Revert "oom: oom_kill_process: fix the child_points logic"

This reverts the parent commit. I hate doing that, but it's generating
some discussion ("half of it is right"), and since I am planning on
doing the 2.6.38 release later today we can punt it to stable if
required. Let's not rock the boat right now.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Changed files
+3 -8
mm
+3 -8
mm/oom_kill.c
··· 458 458 struct mem_cgroup *mem, nodemask_t *nodemask, 459 459 const char *message) 460 460 { 461 - struct task_struct *victim; 461 + struct task_struct *victim = p; 462 462 struct task_struct *child; 463 - struct task_struct *t; 464 - unsigned int victim_points; 463 + struct task_struct *t = p; 464 + unsigned int victim_points = 0; 465 465 466 466 if (printk_ratelimit()) 467 467 dump_header(p, gfp_mask, order, mem, nodemask); ··· 487 487 * parent. This attempts to lose the minimal amount of work done while 488 488 * still freeing memory. 489 489 */ 490 - victim_points = oom_badness(p, mem, nodemask, totalpages); 491 - victim = p; 492 - t = p; 493 490 do { 494 491 list_for_each_entry(child, &t->children, sibling) { 495 492 unsigned int child_points; 496 493 497 - if (child->mm == t->mm) 498 - continue; 499 494 /* 500 495 * oom_badness() returns 0 if the thread is unkillable 501 496 */