Skip to contentSkip to conversation

kernel/audit_tree.c: tree will memory leak when failure occurs for audit_trim_trees()

PATCH In review Chen Gang 2 participants 10 messages Apr 19, 2013 – May 13, 2013 Audit

lore
Share & tools
lore

Chen Gang

author
#0 lore raw

PATCH 1/1 kernel/audit_tree.c: tree will memory leak when failure occurs for audit_trim_trees()

in audit_trim_trees(), has called get_tree() before failure occurs, so need also call put_tree after go to skip_it:

---
 kernel/audit_tree.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
kernel/audit_tree.c +1 -1
index 642a89c..de46ec0 100644
@@ -617,10 +617,10 @@ void audit_trim_trees(void)
 		}
 		spin_unlock(&hash_lock);
 		trim_marked(tree);
-		put_tree(tree);
 		drop_collected_mounts(root_mnt);
 skip_it:
 		mutex_lock(&audit_filter_mutex);
+		put_tree(tree);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);
Signature · 1 line
1.7.7.6
#1 lore raw
Quoted 10 lines · On Fri, 19 Apr 2013 17:39:06 +0800 Chen Gang <[email protected]> wrote:

in audit_trim_trees(), has called get_tree() before failure occurs, so need also call put_tree after go to skip_it:

...

kernel/audit_tree.c +0 -0
@@ -617,10 +617,10 @@ void audit_trim_trees(void)
 		}

kernel/audit_tree.c void audit_trim_trees(void) · line 626

Quoted 10 lines
kernel/audit_tree.c +1 -1
 		spin_unlock(&hash_lock);
 		trim_marked(tree);
-		put_tree(tree);
 		drop_collected_mounts(root_mnt);
 skip_it:
 		mutex_lock(&audit_filter_mutex);
+		put_tree(tree);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);

That looks right to me.

I think we can micro-optimise the code by performing the put_tree() before taking the mutex, to slightly reduce mutex hold times?

kernel/audit_tree.c renamed from kernel/audit_tree.c~kernel-audit_treec-tree-will-leak-memory-when-failure-occurs-in-audit_trim_trees-fix +1 -1
@@ -619,8 +619,8 @@ void audit_trim_trees(void)
 		trim_marked(tree);
 		drop_collected_mounts(root_mnt);
 skip_it:
-		mutex_lock(&audit_filter_mutex);
 		put_tree(tree);
+		mutex_lock(&audit_filter_mutex);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);

_

author
#2 lore raw
Quoted 25 lines · Andrew Morton
Quoted 10 lines · On Fri, 19 Apr 2013 17:39:06 +0800 Chen Gang <[email protected]> wrote:

in audit_trim_trees(), has called get_tree() before failure occurs, so need also call put_tree after go to skip_it:

...

kernel/audit_tree.c +0 -0
@@ -617,10 +617,10 @@ void audit_trim_trees(void)
 		}

kernel/audit_tree.c void audit_trim_trees(void) · line 626

Quoted 10 lines
kernel/audit_tree.c +1 -1
 		spin_unlock(&hash_lock);
 		trim_marked(tree);
-		put_tree(tree);
 		drop_collected_mounts(root_mnt);
 skip_it:
 		mutex_lock(&audit_filter_mutex);
+		put_tree(tree);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);

That looks right to me.

Quoted 3 lines

I think we can micro-optimise the code by performing the put_tree() before taking the mutex, to slightly reduce mutex hold times?

ok, thanks.

Quoted 17 lines
kernel/audit_tree.c renamed from kernel/audit_tree.c~kernel-audit_treec-tree-will-leak-memory-when-failure-occurs-in-audit_trim_trees-fix +1 -1
@@ -619,8 +619,8 @@ void audit_trim_trees(void)
 		trim_marked(tree);
 		drop_collected_mounts(root_mnt);
 skip_it:
-		mutex_lock(&audit_filter_mutex);
 		put_tree(tree);
+		mutex_lock(&audit_filter_mutex);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);

_

Signature · 3 lines
Chen Gang

Asianux Corporation
author
#3 lore raw

PATCH 1/1 kernel/audit_tree.c: fix the original version merging issue for put_tree().

In reply to Andrew Morton

During version merging, added another redundant put_tree().

The related original patch: 1b4f5c2 kernel/audit_tree.c: tree will leak memory when failure occurs in audit_trim_trees().

