ACPICA: Mark acpi_ut_create_internal_object_dbg() memory allocations as non-leaks
authorLarry Finger <Larry.Finger@lwfinger.net>
Sat, 12 May 2018 14:59:38 +0000 (09:59 -0500)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Sun, 27 May 2018 10:23:03 +0000 (12:23 +0200)
commit087ec15606b4a2ffef6fa3e2da4070e297f8b8f8
treefb260ce3e65247539518aac2f2d26ab196c784ff
parentdc689c5b336b9db9745654791f5cdf3ebbb54f4a
ACPICA: Mark acpi_ut_create_internal_object_dbg() memory allocations as non-leaks

In kernel 4.17.0-rcX, kmemleak reports 9 leaks with tracebacks similar to
the following:

unreferenced object 0xffff880224a077e0 (size 72):
  comm "swapper/0", pid 1, jiffies 4294892358 (age 1022.636s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 0e 01 01 00 00 00 00 01  ................
    00 00 00 80 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<000000004f506615>] acpi_ut_create_internal_object_dbg+0x4d/0x10e
    [<000000006e7730e3>] acpi_ds_build_internal_object+0xed/0x1cd
    [<00000000272b7c73>] acpi_ds_build_internal_package_obj+0x245/0x3a2
    [<000000000b64c50e>] acpi_ds_eval_data_object_operands+0x17b/0x21b
    [<00000000589647ac>] acpi_ds_exec_end_op+0x433/0x6c1
    [<000000001d69bcbf>] acpi_ps_parse_loop+0x926/0x9be
    [<000000005d6fa97d>] acpi_ps_parse_aml+0x1a2/0x4af
    [<00000000c4bef823>] acpi_ps_execute_table+0xbb/0x119
    [<00000000fd9632e4>] acpi_ns_execute_table+0x20c/0x260
    [<00000000e6ae17ac>] acpi_ns_parse_table+0x7d/0x1b3
    [<0000000008e1e148>] acpi_ns_load_table+0x8d/0x1c0
    [<000000009fc8346f>] acpi_tb_load_namespace+0x176/0x278
    [<0000000073f98b3b>] acpi_load_tables+0x6e/0xfd
    [<00000000d2ef13d2>] acpi_init+0x8c/0x340
    [<000000007da19d8d>] do_one_initcall+0x46/0x1fa
    [<0000000024681a1d>] kernel_init_freeable+0x1a2/0x237

According to gdb, the offending code is
              object =
                  acpi_ut_allocate_object_desc_dbg(module_name, line_number,
                                                   component_id);

As it is not possible to unload the ACPI code to test that this is a real
leak and not a false positive, and that only these 9 appear no matter how
long the system is up, a kmemleak_not_leak(object) call is inserted.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/acpi/acpica/utobject.c