diff options
Diffstat (limited to 'fs/f2fs/gc.c')
| -rw-r--r-- | fs/f2fs/gc.c | 1009 |
1 files changed, 802 insertions, 207 deletions
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index db8725d473b5..4546e01b2ee0 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -7,35 +7,54 @@ */ #include <linux/fs.h> #include <linux/module.h> -#include <linux/backing-dev.h> #include <linux/init.h> #include <linux/f2fs_fs.h> #include <linux/kthread.h> #include <linux/delay.h> #include <linux/freezer.h> +#include <linux/sched/signal.h> +#include <linux/random.h> +#include <linux/sched/mm.h> #include "f2fs.h" #include "node.h" #include "segment.h" #include "gc.h" +#include "iostat.h" #include <trace/events/f2fs.h> +static struct kmem_cache *victim_entry_slab; + +static unsigned int count_bits(const unsigned long *addr, + unsigned int offset, unsigned int len); + static int gc_thread_func(void *data) { struct f2fs_sb_info *sbi = data; struct f2fs_gc_kthread *gc_th = sbi->gc_thread; wait_queue_head_t *wq = &sbi->gc_thread->gc_wait_queue_head; + wait_queue_head_t *fggc_wq = &sbi->gc_thread->fggc_wq; unsigned int wait_ms; + struct f2fs_gc_control gc_control = { + .victim_segno = NULL_SEGNO, + .should_migrate_blocks = false, + .err_gc_skipped = false }; wait_ms = gc_th->min_sleep_time; set_freezable(); do { + bool sync_mode, foreground = false; + wait_event_interruptible_timeout(*wq, kthread_should_stop() || freezing(current) || + waitqueue_active(fggc_wq) || gc_th->gc_wake, msecs_to_jiffies(wait_ms)); + if (test_opt(sbi, GC_MERGE) && waitqueue_active(fggc_wq)) + foreground = true; + /* give it a try one time */ if (gc_th->gc_wake) gc_th->gc_wake = 0; @@ -55,7 +74,8 @@ static int gc_thread_func(void *data) if (time_to_inject(sbi, FAULT_CHECKPOINT)) { f2fs_show_injection_info(sbi, FAULT_CHECKPOINT); - f2fs_stop_checkpoint(sbi, false); + f2fs_stop_checkpoint(sbi, false, + STOP_CP_REASON_FAULT_INJECT); } if (!sb_start_write_trylock(sbi->sb)) { @@ -76,20 +96,34 @@ static int gc_thread_func(void *data) * invalidated soon after by user update or deletion. * So, I'd like to wait some time to collect dirty segments. */ - if (sbi->gc_mode == GC_URGENT) { + if (sbi->gc_mode == GC_URGENT_HIGH) { + spin_lock(&sbi->gc_urgent_high_lock); + if (sbi->gc_urgent_high_remaining) { + sbi->gc_urgent_high_remaining--; + if (!sbi->gc_urgent_high_remaining) + sbi->gc_mode = GC_NORMAL; + } + spin_unlock(&sbi->gc_urgent_high_lock); + } + + if (sbi->gc_mode == GC_URGENT_HIGH || + sbi->gc_mode == GC_URGENT_MID) { wait_ms = gc_th->urgent_sleep_time; - down_write(&sbi->gc_lock); + f2fs_down_write(&sbi->gc_lock); goto do_gc; } - if (!down_write_trylock(&sbi->gc_lock)) { + if (foreground) { + f2fs_down_write(&sbi->gc_lock); + goto do_gc; + } else if (!f2fs_down_write_trylock(&sbi->gc_lock)) { stat_other_skip_bggc_count(sbi); goto next; } if (!is_idle(sbi, GC_TIME)) { increase_sleep_time(gc_th, &wait_ms); - up_write(&sbi->gc_lock); + f2fs_up_write(&sbi->gc_lock); stat_io_skip_bggc_count(sbi); goto next; } @@ -99,17 +133,34 @@ static int gc_thread_func(void *data) else increase_sleep_time(gc_th, &wait_ms); do_gc: - stat_inc_bggc_count(sbi->stat_info); + if (!foreground) + stat_inc_bggc_count(sbi->stat_info); + + sync_mode = F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC; + + /* foreground GC was been triggered via f2fs_balance_fs() */ + if (foreground) + sync_mode = false; + + gc_control.init_gc_type = sync_mode ? FG_GC : BG_GC; + gc_control.no_bg_gc = foreground; + gc_control.nr_free_secs = foreground ? 1 : 0; /* if return value is not zero, no victim was selected */ - if (f2fs_gc(sbi, test_opt(sbi, FORCE_FG_GC), true, NULL_SEGNO)) - wait_ms = gc_th->no_gc_sleep_time; + if (f2fs_gc(sbi, &gc_control)) { + /* don't bother wait_ms by foreground gc */ + if (!foreground) + wait_ms = gc_th->no_gc_sleep_time; + } + + if (foreground) + wake_up_all(&gc_th->fggc_wq); trace_f2fs_background_gc(sbi->sb, wait_ms, prefree_segments(sbi), free_segments(sbi)); /* balancing f2fs's metadata periodically */ - f2fs_balance_fs_bg(sbi); + f2fs_balance_fs_bg(sbi, true); next: sb_end_write(sbi->sb); @@ -134,15 +185,16 @@ int f2fs_start_gc_thread(struct f2fs_sb_info *sbi) gc_th->max_sleep_time = DEF_GC_THREAD_MAX_SLEEP_TIME; gc_th->no_gc_sleep_time = DEF_GC_THREAD_NOGC_SLEEP_TIME; - gc_th->gc_wake= 0; + gc_th->gc_wake = 0; sbi->gc_thread = gc_th; init_waitqueue_head(&sbi->gc_thread->gc_wait_queue_head); + init_waitqueue_head(&sbi->gc_thread->fggc_wq); sbi->gc_thread->f2fs_gc_task = kthread_run(gc_thread_func, sbi, "f2fs_gc-%u:%u", MAJOR(dev), MINOR(dev)); if (IS_ERR(gc_th->f2fs_gc_task)) { err = PTR_ERR(gc_th->f2fs_gc_task); - kvfree(gc_th); + kfree(gc_th); sbi->gc_thread = NULL; } out: @@ -152,26 +204,41 @@ out: void f2fs_stop_gc_thread(struct f2fs_sb_info *sbi) { struct f2fs_gc_kthread *gc_th = sbi->gc_thread; + if (!gc_th) return; kthread_stop(gc_th->f2fs_gc_task); - kvfree(gc_th); + wake_up_all(&gc_th->fggc_wq); + kfree(gc_th); sbi->gc_thread = NULL; } static int select_gc_type(struct f2fs_sb_info *sbi, int gc_type) { - int gc_mode = (gc_type == BG_GC) ? GC_CB : GC_GREEDY; + int gc_mode; + + if (gc_type == BG_GC) { + if (sbi->am.atgc_enabled) + gc_mode = GC_AT; + else + gc_mode = GC_CB; + } else { + gc_mode = GC_GREEDY; + } switch (sbi->gc_mode) { case GC_IDLE_CB: gc_mode = GC_CB; break; case GC_IDLE_GREEDY: - case GC_URGENT: + case GC_URGENT_HIGH: gc_mode = GC_GREEDY; break; + case GC_IDLE_AT: + gc_mode = GC_AT; + break; } + return gc_mode; } @@ -182,24 +249,41 @@ static void select_policy(struct f2fs_sb_info *sbi, int gc_type, if (p->alloc_mode == SSR) { p->gc_mode = GC_GREEDY; - p->dirty_segmap = dirty_i->dirty_segmap[type]; + p->dirty_bitmap = dirty_i->dirty_segmap[type]; + p->max_search = dirty_i->nr_dirty[type]; + p->ofs_unit = 1; + } else if (p->alloc_mode == AT_SSR) { + p->gc_mode = GC_GREEDY; + p->dirty_bitmap = dirty_i->dirty_segmap[type]; p->max_search = dirty_i->nr_dirty[type]; p->ofs_unit = 1; } else { p->gc_mode = select_gc_type(sbi, gc_type); - p->dirty_segmap = dirty_i->dirty_segmap[DIRTY]; - p->max_search = dirty_i->nr_dirty[DIRTY]; p->ofs_unit = sbi->segs_per_sec; + if (__is_large_section(sbi)) { + p->dirty_bitmap = dirty_i->dirty_secmap; + p->max_search = count_bits(p->dirty_bitmap, + 0, MAIN_SECS(sbi)); + } else { + p->dirty_bitmap = dirty_i->dirty_segmap[DIRTY]; + p->max_search = dirty_i->nr_dirty[DIRTY]; + } } - /* we need to check every dirty segments in the FG_GC case */ + /* + * adjust candidates range, should select all dirty segments for + * foreground GC and urgent GC cases. + */ if (gc_type != FG_GC && - (sbi->gc_mode != GC_URGENT) && + (sbi->gc_mode != GC_URGENT_HIGH) && + (p->gc_mode != GC_AT && p->alloc_mode != AT_SSR) && p->max_search > sbi->max_victim_search) p->max_search = sbi->max_victim_search; /* let's select beginning hot/small space first in no_heap mode*/ - if (test_opt(sbi, NOHEAP) && + if (f2fs_need_rand_seg(sbi)) + p->offset = prandom_u32_max(MAIN_SECS(sbi) * sbi->segs_per_sec); + else if (test_opt(sbi, NOHEAP) && (type == CURSEG_HOT_DATA || IS_NODESEG(type))) p->offset = 0; else @@ -212,10 +296,16 @@ static unsigned int get_max_cost(struct f2fs_sb_info *sbi, /* SSR allocates in a segment unit */ if (p->alloc_mode == SSR) return sbi->blocks_per_seg; + else if (p->alloc_mode == AT_SSR) + return UINT_MAX; + + /* LFS */ if (p->gc_mode == GC_GREEDY) return 2 * sbi->blocks_per_seg * p->ofs_unit; else if (p->gc_mode == GC_CB) return UINT_MAX; + else if (p->gc_mode == GC_AT) + return UINT_MAX; else /* No other gc_mode */ return 0; } @@ -249,13 +339,14 @@ static unsigned int get_cb_cost(struct f2fs_sb_info *sbi, unsigned int segno) unsigned char age = 0; unsigned char u; unsigned int i; + unsigned int usable_segs_per_sec = f2fs_usable_segs_in_sec(sbi, segno); - for (i = 0; i < sbi->segs_per_sec; i++) + for (i = 0; i < usable_segs_per_sec; i++) mtime += get_seg_entry(sbi, start + i)->mtime; vblocks = get_valid_blocks(sbi, segno, true); - mtime = div_u64(mtime, sbi->segs_per_sec); - vblocks = div_u64(vblocks, sbi->segs_per_sec); + mtime = div_u64(mtime, usable_segs_per_sec); + vblocks = div_u64(vblocks, usable_segs_per_sec); u = (vblocks * 100) >> sbi->log_blocks_per_seg; @@ -280,8 +371,11 @@ static inline unsigned int get_gc_cost(struct f2fs_sb_info *sbi, /* alloc_mode == LFS */ if (p->gc_mode == GC_GREEDY) return get_valid_blocks(sbi, segno, true); - else + else if (p->gc_mode == GC_CB) return get_cb_cost(sbi, segno); + + f2fs_bug_on(sbi, 1); + return 0; } static unsigned int count_bits(const unsigned long *addr, @@ -296,6 +390,318 @@ static unsigned int count_bits(const unsigned long *addr, return sum; } +static struct victim_entry *attach_victim_entry(struct f2fs_sb_info *sbi, + unsigned long long mtime, unsigned int segno, + struct rb_node *parent, struct rb_node **p, + bool left_most) +{ + struct atgc_management *am = &sbi->am; + struct victim_entry *ve; + + ve = f2fs_kmem_cache_alloc(victim_entry_slab, + GFP_NOFS, true, NULL); + + ve->mtime = mtime; + ve->segno = segno; + + rb_link_node(&ve->rb_node, parent, p); + rb_insert_color_cached(&ve->rb_node, &am->root, left_most); + + list_add_tail(&ve->list, &am->victim_list); + + am->victim_count++; + + return ve; +} + +static void insert_victim_entry(struct f2fs_sb_info *sbi, + unsigned long long mtime, unsigned int segno) +{ + struct atgc_management *am = &sbi->am; + struct rb_node **p; + struct rb_node *parent = NULL; + bool left_most = true; + + p = f2fs_lookup_rb_tree_ext(sbi, &am->root, &parent, mtime, &left_most); + attach_victim_entry(sbi, mtime, segno, parent, p, left_most); +} + +static void add_victim_entry(struct f2fs_sb_info *sbi, + struct victim_sel_policy *p, unsigned int segno) +{ + struct sit_info *sit_i = SIT_I(sbi); + unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); + unsigned int start = GET_SEG_FROM_SEC(sbi, secno); + unsigned long long mtime = 0; + unsigned int i; + + if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { + if (p->gc_mode == GC_AT && + get_valid_blocks(sbi, segno, true) == 0) + return; + } + + for (i = 0; i < sbi->segs_per_sec; i++) + mtime += get_seg_entry(sbi, start + i)->mtime; + mtime = div_u64(mtime, sbi->segs_per_sec); + + /* Handle if the system time has changed by the user */ + if (mtime < sit_i->min_mtime) + sit_i->min_mtime = mtime; + if (mtime > sit_i->max_mtime) + sit_i->max_mtime = mtime; + if (mtime < sit_i->dirty_min_mtime) + sit_i->dirty_min_mtime = mtime; + if (mtime > sit_i->dirty_max_mtime) + sit_i->dirty_max_mtime = mtime; + + /* don't choose young section as candidate */ + if (sit_i->dirty_max_mtime - mtime < p->age_threshold) + return; + + insert_victim_entry(sbi, mtime, segno); +} + +static struct rb_node *lookup_central_victim(struct f2fs_sb_info *sbi, + struct victim_sel_policy *p) +{ + struct atgc_management *am = &sbi->am; + struct rb_node *parent = NULL; + bool left_most; + + f2fs_lookup_rb_tree_ext(sbi, &am->root, &parent, p->age, &left_most); + + return parent; +} + +static void atgc_lookup_victim(struct f2fs_sb_info *sbi, + struct victim_sel_policy *p) +{ + struct sit_info *sit_i = SIT_I(sbi); + struct atgc_management *am = &sbi->am; + struct rb_root_cached *root = &am->root; + struct rb_node *node; + struct rb_entry *re; + struct victim_entry *ve; + unsigned long long total_time; + unsigned long long age, u, accu; + unsigned long long max_mtime = sit_i->dirty_max_mtime; + unsigned long long min_mtime = sit_i->dirty_min_mtime; + unsigned int sec_blocks = CAP_BLKS_PER_SEC(sbi); + unsigned int vblocks; + unsigned int dirty_threshold = max(am->max_candidate_count, + am->candidate_ratio * + am->victim_count / 100); + unsigned int age_weight = am->age_weight; + unsigned int cost; + unsigned int iter = 0; + + if (max_mtime < min_mtime) + return; + + max_mtime += 1; + total_time = max_mtime - min_mtime; + + accu = div64_u64(ULLONG_MAX, total_time); + accu = min_t(unsigned long long, div_u64(accu, 100), + DEFAULT_ACCURACY_CLASS); + + node = rb_first_cached(root); +next: + re = rb_entry_safe(node, struct rb_entry, rb_node); + if (!re) + return; + + ve = (struct victim_entry *)re; + + if (ve->mtime >= max_mtime || ve->mtime < min_mtime) + goto skip; + + /* age = 10000 * x% * 60 */ + age = div64_u64(accu * (max_mtime - ve->mtime), total_time) * + age_weight; + + vblocks = get_valid_blocks(sbi, ve->segno, true); + f2fs_bug_on(sbi, !vblocks || vblocks == sec_blocks); + + /* u = 10000 * x% * 40 */ + u = div64_u64(accu * (sec_blocks - vblocks), sec_blocks) * + (100 - age_weight); + + f2fs_bug_on(sbi, age + u >= UINT_MAX); + + cost = UINT_MAX - (age + u); + iter++; + + if (cost < p->min_cost || + (cost == p->min_cost && age > p->oldest_age)) { + p->min_cost = cost; + p->oldest_age = age; + p->min_segno = ve->segno; + } +skip: + if (iter < dirty_threshold) { + node = rb_next(node); + goto next; + } +} + +/* + * select candidates around source section in range of + * [target - dirty_threshold, target + dirty_threshold] + */ +static void atssr_lookup_victim(struct f2fs_sb_info *sbi, + struct victim_sel_policy *p) +{ + struct sit_info *sit_i = SIT_I(sbi); + struct atgc_management *am = &sbi->am; + struct rb_node *node; + struct rb_entry *re; + struct victim_entry *ve; + unsigned long long age; + unsigned long long max_mtime = sit_i->dirty_max_mtime; + unsigned long long min_mtime = sit_i->dirty_min_mtime; + unsigned int seg_blocks = sbi->blocks_per_seg; + unsigned int vblocks; + unsigned int dirty_threshold = max(am->max_candidate_count, + am->candidate_ratio * + am->victim_count / 100); + unsigned int cost; + unsigned int iter = 0; + int stage = 0; + + if (max_mtime < min_mtime) + return; + max_mtime += 1; +next_stage: + node = lookup_central_victim(sbi, p); +next_node: + re = rb_entry_safe(node, struct rb_entry, rb_node); + if (!re) { + if (stage == 0) + goto skip_stage; + return; + } + + ve = (struct victim_entry *)re; + + if (ve->mtime >= max_mtime || ve->mtime < min_mtime) + goto skip_node; + + age = max_mtime - ve->mtime; + + vblocks = get_seg_entry(sbi, ve->segno)->ckpt_valid_blocks; + f2fs_bug_on(sbi, !vblocks); + + /* rare case */ + if (vblocks == seg_blocks) + goto skip_node; + + iter++; + + age = max_mtime - abs(p->age - age); + cost = UINT_MAX - vblocks; + + if (cost < p->min_cost || + (cost == p->min_cost && age > p->oldest_age)) { + p->min_cost = cost; + p->oldest_age = age; + p->min_segno = ve->segno; + } +skip_node: + if (iter < dirty_threshold) { + if (stage == 0) + node = rb_prev(node); + else if (stage == 1) + node = rb_next(node); + goto next_node; + } +skip_stage: + if (stage < 1) { + stage++; + iter = 0; + goto next_stage; + } +} +static void lookup_victim_by_age(struct f2fs_sb_info *sbi, + struct victim_sel_policy *p) +{ + f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi, + &sbi->am.root, true)); + + if (p->gc_mode == GC_AT) + atgc_lookup_victim(sbi, p); + else if (p->alloc_mode == AT_SSR) + atssr_lookup_victim(sbi, p); + else + f2fs_bug_on(sbi, 1); +} + +static void release_victim_entry(struct f2fs_sb_info *sbi) +{ + struct atgc_management *am = &sbi->am; + struct victim_entry *ve, *tmp; + + list_for_each_entry_safe(ve, tmp, &am->victim_list, list) { + list_del(&ve->list); + kmem_cache_free(victim_entry_slab, ve); + am->victim_count--; + } + + am->root = RB_ROOT_CACHED; + + f2fs_bug_on(sbi, am->victim_count); + f2fs_bug_on(sbi, !list_empty(&am->victim_list)); +} + +static bool f2fs_pin_section(struct f2fs_sb_info *sbi, unsigned int segno) +{ + struct dirty_seglist_info *dirty_i = DIRTY_I(sbi); + unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); + + if (!dirty_i->enable_pin_section) + return false; + if (!test_and_set_bit(secno, dirty_i->pinned_secmap)) + dirty_i->pinned_secmap_cnt++; + return true; +} + +static bool f2fs_pinned_section_exists(struct dirty_seglist_info *dirty_i) +{ + return dirty_i->pinned_secmap_cnt; +} + +static bool f2fs_section_is_pinned(struct dirty_seglist_info *dirty_i, + unsigned int secno) +{ + return dirty_i->enable_pin_section && + f2fs_pinned_section_exists(dirty_i) && + test_bit(secno, dirty_i->pinned_secmap); +} + +static void f2fs_unpin_all_sections(struct f2fs_sb_info *sbi, bool enable) +{ + unsigned int bitmap_size = f2fs_bitmap_size(MAIN_SECS(sbi)); + + if (f2fs_pinned_section_exists(DIRTY_I(sbi))) { + memset(DIRTY_I(sbi)->pinned_secmap, 0, bitmap_size); + DIRTY_I(sbi)->pinned_secmap_cnt = 0; + } + DIRTY_I(sbi)->enable_pin_section = enable; +} + +static int f2fs_gc_pinned_control(struct inode *inode, int gc_type, + unsigned int segno) +{ + if (!f2fs_is_pinned_file(inode)) + return 0; + if (gc_type != FG_GC) + return -EBUSY; + if (!f2fs_pin_section(F2FS_I_SB(inode), segno)) + f2fs_pin_file_control(inode, true); + return -EAGAIN; +} + /* * This function is called from two paths. * One is garbage collection and the other is SSR segment selection. @@ -305,31 +711,51 @@ static unsigned int count_bits(const unsigned long *addr, * which has minimum valid blocks and removes it from dirty seglist. */ static int get_victim_by_default(struct f2fs_sb_info *sbi, - unsigned int *result, int gc_type, int type, char alloc_mode) + unsigned int *result, int gc_type, int type, + char alloc_mode, unsigned long long age) { struct dirty_seglist_info *dirty_i = DIRTY_I(sbi); struct sit_info *sm = SIT_I(sbi); struct victim_sel_policy p; unsigned int secno, last_victim; unsigned int last_segment; - unsigned int nsearched = 0; + unsigned int nsearched; + bool is_atgc; + int ret = 0; mutex_lock(&dirty_i->seglist_lock); last_segment = MAIN_SECS(sbi) * sbi->segs_per_sec; p.alloc_mode = alloc_mode; - select_policy(sbi, gc_type, type, &p); + p.age = age; + p.age_threshold = sbi->am.age_threshold; +retry: + select_policy(sbi, gc_type, type, &p); p.min_segno = NULL_SEGNO; + p.oldest_age = 0; p.min_cost = get_max_cost(sbi, &p); + is_atgc = (p.gc_mode == GC_AT || p.alloc_mode == AT_SSR); + nsearched = 0; + + if (is_atgc) + SIT_I(sbi)->dirty_min_mtime = ULLONG_MAX; + if (*result != NULL_SEGNO) { - if (get_valid_blocks(sbi, *result, false) && - !sec_usage_check(sbi, GET_SEC_FROM_SEG(sbi, *result))) + if (!get_valid_blocks(sbi, *result, false)) { + ret = -ENODATA; + goto out; + } + + if (sec_usage_check(sbi, GET_SEC_FROM_SEG(sbi, *result))) + ret = -EBUSY; + else p.min_segno = *result; goto out; } + ret = -ENODATA; if (p.max_search == 0) goto out; @@ -357,10 +783,14 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi, } while (1) { - unsigned long cost; - unsigned int segno; - - segno = find_next_bit(p.dirty_segmap, last_segment, p.offset); + unsigned long cost, *dirty_bitmap; + unsigned int unit_no, segno; + + dirty_bitmap = p.dirty_bitmap; + unit_no = find_next_bit(dirty_bitmap, + last_segment / p.ofs_unit, + p.offset / p.ofs_unit); + segno = unit_no * p.ofs_unit; if (segno >= last_segment) { if (sm->last_victim[p.gc_mode]) { last_segment = @@ -373,14 +803,7 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi, } p.offset = segno + p.ofs_unit; - if (p.ofs_unit > 1) { - p.offset -= segno % p.ofs_unit; - nsearched += count_bits(p.dirty_segmap, - p.offset - p.ofs_unit, - p.ofs_unit); - } else { - nsearched++; - } + nsearched++; #ifdef CONFIG_F2FS_CHECK_FS /* @@ -396,14 +819,38 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi, if (sec_usage_check(sbi, secno)) goto next; + /* Don't touch checkpointed data */ - if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) && - get_ckpt_valid_blocks(sbi, segno) && - p.alloc_mode != SSR)) - goto next; + if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { + if (p.alloc_mode == LFS) { + /* + * LFS is set to find source section during GC. + * The victim should have no checkpointed data. + */ + if (get_ckpt_valid_blocks(sbi, segno, true)) + goto next; + } else { + /* + * SSR | AT_SSR are set to find target segment + * for writes which can be full by checkpointed + * and newly written blocks. + */ + if (!f2fs_segment_has_free_slot(sbi, segno)) + goto next; + } + } + if (gc_type == BG_GC && test_bit(secno, dirty_i->victim_secmap)) goto next; + if (gc_type == FG_GC && f2fs_section_is_pinned(dirty_i, secno)) + goto next; + + if (is_atgc) { + add_victim_entry(sbi, &p, segno); + goto next; + } + cost = get_gc_cost(sbi, segno, &p); if (p.min_cost > cost) { @@ -413,14 +860,28 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi, next: if (nsearched >= p.max_search) { if (!sm->last_victim[p.gc_mode] && segno <= last_victim) - sm->last_victim[p.gc_mode] = last_victim + 1; + sm->last_victim[p.gc_mode] = + last_victim + p.ofs_unit; else - sm->last_victim[p.gc_mode] = segno + 1; + sm->last_victim[p.gc_mode] = segno + p.ofs_unit; sm->last_victim[p.gc_mode] %= (MAIN_SECS(sbi) * sbi->segs_per_sec); break; } } + + /* get victim for GC_AT/AT_SSR */ + if (is_atgc) { + lookup_victim_by_age(sbi, &p); + release_victim_entry(sbi); + } + + if (is_atgc && p.min_segno == NULL_SEGNO && + sm->elapsed_time < p.age_threshold) { + p.age_threshold = 0; + goto retry; + } + if (p.min_segno != NULL_SEGNO) { got_it: *result = (p.min_segno / p.ofs_unit) * p.ofs_unit; @@ -432,6 +893,7 @@ got_result: else set_bit(secno, dirty_i->victim_secmap); } + ret = 0; } out: @@ -441,7 +903,7 @@ out: prefree_segments(sbi), free_segments(sbi)); mutex_unlock(&dirty_i->seglist_lock); - return (p.min_segno == NULL_SEGNO) ? 0 : 1; + return ret; } static const struct victim_selection default_v_ops = { @@ -466,7 +928,8 @@ 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); @@ -476,6 +939,7 @@ static void add_gc_inode(struct gc_inode_list *gc_list, struct inode *inode) static void put_gc_inode(struct gc_inode_list *gc_list) { struct inode_entry *ie, *next_ie; + list_for_each_entry_safe(ie, next_ie, &gc_list->ilist, list) { radix_tree_delete(&gc_list->iroot, ie->inode->i_ino); iput(ie->inode); @@ -512,6 +976,7 @@ static int gc_node_segment(struct f2fs_sb_info *sbi, int phase = 0; bool fggc = (gc_type == FG_GC); int submitted = 0; + unsigned int usable_blks_in_seg = f2fs_usable_blks_in_seg(sbi, segno); start_addr = START_BLOCK(sbi, segno); @@ -521,7 +986,7 @@ next_step: if (fggc && phase == 2) atomic_inc(&sbi->wb_sync_req[NODE]); - for (off = 0; off < sbi->blocks_per_seg; off++, entry++) { + for (off = 0; off < usable_blks_in_seg; off++, entry++) { nid_t nid = le32_to_cpu(entry->nid); struct page *node_page; struct node_info ni; @@ -556,7 +1021,7 @@ next_step: continue; } - if (f2fs_get_node_info(sbi, nid, &ni)) { + if (f2fs_get_node_info(sbi, nid, &ni, false)) { f2fs_put_page(node_page, 1); continue; } @@ -599,9 +1064,11 @@ block_t f2fs_start_bidx_of_node(unsigned int node_ofs, struct inode *inode) bidx = node_ofs - 1; } else if (node_ofs <= indirect_blks) { int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1); + bidx = node_ofs - 2 - dec; } else { int dec = (node_ofs - indirect_blks - 3) / (NIDS_PER_BLOCK + 1); + bidx = node_ofs - 5 - dec; } return bidx * ADDRS_PER_BLOCK(inode) + ADDRS_PER_INODE(inode); @@ -612,7 +1079,7 @@ static bool is_alive(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, { struct page *node_page; nid_t nid; - unsigned int ofs_in_node; + unsigned int ofs_in_node, max_addrs; block_t source_blkaddr; nid = le32_to_cpu(sum->nid); @@ -622,7 +1089,7 @@ static bool is_alive(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, if (IS_ERR(node_page)) return false; - if (f2fs_get_node_info(sbi, nid, dni)) { + if (f2fs_get_node_info(sbi, nid, dni, false)) { f2fs_put_page(node_page, 1); return false; } @@ -633,8 +1100,21 @@ static bool is_alive(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, set_sbi_flag(sbi, SBI_NEED_FSCK); } + if (f2fs_check_nid_range(sbi, dni->ino)) { + f2fs_put_page(node_page, 1); + return false; + } + + max_addrs = IS_INODE(node_page) ? DEF_ADDRS_PER_INODE : + DEF_ADDRS_PER_BLOCK; + if (ofs_in_node >= max_addrs) { + f2fs_err(sbi, "Inconsistent ofs_in_node:%u in summary, ino:%u, nid:%u, max:%u", + ofs_in_node, dni->ino, dni->nid, max_addrs); + return false; + } + *nofs = ofs_of_node(node_page); - source_blkaddr = datablock_addr(NULL, node_page, ofs_in_node); + source_blkaddr = data_blkaddr(NULL, node_page, ofs_in_node); f2fs_put_page(node_page, 1); if (source_blkaddr != blkaddr) { @@ -644,9 +1124,9 @@ static bool is_alive(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, if (unlikely(check_valid_map(sbi, segno, offset))) { if (!test_and_set_bit(segno, SIT_I(sbi)->invalid_segmap)) { - f2fs_err(sbi, "mismatched blkaddr %u (source_blkaddr %u) in seg %u\n", - blkaddr, source_blkaddr, segno); - f2fs_bug_on(sbi, 1); + f2fs_err(sbi, "mismatched blkaddr %u (source_blkaddr %u) in seg %u", + blkaddr, source_blkaddr, segno); + set_sbi_flag(sbi, SBI_NEED_FSCK); } } #endif @@ -684,6 +1164,7 @@ static int ra_data_block(struct inode *inode, pgoff_t index) if (unlikely(!f2fs_is_valid_blkaddr(sbi, dn.data_blkaddr, DATA_GENERIC_ENHANCE_READ))) { err = -EFSCORRUPTED; + f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); goto put_page; } goto got_it; @@ -702,6 +1183,7 @@ static int ra_data_block(struct inode *inode, pgoff_t index) if (unlikely(!f2fs_is_valid_blkaddr(sbi, dn.data_blkaddr, DATA_GENERIC_ENHANCE))) { err = -EFSCORRUPTED; + f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR); goto put_page; } got_it: @@ -730,6 +1212,10 @@ got_it: goto put_encrypted_page; f2fs_put_page(fio.encrypted_page, 0); f2fs_put_page(page, 1); + + f2fs_update_iostat(sbi, inode, FS_DATA_READ_IO, F2FS_BLKSIZE); + f2fs_update_iostat(sbi, NULL, FS_GDATA_READ_IO, F2FS_BLKSIZE); + return 0; put_encrypted_page: f2fs_put_page(fio.encrypted_page, 1); @@ -762,7 +1248,10 @@ static int move_data_block(struct inode *inode, block_t bidx, struct page *page, *mpage; block_t newaddr; int err = 0; - bool lfs_mode = test_opt(fio.sbi, LFS); + bool lfs_mode = f2fs_lfs_mode(fio.sbi); + int type = fio.sbi->am.atgc_enabled && (gc_type == BG_GC) && + (fio.sbi->gc_mode != GC_URGENT_HIGH) ? + CURSEG_ALL_DATA_ATGC : CURSEG_COLD_DATA; /* do not read out */ page = f2fs_grab_cache_page(inode->i_mapping, bidx, false); @@ -774,18 +1263,9 @@ static int move_data_block(struct inode *inode, block_t bidx, goto out; } - if (f2fs_is_atomic_file(inode)) { - F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC]++; - F2FS_I_SB(inode)->skipped_atomic_files[gc_type]++; - err = -EAGAIN; + err = f2fs_gc_pinned_control(inode, gc_type, segno); + if (err) goto out; - } - - if (f2fs_is_pinned_file(inode)) { - f2fs_pin_file_control(inode, true); - err = -EAGAIN; - goto out; - } set_new_dnode(&dn, inode, NULL, NULL, 0); err = f2fs_get_dnode_of_data(&dn, bidx, LOOKUP_NODE); @@ -806,23 +1286,23 @@ static int move_data_block(struct inode *inode, block_t bidx, f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); - err = f2fs_get_node_info(fio.sbi, dn.nid, &ni); + err = f2fs_get_node_info(fio.sbi, dn.nid, &ni, false); if (err) goto put_out; - set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); - /* read page */ fio.page = page; fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; if (lfs_mode) - down_write(&fio.sbi->io_order_lock); + f2fs_down_write(&fio.sbi->io_order_lock); mpage = f2fs_grab_cache_page(META_MAPPING(fio.sbi), fio.old_blkaddr, false); - if (!mpage) + if (!mpage) { + err = -ENOMEM; goto up_out; + } fio.encrypted_page = mpage; @@ -833,6 +1313,12 @@ static int move_data_block(struct inode *inode, block_t bidx, f2fs_put_page(mpage, 1); goto up_out; } + + f2fs_update_iostat(fio.sbi, inode, FS_DATA_READ_IO, + F2FS_BLKSIZE); + f2fs_update_iostat(fio.sbi, NULL, FS_GDATA_READ_IO, + F2FS_BLKSIZE); + lock_page(mpage); if (unlikely(mpage->mapping != META_MAPPING(fio.sbi) || !PageUptodate(mpage))) { @@ -842,8 +1328,11 @@ static int move_data_block(struct inode *inode, block_t bidx, } } + set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version); + + /* allocate block address */ f2fs_allocate_data_block(fio.sbi, NULL, fio.old_blkaddr, &newaddr, - &sum, CURSEG_COLD_DATA, NULL, false); + &sum, type, NULL); fio.encrypted_page = f2fs_pagecache_get_page(META_MAPPING(fio.sbi), newaddr, FGP_LOCK | FGP_CREAT, GFP_NOFS); @@ -860,6 +1349,7 @@ static int move_data_block(struct inode *inode, block_t bidx, f2fs_put_page(mpage, 1); invalidate_mapping_pages(META_MAPPING(fio.sbi), fio.old_blkaddr, fio.old_blkaddr); + f2fs_invalidate_compress_page(fio.sbi, fio.old_blkaddr); set_page_dirty(fio.encrypted_page); if (clear_page_dirty_for_io(fio.encrypted_page)) @@ -868,9 +1358,6 @@ static int move_data_block(struct inode *inode, block_t bidx, set_page_writeback(fio.encrypted_page); ClearPageError(page); - /* allocate block address */ - f2fs_wait_on_page_writeback(dn.node_page, NODE, true, true); - fio.op = REQ_OP_WRITE; fio.op_flags = REQ_SYNC; fio.new_blkaddr = newaddr; @@ -882,7 +1369,7 @@ static int move_data_block(struct inode *inode, block_t bidx, goto put_page_out; } - f2fs_update_iostat(fio.sbi, FS_GC_DATA_IO, F2FS_BLKSIZE); + f2fs_update_iostat(fio.sbi, NULL, FS_GC_DATA_IO, F2FS_BLKSIZE); f2fs_update_data_blkaddr(&dn, newaddr); set_inode_flag(inode, FI_APPEND_WRITE); @@ -893,10 +1380,10 @@ put_page_out: recover_block: if (err) f2fs_do_replace_block(fio.sbi, &sum, newaddr, fio.old_blkaddr, - true, true); + true, true, true); up_out: if (lfs_mode) - up_write(&fio.sbi->io_order_lock); + f2fs_up_write(&fio.sbi->io_order_lock); put_out: f2fs_put_dnode(&dn); out: @@ -919,18 +1406,9 @@ static int move_data_page(struct inode *inode, block_t bidx, int gc_type, goto out; } - if (f2fs_is_atomic_file(inode)) { - F2FS_I(inode)->i_gc_failures[GC_FAILURE_ATOMIC]++; - F2FS_I_SB(inode)->skipped_atomic_files[gc_type]++; - err = -EAGAIN; - goto out; - } - if (f2fs_is_pinned_file(inode)) { - if (gc_type == FG_GC) - f2fs_pin_file_control(inode, true); - err = -EAGAIN; + err = f2fs_gc_pinned_control(inode, gc_type, segno); + if (err) goto out; - } if (gc_type == BG_GC) { if (PageWriteback(page)) { @@ -938,7 +1416,7 @@ static int move_data_page(struct inode *inode, block_t bidx, int gc_type, goto out; } set_page_dirty(page); - set_cold_data(page); + set_page_private_gcing(page); } else { struct f2fs_io_info fio = { .sbi = F2FS_I_SB(inode), @@ -964,13 +1442,13 @@ retry: f2fs_remove_dirty_inode(inode); } - set_cold_data(page); + set_page_private_gcing(page); err = f2fs_do_write_data_page(&fio); if (err) { - clear_cold_data(page); + clear_page_private_gcing(page); if (err == -ENOMEM) { - congestion_wait(BLK_RW_ASYNC, HZ/50); + memalloc_retry_wait(GFP_NOFS); goto retry; } if (is_dirty) @@ -990,7 +1468,8 @@ out: * the victim data block is ignored. */ static int gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, - struct