Re: [PATCH 2/5] zram: partial IO refactoring

[Mika PenttilÃ]

On 04/03/2017 09:12 AM, Minchan Kim wrote:
> Hi Mika,
> 
> On Mon, Apr 03, 2017 at 08:52:33AM +0300, Mika Penttilä wrote:
>>
>> Hi!
>>
>> On 04/03/2017 08:17 AM, Minchan Kim wrote:
>>> For architecture(PAGE_SIZE > 4K), zram have supported partial IO.
>>> However, the mixed code for handling normal/partial IO is too mess,
>>> error-prone to modify IO handler functions with upcoming feature
>>> so this patch aims for cleaning up zram's IO handling functions.
>>>
>>> Signed-off-by: Minchan Kim <minchan@xxxxxxxxxx>
>>> ---
>>>  drivers/block/zram/zram_drv.c | 333 +++++++++++++++++++++++-------------------
>>>  1 file changed, 184 insertions(+), 149 deletions(-)
>>>
>>> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
>>> index 28c2836f8c96..7938f4b98b01 100644
>>> --- a/drivers/block/zram/zram_drv.c
>>> +++ b/drivers/block/zram/zram_drv.c
>>> @@ -45,6 +45,8 @@ static const char *default_compressor = "lzo";
>>>  /* Module params (documentation at end) */
>>>  static unsigned int num_devices = 1;
>>>  
>>> +static void zram_free_page(struct zram *zram, size_t index);
>>> +
>>>  static inline bool init_done(struct zram *zram)
>>>  {
>>>  	return zram->disksize;
>>> @@ -98,10 +100,17 @@ static void zram_set_obj_size(struct zram_meta *meta,
>>>  	meta->table[index].value = (flags << ZRAM_FLAG_SHIFT) | size;
>>>  }
>>>  
>>> +#if PAGE_SIZE != 4096
>>>  static inline bool is_partial_io(struct bio_vec *bvec)
>>>  {
>>>  	return bvec->bv_len != PAGE_SIZE;
>>>  }
>>> +#else
>>
>> For page size of 4096 bv_len can still be < 4096 and partial pages should be supported 
>> (uncompress before write etc). ? 
> 
> zram declares this.
> 
>         #define ZRAM_LOGICAL_BLOCK_SIZE (1<<12)
> 
> 	blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE);
> 	blk_queue_logical_block_size(zram->disk->queue,
> 					ZRAM_LOGICAL_BLOCK_SIZE);
> 
> So, I thought there is no such partial IO in 4096 page architecture.
> Am I missing something? Could you tell the scenario if it happens?

I think you're right. At least swap operates with min 4096 sizes.

> 
> Thanks!
>