[dpdk-dev] [PATCH v3 0/5] Xen DOM0 runtime support

[dpdk-dev] [PATCH v3 0/5] Xen DOM0 runtime support

[Stephen Hemminger]

Rebase of patches to current Intel drivers in DPDK 2.2-rc1

Stephen Hemminger (5):
  xen: allow determining DOM0 at runtime
  e1000: use rte_eth_dma_zone_reserve
  ixgbe: use rte_eth_dma_zone_reserve
  i40e: use rte_eth_dma_zone_reserve
  fm10k: use rte_eth_dma_zone_reserve

 drivers/net/e1000/em_rxtx.c                  | 40 ++----------
 drivers/net/e1000/igb_rxtx.c                 | 40 ++----------
 drivers/net/fm10k/fm10k_ethdev.c             | 48 ++------------
 drivers/net/i40e/i40e_ethdev.c               |  9 ---
 drivers/net/i40e/i40e_fdir.c                 |  5 +-
 drivers/net/i40e/i40e_rxtx.c                 | 93 +++++-----------------------
 drivers/net/ixgbe/ixgbe_rxtx.c               | 47 ++------------
 lib/librte_eal/common/include/rte_memory.h   | 30 ++++++++-
 lib/librte_eal/linuxapp/eal/eal_memory.c     |  7 +++
 lib/librte_eal/linuxapp/eal/eal_xen_memory.c |  2 +-
 lib/librte_ether/rte_ethdev.c                | 24 +++++++
 lib/librte_ether/rte_ethdev.h                | 24 +++++++
 lib/librte_mempool/rte_mempool.c             | 48 +++++++++-----
 lib/librte_mempool/rte_mempool.h             |  3 +-
 14 files changed, 162 insertions(+), 258 deletions(-)

-- 
2.1.4

[dpdk-dev] [PATCH 1/5] xen: allow determining DOM0 at runtime

[Stephen Hemminger]

Add memory infrastructure for runtime Xen DOM0 support.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Jijiang Liu <Jijiang.liu@intel.com>

---
 lib/librte_eal/common/include/rte_memory.h   | 30 ++++++++++++++++-
 lib/librte_eal/linuxapp/eal/eal_memory.c     |  7 ++++
 lib/librte_eal/linuxapp/eal/eal_xen_memory.c |  2 +-
 lib/librte_ether/rte_ethdev.c                | 24 ++++++++++++++
 lib/librte_ether/rte_ethdev.h                | 24 ++++++++++++++
 lib/librte_mempool/rte_mempool.c             | 48 ++++++++++++++++++++--------
 lib/librte_mempool/rte_mempool.h             |  3 +-
 7 files changed, 120 insertions(+), 18 deletions(-)

diff --git a/lib/librte_eal/common/include/rte_memory.h b/lib/librte_eal/common/include/rte_memory.h
index 1bed415..067be10 100644
--- a/lib/librte_eal/common/include/rte_memory.h
+++ b/lib/librte_eal/common/include/rte_memory.h
@@ -52,6 +52,8 @@
 extern "C" {
 #endif
 
+#include <rte_common.h>
+
 enum rte_page_sizes {
 	RTE_PGSIZE_4K    = 1ULL << 12,
 	RTE_PGSIZE_64K   = 1ULL << 16,
@@ -180,6 +182,13 @@ unsigned rte_memory_get_nchannel(void);
 unsigned rte_memory_get_nrank(void);
 
 #ifdef RTE_LIBRTE_XEN_DOM0
+
+/**< Internal use only - should DOM0 memory mapping be used */
+extern int is_xen_dom0_supported(void);
+
+/**< Internal use only - phys to virt mapping for xen */
+phys_addr_t rte_xen_mem_phy2mch(uint32_t, const phys_addr_t);
+
 /**
  * Return the physical address of elt, which is an element of the pool mp.
  *
@@ -191,7 +200,14 @@ unsigned rte_memory_get_nrank(void);
  * @return
  *   The physical address or error.
  */
-phys_addr_t rte_mem_phy2mch(uint32_t memseg_id, const phys_addr_t phy_addr);
+static inline phys_addr_t
+rte_mem_phy2mch(uint32_t memseg_id, const phys_addr_t phy_addr)
+{
+	if (is_xen_dom0_supported())
+		return rte_xen_mem_phy2mch(memseg_id, phy_addr);
+	else
+		return phy_addr;
+}
 
 /**
  * Memory init for supporting application running on Xen domain0.
@@ -214,7 +230,19 @@ int rte_xen_dom0_memory_init(void);
  *       negative: error
  */
 int rte_xen_dom0_memory_attach(void);
+#else
+static inline int is_xen_dom0_supported(void)
+{
+	return 0;
+}
+
+static inline phys_addr_t
+rte_mem_phy2mch(uint32_t memseg_id __rte_unused, const phys_addr_t phy_addr)
+{
+	return phy_addr;
+}
 #endif
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c
index 657d19f..0de75cd 100644
--- a/lib/librte_eal/linuxapp/eal/eal_memory.c
+++ b/lib/librte_eal/linuxapp/eal/eal_memory.c
@@ -97,6 +97,13 @@
 #include "eal_filesystem.h"
 #include "eal_hugepages.h"
 
