Comments on: The Linux Kernel, Binary Modules, and ESX Server http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/ The weblog of an IT pro specializing in virtualization, storage, and servers Wed, 08 Feb 2012 17:13:47 +0000 hourly 1 http://wordpress.org/?v= By: Virtualization Short Take #17 - blog.scottlowe.org - The weblog of an IT pro specializing in virtualization, storage, and servers http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-40958 Virtualization Short Take #17 - blog.scottlowe.org - The weblog of an IT pro specializing in virtualization, storage, and servers Thu, 28 Aug 2008 14:12:12 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-40958 [...] the argument about VMware violating the GPL (I wrote about that here quite some time ago) has surfaced again. Gordon Haff promptly and rather cleanly squashes that [...] [...] the argument about VMware violating the GPL (I wrote about that here quite some time ago) has surfaced again. Gordon Haff promptly and rather cleanly squashes that [...]

]]> By: timm http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-36529 timm Sat, 22 Mar 2008 11:35:06 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-36529 >Doesn’t VMWare 3i use GPL’d drivers to >interface with hardware? every esx is using drivers derived from gpl`ed linux drivers. they are ported to the vmkernel. source at http://download3.vmware.com/software/vi/VMware-esx-public-source-3.5.0-61973.tar.gz >Doesn’t VMWare 3i use GPL’d drivers to >interface with hardware?

every esx is using drivers derived from gpl`ed linux drivers. they are ported to the vmkernel.

source at http://download3.vmware.com/software/vi/VMware-esx-public-source-3.5.0-61973.tar.gz

]]> By: fried http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-33786 fried Thu, 18 Oct 2007 01:06:37 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-33786 Doesn't VMWare 3i use GPL'd drivers to interface with hardware? Doesn’t VMWare 3i use GPL’d drivers to interface with hardware?

]]> By: slowe http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-33336 slowe Thu, 13 Sep 2007 14:28:51 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-33336 Carlos, Agreed, ESX Server 3i seems to rather definitely prove that VMkernel doesn't rely on the Linux kernel. Carlos,

Agreed, ESX Server 3i seems to rather definitely prove that VMkernel doesn’t rely on the Linux kernel.

]]> By: Carlos Miguel http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-33308 Carlos Miguel Wed, 12 Sep 2007 03:06:29 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-33308 I think VMware ESX 3i answers every question... Another defection from the Linux Camp, VMWare has extirped the tumor before it became cancer. It couldn't afford any bad press specially after it's recent IPO. VMWare outsmarts everyone again. I think VMware ESX 3i answers every question… Another defection from the Linux Camp, VMWare has extirped the tumor before it became cancer. It couldn’t afford any bad press specially after it’s recent IPO. VMWare outsmarts everyone again.

]]> By: slowe http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-33089 slowe Tue, 21 Aug 2007 15:51:28 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-33089 BTW, If you're really interested in this issue (as I am), subscribing to the comments feed might be useful. It's interesting to me that the argument in the VentureCake article starts out as "requiring Linux to load" puts vmkernel in violation of the GPL. As the discussion continues, the argument seems to shift toward "requiring Linux to run", and finally ends up (in one of the latest comments) "linked against Linux". Indeed, even Alan Cox's latest comment seems to imply that mixing GPL and non-GPL code (during execution) makes the whole entity subject to the GPL, at least as regards redistribution. There seem to be three questions at the heart of this issue: 1. Does vmkernel require the Linux kernel in order to load, i.e., does there exist a way to boostrap the vmkernel that does not involve the Linux kernel? 2. Does vmkernel require any services provided by the Linux kernel after it has loaded? 3. Does vmkernel utilize any code from the Linux kernel in its own code base? Question #2, I believe, has been pretty clearly answered by Zach and others. In addition, as you point out, John, proprietary applications written for Linux all require services provided by the Linux kernel; does this, then, make them subject to the GPL? If so, it dooms the idea of proprietary software development on Linux. Or perhaps this argument only applies to kernel modules and kernel-mode software? Question #3 has been pretty clearly answered by Zach and others. Obviously, only a code review of vmkernel would absolutely PROVE without any doubt that this is the case, but there does not appear to be any reason to believe otherwise. Only question #1 has not been clearly answered, at least in my mind. Zach clearly states that vmkernel does not need the Linux kernel after it loads (that goes back to question #2), but does a method exist of booting vmkernel that does not require the Linux kernel? If so, then the matter is closed--the entire argument rests upon "requiring the Linux kernel in order to load," and if that issue is taken away, the wind in the sails of this matter disappears. Personally, I would have to see some REALLY compelling proof that vmkernel depends heavily upon code taken from the Linux kernel or upon services that exist ONLY within the Linux kernel (thereby implying that vmkernel was written to run WITHIN the Linux kernel ONLY) in order to think that this argument has any validity to it. Otherwise, it looks like political posturing and an effort to capitalize on VMware's (now very public) success. Just my 2 cents, of course... BTW, If you’re really interested in this issue (as I am), subscribing to the comments feed might be useful.

