[dpdk-dev] [PATCH v5 14/25] raw/ioat: create rawdev instances on idxd PCI probe

[Bruce Richardson <bruce.richardson@intel.com>]

When a matching device is found via PCI probe create a rawdev instance for
each queue on the hardware. Use empty self-test function for these devices
so that the overall rawdev_autotest does not report failures.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Acked-by: Radu Nicolau <radu.nicolau@intel.com>
---
 drivers/raw/ioat/idxd_pci.c            | 237 ++++++++++++++++++++++++-
 drivers/raw/ioat/ioat_common.c         |  68 +++++++
 drivers/raw/ioat/ioat_private.h        |  33 ++++
 drivers/raw/ioat/ioat_rawdev_test.c    |   7 +
 drivers/raw/ioat/ioat_spec.h           |  64 +++++++
 drivers/raw/ioat/meson.build           |   1 +
 drivers/raw/ioat/rte_ioat_rawdev_fns.h |  35 +++-
 7 files changed, 442 insertions(+), 3 deletions(-)
 create mode 100644 drivers/raw/ioat/ioat_common.c

diff --git a/drivers/raw/ioat/idxd_pci.c b/drivers/raw/ioat/idxd_pci.c
index 1a30e9c31..c3fec56d5 100644
--- a/drivers/raw/ioat/idxd_pci.c
+++ b/drivers/raw/ioat/idxd_pci.c
@@ -3,8 +3,10 @@
  */
 
 #include <rte_bus_pci.h>
+#include <rte_memzone.h>
 
 #include "ioat_private.h"
+#include "ioat_spec.h"
 
 #define IDXD_VENDOR_ID		0x8086
 #define IDXD_DEVICE_ID_SPR	0x0B25
@@ -16,17 +18,246 @@ const struct rte_pci_id pci_id_idxd_map[] = {
 	{ .vendor_id = 0, /* sentinel */ },
 };
 
