[PATCH 00/12] net-next: mediatek: bug fixes and tx stall fixes

[PATCH 00/12] net-next: mediatek: bug fixes and tx stall fixes

[John Crispin]

This series is a collection of fixes for bugs that we stumble across while
doing more performance testing and code level review.

John Crispin (12):
  net-next: mediatek: fix DQL support
  net-next: mediatek: add missing return code check
  net-next: mediatek: fix missing free of scratch memory
  net-next: mediatek: invalid buffer lookup in mtk_tx_map()
  net-next: mediatek: dropped rx packets are not being counted properly
  net-next: mediatek: add next data pointer coherency protection
  net-next: mediatek: disable all interrupts during probe
  net-next: mediatek: fix threshold value
  net-next: mediatek: increase watchdog_timeo
  net-next: mediatek: fix off by one in the TX ring allocation
  net-next: mediatek: only wake the queue if it is stopped
  net-next: mediatek: remove superfluous queue wake up call

 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   94 ++++++++++++++++++---------
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |    3 +
 2 files changed, 65 insertions(+), 32 deletions(-)

-- 
1.7.10.4

[PATCH 01/12] net-next: mediatek: fix DQL support

[John Crispin]

The MTK ethernet core has 2 MACs both sitting on the same DMA ring. The
current code will assign the TX traffic of each MAC to its own DQL. This
results in the amount of data, that DQL says is in the queue incorrect. As
the data from multiple devices is infact enqueued. This makes any decision
based on these value non deterministic. Fix this by tracking all TX
traffic, regardless of the MAC it belongs to in the DQL of all devices
using the DMA.

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   33 ++++++++++++++++-----------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 41cdc0d..07674e5 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -641,7 +641,16 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 	WRITE_ONCE(itxd->txd3, (TX_DMA_SWC | TX_DMA_PLEN0(skb_headlen(skb)) |
 				(!nr_frags * TX_DMA_LS0)));
 
-	netdev_sent_queue(dev, skb->len);
+	/* we have a single DMA ring so BQL needs to be updated for all devices
+	 * sitting on this ring
+	 */
+	for (i = 0; i < MTK_MAC_COUNT; i++) {
+		if (!eth->netdev[i])
+			continue;
+
+		netdev_sent_queue(eth->netdev[i], skb->len);
+	}
+
 	skb_tx_timestamp(skb);
 
 	ring->next_free = mtk_qdma_phys_to_virt(ring, txd->txd2);
@@ -869,21 +878,18 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget, bool *tx_again)
 	struct mtk_tx_dma *desc;
 	struct sk_buff *skb;
 	struct mtk_tx_buf *tx_buf;
-	int total = 0, done[MTK_MAX_DEVS];
-	unsigned int bytes[MTK_MAX_DEVS];
+	int total = 0, done = 0;
+	unsigned int bytes = 0;
 	u32 cpu, dma;
 	static int condition;
 	int i;
 
-	memset(done, 0, sizeof(done));
-	memset(bytes, 0, sizeof(bytes));
-
 	cpu = mtk_r32(eth, MTK_QTX_CRX_PTR);
 	dma = mtk_r32(eth, MTK_QTX_DRX_PTR);
 
 	desc = mtk_qdma_phys_to_virt(ring, cpu);
 
-	while ((cpu != dma) && budget) {
+	while ((cpu != dma) && done < budget) {
 		u32 next_cpu = desc->txd2;
 		int mac;
 
@@ -903,9 +909,8 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget, bool *tx_again)
 		}
 
 		if (skb != (struct sk_buff *)MTK_DMA_DUMMY_DESC) {
-			bytes[mac] += skb->len;
-			done[mac]++;
-			budget--;
+			bytes += skb->len;
+			done++;
 		}
 		mtk_tx_unmap(eth->dev, tx_buf);
 
@@ -918,11 +923,13 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget, bool *tx_again)
 
 	mtk_w32(eth, cpu, MTK_QTX_CRX_PTR);
 
