]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
cgroup: don't call migration methods if there are no tasks to migrate
authorTejun Heo <tj@kernel.org>
Sat, 8 Jul 2017 11:17:02 +0000 (07:17 -0400)
committerTejun Heo <tj@kernel.org>
Sat, 8 Jul 2017 11:37:50 +0000 (07:37 -0400)
Subsystem migration methods shouldn't be called for empty migrations.
cgroup_migrate_execute() implements this guarantee by bailing early if
there are no source css_sets.  This used to be correct before
a79a908fd2b0 ("cgroup: introduce cgroup namespaces"), but no longer
since the commit because css_sets can stay pinned without tasks in
them.

This caused cgroup_migrate_execute() call into cpuset migration
methods with an empty cgroup_taskset.  cpuset migration methods
correctly assume that cgroup_taskset_first() never returns NULL;
however, due to the bug, it can, leading to the following oops.

  Unable to handle kernel paging request for data at address 0x00000960
  Faulting instruction address: 0xc0000000001d6868
  Oops: Kernel access of bad area, sig: 11 [#1]
  ...
  CPU: 14 PID: 16947 Comm: kworker/14:0 Tainted: G        W
  4.12.0-rc4-next-20170609 #2
  Workqueue: events cpuset_hotplug_workfn
  task: c00000000ca60580 task.stack: c00000000c728000
  NIP: c0000000001d6868 LR: c0000000001d6858 CTR: c0000000001d6810
  REGS: c00000000c72b720 TRAP: 0300   Tainted: GW (4.12.0-rc4-next-20170609)
  MSR: 8000000000009033 <SF,EE,ME,IR,DR,RI,LE>  CR: 44722422  XER: 20000000
  CFAR: c000000000008710 DAR: 0000000000000960 DSISR: 40000000 SOFTE: 1
  GPR00: c0000000001d6858 c00000000c72b9a0 c000000001536e00 0000000000000000
  GPR04: c00000000c72b9c0 0000000000000000 c00000000c72bad0 c000000766367678
  GPR08: c000000766366d10 c00000000c72b958 c000000001736e00 0000000000000000
  GPR12: c0000000001d6810 c00000000e749300 c000000000123ef8 c000000775af4180
  GPR16: 0000000000000000 0000000000000000 c00000075480e9c0 c00000075480e9e0
  GPR20: c00000075480e8c0 0000000000000001 0000000000000000 c00000000c72ba20
  GPR24: c00000000c72baa0 c00000000c72bac0 c000000001407248 c00000000c72ba20
  GPR28: c00000000141fc80 c00000000c72bac0 c00000000c6bc790 0000000000000000
  NIP [c0000000001d6868] cpuset_can_attach+0x58/0x1b0
  LR [c0000000001d6858] cpuset_can_attach+0x48/0x1b0
  Call Trace:
  [c00000000c72b9a0] [c0000000001d6858] cpuset_can_attach+0x48/0x1b0 (unreliable)
  [c00000000c72ba00] [c0000000001cbe80] cgroup_migrate_execute+0xb0/0x450
  [c00000000c72ba80] [c0000000001d3754] cgroup_transfer_tasks+0x1c4/0x360
  [c00000000c72bba0] [c0000000001d923c] cpuset_hotplug_workfn+0x86c/0xa20
  [c00000000c72bca0] [c00000000011aa44] process_one_work+0x1e4/0x580
  [c00000000c72bd30] [c00000000011ae78] worker_thread+0x98/0x5c0
  [c00000000c72bdc0] [c000000000124058] kthread+0x168/0x1b0
  [c00000000c72be30] [c00000000000b2e8] ret_from_kernel_thread+0x5c/0x74
  Instruction dump:
  f821ffa1 7c7d1b78 60000000 60000000 38810020 7fa3eb78 3f42ffed 4bff4c25
  60000000 3b5a0448 3d420020 eb610020 <e92309607f43d378 e9290000 f92af200
  ---[ end trace dcaaf98fb36d9e64 ]---

