x86,static_call: Use alternative RET encoding

In addition to teaching static_call about the new way to spell 'RET',
there is an added complication in that static_call() is allowed to
rewrite text before it is known which particular spelling is required.

In order to deal with this; have a static_call specific fixup in the
apply_return() 'alternative' patching routine that will rewrite the
static_call trampoline to match the definite sequence.

This in turn creates the problem of uniquely identifying static call
trampolines. Currently trampolines are 8 bytes, the first 5 being the
jmp.d32/ret sequence and the final 3 a byte sequence that spells out
'SCT'.

This sequence is used in __static_call_validate() to ensure it is
patching a trampoline and not a random other jmp.d32. That is,
false-positives shouldn't be plenty, but aren't a big concern.

OTOH the new __static_call_fixup() must not have false-positives, and
'SCT' decodes to the somewhat weird but semi plausible sequence:

  push %rbx
  rex.XB push %r12

Additionally, there are SLS concerns with immediate jumps. Combined it
seems like a good moment to change the signature to a single 3 byte
trap instruction that is unique to this usage and will not ever get
generated by accident.

As such, change the signature to: '0x0f, 0xb9, 0xcc', which decodes
to:

  ud1 %esp, %ecx

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>

diff --git a/arch/x86/include/asm/static_call.h b/arch/x86/include/asm/static_call.h
index 2d8dacd..70cc9cc 100644 (file)
--- a/arch/x86/include/asm/static_call.h
+++ b/arch/x86/include/asm/static_call.h
@@ -21,6 +21,16 @@
  * relative displacement across sections.
  */
 
+/*
+ * The trampoline is 8 bytes and of the general form:
+ *
+ *   jmp.d32 \func
+ *   ud1 %esp, %ecx
+ *
+ * That trailing #UD provides both a speculation stop and serves as a unique
+ * 3 byte signature identifying static call trampolines. Also see tramp_ud[]
+ * and __static_call_fixup().
+ */
 #define __ARCH_DEFINE_STATIC_CALL_TRAMP(name, insns)                   \
        asm(".pushsection .static_call.text, \"ax\"             \n"     \
            ".align 4                                           \n"     \
@@ -28,7 +38,7 @@
            STATIC_CALL_TRAMP_STR(name) ":                      \n"     \
            ANNOTATE_NOENDBR                                            \
            insns "                                             \n"     \
-           ".byte 0x53, 0x43, 0x54                             \n"     \
+           ".byte 0x0f, 0xb9, 0xcc                             \n"     \
            ".type " STATIC_CALL_TRAMP_STR(name) ", @function   \n"     \
            ".size " STATIC_CALL_TRAMP_STR(name) ", . - " STATIC_CALL_TRAMP_STR(name) " \n" \
            ".popsection                                        \n")
@@ -36,8 +46,13 @@
 #define ARCH_DEFINE_STATIC_CALL_TRAMP(name, func)                      \
        __ARCH_DEFINE_STATIC_CALL_TRAMP(name, ".byte 0xe9; .long " #func " - (. + 4)")
 
+#ifdef CONFIG_RETPOLINE
+#define ARCH_DEFINE_STATIC_CALL_NULL_TRAMP(name)                       \
+       __ARCH_DEFINE_STATIC_CALL_TRAMP(name, "jmp __x86_return_thunk")
+#else
 #define ARCH_DEFINE_STATIC_CALL_NULL_TRAMP(name)                       \
        __ARCH_DEFINE_STATIC_CALL_TRAMP(name, "ret; int3; nop; nop; nop")
+#endif
 
 #define ARCH_DEFINE_STATIC_CALL_RET0_TRAMP(name)                       \
        ARCH_DEFINE_STATIC_CALL_TRAMP(name, __static_call_return0)
@@ -48,4 +63,6 @@
            ".long " STATIC_CALL_KEY_STR(name) " - .            \n"     \
            ".popsection                                        \n")
 
+extern bool __static_call_fixup(void *tramp, u8 op, void *dest);
+
 #endif /* _ASM_STATIC_CALL_H */

