spi: new orion_spi driver
authorShadi Ammouri <shadi@marvell.com>
Tue, 5 Aug 2008 20:01:09 +0000 (13:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 5 Aug 2008 21:33:46 +0000 (14:33 -0700)
This adds an SPI driver for the SPI controller found in various Marvell
Orion ARM SoCs.  It currently supports only one slave, which must use SPI
mode 0.

[dbrownell@users.sourceforge.net: cleanups, meet specs, pass "sparse"]
Signed-off-by: Shadi Ammouri <shadi@marvell.com>
Signed-off-by: Saeed Bishara <saeed@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/spi/Kconfig
drivers/spi/Makefile
drivers/spi/orion_spi.c [new file with mode: 0644]
include/linux/spi/orion_spi.h [new file with mode: 0644]

index 2303521b4f0984efa7fcc7c1719de1d33a71e74e..b9d0efb6803f2919695c21ed6ba186303c51353c 100644 (file)
@@ -149,6 +149,12 @@ config SPI_OMAP24XX
          SPI master controller for OMAP24xx/OMAP34xx Multichannel SPI
          (McSPI) modules.
 
+config SPI_ORION
+       tristate "Orion SPI master (EXPERIMENTAL)"
+       depends on PLAT_ORION && EXPERIMENTAL
+       help
+         This enables using the SPI master controller on the Orion chips.
+
 config SPI_PXA2XX
        tristate "PXA2xx SSP SPI master"
        depends on ARCH_PXA && EXPERIMENTAL
index 7fca043ce723a77698d1946e1f88947c9a1d5008..ccf18de34e1e0df26a697b2939d58c7e62cf231d 100644 (file)
@@ -21,6 +21,7 @@ obj-$(CONFIG_SPI_LM70_LLP)            += spi_lm70llp.o
 obj-$(CONFIG_SPI_PXA2XX)               += pxa2xx_spi.o
 obj-$(CONFIG_SPI_OMAP_UWIRE)           += omap_uwire.o
 obj-$(CONFIG_SPI_OMAP24XX)             += omap2_mcspi.o
+obj-$(CONFIG_SPI_ORION)                        += orion_spi.o
 obj-$(CONFIG_SPI_MPC52xx_PSC)          += mpc52xx_psc_spi.o
 obj-$(CONFIG_SPI_MPC83xx)              += spi_mpc83xx.o
 obj-$(CONFIG_SPI_S3C24XX_GPIO)         += spi_s3c24xx_gpio.o
diff --git a/drivers/spi/orion_spi.c b/drivers/spi/orion_spi.c
new file mode 100644 (file)
index 0000000..c4eaacd
--- /dev/null
@@ -0,0 +1,574 @@
+/*
+ * orion_spi.c -- Marvell Orion SPI controller driver
+ *
+ * Author: Shadi Ammouri <shadi@marvell.com>
+ * Copyright (C) 2007-2008 Marvell Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/delay.h>
+#include <linux/platform_device.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/spi/spi.h>
+#include <linux/spi/orion_spi.h>
+#include <asm/unaligned.h>
+
+#define DRIVER_NAME                    "orion_spi"
+
+#define ORION_NUM_CHIPSELECTS          1 /* only one slave is supported*/
+#define ORION_SPI_WAIT_RDY_MAX_LOOP    2000 /* in usec */
+
+#define ORION_SPI_IF_CTRL_REG          0x00
+#define ORION_SPI_IF_CONFIG_REG                0x04
+#define ORION_SPI_DATA_OUT_REG         0x08
+#define ORION_SPI_DATA_IN_REG          0x0c
+#define ORION_SPI_INT_CAUSE_REG                0x10
+
+#define ORION_SPI_IF_8_16_BIT_MODE     (1 << 5)
+#define ORION_SPI_CLK_PRESCALE_MASK    0x1F
+
+struct orion_spi {
+       struct work_struct      work;
+
+       /* Lock access to transfer list.        */
+       spinlock_t              lock;
+
+       struct list_head        msg_queue;
+       struct spi_master       *master;
+       void __iomem            *base;
+       unsigned int            max_speed;
+       unsigned int            min_speed;
+       struct orion_spi_info   *spi_info;
+};
+
+static struct workqueue_struct *orion_spi_wq;
+
+static inline void __iomem *spi_reg(struct orion_spi *orion_spi, u32 reg)
+{
+       return orion_spi->base + reg;
+}
+
+static inline void
+orion_spi_setbits(struct orion_spi *orion_spi, u32 reg, u32 mask)
+{
+       void __iomem *reg_addr = spi_reg(orion_spi, reg);
+       u32 val;
+
+       val = readl(reg_addr);
+       val |= mask;
+       writel(val, reg_addr);
+}
+
+static inline void
+orion_spi_clrbits(struct orion_spi *orion_spi, u32 reg, u32 mask)
+{
+       void __iomem *reg_addr = spi_reg(orion_spi, reg);
+       u32 val;
+
+       val = readl(reg_addr);
