Re: [PATCH 0/2] add columns for zoned parameters

[Damien Le Moal <Damien.LeMoal@wdc.com>]

On 2021/08/25 11:12, Naohiro Aota wrote:
> On Tue, Aug 24, 2021 at 02:47:29PM +0200, Karel Zak wrote:
>> On Tue, Aug 24, 2021 at 10:17:16AM +0900, Naohiro Aota wrote:
>>> Several parameters for zoned devices are missing from lsblk's columns. This
>>> series introduces them as following.
>>
>> Thanks! I thought about it a few weeks ago ;-)
>>
>>>
>>>  Column Name      Description             Sysfs path
>>>  ---------------------------------------------------------------------
>>>  ZONE-SIZE        zone size               queue/chunk_sectors
>>>  ZONE-WGRAN       zone write granularity  queue/zone_write_granularity
>>>  ZONE-APPEND      zone append max bytes   queue/zone_append_max_bytes
>>>  ZONES-NR         number of zones         queue/nr_zones
>>>  ZONES-OMAX       max open zones          queue/max_open_zones
>>>  ZONES-AMAX       max active zones        queue/max_active_zones
>>>
>>> Sample output:
>>>
>>> $ lsblk -o NAME,ZONED,ZONE-SIZE,ZONE-WGRAN,ZONE-APPEND,ZONES-NR,ZONES-OMAX,ZONES-AMAX -i
>>
>> What about to add all the columns to "--zoned" output too? 
>>
>> We do not keep backward compatibility for set of columns in outputs
>> like --zoned, so it's open for changes. I guess we can use --zoned as
>> a generic option for all zoned stuff and not for a model only.
> 
> Nice idea. I'll add a patch for that.

+1

> 
>>> NAME        ZONED        ZONE-SIZE ZONE-WGRAN ZONE-APPEND ZONES-NR ZONES-OMAX ZONES-AMAX
>>> sda         host-managed      256M         4K        672K    55880        128          0
>>> sdb         host-managed      256M         4K        672K    55880        128          0
>>> zram0       none                0B         0B          0B        0          0          0
>>> nvme2n1     none                0B         0B          0B        0          0          0
>>> |-nvme2n1p1 none                0B         0B          0B        0          0          0
>>> |-nvme2n1p2 none                0B         0B          0B        0          0          0
>>> `-nvme2n1p3 none                0B         0B          0B        0          0          0
>>> nvme0n1     none                0B         0B          0B        0          0          0
>>> nvme1n1     none                0B         0B          0B        0          0          0
>>> nvme0n2     host-managed        2G         4K          4M     1844         14         14
>>> nvme1n2     host-managed        2G         4K          4M     1844         14         14
>>>
>>> $ lsblk --help|grep -i zone
>>>  -z, --zoned          print zone model
>>>         ZONED  zone model
>>>     ZONE-SIZE  zone size
>>>    ZONE-WGRAN  zone write granularity
>>>   ZONE-APPEND  zone append max bytes
>>>      ZONES-NR  number of zones
>>>    ZONES-OMAX  max open zones
>>>    ZONES-AMAX  max active zones
>>
>> It would be nice to have shorter column names, but I do not have any
>> sane suggestion (Z-SIZE or Z-APPEND seems strange).
> 
> Hmm, considering we already have DISC-{ALN,GRAN,MAX,ZERO}, how about
> ZONE-{SZ,GRAN,APP,NR,OMAX,AMAX}?

That work for me. One nit though: having the ZONE-NR and ZONE-SZ columns come
first would be nice since these are the 2 most important properties that users
will look at before digging into details of other limits.

I also wonder if we want to add max_zone_append_bytes here... We should only if
attributes such as max_sectors etc are listed too (I have not checked).


> 
>>   Karel
>>
>> -- 
>>  Karel Zak  <kzak@redhat.com>
>>  http://karelzak.blogspot.com
>>


-- 
Damien Le Moal
Western Digital Research