Skip to contentSkip to conversation

perf tools: Fix /proc/kallsyms map split

PATCH Applied Namhyung Kim 2 participants 8 messages Dec 2, 2025 – Dec 3, 2025 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

v2 cover perf tools: Fix /proc/kallsyms map split

Hello,

I found a weird bug in symbol handling for kallsyms. My system has a live patch which is a module and it has some symbols that conflict with the main kernel map.

* v2 changes)
- use map__kmaps() (Ian)
- use temporary root directory (Ian)
- Add Ian's Reviewed-by

For example, the symbols are common functions defined in the kernel like kmalloc and kfree. They have 'u' type which is a unique global symbol.

  $ grep ' u ' /proc/kallsyms
  ffffffff98798dd0 u kmalloc_caches	[livepatch]
  ffffffff97afb6c0 u kmalloc_trace_noprof	[livepatch]
  ffffffff9739e7a0 u __list_add_valid_or_report	[livepatch]
  ffffffff97afb240 u __kmalloc_noprof	[livepatch]
  ffffffff970597f0 u klp_enable_patch	[livepatch]
  ffffffff979939f0 u kfree	[livepatch]
  ...

They are duplicate symbols and will be removed by the fixup routine. But if symbol_conf.allow_aliases is set, they remain. This is the case for perf lock contention, and it caused a trouble with the kallsyms split code.

  $ grep ' kfree' /proc/kallsyms
  ffffffff97057a30 t kfree_rcu_shrink_scan
  ffffffff97394380 T kfree_strarray
  ffffffff9779b890 T kfree_skb_list_reason
  ffffffff9787be60 t kfree_pmc
  ffffffff979939f0 T kfree                         <<<--- here
  ffffffff979bbc50 T kfree_skb_partial
  ffffffff97a8f110 t kfree_rcu_work
  ffffffff97a8f2f0 t kfree_rcu_monitor
  ffffffff97a8f910 t kfree_rcu_shrink_count
  ffffffff97af67f0 T kfree_const
  ffffffff97afbbc0 T kfree_sensitive
  ffffffff97b5c4a0 T kfree_link
  ffffffff99255908 d kfree_rcu_shrinker
  ffffffff998beec0 T kfree_rcu_scheduler_running
  ffffffff979939f0 u kfree	[livepatch]        <<<--- duplicate

As the kfree function is in the livepatch module, any functions in the main kernel map that come later than 'kfree' will now to be splitted. This will create a lot of new kernel maps and loading them again will go to the routines to load kallsyms and split. So the process was in an infinite loop creating new maps and eventually gets killed.

I've added some defensive measures to prevent such situations and a test case to verify it. But maybe we need to do something for 'u' type symbols.

Thanks,
Namhyung

Namhyung Kim (5):
perf tools: Mark split kallsyms DSOs as loaded perf tools: Fix split kallsyms DSO counting perf tools: Fallback to initial kernel map properly perf tools: Use machine->root_dir to find /proc/kallsyms perf test: Add kallsyms split test

 tools/perf/tests/Build            |   1 +
 tools/perf/tests/builtin-test.c   |   1 +
 tools/perf/tests/kallsyms-split.c | 156 ++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h          |   1 +
 tools/perf/util/symbol.c          |  16 ++-
 5 files changed, 171 insertions(+), 4 deletions(-)
 create mode 100644 tools/perf/tests/kallsyms-split.c
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
author maintainer
#1 lore raw

PATCH v2 1/5 perf tools: Mark split kallsyms DSOs as loaded

The maps__split_kallsyms() will split symbols to module DSOs if it comes from a module. It also handled some unusual kernel symbols after modules by creating new kernel maps like "[kernel].0".

But they are pseudo DSOs to have those unexpected symbols. They should not be considered as unloaded kernel DSOs. Otherwise the dso__load() for them will end up calling dso__load_kallsyms() and then maps__split_kallsyms() again and again.

