On 2022/5/10 0:55, Jaegeuk Kim wrote:
On 05/08, Chao Yu wrote:On 2022/5/7 7:20, Jaegeuk Kim wrote:The f2fs_gc uses a bitmap to indicate pinned sections, but when disabling chckpoint, we call f2fs_gc() with NULL_SEGNO which selects the same dirty segment as a victim all the time, resulting in checkpoint=disable failure, for example. Let's pick another one, if we fail to collect it.Signed-off-by: Jaegeuk Kim <[email protected]> --- fs/f2fs/f2fs.h | 1 + fs/f2fs/file.c | 12 ++++++++---- fs/f2fs/gc.c | 11 +++++++---- fs/f2fs/segment.c | 3 ++- fs/f2fs/super.c | 3 ++- include/trace/events/f2fs.h | 11 ++++++++--- 6 files changed, 28 insertions(+), 13 deletions(-) diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index d49b9b476592..27871f6efb01 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1282,6 +1282,7 @@ struct f2fs_gc_control { bool no_bg_gc; /* check the space and stop bg_gc */ bool should_migrate_blocks; /* should migrate blocks */ bool err_gc_skipped; /* return EAGAIN if GC skipped */ + unsigned int nr_free_secs; /* # of free sections to do GC */ }; /* For s_flag in struct f2fs_sb_info */ diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 0e7d101c3e65..7072c2b86b2f 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -1650,7 +1650,8 @@ static int expand_inode_data(struct inode *inode, loff_t offset, struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, .init_gc_type = FG_GC, .should_migrate_blocks = false, - .err_gc_skipped = true }; + .err_gc_skipped = true, + .nr_free_secs = 0 }; pgoff_t pg_start, pg_end; loff_t new_size = i_size_read(inode); loff_t off_end; @@ -2453,7 +2454,8 @@ static int f2fs_ioc_gc(struct file *filp, unsigned long arg) struct f2fs_sb_info *sbi = F2FS_I_SB(inode); struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, .no_bg_gc = false, - .should_migrate_blocks = false }; + .should_migrate_blocks = false, + .nr_free_secs = 0 }; __u32 sync; int ret; @@ -2494,7 +2496,8 @@ static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) .init_gc_type = range->sync ? FG_GC : BG_GC, .no_bg_gc = false, .should_migrate_blocks = false, - .err_gc_skipped = range->sync }; + .err_gc_skipped = range->sync, + .nr_free_secs = 0 }; u64 end; int ret; @@ -2940,7 +2943,8 @@ static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) struct f2fs_gc_control gc_control = { .init_gc_type = FG_GC, .should_migrate_blocks = true, - .err_gc_skipped = true }; + .err_gc_skipped = true, + .nr_free_secs = 0 }; int ret; if (!capable(CAP_SYS_ADMIN)) diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index ec3f6f876e76..f63576ff1c2d 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -147,6 +147,7 @@ static int gc_thread_func(void *data) 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 init_gc_type is BG_GC, sec_freed won't increase for background GC due to below statement: if (gc_type == FG_GC && get_valid_blocks(sbi, segno, false) == 0) seg_freed++; It may cause gc thread migrates lots of segments in each round?BG_GC include two cases, f2fs_balance_fs and gc thread for GC_MERGE, both of which are no_bg_gc=true. So, I think this would be enough. Other cases that sets nr_free_secs to 1 should be FG_GC only.
What I mean is, in below check condition, for BG_GC cases, if !has_not_enough_free_secs(sbi, sec_freed, 0) is true, since sec_freed will never be increased due to above condition, so it will goto gc_more label all the time, result in looping until migrating all dirty segments. Thanks,
if ((gc_control->init_gc_type == FG_GC || !has_not_enough_free_secs(sbi, sec_freed, 0))) { if (sec_freed < gc_control->nr_free_secs) goto gc_more; goto stop; }/* if return value is not zero, no victim was selected */ if (f2fs_gc(sbi, &gc_control)) @@ -1776,6 +1777,7 @@ int f2fs_gc(struct f2fs_sb_info *sbi, struct f2fs_gc_control *gc_control) unsigned int skipped_round = 0, round = 0; 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), @@ -1848,11 +1850,12 @@ int f2fs_gc(struct f2fs_sb_info *sbi, struct f2fs_gc_control *gc_control) if (gc_type == FG_GC) sbi->cur_victim_sec = NULL_SEGNO; - if (gc_control->init_gc_type == FG_GC) - goto stop; - - if (!has_not_enough_free_secs(sbi, sec_freed, 0)) + if ((gc_control->init_gc_type == FG_GC || + !has_not_enough_free_secs(sbi, sec_freed, 0))) { + if (sec_freed < gc_control->nr_free_secs) + goto gc_more; goto stop; + } if (skipped_round <= MAX_SKIP_GC_COUNT || skipped_round * 2 < round) { diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index bc63f0572c64..d6b9231ab0e1 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -528,7 +528,8 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need) .init_gc_type = BG_GC, .no_bg_gc = true, .should_migrate_blocks = false, - .err_gc_skipped = false }; + .err_gc_skipped = false, + .nr_free_secs = 1 };.init_gc_type is BG_GC, so !has_not_enough_free_secs(sbi, sec_freed, 0) condition should be enough to exit? if ((gc_control->init_gc_type == FG_GC || !has_not_enough_free_secs(sbi, sec_freed, 0))) { if (sec_freed < gc_control->nr_free_secs) goto gc_more; goto stop; } Thanks,f2fs_down_write(&sbi->gc_lock); f2fs_gc(sbi, &gc_control); } diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 8b23fa6fc6b7..5d5b35067c3d 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -2084,7 +2084,8 @@ static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) .victim_segno = NULL_SEGNO, .init_gc_type = FG_GC, .should_migrate_blocks = false, - .err_gc_skipped = true }; + .err_gc_skipped = true, + .nr_free_secs = 1 }; f2fs_down_write(&sbi->gc_lock); err = f2fs_gc(sbi, &gc_control); diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h index 6699174977a3..349679a72301 100644 --- a/include/trace/events/f2fs.h +++ b/include/trace/events/f2fs.h @@ -653,18 +653,21 @@ TRACE_EVENT(f2fs_background_gc, TRACE_EVENT(f2fs_gc_begin, TP_PROTO(struct super_block *sb, int gc_type, bool no_bg_gc, + unsigned int nr_free_secs, long long dirty_nodes, long long dirty_dents, long long dirty_imeta, unsigned int free_sec, unsigned int free_seg, int reserved_seg, unsigned int prefree_seg), - TP_ARGS(sb, gc_type, no_bg_gc, dirty_nodes, dirty_dents, dirty_imeta, + TP_ARGS(sb, gc_type, no_bg_gc, nr_free_secs, dirty_nodes, + dirty_dents, dirty_imeta, free_sec, free_seg, reserved_seg, prefree_seg), TP_STRUCT__entry( __field(dev_t, dev) __field(int, gc_type) __field(bool, no_bg_gc) + __field(unsigned int, nr_free_secs) __field(long long, dirty_nodes) __field(long long, dirty_dents) __field(long long, dirty_imeta) @@ -678,6 +681,7 @@ TRACE_EVENT(f2fs_gc_begin, __entry->dev = sb->s_dev; __entry->gc_type = gc_type; __entry->no_bg_gc = no_bg_gc; + __entry->nr_free_secs = nr_free_secs; __entry->dirty_nodes = dirty_nodes; __entry->dirty_dents = dirty_dents; __entry->dirty_imeta = dirty_imeta; @@ -687,12 +691,13 @@ TRACE_EVENT(f2fs_gc_begin, __entry->prefree_seg = prefree_seg; ), - TP_printk("dev = (%d,%d), gc_type = %s, no_background_GC = %d, nodes = %lld, " - "dents = %lld, imeta = %lld, free_sec:%u, free_seg:%u, " + TP_printk("dev = (%d,%d), gc_type = %s, no_background_GC = %d, nr_free_secs = %u, " + "nodes = %lld, dents = %lld, imeta = %lld, free_sec:%u, free_seg:%u, " "rsv_seg:%d, prefree_seg:%u", show_dev(__entry->dev), show_gc_type(__entry->gc_type), __entry->no_bg_gc, + __entry->nr_free_secs, __entry->dirty_nodes, __entry->dirty_dents, __entry->dirty_imeta,
_______________________________________________ Linux-f2fs-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
