[Ocfs2-tools-devel] [RFC 2/8] dx_dirs v4: indexed dirs core code in libocfs2

[Tao Ma]

Coly Li wrote:
> 
> On 2010年01月22日 16:21, Tao Ma Wrote:
>>
>> Coly Li wrote:
>>> This patch contains the core part of current indexed dirs support in
>>> libocfs2, includes,
>>> - Indexed tree truncate.
>>> - Build indexed tree.
>>> - Insert dx record into indexed tree.
>>> - Expand inlined dx_root to an extent tree.
>>> - Rebalance indexed tree.
>>>
>>> I worry about whether I use the extent tree code correctly, though I
>>> tried my best to understand the code, IMHO it's
>>> still a gray area to me so far. Any review/comment is helpful !
>>>
>>> Signed-off-by: Coly Li <coly.li at suse.de>
>>> Cc: Tao Ma <tao.ma at oracle.com>
>>> ---
> [snip]
>>> --- a/libocfs2/inode.c
>>> +++ b/libocfs2/inode.c
>>> @@ -139,6 +139,10 @@ static void ocfs2_swap_inode_second(struct
>>> ocfs2_dinode *di)
>>>          sb->s_uuid_hash           = bswap_32(sb->s_uuid_hash);
>>>          sb->s_first_cluster_group = bswap_64(sb->s_first_cluster_group);
>>>          sb->s_xattr_inline_size   = bswap_16(sb->s_xattr_inline_size);
>>> +        sb->s_dx_seed[0]          = bswap_32(sb->s_dx_seed[0]);
>>> +        sb->s_dx_seed[1]          = bswap_32(sb->s_dx_seed[1]);
>>> +        sb->s_dx_seed[2]          = bswap_32(sb->s_dx_seed[2]);
>>> +        sb->s_dx_seed[3]          = bswap_32(sb->s_dx_seed[3]);
>> I just noticed that in ocfs2_super_block we define
>>     __le32 s_dx_seed[3];
>> So swap s_dx_seed[3] is wrong?
> 
> Why do you think it's wrong ?
er, __le32 s_dx_seed[3] means the array has only three members which is 
0,1,2. ;)

Regards,
Tao