On Tue, 2018-10-16 at 13:19 +1100, Stephen Rothwell wrote:
> Hi all,
>
> On Tue, 16 Oct 2018 13:02:16 +1100 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> >
> > Reverting fe3d2a45e8079fdd7d4da1ff07f4b40bc3cb499f (and the following 2
> > commits) produces a kernel that boots.
>
> Instead of that, I applied this patch on top of linux-next and it boots
> and produces a stack trace ...
>
> From: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> Date: Tue, 16 Oct 2018 13:07:01 +1100
> Subject: [PATCH] mm/memblock.c: use dump_stack() instead of WARN_ON_ONCE for
> the alignment checks
>
> Using WARN_ON_ONCE too early causes the PowerPC kernel to fail.
Interesting ... I thought I had fixed that. Might need to be re-fixed.
> Signed-off-by: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> ---
> mm/memblock.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 5fefc70253ee..f2ef3915a356 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -1298,8 +1298,10 @@ static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
> {
> phys_addr_t found;
>
> - if (WARN_ON_ONCE(!align))
> + if (!align) {
> + dump_stack();
> align = SMP_CACHE_BYTES;
> + }
>
> found = memblock_find_in_range_node(size, align, start, end, nid,
> flags);
> @@ -1423,8 +1425,10 @@ static void * __init memblock_alloc_internal(
> if (WARN_ON_ONCE(slab_is_available()))
> return kzalloc_node(size, GFP_NOWAIT, nid);
>
> - if (WARN_ON_ONCE(!align))
> + if (!align) {
> + dump_stack();
> align = SMP_CACHE_BYTES;
> + }
>
> if (max_addr > memblock.current_limit)
> max_addr = memblock.current_limit;
> --
> 2.18.0
>
> So, patch "memblock: stop using implicit alignment to SMP_CACHE_BYTES"
> should *not* remove the 0 -> SMP_CACHE_BYTES update from mm/memblock.c
> and just add the dump_stack().
