MIPS/OCTEON/ata: Convert pata_octeon_cf.c to use device tree.
authorDavid Daney <david.daney@cavium.com>
Thu, 26 Apr 2012 18:10:28 +0000 (11:10 -0700)
committerRalf Baechle <ralf@linux-mips.org>
Thu, 13 Dec 2012 17:15:24 +0000 (18:15 +0100)
The patch needs to eliminate the definition of OCTEON_IRQ_BOOTDMA so
that the device tree code can map the interrupt, so in order to not
temporarily break things, we do a single patch to both the interrupt
registration code and the pata_octeon_cf driver.

Also rolled in is a conversion to use hrtimers and corrections to the
timing calculations.

Acked-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: David Daney <david.daney@cavium.com>
arch/mips/cavium-octeon/octeon-irq.c
arch/mips/cavium-octeon/octeon-platform.c
arch/mips/include/asm/mach-cavium-octeon/irq.h
arch/mips/include/asm/octeon/octeon.h
drivers/ata/pata_octeon_cf.c

index 02b15ee..46f5dbc 100644 (file)
@@ -1266,7 +1266,6 @@ static void __init octeon_irq_init_ciu(void)
                octeon_irq_force_ciu_mapping(ciu_domain, i + OCTEON_IRQ_TIMER0, 0, i + 52);
 
        octeon_irq_force_ciu_mapping(ciu_domain, OCTEON_IRQ_USB0, 0, 56);
-       octeon_irq_force_ciu_mapping(ciu_domain, OCTEON_IRQ_BOOTDMA, 0, 63);
 
        /* CIU_1 */
        for (i = 0; i < 16; i++)
index 0938df1..3c1b625 100644 (file)
 #include <asm/octeon/cvmx-helper.h>
 #include <asm/octeon/cvmx-helper-board.h>
 
-static struct octeon_cf_data octeon_cf_data;
-
-static int __init octeon_cf_device_init(void)
-{
-       union cvmx_mio_boot_reg_cfgx mio_boot_reg_cfg;
-       unsigned long base_ptr, region_base, region_size;
-       struct platform_device *pd;
-       struct resource cf_resources[3];
-       unsigned int num_resources;
-       int i;
-       int ret = 0;
-
-       /* Setup octeon-cf platform device if present. */
-       base_ptr = 0;
-       if (octeon_bootinfo->major_version == 1
-               && octeon_bootinfo->minor_version >= 1) {
-               if (octeon_bootinfo->compact_flash_common_base_addr)
-                       base_ptr =
-                               octeon_bootinfo->compact_flash_common_base_addr;
-       } else {
-               base_ptr = 0x1d000800;
-       }
-
-       if (!base_ptr)
-               return ret;
-
-       /* Find CS0 region. */
-       for (i = 0; i < 8; i++) {
-               mio_boot_reg_cfg.u64 = cvmx_read_csr(CVMX_MIO_BOOT_REG_CFGX(i));
-               region_base = mio_boot_reg_cfg.s.base << 16;
-               region_size = (mio_boot_reg_cfg.s.size + 1) << 16;
-               if (mio_boot_reg_cfg.s.en && base_ptr >= region_base
-                   && base_ptr < region_base + region_size)
-                       break;
-       }
-       if (i >= 7) {
-               /* i and i + 1 are CS0 and CS1, both must be less than 8. */
-               goto out;
-       }
-       octeon_cf_data.base_region = i;
-       octeon_cf_data.is16bit = mio_boot_reg_cfg.s.width;
-       octeon_cf_data.base_region_bias = base_ptr - region_base;
-       memset(cf_resources, 0, sizeof(cf_resources));
-       num_resources = 0;
-       cf_resources[num_resources].flags       = IORESOURCE_MEM;
-       cf_resources[num_resources].start       = region_base;
-       cf_resources[num_resources].end = region_base + region_size - 1;
-       num_resources++;
-
-
-       if (!(base_ptr & 0xfffful)) {
-               /*
-                * Boot loader signals availability of DMA (true_ide
-                * mode) by setting low order bits of base_ptr to
-                * zero.
-                */
-
-               /* Assume that CS1 immediately follows. */
-               mio_boot_reg_cfg.u64 =
-                       cvmx_read_csr(CVMX_MIO_BOOT_REG_CFGX(i + 1));
-               region_base = mio_boot_reg_cfg.s.base << 16;
-               region_size = (mio_boot_reg_cfg.s.size + 1) << 16;
-               if (!mio_boot_reg_cfg.s.en)
-                       goto out;
-
-               cf_resources[num_resources].flags       = IORESOURCE_MEM;
-               cf_resources[num_resources].start       = region_base;
-               cf_resources[num_resources].end = region_base + region_size - 1;
-               num_resources++;
-
-               octeon_cf_data.dma_engine = 0;
-               cf_resources[num_resources].flags       = IORESOURCE_IRQ;
-               cf_resources[num_resources].start       = OCTEON_IRQ_BOOTDMA;
-               cf_resources[num_resources].end = OCTEON_IRQ_BOOTDMA;
-               num_resources++;
-       } else {
-               octeon_cf_data.dma_engine = -1;
-       }
-
-       pd = platform_device_alloc("pata_octeon_cf", -1);
-       if (!pd) {
-               ret = -ENOMEM;
-               goto out;
-       }
-       pd->dev.platform_data = &octeon_cf_data;
-
-       ret = platform_device_add_resources(pd, cf_resources, num_resources);
-       if (ret)
-               goto fail;
-
-       ret = platform_device_add(pd);
-       if (ret)
-               goto fail;
-
-       return ret;
-fail:
-       platform_device_put(pd);
-out:
-       return ret;
-}
-device_initcall(octeon_cf_device_init);
-
 /* Octeon Random Number Generator.  */
 static int __init octeon_rng_device_init(void)
 {
index ff0d490..502bb18 100644 (file)
@@ -42,7 +42,6 @@ enum octeon_irq {
        OCTEON_IRQ_TIMER3,
        OCTEON_IRQ_USB0,
        OCTEON_IRQ_USB1,
-       OCTEON_IRQ_BOOTDMA,
 #ifndef CONFIG_PCI_MSI
        OCTEON_IRQ_LAST = 127
 #endif
index 790939d..254e995 100644 (file)
@@ -209,13 +209,6 @@ union octeon_cvmemctl {
        } s;
 };
 
