Re: PROBLEM: consolidated IDT invalidation causes kexec to reboot

[Alexandru Chirvasitu]

Thank you for the swift reply!

On Sat, Dec 23, 2017 at 07:30:21PM -0800, Linus Torvalds wrote:
> On Sat, Dec 23, 2017 at 5:44 PM, Alexandru Chirvasitu
> <achirvasub@xxxxxxxxx> wrote:
> >
> > For testing purposes, I've altered machine_kexec_32.c making the
> > following toy commit. It naively undoes part of e802a51, solely to
> > confirm that's where it goes awry in my setup.
> 
> That's really funky.
> 
> The idt_invalidate() seems to do *exactly* the same thing. It uses
> "load_idt()" on an IDT with size 0, and the supplied address.
> 
> Can you disassemble your "set_idt()" code vs the "idt_invalidate()"?
>

I seem to have done some such thing just now, but please excuse some
poking around in the dark here (I've disassembled code exactly once
before: yesterday, in answering a similar request for more info in
another lkml thread..).

I'm actually not even certain the sequel is what you are asking.

I couldn't find the set_idt symbol in
arch/x86/kernel/machine_kexec_32.o. Google seemed to think this has
something to do with the 'static' marker for that function, so I made
another commit differing from the previous one only in that it removes
that marker (i.e. set_idt is now 'void' rather than 'static void').

I can now see that function with objdump. The relevant sections of
objdump -D on the two files are:

--- arch/x86/kernel/machine_kexec_32.o ---

00000180 <set_idt>:
 180:   e8 fc ff ff ff          call   181 <set_idt+0x1>
 185:   55                      push   %ebp
 186:   89 e5                   mov    %esp,%ebp
 188:   83 ec 0c                sub    $0xc,%esp
 18b:   89 45 f8                mov    %eax,-0x8(%ebp)
 18e:   66 89 55 f6             mov    %dx,-0xa(%ebp)
 192:   8d 45 f6                lea    -0xa(%ebp),%eax
 195:   65 8b 0d 14 00 00 00    mov    %gs:0x14,%ecx
 19c:   89 4d fc                mov    %ecx,-0x4(%ebp)
 19f:   31 c9                   xor    %ecx,%ecx
 1a1:   ff 15 20 00 00 00       call   *0x20
 1a7:   8b 45 fc                mov    -0x4(%ebp),%eax
 1aa:   65 33 05 14 00 00 00    xor    %gs:0x14,%eax
 1b1:   75 02                   jne    1b5 <set_idt+0x35>
 1b3:   c9                      leave  
 1b4:   c3                      ret    
 1b5:   e8 fc ff ff ff          call   1b6 <set_idt+0x36>
 1ba:   8d b6 00 00 00 00       lea    0x0(%esi),%esi

----------------------------------------------

and

--- arch/x86/kernel/idt.o  ---

00000000 <idt_invalidate>:
   0:   e8 fc ff ff ff          call   1 <idt_invalidate+0x1>
   5:   55                      push   %ebp
   6:   89 e5                   mov    %esp,%ebp
   8:   83 ec 0c                sub    $0xc,%esp
   b:   65 8b 15 14 00 00 00    mov    %gs:0x14,%edx
  12:   89 55 fc                mov    %edx,-0x4(%ebp)
  15:   31 d2                   xor    %edx,%edx
  17:   31 d2                   xor    %edx,%edx
  19:   89 45 f8                mov    %eax,-0x8(%ebp)
  1c:   8d 45 f6                lea    -0xa(%ebp),%eax
  1f:   66 89 55 f6             mov    %dx,-0xa(%ebp)
  23:   ff 15 20 00 00 00       call   *0x20
  29:   8b 45 fc                mov    -0x4(%ebp),%eax
  2c:   65 33 05 14 00 00 00    xor    %gs:0x14,%eax
  33:   75 02                   jne    37 <idt_invalidate+0x37>
  35:   c9                      leave  
  36:   c3                      ret    
  37:   e8 fc ff ff ff          call   38 <idt_invalidate+0x38>

-------------------------------

I've also checked again that this newer compilation still gives a
well-behaved kexec.

> > Is this expected behaviour?
> 
> No. The code literally seems identical. The only difference is
> 
>  (a) where the 0 limit comes from
> 
>  (b) perhaps build flags and whether it is inlined or not due to being
> in a different file
> 
> and neither of those should matter, but maybe they do.
> 
> Which is why I'd like you to actually look at the generated code and
> see if you can see any difference..
> 
>             Linus