[PATCH] kernel/crash_core: suppress unknown crashkernel parameter warning

[Philipp Rudo]

Hi Baoquan,

On Tue, 7 Dec 2021 11:34:46 +0800
Baoquan He <bhe at redhat.com> wrote:

> On 12/06/21 at 12:17pm, Philipp Rudo wrote:
> > When booting with crashkernel= on the kernel command line a warning
> > similar to
> > 
> > [    0.038294] Kernel command line: ro console=ttyS0 crashkernel=256M
> > [    0.038353] Unknown kernel command line parameters "crashkernel=256M", will be passed to user space.
> > 
> > is printed. This originates from crashkernel= being parsed independent from
> > the early_param() mechanism. So the code in init/main.c doesn't know  
> 
> Not only the early_param(), __setup() also takes the same mechanism.
> It's just handled in different stage. You might need to call it kernel
> param handling mechanism, not sure if it's accurate.

you are right, "kernel param handling" is better. I used early_param as
that's where we would need to hook into if we wanted to use the common
kernel param handling. But I don't think it is worth it.

@akpm: do you update the commit message before sending the patch to
       Linus or shall I send a v2?

> > that crashkernel= is a valid kernel parameter and prints this incorrect
> > warning. Suppress the warning by adding a dummy early_param handler for
> > crashkernel=.  
> 
> The fix looks good to me, thanks.
> 
> Acked-by: Baoquan He <bhe at redhat.com>

Thanks

> By the way, on which arch did you find this issue? Ask because I am
> wondering whether there's any other similiar independent kernel cmdline
> handling from __setup_param(). If have, is there a chance to take a
> common method to handle them, e.g a generic function or a place to
> identify them. Just wild thought, I have no idea yet. Otherwise, we may
> need several this kind of dummy handler for each one.

The issue was first reported on s390 but I used x86 to test the fix.
The only other reported parameter I encountered was BOOT_IMAGE= which
is not a kernel parameter and thus correct. But in the corresponding
bugzilla Andrew (on cc) said "Gah! I thought I had squashed all of
these interesting uses of the kernel command line, it is like playing whack-a-mole."
So I believe there were multiple other parameters that had the same problem.

Thanks
Philipp

> > Fixes: 86d1919a4fb0 ("init: print out unknown kernel parameters")
> > Signed-off-by: Philipp Rudo <prudo at redhat.com>
> > ---
> >  kernel/crash_core.c | 11 +++++++++++
> >  1 file changed, 11 insertions(+)
> > 
> > diff --git a/kernel/crash_core.c b/kernel/crash_core.c
> > index eb53f5ec62c9..256cf6db573c 100644
> > --- a/kernel/crash_core.c
> > +++ b/kernel/crash_core.c
> > @@ -6,6 +6,7 @@
> >  
> >  #include <linux/buildid.h>
> >  #include <linux/crash_core.h>
> > +#include <linux/init.h>
> >  #include <linux/utsname.h>
> >  #include <linux/vmalloc.h>
> >  
> > @@ -295,6 +296,16 @@ int __init parse_crashkernel_low(char *cmdline,
> >  				"crashkernel=", suffix_tbl[SUFFIX_LOW]);
> >  }
> >  
> > +/*
> > + * Add a dummy early_param handler to mark crashkernel= as a known command line
> > + * parameter and suppress incorrect warnings in init/main.c.
> > + */
> > +static int __init parse_crashkernel_dummy(char *arg)
> > +{
> > +	return 0;
> > +}
> > +early_param("crashkernel", parse_crashkernel_dummy);
> > +
> >  Elf_Word *append_elf_note(Elf_Word *buf, char *name, unsigned int type,
> >  			  void *data, size_t data_len)
> >  {
> > -- 
> > 2.31.1
> > 
> > 
> > _______________________________________________
> > kexec mailing list
> > kexec at lists.infradead.org
> > http://lists.infradead.org/mailman/listinfo/kexec
> >   
>