Re: [PATCH] x86/mm/dump_pagetables: Change to use DEFINE_SHOW_ATTRIBUTE macro

[Frank Lee]

On Wed, Nov 28, 2018 at 6:50 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>
> On Mon, Nov 19, 2018 at 9:06 AM, Dave Hansen <dave.hansen@xxxxxxxxx> wrote:
> > On 11/19/18 7:43 AM, Yangtao Li wrote:
> >> -static const struct file_operations ptdump_curusr_fops = {
> >> -     .owner          = THIS_MODULE,
> >> -     .open           = ptdump_open_curusr,
> >> -     .read           = seq_read,
> >> -     .llseek         = seq_lseek,
> >> -     .release        = single_release,
> >> -};
> >> +DEFINE_SHOW_ATTRIBUTE(ptdump_curusr);
> >
> > FWIW, I rather dislike this conversion and the DEFINE_SHOW_ATTRIBUTE()
> > approach in general.  It makes it basically impossible to go from
> > ptdump_curusr to ptdump_open_curusr without opening up the macro and
> > reverse-engineering it.
> >
> > My test is that for these macros to be sane, I need to be able to find
> > "ptdump_open_curusr" by grepping for "ptdump_curusr".  This fails the test.
>
> Er, "ptdump_curusr" matches the generated name "ptdump_curusr_show",
> is that what you mean?
>
> > I don't think saving a few lines of code is worth the obfuscation.
>
> This is the standard boilerplate for attributes, though. I'd be nice
> to drop all the copy/pasted code...
>
> --
> Kees Cook
I also think that we should delete the duplicate code which doesn't
make much sense.

MBR,
Yangtao