All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Alexey Charkov" <alchark@flipper.net>
Cc: linux-usb@vger.kernel.org
Subject: Re: [PATCH 3/3] usb: typec: tcpm: Keep cable plug messages away from the port partner
Date: Tue, 17 Nov 2026 11:42:58 +0000	[thread overview]
Message-ID: <sashiko-outbox-218719@kernel.org> (raw)
In-Reply-To: <20261117-tcpm-soft-reset-sop-v1-3-c7e1543828c0@flipper.net>

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Medium] usb: typec: tcpm: SOP' VDMs in tcpm_pd_cable_msg_request()
  affect partner
- [Medium] usb: typec: tcpm: SOP' PD_CTRL_REJECT in
  tcpm_pd_ctrl_request() resets partner
It also noticed 1 related pre-existing issue this patch does not
introduce:
- [High] usb: typec: tcpm: infinite discovery retry loop in
  tcpm_pd_ctrl_request() [1]
--

commit a17153f1f38f59679f1ba169d5aec3d715a00c73
Author: Alexey Charkov <alchark@flipper.net>

usb: typec: tcpm: Keep cable plug messages away from the port partner

This commit filters incoming SOP' messages in tcpm_pd_rx_handler() through
tcpm_pd_cable_msg_request() so that only VDMs and expected control responses
from a cable plug are processed, while dropping unsupported SOP* types before
updating MessageID tracking.

> diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c
> index 947fca918414e..2a909ca96811d 100644
> --- a/drivers/usb/typec/tcpm/tcpm.c
> +++ b/drivers/usb/typec/tcpm/tcpm.c
[ ... ]
> @@ -4164,6 +4166,52 @@ static void tcpm_pd_ext_msg_request(struct tcpm_port *port,
>  	}
>  }
>  
> +/* Whether a cable plug control message responds to what the port sent it */
> +static bool tcpm_cable_ctrl_msg_expected(struct tcpm_port *port,
> +					 enum pd_ctrl_msg_type type)
> +{
> +	switch (type) {
> +	case PD_CTRL_ACCEPT:
> +		/* the Soft_Reset after a VCONN Swap */
> +		return port->state == SOFT_RESET_SEND &&
> +		       port->soft_reset_sent_prime;
> +	case PD_CTRL_REJECT:
> +	case PD_CTRL_NOT_SUPP:
> +		/* a VDM, in the ready state, which takes it as the answer */
> +		return port->vdm_sop_type == TCPC_TX_SOP_PRIME &&
> +		       port->vdm_state > VDM_STATE_READY &&
> +		       port->state == ready_state(port);
> +	default:
> +		return false;
> +	}
> +}

[Severity: Medium]
When a cable plug replies to a structured VDM with PD_CTRL_REJECT in the
ready state, does forwarding it to tcpm_pd_ctrl_request() still soft-reset
the port partner?

At the top of tcpm_pd_ctrl_request(), any control message other than
PD_CTRL_NOT_SUPP and PD_CTRL_GOOD_CRC aborts an active VDM AMS before the
switch statement:

drivers/usb/typec/tcpm/tcpm.c:tcpm_pd_ctrl_request() {
    ...
	if (tcpm_vdm_ams(port) && type != PD_CTRL_NOT_SUPP && type != PD_CTRL_GOOD_CRC) {
		port->vdm_state = VDM_STATE_ERR_BUSY;
		tcpm_ams_finish(port);
		mod_vdm_delayed_work(port, 0);
	}
    ...
}

Because port->vdm_state is overwritten with VDM_STATE_ERR_BUSY (-3), the
subsequent port->vdm_state > VDM_STATE_READY check in the SRC_READY and
SNK_READY cases evaluates to false, falling through to the default branch
and triggering a soft reset against the port partner on SOP.

> +
> +/*
> + * A cable plug only responds to the VCONN source: to its VDMs, and to its
> + * Soft_Reset. Hand those to the usual handlers, which tell SOP' apart where
> + * it matters, and drop anything else on SOP': it comes from a cable plug
> + * that breaks the rules, and no sequence has the port answer a cable plug.
> + */
> +static void tcpm_pd_cable_msg_request(struct tcpm_port *port,
> +				      const struct pd_message *msg)
> +{
> +	unsigned int type = pd_header_type_le(msg->header);
> +	bool ext = le16_to_cpu(msg->header) & PD_HEADER_EXT_HDR;
> +	bool ctrl = !ext && !pd_header_cnt_le(msg->header);
> +
> +	if (!ext && !ctrl && type == PD_DATA_VENDOR_DEF) {
> +		tcpm_pd_data_request(port, msg, TCPC_TX_SOP_PRIME);
> +		return;
> +	}

[Severity: Medium]
Can an unexpected or malformed SOP' VDM passed to tcpm_pd_data_request()
still cause the port to send messages to or reset the port partner on SOP?

