| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAKZCeSVJQAsYVSDBHOmUyBMbQqBA1D7ZP4KK-azWUjb+p-vpJg@mail.gmail.com>
Date: Tue, 10 Feb 2026 15:08:12 +0900
From: kawada <youjingxiaogao2@...il.com>
To: Hillf Danton <hdanton@...a.com>
Cc: linux-bluetooth@...r.kernel.org, luiz.dentz@...il.com, marcel@...tmann.org,
johan.hedberg@...il.com, linux-kernel@...r.kernel.org,
syzbot+3609b9b48e68e1fe47fd@...kaller.appspotmail.com
Subject: Re: [PATCH] Bluetooth: fix use-after-free in hci_conn_drop
> but the race still exists after this patch.
I agree that the TOCTOU issue remains. I looked into all the functions
that can be called as entry->destroy in hci_cmd_sync_work where the
second argument (data) is used as hci_conn*. I found five such
functions across two files:
In hci_sync.c:
- create_le_conn_complete
- create_pa_complete
- create_big_complete
- le_read_features_complete
In hci_conn.c:
- create_big_complete (separate static function)
Of these, create_le_conn_complete and create_pa_complete already use
hci_dev_lock, which effectively prevents this TOCTOU issue.
The remaining three are vulnerable:
- create_big_complete in hci_sync.c calls hci_conn_valid without
holding hci_dev_lock (TOCTOU)
- le_read_features_complete calls hci_conn_drop with no validity
check at all
- create_big_complete in hci_conn.c calls hci_connect_cfm and
hci_conn_del with no validity check at all
Given this, I believe the following set of patches would be
appropriate:
1. Reorder hci_cmd_sync_dequeue before hci_conn_cleanup in
hci_conn_del
2. Wrap all three vulnerable callbacks with hci_dev_lock and
hci_conn_valid, following the same pattern used by
create_le_conn_complete and create_pa_complete
If this approach sounds reasonable, I will prepare updated patches.
2026年2月9日(月) 22:25 Hillf Danton <hdanton@...a.com>:
>
> On Mon, 9 Feb 2026 19:02:11 +0900 Masahiro Kawada wrote:
> > Fix a use-after-free in hci_conn_drop triggered via hci_cmd_sync_work.
> >
> > In hci_conn_del(), hci_cmd_sync_dequeue() is called after
> > hci_conn_cleanup() which may have already freed the conn pointer.
> > Fix by moving the dequeue before cleanup.
> >
> > Additionally, le_read_features_complete() calls hci_conn_drop(conn)
> > without checking whether conn is still valid. When
> > hci_le_read_remote_features_sync() blocks waiting for an HCI event,
> > another thread can free conn through hci_conn_del(). Fix by adding
> > a hci_conn_valid() check before calling hci_conn_drop().
> >
> > Fixes: 881559af5f5c ("Bluetooth: hci_sync: Attempt to dequeue connection attempt")
> > Fixes: a106e50be74b ("Bluetooth: HCI: Add support for LL Extended Feature Set")
> > Reported-by: syzbot+3609b9b48e68e1fe47fd@...kaller.appspotmail.com
> > Closes: https://syzkaller.appspot.com/bug?extid=3609b9b48e68e1fe47fd
> > Tested-by: syzbot+3609b9b48e68e1fe47fd@...kaller.appspotmail.com
> > Signed-off-by: Masahiro Kawada <youjingxiaogao2@...il.com>
> > ---
> > net/bluetooth/hci_conn.c | 6 +++---
> > net/bluetooth/hci_sync.c | 3 +++
> > 2 files changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/net/bluetooth/hci_conn.c b/net/bluetooth/hci_conn.c
> > index 0795818963a..aa3607327ad 100644
> > --- a/net/bluetooth/hci_conn.c
> > +++ b/net/bluetooth/hci_conn.c
> > @@ -1232,15 +1232,15 @@ void hci_conn_del(struct hci_conn *conn)
> > skb_queue_purge(&conn->data_q);
> > skb_queue_purge(&conn->tx_q.queue);
> >
> > + /* Dequeue callbacks using connection pointer as data */
> > + hci_cmd_sync_dequeue(hdev, NULL, conn, NULL);
> > +
> > /* Remove the connection from the list and cleanup its remaining
> > * state. This is a separate function since for some cases like
> > * BT_CONNECT_SCAN we *only* want the cleanup part without the
> > * rest of hci_conn_del.
> > */
> > hci_conn_cleanup(conn);
> > -
> > - /* Dequeue callbacks using connection pointer as data */
> > - hci_cmd_sync_dequeue(hdev, NULL, conn, NULL);
> > }
> >
> > struct hci_dev *hci_get_route(bdaddr_t *dst, bdaddr_t *src, uint8_t src_type)
> > diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
> > index f04a90bce4a..f31086c187f 100644
> > --- a/net/bluetooth/hci_sync.c
> > +++ b/net/bluetooth/hci_sync.c
> > @@ -7371,6 +7371,9 @@ static void le_read_features_complete(struct hci_dev *hdev, void *data, int err)
> > if (err == -ECANCELED)
> > return;
> >
> > + if (!hci_conn_valid(hdev, conn))
> > + return;
> > +
> > hci_conn_drop(conn);
> > }
> >
> > --
> > 2.43.0
> >
> The uaf [1] is due to the following race,
>
> cpu1 cpu2
> hci_cmd_sync_work() hci_rx_work()
> mutex_lock(&hdev->cmd_sync_work_lock);
> entry = list_first_entry_or_null(&hdev->cmd_sync_work_list,
> struct hci_cmd_sync_work_entry,
> list);
> if (entry)
> list_del(&entry->list);
> mutex_unlock(&hdev->cmd_sync_work_lock);
>
> hci_conn_del()
> hci_conn_hash_del(hdev, conn);
> hci_conn_cleanup(conn) // free conn
> hci_cmd_sync_dequeue()
> mutex_lock(&hdev->cmd_sync_work_lock);
> while ((entry = _hci_cmd_sync_lookup_entry(hdev, func, data,
> destroy))) {
> _hci_cmd_sync_cancel_entry(hdev, entry, -ECANCELED);
> ret = true;
> }
> mutex_unlock(&hdev->cmd_sync_work_lock);
>
> hci_req_sync_lock(hdev);
> err = entry->func(hdev, entry->data);
> if (entry->destroy)
> entry->destroy(hdev, entry->data, err);
> hci_conn_drop(conn) // uaf
> hci_req_sync_unlock(hdev);
>
> but the race still exists after this patch.
>
> cpu1 cpu2
> hci_conn_valid(hdev, conn)
> hci_conn_hash_del(hdev, conn);
> hci_cmd_sync_dequeue()
> hci_conn_cleanup(conn) // free conn
> hci_conn_drop(conn); // uaf
>
> [1] Subject: [syzbot] [bluetooth?] KASAN: slab-use-after-free Write in hci_conn_drop (3)
> https://lore.kernel.org/lkml/69301edd.a70a0220.2ea503.00cf.GAE@google.com/
Powered by blists - more mailing lists