staging: android: lowmemorykiller: Don't try to kill the same pid over and over
authorSan Mehat <san@google.com>
Mon, 26 Apr 2010 22:11:04 +0000 (15:11 -0700)
committerGreg Kroah-Hartman <gregkh@suse.de>
Wed, 30 Nov 2011 11:51:36 +0000 (20:51 +0900)
  Under certain circumstances, a process can take awhile to
handle a sig-kill (especially if it's in a scheduler group with
a very low share ratio). When this occurs, lowmemkiller returns
to vmscan indicating the process memory has been freed - even
though the process is still waiting to die. Since the memory
hasn't actually freed, lowmemkiller is called again shortly after,
and picks the same process to die; regardless of the fact that
it has already been 'scheduled' to die and the memory has already
been reported to vmscan as having been freed.

  Solution is to check fatal_signal_pending() on the selected
task, and if it's already pending destruction return; indicating
to vmscan that no resources were freed on this pass.

Signed-off-by: San Mehat <san@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
drivers/staging/android/lowmemorykiller.c

index 32314e85e2b7615cbe3770260523dc4adbed6ed6..83721786dead556ae2f084dc5f476f1471474d2a 100644 (file)
@@ -133,6 +133,12 @@ static int lowmem_shrink(int nr_to_scan, gfp_t gfp_mask)
                             p->pid, p->comm, oom_adj, tasksize);
        }
        if (selected) {
+               if (fatal_signal_pending(selected)) {
+                       pr_warning("process %d is suffering a slow death\n",
+                                  selected->pid);
+                       read_unlock(&tasklist_lock);
+                       return rem;
+               }
                lowmem_print(1, "send sigkill to %d (%s), adj %d, size %d\n",
                             selected->pid, selected->comm,
                             selected_oom_adj, selected_tasksize);