drivers: Final irq namespace conversion
[linux-2.6.git] / drivers / pcmcia / pcmcia_resource.c
index c4612c5..42fbf1a 100644 (file)
@@ -6,7 +6,7 @@
  * are Copyright (C) 1999 David A. Hinds.  All Rights Reserved.
  *
  * Copyright (C) 1999       David A. Hinds
- * Copyright (C) 2004-2005   Dominik Brodowski
+ * Copyright (C) 2004-2010   Dominik Brodowski
  *
  * 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
 #include <linux/pci.h>
 #include <linux/device.h>
 #include <linux/netdevice.h>
+#include <linux/slab.h>
+
+#include <asm/irq.h>
 
-#include <pcmcia/cs_types.h>
 #include <pcmcia/ss.h>
-#include <pcmcia/cs.h>
 #include <pcmcia/cistpl.h>
 #include <pcmcia/cisreg.h>
 #include <pcmcia/ds.h>
@@ -37,29 +38,6 @@ static int io_speed;
 module_param(io_speed, int, 0444);
 
 
-#ifdef CONFIG_PCMCIA_PROBE
-#include <asm/irq.h>
-/* mask of IRQs already reserved by other cards, we should avoid using them */
-static u8 pcmcia_used_irq[NR_IRQS];
-#endif
-
-static int pcmcia_adjust_io_region(struct resource *res, unsigned long start,
-                                  unsigned long end, struct pcmcia_socket *s)
-{
-       if (s->resource_ops->adjust_io_region)
-               return s->resource_ops->adjust_io_region(res, start, end, s);
-       return -ENOMEM;
-}
-
-static struct resource *pcmcia_find_io_region(unsigned long base, int num,
-                                             unsigned long align,
-                                             struct pcmcia_socket *s)
-{
-       if (s->resource_ops->find_io)
-               return s->resource_ops->find_io(base, num, align, s);
-       return NULL;
-}
-
 int pcmcia_validate_mem(struct pcmcia_socket *s)
 {
        if (s->resource_ops->validate_mem)
@@ -77,90 +55,29 @@ struct resource *pcmcia_find_mem_region(u_long base, u_long num, u_long align,
 }
 
 
-/** alloc_io_space
+/**
+ * release_io_space() - release IO ports allocated with alloc_io_space()
+ * @s: pcmcia socket
+ * @res: resource to release
  *
- * Special stuff for managing IO windows, because they are scarce
  */
-
-static int alloc_io_space(struct pcmcia_socket *s, u_int attr,
-                         unsigned int *base, unsigned int num, u_int lines)
+static void release_io_space(struct pcmcia_socket *s, struct resource *res)
 {
+       resource_size_t num = resource_size(res);
        int i;
-       unsigned int try, align;
-
-       align = (*base) ? (lines ? 1<<lines : 0) : 1;
-       if (align && (align < num)) {
-               if (*base) {
-                       dev_dbg(&s->dev, "odd IO request: num %#x align %#x\n",
-                              num, align);
-                       align = 0;
-               } else
-                       while (align && (align < num))
-                               align <<= 1;
-       }
-       if (*base & ~(align-1)) {
-               dev_dbg(&s->dev, "odd IO request: base %#x align %#x\n",
-                      *base, align);
-               align = 0;
-       }
-       if ((s->features & SS_CAP_STATIC_MAP) && s->io_offset) {
-               *base = s->io_offset | (*base & 0x0fff);
-               return 0;
-       }
-       /* Check for an already-allocated window that must conflict with
-        * what was asked for.  It is a hack because it does not catch all
-        * potential conflicts, just the most obvious ones.
-        */
-       for (i = 0; i < MAX_IO_WIN; i++)
-               if ((s->io[i].res) && *base &&
-                   ((s->io[i].res->start & (align-1)) == *base))
-                       return 1;
-       for (i = 0; i < MAX_IO_WIN; i++) {
-               if (!s->io[i].res) {
-                       s->io[i].res = pcmcia_find_io_region(*base, num, align, s);
-                       if (s->io[i].res) {
-                               *base = s->io[i].res->start;
-                               s->io[i].res->flags = (s->io[i].res->flags & ~IORESOURCE_BITS) | (attr & IORESOURCE_BITS);
-                               s->io[i].InUse = num;
-                               break;
-                       } else
-                               return 1;
-               } else if ((s->io[i].res->flags & IORESOURCE_BITS) != (attr & IORESOURCE_BITS))
-                       continue;
-               /* Try to extend top of window */
-               try = s->io[i].res->end + 1;
-               if ((*base == 0) || (*base == try))
-                       if (pcmcia_adjust_io_region(s->io[i].res, s->io[i].res->start,
-                                                   s->io[i].res->end + num, s) == 0) {
-                               *base = try;
-                               s->io[i].InUse += num;
-                               break;
-                       }
-               /* Try to extend bottom of window */
-               try = s->io[i].res->start - num;
-               if ((*base == 0) || (*base == try))
-                       if (pcmcia_adjust_io_region(s->io[i].res, s->io[i].res->start - num,
-                                                   s->io[i].res->end, s) == 0) {
-                               *base = try;
-                               s->io[i].InUse += num;
-                               break;
-                       }
-       }
-       return (i == MAX_IO_WIN);
-} /* alloc_io_space */
 
