[PATCH] sctp: fix refcount bug in sctp_wfree

[PATCH] sctp: fix refcount bug in sctp_wfree

[Qiujun Huang]

Do accounting for skb's real sk.
In some case skb->sk != asoc->base.sk.

Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
---
 net/sctp/socket.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index 1b56fc4..5f5c28b 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -9080,7 +9080,7 @@ static void sctp_wfree(struct sk_buff *skb)
 {
 	struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
 	struct sctp_association *asoc = chunk->asoc;
-	struct sock *sk = asoc->base.sk;
+	struct sock *sk = skb->sk;
 
 	sk_mem_uncharge(sk, skb->truesize);
 	sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
@@ -9109,7 +9109,7 @@ static void sctp_wfree(struct sk_buff *skb)
 	}
 
 	sock_wfree(skb);
-	sctp_wake_up_waiters(sk, asoc);
+	sctp_wake_up_waiters(asoc->base.sk, asoc);
 
 	sctp_association_put(asoc);
 }
-- 
1.8.3.1

[PATCH] sctp: fix refcount bug in sctp_wfree

[Qiujun Huang]

Do accounting for skb's real sk.
In some case skb->sk != asoc->base.sk.

Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
---
 net/sctp/socket.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index 1b56fc4..5f5c28b 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -9080,7 +9080,7 @@ static void sctp_wfree(struct sk_buff *skb)
 {
 	struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
 	struct sctp_association *asoc = chunk->asoc;
-	struct sock *sk = asoc->base.sk;
+	struct sock *sk = skb->sk;
 
 	sk_mem_uncharge(sk, skb->truesize);
 	sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
@@ -9109,7 +9109,7 @@ static void sctp_wfree(struct sk_buff *skb)
 	}
 
 	sock_wfree(skb);
-	sctp_wake_up_waiters(sk, asoc);
+	sctp_wake_up_waiters(asoc->base.sk, asoc);
 
 	sctp_association_put(asoc);
 }
-- 
1.8.3.1

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[David Miller]

From: Qiujun Huang <hqjagain@gmail.com>
Date: Mon, 16 Mar 2020 11:53:24 +0800

> Do accounting for skb's real sk.
> In some case skb->sk != asoc->base.sk.
> 
> Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
> Signed-off-by: Qiujun Huang <hqjagain@gmail.com>

SCTP folks, please review.

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[David Miller]

From: Qiujun Huang <hqjagain@gmail.com>
Date: Mon, 16 Mar 2020 11:53:24 +0800

> Do accounting for skb's real sk.
> In some case skb->sk != asoc->base.sk.
> 
> Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
> Signed-off-by: Qiujun Huang <hqjagain@gmail.com>

SCTP folks, please review.

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[Marcelo Ricardo Leitner]

On Mon, Mar 16, 2020 at 11:53:24AM +0800, Qiujun Huang wrote:
> Do accounting for skb's real sk.
> In some case skb->sk != asoc->base.sk.

This is a too simple description.  Please elaborate how this can
happen in sctp_wfree. Especially considering the construct for
migrating the tx queue on sctp_sock_migrate(), as both sockets are
locked while moving the chunks around and the asoc itself is only
moved in between decrementing and incrementing the refcount:

        lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
        sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
        sctp_assoc_migrate(assoc, newsk);
        sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
	...

> 
> Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com

I can't see a positive test result, though. If I didn't loose any
email, your last test with a patch similar to this one actually
failed.
I'm talking about syzbot test result at Message-ID: <000000000000e7736205a0e041f5@google.com>

> Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
> ---
>  net/sctp/socket.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
> index 1b56fc4..5f5c28b 100644
> --- a/net/sctp/socket.c
> +++ b/net/sctp/socket.c
> @@ -9080,7 +9080,7 @@ static void sctp_wfree(struct sk_buff *skb)
>  {
>  	struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
>  	struct sctp_association *asoc = chunk->asoc;
> -	struct sock *sk = asoc->base.sk;
> +	struct sock *sk = skb->sk;
>  
>  	sk_mem_uncharge(sk, skb->truesize);
>  	sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
> @@ -9109,7 +9109,7 @@ static void sctp_wfree(struct sk_buff *skb)
>  	}
>  
>  	sock_wfree(skb);
> -	sctp_wake_up_waiters(sk, asoc);
> +	sctp_wake_up_waiters(asoc->base.sk, asoc);
>  
>  	sctp_association_put(asoc);
>  }
> -- 
> 1.8.3.1
> 

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[Marcelo Ricardo Leitner]

On Mon, Mar 16, 2020 at 11:53:24AM +0800, Qiujun Huang wrote:
> Do accounting for skb's real sk.
> In some case skb->sk != asoc->base.sk.

