Re: [f2fs-dev] [PATCH 2/2 v2] f2fs: avoid f2fs_bug_on if f2fs_get_meta_page_nofail got EIO

[Chao Yu <yuchao0@huawei.com>]

On 2018/9/21 5:46, Jaegeuk Kim wrote:
> On 09/20, Chao Yu wrote:
>> On 2018/9/19 1:56, Jaegeuk Kim wrote:
>>> This patch avoids BUG_ON when f2fs_get_meta_page_nofail got EIO during
>>> xfstests/generic/475.
>>>
>>> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
>>> ---
>>> Change log from v1:
>>>  - propagate errno
>>>  - drop sum_pages
>>>
>>>  fs/f2fs/checkpoint.c | 10 +++++-----
>>>  fs/f2fs/f2fs.h       |  2 +-
>>>  fs/f2fs/gc.c         |  9 +++++++++
>>>  fs/f2fs/node.c       | 32 ++++++++++++++++++++++++--------
>>>  fs/f2fs/recovery.c   |  2 ++
>>>  fs/f2fs/segment.c    |  3 +++
>>>  6 files changed, 44 insertions(+), 14 deletions(-)
>>>
>>> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
>>> index 01e0d8f5bbbe..1ddf332ce4b2 100644
>>> --- a/fs/f2fs/checkpoint.c
>>> +++ b/fs/f2fs/checkpoint.c
>>> @@ -119,11 +119,8 @@ struct page *f2fs_get_meta_page_nofail(struct f2fs_sb_info *sbi, pgoff_t index)
>>>  		if (PTR_ERR(page) == -EIO &&
>>>  				++count <= DEFAULT_RETRY_IO_COUNT)
>>>  			goto retry;
>>> -
>>>  		f2fs_stop_checkpoint(sbi, false);
>>> -		f2fs_bug_on(sbi, 1);
>>>  	}
>>> -
>>>  	return page;
>>>  }
>>>  
>>> @@ -1512,7 +1509,10 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  	ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver);
>>>  
>>>  	/* write cached NAT/SIT entries to NAT/SIT area */
>>> -	f2fs_flush_nat_entries(sbi, cpc);
>>> +	err = f2fs_flush_nat_entries(sbi, cpc);
>>> +	if (err)
>>> +		goto stop;
>>
>> To make sure, if partial NAT pages become dirty, flush them back to device
>> outside checkpoint() is not harmful, right?
> 
> I think so.

Oh, one more case, now we use NAT #0, if partial NAT pages were
writebacked, then we start to use NAT #1, later, in another checkpoint(),
we update those NAT pages again, we will write them to NAT #0, which is
belong to last valid checkpoint(), it's may cause corrupted checkpoint in
scenario of SPO. So it's harmfull, right?

