Skip to contentSkip to conversation

kunit: fix use-after-free in debugfs when using kunit.filter

PATCH In review Florian Schmaus 3 participants 3 messages May 7 – May 14

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 kunit: fix use-after-free in debugfs when using kunit.filter

When the kernel is booted with a kunit filter (e.g., kunit.filter="speed!=slow"), the kunit executor dynamically allocates copies of the filtered test suites using kmalloc/kmemdup.

During the initial boot execution, kunit_debugfs_create_suite() creates debugfs files (such as /sys/kernel/debug/kunit/<suite>/run) and permanently stores a pointer to the dynamically allocated suite in the inode's i_private field.

Previously, the executor freed this dynamically allocated suite_set immediately after executing the boot-time tests. Because the debugfs nodes were not destroyed, any subsequent interaction with the debugfs run file from userspace triggered a use-after-free (UAF). On systems with architectural capabilities, like CHERI RISC-V, this resulted in an immediate fatal hardware exception due to the invalidation of the capability tags on the reclaimed memory. On other architectures, it resulted in silent memory corruption.

Fix this UAF by properly coupling the lifetime of the filtered suite memory allocation to the lifetime of the kunit subsystem and its associated VFS nodes. Ownership of the boot-time suite_set is now transferred to a global tracker ('kunit_boot_suites'), and the memory is cleanly released in kunit_exit() during module teardown.

---
 include/kunit/test.h |  1 +
 lib/kunit/executor.c | 19 ++++++++++++++++---
 lib/kunit/test.c     |  1 +
 3 files changed, 18 insertions(+), 3 deletions(-)
include/kunit/test.h +1 -0
index 9cd1594ab697..ce0573e196ce 100644
@@ -613,6 +613,7 @@ unsigned long kunit_vm_mmap(struct kunit *test, struct file *file,
 			    unsigned long offset);
 
 void kunit_cleanup(struct kunit *test);
+void kunit_free_boot_suites(void);
 
 void __printf(2, 3) kunit_log_append(struct string_stream *log, const char *fmt, ...);
 
lib/kunit/executor.c +16 -3
index 1fef217de11d..b0f8a41d61d3 100644
@@ -15,6 +15,16 @@ extern struct kunit_suite * const __kunit_suites_end[];
 extern struct kunit_suite * const __kunit_init_suites_start[];
 extern struct kunit_suite * const __kunit_init_suites_end[];
 
+static struct kunit_suite_set kunit_boot_suites;
+
+void kunit_free_boot_suites(void)
+{
+	if (kunit_boot_suites.start) {
+		kunit_free_suite_set(kunit_boot_suites);
+		kunit_boot_suites = (struct kunit_suite_set){ NULL, NULL };
+	}
+}
+
 static char *action_param;
 
 module_param_named(action, action_param, charp, 0400);
@@ -411,9 +421,12 @@ int kunit_run_all_tests(void)
 		pr_err("kunit executor: unknown action '%s'\n", action_param);
 
 free_out:
-	if (filter_glob_param || filter_param)
-		kunit_free_suite_set(suite_set);
-	else if (init_num_suites > 0)
+	if (filter_glob_param || filter_param) {
+		if (err)
+			kunit_free_suite_set(suite_set);
+		else
+			kunit_boot_suites = suite_set;
+	} else if (init_num_suites > 0)
 		/* Don't use kunit_free_suite_set because suites aren't individually allocated */
 		kfree(suite_set.start);
 
lib/kunit/test.c +1 -0
index 41e1c89799b6..99773e000e1b 100644
@@ -1075,6 +1075,7 @@ static void __exit kunit_exit(void)
 	kunit_bus_shutdown();
 
 	kunit_debugfs_cleanup();
+	kunit_free_boot_suites();
 }
 module_exit(kunit_exit);
 
Signature · 1 line
2.53.0
Reviewed-by on v1 1/1
#1 lore raw

Thus wrote Florian Schmaus ([email protected]):

Quoted 18 lines

When the kernel is booted with a kunit filter (e.g., kunit.filter="speed!=slow"), the kunit executor dynamically allocates copies of the filtered test suites using kmalloc/kmemdup.

During the initial boot execution, kunit_debugfs_create_suite() creates debugfs files (such as /sys/kernel/debug/kunit/<suite>/run) and permanently stores a pointer to the dynamically allocated suite in the inode's i_private field.

Previously, the executor freed this dynamically allocated suite_set immediately after executing the boot-time tests. Because the debugfs nodes were not destroyed, any subsequent interaction with the debugfs run file from userspace triggered a use-after-free (UAF). On systems with architectural capabilities, like CHERI RISC-V, this resulted in an immediate fatal hardware exception due to the invalidation of the capability tags on the reclaimed memory. On other architectures, it resulted in silent memory corruption.

