[PATCH 2/2] nvme: fix atomic write boundary validation

John Garry john.g.garry at oracle.com
Mon Jun 16 04:49:56 PDT 2025


On 16/06/2025 12:37, Christoph Hellwig wrote:
>>> +	if (le16_to_cpu(id->awupf) != ctrl->subsys->awupf) {
>>> +		dev_err_ratelimited(ctrl->device,
>>> +			"inconsistent AWUPF, controller not added (%u/%u).\n",
>>> +			le16_to_cpu(id->awupf), ctrl->subsys->awupf);
>> Could we just disable atomic writes instead of doing this?
>>
>> Or are there bigger issues, like the value returned from
>> nvme_update_disk_info() for setting the physical block size is just not
>> valid?
> This brings up back to the old problem of how do we tell the file system
> and/or application that the atomic write size it probed suddenly went
> away entirely because a new controller disappeared.

Yeah, xfs does store those values at mount time.

If spinning another version, it would be good to explicitly mention this 
for benefit of forgetful people like me.

cheers



More information about the Linux-nvme mailing list