Hi,
From: David Marchand [mailto:david.marchand@6wind.com]
Sent: Friday, February 13, 2015 7:08 PM
To: Chen, Jing D
Cc: dev@dpdk.org; Zhang, Helin; Qiu, Michael; Neil Horman; Thomas Monjalon; Shaw, Jeffrey B
Subject: Re: [PATCH v5 06/17] fm10k: add rx_queue_setup/release function
Hello,
On Fri, Feb 13, 2015 at 9:19 AM, Chen Jing D(Mark) <jing.d.chen@intel.com> wrote:
[snip]
+static int
+fm10k_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
+ uint16_t nb_desc, unsigned int socket_id,
+ const struct rte_eth_rxconf *conf, struct rte_mempool *mp)
+{
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct fm10k_rx_queue *q;
+ const struct rte_memzone *mz;
+
+ PMD_INIT_FUNC_TRACE();
+
+ /* make sure the mempool element size can account for alignment. Use
+ * RTE_LOG directly to make sure this error is seen. */
Comment is not valid anymore since you call PMD_INIT_LOG.
[Mark] Thanks! I'll change the comments.
+ if (!mempool_element_size_valid(mp)) {
+ PMD_INIT_LOG(ERR, "Error : Mempool element size is too small");
+ return (-EINVAL);
+ }
+
--
David Marchand
@@ -41,6 +41,7 @@
#include "fm10k.h"
#include "base/fm10k_api.h"
+#define FM10K_RX_BUFF_ALIGN 512
/* Default delay to acquire mailbox lock */
#define FM10K_MBXLOCK_DELAY_US 20
@@ -67,6 +68,46 @@ fm10k_mbx_unlock(struct fm10k_hw *hw)
rte_spinlock_unlock(FM10K_DEV_PRIVATE_TO_MBXLOCK(hw->back));
}
+/*
+ * clean queue, descriptor rings, free software buffers used when stopping
+ * device.
+ */
+static inline void
+rx_queue_clean(struct fm10k_rx_queue *q)
+{
+ union fm10k_rx_desc zero = {.q = {0, 0, 0, 0} };
+ uint32_t i;
+ PMD_INIT_FUNC_TRACE();
+
+ /* zero descriptor rings */
+ for (i = 0; i < q->nb_desc; ++i)
+ q->hw_ring[i] = zero;
+
+ /* free software buffers */
+ for (i = 0; i < q->nb_desc; ++i) {
+ if (q->sw_ring[i]) {
+ rte_pktmbuf_free_seg(q->sw_ring[i]);
+ q->sw_ring[i] = NULL;
+ }
+ }
+}
+
+/*
+ * free all queue memory used when releasing the queue (i.e. configure)
+ */
+static inline void
+rx_queue_free(struct fm10k_rx_queue *q)
+{
+ PMD_INIT_FUNC_TRACE();
+ if (q) {
+ PMD_INIT_LOG(DEBUG, "Freeing rx queue %p", q);
+ rx_queue_clean(q);
+ if (q->sw_ring)
+ rte_free(q->sw_ring);
+ rte_free(q);
+ }
+}
+
static int
fm10k_dev_configure(struct rte_eth_dev *dev)
{
@@ -186,6 +227,218 @@ fm10k_dev_infos_get(struct rte_eth_dev *dev,
}
+static inline int
+check_nb_desc(uint16_t min, uint16_t max, uint16_t mult, uint16_t request)
+{
+ if ((request < min) || (request > max) || ((request % mult) != 0))
+ return -1;
+ else
+ return 0;
+}
+
+/*
+ * Create a memzone for hardware descriptor rings. Malloc cannot be used since
+ * the physical address is required. If the memzone is already created, then
+ * this function returns a pointer to the existing memzone.
