Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
In-Reply-To: <1dcd116c-9384-6885-2d9e-f620c14bd955@osg.samsung.com>
References: <1dcd116c-9384-6885-2d9e-f620c14bd955@osg.samsung.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon, 16 Apr 2018 10:43:07 -0700
Message-ID: <CA+55aFwE1SR87+Xkkmfr+D0Yq3R0HRkdf1GdgScEfXXWN4ba6w@mail.gmail.com>
Subject: Re: Linux 4.17-rc1 - kernel paging errors running x86 selftests
To:     Shuah Khan <shuahkh@osg.samsung.com>
Cc:     Andy Lutomirski <luto@kernel.org>, Ingo Molnar <mingo@kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "open list:KERNEL SELFTEST FRAMEWORK" 
        <linux-kselftest@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>,
        "# .39.x" <stable@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Apr 16, 2018 at 10:01 AM, Shuah Khan <shuahkh@osg.samsung.com> wrote:
>
> [  884.496588] BUG: unable to handle kernel paging request at fffffe8000010030

This is the LDT remap area.

> [  884.496614] Oops: 0009 [#1] SMP KASAN PTI

This is RSVD + P, so it's a system read access that got a protection
fault due to reserved bits.

> [  884.496741] RIP: 0033:0x4031c2
> [  884.496745] RSP: 002b:00007ffd805b56d8 EFLAGS: 00010246

This is not actually a kernel paging request, it's all user space, but
it's user space that does a system access.

That's normal - something loading a segment in user space, and thus
accessing the system LDT.

But:

> [  884.496601] PGD 372870067 P4D 372870067 PUD 346e84067 PMD 34005f067 PTE ffffffffffffffff

WTF? What's that odd bogus PTE entry?

That's also why it gets a RSVD fault. That's just garbage. All-ones is
not a valid PTE.

The other levels look valid, although it strikes me that maybe we
shouldn't have the user bit set in the kernel page tables. I realize
that we clear it at the leaf node, but..

So the user page table is somehow badly set up.

I don't see *why* it would be badly set up, and that test works fine
for me, though.

It doesn't seem to have anything to do with KASAN, although

> [  884.650095] BUG: unable to handle kernel paging request at fffffe8000000000
> [  884.650103] PGD 363699067 P4D 363699067 PUD 3371c6067 PMD 37cfbc067 PTE ffffffffffffffff
> [  884.650112] Oops: 0009 [#2] SMP KASAN PTI
> [  884.650200] RIP: 0033:0x401471
> [  884.650203] RSP: 002b:00007fc8e6775eb0 EFLAGS: 00010206

The other one is exactly the same thing.

                 Linus