Mel,

I noticed the below BUG dmesgs, they can be reproduce reasonably easily.

test case: brickland2/micro/vm-scalability/300s-lru-file-mmap-read-rand

[  380.836690] BUG: Bad page map in process usemem  pte:310103e3c pmd:7b3c2a067
[  380.844622] addr:00007f7e9408a000 vm_flags:000000d1 anon_vma:          (null) mapping:ffff8807d5ce8ae0 index:454c75
[  380.856396] vma->vm_ops->fault: filemap_fault+0x0/0x394
[  380.862326] vma->vm_file->f_op->mmap: xfs_file_mmap+0x0/0x2a
[  380.868705] CPU: 47 PID: 7372 Comm: usemem Not tainted 3.14.0-wl-03657-ge014c34 #1
[  380.877243] Hardware name: Intel Corporation BRICKLAND/BRICKLAND, BIOS BKLDSDP1.86B.0031.R01.1304221600 04/22/2013
[  380.888916]  0000000000000000 ffff8807cd3dbc98 ffffffff819bd4c4 00007f7e9408a000
[  380.897581]  ffff8807cd3dbce0 ffffffff811a66d7 0000000843354067 0000000310103e3c
[  380.906096]  0000000000000000 ffff8808419cfb50 ffffea000d4747c0 0000000000000000
[  380.914549] Call Trace:
[  380.917340]  [<ffffffff819bd4c4>] dump_stack+0x4d/0x66
[  380.923160]  [<ffffffff811a66d7>] print_bad_pte+0x215/0x231
[  380.929475]  [<ffffffff811a776e>] vm_normal_page+0x43/0x77
[  380.935679]  [<ffffffff811a7c08>] unmap_single_vma+0x466/0x7bb
[  380.942263]  [<ffffffff811a8e3f>] unmap_vmas+0x55/0x81
[  380.948098]  [<ffffffff811afca5>] exit_mmap+0x76/0x12c
[  380.953952]  [<ffffffff810dba5b>] mmput+0x74/0x109
[  380.959358]  [<ffffffff810e0351>] do_exit+0x395/0x99b
[  380.965073]  [<ffffffff8148d1b7>] ? trace_hardirqs_off_thunk+0x3a/0x6c
[  380.972449]  [<ffffffff810e09d1>] do_group_exit+0x44/0xac
[  380.978522]  [<ffffffff810e0a4d>] SyS_exit_group+0x14/0x14
[  380.984686]  [<ffffffff819cc1e9>] system_call_fastpath+0x16/0x1b
[  380.991444] Disabling lock debugging due to kernel taint

Thanks,
Fengguang