/sys/kernel/realtime in mainline

/sys/kernel/realtime in mainline

[Alexander Werner]

Hello,

this is about testing if a kernel has been compiled with PREEMPT_RT.
Previously, the PREEMPT_RT patch created a file /sys/kernel/realtime to
enable user space tools to detect this. However, this patch has not
been merged (yet?). The commit which creates this file is
	3e7545de923c4e8184c2be0b95341fdc7ef1c193 as pointed out in thread [1]
on the LinuxCNC forum. Many applications in the robotics space also use
the existence of this file to check for PREEMPT_RT such as [2],[3].

My question is if this will be merged later on or the user space will
have to accomodate this change.


Cheers,

Alex



[1]
https://forum.linuxcnc.org/38-general-linuxcnc-questions/54542-real-time-kerel-not-detected-on-patched-6-12?start=10
[2]https://github.com/ros-controls/realtime_tools/blob/5b85370c75371a5748b790101802910e53bb12f0/src/realtime_helpers.cpp#L48
[3]
https://github.com/frankaemika/libfranka/blob/fd2951bc76e61d32d835241c40c26562148cd457/src/control_tools.cpp#L29

Re: /sys/kernel/realtime in mainline

[Sebastian Andrzej Siewior]

On 2025-01-15 10:24:41 [-0500], Alexander Werner wrote:
> Hello,

Hi,

> this is about testing if a kernel has been compiled with PREEMPT_RT.
> Previously, the PREEMPT_RT patch created a file /sys/kernel/realtime to
> enable user space tools to detect this. However, this patch has not
> been merged (yet?). The commit which creates this file is
> 	3e7545de923c4e8184c2be0b95341fdc7ef1c193 as pointed out in thread [1]
> on the LinuxCNC forum. Many applications in the robotics space also use
> the existence of this file to check for PREEMPT_RT such as [2],[3].
> 
> My question is if this will be merged later on or the user space will
> have to accomodate this change.

I was not sure if we do need this or if this information can be obtained
by some other means. For instance uname provides this information. We
don't provide the PREEMPT vs PREEMPT_DYNAMIC information. With
PREEMPT_RT we can do LAZY or PREEMPT here.

If the robotics space relies on this information I would bring this to
the scheduling people for their opinion.

> Cheers,
> 
> Alex

Sebastian