On Tue, 17 Jan 2017 21:58:31 +0100
Alexander Graf <ag...@suse.de> wrote:
>
>
> On 17/01/2017 21:45, Emmanuel Vadot wrote:
> > On Tue, 17 Jan 2017 21:13:51 +0100
> > Alexander Graf <ag...@suse.de> wrote:
> >
> >>
> >>
> >> On 17/01/2017 16:50, Emmanuel Vadot wrote:
> >>> From: Warner Losh <i...@freebsd.org>
> >>>
> >>> FreeBSD loader(8) just loaded code to some random location that may
> >>> contain stale icache entries.
> >>> FreeBSD Kernel needs the icache and dcache flushed.
> >>> Before running either one of them, flush the icache and dcache.
> >>>
> >>> Signed-off-by: Emmanuel Vadot <m...@bidouilliste.com>
> >>
> >> Is there any downside to doing the flush always? I would prefer to not
> >> conditionalize it on CONFIG_FREEBSD.
> >
> > I don't think there is any downside.
> >
> >>> ---
> >>> api/api.c | 5 +++++
> >>> cmd/boot.c | 5 +++++
> >>> cmd/elf.c | 8 +++++++-
> >>> 3 files changed, 17 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/api/api.c b/api/api.c
> >>> index 8a1433af78..cc25deca09 100644
> >>> --- a/api/api.c
> >>> +++ b/api/api.c
> >>> @@ -290,6 +290,11 @@ static int API_dev_close(va_list ap)
> >>> if (!err)
> >>> di->state = DEV_STA_CLOSED;
> >>>
> >>> +#ifdef CONFIG_FREEBSD
> >>> + flush_dcache_all();
> >>> + invalidate_icache_all();
> >>> +#endif
> >>> +
> >>> return err;
> >>> }
> >>>
> >>> diff --git a/cmd/boot.c b/cmd/boot.c
> >>> index 72f2cf362d..a1a91fbf0a 100644
> >>> --- a/cmd/boot.c
> >>> +++ b/cmd/boot.c
> >>> @@ -19,6 +19,11 @@ __attribute__((weak))
> >>> unsigned long do_go_exec(ulong (*entry)(int, char * const []), int argc,
> >>> char * const argv[])
> >>> {
> >>> +#ifdef CONFIG_FREEBSD
> >>> + flush_dcache_all();
> >>> + invalidate_icache_all();
> >>> +#endif
> >>> +
> >>> return entry (argc, argv);
> >>> }
> >>>
> >>> diff --git a/cmd/elf.c b/cmd/elf.c
> >>> index 5190cc6c0f..b2827fa042 100644
> >>> --- a/cmd/elf.c
> >>> +++ b/cmd/elf.c
> >>> @@ -109,6 +109,7 @@ static unsigned long do_bootelf_exec(ulong
> >>> (*entry)(int, char * const[]),
> >>> {
> >>> unsigned long ret;
> >>>
> >>> +#ifndef CONFIG_FREEBSD
> >>> /*
> >>> * QNX images require the data cache is disabled.
> >>> * Data cache is already flushed, so just turn it off.
> >>> @@ -116,15 +117,20 @@ static unsigned long do_bootelf_exec(ulong
> >>> (*entry)(int, char * const[]),
> >>> int dcache = dcache_status();
> >>> if (dcache)
> >>> dcache_disable();
> >>
> >> Can FreeBSD boot with dcache disabled? Is there some way we can
> >> determine what payload we have from the elf header?
> >>
> >>
> >> Alex
> >
> > FreeBSD can boot with dcache disabled (our loader and kernel can)
> > I'm not sure I get your other question, sorry.
>
> Your patch #ifdefs out the dcache_disable() if CONFIG_FREEBSD is set. So
> if you can boot just fine with dcache disabled, you can remove this hunk
> altogether FWIW.
Loading the kernel from our loader is much faster with dcache enabled
(1 sec instead of 5/7 sec).
I guess we can always flush the cache and only disable it for QNX,
I'll check if there is a CONFIG_QNX or something like that.
> All the other changes in this file are mere cache flushes - and they
> shouldn't hurt any u-boot api user and thus could be done unconditionally.
True, I can remove the #ifdefs on CONFIG_FREEBSD here.
>
> Alex
--
Emmanuel Vadot <m...@bidouilliste.com> <m...@freebsd.org>
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