-struct octeon_cf_data {
-       unsigned long   base_region_bias;
-       unsigned int    base_region;    /* The chip select region used by CF */
-       int             is16bit;        /* 0 - 8bit, !0 - 16bit */
-       int             dma_engine;     /* -1 for no DMA */
-};
-
 extern void octeon_write_lcd(const char *s);
 extern void octeon_check_cpu_bist(void);
 extern int octeon_get_boot_debug_flag(void);
index 1d61d5d..652d035 100644 (file)
@@ -5,17 +5,19 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2005 - 2009 Cavium Networks
+ * Copyright (C) 2005 - 2012 Cavium Inc.
  * Copyright (C) 2008 Wind River Systems
  */
 
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/libata.h>
-#include <linux/irq.h>
+#include <linux/hrtimer.h>
 #include <linux/slab.h>
+#include <linux/irq.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
 #include <linux/platform_device.h>
-#include <linux/workqueue.h>
 #include <scsi/scsi_host.h>
 
 #include <asm/octeon/octeon.h>
  */
 
 #define DRV_NAME       "pata_octeon_cf"
-#define DRV_VERSION    "2.1"
+#define DRV_VERSION    "2.2"
+
+/* Poll interval in nS. */
+#define OCTEON_CF_BUSY_POLL_INTERVAL 500000
 
+#define DMA_CFG 0
+#define DMA_TIM 0x20
+#define DMA_INT 0x38
+#define DMA_INT_EN 0x50
 
 struct octeon_cf_port {
-       struct workqueue_struct *wq;
-       struct delayed_work delayed_finish;
+       struct hrtimer delayed_finish;
        struct ata_port *ap;
        int dma_finished;
+       void            *c0;
+       unsigned int cs0;
+       unsigned int cs1;
+       bool is_true_ide;
+       u64 dma_base;
 };
 
 static struct scsi_host_template octeon_cf_sht = {
        ATA_PIO_SHT(DRV_NAME),
 };
 
+static int enable_dma;
+module_param(enable_dma, int, 0444);
+MODULE_PARM_DESC(enable_dma,
+                "Enable use of DMA on interfaces that support it (0=no dma [default], 1=use dma)");
+
 /**
  * Convert nanosecond based time to setting used in the
  * boot bus timing register, based on timing multiple
@@ -66,12 +84,29 @@ static unsigned int ns_to_tim_reg(unsigned int tim_mult, unsigned int nsecs)
        return val;
 }
 
-static void octeon_cf_set_boot_reg_cfg(int cs)
+static void octeon_cf_set_boot_reg_cfg(int cs, unsigned int multiplier)
 {
        union cvmx_mio_boot_reg_cfgx reg_cfg;
+       unsigned int tim_mult;
+
+       switch (multiplier) {
+       case 8:
+               tim_mult = 3;
+               break;
+       case 4:
+               tim_mult = 0;
+               break;
+       case 2:
+               tim_mult = 2;
+               break;
+       default:
+               tim_mult = 1;
+               break;
+       }
+
        reg_cfg.u64 = cvmx_read_csr(CVMX_MIO_BOOT_REG_CFGX(cs));
        reg_cfg.s.dmack = 0;    /* Don't assert DMACK on access */
-       reg_cfg.s.tim_mult = 2; /* Timing mutiplier 2x */
+       reg_cfg.s.tim_mult = tim_mult;  /* Timing mutiplier */
        reg_cfg.s.rd_dly = 0;   /* Sample on falling edge of BOOT_OE */
        reg_cfg.s.sam = 0;      /* Don't combine write and output enable */
        reg_cfg.s.we_ext = 0;   /* No write enable extension */
@@ -92,12 +127,12 @@ static void octeon_cf_set_boot_reg_cfg(int cs)
  */
 static void octeon_cf_set_piomode(struct ata_port *ap, struct ata_device *dev)
 {
-       struct octeon_cf_data *ocd = ap->dev->platform_data;
+       struct octeon_cf_port *cf_port = ap->private_data;
        union cvmx_mio_boot_reg_timx reg_tim;
-       int cs = ocd->base_region;
        int T;
        struct ata_timing timing;
 
+       unsigned int div;
        int use_iordy;
        int trh;
        int pause;
@@ -106,7 +141,15 @@ static void octeon_cf_set_piomode(struct ata_port *ap, struct ata_device *dev)
        int t2;
        int t2i;
 
-       T = (int)(2000000000000LL / octeon_get_clock_rate());
+       /*
+        * A divisor value of four will overflow the timing fields at
+        * clock rates greater than 800MHz
+        */
+       if (octeon_get_io_clock_rate() <= 800000000)
+               div = 4;
+       else
+               div = 8;
+       T = (int)((1000000000000LL * div) / octeon_get_io_clock_rate());
 
        if (ata_timing_compute(dev, dev->pio_mode, &timing, T, T))
                BUG();
@@ -121,23 +164,26 @@ static void octeon_cf_set_piomode(struct ata_port *ap, struct ata_device *dev)
        if (t2i)
                t2i--;
 
-       trh = ns_to_tim_reg(2, 20);
+       trh = ns_to_tim_reg(div, 20);
        if (trh)
                trh--;
 
-       pause = timing.cycle - timing.active - timing.setup - trh;
+       pause = (int)timing.cycle - (int)timing.active -
+               (int)timing.setup - trh;
+       if (pause < 0)
+               pause = 0;
        if (pause)
                pause--;
 
-       octeon_cf_set_boot_reg_cfg(cs);
-       if (ocd->dma_engine >= 0)
+       octeon_cf_set_boot_reg_cfg(cf_port->cs0, div);
+       if (cf_port->is_true_ide)
                /* True IDE mode, program both chip selects.  */
-               octeon_cf_set_boot_reg_cfg(cs + 1);
+               octeon_cf_set_boot_reg_cfg(cf_port->cs1, div);
 
 
        use_iordy = ata_pio_need_iordy(dev);
 
-       reg_tim.u64 = cvmx_read_csr(CVMX_MIO_BOOT_REG_TIMX(cs));
+       reg_tim.u64 = cvmx_read_csr(CVMX_MIO_BOOT_REG_TIMX(cf_port->cs0));
        /* Disable page mode */
        reg_tim.s.pagem = 0;
        /* Enable dynamic timing */
@@ -161,20 +207,22 @@ static void octeon_cf_set_piomode(struct ata_port *ap, struct ata_device *dev)
        /* How long read enable is asserted */
        reg_tim.s.oe = t2;
        /* Time after CE that read/write starts */
-       reg_tim.s.ce = ns_to_tim_reg(2, 5);
+       reg_tim.s.ce = ns_to_tim_reg(div, 5);
        /* Time before CE that address is valid */
        reg_tim.s.adr = 0;
 
        /* Program the bootbus region timing for the data port chip select. */
-       cvmx_write_csr(CVMX_MIO_BOOT_REG_TIMX(cs), reg_tim.u64);
-       if (ocd->dma_engine >= 0)
+       cvmx_write_csr(CVMX_MIO_BOOT_REG_TIMX(cf_port->cs0), reg_tim.u64);
+       if (cf_port->is_true_ide)
                /* True IDE mode, program both chip selects.  */
-               cvmx_write_csr(CVMX_MIO_BOOT_REG_TIMX(cs + 1), reg_tim.u64);
+               cvmx_write_csr(CVMX_MIO_BOOT_REG_TIMX(cf_port->cs1),
+                              reg_tim.u64);
 }
 
 static void octeon_cf_set_dmamode(struct ata_port *ap, struct ata_device *dev)
 {
-       struct octeon_cf_data *ocd = dev->link->ap->dev->platform_data;
+       struct octeon_cf_port *cf_port = ap->private_data;
+       union cvmx_mio_boot_pin_defs pin_defs;
        union cvmx_mio_boot_dma_timx dma_tim;
        unsigned int oe_a;
        unsigned int oe_n;
@@ -183,6 +231,7 @@ static void octeon_cf_set_dmamode(struct ata_port *ap, struct ata_device *dev)
        unsigned int pause;
        unsigned int T0, Tkr, Td;
        unsigned int tim_mult;
+       int c;
 
        const struct ata_timing *timing;
 
