Skip to contentSkip to conversation

objtool/klp: sympos/module/alternative/etc fixes

PATCH Applied Josh Poimboeuf 6 participants 28 messages Aug 7 – Aug 15

Subsystems (2)
lore
Share & tools
lore

Josh Poimboeuf

author maintainer
#0 lore raw

v3 cover objtool/klp: sympos/module/alternative/etc fixes

v3:
- remove "size = 0" and add comments to create_fake_symbols()
- add patches 8 & 9

v2: https://lore.kernel.org/[email protected]
- rebased on tip/master (first 6 patches were merged)
- dropped original patches 7-8 (will post followup)
- added .klp.symid fix

v1: https://lore.kernel.org/[email protected]

This consolidates fixes for the klp-build issues reported by Joe over the last several weeks, plus some more things I found while testing/reviewing.

Joe Lawrence (1):
objtool/klp: Allow new references to module exports

Josh Poimboeuf (8):
objtool/klp: Fix vmlinux .klp.symid link error for .no_trim_symbol
symbols
objtool/klp: Fix size of empty special section entries objtool/klp: Ignore replacement offset of empty x86 alternatives objtool/klp: Explicitly disallow patching or referencing init
code/data
objtool/klp: Fix cross-module klp relocation section naming objtool/klp: Don't match local symbols against exports objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols objtool/klp: Fix line numbers in Module.symvers parse errors

 tools/objtool/arch/x86/special.c        |  27 ++++++
 tools/objtool/include/objtool/klp.h     |  10 ++-
 tools/objtool/include/objtool/special.h |   7 ++
 tools/objtool/klp-diff.c                | 106 ++++++++++++++++++++----
 tools/objtool/klp-post-link.c           |  53 +++++++-----
 tools/objtool/klp-symid.c               |   1 +
 tools/objtool/klp-sympos.c              |  10 +++
 7 files changed, 175 insertions(+), 39 deletions(-)
Signature · 1 line
2.54.0
author maintainer
#1 lore raw

PATCH v3 1/9 objtool/klp: Fix vmlinux .klp.symid link error for .no_trim_symbol symbols

Testing klp-build with arm64 produced the following linker error during the original kernel build:

__notrim.1' referenced in section .klp.symid' of vmlinux.o: defined in discarded section `.no_trim_symbol' of vmlinux.o

symbol_get() puts a static __notrim[] in .no_trim_symbol, which GCC names __notrim.1, __notrim.2, etc. Two or more built-in translation units calling symbol_get() thus produce duplicate names, resulting in corresponding .klp.symid references which trigger the above error.

Add .no_trim_symbol to the discarded section list so its symbols don't get symids.

Note this issue is not specific to arm64: it just needs two built-in symbol_get() callers. arm64 trips over it easily because it has KVM always compiled in vmlinux, whereas on x86 it's typically a module.

Fixes 029223d30162 ("objtool/klp: Add .klp.symid for sympos disambiguation")
---
 tools/objtool/klp-symid.c | 1 +
 1 file changed, 1 insertion(+)
tools/objtool/klp-symid.c +1 -0
index cf188cdfa6079..21d8708013aba 100644
@@ -31,6 +31,7 @@
 static const char * const discarded_secs[] = {
 	".discard",
 	".modinfo",
+	".no_trim_symbol",
 	"__tracepoint_check",
 };
 
Signature · 1 line
2.54.0
author maintainer
#3 lore raw

PATCH v3 2/9 objtool/klp: Fix size of empty special section entries

In reply to Josh Poimboeuf

create_fake_symbols() sizes each ANNOTATE_DATA_SPECIAL entry from the offset of the next annotation, falling back to the end of the section for the last entry. But the last entry is detected by a zero size, which also happens for an *empty* entry: ALTERNATIVE(oldinstr, "", ft) still annotates its zero-length replacement, at the same offset as the next entry's annotation.

So every empty replacement gets a fake symbol spanning the entire rest of .altinstr_replacement. That's harmless today only because find_symbol_containing() picks the smaller of two overlapping symbols.

Track whether a next annotation was found rather than inferring it from the size. A zero-length fake symbol is fine: find_symbol_containing() skips those, so the properly sized symbol at the same offset still wins.

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/klp-diff.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/objtool/klp-diff.c +11 -2
index 492d7a012cffe..38fae861d12c7 100644
@@ -1627,13 +1627,17 @@ static int create_fake_symbols(struct elf *elf)
 	for_each_reloc(sec->rsec, reloc) {
 		unsigned long offset, size;
 		struct reloc *next_reloc;
+		bool last = true;
 
 		if (annotype(elf, sec, reloc) != ANNOTYPE_DATA_SPECIAL)
 			continue;
 
 		offset = reloc_addend(reloc);
 
-		size = 0;
+		/*
+		 * Find the start of the next entry so the fake symbol size can
+		 * be calculated.
+		 */
 		next_reloc = reloc;
 		for_each_reloc_continue(sec->rsec, next_reloc) {
 			if (annotype(elf, sec, next_reloc) != ANNOTYPE_DATA_SPECIAL ||
@@ -1641,10 +1645,15 @@ static int create_fake_symbols(struct elf *elf)
 				continue;
 
 			size = reloc_addend(next_reloc) - offset;
+			last = false;
 			break;
 		}
 
-		if (!size)
+		/*
+		 * If no next entry found, this is the last entry, so its size
+		 * is from the current offset to the end of the section.
+		 */
+		if (last)
 			size = sec_size(reloc->sym->sec) - offset;
 
 		if (create_fake_symbol(elf, reloc->sym->sec, offset, size))
Signature · 1 line
2.54.0
author maintainer
#5 lore raw

PATCH v3 3/9 objtool/klp: Ignore replacement offset of empty x86 alternatives

In reply to Josh Poimboeuf

An x86 alternative with an empty replacement, e.g. the second entry of

ALTERNATIVE_2("orig", "repl", ft1, "", ft2)

has a replacementlen of zero. Its replacement offset still gets a relocation, but the label it points at is the end of the previous replacement, which is also the beginning of the *next* alternative's replacement. The value is meaningless; get_alt_entry() already ignores it for that reason.

klp diff doesn't ignore it. When such an alternative belongs to a changed function, cloning its relocations drags in the unrelated neighboring replacement, along with everything that replacement references. On an x86 clang/lto build an empty alternative in meminfo_proc_show() pulled in the replacement of an alternative in proc_kcore_init(), silently emitting a klp relocation against init text which has long since been freed by the time the patch is applied.

Add arch_alt_ignore_new_reloc() and skip such relocations when cloning. This has to be arch specific: on arm64 a zero-length replacement instead identifies an alternative callback, whose replacement offset points at the callback function and must be preserved.

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/arch/x86/special.c        | 27 +++++++++++++++++++++++++
 tools/objtool/include/objtool/special.h |  7 +++++++
 tools/objtool/klp-diff.c                |  6 +++++-
 3 files changed, 39 insertions(+), 1 deletion(-)
tools/objtool/arch/x86/special.c +27 -0
index e817a3fff4491..1e84c81bfcd81 100644
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 #include <string.h>
 
+#include <arch/special.h>
 #include <objtool/special.h>
 #include <objtool/builtin.h>
 #include <objtool/warn.h>
@@ -9,6 +10,32 @@
 /* cpu feature name array generated from cpufeatures.h */
 #include "cpu-feature-names.c"
 
+/*
+ * An alternative with an empty replacement, e.g. the second entry of
+ *
+ *   ALTERNATIVE_2("orig", "repl", ft1, "", ft2)
+ *
+ * still gets a relocation for its replacement offset.  But the label it points
+ * at is the end of the previous entry's replacement, which is also the
+ * beginning of the *next* entry's replacement.  The value is meaningless: it's
+ * only ever used with a length of zero.
+ */
+bool arch_alt_ignore_new_reloc(struct section *sec, unsigned long offset)
+{
+	unsigned long entry_off;
+
+	if (strcmp(sec->name, ".altinstructions"))
+		return false;
+
+	entry_off = offset - (offset % ALT_ENTRY_SIZE);
+
+	if (offset - entry_off != ALT_NEW_OFFSET)
+		return false;
+
+	return !*(unsigned char *)(sec->data->d_buf + entry_off +
+				   ALT_NEW_LEN_OFFSET);
+}
+
 void arch_handle_alternative(struct special_alt *alt)
 {
 	static struct special_alt *group, *prev;
tools/objtool/include/objtool/special.h +7 -0
index 121c3761899c1..620dbf6cb0e58 100644
@@ -32,6 +32,13 @@ int special_get_alts(struct elf *elf, struct list_head *alts);
 
 void arch_handle_alternative(struct special_alt *alt);
 