So need delete it, or it will cause issue.

---
 kernel/audit_tree.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)
kernel/audit_tree.c +0 -1
index 9fd17f0..a291aa2 100644
@@ -621,7 +621,6 @@ void audit_trim_trees(void)
 skip_it:
 		put_tree(tree);
 		mutex_lock(&audit_filter_mutex);
-		put_tree(tree);
 	}
 	list_del(&cursor);
 	mutex_unlock(&audit_filter_mutex);
Signature · 1 line
1.7.7.6
author
#4 lore raw

PATCH v2 1/1 kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

In reply to Andrew Morton
Quoted 4 lines · Chen Gang
Quoted 3 lines · Chen Gang F T
Quoted 3 lines · Andrew Morton
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines · On Wed, 17 Apr 2013 12:04:02 +0800 Chen Gang <[email protected]> wrote:
Quoted 9 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

since "normally audit_add_tree_rule() will free it on failure",
need free it completely, when failure occures.

need additional put_tree before return, since get_tree was called.
always need goto error processing area for list_del_init.

Isn't that get_tree() in audit_add_tree_rule() simply unneeded? In other words, is this patch correct:

Hell all:

get_tree() in audit_add_tree_rule() is needed (my original conclusion is incorrect).

the reason is when failed, trim_marked() will call put_tree() to free the tree and the related entry.

    trim_marked() ->
      ...
      tree->goner = 1
      ...
      kill_rules()
      ...
      prune_one() ->
        put_tree()
      ...

after trim_marked() returns, should not call put_tree() again.
but after trim_marked() returns, it has to 'goto Err' to call additional put_tree(). so it has to call additional get_tree() firstly.

And get_tree() need not be in audit_filter_mutex lock protected: it is only for self task to call trim_marked().

All get_tree() and put_tree() in audit_add_tree_rule() are matched with each other.

But we need let 'rule->tree = NULL;' firstly, so can protect rule itself freed in kill_rules().

the reason is when it is killed, the 'rule' itself may have already released, we should not access it. one example: we add a rule to an inode, just at the same time the other task is deleting this inode.

the work flow for add a rule:

    audit_receive() -> (need audit_cmd_mutex lock)
      audit_receive_skb() ->
        audit_receive_msg() ->
          audit_receive_filter() ->
            audit_add_rule() ->
              audit_add_tree_rule() -> (need audit_filter_mutex lock)
                ...
                unlock audit_filter_mutex
                get_tree()
                ...
                iterate_mounts() -> (iterate all related inodes)
                  tag_mount() ->
                    tag_trunk() ->
                      create_trunk() -> (assume it is 1st rule)
                        fsnotify_add_mark() -> 
                          fsnotify_add_inode_mark() ->  (add mark to inode->i_fsnotify_marks)
                        ...
                        get_tree(); (each inode will get one)
                ...
                lock audit_filter_mutex

the work flow for delete inode:

    __destroy_inode() ->
     fsnotify_inode_delete() ->
       __fsnotify_inode_delete() ->
        fsnotify_clear_marks_by_inode() ->  (get mark from inode->i_fsnotify_marks)
          fsnotify_destroy_mark() ->
           fsnotify_destroy_mark_locked() ->
             audit_tree_freeing_mark() ->
               evict_chunk() ->
                 ...
                 tree->goner = 1
                 ...
                 kill_rules() ->   (assume current->audit_context == NULL)
                   call_rcu() ->   (rule->tree != NULL)
                     audit_free_rule_rcu() ->
                       audit_free_rule()
                 ...
                 audit_schedule_prune() ->  (assume current->audit_context == NULL)
                   kthread_run() ->    (need audit_cmd_mutex and audit_filter_mutex lock)
                     prune_one() ->    (delete it from prue_list)
                       put_tree(); (match the original get_tree above)

-----------------------------diff begin---------------------------------

kernel/audit_tree.c +1 -0
index 9fd17f0..85eb26c 100644
@@ -659,6 +659,7 @@ int audit_add_tree_rule(struct audit_krule *rule)
 	struct vfsmount *mnt;
 	int err;
 
+	rule->tree = NULL;
 	list_for_each_entry(tree, &tree_list, list) {
 		if (!strcmp(seed->pathname, tree->pathname)) {
 			put_tree(seed);

-----------------------------diff end-----------------------------------

For me, after 'rule->tree = NULL', all things seems fine !!

