linux-kernel - Re: [PATCH 6/9] mm, page_alloc: simplify zonelist initialization

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170714124645.i3duhuie6cczlybr@suse.de>
Date:   Fri, 14 Jul 2017 13:46:46 +0100
From:   Mel Gorman <mgorman@...e.de>
To:     Michal Hocko <mhocko@...nel.org>
Cc:     linux-mm@...ck.org, Andrew Morton <akpm@...ux-foundation.org>,
        Johannes Weiner <hannes@...xchg.org>,
        Vlastimil Babka <vbabka@...e.cz>,
        LKML <linux-kernel@...r.kernel.org>,
        Michal Hocko <mhocko@...e.com>
Subject: Re: [PATCH 6/9] mm, page_alloc: simplify zonelist initialization

On Fri, Jul 14, 2017 at 10:00:03AM +0200, Michal Hocko wrote:
> From: Michal Hocko <mhocko@...e.com>
> 
> build_zonelists gradually builds zonelists from the nearest to the most
> distant node. As we do not know how many populated zones we will have in
> each node we rely on the _zoneref to terminate initialized part of the
> zonelist by a NULL zone. While this is functionally correct it is quite
> suboptimal because we cannot allow updaters to race with zonelists
> users because they could see an empty zonelist and fail the allocation
> or hit the OOM killer in the worst case.
> 
> We can do much better, though. We can store the node ordering into an
> already existing node_order array and then give this array to
> build_zonelists_in_node_order and do the whole initialization at once.
> zonelists consumers still might see halfway initialized state but that
> should be much more tolerateable because the list will not be empty and
> they would either see some zone twice or skip over some zone(s) in the
> worst case which shouldn't lead to immediate failures.
> 
> This patch alone doesn't introduce any functional change yet, though, it
> is merely a preparatory work for later changes.
> 
> Signed-off-by: Michal Hocko <mhocko@...e.com>
> ---
>  mm/page_alloc.c | 42 ++++++++++++++++++------------------------
>  1 file changed, 18 insertions(+), 24 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 00e117922b3f..78bd62418380 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -4913,17 +4913,20 @@ static int find_next_best_node(int node, nodemask_t *used_node_mask)
>   * This results in maximum locality--normal zone overflows into local
>   * DMA zone, if any--but risks exhausting DMA zone.
>   */
> -static void build_zonelists_in_node_order(pg_data_t *pgdat, int node)
> +static void build_zonelists_in_node_order(pg_data_t *pgdat, int *node_order)
>  {
> -	int j;
>  	struct zonelist *zonelist;
> +	int i, zoneref_idx = 0;
>  
>  	zonelist = &pgdat->node_zonelists[ZONELIST_FALLBACK];
> -	for (j = 0; zonelist->_zonerefs[j].zone != NULL; j++)
> -		;
> -	j = build_zonelists_node(NODE_DATA(node), zonelist, j);
> -	zonelist->_zonerefs[j].zone = NULL;
> -	zonelist->_zonerefs[j].zone_idx = 0;
> +
> +	for (i = 0; i < MAX_NUMNODES; i++) {
> +		pg_data_t *node = NODE_DATA(node_order[i]);
> +
> +		zoneref_idx = build_zonelists_node(node, zonelist, zoneref_idx);
> +	}

The naming here is weird to say the least and makes this a lot more
confusing than it needs to be. Primarily, it's because the zoneref_idx
parameter gets renamed to nr_zones in build_zonelists_node where it's
nothing to do with the number of zones at all.

It also iterates for longer than it needs to. MAX_NUMNODES can be a
large value of mostly empty nodes but it happily goes through them
anyway. Pass zoneref_idx in as a pointer that is updated by the function
and use the return value to break the loop when an empty node is
encountered?

> +	zonelist->_zonerefs[zoneref_idx].zone = NULL;
> +	zonelist->_zonerefs[zoneref_idx].zone_idx = 0;
>  }
>  

It *might* be safer given the next patch to zero out the remainder of
the _zonerefs to that there is no combination of node add/remove that has
an iterator working with a semi-valid _zoneref which is beyond the last
correct value. It *should* be safe as the very last entry will always
be null but if you don't zero it out, it is possible for iterators to be
working beyond the "end" of the zonelist for a short window.

Otherwise think it's ok including my stupid comment about node_order
stack usage.

-- 
Mel Gorman
SUSE Labs