Re: [PATCH v2] null_blk: add support for max open/active zone limit for zoned devices

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2020/08/28 16:47, Klaus Jensen wrote:
> On Aug 28 07:36, Damien Le Moal wrote:
>> On 2020/08/28 16:23, Klaus Jensen wrote:
>>> On Aug 28 07:06, Damien Le Moal wrote:
>>>> On 2020/08/27 22:50, Niklas Cassel wrote:
>>>>> +static blk_status_t null_finish_zone(struct nullb_device *dev, struct blk_zone *zone)
>>>>> +{
>>>>> +	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
>>>>> +		return BLK_STS_IOERR;
>>>>> +
>>>>> +	switch (zone->cond) {
>>>>> +	case BLK_ZONE_COND_FULL:
>>>>> +		/* finish operation on full is not an error */
>>>>> +		return BLK_STS_OK;
>>>>> +	case BLK_ZONE_COND_EMPTY:
>>>>> +		if (!null_manage_zone_resources(dev, zone))
>>>>
>>>> OK. So you are hitting a fuzzy case here that is not actually well described in
>>>> the standards. That is, does finishing an empty zone necessarilly imply a
>>>> temporary transition through imp open ? Which you are assuming is a yes here.
>>>> Personally, I would say that is not necessary, but no strong feeling either way.
>>>>
>>>
>>> For ZNS, the spec is pretty clear that ZSE to ZSF is a legal direct
>>> transition. So I don't think the transition should be allowed to fail
>>> due to a lack of resources.
>>
>> I had a doubt and checked again ZBC & ZAC. I section 4.4.3.2.4 it says:
>>
>> The Zone Condition state machine (see 4.4.3.5) requires the specified zone to
>> have a Zone Condition of EXPLICITLY OPENED or IMPLICITLY OPENED before a finish
>> zone operation is performed. If a zone with a Zone Condition of EMPTY or CLOSED
>> is specified for a finish zone operation, prior to processing the finish zone
>> operation, then the Zone Condition state machine requires that:
>> a) a manage open zone resources operation (see 4.4.3.2.6) be performed; and
>> b) the Zone Condition becomes IMPLICITLY OPENED.
>>
>> And section 5.3 describing the zone finish command points to this section.
>> So this is not the same as ZNS.
>>
>> As Niklas mentioned, nullblk tends to follow more ZBC than ZNS, so the code is
>> correct in this respect. We could also lean toward ZNS on this one. I personally
>>  have no strong opinion either way since there is not real good reasons for
>> finishing an empty zone that I can think of.
>>
>>
> 
> Alrighty then; thanks for looking it up! I won't fight for the ZNS
> behavior then :)

If it becomes a problem, we could add another option to force one way or the
other, something like "zone_specs=zbc" or "zone_specs=zns". That actually may
prove useful for testing file systems etc.


-- 
Damien Le Moal
Western Digital Research




[Index of Archives]     [Linux RAID]     [Linux SCSI]     [Linux ATA RAID]     [IDE]     [Linux Wireless]     [Linux Kernel]     [ATH6KL]     [Linux Bluetooth]     [Linux Netdev]     [Kernel Newbies]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Device Mapper]

  Powered by Linux