>Guys, I'm planning on sending this Linuswards in about two days time. >There's never a right time for this sort of thing and there may be some >relatively trivial breakage for architectures I don't test for. But it >kills 1300 lines of code. > >So. Last call, please...
Current ia64 code in Linus' tree includes code in arch/ia64/kernel/mca.c
like this:
#ifdef CONFIG_SMP
/* read_trylock() would be handy... */
if (!tasklist_lock.write_lock)
read_lock(&tasklist_lock);
#endif
which breaks all the SMP ia64 builds with your patch applied. Keith
Owens has already cleaned up this code to use read_trylock() in my
test tree (and thus in -mm). But that is part of a much bigger change
that adds a couple of new functions to kernel/sched.c that Keith
only posted to LKML yesterday (http://tinyurl.com/8urfm), since Zwane
has already posed a question about this change it seems unlikely that
this will be resolved before the cut-off for big merges.
So you need to add the attached patch to yours before sending to Linus.
-Tony
ia64-fix-spinlock.patch
Description: ia64-fix-spinlock.patch
