Re: [Tsvwg] I-D ACTION:draft-ietf-tsvwg-sctpcsum-00.txt

[Randall Stewart <[e-mail]>]

Doug:

Thanks for your response.. I think we should be able
to get a pint in london..

As far as why I specified CRC-32 ITU ... this is what
I beleive the WG was discussing... Before I think
we could adopt CRC-32c the WG would need to say this
is what they really want...

So?? what do all of you feel out there??

The ITU CRC-32 aka same as ethernet OR is CRC-32c the
one we should be going with??

R

Douglas Otis wrote:
> 
> Randall,
> 
> Sorry for my slow response as well.  The use of a reflected table means that
> either we intentionally store the crc result in reversed order which will
> occur if we blindly use the htonl() function, or we store the crc result in
> the correct byte order by using a flip_htonl() and flip_ntoh() function.
> This simple is a matter of documentation and only a small problem for those
> wishing to implement hardware.
> 
> With respect to the polynomial selection, there was a discussion on ips
> reflector as to the justification for the CRC32c rather than the standard
> polynomial now used on Ethernet, FDDI, Fibre-channel, etc.  Once larger
> packets are allowed, the reliability of this polynomial drops.  By combining
> the Ethernet polynomial found on the medium together with the Castagnoli
> polynomial, a much better result was obtained.
> 
> I have included the suggested sample code together with sample code for
> generating the polynomial table and a serial method of testing the
> polynomial.  Once you recognize that the table results in a flipped byte
> order with respect to bytes, then how these values are stored is simply a
> matter of documentation regardless of what seems correct.  I would hope that
> there could be a benefit derived from the investigations done by the IPS
> work group.  The polynomial has nothing to do with IP storage and everything
> to do with dealing with larger packets.  Do you envision the possibility
> that SCTP may be employed on jumbo frames?
> 
> Doug
> 
> /* Example CRC32 use */
> 
> #include "crc32_tab.h"
> 
> int insert_crc32(unsigned char *buffer, int length)
> {
> SCTP_message *message;
> unsigned int i;
> unsigned long original_crc32, crc32;
> 
> /* save crc value from PDU */
> if (length > NMAX  || length < NMIN)
>   return -1;
> 
> message = (SCTP_message *) buffer;
> message->common_header.crc32 = 0L;
> crc32 = CRC_INIT;
> 
> for (i = 0; i < length; i++)
> {
>   CRC32(crc32, buffer[i]);
> }
> 
> /* and insert it into the message */
> message->common_header.checksum = flip_htonl(crc32);
>     return 1;
> }
> 
> int validate_crc32(unsigned char *buffer, int length)
> {
> SCTP_message *message;
> unsigned int i;
> unsigned long original_crc32, crc32;
> 
> /* save crc value from PDU */
> if (length > NMAX  || length < NMIN)
>   return -1;
> 
> message = (SCTP_message *) buffer;
> original_crc32 = flip_ntohl(message->common_header.crc32);
> message->common_header.crc32 = 0L;
> crc32 = CRC_INIT;
> 
> for (i = 0; i < length; i++)
> {
>   CRC32(crc32, buffer[i]);
> }
> 
> if (original_crc32 != crc32)
>         return -1;
> 
> /* and insert it into the message */
> message->common_header.checksum = flip_htonl(crc32);
>     return 1;
> }
> 
> /* Example CRC32 reflected table creation */
> 
> #include <stdio.h>
> #include <stdlib.h>
> 
> #define OUTPUT_FILE   "crc32_tab.h"
> #define ETHER_POLY                      1
> 
> #ifdef ETHER_POLY
> #define CRC32_POLY                      0x04C11DB7L
> #define CRC_TYPE  "\
> /* Ethernet, FDDI, & Fibre-Channel                                     */\n\
> /* x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x^1+x^0 */\n"
> #else
> #define CRC32_POLY                      0x1EDC6F41L
> #define CRC_TYPE  "\
> /* Castagnoli93                                                        */\n\
> /* x^32+x^28+x^27+x^26+x^25+x^23+x^22+x^20+x^19+x^18+x^14+x^13+x^11+   */\n\
> /* x^10+* x^9+x^8+x^6+x^0                                              */\n\
> /* Guy Castagnoli Stefan Braeuer and Martin Herrman                    */\n\
> /* \"Optimization of Cyclic Redundancy-Check Codes
> */\n\
> /* with 24 and 32 Parity Bits\",
> */\n\
> /* IEEE Transactions on Communications, Vol. 41, No. 6, June 1993      */\n"
> #endif
> 
> FILE *tf;
> 
> unsigned long
>   reflect_8 (int b)
> {
>   int i;
>   unsigned long rb = 0;
> 
>   for (i = 0; i < 8; i++)
>     {
>       if (b & 1)
>         rb |= 1 << (7 - i);
>       b >>= 1;
>     }
>   return (rb);
> }
> 
> unsigned long
>   reflect_32 (unsigned long b)
> {
>   int i;
>   unsigned long rw = 0;
> 
>   for (i = 0; i < 32; i++)
>     {
>       if (b & 1)
>         rw |= 1 << (31 - i);
>       b >>= 1;
>     }
>   return (rw);
> }
> 
> unsigned long
>   build_crc_table (int index)
> {
>   int i;
>   unsigned long rb;
> 
>   rb = reflect_8 (index);
>   rb <<= 24;
> 
>   for (i = 0; i < 8; i++)
>     {
>       if (rb & 0x80000000L)
>         rb = (rb << 1) ^ CRC32_POLY;
>       else
>         rb <<= 1;
>     }
> 
>   return (reflect_32 (rb));
> }
> 
> /* here is a means to verify the bype order via serial equivalent */
> 
> unsigned long