Please help check.

Thanks.

Quoted 142 lines

after reading code again, I think:

we can remove the pair of get_tree() and put_tree(),

a. the author want to pretect tree not to be freed after unlock audit_filter_mutex
when tree is really freed by others, we have chance to check tree->goner

b. it just like another functions done (audit_tag_tree, audit_trim_trees)
for audit_add_tree_rule, also need let get_tree() protected by audit_filter_mutex.
(move 'get_tree' before unlock audit_filter_mutex)

c. but after read code (at least for audit_add_tree_rule)
during unlock audit_filter_mutex in audit_add_tree_rule:
I find only one posible related work flow which may happen (maybe it can not happen either).
fsnotify_destroy_mark_locked ->
audit_tree_freeing_mark ->
evict_chunk ->
kill_rules ->
call_rcu ->
audit_free_rule_rcu ->
audit_free_rule
it will free rules and entry, but it does not call put_tree().

so I think, at least for audit_add_tree_rule, we can remove the pair of get_tree() and put_tree()
(maybe also need give a check for audit_tag_tree and audit_trim_trees, whether can remove 'get_tree' too)

and the process is incorrect after lock audit_filter_mutex again (line 701..705)

    a. if rule->rlist was really empty
        the 'rule' itself would already be freed.
        the caller and the caller of caller, need notice this (not double free)
        instead, we need check tree->gonar (and also need spin_lock protected).

    b. we do not need set "rule->tree = tree" again.
         i. if 'rule' is not touched by any other thread
              it should be rule->tree == tree.

        ii. else if rule->tree == NULL (freed by other thread)
              'rule' itself might be freed too, we'd better return by failure.

       iii. else (!rule->tree && rule->tree != tree) (reused by other thread)
              firstly, it should not happen.
              if could happen, 'rule->tree = tree' would cause original rule->tree memory leak.

my conclusion is only by reading code (and still I am not quite expert about it, either) so welcome experts (especially maintainers) to providing suggestions or completions.

thanks.

if no reply within a week (2013-04-28), I should send related patch.

gchen.

653 /* called with audit_filter_mutex */
654 int audit_add_tree_rule(struct audit_krule *rule) 655 {

656         struct audit_tree *seed = rule->tree, *tree;
657         struct path path;
658         struct vfsmount *mnt;
659         int err;
660 
661         list_for_each_entry(tree, &tree_list, list) {
662                 if (!strcmp(seed->pathname, tree->pathname)) {
663                         put_tree(seed);
664                         rule->tree = tree;
665                         list_add(&rule->rlist, &tree->rules);
666                         return 0;
667                 }
668         }
669         tree = seed;
670         list_add(&tree->list, &tree_list);
671         list_add(&rule->rlist, &tree->rules);
672         /* do not set rule->tree yet */
673         mutex_unlock(&audit_filter_mutex);

674
675 err = kern_path(tree->pathname, 0, &path);
676 if (err)
677 goto Err;

678         mnt = collect_mounts(&path);
679         path_put(&path);
680         if (IS_ERR(mnt)) {

681 err = PTR_ERR(mnt);
682 goto Err;
683 }
684

685         get_tree(tree);
686         err = iterate_mounts(tag_mount, tree, mnt);
687         drop_collected_mounts(mnt);
688 
689         if (!err) {
690                 struct node *node;
691                 spin_lock(&hash_lock);
692                 list_for_each_entry(node, &tree->chunks, list)

693 node->index &= ~(1U<<31);
694 spin_unlock(&hash_lock);
695 } else {
696 trim_marked(tree);
697 goto Err;
698 }
699
700 mutex_lock(&audit_filter_mutex);
701 if (list_empty(&rule->rlist)) {
702 put_tree(tree);
703 return -ENOENT;

704         }
705         rule->tree = tree;
706         put_tree(tree);
707 
708         return 0;
709 Err:
710         mutex_lock(&audit_filter_mutex);
711         list_del_init(&tree->list);
712         list_del_init(&tree->rules);
713         put_tree(tree);
714         return err;

715 }

Quoted 8 lines
Quoted 8 lines

excuse me:
I am not quite familiar with it, and also have to do another things. so I have to spend additional time resource to make sure about it.

is it ok ?
I should make sure about it within this week (2013-04-21) I should finish related test (if need), within next week (2013-4-28)

Quoted 8 lines
Quoted 8 lines

if have additional suggestions or completions, please reply.
(if no reply, I will follow the time point above)

thanks.

gchen.

