Re: [PATCH 5/5] drivers/pci: make most of the PCI_DW drivers modular

[Paul Gortmaker]

[Re: [PATCH 5/5] drivers/pci: make most of the PCI_DW drivers modular] On 08/02/2016 (Mon 11:00) Arnd Bergmann wrote:

> On Sunday 07 February 2016 19:00:44 Paul Gortmaker wrote:
> > We've exported the symbols that we know these specific drivers
> > will need as tristate, so now we can make the conversion from
> > bool to tristate w/o concern for build regressions.
> > 
> > Cc: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
> > Cc: Ley Foon Tan <lftan@xxxxxxxxxx>
> > Cc: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> > Cc: Stanimir Varbanov <svarbanov@xxxxxxxxxx>
> > Cc: Thierry Reding <thierry.reding@xxxxxxxxx>
> > Cc: Arnd Bergmann <arnd@xxxxxxxx>
> > Cc: linux-pci@xxxxxxxxxxxxxxx
> > Cc: linux-arm-msm@xxxxxxxxxxxxxxx
> > Signed-off-by: Paul Gortmaker <paul.gortmaker@xxxxxxxxxxxxx>
> 
> Looks good, but I'm not sure what happens on unload. Did you check?

Per the 0/N I've not got the capability to do run time testing for any
of this stuff.  And from previous discussions, ISTR that people didn't
want to see a blanket block-unload policy implemented.

> Until this is tested, maybe it's best to prevent module unloading?

If that is what people want, I can add that.

P.
-- 

> 
> 	Arnd