mflash: initial support
unsik Kim [Thu, 2 Apr 2009 19:50:58 +0000 (12:50 -0700)]
This driver supports mflash IO mode for linux.

Mflash is embedded flash drive and mainly targeted mobile and consumer
electronic devices.

Internally, mflash has nand flash and other hardware logics and supports 2
different operation (ATA, IO) modes.  ATA mode doesn't need any new driver
and currently works well under standard IDE subsystem.  Actually it's one
chip SSD.  IO mode is ATA-like custom mode for the host that doesn't have
IDE interface.

Followings are brief descriptions about IO mode.
A. IO mode based on ATA protocol and uses some custom command. (read confirm,
write confirm)
B. IO mode uses SRAM bus interface.
C. IO mode supports 4kB boot area, so host can boot from mflash.

This driver is quitely similar to a standard ATA driver, but because of
following reasons it is currently seperated with ATA layer.

1. ATA layer deals standard ATA protocol.  ATA layer have many low-
   level device specific interface, but data transfer keeps ATA rule.
   But, mflash IO mode doesn't.

2. Even though currently not used in mflash driver code, mflash has
   some custom command and modes.  (nand fusing, firmware patch, etc) If
   this feature supported in linux kernel, ATA layer more altered.

3. Currently PATA platform device driver doesn't support interrupt.
   (I'm not sure) But, mflash uses interrupt (polling mode is just for
   debug).

4. mflash is somewhat under-develop product.  Even though some company
   already using mflash their own product, I think more time is needed for
   standardization of custom command and mode.  That time (maybe October)
   I will talk to with ATA people.  If they accept integration, I will
   integrate.

Signed-off-by: unsik Kim <donari75@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Documentation/blockdev/00-INDEX
Documentation/blockdev/mflash.txt [new file with mode: 0644]
drivers/block/Kconfig
drivers/block/Makefile
drivers/block/mg_disk.c [new file with mode: 0644]
include/linux/mg_disk.h [new file with mode: 0644]

index 86f054c..c08df56 100644 (file)
@@ -8,6 +8,8 @@ cpqarray.txt
        - info on using Compaq's SMART2 Intelligent Disk Array Controllers.
 floppy.txt
        - notes and driver options for the floppy disk driver.
+mflash.txt
+       - info on mGine m(g)flash driver for linux.
 nbd.txt
        - info on a TCP implementation of a network block device.
 paride.txt
diff --git a/Documentation/blockdev/mflash.txt b/Documentation/blockdev/mflash.txt
new file mode 100644 (file)
index 0000000..1f610ec
--- /dev/null
@@ -0,0 +1,84 @@
+This document describes m[g]flash support in linux.
+
+Contents
+  1. Overview
+  2. Reserved area configuration
+  3. Example of mflash platform driver registration
+
+1. Overview
+
+Mflash and gflash are embedded flash drive. The only difference is mflash is
+MCP(Multi Chip Package) device. These two device operate exactly same way.
+So the rest mflash repersents mflash and gflash altogether.
+
+Internally, mflash has nand flash and other hardware logics and supports
+2 different operation (ATA, IO) modes. ATA mode doesn't need any new
+driver and currently works well under standard IDE subsystem. Actually it's
+one chip SSD. IO mode is ATA-like custom mode for the host that doesn't have
+IDE interface.
+
+Followings are brief descriptions about IO mode.
+A. IO mode based on ATA protocol and uses some custom command. (read confirm,
+write confirm)
+B. IO mode uses SRAM bus interface.
+C. IO mode supports 4kB boot area, so host can boot from mflash.
+
+2. Reserved area configuration
+If host boot from mflash, usually needs raw area for boot loader image. All of
+the mflash's block device operation will be taken this value as start offset.
+Note that boot loader's size of reserved area and kernel configuration value
+must be same.
+
+3. Example of mflash platform driver registration
+Working mflash is very straight forward. Adding platform device stuff to board
+configuration file is all. Here is some pseudo example.
+
+static struct mg_drv_data mflash_drv_data = {
+       /* If you want to polling driver set to 1 */
+       .use_polling = 0,
+       /* device attribution */
+       .dev_attr = MG_BOOT_DEV
+};
+
+static struct resource mg_mflash_rsc[] = {
+       /* Base address of mflash */
+       [0] = {
+               .start = 0x08000000,
+               .end = 0x08000000 + SZ_64K - 1,
+               .flags = IORESOURCE_MEM
+       },
+       /* mflash interrupt pin */
+       [1] = {
+               .start = IRQ_GPIO(84),
+               .end = IRQ_GPIO(84),
+               .flags = IORESOURCE_IRQ
+       },
+       /* mflash reset pin */
+       [2] = {
+               .start = 43,
+               .end = 43,
+               .name = MG_RST_PIN,
+               .flags = IORESOURCE_IO
+       },
+       /* mflash reset-out pin
+        * If you use mflash as storage device (i.e. other than MG_BOOT_DEV),
+        * should assign this */
+       [3] = {
+               .start = 51,
+               .end = 51,
+               .name = MG_RSTOUT_PIN,
+               .flags = IORESOURCE_IO
+       }
+};
+
+static struct platform_device mflash_dev = {
+       .name = MG_DEV_NAME,
+       .id = -1,
+       .dev = {
+               .platform_data = &mflash_drv_data,
+       },
+       .num_resources = ARRAY_SIZE(mg_mflash_rsc),
+       .resource = mg_mflash_rsc
+};
+
+platform_device_register(&mflash_dev);
index e7b8aa0..ddea8e4 100644 (file)
@@ -410,6 +410,23 @@ config ATA_OVER_ETH
        This driver provides Support for ATA over Ethernet block
        devices like the Coraid EtherDrive (R) Storage Blade.
 
+config MG_DISK
+       tristate "mGine mflash, gflash support"
+       depends on ARM && ATA && GPIOLIB
+       help
+         mGine mFlash(gFlash) block device driver
+
+config MG_DISK_RES
+       int "Size of reserved area before MBR"
+       depends on MG_DISK
+       default 0
+       help
+         Define size of reserved area that usually used for boot. Unit is KB.
