Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200428162227.687978-1-daniel.thompson@linaro.org>
In-Reply-To: <20200428162227.687978-1-daniel.thompson@linaro.org>
From:   Doug Anderson <dianders@chromium.org>
Date:   Wed, 29 Apr 2020 17:40:10 -0700
Message-ID: <CAD=FV=VTkO5AULJe1ksyhKLLxpywqX-RLC_T4ENyDf3ONJNMKg@mail.gmail.com>
Subject: Re: [PATCH] serial: earlycon: Allow earlier DT scan is acpi=off
To:     Daniel Thompson <daniel.thompson@linaro.org>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Jiri Slaby <jslaby@suse.com>,
        Sumit Garg <sumit.garg@linaro.org>,
        linux-serial@vger.kernel.org, LKML <linux-kernel@vger.kernel.org>,
        Patch Tracking <patches@linaro.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,

On Tue, Apr 28, 2020 at 9:22 AM Daniel Thompson
<daniel.thompson@linaro.org> wrote:
>
> Currently if the kernel has support for ACPI SPCR parsing then earlycon
> without arguments is processed later than the full earlycon=...
> alternative.
>
> If ACPI has been explicitly disabled on the kernel command line then
> there is not need to defer since the ACPI code (both x86 and arm64)
> will never actually run.
>
> Or, put another way it allows lazy people to throw "acpi=off earlycon"
> onto the command line of a DT systems and be confident the console will
> start as early as possible without them having to lookup the driver
> and address needed for a specific platform.
>
> Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
> ---
>  drivers/tty/serial/earlycon.c | 28 +++++++++++++++++++++++++++-
>  1 file changed, 27 insertions(+), 1 deletion(-)

I wonder if a cleaner option is to just add a special "earlycon" value
like "earlycon=not_acpi".  This wouldn't require any special peeking
and would just be a sentinel that just says "you should autodetect the
earlycon, but don't worry about waiting for ACPI".  ...that in itself
is a bit of a hack, but at least it's more self contained in the
earlycon driver and maybe more discoverable when someone is figuring
out how to setup earlycon?

-Doug


>
> diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
> index 2ae9190b64bb..ebb648aacb47 100644
> --- a/drivers/tty/serial/earlycon.c
> +++ b/drivers/tty/serial/earlycon.c
> @@ -215,6 +215,31 @@ int __init setup_earlycon(char *buf)
>   */
>  bool earlycon_acpi_spcr_enable __initdata;
>
> +/*
> + * This takes a sneaky peek at other boot arguments (which may not have
> + * been parsed at this point in the boot) to check whether ACPI has
> + * been explicitly disabled. If it is explicitly disabled then there is
> + * no reason to defer initialization of the early console.
> + */
> +static bool earlycon_check_for_acpi_off(void)
> +{
> +       static const char token[] = "acpi=off";
> +       const char *arg;
> +       char before, after;
> +
> +       arg = strstr(boot_command_line, token);
> +       while (arg) {
> +               before = arg == boot_command_line ? ' ' : arg[-1];
> +               after = arg[sizeof(token)-1];
> +               if (isspace(before) && (isspace(after) || after == '\0'))
> +                       return true;
> +
> +               arg = strstr(arg+1, token);
> +       }
> +
> +       return false;
> +}
> +
>  /* early_param wrapper for setup_earlycon() */
>  static int __init param_setup_earlycon(char *buf)
>  {
> @@ -222,7 +247,8 @@ static int __init param_setup_earlycon(char *buf)
>
>         /* Just 'earlycon' is a valid param for devicetree and ACPI SPCR. */
>         if (!buf || !buf[0]) {
> -               if (IS_ENABLED(CONFIG_ACPI_SPCR_TABLE)) {
> +               if (IS_ENABLED(CONFIG_ACPI_SPCR_TABLE) &&
> +                   !earlycon_check_for_acpi_off()) {
>                         earlycon_acpi_spcr_enable = true;
>                         return 0;
>                 } else if (!buf) {
> --
> 2.25.1
>