wifi: cfg80211: cancel sched scan results work on unregister

cfg80211_sched_scan_results() can queue rdev->sched_scan_res_wk from a
driver result notification while a scheduled scan request is present. The
work callback recovers the containing cfg80211_registered_device and then
locks the wiphy and walks the scheduled-scan request list.

wiphy_unregister() already makes the wiphy unreachable and drains rdev work
items before cfg80211_dev_free() can release the object, but it does not
drain sched_scan_res_wk. A queued or running result work item can therefore
cross the unregister/free boundary and access freed rdev state.

The buggy scenario involves two paths, with each column showing the order
within that path:

scheduled-scan result path:        unregister/free path:
1. cfg80211_sched_scan_results()   1. interface teardown stops and
   queues rdev->sched_scan_res_wk.    removes the scheduled scan request.
2. cfg80211_wq starts the work     2. wiphy_unregister() drains other
   item and recovers rdev.            rdev work items.
3. The worker locks rdev->wiphy    3. cfg80211_dev_free() destroys and
   and walks rdev state.              frees rdev.

Cancel sched_scan_res_wk in wiphy_unregister() alongside the other rdev
work items. cancel_work_sync() removes a pending result notification and
waits for an already running callback, so cfg80211_dev_free() cannot free
rdev while this work item is still active.

Validation reproduced this kernel report:
BUG: KASAN: use-after-free in cfg80211_sched_scan_results_wk+0x4a6/0x530
Workqueue: cfg80211 cfg80211_sched_scan_results_wk [cfg80211]
Read of size 8
Call trace:
  dump_stack_lvl+0x66/0xa0
  print_report+0xce/0x630
  cfg80211_sched_scan_results_wk+0x4a6/0x530
  srso_alias_return_thunk+0x5/0xfbef5
  __virt_addr_valid+0x224/0x430
  kasan_report+0xac/0xe0
  lockdep_hardirqs_on_prepare+0xea/0x1a0
  process_one_work+0x8d0/0x18f0 (kernel/workqueue.c:3212)
  lock_is_held_type+0x8f/0x100
  worker_thread+0x5ad/0xfd0
  __kthread_parkme+0xc6/0x200
  kthread+0x31e/0x410
  trace_hardirqs_on+0x1a/0x170
  ret_from_fork+0x576/0x810
  __switch_to+0x57e/0xe20
  __switch_to_asm+0x33/0x70
  ret_from_fork_asm+0x1a/0x30

Fixes: 807f8a8c3004 ("cfg80211/nl80211: add support for scheduled scans")
Assisted-by: Codex:gpt-5.5
Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
Link: https://patch.msgid.link/20260619162542.3878296-1-zzzccc427@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

diff --git a/net/wireless/core.c b/net/wireless/core.c
index 3dcf63b04c41db43a6e626f4ee8f9ceb1fee220e..2c729a7aca12dd797d1b45e604b8ee8934895669 100644 (file)
--- a/net/wireless/core.c
+++ b/net/wireless/core.c
@@ -1335,6 +1335,7 @@ void wiphy_unregister(struct wiphy *wiphy)
        /* this has nothing to do now but make sure it's gone */
        cancel_work_sync(&rdev->wiphy_work);
 
+       cancel_work_sync(&rdev->sched_scan_res_wk);
        cancel_work_sync(&rdev->rfkill_block);
        cancel_work_sync(&rdev->conn_work);
        flush_work(&rdev->event_work);