[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-block
Subject:    Re: [PATCH linux-4.7-rc7] blk_stack_limits() setting discard_granularity
From:       Florian-Ewald_Müller  <florian-ewald.mueller () profitbricks ! com
Date:       2016-07-29 7:59:54
Message-ID: CAHJVUeik5+3mYSFqaONu50XF9sV5+tMpTXqM-Kcijzwj+f6vDg () mail ! gmail ! com
[Download RAW message or body]

Hi Martin,

I'm trying to solve this "squaring the circle" problem:
- my distributed block device "distributes" the data with the
configured block size (e.g. 64K) to different of the many underlying
slave block devices (here: dm-cache).
- my block device evaluates the discard for logically discarding the
specified block (it doesn't have more than 64K in a row) for some
specific (logical) slave device.
- that's why, I have to consider splitting the discard myself (if the
block layer doesn't do it) for treating each 64K chunk on different
(logical) slave devices.
- on the underlying block device (e.g. dm-cache) the data is placed
also in 64K chunks (there is not more in row).
- that's why, ignoring this size and sending the original big size to
it will probably delete unrelated data.

Now my experiments show that, at least, dm-cache doesn't complain nor
rejects those smaller discards than its discard_granularity, but
possibly turning them into noop (?).
May be that the needed functionality of accumulating small discards to
a big one covering its own granularity (similar to SSDs block erasure)
should be done at that driver level.

What do you think ?

Florian

On Fri, Jul 29, 2016 at 3:39 AM, Martin K. Petersen
<martin.petersen@oracle.com> wrote:
>>>>>> "Florian-Ewald" == Florian-Ewald Müller <florian-ewald.mueller@profitbricks.com> writes:
>
> Florian-Ewald,
>
>> My (new developed) distributed block device uses a block size of 64K.
>
>> So, because of the distribution functionality to different underlying
>> slave (dm-cache) block devices, it has:
>> - max_discard_sectors: 128 (64K)
>> - discard_granularity: 65536 (64K)
>
>> The discard sizes of the (underlying) dm-cache devices are:
>> - max_discard_sectors: 536870912 (256G)
>> - discard_granularity: 268435456 (256M)
>
> Why do you set max_discard_sectors for your device to 128? To my
> knowledge dm-cache does not do partial allocation. And consequently any
> discard less than an aligned 256MB will be a nop anyway.
>
> --
> Martin K. Petersen      Oracle Linux Engineering



-- 
Florian-Ewald Mueller

Architecture Board


ProfitBricks GmbH
Greifswalder Str. 207
D - 10405 Berlin


Tel:       +49 30 577 008 331
Fax:      +49 30 577 008 598
Email:   florian-ewald.mueller@profitbricks.com
URL:     http://www.profitbricks.de

Sitz der Gesellschaft: Berlin.
Registergericht: Amtsgericht Charlottenburg, HRB 125506 B.
Geschäftsführer: Andreas Gauger, Achim Weiss.

Please consider the environment before printing this email.
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
[prev in list] [next in list] [prev in thread] [next in thread]