+       val &= ~mask;
+       writel(val, reg_addr);
+}
+
+static int orion_spi_set_transfer_size(struct orion_spi *orion_spi, int size)
+{
+       if (size == 16) {
+               orion_spi_setbits(orion_spi, ORION_SPI_IF_CONFIG_REG,
+                                 ORION_SPI_IF_8_16_BIT_MODE);
+       } else if (size == 8) {
+               orion_spi_clrbits(orion_spi, ORION_SPI_IF_CONFIG_REG,
+                                 ORION_SPI_IF_8_16_BIT_MODE);
+       } else {
+               pr_debug("Bad bits per word value %d (only 8 or 16 are "
+                        "allowed).\n", size);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int orion_spi_baudrate_set(struct spi_device *spi, unsigned int speed)
+{
+       u32 tclk_hz;
+       u32 rate;
+       u32 prescale;
+       u32 reg;
+       struct orion_spi *orion_spi;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+
+       tclk_hz = orion_spi->spi_info->tclk;
+
+       /*
+        * the supported rates are: 4,6,8...30
+        * round up as we look for equal or less speed
+        */
+       rate = DIV_ROUND_UP(tclk_hz, speed);
+       rate = roundup(rate, 2);
+
+       /* check if requested speed is too small */
+       if (rate > 30)
+               return -EINVAL;
+
+       if (rate < 4)
+               rate = 4;
+
+       /* Convert the rate to SPI clock divisor value. */
+       prescale = 0x10 + rate/2;
+
+       reg = readl(spi_reg(orion_spi, ORION_SPI_IF_CONFIG_REG));
+       reg = ((reg & ~ORION_SPI_CLK_PRESCALE_MASK) | prescale);
+       writel(reg, spi_reg(orion_spi, ORION_SPI_IF_CONFIG_REG));
+
+       return 0;
+}
+
+/*
+ * called only when no transfer is active on the bus
+ */
+static int
+orion_spi_setup_transfer(struct spi_device *spi, struct spi_transfer *t)
+{
+       struct orion_spi *orion_spi;
+       unsigned int speed = spi->max_speed_hz;
+       unsigned int bits_per_word = spi->bits_per_word;
+       int     rc;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+
+       if ((t != NULL) && t->speed_hz)
+               speed = t->speed_hz;
+
+       if ((t != NULL) && t->bits_per_word)
+               bits_per_word = t->bits_per_word;
+
+       rc = orion_spi_baudrate_set(spi, speed);
+       if (rc)
+               return rc;
+
+       return orion_spi_set_transfer_size(orion_spi, bits_per_word);
+}
+
+static void orion_spi_set_cs(struct orion_spi *orion_spi, int enable)
+{
+       if (enable)
+               orion_spi_setbits(orion_spi, ORION_SPI_IF_CTRL_REG, 0x1);
+       else
+               orion_spi_clrbits(orion_spi, ORION_SPI_IF_CTRL_REG, 0x1);
+}
+
+static inline int orion_spi_wait_till_ready(struct orion_spi *orion_spi)
+{
+       int i;
+
+       for (i = 0; i < ORION_SPI_WAIT_RDY_MAX_LOOP; i++) {
+               if (readl(spi_reg(orion_spi, ORION_SPI_INT_CAUSE_REG)))
+                       return 1;
+               else
+                       udelay(1);
+       }
+
+       return -1;
+}
+
+static inline int
+orion_spi_write_read_8bit(struct spi_device *spi,
+                         const u8 **tx_buf, u8 **rx_buf)
+{
+       void __iomem *tx_reg, *rx_reg, *int_reg;
+       struct orion_spi *orion_spi;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+       tx_reg = spi_reg(orion_spi, ORION_SPI_DATA_OUT_REG);
+       rx_reg = spi_reg(orion_spi, ORION_SPI_DATA_IN_REG);
+       int_reg = spi_reg(orion_spi, ORION_SPI_INT_CAUSE_REG);
+
+       /* clear the interrupt cause register */
+       writel(0x0, int_reg);
+
+       if (tx_buf && *tx_buf)
+               writel(*(*tx_buf)++, tx_reg);
+       else
+               writel(0, tx_reg);
+
+       if (orion_spi_wait_till_ready(orion_spi) < 0) {
+               dev_err(&spi->dev, "TXS timed out\n");
+               return -1;
+       }
+
+       if (rx_buf && *rx_buf)
+               *(*rx_buf)++ = readl(rx_reg);
+
+       return 1;
+}
+
+static inline int
+orion_spi_write_read_16bit(struct spi_device *spi,