gc_inode_list *gc_list, unsigned int segno, int gc_type) + struct gc_inode_list *gc_list, unsigned int segno, int gc_type, + bool force_migrate) { struct super_block *sb = sbi->sb; struct f2fs_summary *entry; @@ -998,13 +1477,14 @@ static int gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum, int off; int phase = 0; int submitted = 0; + unsigned int usable_blks_in_seg = f2fs_usable_blks_in_seg(sbi, segno); start_addr = START_BLOCK(sbi, segno); next_step: entry = sum; - for (off = 0; off < sbi->blocks_per_seg; off++, entry++) { + for (off = 0; off < usable_blks_in_seg; off++, entry++) { struct page *data_page; struct inode *inode; struct node_info dni; /* dnode info for the data */ @@ -1018,8 +1498,8 @@ next_step: * race condition along with SSR block allocation. */ if ((gc_type == BG_GC && has_not_enough_free_secs(sbi, 0, 0)) || - get_valid_blocks(sbi, segno, false) == - sbi->blocks_per_seg) + (!force_migrate && get_valid_blocks(sbi, segno, true) == + CAP_BLKS_PER_SEC(sbi))) return submitted; if (check_valid_map(sbi, segno, off) == 0) @@ -1048,13 +1528,20 @@ next_step: ofs_in_node = le16_to_cpu(entry->ofs_in_node); if (phase == 3) { + int err; + inode = f2fs_iget(sb, dni.ino); - if (IS_ERR(inode) || is_bad_inode(inode)) { - set_sbi_flag(sbi, SBI_NEED_FSCK); + if (IS_ERR(inode) || is_bad_inode(inode) || + special_file(inode->i_mode)) continue; + + err = f2fs_gc_pinned_control(inode, gc_type, segno); + if (err == -EAGAIN) { + iput(inode); + return submitted; } - if (!down_write_trylock( + if (!f2fs_down_write_trylock( &F2FS_I(inode)->i_gc_rwsem[WRITE])) { iput(inode); sbi->skipped_gc_rwsem++; @@ -1067,7 +1554,7 @@ next_step: if (f2fs_post_read_required(inode)) { int err = ra_data_block(inode, start_bidx); - up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); + f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); if (err) { iput(inode); continue; @@ -1078,7 +1565,7 @@ next_step: data_page = f2fs_get_read_data_page(inode, start_bidx, REQ_RAHEAD, true); - up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); + f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); if (IS_ERR(data_page)) { iput(inode); continue; @@ -1097,12 +1584,14 @@ next_step: int err; if (S_ISREG(inode->i_mode)) { - if (!down_write_trylock(&fi->i_gc_rwsem[READ])) + if (!f2fs_down_write_trylock(&fi->i_gc_rwsem[READ])) { + sbi->skipped_gc_rwsem++; continue; - if (!down_write_trylock( + } + if (!f2fs_down_write_trylock( &fi->i_gc_rwsem[WRITE])) { sbi->skipped_gc_rwsem++; - up_write(&fi->i_gc_rwsem[READ]); + f2fs_up_write(&fi->i_gc_rwsem[READ]); continue; } locked = true; @@ -1125,8 +1614,8 @@ next_step: submitted++; if (locked) { - up_write(&fi->i_gc_rwsem[WRITE]); - up_write(&fi->i_gc_rwsem[READ]); + f2fs_up_write(&fi->i_gc_rwsem[WRITE]); + f2fs_up_write(&fi->i_gc_rwsem[READ]); } stat_inc_data_blk_count(sbi, 1, gc_type); @@ -1147,14 +1636,15 @@ static int __get_victim(struct f2fs_sb_info *sbi, unsigned int *victim, down_write(&sit_i->sentry_lock); ret = DIRTY_I(sbi)->v_ops->get_victim(sbi, victim, gc_type, - NO_CHECK_TYPE, LFS); + NO_CHECK_TYPE, LFS, 0); up_write(&sit_i->sentry_lock); return ret; } static int do_garbage_collect(struct f2fs_sb_info *sbi, unsigned int start_segno, - struct gc_inode_list *gc_list, int gc_type) + struct gc_inode_list *gc_list, int gc_type, + bool force_migrate) { struct page *sum_page; struct f2fs_summary_block *sum; @@ -1169,6 +1659,17 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi, if (__is_large_section(sbi)) end_segno = rounddown(end_segno, sbi->segs_per_sec); + /* + * zone-capacity can be less than zone-size in zoned devices, + * resulting in less than expected usable segments in the zone, + * calculate the end segno in the zone which can be garbage collected + */ + if (f2fs_sb_has_blkzoned(sbi)) + end_segno -= sbi->segs_per_sec - + f2fs_usable_segs_in_sec(sbi, segno); + + sanity_check_seg_type(sbi, get_seg_entry(sbi, segno)->type); + /* readahead multi ssa blocks those have contiguous address */ if (__is_large_section(sbi)) f2fs_ra_meta_pages(sbi, GET_SUM_BLOCK(sbi, segno), @@ -1203,7 +1704,7 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi, if (get_valid_blocks(sbi, segno, false) == 0) goto freed; - if (__is_large_section(sbi) && + if (gc_type == BG_GC && __is_large_section(sbi) && migrated >= sbi->migration_granularity) goto skip; if (!PageUptodate(sum_page) || unlikely(f2fs_cp_error(sbi))) @@ -1214,7 +1715,8 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi, f2fs_err(sbi, "Inconsistent segment (%u) type [%d, %d] in SSA and SIT", segno, type, GET_SUM_TYPE((&sum->footer))); set_sbi_flag(sbi, SBI_NEED_FSCK); - f2fs_stop_checkpoint(sbi, false); + f2fs_stop_checkpoint(sbi, false, + STOP_CP_REASON_CORRUPTED_SUMMARY); goto skip; } @@ -1230,15 +1732,17 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi, gc_type); else submitted += gc_data_segment(sbi, sum->entries, gc_list, - segno, gc_type); + segno, gc_type, + force_migrate); stat_inc_seg_count(sbi, type, gc_type); + sbi->gc_reclaimed_segs[sbi->gc_mode]++; + migrated++; freed: if (gc_type == FG_GC && get_valid_blocks(sbi, segno, false) == 0) seg_freed++; - migrated++; if (__is_large_section(sbi) && segno + 1 < end_segno) sbi->next_victim_seg[gc_type] = segno + 1; @@ -1257,23 +1761,21 @@ skip: return seg_freed; } -int f2fs_gc(struct f2fs_sb_info *sbi, bool sync, - bool background, unsigned int segno) +int f2fs_gc(struct f2fs_sb_info *sbi, struct f2fs_gc_control *gc_control) { - int gc_type = sync ? FG_GC : BG_GC; + int gc_type = gc_control->init_gc_type; + unsigned int segno = gc_control->victim_segno; int sec_freed = 0, seg_freed = 0, total_freed = 0; int ret = 0; struct cp_control cpc; - unsigned int init_segno = segno; struct gc_inode_list gc_list = { .ilist = LIST_HEAD_INIT(gc_list.ilist), .iroot = RADIX_TREE_INIT(gc_list.iroot, GFP_NOFS), }; - unsigned long long last_skipped = sbi->skipped_atomic_files[FG_GC]; - unsigned long long first_skipped; unsigned int skipped_round = 0, round = 0; - trace_f2fs_gc_begin(sbi->sb, sync, background, + trace_f2fs_gc_begin(sbi->sb, gc_type, gc_control->no_bg_gc, + gc_control->nr_free_secs, get_pages(sbi, F2FS_DIRTY_NODES), get_pages(sbi, F2FS_DIRTY_DENTS), get_pages(sbi, F2FS_DIRTY_IMETA), @@ -1284,7 +1786,6 @@ int f2fs_gc(struct f2fs_sb_info *sbi, bool sync, cpc.reason = __get_cp_reason(sbi); sbi->skipped_gc_rwsem = 0; - first_skipped = last_skipped; gc_more: if (unlikely(!