From: Andy Gospodarek <gospo@redhat.com>
Date: Tue, 29 Jan 2008 17:46:56 -0500
Subject: [net] e1000e: disable hw crc stripping
Message-id: 20080129224656.GA25430@gospo.usersys.redhat.com
O-Subject: [RHEL5.2 PATCH] e1000e: disable hw crc stripping
Bugzilla: 430722

It appears that this commit:

commit 140a74802894e9db57e5cd77ccff77e590ece5f3
Author: Auke Kok <auke-jan.h.kok@intel.com>
Date:   Thu Oct 25 13:57:58 2007 -0700

    e1000e: Re-enable SECRC - crc stripping

    This workaround code performed software stripping instead of the
    hardware which can do it much faster. None of the e1000e target
    hardware has issues with this feature and should work fine. This
    gives us some performance back on receive, and removes some
    kludging stripping the 4 bytes.

    Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
    Signed-off-by: Jeff Garzik <jeff@garzik.org>

isn't working quite as well as Intel hoped.  It was discovered that
when running a Xen kernel and using bridging they end up 4 bytes larger
than they should be.  I've confirmed with Intel that there is only a
performance penalty with this patch, so it seems safe to remove.

Herbert was the one that did the leg work to figure out the e1000e
update I posted earlier was the problem and that an 'extra' FCS appeared
on frames coming through the xen bridge.

With that in mind, I tried reverting the above patch, and found it
resolves our issue on one of the systems in the Westford lab.

This resolves BZ 430722.

Acked-by: "John W. Linville" <linville@redhat.com>

diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
index 9914578..27db90a 100644
--- a/drivers/net/e1000e/netdev.c
+++ b/drivers/net/e1000e/netdev.c
@@ -406,6 +406,10 @@ static bool e1000_clean_rx_irq(struct e1000_adapter *adapter,
 			goto next_desc;
 		}
 
+		/* adjust length to remove Ethernet CRC */
+		length -= 4;
+
+		/* probably a little skewed due to removing CRC */
 		total_rx_bytes += length;
 		total_rx_packets++;
 
@@ -699,7 +703,8 @@ static bool e1000_clean_rx_irq_ps(struct e1000_adapter *adapter,
 			kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
 			pci_dma_sync_single_for_device(pdev, ps_page->dma,
 				PAGE_SIZE, PCI_DMA_FROMDEVICE);
-
+			/* remove the CRC */
+			l1 -= 4;
 			skb_put(skb, l1);
 			goto copydone;
 		} /* if */
@@ -721,6 +726,10 @@ static bool e1000_clean_rx_irq_ps(struct e1000_adapter *adapter,
 			skb->truesize += length;
 		}
 
+		/* strip the ethernet crc, problem is we're using pages now so
+		 * this whole operation can get a little cpu intensive */
+		pskb_trim(skb, skb->len - 4);
+
 copydone:
 		total_rx_bytes += skb->len;
 		total_rx_packets++;
@@ -1764,11 +1773,9 @@ static void e1000_setup_rctl(struct e1000_adapter *adapter)
 
 		ew32(RFCTL, rfctl);
 
-		/* Enable Packet split descriptors */
-		rctl |= E1000_RCTL_DTYP_PS;
-		
-		/* Enable hardware CRC frame stripping */
-		rctl |= E1000_RCTL_SECRC;
+		/* disable the stripping of CRC because it breaks
+		 * BMC firmware connected over SMBUS */
+		rctl |= E1000_RCTL_DTYP_PS /* | E1000_RCTL_SECRC */;
 
 		psrctl |= adapter->rx_ps_bsize0 >>
 			E1000_PSRCTL_BSIZE0_SHIFT;