+         All of the block device operation will be taken this value as start
+         offset
+         Examples:
+                       1024 => 1 MB
+
 config SUNVDC
        tristate "Sun Virtual Disk Client support"
        depends on SUN_LDOMS
index 3145141..7755a5e 100644 (file)
@@ -21,6 +21,7 @@ obj-$(CONFIG_BLK_CPQ_CISS_DA)  += cciss.o
 obj-$(CONFIG_BLK_DEV_DAC960)   += DAC960.o
 obj-$(CONFIG_XILINX_SYSACE)    += xsysace.o
 obj-$(CONFIG_CDROM_PKTCDVD)    += pktcdvd.o
+obj-$(CONFIG_MG_DISK)          += mg_disk.o
 obj-$(CONFIG_SUNVDC)           += sunvdc.o
 
 obj-$(CONFIG_BLK_DEV_UMEM)     += umem.o
diff --git a/drivers/block/mg_disk.c b/drivers/block/mg_disk.c
new file mode 100644 (file)
index 0000000..fb39d9a
--- /dev/null
@@ -0,0 +1,1005 @@
+/*
+ *  drivers/block/mg_disk.c
+ *
+ *  Support for the mGine m[g]flash IO mode.
+ *  Based on legacy hd.c
+ *
+ * (c) 2008 mGine Co.,LTD
+ * (c) 2008 unsik Kim <donari75@gmail.com>
+ *
+ *  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/kernel.h>
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/blkdev.h>
+#include <linux/hdreg.h>
+#include <linux/libata.h>
+#include <linux/interrupt.h>
+#include <linux/delay.h>
+#include <linux/platform_device.h>
+#include <linux/gpio.h>
+#include <linux/mg_disk.h>
+
+#define MG_RES_SEC (CONFIG_MG_DISK_RES << 1)
+
+static void mg_request(struct request_queue *);
+
+static void mg_dump_status(const char *msg, unsigned int stat,
+               struct mg_host *host)
+{
+       char *name = MG_DISK_NAME;
+       struct request *req;
+
+       if (host->breq) {
+               req = elv_next_request(host->breq);
+               if (req)
+                       name = req->rq_disk->disk_name;
+       }
+
+       printk(KERN_ERR "%s: %s: status=0x%02x { ", name, msg, stat & 0xff);
+       if (stat & MG_REG_STATUS_BIT_BUSY)
+               printk("Busy ");
+       if (stat & MG_REG_STATUS_BIT_READY)
+               printk("DriveReady ");
+       if (stat & MG_REG_STATUS_BIT_WRITE_FAULT)
+               printk("WriteFault ");
+       if (stat & MG_REG_STATUS_BIT_SEEK_DONE)
+               printk("SeekComplete ");
+       if (stat & MG_REG_STATUS_BIT_DATA_REQ)
+               printk("DataRequest ");
+       if (stat & MG_REG_STATUS_BIT_CORRECTED_ERROR)
+               printk("CorrectedError ");
+       if (stat & MG_REG_STATUS_BIT_ERROR)
+               printk("Error ");
+       printk("}\n");
+       if ((stat & MG_REG_STATUS_BIT_ERROR) == 0) {
+               host->error = 0;
+       } else {
+               host->error = inb((unsigned long)host->dev_base + MG_REG_ERROR);
+               printk(KERN_ERR "%s: %s: error=0x%02x { ", name, msg,
+                               host->error & 0xff);
+               if (host->error & MG_REG_ERR_BBK)
+                       printk("BadSector ");
+               if (host->error & MG_REG_ERR_UNC)
+                       printk("UncorrectableError ");
+               if (host->error & MG_REG_ERR_IDNF)
+                       printk("SectorIdNotFound ");
+               if (host->error & MG_REG_ERR_ABRT)
+                       printk("DriveStatusError ");
+               if (host->error & MG_REG_ERR_AMNF)
+                       printk("AddrMarkNotFound ");
+               printk("}");
+               if (host->error &
+                               (MG_REG_ERR_BBK | MG_REG_ERR_UNC |
+                                MG_REG_ERR_IDNF | MG_REG_ERR_AMNF)) {
+                       if (host->breq) {
+                               req = elv_next_request(host->breq);
+                               if (req)
+                                       printk(", sector=%ld", req->sector);
+                       }
+
+               }
+               printk("\n");
+       }
+}
+
+static unsigned int mg_wait(struct mg_host *host, u32 expect, u32 msec)
+{
+       u8 status;
+       unsigned long expire, cur_jiffies;
+       struct mg_drv_data *prv_data = host->dev->platform_data;
+
+       host->error = MG_ERR_NONE;
+       expire = jiffies + msecs_to_jiffies(msec);
+
+       status = inb((unsigned long)host->dev_base + MG_REG_STATUS);
+
+       do {
+               cur_jiffies = jiffies;
+               if (status & MG_REG_STATUS_BIT_BUSY) {
+                       if (expect == MG_REG_STATUS_BIT_BUSY)
+                               break;
+               } else {
+                       /* Check the error condition! */
+                       if (status & MG_REG_STATUS_BIT_ERROR) {
+                               mg_dump_status("mg_wait", status, host);
+                               break;
+                       }
+
+                       if (expect == MG_STAT_READY)
+                               if (MG_READY_OK(status))
+                                       break;
+
+                       if (expect == MG_REG_STATUS_BIT_DATA_REQ)
+                               if (status & MG_REG_STATUS_BIT_DATA_REQ)
+                                       break;
+               }
+               if (!msec) {
+                       mg_dump_status("not ready", status, host);
+                       return MG_ERR_INV_STAT;
+               }
+               if (prv_data->use_polling)
+                       msleep(1);
+
+               status = inb((unsigned long)host->dev_base + MG_REG_STATUS);
+       } while (time_before(cur_jiffies, expire));
+
+       if (time_after_eq(cur_jiffies, expire) && msec)
+               host->error = MG_ERR_TIMEOUT;
+
+       return host->error;
+}
+
+static unsigned int mg_wait_rstout(u32 rstout, u32 msec)
+{
+       unsigned long expire;
+
+       expire = jiffies + msecs_to_jiffies(msec);
+       while (time_before(jiffies, expire)) {
+               if (gpio_get_value(rstout) == 1)
+                       return MG_ERR_NONE;
+               msleep(10);
+       }
+
+       return MG_ERR_RSTOUT;
+}
+
+static void mg_unexpected_intr(struct mg_host *host)