-- Chen Gang Asianux Corporation

Signature · 3 lines
Chen Gang

Asianux Corporation
#5 lore raw

Re: [PATCH v2] kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

Quoted 1 line · On Thu, 09 May 2013 20:53:06 +0800 Chen Gang <[email protected]> wrote:

get_tree() in audit_add_tree_rule() is needed (my original conclusion is incorrect).

OK..

Quoted 1 line

But we need let 'rule->tree = NULL;' firstly, so can protect rule itself freed in kill_rules().

I'll believe you ;) I turned this into a proper patch and added your (missed) Signed-off-by:.

Quoted 1 line

For me, after 'rule->tree = NULL', all things seems fine !!

Well, what was wrong before? Is there some user-triggerable misbehaviour which you observed? If so, please describe it.

author
#6 lore raw

Re: [PATCH v2] kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

Quoted 4 lines · Andrew Morton
Quoted 1 line
Quoted 1 line

But we need let 'rule->tree = NULL;' firstly, so can protect rule itself freed in kill_rules().

I'll believe you ;) I turned this into a proper patch and added your (missed) Signed-off-by:.

Thanks.

At least, let 'rule->tree = NULL;' can:

a. it matches 'rule->tree = tree;' which is before successful return.

       also can make 'if (list_empty(&rule->rlist))' reasonable.

b. protect rule itself freed in kill_rules(), if it could happen.

       just like all 'rule->tree = NULL;' in audit_remove_tree_rule().

c. it will no negative effect.

Quoted 6 lines
Quoted 1 line
Quoted 1 line

For me, after 'rule->tree = NULL', all things seems fine !!

Well, what was wrong before? Is there some user-triggerable misbehaviour which you observed? If so, please describe it.

I think, it will cause issue (randomly): if when we are using auditctl to add rule to monitor one file, and at the same time, the other user is just deleting this file.

I guess, it is why original code need 'if (list_empty(&rule->rlist))' after lock 'audit_filter_mutex' again.

Currently, I am just testing for it (and should give a test), and I will send the test plan and test result within this week (2013-05-12).

Thanks.

Signature · 3 lines
Chen Gang

Asianux Corporation
author
#7 lore raw

Re: [PATCH v2] kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

Quoted 7 lines · Chen Gang
Quoted 6 lines · Andrew Morton
Quoted 6 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

For me, after 'rule->tree = NULL', all things seems fine !!

Well, what was wrong before? Is there some user-triggerable misbehaviour which you observed? If so, please describe it.

Oh, sorry, after have a test, the original code is no issue (it is my fault).

When the deleting work flow call evict_chunk(), I assume that the 'postponed' can be NULL (at least, in some condition, it can), so kill_rules() can be called directly. But in fact, 'postponed' will never be NULL:

audit_tree depend on CONFIG_AUDIT_TREE which depend on CONFIG_AUDITSYSCALL. if CONFIG_AUDITSYSCALL defined.
do_fork() -> copy_process() -> audit_alloc() -> alloc 'audit_context'. so the audit_killed_tree() will return valid pointer to 'postponed'.

although already have quite a few code for 'postponed == NULL', they are really useless now.

I also read all other work flow which related with kill_rule(), I can not find any of them can lead audit_add_tree_rule() to cause issue: all work flow related with kill_rule() are protected by audit_cmd_mutex now.

Test plan:
code preparation:
define a flag varaible.
wait the flag to be true, before lock 'audit_filter_mutex' again. in audit_add_tree_rule() when evict_trunc() finish, set the flag true.
firstly start: 'rm -rvf /tmp/gchen/linux-next'
then start: 'audit -w /tmp/gchen/linux-next/drivers/char'
(notice the order should not be changed, or all system call will be locked)

Test result:
the evict_chunk() will not call kill_rule() directly, so no issues. the output sample result like this: ('printk' the related information)

