage It is more reasonable to determine the reclaiming rate of prefree segments according to the volume size, which is set to 5% by default. For example, if the volume is 128GB, the prefree segments are reclaimed when the number reaches to 6.4GB. Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.txt | 8 +++++--- fs/f2fs/segment.c | 3 ++- fs/f2fs/segment.h | 2 +- 3 files changed, 8 insertions(+), 5 deletions(-) (limited to 'Documentation/filesystems') diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt index f415b9fc4cc8..2f6d0218dd22 100644 --- a/Documentation/filesystems/f2fs.txt +++ b/Documentation/filesystems/f2fs.txt @@ -169,9 +169,11 @@ Files in /sys/fs/f2fs/ reclaim_segments This parameter controls the number of prefree segments to be reclaimed. If the number of prefree - segments is larger than this number, f2fs tries to - conduct checkpoint to reclaim the prefree segments - to free segments. By default, 100 segments, 200MB. + segments is larger than the number of segments + in the proportion to the percentage over total + volume size, f2fs tries to conduct checkpoint to + reclaim the prefree segments to free segments. + By default, 5% over total # of segments. max_small_discards This parameter controls the number of discard commands that consist small blocks less than 2MB. diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 6c5a4f0218ca..e7ff23a536a4 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -1758,7 +1758,8 @@ int build_segment_manager(struct f2fs_sb_info *sbi) sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main); sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); - sm_info->rec_prefree_segments = DEF_RECLAIM_PREFREE_SEGMENTS; + sm_info->rec_prefree_segments = sm_info->main_segments * + DEF_RECLAIM_PREFREE_SEGMENTS / 100; sm_info->ipu_policy = F2FS_IPU_DISABLE; sm_info->min_ipu_util = DEF_MIN_IPU_UTIL; diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h index 9fc46ee27bb8..7091204680f4 100644 --- a/fs/f2fs/segment.h +++ b/fs/f2fs/segment.h @@ -14,7 +14,7 @@ #define NULL_SEGNO ((unsigned int)(~0)) #define NULL_SECNO ((unsigned int)(~0)) -#define DEF_RECLAIM_PREFREE_SEGMENTS 100 /* 200MB of prefree segments */ +#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */ /* L: Logical segment # in volume, R: Relative segment # in main area */ #define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno) -- cgit v1.2.3 From 6b4afdd794783fe515b50838aa36591e3feea990 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Wed, 2 Apr 2014 15:34:36 +0900 Subject: f2fs: introduce f2fs_issue_flush to avoid redundant flush issue Some storage devices show relatively high latencies to complete cache_flush commands, even though their normal IO speed is prettry much high. In such the case, it needs to merge cache_flush commands as much as possible to avoid issuing them redundantly. So, this patch introduces a mount option, "-o flush_merge", to mitigate such the overhead. If this option is enabled by user, F2FS merges the cache_flush commands and then issues just one cache_flush on behalf of them. Once the single command is finished, F2FS sends a completion signal to all the pending threads. Note that, this option can be used under a workload consisting of very intensive concurrent fsync calls, while the storage handles cache_flush commands slowly. Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.txt | 4 ++ fs/f2fs/f2fs.h | 16 +++++++ fs/f2fs/file.c | 2 +- fs/f2fs/segment.c | 89 ++++++++++++++++++++++++++++++++++++++ fs/f2fs/super.c | 7 +++ 5 files changed, 117 insertions(+), 1 deletion(-) (limited to 'Documentation/filesystems') diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt index 2f6d0218dd22..25311e113e75 100644 --- a/Documentation/filesystems/f2fs.txt +++ b/Documentation/filesystems/f2fs.txt @@ -122,6 +122,10 @@ disable_ext_identify Disable the extension list configured by mkfs, so f2fs inline_xattr Enable the inline xattrs feature. inline_data Enable the inline data feature: New created small(<~3.4k) files can be written into inode block. +flush_merge Merge concurrent cache_flush commands as much as possible + to eliminate redundant command issues. If the underlying + device handles the cache_flush command relatively slowly, + recommend to enable this option. ================================================================================ DEBUGFS ENTRIES diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 1e3d869b60cd..2ecac8312359 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -40,6 +40,7 @@ #define F2FS_MOUNT_DISABLE_EXT_IDENTIFY 0x00000040 #define F2FS_MOUNT_INLINE_XATTR 0x00000080 #define F2FS_MOUNT_INLINE_DATA 0x00000100 +#define F2FS_MOUNT_FLUSH_MERGE 0x00000200 #define clear_opt(sbi, option) (sbi->mount_opt.opt &= ~F2FS_MOUNT_##option) #define set_opt(sbi, option) (sbi->mount_opt.opt |= F2FS_MOUNT_##option) @@ -316,6 +317,12 @@ enum { NO_CHECK_TYPE }; +struct flush_cmd { + struct flush_cmd *next; + struct completion wait; + int ret; +}; + struct f2fs_sm_info { struct sit_info *sit_info; /* whole segment information */ struct free_segmap_info *free_info; /* free segment information */ @@ -344,6 +351,14 @@ struct f2fs_sm_info { unsigned int ipu_policy; /* in-place-update policy */ unsigned int min_ipu_util; /* in-place-update threshold */ + + /* for flush command control */ + struct task_struct *f2fs_issue_flush; /* flush thread */ + wait_queue_head_t flush_wait_queue; /* waiting queue for wake-up */ + struct flush_cmd *issue_list; /* list for command issue */ + struct flush_cmd *dispatch_list; /* list for command dispatch */ + spinlock_t issue_lock; /* for issue list lock */ + struct flush_cmd *issue_tail; /* list tail of issue list */ }; /* @@ -1160,6 +1175,7 @@ void destroy_node_manager_caches(void); */ void f2fs_balance_fs(struct f2fs_sb_info *); void f2fs_balance_fs_bg(struct f2fs_sb_info *); +int f2fs_issue_flush(struct f2fs_sb_info *); void invalidate_blocks(struct f2fs_sb_info *, block_t); void refresh_sit_entry(struct f2fs_sb_info *, block_t, block_t); void clear_prefree_segments(struct f2fs_sb_info *); diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 6ba26680c468..302d552afea5 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -186,7 +186,7 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) ret = wait_on_node_pages_writeback(sbi, inode->i_ino); if (ret) goto out; - ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); + ret = f2fs_issue_flush(F2FS_SB(inode->i_sb)); } out: trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret); diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index f799c6a34c39..085f548be7a3 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #include @@ -24,6 +25,7 @@ #define __reverse_ffz(x) __reverse_ffs(~(x)) static struct kmem_cache *discard_entry_slab; +static struct kmem_cache *flush_cmd_slab; /* * __reverse_ffs is copied from include/asm-generic/bitops/__ffs.h since @@ -195,6 +197,73 @@ void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi) f2fs_sync_fs(sbi->sb, true); } +static int issue_flush_thread(void *data) +{ + struct f2fs_sb_info *sbi = data; + struct f2fs_sm_info *sm_i = SM_I(sbi); + wait_queue_head_t *q = &sm_i->flush_wait_queue; +repeat: + if (kthread_should_stop()) + return 0; + + spin_lock(&sm_i->issue_lock); + if (sm_i->issue_list) { + sm_i->dispatch_list = sm_i->issue_list; + sm_i->issue_list = sm_i->issue_tail = NULL; + } + spin_unlock(&sm_i->issue_lock); + + if (sm_i->dispatch_list) { + struct bio *bio = bio_alloc(GFP_NOIO, 0); + struct