Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:8 as permitted sender) client-ip=2620:137:e000::3:8;
MIME-Version: 1.0
References: <20231206004654.2986026-1-mhal@rbox.co> <CAMzpN2jcEGBcEKbNjwMJ+VCMc-_N1GcpVkGgAhy=XzomJP-Ogw@mail.gmail.com>
In-Reply-To: <CAMzpN2jcEGBcEKbNjwMJ+VCMc-_N1GcpVkGgAhy=XzomJP-Ogw@mail.gmail.com>
From:   Linus Torvalds <torvalds@linuxfoundation.org>
Date:   Sat, 9 Dec 2023 12:08:30 -0800
Message-ID: <CAHk-=wiB2FSWb0HxgzHGJKaDRCaJ6tGSc0OVvodAiZ_QU=05NQ@mail.gmail.com>
Subject: Re: [PATCH 0/2] x86: UMIP emulation leaking kernel addresses
To:     Brian Gerst <brgerst@gmail.com>
Cc:     Michal Luczaj <mhal@rbox.co>, x86@kernel.org, tglx@linutronix.de,
        mingo@redhat.com, bp@alien8.de, dave.hansen@linux.intel.com,
        shuah@kernel.org, luto@kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sat, 9 Dec 2023 at 09:16, Brian Gerst <brgerst@gmail.com> wrote:
>
> A different way to plug this is to harden ptrace (and sigreturn) to
> verify that the segments are code or data type segments instead of
> relying on an IRET fault.

I think that is likely a good idea regardless of this particular issue.

And I don't think you need to even check the segment for any kind of
validity - all you need to check that it's a valid selector.

And we *kind* of do that already, with the x86 ptrace code checking

  static inline bool invalid_selector(u16 value)
  {
        return unlikely(value != 0 && (value & SEGMENT_RPL_MASK) != USER_RPL);
  }

but the thing is, I think we could limit that a lot more.

I think the only valid GDT entries are 0-15 (that includes the default
kernel segments, but they don't contain anything interesting), so we
could tighten that selector check to say that it has to be either a
LDT entry or a selector < 15.

So add some kind of requirement for "(value & 4) || (value < 8*16)", perhaps?

              Linus