x86/alternative: Patch a single alternative location only once

Instead of patching a single location potentially multiple times in case of
nested ALTERNATIVE()s, do the patching only after having evaluated all
alt_instr instances for that location.

This has multiple advantages:

- In case of replacing an indirect with a direct call using the
  ALT_FLAG_DIRECT_CALL flag, there is no longer the need to have that
  instance before any other instances at the same location (the
  original instruction is needed for finding the target of the direct
  call).
  This issue has been hit when trying to do paravirt patching similar
  to the following:
    ALTERNATIVE_2(PARAVIRT_CALL,    // indirect call
                  instr, feature,   // native instruction
                  ALT_CALL_INSTR, X86_FEATURE_XENPV)  // Xen function
  In case "feature" was true, "instr" replaced the indirect call. Under
  Xen PV the patching to have a direct call failed, as the original
  indirect call was no longer there to find the call target.

- In case of nested ALTERNATIVE()s there is no intermediate replacement
  visible. This avoids any problems in case e.g. an interrupt is
  happening between the single instances and the patched location is
  used during handling the interrupt.

Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://patch.msgid.link/20260105080452.5064-3-jgross@suse.com

Author: jgross1

arch/x86/kernel/alternative.c

@@ -593,39 +593,38 @@ struct patch_site {
 	u8 len;
 };
 
-static void __init_or_module analyze_patch_site(struct patch_site *ps,
-						struct alt_instr *start,
-						struct alt_instr *end)
+static struct alt_instr * __init_or_module analyze_patch_site(struct patch_site *ps,
+							     struct alt_instr *start,
+							     struct alt_instr *end)
 {
-	struct alt_instr *r;
+	struct alt_instr *alt = start;
 
 	ps->instr = instr_va(start);
-	ps->len = start->instrlen;
 
 	/*
 	 * In case of nested ALTERNATIVE()s the outer alternative might add
 	 * more padding. To ensure consistent patching find the max padding for
 	 * all alt_instr entries for this site (nested alternatives result in
 	 * consecutive entries).
+	 * Find the last alt_instr eligible for patching at the site.
 	 */
-	for (r = start+1; r < end && instr_va(r) == ps->instr; r++) {
-		ps->len = max(ps->len, r->instrlen);
-		start->instrlen = r->instrlen = ps->len;
+	for (; alt < end && instr_va(alt) == ps->instr; alt++) {
+		ps->len = max(ps->len, alt->instrlen);
+
+		BUG_ON(alt->cpuid >= (NCAPINTS + NBUGINTS) * 32);
+		/*
+		 * Patch if either:
+		 * - feature is present
+		 * - feature not present but ALT_FLAG_NOT is set to mean,
+		 *   patch if feature is *NOT* present.
+		 */
+		if (!boot_cpu_has(alt->cpuid) != !(alt->flags & ALT_FLAG_NOT))
+			ps->alt = alt;
 	}
 
 	BUG_ON(ps->len > sizeof(ps->buff));
-	BUG_ON(start->cpuid >= (NCAPINTS + NBUGINTS) * 32);
 
-	/*
-	 * Patch if either:
-	 * - feature is present
-	 * - feature not present but ALT_FLAG_NOT is set to mean,
-	 *   patch if feature is *NOT* present.
-	 */
-	if (!boot_cpu_has(start->cpuid) == !(start->flags & ALT_FLAG_NOT))
-		ps->alt = NULL;
-	else
-		ps->alt = start;
+	return alt;
 }
 
 static void __init_or_module prep_patch_site(struct patch_site *ps)
@@ -704,10 +703,14 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start,
 	 * So be careful if you want to change the scan order to any other
 	 * order.
 	 */
-	for (a = start; a < end; a++) {
-		struct patch_site ps;
-
-		analyze_patch_site(&ps, a, end);
+	a = start;
+	while (a < end) {
+		struct patch_site ps = {
+			.alt = NULL,
+			.len = 0
+		};
+
+		a = analyze_patch_site(&ps, a, end);
 		prep_patch_site(&ps);
 		patch_site(&ps);
 	}