From: Yu Junzhe <junzheyu1@gmail.com>
To: stable@vger.kernel.org
Cc: Yu Junzhe <junzheyu1@gmail.com>,
Marcel Holtmann <marcel@holtmann.org>,
Johan Hedberg <johan.hedberg@gmail.com>,
Luiz Augusto von Dentz <luiz.dentz@gmail.com>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Sasha Levin <sashal@kernel.org>
Subject: [PATCH 6.6 v2] Bluetooth: hci_core: lookup hci_conn on the ACL RX path
Date: Thu, 19 Nov 2026 03:57:58 +0000 [thread overview]
Message-ID: <20261119035758.45-1-junzheyu1@gmail.com> (raw)
In-Reply-To: <20261118123533.25-1-junzheyu1@gmail.com>
Bluetooth: hci_core: lookup hci_conn on the ACL RX path
Backport of 79a2d4678ba9 ("Bluetooth: hci_core: lookup hci_conn on RX
path on protocol side") for linux-6.6.y, L2CAP only.
6.6 already has l2cap_conn_hold_unless_zero(). Look the handle up under
hci_dev_lock, enter active mode, hold the l2cap_conn, then drop the
lock. hci_acldata_packet() used to drop hci_dev_lock and pass the raw
hci_conn pointer into l2cap_recv_acldata(), whose first access is
hci_dev_lock(hcon->hdev). The hash lookup takes no reference, and
bt_link_release() frees the object with kfree() from hci_cmd_sync_work.
v1 also rewrote SCO and ISO to call hci_conn_get(). That does not match
79a2d4678ba9. hci_conn_get() is get_device() and only delays freeing
hci_conn. On 6.6, sco_conn_del() and iso_conn_del() still kfree the
protocol object. sco_conn_hold_unless_zero() and
iso_conn_hold_unless_zero() arrived in e6720779ae61 and dc26097bdb86
(v6.13), so those paths are left unchanged here.
CVE-2025-68304. Only 6.6.162 was booted, and the KASAN report is in
l2cap_recv_acldata. The ACL hunks in this v2 are the ones from that
build. This v2 itself was not rebuilt: the booted bzImage also contained
the v1 SCO and ISO edits, which are removed here.
This is for linux-6.6.y. bluetooth.git HEAD already has 79a2d4678ba9
(v6.18), so this diff does not apply there.
Cc: stable@vger.kernel.org # 6.6.162
Fixes: 5af1f84ed13a ("Bluetooth: hci_sync: Fix UAF on hci_abort_conn_sync")
Signed-off-by: Yu Junzhe <junzheyu1@gmail.com>
---
v2:
- Drop the SCO and ISO edits. hci_conn_get() does not pin sco_conn or
iso_conn on 6.6.
- Address this to stable. The BlueZ CI applies mail to bluetooth.git
HEAD, which already contains the upstream commit.
include/net/bluetooth/hci_core.h | 3 ++-
net/bluetooth/hci_core.c | 25 ++++++++-----------------
net/bluetooth/l2cap_core.c | 23 ++++++++++++++++++-----
3 files changed, 28 insertions(+), 23 deletions(-)
diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
index 62804d98d..e0deaa54b 100644
--- a/include/net/bluetooth/hci_core.h
+++ b/include/net/bluetooth/hci_core.h
@@ -837,7 +837,8 @@ extern struct mutex hci_cb_list_lock;
/* ----- HCI interface to upper protocols ----- */
int l2cap_connect_ind(struct hci_dev *hdev, bdaddr_t *bdaddr);
int l2cap_disconn_ind(struct hci_conn *hcon);
-void l2cap_recv_acldata(struct hci_conn *hcon, struct sk_buff *skb, u16 flags);
+int l2cap_recv_acldata(struct hci_dev *hdev, u16 handle, struct sk_buff *skb,
+ u16 flags);
#if IS_ENABLED(CONFIG_BT_BREDR)
int sco_connect_ind(struct hci_dev *hdev, bdaddr_t *bdaddr, __u8 *flags);
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index 84d9a445e..a54d29bad 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -3798,13 +3798,14 @@ static void hci_tx_work(struct work_struct *work)
static void hci_acldata_packet(struct hci_dev *hdev, struct sk_buff *skb)
{
struct hci_acl_hdr *hdr;
- struct hci_conn *conn;
__u16 handle, flags;
+ int err;
hdr = skb_pull_data(skb, sizeof(*hdr));
if (!hdr) {
bt_dev_err(hdev, "ACL packet too small");
- goto drop;
+ kfree_skb(skb);
+ return;
}
handle = __le16_to_cpu(hdr->handle);
@@ -3816,23 +3817,13 @@ static void hci_acldata_packet(struct hci_dev *hdev, struct sk_buff *skb)
hdev->stat.acl_rx++;
- hci_dev_lock(hdev);
- conn = hci_conn_hash_lookup_handle(hdev, handle);
- hci_dev_unlock(hdev);
-
- if (conn) {
- hci_conn_enter_active_mode(conn, BT_POWER_FORCE_ACTIVE_OFF);
-
- /* Send to upper protocol */
- l2cap_recv_acldata(conn, skb, flags);
- return;
- } else {
+ err = l2cap_recv_acldata(hdev, handle, skb, flags);
+ if (err == -ENOENT)
bt_dev_err(hdev, "ACL packet for unknown connection handle %d",
handle);
- }
-
-drop:
- kfree_skb(skb);
+ else if (err)
+ bt_dev_dbg(hdev, "ACL packet recv for handle %d failed: %d",
+ handle, err);
}
/* SCO data packet */
diff --git a/net/bluetooth/l2cap_core.c b/net/bluetooth/l2cap_core.c
index 94fddc456..5b5a9e763 100644
--- a/net/bluetooth/l2cap_core.c
+++ b/net/bluetooth/l2cap_core.c
@@ -7752,13 +7752,24 @@ struct l2cap_conn *l2cap_conn_hold_unless_zero(struct l2cap_conn *c)
}
EXPORT_SYMBOL(l2cap_conn_hold_unless_zero);
-void l2cap_recv_acldata(struct hci_conn *hcon, struct sk_buff *skb, u16 flags)
+int l2cap_recv_acldata(struct hci_dev *hdev, u16 handle,
+ struct sk_buff *skb, u16 flags)
{
+ struct hci_conn *hcon;
struct l2cap_conn *conn;
int len;
- /* Lock hdev to access l2cap_data to avoid race with l2cap_conn_del */
- hci_dev_lock(hcon->hdev);
+ /* Lock hdev for hci_conn, and race on l2cap_data vs. l2cap_conn_del */
+ hci_dev_lock(hdev);
+
+ hcon = hci_conn_hash_lookup_handle(hdev, handle);
+ if (!hcon) {
+ hci_dev_unlock(hdev);
+ kfree_skb(skb);
+ return -ENOENT;
+ }
+
+ hci_conn_enter_active_mode(hcon, BT_POWER_FORCE_ACTIVE_OFF);
conn = hcon->l2cap_data;
@@ -7766,12 +7777,13 @@ void l2cap_recv_acldata(struct hci_conn *hcon, struct sk_buff *skb, u16 flags)
conn = l2cap_conn_add(hcon);
conn = l2cap_conn_hold_unless_zero(conn);
+ hcon = NULL;
- hci_dev_unlock(hcon->hdev);
+ hci_dev_unlock(hdev);
if (!conn) {
kfree_skb(skb);
- return;
+ return -EINVAL;
}
BT_DBG("conn %p len %u flags 0x%x", conn, skb->len, flags);
@@ -7885,6 +7897,7 @@ void l2cap_recv_acldata(struct hci_conn *hcon, struct sk_buff *skb, u16 flags)
unlock:
mutex_unlock(&conn->lock);
l2cap_conn_put(conn);
+ return 0;
}
static struct hci_cb l2cap_cb = {
--
2.53.0
prev parent reply other threads:[~2026-11-19 3:58 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-18 12:35 [PATCH 6.6] Bluetooth: hci_core: lookup hci_conn on RX path Yu Junzhe
2026-11-19 3:57 ` Yu Junzhe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20261119035758.45-1-junzheyu1@gmail.com \
--to=junzheyu1@gmail.com \
--cc=gregkh@linuxfoundation.org \
--cc=johan.hedberg@gmail.com \
--cc=luiz.dentz@gmail.com \
--cc=marcel@holtmann.org \
--cc=sashal@kernel.org \
--cc=stable@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox