[RFC] shipping kernels with default modules?

[Doug Barton]

On 6/11/2011 1:17 PM, Kostik Belousov wrote:
> On Sat, Jun 11, 2011 at 02:00:20PM -0600, Warner Losh wrote:
>>
>> On Jun 11, 2011, at 12:34 PM, Doug Barton wrote:
>>> On 6/11/2011 2:21 AM, Adrian Chadd wrote:
>>>> Hi guys,
>>>>
>>>> Has there been any further thought as of late about shipping kernels
>>>> with modules only by default, rather than monolithic kernels?
>>>>
>>>> I tried this experiment a couple years ago and besides a little
>>>> trickery with ACPI module loading, it worked out fine.
>>>>
>>>> Is there any reason we aren't doing this at the moment? Eg by having a
>>>> default loader modules list populated from the kernel config file?
>>>
>>> Has anyone benchmarked monolithic vs. modular? I think that should be done before we move in this direction.
>>
>> I haven't noticed a difference, but I haven't done any specific benchmarking.
>
> There might be some measurable difference on i386, where we use dso for
> modules. As a consequence, the overhead of GOT/PLT indirection, and, more
> important, stolen %ebx on the register-starved architecture, may make
> a difference. I doubt that any difference can be measured on amd64.

And no one has ever been surprised that assumptions proved invalid in 
the light of actual testing? :)  Theorizing on this point is of 
less-than-zero utility. Who is going to volunteer to do the actual 
benchmarking?


Doug

-- 

	Nothin' ever doesn't change, but nothin' changes much.
			-- OK Go

	Breadth of IT experience, and depth of knowledge in the DNS.
	Yours for the right price.  :)  http://SupersetSolutions.com/