[PATCH 2/3] objtool: Ignore retpoline alternatives
From: Josh Poimboeuf
Date: Wed Jan 10 2018 - 20:49:30 EST
Getting objtool to understand retpolines is going to be a bit of a
challenge. For now, take advantage of the fact that retpolines are
patched in with alternatives. Just read the original (sane)
non-alternative instruction, and ignore the patched-in retpoline.
This allows objtool to understand the control flow *around* the
retpoline, even if it can't yet follow what's inside. This means the
ORC unwinder will fail to unwind from inside a retpoline, but will work
fine otherwise.
Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
---
arch/x86/include/asm/nospec-branch.h | 27 +++++++++++++++-
tools/objtool/check.c | 62 ++++++++++++++++++++++++++++++++----
tools/objtool/check.h | 2 +-
3 files changed, 83 insertions(+), 8 deletions(-)
diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 7d70ea977fbe..9c4e1d985e40 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -9,6 +9,19 @@
#ifdef __ASSEMBLY__
+/*
+ * This should be used immediately before a retpoline alternative. It tells
+ * objtool where the retpolines are so that it can make sense of the control
+ * flow by just reading the original instruction(s) and ignoring the
+ * alternatives.
+ */
+.macro ANNOTATE_NOSPEC_ALTERNATIVE
+ .Lannotate_\@:
+ .pushsection .discard.nospec
+ .long .Lannotate_\@ - .
+ .popsection
+.endm
+
/*
* These are the bare retpoline primitives for indirect jmp and call.
* Do not use these directly; they only exist to make the ALTERNATIVE
@@ -43,6 +56,7 @@
*/
.macro JMP_NOSPEC reg:req
#ifdef CONFIG_RETPOLINE
+ ANNOTATE_NOSPEC_ALTERNATIVE
ALTERNATIVE_2 __stringify(jmp *\reg), \
__stringify(RETPOLINE_JMP \reg), X86_FEATURE_RETPOLINE, \
__stringify(lfence; jmp *\reg), X86_FEATURE_RETPOLINE_AMD
@@ -53,6 +67,7 @@
.macro CALL_NOSPEC reg:req
#ifdef CONFIG_RETPOLINE
+ ANNOTATE_NOSPEC_ALTERNATIVE
ALTERNATIVE_2 __stringify(call *\reg), \
__stringify(RETPOLINE_CALL \reg), X86_FEATURE_RETPOLINE,\
__stringify(lfence; call *\reg), X86_FEATURE_RETPOLINE_AMD
@@ -64,15 +79,25 @@
#else /* __ASSEMBLY__ */
#if defined(CONFIG_X86_64) && defined(RETPOLINE)
+
+#define ANNOTATE_NOSPEC_ALTERNATIVE \
+ "999:\n\t" \
+ ".pushsection .discard.nospec\n\t" \
+ ".long 999b - .\n\t" \
+ ".popsection\n\t"
+
/*
* Since the inline asm uses the %V modifier which is only in newer GCC,
* the 64-bit one is dependent on RETPOLINE not CONFIG_RETPOLINE.
*/
-# define CALL_NOSPEC ALTERNATIVE( \
+# define CALL_NOSPEC \
+ ANNOTATE_NOSPEC_ALTERNATIVE \
+ ALTERNATIVE( \
"call *%[thunk_target]\n", \
"call __x86_indirect_thunk_%V[thunk_target]\n", \
X86_FEATURE_RETPOLINE)
# define THUNK_TARGET(addr) [thunk_target] "r" (addr)
+
#elif defined(CONFIG_X86_32) && defined(CONFIG_RETPOLINE)
/*
* For i386 we use the original ret-equivalent retpoline, because
diff --git a/tools/objtool/check.c b/tools/objtool/check.c
index de053fb7049b..f40d46e24bcc 100644
--- a/tools/objtool/check.c
+++ b/tools/objtool/check.c
@@ -427,6 +427,40 @@ static void add_ignores(struct objtool_file *file)
}
}
+/*
+ * FIXME: For now, just ignore any alternatives which add retpolines. This is
+ * a temporary hack, as it doesn't allow ORC to unwind from inside a retpoline.
