From 94afd6d6e5253179c9b891d02081cc8355a11768 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Wed, 4 Aug 2021 10:23:48 +0800 Subject: f2fs: extent cache: support unaligned extent Compressed inode may suffer read performance issue due to it can not use extent cache, so I propose to add this unaligned extent support to improve it. Currently, it only works in readonly format f2fs image. Unaligned extent: in one compressed cluster, physical block number will be less than logical block number, so we add an extra physical block length in extent info in order to indicate such extent status. The idea is if one whole cluster blocks are contiguous physically, once its mapping info was readed at first time, we will cache an unaligned (or aligned) extent info entry in extent cache, it expects that the mapping info will be hitted when rereading cluster. Merge policy: - Aligned extents can be merged. - Aligned extent and unaligned extent can not be merged. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/extent_cache.c | 41 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 41 insertions(+) (limited to 'fs/f2fs/extent_cache.c') diff --git a/fs/f2fs/extent_cache.c b/fs/f2fs/extent_cache.c index 3ebf976a682d..b120589d8517 100644 --- a/fs/f2fs/extent_cache.c +++ b/fs/f2fs/extent_cache.c @@ -661,6 +661,47 @@ static void f2fs_update_extent_tree_range(struct inode *inode, f2fs_mark_inode_dirty_sync(inode, true); } +#ifdef CONFIG_F2FS_FS_COMPRESSION +void f2fs_update_extent_tree_range_compressed(struct inode *inode, + pgoff_t fofs, block_t blkaddr, unsigned int llen, + unsigned int c_len) +{ + struct f2fs_sb_info *sbi = F2FS_I_SB(inode); + struct extent_tree *et = F2FS_I(inode)->extent_tree; + struct extent_node *en = NULL; + struct extent_node *prev_en = NULL, *next_en = NULL; + struct extent_info ei; + struct rb_node **insert_p = NULL, *insert_parent = NULL; + bool leftmost = false; + + trace_f2fs_update_extent_tree_range(inode, fofs, blkaddr, llen); + + /* it is safe here to check FI_NO_EXTENT w/o et->lock in ro image */ + if (is_inode_flag_set(inode, FI_NO_EXTENT)) + return; + + write_lock(&et->lock); + + en = (struct extent_node *)f2fs_lookup_rb_tree_ret(&et->root, + (struct rb_entry *)et->cached_en, fofs, + (struct rb_entry **)&prev_en, + (struct rb_entry **)&next_en, + &insert_p, &insert_parent, false, + &leftmost); + if (en) + goto unlock_out; + + set_extent_info(&ei, fofs, blkaddr, llen); + ei.c_len = c_len; + + if (!__try_merge_extent_node(sbi, et, &ei, prev_en, next_en)) + __insert_extent_tree(sbi, et, &ei, + insert_p, insert_parent, leftmost); +unlock_out: + write_unlock(&et->lock); +} +#endif + unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) { struct extent_tree *et, *next; -- cgit v1.2.3 From 324105775c1982aacbd2972b7024d8cc06474abe Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Mon, 9 Aug 2021 08:24:48 +0800 Subject: f2fs: support fault injection for f2fs_kmem_cache_alloc() This patch supports to inject fault into f2fs_kmem_cache_alloc(). Usage: a) echo 32768 > /sys/fs/f2fs//inject_type or b) mount -o fault_type=32768 Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- Documentation/filesystems/f2fs.rst | 1 + fs/f2fs/checkpoint.c | 3 ++- fs/f2fs/compress.c | 8 +++++--- fs/f2fs/data.c | 2 +- fs/f2fs/dir.c | 4 ++-- fs/f2fs/extent_cache.c | 5 +++-- fs/f2fs/f2fs.h | 17 ++++++++++++++++- fs/f2fs/gc.c | 6 ++++-- fs/f2fs/node.c | 23 ++++++++++++----------- fs/f2fs/recovery.c | 3 ++- fs/f2fs/segment.c | 10 ++++++---- fs/f2fs/super.c | 4 +++- fs/f2fs/xattr.c | 3 ++- 13 files changed, 59 insertions(+), 30 deletions(-) (limited to 'fs/f2fs/extent_cache.c') diff --git a/Documentation/filesystems/f2fs.rst b/Documentation/filesystems/f2fs.rst index 9b0517d90063..21d40e3cfd7a 100644 --- a/Documentation/filesystems/f2fs.rst +++ b/Documentation/filesystems/f2fs.rst @@ -195,6 +195,7 @@ fault_type=%d Support configuring fault injection type, should be FAULT_CHECKPOINT 0x000001000 FAULT_DISCARD 0x000002000 FAULT_WRITE_IO 0x000004000 + FAULT_SLAB_ALLOC 0x000008000 =================== =========== mode=%s Control block allocation mode which supports "adaptive" and "lfs". In "lfs" mode, there should be no random diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 5b6ddeae1107..41960c55c343 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -475,7 +475,8 @@ static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, retry: if (!e) - new = f2fs_kmem_cache_alloc(ino_entry_slab, GFP_NOFS); + new = f2fs_kmem_cache_alloc(ino_entry_slab, + GFP_NOFS, true, NULL); radix_tree_preload(GFP_NOFS | __GFP_NOFAIL); diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c index 7dbfd6965b97..afb79480c6a3 100644 --- a/fs/f2fs/compress.c +++ b/fs/f2fs/compress.c @@ -28,7 +28,8 @@ static void *page_array_alloc(struct inode *inode, int nr) unsigned int size = sizeof(struct page *) * nr; if (likely(size <= sbi->page_array_slab_size)) - return kmem_cache_zalloc(sbi->page_array_slab, GFP_NOFS); + return f2fs_kmem_cache_alloc(sbi->page_array_slab, + GFP_F2FS_ZERO, false, F2FS_I_SB(inode)); return f2fs_kzalloc(sbi, size, GFP_NOFS); } @@ -1228,7 +1229,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc, fio.version = ni.version; - cic = kmem_cache_zalloc(cic_entry_slab, GFP_NOFS); + cic = f2fs_kmem_cache_alloc(cic_entry_slab, GFP_F2FS_ZERO, false, sbi); if (!cic) goto out_put_dnode; @@ -1506,7 +1507,8 @@ struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc) pgoff_t start_idx = start_idx_of_cluster(cc); int i; - dic = kmem_cache_zalloc(dic_entry_slab, GFP_NOFS); + dic = f2fs_kmem_cache_alloc(dic_entry_slab, GFP_F2FS_ZERO, + false, F2FS_I_SB(cc->inode)); if (!dic) return ERR_PTR(-ENOMEM); diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index cec084806725..12cd63603137 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -724,7 +724,7 @@ static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio, struct f2fs_bio_info *io = sbi->write_io[DATA] + temp; struct bio_entry *be; - be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS); + be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS, true, NULL); be->bio = bio; bio_get(bio); diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c index c250bf46ef5e..1820e9c106f7 100644 --- a/fs/f2fs/dir.c +++ b/fs/f2fs/dir.c @@ -83,8 +83,8 @@ int f2fs_init_casefolded_name(const struct inode *dir, struct super_block *sb = dir->i_sb; if (IS_CASEFOLDED(dir)) { - fname->cf_name.name = kmem_cache_alloc(f2fs_cf_name_slab, - GFP_NOFS); + fname->cf_name.name = f2fs_kmem_cache_alloc(f2fs_cf_name_slab, + GFP_NOFS, false, F2FS_SB(sb)); if (!fname->cf_name.name) return -ENOMEM; fname->cf_name.len = utf8_casefold(sb->s_encoding, diff --git a/fs/f2fs/extent_cache.c b/fs/f2fs/extent_cache.c index b120589d8517..866e72b29bd5 100644 --- a/fs/f2fs/extent_cache.c +++ b/fs/f2fs/extent_cache.c @@ -239,7 +239,7 @@ static struct extent_node *__attach_extent_node(struct f2fs_sb_info *sbi, { struct extent_node *en; - en = kmem_cache_alloc(extent_node_slab, GFP_ATOMIC); + en = f2fs_kmem_cache_alloc(extent_node_slab, GFP_ATOMIC, false, sbi); if (!en) return NULL; @@ -292,7 +292,8 @@ static struct extent_tree *__grab_extent_tree(struct inode *inode) mutex_lock(&sbi->extent_tree_lock); et = radix_tree_lookup(&sbi->extent_tree_root, ino); if (!et) { - et = f2fs_kmem_cache_alloc(extent_tree_slab, GFP_NOFS); + et = f2fs_kmem_cache_alloc(extent_tree_slab, + GFP_NOFS, true, NULL); f2fs_radix_tree_insert(&sbi->extent_tree_root, ino, et); memset(et, 0, sizeof(struct extent_tree)); et->ino = ino; diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index