---------------------------sample begin-----------------------------

[   85.455891] gchen_tag: ffff880099f0ddc0, audit_init_entry(): create entry: ffff880097ca2800
[   85.455900] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): before call, type: 1011
[   85.455903] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): enter function
[   85.455927] ida_remove called for id=0 which is not allocated.
[   85.455935] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): begin waiting 100...., rule: ffff880097ca2820
[   91.425947] gchen_tag: ffff880097995dc0, fsnotify_clear_marks_by_inode(): set audit_test_count = true
[   91.425960] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): end waiting, rule: ffff880097ca2820
[   91.426055] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): after call for succeed, type: 1011
[   91.426558] gchen_tag: ffff880097995dc0, kill_rules(): list_del_init, rule: ffff880097ca2820, tree: ffff880099dfff00
[   91.426564] gchen_tag: ffff880097995dc0, kill_rules(): remove entry: ffff880097ca2800
[   91.431023] gchen_tag: ffff880097995dc0, audit_free_rule(): remove entry: ffff880097ca2800

---------------------------sample end-------------------------------

Now, my original fix makes the related code consistent, but the related code maybe be useless now (if what I said is true, in audit, quite a few of code are useless for this reason).

I can not be sure whether these useless code will be used, in the future (whether let AUDIT_TREE and AUDIT_WATCH independent on AUDIT_SYSCALL in the future).

If it will be used in the future, my fix is useful too, else we'd better to delete the related useless code.

Thanks.

Quoted 15 lines

I think, it will cause issue (randomly): if when we are using auditctl to add rule to monitor one file, and at the same time, the other user is just deleting this file.

I guess, it is why original code need 'if (list_empty(&rule->rlist))' after lock 'audit_filter_mutex' again.

Currently, I am just testing for it (and should give a test), and I will send the test plan and test result within this week (2013-05-12).

Thanks.

-- Chen Gang Asianux Corporation

Signature · 3 lines
Chen Gang

Asianux Corporation
author
#8 lore raw

Re: [PATCH v2] kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

Quoted 9 lines · Chen Gang
Quoted 7 lines · Chen Gang
Quoted 6 lines · Andrew Morton
Quoted 6 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

For me, after 'rule->tree = NULL', all things seems fine !!

Well, what was wrong before? Is there some user-triggerable misbehaviour which you observed? If so, please describe it.

Oh, sorry again, the 'postponed' in evict_chunk() still has a chance to be NULL: firstly, 'audit_context->in_syscall' also checked in audit_killed_trees(), and also not all tasks are generated by do_fork().

But really, for most cases, the 'postponed' is not NULL, so my test case can not cause issue.

Currently, I just force 'postponed' to be NULL to see the test result... :-)

It seems my original fix is still useful ! ;-)

Thanks.

Quoted 80 lines

Oh, sorry, after have a test, the original code is no issue (it is my fault).

When the deleting work flow call evict_chunk(), I assume that the 'postponed' can be NULL (at least, in some condition, it can), so kill_rules() can be called directly. But in fact, 'postponed' will never be NULL:

audit_tree depend on CONFIG_AUDIT_TREE which depend on CONFIG_AUDITSYSCALL. if CONFIG_AUDITSYSCALL defined.
do_fork() -> copy_process() -> audit_alloc() -> alloc 'audit_context'. so the audit_killed_tree() will return valid pointer to 'postponed'.

although already have quite a few code for 'postponed == NULL', they are really useless now.

I also read all other work flow which related with kill_rule(), I can not find any of them can lead audit_add_tree_rule() to cause issue: all work flow related with kill_rule() are protected by audit_cmd_mutex now.

Test plan:
code preparation:
define a flag varaible.
wait the flag to be true, before lock 'audit_filter_mutex' again. in audit_add_tree_rule() when evict_trunc() finish, set the flag true.
firstly start: 'rm -rvf /tmp/gchen/linux-next'
then start: 'audit -w /tmp/gchen/linux-next/drivers/char'
(notice the order should not be changed, or all system call will be locked)

Test result:
the evict_chunk() will not call kill_rule() directly, so no issues. the output sample result like this: ('printk' the related information)

