crypto: cavium - Add support for CNN55XX adapters.

Add Physical Function driver support for CNN55XX crypto adapters.
CNN55XX adapters belongs to Cavium NITROX family series,
which accelerate both Symmetric and Asymmetric crypto workloads.

These adapters have crypto engines that need firmware
to become operational.
Signed-off-by: Srikanth Jampala <Jampala.Srikanth@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

drivers/crypto/Kconfig

@@ -541,6 +541,7 @@ config CRYPTO_DEV_MXS_DCP
 
 source "drivers/crypto/qat/Kconfig"
 source "drivers/crypto/cavium/cpt/Kconfig"
+source "drivers/crypto/cavium/nitrox/Kconfig"
 
 config CRYPTO_DEV_CAVIUM_ZIP
 	tristate "Cavium ZIP driver"

drivers/crypto/Makefile

@@ -6,6 +6,7 @@ obj-$(CONFIG_CRYPTO_DEV_CAVIUM_ZIP) += cavium/
 obj-$(CONFIG_CRYPTO_DEV_CCP) += ccp/
 obj-$(CONFIG_CRYPTO_DEV_CHELSIO) += chelsio/
 obj-$(CONFIG_CRYPTO_DEV_CPT) += cavium/cpt/
+obj-$(CONFIG_CRYPTO_DEV_NITROX) += cavium/nitrox/
 obj-$(CONFIG_CRYPTO_DEV_EXYNOS_RNG) += exynos-rng.o
 obj-$(CONFIG_CRYPTO_DEV_FSL_CAAM) += caam/
 obj-$(CONFIG_CRYPTO_DEV_GEODE) += geode-aes.o

drivers/crypto/cavium/nitrox/Kconfig

+#
+# Cavium NITROX Crypto Device configuration
+#
+config CRYPTO_DEV_NITROX
+	tristate
+	select CRYPTO_BLKCIPHER
+	select CRYPTO_AES
+	select CRYPTO_DES
+	select FW_LOADER
+
+config CRYPTO_DEV_NITROX_CNN55XX
+	tristate "Support for Cavium CNN55XX driver"
+	depends on PCI_MSI && 64BIT
+	select CRYPTO_DEV_NITROX
+	default m
+	help
+	  Support for Cavium NITROX family CNN55XX driver
+	  for accelerating crypto workloads.
+
+	  To compile this as a module, choose M here: the module
+	  will be called n5pf.

drivers/crypto/cavium/nitrox/Makefile

+obj-$(CONFIG_CRYPTO_DEV_NITROX_CNN55XX) += n5pf.o
+
+n5pf-objs := nitrox_main.o \
+	nitrox_isr.o \
+	nitrox_lib.o \
+	nitrox_hal.o \
+	nitrox_reqmgr.o

drivers/crypto/cavium/nitrox/nitrox_common.h

+#ifndef __NITROX_COMMON_H
+#define __NITROX_COMMON_H
+
+#include "nitrox_dev.h"
+#include "nitrox_req.h"
+
+void nitrox_pf_cleanup_isr(struct nitrox_device *ndev);
+int nitrox_pf_init_isr(struct nitrox_device *ndev);
+
+int nitrox_common_sw_init(struct nitrox_device *ndev);
+void nitrox_common_sw_cleanup(struct nitrox_device *ndev);
+
+void pkt_slc_resp_handler(unsigned long data);
+int nitrox_process_se_request(struct nitrox_device *ndev,
+			      struct se_crypto_request *req,
+			      completion_t cb,
+			      struct skcipher_request *skreq);
+void backlog_qflush_work(struct work_struct *work);
+
+void nitrox_config_emu_unit(struct nitrox_device *ndev);
+void nitrox_config_pkt_input_rings(struct nitrox_device *ndev);
+void nitrox_config_pkt_solicit_ports(struct nitrox_device *ndev);
+void nitrox_config_vfmode(struct nitrox_device *ndev, int mode);
+void nitrox_config_nps_unit(struct nitrox_device *ndev);
+void nitrox_config_pom_unit(struct nitrox_device *ndev);
+void nitrox_config_rand_unit(struct nitrox_device *ndev);
+void nitrox_config_efl_unit(struct nitrox_device *ndev);
+void nitrox_config_bmi_unit(struct nitrox_device *ndev);
+void nitrox_config_bmo_unit(struct nitrox_device *ndev);
+void nitrox_config_lbc_unit(struct nitrox_device *ndev);
+void invalidate_lbc(struct nitrox_device *ndev);
+void enable_pkt_input_ring(struct nitrox_device *ndev, int ring);
+void enable_pkt_solicit_port(struct nitrox_device *ndev, int port);
+
+#endif /* __NITROX_COMMON_H */

