Re: [dm-devel] [PATCH v5] fault-injection: introduce kvmalloc fallback options

[Mikulas Patocka]

On Wed, 25 Apr 2018, David Rientjes wrote:

> On Wed, 25 Apr 2018, Mikulas Patocka wrote:
> 
> > You need to enable it on boot. Enabling it when the kernel starts to 
> > execute userspace code is already too late (because you would miss 
> > kvmalloc calls in the kernel boot path).
> 
> Is your motivation that since kvmalloc() never falls back to vmalloc() on 
> boot because fragmentation is not be an issue at boot that we should catch 
> bugs where it would matter if it had fallen back?  If we are worrying 
> about falling back to vmalloc before even initscripts have run I think we 
> have bigger problems.

The same driver can be compiled directly into the kernel or be loaded as a 
module. If the user (or the person preparing distro kernel) compiles the 
driver directly into the kernel, kvmalloc should be tested on that driver, 
because a different user or distribution can compile that driver as a 
module.

Mikulas