Re: [casper] ROACH 1 udp bad checksum
No probs! On Mon, 5 Sep 2016 at 11:27 Kaustubh Rajwadewrote: > Hi Jack, > > Thanks for pointing me in the right direction! > > Cheers, > Kaustubh > > On Mon, Sep 5, 2016 at 8:14 PM, Jack Hickish > wrote: > >> Hi Kaustubh, >> >> I think you missed out on a fix by just a couple of weeks -- see >> https://github.com/casper-astro/mlib_devel/commit/129937f7dba65f53db39b3fc0a0c968d2cacf9f6 >> >> Cheers >> Jack >> >> On Mon, 5 Sep 2016 at 10:00 Kaustubh Rajwade >> wrote: >> >>> Hi Jack, >>> >>> >>> Here are commits that I am using: >>> >>> Old library: commit a88c343 (16 Nov 2010) >>> >>> New library: commit ecab6f (2 Jan 2015) >>> >>> Cheers, >>> Kaustubh >>> >>> On Mon, Sep 5, 2016 at 6:07 PM, Jack Hickish >>> wrote: >>> Hi Kaustubh, What version of the mlib_devel libraries are you using? Cheers, Jack On Mon, 5 Sep 2016 at 08:11 Kaustubh Rajwade wrote: > Hi Casperites, > > I have a simple ROACH 1 design streaming packets via two 10GbE ports > (clock:160 MHz). When I compile the design using old libraries (XSG: 11.4, > Matlab 2009), I have no issues in receiving packets via sockets. When I > compile the same design using newer libraries (XSG:14.7, Matlab2013a), I > can receive packets via tcpdump but I get bad checksum and I cannot > capture > the packets using standard methods (e.g. recvfrom()). An example of the > checksum error is shown below: > > > > > > > > > *tcpdump: listening on eth2, link-type EN10MB (Ethernet), capture size > 262144 bytes16:04:58.998384 02:02:0a:00:00:1f > 00:25:90:63:13:ac, > ethertype IPv4 (0x0800), length 8242: (tos 0x0, ttl 255, id 0, offset 0, > flags [DF], proto UDP (17), length 8228, bad cksum 63a8 (->47a0)!) > 10.0.0.31.48400 > 10.0.0.10.48500: [no cksum] UDP, length 8200 > 0x: 0025 9063 13ac 0202 0a00 001f 0800 45000x0010: 2024 > 4000 ff11 63a8 0a00 001f 0a000x0020: 000a bd10 bd74 2010 > > 0027 8734* > I validated the payload in the udp packets by sending counter values. > > Has anyone experienced this issue before? > > Cheers, > Kaustubh > >>> >
Re: [casper] ROACH 1 udp bad checksum
Hi Jack, Thanks for pointing me in the right direction! Cheers, Kaustubh On Mon, Sep 5, 2016 at 8:14 PM, Jack Hickishwrote: > Hi Kaustubh, > > I think you missed out on a fix by just a couple of weeks -- see > https://github.com/casper-astro/mlib_devel/commit/ > 129937f7dba65f53db39b3fc0a0c968d2cacf9f6 > > Cheers > Jack > > On Mon, 5 Sep 2016 at 10:00 Kaustubh Rajwade > wrote: > >> Hi Jack, >> >> >> Here are commits that I am using: >> >> Old library: commit a88c343 (16 Nov 2010) >> >> New library: commit ecab6f (2 Jan 2015) >> >> Cheers, >> Kaustubh >> >> On Mon, Sep 5, 2016 at 6:07 PM, Jack Hickish >> wrote: >> >>> Hi Kaustubh, >>> >>> What version of the mlib_devel libraries are you using? >>> >>> Cheers, >>> >>> Jack >>> >>> On Mon, 5 Sep 2016 at 08:11 Kaustubh Rajwade >>> wrote: >>> Hi Casperites, I have a simple ROACH 1 design streaming packets via two 10GbE ports (clock:160 MHz). When I compile the design using old libraries (XSG: 11.4, Matlab 2009), I have no issues in receiving packets via sockets. When I compile the same design using newer libraries (XSG:14.7, Matlab2013a), I can receive packets via tcpdump but I get bad checksum and I cannot capture the packets using standard methods (e.g. recvfrom()). An example of the checksum error is shown below: *tcpdump: listening on eth2, link-type EN10MB (Ethernet), capture size 262144 bytes16:04:58.998384 02:02:0a:00:00:1f > 00:25:90:63:13:ac, ethertype IPv4 (0x0800), length 8242: (tos 0x0, ttl 255, id 0, offset 0, flags [DF], proto UDP (17), length 8228, bad cksum 63a8 (->47a0)!) 10.0.0.31.48400 > 10.0.0.10.48500: [no cksum] UDP, length 8200 0x: 0025 9063 13ac 0202 0a00 001f 0800 45000x0010: 2024 4000 ff11 63a8 0a00 001f 0a000x0020: 000a bd10 bd74 2010 0027 8734* I validated the payload in the udp packets by sending counter values. Has anyone experienced this issue before? Cheers, Kaustubh >>> >>
Re: [casper] ROACH 1 udp bad checksum