+                          const u16 **tx_buf, u16 **rx_buf)
+{
+       void __iomem *tx_reg, *rx_reg, *int_reg;
+       struct orion_spi *orion_spi;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+       tx_reg = spi_reg(orion_spi, ORION_SPI_DATA_OUT_REG);
+       rx_reg = spi_reg(orion_spi, ORION_SPI_DATA_IN_REG);
+       int_reg = spi_reg(orion_spi, ORION_SPI_INT_CAUSE_REG);
+
+       /* clear the interrupt cause register */
+       writel(0x0, int_reg);
+
+       if (tx_buf && *tx_buf)
+               writel(__cpu_to_le16(get_unaligned((*tx_buf)++)), tx_reg);
+       else
+               writel(0, tx_reg);
+
+       if (orion_spi_wait_till_ready(orion_spi) < 0) {
+               dev_err(&spi->dev, "TXS timed out\n");
+               return -1;
+       }
+
+       if (rx_buf && *rx_buf)
+               put_unaligned(__le16_to_cpu(readl(rx_reg)), (*rx_buf)++);
+
+       return 1;
+}
+
+static unsigned int
+orion_spi_write_read(struct spi_device *spi, struct spi_transfer *xfer)
+{
+       struct orion_spi *orion_spi;
+       unsigned int count;
+       int word_len;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+       word_len = spi->bits_per_word;
+       count = xfer->len;
+
+       if (word_len == 8) {
+               const u8 *tx = xfer->tx_buf;
+               u8 *rx = xfer->rx_buf;
+
+               do {
+                       if (orion_spi_write_read_8bit(spi, &tx, &rx) < 0)
+                               goto out;
+                       count--;
+               } while (count);
+       } else if (word_len == 16) {
+               const u16 *tx = xfer->tx_buf;
+               u16 *rx = xfer->rx_buf;
+
+               do {
+                       if (orion_spi_write_read_16bit(spi, &tx, &rx) < 0)
+                               goto out;
+                       count -= 2;
+               } while (count);
+       }
+
+out:
+       return xfer->len - count;
+}
+
+
+static void orion_spi_work(struct work_struct *work)
+{
+       struct orion_spi *orion_spi =
+               container_of(work, struct orion_spi, work);
+
+       spin_lock_irq(&orion_spi->lock);
+       while (!list_empty(&orion_spi->msg_queue)) {
+               struct spi_message *m;
+               struct spi_device *spi;
+               struct spi_transfer *t = NULL;
+               int par_override = 0;
+               int status = 0;
+               int cs_active = 0;
+
+               m = container_of(orion_spi->msg_queue.next, struct spi_message,
+                                queue);
+
+               list_del_init(&m->queue);
+               spin_unlock_irq(&orion_spi->lock);
+
+               spi = m->spi;
+
+               /* Load defaults */
+               status = orion_spi_setup_transfer(spi, NULL);
+
+               if (status < 0)
+                       goto msg_done;
+
+               list_for_each_entry(t, &m->transfers, transfer_list) {
+                       if (par_override || t->speed_hz || t->bits_per_word) {
+                               par_override = 1;
+                               status = orion_spi_setup_transfer(spi, t);
+                               if (status < 0)
+                                       break;
+                               if (!t->speed_hz && !t->bits_per_word)
+                                       par_override = 0;
+                       }
+
+                       if (!cs_active) {
+                               orion_spi_set_cs(orion_spi, 1);
+                               cs_active = 1;
+                       }
+
+                       if (t->len)
+                               m->actual_length +=
+                                       orion_spi_write_read(spi, t);
+
+                       if (t->delay_usecs)
+                               udelay(t->delay_usecs);
+
+                       if (t->cs_change) {
+                               orion_spi_set_cs(orion_spi, 0);
