[LU-12299] avoid panic for too large cpu partitions

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.13.0, Lustre 2.12.5
• Severity: 3
• Assignee: Wang Shilong (Inactive)
• Reporter: Wang Shilong (Inactive)

Description

We still possibly hit following panic if CPU partitions is larger than online cpus:

[ 3574.857103] LNet: Removed LNI 10.0.1.81@tcp
[ 3576.527707] LNet: 28138:0:(linux-cpu.c:677:cfs_cpt_table_create()) CPU partition number 2 is larger than suggested value (1), your system may have performance issue or run out of memory while under pressure
[ 3576.529391] LNetError: 28138:0:(linux-cpu.c:571:cfs_cpt_choose_ncpus()) ASSERTION( number > 0 ) failed:
[ 3576.530270] LNetError: 28138:0:(linux-cpu.c:571:cfs_cpt_choose_ncpus()) LBUG
[ 3576.530907] Pid: 28138, comm: modprobe
[ 3576.531256]
Call Trace:
[ 3576.531628]  [<ffffffffc08d680e>] libcfs_call_trace+0x4e/0x60 [libcfs]
[ 3576.532228]  [<ffffffffc08d6dcc>] lbug_with_loc+0x4c/0xc0 [libcfs]
[ 3576.532792]  [<ffffffffc08da70a>] cfs_cpt_choose_ncpus+0x81a/0x820 [libcfs]
[ 3576.533438]  [<ffffffff811e42b6>] ? kmem_cache_alloc_trace+0x1d6/0x200
[ 3576.534033]  [<ffffffffc08daa36>] cfs_cpu_init+0x2e6/0x12d0 [libcfs]
[ 3576.534613]  [<ffffffffc08e5220>] ? init_libcfs_module+0x0/0x3c0 [libcfs]
[ 3576.535232]  [<ffffffffc08e52cb>] init_libcfs_module+0xab/0x3c0 [libcfs]
[ 3576.535841]  [<ffffffff810020ea>] do_one_initcall+0xba/0x240
[ 3576.536360]  [<ffffffff81104414>] load_module+0x1f84/0x2a10
[ 3576.536865]  [<ffffffff81353070>] ? ddebug_dyndbg_module_param_cb+0x0/0x60
[ 3576.537491]  [<ffffffff81100a73>] ? copy_module_from_fd.isra.42+0x53/0x150
[ 3576.538116]  [<ffffffff81105056>] SyS_finit_module+0xa6/0xd0
[ 3576.538787]  [<ffffffff816c16d5>] system_call_fastpath+0x1c/0x21
[ 3576.539377]  [<ffffffff816c1621>] ? system_call_after_swapgs+0xae/0x146
[ 3576.539961]