Received-SPF: pass (google.com: domain of linux-kernel+bounces-130104-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Subject: [PATCH 4/4] x86/xen: Enumerate NX from CPUID directly
To: linux-kernel@vger.kernel.org
Cc: jgross@suse.com,tglx@linutronix.de,x86@kernel.org,bp@alien8.de,Dave Hansen <dave.hansen@linux.intel.com>
From: Dave Hansen <dave.hansen@linux.intel.com>
Date: Wed, 03 Apr 2024 08:35:14 -0700
References: <20240403153508.7328E749@davehans-spike.ostc.intel.com>
In-Reply-To: <20240403153508.7328E749@davehans-spike.ostc.intel.com>
Message-Id: <20240403153514.CC2C9D13@davehans-spike.ostc.intel.com>
Precedence: bulk


From: Dave Hansen <dave.hansen@linux.intel.com>

xen_start_kernel() is some of the first C code run "Xen PV" systems.
It precedes things like setup_arch() or processor initialization code
that would be thought of as "very early".

That means that 'boot_cpu_data' is garbage.  It has not even
established the utter basics like if the CPU supports the CPUID
instruction.  Unfortunately get_cpu_cap() requires this exact
information.

Nevertheless, xen_start_kernel() calls get_cpu_cap().  But it
works out in practice because it's looking for the NX bit which
comes from an extended CPUID leaf that doesn't depend on
c->cpuid_level being set.

Stop calling get_cpu_cap().  Use CPUID directly.  Add a little
helper to avoid cluttering up the xen_start_kernel() flow.

This removes the risky, barely-working call to get_cpu_cap().

Note: The '& 31' strips the "capability" word off of an
      X86_FEATURE_* value.  It's a magic number, but there are
      a few of them sparsely scattered around and it's way
      shorter than any helper.

Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Juergen Gross <jgross@suse.com>
---

 b/arch/x86/xen/enlighten_pv.c |   19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff -puN arch/x86/xen/enlighten_pv.c~xen-explain1 arch/x86/xen/enlighten_pv.c
--- a/arch/x86/xen/enlighten_pv.c~xen-explain1	2024-04-02 15:23:00.046913557 -0700
+++ b/arch/x86/xen/enlighten_pv.c	2024-04-02 15:23:00.050913562 -0700
@@ -1306,6 +1306,21 @@ static void __init xen_domu_set_legacy_f
 
 extern void early_xen_iret_patch(void);
 
+/*
+ * It is too early for boot_cpu_has() and friends to work.
+ * Use CPUID to directly enumerate NX support.
+ */
+static inline void xen_configure_nx(void)
+{
+	bool nx_supported;
+	u32 eax = 0;
+
+	get_cpuid_region_leaf(0x80000001, CPUID_EDX, &eax);
+
+	nx_supported = eax & (X86_FEATURE_NX & 31);
+	x86_configure_nx(nx_supported);
+}
+
 /* First C function to be called on Xen boot */
 asmlinkage __visible void __init xen_start_kernel(struct start_info *si)
 {
@@ -1369,9 +1384,7 @@ asmlinkage __visible void __init xen_sta
 	/* Get mfn list */
 	xen_build_dynamic_phys_to_machine();
 
-	/* Work out if we support NX */
-	get_cpu_cap(&boot_cpu_data);
-	x86_configure_nx(boot_cpu_has(X86_FEATURE_NX));
+	xen_configure_nx();
 
 	/*
 	 * Set up kernel GDT and segment registers, mainly so that
_