+	/* we have a single DMA ring so BQL needs to be updated for all devices
+	 * sitting on this ring
+	 */
 	for (i = 0; i < MTK_MAC_COUNT; i++) {
-		if (!eth->netdev[i] || !done[i])
+		if (!eth->netdev[i])
 			continue;
-		netdev_completed_queue(eth->netdev[i], done[i], bytes[i]);
-		total += done[i];
+		netdev_completed_queue(eth->netdev[i], done, bytes);
 	}
 
 	/* read hw index again make sure no new tx packet */
-- 
1.7.10.4

[PATCH 02/12] net-next: mediatek: add missing return code check

[John Crispin]

The code fails to check if the scratch memory was properly allocated. Add
this check and return with an error if the allocation failed.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 07674e5..0ed49a4 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -483,6 +483,9 @@ static int mtk_init_fq_dma(struct mtk_eth *eth)
 
 	eth->scratch_head = kcalloc(cnt, MTK_QDMA_PAGE_SIZE,
 				    GFP_KERNEL);
+	if (unlikely(!eth->scratch_head))
+		return -ENOMEM;
+
 	dma_addr = dma_map_single(eth->dev,
 				  eth->scratch_head, cnt * MTK_QDMA_PAGE_SIZE,
 				  DMA_FROM_DEVICE);
-- 
1.7.10.4

[PATCH 03/12] net-next: mediatek: fix missing free of scratch memory

[John Crispin]

Scratch memory gets allocated in mtk_init_fq_dma() but the corresponding
code to free it is missing inside mtk_dma_free() causing a memory leak.
With this patch applied, we can run ifconfig/up/down several thousand
times without any problems.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   18 +++++++++++++-----
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |    2 ++
 2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 0ed49a4..e415986 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -469,14 +469,14 @@ static inline void mtk_rx_get_desc(struct mtk_rx_dma *rxd,
 /* the qdma core needs scratch memory to be setup */
 static int mtk_init_fq_dma(struct mtk_eth *eth)
 {
-	dma_addr_t phy_ring_head, phy_ring_tail;
+	dma_addr_t phy_ring_tail;
 	int cnt = MTK_DMA_SIZE;
 	dma_addr_t dma_addr;
 	int i;
 
 	eth->scratch_ring = dma_alloc_coherent(eth->dev,
 					       cnt * sizeof(struct mtk_tx_dma),
-					       &phy_ring_head,
+					       &eth->phy_scratch_ring,
 					       GFP_ATOMIC | __GFP_ZERO);
 	if (unlikely(!eth->scratch_ring))
 		return -ENOMEM;
@@ -493,19 +493,19 @@ static int mtk_init_fq_dma(struct mtk_eth *eth)
 		return -ENOMEM;
 
 	memset(eth->scratch_ring, 0x0, sizeof(struct mtk_tx_dma) * cnt);
-	phy_ring_tail = phy_ring_head +
+	phy_ring_tail = eth->phy_scratch_ring +
 			(sizeof(struct mtk_tx_dma) * (cnt - 1));
 
 	for (i = 0; i < cnt; i++) {
 		eth->scratch_ring[i].txd1 =
 					(dma_addr + (i * MTK_QDMA_PAGE_SIZE));
 		if (i < cnt - 1)
-			eth->scratch_ring[i].txd2 = (phy_ring_head +
+			eth->scratch_ring[i].txd2 = (eth->phy_scratch_ring +
 				((i + 1) * sizeof(struct mtk_tx_dma)));
 		eth->scratch_ring[i].txd3 = TX_DMA_SDL(MTK_QDMA_PAGE_SIZE);
 	}
 
