arm64: alternative: Allow immediate branch as alternative instruction

Since all branches are PC-relative on AArch64, these instructions
cannot be used as an alternative with the simplistic approach
we currently have (the immediate has been computed from
the .altinstr_replacement section, and end-up being completely off
if the target is outside of the replacement sequence).

This patch handles the branch instructions in a different way,
using the insn framework to recompute the immediate, and generate
the right displacement in the above case.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

arch/arm64/kernel/alternative.c

@@ -24,8 +24,13 @@
 #include <asm/cacheflush.h>
 #include <asm/alternative.h>
 #include <asm/cpufeature.h>
+#include <asm/insn.h>
 #include <linux/stop_machine.h>
 
+#define __ALT_PTR(a,f)		(u32 *)((void *)&(a)->f + (a)->f)
+#define ALT_ORIG_PTR(a)		__ALT_PTR(a, orig_offset)
+#define ALT_REPL_PTR(a)		__ALT_PTR(a, alt_offset)
+
 extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
 
 struct alt_region {
@@ -33,13 +38,63 @@ struct alt_region {
 	struct alt_instr *end;
 };
 
+/*
+ * Check if the target PC is within an alternative block.
+ */
+static bool branch_insn_requires_update(struct alt_instr *alt, unsigned long pc)
+{
+	unsigned long replptr;
+
+	if (kernel_text_address(pc))
+		return 1;
+
+	replptr = (unsigned long)ALT_REPL_PTR(alt);
+	if (pc >= replptr && pc <= (replptr + alt->alt_len))
+		return 0;
+
+	/*
+	 * Branching into *another* alternate sequence is doomed, and
+	 * we're not even trying to fix it up.
+	 */
+	BUG();
+}
+
+static u32 get_alt_insn(struct alt_instr *alt, u32 *insnptr, u32 *altinsnptr)
+{
+	u32 insn;
+
+	insn = le32_to_cpu(*altinsnptr);
+
+	if (aarch64_insn_is_branch_imm(insn)) {
+		s32 offset = aarch64_get_branch_offset(insn);
+		unsigned long target;
+
+		target = (unsigned long)altinsnptr + offset;
+
+		/*
+		 * If we're branching inside the alternate sequence,
+		 * do not rewrite the instruction, as it is already
+		 * correct. Otherwise, generate the new instruction.
+		 */
+		if (branch_insn_requires_update(alt, target)) {
+			offset = target - (unsigned long)insnptr;
+			insn = aarch64_set_branch_offset(insn, offset);
+		}
+	}
+
+	return insn;
+}
+
 static int __apply_alternatives(void *alt_region)
 {
 	struct alt_instr *alt;
 	struct alt_region *region = alt_region;
-	u8 *origptr, *replptr;
+	u32 *origptr, *replptr;
 
 	for (alt = region->begin; alt < region->end; alt++) {
+		u32 insn;
+		int i, nr_inst;
+
 		if (!cpus_have_cap(alt->cpufeature))
 			continue;
 
@@ -47,11 +102,17 @@ static int __apply_alternatives(void *alt_region)
 
 		pr_info_once("patching kernel code\n");
 
-		origptr = (u8 *)&alt->orig_offset + alt->orig_offset;
-		replptr = (u8 *)&alt->alt_offset + alt->alt_offset;
-		memcpy(origptr, replptr, alt->alt_len);
+		origptr = ALT_ORIG_PTR(alt);
+		replptr = ALT_REPL_PTR(alt);
+		nr_inst = alt->alt_len / sizeof(insn);
+
+		for (i = 0; i < nr_inst; i++) {
+			insn = get_alt_insn(alt, origptr + i, replptr + i);
+			*(origptr + i) = cpu_to_le32(insn);
+		}
+
 		flush_icache_range((uintptr_t)origptr,
-				   (uintptr_t)(origptr + alt->alt_len));
+				   (uintptr_t)(origptr + nr_inst));
 	}
 
 	return 0;