Thank you Paul for your comments. I have accepted most of your comments.
Please see in-line. Have also updated the webrev at

http://cr.opensolaris.org/~girishmg/tcpsig/

thanks
~Girish

Paul Wernau wrote:
>
> -----
> ipsecconf.c
>
> 840: s/alogrithm/algorithm/
>
> 3688: I know you're probably using the text from above, but this error 
> message is misleading.  In the bypass policy case, it is obvious that 
> putting an algorithm in a bypass rule is wrong.  In the TCP signature 
> case, it is not self evident.
>
> How about "tcp signature algorithm found in combination with ipsec 
> action, not allowed.
>
> Nit - be consistent with capitalization of tcp in user visible 
> messages.  I prefer the TCP to be capitalized. (Users have scripts 
> that grep for things and consistency is helpful.)
>
> 4861: s/and negotiate/or negotiate/ as any one of these parameters 
> will invalidate rule.  Also, ulp should really say ulp/proto since 
> these are synonyms in the parser.
Accepted.
>
> ----
> ipseckey.c
>
> 896: Can you expand this comment to include the bounds.  How long can 
> this string be, is it variable length, etc.
>
> 912: s/a auth string/an auth string/
Accepted.
>
> 930: You detect a value is not a printable character, then you try to 
> print it as a string.  This can potentially crash.  Doesn't this emit 
> ugly garbage at the very least?  Try entering in 
> "[CTRL-V][Backspace][CTRL-V]C" and see what happens.
Actually was allowing all the ASCII characters. Now have restricted to 
printable ASCII characters (codes 33 - 126) as specified in RFC.
>
> 3060: s/as SPI do no matter/as the SPI value does not matter/
> ----
>
> keysock.c:
>
> 2355: Please top post a small comment about the context and usage of 
> this function. (i.e. explain that it is stripping out keying material 
> and it is used for TCP signature's keysock reply)
>
> 2457: ISTR that ASSERT(0) is frowned upon.  I don't have anything to 
> back that up, though.
> ----
>
> ipsec_impl.h
>
> 231: throw a comment at the end of the line for consistency
Accepted.
>
>
> ----
>
> inet/tcp.h: 749
> tcp.c: 655, 13456
>
> tcp_opt_sig_digest and TCP_SIG_OK use a magic 16.  This should be a 
> defined constant or use TCP_OPT_SIG_LEN - 2.  It is probably worth 
> explaining the -2 if you do that.
>
> tcp.c
>
> 16224: nit. It is really a bit and not a boolean :)
Accepted.
>
> 24054-24075: again, the 20, 18, etc should really be using defined 
> constants. (TCP_MIN_HEADER_LENGTH?)
Am now using TCPOPT_SIG_LEN as those numbers refer to it.
>
> -----
>
> pfkeyv2.h
>
> Where did you come up with these numbers?  Definitely run them by 
> danmcd first (the author of pfkey v2 RFC).  I'm especially concerned 
> about SADB_AALG_MD5 being 1.
>
> See openbsd header file for instance...
>
> http://fxr.watson.org/fxr/source/net/pfkeyv2.h?v=OPENBSD#L275
Will talk with Dan.
>
>
> ---
> ipsec_util.c
> Makefile.files
> ip.h
> ip.c
> spd.c
> spdsock.c
> ip_stack.h
> tcp_sig.c
> pfpolicy.h
> netinet/tcp.h
>
> No comments from me.
>
> ----
>
> -Paul

_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to