+{
+       u32 status = inb((unsigned long)host->dev_base + MG_REG_STATUS);
+
+       mg_dump_status("mg_unexpected_intr", status, host);
+}
+
+static irqreturn_t mg_irq(int irq, void *dev_id)
+{
+       struct mg_host *host = dev_id;
+       void (*handler)(struct mg_host *) = host->mg_do_intr;
+
+       host->mg_do_intr = 0;
+       del_timer(&host->timer);
+       if (!handler)
+               handler = mg_unexpected_intr;
+       handler(host);
+       return IRQ_HANDLED;
+}
+
+static int mg_get_disk_id(struct mg_host *host)
+{
+       u32 i;
+       s32 err;
+       const u16 *id = host->id;
+       struct mg_drv_data *prv_data = host->dev->platform_data;
+       char fwrev[ATA_ID_FW_REV_LEN + 1];
+       char model[ATA_ID_PROD_LEN + 1];
+       char serial[ATA_ID_SERNO_LEN + 1];
+
+       if (!prv_data->use_polling)
+               outb(MG_REG_CTRL_INTR_DISABLE,
+                               (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+
+       outb(MG_CMD_ID, (unsigned long)host->dev_base + MG_REG_COMMAND);
+       err = mg_wait(host, MG_REG_STATUS_BIT_DATA_REQ, MG_TMAX_WAIT_RD_DRQ);
+       if (err)
+               return err;
+
+       for (i = 0; i < (MG_SECTOR_SIZE >> 1); i++)
+               host->id[i] = le16_to_cpu(inw((unsigned long)host->dev_base +
+                                       MG_BUFF_OFFSET + i * 2));
+
+       outb(MG_CMD_RD_CONF, (unsigned long)host->dev_base + MG_REG_COMMAND);
+       err = mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD);
+       if (err)
+               return err;
+
+       if ((id[ATA_ID_FIELD_VALID] & 1) == 0)
+               return MG_ERR_TRANSLATION;
+
+       host->n_sectors = ata_id_u32(id, ATA_ID_LBA_CAPACITY);
+       host->cyls = id[ATA_ID_CYLS];
+       host->heads = id[ATA_ID_HEADS];
+       host->sectors = id[ATA_ID_SECTORS];
+
+       if (MG_RES_SEC && host->heads && host->sectors) {
+               /* modify cyls, n_sectors */
+               host->cyls = (host->n_sectors - MG_RES_SEC) /
+                       host->heads / host->sectors;
+               host->nres_sectors = host->n_sectors - host->cyls *
+                       host->heads * host->sectors;
+               host->n_sectors -= host->nres_sectors;
+       }
+
+       ata_id_c_string(id, fwrev, ATA_ID_FW_REV, sizeof(fwrev));
+       ata_id_c_string(id, model, ATA_ID_PROD, sizeof(model));
+       ata_id_c_string(id, serial, ATA_ID_SERNO, sizeof(serial));
+       printk(KERN_INFO "mg_disk: model: %s\n", model);
+       printk(KERN_INFO "mg_disk: firm: %.8s\n", fwrev);
+       printk(KERN_INFO "mg_disk: serial: %s\n", serial);
+       printk(KERN_INFO "mg_disk: %d + reserved %d sectors\n",
+                       host->n_sectors, host->nres_sectors);
+
+       if (!prv_data->use_polling)
+               outb(MG_REG_CTRL_INTR_ENABLE, (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+
+       return err;
+}
+
+
+static int mg_disk_init(struct mg_host *host)
+{
+       struct mg_drv_data *prv_data = host->dev->platform_data;
+       s32 err;
+       u8 init_status;
+
+       /* hdd rst low */
+       gpio_set_value(host->rst, 0);
+       err = mg_wait(host, MG_REG_STATUS_BIT_BUSY, MG_TMAX_RST_TO_BUSY);
+       if (err)
+               return err;
+
+       /* hdd rst high */
+       gpio_set_value(host->rst, 1);
+       err = mg_wait(host, MG_STAT_READY, MG_TMAX_HDRST_TO_RDY);
+       if (err)
+               return err;
+
+       /* soft reset on */
+       outb(MG_REG_CTRL_RESET |
+                       (prv_data->use_polling ? MG_REG_CTRL_INTR_DISABLE :
+                        MG_REG_CTRL_INTR_ENABLE),
+                       (unsigned long)host->dev_base + MG_REG_DRV_CTRL);
+       err = mg_wait(host, MG_REG_STATUS_BIT_BUSY, MG_TMAX_RST_TO_BUSY);
+       if (err)
+               return err;
+
+       /* soft reset off */
+       outb(prv_data->use_polling ? MG_REG_CTRL_INTR_DISABLE :
+                       MG_REG_CTRL_INTR_ENABLE,
+                       (unsigned long)host->dev_base + MG_REG_DRV_CTRL);
+       err = mg_wait(host, MG_STAT_READY, MG_TMAX_SWRST_TO_RDY);
+       if (err)
+               return err;
+
+       init_status = inb((unsigned long)host->dev_base + MG_REG_STATUS) & 0xf;
+
+       if (init_status == 0xf)
+               return MG_ERR_INIT_STAT;
+
+       return err;
+}
+
+static void mg_bad_rw_intr(struct mg_host *host)
+{
+       struct request *req = elv_next_request(host->breq);
+       if (req != NULL)
+               if (++req->errors >= MG_MAX_ERRORS ||
+                               host->error == MG_ERR_TIMEOUT)
+                       end_request(req, 0);
+}
+
+static unsigned int mg_out(struct mg_host *host,
+               unsigned int sect_num,
+               unsigned int sect_cnt,
+               unsigned int cmd,
+               void (*intr_addr)(struct mg_host *))
+{
+       struct mg_drv_data *prv_data = host->dev->platform_data;
+
+       if (mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD))
+               return host->error;
+
+       if (!prv_data->use_polling) {
+               host->mg_do_intr = intr_addr;
+               mod_timer(&host->timer, jiffies + 3 * HZ);
+       }
+       if (MG_RES_SEC)
+               sect_num += MG_RES_SEC;
+       outb((u8)sect_cnt, (unsigned long)host->dev_base + MG_REG_SECT_CNT);
+       outb((u8)sect_num, (unsigned long)host->dev_base + MG_REG_SECT_NUM);
+       outb((u8)(sect_num >> 8), (unsigned long)host->dev_base +
+                       MG_REG_CYL_LOW);
+       outb((u8)(sect_num >> 16), (unsigned long)host->dev_base +
+                       MG_REG_CYL_HIGH);
+       outb((u8)((sect_num >> 24) | MG_REG_HEAD_LBA_MODE),
+                       (unsigned long)host->dev_base + MG_REG_DRV_HEAD);
+       outb(cmd, (unsigned long)host->dev_base + MG_REG_COMMAND);
+       return MG_ERR_NONE;
+}
+
+static void mg_read(struct request *req)
