Re: Write to srvio_numvfs triggers kernel panic

[Alex Williamson <alex.williamson@redhat.com>]

On Mon, 9 May 2022 11:49:29 -0500
Bjorn Helgaas <helgaas@kernel.org> wrote:

> On Sun, May 08, 2022 at 11:07:40AM +0000, Volodymyr Babchuk wrote:
> > 
> > As you can see, output for func 0 and func 2 is identical, so yeah,
> > looks like my system reads config space for func 0 in both cases.  
> 
> They are not identical:
> 
>   01:00.0 Non-Volatile memory controller
>     Region 0: Memory at 30010000
> 
>   01:00.2 Non-Volatile memory controller
>     Region 0: Memory at 30018000
> 
> > On other hand, I'm wondering if it is correct to have both is_virtfn and
> > is_physfn in the first place, as there can 4 combinations and only two
> > (or three?) of them are valid. Maybe it is worth to replace them with
> > enum?  
> 
> Good question.  I think there was a reason, but I can't remember it
> right now.

AIUI, three combinations are valid:

is_physfn = 0, is_virtfn = 0: A non-SR-IOV function
is_physfn = 1, is_virtfn = 0: An SR-IOV PF
is_physfn = 0, is_virtfn = 1: An SR-IOV VF

As implemented with bit fields this is 2 bits, which is more space
efficient than an enum.  Thanks,

Alex