+static inline int
+idxd_pci_dev_command(struct idxd_rawdev *idxd, enum rte_idxd_cmds command)
+{
+	uint8_t err_code;
+	uint16_t qid = idxd->qid;
+	int i = 0;
+
+	if (command >= idxd_disable_wq && command <= idxd_reset_wq)
+		qid = (1 << qid);
+	rte_spinlock_lock(&idxd->u.pci->lk);
+	idxd->u.pci->regs->cmd = (command << IDXD_CMD_SHIFT) | qid;
+
+	do {
+		rte_pause();
+		err_code = idxd->u.pci->regs->cmdstatus;
+		if (++i >= 1000) {
+			IOAT_PMD_ERR("Timeout waiting for command response from HW");
+			rte_spinlock_unlock(&idxd->u.pci->lk);
+			return err_code;
+		}
+	} while (idxd->u.pci->regs->cmdstatus & CMDSTATUS_ACTIVE_MASK);
+	rte_spinlock_unlock(&idxd->u.pci->lk);
+
+	return err_code & CMDSTATUS_ERR_MASK;
+}
+
+static int
+idxd_is_wq_enabled(struct idxd_rawdev *idxd)
+{
+	uint32_t state = idxd->u.pci->wq_regs[idxd->qid].wqcfg[WQ_STATE_IDX];
+	return ((state >> WQ_STATE_SHIFT) & WQ_STATE_MASK) == 0x1;
+}
+
+static const struct rte_rawdev_ops idxd_pci_ops = {
+		.dev_close = idxd_rawdev_close,
+		.dev_selftest = idxd_rawdev_test,
+};
+
+/* each portal uses 4 x 4k pages */
+#define IDXD_PORTAL_SIZE (4096 * 4)
+
+static int
+init_pci_device(struct rte_pci_device *dev, struct idxd_rawdev *idxd)
+{
+	struct idxd_pci_common *pci;
+	uint8_t nb_groups, nb_engines, nb_wqs;
+	uint16_t grp_offset, wq_offset; /* how far into bar0 the regs are */
+	uint16_t wq_size, total_wq_size;
+	uint8_t lg2_max_batch, lg2_max_copy_size;
+	unsigned int i, err_code;
+
+	pci = malloc(sizeof(*pci));
+	if (pci == NULL) {
+		IOAT_PMD_ERR("%s: Can't allocate memory", __func__);
+		goto err;
+	}
+	rte_spinlock_init(&pci->lk);
+
+	/* assign the bar registers, and then configure device */
+	pci->regs = dev->mem_resource[0].addr;
+	grp_offset = (uint16_t)pci->regs->offsets[0];
+	pci->grp_regs = RTE_PTR_ADD(pci->regs, grp_offset * 0x100);
+	wq_offset = (uint16_t)(pci->regs->offsets[0] >> 16);
+	pci->wq_regs = RTE_PTR_ADD(pci->regs, wq_offset * 0x100);
+	pci->portals = dev->mem_resource[2].addr;
+
+	/* sanity check device status */
+	if (pci->regs->gensts & GENSTS_DEV_STATE_MASK) {
+		/* need function-level-reset (FLR) or is enabled */
+		IOAT_PMD_ERR("Device status is not disabled, cannot init");
+		goto err;
+	}
+	if (pci->regs->cmdstatus & CMDSTATUS_ACTIVE_MASK) {
+		/* command in progress */
+		IOAT_PMD_ERR("Device has a command in progress, cannot init");
+		goto err;
+	}
+
+	/* read basic info about the hardware for use when configuring */
+	nb_groups = (uint8_t)pci->regs->grpcap;
+	nb_engines = (uint8_t)pci->regs->engcap;
+	nb_wqs = (uint8_t)(pci->regs->wqcap >> 16);
+	total_wq_size = (uint16_t)pci->regs->wqcap;
+	lg2_max_copy_size = (uint8_t)(pci->regs->gencap >> 16) & 0x1F;
+	lg2_max_batch = (uint8_t)(pci->regs->gencap >> 21) & 0x0F;
+
+	IOAT_PMD_DEBUG("nb_groups = %u, nb_engines = %u, nb_wqs = %u",
+			nb_groups, nb_engines, nb_wqs);
+
+	/* zero out any old config */
+	for (i = 0; i < nb_groups; i++) {
+		pci->grp_regs[i].grpengcfg = 0;
+		pci->grp_regs[i].grpwqcfg[0] = 0;
+	}
+	for (i = 0; i < nb_wqs; i++)
+		pci->wq_regs[i].wqcfg[0] = 0;
+
+	/* put each engine into a separate group to avoid reordering */
+	if (nb_groups > nb_engines)
+		nb_groups = nb_engines;
+	if (nb_groups < nb_engines)
+		nb_engines = nb_groups;
+
+	/* assign engines to groups, round-robin style */
+	for (i = 0; i < nb_engines; i++) {
+		IOAT_PMD_DEBUG("Assigning engine %u to group %u",
+				i, i % nb_groups);
+		pci->grp_regs[i % nb_groups].grpengcfg |= (1ULL << i);
+	}
+
+	/* now do the same for queues and give work slots to each queue */
+	wq_size = total_wq_size / nb_wqs;
+	IOAT_PMD_DEBUG("Work queue size = %u, max batch = 2^%u, max copy = 2^%u",
+			wq_size, lg2_max_batch, lg2_max_copy_size);
+	for (i = 0; i < nb_wqs; i++) {
+		/* add engine "i" to a group */
+		IOAT_PMD_DEBUG("Assigning work queue %u to group %u",
+				i, i % nb_groups);
+		pci->grp_regs[i % nb_groups].grpwqcfg[0] |= (1ULL << i);
+		/* now configure it, in terms of size, max batch, mode */
+		pci->wq_regs[i].wqcfg[WQ_SIZE_IDX] = wq_size;
+		pci->wq_regs[i].wqcfg[WQ_MODE_IDX] = (1 << WQ_PRIORITY_SHIFT) |
+				WQ_MODE_DEDICATED;
+		pci->wq_regs[i].wqcfg[WQ_SIZES_IDX] = lg2_max_copy_size |
+				(lg2_max_batch << WQ_BATCH_SZ_SHIFT);
+	}
+
+	/* dump the group configuration to output */
+	for (i = 0; i < nb_groups; i++) {
+		IOAT_PMD_DEBUG("## Group %d", i);
+		IOAT_PMD_DEBUG("    GRPWQCFG: %"PRIx64, pci->grp_regs[i].grpwqcfg[0]);
+		IOAT_PMD_DEBUG("    GRPENGCFG: %"PRIx64, pci->grp_regs[i].grpengcfg);
+		IOAT_PMD_DEBUG("    GRPFLAGS: %"PRIx32, pci->grp_regs[i].grpflags);
+	}
+
+	idxd->u.pci = pci;
+	idxd->max_batches = wq_size;
+
+	/* enable the device itself */
+	err_code = idxd_pci_dev_command(idxd, idxd_enable_dev);
+	if (err_code) {
+		IOAT_PMD_ERR("Error enabling device: code %#x", err_code);
+		return err_code;
+	}
+	IOAT_PMD_DEBUG("IDXD Device enabled OK");
+
+	return nb_wqs;
+
+err:
+	free(pci);
+	return -1;
+}
+
 static int
 idxd_rawdev_probe_pci(struct rte_pci_driver *drv, struct rte_pci_device *dev)
 {
-	int ret = 0;
+	struct idxd_rawdev idxd = {{0}}; /* Double {} to avoid error on BSD12 */
+	uint8_t nb_wqs;
+	int qid, ret = 0;
 	char name[PCI_PRI_STR_SIZE];
 