+#ifdef RTE_LIBRTE_XEN_DOM0
+int is_xen_dom0_supported(void)
+{
+	return internal_config.xen_dom0_support;
+}
+#endif
+
 /**
  * @file
  * Huge page mapping under linux
diff --git a/lib/librte_eal/linuxapp/eal/eal_xen_memory.c b/lib/librte_eal/linuxapp/eal/eal_xen_memory.c
index d228a9d..7fd9e83 100644
--- a/lib/librte_eal/linuxapp/eal/eal_xen_memory.c
+++ b/lib/librte_eal/linuxapp/eal/eal_xen_memory.c
@@ -156,7 +156,7 @@ get_xen_memory_size(void)
  * Based on physical address to caculate MFN in Xen Dom0.
  */
 phys_addr_t
-rte_mem_phy2mch(uint32_t memseg_id, const phys_addr_t phy_addr)
+rte_xen_mem_phy2mch(uint32_t memseg_id, const phys_addr_t phy_addr)
 {
 	int mfn_id;
 	uint64_t mfn, mfn_offset;
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index e0e1dca..756e894 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -2699,6 +2699,30 @@ rte_eth_dev_rx_intr_ctl(uint8_t port_id, int epfd, int op, void *data)
 
 	return 0;
 }
+ 
+const struct rte_memzone *
+rte_eth_dma_zone_reserve(const struct rte_eth_dev *dev, const char *ring_name,
+			 uint16_t queue_id, size_t size, unsigned align,
+			 int socket_id)
+{
+	char z_name[RTE_MEMZONE_NAMESIZE];
+	const struct rte_memzone *mz;
+
+	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
+		 dev->driver->pci_drv.name, ring_name,
+		 dev->data->port_id, queue_id);
+
+	mz = rte_memzone_lookup(z_name);
+	if (mz)
+		return mz;
+
+	if (is_xen_dom0_supported())
+		return rte_memzone_reserve_bounded(z_name, size, socket_id,
+						   0, align, RTE_PGSIZE_2M);
+	else
+		return rte_memzone_reserve_aligned(z_name, size, socket_id,
+						   0, align);
+}
 
 int
 rte_eth_dev_rx_intr_ctl_q(uint8_t port_id, uint16_t queue_id,
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 48a540d..785f482 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -3768,6 +3768,30 @@ extern int rte_eth_timesync_read_tx_timestamp(uint8_t port_id,
 extern void rte_eth_copy_pci_info(struct rte_eth_dev *eth_dev, struct rte_pci_device *pci_dev);
 
 
+/**
+ * Create memzone for HW rings.
+ * malloc can't be used as the physical address is needed.
+ * If the memzone is already created, then this function returns a ptr
+ * to the old one.
+ *
+ * @param eth_dev
+ *   The *eth_dev* pointer is the address of the *rte_eth_dev* structure
+ * @param name
+ *   The name of the memory zone
+ * @param queue_id
+ *   The index of the queue to add to name
+ * @param size
+ *   The sizeof of the memory area
+ * @param align
+ *   Alignment for resulting memzone. Must be a power of 2.
+ * @param socket_id
+ *   The *socket_id* argument is the socket identifier in case of NUMA.
+ */
+const struct rte_memzone *
+rte_eth_dma_zone_reserve(const struct rte_eth_dev *eth_dev, const char *name,
+			 uint16_t queue_id, size_t size,
+			 unsigned align, int socket_id);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index e57cbbd..f53076a 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -375,6 +375,26 @@ rte_mempool_xmem_usage(void *vaddr, uint32_t elt_num, size_t elt_sz,
 	return usz;
 }
 
+#ifndef RTE_LIBRTE_XEN_DOM0
+/* stub if DOM0 support not configured */
+struct rte_mempool *
+rte_dom0_mempool_create(const char *name __rte_unused,
+			unsigned n __rte_unused,
+			unsigned elt_size __rte_unused,
+			unsigned cache_size __rte_unused,
+			unsigned private_data_size __rte_unused,
+			rte_mempool_ctor_t *mp_init __rte_unused,
+			void *mp_init_arg __rte_unused,
+			rte_mempool_obj_ctor_t *obj_init __rte_unused,
+			void *obj_init_arg __rte_unused,
+			int socket_id __rte_unused,
+			unsigned flags __rte_unused)
+{
+	rte_errno = EINVAL;
+	return NULL;
+}
+#endif
+
 /* create the mempool */
 struct rte_mempool *
 rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
@@ -383,20 +403,20 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
 		   rte_mempool_obj_ctor_t *obj_init, void *obj_init_arg,
 		   int socket_id, unsigned flags)
 {
-#ifdef RTE_LIBRTE_XEN_DOM0
-	return rte_dom0_mempool_create(name, n, elt_size,
-		cache_size, private_data_size,
-		mp_init, mp_init_arg,
-		obj_init, obj_init_arg,
-		socket_id, flags);
-#else
-	return rte_mempool_xmem_create(name, n, elt_size,
-		cache_size, private_data_size,
-		mp_init, mp_init_arg,
-		obj_init, obj_init_arg,
-		socket_id, flags,
-		NULL, NULL, MEMPOOL_PG_NUM_DEFAULT, MEMPOOL_PG_SHIFT_MAX);
-#endif
+	if (is_xen_dom0_supported())
+		return rte_dom0_mempool_create(name, n, elt_size,
+					       cache_size, private_data_size,
+					       mp_init, mp_init_arg,
+					       obj_init, obj_init_arg,
+					       socket_id, flags);
+	else
+		return rte_mempool_xmem_create(name, n, elt_size,
+					       cache_size, private_data_size,
+					       mp_init, mp_init_arg,
+					       obj_init, obj_init_arg,
+					       socket_id, flags,
+					       NULL, NULL, MEMPOOL_PG_NUM_DEFAULT,
+					       MEMPOOL_PG_SHIFT_MAX);
 }
 
