[PATCH net v3 1/2] xfrm: iptfs: track independent drop deadlines

[Roshan Kumar]

IP-TFS uses one hrtimer for two independently queued states: an
incomplete inner packet and the receive reorder window. Completing or
aborting reassembly cancels that shared timer unconditionally, so packets
in the reorder window can remain queued indefinitely.

Merely leaving the timer armed is insufficient. If it was armed for a
completed reassembly, its old deadline can expire a subsequent reassembly
before that packet's own drop interval has elapsed. A reassembly created
from a runt also does not arm the timer at all.

Record an absolute deadline for an in-progress reassembly. Whenever either
kind of queued state changes, arm the shared timer for the earliest active
deadline. On expiry, drop only state whose own deadline has passed and
rearm the timer for anything that remains.

Reported-by: Lilly Aronleigh <lilly@xxxxxxxxxxxx>
Link: https://lore.kernel.org/netdev/20260824072851.301644-3-lilly@xxxxxxxxxxxx/
Link: https://lore.kernel.org/netdev/apaRiWQn54Pr9hpm@xxxxxxxxxxx/
Fixes: 075694765446 ("xfrm: iptfs: handle received fragmented inner packets")
Cc: stable@xxxxxxxxxxxxxxx
Assisted-by: LLM
Signed-off-by: Roshan Kumar <roshaen09@xxxxxxxxx>
---
 net/xfrm/xfrm_iptfs.c | 81 +++++++++++++++++++++++++++----------------
 1 file changed, 52 insertions(+), 29 deletions(-)

diff --git a/net/xfrm/xfrm_iptfs.c b/net/xfrm/xfrm_iptfs.c
index 6920940a35b4..e538cc98e257 100644
--- a/net/xfrm/xfrm_iptfs.c
+++ b/net/xfrm/xfrm_iptfs.c
@@ -143,6 +143,7 @@ struct skb_wseq {
  * @drop_time_ns: timer intervan in nanoseconds.
  * @ra_newskb: new pkt being reassembled.
  * @ra_wantseq: expected next sequence for reassembly.
+ * @ra_drop_time: deadline for dropping @ra_newskb.
  * @ra_runt: last pkt bytes from very end of last skb.
  * @ra_runtlen: size of ra_runt.
  */
@@ -172,6 +173,7 @@ struct xfrm_iptfs_data {
 	/* Tunnel input reassembly */
 	struct sk_buff *ra_newskb; /* new pkt being reassembled */
 	u64 ra_wantseq;		   /* expected next sequence */
+	u64 ra_drop_time;	   /* reassembly drop deadline */
 	u8 ra_runt[6];		   /* last pkt bytes from last skb */
 	u8 ra_runtlen;		   /* count of ra_runt */
 };
@@ -703,15 +705,38 @@ static void iptfs_complete_inner_skb(struct xfrm_state *x, struct sk_buff *skb)
 	}
 }
 
+/* Arm the shared timer for the earliest reassembly or reorder deadline. */
+static void iptfs_reset_drop_timer(struct xfrm_iptfs_data *xtfs)
+{
+	u64 expires = 0;
+	u64 now;
+
+	assert_spin_locked(&xtfs->drop_lock);
+
+	if (xtfs->ra_newskb)
+		expires = xtfs->ra_drop_time;
+	if (xtfs->w_savedlen &&
+	    (!expires || xtfs->w_saved[0].drop_time < expires))
+		expires = xtfs->w_saved[0].drop_time;
+	if (!expires) {
+		hrtimer_try_to_cancel(&xtfs->drop_timer);
+		return;
+	}
+
+	now = ktime_get_raw_fast_ns();
+	hrtimer_start(&xtfs->drop_timer, expires > now ? expires - now : 0,
+		      IPTFS_HRTIMER_MODE);
+}
+
 static void __iptfs_reassem_done(struct xfrm_iptfs_data *xtfs, bool free)
 {
 	assert_spin_locked(&xtfs->drop_lock);
 
-	/* We don't care if it works locking takes care of things */
-	hrtimer_try_to_cancel(&xtfs->drop_timer);
 	if (free)
 		kfree_skb(xtfs->ra_newskb);
 	xtfs->ra_newskb = NULL;
+	xtfs->ra_drop_time = 0;
+	iptfs_reset_drop_timer(xtfs);
 }
 
 /**
@@ -845,6 +870,9 @@ static u32 iptfs_reassem_cont(struct xfrm_iptfs_data *xtfs, u64 seq,
 			goto abandon;
 		}
 		xtfs->ra_newskb = newskb;
+		xtfs->ra_drop_time = ktime_get_raw_fast_ns() +
+				     xtfs->drop_time_ns;
+		iptfs_reset_drop_timer(xtfs);
 