 	rte_pci_device_name(&dev->addr, name, sizeof(name));
 	IOAT_PMD_INFO("Init %s on NUMA node %d", name, dev->device.numa_node);
 	dev->device.driver = &drv->driver;
 
-	return ret;
+	ret = init_pci_device(dev, &idxd);
+	if (ret < 0) {
+		IOAT_PMD_ERR("Error initializing PCI hardware");
+		return ret;
+	}
+	nb_wqs = (uint8_t)ret;
+
+	/* set up one device for each queue */
+	for (qid = 0; qid < nb_wqs; qid++) {
+		char qname[32];
+
+		/* add the queue number to each device name */
+		snprintf(qname, sizeof(qname), "%s-q%d", name, qid);
+		idxd.qid = qid;
+		idxd.public.portal = RTE_PTR_ADD(idxd.u.pci->portals,
+				qid * IDXD_PORTAL_SIZE);
+		if (idxd_is_wq_enabled(&idxd))
+			IOAT_PMD_ERR("Error, WQ %u seems enabled", qid);
+		ret = idxd_rawdev_create(qname, &dev->device,
+				&idxd, &idxd_pci_ops);
+		if (ret != 0) {
+			IOAT_PMD_ERR("Failed to create rawdev %s", name);
+			if (qid == 0) /* if no devices using this, free pci */
+				free(idxd.u.pci);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static int
+idxd_rawdev_destroy(const char *name)
+{
+	int ret;
+	uint8_t err_code;
+	struct rte_rawdev *rdev;
+	struct idxd_rawdev *idxd;
+
+	if (!name) {
+		IOAT_PMD_ERR("Invalid device name");
+		return -EINVAL;
+	}
+
+	rdev = rte_rawdev_pmd_get_named_dev(name);
+	if (!rdev) {
+		IOAT_PMD_ERR("Invalid device name (%s)", name);
+		return -EINVAL;
+	}
+
+	idxd = rdev->dev_private;
+
+	/* disable the device */
+	err_code = idxd_pci_dev_command(idxd, idxd_disable_dev);
+	if (err_code) {
+		IOAT_PMD_ERR("Error disabling device: code %#x", err_code);
+		return err_code;
+	}
+	IOAT_PMD_DEBUG("IDXD Device disabled OK");
+
+	/* free device memory */
+	if (rdev->dev_private != NULL) {
+		IOAT_PMD_DEBUG("Freeing device driver memory");
+		rdev->dev_private = NULL;
+		rte_free(idxd->public.batch_ring);
+		rte_free(idxd->public.hdl_ring);
+		rte_memzone_free(idxd->mz);
+	}
+
+	/* rte_rawdev_close is called by pmd_release */
+	ret = rte_rawdev_pmd_release(rdev);
+	if (ret)
+		IOAT_PMD_DEBUG("Device cleanup failed");
+
+	return 0;
 }
 
 static int
@@ -40,6 +271,8 @@ idxd_rawdev_remove_pci(struct rte_pci_device *dev)
 	IOAT_PMD_INFO("Closing %s on NUMA node %d",
 			name, dev->device.numa_node);
 
