On 11/06/2020 15:40, David Sterba wrote:
> On Wed, Jun 10, 2020 at 09:32:49PM +0900, Johannes Thumshirn wrote:
>> Introduce a table holding the paramenters for chunk allocation per RAID
>> profile.
>>
>> Also convert all assignments of hardcoded numbers to table lookups in this
>> process. Further changes will reduce code duplication even more.
>>
>> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@xxxxxxx>
>> ---
>> volumes.c | 95 +++++++++++++++++++++++++++++++++++++++++++++----------
>> 1 file changed, 79 insertions(+), 16 deletions(-)
>>
>> diff --git a/volumes.c b/volumes.c
>> index 04bc3d19a025..fc14283db2bb 100644
>> --- a/volumes.c
>> +++ b/volumes.c
>> @@ -1005,6 +1005,68 @@ error:
>> - 2 * sizeof(struct btrfs_chunk)) \
>> / sizeof(struct btrfs_stripe) + 1)
>>
>> +static const struct btrfs_raid_profile {
>> + int num_stripes;
>> + int max_stripes;
>> + int min_stripes;
>> + int sub_stripes;
>> +} btrfs_raid_profile_table[BTRFS_NR_RAID_TYPES] = {
>> + [BTRFS_RAID_RAID10] = {
>> + .num_stripes = 0,
>> + .max_stripes = 0,
>> + .min_stripes = 4,
>> + .sub_stripes = 2,
>> + },
> ...
>
> This duplicates btrfs_raid_array values, the member variable names don't
> match so this makes it hard to compare. I wouldn't mind to create this
> table as an intermediate step to clean up the code but in the end we
> really don't want to keep btrfs_raid_profile, in the same file even.
>
Yes I got confused in the middle and opted for this intermediate step.
I hope I can reduce it even more and have it slowly converged to the
kernel's implementation.
