[v2,0/3] Bluetooth: ISO-related concurrency fixes

Message

Pauli Virtanen June 13, 2023, 6:06 p.m. UTC

This series addresses some concurrency issues (NULL / GPF) in ISO
sockets or related.

v2:
- Use RCU for the pend_le_* lists, avoid using hci_dev_lock.
- Always call disconn_cfm before hci_conn_del (L2CAP also needs it).

These were found while testing patches that make hci_le_set_cig_params
check the validity of the configuration and return false if incorrect.
This causes dropping of hci_conn just created, which apparently makes
hitting race conditions easier.

The test setup was primitive

while true; do bluetoothctl power on; sleep 12; bluetoothctl power off; sleep 1.5; bluetoothctl power off; sleep 2.5; done;
while true; do sudo systemctl restart bluetooth; sleep 110; done
while true; do systemctl --user restart pipewire wireplumber pipewire-pulse; sleep 91; done
while true; do paplay sample.flac & sleep 2; kill %1; sleep 0.7; done

and equivalent operations manually, on VM + connect to TWS earbuds. This
eventually hit the NULL / GFP errors here, but they are hard to
reproduce aside from the first one that appears in iso-tester.

Pauli Virtanen (3):
  Bluetooth: use RCU for hci_conn_params and iterate safely in hci_sync
  Bluetooth: hci_event: call disconnect callback before deleting conn
  Bluetooth: ISO: fix iso_conn related locking and validity issues

 include/net/bluetooth/hci_core.h |  5 ++
 net/bluetooth/hci_conn.c         |  9 ++--
 net/bluetooth/hci_core.c         | 34 +++++++++---
 net/bluetooth/hci_event.c        | 15 +++---
 net/bluetooth/hci_sync.c         | 93 ++++++++++++++++++++++++++++----
 net/bluetooth/iso.c              | 53 ++++++++++--------
 net/bluetooth/mgmt.c             | 30 +++++------
 7 files changed, 175 insertions(+), 64 deletions(-)