+/*
+ * Should the reloc at @offset -- the "new" (replacement) field of a special
+ * section group entry -- be ignored?  The meaning of a zero-length replacement
+ * is arch specific, so the arch decides.
+ */
+bool arch_alt_ignore_new_reloc(struct section *sec, unsigned long offset);
+
 bool arch_support_alt_relocation(struct special_alt *special_alt,
 				 struct instruction *insn,
 				 struct reloc *reloc);
tools/objtool/klp-diff.c +5 -1
index 38fae861d12c7..3923fabc13331 100644
@@ -12,7 +12,7 @@
 #include <objtool/arch.h>
 #include <objtool/klp.h>
 #include <objtool/util.h>
-#include <arch/special.h>
+#include <objtool/special.h>
 
 #include <linux/align.h>
 #include <linux/objtool_types.h>
@@ -1537,6 +1537,10 @@ static int clone_sym_relocs(struct elfs *e, struct symbol *patched_sym)
 		    !strcmp(patched_reloc->sym->sec->name, ".altinstr_aux"))
 			continue;
 
+		if (arch_alt_ignore_new_reloc(patched_sym->sec,
+					      reloc_offset(patched_reloc)))
+			continue;
+
 		ret = convert_reloc_sym(e->patched, patched_reloc);
 		if (ret < 0) {
 			ERROR_FUNC(patched_rsec->base, reloc_offset(patched_reloc),
Signature · 1 line
2.54.0
author maintainer
#7 lore raw

PATCH v3 4/9 objtool/klp: Explicitly disallow patching or referencing init code/data

In reply to Josh Poimboeuf

Explicitly disallow the patching and referencing of init code/data. Otherwise it could potentially introduce some odd edge cases depending on whether the target object's init section has been freed yet (note that the init code still exists in the target module when doing late module patching).

Such edge cases include sympos calculation and the patching and/or referencing of non-existent (init-freed) code/data. Not to mention the inherent differences in behavior that occur when the init code is only patched *some* of the time depending on module loading order or kernel config.

---
 tools/objtool/klp-sympos.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
tools/objtool/klp-sympos.c +10 -0
index bbfae516d3395..dfca9dd746812 100644
@@ -367,6 +367,11 @@ static unsigned long find_vmlinux_sympos(struct symbol *sym)
 	return sympos;
 }
 
+static bool is_init_sym(struct symbol *sym)
+{
+	return strstarts(sym->sec->name, ".init");
+}
+
 /*
  * "sympos" is used by livepatch to disambiguate duplicate symbol names.
  */
@@ -376,6 +381,11 @@ unsigned long klp_find_sympos(struct elf *elf, struct symbol *sym)
 	bool has_dup = false;
 	struct symbol *s;
 
+	if (is_init_sym(sym)) {
+		ERROR("%s: can't patch or reference init code/data", sym->name);
+		return ULONG_MAX;
+	}
+
 	if (sym->bind != STB_LOCAL)
 		return 0;
 
Signature · 1 line
2.54.0
author maintainer
#10 lore raw

PATCH v3 5/9 objtool/klp: Fix cross-module klp relocation section naming

In reply to Josh Poimboeuf

A klp relocation section is .klp.rela.<objname>.<secname>, where objname is the object being patched.

klp-build wrongly derives objname from where the referenced symbol lives, not where it's referenced. For a cross-module reference like patched can_isotp code calling can.ko's can_rx_unregister(), that gives .klp.rela.can..text rather than .klp.rela.can_isotp..text. Unless the patch happens to patch can.ko as well, the relocation never gets applied and the call goes off into the weeds.

Name the intermediate section __klp_relocs.<objname> so post-link can read the patched object's name from there.

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/include/objtool/klp.h | 10 ++++--
 tools/objtool/klp-diff.c            | 17 +++++++--
 tools/objtool/klp-post-link.c       | 53 +++++++++++++++++------------
 3 files changed, 52 insertions(+), 28 deletions(-)
tools/objtool/include/objtool/klp.h +7 -3
index 0118c2c170c3f..646d8e1f12eff 100644
@@ -14,11 +14,15 @@
 #define KLP_FUNCS_SEC	".init.klp_funcs"
 
 /*
- * __klp_relocs is an intermediate section which are created by klp diff and
- * converted into KLP symbols/relas by "objtool klp post-link".  This is needed
- * to work around the linker, which doesn't preserve SHN_LIVEPATCH or
+ * __klp_relocs.<objname> are intermediate sections which are created by klp
+ * diff and converted into KLP symbols/relas by "objtool klp post-link".  This
+ * is needed to work around the linker, which doesn't preserve SHN_LIVEPATCH or
  * SHF_RELA_LIVEPATCH, nor does it support having two RELA sections for a
  * single PROGBITS section.
+ *
+ * "objname" is the name of the object being patched ("vmlinux" or a module
+ * name).  post-link uses it to name the resulting
+ * .klp.rela.objname.section_name sections.
  */
 #define KLP_RELOCS_SEC	"__klp_relocs"
 #define KLP_STRINGS_SEC	".rodata.klp.str1.1"
tools/objtool/klp-diff.c +14 -3
index 3923fabc13331..e2c6c69dbb4f0 100644
@@ -1380,8 +1380,8 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	}
 
 	/*
-	 * Create the __klp_relocs entry.  This will be converted to an actual
-	 * KLP rela by "objtool klp post-link".
+	 * Create the __klp_relocs.<objname> entry.  This will be converted to
+	 * an actual KLP rela by "objtool klp post-link".
 	 *
 	 * This intermediate step is necessary to prevent corruption by the
 	 * linker, which doesn't know how to properly handle two rela sections
@@ -1389,7 +1389,18 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	 */
 
 	if (!klp_relocs) {
-		klp_relocs = elf_create_section(e->out, KLP_RELOCS_SEC, 0,
+		const char *objname = find_modname(e);
+		char sec_name[SEC_NAME_LEN];
+
+		if (!objname)
+			return -1;
+
+		/* section format: __klp_relocs.objname */
+		if (snprintf_check(sec_name, SEC_NAME_LEN,
+				   KLP_RELOCS_SEC ".%s", objname))
+			return -1;
+
+		klp_relocs = elf_create_section(e->out, sec_name, 0,
 						0, SHT_PROGBITS, 8, SHF_ALLOC);
 		if (!klp_relocs)
 			return -1;
tools/objtool/klp-post-link.c +31 -22

98 lines not on this page · raw message

Signature · 1 line
2.54.0
author maintainer
#12 lore raw

PATCH v3 6/9 objtool/klp: Don't match local symbols against exports

In reply to Josh Poimboeuf

While cloning a reloc, klp diff calls find_export() to determine whether the referenced symbol is exported. That decides whether the reference needs a klp reloc, which object the klp symbol belongs to, and whether the symbol's data needs to be copied into the patch module.

But find_export() matches purely on symbol name, so a static function or variable which happens to share its name with an export is mistaken for a reference to that export:

- klp_reloc_needed() creates a klp reloc pointing at the exporting module's symbol rather than the local one. For a vmlinux export it skips the klp reloc altogether, leaving a normal reloc which the module loader resolves to the vmlinux symbol.

- clone_reloc() treats the symbol as external and clones it without its data, leaving a dangling reference.

- validate_special_section_klp_reloc() attributes a static branch or call key to the wrong module, and for a vmlinux export skips the unsupported-key check entirely.

Exports are always global, so ignore local symbols in find_export().

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/klp-diff.c | 3 +++
 1 file changed, 3 insertions(+)
tools/objtool/klp-diff.c +3 -0
index e2c6c69dbb4f0..f5d5711623f03 100644
@@ -1101,6 +1101,9 @@ static struct export *find_export(struct symbol *sym)
 {
 	struct export *export;
 
+	if (is_local_sym(sym))
+		return NULL;
+
 	hash_for_each_possible(exports, export, hash, str_hash(sym->name)) {
 		if (!strcmp(export->sym, sym->name))
 			return export;
Signature · 1 line
2.54.0
author maintainer
#16 lore raw

PATCH v3 7/9 objtool/klp: Allow new references to module exports

In reply to Josh Poimboeuf

From: Joe Lawrence <[email protected]>

klp_reloc_needed() returns true for module exports to support late-module patching. However, clone_reloc_klp() unconditionally rejects symbols without a twin (i.e., new references added by the patch), even when the symbol is a known export from Module.symvers.