---------------------------sample begin-----------------------------

[   85.455891] gchen_tag: ffff880099f0ddc0, audit_init_entry(): create entry: ffff880097ca2800
[   85.455900] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): before call, type: 1011
[   85.455903] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): enter function
[   85.455927] ida_remove called for id=0 which is not allocated.
[   85.455935] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): begin waiting 100...., rule: ffff880097ca2820
[   91.425947] gchen_tag: ffff880097995dc0, fsnotify_clear_marks_by_inode(): set audit_test_count = true
[   91.425960] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): end waiting, rule: ffff880097ca2820
[   91.426055] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): after call for succeed, type: 1011
[   91.426558] gchen_tag: ffff880097995dc0, kill_rules(): list_del_init, rule: ffff880097ca2820, tree: ffff880099dfff00
[   91.426564] gchen_tag: ffff880097995dc0, kill_rules(): remove entry: ffff880097ca2800
[   91.431023] gchen_tag: ffff880097995dc0, audit_free_rule(): remove entry: ffff880097ca2800

---------------------------sample end-------------------------------

Now, my original fix makes the related code consistent, but the related code maybe be useless now (if what I said is true, in audit, quite a few of code are useless for this reason).

I can not be sure whether these useless code will be used, in the future (whether let AUDIT_TREE and AUDIT_WATCH independent on AUDIT_SYSCALL in the future).

If it will be used in the future, my fix is useful too, else we'd better to delete the related useless code.

Thanks.

Quoted 15 lines

I think, it will cause issue (randomly): if when we are using auditctl to add rule to monitor one file, and at the same time, the other user is just deleting this file.

I guess, it is why original code need 'if (list_empty(&rule->rlist))' after lock 'audit_filter_mutex' again.

Currently, I am just testing for it (and should give a test), and I will send the test plan and test result within this week (2013-05-12).

Thanks.

-- Chen Gang Asianux Corporation

Signature · 3 lines
Chen Gang

Asianux Corporation
author
#9 lore raw

Re: [PATCH v2] kernel: audit_tree: resource management: need put_tree and goto Err when failure occures

Quoted 11 lines · Chen Gang
Quoted 9 lines · Chen Gang
Quoted 7 lines · Chen Gang
Quoted 6 lines · Andrew Morton
Quoted 6 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

For me, after 'rule->tree = NULL', all things seems fine !!

Well, what was wrong before? Is there some user-triggerable misbehaviour which you observed? If so, please describe it.

If we force 'postponed' to be NUL, and still use the original test plan:

Quoted 8 lines
Quoted 8 lines

Test plan:
code preparation:
define a flag varaible.
wait the flag to be true, before lock 'audit_filter_mutex' again. in audit_add_tree_rule() when evict_trunc() finish, set the flag true.
firstly start: 'rm -rvf /tmp/gchen/linux-next'
then start: 'audit -w /tmp/gchen/linux-next/drivers/char'
(notice the order should not be changed, or all system call will be locked)

If not set 'rule->tree = NULL', it will cause issue: kernel will die !!.

But if set 'rule->tree = NULL', all things seems OK, and the output log is just what we expected.

---------------------------test result begin--------------------------------

task_struct ptr: function(): action: related value:

[  627.422698] gchen_tag: ffff88009b8787a0, audit_init_entry(): create entry: ffff88008bc75600
[  627.422712] gchen_tag: ffff88009b8787a0, audit_receive_filter(): before call, type: 1011
[  627.422718] gchen_tag: ffff88009b8787a0, audit_add_tree_rule(): enter function
[  627.422822] ida_remove called for id=0 which is not allocated.
[  627.422834] gchen_tag: ffff88009b8787a0, audit_add_tree_rule(): begin waiting 100...., rule: ffff88008bc75620
[  639.421114] gchen_tag: ffff88009b878000, evict_chunk(): enter function postponed:           (null)
[  639.421197] gchen_tag: ffff88009b878000, evict_chunk(): kill_rull postponed:           (null)
[  639.421209] gchen_tag: ffff88009b878000, kill_rules(): list_del_init, rule: ffff88008bc75620, tree:           (null)
[  639.421213] gchen_tag: ffff88009b878000, evict_chunk(): audit_schedule_prune postponed:           (null)
[  639.421274] gchen_tag: ffff88009b878000, evict_chunk(): set audit_test_count = true, postponed:           (null)
[  639.421282] gchen_tag: ffff88009b8787a0, audit_add_tree_rule(): end waiting, rule: ffff88008bc75620
[  639.421289] gchen_tag: ffff88009b8787a0, audit_add_tree_rule(): list empty for rule->rlist and return fail: ffff88008bc75620
[  639.421364] gchen_tag: ffff88009b8787a0, audit_receive_filter(): after call for failure, type: 1011
[  639.421369] gchen_tag: ffff88009b8787a0, audit_free_rule(): remove entry: ffff88008bc75600

---------------------------test result end----------------------------------

Next, I will try to find a test case to let 'postponed' to NUL by itself (not with hard coded by force), then test it again :-).