+{
+       u32 remains, j;
+       struct mg_host *host = req->rq_disk->private_data;
+
+       remains = req->nr_sectors;
+
+       if (mg_out(host, req->sector, req->nr_sectors, MG_CMD_RD, 0) !=
+                       MG_ERR_NONE)
+               mg_bad_rw_intr(host);
+
+       MG_DBG("requested %d sects (from %ld), buffer=0x%p\n",
+                       remains, req->sector, req->buffer);
+
+       while (remains) {
+               if (mg_wait(host, MG_REG_STATUS_BIT_DATA_REQ,
+                                       MG_TMAX_WAIT_RD_DRQ) != MG_ERR_NONE) {
+                       mg_bad_rw_intr(host);
+                       return;
+               }
+               for (j = 0; j < MG_SECTOR_SIZE >> 1; j++) {
+                       *(u16 *)req->buffer =
+                               inw((unsigned long)host->dev_base +
+                                               MG_BUFF_OFFSET + (j << 1));
+                       req->buffer += 2;
+               }
+
+               req->sector++;
+               req->errors = 0;
+               remains = --req->nr_sectors;
+               --req->current_nr_sectors;
+
+               if (req->current_nr_sectors <= 0) {
+                       MG_DBG("remain : %d sects\n", remains);
+                       end_request(req, 1);
+                       if (remains > 0)
+                               req = elv_next_request(host->breq);
+               }
+
+               outb(MG_CMD_RD_CONF, (unsigned long)host->dev_base +
+                               MG_REG_COMMAND);
+       }
+}
+
+static void mg_write(struct request *req)
+{
+       u32 remains, j;
+       struct mg_host *host = req->rq_disk->private_data;
+
+       remains = req->nr_sectors;
+
+       if (mg_out(host, req->sector, req->nr_sectors, MG_CMD_WR, 0) !=
+                       MG_ERR_NONE) {
+               mg_bad_rw_intr(host);
+               return;
+       }
+
+
+       MG_DBG("requested %d sects (from %ld), buffer=0x%p\n",
+                       remains, req->sector, req->buffer);
+       while (remains) {
+               if (mg_wait(host, MG_REG_STATUS_BIT_DATA_REQ,
+                                       MG_TMAX_WAIT_WR_DRQ) != MG_ERR_NONE) {
+                       mg_bad_rw_intr(host);
+                       return;
+               }
+               for (j = 0; j < MG_SECTOR_SIZE >> 1; j++) {
+                       outw(*(u16 *)req->buffer,
+                                       (unsigned long)host->dev_base +
+                                       MG_BUFF_OFFSET + (j << 1));
+                       req->buffer += 2;
+               }
+               req->sector++;
+               remains = --req->nr_sectors;
+               --req->current_nr_sectors;
+
+               if (req->current_nr_sectors <= 0) {
+                       MG_DBG("remain : %d sects\n", remains);
+                       end_request(req, 1);
+                       if (remains > 0)
+                               req = elv_next_request(host->breq);
+               }
+
+               outb(MG_CMD_WR_CONF, (unsigned long)host->dev_base +
+                               MG_REG_COMMAND);
+       }
+}
+
+static void mg_read_intr(struct mg_host *host)
+{
+       u32 i;
+       struct request *req;
+
+       /* check status */
+       do {
+               i = inb((unsigned long)host->dev_base + MG_REG_STATUS);
+               if (i & MG_REG_STATUS_BIT_BUSY)
+                       break;
+               if (!MG_READY_OK(i))
+                       break;
+               if (i & MG_REG_STATUS_BIT_DATA_REQ)
+                       goto ok_to_read;
+       } while (0);
+       mg_dump_status("mg_read_intr", i, host);
+       mg_bad_rw_intr(host);
+       mg_request(host->breq);
+       return;
+
+ok_to_read:
+       /* get current segment of request */
+       req = elv_next_request(host->breq);
+
+       /* read 1 sector */
+       for (i = 0; i < MG_SECTOR_SIZE >> 1; i++) {
+               *(u16 *)req->buffer =
+                       inw((unsigned long)host->dev_base + MG_BUFF_OFFSET +
+                                       (i << 1));
+               req->buffer += 2;
+       }
+
+       /* manipulate request */
+       MG_DBG("sector %ld, remaining=%ld, buffer=0x%p\n",
+                       req->sector, req->nr_sectors - 1, req->buffer);
+
+       req->sector++;
+       req->errors = 0;
+       i = --req->nr_sectors;
+       --req->current_nr_sectors;
+
+       /* let know if current segment done */
+       if (req->current_nr_sectors <= 0)
+               end_request(req, 1);
+
+       /* set handler if read remains */
+       if (i > 0) {
+               host->mg_do_intr = mg_read_intr;
+               mod_timer(&host->timer, jiffies + 3 * HZ);
+       }
+
+       /* send read confirm */
+       outb(MG_CMD_RD_CONF, (unsigned long)host->dev_base + MG_REG_COMMAND);
+
+       /* goto next request */
+       if (!i)
+               mg_request(host->breq);
+}
+
+static void mg_write_intr(struct mg_host *host)
+{
+       u32 i, j;
+       u16 *buff;
+       struct request *req;
+
+       /* get current segment of request */
+       req = elv_next_request(host->breq);
+
+       /* check status */
+       do {
+               i = inb((unsigned long)host->dev_base + MG_REG_STATUS);
+               if (i & MG_REG_STATUS_BIT_BUSY)
+                       break;
+               if (!MG_READY_OK(i))
+                       break;
+               if ((req->nr_sectors <= 1) || (i & MG_REG_STATUS_BIT_DATA_REQ))
+                       goto ok_to_write;
+       } while (0);
+       mg_dump_status("mg_write_intr", i, host);
+       mg_bad_rw_intr(host);
+       mg_request(host->breq);
+       return;
+
+ok_to_write:
+       /* manipulate request */
+       req->sector++;
+       i = --req->nr_sectors;
+       --req->current_nr_sectors;
+       req->buffer += MG_SECTOR_SIZE;
+
