leong%CMU-ITC-LINUS@PT.CS.CMU.EDU (John Leong) (11/08/85)
We have encountered a nasty little UNIX 4.2 bug in the way it processes IP options. If the option length field is 0, it will loop forever. It happened that someone (with a PC) accidentally generated an IP packet with such incorrect option parameter and it also happened that the packet was sent as an IP broadcast .... quite spectacularly, all our 30 odd SUN's and miscellaneous 4.2 machines died instantly ...... Is that a known bug and will that be fixed in 4.3 ???? John Leong@* leong%cmu-itc-linus@@cmu-cs-h
JNC@MIT-XX.ARPA ("J. Noel Chiappa") (11/10/85)
Larry Allen found this several years ago and reported it to Berkeley. Perhaps they have (or should have) set up a mailing list for 4.2 bug reports that people with 4.2 systems can get on, so that everyone wouldn't have to rediscover things like this? Noel -------
romkey@MIT-BORAX.ARPA (John Romkey) (11/10/85)
The bug in the PC code which sent these bogus packets was also fixed in the January '85 release. We found it one day when a VAX went away while someone was trying to TFTP files to it. Then she moved to another VAX and it happened again and we "that's absurd". Then it was a third VAX... - john romkey
bzs@BOSTONU.CSNET (Barry Shein) (11/11/85)
>From: "J. Noel Chiappa" <JNC@mit-xx.ARPA> > > Larry Allen found this several years ago and reported it to >Berkeley. Perhaps they have (or should have) set up a mailing list for >4.2 bug reports that people with 4.2 systems can get on, so that everyone >wouldn't have to rediscover things like this? > Noel There are in fact a few ways to keep up with bugs on 4.2 systems. All 4.2 tapes came with the program 'sendbug' which posts bugs to Berkeley. In addition it is generally reasonable practice to also post such bugs/fixes to unix-wizards*. Anyone responsible for 4.2 administration should be reading unix-wizards for bug reports/fixes. Further, Mt. Xinu has provided a service where either they will support your 4.2 system or provide you with bugreports. I am surprised after "several years" people still aren't aware of these things. Of course, bugs and fixes still slip through the cracks as they will on any system, and they will get re-discovered, I doubt there is any fix for that ultimately (if you think vendor supported O/S's are better, think again.) Perhaps what is really needed here is some tracking of Internet implementations by some more global organization (such as the NIC) as many of these bugs only exhibit themselves when heterogeneous environments are tried (that is, beyond a list of implementations.) One also cannot help but note that almost all internet implementations (with perhaps the exception of SUN) are barely supported (or not at all, eg: vms) by the vendors of the hardware on which the software runs, and the third party vendors are often too small to really do the job well. -Barry Shein, Boston University * USENET also has a separate address for this, net.bugs.4bsd, tho not nearly as used as unix-wizards.