On Wed, 9 Oct 2019 18:49:34 +0200
"Uladzislau Rezki (Sony)" <urezki@xxxxxxxxx> wrote:
> Get rid of preempt_disable() and preempt_enable() when the
> preload is done for splitting purpose. The reason is that
> calling spin_lock() with disabled preemtion is forbidden in
> CONFIG_PREEMPT_RT kernel.
>
> Therefore, we do not guarantee that a CPU is preloaded, instead
> we minimize the case when it is not with this change.
>
> For example i run the special test case that follows the preload
> pattern and path. 20 "unbind" threads run it and each does
> 1000000 allocations. Only 3.5 times among 1000000 a CPU was
> not preloaded thus. So it can happen but the number is rather
> negligible.
Thanks for the analysis.
>
> Fixes: 82dd23e84be3 ("mm/vmalloc.c: preload a CPU with one object for split purpose")
> Signed-off-by: Uladzislau Rezki (Sony) <urezki@xxxxxxxxx>
> ---
> mm/vmalloc.c | 17 ++++++++---------
> 1 file changed, 8 insertions(+), 9 deletions(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index e92ff5f7dd8b..2ed6fef86950 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -1078,9 +1078,12 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
>
> retry:
> /*
> - * Preload this CPU with one extra vmap_area object to ensure
> - * that we have it available when fit type of free area is
> - * NE_FIT_TYPE.
> + * Preload this CPU with one extra vmap_area object. It is used
> + * when fit type of free area is NE_FIT_TYPE. Please note, it
> + * does not guarantee that an allocation occurs on a CPU that
> + * is preloaded, instead we minimize the case when it is not.
> + * It can happen because of migration, because there is a race
> + * until the below spinlock is taken.
> *
> * The preload is done in non-atomic context, thus it allows us
> * to use more permissive allocation masks to be more stable under
> @@ -1089,20 +1092,16 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
> * Even if it fails we do not really care about that. Just proceed
> * as it is. "overflow" path will refill the cache we allocate from.
> */
> - preempt_disable();
> - if (!__this_cpu_read(ne_fit_preload_node)) {
> - preempt_enable();
As the original code enables preemption here regardless, there's no
guarantee that the original patch would allocate the pva to the CPU in
question.
I agree with this patch, the preempt_disable() here only narrows an
already narrow window, with no real help in what it was doing.
> + if (!this_cpu_read(ne_fit_preload_node)) {
> pva = kmem_cache_alloc_node(vmap_area_cachep, GFP_KERNEL, node);
If the memory allocation failed here, we still may not have a pva for
the current CPU's ne_fit_preload_node, rare as that may be.
> - preempt_disable();
>
> - if (__this_cpu_cmpxchg(ne_fit_preload_node, NULL, pva)) {
> + if (this_cpu_cmpxchg(ne_fit_preload_node, NULL, pva)) {
Reviewed-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
-- Steve
> if (pva)
> kmem_cache_free(vmap_area_cachep, pva);
> }
> }
>
> spin_lock(&vmap_area_lock);
> - preempt_enable();
>
> /*
> * If an allocation fails, the "vend" address is