e97b4d8c5efc..13a7cfe9b23f 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -53,6 +53,7 @@ enum { FAULT_CHECKPOINT, FAULT_DISCARD, FAULT_WRITE_IO, + FAULT_SLAB_ALLOC, FAULT_MAX, }; @@ -2618,7 +2619,7 @@ static inline struct kmem_cache *f2fs_kmem_cache_create(const char *name, return kmem_cache_create(name, size, 0, SLAB_RECLAIM_ACCOUNT, NULL); } -static inline void *f2fs_kmem_cache_alloc(struct kmem_cache *cachep, +static inline void *f2fs_kmem_cache_alloc_nofail(struct kmem_cache *cachep, gfp_t flags) { void *entry; @@ -2629,6 +2630,20 @@ static inline void *f2fs_kmem_cache_alloc(struct kmem_cache *cachep, return entry; } +static inline void *f2fs_kmem_cache_alloc(struct kmem_cache *cachep, + gfp_t flags, bool nofail, struct f2fs_sb_info *sbi) +{ + if (nofail) + return f2fs_kmem_cache_alloc_nofail(cachep, flags); + + if (time_to_inject(sbi, FAULT_SLAB_ALLOC)) { + f2fs_show_injection_info(sbi, FAULT_SLAB_ALLOC); + return NULL; + } + + return kmem_cache_alloc(cachep, flags); +} + static inline bool is_inflight_io(struct f2fs_sb_info *sbi, int type) { if (get_pages(sbi, F2FS_RD_DATA) || get_pages(sbi, F2FS_RD_NODE) || diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index 9dce44619069..3bc0f0162e31 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -371,7 +371,8 @@ static struct victim_entry *attach_victim_entry(struct f2fs_sb_info *sbi, struct atgc_management *am = &sbi->am; struct victim_entry *ve; - ve = f2fs_kmem_cache_alloc(victim_entry_slab, GFP_NOFS); + ve = f2fs_kmem_cache_alloc(victim_entry_slab, + GFP_NOFS, true, NULL); ve->mtime = mtime; ve->segno = segno; @@ -849,7 +850,8 @@ static void add_gc_inode(struct gc_inode_list *gc_list, struct inode *inode) iput(inode); return; } - new_ie = f2fs_kmem_cache_alloc(f2fs_inode_entry_slab, GFP_NOFS); + new_ie = f2fs_kmem_cache_alloc(f2fs_inode_entry_slab, + GFP_NOFS, true, NULL); new_ie->inode = inode; f2fs_radix_tree_insert(&gc_list->iroot, inode->i_ino, new_ie); diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c index 9d838a7929fb..161173de5a2d 100644 --- a/fs/f2fs/node.c +++ b/fs/f2fs/node.c @@ -162,14 +162,13 @@ static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) return dst_page; } -static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail) +static struct nat_entry *__alloc_nat_entry(struct f2fs_sb_info *sbi, + nid_t nid, bool no_fail) { struct nat_entry *new; - if (no_fail) - new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO); - else - new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO); + new = f2fs_kmem_cache_alloc(nat_entry_slab, + GFP_F2FS_ZERO, no_fail, sbi); if (new) { nat_set_nid(new, nid); nat_reset_flag(new); @@ -242,7 +241,8 @@ static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i, head = radix_tree_lookup(&nm_i->nat_set_root, set); if (!head) { - head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); + head = f2fs_kmem_cache_alloc(nat_entry_set_slab, + GFP_NOFS, true, NULL); INIT_LIST_HEAD(&head->entry_list); INIT_LIST_HEAD(&head->set_list); @@ -329,7 +329,8 @@ static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi, unsigned long flags; unsigned int seq_id; - fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, GFP_NOFS); + fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, + GFP_NOFS, true, NULL); get_page(page); fn->page = page; @@ -428,7 +429,7 @@ static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid, struct f2fs_nm_info *nm_i = NM_I(sbi); struct nat_entry *new, *e; - new = __alloc_nat_entry(nid, false); + new = __alloc_nat_entry(sbi, nid, false); if (!new) return; @@ -451,7 +452,7 @@ static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, { struct f2fs_nm_info *nm_i = NM_I(sbi); struct nat_entry *e; - struct nat_entry *new = __alloc_nat_entry(ni->nid, true); + struct