Re: [PATCH] Bluetooth: L2CAP: don't reserve an ident for L2CAP_LE_CREDITS
From: Luiz Augusto von Dentz
Date: Mon Nov 16 2026 - 12:04:42 EST
Hi Nguyen,
On Sun, Nov 15, 2026 at 6:53 AM Nguyen Ngoc Thang
<ngocthang2710.1999@xxxxxxxxx> wrote:
>
> l2cap_chan_le_send_credits() allocates an ident from conn->tx_ida, sends
> L2CAP_LE_CREDITS and then frees the ident unconditionally, since the
> command has no response.
>
> But the ident is visible in tx_ida while the command is in flight, and
> the remote may answer it with L2CAP_COMMAND_REJ anyway. The RX path then
> releases it in l2cap_put_ident() under conn->lock, while the credits path
> runs from recvmsg() with only the socket lock held. Whichever ida_free()
> comes second hits:
>
> ida_free called for id=3 which is not allocated.
> WARNING: lib/idr.c:594 at ida_free+0x24c/0x2d4 lib/idr.c:594
> Call trace:
> ida_free+0x24c/0x2d4 lib/idr.c:594
> l2cap_chan_le_send_credits+0x17c/0x29c net/bluetooth/l2cap_core.c:7113
> l2cap_chan_rx_avail+0xb4/0x134 net/bluetooth/l2cap_core.c:7126
> l2cap_publish_rx_avail+0x110/0x168 net/bluetooth/l2cap_sock.c
> l2cap_sock_recvmsg+0x280/0x7b8 net/bluetooth/l2cap_sock.c:1317
>
> Worse, if the ident is reallocated for a real request in between, the
> credits path frees that request's ident instead.
>
> tx_ida only exists to track requests awaiting a response, so a command
> without one has no business in it. Give L2CAP_LE_CREDITS a plain cyclic
> ident instead, which never needs releasing. The ident is still never 0
> and never exhausts the range, which is what commit 6e1930ece855
> ("Bluetooth: L2CAP: fix tx ident leak for commands without a response")
> set out to fix.
>
> Fixes: 6e1930ece855 ("Bluetooth: L2CAP: fix tx ident leak for commands without a response")
> Reported-by: syzbot+9dc8780713b11155dd74@xxxxxxxxxxxxxxxxxxxxxxxxx
> Closes: https://syzkaller.appspot.com/bug?extid=9dc8780713b11155dd74
> Signed-off-by: Nguyen Ngoc Thang <ngocthang2710.1999@xxxxxxxxx>
> ---
> Tested in QEMU (x86_64, KASAN + lockdep) with the syzbot C reproducer,
> 15 runs per kernel:
> base (origin/master 23c62304dde2): 31x "ida_free called for id=N ..."
> patched: 0 warnings, 1200 LE credits sent,
> none with ident 0 (ident wraps fine)
>
> net/bluetooth/l2cap_core.c | 44 +++++++++++++++++++++-----------------
> 1 file changed, 24 insertions(+), 20 deletions(-)
>
> diff --git a/net/bluetooth/l2cap_core.c b/net/bluetooth/l2cap_core.c
> index 7dcb0586a10b..0b861ced5703 100644
> --- a/net/bluetooth/l2cap_core.c
> +++ b/net/bluetooth/l2cap_core.c
> @@ -1028,21 +1028,36 @@ int l2cap_chan_check_security(struct l2cap_chan *chan, bool initiator)
> initiator);
> }
>
> -static int l2cap_get_ident(struct l2cap_conn *conn)
> +static u8 l2cap_ident_max(struct l2cap_conn *conn)
> {
> - u8 max;
> - int ident;
> -
> /* LE link does not support tools like l2ping so use the full range */
> if (conn->hcon->type == LE_LINK)
> - max = 255;
> + return 255;
> +
> /* Get next available identificator.
> * 1 - 128 are used by kernel.
> * 129 - 199 are reserved.
> * 200 - 254 are used by utilities like l2ping, etc.
> */
> - else
> - max = 128;
> + return 128;
> +}
> +
> +/* Ident for a command that has no response: it is not reserved in tx_ida,
> + * so nothing has to release it and it can't be released twice.
> + */
> +static u8 l2cap_next_ident(struct l2cap_conn *conn)
> +{
> + u8 ident = READ_ONCE(conn->tx_ident) % l2cap_ident_max(conn) + 1;
> +
> + WRITE_ONCE(conn->tx_ident, ident);
> +
> + return ident;
> +}
> +
> +static int l2cap_get_ident(struct l2cap_conn *conn)
> +{
> + u8 max = l2cap_ident_max(conn);
> + int ident;
>
> /* Allocate ident using min as last used + 1 (cyclic) */
> ident = ida_alloc_range(&conn->tx_ida, READ_ONCE(conn->tx_ident) + 1,
> @@ -7081,7 +7096,6 @@ static void l2cap_chan_le_send_credits(struct l2cap_chan *chan)
> struct l2cap_conn *conn = chan->conn;
> struct l2cap_le_credits pkt;
> u16 return_credits = l2cap_le_rx_credits(chan);
> - int ident;
>
> if (chan->mode != L2CAP_MODE_LE_FLOWCTL &&
> chan->mode != L2CAP_MODE_EXT_FLOWCTL)
> @@ -7099,18 +7113,8 @@ static void l2cap_chan_le_send_credits(struct l2cap_chan *chan)
> pkt.cid = cpu_to_le16(chan->scid);
> pkt.credits = cpu_to_le16(return_credits);
>
> - ident = l2cap_get_ident(conn);
> -
> - l2cap_send_cmd(conn, ident, L2CAP_LE_CREDITS, sizeof(pkt), &pkt);
> -
> - /* L2CAP_LE_CREDITS has no response so the ident is never released by
> - * l2cap_put_ident() - release it right away, otherwise the tx_ida
> - * range is exhausted after 254 packets and from then on credits are
> - * sent with the invalid ident 0, which some remote stacks ignore,
> - * stalling the channel.
> - */
> - if (ident > 0)
> - ida_free(&conn->tx_ida, ident);
> + l2cap_send_cmd(conn, l2cap_next_ident(conn), L2CAP_LE_CREDITS,
> + sizeof(pkt), &pkt);
> }
>
> void l2cap_chan_rx_avail(struct l2cap_chan *chan, ssize_t rx_avail)
> --
> 2.43.0
Sashiko flagged a problem:
https://sashiko.dev/#/patchset/20261115115341.316484-1-ngocthang2710.1999%40gmail.com
--
Luiz Augusto von Dentz