Thanks.

Quoted 95 lines

Oh, sorry again, the 'postponed' in evict_chunk() still has a chance to be NULL: firstly, 'audit_context->in_syscall' also checked in audit_killed_trees(), and also not all tasks are generated by do_fork().

But really, for most cases, the 'postponed' is not NULL, so my test case can not cause issue.

Currently, I just force 'postponed' to be NULL to see the test result... :-)

It seems my original fix is still useful ! ;-)

Thanks.

Quoted 80 lines

Oh, sorry, after have a test, the original code is no issue (it is my fault).

When the deleting work flow call evict_chunk(), I assume that the 'postponed' can be NULL (at least, in some condition, it can), so kill_rules() can be called directly. But in fact, 'postponed' will never be NULL:

audit_tree depend on CONFIG_AUDIT_TREE which depend on CONFIG_AUDITSYSCALL. if CONFIG_AUDITSYSCALL defined.
do_fork() -> copy_process() -> audit_alloc() -> alloc 'audit_context'. so the audit_killed_tree() will return valid pointer to 'postponed'.

although already have quite a few code for 'postponed == NULL', they are really useless now.

I also read all other work flow which related with kill_rule(), I can not find any of them can lead audit_add_tree_rule() to cause issue: all work flow related with kill_rule() are protected by audit_cmd_mutex now.

Test plan:
code preparation:
define a flag varaible.
wait the flag to be true, before lock 'audit_filter_mutex' again. in audit_add_tree_rule() when evict_trunc() finish, set the flag true.
firstly start: 'rm -rvf /tmp/gchen/linux-next'
then start: 'audit -w /tmp/gchen/linux-next/drivers/char'
(notice the order should not be changed, or all system call will be locked)

Test result:
the evict_chunk() will not call kill_rule() directly, so no issues. the output sample result like this: ('printk' the related information)

---------------------------sample begin-----------------------------

[   85.455891] gchen_tag: ffff880099f0ddc0, audit_init_entry(): create entry: ffff880097ca2800
[   85.455900] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): before call, type: 1011
[   85.455903] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): enter function
[   85.455927] ida_remove called for id=0 which is not allocated.
[   85.455935] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): begin waiting 100...., rule: ffff880097ca2820
[   91.425947] gchen_tag: ffff880097995dc0, fsnotify_clear_marks_by_inode(): set audit_test_count = true
[   91.425960] gchen_tag: ffff880099f0ddc0, audit_add_tree_rule(): end waiting, rule: ffff880097ca2820
[   91.426055] gchen_tag: ffff880099f0ddc0, audit_receive_filter(): after call for succeed, type: 1011
[   91.426558] gchen_tag: ffff880097995dc0, kill_rules(): list_del_init, rule: ffff880097ca2820, tree: ffff880099dfff00
[   91.426564] gchen_tag: ffff880097995dc0, kill_rules(): remove entry: ffff880097ca2800
[   91.431023] gchen_tag: ffff880097995dc0, audit_free_rule(): remove entry: ffff880097ca2800

---------------------------sample end-------------------------------

Now, my original fix makes the related code consistent, but the related code maybe be useless now (if what I said is true, in audit, quite a few of code are useless for this reason).

I can not be sure whether these useless code will be used, in the future (whether let AUDIT_TREE and AUDIT_WATCH independent on AUDIT_SYSCALL in the future).

If it will be used in the future, my fix is useful too, else we'd better to delete the related useless code.

Thanks.

Quoted 15 lines

I think, it will cause issue (randomly): if when we are using auditctl to add rule to monitor one file, and at the same time, the other user is just deleting this file.

I guess, it is why original code need 'if (list_empty(&rule->rlist))' after lock 'audit_filter_mutex' again.

Currently, I am just testing for it (and should give a test), and I will send the test plan and test result within this week (2013-05-12).

Thanks.

-- Chen Gang Asianux Corporation

Signature · 3 lines
Chen Gang

Asianux Corporation