+ * But it at least allows objtool to understand the control flow *around* the
+ * retpoline.
+ */
+static int add_nospec_ignores(struct objtool_file *file)
+{
+ struct section *sec;
+ struct rela *rela;
+ struct instruction *insn;
+
+ sec = find_section_by_name(file->elf, ".rela.discard.nospec");
+ if (!sec)
+ return 0;
+
+ list_for_each_entry(rela, &sec->rela_list, list) {
+ if (rela->sym->type != STT_SECTION) {
+ WARN("unexpected relocation symbol type in %s", sec->name);
+ return -1;
+ }
+
+ insn = find_insn(file, rela->sym->sec, rela->addend);
+ if (!insn) {
+ WARN("bad .discard.nospec entry");
+ return -1;
+ }
+
+ insn->ignore_alts = true;
+ }
+
+ return 0;
+}
+
/*
* Find the destination instructions for all jumps.
*/
@@ -509,11 +543,18 @@ static int add_call_destinations(struct objtool_file *file)
dest_off = insn->offset + insn->len + insn->immediate;
insn->call_dest = find_symbol_by_offset(insn->sec,
dest_off);
+ /*
+ * FIXME: Thanks to retpolines, it's now considered
+ * normal for a function to call within itself. So
+ * disable this warning for now.
+ */
+#if 0
if (!insn->call_dest) {
WARN_FUNC("can't find call dest symbol at offset 0x%lx",
insn->sec, insn->offset, dest_off);
return -1;
}
+#endif
} else if (rela->sym->type == STT_SECTION) {
insn->call_dest = find_symbol_by_offset(rela->sym->sec,
rela->addend+4);
@@ -678,12 +719,6 @@ static int add_special_section_alts(struct objtool_file *file)
return ret;
list_for_each_entry_safe(special_alt, tmp, &special_alts, list) {
- alt = malloc(sizeof(*alt));
- if (!alt) {
- WARN("malloc failed");
- ret = -1;
- goto out;
- }
orig_insn = find_insn(file, special_alt->orig_sec,
special_alt->orig_off);
@@ -694,6 +729,10 @@ static int add_special_section_alts(struct objtool_file *file)
goto out;
}
+ /* Ignore retpoline alternatives. */
+ if (orig_insn->ignore_alts)
+ continue;
+
new_insn = NULL;
if (!special_alt->group || special_alt->new_len) {
new_insn = find_insn(file, special_alt->new_sec,
@@ -719,6 +758,13 @@ static int add_special_section_alts(struct objtool_file *file)
goto out;
}
+ alt = malloc(sizeof(*alt));
+ if (!alt) {
+ WARN("malloc failed");
+ ret = -1;
+ goto out;
+ }
+
alt->insn = new_insn;
list_add_tail(&alt->list, &orig_insn->alts);
@@ -1035,6 +1081,10 @@ static int decode_sections(struct objtool_file *file)
add_ignores(file);
+ ret = add_nospec_ignores(file);
+ if (ret)
+ return ret;
+
ret = add_jump_destinations(file);
if (ret)
return ret;
diff --git a/tools/objtool/check.h b/tools/objtool/check.h
index 47d9ea70a83d..dbadb304a410 100644
--- a/tools/objtool/check.h
+++ b/tools/objtool/check.h
@@ -44,7 +44,7 @@ struct instruction {
unsigned int len;
unsigned char type;
unsigned long immediate;
- bool alt_group, visited, dead_end, ignore, hint, save, restore;
+ bool alt_group, visited, dead_end, ignore, hint, save, restore, ignore_alts;
struct symbol *call_dest;
struct instruction *jump_dest;
struct list_head alts;
--
2.14.3