Re: [PATCHv2] kernel/crash: make parse_crashkernel()'s return value more indicant

[Dave Young <dyoung@redhat.com>]

On 04/29/19 at 12:48pm, Pingfan Liu wrote:
> On Mon, Apr 29, 2019 at 11:04 AM Pingfan Liu <kernelfans@gmail.com> wrote:
> >
> > On Sun, Apr 28, 2019 at 4:37 PM Dave Young <dyoung@redhat.com> wrote:
> > >
> > > On 04/25/19 at 04:20pm, Pingfan Liu wrote:
> > > > On Wed, Apr 24, 2019 at 4:31 PM Matthias Brugger <mbrugger@suse.com> wrote:
> > > > >
> > > > >
> > > > [...]
> > > > > > @@ -139,6 +141,8 @@ static int __init parse_crashkernel_simple(char *cmdline,
> > > > > >               pr_warn("crashkernel: unrecognized char: %c\n", *cur);
> > > > > >               return -EINVAL;
> > > > > >       }
> > > > > > +     if (*crash_size == 0)
> > > > > > +             return -EINVAL;
> > > > >
> > > > > This covers the case where I pass an argument like "crashkernel=0M" ?
> > > > > Can't we fix that by using kstrtoull() in memparse and check if the return value
> > > > > is < 0? In that case we could return without updating the retptr and we will be
> > > > > fine.
> > > > >
> > > > It seems that kstrtoull() treats 0M as invalid parameter, while
> > > > simple_strtoull() does not.
> > > >
> > > > If changed like your suggestion, then all the callers of memparse()
> > > > will treats 0M as invalid parameter. This affects many components
> > > > besides kexec.  Not sure this can be done or not.
> > >
> > > simple_strtoull is obsolete, move to kstrtoull is the right way.
> > >
> > > $ git grep memparse|wc
> > >     158     950   10479
> > >
> > > Except some documentation/tools etc there are still a log of callers
> > > which directly use the return value as the ull number without error
> > > checking.
> > >
> > > So it would be good to mark memparse as obsolete as well in
> > > lib/cmdline.c, and introduce a new function eg. kmemparse() to use
> > > kstrtoull,  and return a real error code, and save the size in an
> > > argument like &size.  Then update X86 crashkernel code to use it.
> > >
> > Thank for your good suggestion.
> >
> Go through the v5.0 kernel code, I think it will be a huge job.
> 
> The difference between unsigned long long simple_strtoull(const char
> *cp, char **endp, unsigned int base) and int _kstrtoull(const char *s,
> unsigned int base, unsigned long long *res) is bigger than expected,
> especially the output parameter @res. Many references to
> memparse(const char *ptr, char **retptr) rely on @retptr to work. A
> typical example from arch/x86/kernel/e820.c
>         mem_size = memparse(p, &p);
>         if (p == oldp)
>                 return -EINVAL;
> 
>         userdef = 1;
>         if (*p == '@') {  <----------- here
>                 start_at = memparse(p+1, &p);
>                 e820__range_add(start_at, mem_size, E820_TYPE_RAM);
>         } else if (*p == '#') {
>                 start_at = memparse(p+1, &p);
>                 e820__range_add(start_at, mem_size, E820_TYPE_ACPI);
>         } else if (*p == '$') {
>                 start_at = memparse(p+1, &p);
>                 e820__range_add(start_at, mem_size, E820_TYPE_RESERVED);
>         }
> 
> So we need to resolve the prototype of kstrtoull() firstly, and maybe
> kstrtouint() etc too. All of them have lots of references in kernel.
> 
> Any idea about this?


Not only this place, a lot of other places, I think no hurry to fix them
all at one time.

As we talked just do it according to previous reply,  mark memparse as
obsolete, and create a new function to use kstrtoull, and make it used
in crashkernel code first.

Thanks
Dave