Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.80.23.2.2\))
Subject: Re: AMD boot woe due to "x86/mm: Cleanup pgprot_4k_2_large() and
 pgprot_large_2_4k()"
From:   Qian Cai <cai@lca.pw>
In-Reply-To: <20200422170116.GA28345@lst.de>
Date:   Wed, 22 Apr 2020 17:32:00 -0400
Cc:     Borislav Petkov <bp@suse.de>,
        "Peter Zijlstra (Intel)" <peterz@infradead.org>,
        x86 <x86@kernel.org>, LKML <linux-kernel@vger.kernel.org>,
        kasan-dev <kasan-dev@googlegroups.com>
Content-Transfer-Encoding: quoted-printable
Message-Id: <2568586B-B1F7-47F9-8B6F-6A4C0E5280A8@lca.pw>
References: <1ED37D02-125F-4919-861A-371981581D9E@lca.pw>
 <20200422170116.GA28345@lst.de>
To:     Christoph Hellwig <hch@lst.de>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Apr 22, 2020, at 1:01 PM, Christoph Hellwig <hch@lst.de> wrote:
>=20
> On Wed, Apr 22, 2020 at 11:55:54AM -0400, Qian Cai wrote:
>> Reverted the linux-next commit and its dependency,
>>=20
>> a85573f7e741 ("x86/mm: Unexport __cachemode2pte_tbl=E2=80=9D)
>> 9e294786c89a (=E2=80=9Cx86/mm: Cleanup pgprot_4k_2_large() and =
pgprot_large_2_4k()=E2=80=9D)
>>=20
>> fixed crashes or hard reset on AMD machines during boot that have =
been flagged by
>> KASAN in different forms indicating some sort of memory corruption =
with this config,
>=20
> Interesting.  Your config seems to boot fine in my VM until the point
> where the lack of virtio-blk support stops it from mounting the root
> file system.
>=20
> Looking at the patch I found one bug, although that should not affect
> your config (it should use the pgprotval_t type), and one difference
> that could affect code generation, although I prefer the new version
> (use of __pgprot vs a local variable + pgprot_val()).
>=20
> Two patches attached, can you try them?
> =
<0001-x86-Use-pgprotval_t-in-protval_4k_2_large-and-pgprot.patch><0002-foo=
.patch>

This fixed the sucker,

diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
index edf9cea4871f..c54d1d0a8e3b 100644
--- a/arch/x86/mm/pgtable.c
+++ b/arch/x86/mm/pgtable.c
@@ -708,7 +708,7 @@ int pud_set_huge(pud_t *pud, phys_addr_t addr, =
pgprot_t prot)
=20
        set_pte((pte_t *)pud, pfn_pte(
                (u64)addr >> PAGE_SHIFT,
-               __pgprot(protval_4k_2_large(pgprot_val(prot) | =
_PAGE_PSE))));
+               __pgprot(protval_4k_2_large(pgprot_val(prot)) | =
_PAGE_PSE)));
=20
        return 1;
 }=