flush_cmd *cmd, *next; + int ret; + + bio->bi_bdev = sbi->sb->s_bdev; + ret = submit_bio_wait(WRITE_FLUSH, bio); + + for (cmd = sm_i->dispatch_list; cmd; cmd = next) { + cmd->ret = ret; + next = cmd->next; + complete(&cmd->wait); + } + sm_i->dispatch_list = NULL; + } + + wait_event_interruptible(*q, kthread_should_stop() || sm_i->issue_list); + goto repeat; +} + +int f2fs_issue_flush(struct f2fs_sb_info *sbi) +{ + struct f2fs_sm_info *sm_i = SM_I(sbi); + struct flush_cmd *cmd; + int ret; + + if (!test_opt(sbi, FLUSH_MERGE)) + return blkdev_issue_flush(sbi->sb->s_bdev, GFP_KERNEL, NULL); + + cmd = f2fs_kmem_cache_alloc(flush_cmd_slab, GFP_ATOMIC); + cmd->next = NULL; + cmd->ret = 0; + init_completion(&cmd->wait); + + spin_lock(&sm_i->issue_lock); + if (sm_i->issue_list) + sm_i->issue_tail->next = cmd; + else + sm_i->issue_list = cmd; + sm_i->issue_tail = cmd; + spin_unlock(&sm_i->issue_lock); + + if (!sm_i->dispatch_list) + wake_up(&sm_i->flush_wait_queue); + + wait_for_completion(&cmd->wait); + ret = cmd->ret; + kmem_cache_free(flush_cmd_slab, cmd); + return ret; +} + static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno, enum dirty_type dirty_type) { @@ -1763,6 +1832,7 @@ int build_segment_manager(struct f2fs_sb_info *sbi) { struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); + dev_t dev = sbi->sb->s_bdev->bd_dev; struct f2fs_sm_info *sm_info; int err; @@ -1790,6 +1860,16 @@ int build_segment_manager(struct f2fs_sb_info *sbi) sm_info->nr_discards = 0; sm_info->max_discards = 0; + if (test_opt(sbi, FLUSH_MERGE)) { + spin_lock_init(&sm_info->issue_lock); + init_waitqueue_head(&sm_info->flush_wait_queue); + + sm_info->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi, + "f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev)); + if (IS_ERR(sm_info->f2fs_issue_flush)) + return PTR_ERR(sm_info->f2fs_issue_flush); + } + err = build_sit_info(sbi); if (err) return err; @@ -1898,6 +1978,8 @@ void destroy_segment_manager(struct f2fs_sb_info *sbi) struct f2fs_sm_info *sm_info = SM_I(sbi); if (!sm_info) return; + if (sm_info->f2fs_issue_flush) + kthread_stop(sm_info->f2fs_issue_flush); destroy_dirty_segmap(sbi); destroy_curseg(sbi); destroy_free_segmap(sbi); @@ -1912,10 +1994,17 @@ int __init create_segment_manager_caches(void) sizeof(struct discard_entry)); if (!discard_entry_slab) return -ENOMEM; + flush_cmd_slab = f2fs_kmem_cache_create("flush_command", + sizeof(struct flush_cmd)); + if (!flush_cmd_slab) { + kmem_cache_destroy(discard_entry_slab); + return -ENOMEM; + } return 0; } void destroy_segment_manager_caches(void) { kmem_cache_destroy(discard_entry_slab); + kmem_cache_destroy(flush_cmd_slab); } diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 959834066d60..d31b767fde73 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -51,6 +51,7 @@ enum { Opt_disable_ext_identify, Opt_inline_xattr, Opt_inline_data, + Opt_flush_merge, Opt_err, }; @@ -67,6 +68,7 @@ static match_table_t f2fs_tokens = { {Opt_disable_ext_identify, "disable_ext_identify"}, {Opt_inline_xattr, "inline_xattr"}, {Opt_inline_data, "inline_data"}, + {Opt_flush_merge, "flush_merge"}, {Opt_err, NULL}, }; @@ -334,6 +336,9 @@ static int parse_options(struct super_block *sb, char *options) case Opt_inline_data: set_opt(sbi, INLINE_DATA); break; + case Opt_flush_merge: + set_opt(sbi, FLUSH_MERGE); + break; default: f2fs_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" or missing value", @@ -537,6 +542,8 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root) seq_puts(seq, ",disable_ext_identify"); if (test_opt(sbi, INLINE_DATA)) seq_puts(seq, ",inline_data"); + if (test_opt(sbi, FLUSH_MERGE)) + seq_puts(seq, ",flush_merge"); seq_printf(seq, ",active_logs=%u", sbi->active_logs); return 0; -- cgit v1.2.3