Relax the check: allow new references to exported symbols by only erroring on !twin when there is no export. The export metadata from Module.symvers provides sufficient context to emit the klp-relocation without a twin.

For a module export that isn't sufficient on its own though, as the resulting klp relocation will only be resolved at patch-enable time if the exporting module is loaded.

If the original (unpatched) module already depends on the exporting module, the dependency is safe: the module loader ensures the dependency is satisfied before the patched module can be loaded, so the klp relocation target will exist.

However, if the patch introduces a reference to a module that the original doesn't depend on, there is no such guarantee. The exporting module could be absent or could be unloaded at any time, leading to a relocation failure or use-after-free.

So also add a build-time check: when a new symbol reference (no twin) targets a module export, verify that the original module already has at least one UNDEF symbol resolving to that same exporting module. If not, error out with a diagnostic message.

---
 tools/objtool/klp-diff.c | 35 +++++++++++++++++++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)
tools/objtool/klp-diff.c +33 -2
index f5d5711623f03..6d34186d8b24c 100644
@@ -1294,6 +1294,28 @@ static int convert_reloc_sym(struct elf *elf, struct reloc *reloc)
 	return convert_reloc_secsym_to_sym(elf, reloc);
 }
 
+/*
+ * Check if the original module already has a dependency on dep_mod, i.e. it
+ * already references at least one export from that module.
+ */
+static bool has_module_dep(struct elfs *e, const char *dep_mod)
+{
+	struct symbol *sym;
+
+	for_each_sym(e->orig, sym) {
+		struct export *exp;
+
+		if (!is_undef_sym(sym) || is_weak_sym(sym))
+			continue;
+
+		exp = find_export(sym);
+		if (exp && !strcmp(exp->mod, dep_mod))
+			return true;
+	}
+
+	return false;
+}
+
 /*
  * Convert a regular relocation to a klp relocation (sort of).
  */
@@ -1313,8 +1335,17 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	unsigned long sympos;
 
 	if (!patched_sym->twin) {
-		ERROR("unexpected klp reloc for new symbol %s", patched_sym->name);
-		return -1;
+		if (!export) {
+			ERROR("unexpected klp reloc for new symbol %s", patched_sym->name);
+			return -1;
+		}
+
+		if (strcmp(export->mod, "vmlinux") &&
+		    !has_module_dep(e, export->mod)) {
+			ERROR("%s: new reference to %s (exported by %s) would create an undeclared module dependency",
+			      patched_sym->name, export->sym, export->mod);
+			return -1;
+		}
 	}
 
 	/*
Signature · 1 line
2.54.0
author maintainer
#18 lore raw

PATCH v3 8/9 objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

In reply to Josh Poimboeuf

EXPORT_SYMBOL_FOR_MODULES() puts a symbol in a "module:<names>" namespace, which the module loader grants access to by matching the importing module's name against that list.

klp_reloc_needed() only creates a klp reloc for module-owned exports; a vmlinux export gets a normal reloc. For a vmlinux symbol exported with EXPORT_SYMBOL_FOR_MODULES(), using a normal reloc results in a modpost failure in klp-build:

ERROR: modpost: module livepatch-foo uses symbol mpol_shared_policy_lookup from namespace module:kvm, but does not import it.

And the modpost error is correct: even with that error removed, the patch module would fail to load:

livepatch_foo: module uses symbol (mpol_shared_policy_lookup) from namespace module:kvm, but does not import it.
livepatch_foo: Unknown symbol mpol_shared_policy_lookup (err -22)

Treat it like an unexported symbol by using a klp reloc.

Note this only affects "module:" namespaces. Ordinary namespaced exports continue to work with normal relocs thanks to copy_import_ns(), which propagates the patched object's import_ns tags to the patch module.

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/klp-diff.c | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)
tools/objtool/klp-diff.c +23 -5
index 6d34186d8b24c..0f135b74a5b0c 100644
@@ -30,7 +30,9 @@ struct elfs {
 
 struct export {
 	struct hlist_node hash;
-	char *mod, *sym;
+	char *mod;
+	char *sym;
+	bool mod_ns;
 };
 
 bool debug, debug_correlate, debug_clone;
@@ -135,7 +137,7 @@ static int read_exports(void)
 	}
 
 	while (fgets(line, 1024, file)) {
-		char *sym, *mod, *type;
+		char *sym, *mod, *type, *namespace;
 		struct export *export;
 
 		sym = strchr(line, '\t');
@@ -162,6 +164,14 @@ static int read_exports(void)
 
 		*type++ = '\0';
 
+		namespace = strchr(type, '\t');
+		if (!namespace) {
+			ERROR("malformed Module.symvers (namespace) at line %d", line_num);
+			return -1;
+		}
+
+		*namespace++ = '\0';
+
 		if (*sym == '\0' || *mod == '\0') {
 			ERROR("malformed Module.symvers at line %d", line_num);
 			return -1;
@@ -188,6 +198,9 @@ static int read_exports(void)
 			return -1;
 		}
 
+		/* EXPORT_SYMBOL_FOR_MODULES() */
+		export->mod_ns = strstarts(namespace, "module:");
+
 		hash_add(exports, &export->hash, str_hash(sym));
 	}
 
@@ -1174,11 +1187,16 @@ static bool klp_reloc_needed(struct reloc *patched_reloc)
 	 * clusterfunk that is late module patching, the patch module is
 	 * allowed to be loaded before any modules it depends on.
 	 *
-	 * If exported by vmlinux, a normal reloc will do.
+	 * If exported by vmlinux to all modules, a normal reloc will do.
 	 */
 	export = find_export(patched_sym);
-	if (export)
-		return strcmp(export->mod, "vmlinux");
+	if (export) {
+		if (strcmp(export->mod, "vmlinux"))
+			return true;
+
+		/* EXPORT_SYMBOL_FOR_MODULES() gets a klp reloc */
+		return export->mod_ns;
+	}
 
 	if (!patched_sym->twin) {
 		/*
Signature · 1 line
2.54.0
author maintainer
#24 lore raw

PATCH v3 9/9 objtool/klp: Fix line numbers in Module.symvers parse errors

In reply to Josh Poimboeuf

read_exports() reports the offending line number when it fails to parse Module.symvers. The counter is initialized to 1 but never incremented, so every error blames line 1 regardless of where the bad line is.

---
 tools/objtool/klp-diff.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tools/objtool/klp-diff.c +3 -1
index 0f135b74a5b0c..b6b72ed71bf02 100644
@@ -118,7 +118,7 @@ static int read_exports(void)
 {
 	const char *symvers = "Module.symvers";
 	char line[1024], *path = NULL;
-	unsigned int line_num = 1;
+	unsigned int line_num = 0;
 	FILE *file;
 
 	file = fopen(symvers, "r");
@@ -140,6 +140,8 @@ static int read_exports(void)
 		char *sym, *mod, *type, *namespace;
 		struct export *export;
 
+		line_num++;
+
 		sym = strchr(line, '\t');
 		if (!sym) {
 			ERROR("malformed Module.symvers (sym) at line %d", line_num);
Signature · 1 line
2.54.0
#13 lore raw

Re: [PATCH v3 6/9] objtool/klp: Don't match local symbols against exports

In reply to Josh Poimboeuf
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Unexported global symbols that share a name with an export are incorrectly matched against the global exports list, causing misrouted livepatch relocations. --

--- Patch [6]: [PATCH v3 6/9] objtool/klp: Don't match local symbols against exports --- commit 91941e947fdde72be13c7760463244a153a175ce
Author: Josh Poimboeuf <[email protected]>

objtool/klp: Don't match local symbols against exports

This commit updates find_export() in klp-diff.c to ignore local symbols. This addresses an issue where static functions or variables that share a name with an export are incorrectly treated as export references during livepatch generation.

tools/objtool/klp-diff.c static struct export *find_export(struct symbol *sym) · line 1109

Quoted 13 lines
tools/objtool/klp-diff.c +3 -0
@@ -1101,6 +1101,9 @@ static struct export *find_export(struct symbol *sym)
 {
 	struct export *export;
 
+	if (is_local_sym(sym))
+		return NULL;
+
 	hash_for_each_possible(exports, export, hash, str_hash(sym->name)) {
 		if (!strcmp(export->sym, sym->name))
 			return export;

[Severity: High]
This is a pre-existing issue, but does this still leave unexported global symbols vulnerable to incorrect matching?

If a module contains an unexported global symbol that shares its name with an unrelated exported symbol, it will bypass the is_local_sym() check.

When the module is livepatched and klp-diff processes a relocation to this internal symbol, find_export() will perform a string match and return the unrelated export. Would this cause a KLP relocation to incorrectly point to the external module's export instead of the internal symbol?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=6
author maintainer
#14 lore raw

Re: [PATCH v3 6/9] objtool/klp: Don't match local symbols against exports

Quoted 35 lines · On Fri, Aug 07, 2026 at 09:46:11PM +0000, [email protected] wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Unexported global symbols that share a name with an export are incorrectly matched against the global exports list, causing misrouted livepatch relocations. --

--- Patch [6]: [PATCH v3 6/9] objtool/klp: Don't match local symbols against exports --- commit 91941e947fdde72be13c7760463244a153a175ce
Author: Josh Poimboeuf <[email protected]>

objtool/klp: Don't match local symbols against exports

This commit updates find_export() in klp-diff.c to ignore local symbols. This addresses an issue where static functions or variables that share a name with an export are incorrectly treated as export references during livepatch generation.

tools/objtool/klp-diff.c static struct export *find_export(struct symbol *sym) · line 1109

Quoted 13 lines
tools/objtool/klp-diff.c +3 -0
@@ -1101,6 +1101,9 @@ static struct export *find_export(struct symbol *sym)
 {
 	struct export *export;
 
+	if (is_local_sym(sym))
+		return NULL;
+
 	hash_for_each_possible(exports, export, hash, str_hash(sym->name)) {
 		if (!strcmp(export->sym, sym->name))
 			return export;

[Severity: High]
This is a pre-existing issue, but does this still leave unexported global symbols vulnerable to incorrect matching?

Quoted 7 lines

If a module contains an unexported global symbol that shares its name with an unrelated exported symbol, it will bypass the is_local_sym() check.

When the module is livepatched and klp-diff processes a relocation to this internal symbol, find_export() will perform a string match and return the unrelated export. Would this cause a KLP relocation to incorrectly point to the external module's export instead of the internal symbol?

I'm not convinced this is a realistic scenario (module having a global symbol with the same name as an exported symbol), but as mentioned it's a preexisting issue that can be fixed in a followup.

Signature · 1 line
Josh
maintainer Acked-by on the v3 series
#27 lore raw
In reply to Josh Poimboeuf
Quoted 30 lines · Josh Poimboeuf

v3:
- remove "size = 0" and add comments to create_fake_symbols()
- add patches 8 & 9

v2: https://lore.kernel.org/[email protected]
- rebased on tip/master (first 6 patches were merged)
- dropped original patches 7-8 (will post followup)
- added .klp.symid fix

v1: https://lore.kernel.org/[email protected]

This consolidates fixes for the klp-build issues reported by Joe over the last several weeks, plus some more things I found while testing/reviewing.

Joe Lawrence (1):
objtool/klp: Allow new references to module exports

Josh Poimboeuf (8):
objtool/klp: Fix vmlinux .klp.symid link error for .no_trim_symbol
symbols
objtool/klp: Fix size of empty special section entries objtool/klp: Ignore replacement offset of empty x86 alternatives objtool/klp: Explicitly disallow patching or referencing init
code/data
objtool/klp: Fix cross-module klp relocation section naming objtool/klp: Don't match local symbols against exports objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols objtool/klp: Fix line numbers in Module.symvers parse errors

Quoted 9 lines
 tools/objtool/arch/x86/special.c        |  27 ++++++
 tools/objtool/include/objtool/klp.h     |  10 ++-
 tools/objtool/include/objtool/special.h |   7 ++
 tools/objtool/klp-diff.c                | 106 ++++++++++++++++++++----
 tools/objtool/klp-post-link.c           |  53 +++++++-----
 tools/objtool/klp-symid.c               |   1 +
 tools/objtool/klp-sympos.c              |  10 +++
 7 files changed, 175 insertions(+), 39 deletions(-)

For the series:

Previous test cases from the earlier parent patchsets and reports run positive now.

Thanks,

Signature · 1 line
Joe
maintainer Acked-by on v3 4/9
#8 lore raw

Re: [PATCH v3 4/9] objtool/klp: Explicitly disallow patching or referencing init code/data

In reply to Josh Poimboeuf
Quoted 7 lines · Josh Poimboeuf

Explicitly disallow the patching and referencing of init code/data. Otherwise it could potentially introduce some odd edge cases depending on whether the target object's init section has been freed yet (note that the init code still exists in the target module when doing late module patching).

Quoted 7 lines

Such edge cases include sympos calculation and the patching and/or referencing of non-existent (init-freed) code/data. Not to mention the inherent differences in behavior that occur when the init code is only patched *some* of the time depending on module loading order or kernel config.

maintainer Acked-by on v3 8/9
#19 lore raw

Re: [PATCH v3 8/9] objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

In reply to Josh Poimboeuf
Quoted 25 lines · Josh Poimboeuf

EXPORT_SYMBOL_FOR_MODULES() puts a symbol in a "module:<names>" namespace, which the module loader grants access to by matching the importing module's name against that list.

klp_reloc_needed() only creates a klp reloc for module-owned exports; a vmlinux export gets a normal reloc. For a vmlinux symbol exported with EXPORT_SYMBOL_FOR_MODULES(), using a normal reloc results in a modpost failure in klp-build:

ERROR: modpost: module livepatch-foo uses symbol mpol_shared_policy_lookup from namespace module:kvm, but does not import it.

And the modpost error is correct: even with that error removed, the patch module would fail to load:

livepatch_foo: module uses symbol (mpol_shared_policy_lookup) from namespace module:kvm, but does not import it.
livepatch_foo: Unknown symbol mpol_shared_policy_lookup (err -22)

Treat it like an unexported symbol by using a klp reloc.

Note this only affects "module:" namespaces. Ordinary namespaced exports continue to work with normal relocs thanks to copy_import_ns(), which propagates the patched object's import_ns tags to the patch module.

Quoted 4 lines
Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
maintainer Acked-by on v3 9/9
#25 lore raw

Re: [PATCH v3 9/9] objtool/klp: Fix line numbers in Module.symvers parse errors

In reply to Josh Poimboeuf
Quoted 6 lines · Josh Poimboeuf

read_exports() reports the offending line number when it fails to parse Module.symvers. The counter is initialized to 1 but never incremented, so every error blames line 1 regardless of where the bad line is.

merge bot Applied
#26 lore raw

[tip: objtool/core] objtool/klp: Fix line numbers in Module.symvers parse errors

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     51c1de13486315ad46a74b40ec17124916a08277
Gitweb:        https://git.kernel.org/tip/51c1de13486315ad46a74b40ec17124916a08277
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:54 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:58 -07:00

objtool/klp: Fix line numbers in Module.symvers parse errors

read_exports() reports the offending line number when it fails to parse Module.symvers. The counter is initialized to 1 but never incremented, so every error blames line 1 regardless of where the bad line is.

---
 tools/objtool/klp-diff.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tools/objtool/klp-diff.c +3 -1
index 1f05646..cdc63ab 100644
@@ -118,7 +118,7 @@ static int read_exports(void)
 {
 	const char *symvers = "Module.symvers";
 	char line[1024], *path = NULL;
-	unsigned int line_num = 1;
+	unsigned int line_num = 0;
 	FILE *file;
 
 	file = fopen(symvers, "r");
@@ -140,6 +140,8 @@ static int read_exports(void)
 		char *sym, *mod, *type, *namespace;
 		struct export *export;
 
+		line_num++;
+
 		sym = strchr(line, '\t');
 		if (!sym) {
 			ERROR("malformed Module.symvers (sym) at line %d", line_num);
merge bot Applied
#20 lore raw

[tip: objtool/core] objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     4cd3cfb8b54feca89a682720434092a87bfa4967
Gitweb:        https://git.kernel.org/tip/4cd3cfb8b54feca89a682720434092a87bfa4967
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:53 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:51 -07:00

objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

EXPORT_SYMBOL_FOR_MODULES() puts a symbol in a "module:<names>" namespace, which the module loader grants access to by matching the importing module's name against that list.

klp_reloc_needed() only creates a klp reloc for module-owned exports; a vmlinux export gets a normal reloc. For a vmlinux symbol exported with EXPORT_SYMBOL_FOR_MODULES(), using a normal reloc results in a modpost failure in klp-build:

ERROR: modpost: module livepatch-foo uses symbol mpol_shared_policy_lookup from namespace module:kvm, but does not import it.

