NetBSD and Linux for Embedded Systems

Mahendra M
Mahendra_M@infosys.com
http://www.infosys.comThis work is licensed under a Creative Commons License
http://creativecommons.org/licenses/by-sa/2.5/
About the talk
This talk is NOT about to compare NetBSD and Linux and find out which is better.
–We shall compare – but not criticize !!
We shall try to understand both systems in their own merit
It is about sharing our experiences with both the systems – more from a embedded & telecom OS perspective.
We hope to help you choose.
Agenda
A short introduction to NetBSD and Linux based systems
Suitability for embedded systems
Introduction to the features available in NB and Linux
–Specifically for Embedded Systems
Feature wise comparison of the two kernels
–Core architecture, Threading model, Real time capabilities, modularity, performance, ease of deployment, portability
Linux capabilities.
NetBSD Capabilities
–Kqueue : Event notification framework
Availability of Code for future needs
Feasibility and time to market along with License issues.
About the Systems
Linux
–Refers only to the kernel
–Distributions are built around it. ( distros available for embedded systems )
–Has a history of around 15 years
–Is under the GPL license ( v2.0 ) and supports binary modules.
NetBSD
–A BSD “distribution” targeted at portability.
–Includes a kernel, libraries, config tools, scripts and build systems.
–Targeted at portability
–The BSD systems have had a history of around 25 years.
–Is under a BSD license.
Common stuff ...
Highly portable – ports exist for a large number of architectures and reference boards.
Both are POSIX compliant
Good VM, Networking, Threading subsystems
Both have been time tested in the market
Active development happening
–Directly on Linux kernel – on multiple trees maintained by different people.
–NetBSD tend to let features stabilize on FreeBSD/OpenBSD and then pick it up :-)
Let us get on with the differences....
Memory foot print
Linux is slowly getting bloated.
–There are projects that are working on reducing the size of Linux kernel
–“Linux Tiny” project provides a set of patches that can be applied to the Linux kernel to reduce the kernel size (built image and run time memory requirements)
NetBSD projects boast of reduced kernel size.
–TINY configuration file is distributed along with the source.
Both kernels can be easily reduced to a size of less than 1MB ( uncompressed : ~900k, compressed : ~400k )
With sufficient hacking around, it can be reduced even further.
Process Scheduling
Both OS follow a similar model in running user space processes.
–Time slice based – and O(1) in nature.
–They have priority based levels and processes are placed on these levels based on their priority
–Both support dynamic priority modification based on interactiveness of a process
Not needed for embedded systems – and can be turned off at compile time.
–Both support the same user space priorities (40)
Linux has two run-queues : An active and an expired run-queue.
Linux also has a run-queue per CPU – for SMP boxes.
Process Run-queue
Active
Expired
Task 1
Task 2
Task N
Task 1
Task 2
Task M
RunQueueDoubly linked lists of tasks
[Priority: 1]
[Priority:140]Migration Thread...
...
...
...
From kernel/sched.c
Other stuff ...
Kernel pre-emption.
–Linux kernel supports kernel pre-emption, but only at pre-emptible points inside in the kernel.
–There are still areas within the kernel which cannot be pre-empted.
–NetBSD does not support kernel pre-emption.
Hence real-time support is pretty poor on NetBSD systems
SMP Support ( not really critical in embedded systems )
–Very good in Linux.
–Worked upon in NetBSD.
Real-time performance
Hard real-time performance
–Not available in both kernels by default.
–FSM labs provides patches to the Linux kernel ( under GPL ) and commercial patches to NetBSD.
Works by running Linux/NetBSD kernel as low priority tasks under a real time kernel.
Soft real-time performance.
–Linux is pretty good in providing soft real time facilities.
–In addition to the 40 user land priorities, Linux provides an additional 100 real-time priorities
–Linux supports two soft real time scheduling options
SCHED_FIFO – First in first out
SCHED_RR – Round robin
–Real time tasks cannot be pre-empted.
Threading models
Both kernels are thread aware and are POSIX compliant
They support different threading models
–This is where the difference starts and things get interesting..
Linux
–Supports 1:1 model of threading ( NPTL )
–The kernel does not distinguish between threads and processes
A process is a group of thread ids – thats it.
–All threads in a process are visible as tasks to the kernel and active threads are allocated time-slices for scheduling.
All active threads will get their time-slices
–Can set real time priorities to tasks.
–APIs are available for real time threads handling
Threading models ( contd .. )
NetBSD
–Treats threads (lwp) and processes differently
–Supports Scheduler Activations.
m:n model of threading
–Not all threads are visible to the kernel scheduler. User space code takes part in telling the kernel which thread to schedule.