-
-static void release_io_space(struct pcmcia_socket *s, unsigned int base,
-                            unsigned int num)
-{
-       int i;
+       dev_dbg(&s->dev, "release_io_space for %pR\n", res);
 
        for (i = 0; i < MAX_IO_WIN; i++) {
                if (!s->io[i].res)
                        continue;
-               if ((s->io[i].res->start <= base) &&
-                   (s->io[i].res->end >= base+num-1)) {
+               if ((s->io[i].res->start <= res->start) &&
+                   (s->io[i].res->end >= res->end)) {
                        s->io[i].InUse -= num;
+                       if (res->parent)
+                               release_resource(res);
+                       res->start = res->end = 0;
+                       res->flags = IORESOURCE_IO;
                        /* Free the window if no one else is using it */
                        if (s->io[i].InUse == 0) {
                                release_resource(s->io[i].res);
@@ -169,26 +86,86 @@ static void release_io_space(struct pcmcia_socket *s, unsigned int base,
                        }
                }
        }
-} /* release_io_space */
+}
 
 
-/** pccard_access_configuration_register
+/**
+ * alloc_io_space() - allocate IO ports for use by a PCMCIA device
+ * @s: pcmcia socket
+ * @res: resource to allocate (begin: begin, end: size)
+ * @lines: number of IO lines decoded by the PCMCIA card
  *
- * Access_configuration_register() reads and writes configuration
- * registers in attribute memory.  Memory window 0 is reserved for
- * this and the tuple reading services.
+ * Special stuff for managing IO windows, because they are scarce
  */
+static int alloc_io_space(struct pcmcia_socket *s, struct resource *res,
+                       unsigned int lines)
+{
+       unsigned int align;
+       unsigned int base = res->start;
+       unsigned int num = res->end;
+       int ret;
+
+       res->flags |= IORESOURCE_IO;
+
+       dev_dbg(&s->dev, "alloc_io_space request for %pR, %d lines\n",
+               res, lines);
+
+       align = base ? (lines ? 1<<lines : 0) : 1;
+       if (align && (align < num)) {
+               if (base) {
+                       dev_dbg(&s->dev, "odd IO request\n");
+                       align = 0;
+               } else
+                       while (align && (align < num))
+                               align <<= 1;
+       }
+       if (base & ~(align-1)) {
+               dev_dbg(&s->dev, "odd IO request\n");
+               align = 0;
+       }
+
+       ret = s->resource_ops->find_io(s, res->flags, &base, num, align,
+                               &res->parent);
+       if (ret) {
+               dev_dbg(&s->dev, "alloc_io_space request failed (%d)\n", ret);
+               return -EINVAL;
+       }
+
+       res->start = base;
+       res->end = res->start + num - 1;
+
+       if (res->parent) {
+               ret = request_resource(res->parent, res);
+               if (ret) {
+                       dev_warn(&s->dev,
+                               "request_resource %pR failed: %d\n", res, ret);
+                       res->parent = NULL;
+                       release_io_space(s, res);
+               }
+       }
+       dev_dbg(&s->dev, "alloc_io_space request result %d: %pR\n", ret, res);
+       return ret;
+}
+
 
