In article <00c701c1694f$b448d750$3709a8c0@netbeastie.com> you wrote: > Red Hat 7.2 is based on the 2.4.7 kernel. It's based on 2.4.7-ac (the -ac patch is also on kernel.org). However due to several security issues discovered after the release, the current 7.2 kernel is 2.4.9-ac10 based. > So I compiled my LKM for a 2.4.7 > kernel from kernel.org and thought that it would work on the same 2.4.7 > kernel found in Red Hat 7.2. I was wrong. > When my LKM loads on the Red Hat 7.2 box using insmod -f mylkm.o I get an > unresolved symbol up_and_exit. That's because the Red Hat kernel has symbol versioning; you must compile against the exact kernel version (eg the same value of CONFIG_SMP, the same CPU settings etc etc). It will refuse to load modules that are compiled against the wrong kernel version. > Further, the 2.4.8 kernel from kernel.org still has up_and_exit() defined > in kernel.h. The Red Hat Linux kernel has it too, but check the symbol versioning... > Is this a common practice of companies like Red Hat...? I would think > that if they built a 2.4.7-10 kernel that it would at least be > function-level compatable with the approved 2.4.7 kernel found at > kernel.org and certainly not take out kernel functions. It _mostly_ is. The stable/exported interface of the kernel certainly is (eg the system calls and the ioctls etc); however internal functions to the kernel nobody guarantees... > Further, given the above, how would I/others deal with cases where 2.4.7 > has a function that is not found in RedHat's own 2.4.7-10 but is found in > 2.4.8 and then goes away in 2.4.9 and above? check for the existence of the function; no way around that. Greetings, Arjan van de Ven Red Hat Linux kernel maintainer -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ IRC Channel: irc.openprojects.net / #kernelnewbies Web Page: http://www.kernelnewbies.org/
