[PATCH 1/3] x86/pv: Options to disable and/or compile out 32bit PV support

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

This is the start of some performance and security-hardening improvements,
based on the fact that 32bit PV guests are few and far between these days.

Ring1 is full or architectural corner cases, such as counting as supervisor
from a paging point of view.  This accounts for a substantial performance hit
on processors from the last 8 years (adjusting SMEP/SMAP on every privilege
transition), and the gap is only going to get bigger with new hardware
features.

Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
---
CC: Jan Beulich <JBeulich@xxxxxxxx>
CC: Wei Liu <wl@xxxxxxx>
CC: Roger Pau Monné <roger.pau@xxxxxxxxxx>

There is a series I can't quite post yet which wants to conditionally turn
opt_pv32 off, which is why I've put it straight in in an int8_t form rather
than a straight boolean form.
---
 docs/misc/xen-command-line.pandoc | 12 +++++++++++-
 xen/arch/x86/Kconfig              | 16 ++++++++++++++++
 xen/arch/x86/pv/domain.c          | 35 +++++++++++++++++++++++++++++++++++
 xen/arch/x86/setup.c              |  9 +++++++--
 xen/include/asm-x86/pv/domain.h   |  6 ++++++
 5 files changed, 75 insertions(+), 3 deletions(-)

diff --git a/docs/misc/xen-command-line.pandoc 
b/docs/misc/xen-command-line.pandoc
index acd0b3d994..ee12b0f53f 100644
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -1694,7 +1694,17 @@ The following resources are available:
     CDP, one COS will corespond two CBMs other than one with CAT, due to the
     sum of CBMs is fixed, that means actual `cos_max` in use will automatically
     reduce to half when CDP is enabled.
-       
+
+### pv
+    = List of [ 32=<bool> ]
+
+    Applicability: x86
+
+Controls for aspects of PV guest support.
+
+*   The `32` boolean controls whether 32bit PV guests can be created.  It
+    defaults to `true`, and is ignored when `CONFIG_PV32` is compiled out.
+
 ### pv-linear-pt (x86)
 > `= <boolean>`
 
diff --git a/xen/arch/x86/Kconfig b/xen/arch/x86/Kconfig
index 8149362bde..4c52197de3 100644
--- a/xen/arch/x86/Kconfig
+++ b/xen/arch/x86/Kconfig
@@ -49,6 +49,22 @@ config PV
 
          If unsure, say Y.
 
+config PV32
+       bool "Support for 32bit PV guests"
+       depends on PV
+       default y
+       ---help---
+         The 32bit PV ABI uses Ring1, an area of the x86 architecture which
+         was deprecated and mostly removed in the AMD64 spec.  As a result,
+         it occasionally conflicts with newer x86 hardware features, causing
+         overheads for Xen to maintain backwards compatibility.
+
+         People may wish to disable 32bit PV guests for attack surface
+         reduction, or performance reasons.  Backwards compatibility can be
+         provided via the PV Shim mechanism.
+
+         If unsure, say Y.
+
 config PV_LINEAR_PT
        bool "Support for PV linear pagetables"
        depends on PV
diff --git a/xen/arch/x86/pv/domain.c b/xen/arch/x86/pv/domain.c
index 70fae43965..47a0db082f 100644
--- a/xen/arch/x86/pv/domain.c
+++ b/xen/arch/x86/pv/domain.c
@@ -16,6 +16,39 @@
 #include <asm/pv/domain.h>
 #include <asm/shadow.h>
 
+#ifdef CONFIG_PV32
+int8_t __read_mostly opt_pv32 = -1;
+#endif
+
+static int parse_pv(const char *s)
+{
+    const char *ss;
+    int val, rc = 0;
+
+    do {
+        ss = strchr(s, ',');
+        if ( !ss )
+            ss = strchr(s, '\0');
+
+        if ( (val = parse_boolean("32", s, ss)) >= 0 )
+        {
+#ifdef CONFIG_PV32
+            opt_pv32 = val;
+#else
+            printk(XENLOG_INFO
+                   "CONFIG_PV32 disabled - ignoring 'pv=32' setting\n");
+#endif
+        }
+        else
+            rc = -EINVAL;
+
+        s = ss + 1;
+    } while ( *ss );
+
+    return rc;
+}
+custom_param("pv", parse_pv);
+
 static __read_mostly enum {
     PCID_OFF,
     PCID_ALL,
@@ -174,6 +207,8 @@ int switch_compat(struct domain *d)
 
     BUILD_BUG_ON(offsetof(struct shared_info, vcpu_info) != 0);
 
+    if ( !opt_pv32 )
+        return -EOPNOTSUPP;
     if ( is_hvm_domain(d) || domain_tot_pages(d) != 0 )
         return -EACCES;
     if ( is_pv_32bit_domain(d) )
diff --git a/xen/arch/x86/setup.c b/xen/arch/x86/setup.c
index 885919d5c3..c50aefb2de 100644
--- a/xen/arch/x86/setup.c
+++ b/xen/arch/x86/setup.c
@@ -53,6 +53,7 @@
 #include <asm/spec_ctrl.h>
 #include <asm/guest.h>
 #include <asm/microcode.h>
+#include <asm/pv/domain.h>
 
 /* opt_nosmp: If true, secondary processors are ignored. */
 static bool __initdata opt_nosmp;
@@ -1875,8 +1876,12 @@ void arch_get_xen_caps(xen_capabilities_info_t *info)
     {
         snprintf(s, sizeof(s), "xen-%d.%d-x86_64 ", major, minor);
         safe_strcat(*info, s);
-        snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
-        safe_strcat(*info, s);
+
+        if ( opt_pv32 )
+        {
+            snprintf(s, sizeof(s), "xen-%d.%d-x86_32p ", major, minor);
+            safe_strcat(*info, s);
+        }
     }
     if ( hvm_enabled )
     {
diff --git a/xen/include/asm-x86/pv/domain.h b/xen/include/asm-x86/pv/domain.h
index 7a69bfb303..df9716ff26 100644
--- a/xen/include/asm-x86/pv/domain.h
+++ b/xen/include/asm-x86/pv/domain.h
@@ -23,6 +23,12 @@
 
 #include <xen/sched.h>
 
+#ifdef CONFIG_PV32
+extern int8_t opt_pv32;
+#else
+# define opt_pv32 false
+#endif
+
 /*
  * PCID values for the address spaces of 64-bit pv domains:
  *
-- 
2.11.0