Received-SPF: pass (google.com: domain of linux-kernel+bounces-174791-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Date: Thu, 9 May 2024 19:48:06 +0100
From: Al Viro <viro@zeniv.linux.org.uk>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Theodore Ts'o <tytso@mit.edu>, Kees Cook <keescook@chromium.org>,
	Justin Stitt <justinstitt@google.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Mark Rutland <mark.rutland@arm.com>,
	linux-hardening@vger.kernel.org, linux-kernel@vger.kernel.org,
	llvm@lists.linux.dev
Subject: Re: [RFC] Mitigating unexpected arithmetic overflow
Message-ID: <20240509184806.GS2118490@ZenIV>
References: <202405081144.D5FCC44A@keescook>
 <CAHk-=wjeiGb1UxCy6Q8aif50C=wWDX9Pgp+WbZYrO72+B1f_QA@mail.gmail.com>
 <202405081354.B0A8194B3C@keescook>
 <CAHk-=wgoE5EkH+sQwi4KhRhCZizUxwZAnC=+9RbZcw7g6016LQ@mail.gmail.com>
 <202405081949.0565810E46@keescook>
 <20240509140854.GF3620298@mit.edu>
 <CAHk-=wgKyP2ffZPa6aKYtytzzFibiNVN5MS=D2cn7_UGCECKdw@mail.gmail.com>
 <20240509175417.GR2118490@ZenIV>
 <CAHk-=wgXm15gZHvt8waSFhXS9yZTfvMV95xyvNtPc6FSEA_rvA@mail.gmail.com>
 <CAHk-=wgBprh=8Us-MtwH9sVNELZK2hdOkFn3EoauwecYgtXOCQ@mail.gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAHk-=wgBprh=8Us-MtwH9sVNELZK2hdOkFn3EoauwecYgtXOCQ@mail.gmail.com>
Sender: Al Viro <viro@ftp.linux.org.uk>

On Thu, May 09, 2024 at 11:39:04AM -0700, Linus Torvalds wrote:

> .. it might also actually be a good idea *IF* we were to have some
> kind of "implicit cast drops bits" warning, in that the compiler for
> that case wouldn't remove the upper bits calculation, but would
> trigger a warning if they are non-zero.
> 
> So there are actually potential advantages to just always apparently
> doing the full 64-bit arithmetic.
> 
> Without debug warnings, it's a no-op that the compiler will just skip.
> And with some hypothetical debug flag, it would be a "you are now
> losing the high bits of the time value when assigning the result to a
> limited 32-bit time_t" warning.

FWIW, the thing that somewhat worries me about having a helper along
the lines of combine_to_u64(low, high) is that
	foo->splat = combine_to_u64(something, something_else);
would be inviting hard-to-catch brainos -
	foo->splat = combine_to_u64(something_else, something);
would be very hard to catch on RTFS, especially when you'd been
staring at that code for a long time.  Explicitly spelled out
it would be obvious which goes into bits 0..31 and which in 32..64.