+       /* let know if current segment or all done */
+       if (!i || (req->bio && req->current_nr_sectors <= 0))
+               end_request(req, 1);
+
+       /* write 1 sector and set handler if remains */
+       if (i > 0) {
+               buff = (u16 *)req->buffer;
+               for (j = 0; j < MG_STORAGE_BUFFER_SIZE >> 1; j++) {
+                       outw(*buff, (unsigned long)host->dev_base +
+                                       MG_BUFF_OFFSET + (j << 1));
+                       buff++;
+               }
+               MG_DBG("sector %ld, remaining=%ld, buffer=0x%p\n",
+                               req->sector, req->nr_sectors, req->buffer);
+               host->mg_do_intr = mg_write_intr;
+               mod_timer(&host->timer, jiffies + 3 * HZ);
+       }
+
+       /* send write confirm */
+       outb(MG_CMD_WR_CONF, (unsigned long)host->dev_base + MG_REG_COMMAND);
+
+       if (!i)
+               mg_request(host->breq);
+}
+
+void mg_times_out(unsigned long data)
+{
+       struct mg_host *host = (struct mg_host *)data;
+       char *name;
+       struct request *req;
+
+       req = elv_next_request(host->breq);
+       if (!req)
+               return;
+
+       host->mg_do_intr = NULL;
+
+       name = req->rq_disk->disk_name;
+       printk(KERN_DEBUG "%s: timeout\n", name);
+
+       host->error = MG_ERR_TIMEOUT;
+       mg_bad_rw_intr(host);
+
+       mg_request(host->breq);
+}
+
+static void mg_request_poll(struct request_queue *q)
+{
+       struct request *req;
+       struct mg_host *host;
+
+       while ((req = elv_next_request(q)) != NULL) {
+               host = req->rq_disk->private_data;
+               if (blk_fs_request(req)) {
+                       switch (rq_data_dir(req)) {
+                       case READ:
+                               mg_read(req);
+                               break;
+                       case WRITE:
+                               mg_write(req);
+                               break;
+                       default:
+                               printk(KERN_WARNING "%s:%d unknown command\n",
+                                               __func__, __LINE__);
+                               end_request(req, 0);
+                               break;
+                       }
+               }
+       }
+}
+
+static unsigned int mg_issue_req(struct request *req,
+               struct mg_host *host,
+               unsigned int sect_num,
+               unsigned int sect_cnt)
+{
+       u16 *buff;
+       u32 i;
+
+       switch (rq_data_dir(req)) {
+       case READ:
+               if (mg_out(host, sect_num, sect_cnt, MG_CMD_RD, &mg_read_intr)
+                               != MG_ERR_NONE) {
+                       mg_bad_rw_intr(host);
+                       return host->error;
+               }
+               break;
+       case WRITE:
+               /* TODO : handler */
+               outb(MG_REG_CTRL_INTR_DISABLE,
+                               (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+               if (mg_out(host, sect_num, sect_cnt, MG_CMD_WR, &mg_write_intr)
+                               != MG_ERR_NONE) {
+                       mg_bad_rw_intr(host);
+                       return host->error;
+               }
+               del_timer(&host->timer);
+               mg_wait(host, MG_REG_STATUS_BIT_DATA_REQ, MG_TMAX_WAIT_WR_DRQ);
+               outb(MG_REG_CTRL_INTR_ENABLE, (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+               if (host->error) {
+                       mg_bad_rw_intr(host);
+                       return host->error;
+               }
+               buff = (u16 *)req->buffer;
+               for (i = 0; i < MG_SECTOR_SIZE >> 1; i++) {
+                       outw(*buff, (unsigned long)host->dev_base +
+                                       MG_BUFF_OFFSET + (i << 1));
+                       buff++;
+               }
+               mod_timer(&host->timer, jiffies + 3 * HZ);
+               outb(MG_CMD_WR_CONF, (unsigned long)host->dev_base +
+                               MG_REG_COMMAND);
+               break;
+       default:
+               printk(KERN_WARNING "%s:%d unknown command\n",
+                               __func__, __LINE__);
+               end_request(req, 0);
+               break;
+       }
+       return MG_ERR_NONE;
+}
+
+/* This function also called from IRQ context */
+static void mg_request(struct request_queue *q)
+{
+       struct request *req;
+       struct mg_host *host;
+       u32 sect_num, sect_cnt;
+
+       while (1) {
+               req = elv_next_request(q);
+               if (!req)
+                       return;
+
+               host = req->rq_disk->private_data;
+
+               /* check unwanted request call */
+               if (host->mg_do_intr)
+                       return;
+
+               del_timer(&host->timer);
+
+               sect_num = req->sector;
+               /* deal whole segments */
+               sect_cnt = req->nr_sectors;
+
+               /* sanity check */
+               if (sect_num >= get_capacity(req->rq_disk) ||
+                               ((sect_num + sect_cnt) >
+                                get_capacity(req->rq_disk))) {
+                       printk(KERN_WARNING
+                                       "%s: bad access: sector=%d, count=%d\n",
+                                       req->rq_disk->disk_name,
+                                       sect_num, sect_cnt);
+                       end_request(req, 0);
+                       continue;
+               }
+
+               if (!blk_fs_request(req))
+                       return;
+
+               if (!mg_issue_req(req, host, sect_num, sect_cnt))
+                       return;
+       }
+}
+
+static int mg_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+{
+       struct mg_host *host = bdev->bd_disk->private_data;
