Re: Please backport 2ae5c9248e06 ("wifi: mac80211: Use flexible array in struct ieee80211_tim_ie")

[Kees Cook <kees@kernel.org>]

On Thu, Jun 27, 2024 at 07:44:48AM -0700, Jeff Johnson wrote:
> On 6/27/2024 12:18 AM, Greg KH wrote:
> > On Wed, Jun 26, 2024 at 11:32:22AM -0700, Jeff Johnson wrote:
> >> Refer to:
> >> https://lore.kernel.org/all/CAPh3n83zb1PwFBFijJKChBqY95zzpYh=2iPf8tmh=YTS6e3xPw@mail.gmail.com/
> > 
> > Please provide the information in the email, for those of us traveling
> > with store-and-forward email systems, links don't always work.
> 
> Apologies. I was trying to be concise, but over did it.
> 
> The issue reported is a splat in 6.6 due to FORTIFY_SOURCE complaining about
> access to an "old style" variable array declared with size 1:
>             u8 virtual_map[1];
> 
> >> Looks like this should be backported to all LTS kernels where FORTIFY_SOURCE
> >> is expected to be enabled.
> > 
> > And where would that exactly be?  Have you verified that it will apply
> > properly to those unnamed kernel trees?
> 
> I don't know which trees are actively enabling FORTIFY_SOURCE. Since the
> report is coming from 6.6, I would expect we should backport at least to there.

FWIW, FORTIFY_SOURCE is enabled in every distro kernel config I know of
(and has been for a very long time).

> I've cc'd Kees in case he has more definite information, as well as the
> wireless maintainers and the reporter in case they have anything else to add.
> 
> The actual fix is trivial, containing just a change to a data structure along
> with a related documentation change. Any conflicts would more likely occur due
> to the documentation rather than the code, since that code definition had been
> unchanged since 2012 but the documentation has been refined over time.

Right, this is fixing the splat -- no other behavioral changes were
seen.

-Kees

-- 
Kees Cook