DPDK patches and discussions
 help / color / mirror / Atom feed
From: Xueming Li <xuemingl@mellanox.com>
To: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Cc: Xueming Li <xuemingl@mellanox.com>,
	Shahaf Shuler <shahafs@mellanox.com>,
	 dev@dpdk.org
Subject: [dpdk-dev] [PATCH] net/mlx5: remmap UAR address for multiple process
Date: Fri, 19 Jan 2018 23:08:54 +0800	[thread overview]
Message-ID: <20180119150854.89828-1-xuemingl@mellanox.com> (raw)

UAR(doorbell) is hw resources that have to be same address between
primary and secondary process, failed to mmap UAR will make TX packets
invisible to HW.
Today, UAR address returned from verbs api is mixed in heap and loaded
library address space, prone to be occupied in secondary process.
This patch reserves a dedicate UAR address space, both primary and
secondary process re-mmap UAR pages into this space.
Below is a brief picture of dpdk app address space allocation:
	Before			This patch
	------			----------
	[stack]			[stack]
	[.so, uar, heap]	[.so, heap]
	[(empty)]		[(empty)]
	[hugepage]		[hugepage]
	[? others]		[? others]
	[(empty)]		[(empty)]
				[uar]
				[(empty)]
To minimize conflicts, UAR address space comes after hugepage space with
an offset to skip potential usage from other drivers.

Once UAR space reserved successfully, UAR pages are re-mmapped into new
area to keep UAR address aligned between primary and secondary process.

Signed-off-by: Xueming Li <xuemingl@mellanox.com>
---
 drivers/net/mlx5/mlx5.c         | 107 ++++++++++++++++++++++++++++++++++++++++
 drivers/net/mlx5/mlx5.h         |   1 +
 drivers/net/mlx5/mlx5_defs.h    |  10 ++++
 drivers/net/mlx5/mlx5_rxtx.h    |   3 +-
 drivers/net/mlx5/mlx5_trigger.c |   7 ++-
 drivers/net/mlx5/mlx5_txq.c     |  51 +++++++++++++------
 6 files changed, 163 insertions(+), 16 deletions(-)

diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index fc2d59fee..1539ef608 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -39,6 +39,7 @@
 #include <stdlib.h>
 #include <errno.h>
 #include <net/if.h>
+#include <sys/mman.h>
 
 /* Verbs header. */
 /* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
@@ -56,6 +57,7 @@
 #include <rte_pci.h>
 #include <rte_bus_pci.h>
 #include <rte_common.h>
+#include <rte_eal_memconfig.h>
 #include <rte_kvargs.h>
 
 #include "mlx5.h"
@@ -466,6 +468,101 @@ mlx5_args(struct mlx5_dev_config *config, struct rte_devargs *devargs)
 
 static struct rte_pci_driver mlx5_driver;
 
