Saw this warning again, and this time from the ret_from_fork path.
It seems we could clear the back chain earlier in copy_thread(), which
could cover both path, and also fix potential lockdep usage in
schedule_tail(), or exception occurred before we clear the back chain.
Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
 ret_from_kernel_thread:
        REST_NVGPRS(r1)
        bl      schedule_tail
-       li      r3,0
-       stw     r3,0(r1)
        mtlr    r14
        mr      r3,r15
        PPC440EP_ERR42
 
 _GLOBAL(ret_from_kernel_thread)
        bl      .schedule_tail
        REST_NVGPRS(r1)
-       li      r3,0
-       std     r3,0(r1)
        ld      r14, 0(r14)
        mtlr    r14
        mr      r3,r15
 
         * do some house keeping and then return from the fork or clone
         * system call, using the stack frame created above.
         */
+       ((unsigned long *)sp)[0] = 0;
        sp -= sizeof(struct pt_regs);
        kregs = (struct pt_regs *) sp;
        sp -= STACK_FRAME_OVERHEAD;