And the modpost error is correct: even with that error removed, the patch module would fail to load:

livepatch_foo: module uses symbol (mpol_shared_policy_lookup) from namespace module:kvm, but does not import it.
livepatch_foo: Unknown symbol mpol_shared_policy_lookup (err -22)

Treat it like an unexported symbol by using a klp reloc.

Note this only affects "module:" namespaces. Ordinary namespaced exports continue to work with normal relocs thanks to copy_import_ns(), which propagates the patched object's import_ns tags to the patch module.

---
 tools/objtool/klp-diff.c | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)
tools/objtool/klp-diff.c +23 -5
index 0211e5c..1f05646 100644
@@ -30,7 +30,9 @@ struct elfs {
 
 struct export {
 	struct hlist_node hash;
-	char *mod, *sym;
+	char *mod;
+	char *sym;
+	bool mod_ns;
 };
 
 bool debug, debug_correlate, debug_clone;
@@ -135,7 +137,7 @@ static int read_exports(void)
 	}
 
 	while (fgets(line, 1024, file)) {
-		char *sym, *mod, *type;
+		char *sym, *mod, *type, *namespace;
 		struct export *export;
 
 		sym = strchr(line, '\t');
@@ -162,6 +164,14 @@ static int read_exports(void)
 
 		*type++ = '\0';
 
+		namespace = strchr(type, '\t');
+		if (!namespace) {
+			ERROR("malformed Module.symvers (namespace) at line %d", line_num);
+			return -1;
+		}
+
+		*namespace++ = '\0';
+
 		if (*sym == '\0' || *mod == '\0') {
 			ERROR("malformed Module.symvers at line %d", line_num);
 			return -1;
@@ -188,6 +198,9 @@ static int read_exports(void)
 			return -1;
 		}
 
+		/* EXPORT_SYMBOL_FOR_MODULES() */
+		export->mod_ns = strstarts(namespace, "module:");
+
 		hash_add(exports, &export->hash, str_hash(sym));
 	}
 
@@ -1175,11 +1188,16 @@ static bool klp_reloc_needed(struct reloc *patched_reloc)
 	 * clusterfunk that is late module patching, the patch module is
 	 * allowed to be loaded before any modules it depends on.
 	 *
-	 * If exported by vmlinux, a normal reloc will do.
+	 * If exported by vmlinux to all modules, a normal reloc will do.
 	 */
 	export = find_export(patched_sym);
-	if (export)
-		return strcmp(export->mod, "vmlinux");
+	if (export) {
+		if (strcmp(export->mod, "vmlinux"))
+			return true;
+
+		/* EXPORT_SYMBOL_FOR_MODULES() gets a klp reloc */
+		return export->mod_ns;
+	}
 
 	if (!patched_sym->twin) {
 		/*
merge bot Applied
#17 lore raw

[tip: objtool/core] objtool/klp: Allow new references to module exports

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     72d76d0c18ebd40f6a32df875ff40528564849c1
Gitweb:        https://git.kernel.org/tip/72d76d0c18ebd40f6a32df875ff40528564849c1
Author:        Joe Lawrence <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:52 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:51 -07:00

objtool/klp: Allow new references to module exports

klp_reloc_needed() returns true for module exports to support late-module patching. However, clone_reloc_klp() unconditionally rejects symbols without a twin (i.e., new references added by the patch), even when the symbol is a known export from Module.symvers.

Relax the check: allow new references to exported symbols by only erroring on !twin when there is no export. The export metadata from Module.symvers provides sufficient context to emit the klp-relocation without a twin.

For a module export that isn't sufficient on its own though, as the resulting klp relocation will only be resolved at patch-enable time if the exporting module is loaded.

If the original (unpatched) module already depends on the exporting module, the dependency is safe: the module loader ensures the dependency is satisfied before the patched module can be loaded, so the klp relocation target will exist.

However, if the patch introduces a reference to a module that the original doesn't depend on, there is no such guarantee. The exporting module could be absent or could be unloaded at any time, leading to a relocation failure or use-after-free.

So also add a build-time check: when a new symbol reference (no twin) targets a module export, verify that the original module already has at least one UNDEF symbol resolving to that same exporting module. If not, error out with a diagnostic message.

---
 tools/objtool/klp-diff.c | 35 +++++++++++++++++++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)
tools/objtool/klp-diff.c +33 -2
index a7cdba0..0211e5c 100644
@@ -1296,6 +1296,28 @@ static int convert_reloc_sym(struct elf *elf, struct reloc *reloc)
 }
 
 /*
+ * Check if the original module already has a dependency on dep_mod, i.e. it
+ * already references at least one export from that module.
+ */
+static bool has_module_dep(struct elfs *e, const char *dep_mod)
+{
+	struct symbol *sym;
+
+	for_each_sym(e->orig, sym) {
+		struct export *exp;
+
+		if (!is_undef_sym(sym) || is_weak_sym(sym))
+			continue;
+
+		exp = find_export(sym);
+		if (exp && !strcmp(exp->mod, dep_mod))
+			return true;
+	}
+
+	return false;
+}
+
+/*
  * Convert a regular relocation to a klp relocation (sort of).
  */
 static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
@@ -1314,8 +1336,17 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	unsigned long sympos;
 
 	if (!patched_sym->twin) {
-		ERROR("unexpected klp reloc for new symbol %s", patched_sym->name);
-		return -1;
+		if (!export) {
+			ERROR("unexpected klp reloc for new symbol %s", patched_sym->name);
+			return -1;
+		}
+
+		if (strcmp(export->mod, "vmlinux") &&
+		    !has_module_dep(e, export->mod)) {
+			ERROR("%s: new reference to %s (exported by %s) would create an undeclared module dependency",
+			      patched_sym->name, export->sym, export->mod);
+			return -1;
+		}
 	}
 
 	/*
merge bot Applied
#15 lore raw

[tip: objtool/core] objtool/klp: Don't match local symbols against exports

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     86a697572c6271083694025eefa8de518eeef623
Gitweb:        https://git.kernel.org/tip/86a697572c6271083694025eefa8de518eeef623
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:51 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:51 -07:00

objtool/klp: Don't match local symbols against exports

While cloning a reloc, klp diff calls find_export() to determine whether the referenced symbol is exported. That decides whether the reference needs a klp reloc, which object the klp symbol belongs to, and whether the symbol's data needs to be copied into the patch module.

But find_export() matches purely on symbol name, so a static function or variable which happens to share its name with an export is mistaken for a reference to that export:

- klp_reloc_needed() creates a klp reloc pointing at the exporting module's symbol rather than the local one. For a vmlinux export it skips the klp reloc altogether, leaving a normal reloc which the module loader resolves to the vmlinux symbol.

- clone_reloc() treats the symbol as external and clones it without its data, leaving a dangling reference.

- validate_special_section_klp_reloc() attributes a static branch or call key to the wrong module, and for a vmlinux export skips the unsupported-key check entirely.

Exports are always global, so ignore local symbols in find_export().

---
 tools/objtool/klp-diff.c | 3 +++
 1 file changed, 3 insertions(+)
tools/objtool/klp-diff.c +3 -0
index e266a54..a7cdba0 100644
@@ -1102,6 +1102,9 @@ static struct export *find_export(struct symbol *sym)
 {
 	struct export *export;
 
+	if (is_local_sym(sym))
+		return NULL;
+
 	hash_for_each_possible(exports, export, hash, str_hash(sym->name)) {
 		if (!strcmp(export->sym, sym->name))
 			return export;
merge bot Applied
#9 lore raw

[tip: objtool/core] objtool/klp: Explicitly disallow patching or referencing init code/data

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     69f361b8a7a2f65c1bb236ea0899cdaad7267653
Gitweb:        https://git.kernel.org/tip/69f361b8a7a2f65c1bb236ea0899cdaad7267653
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:49 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:38 -07:00

objtool/klp: Explicitly disallow patching or referencing init code/data

Explicitly disallow the patching and referencing of init code/data. Otherwise it could potentially introduce some odd edge cases depending on whether the target object's init section has been freed yet (note that the init code still exists in the target module when doing late module patching).

Such edge cases include sympos calculation and the patching and/or referencing of non-existent (init-freed) code/data. Not to mention the inherent differences in behavior that occur when the init code is only patched *some* of the time depending on module loading order or kernel config.

