>-----Original Message----- >From: Jean-Mickael Guerin [mailto:[email protected]] >Sent: Thursday, April 21, 2011 12:04 AM >To: Tantilov, Emil S >Cc: [email protected] >Subject: Re: [E1000-devel] CRC stripping on 82598 and 82599 > >On 4/15/2011 6:41 PM, Jean-Mickael Guerin wrote: >> On 4/12/2011 7:32 PM, Tantilov, Emil S wrote: >>>> -----Original Message----- >>>> From: Jean-Mickael Guerin [mailto:[email protected]] >>>> Sent: Thursday, April 07, 2011 12:17 AM >>>> To: [email protected] >>>> Subject: [E1000-devel] CRC stripping on 82598 and 82599 >>>> >>>> Hello, >>>> >>>> It seems CRC is not stripped out on 82599, neither on 82598. >>>> The ixgbe driver is setting RDRXCTL for 82599, nothing else. >>>> >From datasheet I found that RDRXCTL.CRCStrip and HLREG0.RXCRCSTRP >>>> must be set to the same value. >>>> >>>> I tested this patch (ixgbe-3.2.10) on 82598: always set >>>> IXGBE_HLREG0_RXCRCSTRP, >>>> and it CRC is stripped now: >>>> >>>> --- a/src/ixgbe_main.c >>>> +++ b/src/ixgbe_main.c >>>> @@ -3439,6 +3439,8 @@ static void ixgbe_set_rx_buffer_len(struct >>>> ixgbe_adapter *adapter) >>>> hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0); >>>> /* set jumbo enable since MHADD.MFS is keeping size locked at >>>> max_frame */ >>>> hlreg0 |= IXGBE_HLREG0_JUMBOEN; >>>> + /* Enable CRC stripping */ >>>> + hlreg0 |= IXGBE_HLREG0_RXCRCSTRP; >>>> IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0); >>>> >>>> >>>> I'm going to test on 82599 too. >>>> I can't test for other NIC than 82598 and 82599, may a test for (82598 >|| >>>> 82599) should >>>> be added. >>> >>> Thanks for the patch. Are you seeing an issue where the CRC is not being >stripped, or is the patch just to satisfy the requirement in the datasheet? >>> >>> If you are seeing a problem could you provide the values of the HLREG0 >and RDRXCTL? You can use the ethregs tool available for download from >e1000.sf.net. A detailed information about your HW would also be helpful >such as lspci -vvv and ethtool -e. >> >> I have seen skb->len 4 bytes too long, so at some point the IP stack >needs to trim it, I wanted to eliminate this step. >> But I will trace to confirm it as soon as I have the hand on the board >again, probably next week. > >Finally CRC is correctly stripped by driver, because IXGBE_HLREG0_RXCRCSTRP >= 1 on startup. >I was mistaken by some short packets. >I patched to reset this register to 0, CRC is not stripped as expected, and >after reboot >IXGBE_HLREG0_RXCRCSTRP starts at 1 whatever the previous value is. > >So the patch does not bring anything, or maybe it gives an explicit >statement about CRC >so that no doubt is permitted.
Alright - thanks for the follow-up. My results match with your observation, so we'll leave it as is. Emil > >Regards, >Jean-Mickael ------------------------------------------------------------------------------ Fulfilling the Lean Software Promise Lean software platforms are now widely adopted and the benefits have been demonstrated beyond question. Learn why your peers are replacing JEE containers with lightweight application servers - and what you can gain from the move. http://p.sf.net/sfu/vmware-sfemails _______________________________________________ E1000-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