–The threads in a process have co-operative scheduling.
A thread can keep running for most of the time – careful programing required.
Both groups had debated the models before taking up their approaches. Both methods are “claimed” to be better than the other.
Today, Linux seems to have better thread/process creation, spawning and context switching times.
Debugging support
NetBSD has much better debugging support.
–DDB – an in-kernel debugger
–Supports kernel crash dumps
–Supports KGDB (source level debug) – remote debugging.
Linux
–Patches available for crash dumps and KDB & KGDB (not available for all architectures)
–Linux Trace Toolkit, UML etc. are also available methods.
–Has “Kernel Hacking” Option in kernel configuration
Allows debugging using “Magic SysRQs”, Debugging of actions like memory allocations, file systems, highmem, stack overflows, page allocations etc.
Also allows stats collection for some modules.
–Also, early printk() support is pretty recent.
Flash Devices and File Systems
Support is poor in NetBSD
Linux supports
–Advanced MTD support ( device concatenation, FTL, NFTL )
–JFFS2 ( JFFS3 is being worked on )
–RAM file systems and ramdisk support
–Supports initrd
In NetBSD
–OpenSource implementations of the above are not available. ( Commercial products are available )
–Results in considerable lead time in development.
–Supports MFS ( not memory efficient ) : tmpfs is being worked upon
–For boot time, it allows embedding a file-system into the kernel
KQueue : Event notification mechanism
NetBSD supports a generic event notification framework – kqueues.
Excellent replacement for select() / poll() APIs.
–No need to pass entire descriptor set to each call.
–On return, applications need not check all file descriptors for updates.
–Reduces data copying ( fd list from kernel to user space and vice-versa )
Can handle multiple types of events.
–Signals, Vnode/Process monitoring, Timer events etc.
Can club multiple occurrences of an event into a single event
New event types can be easily added.
All with just two system calls !!
Build systems and configuration
NetBSD is better
–The entire system : kernel, compilers and tools, libraries and applications can be compiled ( native or cross platform ) using a single script – build.sh
–The same script can build distributions, tarballs, archives and can also update existing systems and install fresh systems.
–Adding new components to the build framework is extremely easy.
Linux – Such systems exist but it is not as flexible as BSD
Both NetBSD and Linux provide different configuration mechanisms.
–NetBSD builds a device tree format.
–Linux provides a GUI interface for configuring the kernel...
Other differences.
Linux kernel code throws out lot of warnings during compilation. NetBSD is very clean in this aspect.
Linux supports MMU Less operations.
Loadable kernel modules
–Better supported in Linux.
Better device support in Linux
With 2.6 kernel, the device driver model has vastly improved ( attend other talks at FOSS.in for more details )
Portability has been a design goal for NetBSD.
Development models are vastly different :-)
–NetBSD is more “cathedral” like !!
In Linux, you get answers easily :-)
Business related
License (violation) is a serious cause of concern
–BSD License is very liberal
–One of the main reasons why telecom companies go for BSD – eg: Juniper ( JUNOS )
Protocol stacks and third party code
–Are available usually for most BSDs and Linux.
–To be more portable, they tend to ignore benefits of one OS
New device support
–Vendors of new devices like Network Processors, etc. release code only/mainly for Linux ( kernel modules etc. ).
–Extra effort is required in such cases to port things to NetBSD.
In case of Linux, extra time is required in selecting and integrating components of the system.
Links
http://linux-bangalore.org/blug/meetings/200401/scheduler-2.6.pdf
http://www.netbsd.org/ - One site for all NetBSD stuff
http://people.freebsd.org/~jlemon/papers/kqueue.pdf- Kqueue design
http://camars.kaist.ac.kr/courses/530/97/Readings/scheduler-act.pdf- SA design
http://josh.trancesoftware.com/linux/- Details of Linux 2.6 scheduler
http://www.wasabisystems.com/gpl/linux.htm- A comparison
http://bulk.fefe.de/scalability/- Started a Battle and cleanup !!
http://lkcd.sourceforge.net/
http://kgdb.sf.net/
http://oss.sgi.com/projects/kdb/
http://www.linuxjournal.com/article/3829- Using Linux Trace Toolkit
http://www.opersys.com/LTT
http://netbsd-soc.sourceforge.net/projects/tmpfs/- TMPFS project page
http://www.selenic.com/linux-tiny- Linux Tiny home page
Finally ...
Questions ??
Thanks to
–Organizers for giving me a chance to speak at FOSS.in
–NetBSD and Linux developers who helped me during my work
–Linux kernel team for doing a wonderful job !!
Special thanks to YOU for listening...
You can contact me at :
Mahendra_M@infosys.com