Re: [PATCH 2/2] remoteproc: imx_dsp_rproc: use common rproc_elf_load_segments

[Daniel Baluta]

On Wed, Apr 6, 2022 at 2:25 PM Daniel Baluta <daniel.baluta@xxxxxxxxx> wrote:
>
> On Wed, Apr 6, 2022 at 1:58 PM Peng Fan <peng.fan@xxxxxxx> wrote:
> >
> > > Subject: Re: [PATCH 2/2] remoteproc: imx_dsp_rproc: use common
> > > rproc_elf_load_segments
> > >
> > > On Thu, Mar 24, 2022 at 1:34 AM Peng Fan (OSS) <peng.fan@xxxxxxxxxxx>
> > > wrote:
> > > >
> > > > From: Peng Fan <peng.fan@xxxxxxx>
> > > >
> > > > remoteproc elf loader supports the specific case that segments have
> > > > PT_LOAD and memsz/filesz set to zero, so no duplicate code.
> > > >
> > > > Signed-off-by: Peng Fan <peng.fan@xxxxxxx>

Acked-by: Daniel Baluta <daniel.baluta@xxxxxxx>

Peng,

I'm fine going on with this now. Next we need to replace the boolean is_iomem
with a flags parameter to hold ATT_IOMEM, ATT_IOMEM32, etc.

> > >
> > > I think this change OK, but we have a case with the DSP were reads/writes
> > > should be done in multiples of 32/64.
> > >
> > > We need a way to provide our own "memcpy" function to be used by
> > > rproc_elf_load_segments.
> >
> > I think when generating elf file, the sections needs to be 32/64bits aligned.
>
> Sure, that could be a fix. But some malicious user can crash the kernel
> by crafting an elf with unaligned sections.