Re: [PATCH] init: Don't proxy console= to earlycon

From: Raul Rangel
Date: Fri Jul 28 2023 - 13:57:26 EST

Next message: Ian Rogers: "Re: [PATCH] perf script: print cgroup on the same line as comm"
Previous message: Jason Gunthorpe: "Re: [PATCH v3 08/17] iommufd: IOMMU_HWPT_ALLOC allocation with user data"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Sorry, sending again since my original reply wasn't in plain text mode
and got bounced from the mailing list.

On Tue, Jul 18, 2023 at 3:50 AM Petr Mladek <pmladek@xxxxxxxx> wrote:
>
> On Fri 2023-07-14 15:42:36, Raul Rangel wrote:
> > On Fri, Jul 14, 2023 at 12:35 PM Petr Mladek <pmladek@xxxxxxxx> wrote:
> > > On Fri 2023-07-14 11:21:09, Raul Rangel wrote:
> > > > On Fri, Jul 14, 2023 at 10:38 AM Petr Mladek <pmladek@xxxxxxxx> wrote:
> > > > > On Mon 2023-07-10 09:30:19, Raul Rangel wrote:
> > > > > > On Sun, Jul 9, 2023 at 8:43 PM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> > > > > > > On 7/9/23 18:15, Mario Limonciello wrote:
> > > > > > > > On 7/9/23 18:46, Randy Dunlap wrote:
> > > > > > > >> On 7/7/23 18:17, Raul E Rangel wrote:
> > > > > > > >>> Right now we are proxying the `console=XXX` command line args to the
> > > > > > > >>> param_setup_earlycon. This is done because the following are
> > > > > > > >>> equivalent:
> > > > > > > >>>
> > > > > > > >>> console=uart[8250],mmio,<addr>[,options]
> > > > > > > >>> earlycon=uart[8250],mmio,<addr>[,options]
>
> I have finally got the meaning of the above paragraph. I thought that
> it was talking about that the format was equivalent. But it was about
> that also the effect was equivalent.
>
> > > > > > > >>> In addition, when `earlycon=` or just `earlycon` is specified on the
> > > > > > > >>> command line, we look at the SPCR table or the DT to extract the device
> > > > > > > >>> options.
> > > > > > > >>>
> > > > > > > >>> When `console=` is specified on the command line, it's intention is to
> > > > > > > >>> disable the console. Right now since we are proxying the `console=`
> > > > > > > >>
> > > > > > > >> How do you figure this (its intention is to disable the console)?
> > > > > > > >
> > > > > >
> > > > > > https://www.kernel.org/doc/html/v6.1/admin-guide/kernel-parameters.html
> > > > > > says the following:
> > > > > > console=
> > > > > > { null | "" }
> > > > > > Use to disable console output, i.e., to have kernel
> > > > > > console messages discarded.
> > > > > > This must be the only console= parameter used on the
> > > > > > kernel command line.
> > > > > >
> > > > > > earlycon= [KNL] Output early console device and options.
> > > > > >
> > > > > > When used with no options, the early console is
> > > > > > determined by stdout-path property in device tree's
> > > > > > chosen node or the ACPI SPCR table if supported by
> > > > > > the platform.
> > > > >
> > > > > Sigh, I wasn't aware of this when we discussed the console= handling.
> > > >
> > > > It took a bit of digging to figure out what the actual intention was :)
> > > >
> > > > >
> > > > > > The reason this bug showed up is that ChromeOS has set `console=` for a
> > > > > > very long time:
> > > > > > https://chromium.googlesource.com/chromiumos/platform/crosutils/+/main/build_kernel_image.sh#282
> > > > > > I'm not sure on the exact history, but AFAIK, we don't have the ttyX devices.
> > > > > >
> > > > > > Coreboot recently added support for the ACPI SPCR table which in
> > > > > > combination with the
> > > > > > `console=` arg, we are now seeing earlycon enabled when it shouldn't be.
> > > > >
> > > > > But this happens only when both "earlycon" and "console=" parameters
> > > > > are used together. Do I get it correctly?
> > > >
> > > > The bug shows up when an SPCR table is present and the `console=`
> > > > parameter is set. No need to specify `earlycon` on the command line.
> > >
> > > Strange, see below.
> > >
> > > > > This combination is ambiguous on its own. Why would anyone add
> > > > > "earlycon" parameter and wanted to keep it disabled?
> > > >
> > > > This is not the case I'm hitting. I'm honestly not sure what the
> > > > behavior should be in the `earlycon console=` case?
> > > >
> > > > >
> > > > > > > >>> diff --git a/init/main.c b/init/main.c
> > > > > > > >>> index aa21add5f7c54..f72bf644910c1 100644
> > > > > > > >>> --- a/init/main.c
> > > > > > > >>> +++ b/init/main.c
> > > > > > > >>> @@ -738,8 +738,7 @@ static int __init do_early_param(char *param, char *val,
> > > > > > > >>> for (p = __setup_start; p < __setup_end; p++) {
> > > > > > > >>> if ((p->early && parameq(param, p->str)) ||
> > > > > > > >>> (strcmp(param, "console") == 0 &&
> > > > > > > >>> - strcmp(p->str, "earlycon") == 0)
> > > > > > > >>> - ) {
> > > > > > > >>> + strcmp(p->str, "earlycon") == 0 && val && val[0])) {
> > > > > > > >>> if (p->setup_func(val) != 0)
> > > > > > > >>> pr_warn("Malformed early option '%s'\n", param);
> > > > > > > >>> }
> > >
> > This contradicts your first point. We need to call
> > `param_setup_earlycon` so it can handle `console=uart,mmio,XXXX`.
> > That's why this block of code is here. IMO it's very confusing
> > behavior that `earlycon=uart,mmio,XXXX` and `console=uart,mmio,XXXX`
> > are the same thing.
>
> Urgh, I didn't know that "console=uart*" started early console.
> I always thought that it was just another way how to define
> the normal console.