+/*
+ * Reserved UAR address space for TXQ UAR(hw doorbell) mapping, process
+ * local resource used by both primary and secondary to avoid duplicate
+ * reservation.
+ * The space has to be available on both primary and secondary process,
+ * TXQ UAR maps to this area using fixed mmap w/o double check.
+ */
+static void *uar_base;
+
+/**
+ * Reserve UAR address space for primary process
+ *
+ * @param[in] priv
+ *   Pointer to private structure.
+ *
+ * @return
+ *   0 on success, negative errno value on failure.
+ */
+static int
+mlx5_uar_init_primary(struct priv *priv)
+{
+	void *addr = (void *)0;
+	int i;
+	const struct rte_mem_config *mcfg;
+
+	if (uar_base) { /* UAR address space mapped */
+		priv->uar_base = uar_base;
+		return 0;
+	}
+	/* find out lower bound of hugepage segments */
+	mcfg = rte_eal_get_configuration()->mem_config;
+	for (i = 0; i < RTE_MAX_MEMSEG && mcfg->memseg[i].addr; i++) {
+		if (addr)
+			addr = RTE_MIN(addr, mcfg->memseg[i].addr);
+		else
+			addr = mcfg->memseg[i].addr;
+	}
+	/* offset down UAR area */
+	addr = RTE_PTR_SUB(addr, MLX5_UAR_OFFSET + MLX5_UAR_SIZE);
+	/* anonymous mmap, no real memory consumption */
+	addr = mmap(addr, MLX5_UAR_SIZE,
+		    PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+	if (addr == MAP_FAILED) {
+		ERROR("Failed to reserve UAR address space, please adjust "
+		      "MLX5_UAR_SIZE or try --base-virtaddr");
+		return -ENOMEM;
+	}
+	/* Accept either same addr or a new addr returned from mmap if target
+	 * range occupied.
+	 */
+	INFO("Reserved UAR address space: 0x%p", addr);
+	priv->uar_base = addr; /* for primary and secondary UAR re-mmap */
+	uar_base = addr; /* process local, don't reserve again */
+	return 0;
+}
+
+/**
+ * Reserve UAR address space for secondary process, align with
+ * primary process.
+ *
+ * @param[in] priv
+ *   Pointer to private structure.
+ *
+ * @return
+ *   0 on success, negative errno value on failure.
+ */
+static int
+mlx5_uar_init_secondary(struct priv *priv)
+{
+	void *addr;
+
+	assert(priv->uar_base);
+	if (uar_base) { /* already reserved */
+		assert(uar_base == priv->uar_base);
+		return 0;
+	}
+	/* anonymous mmap, no real memory consumption */
+	addr = mmap(priv->uar_base, MLX5_UAR_SIZE,
+		    PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+	if (addr == MAP_FAILED) {
+		ERROR("UAR mmap failed: 0x%p size: %llu",
+		      priv->uar_base, MLX5_UAR_SIZE);
+		return -ENXIO;
+	}
+	if (priv->uar_base != addr) {
+		ERROR("UAR address 0x%p size %llu occupied, please adjust "
+		      "MLX5_UAR_OFFSET or try EAL parameter --base-virtaddr",
+		      priv->uar_base, MLX5_UAR_SIZE);
+		return -ENXIO;
+	}
+	uar_base = addr; /* process local, don't reserve again */
+	INFO("Reserved UAR address space: 0x%p", addr);
+	return 0;
+}
+
 /**
  * DPDK callback to register a PCI device.
  *
@@ -648,6 +745,11 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
 			eth_dev->device = &pci_dev->device;
 			eth_dev->dev_ops = &mlx5_dev_sec_ops;
 			priv = eth_dev->data->dev_private;
+			err = mlx5_uar_init_secondary(priv);
+			if (err < 0) {
+				err = -err;
+				goto error;
+			}
 			/* Receive command fd from primary process */
 			err = priv_socket_connect(priv);
 			if (err < 0) {
@@ -805,6 +907,11 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
 			WARN("Rx CQE compression isn't supported");
 			config.cqe_comp = 0;
 		}
+		err = mlx5_uar_init_primary(priv);
+		if (err < 0) {
+			err = -err;
+			goto port_error;
+		}
 		/* Configure the first MAC address by default. */
 		if (priv_get_mac(priv, &mac.addr_bytes)) {
 			ERROR("cannot get MAC address, is mlx5_en loaded?"
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index 9a4a59bd4..6a234c14f 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -162,6 +162,7 @@ struct priv {
 	struct mlx5_xstats_ctrl xstats_ctrl; /* Extended stats control. */
 	rte_spinlock_t lock; /* Lock for control functions. */
 	int primary_socket; /* Unix socket for primary process. */
+	void *uar_base; /* Reserved address space for UAR mapping */
 	struct rte_intr_handle intr_handle_socket; /* Interrupt handler. */
 	struct mlx5_dev_config config; /* Device configuration. */
 };
diff --git a/drivers/net/mlx5/mlx5_defs.h b/drivers/net/mlx5/mlx5_defs.h
index 64bb0712e..26bb39720 100644
--- a/drivers/net/mlx5/mlx5_defs.h
+++ b/drivers/net/mlx5/mlx5_defs.h
@@ -110,4 +110,14 @@
 /* Supported RSS */
 #define MLX5_RSS_HF_MASK (~(ETH_RSS_IP | ETH_RSS_UDP | ETH_RSS_TCP))
 
