On Wed, 27 Apr 2011, Mr Dash Four wrote: > > > 2. Take a note when I try to add 4608 elements to the test set - I did > > > that 3 > > > times with completely different results - it seems that ipset triggers an > > > error when it feels like it. > > > > The hashes are created with a random hash key calculation parameter: if the > > hash is lucky enough, you can stuff the data into it without reaching the > > clashing limit and without the need to resize the hash. > > > I see, so the error is triggered at random because the randomly generated hash > sometimes is able to "stuff the data", as you put it, without reaching the > limit, but sometimes it doesn't and this, then, triggers the error, is that > correct? Yes, exactly. > > > 3. Following on from the previous point, when I try to add more elements > > > than > > > the established "error threshold" in point 2 above, I also get mixed > > > results > > > (attempting to add 4864 elements gives me 2 successes and 2 failures after > > > 4 > > > attempts made when I re-created the test set from scratch). > > > > > > > The same as above: if the clash limit is not reached, the hash happily can > > accept more data. If it's reached, resizing is triggered which is > > successfully done, but adding elements starts again with the already added > > element. That needs to be fixed. > > > I think I understand it now, thanks for the explanation. I hope that you would > be able to fix this bug. I'm attending an annual conference this week, so it seems I won't have time to work on the bugfix before the weekend, sorry. > On a separate note, two days ago I was finally able to split ipset from > xtables-addons and build it as a separate package (i.e. construct a .spec > file, then build and install rpm based on that) while building only the > userspace part (assuming the kernel part will stay with the kernel). This is > by no means finished (there are some things still left to be ironed out - > kernel modules dependency check, changelog history to be added etc), but I > thought you may be interested in this. > > If so, let me know and I'll send it to you (the important part is the .spec > file and a separate patch which removes the build routines for the kernel > modules). I don't really see why you have to split ipset from xtables-addons. What's wrong with the standalone ipset package, what forces you to do so? > I am in a process of doing the same with version 4.5 of ipset (I was able to > successfully build the .35-88 version of the kernel with ipset 4.5 kernel > modules happily integrated into it yesterday), even though ipset 4.5 does not > have pkgconfig files I intend to create these - hopefully today. Best regards, Jozsef - E-mail : kadlec@xxxxxxxxxxxxxxxxx, kadlec@xxxxxxxxxxxx PGP key : http://www.kfki.hu/~kadlec/pgp_public_key.txt Address : KFKI Research Institute for Particle and Nuclear Physics H-1525 Budapest 114, POB. 49, Hungary -- To unsubscribe from this list: send the line "unsubscribe netfilter" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