Yeah, it's all very confusing. We should probably remove the
`console=uart*` from the kernel documentation so people use
`earlycon=uart*` instead.

>
> (I feel shame as a printk maintainer. But I have never used it.
> And this has been the first patch in the related code since
> I started watching printk 10 years ago.)

lol, please don't shame yourself. There are plenty of odd quirks to be
found in maintaining such a large project.

>
> Looking into the history, "earlycon" parameter was added by
> the commit 18a8bd949d6adb311 ("serial: convert early_uart to
> earlycon for 8250") in 2007. The parameter "console=uart,mmio,XXX"
> started the console earlier even before.
>
> The "earlycon" parameter allowed to define the early console
> an explicit way on the command line. But it was not strictly
> necessary. The same effect could have been achieved by:
>
> static int __init do_early_param(char *param, char *val,
> const char *unused, void *arg)
> {
> struct obs_kernel_param *p;
>
> for (p = __setup_start; p < __setup_end; p++) {
> if (p->early && strcmp(param, p->str) == 0) {
> if (p->setup_func(val) != 0)
> printk(KERN_WARNING
> "Malformed early option '%s'\n", param);
> }
>
> if (strcmp(param, "console") == 0) {
> if (setup_early_serial8250_console(val) != 0)
> printk(KERN_WARNING, "Failed to setup early console");
> }
> }
>
> > The reason my patch checks for a NULL or empty val is because
> > `param_setup_earlycon` has a special case to handle the
> > `earlycon`/`earlycon=` case:
> > https://chromium.googlesource.com/chromiumos/third_party/kernel/+/refs/heads/chromeos-6.1/drivers/tty/serial/earlycon.c#223
> >
> > ```
> > /* Just 'earlycon' is a valid param for devicetree and ACPI SPCR. */
> > if (!buf || !buf[0]) {
> > if (IS_ENABLED(CONFIG_ACPI_SPCR_TABLE)) {
> > earlycon_acpi_spcr_enable = true;
> > return 0;
> > } else if (!buf) {
> > return early_init_dt_scan_chosen_stdout();
> > }
> > }
> > ```
>
> I see. It all makes sense now.

