https://bugs.linaro.org/show_bug.cgi?id=3867
--- Comment #1 from Josep Puigdemont <josep.puigdem...@linaro.org> --- I tried mapping the pages with the MAP_POPULATE flag, which fixed the timer test on 1GB huge pages. On my laptop all tests passed fine, however, on the lab I observe that the l2fwd_simple test hangs. The stack trace shows that the test is stuck at futex_wait_queue_me in the kernel, and consuming quite a lot of CPU: [<ffffffff8e50feca>] futex_wait_queue_me+0xca/0x130 [<ffffffff8e5107fe>] futex_wait+0x10e/0x250 [<ffffffff8e5132cf>] do_futex+0x32f/0x510 [<ffffffff8e5135eb>] SyS_futex+0x13b/0x180 [<ffffffff8ed0e8bb>] entry_SYSCALL_64_fastpath+0x1e/0xa9 [<ffffffffffffffff>] 0xffffffffffffffff The other process in the test is waiting at __skb_wait_for_more_packets: [<ffffffff8ebc263f>] __skb_wait_for_more_packets+0x11f/0x170 [<ffffffff8ebc38bf>] __skb_recv_datagram+0x6f/0xc0 [<ffffffff8ebc394f>] skb_recv_datagram+0x3f/0x60 [<ffffffff8ec8416f>] unix_accept+0x8f/0x170 [<ffffffff8ebb1a89>] SYSC_accept4+0xf9/0x210 [<ffffffff8ebb4090>] SyS_accept+0x10/0x20 [<ffffffff8ed0e8bb>] entry_SYSCALL_64_fastpath+0x1e/0xa9 [<ffffffffffffffff>] 0xffffffffffffffff I am not sure if this situation is due to the flag being used, or if using it just exacerbates another problem, because I see this behavior also when adding this simple read after mapping a page: https://github.com/joseppc/odp/commit/3093bde3fad7db700fd6e538d68c39b4f6f3558c I saw a test hanging at futex_wait_queue_me in some other test, but after a while the test continued normally... -- You are receiving this mail because: You are on the CC list for the bug. You are the assignee for the bug.