-	mtk_w32(eth, phy_ring_head, MTK_QDMA_FQ_HEAD);
+	mtk_w32(eth, eth->phy_scratch_ring, MTK_QDMA_FQ_HEAD);
 	mtk_w32(eth, phy_ring_tail, MTK_QDMA_FQ_TAIL);
 	mtk_w32(eth, (cnt << 16) | cnt, MTK_QDMA_FQ_CNT);
 	mtk_w32(eth, MTK_QDMA_PAGE_SIZE << 16, MTK_QDMA_FQ_BLEN);
@@ -1205,6 +1205,14 @@ static void mtk_dma_free(struct mtk_eth *eth)
 	for (i = 0; i < MTK_MAC_COUNT; i++)
 		if (eth->netdev[i])
 			netdev_reset_queue(eth->netdev[i]);
+	if (eth->scratch_ring) {
+		dma_free_coherent(eth->dev,
+				  MTK_DMA_SIZE * sizeof(struct mtk_tx_dma),
+				  eth->scratch_ring,
+				  eth->phy_scratch_ring);
+		eth->scratch_ring = NULL;
+		eth->phy_scratch_ring = 0;
+	}
 	mtk_tx_clean(eth);
 	mtk_rx_clean(eth);
 	kfree(eth->scratch_head);
diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index eed626d..57f7e8a 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -357,6 +357,7 @@ struct mtk_rx_ring {
  * @rx_ring:		Pointer to the memore holding info about the RX ring
  * @rx_napi:		The NAPI struct
  * @scratch_ring:	Newer SoCs need memory for a second HW managed TX ring
+ * @phy_scratch_ring:	physical address of scratch_ring
  * @scratch_head:	The scratch memory that scratch_ring points to.
  * @clk_ethif:		The ethif clock
  * @clk_esw:		The switch clock
@@ -384,6 +385,7 @@ struct mtk_eth {
 	struct mtk_rx_ring		rx_ring;
 	struct napi_struct		rx_napi;
 	struct mtk_tx_dma		*scratch_ring;
+	dma_addr_t			phy_scratch_ring;
 	void				*scratch_head;
 	struct clk			*clk_ethif;
 	struct clk			*clk_esw;
-- 
1.7.10.4

[PATCH 04/12] net-next: mediatek: invalid buffer lookup in mtk_tx_map()

[John Crispin]

The lookup of the tx_buffer in the error path inside mtk_tx_map() uses the
wrong descriptor pointer. This looks like a copy & paste error. Change the
code to use the correct pointer.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index e415986..7fdd1a3 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -671,7 +671,7 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 
 err_dma:
 	do {
-		tx_buf = mtk_desc_to_tx_buf(ring, txd);
+		tx_buf = mtk_desc_to_tx_buf(ring, itxd);
 
 		/* unmap dma */
 		mtk_tx_unmap(&dev->dev, tx_buf);
-- 
1.7.10.4

[PATCH 05/12] net-next: mediatek: dropped rx packets are not being counted properly

[John Crispin]

There are 2 places inside mtk_poll_rx where rx_dropped is not being
incremented properly. Fix this by adding the missing code to increment
the counter.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 7fdd1a3..2e79af9 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -826,6 +826,7 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 					  DMA_FROM_DEVICE);
 		if (unlikely(dma_mapping_error(&netdev->dev, dma_addr))) {
 			skb_free_frag(new_data);
+			netdev->stats.rx_dropped++;
 			goto release_desc;
 		}
 
@@ -833,6 +834,7 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 		skb = build_skb(data, ring->frag_size);
 		if (unlikely(!skb)) {
 			put_page(virt_to_head_page(new_data));
+			netdev->stats.rx_dropped++;
 			goto release_desc;
 		}
 		skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
-- 
1.7.10.4

[PATCH 06/12] net-next: mediatek: add next data pointer coherency protection

[John Crispin]

