From: John Miller <john.miller@atomicrules.com>
To: dev@dpdk.org, ferruh.yigit@intel.com
Cc: shepard.siegel@atomicrules.com,
John Miller <john.miller@atomicrules.com>
Subject: [PATCH v3 5/7] net/ark: support chunk DMA transfers
Date: Tue, 15 Feb 2022 16:19:24 -0600 [thread overview]
Message-ID: <20220215221926.409039-5-john.miller@atomicrules.com> (raw)
In-Reply-To: <20220215221926.409039-1-john.miller@atomicrules.com>
Add support for chunk DMA transfers.
Chunk mpu transfer use 64 objects (512 byte) to maintain memory
read alignment.
Align mpu memory allocation to be at 512 byte boundaries.
Reduce force-close allocation from 10000 objects to 64 objects.
Signed-off-by: John Miller <john.miller@atomicrules.com>
---
drivers/net/ark/ark_ethdev_rx.c | 16 +++++++++-------
1 file changed, 9 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ark/ark_ethdev_rx.c b/drivers/net/ark/ark_ethdev_rx.c
index 0478702cbe..0fbb2603db 100644
--- a/drivers/net/ark/ark_ethdev_rx.c
+++ b/drivers/net/ark/ark_ethdev_rx.c
@@ -12,6 +12,7 @@
#define ARK_RX_META_SIZE 32
#define ARK_RX_META_OFFSET (RTE_PKTMBUF_HEADROOM - ARK_RX_META_SIZE)
+#define ARK_RX_MPU_CHUNK (64U)
/* Forward declarations */
struct ark_rx_queue;
@@ -104,7 +105,7 @@ static inline void
eth_ark_rx_update_cons_index(struct ark_rx_queue *queue, uint32_t cons_index)
{
queue->cons_index = cons_index;
- if ((cons_index + queue->queue_size - queue->seed_index) >= 64U) {
+ if ((cons_index + queue->queue_size - queue->seed_index) >= ARK_RX_MPU_CHUNK) {
eth_ark_rx_seed_mbufs(queue);
ark_mpu_set_producer(queue->mpu, queue->seed_index);
}
@@ -179,12 +180,12 @@ eth_ark_dev_rx_queue_setup(struct rte_eth_dev *dev,
queue->reserve_q =
rte_zmalloc_socket("Ark_rx_queue mbuf",
nb_desc * sizeof(struct rte_mbuf *),
- 64,
+ 512,
socket_id);
queue->paddress_q =
rte_zmalloc_socket("Ark_rx_queue paddr",
nb_desc * sizeof(rte_iova_t),
- 64,
+ 512,
socket_id);
if (queue->reserve_q == 0 || queue->paddress_q == 0) {
@@ -446,7 +447,8 @@ eth_ark_rx_stop_queue(struct rte_eth_dev *dev, uint16_t queue_id)
static inline int
eth_ark_rx_seed_mbufs(struct ark_rx_queue *queue)
{
- uint32_t limit = queue->cons_index + queue->queue_size;
+ uint32_t limit = (queue->cons_index & ~(ARK_RX_MPU_CHUNK - 1)) +
+ queue->queue_size;
uint32_t seed_index = queue->seed_index;
uint32_t count = 0;
@@ -609,14 +611,14 @@ eth_ark_udm_force_close(struct rte_eth_dev *dev)
ark_mpu_start(queue->mpu);
/* Add some buffers */
- index = 100000 + queue->seed_index;
+ index = ARK_RX_MPU_CHUNK + queue->seed_index;
ark_mpu_set_producer(queue->mpu, index);
}
/* Wait to allow data to pass */
usleep(100);
- ARK_PMD_LOG(DEBUG, "UDM forced flush attempt, stopped = %d\n",
- ark_udm_is_flushed(ark->udm.v));
+ ARK_PMD_LOG(NOTICE, "UDM forced flush attempt, stopped = %d\n",
+ ark_udm_is_flushed(ark->udm.v));
}
ark_udm_reset(ark->udm.v);
}
--
2.25.1
next prev parent reply other threads:[~2022-02-15 22:19 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20220119191255.273988-ed.czeck@atomicrules.com>
2022-02-10 14:34 ` [PATCH v2 1/3] net/ark: add device capabilities record John Miller
2022-02-10 14:34 ` [PATCH v2 2/3] net/ark: support arbitrary mbuf size John Miller
2022-02-10 14:34 ` [PATCH v2 3/3] net/ark: support chunk DMA transfers John Miller
2022-02-11 11:39 ` [PATCH v2 1/3] net/ark: add device capabilities record John Miller
2022-02-11 11:39 ` [PATCH v2 2/3] net/ark: support arbitrary mbuf size John Miller
2022-02-11 11:39 ` [PATCH v2 3/3] net/ark: support chunk DMA transfers John Miller
2022-02-14 13:59 ` Ferruh Yigit
2022-02-14 14:00 ` [PATCH v2 1/3] net/ark: add device capabilities record Ferruh Yigit
2022-02-15 22:19 ` [PATCH v3 1/7] " John Miller
2022-02-15 22:19 ` [PATCH v3 2/7] net/ark: add support for new devices John Miller
2022-02-15 22:19 ` [PATCH v3 3/7] net/ark: support arbitrary mbuf size John Miller
2022-02-15 22:19 ` [PATCH v3 4/7] net/ark: packet generator and checker status update John Miller
2022-02-15 22:19 ` John Miller [this message]
2022-02-15 22:19 ` [PATCH v3 6/7] net/ark: add memory write barriers in critical code paths John Miller
2022-02-15 22:19 ` [PATCH v3 7/7] net/ark: add performance optimizations John Miller
2022-02-16 10:25 ` [PATCH v3 1/7] net/ark: add device capabilities record Ferruh Yigit
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=20220215221926.409039-5-john.miller@atomicrules.com \
--to=john.miller@atomicrules.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=shepard.siegel@atomicrules.com \
/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).