From f51de24356e49e4dcb5095e87717065580912120 Mon Sep 17 00:00:00 2001 From: Zoltan Kiss Date: Tue, 8 Jul 2014 19:49:14 +0100 Subject: xen-netback: Adding debugfs "io_ring_qX" files This patch adds debugfs capabilities to netback. There used to be a similar patch floating around for classic kernel, but it used procfs. It is based on a very similar blkback patch. It creates xen-netback/[vifname]/io_ring_q[queueno] files, reading them output various ring variables etc. Writing "kick" into it imitates an interrupt happened, it can be useful to check whether the ring is just stalled due to a missed interrupt. Signed-off-by: Zoltan Kiss Cc: netdev@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: xen-devel@lists.xenproject.org Signed-off-by: David S. Miller --- drivers/net/xen-netback/netback.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'drivers/net/xen-netback/netback.c') diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index 1844a47636b6..77127ca08ca4 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -1987,6 +1987,13 @@ static int __init netback_init(void) rx_drain_timeout_jiffies = msecs_to_jiffies(rx_drain_timeout_msecs); +#ifdef CONFIG_DEBUG_FS + xen_netback_dbg_root = debugfs_create_dir("xen-netback", NULL); + if (IS_ERR_OR_NULL(xen_netback_dbg_root)) + pr_warn("Init of debugfs returned %ld!\n", + PTR_ERR(xen_netback_dbg_root)); +#endif /* CONFIG_DEBUG_FS */ + return 0; failed_init: @@ -1997,6 +2004,10 @@ module_init(netback_init); static void __exit netback_fini(void) { +#ifdef CONFIG_DEBUG_FS + if (!IS_ERR_OR_NULL(xen_netback_dbg_root)) + debugfs_remove_recursive(xen_netback_dbg_root); +#endif /* CONFIG_DEBUG_FS */ xenvif_xenbus_fini(); } module_exit(netback_fini); -- cgit v1.2.3 From 3d1af1df9762e56e563e8fd088a1b4ce2bcfaf8b Mon Sep 17 00:00:00 2001 From: Zoltan Kiss Date: Mon, 4 Aug 2014 16:20:57 +0100 Subject: xen-netback: Using a new state bit instead of carrier This patch introduces a new state bit VIF_STATUS_CONNECTED to track whether the vif is in a connected state. Using carrier will not work with the next patch in this series, which aims to turn the carrier temporarily off if the guest doesn't seem to be able to receive packets. Signed-off-by: Zoltan Kiss Signed-off-by: David Vrabel Cc: netdev@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: xen-devel@lists.xenproject.org v2: - rename the bitshift type to "enum state_bit_shift" here, not in the next patch Signed-off-by: David S. Miller --- drivers/net/xen-netback/netback.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/xen-netback/netback.c') diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index 769e553d3f45..6c4cc0f44da5 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -1953,7 +1953,7 @@ int xenvif_kthread_guest_rx(void *data) * context so we defer it here, if this thread is * associated with queue 0. */ - if (unlikely(queue->vif->disabled && netif_carrier_ok(queue->vif->dev) && queue->id == 0)) + if (unlikely(queue->vif->disabled && queue->id == 0)) xenvif_carrier_off(queue->vif); if (kthread_should_stop()) -- cgit v1.2.3 From f34a4cf9c9b4fd35ba7f9a596cedb011879a1a4d Mon Sep 17 00:00:00 2001 From: Zoltan Kiss Date: Mon, 4 Aug 2014 16:20:58 +0100 Subject: xen-netback: Turn off the carrier if the guest is not able to receive Currently when the guest is not able to receive more packets, qdisc layer starts a timer, and when it goes off, qdisc is started again to deliver a packet again. This is a very slow way to drain the queues, consumes unnecessary resources and slows down other guests shutdown. This patch change the behaviour by turning the carrier off when that timer fires, so all the packets are freed up which were stucked waiting for that vif. Instead of the rx_queue_purge bool it uses the VIF_STATUS_RX_PURGE_EVENT bit to signal the thread that either the timeout happened or an RX interrupt arrived, so the thread can check what it should do. It also disables NAPI, so the guest can't transmit, but leaves the interrupts on, so it can resurrect. Only the queues which brought down the interface can enable it again, the bit QUEUE_STATUS_RX_STALLED makes sure of that. Signed-off-by: Zoltan Kiss Signed-off-by: David Vrabel Cc: netdev@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: xen-devel@lists.xenproject.org Signed-off-by: David S. Miller --- drivers/net/xen-netback/netback.c | 97 +++++++++++++++++++++++++++++++++------ 1 file changed, 83 insertions(+), 14 deletions(-) (limited to 'drivers/net/xen-netback/netback.c') diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index 