[PATCH -next 0/3] fix possible memleak in '__blk_trace_remove'

[PATCH -next 0/3] fix possible memleak in '__blk_trace_remove'

[Ye Bin]

Ye Bin (3):
  blktrace: introduce 'blk_trace_swicth_state' helper
  blktrace: fix possible memleak in '__blk_trace_remove'
  blktrace: remove unnessary stop block trace in 'blk_trace_shutdown'

 kernel/trace/blktrace.c | 84 ++++++++++++++++++++---------------------
 1 file changed, 41 insertions(+), 43 deletions(-)

-- 
2.31.1

[PATCH -next 1/3] blktrace: introduce 'blk_trace_swicth_state' helper

[Ye Bin]

Introduce 'blk_trace_swicth_state' helper. No functional changed.

Signed-off-by: Ye Bin <yebin10@huawei.com>
---
 kernel/trace/blktrace.c | 74 ++++++++++++++++++++---------------------
 1 file changed, 36 insertions(+), 38 deletions(-)

diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index 7f5eb295fe19..edd83e213580 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -353,6 +353,39 @@ static void blk_trace_cleanup(struct request_queue *q, struct blk_trace *bt)
 	put_probe_ref();
 }
 
+static int blk_trace_switch_state(struct blk_trace *bt, int start)
+{
+	/*
+	 * For starting a trace, we can transition from a setup or stopped
+	 * trace. For stopping a trace, the state must be running
+	 */
+	if (start) {
+		if (bt->trace_state == Blktrace_setup ||
+		    bt->trace_state == Blktrace_stopped) {
+			blktrace_seq++;
+			smp_mb();
+			bt->trace_state = Blktrace_running;
+			raw_spin_lock_irq(&running_trace_lock);
+			list_add(&bt->running_list, &running_trace_list);
+			raw_spin_unlock_irq(&running_trace_lock);
+
+			trace_note_time(bt);
+			return 0;
+		}
+	} else {
+		if (bt->trace_state == Blktrace_running) {
+			bt->trace_state = Blktrace_stopped;
+			raw_spin_lock_irq(&running_trace_lock);
+			list_del_init(&bt->running_list);
+			raw_spin_unlock_irq(&running_trace_lock);
+			relay_flush(bt->rchan);
+			return 0;
+		}
+	}
+
+	return -EINVAL;
+}
+
 static int __blk_trace_remove(struct request_queue *q)
 {
 	struct blk_trace *bt;
@@ -658,7 +691,6 @@ static int compat_blk_trace_setup(struct request_queue *q, char *name,
 
 static int __blk_trace_startstop(struct request_queue *q, int start)
 {
-	int ret;
 	struct blk_trace *bt;
 
 	bt = rcu_dereference_protected(q->blk_trace,
@@ -666,36 +698,7 @@ static int __blk_trace_startstop(struct request_queue *q, int start)
 	if (bt == NULL)
 		return -EINVAL;
 
-	/*
-	 * For starting a trace, we can transition from a setup or stopped
-	 * trace. For stopping a trace, the state must be running
-	 */
-	ret = -EINVAL;
-	if (start) {
-		if (bt->trace_state == Blktrace_setup ||
-		    bt->trace_state == Blktrace_stopped) {
-			blktrace_seq++;
-			smp_mb();
-			bt->trace_state = Blktrace_running;
-			raw_spin_lock_irq(&running_trace_lock);
-			list_add(&bt->running_list, &running_trace_list);
-			raw_spin_unlock_irq(&running_trace_lock);
-
-			trace_note_time(bt);
-			ret = 0;
-		}
-	} else {
-		if (bt->trace_state == Blktrace_running) {
-			bt->trace_state = Blktrace_stopped;
-			raw_spin_lock_irq(&running_trace_lock);
-			list_del_init(&bt->running_list);
-			raw_spin_unlock_irq(&running_trace_lock);
-			relay_flush(bt->rchan);
-			ret = 0;
-		}
-	}
-
-	return ret;
+	return blk_trace_switch_state(bt, start);
 }
 
 int blk_trace_startstop(struct request_queue *q, int start)
@@ -1614,13 +1617,8 @@ static int blk_trace_remove_queue(struct request_queue *q)
 	if (bt == NULL)
 		return -EINVAL;
 
-	if (bt->trace_state == Blktrace_running) {
-		bt->trace_state = Blktrace_stopped;
-		raw_spin_lock_irq(&running_trace_lock);
-		list_del_init(&bt->running_list);
-		raw_spin_unlock_irq(&running_trace_lock);
-		relay_flush(bt->rchan);
-	}
+	if (bt->trace_state == Blktrace_running)
+		blk_trace_switch_state(bt, 0);
 
 	put_probe_ref();
 	synchronize_rcu();
-- 
2.31.1

Re: [PATCH -next 1/3] blktrace: introduce 'blk_trace_swicth_state' helper

[Christoph Hellwig]

On Mon, Oct 17, 2022 at 02:53:19PM +0800, Ye Bin wrote:
> +static int blk_trace_switch_state(struct blk_trace *bt, int start)

This is a very confusing helper.  Please split it into two helpers
(e.g. blk_trace_{start,stop}) instead of the awakward int used as
boolean argument that makes it do two entirely unrelated things.