+	ret = idxd_rawdev_destroy(name);
+
 	return ret;
 }
 
diff --git a/drivers/raw/ioat/ioat_common.c b/drivers/raw/ioat/ioat_common.c
new file mode 100644
index 000000000..c3aa015ed
--- /dev/null
+++ b/drivers/raw/ioat/ioat_common.c
@@ -0,0 +1,68 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Intel Corporation
+ */
+
+#include <rte_rawdev_pmd.h>
+#include <rte_memzone.h>
+#include <rte_common.h>
+
+#include "ioat_private.h"
+
+int
+idxd_rawdev_close(struct rte_rawdev *dev __rte_unused)
+{
+	return 0;
+}
+
+int
+idxd_rawdev_create(const char *name, struct rte_device *dev,
+		   const struct idxd_rawdev *base_idxd,
+		   const struct rte_rawdev_ops *ops)
+{
+	struct idxd_rawdev *idxd;
+	struct rte_rawdev *rawdev = NULL;
+	const struct rte_memzone *mz = NULL;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	int ret = 0;
+
+	if (!name) {
+		IOAT_PMD_ERR("Invalid name of the device!");
+		ret = -EINVAL;
+		goto cleanup;
+	}
+
+	/* Allocate device structure */
+	rawdev = rte_rawdev_pmd_allocate(name, sizeof(struct idxd_rawdev),
+					 dev->numa_node);
+	if (rawdev == NULL) {
+		IOAT_PMD_ERR("Unable to allocate raw device");
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	snprintf(mz_name, sizeof(mz_name), "rawdev%u_private", rawdev->dev_id);
+	mz = rte_memzone_reserve(mz_name, sizeof(struct idxd_rawdev),
+			dev->numa_node, RTE_MEMZONE_IOVA_CONTIG);
+	if (mz == NULL) {
+		IOAT_PMD_ERR("Unable to reserve memzone for private data\n");
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+	rawdev->dev_private = mz->addr;
+	rawdev->dev_ops = ops;
+	rawdev->device = dev;
+	rawdev->driver_name = IOAT_PMD_RAWDEV_NAME_STR;
+
+	idxd = rawdev->dev_private;
+	*idxd = *base_idxd; /* copy over the main fields already passed in */
+	idxd->rawdev = rawdev;
+	idxd->mz = mz;
+
+	return 0;
+
+cleanup:
+	if (rawdev)
+		rte_rawdev_pmd_release(rawdev);
+
+	return ret;
+}
diff --git a/drivers/raw/ioat/ioat_private.h b/drivers/raw/ioat/ioat_private.h
index d87d4d055..53f00a9f3 100644
--- a/drivers/raw/ioat/ioat_private.h
+++ b/drivers/raw/ioat/ioat_private.h
@@ -14,6 +14,10 @@
  * @b EXPERIMENTAL: these structures and APIs may change without prior notice
  */
 
+#include <rte_spinlock.h>
+#include <rte_rawdev_pmd.h>
+#include "rte_ioat_rawdev.h"
+
 extern int ioat_pmd_logtype;
 
 #define IOAT_PMD_LOG(level, fmt, args...) rte_log(RTE_LOG_ ## level, \
@@ -24,4 +28,33 @@ extern int ioat_pmd_logtype;
 #define IOAT_PMD_ERR(fmt, args...)    IOAT_PMD_LOG(ERR, fmt, ## args)
 #define IOAT_PMD_WARN(fmt, args...)   IOAT_PMD_LOG(WARNING, fmt, ## args)
 