The QDMA engine can fail to update the register pointing to the next TX
descriptor if this bit does not get set in the QDMA configuration register.
Not setting this bit can result in invalid values inside the TX rings
registers which will causes TX stalls.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    2 +-
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |    1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 2e79af9..60095be 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -1277,7 +1277,7 @@ static int mtk_start_dma(struct mtk_eth *eth)
 	mtk_w32(eth,
 		MTK_TX_WB_DDONE | MTK_RX_DMA_EN | MTK_TX_DMA_EN |
 		MTK_RX_2B_OFFSET | MTK_DMA_SIZE_16DWORDS |
-		MTK_RX_BT_32DWORDS,
+		MTK_RX_BT_32DWORDS | MTK_NDP_CO_PRO,
 		MTK_QDMA_GLO_CFG);
 
 	return 0;
diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index 57f7e8a..a5eb7c6 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -91,6 +91,7 @@
 #define MTK_QDMA_GLO_CFG	0x1A04
 #define MTK_RX_2B_OFFSET	BIT(31)
 #define MTK_RX_BT_32DWORDS	(3 << 11)
+#define MTK_NDP_CO_PRO		BIT(10)
 #define MTK_TX_WB_DDONE		BIT(6)
 #define MTK_DMA_SIZE_16DWORDS	(2 << 4)
 #define MTK_RX_DMA_BUSY		BIT(3)
-- 
1.7.10.4

[PATCH 07/12] net-next: mediatek: disable all interrupts during probe

[John Crispin]

The current code only disables those IRQs that we will later use. To
ensure that we have a predefined state, we really want to disable all IRQs.
Change the code to disable all IRQs to achieve this.

Signed-off-by: John Crispin <john-Pj+rj9U5foFAfugRpC6u6w@public.gmane.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 60095be..08f3df4 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -1391,7 +1391,7 @@ static int __init mtk_hw_init(struct mtk_eth *eth)
 
 	/* disable delay and normal interrupt */
 	mtk_w32(eth, 0, MTK_QDMA_DELAY_INT);
-	mtk_irq_disable(eth, MTK_TX_DONE_INT | MTK_RX_DONE_INT);
+	mtk_irq_disable(eth, ~0);
 	mtk_w32(eth, RST_GL_PSE, MTK_RST_GL);
 	mtk_w32(eth, 0, MTK_RST_GL);
 
-- 
1.7.10.4

[PATCH 08/12] net-next: mediatek: fix threshold value

[John Crispin]

The logic to calculate the threshold value for stopping the TX queue is
bad. Currently it will always use 1/2 of the rings size, which is way too
much. Set the threshold to MAX_SKB_FRAGS. This makes sure that the queue
is stopped when there is not enough room to accept an additional segment. 

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 08f3df4..4439991 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -1028,8 +1028,7 @@ static int mtk_tx_alloc(struct mtk_eth *eth)
 	atomic_set(&ring->free_count, MTK_DMA_SIZE - 2);
 	ring->next_free = &ring->dma[0];
 	ring->last_free = &ring->dma[MTK_DMA_SIZE - 2];
-	ring->thresh = max((unsigned long)MTK_DMA_SIZE >> 2,
-			      MAX_SKB_FRAGS);
+	ring->thresh = MAX_SKB_FRAGS;
 
 	/* make sure that all changes to the dma ring are flushed before we
 	 * continue
-- 
1.7.10.4


_______________________________________________
Linux-mediatek mailing list
Linux-mediatek@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-mediatek

[PATCH 09/12] net-next: mediatek: increase watchdog_timeo

[John Crispin]

During stress testing, after reducing the threshold value, we have seen
TX timeouts that were caused by the watchdog_timeo value being too low.
Increase the value to 5 * HZ which is a value commonly used by many other
drivers.

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 4439991..12e9f1a 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -1704,7 +1704,7 @@ static int mtk_add_mac(struct mtk_eth *eth, struct device_node *np)
 	mac->hw_stats->reg_offset = id * MTK_STAT_OFFSET;
 
 	SET_NETDEV_DEV(eth->netdev[id], eth->dev);
-	eth->netdev[id]->watchdog_timeo = HZ;
+	eth->netdev[id]->watchdog_timeo = 5 * HZ;
 	eth->netdev[id]->netdev_ops = &mtk_netdev_ops;
 	eth->netdev[id]->base_addr = (unsigned long)eth->base;
 	eth->netdev[id]->vlan_features = MTK_HW_FEATURES &
-- 
1.7.10.4

[PATCH 10/12] net-next: mediatek: fix off by one in the TX ring allocation

[John Crispin]

The TX ring setup has an off by one error causing it to not utilise all
descriptors. This has the side effect that we need to reset the next
pointer at runtime to make it work. Fix the off by one and remove the
code fixing the ring at runtime.

