Re: [RFC 4/4] nvme: enable logical block size > PAGE_SIZE

[Pankaj Raghav]

On 2023-06-21 11:07, Hannes Reinecke wrote:
> On 6/21/23 10:38, Pankaj Raghav wrote:
>> Don't set the capacity to zero for when logical block size > PAGE_SIZE
>> as the block device with iomap aops support allocating block cache with
>> a minimum folio order.
>>
>> Signed-off-by: Pankaj Raghav <p.raghav@xxxxxxxxxxx>
>> ---
>>   drivers/nvme/host/core.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
>> index 98bfb3d9c22a..36cf610f938c 100644
>> --- a/drivers/nvme/host/core.c
>> +++ b/drivers/nvme/host/core.c
>> @@ -1886,7 +1886,7 @@ static void nvme_update_disk_info(struct gendisk *disk,
>>        * The block layer can't support LBA sizes larger than the page size
>>        * yet, so catch this early and don't allow block I/O.
>>        */
>> -    if (ns->lba_shift > PAGE_SHIFT) {
>> +    if ((ns->lba_shift > PAGE_SHIFT) && IS_ENABLED(CONFIG_BUFFER_HEAD)) {
>>           capacity = 0;
>>           bs = (1 << 9);
>>       }
> Again, I can't see why this would be contingent on CONFIG_BUFFER_HEAD.
> I'll be rebasing my patchset on your mapping_set_orders() patches and repost.
> 

As I explained in the previous email, I get a BUG from buffer.c when I don't make it conditional.
The hope is when we move to iomap based aops for the block cache, we can just get rid of
`if (ns->lba_shift > PAGE_SHIFT)` block and no dependence on CONFIG_BUFFER_HEAD.

> Cheers,
> 
> Hannes
>