Hi Kaustubh, I think you missed out on a fix by just a couple of weeks -- see https://github.com/casper-astro/mlib_devel/commit/129937f7dba65f53db39b3fc0a0c968d2cacf9f6 Cheers Jack On Mon, 5 Sep 2016 at 10:00 Kaustubh Rajwadewrote: > Hi Jack, > > > Here are commits that I am using: > > Old library: commit a88c343 (16 Nov 2010) > > New library: commit ecab6f (2 Jan 2015) > > Cheers, > Kaustubh > > On Mon, Sep 5, 2016 at 6:07 PM, Jack Hickish > wrote: > >> Hi Kaustubh, >> >> What version of the mlib_devel libraries are you using? >> >> Cheers, >> >> Jack >> >> On Mon, 5 Sep 2016 at 08:11 Kaustubh Rajwade >> wrote: >> >>> Hi Casperites, >>> >>> I have a simple ROACH 1 design streaming packets via two 10GbE ports >>> (clock:160 MHz). When I compile the design using old libraries (XSG: 11.4, >>> Matlab 2009), I have no issues in receiving packets via sockets. When I >>> compile the same design using newer libraries (XSG:14.7, Matlab2013a), I >>> can receive packets via tcpdump but I get bad checksum and I cannot capture >>> the packets using standard methods (e.g. recvfrom()). An example of the >>> checksum error is shown below: >>> >>> >>> >>> >>> >>> >>> >>> >>> *tcpdump: listening on eth2, link-type EN10MB (Ethernet), capture size >>> 262144 bytes16:04:58.998384 02:02:0a:00:00:1f > 00:25:90:63:13:ac, >>> ethertype IPv4 (0x0800), length 8242: (tos 0x0, ttl 255, id 0, offset 0, >>> flags [DF], proto UDP (17), length 8228, bad cksum 63a8 (->47a0)!) >>> 10.0.0.31.48400 > 10.0.0.10.48500: [no cksum] UDP, length 8200 >>> 0x: 0025 9063 13ac 0202 0a00 001f 0800 45000x0010: 2024 >>> 4000 ff11 63a8 0a00 001f 0a000x0020: 000a bd10 bd74 2010 >>> 0027 8734* >>> I validated the payload in the udp packets by sending counter values. >>> >>> Has anyone experienced this issue before? >>> >>> Cheers, >>> Kaustubh >>> >> >
Re: [casper] ROACH 1 udp bad checksum
Hi Kaustubh, What version of the mlib_devel libraries are you using? Cheers, Jack On Mon, 5 Sep 2016 at 08:11 Kaustubh Rajwadewrote: > Hi Casperites, > > I have a simple ROACH 1 design streaming packets via two 10GbE ports > (clock:160 MHz). When I compile the design using old libraries (XSG: 11.4, > Matlab 2009), I have no issues in receiving packets via sockets. When I > compile the same design using newer libraries (XSG:14.7, Matlab2013a), I > can receive packets via tcpdump but I get bad checksum and I cannot capture > the packets using standard methods (e.g. recvfrom()). An example of the > checksum error is shown below: > > > > > > > > > *tcpdump: listening on eth2, link-type EN10MB (Ethernet), capture size > 262144 bytes16:04:58.998384 02:02:0a:00:00:1f > 00:25:90:63:13:ac, > ethertype IPv4 (0x0800), length 8242: (tos 0x0, ttl 255, id 0, offset 0, > flags [DF], proto UDP (17), length 8228, bad cksum 63a8 (->47a0)!) > 10.0.0.31.48400 > 10.0.0.10.48500: [no cksum] UDP, length 8200 > 0x: 0025 9063 13ac 0202 0a00 001f 0800 45000x0010: 2024 > 4000 ff11 63a8 0a00 001f 0a000x0020: 000a bd10 bd74 2010 > 0027 8734* > I validated the payload in the udp packets by sending counter values. > > Has anyone experienced this issue before? > > Cheers, > Kaustubh >
[casper] ROACH 1 udp bad checksum
Hi Casperites, I have a simple ROACH 1 design streaming packets via two 10GbE ports (clock:160 MHz). When I compile the design using old libraries (XSG: 11.4, Matlab 2009), I have no issues in receiving packets via sockets. When I compile the same design using newer libraries (XSG:14.7, Matlab2013a), I can receive packets via tcpdump but I get bad checksum and I cannot capture the packets using standard methods (e.g. recvfrom()). An example of the checksum error is shown below: *tcpdump: listening on eth2, link-type EN10MB (Ethernet), capture size 262144 bytes16:04:58.998384 02:02:0a:00:00:1f > 00:25:90:63:13:ac, ethertype IPv4 (0x0800), length 8242: (tos 0x0, ttl 255, id 0, offset 0, flags [DF], proto UDP (17), length 8228, bad cksum 63a8 (->47a0)!) 10.0.0.31.48400 > 10.0.0.10.48500: [no cksum] UDP, length 8200 0x: 0025 9063 13ac 0202 0a00 001f 0800 45000x0010: 2024 4000 ff11 63a8 0a00 001f 0a000x0020: 000a bd10 bd74 2010 0027 8734* I validated the payload in the udp packets by sending counter values. Has anyone experienced this issue before? Cheers, Kaustubh