[dm-devel] [PATCH] kvmalloc: always use vmalloc if CONFIG_DEBUG_VM

[Matthew Wilcox]

On Fri, Apr 20, 2018 at 04:54:53PM -0400, Mikulas Patocka wrote:
> On Fri, 20 Apr 2018, Michal Hocko wrote:
> > No way. This is just wrong! First of all, you will explode most likely
> > on many allocations of small sizes. Second, CONFIG_DEBUG_VM tends to be
> > enabled quite often.
> 
> You're an evil person who doesn't want to fix bugs.

Steady on.  There's no need for that.  Michal isn't evil.  Please
apologise.

> You refused to fix vmalloc(GFP_NOIO) misbehavior a year ago (did you make 
> some progress with it since that time?) and you refuse to fix kvmalloc 
> misuses.

I understand you're frustrated, but this is not the way to get the problems
fixed.

> I tried this patch on text-only virtual machine and /proc/vmallocinfo 
> shows 614kB more memory. I tried it on a desktop machine with the chrome 
> browser open and /proc/vmallocinfo space is increased by 7MB. So no - this 
> won't exhaust memory and kill the machine.

This is good data, thank you for providing it.

> Arguing that this increases memory consumption is as bogus as arguing that 
> CONFIG_LOCKDEP increses memory consumption. No one is forcing you to 
> enable CONFIG_LOCKDEP and no one is forcing you to enable this kvmalloc 
> test too.

I think there's a real problem which is that CONFIG_DEBUG_VM is too broad.
It inserts code in a *lot* of places, some of which is quite expensive.
We would do better to split it into more granular pieces ... although
an explosion of configuration options isn't great either.  Maybe just
CONFIG_DEBUG_VM and CONFIG_DEBUG_VM_EXPENSIVE.

Michal may be wrong, but he's not evil.