[dpdk-dev] [PATCH 1/3] kni: remove single mempool, single mem_chunk restriction

[dpdk-dev] [PATCH 1/3] kni: remove single mempool, single mem_chunk restriction

[Ferruh Yigit]

Use mempool buf_addr and buf_physaddr fields for address translation.

Since each mbuf address calculated separately, the restriction of all
mbufs should come from a continuous memory restriction is no more valid.

mbuf related FIFO's content changed, rx_q and alloc_q now carries
physical address of mbufs. tx_q and free_q content not changed, they
still carries virtual address of mbufs.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 .../linuxapp/eal/include/exec-env/rte_kni_common.h |   3 +-
 lib/librte_eal/linuxapp/kni/kni_net.c              | 123 ++++++++++++++-------
 lib/librte_kni/rte_kni.c                           |  39 ++++++-
 3 files changed, 118 insertions(+), 47 deletions(-)

diff --git a/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h b/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h
index 2acdfd9..ea1cd0b 100644
--- a/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h
+++ b/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h
@@ -111,7 +111,8 @@ struct rte_kni_fifo {
  */
 struct rte_kni_mbuf {
 	void *buf_addr __attribute__((__aligned__(RTE_CACHE_LINE_SIZE)));
-	char pad0[10];
+	uint64_t buf_physaddr;
+	char pad0[2];
 	uint16_t data_off;      /**< Start address of data in segment buffer. */
 	char pad1[2];
 	uint8_t nb_segs;        /**< Number of segments. */
diff --git a/lib/librte_eal/linuxapp/kni/kni_net.c b/lib/librte_eal/linuxapp/kni/kni_net.c
index fc82193..7d411b4 100644
--- a/lib/librte_eal/linuxapp/kni/kni_net.c
+++ b/lib/librte_eal/linuxapp/kni/kni_net.c
@@ -61,6 +61,44 @@ static int kni_net_process_request(struct kni_dev *kni,
 /* kni rx function pointer, with default to normal rx */
 static kni_net_rx_t kni_net_rx_func = kni_net_rx_normal;
 
+/* physical address to kernel virtual address */
+static void *
+pa2kva(void *pa)
+{
+	return phys_to_virt((unsigned long)pa);
+}
+
+/* physical address to virtual address */
+static void *
+pa2va(void *pa, struct rte_kni_mbuf *m)
+{
+	void *va;
+
+	va = (void *)((unsigned long)pa +
+			(unsigned long)m->buf_addr -
+			(unsigned long)m->buf_physaddr);
+	return va;
+}
+
+/* mbuf data kernel virtual address from mbuf kernel virtual address */
+static void *
+kva2data_kva(struct rte_kni_mbuf *m)
+{
+	return phys_to_virt(m->buf_physaddr + m->data_off);
+}
+
+/* virtual address to physical address */
+static void *
+va2pa(void *va, struct rte_kni_mbuf *m)
+{
+	void *pa;
+
+	pa = (void *)((unsigned long)va -
+			((unsigned long)m->buf_addr -
+			 (unsigned long)m->buf_physaddr));
+	return pa;
+}
+
 /*
  * Open and close
  */
@@ -125,8 +163,9 @@ kni_net_rx_normal(struct kni_dev *kni)
 	uint32_t len;
 	unsigned i, num_rx, num_fq;
 	struct rte_kni_mbuf *kva;
-	struct rte_kni_mbuf *va[MBUF_BURST_SZ];
-	void * data_kva;
+	void *pa[MBUF_BURST_SZ];
+	void *va[MBUF_BURST_SZ];
+	void *data_kva;
 
 	struct sk_buff *skb;
 	struct net_device *dev = kni->net_dev;
@@ -142,17 +181,16 @@ kni_net_rx_normal(struct kni_dev *kni)
 	num_rx = min(num_fq, (unsigned)MBUF_BURST_SZ);
 
 	/* Burst dequeue from rx_q */
-	num_rx = kni_fifo_get(kni->rx_q, (void **)va, num_rx);
+	num_rx = kni_fifo_get(kni->rx_q, pa, num_rx);
 	if (num_rx == 0)
 		return;
 