---
 tools/objtool/klp-sympos.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
tools/objtool/klp-sympos.c +10 -0
index bbfae51..dfca9dd 100644
@@ -367,6 +367,11 @@ static unsigned long find_vmlinux_sympos(struct symbol *sym)
 	return sympos;
 }
 
+static bool is_init_sym(struct symbol *sym)
+{
+	return strstarts(sym->sec->name, ".init");
+}
+
 /*
  * "sympos" is used by livepatch to disambiguate duplicate symbol names.
  */
@@ -376,6 +381,11 @@ unsigned long klp_find_sympos(struct elf *elf, struct symbol *sym)
 	bool has_dup = false;
 	struct symbol *s;
 
+	if (is_init_sym(sym)) {
+		ERROR("%s: can't patch or reference init code/data", sym->name);
+		return ULONG_MAX;
+	}
+
 	if (sym->bind != STB_LOCAL)
 		return 0;
 
merge bot Applied
#11 lore raw

[tip: objtool/core] objtool/klp: Fix cross-module klp relocation section naming

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     07f14d6af9d7791fdc2b44a3dd0693e42c71d2b8
Gitweb:        https://git.kernel.org/tip/07f14d6af9d7791fdc2b44a3dd0693e42c71d2b8
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:50 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 16:10:51 -07:00

objtool/klp: Fix cross-module klp relocation section naming

A klp relocation section is .klp.rela.<objname>.<secname>, where objname is the object being patched.

klp-build wrongly derives objname from where the referenced symbol lives, not where it's referenced. For a cross-module reference like patched can_isotp code calling can.ko's can_rx_unregister(), that gives .klp.rela.can..text rather than .klp.rela.can_isotp..text. Unless the patch happens to patch can.ko as well, the relocation never gets applied and the call goes off into the weeds.

Name the intermediate section __klp_relocs.<objname> so post-link can read the patched object's name from there.

---
 tools/objtool/include/objtool/klp.h | 10 +++--
 tools/objtool/klp-diff.c            | 17 +++++++--
 tools/objtool/klp-post-link.c       | 53 ++++++++++++++++------------
 3 files changed, 52 insertions(+), 28 deletions(-)
tools/objtool/include/objtool/klp.h +7 -3
index 0118c2c..646d8e1 100644
@@ -14,11 +14,15 @@
 #define KLP_FUNCS_SEC	".init.klp_funcs"
 
 /*
- * __klp_relocs is an intermediate section which are created by klp diff and
- * converted into KLP symbols/relas by "objtool klp post-link".  This is needed
- * to work around the linker, which doesn't preserve SHN_LIVEPATCH or
+ * __klp_relocs.<objname> are intermediate sections which are created by klp
+ * diff and converted into KLP symbols/relas by "objtool klp post-link".  This
+ * is needed to work around the linker, which doesn't preserve SHN_LIVEPATCH or
  * SHF_RELA_LIVEPATCH, nor does it support having two RELA sections for a
  * single PROGBITS section.
+ *
+ * "objname" is the name of the object being patched ("vmlinux" or a module
+ * name).  post-link uses it to name the resulting
+ * .klp.rela.objname.section_name sections.
  */
 #define KLP_RELOCS_SEC	"__klp_relocs"
 #define KLP_STRINGS_SEC	".rodata.klp.str1.1"
tools/objtool/klp-diff.c +14 -3
index df2fe01..e266a54 100644
@@ -1381,8 +1381,8 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	}
 
 	/*
-	 * Create the __klp_relocs entry.  This will be converted to an actual
-	 * KLP rela by "objtool klp post-link".
+	 * Create the __klp_relocs.<objname> entry.  This will be converted to
+	 * an actual KLP rela by "objtool klp post-link".
 	 *
 	 * This intermediate step is necessary to prevent corruption by the
 	 * linker, which doesn't know how to properly handle two rela sections
@@ -1390,7 +1390,18 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	 */
 
 	if (!klp_relocs) {
-		klp_relocs = elf_create_section(e->out, KLP_RELOCS_SEC, 0,
+		const char *objname = find_modname(e);
+		char sec_name[SEC_NAME_LEN];
+
+		if (!objname)
+			return -1;
+
+		/* section format: __klp_relocs.objname */
+		if (snprintf_check(sec_name, SEC_NAME_LEN,
+				   KLP_RELOCS_SEC ".%s", objname))
+			return -1;
+
+		klp_relocs = elf_create_section(e->out, sec_name, 0,
 						0, SHT_PROGBITS, 8, SHF_ALLOC);
 		if (!klp_relocs)
 			return -1;
tools/objtool/klp-post-link.c +31 -22

98 lines not on this page · raw message

merge bot Applied
#6 lore raw

[tip: objtool/core] objtool/klp: Ignore replacement offset of empty x86 alternatives

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     636f230ce21e2730072c58280f1fa0bed4a00eed
Gitweb:        https://git.kernel.org/tip/636f230ce21e2730072c58280f1fa0bed4a00eed
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:48 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 15:04:57 -07:00

objtool/klp: Ignore replacement offset of empty x86 alternatives

An x86 alternative with an empty replacement, e.g. the second entry of

ALTERNATIVE_2("orig", "repl", ft1, "", ft2)

has a replacementlen of zero. Its replacement offset still gets a relocation, but the label it points at is the end of the previous replacement, which is also the beginning of the *next* alternative's replacement. The value is meaningless; get_alt_entry() already ignores it for that reason.

klp diff doesn't ignore it. When such an alternative belongs to a changed function, cloning its relocations drags in the unrelated neighboring replacement, along with everything that replacement references. On an x86 clang/lto build an empty alternative in meminfo_proc_show() pulled in the replacement of an alternative in proc_kcore_init(), silently emitting a klp relocation against init text which has long since been freed by the time the patch is applied.

Add arch_alt_ignore_new_reloc() and skip such relocations when cloning. This has to be arch specific: on arm64 a zero-length replacement instead identifies an alternative callback, whose replacement offset points at the callback function and must be preserved.

---
 tools/objtool/arch/x86/special.c        | 27 ++++++++++++++++++++++++-
 tools/objtool/include/objtool/special.h |  7 ++++++-
 tools/objtool/klp-diff.c                |  6 ++++-
 3 files changed, 39 insertions(+), 1 deletion(-)
tools/objtool/arch/x86/special.c +27 -0
index e817a3f..1e84c81 100644
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 #include <string.h>
 
+#include <arch/special.h>
 #include <objtool/special.h>
 #include <objtool/builtin.h>
 #include <objtool/warn.h>
@@ -9,6 +10,32 @@
 /* cpu feature name array generated from cpufeatures.h */
 #include "cpu-feature-names.c"
 
+/*
+ * An alternative with an empty replacement, e.g. the second entry of
+ *
+ *   ALTERNATIVE_2("orig", "repl", ft1, "", ft2)
+ *
+ * still gets a relocation for its replacement offset.  But the label it points
+ * at is the end of the previous entry's replacement, which is also the
+ * beginning of the *next* entry's replacement.  The value is meaningless: it's
+ * only ever used with a length of zero.
+ */
+bool arch_alt_ignore_new_reloc(struct section *sec, unsigned long offset)
+{
+	unsigned long entry_off;
+
+	if (strcmp(sec->name, ".altinstructions"))
+		return false;
+
+	entry_off = offset - (offset % ALT_ENTRY_SIZE);
+
+	if (offset - entry_off != ALT_NEW_OFFSET)
+		return false;
+
+	return !*(unsigned char *)(sec->data->d_buf + entry_off +
+				   ALT_NEW_LEN_OFFSET);
+}
+
 void arch_handle_alternative(struct special_alt *alt)
 {
 	static struct special_alt *group, *prev;
tools/objtool/include/objtool/special.h +7 -0
index 121c376..620dbf6 100644
@@ -32,6 +32,13 @@ int special_get_alts(struct elf *elf, struct list_head *alts);
 
 void arch_handle_alternative(struct special_alt *alt);
 
+/*
+ * Should the reloc at @offset -- the "new" (replacement) field of a special
+ * section group entry -- be ignored?  The meaning of a zero-length replacement
+ * is arch specific, so the arch decides.
+ */
+bool arch_alt_ignore_new_reloc(struct section *sec, unsigned long offset);
+
 bool arch_support_alt_relocation(struct special_alt *special_alt,
 				 struct instruction *insn,
 				 struct reloc *reloc);
tools/objtool/klp-diff.c +5 -1
index b4220e1..df2fe01 100644
@@ -12,7 +12,7 @@
 #include <objtool/arch.h>
 #include <objtool/klp.h>
 #include <objtool/util.h>
-#include <arch/special.h>
+#include <objtool/special.h>
 