Quoted 7 lines

Fix this UAF by properly coupling the lifetime of the filtered suite memory allocation to the lifetime of the kunit subsystem and its associated VFS nodes. Ownership of the boot-time suite_set is now transferred to a global tracker ('kunit_boot_suites'), and the memory is cleanly released in kunit_exit() during module teardown.

maintainer Reviewed-by on v1 1/1
#2 lore raw
In reply to Florian Schmaus

Le 07/05/2026 à 4:48 PM, 'Florian Schmaus' via KUnit Development a écrit :

Quoted 18 lines

When the kernel is booted with a kunit filter (e.g., kunit.filter="speed!=slow"), the kunit executor dynamically allocates copies of the filtered test suites using kmalloc/kmemdup.

During the initial boot execution, kunit_debugfs_create_suite() creates debugfs files (such as /sys/kernel/debug/kunit/<suite>/run) and permanently stores a pointer to the dynamically allocated suite in the inode's i_private field.

Previously, the executor freed this dynamically allocated suite_set immediately after executing the boot-time tests. Because the debugfs nodes were not destroyed, any subsequent interaction with the debugfs run file from userspace triggered a use-after-free (UAF). On systems with architectural capabilities, like CHERI RISC-V, this resulted in an immediate fatal hardware exception due to the invalidation of the capability tags on the reclaimed memory. On other architectures, it resulted in silent memory corruption.

Quoted 8 lines

Fix this UAF by properly coupling the lifetime of the filtered suite memory allocation to the lifetime of the kunit subsystem and its associated VFS nodes. Ownership of the boot-time suite_set is now transferred to a global tracker ('kunit_boot_suites'), and the memory is cleanly released in kunit_exit() during module teardown.

---

Nice catch. Alas, I don't have any CHERI-style setups at the moment to actually test it out, but it looks right to me, and doesn't break anything.

It might be useful to add a Fixes tag to this if you do a future
version: I think that the culprit commit is probably the original debugfs one:

Fixes e2219db280e3 ("kunit: add debugfs /sys/kernel/debug/kunit/<suite>/results display")

Cheers,
-- David

Quoted 66 lines
 include/kunit/test.h |  1 +
 lib/kunit/executor.c | 19 ++++++++++++++++---
 lib/kunit/test.c     |  1 +
 3 files changed, 18 insertions(+), 3 deletions(-)
include/kunit/test.h +1 -0
index 9cd1594ab697..ce0573e196ce 100644
@@ -613,6 +613,7 @@ unsigned long kunit_vm_mmap(struct kunit *test, struct file *file,
 			    unsigned long offset);
 
 void kunit_cleanup(struct kunit *test);
+void kunit_free_boot_suites(void);
 
 void __printf(2, 3) kunit_log_append(struct string_stream *log, const char *fmt, ...);
 
lib/kunit/executor.c +16 -3
index 1fef217de11d..b0f8a41d61d3 100644
@@ -15,6 +15,16 @@ extern struct kunit_suite * const __kunit_suites_end[];
 extern struct kunit_suite * const __kunit_init_suites_start[];
 extern struct kunit_suite * const __kunit_init_suites_end[];
 
+static struct kunit_suite_set kunit_boot_suites;
+
+void kunit_free_boot_suites(void)
+{
+	if (kunit_boot_suites.start) {
+		kunit_free_suite_set(kunit_boot_suites);
+		kunit_boot_suites = (struct kunit_suite_set){ NULL, NULL };
+	}
+}
+
 static char *action_param;
 
 module_param_named(action, action_param, charp, 0400);
@@ -411,9 +421,12 @@ int kunit_run_all_tests(void)
 		pr_err("kunit executor: unknown action '%s'\n", action_param);
 
 free_out:
-	if (filter_glob_param || filter_param)
-		kunit_free_suite_set(suite_set);
-	else if (init_num_suites > 0)
+	if (filter_glob_param || filter_param) {
+		if (err)
+			kunit_free_suite_set(suite_set);
+		else
+			kunit_boot_suites = suite_set;
+	} else if (init_num_suites > 0)
 		/* Don't use kunit_free_suite_set because suites aren't individually allocated */
 		kfree(suite_set.start);
 
lib/kunit/test.c +1 -0
index 41e1c89799b6..99773e000e1b 100644
@@ -1075,6 +1075,7 @@ static void __exit kunit_exit(void)
 	kunit_bus_shutdown();
 
 	kunit_debugfs_cleanup();
+	kunit_free_boot_suites();
 }
 module_exit(kunit_exit);