 	/* Transfer received packets to netif */
 	for (i = 0; i < num_rx; i++) {
-		kva = (void *)va[i] - kni->mbuf_va + kni->mbuf_kva;
+		kva = pa2kva(pa[i]);
 		len = kva->pkt_len;
-
-		data_kva = kva->buf_addr + kva->data_off - kni->mbuf_va
-				+ kni->mbuf_kva;
+		data_kva = kva2data_kva(kva);
+		va[i] = pa2va(pa[i], kva);
 
 		skb = dev_alloc_skb(len + 2);
 		if (!skb) {
@@ -178,9 +216,8 @@ kni_net_rx_normal(struct kni_dev *kni)
 				if (!kva->next)
 					break;
 
-				kva = kva->next - kni->mbuf_va + kni->mbuf_kva;
-				data_kva = kva->buf_addr + kva->data_off
-					- kni->mbuf_va + kni->mbuf_kva;
+				kva = pa2kva(va2pa(kva->next, kva));
+				data_kva = kva2data_kva(kva);
 			}
 		}
 
@@ -197,7 +234,7 @@ kni_net_rx_normal(struct kni_dev *kni)
 	}
 
 	/* Burst enqueue mbufs into free_q */
-	ret = kni_fifo_put(kni->free_q, (void **)va, num_rx);
+	ret = kni_fifo_put(kni->free_q, va, num_rx);
 	if (ret != num_rx)
 		/* Failing should not happen */
 		KNI_ERR("Fail to enqueue entries into free_q\n");
@@ -213,11 +250,13 @@ kni_net_rx_lo_fifo(struct kni_dev *kni)
 	uint32_t len;
 	unsigned i, num, num_rq, num_tq, num_aq, num_fq;
 	struct rte_kni_mbuf *kva;
-	struct rte_kni_mbuf *va[MBUF_BURST_SZ];
+	void *pa[MBUF_BURST_SZ];
+	void *va[MBUF_BURST_SZ];
 	void * data_kva;
 
 	struct rte_kni_mbuf *alloc_kva;
-	struct rte_kni_mbuf *alloc_va[MBUF_BURST_SZ];
+	void *alloc_pa[MBUF_BURST_SZ];
+	void *alloc_va[MBUF_BURST_SZ];
 	void *alloc_data_kva;
 
 	/* Get the number of entries in rx_q */
@@ -243,26 +282,25 @@ kni_net_rx_lo_fifo(struct kni_dev *kni)
 		return;
 
 	/* Burst dequeue from rx_q */
-	ret = kni_fifo_get(kni->rx_q, (void **)va, num);
+	ret = kni_fifo_get(kni->rx_q, pa, num);
 	if (ret == 0)
 		return; /* Failing should not happen */
 
 	/* Dequeue entries from alloc_q */
-	ret = kni_fifo_get(kni->alloc_q, (void **)alloc_va, num);
+	ret = kni_fifo_get(kni->alloc_q, alloc_pa, num);
 	if (ret) {
 		num = ret;
 		/* Copy mbufs */
 		for (i = 0; i < num; i++) {
-			kva = (void *)va[i] - kni->mbuf_va + kni->mbuf_kva;
+			kva = pa2kva(pa[i]);
 			len = kva->pkt_len;
-			data_kva = kva->buf_addr + kva->data_off -
-					kni->mbuf_va + kni->mbuf_kva;
-
-			alloc_kva = (void *)alloc_va[i] - kni->mbuf_va +
-							kni->mbuf_kva;
-			alloc_data_kva = alloc_kva->buf_addr +
-					alloc_kva->data_off - kni->mbuf_va +
-							kni->mbuf_kva;
+			data_kva = kva2data_kva(kva);
+			va[i] = pa2va(pa[i], kva);
+
+			alloc_kva = pa2kva(alloc_pa[i]);
+			alloc_data_kva = kva2data_kva(alloc_kva);
+			alloc_va[i] = pa2va(alloc_pa[i], alloc_kva);
+
 			memcpy(alloc_data_kva, data_kva, len);
 			alloc_kva->pkt_len = len;
 			alloc_kva->data_len = len;
@@ -272,14 +310,14 @@ kni_net_rx_lo_fifo(struct kni_dev *kni)
 		}
 
 		/* Burst enqueue mbufs into tx_q */
-		ret = kni_fifo_put(kni->tx_q, (void **)alloc_va, num);
+		ret = kni_fifo_put(kni->tx_q, alloc_va, num);
 		if (ret != num)
 			/* Failing should not happen */
 			KNI_ERR("Fail to enqueue mbufs into tx_q\n");
 	}
 