+/* Reserved address space for UAR mapping. */
+#define MLX5_UAR_SIZE (1ULL << 32)
+
+/* Offset of reserved UAR address space to hugepage memory. Offset is used here
+ * to minimize possibility of address next to hugepage being used by other code
+ * in either primary or secondary process, failing to map TX UAR would make TX
+ * packets invisible to HW.
+ */
+#define MLX5_UAR_OFFSET (1ULL << 32)
+
 #endif /* RTE_PMD_MLX5_DEFS_H_ */
diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
index a239642ac..be53dd662 100644
--- a/drivers/net/mlx5/mlx5_rxtx.h
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -205,7 +205,7 @@ struct mlx5_txq_data {
 	volatile void *wqes; /* Work queue (use volatile to write into). */
 	volatile uint32_t *qp_db; /* Work queue doorbell. */
 	volatile uint32_t *cq_db; /* Completion queue doorbell. */
-	volatile void *bf_reg; /* Blueflame register. */
+	volatile void *bf_reg; /* Blueflame register remapped. */
 	struct mlx5_mr *mp2mr[MLX5_PMD_TX_MP_CACHE]; /* MR translation table. */
 	struct rte_mbuf *(*elts)[]; /* TX elements. */
 	struct mlx5_txq_stats stats; /* TX queue counters. */
@@ -230,6 +230,7 @@ struct mlx5_txq_ctrl {
 	struct mlx5_txq_ibv *ibv; /* Verbs queue object. */
 	struct mlx5_txq_data txq; /* Data path structure. */
 	off_t uar_mmap_offset; /* UAR mmap offset for non-primary process. */
+	volatile void *bf_reg_orig; /* Blueflame register from verbs. */
 };
 
 /* mlx5_rxq.c */
diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
index 6f5c799b5..9acfa394f 100644
--- a/drivers/net/mlx5/mlx5_trigger.c
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -76,7 +76,12 @@ priv_txq_start(struct priv *priv)
 			goto error;
 		}
 	}
-	return -ret;
+	ret = priv_tx_uar_remap(priv, priv->ctx->cmd_fd);
+	if (ret) {
+		ret = -ret;
+		goto error;
+	}
+	return ret;
 error:
 	priv_txq_stop(priv);
 	return -ret;
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index 26db15a4f..82f1b5ae8 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_txq.c
@@ -288,7 +288,9 @@ mlx5_tx_queue_release(void *dpdk_txq)
 