 #include <linux/align.h>
 #include <linux/objtool_types.h>
@@ -1538,6 +1538,10 @@ static int clone_sym_relocs(struct elfs *e, struct symbol *patched_sym)
 		    !strcmp(patched_reloc->sym->sec->name, ".altinstr_aux"))
 			continue;
 
+		if (arch_alt_ignore_new_reloc(patched_sym->sec,
+					      reloc_offset(patched_reloc)))
+			continue;
+
 		ret = convert_reloc_sym(e->patched, patched_reloc);
 		if (ret < 0) {
 			ERROR_FUNC(patched_rsec->base, reloc_offset(patched_reloc),
merge bot Applied
#4 lore raw

[tip: objtool/core] objtool/klp: Fix size of empty special section entries

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     1ea786097cd79522b76cbd41beeb8f84ef3a76f4
Gitweb:        https://git.kernel.org/tip/1ea786097cd79522b76cbd41beeb8f84ef3a76f4
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:47 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 15:04:57 -07:00

objtool/klp: Fix size of empty special section entries

create_fake_symbols() sizes each ANNOTATE_DATA_SPECIAL entry from the offset of the next annotation, falling back to the end of the section for the last entry. But the last entry is detected by a zero size, which also happens for an *empty* entry: ALTERNATIVE(oldinstr, "", ft) still annotates its zero-length replacement, at the same offset as the next entry's annotation.

So every empty replacement gets a fake symbol spanning the entire rest of .altinstr_replacement. That's harmless today only because find_symbol_containing() picks the smaller of two overlapping symbols.

Track whether a next annotation was found rather than inferring it from the size. A zero-length fake symbol is fine: find_symbol_containing() skips those, so the properly sized symbol at the same offset still wins.

---
 tools/objtool/klp-diff.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/objtool/klp-diff.c +11 -2
index c5284d2..b4220e1 100644
@@ -1628,13 +1628,17 @@ static int create_fake_symbols(struct elf *elf)
 	for_each_reloc(sec->rsec, reloc) {
 		unsigned long offset, size;
 		struct reloc *next_reloc;
+		bool last = true;
 
 		if (annotype(elf, sec, reloc) != ANNOTYPE_DATA_SPECIAL)
 			continue;
 
 		offset = reloc_addend(reloc);
 
-		size = 0;
+		/*
+		 * Find the start of the next entry so the fake symbol size can
+		 * be calculated.
+		 */
 		next_reloc = reloc;
 		for_each_reloc_continue(sec->rsec, next_reloc) {
 			if (annotype(elf, sec, next_reloc) != ANNOTYPE_DATA_SPECIAL ||
@@ -1642,10 +1646,15 @@ static int create_fake_symbols(struct elf *elf)
 				continue;
 
 			size = reloc_addend(next_reloc) - offset;
+			last = false;
 			break;
 		}
 
-		if (!size)
+		/*
+		 * If no next entry found, this is the last entry, so its size
+		 * is from the current offset to the end of the section.
+		 */
+		if (last)
 			size = sec_size(reloc->sym->sec) - offset;
 
 		if (create_fake_symbol(elf, reloc->sym->sec, offset, size))
merge bot Applied
#2 lore raw

[tip: objtool/core] objtool/klp: Fix vmlinux .klp.symid link error for .no_trim_symbol symbols

In reply to Josh Poimboeuf
The following commit has been merged into the objtool/core branch of tip:

The following commit has been merged into the objtool/core branch of tip:

Commit-ID:     3a83d9cd3d0e66c0ef99ce84f972ed26f50c990f
Gitweb:        https://git.kernel.org/tip/3a83d9cd3d0e66c0ef99ce84f972ed26f50c990f
Author:        Josh Poimboeuf <[email protected]>
AuthorDate:    Fri, 07 Aug 2026 14:37:46 -07:00
Committer:     Josh Poimboeuf <[email protected]>
CommitterDate: Tue, 11 Aug 2026 15:04:57 -07:00

objtool/klp: Fix vmlinux .klp.symid link error for .no_trim_symbol symbols

Testing klp-build with arm64 produced the following linker error during the original kernel build:

__notrim.1' referenced in section .klp.symid' of vmlinux.o: defined in discarded section `.no_trim_symbol' of vmlinux.o

symbol_get() puts a static __notrim[] in .no_trim_symbol, which GCC names __notrim.1, __notrim.2, etc. Two or more built-in translation units calling symbol_get() thus produce duplicate names, resulting in corresponding .klp.symid references which trigger the above error.

Add .no_trim_symbol to the discarded section list so its symbols don't get symids.

Note this issue is not specific to arm64: it just needs two built-in symbol_get() callers. arm64 trips over it easily because it has KVM always compiled in vmlinux, whereas on x86 it's typically a module.

---
 tools/objtool/klp-symid.c | 1 +
 1 file changed, 1 insertion(+)
tools/objtool/klp-symid.c +1 -0
index cf188cd..21d8708 100644
@@ -31,6 +31,7 @@
 static const char * const discarded_secs[] = {
 	".discard",
 	".modinfo",
+	".no_trim_symbol",
 	"__tracepoint_check",
 };
 
#21 lore raw

Re: [PATCH v3 8/9] objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

In reply to Josh Poimboeuf

Hi Josh,

Quoted 103 lines · Josh Poimboeuf

EXPORT_SYMBOL_FOR_MODULES() puts a symbol in a "module:<names>" namespace, which the module loader grants access to by matching the importing module's name against that list.

klp_reloc_needed() only creates a klp reloc for module-owned exports; a vmlinux export gets a normal reloc. For a vmlinux symbol exported with EXPORT_SYMBOL_FOR_MODULES(), using a normal reloc results in a modpost failure in klp-build:

ERROR: modpost: module livepatch-foo uses symbol mpol_shared_policy_lookup from namespace module:kvm, but does not import it.

And the modpost error is correct: even with that error removed, the patch module would fail to load:

livepatch_foo: module uses symbol (mpol_shared_policy_lookup) from namespace module:kvm, but does not import it.
livepatch_foo: Unknown symbol mpol_shared_policy_lookup (err -22)

Treat it like an unexported symbol by using a klp reloc.

Note this only affects "module:" namespaces. Ordinary namespaced exports continue to work with normal relocs thanks to copy_import_ns(), which propagates the patched object's import_ns tags to the patch module.

Fixes dd590d4d57eb ("objtool/klp: Introduce klp diff subcommand for diffing object files")
---
 tools/objtool/klp-diff.c | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)
tools/objtool/klp-diff.c +23 -6
index 6d34186d8b24c..0f135b74a5b0c 100644
@@ -30,7 +30,9 @@ struct elfs {
 
 struct export {
        struct hlist_node hash;
-       char *mod, *sym;
+       char *mod;
+       char *sym;
+       bool mod_ns;
 };
 
 bool debug, debug_correlate, debug_clone;
@@ -135,7 +137,7 @@ static int read_exports(void)
        }
 
        while (fgets(line, 1024, file)) {
-               char *sym, *mod, *type;
+               char *sym, *mod, *type, *namespace;
                struct export *export;
 
                sym = strchr(line, '\t');
@@ -162,6 +164,14 @@ static int read_exports(void)
 
                *type++ = '\0';
 
+               namespace = strchr(type, '\t');
+               if (!namespace) {
+                       ERROR("malformed Module.symvers (namespace) at line %d", line_num);
+                       return -1;
+               }
+
+               *namespace++ = '\0';
+
                if (*sym == '\0' || *mod == '\0') {
                        ERROR("malformed Module.symvers at line %d", line_num);
                        return -1;
@@ -188,6 +198,9 @@ static int read_exports(void)
                        return -1;
                }
 
+               /* EXPORT_SYMBOL_FOR_MODULES() */
+               export->mod_ns = strstarts(namespace, "module:");
+
                hash_add(exports, &export->hash, str_hash(sym));
        }
 