 /*
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 8abeca9..6e2390a 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -640,7 +640,6 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 		int socket_id, unsigned flags, void *vaddr,
 		const phys_addr_t paddr[], uint32_t pg_num, uint32_t pg_shift);
 
-#ifdef RTE_LIBRTE_XEN_DOM0
 /**
  * Create a new mempool named *name* in memory on Xen Dom0.
  *
@@ -728,7 +727,7 @@ rte_dom0_mempool_create(const char *name, unsigned n, unsigned elt_size,
 		rte_mempool_ctor_t *mp_init, void *mp_init_arg,
 		rte_mempool_obj_ctor_t *obj_init, void *obj_init_arg,
 		int socket_id, unsigned flags);
-#endif
+
 
 /**
  * Dump the status of the mempool to the console.
-- 
2.1.4

[dpdk-dev] [PATCH 2/5] e1000: use rte_eth_dma_zone_reserve

[Stephen Hemminger]

Use common code to handle Xen support at runtime.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/e1000/em_rxtx.c  | 40 ++++++----------------------------------
 drivers/net/e1000/igb_rxtx.c | 40 ++++++----------------------------------
 2 files changed, 12 insertions(+), 68 deletions(-)

diff --git a/drivers/net/e1000/em_rxtx.c b/drivers/net/e1000/em_rxtx.c
index 03e1bc2..d8fb252 100644
--- a/drivers/net/e1000/em_rxtx.c
+++ b/drivers/net/e1000/em_rxtx.c
@@ -1084,28 +1084,6 @@ eth_em_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 #define	EM_MAX_BUF_SIZE     16384
 #define EM_RCTL_FLXBUF_STEP 1024
 
-static const struct rte_memzone *
-ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
-		uint16_t queue_id, uint32_t ring_size, int socket_id)
-{
-	const struct rte_memzone *mz;
-	char z_name[RTE_MEMZONE_NAMESIZE];
-
-	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
-		dev->driver->pci_drv.name, ring_name, dev->data->port_id,
-		queue_id);
-
-	if ((mz = rte_memzone_lookup(z_name)) != 0)
-		return (mz);
-
-#ifdef RTE_LIBRTE_XEN_DOM0
-	return rte_memzone_reserve_bounded(z_name, ring_size,
-			socket_id, 0, RTE_CACHE_LINE_SIZE, RTE_PGSIZE_2M);
-#else
-	return rte_memzone_reserve(z_name, ring_size, socket_id, 0);
-#endif
-}
-
 static void
 em_tx_queue_release_mbufs(struct em_tx_queue *txq)
 {
@@ -1253,8 +1231,9 @@ eth_em_tx_queue_setup(struct rte_eth_dev *dev,
 	 * resizing in later calls to the queue setup function.
 	 */
 	tsize = sizeof(txq->tx_ring[0]) * E1000_MAX_RING_DESC;
-	if ((tz = ring_dma_zone_reserve(dev, "tx_ring", queue_idx, tsize,
-			socket_id)) == NULL)
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx, tsize,
+				      RTE_CACHE_LINE_SIZE, socket_id);
+	if (tz == NULL)
 		return (-ENOMEM);
 
 	/* Allocate the tx queue data structure. */
@@ -1280,11 +1259,7 @@ eth_em_tx_queue_setup(struct rte_eth_dev *dev,
 	txq->port_id = dev->data->port_id;
 
 	txq->tdt_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_TDT(queue_idx));
-#ifndef RTE_LIBRTE_XEN_DOM0
-	txq->tx_ring_phys_addr = (uint64_t) tz->phys_addr;
-#else
 	txq->tx_ring_phys_addr = rte_mem_phy2mch(tz->memseg_id, tz->phys_addr);
-#endif
 	txq->tx_ring = (struct e1000_data_desc *) tz->addr;
 