+
+       geo->cylinders = (unsigned short)host->cyls;
+       geo->heads = (unsigned char)host->heads;
+       geo->sectors = (unsigned char)host->sectors;
+       return 0;
+}
+
+static struct block_device_operations mg_disk_ops = {
+       .getgeo = mg_getgeo
+};
+
+static int mg_suspend(struct platform_device *plat_dev, pm_message_t state)
+{
+       struct mg_drv_data *prv_data = plat_dev->dev.platform_data;
+       struct mg_host *host = prv_data->host;
+
+       if (mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD))
+               return -EIO;
+
+       if (!prv_data->use_polling)
+               outb(MG_REG_CTRL_INTR_DISABLE,
+                               (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+
+       outb(MG_CMD_SLEEP, (unsigned long)host->dev_base + MG_REG_COMMAND);
+       /* wait until mflash deep sleep */
+       msleep(1);
+
+       if (mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD)) {
+               if (!prv_data->use_polling)
+                       outb(MG_REG_CTRL_INTR_ENABLE,
+                                       (unsigned long)host->dev_base +
+                                       MG_REG_DRV_CTRL);
+               return -EIO;
+       }
+
+       return 0;
+}
+
+static int mg_resume(struct platform_device *plat_dev)
+{
+       struct mg_drv_data *prv_data = plat_dev->dev.platform_data;
+       struct mg_host *host = prv_data->host;
+
+       if (mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD))
+               return -EIO;
+
+       outb(MG_CMD_WAKEUP, (unsigned long)host->dev_base + MG_REG_COMMAND);
+       /* wait until mflash wakeup */
+       msleep(1);
+
+       if (mg_wait(host, MG_STAT_READY, MG_TMAX_CONF_TO_CMD))
+               return -EIO;
+
+       if (!prv_data->use_polling)
+               outb(MG_REG_CTRL_INTR_ENABLE, (unsigned long)host->dev_base +
+                               MG_REG_DRV_CTRL);
+
+       return 0;
+}
+
+static int mg_probe(struct platform_device *plat_dev)
+{
+       struct mg_host *host;
+       struct resource *rsc;
+       struct mg_drv_data *prv_data = plat_dev->dev.platform_data;
+       int err = 0;
+
+       if (!prv_data) {
+               printk(KERN_ERR "%s:%d fail (no driver_data)\n",
+                               __func__, __LINE__);
+               err = -EINVAL;
+               goto probe_err;
+       }
+
+       /* alloc mg_host */
+       host = kzalloc(sizeof(struct mg_host), GFP_KERNEL);
+       if (!host) {
+               printk(KERN_ERR "%s:%d fail (no memory for mg_host)\n",
+                               __func__, __LINE__);
+               err = -ENOMEM;
+               goto probe_err;
+       }
+       host->major = MG_DISK_MAJ;
+
+       /* link each other */
+       prv_data->host = host;
+       host->dev = &plat_dev->dev;
+
+       /* io remap */
+       rsc = platform_get_resource(plat_dev, IORESOURCE_MEM, 0);
+       if (!rsc) {
+               printk(KERN_ERR "%s:%d platform_get_resource fail\n",
+                               __func__, __LINE__);
+               err = -EINVAL;
+               goto probe_err_2;
+       }
+       host->dev_base = ioremap(rsc->start , rsc->end + 1);
+       if (!host->dev_base) {
+               printk(KERN_ERR "%s:%d ioremap fail\n",
+                               __func__, __LINE__);
+               err = -EIO;
+               goto probe_err_2;
+       }
+       MG_DBG("dev_base = 0x%x\n", (u32)host->dev_base);
+
+       /* get reset pin */
+       rsc = platform_get_resource_byname(plat_dev, IORESOURCE_IO,
+                       MG_RST_PIN);
+       if (!rsc) {
+               printk(KERN_ERR "%s:%d get reset pin fail\n",
+                               __func__, __LINE__);
+               err = -EIO;
+               goto probe_err_3;
+       }
+       host->rst = rsc->start;
+
+       /* init rst pin */
+       err = gpio_request(host->rst, MG_RST_PIN);
+       if (err)
+               goto probe_err_3;
+       gpio_direction_output(host->rst, 1);
+
+       /* reset out pin */
+       if (!(prv_data->dev_attr & MG_DEV_MASK))
+               goto probe_err_3a;
+
+       if (prv_data->dev_attr != MG_BOOT_DEV) {
+               rsc = platform_get_resource_byname(plat_dev, IORESOURCE_IO,
+                               MG_RSTOUT_PIN);
+               if (!rsc) {
+                       printk(KERN_ERR "%s:%d get reset-out pin fail\n",
+                                       __func__, __LINE__);
+                       err = -EIO;
+                       goto probe_err_3a;
+               }
+               host->rstout = rsc->start;
+               err = gpio_request(host->rstout, MG_RSTOUT_PIN);
+               if (err)
+                       goto probe_err_3a;
+               gpio_direction_input(host->rstout);
+       }
+
+       /* disk reset */
+       if (prv_data->dev_attr == MG_STORAGE_DEV) {
+               /* If POR seq. not yet finised, wait */
+               err = mg_wait_rstout(host->rstout, MG_TMAX_RSTOUT);
+               if (err)
+                       goto probe_err_3b;
+               err = mg_disk_init(host);
+               if (err) {
+                       printk(KERN_ERR "%s:%d fail (err code : %d)\n",
+                                       __func__, __LINE__, err);
+                       err = -EIO;
+                       goto probe_err_3b;
+               }
+       }
+
+       /* get irq resource */
+       if (!prv_data->use_polling) {
+               host->irq = platform_get_irq(plat_dev, 0);
+               if (host->irq == -ENXIO) {
+                       err = host->irq;
+                       goto probe_err_3b;
+               }
+               err = request_irq(host->irq, mg_irq,
+                               IRQF_DISABLED | IRQF_TRIGGER_RISING,
+                               MG_DEV_NAME, host);
+               if (err) {