@@ -199,13 +248,19 @@ static void octeon_cf_set_dmamode(struct ata_port *ap, struct ata_device *dev)
        /* not spec'ed, value in eclocks, not affected by tim_mult */
        dma_arq = 8;
        pause = 25 - dma_arq * 1000 /
-               (octeon_get_clock_rate() / 1000000); /* Tz */
+               (octeon_get_io_clock_rate() / 1000000); /* Tz */
 
        oe_a = Td;
        /* Tkr from cf spec, lengthened to meet T0 */
        oe_n = max(T0 - oe_a, Tkr);
 
-       dma_tim.s.dmack_pi = 1;
+       pin_defs.u64 = cvmx_read_csr(CVMX_MIO_BOOT_PIN_DEFS);
+
+       /* DMA channel number. */
+       c = (cf_port->dma_base & 8) >> 3;
+
+       /* Invert the polarity if the default is 0*/
+       dma_tim.s.dmack_pi = (pin_defs.u64 & (1ull << (11 + c))) ? 0 : 1;
 
        dma_tim.s.oe_n = ns_to_tim_reg(tim_mult, oe_n);
        dma_tim.s.oe_a = ns_to_tim_reg(tim_mult, oe_a);
@@ -228,14 +283,11 @@ static void octeon_cf_set_dmamode(struct ata_port *ap, struct ata_device *dev)
 
        pr_debug("ns to ticks (mult %d) of %d is: %d\n", tim_mult, 60,
                 ns_to_tim_reg(tim_mult, 60));
-       pr_debug("oe_n: %d, oe_a: %d, dmack_s: %d, dmack_h: "
-                "%d, dmarq: %d, pause: %d\n",
+       pr_debug("oe_n: %d, oe_a: %d, dmack_s: %d, dmack_h: %d, dmarq: %d, pause: %d\n",
                 dma_tim.s.oe_n, dma_tim.s.oe_a, dma_tim.s.dmack_s,
                 dma_tim.s.dmack_h, dma_tim.s.dmarq, dma_tim.s.pause);
 
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_TIMX(ocd->dma_engine),
-                      dma_tim.u64);
-
+       cvmx_write_csr(cf_port->dma_base + DMA_TIM, dma_tim.u64);
 }
 
 /**
@@ -489,15 +541,10 @@ static void octeon_cf_exec_command16(struct ata_port *ap,
        ata_wait_idle(ap);
 }
 
-static void octeon_cf_irq_on(struct ata_port *ap)
+static void octeon_cf_ata_port_noaction(struct ata_port *ap)
 {
 }
 
-static void octeon_cf_irq_clear(struct ata_port *ap)
-{
-       return;
-}
-
 static void octeon_cf_dma_setup(struct ata_queued_cmd *qc)
 {
        struct ata_port *ap = qc->ap;
@@ -519,7 +566,7 @@ static void octeon_cf_dma_setup(struct ata_queued_cmd *qc)
  */
 static void octeon_cf_dma_start(struct ata_queued_cmd *qc)
 {
-       struct octeon_cf_data *ocd = qc->ap->dev->platform_data;
+       struct octeon_cf_port *cf_port = qc->ap->private_data;
        union cvmx_mio_boot_dma_cfgx mio_boot_dma_cfg;
        union cvmx_mio_boot_dma_intx mio_boot_dma_int;
        struct scatterlist *sg;
@@ -535,12 +582,10 @@ static void octeon_cf_dma_start(struct ata_queued_cmd *qc)
         */
        mio_boot_dma_int.u64 = 0;
        mio_boot_dma_int.s.done = 1;
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_INTX(ocd->dma_engine),
-                      mio_boot_dma_int.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_INT, mio_boot_dma_int.u64);
 
        /* Enable the interrupt.  */
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_INT_ENX(ocd->dma_engine),
-                      mio_boot_dma_int.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_INT_EN, mio_boot_dma_int.u64);
 
        /* Set the direction of the DMA */
        mio_boot_dma_cfg.u64 = 0;
@@ -569,8 +614,7 @@ static void octeon_cf_dma_start(struct ata_queued_cmd *qc)
                (mio_boot_dma_cfg.s.rw) ? "write" : "read", sg->length,
                (void *)(unsigned long)mio_boot_dma_cfg.s.adr);
 
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_CFGX(ocd->dma_engine),
-                      mio_boot_dma_cfg.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_CFG, mio_boot_dma_cfg.u64);
 }
 
