Message-ID: <2b94afd608303f104376e6a775b211714e34bc7e.camel@intel.com>
Date:   Tue, 28 Jun 2022 10:34:11 +1200
From:   Kai Huang <kai.huang@...el.com>
To:     Dave Hansen <dave.hansen@...el.com>, linux-kernel@...r.kernel.org,
        kvm@...r.kernel.org
Cc:     seanjc@...gle.com, pbonzini@...hat.com, len.brown@...el.com,
        tony.luck@...el.com, rafael.j.wysocki@...el.com,
        reinette.chatre@...el.com, dan.j.williams@...el.com,
        peterz@...radead.org, ak@...ux.intel.com,
        kirill.shutemov@...ux.intel.com,
        sathyanarayanan.kuppuswamy@...ux.intel.com,
        isaku.yamahata@...el.com
Subject: Re: [PATCH v5 08/22] x86/virt/tdx: Shut down TDX module in case of
 error

On Mon, 2022-06-27 at 13:46 -0700, Dave Hansen wrote:
> On 6/26/22 22:26, Kai Huang wrote:
> > On Fri, 2022-06-24 at 11:50 -0700, Dave Hansen wrote:
> > > So, the last patch was called:
> > > 
> > > 	Implement SEAMCALL function
> > > 
> > > and yet, in this patch, we have a "seamcall()" function.  That's a bit
> > > confusing and not covered at *all* in this subject.
> > > 
> > > Further, seamcall() is the *ONLY* caller of __seamcall() that I see in
> > > this series.  That makes its presence here even more odd.
> > > 
> > > The seamcall() bits should either be in their own patch, or mashed in
> > > with __seamcall().
> > 
> > Right.  The reason I didn't put the seamcall() into previous patch was it is
> > only used in this tdx.c, so it should be static.  But adding a static function
> > w/o using it in previous patch will trigger a compile warning.  So I introduced
> > here where it is first used.
> > 
> > One option is I can introduce seamcall() as a static inline function in tdx.h in
> > previous patch so there won't be a warning.  I'll change to use this way. 
> > Please let me know if you have any comments.
> 
> Does a temporary __unused get rid of the warning?

Yes, and both __maybe_unused and __always_unused also git rid of the warning
too.

__unused is not defined in compiler_attributes.h, so we need to use
__attribute__((__unused__)) explicitly, or have __unused defined to it as a
macro.

I think I can just use __always_unused for this purpose?

So I think we put seamcall() implementation to the patch which implements
__seamcall().  And we can inline for seamcall() and put it in either tdx.h or
tdx.c, or we can use __always_unused  (or the one you prefer) to get rid of the
warning.

What's your opinion?
> 
> > > >  /*
> > > >   * Detect and initialize the TDX module.
> > > >   *
> > > > @@ -138,7 +195,10 @@ static int init_tdx_module(void)
> > > >  
> > > >  static void shutdown_tdx_module(void)
> > > >  {
> > > > -	/* TODO: Shut down the TDX module */
> > > > +	struct seamcall_ctx sc = { .fn = TDH_SYS_LP_SHUTDOWN };
> > > > +
> > > > +	seamcall_on_each_cpu(&sc);
> > > > +
> > > >  	tdx_module_status = TDX_MODULE_SHUTDOWN;
> > > >  }
> > > >  
> > > > @@ -221,6 +281,9 @@ bool platform_tdx_enabled(void)
> > > >   * CPU hotplug is temporarily disabled internally to prevent any cpu
> > > >   * from going offline.
> > > >   *
> > > > + * Caller also needs to guarantee all CPUs are in VMX operation during
> > > > + * this function, otherwise Oops may be triggered.
> > > 
> > > I would *MUCH* rather have this be a:
> > > 
> > > 	if (!cpu_feature_enabled(X86_FEATURE_VMX))
> > > 		WARN_ONCE("VMX should be on blah blah\n");
> > > 
> > > than just plain oops.  Even a pr_err() that preceded the oops would be
> > > nicer than an oops that someone has to go decode and then grumble when
> > > their binutils is too old that it can't disassemble the TDCALL.
> > 
> > I can add this to seamcall():
> > 
> > 	/*
> > 	 * SEAMCALL requires CPU being in VMX operation otherwise it causes
> > #UD.
> > 	 * Sanity check and return early to avoid Oops.  Note cpu_vmx_enabled()
> > 	 * actually only checks whether VMX is enabled but doesn't check
> > whether
> > 	 * CPU is in VMX operation (VMXON is done).  There's no way to check
> > 	 * whether VMXON has been done, but currently enabling VMX and doing
> > 	 * VMXON are always done together.
> > 	 */
> > 	if (!cpu_vmx_enabled())	 {
> > 		WARN_ONCE("CPU is not in VMX operation before making
> > SEAMCALL");
> > 		return -EINVAL;
> > 	}
> > 
> > The reason I didn't do is I'd like to make seamcall() simple, that it only
> > returns TDX_SEAMCALL_VMFAILINVALID or the actual SEAMCALL leaf error.  With
> > above, this function also returns kernel error code, which isn't good.
> 
> I think you're missing the point.  You wasted two lines of code on a
> *COMMENT* that doesn't actually help anyone decode an oops.  You could
> have, instead, spent two lines on actual code that would have been just
> as good or better than a comment *AND* help folks looking at an oops.
> 
> It's almost always better to do something actionable in code than to
> comment it, unless it's in some crazy fast path.