This is a too simple description.  Please elaborate how this can
happen in sctp_wfree. Especially considering the construct for
migrating the tx queue on sctp_sock_migrate(), as both sockets are
locked while moving the chunks around and the asoc itself is only
moved in between decrementing and incrementing the refcount:

        lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
        sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
        sctp_assoc_migrate(assoc, newsk);
        sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
	...

> 
> Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com

I can't see a positive test result, though. If I didn't loose any
email, your last test with a patch similar to this one actually
failed.
I'm talking about syzbot test result at Message-ID: <000000000000e7736205a0e041f5@google.com>

> Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
> ---
>  net/sctp/socket.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
> index 1b56fc4..5f5c28b 100644
> --- a/net/sctp/socket.c
> +++ b/net/sctp/socket.c
> @@ -9080,7 +9080,7 @@ static void sctp_wfree(struct sk_buff *skb)
>  {
>  	struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
>  	struct sctp_association *asoc = chunk->asoc;
> -	struct sock *sk = asoc->base.sk;
> +	struct sock *sk = skb->sk;
>  
>  	sk_mem_uncharge(sk, skb->truesize);
>  	sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
> @@ -9109,7 +9109,7 @@ static void sctp_wfree(struct sk_buff *skb)
>  	}
>  
>  	sock_wfree(skb);
> -	sctp_wake_up_waiters(sk, asoc);
> +	sctp_wake_up_waiters(asoc->base.sk, asoc);
>  
>  	sctp_association_put(asoc);
>  }
> -- 
> 1.8.3.1
> 

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[Qiujun Huang]

On Tue, Mar 17, 2020 at 12:15 PM Marcelo Ricardo Leitner
<marcelo.leitner@gmail.com> wrote:
>
> On Mon, Mar 16, 2020 at 11:53:24AM +0800, Qiujun Huang wrote:
> > Do accounting for skb's real sk.
> > In some case skb->sk != asoc->base.sk.
>
> This is a too simple description.  Please elaborate how this can
> happen in sctp_wfree. Especially considering the construct for
> migrating the tx queue on sctp_sock_migrate(), as both sockets are
> locked while moving the chunks around and the asoc itself is only
> moved in between decrementing and incrementing the refcount:
>
>         lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
>         sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
>         sctp_assoc_migrate(assoc, newsk);
>         sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
>         ...

Yeah, the description is too simple. I'll send v2.

>
> >
> > Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
>
> I can't see a positive test result, though. If I didn't loose any
> email, your last test with a patch similar to this one actually
> failed.
> I'm talking about syzbot test result at Message-ID: <000000000000e7736205a0e041f5@google.com>

I told with syzbot privately avoiding noise :p
Thanks!

Re: [PATCH] sctp: fix refcount bug in sctp_wfree

[Qiujun Huang]

On Tue, Mar 17, 2020 at 12:15 PM Marcelo Ricardo Leitner
<marcelo.leitner@gmail.com> wrote:
>
> On Mon, Mar 16, 2020 at 11:53:24AM +0800, Qiujun Huang wrote:
> > Do accounting for skb's real sk.
> > In some case skb->sk != asoc->base.sk.
>
> This is a too simple description.  Please elaborate how this can
> happen in sctp_wfree. Especially considering the construct for
> migrating the tx queue on sctp_sock_migrate(), as both sockets are
> locked while moving the chunks around and the asoc itself is only
> moved in between decrementing and incrementing the refcount:
>
>         lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
>         sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
>         sctp_assoc_migrate(assoc, newsk);
>         sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
>         ...

Yeah, the description is too simple. I'll send v2.

>
> >
> > Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
>
> I can't see a positive test result, though. If I didn't loose any
> email, your last test with a patch similar to this one actually
> failed.
> I'm talking about syzbot test result at Message-ID: <000000000000e7736205a0e041f5@google.com>

I told with syzbot privately avoiding noise :p
Thanks!