This patch fixes the bug by adding an explicit nr_tasks counter to
cgroup_taskset and skipping calling the migration methods if the
counter is zero.  While at it, remove the now spurious check on no
source css_sets.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: stable@vger.kernel.org # v4.6+
Fixes: a79a908fd2b0 ("cgroup: introduce cgroup namespaces")
Link: http://lkml.kernel.org/r/1497266622.15415.39.camel@abdul.in.ibm.com
kernel/cgroup/cgroup-internal.h
kernel/cgroup/cgroup.c

index 793565c057426656312ab7b31471f6c523b7ae59..8b4c3c2f2509d2b49666410b011bf604926194f3 100644 (file)
@@ -33,6 +33,9 @@ struct cgroup_taskset {
        struct list_head        src_csets;
        struct list_head        dst_csets;
 
+       /* the number of tasks in the set */
+       int                     nr_tasks;
+
        /* the subsys currently being processed */
        int                     ssid;
 
index 620794a20a339c7e10948da829fae5a6818e598b..cc53111072d84e873a873727f51034b523b79c4e 100644 (file)
@@ -2006,6 +2006,8 @@ static void cgroup_migrate_add_task(struct task_struct *task,
        if (!cset->mg_src_cgrp)
                return;
 
+       mgctx->tset.nr_tasks++;
+
        list_move_tail(&task->cg_list, &cset->mg_tasks);
        if (list_empty(&cset->mg_node))
                list_add_tail(&cset->mg_node,
@@ -2094,21 +2096,19 @@ static int cgroup_migrate_execute(struct cgroup_mgctx *mgctx)
        struct css_set *cset, *tmp_cset;
        int ssid, failed_ssid, ret;
 
-       /* methods shouldn't be called if no task is actually migrating */
-       if (list_empty(&tset->src_csets))
-               return 0;
-
        /* check that we can legitimately attach to the cgroup */
-       do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
-               if (ss->can_attach) {
-                       tset->ssid = ssid;
-                       ret = ss->can_attach(tset);
-                       if (ret) {
-                               failed_ssid = ssid;
-                               goto out_cancel_attach;
+       if (tset->nr_tasks) {
+               do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
+                       if (ss->can_attach) {
+                               tset->ssid = ssid;
+                               ret = ss->can_attach(tset);
+                               if (ret) {
+                                       failed_ssid = ssid;
+                                       goto out_cancel_attach;
+                               }
                        }
-               }
-       } while_each_subsys_mask();
+               } while_each_subsys_mask();
+       }
 
        /*
         * Now that we're guaranteed success, proceed to move all tasks to
@@ -2137,25 +2137,29 @@ static int cgroup_migrate_execute(struct cgroup_mgctx *mgctx)
         */
        tset->csets = &tset->dst_csets;
 
-       do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
-               if (ss->attach) {
-                       tset->ssid = ssid;
-                       ss->attach(tset);
-               }
-       } while_each_subsys_mask();
+       if (tset->nr_tasks) {
+               do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
+                       if (ss->attach) {
+                               tset->ssid = ssid;
+                               ss->attach(tset);
+                       }
+               } while_each_subsys_mask();
+       }
 
        ret = 0;
        goto out_release_tset;
 
 out_cancel_attach:
-       do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
-               if (ssid == failed_ssid)
-                       break;
-               if (ss->cancel_attach) {
-                       tset->ssid = ssid;
-                       ss->cancel_attach(tset);
-               }
-       } while_each_subsys_mask();
+       if (tset->nr_tasks) {
+               do_each_subsys_mask(ss, ssid, mgctx->ss_mask) {
+                       if (ssid == failed_ssid)
+                               break;
+                       if (ss->cancel_attach) {
+                               tset->ssid = ssid;
+                               ss->cancel_attach(tset);
+                       }
+               } while_each_subsys_mask();
+       }
 out_release_tset:
        spin_lock_irq(&css_set_lock);
        list_splice_init(&tset->dst_csets, &tset->src_csets);