x86: Introduce alternative indirect thunks

Depending on hardware and microcode availability, we will want to replace
IND_THUNK_REPOLINE with other implementations.

For AMD hardware, choose IND_THUNK_LFENCE in preference to retpoline if lfence
is known to be (or was successfully made) dispatch serialising.

This is part of XSA-254.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 858cba0d4c6b6b45180afcb41561fd6585ad51a3
master date: 2018-01-16 17:45:50 +0000

diff --git a/docs/misc/xen-command-line.markdown b/docs/misc/xen-command-line.markdown
index 768d4f5f5593a2ed7c36740a257b9d250014dc37..5ecb5eb1f8b1f2c92c4569bf6d5ff7570d9a83ab 100644 (file)
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -252,6 +252,22 @@ and not running softirqs. Reduce this if softirqs are not being run frequently
 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 ]`
+
+Branch Target Injection controls.  By default, Xen will pick the most
+appropriate BTI mitigations based on compiled in support, loaded microcode,
+and hardware details.
+
+**WARNING: Any use of this option may interfere with heuristics.  Use with
+extreme care.**
+
+If Xen was compiled with INDIRECT_THUNK support, `thunk=` can be used to
+select which of the thunks gets patched into the `__x86_indirect_thunk_%reg`
+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).
+
 ### xenheap\_megabytes (arm32)
 > `= <size>`
 

diff --git a/xen/arch/x86/indirect-thunk.S b/xen/arch/x86/indirect-thunk.S
index 3eaf505d0ee45aa1a2118c5704b511a9a117a393..7d34707218a8eeac81980e3d0d019c89cb048ca7 100644 (file)
--- a/xen/arch/x86/indirect-thunk.S
+++ b/xen/arch/x86/indirect-thunk.S
@@ -21,15 +21,26 @@
         ret
 .endm
 
+.macro IND_THUNK_LFENCE reg:req
+        lfence
+        jmp *%\reg
+.endm
+
+.macro IND_THUNK_JMP reg:req
+        jmp *%\reg
+.endm
+
 /*
- * Build the __x86_indirect_thunk_* symbols.  Currently implement the
- * retpoline thunk only.
+ * Build the __x86.indirect_thunk.* symbols.  Execution lands on an
+ * alternative patch point which implements one of the above THUNK_*'s
  */
 .macro GEN_INDIRECT_THUNK reg:req
         .section .text.__x86_indirect_thunk_\reg, "ax", @progbits
 
 ENTRY(__x86_indirect_thunk_\reg)
-        IND_THUNK_RETPOLINE \reg
+        ALTERNATIVE_2 __stringify(IND_THUNK_RETPOLINE \reg),              \
+        __stringify(IND_THUNK_LFENCE \reg), X86_FEATURE_IND_THUNK_LFENCE, \
+        __stringify(IND_THUNK_JMP \reg),    X86_FEATURE_IND_THUNK_JMP
 .endm
 
 /* Instantiate GEN_INDIRECT_THUNK for each register except %rsp. */

diff --git a/xen/arch/x86/spec_ctrl.c b/xen/arch/x86/spec_ctrl.c
index 256701a43c02f4483a24cf8b27e75537af0f20a9..d601c028d850352b7f3b4ec855ec03249c41a91a 100644 (file)
--- a/xen/arch/x86/spec_ctrl.c
+++ b/xen/arch/x86/spec_ctrl.c
@@ -16,18 +16,54 @@
  *
  * Copyright (c) 2017-2018 Citrix Systems Ltd.
  */
+#include <xen/errno.h>
 #include <xen/init.h>
 #include <xen/lib.h>
 
 #include <asm/processor.h>
 #include <asm/spec_ctrl.h>
 
-enum ind_thunk {
+static enum ind_thunk {
     THUNK_DEFAULT, /* Decide which thunk to use at boot time. */
     THUNK_NONE,    /* Missing compiler support for thunks. */
 
