DPDK patches and discussions
 help / color / mirror / Atom feed
From: Olivier Matz <olivier.matz@6wind.com>
To: "Morten Brørup" <mb@smartsharesystems.com>
Cc: dev@dpdk.org, stephen@networkplumber.org,
	harry.van.haaren@intel.com, konstantin.ananyev@intel.com,
	mattias.ronnblom@ericsson.com, bruce.richardson@intel.com,
	arybchenko@solarflare.com
Subject: Re: [dpdk-dev] [PATCH v8 2/2] mbuf: add unit test for bulk alloc/free functions
Date: Wed, 16 Oct 2019 09:54:33 +0200	[thread overview]
Message-ID: <20191016075433.coyg3eghmj2h2ubi@platinum> (raw)
In-Reply-To: <20191011121458.56388-3-mb@smartsharesystems.com>

Hi Morten,

On Fri, Oct 11, 2019 at 12:14:58PM +0000, Morten Brørup wrote:
> Add unit test for functions for allocating and freeing a bulk of mbufs.
> 
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>

Few minor comments below. You can add my ack in the next version.

> ---
>  app/test/test_mbuf.c | 171 +++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 171 insertions(+)
> 
> diff --git a/app/test/test_mbuf.c b/app/test/test_mbuf.c
> index 2a97afe20..fbce0ca7c 100644
> --- a/app/test/test_mbuf.c
> +++ b/app/test/test_mbuf.c
> @@ -12,6 +12,7 @@
>  #include <sys/queue.h>
>  
>  #include <rte_common.h>
> +#include <rte_errno.h>
>  #include <rte_debug.h>
>  #include <rte_log.h>
>  #include <rte_memory.h>
> @@ -545,6 +546,170 @@ test_pktmbuf_pool(struct rte_mempool *pktmbuf_pool)
>  	return ret;
>  }
>  
> +/*
> + * test bulk allocation and bulk free of mbufs
> + */
> +static int
> +test_pktmbuf_pool_bulk(void)
> +{
> +	struct rte_mempool *pool = NULL;
> +	struct rte_mempool *pool2 = NULL;
> +	unsigned int i;
> +	struct rte_mbuf *m;
> +	struct rte_mbuf *mbufs[NB_MBUF];
> +	int ret = 0;
> +
> +	/* We cannot use the preallocated mbuf pools because their caches
> +	 * prevent us from bulk allocating all objects in them.
> +	 * So we create our own mbuf pools without caches.
> +	 */
> +	printf("Create mbuf pools for bulk allocation.\n");
> +	pool = rte_pktmbuf_pool_create("test_pktmbuf_bulk",
> +			NB_MBUF, 0, 0, MBUF_DATA_SIZE, SOCKET_ID_ANY);
> +	if (pool == NULL) {
> +		printf("rte_pktmbuf_pool_create() failed. rte_errno %d\n",
> +		       rte_errno);
> +		goto err;
> +	}
> +	pool2 = rte_pktmbuf_pool_create("test_pktmbuf_bulk2",
> +			NB_MBUF, 0, 0, MBUF_DATA_SIZE, SOCKET_ID_ANY);
> +	if (pool2 == NULL) {
> +		printf("rte_pktmbuf_pool_create() failed. rte_errno %d\n",
> +		       rte_errno);
> +		goto err;
> +	}
> +
> +	/* Preconditions: Mempools must be full. */
> +	if (!(rte_mempool_full(pool) && rte_mempool_full(pool2))) {
> +		printf("Test precondition failed: mempools not full\n");
> +		goto err;
> +	}
> +	if (!(rte_mempool_avail_count(pool) == NB_MBUF &&
> +			rte_mempool_avail_count(pool2) == NB_MBUF)) {
> +		printf("Test precondition failed: mempools: %u+%u != %u+%u",
> +		       rte_mempool_avail_count(pool),
> +		       rte_mempool_avail_count(pool2),
> +		       NB_MBUF, NB_MBUF);
> +		goto err;
> +	}
> +
> +	printf("Test single bulk alloc, followed by multiple bulk free.\n");
> +
> +	/* Bulk allocate all mbufs in the pool, in one go. */
> +	ret = rte_pktmbuf_alloc_bulk(pool, mbufs, NB_MBUF);
> +	if (ret != 0) {
> +		printf("rte_pktmbuf_alloc_bulk() failed: %d\n", ret);
> +		goto err;
> +	}
> +	/* Test that they have been removed from the pool. */
> +	if (!rte_mempool_empty(pool)) {
> +		printf("mempool not empty\n");
> +		goto err;
> +	}
> +	/* Bulk free all mbufs, in four steps. */
> +	for (i = 0; i < NB_MBUF; i += NB_MBUF / 4) {

Maybe add a RTE_BUILD_BUG_ON(NB_MBUF % 4 == 0)

