June 2004 - ixp1200 - lists.cs.princeton.edu

Microengine measurments?
by Øyvind Hvamstad 30 Jun '04

30 Jun '04

Hi, it's me again. Does anyone have thoughts on what is the best practice to do precice measurements on a microengine level? I have done some measurments using the cycle_cnt register and it looks valid compared to results from Spalink et al's paper[1]. I have some questions, though (As usual, potentially stupid). Does the Intel Ingress ACE have any impact on performance, or is it merly controlling the interfaces? The reason I ask is because the design document vaguely implies that it is only used for control purposes. I read the cycle_cnt on the top of my dispatch loop and I store the difference after having sent it to the Intel Egress ACE. From the source of DL_MESink, the packet is enqueued for the next ACE. This probably means that I'm unable to measure how long it takes the Egress to process the pakcet. Is this correct? Are there any existing measurements on the performance of the egress? Info: ENP-2505, embedded linux, Intel SDK 2.01 [1] Spaling et al. Building a Robust Software-Based Router Using Network Processors -- Øyvind Hvamstad <oyvindh(a)ifi.uio.no>

1 0

Hellow
by nitin makhija 24 Jun '04

24 Jun '04

Hellow to all, i'm new in this mailing list. I'm an M.tech(CS) student at IIT-Kharagpur. and my MTP is "To Design a Firewall On IXP1200". currently i was stuck in the installation itself, Previously i have installed SDK2.0.1 and Tornado 2.0, but there was some compatibility problem. Now, i have installed SDK2.0.1 but Tornado2.2 is giving some licence problems. And i do wanan know from where to start wd. IXP, i mean wht. shd i read first.. Regards, Nitin Makhija M.tech(CS) IIT-Kharagpur

1 0

Uca Compilation failes
by Øyvind Hvamstad 23 Jun '04

23 Jun '04

wine -- /mnt/ixp-share/workbench/bin/uca.exe failes with this message: WARNING: Old register allocator will try for 45 more seconds before giving up. WARNING: Old register allocator will try for 30 more seconds before giving up. WARNING: Old register allocator will try for 15 more seconds before giving up. ERROR: Timeout occurred while trying to allocate GPRs to A/B Banks. Error report written to file "SlowIngressPol.uci". Is this due to having to many GPR's? Can it be solved by using .local gpr allocations? -- Øyvind Hvamstad <oyvindh(a)ifi.uio.no>

1 1

Ip checksum?
by Øyvind Hvamstad 21 Jun '04

21 Jun '04

I'm sorry for all the fuzz. I'm really nagging the list. I'm wondering what is the best way to calculate the IP checksum on the microengines? As I'm aware, there are ways to optimize this calculation for routers, as they usually only lower the ttl field, but I'm altering the ip addresses. So I guess I'll have to recalculate the whole thing? This function does this on the StrongARM: ix_16bit_t CalcChecksum(ix_8bit_t *addr, ix_16bit_t size) { ix_32bit_t sum = 0; while ( size > 1 ) { sum += *(unsigned short *)(void *)addr; addr += 2; size -= 2; } if ( size > 0 ) { sum += *(unsigned char *)addr; } while ( sum >> 16 ) { sum = ( sum & 0xffff ) + ( sum >> 16 ); } return (~sum); }; Now, how do I translate this function to microcode? Should I write my values to memory, reread the header and calculate the checksum and then write the checksum? Or should I calculate the checksum from GPRs holding the values I want to write? Sorry for another, prossibly stupid, question. I'm in a bit of a hurry u see. -- Øyvind Hvamstad <oyvindh(a)ifi.uio.no>

2 3

RE: [ixp1200] Ip checksum?
by Hussain, Sajjad 21 Jun '04

21 Jun '04

Hi Øyvind Hvamstad, There is a file called cksum32.uc that comes with the IXP Workbench(Microcode\examples\cksum32.uc). It contains code for verifying IP checksum but you can easily modify it according to your need. I am pasting the code in case you don't have access to the example file. ; Prerequisite Register Usage: ; symbol ; iphdr1 - 1st longword of IP header ; iphdr2 - 2nd longword of IP header ; iphdr3 - 3rd longword of IP header ; iphdr4 - 4th longword of IP header (IP Source Address) ; iphdr5 - 5th longword of IP header (IP Dest Address) ; ; Description: ; The iphdr is assumed to be longword aligned in advance ; If the checksum is bad, a branch to bad_header_checksum# occurs ; alu[accum, iphdr1, +, iphdr2] ; Add 1st and 2nd IP longword alu[accum, accum, +carry, iphdr3] ; Add 3rd IP longword alu[accum, accum, +carry, iphdr4] ; Add 4th IP longword alu[accum, accum, +carry, iphdr5] ; Add 5th IP longword alu[accum, accum, +carry, 0] ; Add any carry ld_field_w_clr[temp, 1100, accum] ; Get high 16 of the total alu_shf[accum, temp, +, accum, <<16] ; Add low 16 bits alu[temp, 1, +carry, temp, <<16] ; add last carry +1, temp B op=0 alu[accum, accum, +, temp, <<16] ; add (1+carry)<<16 to 0xffff to get zero result br!=0[bad_header_checksum#] br[end#] ; bad_header_checksum#: immed[bad_header, 0xFFFF] nop ; transactor - consume cycles end#: ctx_arb[voluntary] br[end#]