+                               cs_active = 0;
+                       }
+               }
+
+msg_done:
+               if (cs_active)
+                       orion_spi_set_cs(orion_spi, 0);
+
+               m->status = status;
+               m->complete(m->context);
+
+               spin_lock_irq(&orion_spi->lock);
+       }
+
+       spin_unlock_irq(&orion_spi->lock);
+}
+
+static int __init orion_spi_reset(struct orion_spi *orion_spi)
+{
+       /* Verify that the CS is deasserted */
+       orion_spi_set_cs(orion_spi, 0);
+
+       return 0;
+}
+
+static int orion_spi_setup(struct spi_device *spi)
+{
+       struct orion_spi *orion_spi;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+
+       if (spi->mode) {
+               dev_err(&spi->dev, "setup: unsupported mode bits %x\n",
+                       spi->mode);
+               return -EINVAL;
+       }
+
+       if (spi->bits_per_word == 0)
+               spi->bits_per_word = 8;
+
+       if ((spi->max_speed_hz == 0)
+                       || (spi->max_speed_hz > orion_spi->max_speed))
+               spi->max_speed_hz = orion_spi->max_speed;
+
+       if (spi->max_speed_hz < orion_spi->min_speed) {
+               dev_err(&spi->dev, "setup: requested speed too low %d Hz\n",
+                       spi->max_speed_hz);
+               return -EINVAL;
+       }
+
+       /*
+        * baudrate & width will be set orion_spi_setup_transfer
+        */
+       return 0;
+}
+
+static int orion_spi_transfer(struct spi_device *spi, struct spi_message *m)
+{
+       struct orion_spi *orion_spi;
+       struct spi_transfer *t = NULL;
+       unsigned long flags;
+
+       m->actual_length = 0;
+       m->status = 0;
+
+       /* reject invalid messages and transfers */
+       if (list_empty(&m->transfers) || !m->complete)
+               return -EINVAL;
+
+       orion_spi = spi_master_get_devdata(spi->master);
+
+       list_for_each_entry(t, &m->transfers, transfer_list) {
+               unsigned int bits_per_word = spi->bits_per_word;
+
+               if (t->tx_buf == NULL && t->rx_buf == NULL && t->len) {
+                       dev_err(&spi->dev,
+                               "message rejected : "
+                               "invalid transfer data buffers\n");
+                       goto msg_rejected;
+               }
+
+               if ((t != NULL) && t->bits_per_word)
+                       bits_per_word = t->bits_per_word;
+
+               if ((bits_per_word != 8) && (bits_per_word != 16)) {
+                       dev_err(&spi->dev,
+                               "message rejected : "
+                               "invalid transfer bits_per_word (%d bits)\n",
+                               bits_per_word);
+                       goto msg_rejected;
+               }
+               /*make sure buffer length is even when working in 16 bit mode*/
+               if ((t != NULL) && (t->bits_per_word == 16) && (t->len & 1)) {
+                       dev_err(&spi->dev,
+                               "message rejected : "
+                               "odd data length (%d) while in 16 bit mode\n",
+                               t->len);
+                       goto msg_rejected;
+               }
+
+               if (t->speed_hz < orion_spi->min_speed) {
+                       dev_err(&spi->dev,
+                               "message rejected : "
+                               "device min speed (%d Hz) exceeds "
+                               "required transfer speed (%d Hz)\n",
+                               orion_spi->min_speed, t->speed_hz);
+                       goto msg_rejected;
+               }
+       }
+
+
+       spin_lock_irqsave(&orion_spi->lock, flags);
+       list_add_tail(&m->queue, &orion_spi->msg_queue);
+       queue_work(orion_spi_wq, &orion_spi->work);
+       spin_unlock_irqrestore(&orion_spi->lock, flags);
+
+       return 0;
+msg_rejected:
+       /* Message rejected and not queued */
+       m->status = -EINVAL;
+       if (m->complete)
+               m->complete(m->context);
+       return -EINVAL;
+}
+
+static int __init orion_spi_probe(struct platform_device *pdev)
+{
+       struct spi_master *master;
+       struct orion_spi *spi;
+       struct resource *r;
+       struct orion_spi_info *spi_info;
+       int status = 0;
+
+       spi_info = pdev->dev.platform_data;
+
+       master = spi_alloc_master(&pdev->dev, sizeof *spi);
+       if (master == NULL) {
+               dev_dbg(&pdev->dev, "master allocation failed\n");
+               return -ENOMEM;
+       }
+
+       if (pdev->id != -1)
+               master->bus_num = pdev->id;
+
+       master->setup = orion_spi_setup;
+       master->transfer = orion_spi_transfer;
+       master->num_chipselect = ORION_NUM_CHIPSELECTS;
+
+       dev_set_drvdata(&pdev->dev, master);
+
+       spi = spi_master_get_devdata(master);
+       spi->master = master;
+       spi->spi_info = spi_info;
+
+       spi->max_speed = DIV_ROUND_UP(spi_info->tclk, 4);
+       spi->min_speed = DIV_ROUND_UP(spi_info->tclk, 30);
+
+       r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       if (r == NULL) {
+               status = -ENODEV;
+               goto out;
+       }
+
+       if (!request_mem_region(r->start, (r->end - r->start) + 1,
+                               pdev->dev.bus_id)) {
+               status = -EBUSY;
+               goto out;
+       }
+       spi->base = ioremap(r->start, SZ_1K);
+
+       INIT_WORK(&spi->work, orion_spi_work);
+
+       spin_lock_init(&spi->lock);
+       INIT_LIST_HEAD(&spi->msg_queue);
+
+       if (orion_spi_reset(spi) < 0)
+               goto out_rel_mem;
+
+       status = spi_register_master(master);
+       if (status < 0)
+               goto out_rel_mem;
+
+       return status;
+
+out_rel_mem:
+       release_mem_region(r->start, (r->end - r->start) + 1);
+
+out:
+       spi_master_put(master);
+       return status;
+}
+
+
+static int __exit orion_spi_remove(struct platform_device *pdev)
+{
+       struct spi_master *master;
+       struct orion_spi *spi;
+       struct resource *r;
+
+       master = dev_get_drvdata(&pdev->dev);
+       spi = spi_master_get_devdata(master);
+
+       cancel_work_sync(&spi->work);
+
+       r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       release_mem_region(r->start, (r->end - r->start) + 1);
+
+       spi_unregister_master(master);
+
+       return 0;
+}
+
+MODULE_ALIAS("platform:" DRIVER_NAME);
+
+static struct platform_driver orion_spi_driver = {
+       .driver = {
+               .name   = DRIVER_NAME,
+               .owner  = THIS_MODULE,
+       },
+       .remove         = __exit_p(orion_spi_remove),
+};
+
+static int __init orion_spi_init(void)
+{
+       orion_spi_wq = create_singlethread_workqueue(
+                               orion_spi_driver.driver.name);
+       if (orion_spi_wq == NULL)
+               return -ENOMEM;
+
+       return platform_driver_probe(&orion_spi_driver, orion_spi_probe);
+}
+module_init(orion_spi_init);
+
+static void __exit orion_spi_exit(void)
+{
+       flush_workqueue(orion_spi_wq);
+       platform_driver_unregister(&orion_spi_driver);
+
+       destroy_workqueue(orion_spi_wq);
+}
+module_exit(orion_spi_exit);
+
+MODULE_DESCRIPTION("Orion SPI driver");
+MODULE_AUTHOR("Shadi Ammouri <shadi@marvell.com>");
+MODULE_LICENSE("GPL");
diff --git a/include/linux/spi/orion_spi.h b/include/linux/spi/orion_spi.h
new file mode 100644 (file)
index 0000000..b4d9fa6
--- /dev/null
@@ -0,0 +1,17 @@
+/*
+ * orion_spi.h
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2. This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
+
+#ifndef __LINUX_SPI_ORION_SPI_H
+#define __LINUX_SPI_ORION_SPI_H
+
+struct orion_spi_info {
+       u32     tclk;           /* no <linux/clk.h> support yet */
+};
+
+
+#endif