 	/* Burst enqueue mbufs into free_q */
-	ret = kni_fifo_put(kni->free_q, (void **)va, num);
+	ret = kni_fifo_put(kni->free_q, va, num);
 	if (ret != num)
 		/* Failing should not happen */
 		KNI_ERR("Fail to enqueue mbufs into free_q\n");
@@ -302,8 +340,9 @@ kni_net_rx_lo_fifo_skb(struct kni_dev *kni)
 	uint32_t len;
 	unsigned i, num_rq, num_fq, num;
 	struct rte_kni_mbuf *kva;
-	struct rte_kni_mbuf *va[MBUF_BURST_SZ];
-	void * data_kva;
+	void *pa[MBUF_BURST_SZ];
+	void *va[MBUF_BURST_SZ];
+	void *data_kva;
 
 	struct sk_buff *skb;
 	struct net_device *dev = kni->net_dev;
@@ -323,16 +362,16 @@ kni_net_rx_lo_fifo_skb(struct kni_dev *kni)
 		return;
 
 	/* Burst dequeue mbufs from rx_q */
-	ret = kni_fifo_get(kni->rx_q, (void **)va, num);
+	ret = kni_fifo_get(kni->rx_q, pa, num);
 	if (ret == 0)
 		return;
 
 	/* Copy mbufs to sk buffer and then call tx interface */
 	for (i = 0; i < num; i++) {
-		kva = (void *)va[i] - kni->mbuf_va + kni->mbuf_kva;
+		kva = pa2kva(pa[i]);
 		len = kva->pkt_len;
-		data_kva = kva->buf_addr + kva->data_off - kni->mbuf_va +
-				kni->mbuf_kva;
+		data_kva = kva2data_kva(kva);
+		va[i] = pa2va(pa[i], kva);
 
 		skb = dev_alloc_skb(len + 2);
 		if (skb == NULL)
@@ -370,9 +409,8 @@ kni_net_rx_lo_fifo_skb(struct kni_dev *kni)
 				if (!kva->next)
 					break;
 
-				kva = kva->next - kni->mbuf_va + kni->mbuf_kva;
-				data_kva = kva->buf_addr + kva->data_off
-					- kni->mbuf_va + kni->mbuf_kva;
+				kva = pa2kva(va2pa(kva->next, kva));
+				data_kva = kva2data_kva(kva);
 			}
 		}
 
@@ -387,7 +425,7 @@ kni_net_rx_lo_fifo_skb(struct kni_dev *kni)
 	}
 
 	/* enqueue all the mbufs from rx_q into free_q */
-	ret = kni_fifo_put(kni->free_q, (void **)&va, num);
+	ret = kni_fifo_put(kni->free_q, va, num);
 	if (ret != num)
 		/* Failing should not happen */
 		KNI_ERR("Fail to enqueue mbufs into free_q\n");
@@ -426,7 +464,8 @@ kni_net_tx(struct sk_buff *skb, struct net_device *dev)
 	unsigned ret;
 	struct kni_dev *kni = netdev_priv(dev);
 	struct rte_kni_mbuf *pkt_kva = NULL;
-	struct rte_kni_mbuf *pkt_va = NULL;
+	void *pkt_pa = NULL;
+	void *pkt_va = NULL;
 
 	/* save the timestamp */
 #ifdef HAVE_TRANS_START_HELPER
@@ -453,13 +492,13 @@ kni_net_tx(struct sk_buff *skb, struct net_device *dev)
 	}
 
 	/* dequeue a mbuf from alloc_q */
-	ret = kni_fifo_get(kni->alloc_q, (void **)&pkt_va, 1);
+	ret = kni_fifo_get(kni->alloc_q, &pkt_pa, 1);
 	if (likely(ret == 1)) {
 		void *data_kva;
 
-		pkt_kva = (void *)pkt_va - kni->mbuf_va + kni->mbuf_kva;
-		data_kva = pkt_kva->buf_addr + pkt_kva->data_off - kni->mbuf_va
-				+ kni->mbuf_kva;
+		pkt_kva = pa2kva(pkt_pa);
+		data_kva = kva2data_kva(pkt_kva);
+		pkt_va = pa2va(pkt_pa, pkt_kva);
 
 		len = skb->len;
 		memcpy(data_kva, skb->data, len);
@@ -471,7 +510,7 @@ kni_net_tx(struct sk_buff *skb, struct net_device *dev)
 		pkt_kva->data_len = len;
 
