[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH v10 08/11] x86/boot: Calculate the most appropriate BTI mitigation to use



See the logic and comments in init_speculation_mitigations() for further
details.

There are two controls for RSB overwriting, because in principle there are
cases where it might be safe to forego rsb_native (Off the top of my head,
SMEP active, no 32bit PV guests at all, no use of vmevent/paging subsystems
for HVM guests, but I make no guarantees that this list of restrictions is
exhaustive).

Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
---
CC: Jan Beulich <JBeulich@xxxxxxxx>
CC: Asit K Mallick <asit.k.mallick@xxxxxxxxx>
CC: Jun Nakajima <jun.nakajima@xxxxxxxxx>

TODO: Confirm Broadwell microcode details, and Atom safety WRT retpoline.
These details should not block this series.

v7:
 * static, and tweak comment
v10:
 * Simplify init_speculation_mitigations()
 * Merge half of the old "x86/entry: Clobber the Return Stack Buffer/Return
   Address Stack on entry to Xen" due to the IST safety work
---
 docs/misc/xen-command-line.markdown |  10 ++-
 xen/arch/x86/spec_ctrl.c            | 142 +++++++++++++++++++++++++++++++++++-
 xen/include/asm-x86/spec_ctrl.h     |   3 +
 3 files changed, 150 insertions(+), 5 deletions(-)

diff --git a/docs/misc/xen-command-line.markdown 
b/docs/misc/xen-command-line.markdown
index f73990f..11399ce 100644
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -246,7 +246,7 @@ enough. Setting this to a high value may cause boot 
failure, particularly if
 the NMI watchdog is also enabled.
 
 ### bti (x86)
-> `= List of [ thunk=retpoline|lfence|jmp ]`
+> `= List of [ thunk=retpoline|lfence|jmp, ibrs=<bool>, 
rsb_{vmexit,native}=<bool> ]`
 
 Branch Target Injection controls.  By default, Xen will pick the most
 appropriate BTI mitigations based on compiled in support, loaded microcode,
@@ -261,6 +261,14 @@ locations.  The default thunk is `retpoline` (generally 
preferred for Intel
 hardware), with the alternatives being `jmp` (a `jmp *%reg` gadget, minimal
 overhead), and `lfence` (an `lfence; jmp *%reg` gadget, preferred for AMD).
 
+On hardware supporting IBRS, the `ibrs=` option can be used to force or
+prevent Xen using the feature itself.  If Xen is not using IBRS itself,
+functionality is still set up so IBRS can be virtualised for guests.
+
+The `rsb_vmexit=` and `rsb_native=` options can be used to fine tune when the
+RSB gets overwritten.  There are individual controls for an entry from HVM
+context, and an entry from a native (PV or Xen) context.
+
 ### xenheap\_megabytes (arm32)
 > `= <size>`
 
diff --git a/xen/arch/x86/spec_ctrl.c b/xen/arch/x86/spec_ctrl.c
index 89e7287..de5ba1a 100644
--- a/xen/arch/x86/spec_ctrl.c
+++ b/xen/arch/x86/spec_ctrl.c
@@ -20,8 +20,10 @@
 #include <xen/init.h>
 #include <xen/lib.h>
 
+#include <asm/microcode.h>
 #include <asm/processor.h>
 #include <asm/spec_ctrl.h>
+#include <asm/spec_ctrl_asm.h>
 
 static enum ind_thunk {
     THUNK_DEFAULT, /* Decide which thunk to use at boot time. */
@@ -31,11 +33,15 @@ static enum ind_thunk {
     THUNK_LFENCE,
     THUNK_JMP,
 } opt_thunk __initdata = THUNK_DEFAULT;
+static int8_t __initdata opt_ibrs = -1;
+static bool __initdata opt_rsb_native = true;
+static bool __initdata opt_rsb_vmexit = true;
+uint8_t __read_mostly default_bti_ist_info;
 
 static int __init parse_bti(const char *s)
 {
     const char *ss;
-    int rc = 0;
+    int val, rc = 0;
 
     do {
         ss = strchr(s, ',');
@@ -55,6 +61,12 @@ static int __init parse_bti(const char *s)
             else
                 rc = -EINVAL;
         }
+        else if ( (val = parse_boolean("ibrs", s, ss)) >= 0 )
+            opt_ibrs = val;
+        else if ( (val = parse_boolean("rsb_native", s, ss)) >= 0 )
+            opt_rsb_native = val;
+        else if ( (val = parse_boolean("rsb_vmexit", s, ss)) >= 0 )
+            opt_rsb_vmexit = val;
         else
             rc = -EINVAL;
 
