Re: [PATCH v5 3/6] mm/memfd: add MFD_NOEXEC_SEAL and MFD_EXEC

[Jeff Xu]

On Tue, Dec 6, 2022 at 8:35 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Dec 06, 2022 at 08:26:30AM -0800, Jeff Xu wrote:
> > On Tue, Dec 6, 2022 at 8:04 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > On Tue, Dec 06, 2022 at 03:23:55PM +0000, jeffxu@xxxxxxxxxxxx wrote:
> > > > --- a/kernel/pid_namespace.c
> > > > +++ b/kernel/pid_namespace.c
> > > > @@ -110,6 +110,11 @@ static struct pid_namespace *create_pid_namespace(struct user_namespace *user_ns
> > > >       ns->ucounts = ucounts;
> > > >       ns->pid_allocated = PIDNS_ADDING;
> > > >
> > > > +#if defined(CONFIG_SYSCTL) && defined(CONFIG_MEMFD_CREATE)
> > > > +     ns->memfd_noexec_scope =
> > > > +             task_active_pid_ns(current)->memfd_noexec_scope;
> > > > +#endif
> > >
> > > .c files should never have #if in them.  Can't you put this in a .h file
> > > properly so that this does not get really messy over time?
> > >
> > >
> > Thanks for reviewing.
> > It seems to me that checking for CONFIG_XXX is  common in c code in
> > kernel/ path.
>
> Maybe, but please don't make it any worse if at all possible.  It's
> tough to maintain code like that.
>
> > Do you have a sample code pattern (link/function) that I can follow?
>
> Any of the zillions of #if statements in .h files :)
>
Thanks.
I will take the approach of having real/stub implementation in the h
file, and the c file  using it without a compile flag.
Please kindly let me know if this is not right.

Thanks
Jeff

> thanks,
>
> greg k-h