6c4cc0f44da5..aa2093325be1 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -1869,8 +1869,7 @@ void xenvif_idx_unmap(struct xenvif_queue *queue, u16 pending_idx) static inline int rx_work_todo(struct xenvif_queue *queue) { return (!skb_queue_empty(&queue->rx_queue) && - xenvif_rx_ring_slots_available(queue, queue->rx_last_skb_slots)) || - queue->rx_queue_purge; + xenvif_rx_ring_slots_available(queue, queue->rx_last_skb_slots)); } static inline int tx_work_todo(struct xenvif_queue *queue) @@ -1935,6 +1934,75 @@ static void xenvif_start_queue(struct xenvif_queue *queue) xenvif_wake_queue(queue); } +/* Only called from the queue's thread, it handles the situation when the guest + * doesn't post enough requests on the receiving ring. + * First xenvif_start_xmit disables QDisc and start a timer, and then either the + * timer fires, or the guest send an interrupt after posting new request. If it + * is the timer, the carrier is turned off here. + * */ +static void xenvif_rx_purge_event(struct xenvif_queue *queue) +{ + /* Either the last unsuccesful skb or at least 1 slot should fit */ + int needed = queue->rx_last_skb_slots ? + queue->rx_last_skb_slots : 1; + + /* It is assumed that if the guest post new slots after this, the RX + * interrupt will set the QUEUE_STATUS_RX_PURGE_EVENT bit and wake up + * the thread again + */ + set_bit(QUEUE_STATUS_RX_STALLED, &queue->status); + if (!xenvif_rx_ring_slots_available(queue, needed)) { + rtnl_lock(); + if (netif_carrier_ok(queue->vif->dev)) { + /* Timer fired and there are still no slots. Turn off + * everything except the interrupts + */ + netif_carrier_off(queue->vif->dev); + skb_queue_purge(&queue->rx_queue); + queue->rx_last_skb_slots = 0; + if (net_ratelimit()) + netdev_err(queue->vif->dev, "Carrier off due to lack of guest response on queue %d\n", queue->id); + } else { + /* Probably an another queue already turned the carrier + * off, make sure nothing is stucked in the internal + * queue of this queue + */ + skb_queue_purge(&queue->rx_queue); + queue->rx_last_skb_slots = 0; + } + rtnl_unlock(); + } else if (!netif_carrier_ok(queue->vif->dev)) { + unsigned int num_queues = queue->vif->num_queues; + unsigned int i; + /* The carrier was down, but an interrupt kicked + * the thread again after new requests were + * posted + */ + clear_bit(QUEUE_STATUS_RX_STALLED, + &queue->status); + rtnl_lock(); + netif_carrier_on(queue->vif->dev); + netif_tx_wake_all_queues(queue->vif->dev); + rtnl_unlock(); + + for (i = 0; i < num_queues; i++) { + struct xenvif_queue *temp = &queue->vif->queues[i]; + + xenvif_napi_schedule_or_enable_events(temp); + } + if (net_ratelimit()) + netdev_err(queue->vif->dev, "Carrier on again\n"); + } else { + /* Queuing were stopped, but the guest posted + * new requests and sent an interrupt + */ + clear_bit(QUEUE_STATUS_RX_STALLED, + &queue->status); + del_timer_sync(&queue->rx_stalled); + xenvif_start_queue(queue); + } +} + int xenvif_kthread_guest_rx(void *data) { struct xenvif_queue *queue = data; @@ -1944,8 +2012,12 @@ int xenvif_kthread_guest_rx(void *data) wait_event_interruptible(queue->wq, rx_work_todo(queue) || queue->vif->disabled || + test_bit(QUEUE_STATUS_RX_PURGE_EVENT, &queue->status) || kthread_should_stop()); + if (kthread_should_stop()) + break; + /* This frontend is found to be rogue, disable it in * kthread context. Currently this is only set when * netback finds out frontend sends malformed packet, @@ -1955,24 +2027,21 @@ int xenvif_kthread_guest_rx(void *data) */ if (unlikely(queue->vif->disabled && queue->id == 0)) xenvif_carrier_off(queue->vif); - - if (kthread_should_stop()) - break; - - if (queue->rx_queue_purge) { + else if (unlikely(test_and_clear_bit(QUEUE_STATUS_RX_PURGE_EVENT, + &queue->status))) { + xenvif_rx_purge_event(queue); + } else if (!netif_carrier_ok(queue->vif->dev)) { + /* Another queue stalled and turned the carrier off, so + * purge the internal queue of queues which were not + * blocked + */ skb_queue_purge(&queue->rx_queue); - queue->rx_queue_purge = false; + queue->rx_last_skb_slots = 0; } if (!skb_queue_empty(&queue->rx_queue)) xenvif_rx_action(queue); - if (skb_queue_empty(&queue->rx_queue) && - xenvif_queue_stopped(queue)) { - del_timer_sync(&queue->wake_queue); - xenvif_start_queue(queue); - } - cond_resched(); } -- cgit v1.2.3 From 743b0a92b92a0e1b6a68497ccd18a0d60a4b6082 Mon Sep 17 00:00:00 2001 From: Zoltan Kiss Date: Thu, 7 Aug 2014 13:38:38 +0100 Subject: xen-netback: Fix vif->disable handling In the patch called "xen-netback: Turn off the carrier if the guest is not able to receive" new branches were introduced to this if statement, risking that a queue with non-zero id can reenable the disabled interface. Signed-off-by: Zoltan Kiss Signed-off-by: David Vrabel Cc: netdev@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: xen-devel@lists.xenproject.org Signed-off-by: David S. Miller --- drivers/net/xen-netback/netback.