Talk

Challenges and techniques to use Linux in Real-Time Systems

Watch Now

45:47

EOC 2023

Login to Watch

Sergio Prado

Sergio Prado has been working with embedded systems for more than 25 years, providing consulting and training services for companies worldwide. He also writes on his blog at sergioprado.blog and contributes to several free and open-source projects, including Buildroot, Yocto Project, and the Linux kernel.

View full profile

Topics Covered

RTOS Embedded Linux Software Design Open Source

Talk

Challenges and techniques to use Linux in Real-Time Systems

Sergio Prado

45:47 EOC 2023

Sergio Prado

45:47

Is the Linux kernel a good fit for real-time systems? Can we consider Linux a deterministic and fully-preemptible kernel so it can be used in real-time applications? If so, how is it possible, and what we should be aware of? Although Linux (with the PREEMPT_RT patch) is designed to be used on Real-Time Systems, it takes more than just a kernel to make sure you can meet all real-time requirements. This talk will cover the main aspects related to designing real-time applications on Linux-based operating systems, from configuring the kernel to writing real-time user space applications and measuring latencies.

1 / 4

Please log in or create an account to test your knowledge and see the answers.

According to Sergio Prado's talk, what is the defining characteristic of a real-time system?

A That the system produces correct outputs within a specified deadline (determinism), not necessarily that it is fast.

B That the system always has the highest possible throughput to process many events per second.

C That the system minimizes average latency even if worst-case latency is unbounded.

D That the system runs as fast as possible, even if deadlines can be missed occasionally.

E That the system consumes the least possible power while processing events.

Which approach from the talk is described as letting you achieve hard real-time deadlines by running Linux as a task under a real-time kernel?

A Using a real-time extension (e.g., Xenomai) that runs Linux as a task on a real-time kernel.

B Applying PREEMPT_RT patches to stock Linux and assuming hard real-time is guaranteed.

C Running stock Linux without patches and relying on high CPU frequency for hard real-time.

D Using only user-space priority tuning (nice and pthread priorities) without kernel changes.

E Disabling interrupts globally so Linux never gets preempted.

What is the primary objective of the PREEMPT_RT patches as described in the talk?

A Make the kernel as fully preemptible as possible by minimizing time interrupts and the scheduler are disabled, thereby improving latency.

B Increase system throughput by batching interrupts and disabling preemption for longer intervals.

C Provide formal verification of the kernel so every deadline is mathematically guaranteed.

D Replace all mutexes with spinlocks to eliminate context-switch overhead and improve latency.

E Move all interrupt handling into user space to avoid kernel-induced latency.

Which practice did Sergio Prado recommend to avoid page-fault induced latency in a real-time Linux application?

A Lock application memory at startup (use mlock/mlockall and preallocate/touch stack & heap) so pages are resident before runtime.

B Rely on lazy allocation and let the kernel handle page faults during runtime to save physical memory.

C Use CLOCK_REALTIME for all timing because it is labeled 'real-time' and gives best accuracy.

D Prefer semaphores over mutexes because semaphores are simpler and always faster in real-time.

E Implement tight polling loops in user space to avoid wake-up jitter from the scheduler.

Formatting help

italics	surround text with asterisks
bold	surround text with two asterisks
hyperlink	[hyperlink](https://example.com) or just a bare URL
inline code	surround text with single `backticks`
code block (multi-line)	wrap on its own lines with three backticks: ``` your code here ```
~~strikethrough~~	surround text with ~~two tilde characters~~
quote	prefix with >

Upvotes Newest Oldest

uapeiron

Score: 2 | 3 years ago | 1 reply

Hello Sergio, thank you for the nice presentation. Excuse me If I'm making a wrong conception here but I'm wondering why the developers of the PREEMPT_RT patch decided to implement RR, FIFO, DEADLINE but no Rate-Monotonic (RM) Algorithm for real time scheduling as far as I understand RM guarantees determinism in RTOS while DEADLINE or Earliest Deadline First does not. Thanks for your answer!

sergio_pradoSpeaker

Score: 0 | 3 years ago | 1 reply

I am glad you enjoyed the talk!

Yeah, RMS is one option for real-time scheduling, but I am not sure if it guarantees determinism.

Most algorithms depend on the developer to provide some input (priority, deadline, period, etc) so the algorithm can take decisions - at least until we decide to start adding AI to schedulers! :-)

Having said that, an RMS analysis can be applied to any real-time operating system (including Linux), independent of whether the kernel has an RMS scheduler or not. RMS just says you should define the priority of tasks based on their period (more frequent tasks get a higher priority), and you can do that upfront (at design time). And on Linux, you could use SCHED_FIFO or SCHED_RR to schedule those tasks.

Let me know if you have any further questions.

uapeiron

Score: 0 | 3 years ago | no reply

Thank you so much for your answer it clarifies my doubt!!!

Partners & Sponsors

STMicroelectronics

Amazon AWS

Golioth

The Qt Company

Memfault

NXP

SEGGER Microcontroller

PE Micro

PX5

Become a Sponsor