-int pcmcia_access_configuration_register(struct pcmcia_device *p_dev,
-                                        conf_reg_t *reg)
+/**
+ * pcmcia_access_config() - read or write card configuration registers
+ *
+ * pcmcia_access_config() reads and writes configuration registers in
+ * attribute memory.  Memory window 0 is reserved for this and the tuple
+ * reading services. Drivers must use pcmcia_read_config_byte() or
+ * pcmcia_write_config_byte().
+ */
+static int pcmcia_access_config(struct pcmcia_device *p_dev,
+                               off_t where, u8 *val,
+                               int (*accessf) (struct pcmcia_socket *s,
+                                               int attr, unsigned int addr,
+                                               unsigned int len, void *ptr))
 {
        struct pcmcia_socket *s;
        config_t *c;
        int addr;
-       u_char val;
-
-       if (!p_dev || !p_dev->function_config)
-               return -EINVAL;
+       int ret = 0;
 
        s = p_dev->socket;
 
@@ -196,153 +173,180 @@ int pcmcia_access_configuration_register(struct pcmcia_device *p_dev,
        c = p_dev->function_config;
 
        if (!(c->state & CONFIG_LOCKED)) {
-               dev_dbg(&s->dev, "Configuration isnt't locked\n");
+               dev_dbg(&p_dev->dev, "Configuration isnt't locked\n");
                mutex_unlock(&s->ops_mutex);
                return -EACCES;
        }
 
-       addr = (c->ConfigBase + reg->Offset) >> 1;
+       addr = (p_dev->config_base + where) >> 1;
+
+       ret = accessf(s, 1, addr, 1, val);
+
        mutex_unlock(&s->ops_mutex);
 
-       switch (reg->Action) {
-       case CS_READ:
-               pcmcia_read_cis_mem(s, 1, addr, 1, &val);
-               reg->Value = val;
-               break;
-       case CS_WRITE:
-               val = reg->Value;
-               pcmcia_write_cis_mem(s, 1, addr, 1, &val);
-               break;
-       default:
-               dev_dbg(&s->dev, "Invalid conf register request\n");
-               return -EINVAL;
-               break;
-       }
-       return 0;
-} /* pcmcia_access_configuration_register */
-EXPORT_SYMBOL(pcmcia_access_configuration_register);
+       return ret;
+}
 
 
-int pcmcia_map_mem_page(struct pcmcia_device *p_dev, window_handle_t wh,
-                       memreq_t *req)
+/**
+ * pcmcia_read_config_byte() - read a byte from a card configuration register
+ *
+ * pcmcia_read_config_byte() reads a byte from a configuration register in
+ * attribute memory.
+ */
+int pcmcia_read_config_byte(struct pcmcia_device *p_dev, off_t where, u8 *val)
+{
+       return pcmcia_access_config(p_dev, where, val, pcmcia_read_cis_mem);
+}
+EXPORT_SYMBOL(pcmcia_read_config_byte);
+
+
+/**
+ * pcmcia_write_config_byte() - write a byte to a card configuration register
+ *
+ * pcmcia_write_config_byte() writes a byte to a configuration register in
+ * attribute memory.
+ */
+int pcmcia_write_config_byte(struct pcmcia_device *p_dev, off_t where, u8 val)
+{
+       return pcmcia_access_config(p_dev, where, &val, pcmcia_write_cis_mem);
+}
+EXPORT_SYMBOL(pcmcia_write_config_byte);
+
+
+/**
+ * pcmcia_map_mem_page() - modify iomem window to point to a different offset
+ * @p_dev: pcmcia device
+ * @res: iomem resource already enabled by pcmcia_request_window()
+ * @offset: card_offset to map
+ *
+ * pcmcia_map_mem_page() modifies what can be read and written by accessing
+ * an iomem range previously enabled by pcmcia_request_window(), by setting
+ * the card_offset value to @offset.
+ */
+int pcmcia_map_mem_page(struct pcmcia_device *p_dev, struct resource *res,
+                       unsigned int offset)
 {
        struct pcmcia_socket *s = p_dev->socket;
+       unsigned int w;
        int ret;
 
