On Thu, 8 Oct 2009 14:22:21 -0700 (PDT)
David Rientjes <rientjes@xxxxxxxxxx> wrote:
> Objects passed to NODEMASK_ALLOC() are relatively small in size and are
> backed by slab caches that are not of large order, traditionally never
> greater than PAGE_ALLOC_COSTLY_ORDER.
>
> Thus, using GFP_KERNEL for these allocations on large machines when
> CONFIG_NODES_SHIFT > 8 will cause the page allocator to loop endlessly in
> the allocation attempt, each time invoking both direct reclaim or the oom
> killer.
>
> This is of particular interest when using NODEMASK_ALLOC() from a
> mempolicy context (either directly in mm/mempolicy.c or the mempolicy
> constrained hugetlb allocations) since the oom killer always kills
> current when allocations are constrained by mempolicies. So for all
> present use cases in the kernel, current would end up being oom killed
> when direct reclaim fails. That would allow the NODEMASK_ALLOC() to
> succeed but current would have sacrificed itself upon returning.
>
> This patch adds gfp flags to NODEMASK_ALLOC() to pass to kmalloc() on
> CONFIG_NODES_SHIFT > 8; this parameter is a nop on other configurations.
> All current use cases either directly from hugetlb code or indirectly via
> NODEMASK_SCRATCH() union __GFP_NORETRY to avoid direct reclaim and the
> oom killer when the slab allocator needs to allocate additional pages.
>
> The side-effect of this change is that all current use cases of either
> NODEMASK_ALLOC() or NODEMASK_SCRATCH() need appropriate -ENOMEM handling
> when the allocation fails (never for CONFIG_NODES_SHIFT <= 8). All
> current use cases were audited and do have appropriate error handling at
> this time.
>
> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> ---
> Andrew, this was written on mmotm-09251435 plus Lee's entire patchset.
>
> include/linux/nodemask.h | 21 ++++++++++++---------
> mm/hugetlb.c | 5 +++--
> 2 files changed, 15 insertions(+), 11 deletions(-)
>
> diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h
> --- a/include/linux/nodemask.h
> +++ b/include/linux/nodemask.h
> @@ -485,15 +485,17 @@ static inline int num_node_state(enum node_states state)
> #define for_each_online_node(node) for_each_node_state(node, N_ONLINE)
>
> /*
> - * For nodemask scrach area.(See CPUMASK_ALLOC() in cpumask.h)
> - * NODEMASK_ALLOC(x, m) allocates an object of type 'x' with the name 'm'.
> + * For nodemask scrach area.
> + * NODEMASK_ALLOC(type, name) allocates an object with a specified type and
> + * name.
> */
> -#if NODES_SHIFT > 8 /* nodemask_t > 64 bytes */
> -#define NODEMASK_ALLOC(x, m) x *m = kmalloc(sizeof(*m), GFP_KERNEL)
> -#define NODEMASK_FREE(m) kfree(m)
> +#if NODES_SHIFT > 8 /* nodemask_t > 256 bytes */
> +#define NODEMASK_ALLOC(type, name, gfp_flags) \
> + type *name = kmalloc(sizeof(*name), gfp_flags)
> +#define NODEMASK_FREE(m) kfree(m)
> #else
> -#define NODEMASK_ALLOC(x, m) x _m, *m = &_m
> -#define NODEMASK_FREE(m) do {} while (0)
> +#define NODEMASK_ALLOC(type, name, gfp_flags) type _name, *name = &_name
> +#define NODEMASK_FREE(m) do {} while (0)
> #endif
>
> /* A example struture for using NODEMASK_ALLOC, used in mempolicy. */
> @@ -502,8 +504,9 @@ struct nodemask_scratch {
> nodemask_t mask2;
> };
>
> -#define NODEMASK_SCRATCH(x) \
> - NODEMASK_ALLOC(struct nodemask_scratch, x)
> +#define NODEMASK_SCRATCH(x) \
> + NODEMASK_ALLOC(struct nodemask_scratch, x, \
> + GFP_KERNEL | __GFP_NORETRY)
> #define NODEMASK_SCRATCH_FREE(x) NODEMASK_FREE(x)
>
>
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -1361,7 +1361,7 @@ static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
> int nid;
> unsigned long count;
> struct hstate *h;
> - NODEMASK_ALLOC(nodemask_t, nodes_allowed);
> + NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
>
> err = strict_strtoul(buf, 10, &count);
> if (err)
> @@ -1857,7 +1857,8 @@ static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
> proc_doulongvec_minmax(table, write, buffer, length, ppos);
>
> if (write) {
> - NODEMASK_ALLOC(nodemask_t, nodes_allowed);
> + NODEMASK_ALLOC(nodemask_t, nodes_allowed,
> + GFP_KERNEL | __GFP_NORETRY);
> if (!(obey_mempolicy &&
> init_nodemask_of_mempolicy(nodes_allowed))) {
> NODEMASK_FREE(nodes_allowed);
>
--
To unsubscribe from this list: send the line "unsubscribe linux-numa" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