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 76b7459..cf447ee 100644 (file)
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -539,18 +539,22 @@ void __init_or_module noinline apply_returns(s32 *start, s32 *end)
        s32 *s;
 
        for (s = start; s < end; s++) {
-               void *addr = (void *)s + *s;
+               void *dest = NULL, *addr = (void *)s + *s;
                struct insn insn;
                int len, ret;
                u8 bytes[16];
-               u8 op1;
+               u8 op;
 
                ret = insn_decode_kernel(&insn, addr);
                if (WARN_ON_ONCE(ret < 0))
                        continue;
 
-               op1 = insn.opcode.bytes[0];
-               if (WARN_ON_ONCE(op1 != JMP32_INSN_OPCODE))
+               op = insn.opcode.bytes[0];
+               if (op == JMP32_INSN_OPCODE)
+                       dest = addr + insn.length + insn.immediate.value;
+
+               if (__static_call_fixup(addr, op, dest) ||
+                   WARN_ON_ONCE(dest != &__x86_return_thunk))
                        continue;
 
                DPRINTK("return thunk at: %pS (%px) len: %d to: %pS",

diff --git a/arch/x86/kernel/static_call.c b/arch/x86/kernel/static_call.c
index aa72cef..fe21fe7 100644 (file)
--- a/arch/x86/kernel/static_call.c
+++ b/arch/x86/kernel/static_call.c
@@ -12,6 +12,13 @@ enum insn_type {
 };
 
 /*
+ * ud1 %esp, %ecx - a 3 byte #UD that is unique to trampolines, chosen such
+ * that there is no false-positive trampoline identification while also being a
+ * speculation stop.
+ */
+static const u8 tramp_ud[] = { 0x0f, 0xb9, 0xcc };
+
+/*
  * cs cs cs xorl %eax, %eax - a single 5 byte instruction that clears %[er]ax
  */
 static const u8 xor5rax[] = { 0x2e, 0x2e, 0x2e, 0x31, 0xc0 };
@@ -43,7 +50,10 @@ static void __ref __static_call_transform(void *insn, enum insn_type type, void
                break;
 
        case RET:
-               code = &retinsn;
+               if (cpu_feature_enabled(X86_FEATURE_RETHUNK))
+                       code = text_gen_insn(JMP32_INSN_OPCODE, insn, &__x86_return_thunk);
+               else
+                       code = &retinsn;
                break;
        }
 
@@ -60,7 +70,7 @@ static void __static_call_validate(void *insn, bool tail, bool tramp)
 {
        u8 opcode = *(u8 *)insn;
 
-       if (tramp && memcmp(insn+5, "SCT", 3)) {
+       if (tramp && memcmp(insn+5, tramp_ud, 3)) {
                pr_err("trampoline signature fail");
                BUG();
        }
@@ -115,3 +125,29 @@ void arch_static_call_transform(void *site, void *tramp, void *func, bool tail)
        mutex_unlock(&text_mutex);
 }
 EXPORT_SYMBOL_GPL(arch_static_call_transform);
+
+#ifdef CONFIG_RETPOLINE
+/*
+ * This is called by apply_returns() to fix up static call trampolines,
+ * specifically ARCH_DEFINE_STATIC_CALL_NULL_TRAMP which is recorded as
+ * having a return trampoline.
+ *
+ * The problem is that static_call() is available before determining
+ * X86_FEATURE_RETHUNK and, by implication, running alternatives.
+ *
+ * This means that __static_call_transform() above can have overwritten the
+ * return trampoline and we now need to fix things up to be consistent.
+ */
+bool __static_call_fixup(void *tramp, u8 op, void *dest)
+{
+       if (memcmp(tramp+5, tramp_ud, 3)) {
+               /* Not a trampoline site, not our problem. */
+               return false;
+       }
+
+       if (op == RET_INSN_OPCODE || dest == &__x86_return_thunk)
+               __static_call_transform(tramp, RET, NULL);
+
+       return true;
+}
+#endif