[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230710150532.hvj7jzc5ep73yq4i@revolver>
Date: Mon, 10 Jul 2023 11:05:32 -0400
From: "Liam R. Howlett" <Liam.Howlett@...cle.com>
To: Peng Zhang <zhangpeng.00@...edance.com>, akpm@...ux-foundation.org,
linux-mm@...ck.org, linux-kernel@...r.kernel.org,
maple-tree@...ts.infradead.org
Subject: Re: [RESEND PATCH 1/8] maple_tree: set the node limit when creating
a new root node
* Liam R. Howlett <Liam.Howlett@...cle.com> [230707 11:18]:
> * Peng Zhang <zhangpeng.00@...edance.com> [230707 06:11]:
> > Set the node limit of the root node so that the last pivot of all nodes
> > is the node limit (if the node is not full).
This patch also fixes a bug in mas_rev_awalk(). Effectively, always
setting a maximum makes mas_logical_pivot() behave as mas_safe_pivot().
Without this fix, it is possible that very small tasks would fail to
find the correct gap. Although this has not been observed with real
tasks, it has been reported to happen in m68k nommu running the maple
tree tests.
Link: https://lore.kernel.org/linux-mm/CAMuHMdV4T53fOw7VPoBgPR7fP6RYqf=CBhD_y_vOg53zZX_DnA@mail.gmail.com/
Cc: <stable@...r.kernel.org>
Cc: Geert Uytterhoeven <geert@...ux-m68k.org>
> >
> > Signed-off-by: Peng Zhang <zhangpeng.00@...edance.com>
>
> This has been on my list of things to do for a while, thanks.
>
> Reviewed-by: Liam R. Howlett <Liam.Howlett@...cle.com>
>
> > ---
> > lib/maple_tree.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/lib/maple_tree.c b/lib/maple_tree.c
> > index d3072858c280..f55e59bd9122 100644
> > --- a/lib/maple_tree.c
> > +++ b/lib/maple_tree.c
> > @@ -3692,7 +3692,8 @@ static inline int mas_root_expand(struct ma_state *mas, void *entry)
> > mas->offset = slot;
> > pivots[slot] = mas->last;
> > if (mas->last != ULONG_MAX)
> > - slot++;
> > + pivots[++slot] = ULONG_MAX;
> > +
> > mas->depth = 1;
> > mas_set_height(mas);
> > ma_set_meta(node, maple_leaf_64, 0, slot);
> > --
> > 2.20.1
> >
> >
Powered by blists - more mailing lists