Fixes 2e538c4a1847291cf ("perf tools: Improve kernel/modules symbol lookup")
---
 tools/perf/util/symbol.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/util/symbol.c +1 -0
index d8fc5ea77f849326..5a38435d90c96092 100644
@@ -972,6 +972,7 @@ static int maps__split_kallsyms(struct maps *kmaps, struct dso *dso, u64 delta,
 				return -1;
 
 			dso__set_kernel(ndso, dso__kernel(dso));
+			dso__set_loaded(ndso);
 
 			curr_map = map__new2(pos->start, ndso);
 			if (curr_map == NULL) {
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
author maintainer
#2 lore raw

PATCH v2 2/5 perf tools: Fix split kallsyms DSO counting

In reply to Namhyung Kim

It's counted twice as it's increased after calling maps__insert(). I guess we want to increase it only after it's added properly.

Fixes 2e538c4a1847291cf ("perf tools: Improve kernel/modules symbol lookup")
---
 tools/perf/util/symbol.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/util/symbol.c +2 -2
index 5a38435d90c96092..8eea49c50453d13a 100644
@@ -960,11 +960,11 @@ static int maps__split_kallsyms(struct maps *kmaps, struct dso *dso, u64 delta,
 			if (dso__kernel(dso) == DSO_SPACE__KERNEL_GUEST)
 				snprintf(dso_name, sizeof(dso_name),
 					"[guest.kernel].%d",
-					kernel_range++);
+					kernel_range);
 			else
 				snprintf(dso_name, sizeof(dso_name),
 					"[kernel].%d",
-					kernel_range++);
+					kernel_range);
 
 			ndso = dso__new(dso_name);
 			map__zput(curr_map);
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
author maintainer
#3 lore raw

PATCH v2 3/5 perf tools: Fallback to initial kernel map properly

In reply to Namhyung Kim

In maps__split_kallsyms(), it assumes new kernel map when it finds a symbol without module after any module and the initial kernel map has some symbols. Because it expects modules are out of the kernel map so modules should not have symbols in the kernel map.

For example, the following memory map shows symbols and maps. Any symbols in the module 1 area will go to the module 1. The main kernel map starts at 0xffffffffbc200000. But if any symbol has a module between the symbols in that area, next symbols after 0xffffffffbd008000 will generate new kernel maps like [kernel].1.

   kernel address   |                     |
                    |                     |
 0xffffffffc0000000 |---------------------|
                    |     (symbols)       |
                    |        ...          |   <---  [kernel].N
 0xffffffffbc400000 |---------------------|
                    |     (symbols)       |
                    |      module 2       |   <---  bad?
 0xffffffffbc380000 |---------------------|
                    |        ...          |
                    |     (symbols)       |
                    |  [kernel.kallsyms]  |   <---  initial map
 0xffffffffbc200000 |---------------------|
                    |                     |
                    |                     |
 0xffffffffabcde000 |---------------------|
                    |     (symbols)       |
                    |      module 1       |
 0xffffffffabcd0000 |---------------------|

This is very fragile when the module has a symbol that falls into the main kernel map for some reason. My system has a livepatch module with such symbols. And it created a lot of new kernel maps after those symbols. But the symbol may have broken addresses and the later symbols can still be found in the initial kernel map.

Let's check the symbol address in the initial map and use it if found.

---
 tools/perf/util/symbol.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
tools/perf/util/symbol.c +2 -1
index 8eea49c50453d13a..b533fbf17a8b19a3 100644
@@ -951,7 +951,8 @@ static int maps__split_kallsyms(struct maps *kmaps, struct dso *dso, u64 delta,
 				pos->end -= delta;
 			}
 
-			if (count == 0) {
+			if (map__start(initial_map) <= (pos->start + delta) &&
+			    (pos->start + delta) < map__end(initial_map)) {
 				map__zput(curr_map);
 				curr_map = map__get(initial_map);
 				goto add_symbol;
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
author maintainer
#4 lore raw

PATCH v2 4/5 perf tools: Use machine->root_dir to find /proc/kallsyms

In reply to Namhyung Kim

This is for test functions to find the kallsyms correctly. It can find the machine from the kernel maps and use its root_dir. This is helpful to setup fake /proc directory for testing.

---
 tools/perf/util/symbol.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
tools/perf/util/symbol.c +7 -1
index b533fbf17a8b19a3..6ddbf4427c60dace 100644
@@ -2002,6 +2002,7 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 	char sbuild_id[SBUILD_ID_SIZE];
 	bool is_host = false;
 	char path[PATH_MAX];
+	struct maps *kmaps = map__kmaps(map);
 
 	if (!dso__has_build_id(dso)) {
 		/*
@@ -2038,8 +2039,13 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 		return strdup(path);
 
 	/* Use current /proc/kallsyms if possible */
-	if (is_host) {
 proc_kallsyms:
+	if (kmaps) {
+		struct machine *machine = maps__machine(kmaps);
+
+		scnprintf(path, sizeof(path), "%s/proc/kallsyms", machine->root_dir);
+		return strdup(path);
+	} else if (is_host) {
 		return strdup("/proc/kallsyms");
 	}
 
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
author maintainer
#5 lore raw

PATCH v2 5/5 perf test: Add kallsyms split test

In reply to Namhyung Kim

Create a fake root directory for /proc/{version,modules,kallsyms} in /tmp for testing. The kallsyms has a bad symbol in the module and it causes the main map splitted. The test ensures it only has two maps - kernel and the module and it finds the initial map after the module without creating the split maps like [kernel].0 and so on.

  $ perf test -vv "split kallsyms"
   69: split kallsyms:
  --- start ---
  test child forked, pid 1016196
  try to create fake root directory
  create kernel maps from the fake root directory
  maps__set_modules_path_dir: cannot open /tmp/perf-test.Zrv6Sy/lib/modules/X.Y.Z dir
  Problems setting modules path maps, continuing anyway...
  Failed to open /tmp/perf-test.Zrv6Sy/proc/kcore. Note /proc/kcore requires CAP_SYS_RAWIO capability to access.
  Using /tmp/perf-test.Zrv6Sy/proc/kallsyms for symbols
  kernel map loaded - check symbol and map
  ---- end(0) ----
   69: split kallsyms                                                  : Ok
---
 tools/perf/tests/Build            |   1 +
 tools/perf/tests/builtin-test.c   |   1 +
 tools/perf/tests/kallsyms-split.c | 156 ++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h          |   1 +
 4 files changed, 159 insertions(+)
 create mode 100644 tools/perf/tests/kallsyms-split.c
tools/perf/tests/Build +1 -0
index af67f8ef74b49c58..c2a67ce459417e0f 100644
@@ -69,6 +69,7 @@ perf-test-y += util.o
 perf-test-y += hwmon_pmu.o
 perf-test-y += tool_pmu.o
 perf-test-y += subcmd-help.o
+perf-test-y += kallsyms-split.o
 
 ifeq ($(SRCARCH),$(filter $(SRCARCH),x86 arm arm64 powerpc))
 perf-test-$(CONFIG_DWARF_UNWIND) += dwarf-unwind.o
tools/perf/tests/builtin-test.c +1 -0
index 9090e8238a447826..bd6ffa8e4578431a 100644
@@ -140,6 +140,7 @@ static struct test_suite *generic_tests[] = {
 	&suite__symbols,
 	&suite__util,
 	&suite__subcmd_help,
+	&suite__kallsyms_split,
 	NULL,
 };
 
tools/perf/tests/kallsyms-split.c new file +156 -0

156 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index f5fba95b6f3ffbd3..cb67ddbd03753531 100644
@@ -178,6 +178,7 @@ DECLARE_SUITE(event_groups);
 DECLARE_SUITE(symbols);
 DECLARE_SUITE(util);
 DECLARE_SUITE(subcmd_help);
+DECLARE_SUITE(kallsyms_split);
 
 /*
  * PowerPC and S390 do not support creation of instruction breakpoints using the
Signature · 1 line
2.52.0.158.g65b55ccf14-goog
reviewer Reviewed-by on v2 5/5
#6 lore raw

Re: [PATCH v2 5/5] perf test: Add kallsyms split test

Quoted 21 lines · Namhyung Kim

Create a fake root directory for /proc/{version,modules,kallsyms} in /tmp for testing. The kallsyms has a bad symbol in the module and it causes the main map splitted. The test ensures it only has two maps - kernel and the module and it finds the initial map after the module without creating the split maps like [kernel].0 and so on.

