Re: [PATCH 2/4] introduce four macros for in-kernel hints

["Javier González" <javier@javigon.com>]

[-- Attachment #1: Type: text/plain, Size: 2345 bytes --]


> On 24 Jan 2019, at 09.35, Jan Kara <jack@suse.cz> wrote:
> 
> On Wed 23-01-19 19:27:12, Javier González wrote:
>>> On 9 Jan 2019, at 16.30, Kanchan Joshi <joshi.k@samsung.com> wrote:
>>> 
>>> Exiting write-hints are exposed to user-mode. There is a possiblity
>>> of conflict if kernel happens to use those. This patch introduces four
>>> write-hints for exclusive kernel-mode use.
>>> 
>>> Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
>>> ---
>>> include/linux/fs.h | 5 +++++
>>> 1 file changed, 5 insertions(+)
>>> 
>>> diff --git a/include/linux/fs.h b/include/linux/fs.h
>>> index 811c777..e8548eb 100644
>>> --- a/include/linux/fs.h
>>> +++ b/include/linux/fs.h
>>> @@ -291,6 +291,11 @@ enum rw_hint {
>>> 	WRITE_LIFE_MEDIUM	= RWH_WRITE_LIFE_MEDIUM,
>>> 	WRITE_LIFE_LONG		= RWH_WRITE_LIFE_LONG,
>>> 	WRITE_LIFE_EXTREME	= RWH_WRITE_LIFE_EXTREME,
>>> +/* below ones are meant for in-kernel use */
>>> +	KERN_WRITE_LIFE_SHORT,
>>> +	KERN_WRITE_LIFE_MEDIUM,
>>> +	KERN_WRITE_LIFE_LONG,
>>> +	KERN_WRITE_LIFE_EXTREME
>>> };
>> 
>> I think Jens and Dave meant kernel hints to go top down. This would also
>> give space for supporting more hints / streams from both ends for user
>> and kernel.
> 
> Yes, that was the idea however if I understand it right, the write hints do
> not really have to be consistent boot-to-boot since they aren't stored
> persistently by the disk, are they? If that's the case, it doesn't really
> matter which numbers we pick.
> 

I guess this is implementation specific. Some times the drive will want
to store this to improve GC. For the current "coldness" hint I does not
matter much, but if the hint were to express other metric it can become
relevant.

Anyway, the comment was more to separate user / kernel hints and allow
them to grow from the ends.


> One thing I don't quite like is the naming of KERN_WRITE_LIFE_SHORT etc.. It
> is upto filesystem to assign meanings to the write hints. So I think it is
> enough to provide something like KERN_WRITE_HINT_MIN which is the first
> hint available to the kernel and then the number of hints available to the
> kernel.
> 

Makes sense to me. Then we can rename the hint for each FS to give it
proper mening.

> 								Honza
> 
> --
> Jan Kara <jack@suse.com>
> SUSE Labs, CR

Javier

[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 833 bytes --]