On 3/17/22 00:43, Pankaj Raghav wrote:
> Hi Naohiro,
>
> On 2022-03-16 05:07, Naohiro Aota wrote:
>>> +# blkzone supports printing zone cap only after version 2.37
>>> +_blkzone_req_major_ver=2
>>> +_blkzone_req_minor_ver=37
>>> +blkzone_ver=$($BLKZONE_PROG --version | cut -d " " -f 4 | cut -d "-" -f 1)
>>> +blkzone_ver=( ${blkzone_ver//./ } )
>>> +blkzone_major_ver=${blkzone_ver[0]}
>>> +blkzone_minor_ver=${blkzone_ver[1]}
>>> +if [[ $blkzone_major_ver -ge $_blkzone_req_major_ver &&\
>>> + $blkzone_minor_ver -ge $_blkzone_req_minor_ver ]]; then
>>
>> I'd like to have it as a helper.
>>
>> Or, how about adding a filter and a wrapper function to add a "cap"
>> filed whose value equal to the zone size for the old version (like
>> with _getcap() and _filter_getcap())? Then, we can just call:
>> zonecap=$(_blkzone_report ... | grep -Po "cap ([0x\d]+)+" | cut -d ' ' -f 2)
>> size=$((zonecap << 9))
>>
> That is a good idea. I will give it a shot. Thanks.
>
>>> + zonecap=$($BLKZONE_PROG report -c 1 $SCRATCH_DEV |\
>>> + grep -Po "cap ([0x\d]+)+" | cut -d ' ' -f 2)
>>
>> But still, zone capacity can vary for each zone. So, we need to read
>> the capacity of a zone where the data BG resides on.
>>
> Do we support variable zone capacity in Linux? IIRC variable zone sizes
> are definitely not supported but I am not sure about variable zone capacity.
No, variable zone capacity is not supported in Linux. By that, I mean that
drive that may change the capacity of a zone after a reset are not
supported. So a zone capacity in Linux is always fixed throughout the life
time of the NS it belongs to.
But nothing mandates that all zones have the same capacity. A drive could
expose zones with different capacities. That is the point Naohiro was making.
>
> But even if we do support, I see that the zone 5 (old_data_zone) and
> zone 6 (new_data_zone) during the test and what if the new_data_zone
> (zone 6) has a smaller cap than old_data_zone (zone 5)?
> The main question: Is there a way to deterministically tell where the
> data BG will reside and where it will relocate before we start the test
> with variable capacity?
>
> My first look indicates that adding variable zone capacity will make the
> test a bit more complex and I am not sure if it is worth the effort if
> there are no use cases for it.
> Let me know your thoughts.
>
--
Damien Le Moal
Western Digital Research