It’s interesting to me that the argument in the VentureCake article starts out as “requiring Linux to load” puts vmkernel in violation of the GPL. As the discussion continues, the argument seems to shift toward “requiring Linux to run”, and finally ends up (in one of the latest comments) “linked against Linux”. Indeed, even Alan Cox’s latest comment seems to imply that mixing GPL and non-GPL code (during execution) makes the whole entity subject to the GPL, at least as regards redistribution.

There seem to be three questions at the heart of this issue:

1. Does vmkernel require the Linux kernel in order to load, i.e., does there exist a way to boostrap the vmkernel that does not involve the Linux kernel?
2. Does vmkernel require any services provided by the Linux kernel after it has loaded?
3. Does vmkernel utilize any code from the Linux kernel in its own code base?

Question #2, I believe, has been pretty clearly answered by Zach and others. In addition, as you point out, John, proprietary applications written for Linux all require services provided by the Linux kernel; does this, then, make them subject to the GPL? If so, it dooms the idea of proprietary software development on Linux. Or perhaps this argument only applies to kernel modules and kernel-mode software?

Question #3 has been pretty clearly answered by Zach and others. Obviously, only a code review of vmkernel would absolutely PROVE without any doubt that this is the case, but there does not appear to be any reason to believe otherwise.

Only question #1 has not been clearly answered, at least in my mind. Zach clearly states that vmkernel does not need the Linux kernel after it loads (that goes back to question #2), but does a method exist of booting vmkernel that does not require the Linux kernel? If so, then the matter is closed–the entire argument rests upon “requiring the Linux kernel in order to load,” and if that issue is taken away, the wind in the sails of this matter disappears.

Personally, I would have to see some REALLY compelling proof that vmkernel depends heavily upon code taken from the Linux kernel or upon services that exist ONLY within the Linux kernel (thereby implying that vmkernel was written to run WITHIN the Linux kernel ONLY) in order to think that this argument has any validity to it. Otherwise, it looks like political posturing and an effort to capitalize on VMware’s (now very public) success.

Just my 2 cents, of course…

]]> By: John Volker http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/comment-page-1/#comment-33080 John Volker Tue, 21 Aug 2007 06:37:55 +0000 http://blog.scottlowe.org/2007/08/19/the-linux-kernel-binary-modules-and-esx-server/#comment-33080 I find the "it depends on Linux to load" argument nonsensical: Google Earth (just like many other proprietary applications) depends on the Linux userland loader (ld.so) to load (and run). Does that mean Google Earth violates the GPL? Of course not. Google Earth does not require Linux: it would work the same on Windows. Very similarly: vmkernel depends on the Linux kernel module loader (kmod) to load (and run). Does that mean vmkernel violates the GPL? Of course not. vmkernel does not require Linux: it would work the same on Windows. In both cases, the loaded entity uses absolutely no code and no data from the loading entity. I find the “it depends on Linux to load” argument nonsensical:

Google Earth (just like many other proprietary applications) depends on the Linux userland loader (ld.so) to load (and run). Does that mean Google Earth violates the GPL? Of course not. Google Earth does not require Linux: it would work the same on Windows.

Very similarly:

vmkernel depends on the Linux kernel module loader (kmod) to load (and run). Does that mean vmkernel violates the GPL? Of course not. vmkernel does not require Linux: it would work the same on Windows.

In both cases, the loaded entity uses absolutely no code and no data from the loading entity.

]]>