-       wh--;
-       if (wh >= MAX_WIN)
+       w = ((res->flags & IORESOURCE_BITS & WIN_FLAGS_REQ) >> 2) - 1;
+       if (w >= MAX_WIN)
                return -EINVAL;
-       if (req->Page != 0) {
-               dev_dbg(&s->dev, "failure: requested page is zero\n");
-               return -EINVAL;
-       }
+
        mutex_lock(&s->ops_mutex);
-       s->win[wh].card_start = req->CardOffset;
-       ret = s->ops->set_mem_map(s, &s->win[wh]);
+       s->win[w].card_start = offset;
+       ret = s->ops->set_mem_map(s, &s->win[w]);
        if (ret)
-               dev_warn(&s->dev, "failed to set_mem_map\n");
+               dev_warn(&p_dev->dev, "failed to set_mem_map\n");
        mutex_unlock(&s->ops_mutex);
        return ret;
-} /* pcmcia_map_mem_page */
+}
 EXPORT_SYMBOL(pcmcia_map_mem_page);
 
 
-/** pcmcia_modify_configuration
+/**
+ * pcmcia_fixup_iowidth() - reduce io width to 8bit
+ * @p_dev: pcmcia device
  *
- * Modify a locked socket configuration
+ * pcmcia_fixup_iowidth() allows a PCMCIA device driver to reduce the
+ * IO width to 8bit after having called pcmcia_enable_device()
+ * previously.
  */
-int pcmcia_modify_configuration(struct pcmcia_device *p_dev,
-                               modconf_t *mod)
+int pcmcia_fixup_iowidth(struct pcmcia_device *p_dev)
 {
-       struct pcmcia_socket *s;
-       config_t *c;
-       int ret;
-
-       s = p_dev->socket;
+       struct pcmcia_socket *s = p_dev->socket;
+       pccard_io_map io_off = { 0, 0, 0, 0, 1 };
+       pccard_io_map io_on;
+       int i, ret = 0;
 
        mutex_lock(&s->ops_mutex);
-       c = p_dev->function_config;
 
-       if (!(s->state & SOCKET_PRESENT)) {
-               dev_dbg(&s->dev, "No card present\n");
-               ret = -ENODEV;
-               goto unlock;
-       }
-       if (!(c->state & CONFIG_LOCKED)) {
-               dev_dbg(&s->dev, "Configuration isnt't locked\n");
+       dev_dbg(&p_dev->dev, "fixup iowidth to 8bit\n");
+
+       if (!(s->state & SOCKET_PRESENT) ||
+               !(p_dev->function_config->state & CONFIG_LOCKED)) {
+               dev_dbg(&p_dev->dev, "No card? Config not locked?\n");
                ret = -EACCES;
                goto unlock;
        }
 
-       if (mod->Attributes & CONF_IRQ_CHANGE_VALID) {
-               if (mod->Attributes & CONF_ENABLE_IRQ) {
-                       c->Attributes |= CONF_ENABLE_IRQ;
-                       s->socket.io_irq = s->irq.AssignedIRQ;
-               } else {
-                       c->Attributes &= ~CONF_ENABLE_IRQ;
-                       s->socket.io_irq = 0;
-               }
-               s->ops->set_socket(s, &s->socket);
+       io_on.speed = io_speed;
+       for (i = 0; i < MAX_IO_WIN; i++) {
+               if (!s->io[i].res)
+                       continue;
+               io_off.map = i;
+               io_on.map = i;
+
+               io_on.flags = MAP_ACTIVE | IO_DATA_PATH_WIDTH_8;
+               io_on.start = s->io[i].res->start;
+               io_on.stop = s->io[i].res->end;
+
+               s->ops->set_io_map(s, &io_off);
+               mdelay(40);
+               s->ops->set_io_map(s, &io_on);
        }
+unlock:
+       mutex_unlock(&s->ops_mutex);
 
