sched: Fix qdisc interactions exposed by using sch_cake as a leaf qdisc
Subsystems (5)
v2 cover sched: Fix qdisc interactions exposed by using sch_cake as a leaf qdisc
This series fixes a couple of issues exposed by running sch_cake as a leaf qdisc in an HFSC tree, which were discovered and reported by Pete Heist. The interaction between CAKE's GSO splitting and the parent qdisc's notion of its own queue length could cause queue stalls. While investigating the report, I also noticed that several qdiscs would dereference the skb pointer after dequeue, which is potentially problematic since the GSO splitting code also frees the original skb.
See the individual patches in the series for details.
Toke Høiland-Jørgensen (3):
sched: Avoid dereferencing skb pointer after child enqueue
sched: Fix detection of empty queues in child qdiscs
sch_cake: Correctly update parent qlen when splitting GSO packets
net/sched/sch_cake.c | 5 +++-- net/sched/sch_cbs.c | 3 ++- net/sched/sch_drr.c | 7 +++++-- net/sched/sch_dsmark.c | 3 ++- net/sched/sch_hfsc.c | 9 +++++---- net/sched/sch_htb.c | 3 ++- net/sched/sch_prio.c | 3 ++- net/sched/sch_qfq.c | 20 ++++++++++++-------- net/sched/sch_tbf.c | 3 ++- 9 files changed, 35 insertions(+), 21 deletions(-)
Signature · 1 line
2.19.1
PATCH v2 1/3 sched: Avoid dereferencing skb pointer after child enqueue
Parent qdiscs may dereference the pointer to the enqueued skb after enqueue. However, both CAKE and TBF call consume_skb() on the original skb when splitting GSO packets, leading to a potential use-after-free in the parent. Fix this by avoiding dereferencing the skb pointer after enqueueing to the child.
--- net/sched/sch_cbs.c | 3 ++- net/sched/sch_drr.c | 3 ++- net/sched/sch_dsmark.c | 3 ++- net/sched/sch_hfsc.c | 5 ++--- net/sched/sch_htb.c | 3 ++- net/sched/sch_prio.c | 3 ++- net/sched/sch_qfq.c | 16 +++++++++------- net/sched/sch_tbf.c | 3 ++- 8 files changed, 23 insertions(+), 16 deletions(-)
net/sched/sch_cbs.c +2 -1
@@ -88,13 +88,14 @@ static int cbs_child_enqueue(struct sk_buff *skb, struct Qdisc *sch,struct Qdisc *child, struct sk_buff **to_free) { + unsigned int len = qdisc_pkt_len(skb);int err; err = child->ops->enqueue (skb, child, to_free);if (err !=NET_XMIT_SUCCESS )return err; -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ;
net/sched/sch_drr.c +2 -1
@@ -350,6 +350,7 @@ static struct drr_class *drr_classify(struct sk_buff *skb, struct Qdisc *sch,static int drr_enqueue (struct sk_buff *skb,struct Qdisc *sch,struct sk_buff **to_free) { +unsigned int len =qdisc_pkt_len (skb);struct drr_sched *q =qdisc_priv (sch);struct drr_class *cl;int err =0 ;
@@ -376,7 +377,7 @@ static int drr_enqueue(struct sk_buff *skb, struct Qdisc *sch, cl->deficit = cl->quantum; } -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return err; }
net/sched/sch_dsmark.c +2 -1
@@ -199,6 +199,7 @@ static struct tcf_block *dsmark_tcf_block(struct Qdisc *sch, unsigned long cl,static int dsmark_enqueue (struct sk_buff *skb,struct Qdisc *sch,struct sk_buff **to_free) { +unsigned int len =qdisc_pkt_len (skb);struct dsmark_qdisc_data *p =qdisc_priv (sch);int err;
@@ -271,7 +272,7 @@ static int dsmark_enqueue(struct sk_buff *skb, struct Qdisc *sch,return err; } -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ;
net/sched/sch_hfsc.c +2 -3
@@ -1539,6 +1539,7 @@ hfsc_dump_qdisc(struct Qdisc *sch, struct sk_buff *skb)static int hfsc_enqueue (struct sk_buff *skb,struct Qdisc *sch,struct sk_buff **to_free) { +unsigned int len =qdisc_pkt_len (skb);struct hfsc_class *cl;int uninitialized_var (err );
@@ -1560,8 +1561,6 @@ hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) }if (cl->qdisc->q.qlen ==1 ) { -unsigned int len =qdisc_pkt_len (skb); -if (cl->cl_flags &HFSC_RSC )init_ed (cl, len);if (cl->cl_flags &HFSC_FSC )