@@ -91,24 +103,84 @@ static void __init print_details(enum ind_thunk thunk)
         printk(XENLOG_DEBUG "  Compiled-in support: INDIRECT_THUNK\n");
 
     printk(XENLOG_INFO
-           "BTI mitigations: Thunk %s\n",
+           "BTI mitigations: Thunk %s, Others:%s%s%s\n",
            thunk == THUNK_NONE      ? "N/A" :
            thunk == THUNK_RETPOLINE ? "RETPOLINE" :
            thunk == THUNK_LFENCE    ? "LFENCE" :
-           thunk == THUNK_JMP       ? "JMP" : "?");
+           thunk == THUNK_JMP       ? "JMP" : "?",
+           boot_cpu_has(X86_FEATURE_XEN_IBRS_SET)    ? " IBRS+" :
+           boot_cpu_has(X86_FEATURE_XEN_IBRS_CLEAR)  ? " IBRS-"      : "",
+           boot_cpu_has(X86_FEATURE_RSB_NATIVE)      ? " RSB_NATIVE" : "",
+           boot_cpu_has(X86_FEATURE_RSB_VMEXIT)      ? " RSB_VMEXIT" : "");
+}
+
+/* Calculate whether Retpoline is known-safe on this CPU. */
+static bool __init retpoline_safe(void)
+{
+    unsigned int ucode_rev = this_cpu(ucode_cpu_info).cpu_sig.rev;
+
+    if ( boot_cpu_data.x86_vendor == X86_VENDOR_AMD )
+        return true;
+
+    if ( boot_cpu_data.x86_vendor != X86_VENDOR_INTEL ||
+         boot_cpu_data.x86 != 6 )
+        return false;
+
+    switch ( boot_cpu_data.x86_model )
+    {
+    case 0x17: /* Penryn */
+    case 0x1d: /* Dunnington */
+    case 0x1e: /* Nehalem */
+    case 0x1f: /* Auburndale / Havendale */
+    case 0x1a: /* Nehalem EP */
+    case 0x2e: /* Nehalem EX */
+    case 0x25: /* Westmere */
+    case 0x2c: /* Westmere EP */
+    case 0x2f: /* Westmere EX */
+    case 0x2a: /* SandyBridge */
+    case 0x2d: /* SandyBridge EP/EX */
+    case 0x3a: /* IvyBridge */
+    case 0x3e: /* IvyBridge EP/EX */
+    case 0x3c: /* Haswell */
+    case 0x3f: /* Haswell EX/EP */
+    case 0x45: /* Haswell D */
+    case 0x46: /* Haswell H */
+        return true;
+
+        /*
+         * Broadwell processors are retpoline-safe after specific microcode
+         * versions.
+         */
+    case 0x3d: /* Broadwell */
+        return ucode_rev >= 0x28;
+    case 0x47: /* Broadwell H */
+        return ucode_rev >= 0x1b;
+    case 0x4f: /* Broadwell EP/EX */
+        return ucode_rev >= 0xb000025;
+    case 0x56: /* Broadwell D */
+        return false; /* TBD. */
+
+        /*
+         * Skylake and later processors are not retpoline-safe.
+         */
+    default:
+        return false;
+    }
 }
 
 void __init init_speculation_mitigations(void)
 {
     enum ind_thunk thunk = THUNK_DEFAULT;
+    bool ibrs = false;
 