1 0

Altering a packet in memory
by Øyvind Hvamstad 21 Jun '04

21 Jun '04

Hi, I have some doubts on how to alter a packet in memory. What I basically want to do is UDP forwarding. I have to change the destination ip and port. I will also have to change the mac addresses and calculate some checksums. I have been reading the L3Fwd code and also the microcode programmers reference guide, but I still have some, potentially stupid, questions: Say I want to alter the IP address and do something like this: sdram[write, $$hdr0, ip_hdr_base, dst_ip_offset, 1] Will this write 64 bits from $$hdr0 to ip_hdr_base + dst_ip_offset in sdram? Then, must I ensure that the 32 first bits of $$hdr0 is the new IP and that the 32 next bits are the same as the original? I just want to make sure I understand this correctly. Thanks. Info: ENP-2505, embedded linux, Intel SDK 2.01 -- Øyvind Hvamstad <oyvindh(a)ifi.uio.no>

1 0

FW: [ixp1200] HELP!!
by ULStudent:Denis.Lyons 17 Jun '04

17 Jun '04

Hi again, Just resending this mail as I'm not sure if it was sent successfully last time. I desperately need some help with the following!! Can anyone provide me with code to do the following.. I want to create a lookup table in SRAM as follows: ------------------------------------------- | packet offset | signature len | signature | N matched pkts| ------------------------------------------- | | | | ------------------------------------------- As packets arrive I want to search this table and increment the N matched packets counter if a packet contains any of the signatures in the table. Any help will be greatly appreciated, Thanks for your time, Denis _______________________________________________ ixp1200 mailing list ixp1200(a)lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/ixp1200

2 1

HELP!!
by ULStudent:Denis.Lyons 14 Jun '04

14 Jun '04

Hi There, Can anyone provide me with code to do the following.. I want to create a lookup table in SRAM as follows: ------------------------------------------- | packet offset | signature len | signature | N matched pkts| ------------------------------------------- | | | | ------------------------------------------- As packets arrive I want to search this table and increment the N matched packets counter if a packet contains any of the signatures in the table. Any help will be greatly appreciated, Thanks for your time, Denis

1 0

fixing module versioning Was: RE: booting over PCI (ENP boards)...
by Willem de Bruijn 12 Jun '04

12 Jun '04

He Ricardo, from what I read your problem really isn't IXP specific. To get modules loaded in the first place I suggest you remove the EXTRAVERSION tag. Linux fails to load modules built for other systems, but in most cases 2.4.18 modules can be easily swapped. This is especially true of the pciDg drivers, as they're fairly simple. Simply rebuild them for 2.4.18 (or switch your kernel back to 2.4.17) Second, when the modules have been loaded succesfully try to look at your kernel message log with dmesg for statements regarding failure of network device detection. If nothing comes up you might want to do an ifconfig -a to see if a new device exists. I hope this helps. For more information on modules you might want to check out the kernel module programming guide or the all round excellent resource Linux Device Drivers published by O'Reilly and freely available at http://www.xml.com/ldd/chapter/book/ Willem -- Willem de Bruijn +31 6 2695 2446 +31 71 517 7174 wdebruij_at_dds.nl http://www.wdebruij.dds.nl/ current project : Fairly Fast Packet Filter (FFPF) http://ffpf.sourceforge.net/ -- Dit bericht is gescand op virussen en andere gevaarlijke inhoud door ULCN MailScanner en het bericht lijkt schoon te zijn. This message has been scanned for viruses and dangerous content by ULCN MailScanner, and is believed to be clean.

1 0

RE: [ixp1200] How to duplicate a packet in memory?
by Johnson, Erik J 10 Jun '04

10 Jun '04

Hi Øyvind I am unaware of any magic instructions. My main suggestion is to consider alternative buffer representations. If it is common for your application to perform this type of "multicast", a buffer representation that had reference counting capabilities (so you can copy by only increasing the reference count) and non-contiguous data layout (for example, like mbufs so that the multiple owners of packets and modify, say, the packet headers independently while sharing the packet payload) will probably provide the best performance. The downside is integrating a new buffer scheme with the existing SDK, which will require touching several of the SDK components (rx, tx, resource manager). If you don't want to write a new buffer data structure, my only other thought would be to dedicate a microengine to buffer copying. This ME could receive packet (handles) on one queue and return the duplicated handle on another queue. The implementation of the buffer copy could obviously use multiple threads in the ME, but could also include a single thread working on copying multiple packets at once if you need to hide even more memory latency. erik ~~~~~~~~~~~~~~~~~~~~~~~~ Erik J. Johnson erik.j.johnson(a)intel.com Not speaking for Intel ~~~~~~~~~~~~~~~~~~~~~~~~ -----Original Message----- From: ixp1200-bounces(a)lists.cs.princeton.edu [mailto:ixp1200-bounces@lists.cs.princeton.edu] On Behalf Of Øyvind Hvamstad Sent: Thursday, June 10, 2004 6:58 AM To: ixp1200(a)lists.cs.princeton.edu Subject: [ixp1200] How to duplicate a packet in memory? Hi, I'm wondering what is the best way to duplicate a packet in SDRAM? The reason for this is that I want to send one copy of the packet to the StrongARM and the other copy will be forwarded to a different host after some modifications. 1. I could loop in microcode and use sdram[read] -> sdram[write] in sequence until I reach the end of the packet. 2. I can copy it using the StrongARM, but this would probably mean that the whole forwarding process will be stalled until the copying is finished. 3. There is some 'magic' instruction that does this for me, that I'm unaware of. I'd much appreciate it if someone could give me a clue about what is the best practice here? Info: ENP-2505, embedded linux, Intel SDK 2.01 -- Øyvind Hvamstad <oyvindh(a)ifi.uio.no> _______________________________________________ ixp1200 mailing list ixp1200(a)lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/ixp1200

1 0