spinlock contention in shrink_zone() even with vm.max_reclaims_in_progress enabled

Issue

• System hanged
• Spinlock contention in shrink_zone()

PID: 51885  TASK: ffff810a98943080  CPU: 1   COMMAND: "myprocess"
 #0 [ffff810116249f20] crash_nmi_callback at ffffffff8007c21a
 #1 [ffff810116249f40] do_nmi at ffffffff800658e5
 #2 [ffff810116249f50] nmi at ffffffff80064ecf
    [exception RIP: .text.lock.spinlock+29]
    RIP: ffffffff80064bf7  RSP: ffff8108e011f9a0  RFLAGS: 00000046
    RAX: ffffffff804aa780  RBX: 000000000003c8f0  RCX: ffffffff804aa800
    RDX: ffff8106e4c39000  RSI: ffff810650001600  RDI: ffff810650001f80
    RBP: ffff810650001600   R8: 0000000000000004   R9: 0000000000000032
    R10: ffff810650001600  R11: 00000000e011fbf8  R12: 000000000001969a
    R13: ffff810650001600  R14: ffff8108e011fbf8  R15: ffff8108e011fbf8
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
--- <NMI exception stack> ---
 #3 [ffff8108e011f9a0] .text.lock.spinlock at ffffffff80064bf7 (via _spin_lock_irq)
 #4 [ffff8108e011f9a0] shrink_inactive_list at ffffffff800ce249
 #5 [ffff8108e011fb90] shrink_zone at ffffffff80013264
 #6 [ffff8108e011fbd0] try_to_free_pages at ffffffff800cf23c
 #7 [ffff8108e011fc60] __alloc_pages at ffffffff8000f58b
 #8 [ffff8108e011fcd0] __do_page_cache_readahead at ffffffff80012fef
 #9 [ffff8108e011fd40] filemap_nopage at ffffffff8001392c
#10 [ffff8108e011fda0] __handle_mm_fault at ffffffff80008972
#11 [ffff8108e011fe60] do_page_fault at ffffffff80067202
#12 [ffff8108e011ff50] error_exit at ffffffff8005dde9
    RIP: 000000000043e320  RSP: 0000000041856c28  RFLAGS: 00010202
    RAX: 0000000000000000  RBX: 00000000000000d4  RCX: 00002aaab4054024
    RDX: 0000000000000004  RSI: 0000000041856c40  RDI: 00002aaab4041080
    RBP: 0000000000000000   R8: 0000000000000000   R9: 000000001e5a3ed8
    R10: 0000000000000000  R11: aaaaaaaaaaaaaaab  R12: 0000000041856c40
    R13: 000000001e58ec08  R14: 000000001e5a3ed8  R15: 0000000000000030
    ORIG_RAX: ffffffffffffffff  CS: 0033  SS: 002b