Description of problem: depmod messes up modules.dep, e.g. it build this entry: /lib/modules/2.4.20-20.1.2013.nptl/kernel/net/ipv4/netfilter/ipt_MASQUERADE.o: /lib/modules/2.4.20-20.1.2013.nptl/kernel/net/ipv4/netfilter/ip_tables.o \ /lib/modules/2.4.20-20.1.2013.nptl/kernel/net/ipv4/netfilter/ip_conntrack.o \ /lib/modules/2.4.20-20.1.2013.nptl/kernel/net/ipv4/netfilter/ipchains.o Obviously, the ipchains dependancy is wrong (it should depend on iptable_nat instead). This causes iptables-restore to fail (on MASQUERADE targets), which in turn leaves the machine without a packet filter. Version-Release number of selected component (if applicable): modutils-2.4.25-6 kernel-2.4.20-20.1.2013.nptl glibc-2.3.2-57 How reproducible: reproducible Steps to Reproduce: 1. depmod -a 2. modprobe ipt_MASQUERADE 3. or: service iptables start Actual results: iptables firewall doesn't get loaded Expected results: iptables firewall gets loaded
The error shows also with modutils as old as 2.4.18-2, but only with newer kernels, e.g. not with 2.4.20-18.9, but with all of these: kernel-2.4.20-20.1.2013.nptl kernel-2.4.20-20.1.2007.nptl kernel-2.4.20-20.1.2005.nptl
Forgot to mention that when insmodding the modules by hand (in the correct order), everything works fine (substituting ipchains with iptables_nat of course).
ipchains is exporting symbols, it probably shouldn't be.
Still the case with 2.4.21-1.2023: /lib/modules/2.4.21-1.2023/kernel/net/ipv4/netfilter/ipt_MASQUERADE.o: /lib/modules/2.4.21-1.2023/kernel/net/ipv4/netfilter/ip_tables.o \ /lib/modules/2.4.21-1.2023/kernel/net/ipv4/netfilter/ipchains.o \ /lib/modules/2.4.21-1.2023/kernel/net/ipv4/netfilter/ip_conntrack.o
*** Bug 90647 has been marked as a duplicate of this bug. ***
As I mentioned in #90647, nuking ipchains.o and ipfwadm.o and rerunning depmod -a is a workaround for this bug.
Why isn't this considered a blocker bug for Cambridge (#100643)? After all this _is_ a security issue.
*** Bug 100428 has been marked as a duplicate of this bug. ***
*** Bug 100763 has been marked as a duplicate of this bug. ***
Fixed in kernel-2.4.22-1.2030.nptl
Err.... -1.2030? That's a lower version number than -20.1.2024.2.36, that still has the problem. Isn't the `20.' missing in this versioning scheme?
It's deliberate. The -20 was bogus.