[dpdk-dev] [PATCH] malloc: respect SIZE_HINT_ONLY when looking for the biggest free elem

[dpdk-dev] [PATCH] malloc: respect SIZE_HINT_ONLY when looking for the biggest free elem

[Darek Stojaczyk]

RTE_MEMZONE_SIZE_HINT_ONLY wasn't checked in any way,
causing size hints to be parsed as hard requirements.
This resulted in some allocations being failed prematurely.

Fixes: 68b6092bd3c7 ("malloc: allow reserving biggest element")
Cc: anatoly.burakov@intel.com
Cc: stable@dpdk.org

Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
---
 lib/librte_eal/common/malloc_heap.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/lib/librte_eal/common/malloc_heap.c b/lib/librte_eal/common/malloc_heap.c
index ac7bbb3ba..d2a8bd8dc 100644
--- a/lib/librte_eal/common/malloc_heap.c
+++ b/lib/librte_eal/common/malloc_heap.c
@@ -165,7 +165,9 @@ find_biggest_element(struct malloc_heap *heap, size_t *size,
 		for (elem = LIST_FIRST(&heap->free_head[idx]);
 				!!elem; elem = LIST_NEXT(elem, free_list)) {
 			size_t cur_size;
-			if (!check_hugepage_sz(flags, elem->msl->page_sz))
+			if ((flags & RTE_MEMZONE_SIZE_HINT_ONLY) == 0 &&
+					!check_hugepage_sz(flags,
+						elem->msl->page_sz))
 				continue;
 			if (contig) {
 				cur_size =
-- 
2.17.1

Re: [dpdk-dev] [PATCH] malloc: respect SIZE_HINT_ONLY when looking for the biggest free elem

[Burakov, Anatoly]

On 07-Oct-18 8:31 PM, Darek Stojaczyk wrote:
> RTE_MEMZONE_SIZE_HINT_ONLY wasn't checked in any way,
> causing size hints to be parsed as hard requirements.
> This resulted in some allocations being failed prematurely.
> 
> Fixes: 68b6092bd3c7 ("malloc: allow reserving biggest element")
> Cc: anatoly.burakov@intel.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
> ---
>   lib/librte_eal/common/malloc_heap.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/librte_eal/common/malloc_heap.c b/lib/librte_eal/common/malloc_heap.c
> index ac7bbb3ba..d2a8bd8dc 100644
> --- a/lib/librte_eal/common/malloc_heap.c
> +++ b/lib/librte_eal/common/malloc_heap.c
> @@ -165,7 +165,9 @@ find_biggest_element(struct malloc_heap *heap, size_t *size,
>   		for (elem = LIST_FIRST(&heap->free_head[idx]);
>   				!!elem; elem = LIST_NEXT(elem, free_list)) {
>   			size_t cur_size;
> -			if (!check_hugepage_sz(flags, elem->msl->page_sz))
> +			if ((flags & RTE_MEMZONE_SIZE_HINT_ONLY) == 0 &&
> +					!check_hugepage_sz(flags,
> +						elem->msl->page_sz))
>   				continue;

Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>

Although to be frank, the whole concept of "reserving biggest available 
memzone" is currently broken because of dynamic memory allocation. There 
is currently no way to allocate "as many hugepages as you can", so we're 
only looking at memory already allocated, which in the general case is 
less than page size long (unless you use legacy mode or memory 
preallocation switches).

-- 
Thanks,
Anatoly

Re: [dpdk-dev] [PATCH] malloc: respect SIZE_HINT_ONLY when looking for the biggest free elem

[Thomas Monjalon]

08/10/2018 11:02, Burakov, Anatoly:
> On 07-Oct-18 8:31 PM, Darek Stojaczyk wrote:
> > RTE_MEMZONE_SIZE_HINT_ONLY wasn't checked in any way,
> > causing size hints to be parsed as hard requirements.
> > This resulted in some allocations being failed prematurely.
> > 
> > Fixes: 68b6092bd3c7 ("malloc: allow reserving biggest element")
> > Cc: anatoly.burakov@intel.com
> > Cc: stable@dpdk.org
> > 
> > Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
> > ---
> >   lib/librte_eal/common/malloc_heap.c | 4 +++-
> >   1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/lib/librte_eal/common/malloc_heap.c b/lib/librte_eal/common/malloc_heap.c
> > index ac7bbb3ba..d2a8bd8dc 100644
> > --- a/lib/librte_eal/common/malloc_heap.c
> > +++ b/lib/librte_eal/common/malloc_heap.c
> > @@ -165,7 +165,9 @@ find_biggest_element(struct malloc_heap *heap, size_t *size,
> >   		for (elem = LIST_FIRST(&heap->free_head[idx]);
> >   				!!elem; elem = LIST_NEXT(elem, free_list)) {
> >   			size_t cur_size;
> > -			if (!check_hugepage_sz(flags, elem->msl->page_sz))
> > +			if ((flags & RTE_MEMZONE_SIZE_HINT_ONLY) == 0 &&
> > +					!check_hugepage_sz(flags,
> > +						elem->msl->page_sz))
> >   				continue;
> 
> Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
> 
> Although to be frank, the whole concept of "reserving biggest available 
> memzone" is currently broken because of dynamic memory allocation. There 
> is currently no way to allocate "as many hugepages as you can", so we're 
> only looking at memory already allocated, which in the general case is 
> less than page size long (unless you use legacy mode or memory 
> preallocation switches).

Applied anyway, thanks