Re: [PATCH] net/tls: Fix slab-use-after-free in tls_encrypt_done

[Juntong Deng]

On 2023/10/16 17:50, Simon Horman wrote:
> On Thu, Oct 12, 2023 at 07:02:51PM +0800, Juntong Deng wrote:
> > In the current implementation, ctx->async_wait.completion is completed
> > after spin_lock_bh, which causes tls_sw_release_resources_tx to
> > continue executing and return to tls_sk_proto_cleanup, then return
>
> Hi Juntong Deng,
>
> I'm slightly confused by "causes tls_sw_release_resources_tx to  continue
> executing".
>
> What I see in tls_sw_release_resources_tx() is:
>
>          /* Wait for any pending async encryptions to complete */
>          spin_lock_bh(&ctx->encrypt_compl_lock);
>          ctx->async_notify = true;
>          pending = atomic_read(&ctx->encrypt_pending);
>          spin_unlock_bh(&ctx->encrypt_compl_lock);
>
> Am I wrong in thinking the above will block because
> (the same) ctx->encrypt_compl_lock is held in tls_encrypt_done?

Hi Simon Horman,

What I mean is that tls_sw_release_resources_tx will pause at
crypto_wait_req(-EINPROGRESS, &ctx->async_wait) because crypto_wait_req
call wait_for_completion.

Then after tls_encrypt_done call complete(&ctx->async_wait.completion),
it will cause tls_sw_release_resources_tx to continue executing.

> > to tls_sk_proto_close, and after that enter tls_sw_free_ctx_tx to kfree
> > the entire struct tls_context (including ctx->encrypt_compl_lock).
> >
> > Since ctx->encrypt_compl_lock has been freed, subsequent spin_unlock_bh
> > will result in slab-use-after-free error. Due to SMP, even using
> > spin_lock_bh does not prevent tls_sw_release_resources_tx from continuing
> > on other CPUs. After tls_sw_release_resources_tx is woken up, there is no
> > attempt to hold ctx->encrypt_compl_lock again, therefore everything
> > described above is possible.
> >
> > The fix is to put complete(&ctx->async_wait.completion) after
> > spin_unlock_bh, making the release after the unlock. Since complete is
> > only executed if pending is 0, which means this is the last record, there
> > is no need to worry about race condition causing duplicate completes.
> >
> > Reported-by: syzbot+29c22ea2d6b2c5fd2eae@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Closes: https://syzkaller.appspot.com/bug?extid=29c22ea2d6b2c5fd2eae
> > Signed-off-by: Juntong Deng <juntong.deng@xxxxxxxxxxx>
> > ---
> >   net/tls/tls_sw.c | 6 ++++--
> >   1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
> > index 270712b8d391..7abe5a6aa989 100644
> > --- a/net/tls/tls_sw.c
> > +++ b/net/tls/tls_sw.c
> > @@ -441,6 +441,7 @@ static void tls_encrypt_done(void *data, int err)
> >   	struct sk_msg *msg_en;
> >   	bool ready = false;
> >   	struct sock *sk;
> > +	int async_notify;
> >   	int pending;
> >   
> >   	msg_en = &rec->msg_encrypted;
> > @@ -482,10 +483,11 @@ static void tls_encrypt_done(void *data, int err)
> >   
> >   	spin_lock_bh(&ctx->encrypt_compl_lock);
> >   	pending = atomic_dec_return(&ctx->encrypt_pending);
> > +	async_notify = ctx->async_notify;
> > +	spin_unlock_bh(&ctx->encrypt_compl_lock);
> >   
> > -	if (!pending && ctx->async_notify)
> > +	if (!pending && async_notify)
> >   		complete(&ctx->async_wait.completion);
> > -	spin_unlock_bh(&ctx->encrypt_compl_lock);
> >   
> >   	if (!ready)
> >   		return;
> > --
> > 2.39.2