@@ -1174,11 +1187,16 @@ static bool klp_reloc_needed(struct reloc *patched_reloc)
         * clusterfunk that is late module patching, the patch module is
         * allowed to be loaded before any modules it depends on.
         *
-        * If exported by vmlinux, a normal reloc will do.
+        * If exported by vmlinux to all modules, a normal reloc will do.
         */
        export = find_export(patched_sym);
-       if (export)
-               return strcmp(export->mod, "vmlinux");
+       if (export) {
+               if (strcmp(export->mod, "vmlinux"))
+                       return true;
+
+               /* EXPORT_SYMBOL_FOR_MODULES() gets a klp reloc */
+               return export->mod_ns;
+       }
 
        if (!patched_sym->twin) {
                /*
--
Quoted 3 lines

2.54.0

Following up on the other thread [1], I noticed that when a patch is touching a module function with a reference to one of these module-exported symbols, the patch/module is rejected because KLP relocs referencing vmlinux symbols are not allowed from module-specific livepatch relocation sections. I was able to reproduce this with a simple module/livepatch combo that depends on one of these symbols [2] (see samples/livepatch/testmod.c and test.patch):

root@debian-vm:~$ insmod livepatch-test.ko
root@debian-vm:~$ insmod testmod.ko
insmod: ERROR: could not insert module testmod.ko: Invalid parameters

With dmesg:

[  655.596876] livepatch_test: loading out-of-tree module taints kernel.
[  655.600961] livepatch_test: tainting kernel with TAINT_LIVEPATCH
[  655.605436] livepatch: enabling patch 'livepatch_test'
[  655.609119] livepatch: 'livepatch_test': starting patching transition
[  656.653454] livepatch: 'livepatch_test': patching complete
[  738.777872] livepatch: invalid access to vmlinux symbol
'get_task_policy' from module-specific livepatch relocation section
[  738.784899] livepatch: failed to initialize patch 'livepatch_test'
for module 'testmod' (-22)
[  738.790371] livepatch: patch 'livepatch_test' failed for module

'testmod', refusing to load module 'testmod'

Do you recommend a strategy for working around this, or is this something that would have to be fixed in the kernel?

Refs:

[1]: https://lore.kernel.org/all/CADBMgpyY6R_YmEqhZSw=88-Bfkv9=s-N-x1siE1uPAgH_fTURw@mail.gmail.com/
[2]: https://github.com/dylanbhatch/linux/tree/mod-ns-lp

Thanks,
Dylan

author maintainer
#22 lore raw

Re: [PATCH v3 8/9] objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

Quoted 23 lines · Dylan Hatch

Following up on the other thread [1], I noticed that when a patch is touching a module function with a reference to one of these module-exported symbols, the patch/module is rejected because KLP relocs referencing vmlinux symbols are not allowed from module-specific livepatch relocation sections. I was able to reproduce this with a simple module/livepatch combo that depends on one of these symbols [2] (see samples/livepatch/testmod.c and test.patch):

root@debian-vm:~$ insmod livepatch-test.ko
root@debian-vm:~$ insmod testmod.ko
insmod: ERROR: could not insert module testmod.ko: Invalid parameters

With dmesg:

[  655.596876] livepatch_test: loading out-of-tree module taints kernel.
[  655.600961] livepatch_test: tainting kernel with TAINT_LIVEPATCH
[  655.605436] livepatch: enabling patch 'livepatch_test'
[  655.609119] livepatch: 'livepatch_test': starting patching transition
[  656.653454] livepatch: 'livepatch_test': patching complete
[  738.777872] livepatch: invalid access to vmlinux symbol
'get_task_policy' from module-specific livepatch relocation section
[  738.784899] livepatch: failed to initialize patch 'livepatch_test'
for module 'testmod' (-22)
[  738.790371] livepatch: patch 'livepatch_test' failed for module
Quoted 4 lines

'testmod', refusing to load module 'testmod'

Do you recommend a strategy for working around this, or is this something that would have to be fixed in the kernel?

Ah, this is another tooling issue, let me work up a patch.

Signature · 1 line
Josh
author maintainer
#23 lore raw

Re: [PATCH v3 8/9] objtool/klp: Fix relocations for EXPORT_SYMBOL_FOR_MODULES() symbols

Quoted 26 lines · Josh Poimboeuf
Quoted 25 lines · Dylan Hatch

Following up on the other thread [1], I noticed that when a patch is touching a module function with a reference to one of these module-exported symbols, the patch/module is rejected because KLP relocs referencing vmlinux symbols are not allowed from module-specific livepatch relocation sections. I was able to reproduce this with a simple module/livepatch combo that depends on one of these symbols [2] (see samples/livepatch/testmod.c and test.patch):

root@debian-vm:~$ insmod livepatch-test.ko
root@debian-vm:~$ insmod testmod.ko
insmod: ERROR: could not insert module testmod.ko: Invalid parameters

With dmesg:

[  655.596876] livepatch_test: loading out-of-tree module taints kernel.
[  655.600961] livepatch_test: tainting kernel with TAINT_LIVEPATCH
[  655.605436] livepatch: enabling patch 'livepatch_test'
[  655.609119] livepatch: 'livepatch_test': starting patching transition
[  656.653454] livepatch: 'livepatch_test': patching complete
[  738.777872] livepatch: invalid access to vmlinux symbol
'get_task_policy' from module-specific livepatch relocation section
[  738.784899] livepatch: failed to initialize patch 'livepatch_test'
for module 'testmod' (-22)
[  738.790371] livepatch: patch 'livepatch_test' failed for module

'testmod', refusing to load module 'testmod'

Quoted 4 lines
Quoted 2 lines

Do you recommend a strategy for working around this, or is this something that would have to be fixed in the kernel?

Ah, this is another tooling issue, let me work up a patch.

Here is an untested diff, I'll post a proper patch once I get a chance to test it.

tools/objtool/include/objtool/klp.h +3 -2
index 646d8e1f12eff..c57775d78c71e 100644
@@ -20,8 +20,9 @@
  * SHF_RELA_LIVEPATCH, nor does it support having two RELA sections for a
  * single PROGBITS section.
  *
- * "objname" is the name of the object being patched ("vmlinux" or a module
- * name).  post-link uses it to name the resulting
+ * "objname" is the object whose loading gates the relocation: "vmlinux" for
+ * references to vmlinux symbols, otherwise the name of the module being
+ * patched.  post-link uses it to name the resulting
  * .klp.rela.objname.section_name sections.
  */
 #define KLP_RELOCS_SEC	"__klp_relocs"
tools/objtool/klp-diff.c +22 -13
index a66049e0726a6..16681a76f13d0 100644
@@ -1344,13 +1344,14 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 			   struct section *sec, unsigned long offset,
 			   struct export *export)
 {
+	const char *sym_modname, *sym_orig_name, *sec_objname;
 	struct symbol *patched_sym = patched_reloc->sym;
 	s64 addend = reloc_addend(patched_reloc);
-	const char *sym_modname, *sym_orig_name;
-	static struct section *klp_relocs;
 	char tombstone_name[SYM_NAME_LEN];
 	struct symbol *sym, *klp_sym;
 	unsigned long klp_reloc_off;
+	struct section *klp_relocs;
+	char sec_name[SEC_NAME_LEN];
 	char sym_name[SYM_NAME_LEN];
 	struct klp_reloc klp_reloc;
 	unsigned long sympos;
@@ -1441,20 +1442,28 @@ static int clone_reloc_klp(struct elfs *e, struct reloc *patched_reloc,
 	 * This intermediate step is necessary to prevent corruption by the
 	 * linker, which doesn't know how to properly handle two rela sections
 	 * applying to the same base section.
+	 *
+	 * The objname decides when the reloc gets applied.  A reference to a
+	 * vmlinux symbol goes in the vmlinux section so it gets applied when
+	 * the patch module loads.  Everything else goes in the patched
+	 * object's section, applied when the patched module is loaded.
 	 */
 
+	if (!strcmp(sym_modname, "vmlinux")) {
+		sec_objname = "vmlinux";
+	} else {
+		sec_objname = find_modname(e);
+		if (!sec_objname)
+			return -1;
+	}
+
+	/* section format: __klp_relocs.objname */
+	if (snprintf_check(sec_name, SEC_NAME_LEN,
+			   KLP_RELOCS_SEC ".%s", sec_objname))
+		return -1;
+
+	klp_relocs = find_section_by_name(e->out, sec_name);
 	if (!klp_relocs) {
-		const char *objname = find_modname(e);
-		char sec_name[SEC_NAME_LEN];
-
-		if (!objname)
-			return -1;
-
-		/* section format: __klp_relocs.objname */
-		if (snprintf_check(sec_name, SEC_NAME_LEN,
-				   KLP_RELOCS_SEC ".%s", objname))
-			return -1;
-
 		klp_relocs = elf_create_section(e->out, sec_name, 0,
 						0, SHT_PROGBITS, 8, SHF_ALLOC);
 		if (!klp_relocs)