     THUNK_RETPOLINE,
-};
+    THUNK_LFENCE,
+    THUNK_JMP,
+} opt_thunk __initdata = THUNK_DEFAULT;
+
+static int __init parse_bti(const char *s)
+{
+    const char *ss;
+    int rc = 0;
+
+    do {
+        ss = strchr(s, ',');
+        if ( !ss )
+            ss = strchr(s, '\0');
+
+        if ( !strncmp(s, "thunk=", 6) )
+        {
+            s += 6;
+
+            if ( !strncmp(s, "retpoline", ss - s) )
+                opt_thunk = THUNK_RETPOLINE;
+            else if ( !strncmp(s, "lfence", ss - s) )
+                opt_thunk = THUNK_LFENCE;
+            else if ( !strncmp(s, "jmp", ss - s) )
+                opt_thunk = THUNK_JMP;
+            else
+                rc = -EINVAL;
+        }
+        else
+            rc = -EINVAL;
+
+        s = ss + 1;
+    } while ( *ss );
+
+    return rc;
+}
+custom_param("bti", parse_bti);
 
 static void __init print_details(enum ind_thunk thunk)
 {
@@ -40,13 +76,40 @@ static void __init print_details(enum ind_thunk thunk)
     printk(XENLOG_INFO
            "BTI mitigations: Thunk %s\n",
            thunk == THUNK_NONE      ? "N/A" :
-           thunk == THUNK_RETPOLINE ? "RETPOLINE" : "?");
+           thunk == THUNK_RETPOLINE ? "RETPOLINE" :
+           thunk == THUNK_LFENCE    ? "LFENCE" :
+           thunk == THUNK_JMP       ? "JMP" : "?");
 }
 
 void __init init_speculation_mitigations(void)
 {
     enum ind_thunk thunk = THUNK_DEFAULT;
 
+    /*
+     * Has the user specified any custom BTI mitigations?  If so, follow their
+     * instructions exactly and disable all heuristics.
+     */
+    if ( opt_thunk != THUNK_DEFAULT )
+    {
+        thunk = opt_thunk;
+    }
+    else
+    {
+        /*
+         * Evaluate the safest Branch Target Injection mitigations to use.
+         * First, begin with compiler-aided mitigations.
+         */
+        if ( IS_ENABLED(CONFIG_INDIRECT_THUNK) )
+        {
+            /*
+             * AMD's recommended mitigation is to set lfence as being dispatch
+             * serialising, and to use IND_THUNK_LFENCE.
+             */
+            if ( cpu_has_lfence_dispatch )
+                thunk = THUNK_LFENCE;
+        }
+    }
+
     /*
      * Supplimentary minor adjustments.  Without compiler support, there are
      * no thunks.
@@ -61,6 +124,12 @@ void __init init_speculation_mitigations(void)
     if ( thunk == THUNK_DEFAULT )
         thunk = THUNK_RETPOLINE;
 
+    /* Apply the chosen settings. */
+    if ( thunk == THUNK_LFENCE )
+        setup_force_cpu_cap(X86_FEATURE_IND_THUNK_LFENCE);
+    else if ( thunk == THUNK_JMP )
+        setup_force_cpu_cap(X86_FEATURE_IND_THUNK_JMP);
+
     print_details(thunk);
 }
 

diff --git a/xen/include/asm-x86/cpufeature.h b/xen/include/asm-x86/cpufeature.h
index 5288bbfdd2f628e03372629ffbb0012b8599b07f..4dbcf863e0d13e7f86561cd4396b5dd1a4697be4 100644 (file)
--- a/xen/include/asm-x86/cpufeature.h
+++ b/xen/include/asm-x86/cpufeature.h
@@ -22,6 +22,8 @@ XEN_CPUFEATURE(XEN_SMAP,        (FSCAPINTS+0)*32+ 11) /* SMAP gets used by Xen i
 XEN_CPUFEATURE(MSR_PLATFORM_INFO, (FSCAPINTS+0)*32+12) /* PLATFORM_INFO MSR present */
 XEN_CPUFEATURE(MSR_MISC_FEATURES, (FSCAPINTS+0)*32+13) /* MISC_FEATURES_ENABLES MSR present */
 XEN_CPUFEATURE(LFENCE_DISPATCH, (FSCAPINTS+0)*32+14) /* lfence set as Dispatch Serialising */
+XEN_CPUFEATURE(IND_THUNK_LFENCE,(FSCAPINTS+0)*32+15) /* Use IND_THUNK_LFENCE */
+XEN_CPUFEATURE(IND_THUNK_JMP,   (FSCAPINTS+0)*32+16) /* Use IND_THUNK_JMP */
 
 #define NCAPINTS (FSCAPINTS + 1) /* N 32-bit words worth of info */