[PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[Toshiharu Okada]

This PCH_GBE driver had an issue that the receiving data is not normal.
This driver had not removed correctly the padding data 
which the DMA include in receiving data. 

This patch fixed this issue.

Signed-off-by: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
---
 drivers/net/pch_gbe/pch_gbe_main.c |   21 ++++++++++++---------
 1 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/drivers/net/pch_gbe/pch_gbe_main.c b/drivers/net/pch_gbe/pch_gbe_main.c
index 03a1d28..1bf2626 100644
--- a/drivers/net/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/pch_gbe/pch_gbe_main.c
@@ -29,6 +29,7 @@ const char pch_driver_version[] = DRV_VERSION;
 #define PCH_GBE_SHORT_PKT		64
 #define DSC_INIT16			0xC000
 #define PCH_GBE_DMA_ALIGN		0
+#define PCH_GBE_DMA_PADDING		2
 #define PCH_GBE_WATCHDOG_PERIOD		(1 * HZ)	/* watchdog time */
 #define PCH_GBE_COPYBREAK_DEFAULT	256
 #define PCH_GBE_PCI_BAR			1
@@ -1422,8 +1423,8 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
 			pr_err("Receive CRC Error\n");
 		} else {
 			/* get receive length */
-			/* length convert[-3], padding[-2] */
-			length = (rx_desc->rx_words_eob) - 3 - 2;
+			/* length convert[-3] */
+			length = (rx_desc->rx_words_eob) - 3;
 
 			/* Decide the data conversion method */
 			if (!adapter->rx_csum) {
@@ -1443,12 +1444,11 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
 			if (skb_copy_flag) {	/* recycle  skb */
 				struct sk_buff *new_skb;
 				new_skb =
-				    netdev_alloc_skb(netdev,
-						     length + NET_IP_ALIGN);
+				    netdev_alloc_skb(netdev, length);
 				if (new_skb) {
 					if (!skb_padding_flag) {
 						skb_reserve(new_skb,
-								NET_IP_ALIGN);
+							PCH_GBE_DMA_PADDING);
 					}
 					memcpy(new_skb->data, skb->data,
 						length);
@@ -1465,12 +1465,15 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
 			}
 			if (skb_padding_flag) {
 				memcpy(&tmp_packet[0], &skb->data[0], ETH_HLEN);
-				memcpy(&skb->data[NET_IP_ALIGN], &tmp_packet[0],
-					ETH_HLEN);
-				skb_reserve(skb, NET_IP_ALIGN);
-
+				memcpy(&skb->data[PCH_GBE_DMA_PADDING],
+				       &tmp_packet[0], ETH_HLEN);
+				skb_reserve(skb, PCH_GBE_DMA_PADDING);
+				/* The length includes padding length */
+				length = length - PCH_GBE_DMA_PADDING;
 			}
 
+			/* The length includes FCS length */
+			length = length - ETH_FCS_LEN;
 			/* update status of driver */
 			adapter->stats.rx_bytes += length;
 			adapter->stats.rx_packets++;
-- 
1.6.2.5

Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[David Miller]

From: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
Date: Mon, 24 Jan 2011 13:43:31 +0900

> This PCH_GBE driver had an issue that the receiving data is not normal.
> This driver had not removed correctly the padding data 
> which the DMA include in receiving data. 
> 
> This patch fixed this issue.
> 
> Signed-off-by: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>

There are bugs in these changes:

>  			if (skb_copy_flag) {	/* recycle  skb */
>  				struct sk_buff *new_skb;
>  				new_skb =
> -				    netdev_alloc_skb(netdev,
> -						     length + NET_IP_ALIGN);
> +				    netdev_alloc_skb(netdev, length);
>  				if (new_skb) {
>  					if (!skb_padding_flag) {
>  						skb_reserve(new_skb,
> -								NET_IP_ALIGN);
> +							PCH_GBE_DMA_PADDING);
>  					}
>  					memcpy(new_skb->data, skb->data,
>  						length);

If "!skb_padding_flag" then you will write past the end of the SKB
data in that memcpy.