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'drivers/net/xen-netback/netback.c') diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index aa2093325be1..4734472aa620 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -2025,9 +2025,15 @@ int xenvif_kthread_guest_rx(void *data) * context so we defer it here, if this thread is * associated with queue 0. */ - if (unlikely(queue->vif->disabled && queue->id == 0)) + if (unlikely(queue->vif->disabled && queue->id == 0)) { xenvif_carrier_off(queue->vif); - else if (unlikely(test_and_clear_bit(QUEUE_STATUS_RX_PURGE_EVENT, + } else if (unlikely(queue->vif->disabled)) { + /* kthread_stop() would be called upon this thread soon, + * be a bit proactive + */ + skb_queue_purge(&queue->rx_queue); + queue->rx_last_skb_slots = 0; + } else if (unlikely(test_and_clear_bit(QUEUE_STATUS_RX_PURGE_EVENT, &queue->status))) { xenvif_rx_purge_event(queue); } else if (!netif_carrier_ok(queue->vif->dev)) { -- cgit v1.2.3 From a64bd934528e26e8956112e43a279fba2ee0634e Mon Sep 17 00:00:00 2001 From: Wei Liu Date: Tue, 12 Aug 2014 11:48:07 +0100 Subject: xen-netback: don't stop dealloc kthread too early Reference count the number of packets in host stack, so that we don't stop the deallocation thread too early. If not, we can end up with xenvif_free permanently waiting for deallocation thread to unmap grefs. Reported-by: Thomas Leonard Signed-off-by: Wei Liu Cc: Ian Campbell Cc: Zoltan Kiss Signed-off-by: David S. Miller --- drivers/net/xen-netback/netback.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) (limited to 'drivers/net/xen-netback/netback.c') diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index 4734472aa620..08f65996534c 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -1525,10 +1525,12 @@ static int xenvif_handle_frag_list(struct xenvif_queue *queue, struct sk_buff *s /* remove traces of mapped pages and frag_list */ skb_frag_list_init(skb); uarg = skb_shinfo(skb)->destructor_arg; + /* increase inflight counter to offset decrement in callback */ + atomic_inc(&queue->inflight_packets); uarg->callback(uarg, true); skb_shinfo(skb)->destructor_arg = NULL; - skb_shinfo(nskb)->tx_flags |= SKBTX_DEV_ZEROCOPY; + xenvif_skb_zerocopy_prepare(queue, nskb); kfree_skb(nskb); return 0; @@ -1589,7 +1591,7 @@ static int xenvif_tx_submit(struct xenvif_queue *queue) if (net_ratelimit()) netdev_err(queue->vif->dev, "Not enough memory to consolidate frag_list!\n"); - skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; + xenvif_skb_zerocopy_prepare(queue, skb); kfree_skb(skb); continue; } @@ -1609,7 +1611,7 @@ static int xenvif_tx_submit(struct xenvif_queue *queue) "Can't setup checksum in net_tx_action\n"); /* We have to set this flag to trigger the callback */ if (skb_shinfo(skb)->destructor_arg) - skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; + xenvif_skb_zerocopy_prepare(queue, skb); kfree_skb(skb); continue; } @@ -1641,7 +1643,7 @@ static int xenvif_tx_submit(struct xenvif_queue *queue) * skb. E.g. the __pskb_pull_tail earlier can do such thing. */ if (skb_shinfo(skb)->destructor_arg) { - skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; + xenvif_skb_zerocopy_prepare(queue, skb); queue->stats.tx_zerocopy_sent++; } @@ -1681,6 +1683,7 @@ void xenvif_zerocopy_callback(struct ubuf_info *ubuf, bool zerocopy_success) queue->stats.tx_zerocopy_success++; else queue->stats.tx_zerocopy_fail++; + xenvif_skb_zerocopy_complete(queue); } static inline void xenvif_tx_dealloc_action(struct xenvif_queue *queue) @@ -2058,15 +2061,24 @@ int xenvif_kthread_guest_rx(void *data) return 0; } +static bool xenvif_dealloc_kthread_should_stop(struct xenvif_queue *queue) +{ + /* Dealloc thread must remain running until all inflight + * packets complete. + */ + return kthread_should_stop() && + !atomic_read(&queue->inflight_packets); +} + int xenvif_dealloc_kthread(void *data) { struct xenvif_queue *queue = data; - while (!kthread_should_stop()) { + for (;;) { wait_event_interruptible(queue->dealloc_wq, tx_dealloc_work_todo(queue) || - kthread_should_stop()); - if (kthread_should_stop()) + xenvif_dealloc_kthread_should_stop(queue)); + if (xenvif_dealloc_kthread_should_stop(queue)) break; xenvif_tx_dealloc_action(queue); -- cgit v1.2.3