[PATCH -next 2/3] blktrace: fix possible memleak in '__blk_trace_remove'

[Ye Bin]

When test as follows:
step1: ioctl(sda, BLKTRACESETUP, &arg)
step2: ioctl(sda, BLKTRACESTART, NULL)
step3: ioctl(sda, BLKTRACETEARDOWN, NULL)
step4: ioctl(sda, BLKTRACESETUP, &arg)
Got issue as follows:
debugfs: File 'dropped' in directory 'sda' already present!
debugfs: File 'msg' in directory 'sda' already present!
debugfs: File 'trace0' in directory 'sda' already present!

And also find syzkaller report issue like "KASAN: use-after-free Read in relay_switch_subbuf"
"https://syzkaller.appspot.com/bug?id=13849f0d9b1b818b087341691be6cc3ac6a6bfb7"

If remove block trace without stop(BLKTRACESTOP) block trace, '__blk_trace_remove'
will just set 'q->blk_trace' with NULL. However, debugfs file isn't removed, so
will report file already present when call BLKTRACESETUP.
static int __blk_trace_remove(struct request_queue *q)
{
        struct blk_trace *bt;

        bt = rcu_replace_pointer(q->blk_trace, NULL,
                                 lockdep_is_held(&q->debugfs_mutex));
        if (!bt)
                return -EINVAL;

	if (bt->trace_state != Blktrace_running)
        	blk_trace_cleanup(q, bt);

        return 0;
}

If do test as follows:
step1: ioctl(sda, BLKTRACESETUP, &arg)
step2: ioctl(sda, BLKTRACESTART, NULL)
step3: ioctl(sda, BLKTRACETEARDOWN, NULL)
step4: remove sda

There will remove debugfs directory which will remove recursively all file
under directory.
>> blk_release_queue
>>	debugfs_remove_recursive(q->debugfs_dir)
So all files which created in 'do_blk_trace_setup' are removed, and
'dentry->d_inode' is NULL. But 'q->blk_trace' is still in 'running_trace_lock',
'trace_note_tsk' will traverse 'running_trace_lock' all nodes.
>>trace_note_tsk
>>  trace_note
>>    relay_reserve
>>       relay_switch_subbuf
>>        d_inode(buf->dentry)->i_size

To solve above issues, reference commit '5afedf670caf', first stop block trace
when block trace state is 'Blktrace_running' in '__blk_trace_remove'.

Signed-off-by: Ye Bin <yebin10@huawei.com>
---
 kernel/trace/blktrace.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index edd83e213580..0d93a0110ab5 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -395,8 +395,10 @@ static int __blk_trace_remove(struct request_queue *q)
 	if (!bt)
 		return -EINVAL;
 
-	if (bt->trace_state != Blktrace_running)
-		blk_trace_cleanup(q, bt);
+	if (bt->trace_state == Blktrace_running)
+		blk_trace_switch_state(bt, 0);
+
+	blk_trace_cleanup(q, bt);
 
 	return 0;
 }
-- 
2.31.1

Re: [PATCH -next 2/3] blktrace: fix possible memleak in '__blk_trace_remove'

[Yu Kuai]

Hi,

在 2022/10/17 14:53, Ye Bin 写道:
> To solve above issues, reference commit '5afedf670caf', first stop block trace
> when block trace state is 'Blktrace_running' in '__blk_trace_remove'.

Will it be much simpler if we just return -EBUSY in blk_trace_remove if
state is still running?

And similar checking in blk_trace_setup.

Thanks,
Kuai

Re: [PATCH -next 2/3] blktrace: fix possible memleak in '__blk_trace_remove'

[Christoph Hellwig]

On Mon, Oct 17, 2022 at 02:53:20PM +0800, Ye Bin wrote:
> +	if (bt->trace_state == Blktrace_running)
> +		blk_trace_switch_state(bt, 0);

AFAICS blk_trace_switch_state already has that state check, so there
should be no need to duplicate it here.

I think having this call in blk_trace_cleanup itself might be a little
more obvious, too.

[PATCH -next 3/3] blktrace: remove unnessary stop block trace in 'blk_trace_shutdown'

[Ye Bin]

As previous commit, '__blk_trace_remove' will stop block trace if
block trace's state is 'Blktrace_running'.
So remove unnessary stop block trace in 'blk_trace_shutdown'.

Signed-off-by: Ye Bin <yebin10@huawei.com>
---
 kernel/trace/blktrace.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index 0d93a0110ab5..f33ec0f5750a 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -777,10 +777,8 @@ int blk_trace_ioctl(struct block_device *bdev, unsigned cmd, char __user *arg)
 void blk_trace_shutdown(struct request_queue *q)
 {
 	if (rcu_dereference_protected(q->blk_trace,
-				      lockdep_is_held(&q->debugfs_mutex))) {
-		__blk_trace_startstop(q, 0);
+				      lockdep_is_held(&q->debugfs_mutex)))
 		__blk_trace_remove(q);
-	}
 }
 
 #ifdef CONFIG_BLK_CGROUP
-- 
2.31.1