+ */
+static inline const struct rte_memzone *
+allocate_hw_ring(const char *driver_name, const char *ring_name,
+ uint8_t port_id, uint16_t queue_id, int socket_id,
+ uint32_t size, uint32_t align)
+{
+ char name[RTE_MEMZONE_NAMESIZE];
+ const struct rte_memzone *mz;
+
+ snprintf(name, sizeof(name), "%s_%s_%d_%d_%d",
+ driver_name, ring_name, port_id, queue_id, socket_id);
+
+ /* return the memzone if it already exists */
+ mz = rte_memzone_lookup(name);
+ if (mz)
+ return mz;
+
+#ifdef RTE_LIBRTE_XEN_DOM0
+ return rte_memzone_reserve_bounded(name, size, socket_id, 0, align,
+ RTE_PGSIZE_2M);
+#else
+ return rte_memzone_reserve_aligned(name, size, socket_id, 0, align);
+#endif
+}
+
+static inline int
+check_thresh(uint16_t min, uint16_t max, uint16_t div, uint16_t request)
+{
+ if ((request < min) || (request > max) || ((div % request) != 0))
+ return -1;
+ else
+ return 0;
+}
+
+static inline int
+handle_rxconf(struct fm10k_rx_queue *q, const struct rte_eth_rxconf *conf)
+{
+ uint16_t rx_free_thresh;
+
+ if (conf->rx_free_thresh == 0)
+ rx_free_thresh = FM10K_RX_FREE_THRESH_DEFAULT(q);
+ else
+ rx_free_thresh = conf->rx_free_thresh;
+
+ /* make sure the requested threshold satisfies the constraints */
+ if (check_thresh(FM10K_RX_FREE_THRESH_MIN(q),
+ FM10K_RX_FREE_THRESH_MAX(q),
+ FM10K_RX_FREE_THRESH_DIV(q),
+ rx_free_thresh)) {
+ PMD_INIT_LOG(ERR, "rx_free_thresh (%u) must be "
+ "less than or equal to %u, "
+ "greater than or equal to %u, "
+ "and a divisor of %u",
+ rx_free_thresh, FM10K_RX_FREE_THRESH_MAX(q),
+ FM10K_RX_FREE_THRESH_MIN(q),
+ FM10K_RX_FREE_THRESH_DIV(q));
+ return (-EINVAL);
+ }
+
+ q->alloc_thresh = rx_free_thresh;
+ q->drop_en = conf->rx_drop_en;
+ q->rx_deferred_start = conf->rx_deferred_start;
+
+ return 0;
+}
+
+/*
+ * Hardware requires specific alignment for Rx packet buffers. At
+ * least one of the following two conditions must be satisfied.
+ * 1. Address is 512B aligned
+ * 2. Address is 8B aligned and buffer does not cross 4K boundary.
+ *
+ * As such, the driver may need to adjust the DMA address within the
+ * buffer by up to 512B. The mempool element size is checked here
+ * to make sure a maximally sized Ethernet frame can still be wholly
+ * contained within the buffer after 512B alignment.
+ *
+ * return 1 if the element size is valid, otherwise return 0.
+ */
+static int
+mempool_element_size_valid(struct rte_mempool *mp)
+{
+ uint32_t min_size;
+
+ /* elt_size includes mbuf header and headroom */
+ min_size = mp->elt_size - sizeof(struct rte_mbuf) -
+ RTE_PKTMBUF_HEADROOM;
+
+ /* account for up to 512B of alignment */
+ min_size -= FM10K_RX_BUFF_ALIGN;
+
+ /* sanity check for overflow */
+ if (min_size > mp->elt_size)
+ return 0;
+
+ if (min_size < ETHER_MAX_VLAN_FRAME_LEN)
+ return 0;
+
+ /* size is valid */
+ return 1;
+}
+
+static int
+fm10k_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
+ uint16_t nb_desc, unsigned int socket_id,
+ const struct rte_eth_rxconf *conf, struct rte_mempool *mp)
+{
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct fm10k_rx_queue *q;
+ const struct rte_memzone *mz;
+
+ PMD_INIT_FUNC_TRACE();
+
+ /* make sure the mempool element size can account for alignment. Use
+ * RTE_LOG directly to make sure this error is seen. */
+ if (!mempool_element_size_valid(mp)) {
+ PMD_INIT_LOG(ERR, "Error : Mempool element size is too small");
+ return (-EINVAL);
+ }
+
+ /* make sure a valid number of descriptors have been requested */
+ if (check_nb_desc(FM10K_MIN_RX_DESC, FM10K_MAX_RX_DESC,
+ FM10K_MULT_RX_DESC, nb_desc)) {
+ PMD_INIT_LOG(ERR, "Number of Rx descriptors (%u) must be "
+ "less than or equal to %"PRIu32", "
+ "greater than or equal to %u, "
+ "and a multiple of %u",
+ nb_desc, (uint32_t)FM10K_MAX_RX_DESC, FM10K_MIN_RX_DESC,
+ FM10K_MULT_RX_DESC);
+ return (-EINVAL);
+ }
+
+ /*
+ * if this queue existed already, free the associated memory. The
+ * queue cannot be reused in case we need to allocate memory on
+ * different socket than was previously used.
+ */
+ if (dev->data->rx_queues[queue_id] != NULL) {
+ rx_queue_free(dev->data->rx_queues[queue_id]);
+ dev->data->rx_queues[queue_id] = NULL;
+ }
+
+ /* allocate memory for the queue structure */
+ q = rte_zmalloc_socket("fm10k", sizeof(*q), RTE_CACHE_LINE_SIZE,
+ socket_id);
+ if (q == NULL) {
+ PMD_INIT_LOG(ERR, "Cannot allocate queue structure");
+ return (-ENOMEM);
+ }
+
+ /* setup queue */
+ q->mp = mp;
+ q->nb_desc = nb_desc;
+ q->port_id = dev->data->port_id;
+ q->queue_id = queue_id;
+ q->tail_ptr = (volatile uint32_t *)
+ &((uint32_t *)hw->hw_addr)[FM10K_RDT(queue_id)];
+ if (handle_rxconf(q, conf))
+ return (-EINVAL);
+
+ /* allocate memory for the software ring */
+ q->sw_ring = rte_zmalloc_socket("fm10k sw ring",
+ nb_desc * sizeof(struct rte_mbuf *),
+ RTE_CACHE_LINE_SIZE, socket_id);
+ if (q->sw_ring == NULL) {
+ PMD_INIT_LOG(ERR, "Cannot allocate software ring");
+ rte_free(q);
+ return (-ENOMEM);
+ }
+
+ /*
+ * allocate memory for the hardware descriptor ring. A memzone large
+ * enough to hold the maximum ring size is requested to allow for
+ * resizing in later calls to the queue setup function.
+ */
+ mz = allocate_hw_ring(dev->driver->pci_drv.name, "rx_ring",
+ dev->data->port_id, queue_id, socket_id,
+ FM10K_MAX_RX_RING_SZ, FM10K_ALIGN_RX_DESC);
+ if (mz == NULL) {
+ PMD_INIT_LOG(ERR, "Cannot allocate hardware ring");
+ rte_free(q->sw_ring);
+ rte_free(q);
+ return (-ENOMEM);
+ }
+ q->hw_ring = mz->addr;
+ q->hw_ring_phys_addr = mz->phys_addr;
+
+ dev->data->rx_queues[queue_id] = q;
+ return 0;
+}
+
+static void
+fm10k_rx_queue_release(void *queue)
+{
+ PMD_INIT_FUNC_TRACE();
+
+ rx_queue_free(queue);
+}
+
static int
fm10k_reta_update(struct rte_eth_dev *dev,
struct rte_eth_rss_reta_entry64 *reta_conf,
@@ -325,6 +578,8 @@ static struct eth_dev_ops fm10k_eth_dev_ops = {
.stats_reset = fm10k_stats_reset,
.link_update = fm10k_link_update,
.dev_infos_get = fm10k_dev_infos_get,
+ .rx_queue_setup = fm10k_rx_queue_setup,
+ .rx_queue_release = fm10k_rx_queue_release,
.reta_update = fm10k_reta_update,
.reta_query = fm10k_reta_query,
};