From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-8.2 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_HELO_NONE,SPF_PASS, UNPARSEABLE_RELAY,USER_AGENT_SANE_2 autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id C2596C47404 for ; Thu, 3 Oct 2019 02:18:06 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 836DA222C2 for ; Thu, 3 Oct 2019 02:18:06 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 836DA222C2 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=mediatek.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id 023446B0005; Wed, 2 Oct 2019 22:18:06 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id F3CC56B0006; Wed, 2 Oct 2019 22:18:05 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id E52FD6B0007; Wed, 2 Oct 2019 22:18:05 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0124.hostedemail.com [216.40.44.124]) by kanga.kvack.org (Postfix) with ESMTP id BDD1A6B0005 for ; Wed, 2 Oct 2019 22:18:05 -0400 (EDT) Received: from smtpin18.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay03.hostedemail.com (Postfix) with SMTP id 6CDF2824CA3D for ; Thu, 3 Oct 2019 02:18:05 +0000 (UTC) X-FDA: 76000863330.18.walk02_c2f593091c29 X-HE-Tag: walk02_c2f593091c29 X-Filterd-Recvd-Size: 12153 Received: from mailgw02.mediatek.com (unknown [210.61.82.184]) by imf23.hostedemail.com (Postfix) with ESMTP for ; Thu, 3 Oct 2019 02:18:03 +0000 (UTC) X-UUID: 6a934931ebee47a7b59356e638e1e5f6-20191003 X-UUID: 6a934931ebee47a7b59356e638e1e5f6-20191003 Received: from mtkcas08.mediatek.inc [(172.21.101.126)] by mailgw02.mediatek.com (envelope-from ) (Cellopoint E-mail Firewall v4.1.10 Build 0809 with TLS) with ESMTP id 356317875; Thu, 03 Oct 2019 10:18:00 +0800 Received: from mtkcas09.mediatek.inc (172.21.101.178) by mtkmbs07n1.mediatek.inc (172.21.101.16) with Microsoft SMTP Server (TLS) id 15.0.1395.4; Thu, 3 Oct 2019 10:17:59 +0800 Received: from [172.21.84.99] (172.21.84.99) by mtkcas09.mediatek.inc (172.21.101.73) with Microsoft SMTP Server id 15.0.1395.4 via Frontend Transport; Thu, 3 Oct 2019 10:17:58 +0800 Message-ID: <1570069078.19702.57.camel@mtksdccf07> Subject: Re: [PATCH] kasan: fix the missing underflow in memmove and memcpy with CONFIG_KASAN_GENERIC=y From: Walter Wu To: Dmitry Vyukov CC: Andrey Ryabinin , Alexander Potapenko , Matthias Brugger , LKML , kasan-dev , Linux-MM , Linux ARM , , wsd_upstream Date: Thu, 3 Oct 2019 10:17:58 +0800 In-Reply-To: References: <20190927034338.15813-1-walter-zh.wu@mediatek.com> <1569594142.9045.24.camel@mtksdccf07> <1569818173.17361.19.camel@mtksdccf07> <1570018513.19702.36.camel@mtksdccf07> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.2.3-0ubuntu6 MIME-Version: 1.0 X-MTK: N Content-Transfer-Encoding: quoted-printable X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On Wed, 2019-10-02 at 15:57 +0200, Dmitry Vyukov wrote: > On Wed, Oct 2, 2019 at 2:15 PM Walter Wu wr= ote: > > > > On Mon, 2019-09-30 at 12:36 +0800, Walter Wu wrote: > > > On Fri, 2019-09-27 at 21:41 +0200, Dmitry Vyukov wrote: > > > > On Fri, Sep 27, 2019 at 4:22 PM Walter Wu wrote: > > > > > > > > > > On Fri, 2019-09-27 at 15:07 +0200, Dmitry Vyukov wrote: > > > > > > On Fri, Sep 27, 2019 at 5:43 AM Walter Wu wrote: > > > > > > > > > > > > > > memmove() and memcpy() have missing underflow issues. > > > > > > > When -7 <=3D size < 0, then KASAN will miss to catch the un= derflow issue. > > > > > > > It looks like shadow start address and shadow end address i= s the same, > > > > > > > so it does not actually check anything. > > > > > > > > > > > > > > The following test is indeed not caught by KASAN: > > > > > > > > > > > > > > char *p =3D kmalloc(64, GFP_KERNEL); > > > > > > > memset((char *)p, 0, 64); > > > > > > > memmove((char *)p, (char *)p + 4, -2); > > > > > > > kfree((char*)p); > > > > > > > > > > > > > > It should be checked here: > > > > > > > > > > > > > > void *memmove(void *dest, const void *src, size_t