@@ -1576,7 +1575,7 @@ hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) } -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ;
net/sched/sch_htb.c +2 -1
@@ -581,6 +581,7 @@ static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) {int uninitialized_var (ret ); +unsigned int len =qdisc_pkt_len (skb);struct htb_sched *q =qdisc_priv (sch);struct htb_class *cl =htb_classify (skb, sch, &ret);
@@ -610,7 +611,7 @@ static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch,htb_activate (q, cl); } -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ; }
net/sched/sch_prio.c +2 -1
@@ -72,6 +72,7 @@ prio_classify(struct sk_buff *skb, struct Qdisc *sch, int *qerr)static int prio_enqueue (struct sk_buff *skb,struct Qdisc *sch,struct sk_buff **to_free) { +unsigned int len =qdisc_pkt_len (skb);struct Qdisc *qdisc;int ret;
@@ -88,7 +89,7 @@ prio_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) ret =qdisc_enqueue (skb, qdisc, to_free);if (ret ==NET_XMIT_SUCCESS ) { -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ; }
net/sched/sch_qfq.c +9 -7
@@ -1210,6 +1210,7 @@ static struct qfq_aggregate *qfq_choose_next_agg(struct qfq_sched *q)static int qfq_enqueue (struct sk_buff *skb,struct Qdisc *sch,struct sk_buff **to_free) { +unsigned int len =qdisc_pkt_len (skb), gso_segs;struct qfq_sched *q =qdisc_priv (sch);struct qfq_class *cl;struct qfq_aggregate *agg;
@@ -1224,17 +1225,17 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, }pr_debug ("qfq_enqueue: cl = %x\n" , cl->common.classid); -if (unlikely (cl->agg->lmax <qdisc_pkt_len (skb))) { +if (unlikely (cl->agg->lmax < len)) {pr_debug ("qfq: increasing maxpkt from %u to %u for class %u" , - cl->agg->lmax,qdisc_pkt_len (skb), cl->common.classid); - err =qfq_change_agg (sch, cl, cl->agg->class_weight, -qdisc_pkt_len (skb)); + cl->agg->lmax, len, cl->common.classid); + err =qfq_change_agg (sch, cl, cl->agg->class_weight, len);if (err) { cl->qstats.drops++;return qdisc_drop (skb, sch, to_free); } } + gso_segs =skb_is_gso (skb) ?skb_shinfo (skb)->gso_segs :1 ; err =qdisc_enqueue (skb, cl->qdisc, to_free);if (unlikely (err !=NET_XMIT_SUCCESS )) {pr_debug ("qfq_enqueue: enqueue failed %d\n" , err);
@@ -1245,8 +1246,9 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch,return err; } -bstats_update (&cl->bstats, skb); -qdisc_qstats_backlog_inc (sch, skb); + cl->bstats.bytes += len; + cl->bstats.packets += gso_segs; + sch->qstats.backlog += len; ++sch->q.qlen; agg = cl->agg;
@@ -1254,7 +1256,7 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch,if (cl->qdisc->q.qlen !=1 ) {if (unlikely (skb == cl->qdisc->ops->peek (cl->qdisc)) && list_first_entry(&agg->active, struct qfq_class, alist) - == cl && cl->deficit <qdisc_pkt_len (skb)) + == cl && cl->deficit < len)list_move_tail (&cl->alist, &agg->active);return err;
net/sched/sch_tbf.c +2 -1
@@ -185,6 +185,7 @@ static int tbf_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) {struct tbf_sched_data *q =qdisc_priv (sch); +unsigned int len =qdisc_pkt_len (skb);int ret;if (qdisc_pkt_len (skb) > q->max_size) {
@@ -200,7 +201,7 @@ static int tbf_enqueue(struct sk_buff *skb, struct Qdisc *sch,return ret; } -qdisc_qstats_backlog_inc (sch, skb); + sch->qstats.backlog += len; sch->q.qlen++;return NET_XMIT_SUCCESS ; }
Signature · 1 line
2.19.1
PATCH v2 2/3 sched: Fix detection of empty queues in child qdiscs
In reply to Toke Høiland-JørgensenSeveral qdiscs check on enqueue whether the packet was enqueued to a class with an empty queue, in which case the class is activated. This is done by checking if the qlen is exactly 1 after enqueue. However, if GSO splitting is enabled in the child qdisc, a single packet can result in a qlen longer than 1. This means the activation check fails, leading to a stalled queue.
Fix this by checking if the queue is empty *before* enqueue, and running the activation logic if this was the case.