 
 /**
- * Map locally UAR used in Tx queues for BlueFlame doorbell.
+ * Mmap TX UAR(HW doorbell) pages into reserved UAR address space.
+ * Both primary and secondary process do mmap to make UAR address
+ * aligned.
  *
  * @param[in] priv
  *   Pointer to private structure.
@@ -296,7 +298,7 @@ mlx5_tx_queue_release(void *dpdk_txq)
  *   Verbs file descriptor to map UAR pages.
  *
  * @return
- *   0 on success, errno value on failure.
+ *   0 on success, negative errno value on failure.
  */
 int
 priv_tx_uar_remap(struct priv *priv, int fd)
@@ -305,7 +307,9 @@ priv_tx_uar_remap(struct priv *priv, int fd)
 	uintptr_t pages[priv->txqs_n];
 	unsigned int pages_n = 0;
 	uintptr_t uar_va;
+	uintptr_t off;
 	void *addr;
+	void *ret;
 	struct mlx5_txq_data *txq;
 	struct mlx5_txq_ctrl *txq_ctrl;
 	int already_mapped;
@@ -320,8 +324,10 @@ priv_tx_uar_remap(struct priv *priv, int fd)
 	for (i = 0; i != priv->txqs_n; ++i) {
 		txq = (*priv->txqs)[i];
 		txq_ctrl = container_of(txq, struct mlx5_txq_ctrl, txq);
-		uar_va = (uintptr_t)txq_ctrl->txq.bf_reg;
-		uar_va = RTE_ALIGN_FLOOR(uar_va, page_size);
+		/* UAR addr form verbs used to find dup and offset in page */
+		uar_va = (uintptr_t)txq_ctrl->bf_reg_orig;
+		off = uar_va & (page_size - 1); /* offset in page */
+		uar_va = RTE_ALIGN_FLOOR(uar_va, page_size); /* page addr */
 		already_mapped = 0;
 		for (j = 0; j != pages_n; ++j) {
 			if (pages[j] == uar_va) {
@@ -329,16 +335,29 @@ priv_tx_uar_remap(struct priv *priv, int fd)
 				break;
 			}
 		}
-		if (already_mapped)
-			continue;
-		pages[pages_n++] = uar_va;
-		addr = mmap((void *)uar_va, page_size,
-			    PROT_WRITE, MAP_FIXED | MAP_SHARED, fd,
-			    txq_ctrl->uar_mmap_offset);
-		if (addr != (void *)uar_va) {
-			ERROR("call to mmap failed on UAR for txq %d\n", i);
-			return -1;
+		/* new address in reserved UAR address space */
+		addr = RTE_PTR_ADD(priv->uar_base,
+				   uar_va & (MLX5_UAR_SIZE - 1));
+		if (!already_mapped) {
+			pages[pages_n++] = uar_va;
+			/* fixed mmap to specified address in reserved
+			 * address space
+			 */
+			ret = mmap(addr, page_size,
+				   PROT_WRITE, MAP_FIXED | MAP_SHARED, fd,
+				   txq_ctrl->uar_mmap_offset);
+			if (ret != addr) {
+				/* fixed mmap have to return same address */
+				ERROR("call to mmap failed on UAR for txq %d\n",
+				      i);
+				return -ENXIO;
+			}
 		}
+		if (rte_eal_process_type() == RTE_PROC_PRIMARY) /* save once */
+			txq_ctrl->txq.bf_reg = RTE_PTR_ADD((void *)addr, off);
+		else
+			assert(txq_ctrl->txq.bf_reg ==
+			       RTE_PTR_ADD((void *)addr, off));
 	}
 	return 0;
 }
@@ -503,7 +522,7 @@ mlx5_priv_txq_ibv_new(struct priv *priv, uint16_t idx)
 	txq_data->wqes = qp.sq.buf;
 	txq_data->wqe_n = log2above(qp.sq.wqe_cnt);
 	txq_data->qp_db = &qp.dbrec[MLX5_SND_DBR];
-	txq_data->bf_reg = qp.bf.reg;
+	txq_ctrl->bf_reg_orig = qp.bf.reg;
 	txq_data->cq_db = cq_info.dbrec;
 	txq_data->cqes =
 		(volatile struct mlx5_cqe (*)[])
@@ -836,6 +855,7 @@ mlx5_priv_txq_release(struct priv *priv, uint16_t idx)
 {
 	unsigned int i;
 	struct mlx5_txq_ctrl *txq;
+	size_t page_size = sysconf(_SC_PAGESIZE);
 
 	if (!(*priv->txqs)[idx])
 		return 0;
@@ -855,6 +875,9 @@ mlx5_priv_txq_release(struct priv *priv, uint16_t idx)
 			txq->txq.mp2mr[i] = NULL;
 		}
 	}
+	if (priv->uar_base)
+		munmap((void *)RTE_ALIGN_FLOOR((uintptr_t)txq->txq.bf_reg,
+		       page_size), page_size);
 	if (rte_atomic32_dec_and_test(&txq->refcnt)) {
 		txq_free_elts(txq);
 		LIST_REMOVE(txq, next);
-- 
2.13.3

             reply	other threads:[~2018-01-19 15:09 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-01-19 15:08 Xueming Li [this message]
2018-01-22 14:53 ` Nélio Laranjeiro
2018-01-23  9:50   ` Xueming(Steven) Li
2018-01-23 13:31     ` Nélio Laranjeiro
2018-01-23 14:16       ` Xueming(Steven) Li
2018-01-25 15:00 ` [dpdk-dev] [PATCH v2] net/mlx5: mmap uar address around huge pages Xueming Li
2018-01-25 16:01   ` Nélio Laranjeiro
2018-01-25 16:33     ` Shahaf Shuler

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=20180119150854.89828-1-xuemingl@mellanox.com \
    --to=xuemingl@mellanox.com \
    --cc=dev@dpdk.org \
    --cc=nelio.laranjeiro@6wind.com \
    --cc=shahafs@mellanox.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).