On Thu, Mar 12, 2026 at 9:07 PM Simon Schippers <[email protected]> wrote: > > This commit prevents tail-drop when a qdisc is present and the ptr_ring > becomes full. Once an entry is successfully produced and the ptr_ring > reaches capacity, the netdev queue is stopped instead of dropping > subsequent packets. > > If producing an entry fails anyways due to a race, tun_net_xmit returns > NETDEV_TX_BUSY, again avoiding a drop. Such races are expected because > LLTX is enabled and the transmit path operates without the usual locking. > > The existing __tun_wake_queue() function wakes the netdev queue. Races > between this wakeup and the queue-stop logic could leave the queue > stopped indefinitely. To prevent this, a memory barrier is enforced > (as discussed in a similar implementation in [1]), followed by a recheck > that wakes the queue if space is already available. > > If no qdisc is present, the previous tail-drop behavior is preserved.
I wonder if we need a dedicated TUN flag to enable this. With this new flag, we can even prevent TUN from using noqueue (not sure if it's possible or not). > > Benchmarks: > The benchmarks show a slight regression in raw transmission performance, > though no packets are lost anymore. > > The previously introduced threshold to only wake after the queue stopped > and half of the ring was consumed showed to be a descent choice: > Waking the queue whenever a consume made space in the ring strongly > degrades performance for tap, while waking only when the ring is empty > is too late and also hurts throughput for tap & tap+vhost-net. > Other ratios (3/4, 7/8) showed similar results (not shown here), so > 1/2 was chosen for the sake of simplicity for both tun/tap and > tun/tap+vhost-net. > > Test setup: > AMD Ryzen 5 5600X at 4.3 GHz, 3200 MHz RAM, isolated QEMU threads; > Average over 20 runs @ 100,000,000 packets. SRSO and spectre v2 > mitigations disabled. > > Note for tap+vhost-net: > XDP drop program active in VM -> ~2.5x faster, slower for tap due to > more syscalls (high utilization of entry_SYSRETQ_unsafe_stack in perf) > > +--------------------------+--------------+----------------+----------+ > | 1 thread | Stock | Patched with | diff | > | sending | | fq_codel qdisc | | > +------------+-------------+--------------+----------------+----------+ > | TAP | Transmitted | 1.151 Mpps | 1.139 Mpps | -1.1% | > | +-------------+--------------+----------------+----------+ > | | Lost/s | 3.606 Mpps | 0 pps | | > +------------+-------------+--------------+----------------+----------+ > | TAP | Transmitted | 3.948 Mpps | 3.738 Mpps | -5.3% | > | +-------------+--------------+----------------+----------+ > | +vhost-net | Lost/s | 496.5 Kpps | 0 pps | | > +------------+-------------+--------------+----------------+----------+ > > +--------------------------+--------------+----------------+----------+ > | 2 threads | Stock | Patched with | diff | > | sending | | fq_codel qdisc | | > +------------+-------------+--------------+----------------+----------+ > | TAP | Transmitted | 1.133 Mpps | 1.109 Mpps | -2.1% | > | +-------------+--------------+----------------+----------+ > | | Lost/s | 8.269 Mpps | 0 pps | | > +------------+-------------+--------------+----------------+----------+ > | TAP | Transmitted | 3.820 Mpps | 3.513 Mpps | -8.0% | > | +-------------+--------------+----------------+----------+ > | +vhost-net | Lost/s | 4.961 Mpps | 0 pps | | > +------------+-------------+--------------+----------------+----------+ > > [1] Link: https://lore.kernel.org/all/[email protected]/ > > Co-developed-by: Tim Gebauer <[email protected]> > Signed-off-by: Tim Gebauer <[email protected]> > Signed-off-by: Simon Schippers <[email protected]> > --- > drivers/net/tun.c | 30 ++++++++++++++++++++++++++++-- > 1 file changed, 28 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/tun.c b/drivers/net/tun.c > index b86582cc6cb6..9b7daec69acd 100644 > --- a/drivers/net/tun.c > +++ b/drivers/net/tun.c > @@ -1011,6 +1011,8 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, > struct net_device *dev) > struct netdev_queue *queue; > struct tun_file *tfile; > int len = skb->len; > + bool qdisc_present; > + int ret; > > rcu_read_lock(); > tfile = rcu_dereference(tun->tfiles[txq]); > @@ -1063,13 +1065,37 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, > struct net_device *dev) > > nf_reset_ct(skb); > > - if (ptr_ring_produce(&tfile->tx_ring, skb)) { > + queue = netdev_get_tx_queue(dev, txq); > + qdisc_present = !qdisc_txq_has_no_queue(queue); > + > + spin_lock(&tfile->tx_ring.producer_lock); > + ret = __ptr_ring_produce(&tfile->tx_ring, skb); > + if (__ptr_ring_produce_peek(&tfile->tx_ring) && qdisc_present) { So, it's possible that the administrator is switching between noqueue and another qdisc. So ptr_ring_produce() can fail here, do we need to check that or not? > + netif_tx_stop_queue(queue); > + /* Avoid races with queue wake-ups in __tun_wake_queue by > + * waking if space is available in a re-check. > + * The barrier makes sure that the stop is visible before > + * we re-check. > + */ > + smp_mb__after_atomic(); Let's document which barrier is paired with this. > + if (!__ptr_ring_produce_peek(&tfile->tx_ring)) > + netif_tx_wake_queue(queue); > + } > + spin_unlock(&tfile->tx_ring.producer_lock); > + > + if (ret) { > + /* If a qdisc is attached to our virtual device, > + * returning NETDEV_TX_BUSY is allowed. > + */ > + if (qdisc_present) { > + rcu_read_unlock(); > + return NETDEV_TX_BUSY; > + } > drop_reason = SKB_DROP_REASON_FULL_RING; > goto drop; > } > > /* dev->lltx requires to do our own update of trans_start */ > - queue = netdev_get_tx_queue(dev, txq); > txq_trans_cond_update(queue); > > /* Notify and wake up reader process */ > -- > 2.43.0 > Thanks
