Skip to contentSkip to conversation

ext4: Forcing subclasses to have same name pointer as their parent class

PATCH Superseded botta633 3 participants 5 messages Jul 14, 2024 1 message not in the archive

Subsystems (3)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

botta633

author
#0 lore raw

PATCH v2 1/2 ext4: Forcing subclasses to have same name pointer as their parent class

From: Ahmed Ehab <[email protected]>

Preventing lockdep_set_subclass from creating a new instance of the string literal. Hence, we will always have the same class->name among parent and subclasses. This prevents kernel panics when looking up a lock class while comparing class locks and class names.

Fixes fd5e3f5fe27
---
 include/linux/lockdep.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/lockdep.h +1 -1
index 08b0d1d9d78b..df8fa5929de7 100644
@@ -173,7 +173,7 @@ static inline void lockdep_init_map(struct lockdep_map *lock, const char *name,
 			      (lock)->dep_map.lock_type)
 
 #define lockdep_set_subclass(lock, sub)					\
-	lockdep_init_map_type(&(lock)->dep_map, #lock, (lock)->dep_map.key, sub,\
+	lockdep_init_map_type(&(lock)->dep_map, (lock)->dep_map.name, (lock)->dep_map.key, sub,\
 			      (lock)->dep_map.wait_type_inner,		\
 			      (lock)->dep_map.wait_type_outer,		\
 			      (lock)->dep_map.lock_type)
Signature · 1 line
2.45.2
author
#1 lore raw

PATCH 2/2 ext4: Testing lock class and subclass got the same name pointer

Checking if the lockdep_map->name will change when setting the subclass. It shouldn't change so that the lock class and subclass will have the same name

Fixes fd5e3f5fe27
---
 lib/locking-selftest.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
lib/locking-selftest.c +16 -0
index 6f6a5fc85b42..1d7885205f36 100644
@@ -2710,12 +2710,24 @@ static void local_lock_3B(void)
 
 }
 
+static void class_subclass_X1_name(void)
+{
+	const char *name_before_subclass = rwsem_X1.dep_map.name;
+	const char *name_after_subclass;
+
+	WARN_ON(!rwsem_X1.dep_map.name);
+	lockdep_set_subclass(&rwsem_X1, 1);
+	WARN_ON(name_before_subclass != name_after_subclass);
+}
+
 static void local_lock_tests(void)
 {
 	printk("  --------------------------------------------------------------------------\n");
 	printk("  | local_lock tests |\n");
 	printk("  ---------------------\n");
 
+	init_class_X(&lock_X1, &rwlock_X1, &mutex_X1, &rwsem_X1);
+
 	print_testname("local_lock inversion  2");
 	dotest(local_lock_2, SUCCESS, LOCKTYPE_LL);
 	pr_cont("\n");
@@ -2727,6 +2739,10 @@ static void local_lock_tests(void)
 	print_testname("local_lock inversion 3B");
 	dotest(local_lock_3B, FAILURE, LOCKTYPE_LL);
 	pr_cont("\n");
+
+	print_testname("Class and subclass");
+	dotest(class_subclass_X1_name, SUCCESS, LOCKTYPE_RWSEM);
+	pr_cont("\n");
 }
 
 static void hardirq_deadlock_softirq_not_deadlock(void)
Signature · 1 line
2.45.2
#2 lore raw

Re: [PATCH 2/2] ext4: Testing lock class and subclass got the same name pointer

On Sun, Jul 14, 2024 at 08:14:27AM +0300, botta633 wrote:

First, the subsystem tag also needs to be "locking/lockdep" or "lockdep".

Quoted 3 lines

Checking if the lockdep_map->name will change when setting the subclass. It shouldn't change so that the lock class and subclass will have the same name

Could you make the commit log wrapped at 75 columns?

Quoted 4 lines
Fixes fd5e3f5fe27

Since this is only adding test for a bug fix, you don't need to put these tags here.

Quoted 1 line

Again, could you please put your name here?

Also seems that you send two patch #2, one is with the proper version number "v2", but not in-reply-to the patch #1, the other doesn't have the correct version number but has the correct "in-reply-to" field. Could you use the correct version number and "in-reply-to" next time?

Quoted 10 lines
---
 lib/locking-selftest.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
lib/locking-selftest.c +0 -0
index 6f6a5fc85b42..1d7885205f36 100644
@@ -2710,12 +2710,24 @@ static void local_lock_3B(void)
 

lib/locking-selftest.c static void local_lock_3B(void) · line 2720

Quoted 10 lines
lib/locking-selftest.c +8 -0
 }
 
+static void class_subclass_X1_name(void)
+{
+	const char *name_before_subclass = rwsem_X1.dep_map.name;
+	const char *name_after_subclass;
+
+	WARN_ON(!rwsem_X1.dep_map.name);
+	lockdep_set_subclass(&rwsem_X1, 1);
+	WARN_ON(name_before_subclass != name_after_subclass);

Could you add some comments here explaining your test? For example, where name_after_subclass gets set?

lib/locking-selftest.c static void local_lock_3B(void) · line 2724

Quoted 4 lines
lib/locking-selftest.c +2 -0
+}
+
 static void local_lock_tests(void)
 {

Please don't add this test into another test, you could directly call your class_subclass_X1_name() (maybe rename it to *_test()) in lockding_selftest() function.

Besides, make sure you run the test with and without your modification in patch #1, and confirm this is the test that could verify your fix.