 	PMD_INIT_LOG(DEBUG, "sw_ring=%p hw_ring=%p dma_addr=0x%"PRIx64,
@@ -1380,8 +1355,9 @@ eth_em_rx_queue_setup(struct rte_eth_dev *dev,
 
 	/* Allocate RX ring for max possible mumber of hardware descriptors. */
 	rsize = sizeof(rxq->rx_ring[0]) * E1000_MAX_RING_DESC;
-	if ((rz = ring_dma_zone_reserve(dev, "rx_ring", queue_idx, rsize,
-			socket_id)) == NULL)
+	rz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx, rsize,
+				      RTE_CACHE_LINE_SIZE, socket_id);
+	if (rz == NULL)
 		return (-ENOMEM);
 
 	/* Allocate the RX queue data structure. */
@@ -1410,11 +1386,7 @@ eth_em_rx_queue_setup(struct rte_eth_dev *dev,
 
 	rxq->rdt_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_RDT(queue_idx));
 	rxq->rdh_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_RDH(queue_idx));
-#ifndef RTE_LIBRTE_XEN_DOM0
-	rxq->rx_ring_phys_addr = (uint64_t) rz->phys_addr;
-#else
 	rxq->rx_ring_phys_addr = rte_mem_phy2mch(rz->memseg_id, rz->phys_addr);
-#endif
 	rxq->rx_ring = (struct e1000_rx_desc *) rz->addr;
 
 	PMD_INIT_LOG(DEBUG, "sw_ring=%p hw_ring=%p dma_addr=0x%"PRIx64,
diff --git a/drivers/net/e1000/igb_rxtx.c b/drivers/net/e1000/igb_rxtx.c
index 384e4f1..996e7da 100644
--- a/drivers/net/e1000/igb_rxtx.c
+++ b/drivers/net/e1000/igb_rxtx.c
@@ -1209,28 +1209,6 @@ eth_igb_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
  * desscriptors should meet the following condition:
  *      (num_ring_desc * sizeof(struct e1000_rx/tx_desc)) % 128 == 0
  */
-static const struct rte_memzone *
-ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
-		      uint16_t queue_id, uint32_t ring_size, int socket_id)
-{
-	char z_name[RTE_MEMZONE_NAMESIZE];
-	const struct rte_memzone *mz;
-
-	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
-			dev->driver->pci_drv.name, ring_name,
-				dev->data->port_id, queue_id);
-	mz = rte_memzone_lookup(z_name);
-	if (mz)
-		return mz;
-
-#ifdef RTE_LIBRTE_XEN_DOM0
-	return rte_memzone_reserve_bounded(z_name, ring_size,
-			socket_id, 0, E1000_ALIGN, RTE_PGSIZE_2M);
-#else
-	return rte_memzone_reserve_aligned(z_name, ring_size,
-			socket_id, 0, E1000_ALIGN);
-#endif
-}
 
 static void
 igb_tx_queue_release_mbufs(struct igb_tx_queue *txq)
@@ -1365,8 +1343,8 @@ eth_igb_tx_queue_setup(struct rte_eth_dev *dev,
 	 * resizing in later calls to the queue setup function.
 	 */
 	size = sizeof(union e1000_adv_tx_desc) * E1000_MAX_RING_DESC;
-	tz = ring_dma_zone_reserve(dev, "tx_ring", queue_idx,
-					size, socket_id);
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx, size,
+				      E1000_ALIGN, socket_id);
 	if (tz == NULL) {
 		igb_tx_queue_release(txq);
 		return (-ENOMEM);
@@ -1384,12 +1362,9 @@ eth_igb_tx_queue_setup(struct rte_eth_dev *dev,
 	txq->port_id = dev->data->port_id;
 
 	txq->tdt_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_TDT(txq->reg_idx));
-#ifndef RTE_LIBRTE_XEN_DOM0
-	txq->tx_ring_phys_addr = (uint64_t) tz->phys_addr;
-#else
 	txq->tx_ring_phys_addr = rte_mem_phy2mch(tz->memseg_id, tz->phys_addr);
-#endif
-	 txq->tx_ring = (union e1000_adv_tx_desc *) tz->addr;
+
+	txq->tx_ring = (union e1000_adv_tx_desc *) tz->addr;
 	/* Allocate software ring */
 	txq->sw_ring = rte_zmalloc("txq->sw_ring",
 				   sizeof(struct igb_tx_entry) * nb_desc,
@@ -1514,18 +1489,15 @@ eth_igb_rx_queue_setup(struct rte_eth_dev *dev,
 	 *  resizing in later calls to the queue setup function.
 	 */
 	size = sizeof(union e1000_adv_rx_desc) * E1000_MAX_RING_DESC;
-	rz = ring_dma_zone_reserve(dev, "rx_ring", queue_idx, size, socket_id);
+	rz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx, size,
+				      E1000_ALIGN, socket_id);
 	if (rz == NULL) {
 		igb_rx_queue_release(rxq);
 		return (-ENOMEM);
 	}
 	rxq->rdt_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_RDT(rxq->reg_idx));
 	rxq->rdh_reg_addr = E1000_PCI_REG_ADDR(hw, E1000_RDH(rxq->reg_idx));
-#ifndef RTE_LIBRTE_XEN_DOM0
-	rxq->rx_ring_phys_addr = (uint64_t) rz->phys_addr;
-#else
 	rxq->rx_ring_phys_addr = rte_mem_phy2mch(rz->memseg_id, rz->phys_addr);
-#endif
 	rxq->rx_ring = (union e1000_adv_rx_desc *) rz->addr;
 