 		/* Copy the runt data into the buffer, but leave data
 		 * pointers the same as normal non-runt case. The extra `rrem`
@@ -1162,12 +1190,9 @@ static bool __input_process_payload(struct xfrm_state *x, u32 data,
 
 			xtfs->ra_newskb = skb;
 			xtfs->ra_wantseq = seq + 1;
-			if (!hrtimer_is_queued(&xtfs->drop_timer)) {
-				/* softirq blocked lest the timer fire and interrupt us */
-				hrtimer_start(&xtfs->drop_timer,
-					      xtfs->drop_time_ns,
-					      IPTFS_HRTIMER_MODE);
-			}
+			xtfs->ra_drop_time = ktime_get_raw_fast_ns() +
+					     xtfs->drop_time_ns;
+			iptfs_reset_drop_timer(xtfs);
 
 			spin_unlock(&xtfs->drop_lock);
 
@@ -1336,7 +1361,7 @@ static u32 __reorder_drop(struct xfrm_iptfs_data *xtfs, struct list_head *list)
 	u32 scount = 0;
 
 	if (xtfs->w_saved[0].drop_time > now)
-		goto set_timer;
+		return 0;
 
 	++xtfs->w_wantseq;
 
@@ -1363,13 +1388,6 @@ static u32 __reorder_drop(struct xfrm_iptfs_data *xtfs, struct list_head *list)
 		__vec_shift(xtfs, count);
 	}
 
-	if (xtfs->w_savedlen) {
-set_timer:
-		/* Drifting is OK */
-		hrtimer_start(&xtfs->drop_timer,
-			      xtfs->w_saved[0].drop_time - now,
-			      IPTFS_HRTIMER_MODE);
-	}
 	return scount;
 }
 
@@ -1423,10 +1441,7 @@ static void iptfs_set_window_drop_times(struct xfrm_iptfs_data *xtfs, int index)
 	while (index-- > 0 && !s[index].skb)
 		s[index].drop_time = drop_time;
 
-	/* If we walked all the way back, schedule the drop timer if needed */
-	if (index == -1 && !hrtimer_is_queued(&xtfs->drop_timer))
-		hrtimer_start(&xtfs->drop_timer, xtfs->drop_time_ns,
-			      IPTFS_HRTIMER_MODE);
+	iptfs_reset_drop_timer(xtfs);
 }
 
 static void __reorder_future_fits(struct xfrm_iptfs_data *xtfs,
@@ -1660,16 +1675,15 @@ static void iptfs_input_reorder(struct xfrm_iptfs_data *xtfs,
  * The drop timer is set when we start an in progress reassembly, and also when
  * we save a future packet in the window saved array.
  *
- * NOTE packets in the save window are always newer WRT drop times as
- * they get further in the future. i.e. for:
+ * Packets in the save window are always newer WRT drop times as they get
+ * further in the future. i.e. for:
  *
  *    if slots (S0, S1, ... Sn) and `Dn` is the drop time for slot `Sn`,
  *    then D(n-1) <= D(n).
  *
- * So, regardless of why the timer is firing we can always discard any inprogress
- * fragment; either it's the reassembly timer, or slot 0 is going to be
- * dropped as S0 must have the most recent drop time, and slot 0 holds the
- * continuation fragment of the in progress packet.
+ * Reassembly and slot 0 keep independent deadlines. The shared timer is armed
+ * for the earlier one, and this callback expires only the state whose deadline
+ * has passed before rearming for any state that remains.
  *
  * Returns HRTIMER_NORESTART.
  */
@@ -1679,6 +1693,7 @@ static enum hrtimer_restart iptfs_drop_timer(struct hrtimer *me)
 	struct list_head list;
 	struct xfrm_iptfs_data *xtfs;
 	struct xfrm_state *x;
+	u64 now;
 	u32 count;
 
 	xtfs = container_of(me, typeof(*xtfs), drop_timer);
@@ -1687,15 +1702,22 @@ static enum hrtimer_restart iptfs_drop_timer(struct hrtimer *me)
 	INIT_LIST_HEAD(&list);
 
 	spin_lock(&xtfs->drop_lock);
+	now = ktime_get_raw_fast_ns();
 
-	/* Drop any in progress packet */
-	skb = xtfs->ra_newskb;
-	xtfs->ra_newskb = NULL;
+	/* Drop an in-progress packet only after its own deadline. */
+	if (xtfs->ra_newskb && xtfs->ra_drop_time <= now) {
+		skb = xtfs->ra_newskb;
+		xtfs->ra_newskb = NULL;
+		xtfs->ra_drop_time = 0;
+	} else {
+		skb = NULL;
+	}
 
 	/* Now drop as many packets as we should from the reordering window
 	 * saved array
 	 */
 	count = xtfs->w_savedlen ? __reorder_drop(xtfs, &list) : 0;
+	iptfs_reset_drop_timer(xtfs);
 
 	spin_unlock(&xtfs->drop_lock);
 
@@ -2702,6 +2724,7 @@ static int iptfs_clone_state(struct xfrm_state *x, struct xfrm_state *orig)
 	xtfs->w_savedlen = 0;
 	xtfs->ra_newskb = NULL;
 	xtfs->ra_wantseq = 0;
+	xtfs->ra_drop_time = 0;
 	xtfs->ra_runtlen = 0;
 
 	__module_get(x->mode_cbs->owner);
-- 
2.43.0