[PATCH v2 net] nfc: llcp: use disable_work_sync and timer_shutdown_sync in local_cleanup

[PATCH v2 net] nfc: llcp: use disable_work_sync and timer_shutdown_sync in local_cleanup

[Tristan Madani]

From: Tristan Madani <tristan@talencesecurity.com>

local_cleanup() tears down timers and work items before
local_release_work() frees the nfc_llcp_local structure.  The current
code has two ordering problems:

1) timer_delete_sync() does not prevent timer callbacks from re-arming
   the timer via mod_timer().  nfc_llcp_tx_work() unconditionally
   re-arms link_timer, and nfc_llcp_sdreq_timeout_work() conditionally
   re-arms sdreq_timer.  A work callback that runs during
   cancel_work_sync() and calls mod_timer() causes the timer to fire
   after the structure is freed.

2) cancel_work_sync() does not prevent work items from being
   re-queued.  nfc_llcp_rx_work() queues tx_work via
   schedule_work(), and the timer callbacks queue their associated
   work items.  Since tx_work is cancelled before rx_work, a
   running rx_work instance can re-queue tx_work after its
   cancel_work_sync() has returned.  The re-queued tx_work then
   runs on the freed structure.

Switch the timers to timer_shutdown_sync() which rejects subsequent
mod_timer() calls.  Switch the work items to disable_work_sync()
which rejects subsequent schedule_work() calls.

Reorder the teardown sequence so that both timers are shut down
first (their callbacks queue work items), rx_work is disabled
before tx_work (rx_work queues tx_work), and data structures are
freed only after all asynchronous callbacks are quiesced.

Fixes: d646960f7986 ("NFC: Initial LLCP support")
Cc: stable@vger.kernel.org
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
---
v2:
  - Switch work items from cancel_work_sync() to disable_work_sync()
    to prevent re-queuing after cancellation, as identified by the
    automated review and confirmed by Simon Horman
  - Reorder teardown: shut down both timers first (their callbacks
    queue work items), then disable rx_work before tx_work (rx_work
    queues tx_work), then disable remaining work items
  - Move data structure cleanup (skb_queue_purge, kfree_skb) after
    all asynchronous callbacks are quiesced
 net/nfc/llcp_core.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/net/nfc/llcp_core.c b/net/nfc/llcp_core.c
index 74bf817007cf2..984e7088288f0 100644
--- a/net/nfc/llcp_core.c
+++ b/net/nfc/llcp_core.c
@@ -173,15 +173,15 @@ static struct nfc_llcp_local *nfc_llcp_local_get(struct nfc_llcp_local *local)
 static void local_cleanup(struct nfc_llcp_local *local)
 {
 	nfc_llcp_socket_release(local, false, ENXIO);
-	timer_delete_sync(&local->link_timer);
+	timer_shutdown_sync(&local->link_timer);
+	timer_shutdown_sync(&local->sdreq_timer);
+	disable_work_sync(&local->rx_work);
+	disable_work_sync(&local->tx_work);
+	disable_work_sync(&local->timeout_work);
+	disable_work_sync(&local->sdreq_timeout_work);
 	skb_queue_purge(&local->tx_queue);
-	cancel_work_sync(&local->tx_work);
-	cancel_work_sync(&local->rx_work);
-	cancel_work_sync(&local->timeout_work);
 	kfree_skb(local->rx_pending);
 	local->rx_pending = NULL;
-	timer_delete_sync(&local->sdreq_timer);
-	cancel_work_sync(&local->sdreq_timeout_work);
 	nfc_llcp_free_sdp_tlv_list(&local->pending_sdreqs);
 }
 
-- 
2.47.3

Re: [PATCH v2 net] nfc: llcp: use disable_work_sync and timer_shutdown_sync in local_cleanup

[netdev-bot+sinfo]

Hi!

This is an automated message. This series looks like a fix, but its
commit messages seem to be missing some information:

 - How the issue was discovered, e.g. hit in production, hit during
   development, syzbot report, manual code inspection, LLM or static
   analysis tool scan.

 - Whether the issue was actually triggered, or is only theoretical
   (e.g. found by code inspection). If it was triggered please include
   the symptoms, like the stack trace or error messages.

Please do not repost the series just to address the above. Instead,
reply to this email with the missing information, so that reviewers
can take it into account. If the series needs another revision for
other reasons, please include the information in the commit messages
then.

The evaluation is done by an LLM so it may be wrong, if you think
that is the case please reply and explain.