+struct idxd_pci_common {
+	rte_spinlock_t lk;
+	volatile struct rte_idxd_bar0 *regs;
+	volatile struct rte_idxd_wqcfg *wq_regs;
+	volatile struct rte_idxd_grpcfg *grp_regs;
+	volatile void *portals;
+};
+
+struct idxd_rawdev {
+	struct rte_idxd_rawdev public; /* the public members, must be first */
+
+	struct rte_rawdev *rawdev;
+	const struct rte_memzone *mz;
+	uint8_t qid;
+	uint16_t max_batches;
+
+	union {
+		struct idxd_pci_common *pci;
+	} u;
+};
+
+extern int idxd_rawdev_create(const char *name, struct rte_device *dev,
+		       const struct idxd_rawdev *idxd,
+		       const struct rte_rawdev_ops *ops);
+
+extern int idxd_rawdev_close(struct rte_rawdev *dev);
+
+extern int idxd_rawdev_test(uint16_t dev_id);
+
 #endif /* _IOAT_PRIVATE_H_ */
diff --git a/drivers/raw/ioat/ioat_rawdev_test.c b/drivers/raw/ioat/ioat_rawdev_test.c
index 8b665cc9a..87a65b7ae 100644
--- a/drivers/raw/ioat/ioat_rawdev_test.c
+++ b/drivers/raw/ioat/ioat_rawdev_test.c
@@ -7,6 +7,7 @@
 #include <rte_mbuf.h>
 #include "rte_rawdev.h"
 #include "rte_ioat_rawdev.h"
+#include "ioat_private.h"
 
 #define MAX_SUPPORTED_RAWDEVS 64
 #define TEST_SKIPPED 77
@@ -267,3 +268,9 @@ ioat_rawdev_test(uint16_t dev_id)
 	free(ids);
 	return -1;
 }
+
+int
+idxd_rawdev_test(uint16_t dev_id __rte_unused)
+{
+	return 0;
+}
diff --git a/drivers/raw/ioat/ioat_spec.h b/drivers/raw/ioat/ioat_spec.h
index 9645e16d4..1aa768b9a 100644
--- a/drivers/raw/ioat/ioat_spec.h
+++ b/drivers/raw/ioat/ioat_spec.h
@@ -268,6 +268,70 @@ union rte_ioat_hw_desc {
 	struct rte_ioat_pq_update_hw_desc pq_update;
 };
 
