On 5/17/23 21:56, Coly Li wrote:
>
>
>> 2023年5月17日 11:51，Lin Feng <linf@xxxxxxxxxx> 写道：
>>
>> In fact due to this check in cache_alloc:
>> free = roundup_pow_of_two(ca->sb.nbuckets) >> 10;
>
> This was introduced from commit 78365411b344d (“bcache: Rework allocator reserves”).
> - free = roundup_pow_of_two(ca->sb.nbuckets) >> 9;
> - free = max_t(size_t, free, (prio_buckets(ca) + 8) * 2);
> + free = roundup_pow_of_two(ca->sb.nbuckets) >> 10;
>
>> if (!free) {
>> ret = -EPERM;
>> err = "ca->sb.nbuckets is too small";
>> goto err_free;
>> }
>> we can only create bcache device with nbuckets greater than 512,
>> so this patch is to make the codes logic consistent.
>>
>> Signed-off-by: Lin Feng <linf@xxxxxxxxxx>
>> ---
>> drivers/md/bcache/super.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
>> index 7e9d19fd21dd..681a7ea442b9 100644
>> --- a/drivers/md/bcache/super.c
>> +++ b/drivers/md/bcache/super.c
>> @@ -110,7 +110,7 @@ static const char *read_super_common(struct cache_sb *sb, struct block_device *
>> goto err;
>>
>> err = "Not enough buckets";
>> - if (sb->nbuckets < 1 << 7)
>> + if (sb->nbuckets <= 1 << 9)
>
> if (roundup_pow_of_two(ca->sb.nbuckets) < 1<< 10)
>
> Might be more clear?
Yes, I also prefer your change :)
>
>> goto err;
>>
>> err = "Bad block size (not power of 2)";
>
>
>
>
> In bcache-tools the minimum nbuckets is (1<<7) too. It is fine to modify it to 1<<10, I will modify bcache-tools too.
>
bcache-tools also has this flaw, both are aligned to 1<<10 is a good choice!
Thanks,
linfeng