>   serial_gen (unsigned char buf[], int length)
> {
>   int i;
>   int j = 0;
>   unsigned char d;
>   unsigned long crc = 0;
> 
>   while (j < length)
>     {
>       d = buf[j++];
>       for (i = 0; i < 8; i++)
>           {
>           if ((crc >> 31) ^ (d & 1))
>               crc = (crc << 1) ^ CRC32_POLY;
>           else
>               crc <<= 1;
>           d >>= 1;
>           }
>     }
> 
>   return (crc);
> }
> 
> main ()
> {
>   int i;
> 
>   printf ("\nGenerating CRC32 Table file <%s>\n", OUTPUT_FILE);
>   if ((tf = fopen (OUTPUT_FILE, "w")) == NULL)
>     {
>       printf ("Unable to open %s\n", OUTPUT_FILE);
>       exit (1);
>     }
>   fprintf (tf, "#ifndef __crc32_table_h__\n");
>   fprintf (tf, "#define __crc32_table_h__\n\n");
>   fprintf (tf, "#define CRC32_POLY 0x%08lX\n", CRC32_POLY);
>   fprintf (tf, "#define CRC32(c,d) (c=(c>>8)^crc32tab[(c^(d))&0xFF])\n");
>   fprintf (tf, "\
> /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */\n\
> /* 32 Bit Reflected CRC table generation for SCTP.                     */\n\
> /* To accomodate serial byte data being shifted out least              */\n\
> /* significant bit first, the table's 32 bit words are reflected       */\n\
> /* which flips both byte and bit MS and LS positions.  The CRC         */\n\
> /* is calculated MS bits first from the perspective of the serial      */\n\
> /* stream.  The x^32 term is implied and the x^0 term may also         */\n\
> /* be shown as +1.  The polynomial code used is 0x%08lX.            */\n%s\
> /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */\n"
>            , CRC32_POLY, CRC_TYPE);
> 
>   fprintf (tf, "\nunsigned long  crc32tab[256] =\n{\n");
>   for (i = 0; i < 256; i++)
>     {
>       fprintf (tf, "0x%08lXL, ", build_crc_table (i));
>       if ((i & 3) == 3)
>         fprintf (tf, "\n");
>     }
> 
>   fprintf (tf, "};\n\n#endif\n");
> 
>   if (fclose (tf) != 0)
>     printf ("Unable to close <%s>." OUTPUT_FILE);
>   else
>     printf ("\nThe CRC32 table has been written to <%s>.\n", OUTPUT_FILE);
> }
> 
> > -----Original Message-----
> > From: [e-mail] [mailto:[e-mail]]On Behalf Of
> > Randall Stewart
> > Sent: Friday, July 20, 2001 8:02 AM
> > To: Douglas Otis
> > Cc: Tsvwg
> > Subject: Re: [Tsvwg] I-D ACTION:draft-ietf-tsvwg-sctpcsum-00.txt
> >
> >
> > Doug:
> >
> > Now that I am back from my vacation/conference and inbetween my
> > next travel... a response or two..
> >
> > Douglas Otis wrote:
> > >
> > > Randall,
> > >
> > > I noticed a few things regarding this draft.
> > >
> > > One, you failed to also indicate that the table value results
> > in the byte
> > > order being reversed which must be considered when storing.
> > The MS Byte in
> > > the resultant CRC value is actually the LSB byte as result of the
> > > reflection.  You should indicate the correct order for storing
> > by example
> > > with code.
> >
> > The appendix code was COPIED from the ssh implementation on FreeBSD. I
> > am
> > not sure what you are getting at.. to quote the AD's
> >
> > "Send text"
> >
> >
> > >
> > > Two, this is a different polynomial than that selected by the IPS group.
> >
> > This was by choice. The CRC-32 ITU version has been well studied and
> > is well known and used extensively. SCTP may be used by IPS but is
> > NOT an IPS protocol... I think since IPS will be adding there own
> > layer of protection this is ok, since they can use the alternate
> > polynomial
> > and thus get double CRC protection...
> >
> >
> >
> > >
> > > Three, you did not start with an all one's seed which is to
> > improve leading
> > > zero detection.  For some odd reason the Ethernet also stores
> > the inverted
> > > CRC value which again is not being done here.
> >
> > I assume you are again refering to the code in the appendix... again I
> > took this right from the ssh code in FreeBSD since it was free...
> >
> > send suggested text :)
> >
> > R
> >
> >
> > >
> > > Doug
> > >
> > > >       Title           : SCTP Checksum Change
> > > >       Author(s)       : R. Stewart, C. Sharp, J. Stone
> > > >       Filename        : draft-ietf-tsvwg-sctpcsum-00.txt
> > > >       Pages           : 7
> > > >       Date            : 29-Jun-01
> > > >
> > > > SCTP [RFC2960] currently uses an Adler-32 checksum.  For small
> > > > packets, this provides weak protection against the detection of
> > > > errors. This document changes that checksum and updates SCTP to use
> > > > the CRC-32 checksum.
> > > >
> > > > A URL for this Internet-Draft is:
> > > > http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-sctpcsum-00.txt
> > > >
> > >
> > > _______________________________________________
> > > tsvwg mailing list
> > > [e-mail]
> > > http://www1.ietf.org/mailman/listinfo/tsvwg
> >
> > --
> > Randall R. Stewart
> > [e-mail] 815-342-5222 (cell phone)
> >
> > _______________________________________________
> > tsvwg mailing list
> > [e-mail]
> > http://www1.ietf.org/mailman/listinfo/tsvwg
> >

-- 
Randall R. Stewart
[e-mail] 815-342-5222 (cell phone)

_______________________________________________
tsvwg mailing list
[e-mail]
http://www1.ietf.org/mailman/listinfo/tsvwg