 /**
@@ -583,10 +627,9 @@ static unsigned int octeon_cf_dma_finished(struct ata_port *ap,
                                        struct ata_queued_cmd *qc)
 {
        struct ata_eh_info *ehi = &ap->link.eh_info;
-       struct octeon_cf_data *ocd = ap->dev->platform_data;
+       struct octeon_cf_port *cf_port = ap->private_data;
        union cvmx_mio_boot_dma_cfgx dma_cfg;
        union cvmx_mio_boot_dma_intx dma_int;
-       struct octeon_cf_port *cf_port;
        u8 status;
 
        VPRINTK("ata%u: protocol %d task_state %d\n",
@@ -596,9 +639,7 @@ static unsigned int octeon_cf_dma_finished(struct ata_port *ap,
        if (ap->hsm_task_state != HSM_ST_LAST)
                return 0;
 
-       cf_port = ap->private_data;
-
-       dma_cfg.u64 = cvmx_read_csr(CVMX_MIO_BOOT_DMA_CFGX(ocd->dma_engine));
+       dma_cfg.u64 = cvmx_read_csr(cf_port->dma_base + DMA_CFG);
        if (dma_cfg.s.size != 0xfffff) {
                /* Error, the transfer was not complete.  */
                qc->err_mask |= AC_ERR_HOST_BUS;
@@ -608,15 +649,15 @@ static unsigned int octeon_cf_dma_finished(struct ata_port *ap,
        /* Stop and clear the dma engine.  */
        dma_cfg.u64 = 0;
        dma_cfg.s.size = -1;
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_CFGX(ocd->dma_engine), dma_cfg.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_CFG, dma_cfg.u64);
 
        /* Disable the interrupt.  */
        dma_int.u64 = 0;
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_INT_ENX(ocd->dma_engine), dma_int.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_INT_EN, dma_int.u64);
 
        /* Clear the DMA complete status */
        dma_int.s.done = 1;
-       cvmx_write_csr(CVMX_MIO_BOOT_DMA_INTX(ocd->dma_engine), dma_int.u64);
+       cvmx_write_csr(cf_port->dma_base + DMA_INT, dma_int.u64);
 
        status = ap->ops->sff_check_status(ap);
 
@@ -649,69 +690,68 @@ static irqreturn_t octeon_cf_interrupt(int irq, void *dev_instance)
                struct ata_queued_cmd *qc;
                union cvmx_mio_boot_dma_intx dma_int;
                union cvmx_mio_boot_dma_cfgx dma_cfg;
-               struct octeon_cf_data *ocd;
 
                ap = host->ports[i];
-               ocd = ap->dev->platform_data;
                cf_port = ap->private_data;
-               dma_int.u64 =
-                       cvmx_read_csr(CVMX_MIO_BOOT_DMA_INTX(ocd->dma_engine));
-               dma_cfg.u64 =
-                       cvmx_read_csr(CVMX_MIO_BOOT_DMA_CFGX(ocd->dma_engine));
+
+               dma_int.u64 = cvmx_read_csr(cf_port->dma_base + DMA_INT);
+               dma_cfg.u64 = cvmx_read_csr(cf_port->dma_base + DMA_CFG);
 
                qc = ata_qc_from_tag(ap, ap->link.active_tag);
 
-               if (qc && !(qc->tf.flags & ATA_TFLAG_POLLING)) {
-                       if (dma_int.s.done && !dma_cfg.s.en) {
-                               if (!sg_is_last(qc->cursg)) {
-                                       qc->cursg = sg_next(qc->cursg);
-                                       handled = 1;
-                                       octeon_cf_dma_start(qc);
-                                       continue;
-                               } else {
-                                       cf_port->dma_finished = 1;
-                               }
-                       }
-                       if (!cf_port->dma_finished)
-                               continue;
-                       status = ioread8(ap->ioaddr.altstatus_addr);
-                       if (status & (ATA_BUSY | ATA_DRQ)) {
-                               /*
-                                * We are busy, try to handle it
-                                * later.  This is the DMA finished
-                                * interrupt, and it could take a
-                                * little while for the card to be
-                                * ready for more commands.
-                                */
-                               /* Clear DMA irq. */
-                               dma_int.u64 = 0;
-                               dma_int.s.done = 1;
-                               cvmx_write_csr(CVMX_MIO_BOOT_DMA_INTX(ocd->dma_engine),
-                                              dma_int.u64);
-
-                               queue_delayed_work(cf_port->wq,
-                                                  &cf_port->delayed_finish, 1);
+               if (!qc || (qc->tf.flags & ATA_TFLAG_POLLING))
+                       continue;
+
+               if (dma_int.s.done && !dma_cfg.s.en) {
+                       if (!sg_is_last(qc->cursg)) {
+                               qc->cursg = sg_next(qc->cursg);
                                handled = 1;
+                               octeon_cf_dma_start(qc);
+                               continue;
                        } else {
-                               handled |= octeon_cf_dma_finished(ap, qc);
+                               cf_port->dma_finished = 1;
                        }
                }
+               if (!cf_port->dma_finished)
+                       continue;
+               status = ioread8(ap->ioaddr.altstatus_addr);
+               if (status & (ATA_BUSY | ATA_DRQ)) {
+                       /*
+                        * We are busy, try to handle it later.  This
+                        * is the DMA finished interrupt, and it could
+                        * take a little while for the card to be
+                        * ready for more commands.
+                        */
+                       /* Clear DMA irq. */
+                       dma_int.u64 = 0;
+                       dma_int.s.done = 1;
+                       cvmx_write_csr(cf_port->dma_base + DMA_INT,
+                                      dma_int.u64);
+                       hrtimer_start_range_ns(&cf_port->delayed_finish,
+                                              ns_to_ktime(OCTEON_CF_BUSY_POLL_INTERVAL),
+                                              OCTEON_CF_BUSY_POLL_INTERVAL / 5,
+                                              HRTIMER_MODE_REL);
+                       handled = 1;
+               } else {
+                       handled |= octeon_cf_dma_finished(ap, qc);
+               }
        }
        spin_unlock_irqrestore(&host->lock, flags);
        DPRINTK("EXIT\n");
        return IRQ_RETVAL(handled);
 }
 
