tegra: nvmap: fix the possible dead lock

tegra: nvmap: fix the possible dead lock

bug 200107851

When a process call nvmap_iovmm_get_client_mss it will hold mutex_lock(&h->lock)

and try to get down_read(&mm->mmap_sem), at the same time if the related

process is calling nvmap_vma_close, it have hold mm->mmap_sem for write when

it entering vm_munmap, and try to get mutex_lock(&h->lock). Then the problem

happen, many process become "uninterruptable sleep" with the two process.

Then bugreport will not work.

Try to fix the problem and let nvmap_iovmm_get_client_mss get mmap_sem firstly

then it will align with nvmap_vma_close.

The call stack:

<6>[68949.696132] memory_usage.sh D ffffffc000086b1c    0 2771     1 0x00000000
<7>[68949.696140] Call trace:
<7>[68949.696146] [<ffffffc000086b1c>] __switch_to+0x3c/0x48
<7>[68949.696151] [<ffffffc000b90410>] __schedule+0x4b4/0x5e0
<7>[68949.696156] [<ffffffc000b90588>] schedule+0x4c/0x68
<7>[68949.696161] [<ffffffc000b91234>] __down_read+0xb8/0xe0
<7>[68949.696166] [<ffffffc000b8f85c>] down_read+0x28/0x38
<7>[68949.696174] [<ffffffc0004705b8>] nvmap_iovmm_get_client_mss+0xf8/0x1ac
<7>[68949.696180] [<ffffffc0004706fc>] nvmap_debug_iovmm_procrank_show+0x90/0x124
<7>[68949.696185] [<ffffffc0001c8b94>] seq_read+0x178/0x3d4
<7>[68949.696190] [<ffffffc0001a30c8>] vfs_read+0x94/0x158
<7>[68949.696195] [<ffffffc0001a3a78>] SyS_read+0xbc/0x16c
<6>[68949.696870] nvtest64       D ffffffc000086b1c    0 9220     1 0x00000000
<7>[68949.696877] Call trace:
<7>[68949.696883] [<ffffffc000086b1c>] __switch_to+0x3c/0x48
<7>[68949.696888] [<ffffffc000b90410>] __schedule+0x4b4/0x5e0
<7>[68949.696893] [<ffffffc000b90588>] schedule+0x4c/0x68
<7>[68949.696898] [<ffffffc000b90a04>] schedule_preempt_disabled+0x10/0x24
<7>[68949.696904] [<ffffffc000b8f42c>] __mutex_lock_slowpath+0x19c/0x264
<7>[68949.696909] [<ffffffc000b8f534>] mutex_lock+0x40/0x60
<7>[68949.696914] [<ffffffc00046ede8>] nvmap_vma_close+0x64/0x234
<7>[68949.696919] [<ffffffc000182e64>] remove_vma+0x3c/0x6c
<7>[68949.696924] [<ffffffc000184110>] remove_vma_list+0x68/0x9c
<7>[68949.696928] [<ffffffc000184b60>] do_munmap+0x1d8/0x250
<7>[68949.696932] [<ffffffc000184c18>] vm_munmap+0x40/0x64
<7>[68949.696937] [<ffffffc000184cc8>] SyS_munmap+0x8c/0x11c

Change-Id: Icee67329c2dbbd05b10a7ddf0dcc4167191623d3
Signed-off-by: Xianhui Wang <xianhuiw@nvidia.com>
Reviewed-on: http://git-master/r/785721
Reviewed-by: Maneet Maneet Singh <mmaneetsingh@nvidia.com>
Reviewed-by: Vinayak Pane <vpane@nvidia.com>
Reviewed-by: Krishna Reddy <vdumpa@nvidia.com>