+                       printk(KERN_ERR "%s:%d fail (request_irq err=%d)\n",
+                                       __func__, __LINE__, err);
+                       goto probe_err_3b;
+               }
+
+       }
+
+       /* get disk id */
+       err = mg_get_disk_id(host);
+       if (err) {
+               printk(KERN_ERR "%s:%d fail (err code : %d)\n",
+                               __func__, __LINE__, err);
+               err = -EIO;
+               goto probe_err_4;
+       }
+
+       err = register_blkdev(host->major, MG_DISK_NAME);
+       if (err < 0) {
+               printk(KERN_ERR "%s:%d register_blkdev fail (err code : %d)\n",
+                               __func__, __LINE__, err);
+               goto probe_err_4;
+       }
+       if (!host->major)
+               host->major = err;
+
+       spin_lock_init(&host->lock);
+
+       if (prv_data->use_polling)
+               host->breq = blk_init_queue(mg_request_poll, &host->lock);
+       else
+               host->breq = blk_init_queue(mg_request, &host->lock);
+
+       if (!host->breq) {
+               err = -ENOMEM;
+               printk(KERN_ERR "%s:%d (blk_init_queue) fail\n",
+                               __func__, __LINE__);
+               goto probe_err_5;
+       }
+
+       /* mflash is random device, thanx for the noop */
+       elevator_exit(host->breq->elevator);
+       err = elevator_init(host->breq, "noop");
+       if (err) {
+               printk(KERN_ERR "%s:%d (elevator_init) fail\n",
+                               __func__, __LINE__);
+               goto probe_err_6;
+       }
+       blk_queue_max_sectors(host->breq, MG_MAX_SECTS);
+       blk_queue_hardsect_size(host->breq, MG_SECTOR_SIZE);
+
+       init_timer(&host->timer);
+       host->timer.function = mg_times_out;
+       host->timer.data = (unsigned long)host;
+
+       host->gd = alloc_disk(MG_DISK_MAX_PART);
+       if (!host->gd) {
+               printk(KERN_ERR "%s:%d (alloc_disk) fail\n",
+                               __func__, __LINE__);
+               err = -ENOMEM;
+               goto probe_err_7;
+       }
+       host->gd->major = host->major;
+       host->gd->first_minor = 0;
+       host->gd->fops = &mg_disk_ops;
+       host->gd->queue = host->breq;
+       host->gd->private_data = host;
+       sprintf(host->gd->disk_name, MG_DISK_NAME"a");
+
+       set_capacity(host->gd, host->n_sectors);
+
+       add_disk(host->gd);
+
+       return err;
+
+probe_err_7:
+       del_timer_sync(&host->timer);
+probe_err_6:
+       blk_cleanup_queue(host->breq);
+probe_err_5:
+       unregister_blkdev(MG_DISK_MAJ, MG_DISK_NAME);
+probe_err_4:
+       if (!prv_data->use_polling)
+               free_irq(host->irq, host);
+probe_err_3b:
+       gpio_free(host->rstout);
+probe_err_3a:
+       gpio_free(host->rst);
+probe_err_3:
+       iounmap(host->dev_base);
+probe_err_2:
+       kfree(host);
+probe_err:
+       return err;
+}
+
+static int mg_remove(struct platform_device *plat_dev)
+{
+       struct mg_drv_data *prv_data = plat_dev->dev.platform_data;
+       struct mg_host *host = prv_data->host;
+       int err = 0;
+
+       /* delete timer */
+       del_timer_sync(&host->timer);
+
+       /* remove disk */
+       if (host->gd) {
+               del_gendisk(host->gd);
+               put_disk(host->gd);
+       }
+       /* remove queue */
+       if (host->breq)
+               blk_cleanup_queue(host->breq);
+
+       /* unregister blk device */
+       unregister_blkdev(host->major, MG_DISK_NAME);
+
+       /* free irq */
+       if (!prv_data->use_polling)
+               free_irq(host->irq, host);
+
+       /* free reset-out pin */
+       if (prv_data->dev_attr != MG_BOOT_DEV)
+               gpio_free(host->rstout);
+
+       /* free rst pin */
+       if (host->rst)
+               gpio_free(host->rst);
+
+       /* unmap io */
+       if (host->dev_base)
+               iounmap(host->dev_base);
+
+       /* free mg_host */
+       kfree(host);
+
+       return err;
+}
+
+static struct platform_driver mg_disk_driver = {
+       .probe = mg_probe,
+       .remove = mg_remove,
+       .suspend = mg_suspend,
+       .resume = mg_resume,
+       .driver = {
+               .name = MG_DEV_NAME,
+               .owner = THIS_MODULE,
+       }
+};
+
+/****************************************************************************
+ *
+ * Module stuff
+ *
+ ****************************************************************************/
+
+static int __init mg_init(void)
+{
+       printk(KERN_INFO "mGine mflash driver, (c) 2008 mGine Co.\n");
+       return platform_driver_register(&mg_disk_driver);
+}
+
+static void __exit mg_exit(void)
+{
+       printk(KERN_INFO "mflash driver : bye bye\n");
+       platform_driver_unregister(&mg_disk_driver);
+}
+
+module_init(mg_init);
+module_exit(mg_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("unsik Kim <donari75@gmail.com>");
+MODULE_DESCRIPTION("mGine m[g]flash device driver");
diff --git a/include/linux/mg_disk.h b/include/linux/mg_disk.h
new file mode 100644 (file)
index 0000000..1f76b1e
--- /dev/null
@@ -0,0 +1,206 @@
+/*
+ *  include/linux/mg_disk.c
+ *
+ *  Support for the mGine m[g]flash IO mode.
+ *  Based on legacy hd.c
+ *
+ * (c) 2008 mGine Co.,LTD
+ * (c) 2008 unsik Kim <donari75@gmail.com>
+ *
+ *  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.
+ */
+
+#ifndef __MG_DISK_H__
+#define __MG_DISK_H__
+
+#include <linux/blkdev.h>
+#include <linux/ata.h>
+
+/* name for block device */
+#define MG_DISK_NAME "mgd"
+/* name for platform device */
+#define MG_DEV_NAME "mg_disk"
+
+#define MG_DISK_MAJ 0
+#define MG_DISK_MAX_PART 16
+#define MG_SECTOR_SIZE 512
+#define MG_MAX_SECTS 256
+
+/* Register offsets */
+#define MG_BUFF_OFFSET                 0x8000
+#define MG_STORAGE_BUFFER_SIZE         0x200
+#define MG_REG_OFFSET                  0xC000
