[PATCH] mv643xx_eth: Fix a DMA-API error handling warning

[PATCH] mv643xx_eth: Fix a DMA-API error handling warning

[Lubomir Rintel]

We check the failure status just prior to unmap, since it would be too much of
a hassle to roll back commands we already started to enqueue if we handled it
just after the map.

This way we at least avoid a lockup on reclaim, the card presumably didn't
succeed DMA-ing to a bogus address anyway.

------------[ cut here ]------------
WARNING: at lib/dma-debug.c:933 check_unmap+0x75c/0x840()
mv643xx_eth_port mv643xx_eth_port.0: DMA-API: device driver failed to check map error[device address=0x000000001cc3f1c2] [size=90 bytes] [mapped as single]
Modules linked in:
 bnep bluetooth rfkill marvell mv643xx_eth mv_cesa leds_gpio uinput mmc_block sata_mv mvsdio mmc_core usb_storage
[<c000f838>] (unwind_backtrace+0x0/0x124) from [<c001bbec>] (warn_slowpath_common+0x54/0x6c)
[<c001bbec>] (warn_slowpath_common+0x54/0x6c) from [<c001bc9c>] (warn_slowpath_fmt+0x34/0x44)
[<c001bc9c>] (warn_slowpath_fmt+0x34/0x44) from [<c0281040>] (check_unmap+0x75c/0x840)
[<c0281040>] (check_unmap+0x75c/0x840) from [<c02811f0>] (debug_dma_unmap_page+0x64/0x70)
[<c02811f0>] (debug_dma_unmap_page+0x64/0x70) from [<bf068fc0>] (txq_reclaim+0x218/0x264 [mv643xx_eth])
[<bf068fc0>] (txq_reclaim+0x218/0x264 [mv643xx_eth]) from [<bf069854>] (mv643xx_eth_poll+0x2a8/0x6b8 [mv643xx_eth])
[<bf069854>] (mv643xx_eth_poll+0x2a8/0x6b8 [mv643xx_eth]) from [<c040e808>] (net_rx_action+0x9c/0x338)
[<c040e808>] (net_rx_action+0x9c/0x338) from [<c002483c>] (__do_softirq+0x16c/0x398)
[<c002483c>] (__do_softirq+0x16c/0x398) from [<c0024e40>] (irq_exit+0x5c/0xc0)
[<c0024e40>] (irq_exit+0x5c/0xc0) from [<c0009ba4>] (handle_IRQ+0x6c/0x8c)
[<c0009ba4>] (handle_IRQ+0x6c/0x8c) from [<c05167b8>] (__irq_svc+0x38/0x80)
[<c05167b8>] (__irq_svc+0x38/0x80) from [<c02629e4>] (memcpy+0x24/0x3a4)
---[ end trace 0c94555a37c67b61 ]---
Mapped at:
 [<c0281818>] debug_dma_map_page+0x48/0x11c
 [<bf068494>] mv643xx_eth_xmit+0x41c/0x5a8 [mv643xx_eth]
 [<c0410a6c>] dev_hard_start_xmit+0x308/0x6cc
 [<c042c720>] sch_direct_xmit+0x74/0x2d4
 [<c041148c>] dev_queue_xmit+0x4b8/0x8dc

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Cc: Lennert Buytenhek <buytenh@wantstofly.org>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org

---
 drivers/net/ethernet/marvell/mv643xx_eth.c |   26 +++++++++++++++++++-------
 1 files changed, 19 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mv643xx_eth.c b/drivers/net/ethernet/marvell/mv643xx_eth.c
index d1cbfb1..7e14c83 100644
--- a/drivers/net/ethernet/marvell/mv643xx_eth.c
+++ b/drivers/net/ethernet/marvell/mv643xx_eth.c
@@ -519,8 +519,13 @@ static int rxq_process(struct rx_queue *rxq, int budget)
 		if (rxq->rx_curr_desc == rxq->rx_ring_size)
 			rxq->rx_curr_desc = 0;
 
-		dma_unmap_single(mp->dev->dev.parent, rx_desc->buf_ptr,
-				 rx_desc->buf_size, DMA_FROM_DEVICE);
+		if (dma_mapping_error(mp->dev->dev.parent, rx_desc->buf_ptr)) {
+			dev_err(mp->dev->dev.parent,
+				"Receive buffer could not be mapped, skipping unmap.\n");
+		} else {
+			dma_unmap_single(mp->dev->dev.parent, rx_desc->buf_ptr,
+					 rx_desc->buf_size, DMA_FROM_DEVICE);
+		}
 		rxq->rx_desc_count--;
 		rx++;
 
@@ -902,12 +907,19 @@ static int txq_reclaim(struct tx_queue *txq, int budget, int force)
 			mp->dev->stats.tx_errors++;
 		}
 
-		if (cmd_sts & TX_FIRST_DESC) {
-			dma_unmap_single(mp->dev->dev.parent, desc->buf_ptr,
-					 desc->byte_cnt, DMA_TO_DEVICE);
+		if (dma_mapping_error(mp->dev->dev.parent, desc->buf_ptr)) {
+			dev_err(mp->dev->dev.parent,
+				"Transmit buffer could not be mapped, skipping unmap.\n");
 		} else {
-			dma_unmap_page(mp->dev->dev.parent, desc->buf_ptr,
-				       desc->byte_cnt, DMA_TO_DEVICE);
+			if (cmd_sts & TX_FIRST_DESC) {
+				dma_unmap_single(mp->dev->dev.parent,
+						desc->buf_ptr, desc->byte_cnt,
+						DMA_TO_DEVICE);
+			} else {
+				dma_unmap_page(mp->dev->dev.parent,
+						desc->buf_ptr, desc->byte_cnt,
+						DMA_TO_DEVICE);
+			}
 		}
 
 		dev_kfree_skb(skb);
-- 
1.7.1

Re: [PATCH] mv643xx_eth: Fix a DMA-API error handling warning

[David Miller]

From: Lubomir Rintel <lkundrak@v3.sk>
Date: Tue, 18 Jun 2013 19:33:28 +0200

> We check the failure status just prior to unmap, since it would be too much of
> a hassle to roll back commands we already started to enqueue if we handled it
> just after the map.
> 
> This way we at least avoid a lockup on reclaim, the card presumably didn't
> succeed DMA-ing to a bogus address anyway.

You have to handle it at map time, I don't care how complicated it is.

As is, when a DMA mapping failure occurs the chip is crapping into
random memory.  At that point, who cares what you decide to do about
it at unmap time?

This patch is unacceptable, sorry.