Re: [PATCH nvme-7.3 3/4] nvme-fabrics: add helpers for subsystem and host identity options

[Christoph Hellwig <hch@lst.de>]

On Fri, Aug 21, 2026 at 02:25:28PM +0800, raoxu wrote:
> +static int nvmf_parse_subsysnqn(struct nvmf_ctrl_options *opts, substring_t *args)
> +{
> +	char *nqn;
> +
> +	nqn = match_strdup(args);
> +	if (!nqn)
> +		return -ENOMEM;
> +
> +	kfree(opts->subsysnqn);
> +	opts->subsysnqn = nqn;
> +	if (strlen(opts->subsysnqn) >= NVMF_NQN_SIZE) {
> +		pr_err("%s needs to be < %d bytes\n",
> +		       opts->subsysnqn, NVMF_NQN_SIZE);
> +		return -EINVAL;
> +	}

I think we should only update opts->subsysnqn after the sanity check. The
current code doesn't do that, but we should use the chance to fix that
up as well.

> +	nqn = match_strdup(args);
> +	if (!nqn)
> +		return -ENOMEM;
> +
> +	if (strlen(nqn) >= NVMF_NQN_SIZE) {
> +		pr_err("%s needs to be < %d bytes\n", nqn, NVMF_NQN_SIZE);
> +		kfree(nqn);
> +		return -EINVAL;
> +	}
> +
> +	strscpy(hostnqn, nqn, NVMF_NQN_SIZE);
> +	kfree(nqn);

This already gets is right.  Also does anyone rememeber why we handled
the host vs subsys NQN so differently?