Hi Eric,
On 2019/6/26 1:52, Eric Biggers wrote:
> Hi Chao, thanks for the review.
>
> On Tue, Jun 25, 2019 at 03:55:57PM +0800, Chao Yu wrote:
>> Hi Eric,
>>
>> On 2019/6/21 4:50, Eric Biggers wrote:
>>> +static int f2fs_begin_enable_verity(struct file *filp)
>>> +{
>>> + struct inode *inode = file_inode(filp);
>>> + int err;
>>> +
>>
>> I think we'd better add condition here (under inode lock) to disallow enabling
>> verity on atomic/volatile inode, as we may fail to write merkle tree data due to
>> atomic/volatile inode's special writeback method.
>>
>
> Yes, I'll add the following:
>
> if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode))
> return -EOPNOTSUPP;
>
>>> + err = f2fs_convert_inline_inode(inode);
>>> + if (err)
>>> + return err;
>>> +
>>> + err = dquot_initialize(inode);
>>> + if (err)
>>> + return err;
>>
>> We can get rid of dquot_initialize() here, since f2fs_file_open() ->
>> dquot_file_open() should has initialized quota entry previously, right?
>
> We still need it because dquot_file_open() only calls dquot_initialize() if the
> file is being opened for writing. But here the file descriptor is readonly.
> I'll add a comment explaining this here and in the ext4 equivalent.
Ah, you're right.
f2fs_convert_inline_inode() may grab one more block during conversion, so we
need to call dquot_initialize() before inline conversion?
Thanks,
>
> - Eric
> .
>