 	/* Allocate software ring. */
-- 
2.1.4

[dpdk-dev] [PATCH 3/5] ixgbe: use rte_eth_dma_zone_reserve

[Stephen Hemminger]

Adapt DMA memory for Xen at runtime.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/ixgbe/ixgbe_rxtx.c | 47 ++++++------------------------------------
 1 file changed, 6 insertions(+), 41 deletions(-)

diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net/ixgbe/ixgbe_rxtx.c
index 5561195..62b0737 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx.c
+++ b/drivers/net/ixgbe/ixgbe_rxtx.c
@@ -1819,35 +1819,6 @@ ixgbe_recv_pkts_lro_bulk_alloc(void *rx_queue, struct rte_mbuf **rx_pkts,
  *
  **********************************************************************/
 
-/*
- * Create memzone for HW rings. malloc can't be used as the physical address is
- * needed. If the memzone is already created, then this function returns a ptr
- * to the old one.
- */
-static const struct rte_memzone * __attribute__((cold))
-ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
-		      uint16_t queue_id, uint32_t ring_size, int socket_id)
-{
-	char z_name[RTE_MEMZONE_NAMESIZE];
-	const struct rte_memzone *mz;
-
-	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
-			dev->driver->pci_drv.name, ring_name,
-			dev->data->port_id, queue_id);
-
-	mz = rte_memzone_lookup(z_name);
-	if (mz)
-		return mz;
-
-#ifdef RTE_LIBRTE_XEN_DOM0
-	return rte_memzone_reserve_bounded(z_name, ring_size,
-		socket_id, 0, IXGBE_ALIGN, RTE_PGSIZE_2M);
-#else
-	return rte_memzone_reserve_aligned(z_name, ring_size,
-		socket_id, 0, IXGBE_ALIGN);
-#endif
-}
-
 static void __attribute__((cold))
 ixgbe_tx_queue_release_mbufs(struct ixgbe_tx_queue *txq)
 {
@@ -2082,9 +2053,9 @@ ixgbe_dev_tx_queue_setup(struct rte_eth_dev *dev,
 	 * handle the maximum ring size is allocated in order to allow for
 	 * resizing in later calls to the queue setup function.
 	 */
-	tz = ring_dma_zone_reserve(dev, "tx_ring", queue_idx,
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
 			sizeof(union ixgbe_adv_tx_desc) * IXGBE_MAX_RING_DESC,
-			socket_id);
+			IXGBE_ALIGN, socket_id);
 	if (tz == NULL) {
 		ixgbe_tx_queue_release(txq);
 		return (-ENOMEM);
@@ -2114,11 +2085,8 @@ ixgbe_dev_tx_queue_setup(struct rte_eth_dev *dev,
 		txq->tdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_VFTDT(queue_idx));
 	else
 		txq->tdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_TDT(txq->reg_idx));
-#ifndef	RTE_LIBRTE_XEN_DOM0
-	txq->tx_ring_phys_addr = (uint64_t) tz->phys_addr;
-#else
+
 	txq->tx_ring_phys_addr = rte_mem_phy2mch(tz->memseg_id, tz->phys_addr);
-#endif
 	txq->tx_ring = (union ixgbe_adv_tx_desc *) tz->addr;
 
 	/* Allocate software ring */
@@ -2388,8 +2356,8 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
 	 * handle the maximum ring size is allocated in order to allow for
 	 * resizing in later calls to the queue setup function.
 	 */
-	rz = ring_dma_zone_reserve(dev, "rx_ring", queue_idx,
-				   RX_RING_SZ, socket_id);
+	rz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
+				      RX_RING_SZ, IXGBE_ALIGN, socket_id);
 	if (rz == NULL) {
 		ixgbe_rx_queue_release(rxq);
 		return (-ENOMEM);
@@ -2418,11 +2386,8 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
 		rxq->rdh_reg_addr =
 			IXGBE_PCI_REG_ADDR(hw, IXGBE_RDH(rxq->reg_idx));
 	}
-#ifndef RTE_LIBRTE_XEN_DOM0
-	rxq->rx_ring_phys_addr = (uint64_t) rz->phys_addr;
-#else
+
 	rxq->rx_ring_phys_addr = rte_mem_phy2mch(rz->memseg_id, rz->phys_addr);
-#endif
 	rxq->rx_ring = (union ixgbe_adv_rx_desc *) rz->addr;
 
 	/*
-- 
2.1.4

[dpdk-dev] [PATCH 4/5] i40e: use rte_eth_dma_zone_reserve

[Stephen Hemminger]

