Re: [tip:x86/seves] BUILD SUCCESS WITH WARNING e6eb15c9ba3165698488ae5c34920eea20eaa38e

From: Borislav Petkov <bp@alien8.de>
To: Nick Desaulniers <ndesaulniers@google.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>,
	Rong Chen <rong.a.chen@intel.com>,
	kernel test robot <lkp@intel.com>,
	"Li, Philip" <philip.li@intel.com>, x86-ml <x86@kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	clang-built-linux <clang-built-linux@googlegroups.com>,
	Marco Elver <elver@google.com>
Subject: Re: [tip:x86/seves] BUILD SUCCESS WITH WARNING e6eb15c9ba3165698488ae5c34920eea20eaa38e
Date: Tue, 15 Sep 2020 22:49:12 +0200	[thread overview]
Message-ID: <20200915204912.GA14436@zn.tnic> (raw)
In-Reply-To: <CAKwvOdnc8au10g8q8miab89j3tT8UhwnZOMAJdRgkXVrnkhwqQ@mail.gmail.com>

On Tue, Sep 15, 2020 at 01:12:24PM -0700, Nick Desaulniers wrote:
>       1 warning: objtool: ist_exc_vmm_communication()+0x12: unreachable instruction

That looks interesting. So your .o has:

00000000000004c0 <ist_exc_vmm_communication>:
 4c0:   55                      push   %rbp
 4c1:   48 89 e5                mov    %rsp,%rbp
 4c4:   48 c7 c7 00 00 00 00    mov    $0x0,%rdi
 4cb:   31 c0                   xor    %eax,%eax
 4cd:   e8 00 00 00 00          callq  4d2 <ist_exc_vmm_communication+0x12>
 4d2:   0f 0b                   ud2    
 4d4:   66 66 2e 0f 1f 84 00    data16 nopw %cs:0x0(%rax,%rax,1)
 4db:   00 00 00 00 
 4df:   90                      nop

 And the unreachable insn is at 0x4d2. The version I got when building with
 clang12 built from git of today is:

00000000000003e0 <ist_exc_vmm_communication>:
 3e0:   55                      push   %rbp
 3e1:   48 89 e5                mov    %rsp,%rbp
 3e4:   48 c7 c7 00 00 00 00    mov    $0x0,%rdi
 3eb:   31 c0                   xor    %eax,%eax
 3ed:   e8 00 00 00 00          callq  3f2 <ist_exc_vmm_communication+0x12>
 3f2:   66 66 2e 0f 1f 84 00    data16 nopw %cs:0x0(%rax,%rax,1)
 3f9:   00 00 00 00 
 3fd:   0f 1f 00                nopl   (%rax)

and that version is calling a bunch of NOPs.

gcc produces:

00000000000002aa <ist_exc_vmm_communication>:
 2aa:   55                      push   %rbp
 2ab:   48 c7 c7 00 00 00 00    mov    $0x0,%rdi
 2b2:   48 89 e5                mov    %rsp,%rbp
 2b5:   e8 00 00 00 00          callq  2ba <ist_exc_vmm_communication+0x10>
 2ba:   66 0f 1f 44 00 00       nopw   0x0(%rax,%rax,1)

(Btw, clang doesn't need to add that "xor %eax,%eax" - panic() should not be
 returning, ever. :-))

So what that call actually is, is:

# arch/x86/kernel/sev-es.c:1342:        panic("Can't handle #VC exception from unsupported context\n");
        call    panic   #

and the address of panic() gets fixed up by the linker into:

ffffffff83066dca <ist_exc_vmm_communication>:
ffffffff83066dca:       55                      push   %rbp
ffffffff83066dcb:       48 c7 c7 08 4f e2 83    mov    $0xffffffff83e24f08,%rdi
ffffffff83066dd2:       48 89 e5                mov    %rsp,%rbp
ffffffff83066dd5:       e8 52 23 ff ff          callq  ffffffff8305912c <panic>
ffffffff83066dda:       66 0f 1f 44 00 00       nopw   0x0(%rax,%rax,1)

But your compiler generates a call to UD2.

Interesting.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette