exec: make de_thread() freezable
authorChanho Min <chanho.min@lge.com>
Mon, 12 Nov 2018 03:54:45 +0000 (12:54 +0900)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Mon, 19 Nov 2018 10:28:10 +0000 (11:28 +0100)
Suspend fails due to the exec family of functions blocking the freezer.
The casue is that de_thread() sleeps in TASK_UNINTERRUPTIBLE waiting for
all sub-threads to die, and we have the deadlock if one of them is frozen.
This also can occur with the schedule() waiting for the group thread leader
to exit if it is frozen.

In our machine, it causes freeze timeout as bellows.

Freezing of tasks failed after 20.010 seconds (1 tasks refusing to freeze, wq_busy=0):
setcpushares-ls D ffffffc00008ed70     0  5817   1483 0x0040000d
 Call trace:
[<ffffffc00008ed70>] __switch_to+0x88/0xa0
[<ffffffc000d1c30c>] __schedule+0x1bc/0x720
[<ffffffc000d1ca90>] schedule+0x40/0xa8
[<ffffffc0001cd784>] flush_old_exec+0xdc/0x640
[<ffffffc000220360>] load_elf_binary+0x2a8/0x1090
[<ffffffc0001ccff4>] search_binary_handler+0x9c/0x240
[<ffffffc00021c584>] load_script+0x20c/0x228
[<ffffffc0001ccff4>] search_binary_handler+0x9c/0x240
[<ffffffc0001ce8e0>] do_execveat_common.isra.14+0x4f8/0x6e8
[<ffffffc0001cedd0>] compat_SyS_execve+0x38/0x48
[<ffffffc00008de30>] el0_svc_naked+0x24/0x28

To fix this, make de_thread() freezable. It looks safe and works fine.

Suggested-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Chanho Min <chanho.min@lge.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
fs/exec.c

index fc281b738a9822a652f7d19bb60ae15acd7a7ebf..acc3a5536384cf65c6742d918cc2bd9d42305da3 100644 (file)
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -62,6 +62,7 @@
 #include <linux/oom.h>
 #include <linux/compat.h>
 #include <linux/vmalloc.h>
+#include <linux/freezer.h>
 
 #include <linux/uaccess.h>
 #include <asm/mmu_context.h>
@@ -1083,7 +1084,7 @@ static int de_thread(struct task_struct *tsk)
        while (sig->notify_count) {
                __set_current_state(TASK_KILLABLE);
                spin_unlock_irq(lock);
-               schedule();
+               freezable_schedule();
                if (unlikely(__fatal_signal_pending(tsk)))
                        goto killed;
                spin_lock_irq(lock);
@@ -1111,7 +1112,7 @@ static int de_thread(struct task_struct *tsk)
                        __set_current_state(TASK_KILLABLE);
                        write_unlock_irq(&tasklist_lock);
                        cgroup_threadgroup_change_end(tsk);
-                       schedule();
+                       freezable_schedule();
                        if (unlikely(__fatal_signal_pending(tsk)))
                                goto killed;
                }