 		/* enqueue mbuf into tx_q */
-		ret = kni_fifo_put(kni->tx_q, (void **)&pkt_va, 1);
+		ret = kni_fifo_put(kni->tx_q, &pkt_va, 1);
 		if (unlikely(ret != 1)) {
 			/* Failing should not happen */
 			KNI_ERR("Fail to enqueue mbuf into tx_q\n");
diff --git a/lib/librte_kni/rte_kni.c b/lib/librte_kni/rte_kni.c
index f48b72b..0f7c9e5 100644
--- a/lib/librte_kni/rte_kni.c
+++ b/lib/librte_kni/rte_kni.c
@@ -455,6 +455,20 @@ kni_free_fifo(struct rte_kni_fifo *fifo)
 	} while (ret);
 }
 
+static void
+kni_free_fifo_phy(struct rte_kni_fifo *fifo)
+{
+	void *mbuf_phys;
+	int ret;
+
+	do {
+		ret = kni_fifo_get(fifo, &mbuf_phys, 1);
+		/*
+		 * TODO: free mbufs
+		 */
+	} while (ret);
+}
+
 int
 rte_kni_release(struct rte_kni *kni)
 {
@@ -472,8 +486,8 @@ rte_kni_release(struct rte_kni *kni)
 
 	/* mbufs in all fifo should be released, except request/response */
 	kni_free_fifo(kni->tx_q);
-	kni_free_fifo(kni->rx_q);
-	kni_free_fifo(kni->alloc_q);
+	kni_free_fifo_phy(kni->rx_q);
+	kni_free_fifo_phy(kni->alloc_q);
 	kni_free_fifo(kni->free_q);
 
 	slot_id = kni->slot_id;
@@ -537,10 +551,25 @@ rte_kni_handle_request(struct rte_kni *kni)
 	return 0;
 }
 
+static void *
+va2pa(struct rte_mbuf *m)
+{
+	return (void *)((unsigned long)m -
+			((unsigned long)m->buf_addr -
+			 (unsigned long)m->buf_physaddr));
+}
+
 unsigned
 rte_kni_tx_burst(struct rte_kni *kni, struct rte_mbuf **mbufs, unsigned num)
 {
-	unsigned ret = kni_fifo_put(kni->rx_q, (void **)mbufs, num);
+	void *phy_mbufs[num];
+	unsigned int ret;
+	unsigned int i;
+
+	for (i = 0; i < num; i++)
+		phy_mbufs[i] = va2pa(mbufs[i]);
+
+	ret = kni_fifo_put(kni->rx_q, phy_mbufs, num);
 
 	/* Get mbufs from free_q and then free them */
 	kni_free_mbufs(kni);
@@ -578,6 +607,7 @@ kni_allocate_mbufs(struct rte_kni *kni)
 {
 	int i, ret;
 	struct rte_mbuf *pkts[MAX_MBUF_BURST_NUM];
+	void *phys[MAX_MBUF_BURST_NUM];
 
 	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pool) !=
 			 offsetof(struct rte_kni_mbuf, pool));
@@ -607,13 +637,14 @@ kni_allocate_mbufs(struct rte_kni *kni)
 			RTE_LOG(ERR, KNI, "Out of memory\n");
 			break;
 		}
+		phys[i] = va2pa(pkts[i]);
 	}
 
 	/* No pkt mbuf alocated */
 	if (i <= 0)
 		return;
 
-	ret = kni_fifo_put(kni->alloc_q, (void **)pkts, i);
+	ret = kni_fifo_put(kni->alloc_q, phys, i);
 
 	/* Check if any mbufs not put into alloc_q, and then free them */
 	if (ret >= 0 && ret < i && ret < MAX_MBUF_BURST_NUM) {
-- 
2.7.4

[dpdk-dev] [PATCH 2/3] kni: don't use mbuf_va and mbuf_kva

[Ferruh Yigit]