-       if (mod->Attributes & CONF_VCC_CHANGE_VALID) {
-               dev_dbg(&s->dev, "changing Vcc is not allowed at this time\n");
-               ret = -EINVAL;
+       return ret;
+}
+EXPORT_SYMBOL(pcmcia_fixup_iowidth);
+
+
+/**
+ * pcmcia_fixup_vpp() - set Vpp to a new voltage level
+ * @p_dev: pcmcia device
+ * @new_vpp: new Vpp voltage
+ *
+ * pcmcia_fixup_vpp() allows a PCMCIA device driver to set Vpp to
+ * a new voltage level between calls to pcmcia_enable_device()
+ * and pcmcia_disable_device().
+ */
+int pcmcia_fixup_vpp(struct pcmcia_device *p_dev, unsigned char new_vpp)
+{
+       struct pcmcia_socket *s = p_dev->socket;
+       int ret = 0;
+
+       mutex_lock(&s->ops_mutex);
+
+       dev_dbg(&p_dev->dev, "fixup Vpp to %d\n", new_vpp);
+
+       if (!(s->state & SOCKET_PRESENT) ||
+               !(p_dev->function_config->state & CONFIG_LOCKED)) {
+               dev_dbg(&p_dev->dev, "No card? Config not locked?\n");
+               ret = -EACCES;
                goto unlock;
        }
 
-       /* We only allow changing Vpp1 and Vpp2 to the same value */
-       if ((mod->Attributes & CONF_VPP1_CHANGE_VALID) &&
-           (mod->Attributes & CONF_VPP2_CHANGE_VALID)) {
-               if (mod->Vpp1 != mod->Vpp2) {
-                       dev_dbg(&s->dev, "Vpp1 and Vpp2 must be the same\n");
-                       ret = -EINVAL;
-                       goto unlock;
-               }
-               s->socket.Vpp = mod->Vpp1;
-               if (s->ops->set_socket(s, &s->socket)) {
-                       dev_printk(KERN_WARNING, &s->dev,
-                                  "Unable to set VPP\n");
-                       ret = -EIO;
-                       goto unlock;
-               }
-       } else if ((mod->Attributes & CONF_VPP1_CHANGE_VALID) ||
-                  (mod->Attributes & CONF_VPP2_CHANGE_VALID)) {
-               dev_dbg(&s->dev, "changing Vcc is not allowed at this time\n");
-               ret = -EINVAL;
+       s->socket.Vpp = new_vpp;
+       if (s->ops->set_socket(s, &s->socket)) {
+               dev_warn(&p_dev->dev, "Unable to set VPP\n");
+               ret = -EIO;
                goto unlock;
        }
+       p_dev->vpp = new_vpp;
 
-       if (mod->Attributes & CONF_IO_CHANGE_WIDTH) {
-               pccard_io_map io_off = { 0, 0, 0, 0, 1 };
-               pccard_io_map io_on;
-               int i;
-
-               io_on.speed = io_speed;
-               for (i = 0; i < MAX_IO_WIN; i++) {
-                       if (!s->io[i].res)
-                               continue;
-                       io_off.map = i;
-                       io_on.map = i;
-
-                       io_on.flags = MAP_ACTIVE | IO_DATA_PATH_WIDTH_8;
-                       io_on.start = s->io[i].res->start;
-                       io_on.stop = s->io[i].res->end;
-
-                       s->ops->set_io_map(s, &io_off);
-                       mdelay(40);
-                       s->ops->set_io_map(s, &io_on);
-               }
-       }
-       ret = 0;
 unlock:
        mutex_unlock(&s->ops_mutex);
 
        return ret;
-} /* modify_configuration */
-EXPORT_SYMBOL(pcmcia_modify_configuration);
+}
+EXPORT_SYMBOL(pcmcia_fixup_vpp);
 
 
+/**
+ * pcmcia_release_configuration() - physically disable a PCMCIA device
+ * @p_dev: pcmcia device
+ *
+ * pcmcia_release_configuration() is the 1:1 counterpart to
+ * pcmcia_enable_device(): If a PCMCIA device is no longer used by any
+ * driver, the Vpp voltage is set to 0, IRQs will no longer be generated,
+ * and I/O ranges will be disabled. As pcmcia_release_io() and
+ * pcmcia_release_window() still need to be called, device drivers are
+ * expected to call pcmcia_disable_device() instead.
+ */
 int pcmcia_release_configuration(struct pcmcia_device *p_dev)
 {
        pccard_io_map io = { 0, 0, 0, 0, 1 };
@@ -355,7 +359,7 @@ int pcmcia_release_configuration(struct pcmcia_device *p_dev)
        if (p_dev->_locked) {
                p_dev->_locked = 0;
                if (--(s->lock_count) == 0) {
-                       s->socket.flags = SS_OUTPUT_ENA;   /* Is this correct? */
+                       s->socket.flags = SS_OUTPUT_ENA; /* Is this correct? */
                        s->socket.Vpp = 0;
                        s->socket.io_irq = 0;
                        s->ops->set_socket(s, &s->socket);
@@ -377,110 +381,71 @@ int pcmcia_release_configuration(struct pcmcia_device *p_dev)
        mutex_unlock(&s->ops_mutex);
 
