On Wed, Feb 10, 2016 at 02:48:02PM +0100, Michael Matz wrote:
> Hi,
> 
> On Wed, 10 Feb 2016, Borislav Petkov wrote:
> 
> > --- a/arch/x86/include/asm/tlbflush.h
> > +++ b/arch/x86/include/asm/tlbflush.h
> > @@ -23,7 +23,7 @@ static inline void __invpcid(unsigned long pcid, unsigned 
> > long addr,
> >      * invpcid (%rcx), %rax in long mode.
> >      */
> >     asm volatile (".byte 0x66, 0x0f, 0x38, 0x82, 0x01"
> > -                 : : "m" (desc), "a" (type), "c" (desc) : "memory");
> > +                 : : "m" (*desc), "a" (type), "c" (desc) : "memory");
> 
> That still doesn't do what you want.  Arrays in C are funny.  *desc is 
> exactly equivalent to desc[0], _not_ to the whole array,

Doh!

> indeed there's no C syntax to name an lvalue of array type in normal
> expressions. You need to jump through hoops for this:
>
>   "m" (*(struct {unsigned long x[2];} *)desc)

Aha! That's why we wrapped the array in clwb() in a struct too, btw:

static inline void clwb(volatile void *__p)
{
        volatile struct { char x[64]; } *p = __p;
        ...

> It'd probably be easier to simply declare the descriptor as a struct, 
> rather than an array, then the original syntax would have been mostly 
> correct:
> 
>   struct {u64 d[2];} desc = { pcid, addr };
>   asm ... "m" (desc), "c" (&desc)

Sounds better. Done. How does that below look like?

Thanks Micha!

---
From: Borislav Petkov <[email protected]>
Date: Wed, 10 Feb 2016 12:53:48 +0100
Subject: [PATCH -v1.1] x86/mm: Fix INVPCID asm constraint
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

So we want to specify the dependency on both @pcid and @addr so that the
compiler doesn't reorder accesses to them *before* the TLB flush. But
for that to work, we need to express this properly in the inline asm and
deref the whole desc array, not the pointer to it. See clwb() for an
example.

This fixes the build error on 32-bit:

  arch/x86/include/asm/tlbflush.h: In function ‘__invpcid’:
  arch/x86/include/asm/tlbflush.h:26:18: error: memory input 0 is not directly 
addressable

which gcc4.7 caught but 5.x didn't. Which is strange. :-\

Signed-off-by: Borislav Petkov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Luis R. Rodriguez <[email protected]>
Cc: Michael Matz <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Toshi Kani <[email protected]>
Cc: [email protected]
---
 arch/x86/include/asm/tlbflush.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 6f9e27aa2aaf..c24b4224d439 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -11,7 +11,7 @@
 static inline void __invpcid(unsigned long pcid, unsigned long addr,
                             unsigned long type)
 {
-       u64 desc[2] = { pcid, addr };
+       struct { u64 d[2]; } desc = { { pcid, addr } };
 
        /*
         * The memory clobber is because the whole point is to invalidate
@@ -23,7 +23,7 @@ static inline void __invpcid(unsigned long pcid, unsigned 
long addr,
         * invpcid (%rcx), %rax in long mode.
         */
        asm volatile (".byte 0x66, 0x0f, 0x38, 0x82, 0x01"
-                     : : "m" (desc), "a" (type), "c" (desc) : "memory");
+                     : : "m" (desc), "a" (type), "c" (&desc) : "memory");
 }
 
 #define INVPCID_TYPE_INDIV_ADDR                0
-- 
2.3.5

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

Reply via email to