  $ perf test -vv "split kallsyms"
   69: split kallsyms:
  --- start ---
  test child forked, pid 1016196
  try to create fake root directory
  create kernel maps from the fake root directory
  maps__set_modules_path_dir: cannot open /tmp/perf-test.Zrv6Sy/lib/modules/X.Y.Z dir
  Problems setting modules path maps, continuing anyway...
  Failed to open /tmp/perf-test.Zrv6Sy/proc/kcore. Note /proc/kcore requires CAP_SYS_RAWIO capability to access.
  Using /tmp/perf-test.Zrv6Sy/proc/kallsyms for symbols
  kernel map loaded - check symbol and map
  ---- end(0) ----
   69: split kallsyms                                                  : Ok
Quoted 1 line

Thanks,
Ian

Quoted 209 lines
---
 tools/perf/tests/Build            |   1 +
 tools/perf/tests/builtin-test.c   |   1 +
 tools/perf/tests/kallsyms-split.c | 156 ++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h          |   1 +
 4 files changed, 159 insertions(+)
 create mode 100644 tools/perf/tests/kallsyms-split.c
tools/perf/tests/Build +1 -0
index af67f8ef74b49c58..c2a67ce459417e0f 100644
@@ -69,6 +69,7 @@ perf-test-y += util.o
 perf-test-y += hwmon_pmu.o
 perf-test-y += tool_pmu.o
 perf-test-y += subcmd-help.o
+perf-test-y += kallsyms-split.o
 
 ifeq ($(SRCARCH),$(filter $(SRCARCH),x86 arm arm64 powerpc))
 perf-test-$(CONFIG_DWARF_UNWIND) += dwarf-unwind.o
tools/perf/tests/builtin-test.c +1 -0
index 9090e8238a447826..bd6ffa8e4578431a 100644
@@ -140,6 +140,7 @@ static struct test_suite *generic_tests[] = {
        &suite__symbols,
        &suite__util,
        &suite__subcmd_help,
+       &suite__kallsyms_split,
        NULL,
 };
 