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 12e9f1a..81155a1 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -919,7 +919,6 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget, bool *tx_again)
 		}
 		mtk_tx_unmap(eth->dev, tx_buf);
 
-		ring->last_free->txd2 = next_cpu;
 		ring->last_free = desc;
 		atomic_inc(&ring->free_count);
 
@@ -1027,7 +1026,7 @@ static int mtk_tx_alloc(struct mtk_eth *eth)
 
 	atomic_set(&ring->free_count, MTK_DMA_SIZE - 2);
 	ring->next_free = &ring->dma[0];
-	ring->last_free = &ring->dma[MTK_DMA_SIZE - 2];
+	ring->last_free = &ring->dma[MTK_DMA_SIZE - 1];
 	ring->thresh = MAX_SKB_FRAGS;
 
 	/* make sure that all changes to the dma ring are flushed before we
-- 
1.7.10.4

[PATCH 11/12] net-next: mediatek: only wake the queue if it is stopped

[John Crispin]

The current code unconditionally wakes up the queue at the end of each
tx_poll action. Change the code to only wake up the queues if any of
them have actually been stopped before.

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 81155a1..2d3c464 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -701,6 +701,20 @@ static inline int mtk_cal_txd_req(struct sk_buff *skb)
 	return nfrags;
 }
 
+static int mtk_queue_stopped(struct mtk_eth *eth)
+{
+	int i;
+
+	for (i = 0; i < MTK_MAC_COUNT; i++) {
+		if (!eth->netdev[i])
+			continue;
+		if (netif_queue_stopped(eth->netdev[i]))
+			return 1;
+	}
+
+	return 0;
+}
+
 static void mtk_wake_queue(struct mtk_eth *eth)
 {
 	int i;
@@ -945,7 +959,8 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget, bool *tx_again)
 	if (!total)
 		return 0;
 
-	if (atomic_read(&ring->free_count) > ring->thresh)
+	if (mtk_queue_stopped(eth) &&
+	    (atomic_read(&ring->free_count) > ring->thresh))
 		mtk_wake_queue(eth);
 
 	return total;
-- 
1.7.10.4

[PATCH 12/12] net-next: mediatek: remove superfluous queue wake up call

[John Crispin]

The code checks if the queue should be stopped because we are below the
threshold of free descriptors only to check if it should be started again.
If we do end up in a state where we are at the threshold limit, it makes
more sense to just stop the queue and wait for the next IRQ to trigger the
TX housekeeping again. There is no rush in enqueuing the next packet, it
needs to wait for all the others in the queue to be dispatched first
anyway.

Signed-off-by: John Crispin <john@phrozen.org>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 2d3c464..4d0b179 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -780,12 +780,9 @@ static int mtk_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	if (mtk_tx_map(skb, dev, tx_num, ring, gso) < 0)
 		goto drop;
 
-	if (unlikely(atomic_read(&ring->free_count) <= ring->thresh)) {
+	if (unlikely(atomic_read(&ring->free_count) <= ring->thresh))
 		mtk_stop_queue(eth);
-		if (unlikely(atomic_read(&ring->free_count) >
-			     ring->thresh))
-			mtk_wake_queue(eth);
-	}
+
 	spin_unlock_irqrestore(&eth->page_lock, flags);
 
 	return NETDEV_TX_OK;
-- 
1.7.10.4

Re: [PATCH 00/12] net-next: mediatek: bug fixes and tx stall fixes

[Andrew Lunn]

On Thu, May 05, 2016 at 11:26:13AM +0200, John Crispin wrote:
> This series is a collection of fixes for bugs that we stumble across while
> doing more performance testing and code level review.

Hi John

I just noticed the code uses kfree(mii_bus), rather than
mdiobus_free(mii_bus).

	Andrew