> Steven N. Hirsch wrote:
>
> > I've found the underlying problem with daemonless 'kmod' on Alpha.
> > Apparently, glibc header changes are to blame. I became suspicious after
> > noticing that kmod is compiled with __KERNEL__ defined. Turns out to be a
> > kernel vs. user space disagreement as to the storage type and size of
> > sigset_t.
> >
> > The following quick hack brings kmod to life:
> >...
> > ! sigfillset(¤t->blocked);
> >...
> > ! memset (¤t->blocked, '\0xff', sizeof(current->blocked));
>
> Good detective work!
> But if this was the cause of kmod's failures, then surely
> there is something really sinister going on, either in the
> Alpha version of the kernel, or in your machine's setup.
> I mean, when compiling the kernel the compiler shouldn't
> reference user-land include files, especially not the "improved"
> ones supplied with glibc.
>
> A quick look reveals that there are other uses of sigfillset()
> in the kernel, including in fs/lockd/clntproc.c and mm/vmscan.c.
> You could have a really dysfunctional kernel there..
I agree, and I'm not too comfortable with the quick fix. Richard
Henderson is the individual with the most intimate knowledge of low-level
Alpha linux details. Perhaps he can elaborate on this and suggest a more
correct fix?
Steve
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu