cpu cache info in drivers base?

Hi there,

I'm just wondering what the point of the current cpu cache info within drivers/base is?
If I startup linux-next (as of today) all I get is an error message "error detecting cacheinfo..cpu0".

Which naturally comes from:

int __weak init_cache_level(unsigned int cpu)

{
	return -ENOENT;
}

Given that there is no implementation without __weak I'm wondering what the point is? Also it looks like no architecture has been converted to the new infrastructure, even though such patches have been posted in the past?

Hi Heiko,

Quoted 6 lines · Heiko Carstens

Hi there,

I'm just wondering what the point of the current cpu cache info within drivers/base is?
If I startup linux-next (as of today) all I get is an error message "error detecting cacheinfo..cpu0".

That's correct, I did post a patch[1] to suppress that given not all architectures might add support.

Quoted 11 lines

Which naturally comes from:

int __weak init_cache_level(unsigned int cpu)

{
	return -ENOENT;
}

Given that there is no implementation without __weak I'm wondering what the point is? Also it looks like no architecture has been converted to the new infrastructure, even though such patches have been posted in the past?

Yes, I asked Greg to hold off on the architecture patches for next versions as there may be conflicts(though we can have solve it using a common baseline, but I thought it was too late to try that). Also I have not got testing on few other architectures(ppc, ia64, amd). I will post those patches individually on respective arch lists and chase after the v3.19-rc1 release. Is that OK ?

Hi Heiko,

Quoted 6 lines · Heiko Carstens

Hi there,

I'm just wondering what the point of the current cpu cache info within drivers/base is?
If I startup linux-next (as of today) all I get is an error message "error detecting cacheinfo..cpu0".

That's correct, I did post a patch[1] to suppress that given not all architectures might add support.

Quoted 11 lines

Which naturally comes from:

int __weak init_cache_level(unsigned int cpu)

{
	return -ENOENT;
}

Given that there is no implementation without __weak I'm wondering what the point is? Also it looks like no architecture has been converted to the new infrastructure, even though such patches have been posted in the past?

Yes, I asked Greg to hold off on the architecture patches for next versions as there may be conflicts(though we can have solve it using a common baseline, but I thought it was too late to try that). Also I have not got testing on few other architectures(ppc, ia64, amd). I will post those patches individually on respective arch lists and chase after the v3.19-rc1 release. Is that OK ?