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;
MIME-Version: 1.0
References: <20200814181617.442787-1-edumazet@google.com> <784A8221-8E96-4C81-B218-4BC4960AA34A@amacapital.net>
In-Reply-To: <784A8221-8E96-4C81-B218-4BC4960AA34A@amacapital.net>
From:   Jann Horn <jannh@google.com>
Date:   Fri, 14 Aug 2020 21:37:24 +0200
Message-ID: <CAG48ez3m9muZVPbxKUiGhUs8CM+i7C19UwwAqGwcpyxWXoqW+A@mail.gmail.com>
Subject: Re: [PATCH] x86/fsgsbase/64: Fix NULL deref in 86_fsgsbase_read_task
To:     Andy Lutomirski <luto@amacapital.net>
Cc:     Eric Dumazet <edumazet@google.com>, Ingo Molnar <mingo@kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Eric Dumazet <eric.dumazet@gmail.com>,
        syzbot <syzkaller@googlegroups.com>,
        Andy Lutomirski <luto@kernel.org>,
        "Chang S . Bae" <chang.seok.bae@intel.com>,
        Borislav Petkov <bp@alien8.de>,
        Brian Gerst <brgerst@gmail.com>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Denys Vlasenko <dvlasenk@redhat.com>,
        "H . Peter Anvin" <hpa@zytor.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Markus T Metzger <markus.t.metzger@intel.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Ravi Shankar <ravi.v.shankar@intel.com>,
        Rik van Riel <riel@surriel.com>,
        Thomas Gleixner <tglx@linutronix.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Aug 14, 2020 at 9:03 PM Andy Lutomirski <luto@amacapital.net> wrote=
:
> > On Aug 14, 2020, at 11:16 AM, Eric Dumazet <edumazet@google.com> wrote:
> >
> > =EF=BB=BFsyzbot found its way in 86_fsgsbase_read_task() [1]
> >
> > Fix is to make sure ldt pointer is not NULL
>
> Acked-by: Andy Lutomirski <luto@kernel.org>
>
> Maybe add something like this to the changelog:
>
> This can happen if ptrace() or sigreturn() pokes an LDT selector into FS =
or GS for a task with no LDT and something tries to read the base before a =
return to usermode notices the bad selector and fixes it.
>
> I=E2=80=99ll see if I can whip up a test case too.

This is the reproducer I used to test this on 4.20.17:

#include <stdio.h>
#include <unistd.h>
#include <err.h>
#include <sys/wait.h>
#include <sys/ptrace.h>
#include <sys/user.h>
#include <sys/prctl.h>
#include <signal.h>
#include <stddef.h>
#include <errno.h>

#define SEL_LDT 0x4
#define USER_RPL 0x3
#define LDT_SELECTOR(idx) (((idx)<<3) | SEL_LDT | USER_RPL)

int main(void) {
  pid_t child =3D fork();
  if (child =3D=3D -1) err(1, "fork");
  if (child =3D=3D 0) {
    prctl(PR_SET_PDEATHSIG, SIGKILL);
    while (1) pause();
  }
  if (ptrace(PTRACE_ATTACH, child, NULL, NULL)) err(1, "PTRACE_ATTACH");
  int status;
  if (waitpid(child, &status, __WALL) !=3D child) err(1, "waitpid");
  if (ptrace( PTRACE_POKEUSER, child, (void*)offsetof(struct
user_regs_struct, fs), (void*)LDT_SELECTOR(0) ))
    err(1, "PTRACE_POKEUSER");
  errno =3D 0;
  unsigned long val =3D ptrace( PTRACE_PEEKUSER, child,
(void*)offsetof(struct user_regs_struct, fs_base), NULL );
  printf("PTRACE_PEEKUSER returns user_regs_struct.fs_base =3D 0x%lx
(%m)\n", val);
}