Re: [PATCH 3/6] x86/entry: Use PUSH_AND_CLEAR_REGS for compat

[Josh Poimboeuf]

On Thu, May 19, 2022 at 07:00:09PM +0200, Peter Zijlstra wrote:
> On Thu, May 19, 2022 at 09:24:11AM -0700, Guenter Roeck wrote:
> > On Fri, May 06, 2022 at 02:14:34PM +0200, Peter Zijlstra wrote:
> > > Since the upper regs don't exist for ia32 code, preserving them
> > > doesn't hurt and it simplifies the code.
> > > 
> > > This doesn't add any attack surface that would not already be
> > > available through INT80.
> > > 
> > > Notably:
> > > 
> > >  - 32bit SYSENTER: didn't clear si, dx, cx.
> > > 
> > >  - 32bit SYSCALL, INT80: *do* clear si since the C functions don't
> > >    take a second argument.
> > > 
> > >  - 64bit: didn't clear si since the C functions take a second
> > >    argument; except the error_entry path might have only one argument,
> > >    so clearing si was missing here.
> > > 
> > > 32b SYSENTER should be clearing all those 3 registers, nothing uses them
> > > and selftests pass.
> > > 
> > > Unconditionally clear rsi since it simplifies code.
> > > 
> > > Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> > > Reviewed-by: Borislav Petkov <bp@xxxxxxx>
> > 
> > linux-next (next-20220519) crashes due to this patch when booting
> > q35:EPYC-Rome in qemu.
> 
> Could you try backing out each of the hunks one at a time? They're all
> more or less independent.
> 
> My bet with this being a #PF on an AMD machine, it's either the SI clear
> or the SYSCALL change.

I think this should fix it:

diff --git a/arch/x86/entry/calling.h b/arch/x86/entry/calling.h
index a97cc78ecb92..29b36e9e4e74 100644
--- a/arch/x86/entry/calling.h
+++ b/arch/x86/entry/calling.h
@@ -63,7 +63,7 @@ For 32-bit we have the following conventions - kernel is built with
  * for assembly code:
  */
 
-.macro PUSH_REGS rdx=%rdx rax=%rax save_ret=0
+.macro PUSH_REGS rdx=%rdx rcx=%rcx rax=%rax save_ret=0
 	.if \save_ret
 	pushq	%rsi		/* pt_regs->si */
 	movq	8(%rsp), %rsi	/* temporarily store the return address in %rsi */
@@ -73,7 +73,7 @@ For 32-bit we have the following conventions - kernel is built with
 	pushq   %rsi		/* pt_regs->si */
 	.endif
 	pushq	\rdx		/* pt_regs->dx */
-	pushq   %rcx		/* pt_regs->cx */
+	pushq   \rcx		/* pt_regs->cx */
 	pushq   \rax		/* pt_regs->ax */
 	pushq   %r8		/* pt_regs->r8 */
 	pushq   %r9		/* pt_regs->r9 */
@@ -115,8 +115,8 @@ For 32-bit we have the following conventions - kernel is built with
 
 .endm
 
-.macro PUSH_AND_CLEAR_REGS rdx=%rdx rax=%rax save_ret=0
-	PUSH_REGS rdx=\rdx, rax=\rax, save_ret=\save_ret
+.macro PUSH_AND_CLEAR_REGS rdx=%rdx rcx=%rcx rax=%rax save_ret=0
+	PUSH_REGS rdx=\rdx, rcx=\rcx, rax=\rax, save_ret=\save_ret
 	CLEAR_REGS
 .endm
 
diff --git a/arch/x86/entry/entry_64_compat.S b/arch/x86/entry/entry_64_compat.S
index ed2be3615b50..2d40dd132442 100644
--- a/arch/x86/entry/entry_64_compat.S
+++ b/arch/x86/entry/entry_64_compat.S
@@ -200,7 +200,7 @@ SYM_INNER_LABEL(entry_SYSCALL_compat_safe_stack, SYM_L_GLOBAL)
 SYM_INNER_LABEL(entry_SYSCALL_compat_after_hwframe, SYM_L_GLOBAL)
 	movl	%eax, %eax		/* discard orig_ax high bits */
 	pushq	%rax			/* pt_regs->orig_ax */
-	PUSH_AND_CLEAR_REGS rax=$-ENOSYS
+	PUSH_AND_CLEAR_REGS rax=$-ENOSYS rcx=%rbx
 	UNWIND_HINT_REGS
 
 	movq	%rsp, %rdi