Re: [GIT PULL] overflow updates for v5.15-rc1

From: Kees Cook <keescook@chromium.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Arnd Bergmann <arnd@arndb.de>, Daniel Vetter <daniel@ffwll.ch>,
	Dan Williams <dan.j.williams@intel.com>,
	Rasmus Villemoes <linux@rasmusvillemoes.dk>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	"Gustavo A. R. Silva" <gustavoars@kernel.org>,
	Keith Packard <keithp@keithp.com>,
	Nathan Chancellor <nathan@kernel.org>,
	Nick Desaulniers <ndesaulniers@google.com>,
	linux-hardening@vger.kernel.org
Subject: Re: [GIT PULL] overflow updates for v5.15-rc1
Date: Sun, 5 Sep 2021 11:31:44 -0700	[thread overview]
Message-ID: <202109051123.11E4E31@keescook> (raw)
In-Reply-To: <CAHk-=widUkzjVMW99L6OZpJc1wDnZbBbnOOzgXOMypOPoV6mjg@mail.gmail.com>

On Sun, Sep 05, 2021 at 10:36:22AM -0700, Linus Torvalds wrote:
> On Sun, Sep 5, 2021 at 12:38 AM Kees Cook <keescook@chromium.org> wrote:
> >
> > Yeech. Yeah, no, that was not expected at all. I even did test merge builds against your latest tree before sending the Pull Request. This has been in -next for weeks, too.
> 
> Sadly, I don't think linux-next checks for warnings.

Oh, I thought I'd gotten such reports from sfr before, but certainly the
0day bot and others have yelled loudly about new warnings (from earlier
iterations of this series in -next).

> I really want to enable -Werror at some point, but every time I think
> I should, I just end up worrying about another random new compiler (or
> a random old one).
> 
> We do have -Werror in various configurations (and in some sub-trees).

Yup, I think ppc and drm?

> > What was the build environment?
> 
> This is actually just bog-standard gcc-11.2 from F34, and an allmodconfig build.

Ah, fun. Yeah, I'm behind on versions, it seems. Default gcc version on
latest stable Ubuntu release is 10.3. I will go retest on the devel
release.

> > Seeing an unexpected "-Wunused-value" in your output makes me think I've got a compiler version blind-spot, with some different default flags.)
> 
> There were lots of other ones too, I just pasted a small subset. Thne
> full error log was 400+ lines. Most of those lines are just because of
> the very verbose warnings.
> 
> Three errors due to "-Werror=unused-value", but 17 each of variations on
> 
>     error: call to ‘__read_overflow’ declared with attribute error:
> detected read beyond size of object (1st parameter)
> 
> and
> 
>     warning: unsafe xyz() usage lacked '__read_overflow' warning
> 
> warnings.
> 
> Full 400+ lines (25kB) of errors/warnings messages attached in case
> you care about the whole thing and can't easily reproduce.

Yeah, the tests are designed to freak out if it gets an unexpected
warning (since it's trying to check for _expected_ warnings), but
regardless, they were not at all supposed to be spewing like this
immediately! :P

Sorry for the noise; I will get it cleaned up and re-sent.

-Kees

-- 
Kees Cook