Re: [PATCH 8/8] driver-core: allow enabling async probing for all modules and builtins

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, May 20, 2015 at 09:44:59AM -0700, Dmitry Torokhov wrote:
> On Wed, May 20, 2015 at 12:27:34AM -0700, Greg Kroah-Hartman wrote:
> > On Mon, Mar 30, 2015 at 04:20:10PM -0700, Dmitry Torokhov wrote:
> > > From: "Luis R. Rodriguez" <mcgrof@suse.com>
> > > 
> > > Folks wishing to test enabling async probe for all built-in drivers
> > > and/or for all modules can use
> > > __DEBUG__kernel_force_builtin_async_probe or
> > > __DEBUG__kernel_force_modules_async_probe kernel parameters.
> > > 
> > > Activating either one will taint your kernel.
> > > 
> > > Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com>
> > > [Dmitry: split off from another patch, split into 2 parameters, moved
> > > over to core_param_unsafe()]
> > > Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
> > 
> > I've dropped this from my tree as I don't want to add these options,
> > only to have to remove them later on when it's found out that these were
> > a bad idea.
> 
> OK.
> 
> >
> > I don't want to create a user api that we have to keep around for
> > forever, and this would be such a thing (specifying how the kernel
> > probing works.)
> 
> Given that they are marked as __DEBUG and taint the kernel I do not
> believe they shoudl be considered as an API/ABI. We can emphasise this
> in docs and/or kernel messages.

But they are options a user can set on the command line, and changing
command lines is a pain.  Yes, it's a bit odd name, but we don't have
any other such naming scheme for command line options, so I don't know
what to suggest here.

> >  For debugging, can't you just patch up your kernel and
> 
> I can, but I do not have all hardware in my possession to validate the
> behavior.
> 
> > test this out?  What's the real use of this?  Who do you want to enable
> > these?  And why?  What will you do with the information?
> 
> The expectation was that distribution developers might use these
> switches when evaluating whether they are ready to switch to
> asynchronous probing.

Distro developers will never do that, they have to support just too many
different hardware types.  And there's no real gain here for them.

greg k-h