(sbi->sb->s_flags & SB_ACTIVE))) { ret = -EINVAL; @@ -1301,8 +1802,7 @@ gc_more: * threshold, we can make them free by checkpoint. Then, we * secure free segments which doesn't need fggc any more. */ - if (prefree_segments(sbi) && - !is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { + if (prefree_segments(sbi)) { ret = f2fs_write_checkpoint(sbi, &cpc); if (ret) goto stop; @@ -1312,54 +1812,69 @@ gc_more: } /* f2fs_balance_fs doesn't need to do BG_GC in critical path. */ - if (gc_type == BG_GC && !background) { + if (gc_type == BG_GC && gc_control->no_bg_gc) { ret = -EINVAL; goto stop; } - if (!__get_victim(sbi, &segno, gc_type)) { - ret = -ENODATA; +retry: + ret = __get_victim(sbi, &segno, gc_type); + if (ret) { + /* allow to search victim from sections has pinned data */ + if (ret == -ENODATA && gc_type == FG_GC && + f2fs_pinned_section_exists(DIRTY_I(sbi))) { + f2fs_unpin_all_sections(sbi, false); + goto retry; + } goto stop; } - seg_freed = do_garbage_collect(sbi, segno, &gc_list, gc_type); - if (gc_type == FG_GC && seg_freed == sbi->segs_per_sec) - sec_freed++; + seg_freed = do_garbage_collect(sbi, segno, &gc_list, gc_type, + gc_control->should_migrate_blocks); total_freed += seg_freed; - if (gc_type == FG_GC) { - if (sbi->skipped_atomic_files[FG_GC] > last_skipped || - sbi->skipped_gc_rwsem) - skipped_round++; - last_skipped = sbi->skipped_atomic_files[FG_GC]; - round++; - } + if (seg_freed == f2fs_usable_segs_in_sec(sbi, segno)) + sec_freed++; - if (gc_type == FG_GC && seg_freed) + if (gc_type == FG_GC) sbi->cur_victim_sec = NULL_SEGNO; - if (sync) + if (gc_control->init_gc_type == FG_GC || + !has_not_enough_free_secs(sbi, + (gc_type == FG_GC) ? sec_freed : 0, 0)) { + if (gc_type == FG_GC && sec_freed < gc_control->nr_free_secs) + goto go_gc_more; goto stop; + } - if (has_not_enough_free_secs(sbi, sec_freed, 0)) { - if (skipped_round <= MAX_SKIP_GC_COUNT || - skipped_round * 2 < round) { - segno = NULL_SEGNO; - goto gc_more; + /* FG_GC stops GC by skip_count */ + if (gc_type == FG_GC) { + if (sbi->skipped_gc_rwsem) + skipped_round++; + round++; + if (skipped_round > MAX_SKIP_GC_COUNT && + skipped_round * 2 >= round) { + ret = f2fs_write_checkpoint(sbi, &cpc); + goto stop; } + } - if (first_skipped < last_skipped && - (last_skipped - first_skipped) > - sbi->skipped_gc_rwsem) { - f2fs_drop_inmem_pages_all(sbi, true); - segno = NULL_SEGNO; - goto gc_more; - } - if (gc_type == FG_GC && !is_sbi_flag_set(sbi, SBI_CP_DISABLED)) - ret = f2fs_write_checkpoint(sbi, &cpc); + /* Write checkpoint to reclaim prefree segments */ + if (free_sections(sbi) < NR_CURSEG_PERSIST_TYPE && + prefree_segments(sbi)) { + ret = f2fs_write_checkpoint(sbi, &cpc); + if (ret) + goto stop; } +go_gc_more: + segno = NULL_SEGNO; + goto gc_more; + stop: SIT_I(sbi)->last_victim[ALLOC_NEXT] = 0; - SIT_I(sbi)->last_victim[FLUSH_DEVICE] = init_segno; + SIT_I(sbi)->last_victim[FLUSH_DEVICE] = gc_control->victim_segno; + + if (gc_type == FG_GC) + f2fs_unpin_all_sections(sbi, true); trace_f2fs_gc_end(sbi->sb, ret, total_freed, sec_freed, get_pages(sbi, F2FS_DIRTY_NODES), @@ -1370,15 +1885,47 @@ stop: reserved_segments(sbi), prefree_segments(sbi)); - up_write(&sbi->gc_lock); + f2fs_up_write(&sbi->gc_lock); put_gc_inode(&gc_list); - if (sync && !ret) + if (gc_control->err_gc_skipped && !ret) ret = sec_freed ? 0 : -EAGAIN; return ret; } +int __init f2fs_create_garbage_collection_cache(void) +{ + victim_entry_slab = f2fs_kmem_cache_create("f2fs_victim_entry", + sizeof(struct victim_entry)); + if (!victim_entry_slab) + return -ENOMEM; + return 0; +} + +void f2fs_destroy_garbage_collection_cache(void) +{ + kmem_cache_destroy(victim_entry_slab); +} + +static void init_atgc_management(struct f2fs_sb_info *sbi) +{ + struct atgc_management *am = &sbi->am; + + if (test_opt(sbi, ATGC) && + SIT_I(sbi)->elapsed_time >= DEF_GC_THREAD_AGE_THRESHOLD) + am->atgc_enabled = true; + + am->root = RB_ROOT_CACHED; + INIT_LIST_HEAD(&am->victim_list); + am->victim_count = 0; + + am->candidate_ratio = DEF_GC_THREAD_CANDIDATE_RATIO; + am->max_candidate_count = DEF_GC_THREAD_MAX_CANDIDATE_COUNT; + am->age_weight = DEF_GC_THREAD_AGE_WEIGHT; + am->age_threshold = DEF_GC_THREAD_AGE_THRESHOLD; +} + void f2fs_build_gc_manager(struct f2fs_sb_info *sbi) { DIRTY_I(sbi)->v_ops = &default_v_ops; @@ -1389,18 +1936,37 @@ void f2fs_build_gc_manager(struct f2fs_sb_info *sbi) if (f2fs_is_multi_device(sbi) && !__is_large_section(sbi)) SIT_I(sbi)->last_victim[ALLOC_NEXT] = GET_SEGNO(sbi, FDEV(0).end_blk) + 1; + + init_atgc_management(sbi); } -static int free_segment_range(struct f2fs_sb_info *sbi, unsigned int start, - unsigned int end) +static int free_segment_range(struct f2fs_sb_info *sbi, + unsigned int secs, bool gc_only) { - int type; - unsigned int segno, next_inuse; + unsigned int segno, next_inuse, start, end; + struct cp_control cpc = { CP_RESIZE, 0, 0, 0 }; + int gc_mode, gc_type; int err = 0; + int type; + + /* Force block allocation for GC */ + MAIN_SECS(sbi) -= secs; + start = MAIN_SECS(sbi) * sbi->segs_per_sec; + end = MAIN_SEGS(sbi) - 1; + + mutex_lock(&DIRTY_I(sbi)->seglist_lock); + for (gc_mode = 0; gc_mode < MAX_GC_POLICY; gc_mode++) + if (SIT_I(sbi)->last_victim[gc_mode] >= start) + SIT_I(sbi)->last_victim[gc_mode] = 0; + + for (gc_type = BG_GC; gc_type <= FG_GC; gc_type++) + if (sbi->next_victim_seg[gc_type] >= start) + sbi->next_victim_seg[gc_type] = NULL_SEGNO; + mutex_unlock(&DIRTY_I(sbi)->seglist_lock); /* Move out cursegs from the target range */ - for (type = CURSEG_HOT_DATA; type < NR_CURSEG_TYPE; type++) - allocate_segment_for_resize(sbi, type, start, end); + for (type = CURSEG_HOT_DATA; type < NR_CURSEG_PERSIST_TYPE; type++) + f2fs_allocate_segment_for_resize(sbi, type, start, end); /* do