--- net/sched/sch_drr.c | 4 +++- net/sched/sch_hfsc.c | 4 +++- net/sched/sch_qfq.c | 4 +++- 3 files changed, 9 insertions(+), 3 deletions(-)
net/sched/sch_drr.c +3 -1
@@ -354,6 +354,7 @@ static int drr_enqueue(struct sk_buff *skb, struct Qdisc *sch,struct drr_sched *q =qdisc_priv (sch);struct drr_class *cl;int err =0 ; +bool first; cl =drr_classify (skb, sch, &err);if (cl ==NULL ) {
@@ -363,6 +364,7 @@ static int drr_enqueue(struct sk_buff *skb, struct Qdisc *sch,return err; } + first = !cl->qdisc->q.qlen; err =qdisc_enqueue (skb, cl->qdisc, to_free);if (unlikely (err !=NET_XMIT_SUCCESS )) {if (net_xmit_drop_count (err)) {
@@ -372,7 +374,7 @@ static int drr_enqueue(struct sk_buff *skb, struct Qdisc *sch,return err; } -if (cl->qdisc->q.qlen ==1 ) { +if (first) {list_add_tail (&cl->alist, &q->active); cl->deficit = cl->quantum; }
net/sched/sch_hfsc.c +3 -1
@@ -1542,6 +1542,7 @@ hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free)unsigned int len =qdisc_pkt_len (skb);struct hfsc_class *cl;int uninitialized_var (err ); +bool first; cl =hfsc_classify (skb, sch, &err);if (cl ==NULL ) {
@@ -1551,6 +1552,7 @@ hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free)return err; } + first = !cl->qdisc->q.qlen; err =qdisc_enqueue (skb, cl->qdisc, to_free);if (unlikely (err !=NET_XMIT_SUCCESS )) {if (net_xmit_drop_count (err)) {
@@ -1560,7 +1562,7 @@ hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free)return err; } -if (cl->qdisc->q.qlen ==1 ) { +if (first) {if (cl->cl_flags &HFSC_RSC )init_ed (cl, len);if (cl->cl_flags &HFSC_FSC )
net/sched/sch_qfq.c +3 -1
@@ -1215,6 +1215,7 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch,struct qfq_class *cl;struct qfq_aggregate *agg;int err =0 ; +bool first; cl =qfq_classify (skb, sch, &err);if (cl ==NULL ) {
@@ -1236,6 +1237,7 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, } gso_segs =skb_is_gso (skb) ?skb_shinfo (skb)->gso_segs :1 ; + first = !cl->qdisc->q.qlen; err =qdisc_enqueue (skb, cl->qdisc, to_free);if (unlikely (err !=NET_XMIT_SUCCESS )) {pr_debug ("qfq_enqueue: enqueue failed %d\n" , err);
@@ -1253,7 +1255,7 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, agg = cl->agg;/* if the queue was not empty, then done here */ -if (cl->qdisc->q.qlen !=1 ) { +if (!first) {if (unlikely (skb == cl->qdisc->ops->peek (cl->qdisc)) && list_first_entry(&agg->active, struct qfq_class, alist) == cl && cl->deficit < len)
Signature · 1 line
2.19.1
PATCH v2 3/3 sch_cake: Correctly update parent qlen when splitting GSO packets
In reply to Toke Høiland-JørgensenTo ensure parent qdiscs have the same notion of the number of enqueued packets even after splitting a GSO packet, update the qdisc tree with the number of packets that was added due to the split.
--- net/sched/sch_cake.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
net/sched/sch_cake.c +3 -2
@@ -1667,7 +1667,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,if (skb_is_gso (skb) && q->rate_flags &CAKE_FLAG_SPLIT_GSO ) {struct sk_buff *segs, *nskb;netdev_features_t features =netif_skb_features (skb); -unsigned int slen =0 ; +unsigned int slen =0 , numsegs =0 ; segs =skb_gso_segment (skb, features & ~NETIF_F_GSO_MASK );if (IS_ERR_OR_NULL (segs))
@@ -1683,6 +1683,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,flow_queue_add (flow, segs); sch->q.qlen++; + numsegs++; slen += segs->len; q->buffer_used += segs->truesize; b->packets++;
@@ -1696,7 +1697,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch, sch->qstats.backlog += slen; q->avg_window_bytes += slen; -qdisc_tree_reduce_backlog (sch,1 , len); +qdisc_tree_reduce_backlog (sch,1 -numsegs, len-slen);consume_skb (skb); }else {/* not splitting */
Signature · 1 line
2.19.1
David Miller
maintainerFrom: Toke Høiland-Jørgensen <[email protected]>
Date: Wed, 9 Jan 2019 17:09:40 +0100
Quoted 9 lines
This series fixes a couple of issues exposed by running
sch_cakeas a leaf qdisc in an HFSC tree, which were discovered and reported by Pete Heist. The interaction between CAKE's GSO splitting and the parent qdisc's notion of its own queue length could cause queue stalls. While investigating the report, I also noticed that several qdiscs would dereference the skb pointer after dequeue, which is potentially problematic since the GSO splitting code also frees the original skb.See the individual patches in the series for details.
Series applied.