x86/suspend: fix false positive KASAN warning on suspend/resume
authorJosh Poimboeuf <jpoimboe@redhat.com>
Fri, 2 Dec 2016 17:42:21 +0000 (11:42 -0600)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Tue, 6 Dec 2016 01:22:44 +0000 (02:22 +0100)
Resuming from a suspend operation is showing a KASAN false positive
warning:

  BUG: KASAN: stack-out-of-bounds in unwind_get_return_address+0x11d/0x130 at addr ffff8803867d7878
  Read of size 8 by task pm-suspend/7774
  page:ffffea000e19f5c0 count:0 mapcount:0 mapping:          (null) index:0x0
  flags: 0x2ffff0000000000()
  page dumped because: kasan: bad access detected
  CPU: 0 PID: 7774 Comm: pm-suspend Tainted: G    B           4.9.0-rc7+ #8
  Hardware name: Gigabyte Technology Co., Ltd. Z170X-UD5/Z170X-UD5-CF, BIOS F5 03/07/2016
  Call Trace:
    dump_stack+0x63/0x82
    kasan_report_error+0x4b4/0x4e0
    ? acpi_hw_read_port+0xd0/0x1ea
    ? kfree_const+0x22/0x30
    ? acpi_hw_validate_io_request+0x1a6/0x1a6
    __asan_report_load8_noabort+0x61/0x70
    ? unwind_get_return_address+0x11d/0x130
    unwind_get_return_address+0x11d/0x130
    ? unwind_next_frame+0x97/0xf0
    __save_stack_trace+0x92/0x100
    save_stack_trace+0x1b/0x20
    save_stack+0x46/0xd0
    ? save_stack_trace+0x1b/0x20
    ? save_stack+0x46/0xd0
    ? kasan_kmalloc+0xad/0xe0
    ? kasan_slab_alloc+0x12/0x20
    ? acpi_hw_read+0x2b6/0x3aa
    ? acpi_hw_validate_register+0x20b/0x20b
    ? acpi_hw_write_port+0x72/0xc7
    ? acpi_hw_write+0x11f/0x15f
    ? acpi_hw_read_multiple+0x19f/0x19f
    ? memcpy+0x45/0x50
    ? acpi_hw_write_port+0x72/0xc7
    ? acpi_hw_write+0x11f/0x15f
    ? acpi_hw_read_multiple+0x19f/0x19f
    ? kasan_unpoison_shadow+0x36/0x50
    kasan_kmalloc+0xad/0xe0
    kasan_slab_alloc+0x12/0x20
    kmem_cache_alloc_trace+0xbc/0x1e0
    ? acpi_get_sleep_type_data+0x9a/0x578
    acpi_get_sleep_type_data+0x9a/0x578
    acpi_hw_legacy_wake_prep+0x88/0x22c
    ? acpi_hw_legacy_sleep+0x3c7/0x3c7
    ? acpi_write_bit_register+0x28d/0x2d3
    ? acpi_read_bit_register+0x19b/0x19b
    acpi_hw_sleep_dispatch+0xb5/0xba
    acpi_leave_sleep_state_prep+0x17/0x19
    acpi_suspend_enter+0x154/0x1e0
    ? trace_suspend_resume+0xe8/0xe8
    suspend_devices_and_enter+0xb09/0xdb0
    ? printk+0xa8/0xd8
    ? arch_suspend_enable_irqs+0x20/0x20
    ? try_to_freeze_tasks+0x295/0x600
    pm_suspend+0x6c9/0x780
    ? finish_wait+0x1f0/0x1f0
    ? suspend_devices_and_enter+0xdb0/0xdb0
    state_store+0xa2/0x120
    ? kobj_attr_show+0x60/0x60
    kobj_attr_store+0x36/0x70
    sysfs_kf_write+0x131/0x200
    kernfs_fop_write+0x295/0x3f0
    __vfs_write+0xef/0x760
    ? handle_mm_fault+0x1346/0x35e0
    ? do_iter_readv_writev+0x660/0x660
    ? __pmd_alloc+0x310/0x310
    ? do_lock_file_wait+0x1e0/0x1e0
    ? apparmor_file_permission+0x18/0x20
    ? security_file_permission+0x73/0x1c0
    ? rw_verify_area+0xbd/0x2b0
    vfs_write+0x149/0x4a0
    SyS_write+0xd9/0x1c0
    ? SyS_read+0x1c0/0x1c0
    entry_SYSCALL_64_fastpath+0x1e/0xad
  Memory state around the buggy address:
   ffff8803867d7700: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   ffff8803867d7780: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  >ffff8803867d7800: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f4
                                                                  ^
   ffff8803867d7880: f3 f3 f3 f3 00 00 00 00 00 00 00 00 00 00 00 00
   ffff8803867d7900: 00 00 00 f1 f1 f1 f1 04 f4 f4 f4 f3 f3 f3 f3 00

KASAN instrumentation poisons the stack when entering a function and
unpoisons it when exiting the function.  However, in the suspend path,
some functions never return, so their stack never gets unpoisoned,
resulting in stale KASAN shadow data which can cause later false
positive warnings like the one above.

Reported-by: Scott Bauer <scott.bauer@intel.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
arch/x86/kernel/acpi/wakeup_64.S
mm/kasan/kasan.c

index 169963f471bbb2138352c2cdfd24d4f7cf9581a2..50b8ed0317a34bb1a5d0fb9296462775f62a11bc 100644 (file)
@@ -109,6 +109,15 @@ ENTRY(do_suspend_lowlevel)
        movq    pt_regs_r14(%rax), %r14
        movq    pt_regs_r15(%rax), %r15
 
+#ifdef CONFIG_KASAN
+       /*
+        * The suspend path may have poisoned some areas deeper in the stack,
+        * which we now need to unpoison.
+        */
+       movq    %rsp, %rdi
+       call    kasan_unpoison_task_stack_below
+#endif
+
        xorl    %eax, %eax
        addq    $8, %rsp
        FRAME_END
index 70c009741aab705c6bb344df1f1851d9582bc7a7..f7dd9595ef4335aab32279a1b7709f273497895d 100644 (file)
@@ -80,7 +80,14 @@ void kasan_unpoison_task_stack(struct task_struct *task)
 /* Unpoison the stack for the current task beyond a watermark sp value. */
 asmlinkage void kasan_unpoison_task_stack_below(const void *watermark)
 {
-       __kasan_unpoison_stack(current, watermark);
+       /*
+        * Calculate the task stack base address.  Avoid using 'current'
+        * because this function is called by early resume code which hasn't
+        * yet set up the percpu register (%gs).
+        */
+       void *base = (void *)((unsigned long)watermark & ~(THREAD_SIZE - 1));
+
+       kasan_unpoison_shadow(base, watermark - base);
 }
 
 /*