len) > > > > > > > { > > > > > > > check_memory_region((unsigned long)src, len, false,= _RET_IP_); > > > > > > > check_memory_region((unsigned long)dest, len, true,= _RET_IP_); > > > > > > > > > > > > > > return __memmove(dest, src, len); > > > > > > > } > > > > > > > > > > > > > > We fix the shadow end address which is calculated, then gen= eric KASAN > > > > > > > get the right shadow end address and detect this underflow = issue. > > > > > > > > > > > > > > [1] https://bugzilla.kernel.org/show_bug.cgi?id=3D199341 > > > > > > > > > > > > > > Signed-off-by: Walter Wu > > > > > > > Reported-by: Dmitry Vyukov > > > > > > > --- > > > > > > > lib/test_kasan.c | 36 ++++++++++++++++++++++++++++++++++= ++ > > > > > > > mm/kasan/generic.c | 8 ++++++-- > > > > > > > 2 files changed, 42 insertions(+), 2 deletions(-) > > > > > > > > > > > > > > diff --git a/lib/test_kasan.c b/lib/test_kasan.c > > > > > > > index b63b367a94e8..8bd014852556 100644 > > > > > > > --- a/lib/test_kasan.c > > > > > > > +++ b/lib/test_kasan.c > > > > > > > @@ -280,6 +280,40 @@ static noinline void __init kmalloc_oo= b_in_memset(void) > > > > > > > kfree(ptr); > > > > > > > } > > > > > > > > > > > > > > +static noinline void __init kmalloc_oob_in_memmove_underfl= ow(void) > > > > > > > +{ > > > > > > > + char *ptr; > > > > > > > + size_t size =3D 64; > > > > > > > + > > > > > > > + pr_info("underflow out-of-bounds in memmove\n"); > > > > > > > + ptr =3D kmalloc(size, GFP_KERNEL); > > > > > > > + if (!ptr) { > > > > > > > + pr_err("Allocation failed\n"); > > > > > > > + return; > > > > > > > + } > > > > > > > + > > > > > > > + memset((char *)ptr, 0, 64); > > > > > > > + memmove((char *)ptr, (char *)ptr + 4, -2); > > > > > > > + kfree(ptr); > > > > > > > +} > > > > > > > + > > > > > > > +static noinline void __init kmalloc_oob_in_memmove_overflo= w(void) > > > > > > > +{ > > > > > > > + char *ptr; > > > > > > > + size_t size =3D 64; > > > > > > > + > > > > > > > + pr_info("overflow out-of-bounds in memmove\n"); > > > > > > > + ptr =3D kmalloc(size, GFP_KERNEL); > > > > > > > + if (!ptr) { > > > > > > > + pr_err("Allocation failed\n"); > > > > > > > + return; > > > > > > > + } > > > > > > > + > > > > > > > + memset((char *)ptr, 0, 64); > > > > > > > + memmove((char *)ptr + size, (char *)ptr, 2); > > > > > > > + kfree(ptr); > > > > > > > +} > > > > > > > + > > > > > > > static noinline void __init kmalloc_uaf(void) > > > > > > > { > > > > > > > char *ptr; > > > > > > > @@ -734,6 +768,8 @@ static int __init kmalloc_tests_init(vo= id) > > > > > > > kmalloc_oob_memset_4(); > > > > > > > kmalloc_oob_memset_8(); > > > > > > > kmalloc_oob_memset_16(); > > > > > > > + kmalloc_oob_in_memmove_underflow(); > > > > > > > + kmalloc_oob_in_memmove_overflow(); > > > > > > > kmalloc_uaf(); > > > > > > > kmalloc_uaf_memset(); > > > > > > > kmalloc_uaf2(); > > > > > > > diff --git a/mm/kasan/generic.c b/mm/kasan/generic.c > > > > > > > index 616f9dd82d12..34ca23d59e67 100644 > > > > > > > --- a/mm/kasan/generic.c > > > > > > > +++ b/mm/kasan/generic.c > > > > > > > @@ -131,9 +131,13 @@ static __always_inline bool memory_is_= poisoned_n(unsigned long addr, > > > > > > > size_t size= ) > > > > > > > { > > > > > > > unsigned long ret; > > > > > > > + void *shadow_start =3D kasan_mem_to_shadow((void *)= addr); > > > > > > > + void *shadow_end =3D kasan_mem_to_shadow((void *)ad= dr + size - 1) + 1; > > > > > > > > > > > > > > - ret =3D memory_is_nonzero(kasan_mem_to_shadow((void= *)addr), > > > > > > > - kasan_mem_to_shadow((void *)addr + = size - 1) + 1); > > > > > > > + if ((long)size < 0) > > > > > > > + shadow_end =3D kasan_mem_to_shadow((void *)= addr + size); > > > > > > > > > > > > Hi Walter, > > > > > > > > > > > > Thanks for working on this. > > > > > > > > > > > > If size<0, does it make sense to continue at all? We will sti= ll check > > > > > > 1PB of shadow memory? What happens when we pass such huge ran= ge to > > > > > > memory_is_nonzero? > > > > > > Perhaps it's better to produce an error and bail out immediat= ely if size<0? > > > > > > > > > > I agree with what you said. when size<0, it is indeed an unreas= onable > > > > > behavior, it should be blocked from continuing to do. > > > > > > > > > > > > > > > > Also, what's the failure mode of the tests? Didn't they badly= corrupt > > > > > > memory? We tried to keep tests such that they produce the KAS= AN > > > > > > reports, but don't badly corrupt memory b/c/ we need to run a= ll of > > > > > > them. > > > > > > > > > > Maybe we should first produce KASAN reports and then go to exec= ute > > > > > memmove() or do nothing? It looks like it=E2=80=99s doing the f= ollowing.or? > > > > > > > > > > void *memmove(void *dest, const void *src, size_t len) > > > > > { > > > > > + if (long(len) <=3D 0) > > > > > > > > /\/\/\/\/\/\ > > > > > > > > This check needs to be inside of check_memory_region, otherwise w= e > > > > will have similar problems in all other places that use > > > > check_memory_region. > > > Thanks for your reminder. > > > > > > bool check_memory_region(unsigned long addr, size_t size, bool wri= te, > > > unsigned long ret_ip) > > > { > > > + if (long(size) < 0) { > > > + kasan_report_invalid_size(src, dest, len, _RET_IP_)= ; > > > + return false; > > > + } > > > + > > > return check_memory_region_inline(addr, size, write, ret_ip= ); > > > } > > > > > > > But check_memory_region already returns a bool, so we could check= that > > > > bool and return early. > > > > > > When size<0, we should only show one KASAN report, and should we on= ly > > > limit to return when size<0 is true? If yse, then __memmove() will = do > > > nothing. > > > > > > > > > void *memmove(void *dest, const void *src, size_t len) > > > { > > > - check_memory_region((unsigned long)src, len, false, _RET_IP= _); > > > + if(!check_memory_region((unsigned long)src, len, false, > > > _RET_IP_) > > > + && long(size) < 0) > > > + return; > > > + > > > check_memory_region((unsigned long)dest, len, true, _RET_IP= _); > > > > > > return __memmove(dest, src, len); > > > > > > > > > Hi Dmitry, > > > > What do you think the following code is better than the above one. > > In memmmove/memset/memcpy, they need to determine whether size < 0 is > > true. we directly determine whether size is negative in memmove and > > return early. it avoid to generate repeated KASAN report. Is it bette= r? > > > > void *memmove(void *dest, const void *src, size_t len) > > { > > + if (long(size) < 0) { > > + kasan_report_invalid_size(src, dest, len, _RET_IP_); > > + return; > > + } > > + > > check_memory_region((unsigned long)src, len, false, _RET_IP_)= ; > > check_memory_region((unsigned long)dest, len, true, _RET_IP_)= ; > > > > > > check_memory_region() still has to check whether the size is negative= . > > but memmove/memset/memcpy generate invalid size KASAN report will not= be > > there. >=20 >=20 > If check_memory_region() will do the check, why do we need to > duplicate it inside of memmove and all other range functions? >=20 Yes, I know it has duplication, but if we don't have to determine size<0 in memmove, then all check_memory_region return false will do nothing, it includes other memory corruption behaviors, this is my original concern.=20 > I would do: >=20 > void *memmove(void *dest, const void *src, size_t len) > { > if (check_memory_region((unsigned long)src, len, false, _RET_IP= _)) > return; if check_memory_region return TRUE is to do nothing, but it is no memory corruption? Should it return early when check_memory_region return a FALSE? >=20 > This avoids duplicating the check, adds minimal amount of code to > range functions and avoids adding kasan_report_invalid_size. Thanks for your suggestion. We originally want to show complete information(destination address, source address, and its length), but add minimal amount of code into kasan_report(), it should be good.