-static void octeon_cf_delayed_finish(struct work_struct *work)
+static enum hrtimer_restart octeon_cf_delayed_finish(struct hrtimer *hrt)
 {
-       struct octeon_cf_port *cf_port = container_of(work,
+       struct octeon_cf_port *cf_port = container_of(hrt,
                                                      struct octeon_cf_port,
-                                                     delayed_finish.work);
+                                                     delayed_finish);
        struct ata_port *ap = cf_port->ap;
        struct ata_host *host = ap->host;
        struct ata_queued_cmd *qc;
        unsigned long flags;
        u8 status;
+       enum hrtimer_restart rv = HRTIMER_NORESTART;
 
        spin_lock_irqsave(&host->lock, flags);
 
@@ -726,15 +766,17 @@ static void octeon_cf_delayed_finish(struct work_struct *work)
        status = ioread8(ap->ioaddr.altstatus_addr);
        if (status & (ATA_BUSY | ATA_DRQ)) {
                /* Still busy, try again. */
-               queue_delayed_work(cf_port->wq,
-                                  &cf_port->delayed_finish, 1);
+               hrtimer_forward_now(hrt,
+                                   ns_to_ktime(OCTEON_CF_BUSY_POLL_INTERVAL));
+               rv = HRTIMER_RESTART;
                goto out;
        }
        qc = ata_qc_from_tag(ap, ap->link.active_tag);
-       if (qc && !(qc->tf.flags & ATA_TFLAG_POLLING))
+       if (qc && (!(qc->tf.flags & ATA_TFLAG_POLLING)))
                octeon_cf_dma_finished(ap, qc);
 out:
        spin_unlock_irqrestore(&host->lock, flags);
+       return rv;
 }
 
 static void octeon_cf_dev_config(struct ata_device *dev)
@@ -786,8 +828,8 @@ static struct ata_port_operations octeon_cf_ops = {
        .qc_prep                = ata_noop_qc_prep,
        .qc_issue               = octeon_cf_qc_issue,
        .sff_dev_select         = octeon_cf_dev_select,
-       .sff_irq_on             = octeon_cf_irq_on,
-       .sff_irq_clear          = octeon_cf_irq_clear,
+       .sff_irq_on             = octeon_cf_ata_port_noaction,
+       .sff_irq_clear          = octeon_cf_ata_port_noaction,
        .cable_detect           = ata_cable_40wire,
        .set_piomode            = octeon_cf_set_piomode,
        .set_dmamode            = octeon_cf_set_dmamode,
@@ -798,46 +840,113 @@ static int __devinit octeon_cf_probe(struct platform_device *pdev)
 {
        struct resource *res_cs0, *res_cs1;
 
+       bool is_16bit;
+       const __be32 *cs_num;
+       struct property *reg_prop;
+       int n_addr, n_size, reg_len;
+       struct device_node *node;
+       const void *prop;
        void __iomem *cs0;
        void __iomem *cs1 = NULL;
        struct ata_host *host;
        struct ata_port *ap;
-       struct octeon_cf_data *ocd;
        int irq = 0;
        irq_handler_t irq_handler = NULL;
        void __iomem *base;
        struct octeon_cf_port *cf_port;
-       char version[32];
+       int rv = -ENOMEM;
 
-       res_cs0 = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 
-       if (!res_cs0)
+       node = pdev->dev.of_node;
+       if (node == NULL)
                return -EINVAL;
 
-       ocd = pdev->dev.platform_data;
+       cf_port = kzalloc(sizeof(*cf_port), GFP_KERNEL);
+       if (!cf_port)
+               return -ENOMEM;
 