Regards,
Boqun

Quoted 23 lines
lib/locking-selftest.c +6 -0
 	printk("  --------------------------------------------------------------------------\n");
 	printk("  | local_lock tests |\n");
 	printk("  ---------------------\n");
 
+	init_class_X(&lock_X1, &rwlock_X1, &mutex_X1, &rwsem_X1);
+
 	print_testname("local_lock inversion  2");
 	dotest(local_lock_2, SUCCESS, LOCKTYPE_LL);
 	pr_cont("\n");
@@ -2727,6 +2739,10 @@ static void local_lock_tests(void)
 	print_testname("local_lock inversion 3B");
 	dotest(local_lock_3B, FAILURE, LOCKTYPE_LL);
 	pr_cont("\n");
+
+	print_testname("Class and subclass");
+	dotest(class_subclass_X1_name, SUCCESS, LOCKTYPE_RWSEM);
+	pr_cont("\n");
 }
 
 static void hardirq_deadlock_softirq_not_deadlock(void)
Signature · 1 line
2.45.2
reviewer
#3 lore raw
In reply to botta633
Quoted 20 lines · On 7/14/24 01:14, botta633 wrote:

From: Ahmed Ehab <[email protected]>

Preventing lockdep_set_subclass from creating a new instance of the string literal. Hence, we will always have the same class->name among parent and subclasses. This prevents kernel panics when looking up a lock class while comparing class locks and class names.

Fixes fd5e3f5fe27
---
  include/linux/lockdep.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/lockdep.h +0 -0
index 08b0d1d9d78b..df8fa5929de7 100644

include/linux/lockdep.h static inline void lockdep_init_map(struct lockdep_map *lock, const char *name, · line 179

Quoted 8 lines
include/linux/lockdep.h +1 -1
@@ -173,7 +173,7 @@ static inline void lockdep_init_map(struct lockdep_map *lock, const char *name,
  			      (lock)->dep_map.lock_type)
 
  #define lockdep_set_subclass(lock, sub)					\
-	lockdep_init_map_type(&(lock)->dep_map, #lock, (lock)->dep_map.key, sub,\
+	lockdep_init_map_type(&(lock)->dep_map, (lock)->dep_map.name, (lock)->dep_map.key, sub,\
  			      (lock)->dep_map.wait_type_inner,		\
  			      (lock)->dep_map.wait_type_outer,		\
  			      (lock)->dep_map.lock_type)

ext4 is a filesystem. It has nothing to do with locking/lockdep. Could you resend the patches with the proper prefix of "lockdep:" or "locking/lockdep:"?

Thanks,
Longman

A message that is not in the archive (sent off-list or not yet received): CA+6bSasRZ7HRURZcSPEsAyDtNDdx+7UGwuXRG+Dw0Gqo+vs9Ew@mail.gmail.com

#5 lore raw
Quoted 6 lines · ahmed Ehab

Ok, I will.
I just put ext4 because the syzkaller bug was mentioned in the ext4 subsystem.
Thanks,
Ahmed

Please avoid top-posting. And

Quoted 5 lines · Waiman Long
Quoted 4 lines · On 7/14/24 01:14, botta633 wrote:
Quoted 2 lines

From: Ahmed Ehab <[email protected]>

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Preventing lockdep_set_subclass from creating a new instance of the string literal. Hence, we will always have the same class->name among parent and subclasses. This prevents kernel panics when looking up a lock class while comparing class locks and class names.

Fixes fd5e3f5fe27

please add the title of the commit here as well, e.g.

Fixes <sha1> ("<title>")

see

	https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=f7ce5eb2cb7993e4417642ac28713a063123461f	

for example.

Regards,
Boqun

Quoted 32 lines
Quoted 32 lines

include/linux/lockdep.h static inline void lockdep_init_map(struct

Quoted 11 lines
---
  include/linux/lockdep.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/lockdep.h +0 -0
index 08b0d1d9d78b..df8fa5929de7 100644
@@ -173,7 +173,7 @@ static inline void lockdep_init_map(struct

lockdep_map *lock, const char *name,

include/linux/lockdep.h static inline void lockdep_init_map(struct lockdep_map *lock, const char *name,

Quoted 3 lines
include/linux/lockdep.h +0 -0
                            (lock)->dep_map.lock_type)
 
  #define lockdep_set_subclass(lock, sub)
     \

include/linux/lockdep.h static inline void lockdep_init_map(struct · line 176

Quoted 1 line
include/linux/lockdep.h +0 -1
-     lockdep_init_map_type(&(lock)->dep_map, #lock,

(lock)->dep_map.key, sub,\

include/linux/lockdep.h static inline void lockdep_init_map(struct · line 176

Quoted 1 line
include/linux/lockdep.h +1 -0
+     lockdep_init_map_type(&(lock)->dep_map, (lock)->dep_map.name,

(lock)->dep_map.key, sub,\

include/linux/lockdep.h static inline void lockdep_init_map(struct lockdep_map *lock, const char *name, · line 179

Quoted 3 lines
include/linux/lockdep.h +0 -0
                            (lock)->dep_map.wait_type_inner,          \
                            (lock)->dep_map.wait_type_outer,          \
                            (lock)->dep_map.lock_type)

ext4 is a filesystem. It has nothing to do with locking/lockdep. Could you resend the patches with the proper prefix of "lockdep:" or "locking/lockdep:"?

Thanks,
Longman