[dpdk-dev] eal/malloc: fix malloc cookie check

  From: xuemingl <xuemingl@mellanox.com>

DPDK uses it's own memory management, few regular memory profiler tool
support DPDK now. Malloc cookie check provides limited memory corruption
check, better than nothing. This patch fixes the following:
1. Replace broken generated configuration macro RTE_LIBRTE_MALLOC_DEBUG
with RTE_MALLOC_DEBUG
2. Fix malloc size calculation when RTE_MALLOC_DEBUG cookie check
enabled.

From real test, it IS very helpful to detect memory corruption.

A better designed DPDK application should quit nicely with resoure
releasing so that all allocated memory could be checked against cookie.

Fixes: af75078 ("first public release")
Cc: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>

Signed-off-by: Xueming Li <xuemingl@mellanox.com>
---
 lib/librte_eal/common/malloc_elem.c |  8 ++++----
 lib/librte_eal/common/malloc_elem.h |  4 ++--
 test/test/test_malloc.c             | 10 +++++-----
 3 files changed, 11 insertions(+), 11 deletions(-)
  

Hi Xueming,

I think I have understood what you are doing and why the change is 
required, but
I believe we can improve the commit message.

It might be even a good idea to split the patches:
1. RTE_MALLOC_DEBUG fix
2. malloc_elem_free fix


On 23/08/2017 03:29, Xueming Li wrote:
> From: xuemingl <xuemingl@mellanox.com>

'From: ..' not needed.

> DPDK uses it's own memory management, few regular memory profiler tool
> support DPDK now. Malloc cookie check provides limited memory corruption
> check, better than nothing. This patch fixes the following:
> 1. Replace broken generated configuration macro RTE_LIBRTE_MALLOC_DEBUG
> with RTE_MALLOC_DEBUG
> 2. Fix malloc size calculation when RTE_MALLOC_DEBUG cookie check
> enabled.

For the 2nd change, you should mention what is the problem.
The issue is that malloc_elem_free is clearing (setting to 0) the 
trailer cookie when RTE_MALLOC_DEBUG is enabled.

You could also add that you would trigger this issue if the elem being 
freed does not join the next elem.

