On Friday, 22 July 2005 at 23:12:30 -0700, Doug Barton wrote: > Changing venue to -current, since these kinds of protracted discussions > don't belong on the commit mailing lists. > > Greg 'groggy' Lehey wrote: > >> But for whatever reasons, many systems seem to have incorrect >> random(4) initialization. > > I don't think it has anything to do with /dev/random initialization, I > think that there is an error somewhere in the fortune code that is causing > this problem. You should take a look at what I committed. It simply uses the microsecond value returned by getlocaltime() for the automatic seeding by srandomdev(). It fixes the problem. I can see only two explanations: 1. srandomdev(), random(4) or friends are broken. 2. random(4) has been initialized incorrectly. Currently I'm guessing (2), but I don't care much either way. > I have seen the same problem Greg has for a long time now, > but for me it's the fortune with the CVS $FreeBSD$ tag that comes up about > 8 times out of 10. This is still true even on a system that has been up and > running for hours, and has kern.random.sys.seeded: 1 > > I can't tell if it's a problem with how the files are randomized in the > first place, or how they get played back, but I haven't looked at this very > hard yet. The obvious thing is to try this patch and see what happens. >> You'll recall the debate about removing Rush Limbaugh fortunes recently > > I didn't see this debate as I've had e-mail "issues" for the last couple of > days, what list was it on? This was early 2004, it seems. I've told you the details on IRC (and I've dropped the buffer from Emacs, so I've forgotten the details; sorry to the rest of you, but ISTR it was a message from Stephen Mckay to -chat on 4 January 2004). > In any case, I would not be at all happy if these fortunes were > removed. I agreed to add them to the "offensive" fortune database > because the rule there is that you don't add -o or -a to your > command line unless you are willing to run the risk of being > offended. If every committer gets to go through the fortune > database and remove every fortune that they find offensive, it's > going to be a mighty small file. There was no discussion of that. If you go back to the thread I mentioned, you'll see that, though I don't like them, I found it inappropriate to remove them. >> This is the only place where it seems to make any difference, so >> it's easier to use a different seed. > > Since my thesis is that this is not a /dev/random initialization > problem, I tend to agree with you. Just to be on the safe side, I > have bumped up both the size and number of my /var/db/entropy files, > and I'll gladly be proven wrong on this one. However, I sincerely > doubt that this is actually a problem with /dev/random. It's certainly worth following up on. From the PoV of fortune, it's really not worth worrying about the most perfect random number generator; I suspect that the microsecond seed will satisfy everybody except possibly the purists. That doesn't mean that we should forget the potential issues with random(4). Greg -- The virus once contained in this message has lost interest in life, shrivelled up and died. LEMIS anti-virus has given it an appropriate burial. For further details see http://www.lemis.com/grog/lemis-virus.html Finger grog_at_FreeBSD.org for PGP public key. See complete headers for address and phone numbers.
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:39 UTC