* GeHao Kang | 2016-07-01 14:37:12 [+0800]: >Hi Ingo, > >When the kernel built with CONFIG_PREEMPT_RT_FULL and boot up with some >specific isolcpus, e.g. 'isolcpus=1,3', the system cannot enter the >user space and be blocked with the message, "random: nonblocking pool >is initialized". Not all combinations of CPU isolation can cause this >issue. In my test on i.MX6Q, this might happen when isolcpus is set as >'1, 3', or '2, 3'. > >I have tested CPU isolation on various kernels, 4.1.15-rt17, >4.4.12-rt19, 4.1.15, and 4.4.0. This issue does not happen on >4.1.15 and 4.4.0. The following is the detail about the test >environment and kernel message for system blocking. So it happens always with a -RT kernel? You have also nohz_full enabled which does not work (perfectly) on -RT (you should wake up on each tick). I don't have imx6q at hand and I can't reproduce the issue on x86. Do you have the problem with isolcpus= but without nohz? If this is limited to CPUs 1+3 or 2+3 I suggest magic-sysrq to perform a task-dump on all CPUs and locks to figure out what is blocking the system from entering user space. Sebastian -- To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