Thanks,
Sergio

>  From real test, it IS very helpful to detect memory corruption.
>
> A better designed DPDK application should quit nicely with resoure
> releasing so that all allocated memory could be checked against cookie.
>
> Fixes: af75078 ("first public release")
> Cc: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
>
> Signed-off-by: Xueming Li <xuemingl@mellanox.com>
> ---
>   lib/librte_eal/common/malloc_elem.c |  8 ++++----
>   lib/librte_eal/common/malloc_elem.h |  4 ++--
>   test/test/test_malloc.c             | 10 +++++-----
>   3 files changed, 11 insertions(+), 11 deletions(-)
>
> diff --git a/lib/librte_eal/common/malloc_elem.c b/lib/librte_eal/common/malloc_elem.c
> index 150769057..889dffd21 100644
> --- a/lib/librte_eal/common/malloc_elem.c
> +++ b/lib/librte_eal/common/malloc_elem.c
> @@ -275,14 +275,14 @@ malloc_elem_free(struct malloc_elem *elem)
>   		return -1;
>   
>   	rte_spinlock_lock(&(elem->heap->lock));
> -	size_t sz = elem->size - sizeof(*elem);
> +	size_t sz = elem->size - sizeof(*elem) - MALLOC_ELEM_TRAILER_LEN;
>   	uint8_t *ptr = (uint8_t *)&elem[1];
>   	struct malloc_elem *next = RTE_PTR_ADD(elem, elem->size);
>   	if (next->state == ELEM_FREE){
>   		/* remove from free list, join to this one */
>   		elem_free_list_remove(next);
>   		join_elem(elem, next);
> -		sz += sizeof(*elem);
> +		sz += (sizeof(*elem) + MALLOC_ELEM_TRAILER_LEN);
>   	}
>   
>   	/* check if previous element is free, if so join with it and return,
> @@ -291,8 +291,8 @@ malloc_elem_free(struct malloc_elem *elem)
>   	if (elem->prev != NULL && elem->prev->state == ELEM_FREE) {
>   		elem_free_list_remove(elem->prev);
>   		join_elem(elem->prev, elem);
> -		sz += sizeof(*elem);
> -		ptr -= sizeof(*elem);
> +		sz += (sizeof(*elem) + MALLOC_ELEM_TRAILER_LEN);
> +		ptr -= (sizeof(*elem) + MALLOC_ELEM_TRAILER_LEN);
>   		elem = elem->prev;
>   	}
>   	malloc_elem_free_list_insert(elem);
> diff --git a/lib/librte_eal/common/malloc_elem.h b/lib/librte_eal/common/malloc_elem.h
> index f04b2d1e4..ce39129d9 100644
> --- a/lib/librte_eal/common/malloc_elem.h
> +++ b/lib/librte_eal/common/malloc_elem.h
> @@ -53,13 +53,13 @@ struct malloc_elem {
>   	volatile enum elem_state state;
>   	uint32_t pad;
>   	size_t size;
> -#ifdef RTE_LIBRTE_MALLOC_DEBUG
> +#ifdef RTE_MALLOC_DEBUG
>   	uint64_t header_cookie;         /* Cookie marking start of data */
>   	                                /* trailer cookie at start + size */
>   #endif
>   } __rte_cache_aligned;
>   
> -#ifndef RTE_LIBRTE_MALLOC_DEBUG
> +#ifndef RTE_MALLOC_DEBUG
>   static const unsigned MALLOC_ELEM_TRAILER_LEN = 0;
>   
>   /* dummy function - just check if pointer is non-null */
> diff --git a/test/test/test_malloc.c b/test/test/test_malloc.c
> index 013fd4407..5558acda4 100644
> --- a/test/test/test_malloc.c
> +++ b/test/test/test_malloc.c
> @@ -108,7 +108,7 @@ test_align_overlap_per_lcore(__attribute__((unused)) void *arg)
>   		}
>   		for(j = 0; j < 1000 ; j++) {
>   			if( *(char *)p1 != 0) {
> -				printf("rte_zmalloc didn't zero"
> +				printf("rte_zmalloc didn't zero "
>   				       "the allocated memory\n");
>   				ret = -1;
>   			}
> @@ -180,7 +180,7 @@ test_reordered_free_per_lcore(__attribute__((unused)) void *arg)
>   		}
>   		for(j = 0; j < 1000 ; j++) {
>   			if( *(char *)p1 != 0) {
> -				printf("rte_zmalloc didn't zero"
> +				printf("rte_zmalloc didn't zero "
>   				       "the allocated memory\n");
>   				ret = -1;
>   			}
> @@ -293,7 +293,7 @@ test_multi_alloc_statistics(void)
>   	struct rte_malloc_socket_stats pre_stats, post_stats ,first_stats, second_stats;
>   	size_t size = 2048;
>   	int align = 1024;
> -#ifndef RTE_LIBRTE_MALLOC_DEBUG
> +#ifndef RTE_MALLOC_DEBUG
>   	int trailer_size = 0;
>   #else
>   	int trailer_size = RTE_CACHE_LINE_SIZE;
> @@ -623,7 +623,7 @@ test_rte_malloc_validate(void)
>   	const size_t request_size = 1024;
>   	size_t allocated_size;
>   	char *data_ptr = rte_malloc(NULL, request_size, RTE_CACHE_LINE_SIZE);
> -#ifdef RTE_LIBRTE_MALLOC_DEBUG
> +#ifdef RTE_MALLOC_DEBUG
>   	int retval;
>   	char *over_write_vals = NULL;
>   #endif
> @@ -645,7 +645,7 @@ test_rte_malloc_validate(void)
>   	if (allocated_size < request_size)
>   		err_return();
>   
> -#ifdef RTE_LIBRTE_MALLOC_DEBUG
> +#ifdef RTE_MALLOC_DEBUG
>   
>   	/****** change the header to be bad */
>   	char save_buf[64];