> +		rte_pktmbuf_free_bulk(&mbufs[i], NB_MBUF / 4);
> +		/* Test that they have been returned to the pool. */
> +		if (rte_mempool_avail_count(pool) != i + NB_MBUF / 4) {
> +			printf("mempool avail count incorrect\n");
> +			goto err;
> +		}
> +	}
> +
> +	printf("Test multiple bulk alloc, followed by single bulk free.\n");
> +
> +	/* Bulk allocate all mbufs in the pool, in four steps. */
> +	for (i = 0; i < NB_MBUF; i += NB_MBUF / 4) {
> +		ret = rte_pktmbuf_alloc_bulk(pool, &mbufs[i], NB_MBUF / 4);
> +		if (ret != 0) {
> +			printf("rte_pktmbuf_alloc_bulk() failed: %d\n", ret);
> +			goto err;
> +		}
> +	}
> +	/* Test that they have been removed from the pool. */
> +	if (!rte_mempool_empty(pool)) {
> +		printf("mempool not empty\n");
> +		goto err;
> +	}
> +	/* Bulk free all mbufs, in one go. */
> +	rte_pktmbuf_free_bulk(mbufs, NB_MBUF);
> +	/* Test that they have been returned to the pool. */
> +	if (!rte_mempool_full(pool)) {
> +		printf("mempool not full\n");
> +		goto err;
> +	}
> +
> +	printf("Test bulk free of single long chain.\n");
> +
> +	/* Bulk allocate all mbufs in the pool, in one go. */
> +	ret = rte_pktmbuf_alloc_bulk(pool, mbufs, NB_MBUF);
> +	if (ret != 0) {
> +		printf("rte_pktmbuf_alloc_bulk() failed: %d\n", ret);
> +		goto err;
> +	}
> +	/* Create a long mbuf chain. */
> +	for (i = 1; i < NB_MBUF; i++) {
> +		ret = rte_pktmbuf_chain(mbufs[0], mbufs[i]);
> +		if (ret != 0) {
> +			printf("rte_pktmbuf_chain() failed: %d\n", ret);
> +			goto err;
> +		}
> +		mbufs[i] = NULL;
> +	}
> +	/* Free the mbuf chain containing all the mbufs. */
> +	rte_pktmbuf_free_bulk(mbufs, 1);
> +	/* Test that they have been returned to the pool. */
> +	if (!rte_mempool_full(pool)) {
> +		printf("mempool not full\n");
> +		goto err;
> +	}
> +
> +	printf("Test bulk free of multiple chains using multiple pools.\n");
> +
> +	/* Create mbuf chains containing mbufs from different pools. */
> +	for (i = 0; i < NB_MBUF * 2; i++) {
> +		m = rte_pktmbuf_alloc((i & 4) ? pool2 : pool);
> +		if (m == NULL) {
> +			printf("rte_pktmbuf_alloc() failed (%u)\n", i);
> +			goto err;
> +		}
> +		if ((i % 16) == 0) {
> +			mbufs[i / 16] = m;
> +		} else {
> +			rte_pktmbuf_chain(mbufs[i / 16], m);
> +		}
> +	}

The 16 could be a #define CHAIN_LEN.

Maybe another bug_on:
RTE_BUILD_BUG_ON(NB_MBUF % (CHAIN_LEN / 2) == 0)


> +	/* Test that both pools have been emptied. */
> +	if (!(rte_mempool_empty(pool) && rte_mempool_empty(pool2))) {
> +		printf("mempools not empty\n");
> +		goto err;
> +	}
> +	/* Free one mbuf chain. */
> +	rte_pktmbuf_free_bulk(mbufs, 1);
> +	/* Test that the segments have been returned to the pools. */
> +	if (!(rte_mempool_avail_count(pool) == 8 &&
> +			rte_mempool_avail_count(pool2) == 8)) {
> +		printf("all segments of first mbuf have not been returned\n");
> +		goto err;
> +	}
> +	/* Free the remaining mbuf chains. */
> +	rte_pktmbuf_free_bulk(&mbufs[1], NB_MBUF * 2 / 16 - 1);
> +	/* Test that they have been returned to the pools. */
> +	if (!(rte_mempool_full(pool) && rte_mempool_full(pool2))) {
> +		printf("mempools not full\n");
> +		goto err;
> +	}

Use CHAIN_LEN in the code above.

> +
> +	rte_mempool_free(pool);
> +	rte_mempool_free(pool2);
> +	return 0;
> +
> +err:
> +	rte_mempool_free(pool);
> +	rte_mempool_free(pool2);
> +	return -1;
> +}
> +
>  /*
>   * test that the pointer to the data on a packet mbuf is set properly
>   */
> @@ -1162,6 +1327,12 @@ test_mbuf(void)
>  		goto err;
>  	}
>  
> +	/* test bulk mbuf alloc and free */
> +	if (test_pktmbuf_pool_bulk() < 0) {
> +		printf("test_pktmbuf_pool_bulk() failed\n");
> +		goto err;
> +	}
> +
>  	/* test that the pointer to the data on a packet mbuf is set properly */
>  	if (test_pktmbuf_pool_ptr(pktmbuf_pool) < 0) {
>  		printf("test_pktmbuf_pool_ptr() failed\n");
> -- 
> 2.17.1
> 

Thanks!
Olivier

      reply	other threads:[~2019-10-16  7:54 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-11 12:14 [dpdk-dev] [PATCH v8 0/2] mbuf: add bulk free function Morten Brørup
2019-10-11 12:14 ` [dpdk-dev] [PATCH v8 1/2] " Morten Brørup
2019-10-16  7:35   ` Olivier Matz
2019-10-11 12:14 ` [dpdk-dev] [PATCH v8 2/2] mbuf: add unit test for bulk alloc/free functions Morten Brørup
2019-10-16  7:54   ` Olivier Matz [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191016075433.coyg3eghmj2h2ubi@platinum \
    --to=olivier.matz@6wind.com \
    --cc=arybchenko@solarflare.com \
    --cc=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=harry.van.haaren@intel.com \
    --cc=konstantin.ananyev@intel.com \
    --cc=mattias.ronnblom@ericsson.com \
    --cc=mb@smartsharesystems.com \
    --cc=stephen@networkplumber.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).