netdev - Re: [PATCH v3 7/8] nvmet-tcp: Support KeyUpdate

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20251003095434.GB15497@lst.de>
Date: Fri, 3 Oct 2025 11:54:34 +0200
From: Christoph Hellwig <hch@....de>
To: alistair23@...il.com
Cc: chuck.lever@...cle.com, hare@...nel.org,
	kernel-tls-handshake@...ts.linux.dev, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org, linux-doc@...r.kernel.org,
	linux-nvme@...ts.infradead.org, linux-nfs@...r.kernel.org,
	kbusch@...nel.org, axboe@...nel.dk, hch@....de, sagi@...mberg.me,
	kch@...dia.com, hare@...e.de,
	Alistair Francis <alistair.francis@....com>
Subject: Re: [PATCH v3 7/8] nvmet-tcp: Support KeyUpdate

On Fri, Oct 03, 2025 at 02:31:38PM +1000, alistair23@...il.com wrote:
> +#ifdef CONFIG_NVME_TARGET_TCP_TLS
> +static int nvmet_tcp_try_peek_pdu(struct nvmet_tcp_queue *queue);
> +static void nvmet_tcp_tls_handshake_timeout(struct work_struct *w);
> +#endif

Can we find a way to structure the code to do without forward declarations
and either without ifdefs or with stubs when you need them?

> +#ifdef CONFIG_NVME_TARGET_TCP_TLS
> +			if (ret == -EKEYEXPIRED &&
> +				queue->state != NVMET_TCP_Q_DISCONNECTING &&
> +				queue->state != NVMET_TCP_Q_TLS_HANDSHAKE) {
> +					goto done;
> +			}

Wrong indentation and superflous braces here.

> +	ret = nvmet_tcp_tls_handshake(queue, HANDSHAKE_KEY_UPDATE_TYPE_RECEIVED);
> +
> +	if (ret < 0)
> +		return ret;
> +
> +	ret = wait_for_completion_interruptible_timeout(&queue->tls_complete, 10 * HZ);

Please avoid the overly long lines.

> +
> +	if (ret <= 0) {
> +		tls_handshake_cancel(queue->sock->sk);
> +		return ret;
> +	}
> +
> +	queue->state = NVMET_TCP_Q_LIVE;
> +
> +	return ret;

This should be a unconditional ret 0, or am I missing something?

> +#ifdef CONFIG_NVME_TARGET_TCP_TLS
> +			if (ret == -EKEYEXPIRED &&
> +				queue->state != NVMET_TCP_Q_DISCONNECTING &&
> +				queue->state != NVMET_TCP_Q_TLS_HANDSHAKE) {
> +					goto done;
> +			}

Same as above.  And given that we have multiple instances of this
check I suspect we want a little helper for it.

> +#ifdef CONFIG_NVME_TARGET_TCP_TLS
> +			if (ret == -EKEYEXPIRED)
> +				update_tls_keys(queue);
> +			else
> +#endif
> +				return;
> +		}

If you provide a proper stub for update_tls_keys this becomes much
saner:

		if (ret != -EKEYEXPIRED)
			return;
		update_tls_keys(queue);

> +#ifdef CONFIG_NVME_TARGET_TCP_TLS
> +			if (ret == -EKEYEXPIRED)
> +				update_tls_keys(queue);
> +			else
> +#endif
> +				return;

Same here.