Re: [PATCH] firmware: make sure builtin firmware is page alignment

From: gregkh@xxxxxxxxxxxxxxxxxxx
Date: Mon Aug 06 2018 - 10:05:49 EST


On Mon, Aug 06, 2018 at 01:48:44AM +0000, Zhang, Ning A wrote:
> å 2018-08-03äç 12:31 +0200ïgregkh@xxxxxxxxxxxxxxxxxxxåéï
> > On Fri, Aug 03, 2018 at 08:42:25AM +0000, Zhang, Ning A wrote:
> > > å 2018-08-03äç 07:39 +0200ïGreg KHåéï
> > > > On Fri, Aug 03, 2018 at 09:45:21AM +0800, Zhang Ning wrote:
> > > > > when firmware is in filesystem, request_firmware will load it,
> > > > > and copy it to vmalloc memory, that is page align memory.
> > > > >
> > > > > but when firmware is builtin, it is 8 bytes or 4 bytes
> > > > > alignment.
> > > > >
> > > > > make sure builtin firmware is page algnment, that can simplify
> > > > > algorithm
> > > > > to handle firmware.
> > > >
> > > > How is it simplified?ÂÂI don't see any such change like that here
> > > > :(
> > > >
> > >
> > > Thank you for review this patch.
> > >
> > > When driver handles its firmware based onÂÂpage, like below:
> > >
> > > struct page *p;
> > > p = vmalloc_to_page(fw->data);ÂÂ// for filesystem firmware
> > > p = virt_to_page(fw->data); // for builtin firmware
> > >
> > > but if builtin firmware is not page alignment, page pointer for
> > > builtin
> > > firmware is wrong, it contains memory not belong to firmware.
> > > drivers
> > > has to use additional code to handle this.Â
> > >
> > > if builtin firmware is also page alignment, no need additional code
> > > to
> > > handle builtin firmware. simplified.
> >
> > But you did not change anything like this in your code, so why would
> > I
> > know this?
>
> I understand it is very difficult to review this patch without context.
> The driver is not opensource, I can't show the patch for driver.

Then I can not accept your patch. Go talk to your corporate lawyers
about changing core kernel code for a closed source driver and what that
implies about that closed driver (hint, your driver can not be
closed...) :)

Then come back with a proper open sourced driver, that's the only way
Linux drivers can be written sorry.

best of luck,

greg k-h