tools/testing/nvdimm: fix nfit_test buffer overflow
authorYasunori Goto <y-goto@jp.fujitsu.com>
Thu, 15 Jun 2017 05:04:16 +0000 (14:04 +0900)
committerDan Williams <dan.j.williams@intel.com>
Thu, 15 Jun 2017 21:31:41 +0000 (14:31 -0700)
commita117699c6c4a4b1b4e90ed51e393590986567cb4
tree3d939774a4265cbb69bd5792a0e69ee6161cfa1b
parent8990cdf10cf50dc68aaf5a8479b04304d05f1581
tools/testing/nvdimm: fix nfit_test buffer overflow

The root cause of panic is the num_pm of nfit_test1 is wrong.
Though 1 is specified for num_pm at nfit_test_init(), it must be 2,
because nfit_test1->spa_set[] array has 2 elements.

Since the array is smaller than expected, the driver breaks other area.
(it is often the link list of devres).

As a result, panic occurs like the following example.

    CPU: 4 PID: 2233 Comm: lt-libndctl Tainted: G           O    4.12.0-rc1+ #12
    RIP: 0010:__list_del_entry_valid+0x6c/0xa0
    Call Trace:
     release_nodes+0x76/0x260
     devres_release_all+0x3c/0x50
     device_release_driver_internal+0x159/0x200
     device_release_driver+0x12/0x20
     bus_remove_device+0xfd/0x170
     device_del+0x1e8/0x330
     platform_device_del+0x28/0x90
     platform_device_unregister+0x12/0x30
     nfit_test_exit+0x2a/0x93b [nfit_test]

Cc: <stable@vger.kernel.org>
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
tools/testing/nvdimm/test/nfit.c