[U-Boot] FW: [PATCH 18/30] riscv: invalidate the instruction cache before jumping to Linux

[Greentime Hu]

Rick Chen <rickchen36 at gmail.com> 於 2018年10月29日 週一 上午10:25寫道：
>
> Auer, Lukas <lukas.auer at aisec.fraunhofer.de> 於 2018年10月27日 週六 上午12:27寫道：
> >
> > Hi Rick,
> >
> > On Mon, 2018-10-22 at 09:39 +0800, Rick Chen wrote:
> > > > From: Lukas Auer [mailto:lukas.auer at aisec.fraunhofer.de]
> > > > Sent: Saturday, October 20, 2018 6:08 AM
> > > > To: u-boot at lists.denx.de
> > > > Cc: Bin Meng; Lukas Auer; Greentime Hu; Alexander Graf; Rick Jian-
> > > > Zhi Chen(陳建志)
> > > > Subject: [PATCH 18/30] riscv: invalidate the instruction cache
> > > > before jumping to Linux
> > > >
> > > > Signed-off-by: Lukas Auer <lukas.auer at aisec.fraunhofer.de>
> > > > ---
> > > >
> > > >  arch/riscv/lib/bootm.c | 1 +
> > > >  1 file changed, 1 insertion(+)
> > > >
> > > > diff --git a/arch/riscv/lib/bootm.c b/arch/riscv/lib/bootm.c index
> > > > a7a9fb921b..bc1d4b2864 100644
> > > > --- a/arch/riscv/lib/bootm.c
> > > > +++ b/arch/riscv/lib/bootm.c
> > > > @@ -38,6 +38,7 @@ int do_bootm_linux(int flag, int argc, char
> > > > *argv[], bootm_headers_t *images)
> > > >                 return 1;
> > > >
> > > >         kernel = (void (*)(ulong, void *))images->ep;
> > > > +       invalidate_icache_all();
> > >
> > > Hi Likas
> > >
> > > I wull use cleanup_before_linux() which is in cpu.c as below
> > >
> >
> > I would prefer to keep the invalidate_icache_all() in bootm.c since it
> > is important in the context of the function. I do agree that the data
> > and instruction caches should be disabled in cleanup_before_linux().
> >
>
> Hi Lukas
>
> It is ok to keep the invalidate_icache_all() in bootm.c
>
> Rick
>

Hi Rick,

Since cleanup_before_linux() will invalidate icache, I think it will
be better to do the whole cache operations there.
As we can see the document from ARM(doc/README.arm-caches) it also said that
"
Cleanup Before Linux:
- cleanup_before_linux() should flush the D-cache, invalidate I-cache, and
  disable MMU and caches.
"
This may reduce the redundant codes/cost.
What do you think?