MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Alexey Dobriyan]

On Thu, Dec 29, 2005 at 04:04:00PM -0800, Kees Cook wrote:
> Would patches towards adding missing MODULE_DESCRIPTION, MODULE_VERSION,
> and MODULE_AUTHOR stuff be taken?  While these aren't covered in the
> CodingStyle document, I did find reference to their preferred order in
> Documentation/i2c/porting-clients.txt where Greg KH said the order
> should be:
>
> MODULE_AUTHOR
> MODULE_DESCRIPTION
> MODULE_LICENSE /* last line of source */
>
> I'm curious where MODULE_VERSION should go, as well as
> MODULE_PARAM_DESC.
>
> Notably, AUTHOR, DESCRIPTION, and VERSION seem to be missing from the
> various examples in Documention/

I hate MODULE_VERSION. It stops making sense after the following
definition:

	Version of a module is a version of kernel it was shipped with.

Given:	module 8139too version 0.9.27 is buggy in somesuch way.
Question: which one?

There were quite a few nontrivial changes made since transition to git:
-----------------------------------------------------------------------
Christoph Lameter:
      Cleanup patch for process freezing

Jeff Garzik:
      [netdrvr 8139too] replace hand-crafted kernel thread with workqueue
      [netdrvr 8139too] use cancel_rearming_delayed_work() to cancel thread
      [netdrvr 8139too] use rtnl_shlock_nowait() rather than rtnl_lock_interruptible()
      [netdrvr 8139too] fast poll for thread, if an unlikely race occurs

John W. Linville:
      8139too: support ETHTOOL_GPERMADDR
      8139too: fix resume for Realtek 8100B/8139D

Olaf Hering:
      turn many #if $undefined_string into #ifdef $undefined_string

Pekka Enberg:
        8139too: use iomap for pio/mmio
-----------------------------------------------------------------------
None of the above changes touched MODULE_VERSION. It's still 0.9.27.

MODULE_VERSION is almost always outdated. You can't rely on it in
bugreports. All you can rely on is kernel version, be it 2.6.15-git1 or
2.6.15-0aec63e67c69545ca757a73a66f5dcf05fa484bf.

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Carlos Manuel Duclos Vergara]

Hi everyone,

I have two ideas about what to do with MODULE_VERSION:
1.- Defining MODULE_VERSION = KERNEL_VERSION
2.- Schedule it for removal in one or two more versions, and automagically use 
the KERNEL_VERSION as module's version.

Any comments?

-- 
Carlos Manuel Duclos Vergara
http://www.toolchains.com/personal/blog

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Greg KH]

On Sun, Jan 08, 2006 at 06:55:16PM -0300, Carlos Manuel Duclos Vergara wrote:
> Hi everyone,
> 
> I have two ideas about what to do with MODULE_VERSION:
> 1.- Defining MODULE_VERSION = KERNEL_VERSION

No.

> 2.- Schedule it for removal in one or two more versions, and automagically use 
> the KERNEL_VERSION as module's version.

No, just let the authors of the different drivers that want to use
MODULE_VERSION use it (for some people it does matter, and they keep it
up to date.)  Everyone else, just don't add it if you don't care about
it.

> Any comments?

In short, leave it alone :)

thanks,

greg k-h

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Matthew Wilcox]

On Sun, Jan 08, 2006 at 06:55:16PM -0300, Carlos Manuel Duclos Vergara wrote:
> I have two ideas about what to do with MODULE_VERSION:
> 1.- Defining MODULE_VERSION = KERNEL_VERSION
> 2.- Schedule it for removal in one or two more versions, and automagically use 
> the KERNEL_VERSION as module's version.
> 
> Any comments?

Do neither.  Just because some people don't use it properly is no reason
to remove it.

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Matt Domsch]

On Sun, Jan 08, 2006 at 01:58:00PM -0800, Greg KH wrote:
> On Sun, Jan 08, 2006 at 06:55:16PM -0300, Carlos Manuel Duclos Vergara wrote:
> > Hi everyone,
> > 
> > I have two ideas about what to do with MODULE_VERSION:
> > 1.- Defining MODULE_VERSION = KERNEL_VERSION
> 
> No.
> 
> > 2.- Schedule it for removal in one or two more versions, and automagically use 
> > the KERNEL_VERSION as module's version.
> 
> No, just let the authors of the different drivers that want to use
> MODULE_VERSION use it (for some people it does matter, and they keep it
> up to date.)  Everyone else, just don't add it if you don't care about
> it.

DKMS (http://linux.dell.com/dkms/) uses it to tell if one version is
newer than another.  For the couple dozen modules Dell regularly cares
about, the module authors have been doing a good job of keeping
MODULE_VERSION correct, for exactly this reason.

> In short, leave it alone :)

Please!

-- 
Matt Domsch
Software Architect
Dell Linux Solutions linux.dell.com & www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Matt Domsch]

On Sun, Jan 08, 2006 at 01:58:00PM -0800, Greg KH wrote:
> On Sun, Jan 08, 2006 at 06:55:16PM -0300, Carlos Manuel Duclos Vergara wrote:
> > 
> > I have two ideas about what to do with MODULE_VERSION:
> > 1.- Defining MODULE_VERSION = KERNEL_VERSION

vermagic already has KERNEL_VERSION, if that's what someone really
cares about.

-- 
Matt Domsch
Software Architect
Dell Linux Solutions linux.dell.com & www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com

Re: MODULE_VERSION useless? (was Re: [KJ] adding missing MODULE_* stuffs)

[Håkon Løvdal]

On 1/8/06, Carlos Manuel Duclos Vergara <carlos@embedded.cl> wrote:
> Hi everyone,
>
> I have two ideas about what to do with MODULE_VERSION:
> 1.- Defining MODULE_VERSION = KERNEL_VERSION
> 2.- Schedule it for removal in one or two more versions, and automagically use
> the KERNEL_VERSION as module's version.
>
> Any comments?

I think there is another option:

3. Always print KERNEL_VERSION in addition to MODULE_VERSION.

Since for some modules MODULE_VERSION is considered useful and
KERNEL_VERSION always is useful, this should give the best of both, right?

BR Håkon Løvdal