+#define MG_REG_FEATURE                 (MG_REG_OFFSET + 2)     /* write case */
+#define MG_REG_ERROR                   (MG_REG_OFFSET + 2)     /* read case */
+#define MG_REG_SECT_CNT                        (MG_REG_OFFSET + 4)
+#define MG_REG_SECT_NUM                        (MG_REG_OFFSET + 6)
+#define MG_REG_CYL_LOW                 (MG_REG_OFFSET + 8)
+#define MG_REG_CYL_HIGH                        (MG_REG_OFFSET + 0xA)
+#define MG_REG_DRV_HEAD                        (MG_REG_OFFSET + 0xC)
+#define MG_REG_COMMAND                 (MG_REG_OFFSET + 0xE)   /* write case */
+#define MG_REG_STATUS                  (MG_REG_OFFSET + 0xE)   /* read  case */
+#define MG_REG_DRV_CTRL                        (MG_REG_OFFSET + 0x10)
+#define MG_REG_BURST_CTRL              (MG_REG_OFFSET + 0x12)
+
+/* "Drive Select/Head Register" bit values */
+#define MG_REG_HEAD_MUST_BE_ON         0xA0 /* These 2 bits are always on */
+#define MG_REG_HEAD_DRIVE_MASTER       (0x00 | MG_REG_HEAD_MUST_BE_ON)
+#define MG_REG_HEAD_DRIVE_SLAVE                (0x10 | MG_REG_HEAD_MUST_BE_ON)
+#define MG_REG_HEAD_LBA_MODE           (0x40 | MG_REG_HEAD_MUST_BE_ON)
+
+
+/* "Device Control Register" bit values */
+#define MG_REG_CTRL_INTR_ENABLE                        0x0
+#define MG_REG_CTRL_INTR_DISABLE               (0x1<<1)
+#define MG_REG_CTRL_RESET                      (0x1<<2)
+#define MG_REG_CTRL_INTR_POLA_ACTIVE_HIGH      0x0
+#define MG_REG_CTRL_INTR_POLA_ACTIVE_LOW       (0x1<<4)
+#define MG_REG_CTRL_DPD_POLA_ACTIVE_LOW                0x0
+#define MG_REG_CTRL_DPD_POLA_ACTIVE_HIGH       (0x1<<5)
+#define MG_REG_CTRL_DPD_DISABLE                        0x0
+#define MG_REG_CTRL_DPD_ENABLE                 (0x1<<6)
+
+/* Status register bit */
+/* error bit in status register */
+#define MG_REG_STATUS_BIT_ERROR                        0x01
+/* corrected error in status register */
+#define MG_REG_STATUS_BIT_CORRECTED_ERROR      0x04
+/* data request bit in status register */
+#define MG_REG_STATUS_BIT_DATA_REQ             0x08
+/* DSC - Drive Seek Complete */
+#define MG_REG_STATUS_BIT_SEEK_DONE            0x10
+/* DWF - Drive Write Fault */
+#define MG_REG_STATUS_BIT_WRITE_FAULT          0x20
+#define MG_REG_STATUS_BIT_READY                        0x40
+#define MG_REG_STATUS_BIT_BUSY                 0x80
+
+/* handy status */
+#define MG_STAT_READY  (MG_REG_STATUS_BIT_READY | MG_REG_STATUS_BIT_SEEK_DONE)
+#define MG_READY_OK(s) (((s) & (MG_STAT_READY | \
+                               (MG_REG_STATUS_BIT_BUSY | \
+                                MG_REG_STATUS_BIT_WRITE_FAULT | \
+                                MG_REG_STATUS_BIT_ERROR))) == MG_STAT_READY)
+
+/* Error register */
+#define MG_REG_ERR_AMNF                0x01
+#define MG_REG_ERR_ABRT                0x04
+#define MG_REG_ERR_IDNF                0x10
+#define MG_REG_ERR_UNC         0x40
+#define MG_REG_ERR_BBK         0x80
+
+/* error code for others */
+#define MG_ERR_NONE            0
+#define MG_ERR_TIMEOUT         0x100
+#define MG_ERR_INIT_STAT       0x101
+#define MG_ERR_TRANSLATION     0x102
+#define MG_ERR_CTRL_RST                0x103
+#define MG_ERR_INV_STAT                0x104
+#define MG_ERR_RSTOUT          0x105
+
+#define MG_MAX_ERRORS  6       /* Max read/write errors */
+
+/* command */
+#define MG_CMD_RD 0x20
+#define MG_CMD_WR 0x30
+#define MG_CMD_SLEEP 0x99
+#define MG_CMD_WAKEUP 0xC3
+#define MG_CMD_ID 0xEC
+#define MG_CMD_WR_CONF 0x3C
+#define MG_CMD_RD_CONF 0x40
+
+/* operation mode */
+#define MG_OP_CASCADE (1 << 0)
+#define MG_OP_CASCADE_SYNC_RD (1 << 1)
+#define MG_OP_CASCADE_SYNC_WR (1 << 2)
+#define MG_OP_INTERLEAVE (1 << 3)
+
+/* synchronous */
+#define MG_BURST_LAT_4 (3 << 4)
+#define MG_BURST_LAT_5 (4 << 4)
+#define MG_BURST_LAT_6 (5 << 4)
+#define MG_BURST_LAT_7 (6 << 4)
+#define MG_BURST_LAT_8 (7 << 4)
+#define MG_BURST_LEN_4 (1 << 1)
+#define MG_BURST_LEN_8 (2 << 1)
+#define MG_BURST_LEN_16 (3 << 1)
+#define MG_BURST_LEN_32 (4 << 1)
+#define MG_BURST_LEN_CONT (0 << 1)
+
+/* timeout value (unit: ms) */
+#define MG_TMAX_CONF_TO_CMD    1
+#define MG_TMAX_WAIT_RD_DRQ    10
+#define MG_TMAX_WAIT_WR_DRQ    500
+#define MG_TMAX_RST_TO_BUSY    10
+#define MG_TMAX_HDRST_TO_RDY   500
+#define MG_TMAX_SWRST_TO_RDY   500
+#define MG_TMAX_RSTOUT         3000
+
+/* device attribution */
+/* use mflash as boot device */
+#define MG_BOOT_DEV            (1 << 0)
+/* use mflash as storage device */
+#define MG_STORAGE_DEV         (1 << 1)
+/* same as MG_STORAGE_DEV, but bootloader already done reset sequence */
+#define MG_STORAGE_DEV_SKIP_RST        (1 << 2)
+
+#define MG_DEV_MASK (MG_BOOT_DEV | MG_STORAGE_DEV | MG_STORAGE_DEV_SKIP_RST)
+
+/* names of GPIO resource */
+#define MG_RST_PIN     "mg_rst"
+/* except MG_BOOT_DEV, reset-out pin should be assigned */
+#define MG_RSTOUT_PIN  "mg_rstout"
+
+/* private driver data */
+struct mg_drv_data {
+       /* disk resource */
+       u32 use_polling;
+
+       /* device attribution */
+       u32 dev_attr;
+
+       /* internally used */
+       struct mg_host *host;
+};
+
+/* main structure for mflash driver */
+struct mg_host {
+       struct device *dev;
+
+       struct request_queue *breq;
+       spinlock_t lock;
+       struct gendisk *gd;
+
+       struct timer_list timer;
+       void (*mg_do_intr) (struct mg_host *);
+
+       u16 id[ATA_ID_WORDS];
+
+       u16 cyls;
+       u16 heads;
+       u16 sectors;
+       u32 n_sectors;
+       u32 nres_sectors;
+
+       void __iomem *dev_base;
+       unsigned int irq;
+       unsigned int rst;
+       unsigned int rstout;
+
+       u32 major;
+       u32 error;
+};
+
+/*
+ * Debugging macro and defines
+ */
+#undef DO_MG_DEBUG
+#ifdef DO_MG_DEBUG
+#  define MG_DBG(fmt, args...) \
+       printk(KERN_DEBUG "%s:%d "fmt, __func__, __LINE__, ##args)
+#else /* CONFIG_MG_DEBUG */
+#  define MG_DBG(fmt, args...) do { } while (0)
+#endif /* CONFIG_MG_DEBUG */
+
+#endif