From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wr1-x432.google.com (mail-wr1-x432.google.com [IPv6:2a00:1450:4864:20::432]) by sourceware.org (Postfix) with ESMTPS id 081A53858D32 for ; Mon, 18 Sep 2023 15:27:23 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 081A53858D32 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com Received: by mail-wr1-x432.google.com with SMTP id ffacd0b85a97d-32008b5e2d2so1168181f8f.0 for ; Mon, 18 Sep 2023 08:27:22 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1695050842; x=1695655642; darn=gcc.gnu.org; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:from:to:cc:subject:date :message-id:reply-to; bh=WqD7DDE50SoyVLOKYCR8QLBIElZkJfV+UiSFQHNMPZY=; b=WpI6Vs6oIQz+GPFTxOoDubSCWy9EBvw2ZPnQAu5ImYV8R4x905KurRoS9B9kdsgsJ6 zxpO/TOk/hV2+lDSVG2rwaXjFI8bTZkpE4EaqQSjywrHg6mumifnLdKsnUFqm5Y3CbFi JWRZgRoPdiez1YHGLq/dJCF5gVVGiW4Ao64jHWIoANgUrasScmo31q3IhFyGB+4tuQcy 8pUgqdrVrBsNGEI1n4HehaJDSUeSFVnbqbHHid8Z5FHYXIIYjHe0L4qUuBBouX/BXjwe 57jbvltWumgxwVpgbkGuhsVHl38rdv8wZNmlM0nC7I0fiE02Kj9Tu39nU4tWGux8F+vN qBYA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1695050842; x=1695655642; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=WqD7DDE50SoyVLOKYCR8QLBIElZkJfV+UiSFQHNMPZY=; b=ft481/wcklIKkc55TRGaPK6y3NHBSBNQEEIrm0MMFOJYASJP8pCw9WfDZBSIV3skcg WSlyUdTh7KlUVJpHDVveDcuWmHl0aimYhqrd/9/ljKTK+uc7E8rXGKRDIP5DMFRzKQIq cfZz7KBNQKA4JOuICNafYrRTBntqp4nlpOdxo9ZpgkQAKouxl3DweU28jcxMsR7Cjytr /URhmFbsyUGDKtte/hdO95lNbiGCyqD5qN8R4bq1fjJ4HFBx/ZwhRDC/p37rugUBNWT0 k4/4AWB7/l7IwN32VQcZwFVmzxoTOY7FZ1VBcpfkBdW9cgwCapQS2cCcE99Mb0QLv6u/ CJRA== X-Gm-Message-State: AOJu0Yz8LkmH3fCiKk4uDqPf+yauDW8sWcnkAB0lgUZlsoknwi8BNVmk PbUi4TysLlukB2F1kpzcxJx9HXtN0xbCwYL5IGo= X-Google-Smtp-Source: AGHT+IHdXPcuravBDJqSaOWjxauE6PKGRtB9i16G9pEe0fWkg5ly8GyQAo5lGus5Bloqdef9mtNEmdhawg7GXrGwto4= X-Received: by 2002:a5d:498d:0:b0:31f:eb45:462e with SMTP id r13-20020a5d498d000000b0031feb45462emr7531087wrq.28.1695050841333; Mon, 18 Sep 2023 08:27:21 -0700 (PDT) MIME-Version: 1.0 References: <202309151107.C2F7D9A01@keescook> <228be10b14dabe4e8216f9f93da99f629328b3e0.camel@gmail.com> In-Reply-To: From: Andrew Pinski Date: Mon, 18 Sep 2023 08:27:08 -0700 Message-ID: Subject: Re: Question on -fwrapv and -fwrapv-pointer To: Richard Biener Cc: Martin Uecker , keescook@chromium.org, gcc@gcc.gnu.org Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,FREEMAIL_FROM,RCVD_IN_DNSWL_NONE,SPF_HELO_NONE,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: On Mon, Sep 18, 2023 at 12:33=E2=80=AFAM Richard Biener via Gcc wrote: > > On Sat, Sep 16, 2023 at 10:38=E2=80=AFAM Martin Uecker via Gcc wrote: > > > > > > > > (moved to gcc@) > > > > > On Fri, Sep 15, 2023 at 08:18:28AM -0700, Andrew Pinski wrote: > > > > On Fri, Sep 15, 2023 at 8:12=E2=80=AFAM Qing Zhao wrote: > > > > > > > > > > > > > > > > > > > > > On Sep 15, 2023, at 3:43 AM, Xi Ruoyao wro= te: > > > > > > > > > > > > On Thu, 2023-09-14 at 21:41 +0000, Qing Zhao wrote: > > > > > >>>> CLANG already provided -fsanitize=3Dunsigned-integer-overflo= w. GCC > > > > > >>>> might need to do the same. > > > > > >>> > > > > > >>> NO. There is no such thing as unsigned integer overflow. That= option > > > > > >>> is badly designed and the GCC community has rejected a few ti= mes now > > > > > >>> having that sanitizer before. It is bad form to have a saniti= zer for > > > > > >>> well defined code. > > > > > >> > > > > > >> Even though unsigned integer overflow is well defined, it migh= t be > > > > > >> unintentional, shall we warn user about this? > > > > > > > > > > > > *Everything* could be unintentional