[PATCH 1/2] nvme: only allow entering LIVE from CONNECTING state

[Hannes Reinecke]

On 4/29/25 20:13, Keith Busch wrote:
> On Mon, Apr 28, 2025 at 03:21:18PM +0200, Hannes Reinecke wrote:
>>> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
>>> index b502ac07483b..d3c4eacf607f 100644
>>> --- a/drivers/nvme/host/core.c
>>> +++ b/drivers/nvme/host/core.c
>>> @@ -4493,7 +4493,8 @@ static void nvme_fw_act_work(struct work_struct *work)
>>>                   msleep(100);
>>>           }
>>>
>>> -       if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_LIVE))
>>> +       if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_CONNECTING) ||
>>> +           !nvme_change_ctrl_state(ctrl, NVME_CTRL_LIVE))
>>>                   return;
>>>
>>>           nvme_unquiesce_io_queues(ctrl);
>>
>> I would rather have a separate state for firmware activation.
>> (Ab-)using the 'RESETTING' state here has direct implications
>> with the error handler, as for the error handler 'RESETTING'
>> means that the error handler has been scheduled.
>> Which is not true for firmware activation.
> 
> But the point of having firmware activation set the state to RESETTING
> was to fence off error handling from trying to schedule a real reset.
> The fw activation work schedules its own recovery if it times out, but
> we don't want any other recovery action or user requested resets to
> proceed while an activation is still pending.

I know; that was precisely my point. We are overloading the 'RESETTTING'
state to mean either 'reset has started' or 'fw activation is ongoing'.
Which are two _vastly_ different situations, and we should differentiate
them eg by introducing a new state. That new state can (and should) have
the same effects as the RESETTING state, true.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare at suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich