From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michael Tremer To: development@lists.ipfire.org Subject: Re: libnfnetlink and libmnl Date: Thu, 07 Apr 2022 16:53:40 +0100 Message-ID: <7F88BEBB-3E31-4F35-9FA0-A6E785DCA25E@ipfire.org> In-Reply-To: <302e0bee-222a-a61d-1faa-7cf2c39c840b@ipfire.org> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1245258787217655914==" List-Id: --===============1245258787217655914== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable I couldn=E2=80=99t find anything using this either. Happy to be corrected if = I am wrong :) > On 7 Apr 2022, at 16:43, Adolf Belka wrote: >=20 > Hi Tom, >=20 >=20 > On 07/04/2022 17:37, Tom Rymes wrote: >> Adolf, >> Is sdparm and/or lvm2 used by the =E2=80=9Cextra HD=E2=80=9D addon? >=20 > It didn't show up with a grep of the IPFire git repository. I will have a f= urther look into it to see if they are needed to run it. >=20 > Regards, > Adolf. >> Tom >>> On Apr 7, 2022, at 7:19 AM, Adolf Belka wrote: >>>=20 >>> =EF=BB=BFHi Michael & *, >>>=20 >>> Building without >>>=20 >>> libart >>> libdnet >>> libpri >>> libsolv >>> pigz >>> sdparm >>>=20 >>> went without any problems. I have submitted a patch set removing them fro= m IPFire. >>>=20 >>> libnet needs to stay as arping requires it. >>>=20 >>> libaio is needed by lvm2. However when I looked at lvm2 in IPFire it is n= ot actually used by IPFire at all. The commands are all available but there a= re no mapped volume groups , logical groups etc. >>> Is lvm2 provided so that people can use logical volumes on additional har= d disks that might be mounted on IPFire. Presumably if people have already cr= eated lvm structures on other disk drives on the IPFire hardware then we need= to stay with lvm2 being present otherwise its removal would stop people bein= g able to access that additional hard disk(s). >>> If that is the case then libaio will also need to stay. >>>=20 >>> Regards, >>> Adolf. >>>=20 >>>> On 06/04/2022 19:02, Michael Tremer wrote: >>>> Hello, >>>>>> On 6 Apr 2022, at 17:52, Adolf Belka wrote: >>>>>=20 >>>>> Hi Michael, >>>>>=20 >>>>> On 06/04/2022 17:38, Adolf Belka wrote: >>>>>> Hi Michael, >>>>>>=20 >>>>>> On 06/04/2022 16:42, Michael Tremer wrote: >>>>>>> Hello Adolf, >>>>>>>=20 >>>>>>>> On 6 Apr 2022, at 14:18, Adolf Belka wrot= e: >>>>>>>>=20 >>>>>>>> Hi All, >>>>>>>>=20 >>>>>>>> We have both libnfnetlink and libmnl in IPFire. I have just done an = update of libmnl. >>>>>>>> There is an update of libnfnetlink but it also says that libnfnetlin= k is still deprecated and should preferably be replaced by libmnl >>>>>>>=20 >>>>>>> Oh this is very interesting. >>>>>>>=20 >>>>>>> It looks like libmnl was introduced for conntrack-tools in this commi= t: >>>>>>>=20 >>>>>>> commit a10733a5d8580b6ab8cff46235daab6547723781 >>>>>>> Author: Arne Fitzenreiter >>>>>>> Date: Thu Jan 3 14:27:11 2013 +0100 >>>>>>>=20 >>>>>>> conntrack-tools: add conntrack and needed deps. >>>>>>>=20 >>>>>>> You can try to comment out libnfnetlink and libmnl and run a clean bu= ild and see if it goes through. If so, then we can safely drop them. >>>>>>>=20 >>>>> It didn't go clean. It looks like libnetfilter_queue requires both libm= nl and libnfnetlink for compiling/building different parts of libnetfilter_qu= eue. I don't know if it is required for running but it wouldn't surprise me s= o both these will have to stay. >>>>>=20 >>>>> It's interesting that the group that says libmnl should be used in pref= erence to libnfnetlink then go and require both to still be used for one of t= heir programs. >>>> Well, we all have good intentions to keep our software clean, but very o= ften there is some legacy bit that everyone relies on and simply never gets c= leaned up. >>>>> I will continue looking through other files that you highlighted. >>>> Thank you. >>>>>=20 >>>>>=20 >>>>> Regards, >>>>>=20 >>>>> Adolf. >>>>>=20 >>>>>>>> I have run ./make.sh find-dependencies on both libmnl.so.0.2.0 and l= ibnfnetlink.so.0.2.0 and neither came up with anything. Doing a grep on the g= it repository for mnl or nfnetlink also didn't indicate anything using these. >>>>>>>> How are they being used and what needs to be done to change the usag= e of libnfnetlink to libmnl? >>>>>>>=20 >>>>>>> See above. >>>>>>>=20 >>>>>>> I believe there are a couple of other candidates for this which we sh= ould have a look at. There is libdnet that I noticed a little while ago, but = I am sure there will be plenty more when looking through the package list: >>>>>>>=20 >>>>>>> libaio >>>>>>> libart >>>>>>> libnet >>>>>>> libnl (and if something depends on it, can it not use libnl-3?) >>>>>>> libpri (if I remember correctly this was a dependency of asterisk) >>>>>>> libsolv can absolutely go for the moment, we don=E2=80=99t need it >>>>>>> netpbm? >>>>>>> pigz (we don=E2=80=99t use it anywhere since we are compressing image= s with XZ b302b9a695e391477eab0cb2343f3ba1b1ba1989) >>>>>>> sdparm >>>>>>>=20 >>>>>>> If you have the time, maybe you can have a look at what is used somew= here and what can be dropped? >>>>>> I will give that a go and use your inputs as starting points. >>>>>>=20 >>>>>> Regards, >>>>>> Adolf >>>>>>>=20 >>>>>>> This won=E2=80=99t shrink the distribution by a massive amount, but w= hy should we carry around dead code? >>>>>>>=20 >>>>>>> Best, >>>>>>> -Michael >>>>>>>=20 >>>>>>>> Regards, >>>>>>>> Adolf --===============1245258787217655914==--