CODEWITHSUNDEEP
linuxlinux-kernel
user3277360
user3277360

Reputation: 127

How to quickly get all the processes currently being executed by a multicore processor?

I'm working on a kernel module that implements a driver.

When my read() file op is called, I can used current to get the current running process. But, when read() is entered, I also intend to get all the processes that are currently being executed by other cores on a multicore platform.

I can use for_each_process() to iterate each struct task_struct to see if it is TASK_RUNNING. But I know this cannot solve my problem, since TASK_RUNNING cannot give info about whether the process is being executed by CPU (it just means the corresponding process is runnable). Besides, I don't want to iterate all processes to save some overhead.

Any idea on how to solve this problem? Thank you!

Upvotes: 3

Views: 291

Answers (1)

Roman Zaitsev
Roman Zaitsev

Reputation: 1458

There is command echo l > /proc/sysrq-trigger, that shows active tasks on CPU ( https://en.wikipedia.org/wiki/Magic_SysRq_key ). You can send l from your read() function

Or you can try to write your own dump function. That's how it works in current linux kernel:

sysrq_handle_showallcpus ( https://elixir.bootlin.com/linux/v5.0.8/source/drivers/tty/sysrq.c#L234 ) =>

trigger_all_cpu_backtrace ( https://elixir.bootlin.com/linux/v5.0.8/source/include/linux/nmi.h#L144 ) =>

arch_trigger_cpumask_backtrace ( arch dependent )

Upvotes: 1

Related Questions