linux-kernel - Re: [PATCH v2 1/7] serial: qcom-geni: fix premature receiver enable

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]

Message-ID: <ZwaLTRHKXXhq6Qiu@hovoldconsulting.com>
Date: Wed, 9 Oct 2024 15:55:25 +0200
From: Johan Hovold <johan@...nel.org>
To: Doug Anderson <dianders@...omium.org>
Cc: Johan Hovold <johan+linaro@...nel.org>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Jiri Slaby <jirislaby@...nel.org>,
	Bjorn Andersson <andersson@...nel.org>,
	Konrad Dybcio <konradybcio@...nel.org>,
	linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-serial@...r.kernel.org, stable@...r.kernel.org,
	Aniket Randive <quic_arandive@...cinc.com>
Subject: Re: [PATCH v2 1/7] serial: qcom-geni: fix premature receiver enable

On Thu, Oct 03, 2024 at 11:29:58AM -0700, Doug Anderson wrote:
> On Tue, Oct 1, 2024 at 5:51 AM Johan Hovold <johan+linaro@...nel.org> wrote:
> >
> > The receiver should not be enabled until the port is opened so drop the
> > bogus call to start rx from the setup code which is shared with the
> > console implementation.
> >
> > This was added for some confused implementation of hibernation support,
> > but the receiver must not be started unconditionally as the port may not
> > have been open when hibernating the system.
> 
> Could you provide a motivation for your patch in the description? Is
> patch needed for something (perhaps a future patch in the series)? Is
> it fixing a bug? Does it save power? Is the call harmless but cleaner
> to get rid of?

I was trying to bring some order to this driver so that the receiver is
enabled when the port is opened and disabled when it is closed again as
expected, and get rid of the random calls added in places where they do
not belong (e.g., as Bjorn also mentioned, why was the call to start rx
added in the port setup code if it was needed for hibernation?).

Data "received" over the wire before opening the port should not be
processed, but it also turns out that enabling the receiver before the
port is opened can confuse the firmware and break the "stale" rx timer
handling so that data is only forwarded in chunks of 12 bytes instead of
when each char is received.

> > Fixes: 35781d8356a2 ("tty: serial: qcom-geni-serial: Add support for Hibernation feature")
> > Cc: stable@...r.kernel.org      # 6.2
> > Cc: Aniket Randive <quic_arandive@...cinc.com>
> > Signed-off-by: Johan Hovold <johan+linaro@...nel.org>
> > ---
> >  drivers/tty/serial/qcom_geni_serial.c | 1 -
> >  1 file changed, 1 deletion(-)
> >
> > diff --git a/drivers/tty/serial/qcom_geni_serial.c b/drivers/tty/serial/qcom_geni_serial.c
> > index 6f0db310cf69..9ea6bd09e665 100644
> > --- a/drivers/tty/serial/qcom_geni_serial.c
> > +++ b/drivers/tty/serial/qcom_geni_serial.c
> > @@ -1152,7 +1152,6 @@ static int qcom_geni_serial_port_setup(struct uart_port *uport)
> >                                false, true, true);
> >         geni_se_init(&port->se, UART_RX_WM, port->rx_fifo_depth - 2);
> >         geni_se_select_mode(&port->se, port->dev_data->mode);
> > -       qcom_geni_serial_start_rx(uport);
> 
> FWIW, I found at least one thing that's broken by your patch. If you
> enable kgdb (but _not_ "kgdboc_earlycon") and then add "kgdbwait" to
> the kernel command line parameters then things will be broken after
> your patch. You'll drop into the debugger but can't interact with it.
> The "kgdboc_earlycon" path handles this because of
> "qcom_geni_serial_enable_early_read()" but it doesn't seem like
> there's anything that handles it for normal kgdb. If you drop in the
> debugger later it'll probably work if you've got an "agetty" running
> because that'll enable the RX path.

Ok, so the kgdb has started relying on this call since d8851a96ba25
("tty: serial: qcom-geni-serial: Add a poll_init() function"). Thanks
for pointing that out.

The polled console code should not be calling the port setup code
unconditionally anyway so I'll fix this up as well in v3.

Johan