'mount: page allocation failure. order:5' occurred

Issue

• Following "page allocation failure" error was logged when mount command was executed

Jun 24 05:27:01 server1 kernel: mount: page allocation failure. order:5, mode:0xd0
Jun 24 05:27:01 server1 kernel: Pid: 8039, comm: mount Tainted: P           ---------------    2.6.32-358.2.1.el6.x86_64 #1
Jun 24 05:27:01 server1 kernel: Call Trace:
Jun 24 05:27:01 server1 kernel: [<ffffffff8112c207>] ? __alloc_pages_nodemask+0x757/0x8d0
Jun 24 05:27:01 server1 kernel: [<ffffffff81166ab2>] ? kmem_getpages+0x62/0x170
Jun 24 05:27:01 server1 kernel: [<ffffffff811676ca>] ? fallback_alloc+0x1ba/0x270
Jun 24 05:27:01 server1 kernel: [<ffffffff8116711f>] ? cache_grow+0x2cf/0x320
Jun 24 05:27:01 server1 kernel: [<ffffffff81167449>] ? ____cache_alloc_node+0x99/0x160
Jun 24 05:27:01 server1 kernel: [<ffffffffa00c4661>] ? ext4_fill_super+0x11b1/0x27e0 [ext4]
Jun 24 05:27:01 server1 kernel: [<ffffffff81168219>] ? __kmalloc+0x189/0x220
Jun 24 05:27:01 server1 kernel: [<ffffffffa00c4661>] ? ext4_fill_super+0x11b1/0x27e0 [ext4]
Jun 24 05:27:01 server1 kernel: [<ffffffff8150e501>] ? wait_for_completion_interruptible+0x1/0x30
Jun 24 05:27:01 server1 kernel: [<ffffffff811844ae>] ? get_sb_bdev+0x18e/0x1d0
Jun 24 05:27:01 server1 kernel: [<ffffffffa00c34b0>] ? ext4_fill_super+0x0/0x27e0 [ext4]
Jun 24 05:27:01 server1 kernel: [<ffffffffa00be648>] ? ext4_get_sb+0x18/0x20 [ext4]
Jun 24 05:27:01 server1 kernel: [<ffffffff811838fb>] ? vfs_kern_mount+0x7b/0x1b0
Jun 24 05:27:01 server1 kernel: [<ffffffff81183aa2>] ? do_kern_mount+0x52/0x130
Jun 24 05:27:01 server1 kernel: [<ffffffff811a3cf2>] ? do_mount+0x2d2/0x8d0
Jun 24 05:27:01 server1 kernel: [<ffffffff81139f34>] ? strndup_user+0x64/0xc0
Jun 24 05:27:01 server1 kernel: [<ffffffff811a4380>] ? sys_mount+0x90/0xe0
Jun 24 05:27:01 server1 kernel: [<ffffffff8100b072>] ? system_call_fastpath+0x16/0x1b

• A few queries
  1. What is the root cause of page shortage of order:5 or higher? Is it related to the problem described in Failed GFP_ATOMIC allocations (dropped network packets) result in kernel warnings and backtrace?
  2. Why did page allocation failure of order:5 occur although page of order:5 or higher are still available (3128kB, 12048kB)? We also collected /proc/buddyinfo and it seems that page frame of order:5 or higher could be allocated

2013-06-24 05:26:48 Node 1, zone   Normal  47108   6445    239      7      0     *3      0      0      0     *1      0 
2013-06-24 05:27:18 Node 1, zone   Normal  10147      0      6      0      0     *0      0      0      0     *1      0