Message-ID: <CAD=FV=Uo4YAh8zGeU+tsxHEapsPyjAr9AyBEaAgiL3mUnN41=w@mail.gmail.com>
Date: Fri, 24 Oct 2025 09:03:16 -0700
From: Doug Anderson <dianders@...omium.org>
To: Geert Uytterhoeven <geert@...ux-m68k.org>
Cc: linux-kernel@...r.kernel.org, Andrew Morton <akpm@...ux-foundation.org>, 
	linux-s390 <linux-s390@...r.kernel.org>, Randy Dunlap <rdunlap@...radead.org>, 
	Andrew Chant <achant@...gle.com>, Heiko Carstens <hca@...ux.ibm.com>, 
	"Paul E . McKenney" <paulmck@...nel.org>, Sven Schnelle <svens@...ux.ibm.com>, 
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>, Brian Gerst <brgerst@...il.com>, 
	Christian Brauner <brauner@...nel.org>, Francesco Valla <francesco@...la.it>, 
	Guo Weikang <guoweikang.kernel@...il.com>, Huacai Chen <chenhuacai@...nel.org>, 
	Ingo Molnar <mingo@...nel.org>, Jan Hendrik Farr <kernel@...rr.cc>, Jeff Xu <jeffxu@...omium.org>, 
	Kees Cook <kees@...nel.org>, Masahiro Yamada <masahiroy@...nel.org>, 
	Michal Koutný <mkoutny@...e.com>, 
	Miguel Ojeda <ojeda@...nel.org>, "Mike Rapoport (Microsoft)" <rppt@...nel.org>, 
	Nathan Chancellor <nathan@...nel.org>, Peter Zijlstra <peterz@...radead.org>, 
	Shakeel Butt <shakeel.butt@...ux.dev>, Tejun Heo <tj@...nel.org>, 
	Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH v3] init/main.c: Wrap long kernel cmdline when printing to logs

Hi,

On Fri, Oct 24, 2025 at 12:51 AM Geert Uytterhoeven
<geert@...ux-m68k.org> wrote:
>
> > +config CMDLINE_LOG_WRAP_IDEAL_LEN
> > +       int "Length to try to wrap the cmdline when logged at boot"
> > +       default 1021
> > +       range -1 1021
>
> Apparently COMMAND_LINE_SIZE is still smaller than 1021 on several
> architectures (even in asm-generic: 512).  Unfortunately only s390
> still controls it through a config option, so you cannot have a
> "depends on COMMAND_LINE_SIZE > 1021" here...
>
> > +       help
> > +         At boot time, the kernel command line is logged to the console.
> > +         The log message will start with the prefix "Kernel command line: ".
> > +         The log message will attempt to be wrapped (split into multiple log
> > +         messages) at spaces based on CMDLINE_LOG_WRAP_IDEAL_LEN characters.
> > +         If wrapping happens, each log message will start with the prefix and
> > +         all but the last message will end with " \". Messages may exceed the
> > +         ideal length if a place to wrap isn't found before the specified
> > +         number of characters.
> > +
> > +         A value of -1 disables wrapping, though be warned that the maximum
>
> Or zero, right?
> So perhaps just use range 0 1021.

Sure, we can use 0 as the sentinel value. I was thinking -1 would be a
more obvious "wrapping is totally disabled" value but I don't feel
strongly about it. I'll change to 0 in the next patch.


> > +         length of a log message (1021 characters) may cause the cmdline to
> > +         be truncated.
> > +
> >  config INITRAMFS_PRESERVE_MTIME
> >         bool "Preserve cpio archive mtimes in initramfs"
> >         depends on BLK_DEV_INITRD
>
> > --- a/init/main.c
> > +++ b/init/main.c
>
> > +static void print_kernel_cmdline(const char *cmdline)
> > +{
> > +       size_t len;
> > +
> > +       /* Config option of -1 disables wrapping */
> > +       if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN < 0) {
>
> As does zero, right?

As the code is written right now, 0 does not disable wrapping. The
code treats everything less than 23 characters (the length of "Kernel
command line: " plus " \") to mean "wrap everything at the first
space". Yes, it's a little weird to do it this way, but it was either
that or add yet-another KConfig setting to disable wrapping and then
set the minimum to something higher. In v1/v2 I had the minimum set to
40 specifically to avoid the confusing case. There was previous
discussion about this in v2 [1].

...but yes, we can choose to make 0 be the special sentinel to disable
wrapping. I'll assume that's what you want and I'll change it in the
next version.


> You can add a check for "COMMAND_LINE_SIZE <= 1021" here,  so the
> compiler will eliminate the splitting code when it is not needed.

As the Kconfig is described and as the code is written, someone could
still choose wrapping even if COMMAND_LINE_SIZE <= 1021. Someone
could, for instance, choose to wrap these lines at 100 or 200
characters to make the log message look cleaner in their terminal/text
editor. ...but you're right that I can write this to be more optimal
for folks with shorter command lines who haven't touched the config.
I'll change the "if" test to this:

if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 ||
    IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) {
  pr_noitce(...);
  return;
}

That changes the sentinel to 0 (as I think you requested) and should
allow the compiler to optimize things out.

Making sure I got my math correct... Let me know if you see something wrong.

KERNEL_CMDLINE_PREFIX_LEN is 21

Assuming the CONFIG value isn't tiny, IDEAL_CMDLINE_LEN is the CONFIG
value minus 21.

So let's assume COMMAND_LINE_SIZE is 256. That means we can have at
most a cmdline length 255 to handle the '\0' termination.

So if the CONFIG value is (255 + 21) = 276 or more then we should hit
the "if" test and early-out because we can't wrap. If the CONFIG value
is 275 or less then we should proceed to wrapping.

We want ((276 - 21) >= (256 - 1)) to be true. It is.

We want ((275 - 21) >= (256 - 1)) to be false. It is.


I'll wait a few days and send a v4. I'll fold in Andrew's "__init"
patch as well.


[1] http://lore.kernel.org/r/CAD=FV=WFbH6kBMcoHNwQzsay6ecQQ2sZ3qc-=XTboFXK+RSspA@mail.gmail.com


-Doug

Powered by blists - more mailing lists