[Intel-gfx] [PATCH] drm/i915: Prefer struct_size over open coded arithmetic

[Len Baker]

Hi Jani,

On Mon, Oct 18, 2021 at 01:00:01PM +0300, Jani Nikula wrote:
> On Sat, 16 Oct 2021, Len Baker <len.baker at gmx.com> wrote:
> > Hi Daniel and Jani,
> >
> > On Wed, Oct 13, 2021 at 01:51:30PM +0200, Daniel Vetter wrote:
> >> On Wed, Oct 13, 2021 at 02:24:05PM +0300, Jani Nikula wrote:
> >> > On Mon, 11 Oct 2021, Len Baker <len.baker at gmx.com> wrote:
> >> > > Hi,
> >> > >
> >> > > On Sun, Oct 03, 2021 at 12:42:58PM +0200, Len Baker wrote:
> >> > >> As noted in the "Deprecated Interfaces, Language Features, Attributes,
> >> > >> and Conventions" documentation [1], size calculations (especially
> >> > >> multiplication) should not be performed in memory allocator (or similar)
> >> > >> function arguments due to the risk of them overflowing. This could lead
> >> > >> to values wrapping around and a smaller allocation being made than the
> >> > >> caller was expecting. Using those allocations could lead to linear
> >> > >> overflows of heap memory and other misbehaviors.
> >> > >>
> >> > >> In this case these are not actually dynamic sizes: all the operands
> >> > >> involved in the calculation are constant values. However it is better to
> >> > >> refactor them anyway, just to keep the open-coded math idiom out of
> >> > >> code.
> >> > >>
> >> > >> So, add at the end of the struct i915_syncmap a union with two flexible
> >> > >> array members (these arrays share the same memory layout). This is
> >> > >> possible using the new DECLARE_FLEX_ARRAY macro. And then, use the
> >> > >> struct_size() helper to do the arithmetic instead of the argument
> >> > >> "size + count * size" in the kmalloc and kzalloc() functions.
> >> > >>
> >> > >> Also, take the opportunity to refactor the __sync_seqno and __sync_child
> >> > >> making them more readable.
> >> > >>
> >> > >> This code was detected with the help of Coccinelle and audited and fixed
> >> > >> manually.
> >> > >>
> >> > >> [1] https://www.kernel.org/doc/html/latest/process/deprecated.html#open-coded-arithmetic-in-allocator-arguments
> >> > >>
> >> > >> Signed-off-by: Len Baker <len.baker at gmx.com>
> >> > >> ---
> >> > >>  drivers/gpu/drm/i915/i915_syncmap.c | 12 ++++++++----
> >> > >>  1 file changed, 8 insertions(+), 4 deletions(-)
> >> > >
> >> > > I received a mail telling that this patch doesn't build:
> >> > >
> >> > > == Series Details ==
> >> > >
> >> > > Series: drm/i915: Prefer struct_size over open coded arithmetic
> >> > > URL   : https://patchwork.freedesktop.org/series/95408/
> >> > > State : failure
> >> > >
> >> > > But it builds without error against linux-next (tag next-20211001). Against
> >> > > which tree and branch do I need to build?
> >> >
> >> > drm-tip [1]. It's a sort of linux-next for graphics. I think there are
> >> > still some branches that don't feed to linux-next.
> >>
> >> Yeah we need to get gt-next in linux-next asap. Joonas promised to send
> >> out his patch to make that happen in dim.
> >> -Daniel
> >
> > Is there a possibility that you give an "Acked-by" tag? And then this patch
> > goes to the mainline through the Kees' tree or Gustavo's tree?
>
> If this does not apply to drm-intel-gt-next (or drm-tip), applying it to
> some other branch will just cause unnecessary conflicts later on. It's
> unnecessary extra work. It's not an urgent fix or anything, there is no
> reason to do that. So that's a NAK.

Ok. Understood.

> > Or is it better to wait for drm-tip to update?
>
> drm-tip is up to date, it's just that one of the branches that feed to
> it is (was?) not feeding to linux-next.

Sorry, but I'm missing something here. In linux-next this is the commit
history of include/linux/stddef.h file:

3080ea5553cc stddef: Introduce DECLARE_FLEX_ARRAY() helper
50d7bd38c3aa stddef: Introduce struct_group() helper macro
e7f18c22e6be stddef: Fix kerndoc for sizeof_field() and offsetofend()
4229a470175b stddef.h: Introduce sizeof_field()
...

But in drm-tip this is the commit history:

4229a470175b stddef.h: Introduce sizeof_field()
...

For this patch the DECLARE_FLEX_ARRAY() helper is needed. But the build
fails due to the last tree commits for stddef.h file are not present.
So, if I understand correctly, drm-tip is not up to date with linux-next.

Regards,
Len

>
> If you're contributing to drm, please consider basing your patches on
> top of drm-tip.
>
>
> BR,
> Jani.
>
>
> >
> > Regards,
> > Len
> >
> >>
> >> >
> >> > BR,
> >> > Jani.
> >> >
> >> >
> >> > [1] https://cgit.freedesktop.org/drm/drm-tip
> >> >
> >> >
> >> > >
> >> > > Regards,
> >> > > Len
> >> >
> >> > --
> >> > Jani Nikula, Intel Open Source Graphics Center
> >>
> >> --
> >> Daniel Vetter
> >> Software Engineer, Intel Corporation
> >> http://blog.ffwll.ch
>
> --
> Jani Nikula, Intel Open Source Graphics Center