and should be warned then. = GCC is a > > > > > > compiler, not an advanced AI educating the programmers. > > > > > > > > > > Well, you are right in some sense. -:) > > > > > > > > > > However, overflow is one important source for security flaws, it= =E2=80=99s important for compilers to detect > > > > > overflows in the programs in general. > > > > > > > > Except it is NOT an overflow. Rather it is wrapping. That is a big > > > > point here. unsigned wraps and does NOT overflow. Yes there is a ma= jor > > > > difference. > > > > > > Right, yes. I will try to pick my language very carefully. :) > > > > > > The practical problem I am trying to solve in the 30 million lines of > > > Linux kernel code is that of catching arithmetic wrap-around. The > > > problem is one of evolving the code -- I can't just drop -fwrapv and > > > -fwrapv-pointer because it's not possible to fix all the cases at onc= e. > > > (And we really don't want to reintroduce undefined behavior.) > > > > > > So, for signed, pointer, and unsigned types, we need: > > > > > > a) No arithmetic UB -- everything needs to have deterministic behavio= r. > > > The current solution here is "-fno-strict-overflow", which elimina= tes > > > the UB and makes sure everything wraps. > > > > > > b) A way to run-time warn/trap on overflow/underflow/wrap-around. Thi= s > > > would work with -fsanitize=3D[signed-integer|pointer]-overflow exc= ept > > > due to "a)" we always wrap. And there isn't currently coverage lik= e > > > this for unsigned (in GCC). > > > > > > Our problem is that the kernel is filled with a mix of places where t= here > > > is intended wrap-around and unintended wrap-around. We can chip away = at > > > fixing the intended wrap-around that we can find with static analyzer= s, > > > etc, but at the end of the day there is a long tail of finding the pl= aces > > > where intended wrap-around is hiding. But when the refactoring is > > > sufficiently completely, we can move the wrap-around warning to a tra= p, > > > and the kernel will not longer have this class of security flaw. > > > > > > As a real-world example, here is a bug where a u8 wraps around causin= g > > > an under-allocation that allowed for a heap overwrite: > > > > > > https://git.kernel.org/linus/6311071a0562 > > > https://elixir.bootlin.com/linux/v6.5/source/net/wireless/nl80211.c#L= 5422 > > > > > > If there were more than 255 elements in a linked list, the allocation > > > would be too small, and the second loop would write past the end of t= he > > > allocation. This is a pretty classic allocation underflow and linear > > > heap write overflow security flaw. (And it would be trivially stopped= by > > > trapping on the u8 wrap around.) > > > > > > So, I want to be able to catch that at run-time. But we also have cod= e > > > doing things like "if (ulong + offset < ulong) { ... }": > > > > > > https://elixir.bootlin.com/linux/v6.5/source/drivers/crypto/axis/artp= ec6_crypto.c#L1187 > > > > > > This is easy for a static analyzer to find and we can replace it with= a > > > non-wrapping test (e.g. __builtin_add_overflow()), but we'll not find > > > them all immediately, especially for the signed and pointer cases. > > > > > > So, I need to retain the "everything wraps" behavior while still bein= g > > > able to detect when it happens. > > > > > > Hi Kees, > > > > I have a couple of questions: > > > > Currently, my thinking was that you would use signed integers > > if you want the usual integer arithmetic rules we know from > > elementary school and if you overflow this is clearly a bug > > you can diagnose with