Here is much easier test.
(I'm sorry I'll be absent tomorrow.)

the number of cpus is 8. ia64/NUMA.

The hang occurs when the number of tasks is not smaller than available cpus.
Can be a hint ?

==
[root@rhel51GA testpro]# cat yield.c
#include <sched.h>

int main()
{
        while (1)
                sched_yield();
}


[root@rhel51GA testpro]# cat batch-test.sh
#!/bin/bash -x

mount -t cgroup none /opt/cgroup -o cpu
mkdir /opt/cgroup/group_1
mkdir /opt/cgroup/group_2

./yield &
PIDA=$!
./yield &
PIDB=$!

while true; do
        echo $PIDA > /opt/cgroup/group_1/tasks
        echo $PIDB > /opt/cgroup/group_1/tasks
        echo $PIDA > /opt/cgroup/group_2/tasks;
        echo $PIDB > /opt/cgroup/group_2/tasks
done

[root@rhel51GA testpro]#./batech-test.sh
no hang.

[root@rhel51GA testpro]#taskset 0f ./batech-test.sh
no hang

[root@rhel51GA testpro]#taskset 03 ./batech-test.sh
hang.
Pid: 8132, CPU 0, comm:                yield
psr : 00001210085a2010 ifs : 8000000000000206 ip  : [<a000000100067c01>]    Not tainted
ip is at pick_next_task_fair+0x81/0xe0
unat: 0000000000000000 pfs : 0000000000000b1d rsc : 0000000000000003
rnat: 0000000000000000 bsps: 0000000000000000 pr  : 0000000000566959
ldrs: 0000000000000000 ccv : 0000000000000000 fpsr: 0009804c0270033f
csd : 0000000000000000 ssd : 0000000000000000
b0  : a0000001006f6ac0 b6  : a000000100076a60 b7  : a000000100067b80
f6  : 000000000000000000000 f7  : 000000000000000000000
f8  : 1003e00000000a0000007 f9  : 1003e0000004633b23e65
f10 : 1003ee04f68ea89dfb4c3 f11 : 1003e000000000000002b
r1  : a000000100d87a60 r2  : e0000000011082f0 r3  : 0000000000000060
r8  : 0000000000000000 r9  : e000000001108310 r10 : e000004080032018
r11 : 00000000f86ccc70 r12 : e00000408394fe10 r13 : e000004083940000
r14 : 0000000000000001 r15 : 0000000000000064 r16 : e0000000011089f0
r17 : ffffffffffffffff r18 : e000000001108360 r19 : 0000000000000000
r20 : e00000408003ef10 r21 : 0000000001e9555b r22 : 000000af762794d4
r23 : 00000015e1abc70b r24 : ffffffffffffe463 r25 : e00000408003ef10
r26 : 0000000000000002 r27 : e0000000011082f0 r28 : e000000001108288
r29 : a0000001008a5468 r30 : a000000100076a60 r31 : a000000100b726e0

Call Trace:
 [<a000000100013bc0>] show_stack+0x40/0xa0
                                sp=e00000408394f860 bsp=e000004083940f18
 [<a000000100014840>] show_regs+0x840/0x880
                                sp=e00000408394fa30 bsp=e000004083940ec0
 [<a000000100036fa0>] die+0x1a0/0x2a0
                                sp=e00000408394fa30 bsp=e000004083940e78
 [<a0000001000370f0>] die_if_kernel+0x50/0x80
                                sp=e00000408394fa30 bsp=e000004083940e48
 [<a000000100038260>] ia64_fault+0x1140/0x1260
                                sp=e00000408394fa30 bsp=e000004083940de8
 [<a00000010000ae20>] ia64_leave_kernel+0x0/0x270
                                sp=e00000408394fc40 bsp=e000004083940de8
 [<a000000100067c00>] pick_next_task_fair+0x80/0xe0
                                sp=e00000408394fe10 bsp=e000004083940db8
 [<a0000001006f6ac0>] schedule+0x940/0x1280
                                sp=e00000408394fe10 bsp=e000004083940d08
 [<a000000100074e20>] sys_sched_yield+0xe0/0x100
                                sp=e00000408394fe30 bsp=e000004083940ca8
 [<a00000010000aca0>] ia64_ret_from_syscall+0x0/0x20
                                sp=e00000408394fe30 bsp=e000004083940ca8
 [<a000000000010720>] __kernel_syscall_via_break+0x0/0x20
                                sp=e000004083950000 bsp=e000004083940ca8
Thanks,
-Kame

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers