Re: [PATCH v2] f2fs: fix m_may_create to make OPU DIO write correctly

[Chao Yu <yuchao0@huawei.com>]

Ping,

On 2018/11/20 19:58, Chao Yu wrote:
> On 2018-11-20 4:29, Jia Zhu wrote:
>> Previously, we added a parameter @map.m_may_create to trigger OPU
>> allocation and call f2fs_balance_fs() correctly.
>>
>> But in get_more_blocks(), @create has been overwritten by below code.
>> So the function f2fs_map_blocks() will not allocate new block address
>> but directly go out. Meanwile,there are several functions calling
>> f2fs_map_blocks() directly and @map.m_may_create not initialized.
> 
> Oh, I missed to check all f2fs_map_blocks structure referrers, sorry.
> 
>> CODE:
>> create = dio->op == REQ_OP_WRITE;
>> 	if (dio->flags & DIO_SKIP_HOLES) {
>> 		if (fs_startblk <= ((i_size_read(dio->inode) - 1) >>
>> 						i_blkbits))
>> 			create = 0;
>> 	}
>>
>> This patch fixes it.
>>
>> Signed-off-by: Jia Zhu <zhujia13@huawei.com>> ---
> 
> It will be better to add simple change logs here to indicate how you modify your
> patch comparing to previous one, please keep that rule for your next patch. ;)
> 
> Reviewed-by: Chao Yu <yuchao0@huawei.com>
> 
> Thanks,
> 
>>  fs/f2fs/data.c | 5 +++++
>>  fs/f2fs/file.c | 4 +++-
>>  2 files changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>> index aa8843a..7226300 100644
>> --- a/fs/f2fs/data.c
>> +++ b/fs/f2fs/data.c
>> @@ -1052,6 +1052,10 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
>>  	end = pgofs + maxblocks;
>>  
>>  	if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) {
>> +		if (test_opt(sbi, LFS) && flag == F2FS_GET_BLOCK_DIO &&
>> +							map->m_may_create)
>> +			goto next_dnode;
>> +
>>  		map->m_pblk = ei.blk + pgofs - ei.fofs;
>>  		map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs);
>>  		map->m_flags = F2FS_MAP_MAPPED;
>> @@ -1261,6 +1265,7 @@ bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
>>  	map.m_next_pgofs = NULL;
>>  	map.m_next_extent = NULL;
>>  	map.m_seg_type = NO_CHECK_TYPE;
>> +	map.m_may_create = false;
>>  	last_lblk = F2FS_BLK_ALIGN(pos + len);
>>  
>>  	while (map.m_lblk < last_lblk) {
>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>> index 3271830..ff82350 100644
>> --- a/fs/f2fs/file.c
>> +++ b/fs/f2fs/file.c
>> @@ -2201,7 +2201,8 @@ static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
>>  {
>>  	struct inode *inode = file_inode(filp);
>>  	struct f2fs_map_blocks map = { .m_next_extent = NULL,
>> -					.m_seg_type = NO_CHECK_TYPE };
>> +					.m_seg_type = NO_CHECK_TYPE ,
>> +					.m_may_create = false };
>>  	struct extent_info ei = {0, 0, 0};
>>  	pgoff_t pg_start, pg_end, next_pgofs;
>>  	unsigned int blk_per_seg = sbi->blocks_per_seg;
>> @@ -2935,6 +2936,7 @@ int f2fs_precache_extents(struct inode *inode)
>>  	map.m_next_pgofs = NULL;
>>  	map.m_next_extent = &m_next_extent;
>>  	map.m_seg_type = NO_CHECK_TYPE;
>> +	map.m_may_create = false;
>>  	end = F2FS_I_SB(inode)->max_file_blocks;
>>  
>>  	while (map.m_lblk < end) {
>>
> 
> .
>