linux-kernel - Re: [PATCH 3/4] platform/chrome: cros_ec_typec: Configure muxes at start of port update

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

Date:   Mon, 7 Feb 2022 22:12:10 -0800
From:   Prashant Malani <pmalani@...omium.org>
To:     Tzung-Bi Shih <tzungbi@...gle.com>
Cc:     linux-kernel@...r.kernel.org, Benson Leung <bleung@...omium.org>,
        Guenter Roeck <groeck@...omium.org>,
        "open list:CHROMEOS EC USB TYPE-C DRIVER" 
        <chrome-platform@...ts.linux.dev>
Subject: Re: [PATCH 3/4] platform/chrome: cros_ec_typec: Configure muxes at
 start of port update

Hi Tzung-Bi,

On Mon, Feb 7, 2022 at 9:38 PM Tzung-Bi Shih <tzungbi@...gle.com> wrote:
>
> On Mon, Feb 07, 2022 at 09:40:28PM +0000, Prashant Malani wrote:
> > There are situations where the mux state reported by the Embedded
> > Controller (EC), might lag the partner "connected" state. So, the mux
> > state might still suggest that a partner is connected, while the PD
> > "connected" state, being in Try.SNK (for example) suggests that the
> > partner is disconnected.
> >
> > In such a scenario, we will end up sending a disconnect command to the
> > mux driver, followed by a connect command, since the mux is configured
> > later. Avoid this by configuring the mux before
> > registering/disconnecting a partner.
>
> I failed to understand the description.  It looks like some protocol details.
> Could you provide some brief explanation in the commit message?

I'm not sure how else I can better elaborate on this in the commit message than
as currently stated.
Since the EC is an independent controller, the mux state *can* lag the
"connected" state.
So, as described in the commit message, when a disconnect happens, we could have
a disconnect (since PD_CTRL contains the "connected state") followed
by a configure_mux
with the mux state still suggesting a connected device (the drivers
which implement the
mux/switch controls can misconstrue the old mux state) which results
in a connect. This
patch eliminates that.

>
> On a related note, followed up the example scenario, which one of the
> understanding is the most applicable:
> 1) The disconnect followed by a connect is suboptimal.  The patch cleans it.
> 2) The disconnect followed by a connect is a bug.  The patch fixes it.
This one (number 2)

>
> > @@ -965,6 +965,11 @@ static int cros_typec_port_update(struct cros_typec_data *typec, int port_num)
> >       if (ret < 0)
> >               return ret;
> >
> > +     /* Update the switches if they exist, according to requested state */
> > +     ret = cros_typec_configure_mux(typec, port_num, &resp);
> > +     if (ret)
> > +             dev_warn(typec->dev, "Configure muxes failed, err = %d\n", ret);
>
> It used the fact that the function returns `ret` at the end.  After the move,
> the block is no longer the last thing before function returns.
>
> Does it make more sense to return earlier if cros_typec_configure_mux() fails?
> Does the rest of code need to be executed even if cros_typec_configure_mux()
> fails?

Yes, it should still be executed (we still need to update the port
state). That is why the return is eliminated.

>
> > @@ -980,11 +985,6 @@ static int cros_typec_port_update(struct cros_typec_data *typec, int port_num)
> >       if (typec->typec_cmd_supported)
> >               cros_typec_handle_status(typec, port_num);
> >
> > -     /* Update the switches if they exist, according to requested state */
> > -     ret = cros_typec_configure_mux(typec, port_num, &resp);
> > -     if (ret)
> > -             dev_warn(typec->dev, "Configure muxes failed, err = %d\n", ret);
> > -
> >       return ret;
>
> If the function decides to return earlier, it can be `return 0;`.
Sure, I can change this in the next version