tools/perf/tests/kallsyms-split.c new file +156 -0
new file mode 100644
index 0000000000000000..bbbc66957e5d0453
@@ -0,0 +1,156 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/compiler.h>
+#include <fcntl.h>
+#include <signal.h>
+#include <unistd.h>
+#include <sys/stat.h>
+#include "util/dso.h"
+#include "util/map.h"
+#include "util/symbol.h"
+#include "util/debug.h"
+#include "util/machine.h"
+#include "tests.h"
+
+/*
+ * This test is to check whether a bad symbol in a module won't split kallsyms maps.
+ * The main_symbol[1-3] should belong to the main [kernel.kallsyms] map even if the
+ * bad_symbol from the module is found in the middle.
+ */
+static char root_template[] = "/tmp/perf-test.XXXXXX";
+static char *root_dir;
+
+static const char proc_version[] = "Linux version X.Y.Z (just for perf test)\n";
+static const char proc_modules[] = "module 4096 1 - Live 0xffffffffcd000000\n";
+static const char proc_kallsyms[] =
+       "ffffffffab200000 T _stext\n"
+       "ffffffffab200010 T good_symbol\n"
+       "ffffffffab200020 t bad_symbol\n"
+       "ffffffffab200030 t main_symbol1\n"
+       "ffffffffab200040 t main_symbol2\n"
+       "ffffffffab200050 t main_symbol3\n"
+       "ffffffffab200060 T _etext\n"
+       "ffffffffcd000000 T start_module\t[module]\n"
+       "ffffffffab200020 u bad_symbol\t[module]\n"
+       "ffffffffcd000040 T end_module\t[module]\n";
+
+static struct {
+       const char *name;
+       const char *contents;
+       long len;
+} proc_files[] = {
+       { "version", proc_version, sizeof(proc_version) - 1 },
+       { "modules", proc_modules, sizeof(proc_modules) - 1 },
+       { "kallsyms", proc_kallsyms, sizeof(proc_kallsyms) - 1 },
+};
+
+static void remove_proc_dir(int sig __maybe_unused)
+{
+       char buf[128];
+
+       if (root_dir == NULL)
+               return;
+
+       for (unsigned i = 0; i < ARRAY_SIZE(proc_files); i++) {
+               scnprintf(buf, sizeof(buf), "%s/proc/%s", root_dir, proc_files[i].name);
+               remove(buf);
+       }
+
+       scnprintf(buf, sizeof(buf), "%s/proc", root_dir);
+       rmdir(buf);
+
+       rmdir(root_dir);
+       root_dir = NULL;
+}
+
+static int create_proc_dir(void)
+{
+       char buf[128];
+
+       root_dir = mkdtemp(root_template);
+       if (root_dir == NULL)
+               return -1;
+
+       scnprintf(buf, sizeof(buf), "%s/proc", root_dir);
+       if (mkdir(buf, 0700) < 0)
+               goto err;
+
+       for (unsigned i = 0; i < ARRAY_SIZE(proc_files); i++) {
+               int fd, len;
+
+               scnprintf(buf, sizeof(buf), "%s/proc/%s", root_dir, proc_files[i].name);
+               fd = open(buf, O_RDWR | O_CREAT, 0600);
+               if (fd < 0)
+                       goto err;
+
+               len = write(fd, proc_files[i].contents, proc_files[i].len);
+               close(fd);
+               if (len != proc_files[i].len)
+                       goto err;
+       }
+       return 0;
+
+err:
+       remove_proc_dir(0);
+       return -1;
+}
+
+static int test__kallsyms_split(struct test_suite *test __maybe_unused,
+                               int subtest __maybe_unused)
+{
+       struct machine m;
+       struct map *map = NULL;
+       int ret = TEST_FAIL;
+
+       pr_debug("try to create fake root directory\n");
+       if (create_proc_dir() < 0) {
+               pr_debug("SKIP: cannot create a fake root directory\n");
+               return TEST_SKIP;
+       }
+
+       signal(SIGINT, remove_proc_dir);
+       signal(SIGPIPE, remove_proc_dir);
+       signal(SIGSEGV, remove_proc_dir);
+       signal(SIGTERM, remove_proc_dir);
+
+       pr_debug("create kernel maps from the fake root directory\n");
+       machine__init(&m, root_dir, HOST_KERNEL_ID);
+       if (machine__create_kernel_maps(&m) < 0) {
+               pr_debug("FAIL: failed to create kernel maps\n");
+               goto out;
+       }
+
+       /* force to use /proc/kallsyms */
+       symbol_conf.ignore_vmlinux = true;
+       symbol_conf.ignore_vmlinux_buildid = true;
+       symbol_conf.allow_aliases = true;
+
+       if (map__load(machine__kernel_map(&m)) < 0) {
+               pr_debug("FAIL: failed to load kallsyms\n");
+               goto out;
+       }
+
+       pr_debug("kernel map loaded - check symbol and map\n");
+       if (maps__nr_maps(machine__kernel_maps(&m)) != 2) {
+               pr_debug("FAIL: it should have the kernel and a module, but has %u maps\n",
+                        maps__nr_maps(machine__kernel_maps(&m)));
+               goto out;
+       }
+
+       if (machine__find_kernel_symbol_by_name(&m, "main_symbol3", &map) == NULL) {
+               pr_debug("FAIL: failed to find a symbol\n");
+               goto out;
+       }
+
+       if (!RC_CHK_EQUAL(map, machine__kernel_map(&m))) {
+               pr_debug("FAIL: the symbol is not in the kernel map\n");
+               goto out;
+       }
+       ret = TEST_OK;
+
+out:
+       remove_proc_dir(0);
+       machine__exit(&m);
+       return ret;
+}
+
+DEFINE_SUITE("split kallsyms", kallsyms_split);
tools/perf/tests/tests.h +1 -1
index f5fba95b6f3ffbd3..cb67ddbd03753531 100644
@@ -178,6 +178,7 @@ DECLARE_SUITE(event_groups);
 DECLARE_SUITE(symbols);
 DECLARE_SUITE(util);
 DECLARE_SUITE(subcmd_help);
+DECLARE_SUITE(kallsyms_split);
 
 /*
  * PowerPC and S390 do not support creation of instruction breakpoints using the
--

2.52.0.158.g65b55ccf14-goog

author maintainer Applied
#7 lore raw
In reply to Namhyung Kim
Quoted 10 lines · Namhyung Kim

I found a weird bug in symbol handling for kallsyms. My system has a live patch which is a module and it has some symbols that conflict with the main kernel map.

* v2 changes)
- use map__kmaps() (Ian)
- use temporary root directory (Ian)
- Add Ian's Reviewed-by

[...]

Applied to perf-tools-next, thanks!

Best regards,
Namhyung