[RFC PATCH] nvme: use NS head lock to serialize access to NS list

John Garry john.g.garry at oracle.com
Mon Jan 26 10:22:13 PST 2026


On 26/01/2026 17:08, Keith Busch wrote:
> On Mon, Jan 26, 2026 at 04:45:52PM +0000, John Garry wrote:
>> @@ -702,12 +702,12 @@ static void nvme_remove_head_work(struct work_struct *work)
>>   			struct nvme_ns_head, remove_work);
>>   	bool remove = false;
>>   
>> -	mutex_lock(&head->subsys->lock);
>> +	mutex_lock(&head->lock);
>>   	if (list_empty(&head->list)) {
>>   		list_del_init(&head->entry);
> 
> The head->entry is linked into the subsystem's "nsheads" list, so at
> least access to that still needs to happen under the subsys->lock.
> 

Ah, yes, I overlooked that.

> The other lock changes constrained to the head->list usage look fine.

ok, thanks



More information about the Linux-nvme mailing list