Handle Xen support at runtime.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/i40e/i40e_ethdev.c |  9 ----
 drivers/net/i40e/i40e_fdir.c   |  5 +--
 drivers/net/i40e/i40e_rxtx.c   | 93 ++++++++----------------------------------
 3 files changed, 18 insertions(+), 89 deletions(-)

diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index ddf3d38..c8a0b43 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -2919,24 +2919,15 @@ i40e_allocate_dma_mem_d(__attribute__((unused)) struct i40e_hw *hw,
 
 	id++;
 	snprintf(z_name, sizeof(z_name), "i40e_dma_%"PRIu64, id);
-#ifdef RTE_LIBRTE_XEN_DOM0
 	mz = rte_memzone_reserve_bounded(z_name, size, SOCKET_ID_ANY, 0,
 					 alignment, RTE_PGSIZE_2M);
-#else
-	mz = rte_memzone_reserve_aligned(z_name, size, SOCKET_ID_ANY, 0,
-					 alignment);
-#endif
 	if (!mz)
 		return I40E_ERR_NO_MEMORY;
 
 	mem->id = id;
 	mem->size = size;
 	mem->va = mz->addr;
-#ifdef RTE_LIBRTE_XEN_DOM0
 	mem->pa = rte_mem_phy2mch(mz->memseg_id, mz->phys_addr);
-#else
-	mem->pa = mz->phys_addr;
-#endif
 
 	return I40E_SUCCESS;
 }
diff --git a/drivers/net/i40e/i40e_fdir.c b/drivers/net/i40e/i40e_fdir.c
index 7b1aa6e..6ff6702 100644
--- a/drivers/net/i40e/i40e_fdir.c
+++ b/drivers/net/i40e/i40e_fdir.c
@@ -275,11 +275,8 @@ i40e_fdir_setup(struct i40e_pf *pf)
 		goto fail_mem;
 	}
 	pf->fdir.prg_pkt = mz->addr;
-#ifdef RTE_LIBRTE_XEN_DOM0
 	pf->fdir.dma_addr = rte_mem_phy2mch(mz->memseg_id, mz->phys_addr);
-#else
-	pf->fdir.dma_addr = (uint64_t)mz->phys_addr;
-#endif
+
 	pf->fdir.match_counter_index = I40E_COUNTER_INDEX_FDIR(hw->pf_id);
 	PMD_DRV_LOG(INFO, "FDIR setup successfully, with programming queue %u.",
 		    vsi->base_queue);
diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index b60c356..13abd67 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -84,12 +84,6 @@
 #define RTE_MBUF_DATA_DMA_ADDR(mb) \
 	((uint64_t)((mb)->buf_physaddr + (mb)->data_off))
 
-static const struct rte_memzone *
-i40e_ring_dma_zone_reserve(struct rte_eth_dev *dev,
-			   const char *ring_name,
-			   uint16_t queue_id,
-			   uint32_t ring_size,
-			   int socket_id);
 static uint16_t i40e_xmit_pkts_simple(void *tx_queue,
 				      struct rte_mbuf **tx_pkts,
 				      uint16_t nb_pkts);
@@ -2175,11 +2169,8 @@ i40e_dev_rx_queue_setup(struct rte_eth_dev *dev,
 	/* Allocate the maximun number of RX ring hardware descriptor. */
 	ring_size = sizeof(union i40e_rx_desc) * I40E_MAX_RING_DESC;
 	ring_size = RTE_ALIGN(ring_size, I40E_DMA_MEM_ALIGN);
-	rz = i40e_ring_dma_zone_reserve(dev,
-					"rx_ring",
-					queue_idx,
-					ring_size,
-					socket_id);
+	rz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
+			      ring_size, I40E_RING_BASE_ALIGN, socket_id);
 	if (!rz) {
 		i40e_dev_rx_queue_release(rxq);
 		PMD_DRV_LOG(ERR, "Failed to reserve DMA memory for RX");
@@ -2189,12 +2180,7 @@ i40e_dev_rx_queue_setup(struct rte_eth_dev *dev,
 	/* Zero all the descriptors in the ring. */
 	memset(rz->addr, 0, ring_size);
 
-#ifdef RTE_LIBRTE_XEN_DOM0
 	rxq->rx_ring_phys_addr = rte_mem_phy2mch(rz->memseg_id, rz->phys_addr);
-#else
-	rxq->rx_ring_phys_addr = (uint64_t)rz->phys_addr;
-#endif
-
 	rxq->rx_ring = (union i40e_rx_desc *)rz->addr;
 
 #ifdef RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC
@@ -2457,11 +2443,8 @@ i40e_dev_tx_queue_setup(struct rte_eth_dev *dev,
 	/* Allocate TX hardware ring descriptors. */
 	ring_size = sizeof(struct i40e_tx_desc) * I40E_MAX_RING_DESC;
 	ring_size = RTE_ALIGN(ring_size, I40E_DMA_MEM_ALIGN);
-	tz = i40e_ring_dma_zone_reserve(dev,
-					"tx_ring",
-					queue_idx,
-					ring_size,
-					socket_id);
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
+			      ring_size, I40E_RING_BASE_ALIGN, socket_id);
 	if (!tz) {
 		i40e_dev_tx_queue_release(txq);
 		PMD_DRV_LOG(ERR, "Failed to reserve DMA memory for TX");
@@ -2486,11 +2469,7 @@ i40e_dev_tx_queue_setup(struct rte_eth_dev *dev,
 	txq->vsi = vsi;
 	txq->tx_deferred_start = tx_conf->tx_deferred_start;
 
-#ifdef RTE_LIBRTE_XEN_DOM0
 	txq->tx_ring_phys_addr = rte_mem_phy2mch(tz->memseg_id, tz->phys_addr);
-#else
-	txq->tx_ring_phys_addr = (uint64_t)tz->phys_addr;
-#endif
 	txq->tx_ring = (struct i40e_tx_desc *)tz->addr;
 
 	/* Allocate software ring */
@@ -2543,47 +2522,21 @@ i40e_dev_tx_queue_release(void *txq)
 	rte_free(q);
 }
 
-static const struct rte_memzone *
-i40e_ring_dma_zone_reserve(struct rte_eth_dev *dev,
-			   const char *ring_name,
-			   uint16_t queue_id,
-			   uint32_t ring_size,
-			   int socket_id)
-{
-	char z_name[RTE_MEMZONE_NAMESIZE];
-	const struct rte_memzone *mz;
-
-	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
-			dev->driver->pci_drv.name, ring_name,
-				dev->data->port_id, queue_id);
-	mz = rte_memzone_lookup(z_name);
-	if (mz)
-		return mz;
-
-#ifdef RTE_LIBRTE_XEN_DOM0
-	return rte_memzone_reserve_bounded(z_name, ring_size,
-		socket_id, 0, I40E_RING_BASE_ALIGN, RTE_PGSIZE_2M);
-#else
-	return rte_memzone_reserve_aligned(z_name, ring_size,
-				socket_id, 0, I40E_RING_BASE_ALIGN);
-#endif
-}
-
 const struct rte_memzone *
 i40e_memzone_reserve(const char *name, uint32_t len, int socket_id)
 {
-	const struct rte_memzone *mz = NULL;
+	const struct rte_memzone *mz;
 
 	mz = rte_memzone_lookup(name);
 	if (mz)
 		return mz;
-#ifdef RTE_LIBRTE_XEN_DOM0
-	mz = rte_memzone_reserve_bounded(name, len,
-		socket_id, 0, I40E_RING_BASE_ALIGN, RTE_PGSIZE_2M);
-#else
-	mz = rte_memzone_reserve_aligned(name, len,
+
+	if (is_xen_dom0_supported())
+		mz = rte_memzone_reserve_bounded(name, len,
+				socket_id, 0, I40E_RING_BASE_ALIGN, RTE_PGSIZE_2M);
+	else
+		mz = rte_memzone_reserve_aligned(name, len,
 				socket_id, 0, I40E_RING_BASE_ALIGN);
-#endif
 	return mz;
 }
 
@@ -3003,11 +2956,9 @@ i40e_fdir_setup_tx_resources(struct i40e_pf *pf)
 	ring_size = sizeof(struct i40e_tx_desc) * I40E_FDIR_NUM_TX_DESC;
 	ring_size = RTE_ALIGN(ring_size, I40E_DMA_MEM_ALIGN);
 
-	tz = i40e_ring_dma_zone_reserve(dev,
-					"fdir_tx_ring",
-					I40E_FDIR_QUEUE_ID,
-					ring_size,
-					SOCKET_ID_ANY);
+	tz = rte_eth_dma_zone_reserve(dev, "fdir_tx_ring",
+				      I40E_FDIR_QUEUE_ID, ring_size,
+				      I40E_RING_BASE_ALIGN, SOCKET_ID_ANY);
 	if (!tz) {
 		i40e_dev_tx_queue_release(txq);
 		PMD_DRV_LOG(ERR, "Failed to reserve DMA memory for TX.");
@@ -3019,11 +2970,7 @@ i40e_fdir_setup_tx_resources(struct i40e_pf *pf)
 	txq->reg_idx = pf->fdir.fdir_vsi->base_queue;
 	txq->vsi = pf->fdir.fdir_vsi;
 
-#ifdef RTE_LIBRTE_XEN_DOM0
 	txq->tx_ring_phys_addr = rte_mem_phy2mch(tz->memseg_id, tz->phys_addr);
-#else
-	txq->tx_ring_phys_addr = (uint64_t)tz->phys_addr;
-#endif
 	txq->tx_ring = (struct i40e_tx_desc *)tz->addr;
 	/*
 	 * don't need to allocate software ring and reset for the fdir
@@ -3063,11 +3010,9 @@ i40e_fdir_setup_rx_resources(struct i40e_pf *pf)
 	ring_size = sizeof(union i40e_rx_desc) * I40E_FDIR_NUM_RX_DESC;
 	ring_size = RTE_ALIGN(ring_size, I40E_DMA_MEM_ALIGN);
 
-	rz = i40e_ring_dma_zone_reserve(dev,
-					"fdir_rx_ring",
-					I40E_FDIR_QUEUE_ID,
-					ring_size,
-					SOCKET_ID_ANY);
+	rz = rte_eth_dma_zone_reserve(dev, "fdir_rx_ring",
+				      I40E_FDIR_QUEUE_ID, ring_size,
+				      I40E_RING_BASE_ALIGN, SOCKET_ID_ANY);
 	if (!rz) {
 		i40e_dev_rx_queue_release(rxq);
 		PMD_DRV_LOG(ERR, "Failed to reserve DMA memory for RX.");
@@ -3079,11 +3024,7 @@ i40e_fdir_setup_rx_resources(struct i40e_pf *pf)
 	rxq->reg_idx = pf->fdir.fdir_vsi->base_queue;
 	rxq->vsi = pf->fdir.fdir_vsi;
 
-#ifdef RTE_LIBRTE_XEN_DOM0
 	rxq->rx_ring_phys_addr = rte_mem_phy2mch(rz->memseg_id, rz->phys_addr);
-#else
-	rxq->rx_ring_phys_addr = (uint64_t)rz->phys_addr;
-#endif
 	rxq->rx_ring = (union i40e_rx_desc *)rz->addr;
 
 	/*
-- 
2.1.4

[dpdk-dev] [PATCH 5/5] fm10k: use rte_eth_dma_zone_reserve

[Stephen Hemminger]

Adapt to Xen at runtime.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/fm10k/fm10k_ethdev.c | 48 +++++-----------------------------------
 1 file changed, 6 insertions(+), 42 deletions(-)

diff --git a/drivers/net/fm10k/fm10k_ethdev.c b/drivers/net/fm10k/fm10k_ethdev.c
index cf7ada7..441f713 100644
--- a/drivers/net/fm10k/fm10k_ethdev.c
+++ b/drivers/net/fm10k/fm10k_ethdev.c
@@ -1492,34 +1492,6 @@ check_nb_desc(uint16_t min, uint16_t max, uint16_t mult, uint16_t request)
 		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)
@@ -1667,9 +1639,9 @@ fm10k_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
 	 * 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);
+	mz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_id,
+				      FM10K_MAX_RX_RING_SZ, FM10K_ALIGN_RX_DESC,
+				      socket_id);
 	if (mz == NULL) {
 		PMD_INIT_LOG(ERR, "Cannot allocate hardware ring");
 		rte_free(q->sw_ring);
@@ -1677,11 +1649,7 @@ fm10k_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
 		return (-ENOMEM);
 	}
 	q->hw_ring = mz->addr;
-#ifdef RTE_LIBRTE_XEN_DOM0
 	q->hw_ring_phys_addr = rte_mem_phy2mch(mz->memseg_id, mz->phys_addr);
-#else
-	q->hw_ring_phys_addr = mz->phys_addr;
-#endif
 
 	/* Check if number of descs satisfied Vector requirement */
 	if (!rte_is_power_of_2(nb_desc)) {
@@ -1831,9 +1799,9 @@ fm10k_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
 	 * 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, "tx_ring",
-				dev->data->port_id, queue_id, socket_id,
-				FM10K_MAX_TX_RING_SZ, FM10K_ALIGN_TX_DESC);
+	mz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_id,
+				      FM10K_MAX_TX_RING_SZ, FM10K_ALIGN_TX_DESC,
+				      socket_id);
 	if (mz == NULL) {
 		PMD_INIT_LOG(ERR, "Cannot allocate hardware ring");
 		rte_free(q->sw_ring);
@@ -1841,11 +1809,7 @@ fm10k_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_id,
 		return (-ENOMEM);
 	}
 	q->hw_ring = mz->addr;
-#ifdef RTE_LIBRTE_XEN_DOM0
 	q->hw_ring_phys_addr = rte_mem_phy2mch(mz->memseg_id, mz->phys_addr);
-#else
-	q->hw_ring_phys_addr = mz->phys_addr;
-#endif
 
 	/*
 	 * allocate memory for the RS bit tracker. Enough slots to hold the
-- 
2.1.4

Re: [dpdk-dev] [PATCH v3 0/5] Xen DOM0 runtime support

[Thomas Monjalon]

2015-11-05 16:09, Stephen Hemminger:
> Rebase of patches to current Intel drivers in DPDK 2.2-rc1
> 
> Stephen Hemminger (5):
>   xen: allow determining DOM0 at runtime
>   e1000: use rte_eth_dma_zone_reserve
>   ixgbe: use rte_eth_dma_zone_reserve
>   i40e: use rte_eth_dma_zone_reserve
>   fm10k: use rte_eth_dma_zone_reserve

Applied, thanks

I think we need to check "git grep LIBRTE_XEN_DOM0" and see what can be
improved to better reduce the number of ifdefs for Xen.