These variables was for address translation and no more used.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 lib/librte_eal/linuxapp/kni/kni_misc.c | 6 ------
 lib/librte_kni/rte_kni.c               | 2 --
 2 files changed, 8 deletions(-)

diff --git a/lib/librte_eal/linuxapp/kni/kni_misc.c b/lib/librte_eal/linuxapp/kni/kni_misc.c
index 41a331f..1efbcd7 100644
--- a/lib/librte_eal/linuxapp/kni/kni_misc.c
+++ b/lib/librte_eal/linuxapp/kni/kni_misc.c
@@ -458,9 +458,6 @@ kni_ioctl_create(struct net *net,
 	kni->sync_va = dev_info.sync_va;
 	kni->sync_kva = phys_to_virt(dev_info.sync_phys);
 
-	kni->mbuf_kva = phys_to_virt(dev_info.mbuf_phys);
-	kni->mbuf_va = dev_info.mbuf_va;
-
 #ifdef RTE_KNI_VHOST
 	kni->vhost_queue = NULL;
 	kni->vq_status = BE_STOP;
@@ -479,9 +476,6 @@ kni_ioctl_create(struct net *net,
 		(unsigned long long) dev_info.req_phys, kni->req_q);
 	KNI_PRINT("resp_phys:    0x%016llx, resp_q addr:    0x%p\n",
 		(unsigned long long) dev_info.resp_phys, kni->resp_q);
-	KNI_PRINT("mbuf_phys:    0x%016llx, mbuf_kva:       0x%p\n",
-		(unsigned long long) dev_info.mbuf_phys, kni->mbuf_kva);
-	KNI_PRINT("mbuf_va:      0x%p\n", dev_info.mbuf_va);
 	KNI_PRINT("mbuf_size:    %u\n", kni->mbuf_size);
 
 	KNI_DBG("PCI: %02x:%02x.%02x %04x:%04x\n",
diff --git a/lib/librte_kni/rte_kni.c b/lib/librte_kni/rte_kni.c
index 0f7c9e5..21bf9d4 100644
--- a/lib/librte_kni/rte_kni.c
+++ b/lib/librte_kni/rte_kni.c
@@ -418,8 +418,6 @@ rte_kni_alloc(struct rte_mempool *pktmbuf_pool,
 	if (pktmbuf_pool->nb_mem_chunks != 1)
 		goto kni_fail;
 
-	dev_info.mbuf_va = STAILQ_FIRST(&pktmbuf_pool->mem_list)->addr;
-	dev_info.mbuf_phys = STAILQ_FIRST(&pktmbuf_pool->mem_list)->phys_addr;
 	ctx->pktmbuf_pool = pktmbuf_pool;
 	ctx->group_id = conf->group_id;
 	ctx->slot_id = slot->id;
-- 
2.7.4

[dpdk-dev] [PATCH 3/3] kni: remove mempool number of mem chunk check

[Ferruh Yigit]

KNI supports only single mem chunk restriction is no more valid.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 lib/librte_kni/rte_kni.c | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/lib/librte_kni/rte_kni.c b/lib/librte_kni/rte_kni.c
index 21bf9d4..08bd999 100644
--- a/lib/librte_kni/rte_kni.c
+++ b/lib/librte_kni/rte_kni.c
@@ -412,12 +412,6 @@ rte_kni_alloc(struct rte_mempool *pktmbuf_pool,
 	dev_info.sync_va = mz->addr;
 	dev_info.sync_phys = mz->phys_addr;
 
-
-	/* MBUF mempool */
-	/* KNI currently requires to have only one memory chunk */
-	if (pktmbuf_pool->nb_mem_chunks != 1)
-		goto kni_fail;
-
 	ctx->pktmbuf_pool = pktmbuf_pool;
 	ctx->group_id = conf->group_id;
 	ctx->slot_id = slot->id;
-- 
2.7.4

Re: [dpdk-dev] [PATCH 1/3] kni: remove single mempool, single mem_chunk restriction

[Thomas Monjalon]

2016-08-25 12:30, Ferruh Yigit:
> Use mempool buf_addr and buf_physaddr fields for address translation.
> 
> Since each mbuf address calculated separately, the restriction of all
> mbufs should come from a continuous memory restriction is no more valid.
> 
> mbuf related FIFO's content changed, rx_q and alloc_q now carries
> physical address of mbufs. tx_q and free_q content not changed, they
> still carries virtual address of mbufs.
> 
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>

Applied, thanks

The next 2 patches have been squashed with the first one as they are
tightly related.