The livefish checkin had a bug in it. It was fixed on 2891. My mistake. Remove of livefish == live.
Thanks Tzachi From: Smith, Stan [mailto:[email protected]] Sent: Thursday, September 02, 2010 7:27 PM To: Tzachi Dar; Alex Naslednikov; [email protected] Subject: RE: [ofw] [Patch] [IPoIB_NDIS6_CM] Remove wrong ASSERT Hello, Thanks for forwarding your version of ipoib_port.cpp this will assist in debugging IPoIB patches which would not apply correctly. So far the latest SVN commits have produced the following behaviors: IO WorkFlow commit has broken ttcp over an IPoIB interface, svr 2008 R2 x64, svr2003 x86 Removal of LiveFish code causes a BSOD during install for Svr2003 x86, mthca driver. (removal of LiveFish == DeadFish?) Don't get me wrong, the patches are welcome although some upfront testing in the winOFED environment would be good in order to skip the above situations. stan. ________________________________ From: Tzachi Dar [mailto:[email protected]] Sent: Thursday, September 02, 2010 3:59 AM To: Smith, Stan; Alex Naslednikov; [email protected] Subject: RE: [ofw] [Patch] [IPoIB_NDIS6_CM] Remove wrong ASSERT Hi Stan, Attached to this mail you can see our latest version of ipoib_port.cpp I would like to explain the reasons for the different versions that we are having. The goal of our joint work is to have one codebase that we can all share. However, due to different business reasons and to the fact that our work with the open fabrics svn is very slow, we had to create our own repository. We try to keep the different source code as close as possible to each other, otherwise it makes life harder on both sides. As you saw on the last week we have been sending a big number of patches in order to sync both repositories. We have also been taking many of the patches that have been made by the community. Our goal is to have the same code base, and then patches will be applied easily. Please also note that when doing changes to our code, we try to be on the safe side. We are trying very hard not to be in a place where customers will reach a BSOD because of a small bug. This is the reason for checkins that are very defensive like the memory allocation checkin. I hope that you understand. Thanks Tzachi From: [email protected] [mailto:[email protected]] On Behalf Of Smith, Stan Sent: Wednesday, September 01, 2010 7:24 PM To: Alex Naslednikov; [email protected] Subject: Re: [ofw] [Patch] [IPoIB_NDIS6_CM] Remove wrong ASSERT The first patch you supplied has this removal, the 2nd patch did not. When I checked the patch in I removed the ASSERT(FALSE). Please note the line number in your patch is 4607, the same line is @ 4671. Why is there such a skew? Is there a problem sending me your version of ipoib_port.cpp? ________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Alex Naslednikov Sent: Wednesday, September 01, 2010 7:15 AM To: [email protected] Subject: [ofw] [Patch] [IPoIB_NDIS6_CM] Remove wrong ASSERT The ASSERT assume that we never get a DHCP DISCOVERY msg with local MAC, while the original meaning was not to get DHCP_INFORM message with non-local MAD. Signed-Off by: Uri Habusha (urih at mellanox.co.il) Index: B:/users/xalex/MLNX_WinOF-2_1_2/ulp/ipoib_NDIS6_CM/kernel/ipoib_port.cpp =================================================================== --- B:/users/xalex/MLNX_WinOF-2_1_2/ulp/ipoib_NDIS6_CM/kernel/ipoib_port.cpp (revision 6300) +++ B:/users/xalex/MLNX_WinOF-2_1_2/ulp/ipoib_NDIS6_CM/kernel/ipoib_port.cpp (revision 6301) @@ -4607,7 +4607,6 @@ if( p_cid[1] == HW_ADDR_LEN+1 && !cl_memcmp( &p_cid[3], &s_buf->p_port->p_adapter->params.conf_mac.addr, HW_ADDR_LEN ) ) { - ASSERT( FALSE ); /* Make sure there's room to extend it. 22 is the size of * the CID option for IPoIB. (20 is the length, one byte for type and the second for lenght field) */
_______________________________________________ ofw mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