Unlike tcpm_cable_ctrl_msg_expected(), this path does not check whether the
port is waiting for an SOP' VDM response.  In tcpm_handle_vdm_request(), an
unstructured SOP' VDM (or an SVDM when no VDM AMS, partner altmode, or sink
VDO is active) hits the else branch and calls
tcpm_pd_handle_msg(port, PD_MSG_CTRL_NOT_SUPP, NONE_AMS), which transmits
Not_Supported on SOP or initiates a soft or hard reset of the port partner:

drivers/usb/typec/tcpm/tcpm.c:tcpm_handle_vdm_request() {
    ...
	if (PD_VDO_SVDM(p[0]) && (adev || tcpm_vdm_ams(port) || port->nr_snk_vdo)) {
		port->vdm_sm_running = true;
		rlen = tcpm_pd_svdm(port, adev, p, cnt, response, &adev_action,
				    rx_sop_type, &response_tx_sop_type);
	} else {
		if (port->negotiated_rev >= PD_REV30)
			tcpm_pd_handle_msg(port, PD_MSG_CTRL_NOT_SUPP, NONE_AMS);
	}
    ...
}

In addition, if the port is waiting for an SOP VDM response from the partner
(port->vdm_state == VDM_STATE_BUSY && port->vdm_sop_type == TCPC_TX_SOP), an
unsolicited SOP' VDM mutates port->vdm_state and port->vdo_retry for the
partner VDM state machine.  Finally, an SOP' SVDM that hits a CMDT_RSP_NAK
path in tcpm_pd_svdm() leaves response_tx_sop_type at its default of
TCPC_TX_SOP and queues the VDM NAK to the port partner.

> +	if (ctrl && tcpm_cable_ctrl_msg_expected(port, type)) {
> +		tcpm_pd_ctrl_request(port, msg, TCPC_TX_SOP_PRIME);
> +		return;
> +	}

[Severity: High]
This problem wasn't introduced by this patch, but when
tcpm_pd_ctrl_request() handles PD_CTRL_NOT_SUPP or PD_CTRL_REJECT for an
in-flight VDM in SRC_READY or SNK_READY [1], does it leave
port->vdm_discovery_state unadvanced and cause an infinite discovery retry
loop?

In tcpm_pd_ctrl_request(), the SRC_READY and SNK_READY handler sets
port->vdm_state = VDM_STATE_DONE and finishes the VDM AMS without updating
port->vdm_discovery_state:

drivers/usb/typec/tcpm/tcpm.c:tcpm_pd_ctrl_request() {
    ...
		case SRC_READY:
		case SNK_READY:
			if (port->vdm_state > VDM_STATE_READY) {
				port->vdm_state = VDM_STATE_DONE;
				if (tcpm_vdm_ams(port))
					tcpm_ams_finish(port);
				mod_vdm_delayed_work(port, 0);
				break;
			}
    ...
}

When a discovery command such as DISCOVER_SVIDS or DISCOVER_MODES is
answered with PD_CTRL_NOT_SUPP, port->vdm_discovery_state remains in its
intermediate phase instead of advancing, so subsequent state machine runs in
SRC_READY or SNK_READY that schedule mod_vdm_discovery_delayed_work(port, 0)
repeatedly re-send the same rejected VDM command.

[1] https://sashiko.dev/#/bug/linux-2c1789d2-5c2d-407e-af88-42956bee42c7

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117-tcpm-soft-reset-sop-v1-0-c7e1543828c0@flipper.net?part=3

  reply	other threads:[~2026-11-17 11:42 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17 11:15 [PATCH 0/3] usb: typec: tcpm: Fixes for talking to the cable plug as a VCONN source Alexey Charkov
2026-11-17 11:15 ` [PATCH 1/3] usb: typec: tcpm: Keep the Soft_Reset target apart from the VDM one Alexey Charkov
2026-11-17 11:30   ` sashiko-bot
2026-11-17 12:01     ` Alexey Charkov
2026-11-17 11:15 ` [PATCH 2/3] usb: typec: tcpm: Leave SRC_VDM_IDENTITY_REQUEST whatever the cable answers Alexey Charkov
2026-11-17 11:39   ` sashiko-bot
2026-11-17 12:02     ` Alexey Charkov
2026-11-17 11:15 ` [PATCH 3/3] usb: typec: tcpm: Keep cable plug messages away from the port partner Alexey Charkov
2026-11-17 11:42   ` sashiko-bot [this message]
2026-11-17 12:08     ` Alexey Charkov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-218719@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=alchark@flipper.net \
    --cc=linux-usb@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.