[PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[Antoine Tenart]

With CONFIG_DMA_API_DEBUG enabled we now get a warning when using the
mvneta driver:

  mvneta d0030000.ethernet: DMA-API: device driver frees DMA memory with
  wrong function [device address=0x000000001165b000] [size=4096 bytes]
  [mapped as page] [unmapped as single]

This is because when using the s/w buffer management, the Rx descriptor
buffer is mapped with dma_map_page but unmapped with dma_unmap_single.
This patch fixes this by using the right unmapping function.

Fixes: 562e2f467e71 ("net: mvneta: Improve the buffer allocation method for SWBM")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
---
 drivers/net/ethernet/marvell/mvneta.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index bc80a678abc3..2db9708f2e24 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -2008,8 +2008,8 @@ static int mvneta_rx_swbm(struct napi_struct *napi,
 				skb_add_rx_frag(rxq->skb, frag_num, page,
 						frag_offset, frag_size,
 						PAGE_SIZE);
-				dma_unmap_single(dev->dev.parent, phys_addr,
-						 PAGE_SIZE, DMA_FROM_DEVICE);
+				dma_unmap_page(dev->dev.parent, phys_addr,
+					       PAGE_SIZE, DMA_FROM_DEVICE);
 				rxq->left_size -= frag_size;
 			}
 		} else {
-- 
2.17.1

Re: [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[David Miller]

From: Antoine Tenart <antoine.tenart@bootlin.com>
Date: Wed, 19 Sep 2018 15:29:06 +0200

> With CONFIG_DMA_API_DEBUG enabled we now get a warning when using the
> mvneta driver:
> 
>   mvneta d0030000.ethernet: DMA-API: device driver frees DMA memory with
>   wrong function [device address=0x000000001165b000] [size=4096 bytes]
>   [mapped as page] [unmapped as single]
> 
> This is because when using the s/w buffer management, the Rx descriptor
> buffer is mapped with dma_map_page but unmapped with dma_unmap_single.
> This patch fixes this by using the right unmapping function.
> 
> Fixes: 562e2f467e71 ("net: mvneta: Improve the buffer allocation method for SWBM")
> Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>

Applied.

RE: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[Yelena Krivosheev]

Hi Antoine.

Good point.
Please, check and fix all cases of dma_unmap_single() usage.
See mvneta_rxq_drop_pkts()
...
		if (!data || !(rx_desc->buf_phys_addr))
			continue;
		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
		__free_page(data);
...

Thanks.
Yelena


-----Original Message-----
From: Antoine Tenart [mailto:antoine.tenart@bootlin.com] 
Sent: Wednesday, September 19, 2018 4:29 PM
To: davem@davemloft.net; Yelena Krivosheev <yelena@marvell.com>
Cc: Antoine Tenart <antoine.tenart@bootlin.com>; netdev@vger.kernel.org; linux-kernel@vger.kernel.org; thomas.petazzoni@bootlin.com; maxime.chevallier@bootlin.com; gregory.clement@bootlin.com; miquel.raynal@bootlin.com; Nadav Haklai <nadavh@marvell.com>; Stefan Chulski <stefanc@marvell.com>; Yan Markman <ymarkman@marvell.com>; mw@semihalf.com
Subject: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

External Email

----------------------------------------------------------------------
With CONFIG_DMA_API_DEBUG enabled we now get a warning when using the mvneta driver:

  mvneta d0030000.ethernet: DMA-API: device driver frees DMA memory with
  wrong function [device address=0x000000001165b000] [size=4096 bytes]
  [mapped as page] [unmapped as single]

This is because when using the s/w buffer management, the Rx descriptor buffer is mapped with dma_map_page but unmapped with dma_unmap_single.
This patch fixes this by using the right unmapping function.

Fixes: 562e2f467e71 ("net: mvneta: Improve the buffer allocation method for SWBM")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
---
 drivers/net/ethernet/marvell/mvneta.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index bc80a678abc3..2db9708f2e24 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -2008,8 +2008,8 @@ static int mvneta_rx_swbm(struct napi_struct *napi,
 				skb_add_rx_frag(rxq->skb, frag_num, page,
 						frag_offset, frag_size,
 						PAGE_SIZE);
-				dma_unmap_single(dev->dev.parent, phys_addr,
-						 PAGE_SIZE, DMA_FROM_DEVICE);
+				dma_unmap_page(dev->dev.parent, phys_addr,
+					       PAGE_SIZE, DMA_FROM_DEVICE);
 				rxq->left_size -= frag_size;
 			}
 		} else {
--
2.17.1

Re: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[Antoine Tenart]

Hi Yelena,

On Thu, Sep 20, 2018 at 10:14:56AM +0000, Yelena Krivosheev wrote:
> 
> Please, check and fix all cases of dma_unmap_single() usage.
> See mvneta_rxq_drop_pkts()
> ...
> 		if (!data || !(rx_desc->buf_phys_addr))
> 			continue;
> 		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
> 				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
> 		__free_page(data);
> ...

I had a look at the one reported by CONFIG_DMA_API_DEBUG, and at DMA
unmapping calls using PAGE_SIZE. As you pointed out there might be
others parts, thanks!

Antoine

-- 
Antoine Ténart, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Re: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[Gregory CLEMENT]

Hi Antoine,
 
 On jeu., sept. 20 2018, Antoine Tenart <antoine.tenart@bootlin.com> wrote:

> Hi Yelena,
>
> On Thu, Sep 20, 2018 at 10:14:56AM +0000, Yelena Krivosheev wrote:
>> 
>> Please, check and fix all cases of dma_unmap_single() usage.
>> See mvneta_rxq_drop_pkts()
>> ...
>> 		if (!data || !(rx_desc->buf_phys_addr))
>> 			continue;
>> 		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
>> 				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
>> 		__free_page(data);
>> ...
>
> I had a look at the one reported by CONFIG_DMA_API_DEBUG, and at DMA
> unmapping calls using PAGE_SIZE. As you pointed out there might be
> others parts, thanks!

Actually Jisheng had submitted a similar patch few weeks ago and as I
pointed at this time, the dma_unmap in mvneta_rxq_drop_pkts can be
called when the allocation is done in with HWBM in this case which use a
dma_map_single.

I though that in this case using dma_map_single is the things to do even
if in the SWBM case it is less optimal.

Gregory

>
> Antoine
>
> -- 
> Antoine Ténart, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com

-- 
Gregory Clement, Bootlin
Embedded Linux and Kernel engineering
http://bootlin.com

RE: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

[Yelena Krivosheev]

Hi Gregory.

I want to clarify static mvneta_rxq_drop_pkts():

static void mvneta_rxq_drop_pkts(struct mvneta_port *pp,
				 struct mvneta_rx_queue *rxq)
{
	int rx_done, i;

	rx_done = mvneta_rxq_busy_desc_num_get(pp, rxq);
	if (rx_done)
		mvneta_rxq_desc_num_update(pp, rxq, rx_done, rx_done);

	if (pp->bm_priv) { <---------------------------- this is case for HWBM
		for (i = 0; i < rx_done; i++) {
			struct mvneta_rx_desc *rx_desc =
						  mvneta_rxq_next_desc_get(rxq);
			u8 pool_id = MVNETA_RX_GET_BM_POOL_ID(rx_desc);
			struct mvneta_bm_pool *bm_pool;

			bm_pool = &pp->bm_priv->bm_pools[pool_id];
			/* Return dropped buffer to the pool */
			mvneta_bm_pool_put_bp(pp->bm_priv, bm_pool,
					      rx_desc->buf_phys_addr);
		}
		return;
	}

<-------- this is case for SWBM only
	for (i = 0; i < rxq->size; i++) {
		struct mvneta_rx_desc *rx_desc = rxq->descs + i;
		void *data = rxq->buf_virt_addr[i];

		if (!data || !(rx_desc->buf_phys_addr))
			continue;
		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
		__free_page(data);
	}
}

So I suggest to fix dma_unmap_single() call too.

Thanks.
Yelena


-----Original Message-----
From: Gregory CLEMENT [mailto:gregory.clement@bootlin.com] 
Sent: Thursday, September 20, 2018 6:00 PM
To: Antoine Tenart <antoine.tenart@bootlin.com>
Cc: Yelena Krivosheev <yelena@marvell.com>; davem@davemloft.net; netdev@vger.kernel.org; linux-kernel@vger.kernel.org; thomas.petazzoni@bootlin.com; maxime.chevallier@bootlin.com; miquel.raynal@bootlin.com; Nadav Haklai <nadavh@marvell.com>; Stefan Chulski <stefanc@marvell.com>; Yan Markman <ymarkman@marvell.com>; mw@semihalf.com
Subject: Re: [EXT] [PATCH net] net: mvneta: fix the Rx desc buffer DMA unmapping

Hi Antoine,
 
 On jeu., sept. 20 2018, Antoine Tenart <antoine.tenart@bootlin.com> wrote:

> Hi Yelena,
>
> On Thu, Sep 20, 2018 at 10:14:56AM +0000, Yelena Krivosheev wrote:
>> 
>> Please, check and fix all cases of dma_unmap_single() usage.
>> See mvneta_rxq_drop_pkts()
>> ...
>> 		if (!data || !(rx_desc->buf_phys_addr))
>> 			continue;
>> 		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
>> 				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
>> 		__free_page(data);
>> ...
>
> I had a look at the one reported by CONFIG_DMA_API_DEBUG, and at DMA 
> unmapping calls using PAGE_SIZE. As you pointed out there might be 
> others parts, thanks!

Actually Jisheng had submitted a similar patch few weeks ago and as I pointed at this time, the dma_unmap in mvneta_rxq_drop_pkts can be called when the allocation is done in with HWBM in this case which use a dma_map_single.

I though that in this case using dma_map_single is the things to do even if in the SWBM case it is less optimal.

Gregory

>
> Antoine
>
> --
> Antoine Ténart, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com

--
Gregory Clement, Bootlin
Embedded Linux and Kernel engineering
http://bootlin.com