2012-07-19 14:51:52

by Wang Sheng-Hui

[permalink] [raw]
Subject: One question about ext4_fallocate

Dear all,

I have one question about ext4_fallocate:
For its parameters, offset & len, it will preallocate
one space with length >= len, right?

>From the code:
map.m_lblk = offset >> blkbits;
/*
* We can't just convert len to max_blocks because
* If blocksize = 4096 offset = 3072 and len = 2048
*/
max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
- map.m_lblk;

We can see that the target space is enlarged to the (floor, ceiling) of
(offset, offset+len). Right?

If I'm wrong, please correct me.


Regards,


2012-07-19 15:42:18

by Andreas Dilger

[permalink] [raw]
Subject: Re: One question about ext4_fallocate


On 2012-07-19, at 7:51 AM, Wang Sheng-Hui wrote:
> I have one question about ext4_fallocate:
> For its parameters, offset & len, it will preallocate
> one space with length >= len, right?
>
> From the code:
> map.m_lblk = offset >> blkbits;
> /*
> * We can't just convert len to max_blocks because
> * If blocksize = 4096 offset = 3072 and len = 2048
> */
> max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
> - map.m_lblk;
>
> We can see that the target space is enlarged to the (floor, ceiling) of
> (offset, offset+len). Right?
>
> If I'm wrong, please correct me.

A whole block has to be allocated for fallocate, but if KEEP_SIZE is not used, the i_size should be changed to match the actual requested size.

Cheers, Andreas






2012-07-20 06:39:41

by Wang Sheng-Hui

[permalink] [raw]
Subject: Re: One question about ext4_fallocate

On 2012年07月19日 23:42, Andreas Dilger wrote:
>
> On 2012-07-19, at 7:51 AM, Wang Sheng-Hui wrote:
>> I have one question about ext4_fallocate:
>> For its parameters, offset & len, it will preallocate
>> one space with length >= len, right?
>>
>> From the code:
>> map.m_lblk = offset >> blkbits;
>> /*
>> * We can't just convert len to max_blocks because
>> * If blocksize = 4096 offset = 3072 and len = 2048
>> */
>> max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
>> - map.m_lblk;
>>
>> We can see that the target space is enlarged to the (floor, ceiling) of
>> (offset, offset+len). Right?
>>
>> If I'm wrong, please correct me.
>
> A whole block has to be allocated for fallocate, but if KEEP_SIZE is not used, the i_size should be changed to match the actual requested size.
>

Got it.

From the code:
/* Return error if mode is not supported */
if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
return -EOPNOTSUPP;

if (mode & FALLOC_FL_PUNCH_HOLE)
return ext4_punch_hole(file, offset, len);

ext4 only supports FALLOC_FL_KEEP_SIZE & FALLOC_FL_PUNCH_HOLE currently.
For FALLOC_FL_PUNCH_HOLE, it will turn to ext4_punch_hole, so the rest
path deals with only FALLOC_FL_KEEP_SIZE and thus it will use whole blocks.

Thanks, Andreas.
> Cheers, Andreas
>
>
>
>
>