You cannot allocate only "length" then proceed to reserve PCH_GBE_DMA_PADDING
and then add "length" worth of data on top of that.  In such a cause you
must allocate at least "length + PCH_GBE_DMA_PADDING".

Furthermore you _MUST_ respect NET_IP_ALIGN.  Some platforms set this value
to "0", because otherwise performance suffers greatly.

There are two seperate issues, removing the padding bytes provided by
the device, and aligning the IP headers as wanted by the cpu
architecutre.  Therefore they should be handled seperately, and we
therefore should still see references to NET_IP_ALIGN in your patch.

Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[Toshiharu Okada]

Hi David

Thank you for your comment.
I will confirm them  and will submit the patch modified.

Best regards
Toshiharu Okada(OKI semiconductor)
----- Original Message ----- 
From: "David Miller" <davem@davemloft.net>
To: <toshiharu-linux@dsn.okisemi.com>
Cc: <netdev@vger.kernel.org>; <linux-kernel@vger.kernel.org>; 
<qi.wang@intel.com>; <yong.y.wang@intel.com>; 
<andrew.chih.howe.khor@intel.com>; <joel.clark@intel.com>; 
<kok.howg.ewe@intel.com>
Sent: Wednesday, January 26, 2011 6:32 AM
Subject: Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not 
normal.


From: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
Date: Mon, 24 Jan 2011 13:43:31 +0900

> This PCH_GBE driver had an issue that the receiving data is not normal.
> This driver had not removed correctly the padding data
> which the DMA include in receiving data.
>
> This patch fixed this issue.
>
> Signed-off-by: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>

There are bugs in these changes:

>  if (skb_copy_flag) { /* recycle  skb */
>  struct sk_buff *new_skb;
>  new_skb =
> -     netdev_alloc_skb(netdev,
> -      length + NET_IP_ALIGN);
> +     netdev_alloc_skb(netdev, length);
>  if (new_skb) {
>  if (!skb_padding_flag) {
>  skb_reserve(new_skb,
> - NET_IP_ALIGN);
> + PCH_GBE_DMA_PADDING);
>  }
>  memcpy(new_skb->data, skb->data,
>  length);

If "!skb_padding_flag" then you will write past the end of the SKB
data in that memcpy.

You cannot allocate only "length" then proceed to reserve 
PCH_GBE_DMA_PADDING
and then add "length" worth of data on top of that.  In such a cause you
must allocate at least "length + PCH_GBE_DMA_PADDING".

Furthermore you _MUST_ respect NET_IP_ALIGN.  Some platforms set this value
to "0", because otherwise performance suffers greatly.

There are two seperate issues, removing the padding bytes provided by
the device, and aligning the IP headers as wanted by the cpu
architecutre.  Therefore they should be handled seperately, and we
therefore should still see references to NET_IP_ALIGN in your patch.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[Toshiharu Okada]

Hi David

Thank you for your comment.
I will confirm them  and will submit the patch modified.

Best regards
Toshiharu Okada(OKI semiconductor)
----- Original Message ----- 
From: "David Miller" <davem@davemloft.net>
To: <toshiharu-linux@dsn.okisemi.com>
Cc: <netdev@vger.kernel.org>; <linux-kernel@vger.kernel.org>; 
<qi.wang@intel.com>; <yong.y.wang@intel.com>; 
<andrew.chih.howe.khor@intel.com>; <joel.clark@intel.com>; 
<kok.howg.ewe@intel.com>
Sent: Wednesday, January 26, 2011 6:32 AM
Subject: Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not 
normal.


From: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
Date: Mon, 24 Jan 2011 13:43:31 +0900

> This PCH_GBE driver had an issue that the receiving data is not normal.
> This driver had not removed correctly the padding data
> which the DMA include in receiving data.
>
> This patch fixed this issue.
>
> Signed-off-by: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>

There are bugs in these changes:

>  if (skb_copy_flag) { /* recycle  skb */
>  struct sk_buff *new_skb;
>  new_skb =
> -     netdev_alloc_skb(netdev,
> -      length + NET_IP_ALIGN);
> +     netdev_alloc_skb(netdev, length);
>  if (new_skb) {
>  if (!skb_padding_flag) {
>  skb_reserve(new_skb,
> - NET_IP_ALIGN);
> + PCH_GBE_DMA_PADDING);
>  }
>  memcpy(new_skb->data, skb->data,
>  length);

If "!skb_padding_flag" then you will write past the end of the SKB
data in that memcpy.