Great :)

>
> Your patch is good then. Well, would you mind to add a comment into
> the code and make the commit message more clear even for dummies like
> me?
>
> Something like the patch below. It would be better to split it into
> two:
>
> + 1st shuffling the check and adding the first part of the comment
> + 2nd fixing the case with empty console= options.
>
> I could prepare the patchset. I would keep your SOB for the 2nd patch
> if you agreed.

Yes, feel free. Thanks!

>
> Here is the proposal:
>
> From d2fb7c54bed4c67df229c56fcc1b0af83ada5227 Mon Sep 17 00:00:00 2001
> From: Raul E Rangel <rrangel@xxxxxxxxxxxx>
> Date: Fri, 7 Jul 2023 19:17:25 -0600
> Subject: [PATCH] init: Don't proxy the empty console= options to earlycon
>
> Right now we are proxying the `console=XXX` command line args to the
> param_setup_earlycon. This is done because the early consoles used to
> be enabled via the `console` parameter.
>
> The `earlycon` parameter was added later by the commit 18a8bd949d6adb311
> ("serial: convert early_uart to earlycon for 8250"). It allowed to
> setup the early consoles using another callback. And it was indeed
> more self-explanatory and cleaner approach.
>
> As a result, for example, the following parameters have the same effect:
>
> console=uart[8250],mmio,<addr>[,options]
> earlycon=uart[8250],mmio,<addr>[,options]
>
> Nevertheless, `console` and `earlycon` parameters behave different when
> the options do not match an early console. setup_earlycon() accepts only
> console names in __earlycon_table. Also the empty options are handled
> differently:
>
> + When `earlycon=` or just `earlycon` is specified on the command line,
> param_setup_earlycon() tries to enable an early console via the SPCR
> table or the device tree.
>
> + When `console=` is specified on the command line, it's intention is to
> disable the console.
>
> Here comes the problem. The current code calls param_setup_earlycon()
> even when `console=` is used with no options. As a result, it might
> enable the early console when it is defined in the SPCR table or
> a device tree. This is obviously not what users want.
>
> The early console should be enabled via SPCR or DT only when `earlycon`
> is used explicitly with no options.
>
> Signed-off-by: Raul E Rangel <rrangel@xxxxxxxxxxxx>
> [pmladek@xxxxxxxx: Add comments and more details into the commit message]
> ---
> init/main.c | 20 ++++++++++++++++----
> 1 file changed, 16 insertions(+), 4 deletions(-)
>
> diff --git a/init/main.c b/init/main.c
> index ad920fac325c..3b059e316e62 100644
> --- a/init/main.c
> +++ b/init/main.c
> @@ -733,13 +733,25 @@ static int __init do_early_param(char *param, char *val,
> const struct obs_kernel_param *p;
>
> for (p = __setup_start; p < __setup_end; p++) {
> - if ((p->early && parameq(param, p->str)) ||
> - (strcmp(param, "console") == 0 &&
> - strcmp(p->str, "earlycon") == 0)
> - ) {
> + if (p->early && parameq(param, p->str)) {
> if (p->setup_func(val) != 0)
> pr_warn("Malformed early option '%s'\n", param);
> }
> +
> + /*
> + * Early consoles can be historically enabled also when earlycon
> + * specific options are passed via console=[earlycon options]
> + * parameter.
> + *
> + * Do not try it with an empty value. "console=" is used to
> + * disable real normal consoles. While "earlycon" is used to
> + * enable an early console via SPCR or DT.
> + */
> + if (strcmp(param, "console") == 0 && val && val[0] &&
> + strcmp(p->str, "earlycon") == 0) {
> + if (p->setup_func(val) != 0)
> + pr_warn("Failed to setup earlycon via console=%s\n", val);
> + }
> }
> /* We accept everything at this stage. */
> return 0;
> --
> 2.35.3
>

Signed-off-by: Raul E Rangel <rrangel@xxxxxxxxxxxx>

Next message: Ian Rogers: "Re: [PATCH] perf script: print cgroup on the same line as comm"
Previous message: Jason Gunthorpe: "Re: [PATCH v3 08/17] iommufd: IOMMU_HWPT_ALLOC allocation with user data"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]