-       cs0 = devm_ioremap_nocache(&pdev->dev, res_cs0->start,
-                                  resource_size(res_cs0));
+       cf_port->is_true_ide = (of_find_property(node, "cavium,true-ide", NULL) != NULL);
 
-       if (!cs0)
-               return -ENOMEM;
+       prop = of_get_property(node, "cavium,bus-width", NULL);
+       if (prop)
+               is_16bit = (be32_to_cpup(prop) == 16);
+       else
+               is_16bit = false;
 
-       /* Determine from availability of DMA if True IDE mode or not */
-       if (ocd->dma_engine >= 0) {
-               res_cs1 = platform_get_resource(pdev, IORESOURCE_MEM, 1);
-               if (!res_cs1)
-                       return -EINVAL;
+       n_addr = of_n_addr_cells(node);
+       n_size = of_n_size_cells(node);
 
+       reg_prop = of_find_property(node, "reg", &reg_len);
+       if (!reg_prop || reg_len < sizeof(__be32)) {
+               rv = -EINVAL;
+               goto free_cf_port;
+       }
+       cs_num = reg_prop->value;
+       cf_port->cs0 = be32_to_cpup(cs_num);
+
+       if (cf_port->is_true_ide) {
+               struct device_node *dma_node;
+               dma_node = of_parse_phandle(node,
+                                           "cavium,dma-engine-handle", 0);
+               if (dma_node) {
+                       struct platform_device *dma_dev;
+                       dma_dev = of_find_device_by_node(dma_node);
+                       if (dma_dev) {
+                               struct resource *res_dma;
+                               int i;
+                               res_dma = platform_get_resource(dma_dev, IORESOURCE_MEM, 0);
+                               if (!res_dma) {
+                                       of_node_put(dma_node);
+                                       rv = -EINVAL;
+                                       goto free_cf_port;
+                               }
+                               cf_port->dma_base = (u64)devm_ioremap_nocache(&pdev->dev, res_dma->start,
+                                                                        resource_size(res_dma));
+
+                               if (!cf_port->dma_base) {
+                                       of_node_put(dma_node);
+                                       rv = -EINVAL;
+                                       goto free_cf_port;
+                               }
+
+                               irq_handler = octeon_cf_interrupt;
+                               i = platform_get_irq(dma_dev, 0);
+                               if (i > 0)
+                                       irq = i;
+                       }
+                       of_node_put(dma_node);
+               }
+               res_cs1 = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+               if (!res_cs1) {
+                       rv = -EINVAL;
+                       goto free_cf_port;
+               }
                cs1 = devm_ioremap_nocache(&pdev->dev, res_cs1->start,
-                                          resource_size(res_cs1));
+                                          res_cs1->end - res_cs1->start + 1);
 
                if (!cs1)
-                       return -ENOMEM;
+                       goto free_cf_port;
+
+               if (reg_len < (n_addr + n_size + 1) * sizeof(__be32)) {
+                       rv = -EINVAL;
+                       goto free_cf_port;
+               }
+               cs_num += n_addr + n_size;
+               cf_port->cs1 = be32_to_cpup(cs_num);
        }
 
-       cf_port = kzalloc(sizeof(*cf_port), GFP_KERNEL);
-       if (!cf_port)
-               return -ENOMEM;
+       res_cs0 = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+
+       if (!res_cs0) {
+               rv = -EINVAL;
+               goto free_cf_port;
+       }
+
+       cs0 = devm_ioremap_nocache(&pdev->dev, res_cs0->start,
+                                  resource_size(res_cs0));
+
+       if (!cs0)
+               goto free_cf_port;
 
        /* allocate host */
        host = ata_host_alloc(&pdev->dev, 1);
@@ -846,21 +955,22 @@ static int __devinit octeon_cf_probe(struct platform_device *pdev)
 
        ap = host->ports[0];
        ap->private_data = cf_port;
+       pdev->dev.platform_data = cf_port;
        cf_port->ap = ap;
        ap->ops = &octeon_cf_ops;
        ap->pio_mask = ATA_PIO6;
        ap->flags |= ATA_FLAG_NO_ATAPI | ATA_FLAG_PIO_POLLING;
 
-       base = cs0 + ocd->base_region_bias;
-       if (!ocd->is16bit) {
+       if (!is_16bit) {
+               base = cs0 + 0x800;
                ap->ioaddr.cmd_addr     = base;
                ata_sff_std_ports(&ap->ioaddr);
 
                ap->ioaddr.altstatus_addr = base + 0xe;
                ap->ioaddr.ctl_addr     = base + 0xe;
                octeon_cf_ops.sff_data_xfer = octeon_cf_data_xfer8;
-       } else if (cs1) {
-               /* Presence of cs1 indicates True IDE mode.  */
+       } else if (cf_port->is_true_ide) {
+               base = cs0;
                ap->ioaddr.cmd_addr     = base + (ATA_REG_CMD << 1) + 1;
                ap->ioaddr.data_addr    = base + (ATA_REG_DATA << 1);
                ap->ioaddr.error_addr   = base + (ATA_REG_ERR << 1) + 1;
@@ -876,19 +986,15 @@ static int __devinit octeon_cf_probe(struct platform_device *pdev)
                ap->ioaddr.ctl_addr     = cs1 + (6 << 1) + 1;
                octeon_cf_ops.sff_data_xfer = octeon_cf_data_xfer16;
 
