Re: [PATCH v5 5/8] remoteproc: Rename rproc_elf_sanity_check for elf32

[Bjorn Andersson]

On Tue 10 Mar 08:38 PDT 2020, Cl?ment Leger wrote:

> Hi Mathieu,
> 
> ----- On 10 Mar, 2020, at 16:20, Mathieu Poirier mathieu.poirier@xxxxxxxxxx wrote:
> 
> > On Mon, Mar 09, 2020 at 05:00:05PM -0700, Bjorn Andersson wrote:
> >> On Tue 03 Mar 00:02 PST 2020, Cl?ment Leger wrote:
> >> 
> >> > Hi Bjorn,
> >> > 
> >> > ----- On 3 Mar, 2020, at 00:13, Bjorn Andersson bjorn.andersson@xxxxxxxxxx
> >> > wrote:
> >> > 
> >> > > On Mon 02 Mar 01:38 PST 2020, Clement Leger wrote:
> >> > > 
> >> > >> Since this function will be modified to support both elf32 and elf64,
> >> > >> rename the existing one to elf32 (which is the only supported format
> >> > >> at the moment). This will allow not to introduce possible side effect
> >> > >> when adding elf64 support (ie: all backends will still support only
> >> > >> elf32 if not requested explicitely using rproc_elf_sanity_check).
> >> > >> 
> >> > > 
> >> > > Is there a reason for preventing ELF64 binaries be loaded?
> >> > 
> >> > I decided to go this way to let driver maintainer decide if they want
> >> > to support elf64 to avoid problems with 64bits addresses/sizes which do
> >> > not fit in their native type (size_t for instance). This is probably
> >> > not going to happen and there are additionnal checks before calling
> >> > rproc_da_to_va. And addresses should be filtered by rproc_da_to_va.
> >> > So, actually it seems there is no reason to forbid supporting elf32/64
> >> > for all drivers.
> >> > 
> >> 
> >> I was hoping to hear some additional feedback on this from others.
> > 
> > I didn't follow up on this one because I agreed with your assesment and didn't
> > think it was needed.
> > 
> > Simply put I would rather see rproc_elf_sanity_check() gain support for elf64
> > and let the platform code decide what to do with format they don't support
> > rather than spinning a new function.
> > 
> >> 
> >> I've merge the patch as is, but think it would be nice to clean this up
> >> and just have the driver ignore if fed a 32 or 64-elf.
> > 
> > It would be really nice to see this cleaned up in time for the coming merge
> > window...
> 
> I could have sent a V7, but Bjorn was faster than my comment ;)

I figured it had been maturing on the list long enough and expected the
cleanup to be a nice incremental patch.

> Bjorn, Is there any way to revert that or it's already pushed ?
> I already have a clean V7.
> 

Please base your changes on what's in rproc-next (and today's
linux-next).

Thank you,
Bjorn