Hi All !
GNS3 network simulator, that I recently ported to VirtualBox, has few
new requirements, and '16 vNICs per VM' is one of them - to allow virtual Linux routers in big network
topologies.
Current limit of '8 vNICs per VM' -- is 'ok' for medium sized labs. Large topologies won't fit in.
Based on VBox patch from VirtualBox 2.1 era, I have extended it to support 16 vNICs per VM.
Testing:
Tested on Debian Linux 6.0 host (+patched VBox v4.0.6) with Debian
Linux 6.0 guest ( all 16 AMD-pcNET III vNICs worked fine ).
I submitted for review for inclusion into v4.1.0.
--------------------------------------------------------------
"Klaus Espenlaub" from vbox devteam rejected it on technical terms:
"That's easy to review: it doesn't work. Slot 20 is used by the LSILogic SCSI controller, 21 by the BusLogic SCSI controller, 22 by the LSILogic SAS controller. If any of those controllers would be enabled the VM wouldn't start any more.
There's a reason why we didn't go for so many NICs - it would exhaust the PCI slot space in the default chipset. Fixing your code to skip those already assigned slots would leave exactly one slot free. Not the kind of situation I consider desirable for a product which is nowhere near EOL.
Supporting so many NICs only makes sense with the "big iron" setup, using the ICH9 chipset, where PCI bridges allow much more devices."
--------------------------------------------------------------
I decided to keep it here, for those users who *really* need it.
WARNING: It is incompatible with virtual SCSI, so use at your own risk.
-Technologov
Experimental patch: 16 vNICs per VM
-
Technologov
- Volunteer
- Posts: 3342
- Joined: 10. May 2007, 16:59
- Location: Israel
Experimental patch: 16 vNICs per VM
- Attachments
-
- vbox-svn-16nics-2011-04-23.zip
- (1.05 KiB) Downloaded 167 times
-
Perryg
- Site Moderator
- Posts: 34369
- Joined: 6. Sep 2008, 22:55
- Primary OS: Linux other
- VBox Version: OSE self-compiled
- Guest OSses: *NIX
Re: Experimental patch: 16 vNICs per VM
So this works. But it is for a specialised subset of users. Seems to me you have done your job and have a working model for those that need it. Congratulations!
Seeing as this is not needed by 95% or more of the user base I can see where Klaus is coming from especially since the slot count is finite.
Why not compile your own version with all of the necessary improvements and make that available to the ones that need it.
One question though is can this be made to work with VirtIO or Intel?
Seeing as this is not needed by 95% or more of the user base I can see where Klaus is coming from especially since the slot count is finite.
Why not compile your own version with all of the necessary improvements and make that available to the ones that need it.
One question though is can this be made to work with VirtIO or Intel?
-
Technologov
- Volunteer
- Posts: 3342
- Joined: 10. May 2007, 16:59
- Location: Israel
Re: Experimental patch: 16 vNICs per VM
>So this works
I wouldn't post it if it wouldn't work. @_@
Yes, I hope it should work with VirtIO and Intel E1000, too. Not tested.
Also Klaus says that 16 vNICs could be supported in ICH9 chipset, so it shoulnd't cause problems as part of base product.
>Why not compile your own version with all of the necessary improvements and make that available to the ones that need it.
Because I can't compile on Windows. Only on Linux.
I wouldn't post it if it wouldn't work. @_@
Yes, I hope it should work with VirtIO and Intel E1000, too. Not tested.
Also Klaus says that 16 vNICs could be supported in ICH9 chipset, so it shoulnd't cause problems as part of base product.
>Why not compile your own version with all of the necessary improvements and make that available to the ones that need it.
Because I can't compile on Windows. Only on Linux.