drivers/crypto/cavium/nitrox/nitrox_dev.h

+#ifndef __NITROX_DEV_H
+#define __NITROX_DEV_H
+
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+
+#define VERSION_LEN 32
+
+struct nitrox_cmdq {
+	/* command queue lock */
+	spinlock_t cmdq_lock;
+	/* response list lock */
+	spinlock_t response_lock;
+	/* backlog list lock */
+	spinlock_t backlog_lock;
+
+	/* request submitted to chip, in progress */
+	struct list_head response_head;
+	/* hw queue full, hold in backlog list */
+	struct list_head backlog_head;
+
+	/* doorbell address */
+	u8 __iomem *dbell_csr_addr;
+	/* base address of the queue */
+	u8 *head;
+
+	struct nitrox_device *ndev;
+	/* flush pending backlog commands */
+	struct work_struct backlog_qflush;
+
+	/* requests posted waiting for completion */
+	atomic_t pending_count;
+	/* requests in backlog queues */
+	atomic_t backlog_count;
+
+	/* command size 32B/64B */
+	u8 instr_size;
+	u8 qno;
+	u32 qsize;
+
+	/* unaligned addresses */
+	u8 *head_unaligned;
+	dma_addr_t dma_unaligned;
+	/* dma address of the base */
+	dma_addr_t dma;
+};
+
+struct nitrox_hw {
+	/* firmware version */
+	char fw_name[VERSION_LEN];
+
+	u16 vendor_id;
+	u16 device_id;
+	u8 revision_id;
+
+	/* CNN55XX cores */
+	u8 se_cores;
+	u8 ae_cores;
+	u8 zip_cores;
+};
+
+#define MAX_MSIX_VECTOR_NAME	20
+/**
+ * vectors for queues (64 AE, 64 SE and 64 ZIP) and
+ * error condition/mailbox.
+ */
+#define MAX_MSIX_VECTORS	192
+
+struct nitrox_msix {
+	struct msix_entry *entries;
+	char **names;
+	DECLARE_BITMAP(irqs, MAX_MSIX_VECTORS);
+	u32 nr_entries;
+};
+
+struct bh_data {
+	/* slc port completion count address */
+	u8 __iomem *completion_cnt_csr_addr;
+
+	struct nitrox_cmdq *cmdq;
+	struct tasklet_struct resp_handler;
+};
+
+struct nitrox_bh {
+	struct bh_data *slc;
+};
+
+/* NITROX-5 driver state */
+#define NITROX_UCODE_LOADED	0
+#define NITROX_READY		1
+
+/* command queue size */
+#define DEFAULT_CMD_QLEN 2048
+/* command timeout in milliseconds */
+#define CMD_TIMEOUT 2000
+
+#define DEV(ndev) ((struct device *)(&(ndev)->pdev->dev))
+#define PF_MODE 0
+
+#define NITROX_CSR_ADDR(ndev, offset) \
+	((ndev)->bar_addr + (offset))
+
+/**
+ * struct nitrox_device - NITROX Device Information.
+ * @list: pointer to linked list of devices
+ * @bar_addr: iomap address
+ * @pdev: PCI device information
+ * @status: NITROX status
+ * @timeout: Request timeout in jiffies
+ * @refcnt: Device usage count
+ * @idx: device index (0..N)
+ * @node: NUMA node id attached
+ * @qlen: Command queue length
+ * @nr_queues: Number of command queues
+ * @ctx_pool: DMA pool for crypto context
+ * @pkt_cmdqs: SE Command queues
+ * @msix: MSI-X information
+ * @bh: post processing work
+ * @hw: hardware information
+ */
+struct nitrox_device {
+	struct list_head list;
+
+	u8 __iomem *bar_addr;
+	struct pci_dev *pdev;
+
+	unsigned long status;
+	unsigned long timeout;
+	refcount_t refcnt;
+
+	u8 idx;
+	int node;
+	u16 qlen;
+	u16 nr_queues;
+
+	struct dma_pool *ctx_pool;
+	struct nitrox_cmdq *pkt_cmdqs;
+
+	struct nitrox_msix msix;
+	struct nitrox_bh bh;
+
+	struct nitrox_hw hw;
+};
+
+/**
+ * nitrox_read_csr - Read from device register
+ * @ndev: NITROX device
+ * @offset: offset of the register to read
+ *
+ * Returns: value read
+ */
+static inline u64 nitrox_read_csr(struct nitrox_device *ndev, u64 offset)
+{
+	return readq(ndev->bar_addr + offset);
+}
+
+/**
+ * nitrox_write_csr - Write to device register
+ * @ndev: NITROX device
+ * @offset: offset of the register to write
+ * @value: value to write
+ */
+static inline void nitrox_write_csr(struct nitrox_device *ndev, u64 offset,
+				    u64 value)
+{
+	writeq(value, (ndev->bar_addr + offset));
+}
+
+static inline int nitrox_ready(struct nitrox_device *ndev)
+{
+	return test_bit(NITROX_READY, &ndev->status);
+}
+
+#endif /* __NITROX_DEV_H */