You cannot allocate only "length" then proceed to reserve 
PCH_GBE_DMA_PADDING
and then add "length" worth of data on top of that.  In such a cause you
must allocate at least "length + PCH_GBE_DMA_PADDING".

Furthermore you _MUST_ respect NET_IP_ALIGN.  Some platforms set this value
to "0", because otherwise performance suffers greatly.

There are two seperate issues, removing the padding bytes provided by
the device, and aligning the IP headers as wanted by the cpu
architecutre.  Therefore they should be handled seperately, and we
therefore should still see references to NET_IP_ALIGN in your patch.

[PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[Toshiharu Okada]

Hi Devid

I resubmit this patch modified.
Please check them.

Best regards
Toshiharu Okada(OKI semiconductor)
---

This PCH_GBE driver had an issue that the receiving data is not normal.
This driver had not removed correctly the padding data 
which the DMA include in receiving data. 

This patch fixed this issue.

Signed-off-by: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
---
 drivers/net/pch_gbe/pch_gbe_main.c |   77 +++++++++++++++++++++--------------
 1 files changed, 46 insertions(+), 31 deletions(-)

diff --git a/drivers/net/pch_gbe/pch_gbe_main.c b/drivers/net/pch_gbe/pch_gbe_main.c
index 03a1d28..3248313 100644
--- a/drivers/net/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/pch_gbe/pch_gbe_main.c
@@ -29,6 +29,7 @@ const char pch_driver_version[] = DRV_VERSION;
 #define PCH_GBE_SHORT_PKT		64
 #define DSC_INIT16			0xC000
 #define PCH_GBE_DMA_ALIGN		0
+#define PCH_GBE_DMA_PADDING		2
 #define PCH_GBE_WATCHDOG_PERIOD		(1 * HZ)	/* watchdog time */
 #define PCH_GBE_COPYBREAK_DEFAULT	256
 #define PCH_GBE_PCI_BAR			1
@@ -1373,7 +1374,7 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
 	unsigned int i;
 	unsigned int cleaned_count = 0;
 	bool cleaned = false;
-	struct sk_buff *skb;
+	struct sk_buff *skb, *new_skb;
 	u8 dma_status;
 	u16 gbec_status;
 	u32 tcp_ip_status;
@@ -1422,55 +1423,69 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
 			pr_err("Receive CRC Error\n");
 		} else {
 			/* get receive length */
-			/* length convert[-3], padding[-2] */
-			length = (rx_desc->rx_words_eob) - 3 - 2;
+			/* length convert[-3] */
+			length = (rx_desc->rx_words_eob) - 3;
 