UBsan. > > > > There are people who think that signed overflow should be > > defined to wrap, but I think this would be a severe > > mistake because then code would start to rely on it, which > > makes it then difficult to differentiate between bugs and > > intended uses (e.g. the unfortunate situation you have > > with the kernel). > > > > I assume you want to combine UBSan plus wrapping for > > production use? Or only for testing? Or in other words: > > why would testing UBSan and production with wrapping > > not be sufficient to find and fix all bugs? > > > > Wrapping would not be correct because it may lead to > > logic errors or use-after-free etc. I assume it is still > > preferred because it more deterministic than whatever comes > > out of the optimizer assuming that overflow has UB. Is this > > the reasoning applied here? > > > > > > For unsigned the intended use case is modulo arithmetic > > where wrapping is the correct behavior. At least, this > > is what I thought so far.. This seems also to be the > > position of the overall GCC community rejecting > > -fsanitize=3Dunsigned-integer-overflow. > > > > But then there are also people like Richard Seacord that > > have the position that one should use "unsigned" for > > every quantity which can not be negative, which implies > > that then the modulo use case becomes the exception. > > > > Related to this I have the following question: In the > > bug you refer to above, an unsigned variable was used > > for something that is not meant for modulo arithmetic. > > Is this done in the kernel to save space? I.e. because > > 127 would not be enough as maximum but going to i16 takes > > to much space? or is this for compatibility with some > > on-the-wire protocol? > > > > Would an attribute for variables help that tells the > > compiler that if stores to the variable wrap around > > then this is not intended and this is an error? > > > > u8 x [[nowrap]]; > > > > x =3D 256; // can be diagnosed ? > > There's a problem with representing this, it's a long-standing thing stan= ding > in the way of handling -fwrapv in the IL. Maybe C wants to have > 'nonnegative int' in addition to 'unsigned int', or 'bit int' for the > "just some two-complement bits". > > For GCC we think of -fwrapv to perform signed integer arithmetic as if > promoted to a large enough type first and then truncated to the original > type in the implementation defined manner (modulo reducing). But we > end up doing what the actual hardware does (I'm for example unsure > how that handles INT_MIN / -1 - x86 documents it as raising #DE > which it indeed does, even with -fwrapv). > > volatile int a, b; > int main () > { > a =3D -__INT_MAX__ - 1; > b =3D -1; > int z =3D a / b; > __builtin_printf ("%d\n", z); > } > > raises SIGFPE with and without -fwrapv. I'll note that we document > -fwrapv to only affect addition, subtraction and multiplication > (it also affects negation), One small addition, it also affects the function `abs` (and the corresponding builtins); the same way as negation. Thanks, Andrew > matching the libgcc *v functions we have > for -ftrapv. -fsanitize=3Dundefined instruments division but > -fsanitize=3Drecover doesn't help. Maybe there's a bug about this > recorded. The documentation of -fwrapv should likely be clarified. > > For recent work I failed to spot the C standards language on > singed overflow behavior for negation, addition, subtraction > and multiplication. I can find it for division though. Can someone > point me to where that is (moved?)? > > Richard. > > > > > > > > > Martin > > > > > > > > > > > >