Ubuntu Wiki

• Launchpad entry: HERE

• Created: 12-4-2006 by C.Kontros

• Packages affected:

Summary

This is the specification for a proper kernel, intended for use in professional multimedia production.

Rationale

Currently, in order to enable realtime features, Ubuntu users have to use vanilla patched kernels that are vulnerable to security issues, missing hardware support and in general, cause regressions when compared to the stock Ubuntu kernels.

Use cases

• John wants to create a mobile studio setup for audio and/or video casual editing. It should be easy enough to setup so he can focus on the work, rather than the tools themselves.
• Jack is a musician interested in doing serious audio production work. He has a specialized hardware (i.e. a multiple input/output soundcard), and expects his system to provide him with the lowest latency possible, avoiding any hiccups.
• Joan is an avid video producer who needs a responsive and stable workstation. She requires features like: fast I/O, working with big files and perfect sync between audio/video.

Scope

The intention is have an Ubuntu stock kernel built with the necessary patches/options for proper realtime capabilities.

Implementation

Ideally, it should just be a matter of recompiling the stock kernel with a different configuration (taking into account the inclusion of RT capabilities in 2.6.18+). The next is a list with those options that should be changed in this multimedia enhanced kernel:

• Preemptible Kernel (currently in -lowlatency)
• Timer frequency (1000 HZ) (currently in -lowlatency)

This is a list with possible options not yet confirmed:

• Support for Large Block Devices
• Support for tracing block io actions
• Support for Large Single Files
• Support for snd-seq
• Support for snd-seq-midi
• Support for USB Midi/Audio (as a module)

Comments

It is still quite possible that you will need the preempt patch. Especially for older machines this seems to allows them to become studio work horses... - Frink

Several tweaks are necessary for low latency Audio on USB it will require some digging and a lot of testing/tweaking. While this can be done some in modules and user-space programs there are also tweaks in the kernel... - Frink

DSSI, LAPDSA, LASH support should be built-in. For users migrating from Cubase and the like, plug-ins are very important. Rosegarden supports DSSI plugins, but I have found that there are many issues getting plugins installed with most distros. Even 64Studio has major library dependency issues pertaining to the install of DSSI plugins. Currently available DSSI softsynths should be part of the package (Hexter, WhySynth, Xsynth, Nekobee, Om, etc); to build these plugins requires a variety of libraries and dev files not part of many current builds. DSSI softsynth build dependencies should be fulfilled in Ubuntustudio. As for SoundFonts, these items are much easier to find and install; some GNU .sf files should be included in a sensible folder, but it should be expected that users will download SoundFonts to suit their purposes. My experiences with Rosegarden and multiple DSSI instruments leads me to believe that there is great potential for system instability if the builds and libraries aren't just right... Beyond these comments, being part of the Ubuntu metaverse is a huge advantage... some custom Linux builds that currently exist are missing many of the refinements that make Ubuntu the fantastic distro it is... Just because a distro is built for media creation shouldn't mean one has to sacrifice usability in other areas, insofar as that usability doesn't adversely affect the core multimedia components.... screamlab

sorry odnt know here to add this. i am not a Kernel or optimising expert but i do sincerely believe the linux kernel is wired for generic stability which is great but an emphasis could be put on extremely efficient sheduling. well done out of the box support for a non EXT3 based FS. [whichever ones you decide on] true support for ZFS as we move over to HD raw captures and HUGE files... [maybe in '08 april?]

kai che