Agreed.  Thanks.

> 
> > Alternatively, we can always add EXTABLE to TDX_MODULE_CALL macro to handle #UD
> > and #GP by returning dedicated error codes (please also see my reply to previous
> > patch for the code needed to handle), in which case we don't need such check
> > here.
> > 
> > Always handling #UD in TDX_MODULE_CALL macro also has another advantage:  there
> > will be no Oops for #UD regardless the issue that "there's no way to check
> > whether VMXON has been done" in the above comment.
> > 
> > What's your opinion?
> 
> I think you should explore using the EXTABLE.  Let's see how it looks.

I tried to wrote the code before.  I didn't test but it should look like to
something below.  Any comments?

diff --git a/arch/x86/include/asm/tdx.h b/arch/x86/include/asm/tdx.h
index 4b75c930fa1b..4a97ca8eb14c 100644
--- a/arch/x86/include/asm/tdx.h
+++ b/arch/x86/include/asm/tdx.h
@@ -8,6 +8,7 @@
 #include <asm/ptrace.h>
 #include <asm/shared/tdx.h>

+#ifdef CONFIG_INTEL_TDX_HOST
 /*
  * SW-defined error codes.
  *
@@ -18,6 +19,21 @@
 #define TDX_SW_ERROR                   (TDX_ERROR | GENMASK_ULL(47, 40))
 #define TDX_SEAMCALL_VMFAILINVALID     (TDX_SW_ERROR | _UL(0xFFFF0000))

+/*
+ * Special error codes to indicate SEAMCALL #GP and #UD.
+ *
+ * SEAMCALL causes #GP when SEAMRR is not properly enabled by BIOS, and
+ * causes #UD when CPU is not in VMX operation.  Define two separate
+ * error codes to distinguish the two cases so caller can be aware of
+ * what caused the SEAMCALL to fail.
+ *
+ * Bits 61:48 are reserved bits which will never be set by the TDX
+ * module.  Borrow 2 reserved bits to represent #GP and #UD.
+ */
+#define TDX_SEAMCALL_GP                (TDX_ERROR | GENMASK_ULL(48, 48))
+#define TDX_SEAMCALL_UD                (TDX_ERROR | GENMASK_ULL(49, 49))
+#endif
+
 #ifndef __ASSEMBLY__

 /*
diff --git a/arch/x86/virt/vmx/tdx/tdxcall.S b/arch/x86/virt/vmx/tdx/tdxcall.S
index 49a54356ae99..7431c47258d9 100644
--- a/arch/x86/virt/vmx/tdx/tdxcall.S
+++ b/arch/x86/virt/vmx/tdx/tdxcall.S
@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #include <asm/asm-offsets.h>
 #include <asm/tdx.h>
+#include <asm/asm.h>

 /*
  * TDCALL and SEAMCALL are supported in Binutils >= 2.36.
@@ -45,6 +46,7 @@
        /* Leave input param 2 in RDX */

        .if \host
+1:
        seamcall
        /*
         * SEAMCALL instruction is essentially a VMExit from VMX root
@@ -57,9 +59,25 @@
         * This value will never be used as actual SEAMCALL error code as
         * it is from the Reserved status code class.
         */
-       jnc .Lno_vmfailinvalid
+       jnc .Lseamcall_out
        mov $TDX_SEAMCALL_VMFAILINVALID, %rax
-.Lno_vmfailinvalid:
+       jmp .Lseamcall_out
+2:
+       /*
+        * SEAMCALL caused #GP or #UD.  By reaching here %eax contains
+        * the trap number.  Check the trap number and set up the return
+        * value to %rax.
+        */
+       cmp $X86_TRAP_GP, %eax
+       je .Lseamcall_gp
+       mov $TDX_SEAMCALL_UD, %rax
+       jmp .Lseamcall_out
+.Lseamcall_gp:
+       mov $TDX_SEAMCALL_GP, %rax
+       jmp .Lseamcall_out
+
+       _ASM_EXTABLE_FAULT(1b, 2b)
+.Lseamcall_out

Powered by blists - more mailing lists