Dean Anderson <[EMAIL PROTECTED]> writes: > On Tue, 17 Apr 2007, Russ Allbery wrote:
>> Uh, where did you get static linking from? I don't remember ever >> statically linking an AFS kernel module on Solaris. > If solaris,linux,etc remove dynamic system calls, then one will have to > statically link syscalls into the kernel, or use a dynamically loaded > pseudo driver. At least on Linux, building the module statically doesn't help. Either you can add things to the system call table, in which case you can do it dynamically anyway, or you can't, in which case building the module static doesn't help and the only thing you can do is patch the kernel source. >> What are we even talking about at this point? You're describing a >> method of handling kernel interfaces that I don't recognize at all. > Really? What did we do before there were dynamic modules? I recall > having to relink the kernel with vendor-provided objects back in the old > days. Yeah, but that doesn't get you a new system call table, at least on Linux. You'd need to replace portions of the kernel source, not just link in your own objects. My assumption is that Solaris, should they ever disable dynamic addition of system calls, would work the same way. I don't know if openafs will even build as a static module on Linux. > I think you missed the future tense in my statement above. What happens > when Linux, etc removes dynamic system calls? Then we have to stop using system calls completely unless we expect all AFS clients on Linux to use modified kernels, and by modified I mean actually changing the source of the kernel, not just linking in some additional objects at build time. -- Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/> _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
