Hello Vlastimil and Sunghyun,
On Thu, Sep 03, 2020 at 03:46:33PM +0200, Vlastimil Babka wrote:
> On 9/3/20 2:41 PM, mcsmonk wrote:
> > From: Sunghyun Jin <mcsmonk@xxxxxxxxx>
> >
> > Variable populated, which is a member of struct pcpu_chunk, is used as a
> > unit of size of unsigned long.
> > However, size of populated is miscounted. So, I fix this minor part.
>
> +CC folks who touched it last
>
Thanks for CCing me.
> Nice find! Did you observe e.g. a panic that can be used in the commit log? Or
> were we always lucky thanks to alignment?
Well that is indeed awkward. Luckily the first chunk is a bit special
and only holds 7 dynamic pages. Additionally, the allocation rounds to
SMP_CACHE_BYTES so that would give us 8 bytes to play with as
struct pcpu_chunk is 120 bytes. So, while technically (wrong) the
1 byte was sufficient and the additional buffer is why at least I never
got a panic report to today.
> Is there perhaps a commit that introduced the bug and we can use it as Fixes:?
> My brief look suggests 8ab16c43ea79 ("percpu: change the number of pages marked
> in the first_chunk pop bitmap")
>
That is the right commit. I'll pick this up and add the fixes line.
> Thanks!
>
Thanks,
Dennis
> > Signed-off-by: Sunghyun Jin <mcsmonk@xxxxxxxxx>
> > ---
> > mm/percpu.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/mm/percpu.c b/mm/percpu.c
> > index f4709629e6de..1ed1a349eab8 100644
> > --- a/mm/percpu.c
> > +++ b/mm/percpu.c
> > @@ -1316,7 +1316,7 @@ static struct pcpu_chunk * __init pcpu_alloc_first_chunk(unsigned long tmp_addr,
> >
> > /* allocate chunk */
> > alloc_size = sizeof(struct pcpu_chunk) +
> > - BITS_TO_LONGS(region_size >> PAGE_SHIFT);
> > + BITS_TO_LONGS(region_size >> PAGE_SHIFT) * sizeof(unsigned long);
> > chunk = memblock_alloc(alloc_size, SMP_CACHE_BYTES);
> > if (!chunk)
> > panic("%s: Failed to allocate %zu bytes\n", __func__,
> >
>