GC to move out valid blocks in the range */ for (segno = start; segno <= end; segno += sbi->segs_per_sec) { @@ -1409,18 +1975,24 @@ static int free_segment_range(struct f2fs_sb_info *sbi, unsigned int start, .iroot = RADIX_TREE_INIT(gc_list.iroot, GFP_NOFS), }; - down_write(&sbi->gc_lock); - do_garbage_collect(sbi, segno, &gc_list, FG_GC); - up_write(&sbi->gc_lock); + do_garbage_collect(sbi, segno, &gc_list, FG_GC, true); put_gc_inode(&gc_list); - if (get_valid_blocks(sbi, segno, true)) - return -EAGAIN; + if (!gc_only && get_valid_blocks(sbi, segno, true)) { + err = -EAGAIN; + goto out; + } + if (fatal_signal_pending(current)) { + err = -ERESTARTSYS; + goto out; + } } + if (gc_only) + goto out; - err = f2fs_sync_fs(sbi->sb, 1); + err = f2fs_write_checkpoint(sbi, &cpc); if (err) - return err; + goto out; next_inuse = find_next_inuse(FREE_I(sbi), end + 1, start); if (next_inuse <= end) { @@ -1428,18 +2000,27 @@ static int free_segment_range(struct f2fs_sb_info *sbi, unsigned int start, next_inuse); f2fs_bug_on(sbi, 1); } +out: + MAIN_SECS(sbi) += secs; return err; } static void update_sb_metadata(struct f2fs_sb_info *sbi, int secs) { struct f2fs_super_block *raw_sb = F2FS_RAW_SUPER(sbi); - int section_count = le32_to_cpu(raw_sb->section_count); - int segment_count = le32_to_cpu(raw_sb->segment_count); - int segment_count_main = le32_to_cpu(raw_sb->segment_count_main); - long long block_count = le64_to_cpu(raw_sb->block_count); + int section_count; + int segment_count; + int segment_count_main; + long long block_count; int segs = secs * sbi->segs_per_sec; + f2fs_down_write(&sbi->sb_lock); + + section_count = le32_to_cpu(raw_sb->section_count); + segment_count = le32_to_cpu(raw_sb->segment_count); + segment_count_main = le32_to_cpu(raw_sb->segment_count_main); + block_count = le64_to_cpu(raw_sb->block_count); + raw_sb->section_count = cpu_to_le32(section_count + secs); raw_sb->segment_count = cpu_to_le32(segment_count + segs); raw_sb->segment_count_main = cpu_to_le32(segment_count_main + segs); @@ -1453,6 +2034,8 @@ static void update_sb_metadata(struct f2fs_sb_info *sbi, int secs) raw_sb->devs[last_dev].total_segments = cpu_to_le32(dev_segs + segs); } + + f2fs_up_write(&sbi->sb_lock); } static void update_fs_metadata(struct f2fs_sb_info *sbi, int secs) @@ -1464,6 +2047,7 @@ static void update_fs_metadata(struct f2fs_sb_info *sbi, int secs) SM_I(sbi)->segment_count = (int)SM_I(sbi)->segment_count + segs; MAIN_SEGS(sbi) = (int)MAIN_SEGS(sbi) + segs; + MAIN_SECS(sbi) += secs; FREE_I(sbi)->free_sections = (int)FREE_I(sbi)->free_sections + secs; FREE_I(sbi)->free_segments = (int)FREE_I(sbi)->free_segments + segs; F2FS_CKPT(sbi)->user_block_count = cpu_to_le64(user_block_count + blks); @@ -1485,8 +2069,8 @@ static void update_fs_metadata(struct f2fs_sb_info *sbi, int secs) int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count) { __u64 old_block_count, shrunk_blocks; + struct cp_control cpc = { CP_RESIZE, 0, 0, 0 }; unsigned int secs; - int gc_mode, gc_type; int err = 0; __u32 rem; @@ -1521,75 +2105,86 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count) return -EINVAL; } - freeze_bdev(sbi->sb->s_bdev); - shrunk_blocks = old_block_count - block_count; secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi)); + + /* stop other GC */ + if (!f2fs_down_write_trylock(&sbi->gc_lock)) + return -EAGAIN; + + /* stop CP to protect MAIN_SEC in free_segment_range */ + f2fs_lock_op(sbi); + spin_lock(&sbi->stat_lock); if (shrunk_blocks + valid_user_blocks(sbi) + sbi->current_reserved_blocks + sbi->unusable_block_count + F2FS_OPTION(sbi).root_reserved_blocks > sbi->user_block_count) err = -ENOSPC; - else - sbi->user_block_count -= shrunk_blocks; spin_unlock(&sbi->stat_lock); - if (err) { - thaw_bdev(sbi->sb->s_bdev, sbi->sb); - return err; - } - mutex_lock(&sbi->resize_mutex); - set_sbi_flag(sbi, SBI_IS_RESIZEFS); + if (err) + goto out_unlock; - mutex_lock(&DIRTY_I(sbi)->seglist_lock); + err = free_segment_range(sbi, secs, true); - MAIN_SECS(sbi) -= secs; +out_unlock: + f2fs_unlock_op(sbi); + f2fs_up_write(&sbi->gc_lock); + if (err) + return err; - for (gc_mode = 0; gc_mode < MAX_GC_POLICY; gc_mode++) - if (SIT_I(sbi)->last_victim[gc_mode] >= - MAIN_SECS(sbi) * sbi->segs_per_sec) - SIT_I(sbi)->last_victim[gc_mode] = 0; + set_sbi_flag(sbi, SBI_IS_RESIZEFS); - for (gc_type = BG_GC; gc_type <= FG_GC; gc_type++) - if (sbi->next_victim_seg[gc_type] >= - MAIN_SECS(sbi) * sbi->segs_per_sec) - sbi->next_victim_seg[gc_type] = NULL_SEGNO; + freeze_super(sbi->sb); + f2fs_down_write(&sbi->gc_lock); + f2fs_down_write(&sbi->cp_global_sem); - mutex_unlock(&DIRTY_I(sbi)->seglist_lock); + spin_lock(&sbi->stat_lock); + if (shrunk_blocks + valid_user_blocks(sbi) + + sbi->current_reserved_blocks + sbi->unusable_block_count + + F2FS_OPTION(sbi).root_reserved_blocks > sbi->user_block_count) + err = -ENOSPC; + else + sbi->user_block_count -= shrunk_blocks; + spin_unlock(&sbi->stat_lock); + if (err) + goto out_err; - err = free_segment_range(sbi, MAIN_SECS(sbi) * sbi->segs_per_sec, - MAIN_SEGS(sbi) - 1); + err = free_segment_range(sbi, secs, false); if (err) - goto out; + goto recover_out; update_sb_metadata(sbi, -secs); err = f2fs_commit_super(sbi, false); if (err) { update_sb_metadata(sbi, secs); - goto out; + goto recover_out; } update_fs_metadata(sbi, -secs); clear_sbi_flag(sbi, SBI_IS_RESIZEFS); - err = f2fs_sync_fs(sbi->sb, 1); + set_sbi_flag(sbi, SBI_IS_DIRTY); + + err = f2fs_write_checkpoint(sbi, &cpc); if (err) { update_fs_metadata(sbi, secs); update_sb_metadata(sbi, secs); f2fs_commit_super(sbi, false); } -out: +recover_out: if (err) { set_sbi_flag(sbi, SBI_NEED_FSCK); f2fs_err(sbi, "resize_fs failed, should run fsck to repair!"); - MAIN_SECS(sbi) += secs; spin_lock(&sbi->stat_lock); sbi->user_block_count += shrunk_blocks; spin_unlock(&sbi->stat_lock); } +out_err: + f2fs_up_write(&sbi->cp_global_sem); + f2fs_up_write(&sbi->gc_lock); + thaw_super(sbi->sb); clear_sbi_flag(sbi, SBI_IS_RESIZEFS); - mutex_unlock(&sbi->resize_mutex); - thaw_bdev(sbi->sb->s_bdev, sbi->sb); return err; } |