     /*
      * Has the user specified any custom BTI mitigations?  If so, follow their
      * instructions exactly and disable all heuristics.
      */
-    if ( opt_thunk != THUNK_DEFAULT )
+    if ( opt_thunk != THUNK_DEFAULT || opt_ibrs != -1 )
     {
         thunk = opt_thunk;
+        ibrs  = !!opt_ibrs;
     }
     else
     {
@@ -124,7 +196,18 @@ void __init init_speculation_mitigations(void)
              */
             if ( cpu_has_lfence_dispatch )
                 thunk = THUNK_LFENCE;
+            /*
+             * On Intel hardware, we'd like to use retpoline in preference to
+             * IBRS, but only if it is safe on this hardware.
+             */
+            else if ( retpoline_safe() )
+                thunk = THUNK_RETPOLINE;
+            else
+                ibrs = true;
         }
+        /* Without compiler thunk support, use IBRS if available. */
+        else if ( boot_cpu_has(X86_FEATURE_IBRSB) )
+            ibrs = true;
     }
 
     /*
@@ -135,6 +218,13 @@ void __init init_speculation_mitigations(void)
         thunk = THUNK_NONE;
 
     /*
+     * If IBRS is in use and thunks are compiled in, there is no point
+     * suffering extra overhead.  Switch to the least-overhead thunk.
+     */
+    if ( ibrs && thunk == THUNK_DEFAULT )
+        thunk = THUNK_JMP;
+
+    /*
      * If there are still no thunk preferences, the compiled default is
      * actually retpoline, and it is better than nothing.
      */
@@ -147,6 +237,50 @@ void __init init_speculation_mitigations(void)
     else if ( thunk == THUNK_JMP )
         setup_force_cpu_cap(X86_FEATURE_IND_THUNK_JMP);
 
+    if ( boot_cpu_has(X86_FEATURE_IBRSB) )
+    {
+        /*
+         * Even if we've chosen to not have IBRS set in Xen context, we still
+         * need the IBRS entry/exit logic to virtualise IBRS support for
+         * guests.
+         */
+        if ( ibrs )
+            setup_force_cpu_cap(X86_FEATURE_XEN_IBRS_SET);
+        else
+            setup_force_cpu_cap(X86_FEATURE_XEN_IBRS_CLEAR);
+
+        default_bti_ist_info |= BTI_IST_WRMSR | ibrs;
+    }
+
+    /*
+     * PV guests can poison the RSB to any virtual address from which
+     * they can execute a call instruction.  This is necessarily outside
+     * of the Xen supervisor mappings.
+     *
+     * With SMEP enabled, the processor won't speculate into user mappings.
+     * Therefore, in this case, we don't need to worry about poisoned entries
+     * from 64bit PV guests.
+     *
+     * 32bit PV guest kernels run in ring 1, so use supervisor mappings.
+     * If a processors speculates to 32bit PV guest kernel mappings, it is
+     * speculating in 64bit supervisor mode, and can leak data.
+     */
+    if ( opt_rsb_native )
+    {
+        setup_force_cpu_cap(X86_FEATURE_RSB_NATIVE);
+        default_bti_ist_info |= BTI_IST_RSB;
+    }
+
+    /*
+     * HVM guests can always poison the RSB to point at Xen supervisor
+     * mappings.
+     */
+    if ( opt_rsb_vmexit )
+        setup_force_cpu_cap(X86_FEATURE_RSB_VMEXIT);
+
+    /* (Re)init BSP state how that default_bti_ist_info has been calculated. */
+    init_shadow_spec_ctrl_state();
+
     print_details(thunk);
 }
 
diff --git a/xen/include/asm-x86/spec_ctrl.h b/xen/include/asm-x86/spec_ctrl.h
index b451250..6120e4f 100644
--- a/xen/include/asm-x86/spec_ctrl.h
+++ b/xen/include/asm-x86/spec_ctrl.h
@@ -24,11 +24,14 @@
 
 void init_speculation_mitigations(void);
 
+extern uint8_t default_bti_ist_info;
+
 static inline void init_shadow_spec_ctrl_state(void)
 {
     struct cpu_info *info = get_cpu_info();
 
     info->shadow_spec_ctrl = info->use_shadow_spec_ctrl = 0;
+    info->bti_ist_info = default_bti_ist_info;
 }
 
 #endif /* !__X86_SPEC_CTRL_H__ */
-- 
2.1.4


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.