Hi Rasmus,
On Thu, Sep 27, 2018 at 12:55 PM Rasmus Villemoes
<linux@xxxxxxxxxxxxxxxxxx> wrote:
> On 2018-09-27 01:13, Kees Cook wrote:
> > On Mon, Sep 24, 2018 at 3:11 AM, Bartosz Golaszewski <brgl@xxxxxxxx> wrote:
> >> Provide a resource managed version of kstrdup_const(). This variant
> >> internally calls devm_kstrdup() on pointers that are outside of
> >> .rodata section and returns the string as is otherwise.
> >>
> >> Also provide a corresponding version of devm_kfree().
> >>
> >> +/**
> >> + * devm_kfree_const - Resource managed conditional kfree
> >> + * @dev: device this memory belongs to
> >> + * @p: memory to free
> >> + *
> >> + * Function calls devm_kfree only if @p is not in .rodata section.
> >> + */
> >> +void devm_kfree_const(struct device *dev, const void *p)
> >> +{
> >> + if (!is_kernel_rodata((unsigned long)p))
> >> + devm_kfree(dev, p);
> >> +}
> >> +EXPORT_SYMBOL(devm_kfree_const);
> >> +
> >> /**
> >> * devm_kmemdup - Resource-managed kmemdup
> >> * @dev: Device this memory belongs to
> >> diff --git a/include/linux/device.h b/include/linux/device.h
> >> index 33f7cb271fbb..79ccc6eb0975 100644
> >> --- a/include/linux/device.h
> >> +++ b/include/linux/device.h
> >> @@ -693,7 +693,10 @@ static inline void *devm_kcalloc(struct device *dev,
> >> return devm_kmalloc_array(dev, n, size, flags | __GFP_ZERO);
> >> }
> >> extern void devm_kfree(struct device *dev, const void *p);
> >> +extern void devm_kfree_const(struct device *dev, const void *p);
> >
> > With devm_kfree and devm_kfree_const both taking "const", how are
> > devm_kstrdup_const() and devm_kfree_const() going to be correctly
> > paired at compile time? (i.e. I wasn't expecting the prototype change
> > to devm_kfree())
>
> Just drop devm_kfree_const and teach devm_kfree to ignore
> is_kernel_rodata(). That avoids the 50-100 bytes of overhead for adding
> yet another EXPORT_SYMBOL and makes it easier to port drivers to
> devm_kstrdup_const (and avoids the bugs Kees is worried about). devm
> managed resources are almost never freed explicitly, so that single
> extra comparison in devm_kfree shouldn't matter for performance.
I guess we can also teach kfree() to ignore is_kernel_rodata(), and
drop kfree_const()?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
