Re: regarding testing of modular sysv IPC

Dmitry Torokhov (
Sat, 05 Dec 1998 23:22:33 +0300


I tried to compile your patch for x86 SMP system, but, unfortunately,
there is unresolved symbol - smp_flush_tlb (It's used all over pgtable.h).
For SMP this function is in smp.c module and is not present
in kernel's exported symbols. (For UP it's analog defined right in the

Also in Alan's patches find_vma function was removed from mm.h and
declared as external - so there could be another unresolved symbol.

Keep hacking :) I'll be waiting for your next patch.

S. y.

Zack Weinberg wrote:

> I've now gotten a pile of success reports and suggestions for tests I
> can do - thanks people! - but I haven't heard much if anything about
> the cases I'm worried about: races between module load/unload and
> create/delete of IPC objects, semaphore undo requests that linger
> after the module is unloaded, resource leaks, and use miscounts. In
> other words, I'm looking for problems with the module code, not IPC
> itself.
> There are four KERN_ERR printk's in the patch. If the patch is
> correct, it should be impossible to trigger any of those printk's.
> I'd like testers to focus on verifying that this is indeed impossible.
> I have done some testing of this myself, but the more people hit it
> the better.
> I'm particularly interested in results for SMP systems, which I cannot
> test myself.
> Thanks again for all the help.
> zw

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at