        return 0;
-} /* pcmcia_release_configuration */
+}
 
 
-/** pcmcia_release_io
+/**
+ * pcmcia_release_io() - release I/O allocated by a PCMCIA device
+ * @p_dev: pcmcia device
  *
- * Release_io() releases the I/O ranges allocated by a client.  This
- * may be invoked some time after a card ejection has already dumped
- * the actual socket configuration, so if the client is "stale", we
- * don't bother checking the port ranges against the current socket
- * values.
+ * pcmcia_release_io() releases the I/O ranges allocated by a PCMCIA
+ * device.  This may be invoked some time after a card ejection has
+ * already dumped the actual socket configuration, so if the client is
+ * "stale", we don't bother checking the port ranges against the
+ * current socket values.
  */
-static int pcmcia_release_io(struct pcmcia_device *p_dev, io_req_t *req)
+static int pcmcia_release_io(struct pcmcia_device *p_dev)
 {
        struct pcmcia_socket *s = p_dev->socket;
        int ret = -EINVAL;
        config_t *c;
 
        mutex_lock(&s->ops_mutex);
-       c = p_dev->function_config;
-
        if (!p_dev->_io)
                goto out;
 
-       p_dev->_io = 0;
-
-       if ((c->io.BasePort1 != req->BasePort1) ||
-           (c->io.NumPorts1 != req->NumPorts1) ||
-           (c->io.BasePort2 != req->BasePort2) ||
-           (c->io.NumPorts2 != req->NumPorts2))
-               goto out;
-
-       c->state &= ~CONFIG_IO_REQ;
-
-       release_io_space(s, req->BasePort1, req->NumPorts1);
-       if (req->NumPorts2)
-               release_io_space(s, req->BasePort2, req->NumPorts2);
-
-out:
-       mutex_unlock(&s->ops_mutex);
-
-       return ret;
-} /* pcmcia_release_io */
-
-
-static int pcmcia_release_irq(struct pcmcia_device *p_dev, irq_req_t *req)
-{
-       struct pcmcia_socket *s = p_dev->socket;
-       config_t *c;
-       int ret = -EINVAL;
-
-       mutex_lock(&s->ops_mutex);
-
        c = p_dev->function_config;
 
-       if (!p_dev->_irq)
-               goto out;
-
-       p_dev->_irq = 0;
-
-       if (c->state & CONFIG_LOCKED)
-               goto out;
-
-       if (c->irq.Attributes != req->Attributes) {
-               dev_dbg(&s->dev, "IRQ attributes must match assigned ones\n");
-               goto out;
-       }
-       if (s->irq.AssignedIRQ != req->AssignedIRQ) {
-               dev_dbg(&s->dev, "IRQ must match assigned one\n");
-               goto out;
-       }
-       if (--s->irq.Config == 0) {
-               c->state &= ~CONFIG_IRQ_REQ;
-               s->irq.AssignedIRQ = 0;
-       }
+       release_io_space(s, &c->io[0]);
 
-       if (req->Handler)
-               free_irq(req->AssignedIRQ, p_dev->priv);
+       if (c->io[1].end)
+               release_io_space(s, &c->io[1]);
 
-#ifdef CONFIG_PCMCIA_PROBE
-       pcmcia_used_irq[req->AssignedIRQ]--;
-#endif
-       ret = 0;
+       p_dev->_io = 0;
+       c->state &= ~CONFIG_IO_REQ;
 
 out:
        mutex_unlock(&s->ops_mutex);
 
        return ret;
-} /* pcmcia_release_irq */
+} /* pcmcia_release_io */
 
 
-int pcmcia_release_window(struct pcmcia_device *p_dev, window_handle_t wh)
+/**
+ * pcmcia_release_window() - release reserved iomem for PCMCIA devices
+ * @p_dev: pcmcia device
+ * @res: iomem resource to release
+ *
+ * pcmcia_release_window() releases &struct resource *res which was
+ * previously reserved by calling pcmcia_request_window().
+ */
+int pcmcia_release_window(struct pcmcia_device *p_dev, struct resource *res)
 {
        struct pcmcia_socket *s = p_dev->socket;
        pccard_mem_map *win;
+       unsigned int w;
+
+       dev_dbg(&p_dev->dev, "releasing window %pR\n", res);
 
-       wh--;
-       if (wh >= MAX_WIN)
+       w = ((res->flags & IORESOURCE_BITS & WIN_FLAGS_REQ) >> 2) - 1;
+       if (w >= MAX_WIN)
                return -EINVAL;
 
        mutex_lock(&s->ops_mutex);
-       win = &s->win[wh];
+       win = &s->win[w];
 
-       if (!(p_dev->_win & CLIENT_WIN_REQ(wh))) {
-               dev_dbg(&s->dev, "not releasing unknown window\n");
+       if (!(p_dev->_win & CLIENT_WIN_REQ(w))) {
+               dev_dbg(&p_dev->dev, "not releasing unknown window\n");
                mutex_unlock(&s->ops_mutex);
                return -EINVAL;
        }
@@ -488,15 +453,18 @@ int pcmcia_release_window(struct pcmcia_device *p_dev, window_handle_t wh)
        /* Shut down memory window */
        win->flags &= ~MAP_ACTIVE;
        s->ops->set_mem_map(s, win);
-       s->state &= ~SOCKET_WIN_REQ(wh);
+       s->state &= ~SOCKET_WIN_REQ(w);
 
        /* Release system memory */
        if (win->res) {
+               release_resource(res);
                release_resource(win->res);
                kfree(win->res);
                win->res = NULL;
        }
-       p_dev->_win &= ~CLIENT_WIN_REQ(wh);
+       res->start = res->end = 0;
+       res->flags = IORESOURCE_MEM;
+       p_dev->_win &= ~CLIENT_WIN_REQ(w);
        mutex_unlock(&s->ops_mutex);
 
        return 0;
@@ -504,107 +472,124 @@ int pcmcia_release_window(struct pcmcia_device *p_dev, window_handle_t wh)
 EXPORT_SYMBOL(pcmcia_release_window);
 
 
-int pcmcia_request_configuration(struct pcmcia_device *p_dev,
-                                config_req_t *req)
+/**
+ * pcmcia_enable_device() - set up and activate a PCMCIA device
+ * @p_dev: the associated PCMCIA device
+ *
+ * pcmcia_enable_device() physically enables a PCMCIA device. It parses
+ * the flags passed to in @flags and stored in @p_dev->flags and sets up
+ * the Vpp voltage, enables the speaker line, I/O ports and store proper
+ * values to configuration registers.
+ */
+int pcmcia_enable_device(struct pcmcia_device *p_dev)
 {
        int i;
-       u_int base;
+       unsigned int base;
        struct pcmcia_socket *s = p_dev->socket;
        config_t *c;
        pccard_io_map iomap;
+       unsigned char status = 0;
+       unsigned char ext_status = 0;
+       unsigned char option = 0;
+       unsigned int flags = p_dev->config_flags;
 
        if (!(s->state & SOCKET_PRESENT))
                return -ENODEV;
 
-       if (req->IntType & INT_CARDBUS) {
-               dev_dbg(&s->dev, "IntType may not be INT_CARDBUS\n");
-               return -EINVAL;
-       }
-
        mutex_lock(&s->ops_mutex);
        c = p_dev->function_config;
        if (c->state & CONFIG_LOCKED) {
                mutex_unlock(&s->ops_mutex);
-               dev_dbg(&s->dev, "Configuration is locked\n");
+               dev_dbg(&p_dev->dev, "Configuration is locked\n");
                return -EACCES;
        }
 
        /* Do power control.  We don't allow changes in Vcc. */
-       s->socket.Vpp = req->Vpp;
+       s->socket.Vpp = p_dev->vpp;
        if (s->ops->set_socket(s, &s->socket)) {
                mutex_unlock(&s->ops_mutex);
-               dev_printk(KERN_WARNING, &s->dev,
+               dev_printk(KERN_WARNING, &p_dev->dev,
                           "Unable to set socket state\n");
                return -EINVAL;
        }
 