-               ap->mwdma_mask  = ATA_MWDMA4;
-               irq = platform_get_irq(pdev, 0);
-               irq_handler = octeon_cf_interrupt;
-
-               /* True IDE mode needs delayed work to poll for not-busy.  */
-               cf_port->wq = create_singlethread_workqueue(DRV_NAME);
-               if (!cf_port->wq)
-                       goto free_cf_port;
-               INIT_DELAYED_WORK(&cf_port->delayed_finish,
-                                 octeon_cf_delayed_finish);
+               ap->mwdma_mask  = enable_dma ? ATA_MWDMA4 : 0;
 
+               /* True IDE mode needs a timer to poll for not-busy.  */
+               hrtimer_init(&cf_port->delayed_finish, CLOCK_MONOTONIC,
+                            HRTIMER_MODE_REL);
+               cf_port->delayed_finish.function = octeon_cf_delayed_finish;
        } else {
                /* 16 bit but not True IDE */
+               base = cs0 + 0x800;
                octeon_cf_ops.sff_data_xfer     = octeon_cf_data_xfer16;
                octeon_cf_ops.softreset         = octeon_cf_softreset16;
                octeon_cf_ops.sff_check_status  = octeon_cf_check_status16;
@@ -902,28 +1008,71 @@ static int __devinit octeon_cf_probe(struct platform_device *pdev)
                ap->ioaddr.ctl_addr     = base + 0xe;
                ap->ioaddr.altstatus_addr = base + 0xe;
        }
+       cf_port->c0 = ap->ioaddr.ctl_addr;
+
+       pdev->dev.coherent_dma_mask = DMA_BIT_MASK(64);
+       pdev->dev.dma_mask = &pdev->dev.coherent_dma_mask;
 
        ata_port_desc(ap, "cmd %p ctl %p", base, ap->ioaddr.ctl_addr);
 
 
-       snprintf(version, sizeof(version), "%s %d bit%s",
-                DRV_VERSION,
-                (ocd->is16bit) ? 16 : 8,
-                (cs1) ? ", True IDE" : "");
-       ata_print_version_once(&pdev->dev, version);
+       dev_info(&pdev->dev, "version " DRV_VERSION" %d bit%s.\n",
+                is_16bit ? 16 : 8,
+                cf_port->is_true_ide ? ", True IDE" : "");
 
-       return ata_host_activate(host, irq, irq_handler, 0, &octeon_cf_sht);
+       return ata_host_activate(host, irq, irq_handler,
+                                IRQF_SHARED, &octeon_cf_sht);
 
 free_cf_port:
        kfree(cf_port);
-       return -ENOMEM;
+       return rv;
+}
+
+static void octeon_cf_shutdown(struct device *dev)
+{
+       union cvmx_mio_boot_dma_cfgx dma_cfg;
+       union cvmx_mio_boot_dma_intx dma_int;
+
+       struct octeon_cf_port *cf_port = dev->platform_data;
+
+       if (cf_port->dma_base) {
+               /* Stop and clear the dma engine.  */
+               dma_cfg.u64 = 0;
+               dma_cfg.s.size = -1;
+               cvmx_write_csr(cf_port->dma_base + DMA_CFG, dma_cfg.u64);
+
+               /* Disable the interrupt.  */
+               dma_int.u64 = 0;
+               cvmx_write_csr(cf_port->dma_base + DMA_INT_EN, dma_int.u64);
+
+               /* Clear the DMA complete status */
+               dma_int.s.done = 1;
+               cvmx_write_csr(cf_port->dma_base + DMA_INT, dma_int.u64);
+
+               __raw_writeb(0, cf_port->c0);
+               udelay(20);
+               __raw_writeb(ATA_SRST, cf_port->c0);
+               udelay(20);
+               __raw_writeb(0, cf_port->c0);
+               mdelay(100);
+       }
 }
 
+static struct of_device_id octeon_cf_match[] = {
+       {
+               .compatible = "cavium,ebt3000-compact-flash",
+       },
+       {},
+};
+MODULE_DEVICE_TABLE(of, octeon_i2c_match);
+
 static struct platform_driver octeon_cf_driver = {
        .probe          = octeon_cf_probe,
        .driver         = {
                .name   = DRV_NAME,
                .owner  = THIS_MODULE,
+               .of_match_table = octeon_cf_match,
+               .shutdown = octeon_cf_shutdown
        },
 };