drivers/crypto/cavium/nitrox/nitrox_lib.c

+#include <linux/cpumask.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+#include <linux/delay.h>
+#include <linux/gfp.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/pci_regs.h>
+#include <linux/vmalloc.h>
+#include <linux/pci.h>
+
+#include "nitrox_dev.h"
+#include "nitrox_common.h"
+#include "nitrox_req.h"
+#include "nitrox_csr.h"
+
+#define CRYPTO_CTX_SIZE	256
+
+/* command queue alignments */
+#define PKT_IN_ALIGN	16
+
+static int cmdq_common_init(struct nitrox_cmdq *cmdq)
+{
+	struct nitrox_device *ndev = cmdq->ndev;
+	u32 qsize;
+
+	qsize = (ndev->qlen) * cmdq->instr_size;
+	cmdq->head_unaligned = dma_zalloc_coherent(DEV(ndev),
+						   (qsize + PKT_IN_ALIGN),
+						   &cmdq->dma_unaligned,
+						   GFP_KERNEL);
+	if (!cmdq->head_unaligned)
+		return -ENOMEM;
+
+	cmdq->head = PTR_ALIGN(cmdq->head_unaligned, PKT_IN_ALIGN);
+	cmdq->dma = PTR_ALIGN(cmdq->dma_unaligned, PKT_IN_ALIGN);
+	cmdq->qsize = (qsize + PKT_IN_ALIGN);
+
+	spin_lock_init(&cmdq->response_lock);
+	spin_lock_init(&cmdq->cmdq_lock);
+	spin_lock_init(&cmdq->backlog_lock);
+
+	INIT_LIST_HEAD(&cmdq->response_head);
+	INIT_LIST_HEAD(&cmdq->backlog_head);
+	INIT_WORK(&cmdq->backlog_qflush, backlog_qflush_work);
+
+	atomic_set(&cmdq->pending_count, 0);
+	atomic_set(&cmdq->backlog_count, 0);
+	return 0;
+}
+
+static void cmdq_common_cleanup(struct nitrox_cmdq *cmdq)
+{
+	struct nitrox_device *ndev = cmdq->ndev;
+
+	cancel_work_sync(&cmdq->backlog_qflush);
+
+	dma_free_coherent(DEV(ndev), cmdq->qsize,
+			  cmdq->head_unaligned, cmdq->dma_unaligned);
+
+	atomic_set(&cmdq->pending_count, 0);
+	atomic_set(&cmdq->backlog_count, 0);
+
+	cmdq->dbell_csr_addr = NULL;
+	cmdq->head = NULL;
+	cmdq->dma = 0;
+	cmdq->qsize = 0;
+	cmdq->instr_size = 0;
+}
+
+static void nitrox_cleanup_pkt_cmdqs(struct nitrox_device *ndev)
+{
+	int i;
+
+	for (i = 0; i < ndev->nr_queues; i++) {
+		struct nitrox_cmdq *cmdq = &ndev->pkt_cmdqs[i];
+
+		cmdq_common_cleanup(cmdq);
+	}
+	kfree(ndev->pkt_cmdqs);
