[PATCHv3 3/5] nvme: implement I/O Command Sets Command Set support

[Sagi Grimberg]

On 6/23/20 4:25 AM, Niklas Cassel wrote:
> On Tue, Jun 23, 2020 at 01:53:47AM -0700, Sagi Grimberg wrote:
>>
>>
>> On 6/22/20 9:25 AM, Keith Busch wrote:
>>> From: Niklas Cassel <niklas.cassel at wdc.com>
>>>
>>> Implements support for the I/O Command Sets command set. The command set
>>> introduces a method to enumerate multiple command sets per namespace. If
>>> the command set is exposed, this method for enumeration will be used
>>> instead of the traditional method that uses the CC.CSS register command
>>> set register for command set identification.
>>>
>>> For namespaces where the Command Set Identifier is not supported or
>>> recognized, the specific namespace will not be created.
>>>
>>> Reviewed-by: Javier González <javier.gonz at samsung.com>
>>> Reviewed-by: Martin K. Petersen <martin.petersen at oracle.com>
>>> Reviewed-by: Johannes Thumshirn <johannes.thumshirn at wdc.com>
>>> Reviewed-by: Matias Bjørling <matias.bjorling at wdc.com>
>>> Reviewed-by: Daniel Wagner <dwagner at suse.de>
>>> Signed-off-by: Niklas Cassel <niklas.cassel at wdc.com>
>>> ---
>>>    drivers/nvme/host/core.c | 48 +++++++++++++++++++++++++++++++++-------
>>>    drivers/nvme/host/nvme.h |  1 +
>>>    include/linux/nvme.h     | 19 ++++++++++++++--
>>>    3 files changed, 58 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
>>> index 9491dbcfe81a..45a3cb5a35bd 100644
>>> --- a/drivers/nvme/host/core.c
>>> +++ b/drivers/nvme/host/core.c
>>> @@ -1056,8 +1056,13 @@ static int nvme_identify_ctrl(struct nvme_ctrl *dev, struct nvme_id_ctrl **id)
>>>    	return error;
>>>    }
>>> +static bool nvme_multi_css(struct nvme_ctrl *ctrl)
>>> +{
>>> +	return (ctrl->ctrl_config & NVME_CC_CSS_MASK) == NVME_CC_CSS_CSI;
>>> +}
>>> +
>>>    static int nvme_process_ns_desc(struct nvme_ctrl *ctrl, struct nvme_ns_ids *ids,
>>> -		struct nvme_ns_id_desc *cur)
>>> +		struct nvme_ns_id_desc *cur, bool *csi_seen)
>>>    {
>>>    	const char *warn_str = "ctrl returned bogus length:";
>>>    	void *data = cur;
>>> @@ -1087,6 +1092,15 @@ static int nvme_process_ns_desc(struct nvme_ctrl *ctrl, struct nvme_ns_ids *ids,
>>>    		}
>>>    		uuid_copy(&ids->uuid, data + sizeof(*cur));
>>>    		return NVME_NIDT_UUID_LEN;
>>> +	case NVME_NIDT_CSI:
>>> +		if (cur->nidl != NVME_NIDT_CSI_LEN) {
>>> +			dev_warn(ctrl->device, "%s %d for NVME_NIDT_CSI\n",
>>> +				 warn_str, cur->nidl);
>>> +			return -1;
>>> +		}
>>> +		memcpy(&ids->csi, data + sizeof(*cur), NVME_NIDT_CSI_LEN);
>>> +		*csi_seen = true;
>>> +		return NVME_NIDT_CSI_LEN;
>>>    	default:
>>>    		/* Skip unknown types */
>>>    		return cur->nidl;
>>> @@ -1097,10 +1111,9 @@ static int nvme_identify_ns_descs(struct nvme_ctrl *ctrl, unsigned nsid,
>>>    		struct nvme_ns_ids *ids)
>>>    {
>>>    	struct nvme_command c = { };
>>> -	int status;
>>> +	bool csi_seen = false;
>>> +	int status, pos, len;
>>>    	void *data;
>>> -	int pos;
>>> -	int len;
>>>    	c.identify.opcode = nvme_admin_identify;
>>>    	c.identify.nsid = cpu_to_le32(nsid);
>>> @@ -1130,13 +1143,19 @@ static int nvme_identify_ns_descs(struct nvme_ctrl *ctrl, unsigned nsid,
>>>    		if (cur->nidl == 0)
>>>    			break;
>>> -		len = nvme_process_ns_desc(ctrl, ids, cur);
>>> +		len = nvme_process_ns_desc(ctrl, ids, cur, &csi_seen);
>>>    		if (len < 0)
>>>    			goto free_data;
>>>    		len += sizeof(*cur);
>>>    	}
>>>    free_data:
>>> +	if (!status && nvme_multi_css(ctrl) && !csi_seen) {
>>
>> We will clear the status if we detect a path error, that is to
>> avoid needlessly removing the ns for path failures, so you should
>> check at the goto site.
> 
> The problem is that this check has to be done after checking all the ns descs,
> so this check to be done as the final thing, at least after processing all the
> ns descs. No matter if nvme_process_ns_desc() returned an error, or if
> simply NVME_NIDT_CSI wasn't part of the ns desc list, so the loop reached the
> end without error.
> 
> Even if the nvme command failed and the status was cleared:
> 
>                  if (status > 0 && !(status & NVME_SC_DNR))
>                          status = 0;
> 
> we still need to return an error, if (nvme_multi_css(ctrl) && !csi_seen).
> (Not reporting a CSI when nvme_multi_css() is enabled, is fatal.)
> 
> That is why the code looks like it does.
> 
> I guess we could do something like this, which does the same thing,
> but perhaps is a bit clearer:
> 
> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> index e95f0c498a6b..bef687b9a277 100644
> --- a/drivers/nvme/host/core.c
> +++ b/drivers/nvme/host/core.c
> @@ -1160,8 +1160,10 @@ static int nvme_identify_ns_descs(struct nvme_ctrl *ctrl, unsigned nsid,
>                    * Don't treat an error as fatal, as we potentially already
>                    * have a NGUID or EUI-64.
>                    */
> -               if (status > 0 && !(status & NVME_SC_DNR))
> +               if (status > 0 && !(status & NVME_SC_DNR)) {
>                          status = 0;
> +                       goto csi_check;
> +               }

I think its the opposite. If we failed with DNR, you should assume
that either the controller wants the host to retry, or its a
path/transport error. So we need to leave this namespace alone and
assume that when the host is connected back to the controller, the
scan_work will revalidate again.

So you should fail the csi_check only if you see a DNR error status.