[PATCH net,stable 1/1] net: fec: unmap the xmit buffer that are not transferred by DMA

[PATCH net,stable 1/1] net: fec: unmap the xmit buffer that are not transferred by DMA

[Fugang Duan]

The enet IP only support 32 bit, it will use swiotlb buffer to do dma
mapping when xmit buffer DMA memory address is bigger than 4G in i.MX
platform. After stress suspend/resume test, it will print out:

log:
[12826.352864] fec 5b040000.ethernet: swiotlb buffer is full (sz: 191 bytes)
[12826.359676] DMA: Out of SW-IOMMU space for 191 bytes at device 5b040000.ethernet
[12826.367110] fec 5b040000.ethernet eth0: Tx DMA memory map failed

The issue is that the ready xmit buffers that are dma mapped but DMA still
don't copy them into fifo, once MAC restart, these DMA buffers are not unmapped.
So it should check the dma mapping buffer and unmap them.

Signed-off-by: Fugang Duan <fugang.duan@nxp.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 2d1b065..e17d10b 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -818,6 +818,12 @@ static void fec_enet_bd_init(struct net_device *dev)
 		for (i = 0; i < txq->bd.ring_size; i++) {
 			/* Initialize the BD for every fragment in the page. */
 			bdp->cbd_sc = cpu_to_fec16(0);
+			if (bdp->cbd_bufaddr &&
+			    !IS_TSO_HEADER(txq, fec32_to_cpu(bdp->cbd_bufaddr)))
+				dma_unmap_single(&fep->pdev->dev,
+						 fec32_to_cpu(bdp->cbd_bufaddr),
+						 fec16_to_cpu(bdp->cbd_datlen),
+						 DMA_TO_DEVICE);
 			if (txq->tx_skbuff[i]) {
 				dev_kfree_skb_any(txq->tx_skbuff[i]);
 				txq->tx_skbuff[i] = NULL;
-- 
1.9.1

RE: [PATCH net,stable 1/1] net: fec: unmap the xmit buffer that are not transferred by DMA

[Andy Duan]

From: Fugang Duan <fugang.duan@nxp.com> Sent: Friday, December 22, 2017 5:12 PM
>The enet IP only support 32 bit, it will use swiotlb buffer to do dma mapping
>when xmit buffer DMA memory address is bigger than 4G in i.MX platform.
>After stress suspend/resume test, it will print out:
>
>log:
>[12826.352864] fec 5b040000.ethernet: swiotlb buffer is full (sz: 191 bytes)
>[12826.359676] DMA: Out of SW-IOMMU space for 191 bytes at device
>5b040000.ethernet [12826.367110] fec 5b040000.ethernet eth0: Tx DMA
>memory map failed
>
>The issue is that the ready xmit buffers that are dma mapped but DMA still
>don't copy them into fifo, once MAC restart, these DMA buffers are not
>unmapped.
>So it should check the dma mapping buffer and unmap them.
>
>Signed-off-by: Fugang Duan <fugang.duan@nxp.com>
>---
Hi David,

The patch pass two days long time suspend/resume test. Pls apply it to net tree and stable tree.

Thanks.

> drivers/net/ethernet/freescale/fec_main.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
>diff --git a/drivers/net/ethernet/freescale/fec_main.c
>b/drivers/net/ethernet/freescale/fec_main.c
>index 2d1b065..e17d10b 100644
>--- a/drivers/net/ethernet/freescale/fec_main.c
>+++ b/drivers/net/ethernet/freescale/fec_main.c
>@@ -818,6 +818,12 @@ static void fec_enet_bd_init(struct net_device *dev)
> 		for (i = 0; i < txq->bd.ring_size; i++) {
> 			/* Initialize the BD for every fragment in the page. */
> 			bdp->cbd_sc = cpu_to_fec16(0);
>+			if (bdp->cbd_bufaddr &&
>+			    !IS_TSO_HEADER(txq, fec32_to_cpu(bdp-
>>cbd_bufaddr)))
>+				dma_unmap_single(&fep->pdev->dev,
>+						 fec32_to_cpu(bdp-
>>cbd_bufaddr),
>+						 fec16_to_cpu(bdp-
>>cbd_datlen),
>+						 DMA_TO_DEVICE);
> 			if (txq->tx_skbuff[i]) {
> 				dev_kfree_skb_any(txq->tx_skbuff[i]);
> 				txq->tx_skbuff[i] = NULL;
>--
>1.9.1

Re: [PATCH net,stable 1/1] net: fec: unmap the xmit buffer that are not transferred by DMA

[David Miller]

From: Fugang Duan <fugang.duan@nxp.com>
Date: Fri, 22 Dec 2017 17:12:09 +0800

> The enet IP only support 32 bit, it will use swiotlb buffer to do dma
> mapping when xmit buffer DMA memory address is bigger than 4G in i.MX
> platform. After stress suspend/resume test, it will print out:
> 
> log:
> [12826.352864] fec 5b040000.ethernet: swiotlb buffer is full (sz: 191 bytes)
> [12826.359676] DMA: Out of SW-IOMMU space for 191 bytes at device 5b040000.ethernet
> [12826.367110] fec 5b040000.ethernet eth0: Tx DMA memory map failed
> 
> The issue is that the ready xmit buffers that are dma mapped but DMA still
> don't copy them into fifo, once MAC restart, these DMA buffers are not unmapped.
> So it should check the dma mapping buffer and unmap them.
> 
> Signed-off-by: Fugang Duan <fugang.duan@nxp.com>

Applied and queued up for -stable.