nat_entry *new = __alloc_nat_entry(sbi, ni->nid, true); down_write(&nm_i->nat_tree_lock); e = __lookup_nat_cache(nm_i, ni->nid); @@ -2252,7 +2253,7 @@ static bool add_free_nid(struct f2fs_sb_info *sbi, if (unlikely(f2fs_check_nid_range(sbi, nid))) return false; - i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); + i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS, true, NULL); i->nid = nid; i->state = FREE_NID; @@ -2842,7 +2843,7 @@ static void remove_nats_in_journal(struct f2fs_sb_info *sbi) ne = __lookup_nat_cache(nm_i, nid); if (!ne) { - ne = __alloc_nat_entry(nid, true); + ne = __alloc_nat_entry(sbi, nid, true); __init_nat_entry(nm_i, ne, &raw_ne, true); } diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c index 695eacfe776c..04655511d7f5 100644 --- a/fs/f2fs/recovery.c +++ b/fs/f2fs/recovery.c @@ -91,7 +91,8 @@ static struct fsync_inode_entry *add_fsync_inode(struct f2fs_sb_info *sbi, goto err_out; } - entry = f2fs_kmem_cache_alloc(fsync_entry_slab, GFP_F2FS_ZERO); + entry = f2fs_kmem_cache_alloc(fsync_entry_slab, + GFP_F2FS_ZERO, true, NULL); entry->inode = inode; list_add_tail(&entry->list, head); diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index ca9876a6d396..b4dd22134a73 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -188,7 +188,8 @@ void f2fs_register_inmem_page(struct inode *inode, struct page *page) set_page_private_atomic(page); - new = f2fs_kmem_cache_alloc(inmem_entry_slab, GFP_NOFS); + new = f2fs_kmem_cache_alloc(inmem_entry_slab, + GFP_NOFS, true, NULL); /* add atomic page indices to the list */ new->page = page; @@ -1001,7 +1002,7 @@ static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi, pend_list = &dcc->pend_list[plist_idx(len)]; - dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS); + dc = f2fs_kmem_cache_alloc(discard_cmd_slab, GFP_NOFS, true, NULL); INIT_LIST_HEAD(&dc->list); dc->bdev = bdev; dc->lstart = lstart; @@ -1962,7 +1963,7 @@ static bool add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc, if (!de) { de = f2fs_kmem_cache_alloc(discard_entry_slab, - GFP_F2FS_ZERO); + GFP_F2FS_ZERO, true, NULL); de->start_blkaddr = START_BLOCK(sbi, cpc->trim_start); list_add_tail(&de->list, head); } @@ -4099,7 +4100,8 @@ static struct page *get_next_sit_page(struct f2fs_sb_info *sbi, static struct sit_entry_set *grab_sit_entry_set(void) { struct sit_entry_set *ses = - f2fs_kmem_cache_alloc(sit_entry_set_slab, GFP_NOFS); + f2fs_kmem_cache_alloc(sit_entry_set_slab, + GFP_NOFS, true, NULL); ses->entry_cnt = 0; INIT_LIST_HEAD(&ses->set_list); diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 9e0e3c998142..b556ca38f0fb 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -56,6 +56,7 @@ const char *f2fs_fault_name[FAULT_MAX] = { [FAULT_CHECKPOINT] = "checkpoint error", [FAULT_DISCARD] = "discard error", [FAULT_WRITE_IO] = "write IO error", + [FAULT_SLAB_ALLOC] = "slab alloc", }; void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, @@ -1300,7 +1301,8 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb) { struct f2fs_inode_info *fi; - fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); + fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep, + GFP_F2FS_ZERO, false, F2FS_SB(sb)); if (!fi) return NULL; diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c index c8f34decbf8e..1d2d29dcd41c 100644 --- a/fs/f2fs/xattr.c +++ b/fs/f2fs/xattr.c @@ -27,7 +27,8 @@ static void *xattr_alloc(struct f2fs_sb_info *sbi, int size, bool *is_inline) { if (likely(size == sbi->inline_xattr_slab_size)) { *is_inline = true; - return kmem_cache_zalloc(sbi->inline_xattr_slab, GFP_NOFS); + return f2fs_kmem_cache_alloc(sbi->inline_xattr_slab, + GFP_F2FS_ZERO, false, sbi); } *is_inline = false; return f2fs_kzalloc(sbi, size, GFP_NOFS); -- cgit v1.2.3