+/*** Definitions for Intel(R) Data Streaming Accelerator Follow ***/
+
+#define IDXD_CMD_SHIFT 20
+enum rte_idxd_cmds {
+	idxd_enable_dev = 1,
+	idxd_disable_dev,
+	idxd_drain_all,
+	idxd_abort_all,
+	idxd_reset_device,
+	idxd_enable_wq,
+	idxd_disable_wq,
+	idxd_drain_wq,
+	idxd_abort_wq,
+	idxd_reset_wq,
+};
+
+/* General bar0 registers */
+struct rte_idxd_bar0 {
+	uint32_t __rte_cache_aligned version;    /* offset 0x00 */
+	uint64_t __rte_aligned(0x10) gencap;     /* offset 0x10 */
+	uint64_t __rte_aligned(0x10) wqcap;      /* offset 0x20 */
+	uint64_t __rte_aligned(0x10) grpcap;     /* offset 0x30 */
+	uint64_t __rte_aligned(0x08) engcap;     /* offset 0x38 */
+	uint64_t __rte_aligned(0x10) opcap;      /* offset 0x40 */
+	uint64_t __rte_aligned(0x20) offsets[2]; /* offset 0x60 */
+	uint32_t __rte_aligned(0x20) gencfg;     /* offset 0x80 */
+	uint32_t __rte_aligned(0x08) genctrl;    /* offset 0x88 */
+	uint32_t __rte_aligned(0x10) gensts;     /* offset 0x90 */
+	uint32_t __rte_aligned(0x08) intcause;   /* offset 0x98 */
+	uint32_t __rte_aligned(0x10) cmd;        /* offset 0xA0 */
+	uint32_t __rte_aligned(0x08) cmdstatus;  /* offset 0xA8 */
+	uint64_t __rte_aligned(0x20) swerror[4]; /* offset 0xC0 */
+};
+
+struct rte_idxd_wqcfg {
+	uint32_t wqcfg[8] __rte_aligned(32); /* 32-byte register */
+};
+
+#define WQ_SIZE_IDX      0 /* size is in first 32-bit value */
+#define WQ_THRESHOLD_IDX 1 /* WQ threshold second 32-bits */
+#define WQ_MODE_IDX      2 /* WQ mode and other flags */
+#define WQ_SIZES_IDX     3 /* WQ transfer and batch sizes */
+#define WQ_OCC_INT_IDX   4 /* WQ occupancy interrupt handle */
+#define WQ_OCC_LIMIT_IDX 5 /* WQ occupancy limit */
+#define WQ_STATE_IDX     6 /* WQ state and occupancy state */
+
+#define WQ_MODE_SHARED    0
+#define WQ_MODE_DEDICATED 1
+#define WQ_PRIORITY_SHIFT 4
+#define WQ_BATCH_SZ_SHIFT 5
+#define WQ_STATE_SHIFT 30
+#define WQ_STATE_MASK 0x3
+
+struct rte_idxd_grpcfg {
+	uint64_t grpwqcfg[4]  __rte_cache_aligned; /* 64-byte register set */
+	uint64_t grpengcfg;  /* offset 32 */
+	uint32_t grpflags;   /* offset 40 */
+};
+
+#define GENSTS_DEV_STATE_MASK 0x03
+#define CMDSTATUS_ACTIVE_SHIFT 31
+#define CMDSTATUS_ACTIVE_MASK (1 << 31)
+#define CMDSTATUS_ERR_MASK 0xFF
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/drivers/raw/ioat/meson.build b/drivers/raw/ioat/meson.build
index b343b7367..5eff76a1a 100644
--- a/drivers/raw/ioat/meson.build
+++ b/drivers/raw/ioat/meson.build
@@ -6,6 +6,7 @@ reason = 'only supported on x86'
 sources = files(
 	'idxd_pci.c',
 	'idxd_vdev.c',
+	'ioat_common.c',
 	'ioat_rawdev.c',
 	'ioat_rawdev_test.c')
 deps += ['bus_pci',
diff --git a/drivers/raw/ioat/rte_ioat_rawdev_fns.h b/drivers/raw/ioat/rte_ioat_rawdev_fns.h
index c6e0b9a58..fa2eb5334 100644
--- a/drivers/raw/ioat/rte_ioat_rawdev_fns.h
+++ b/drivers/raw/ioat/rte_ioat_rawdev_fns.h
@@ -41,9 +41,20 @@ struct rte_ioat_generic_hw_desc {
 
 /**
  * @internal
- * Structure representing a device instance
+ * Identify the data path to use.
+ * Must be first field of rte_ioat_rawdev and rte_idxd_rawdev structs
+ */
+enum rte_ioat_dev_type {
+	RTE_IOAT_DEV,
+	RTE_IDXD_DEV,
+};
+
+/**
+ * @internal
+ * Structure representing an IOAT device instance
  */
 struct rte_ioat_rawdev {
+	enum rte_ioat_dev_type type;
 	struct rte_rawdev *rawdev;
 	const struct rte_memzone *mz;
 	const struct rte_memzone *desc_mz;
@@ -79,6 +90,28 @@ struct rte_ioat_rawdev {
 #define RTE_IOAT_CHANSTS_HALTED			0x3
 #define RTE_IOAT_CHANSTS_ARMED			0x4
 
+/**
+ * @internal
+ * Structure representing an IDXD device instance
+ */
+struct rte_idxd_rawdev {
+	enum rte_ioat_dev_type type;
+	void *portal; /* address to write the batch descriptor */
+
+	/* counters to track the batches and the individual op handles */
+	uint16_t batch_ring_sz;  /* size of batch ring */
+	uint16_t hdl_ring_sz;    /* size of the user hdl ring */
+
+	uint16_t next_batch;     /* where we write descriptor ops */
+	uint16_t next_completed; /* batch where we read completions */
+	uint16_t next_ret_hdl;   /* the next user hdl to return */
+	uint16_t last_completed_hdl; /* the last user hdl that has completed */
+	uint16_t next_free_hdl;  /* where the handle for next op will go */
+
+	struct rte_idxd_user_hdl *hdl_ring;
+	struct rte_idxd_desc_batch *batch_ring;
+};
+
 /*
  * Enqueue a copy operation onto the ioat device
  */
-- 
2.25.1