+	ndev->pkt_cmdqs = NULL;
+}
+
+static int nitrox_init_pkt_cmdqs(struct nitrox_device *ndev)
+{
+	int i, err, size;
+
+	size = ndev->nr_queues * sizeof(struct nitrox_cmdq);
+	ndev->pkt_cmdqs = kzalloc(size, GFP_KERNEL);
+	if (!ndev->pkt_cmdqs)
+		return -ENOMEM;
+
+	for (i = 0; i < ndev->nr_queues; i++) {
+		struct nitrox_cmdq *cmdq;
+		u64 offset;
+
+		cmdq = &ndev->pkt_cmdqs[i];
+		cmdq->ndev = ndev;
+		cmdq->qno = i;
+		cmdq->instr_size = sizeof(struct nps_pkt_instr);
+
+		offset = NPS_PKT_IN_INSTR_BAOFF_DBELLX(i);
+		/* SE ring doorbell address for this queue */
+		cmdq->dbell_csr_addr = NITROX_CSR_ADDR(ndev, offset);
+
+		err = cmdq_common_init(cmdq);
+		if (err)
+			goto pkt_cmdq_fail;
+	}
+	return 0;
+
+pkt_cmdq_fail:
+	nitrox_cleanup_pkt_cmdqs(ndev);
+	return err;
+}
+
+static int create_crypto_dma_pool(struct nitrox_device *ndev)
+{
+	size_t size;
+
+	/* Crypto context pool, 16 byte aligned */
+	size = CRYPTO_CTX_SIZE + sizeof(struct ctx_hdr);
+	ndev->ctx_pool = dma_pool_create("crypto-context",
+					 DEV(ndev), size, 16, 0);
+	if (!ndev->ctx_pool)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void destroy_crypto_dma_pool(struct nitrox_device *ndev)
+{
+	if (!ndev->ctx_pool)
+		return;
+
+	dma_pool_destroy(ndev->ctx_pool);
+	ndev->ctx_pool = NULL;
+}
+
+/**
+ * nitrox_common_sw_init - allocate software resources.
+ * @ndev: NITROX device
+ *
+ * Allocates crypto context pools and command queues etc.
+ *
+ * Return: 0 on success, or a negative error code on error.
+ */
+int nitrox_common_sw_init(struct nitrox_device *ndev)
+{
+	int err = 0;
+
+	/* per device crypto context pool */
+	err = create_crypto_dma_pool(ndev);
+	if (err)
+		return err;
+
+	err = nitrox_init_pkt_cmdqs(ndev);
+	if (err)
+		destroy_crypto_dma_pool(ndev);
+
+	return err;
+}
+
+/**
+ * nitrox_common_sw_cleanup - free software resources.
+ * @ndev: NITROX device
+ */
+void nitrox_common_sw_cleanup(struct nitrox_device *ndev)
+{
+	nitrox_cleanup_pkt_cmdqs(ndev);
+	destroy_crypto_dma_pool(ndev);
+}