Thanks,

> 
>>
>>> +
>>>  	f2fs_flush_sit_entries(sbi, cpc);
>>>  
>>>  	/* unlock all the fs_lock[] in do_checkpoint() */
>>> @@ -1521,7 +1521,7 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  		f2fs_release_discard_addrs(sbi);
>>>  	else
>>>  		f2fs_clear_prefree_segments(sbi, cpc);
>>> -
>>> +stop:
>>>  	unblock_operations(sbi);
>>>  	stat_inc_cp_count(sbi->stat_info);
>>>  
>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>>> index a0c868127a7c..29021dbc8f2a 100644
>>> --- a/fs/f2fs/f2fs.h
>>> +++ b/fs/f2fs/f2fs.h
>>> @@ -2895,7 +2895,7 @@ int f2fs_recover_xattr_data(struct inode *inode, struct page *page);
>>>  int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page);
>>>  int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
>>>  			unsigned int segno, struct f2fs_summary_block *sum);
>>> -void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc);
>>> +int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc);
>>>  int f2fs_build_node_manager(struct f2fs_sb_info *sbi);
>>>  void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi);
>>>  int __init f2fs_create_node_manager_caches(void);
>>> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
>>> index 4bcc8a59fdef..c7d14282dbf4 100644
>>> --- a/fs/f2fs/gc.c
>>> +++ b/fs/f2fs/gc.c
>>> @@ -1070,6 +1070,15 @@ static int do_garbage_collect(struct f2fs_sb_info *sbi,
>>>  	/* reference all summary page */
>>>  	while (segno < end_segno) {
>>>  		sum_page = f2fs_get_sum_page(sbi, segno++);
>>> +		if (IS_ERR(sum_page)) {
>>> +			end_segno = segno - 1;
>>> +			for (segno = start_segno; segno < end_segno; segno++) {
>>> +				sum_page = find_get_page(META_MAPPING(sbi),
>>> +						GET_SUM_BLOCK(sbi, segno));
>>> +				f2fs_put_page(sum_page, 0);
>>
>> find_get_page() will add one more reference on page, so we need to call
>> f2fs_put_page(sum_page, 0) twice.
> 
> Done.
> 
>>
>> Thanks,
>>
>>> +			}
>>> +			return PTR_ERR(sum_page);
>>> +		}
>>>  		unlock_page(sum_page);
>>>  	}
>>>  
>>> diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
>>> index fa2381c0bc47..79b6fee354f7 100644
>>> --- a/fs/f2fs/node.c
>>> +++ b/fs/f2fs/node.c
>>> @@ -126,6 +126,8 @@ static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
>>>  
>>>  	/* get current nat block page with lock */
>>>  	src_page = get_current_nat_page(sbi, nid);
>>> +	if (IS_ERR(src_page))
>>> +		return src_page;
>>>  	dst_page = f2fs_grab_meta_page(sbi, dst_off);
>>>  	f2fs_bug_on(sbi, PageDirty(src_page));
>>>  
>>> @@ -2265,15 +2267,19 @@ static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
>>>  						nm_i->nat_block_bitmap)) {
>>>  			struct page *page = get_current_nat_page(sbi, nid);
>>>  
>>> -			ret = scan_nat_page(sbi, page, nid);
>>> -			f2fs_put_page(page, 1);
>>> +			if (IS_ERR(page)) {
>>> +				ret = PTR_ERR(page);
>>> +			} else {
>>> +				ret = scan_nat_page(sbi, page, nid);
>>> +				f2fs_put_page(page, 1);
>>> +			}
>>>  
>>>  			if (ret) {
>>>  				up_read(&nm_i->nat_tree_lock);
>>>  				f2fs_bug_on(sbi, !mount);
>>>  				f2fs_msg(sbi->sb, KERN_ERR,
>>>  					"NAT is corrupt, run fsck to fix it");
>>> -				return -EINVAL;
>>> +				return ret;
>>>  			}
>>>  		}
>>>  
>>> @@ -2700,7 +2706,7 @@ static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
>>>  		__clear_bit_le(nat_index, nm_i->full_nat_bits);
>>>  }
>>>  
>>> -static void __flush_nat_entry_set(struct f2fs_sb_info *sbi,
>>> +static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
>>>  		struct nat_entry_set *set, struct cp_control *cpc)
>>>  {
>>>  	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
>>> @@ -2724,6 +2730,9 @@ static void __flush_nat_entry_set(struct f2fs_sb_info *sbi,
>>>  		down_write(&curseg->journal_rwsem);
>>>  	} else {
>>>  		page = get_next_nat_page(sbi, start_nid);
>>> +		if (IS_ERR(page))
>>> +			return PTR_ERR(page);
>>> +
>>>  		nat_blk = page_address(page);
>>>  		f2fs_bug_on(sbi, !nat_blk);
>>>  	}
>>> @@ -2769,12 +2778,13 @@ static void __flush_nat_entry_set(struct f2fs_sb_info *sbi,
>>>  		radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
>>>  		kmem_cache_free(nat_entry_set_slab, set);
>>>  	}
>>> +	return 0;
>>>  }
>>>  
>>>  /*
>>>   * This function is called during the checkpointing process.
>>>   */
>>> -void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>> +int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  {
>>>  	struct f2fs_nm_info *nm_i = NM_I(sbi);
>>>  	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
>>> @@ -2784,6 +2794,7 @@ void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  	unsigned int found;
>>>  	nid_t set_idx = 0;
>>>  	LIST_HEAD(sets);
>>> +	int err = 0;
>>>  
>>>  	/* during unmount, let's flush nat_bits before checking dirty_nat_cnt */
>>>  	if (enabled_nat_bits(sbi, cpc)) {
>>> @@ -2793,7 +2804,7 @@ void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  	}
>>>  
>>>  	if (!nm_i->dirty_nat_cnt)
>>> -		return;
>>> +		return 0;
>>>  
>>>  	down_write(&nm_i->nat_tree_lock);
>>>  
>>> @@ -2816,11 +2827,16 @@ void f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>>  	}
>>>  
>>>  	/* flush dirty nats in nat entry set */
>>> -	list_for_each_entry_safe(set, tmp, &sets, set_list)
>>> -		__flush_nat_entry_set(sbi, set, cpc);
>>> +	list_for_each_entry_safe(set, tmp, &sets, set_list) {
>>> +		err = __flush_nat_entry_set(sbi, set, cpc);
>>> +		if (err)
>>> +			break;
>>> +	}
>>>  
>>>  	up_write(&nm_i->nat_tree_lock);
>>>  	/* Allow dirty nats by node block allocation in write_begin */
>>> +
>>> +	return err;
>>>  }
>>>  
>>>  static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
>>> diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
>>> index 56d34193a74b..ba678efe7cad 100644
>>> --- a/fs/f2fs/recovery.c
>>> +++ b/fs/f2fs/recovery.c
>>> @@ -355,6 +355,8 @@ static int check_index_in_prev_nodes(struct f2fs_sb_info *sbi,
>>>  	}
>>>  
>>>  	sum_page = f2fs_get_sum_page(sbi, segno);
>>> +	if (IS_ERR(sum_page))
>>> +		return PTR_ERR(sum_page);
>>>  	sum_node = (struct f2fs_summary_block *)page_address(sum_page);
>>>  	sum = sum_node->entries[blkoff];
>>>  	f2fs_put_page(sum_page, 1);
>>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
>>> index aa96a371aaf8..ab7386a7e9d3 100644
>>> --- a/fs/f2fs/segment.c
>>> +++ b/fs/f2fs/segment.c
>>> @@ -2487,6 +2487,7 @@ static void change_curseg(struct f2fs_sb_info *sbi, int type)
>>>  	__next_free_blkoff(sbi, curseg, 0);
>>>  
>>>  	sum_page = f2fs_get_sum_page(sbi, new_segno);
>>> +	f2fs_bug_on(sbi, IS_ERR(sum_page));
>>>  	sum_node = (struct f2fs_summary_block *)page_address(sum_page);
>>>  	memcpy(curseg->sum_blk, sum_node, SUM_ENTRY_SIZE);
>>>  	f2fs_put_page(sum_page, 1);
>>> @@ -3971,6 +3972,8 @@ static int build_sit_entries(struct f2fs_sb_info *sbi)
>>>  
>>>  			se = &sit_i->sentries[start];
>>>  			page = get_current_sit_page(sbi, start);
>>> +			if (IS_ERR(page))
>>> +				return PTR_ERR(page);
>>>  			sit_blk = (struct f2fs_sit_block *)page_address(page);
>>>  			sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, start)];
>>>  			f2fs_put_page(page, 1);
>>>
> 
> .
>