 			/* Decide the data conversion method */
 			if (!adapter->rx_csum) {
 				/* [Header:14][payload] */
 				skb_padding_flag = 0;
-				skb_copy_flag = 1;
+				if (NET_IP_ALIGN)
+					/* [NET_IP_ALIGN][Header:14][payload] */
+					skb_copy_flag = 1;
+				else
+					skb_copy_flag = 0;
 			} else {
 				/* [Header:14][padding:2][payload] */
 				skb_padding_flag = 1;
+				/* The length includes padding length */
+				length = length - PCH_GBE_DMA_PADDING;
 				if (length < copybreak)
 					skb_copy_flag = 1;
-				else
-					skb_copy_flag = 0;
+				else {
+					/* [Header:14][padding:2][payload]
+					 *   Chenge to the following
+					 * [NET_IP_ALIGN][Header:14][payload] */
+					if (NET_IP_ALIGN == PCH_GBE_DMA_PADDING)
+						skb_copy_flag = 0;
+					else
+						skb_copy_flag = 1;
+				}
 			}
-
 			/* Data conversion */
-			if (skb_copy_flag) {	/* recycle  skb */
-				struct sk_buff *new_skb;
-				new_skb =
-				    netdev_alloc_skb(netdev,
-						     length + NET_IP_ALIGN);
-				if (new_skb) {
-					if (!skb_padding_flag) {
-						skb_reserve(new_skb,
-								NET_IP_ALIGN);
-					}
-					memcpy(new_skb->data, skb->data,
-						length);
-					/* save the skb
-					 * in buffer_info as good */
-					skb = new_skb;
-				} else if (!skb_padding_flag) {
+			if (skb_copy_flag) {
+				new_skb = netdev_alloc_skb(netdev,
+							 length + NET_IP_ALIGN);
+				if (!new_skb) {
 					/* dorrop error */
 					pr_err("New skb allocation Error\n");
 					goto dorrop;
 				}
+				skb_reserve(new_skb, NET_IP_ALIGN);
+				if (skb_padding_flag) {
+					memcpy(new_skb->data, skb->data,
+					       ETH_HLEN);
+					memcpy(&new_skb->data[ETH_HLEN],
+					       &skb->data[ETH_HLEN +
+					       PCH_GBE_DMA_PADDING],
+					       length - ETH_HLEN);
+				} else {
+					memcpy(new_skb->data, skb->data,
+					       length);
+				}
+				skb = new_skb;
 			} else {
+				if (skb_padding_flag) {
+					memcpy(&tmp_packet[0], &skb->data[0],
+					       ETH_HLEN);
+					memcpy(&skb->data[PCH_GBE_DMA_PADDING],
+					       &tmp_packet[0], ETH_HLEN);
+					skb_reserve(skb, PCH_GBE_DMA_PADDING);
+				}
 				buffer_info->skb = NULL;
 			}
-			if (skb_padding_flag) {
-				memcpy(&tmp_packet[0], &skb->data[0], ETH_HLEN);
-				memcpy(&skb->data[NET_IP_ALIGN], &tmp_packet[0],
-					ETH_HLEN);
-				skb_reserve(skb, NET_IP_ALIGN);
-
-			}
-
+			/* The length includes FCS length */
+			length = length - ETH_FCS_LEN;
 			/* update status of driver */
 			adapter->stats.rx_bytes += length;
 			adapter->stats.rx_packets++;
-- 
1.6.2.5

Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[David Miller]

From: Toshiharu Okada <toshiharu-linux@dsn.okisemi.com>
Date: Tue, 08 Feb 2011 17:24:10 +0900

> Hi Devid
> 
> I resubmit this patch modified.
> Please check them.

Your memcpy+memcpy sequences are equivalent to memmove(), please
use that instead.

I have to say this function is insanely complicated.  There seems
to be 16 different ways RX packets are processed.  I can't believe
that is needs to be like this.

Re: [PATCH] pch_gbe: Fix the issue that the receiving data is not normal.

[Toshiharu Okada]

From: David Miller <davem@davemloft.net>
Date: Tue, 08 Feb 2011 16:35:22 -0800 (PST)
>> Hi Devid
>>
>> I resubmit this patch modified.
>> Please check them.
>
>Your memcpy+memcpy sequences are equivalent to memmove(), please
>use that instead.
>

I use memmove().

>I have to say this function is insanely complicated.  There seems
>to be 16 different ways RX packets are processed.  I can't believe
>that is needs to be like this.

If processing is arranged, There are 4 different ways RX packets are 
processed.

[Case1]
"rx_csum is enable " and
 "NET_IP_ALIGN == 0"

 DMA buffer is used as SKB as it is.

[Case2]
"rx_csum is enable " and
 "NET_IP_ALIGN != 0"

 Because alignment differs, the new_skb is newly allocated, and data is 
copied to new_skb.

[Case3]
"rx_csum is disable " and
 "length < copybreak" or "NET_IP_ALIGN != PCH_GBE_DMA_PADDING"

 Because alignment differs, the new_skb is newly allocated, and data is 
copied to new_skb.
 Padding data is deleted at the time of a copy.

[Case4]
"rx_csum is disable " and
 "length >= copybreak" and "NET_IP_ALIGN == PCH_GBE_DMA_PADDING"

 Padding data is deleted by moving header data.

I rewrite without using skb_copy_flag and skb_padding simply.

Best regards
Toshiharu Okada(OKI semiconductor)