x86/spec-ctrl: Support the "long" BHB loop sequence

Out of an abudnance of caution, implement the long loop too, and allowing for
it to be opted in to.

This is part of XSA-456 / CVE-2024-2201.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>

docs/misc/xen-command-line.pandoc

@@ -2375,7 +2375,7 @@ By default SSBD will be mitigated at runtime (i.e `ssbd=runtime`).
 ### spec-ctrl (x86)
 > `= List of [ <bool>, xen=<bool>, {pv,hvm}=<bool>,
 >              {msr-sc,rsb,verw,{ibpb,bhb}-entry}=<bool>|{pv,hvm}=<bool>,
->              bti-thunk=retpoline|lfence|jmp,bhb-seq=short|tsx,
+>              bti-thunk=retpoline|lfence|jmp,bhb-seq=short|tsx|long,
 >              {ibrs,ibpb,ssbd,psfd,
 >              eager-fpu,l1d-flush,branch-harden,srb-lock,
 >              unpriv-mmio,gds-mit,div-scrub,lock-harden,
@@ -2447,7 +2447,7 @@ On all hardware, `bhb-seq=` can be used to select which of the BHB-clearing
 sequences gets used.  This interacts with the `bhb-entry=` and `bhi-dis-s=`
 options in order to mitigate Branch History Injection on affected hardware.
 The default sequence is `short`, with `tsx` as an alternative available
-capable hardware that can be opted in to.
+capable hardware, and `long` that can be opted in to.
 
 On hardware supporting IBRS (Indirect Branch Restricted Speculation), the
 `ibrs=` option can be used to force or prevent Xen using the feature itself.

xen/arch/x86/bhb-thunk.S

@@ -54,9 +54,13 @@ END(clear_bhb_tsx)
  *
  * The "short" sequence (5 and 5) is for CPUs prior to Alder Lake / Sapphire
  * Rapids (i.e. Cores prior to Golden Cove and/or Gracemont).
+ *
+ * The "long" sequence (12 and 7) is for Alder Lake / Sapphire Rapids
+ * (i.e. Golden Cove and/or Gracemont cores).  However, such CPUs are expected
+ * to use BHI_DIS_S in preference.
  */
 FUNC(clear_bhb_loops)
-        mov     $5, %ecx
+        ALTERNATIVE "mov $5, %ecx", "mov $12, %ecx", X86_SPEC_BHB_LOOPS_LONG
 
         call    1f
         jmp     5f
@@ -68,7 +72,7 @@ FUNC(clear_bhb_loops)
         int3
 
         .align 64
-2:      mov     $5, %eax
+2:      ALTERNATIVE "mov $5, %eax", "mov $7, %eax", X86_SPEC_BHB_LOOPS_LONG
 
 3:      jmp     4f
         int3

xen/arch/x86/include/asm/cpufeatures.h

@@ -58,6 +58,7 @@ XEN_CPUFEATURE(IBPB_ENTRY_HVM,    X86_SYNTH(29)) /* MSR_PRED_CMD used by Xen for
 
 #define X86_SPEC_BHB_TSX          X86_BUG(19) /* Use clear_bhb_tsx for BHI mitigation. */
 #define X86_SPEC_BHB_LOOPS        X86_BUG(20) /* Use clear_bhb_loops for BHI mitigation.*/
+#define X86_SPEC_BHB_LOOPS_LONG   X86_BUG(21) /* Upgrade clear_bhb_loops to the "long" sequence. */
 
 /* Total number of capability words, inc synth and bug words. */
 #define NCAPINTS (FSCAPINTS + X86_NR_SYNTH + X86_NR_BUG) /* N 32-bit words worth of info */

xen/arch/x86/spec_ctrl.c

@@ -41,6 +41,7 @@ static enum bhb_thunk {
     BHB_NONE,
     BHB_TSX,
     BHB_SHORT,
+    BHB_LONG,
 } opt_bhb_seq __initdata;
 
 /* Cmdline controls for Xen's speculative settings. */
@@ -308,6 +309,8 @@ static int __init cf_check parse_spec_ctrl(const char *s)
                 opt_bhb_seq = BHB_TSX;
             else if ( !cmdline_strcmp(s, "short") )
                 opt_bhb_seq = BHB_SHORT;
+            else if ( !cmdline_strcmp(s, "long") )
+                opt_bhb_seq = BHB_LONG;
             else
                 rc = -EINVAL;
         }
@@ -587,7 +590,8 @@ static void __init print_details(enum ind_thunk thunk)
            opt_bhb_seq != BHB_NONE    ? "BHB-Seq: " : "",
            opt_bhb_seq == BHB_NONE    ? "" :
            opt_bhb_seq == BHB_TSX     ? "TSX, " :
-           opt_bhb_seq == BHB_SHORT   ? "SHORT, " : "?, ",
+           opt_bhb_seq == BHB_SHORT   ? "SHORT, " :
+           opt_bhb_seq == BHB_LONG    ? "LONG, " : "?, ",
            (!boot_cpu_has(X86_FEATURE_IBRSB) &&
             !boot_cpu_has(X86_FEATURE_IBRS))         ? "No" :
            (default_xen_spec_ctrl & SPEC_CTRL_IBRS)  ? "IBRS+" :  "IBRS-",
@@ -1730,6 +1734,10 @@ static void __init bhi_calculations(void)
 
     switch ( opt_bhb_seq )
     {
+    case BHB_LONG:
+        setup_force_cpu_cap(X86_SPEC_BHB_LOOPS_LONG);
+        fallthrough;
+
     case BHB_SHORT:
         setup_force_cpu_cap(X86_SPEC_BHB_LOOPS);
         break;