Re: [PATCH] zram: support REQ_DISCARD

[Jerome Marchand <jmarchan@redhat.com>]

On 02/24/2014 04:56 PM, Joonsoo Kim wrote:
> 2014-02-25 0:15 GMT+09:00 Jerome Marchand <jmarchan@redhat.com>:
>> On 02/24/2014 04:02 PM, Joonsoo Kim wrote:
>>> 2014-02-24 22:36 GMT+09:00 Jerome Marchand <jmarchan@redhat.com>:
>>>> On 02/24/2014 06:51 AM, Joonsoo Kim wrote:
>>>>> zram is ram based block device and can be used by backend of filesystem.
>>>>> When filesystem deletes a file, it normally doesn't do anything on data
>>>>> block of that file. It just marks on metadata of that file. This behavior
>>>>> has no problem on disk based block device, but has problems on ram based
>>>>> block device, since we can't free memory used for data block. To overcome
>>>>> this disadvantage, there is REQ_DISCARD functionality. If block device
>>>>> support REQ_DISCARD and filesystem is mounted with discard option,
>>>>> filesystem sends REQ_DISCARD to block device whenever some data blocks are
>>>>> discarded. All we have to do is to handle this request.
>>>>>
>>>>> This patch implements to flag up QUEUE_FLAG_DISCARD and handle this
>>>>> REQ_DISCARD request. With it, we can free memory used by zram if it isn't
>>>>> used.
>>>>>
>>>>> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
>>>>> ---
>>>>> This patch is based on master branch of linux-next tree.
>>>>>
>>>>> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
>>>>> index 5ec61be..cff2c0e 100644
>>>>> --- a/drivers/block/zram/zram_drv.c
>>>>> +++ b/drivers/block/zram/zram_drv.c
>>>>> @@ -501,6 +501,20 @@ static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index,
>>>>>       return ret;
>>>>>  }
>>>>>
>>>>> +static void zram_bio_discard(struct zram *zram, struct bio *bio)
>>>>> +{
>>>>> +     u32 index = bio->bi_iter.bi_sector >> SECTORS_PER_PAGE_SHIFT;
>>>>
>>>> Hi Joonsoo,
>>>>
>>>> If bi_sector is not aligned on a page size, we might end up discarding
>>>> a page that still contain valid data.
>>>>
>>>>
>>>
>>> Hello, Jerome.
>>>
>>> Is it possible that request isn't aligned on a page size if
>>> logical/physical block size
>>> is PAGE_SIZE?
>>
>> Yes, zram has an logical block size of 4k (ZRAM_LOGICAL_BLOCK_SIZE),
>> while its physical block size, which is a page size, can be bigger.
>>
>>> When I tested it, I didn't find any invalid io.
>>> If we meet any misaligned request, it would be filtered by
>>> valid_io_request(). :)
>>
>> zram accepts request aligned on logical blocks. So valid_io_request()
>> wouldn't filter misaligned requests out as long as they are aligned
>> on logical blocks.
>> If your system use 4k pages, your tests would never trigger the issue,
>> but on a system which uses 64k pages, it could.
> 
> Okay. I got it.
> So, how about using PAGE_SIZE as ZRAM_LOGICAL_BLOCK_SIZE?
> Is there any reason to set 4096 to ZRAM_LOGICAL_BLOCK_SIZE,
> instead of setting PAGE_SIZE to ZRAM_LOGICAL_BLOCK_SIZE?
> 

ZRAM_LOGICAL_BLOCK_SIZE was introduced in commit 7b19b8d because the
block layer couldn't handle 64k logical block. Also, some filesytems
(including FAT IRC), can't cope with 64k block either.


> Thanks.
>