        /* Pick memory or I/O card, DMA mode, interrupt */
-       c->IntType = req->IntType;
-       c->Attributes = req->Attributes;
-       if (req->IntType & INT_MEMORY_AND_IO)
+       if (p_dev->_io || flags & CONF_ENABLE_IRQ)
+               flags |= CONF_ENABLE_IOCARD;
+       if (flags & CONF_ENABLE_IOCARD)
                s->socket.flags |= SS_IOCARD;
-       if (req->IntType & INT_ZOOMED_VIDEO)
+       if (flags & CONF_ENABLE_ZVCARD)
                s->socket.flags |= SS_ZVCARD | SS_IOCARD;
-       if (req->Attributes & CONF_ENABLE_DMA)
-               s->socket.flags |= SS_DMA_MODE;
-       if (req->Attributes & CONF_ENABLE_SPKR)
+       if (flags & CONF_ENABLE_SPKR) {
                s->socket.flags |= SS_SPKR_ENA;
-       if (req->Attributes & CONF_ENABLE_IRQ)
-               s->socket.io_irq = s->irq.AssignedIRQ;
+               status = CCSR_AUDIO_ENA;
+               if (!(p_dev->config_regs & PRESENT_STATUS))
+                       dev_warn(&p_dev->dev, "speaker requested, but "
+                                             "PRESENT_STATUS not set!\n");
+       }
+       if (flags & CONF_ENABLE_IRQ)
+               s->socket.io_irq = s->pcmcia_irq;
        else
                s->socket.io_irq = 0;
+       if (flags & CONF_ENABLE_ESR) {
+               p_dev->config_regs |= PRESENT_EXT_STATUS;
+               ext_status = ESR_REQ_ATTN_ENA;
+       }
        s->ops->set_socket(s, &s->socket);
        s->lock_count++;
-       mutex_unlock(&s->ops_mutex);
+
+       dev_dbg(&p_dev->dev,
+               "enable_device: V %d, flags %x, base %x, regs %x, idx %x\n",
+               p_dev->vpp, flags, p_dev->config_base, p_dev->config_regs,
+               p_dev->config_index);
 
        /* Set up CIS configuration registers */
-       base = c->ConfigBase = req->ConfigBase;
-       c->CardValues = req->Present;
-       if (req->Present & PRESENT_COPY) {
-               c->Copy = req->Copy;
-               pcmcia_write_cis_mem(s, 1, (base + CISREG_SCR)>>1, 1, &c->Copy);
-       }
-       if (req->Present & PRESENT_OPTION) {
+       base = p_dev->config_base;
+       if (p_dev->config_regs & PRESENT_COPY) {
+               u16 tmp = 0;
+               dev_dbg(&p_dev->dev, "clearing CISREG_SCR\n");
+               pcmcia_write_cis_mem(s, 1, (base + CISREG_SCR)>>1, 1, &tmp);
+       }
+       if (p_dev->config_regs & PRESENT_PIN_REPLACE) {
+               u16 tmp = 0;
+               dev_dbg(&p_dev->dev, "clearing CISREG_PRR\n");
+               pcmcia_write_cis_mem(s, 1, (base + CISREG_PRR)>>1, 1, &tmp);
+       }
+       if (p_dev->config_regs & PRESENT_OPTION) {
                if (s->functions == 1) {
-                       c->Option = req->ConfigIndex & COR_CONFIG_MASK;
+                       option = p_dev->config_index & COR_CONFIG_MASK;
                } else {
-                       c->Option = req->ConfigIndex & COR_MFC_CONFIG_MASK;
-                       c->Option |= COR_FUNC_ENA|COR_IREQ_ENA;
-                       if (req->Present & PRESENT_IOBASE_0)
-                               c->Option |= COR_ADDR_DECODE;
+                       option = p_dev->config_index & COR_MFC_CONFIG_MASK;
+                       option |= COR_FUNC_ENA|COR_IREQ_ENA;
+                       if (p_dev->config_regs & PRESENT_IOBASE_0)
+                               option |= COR_ADDR_DECODE;
                }
-               if (c->state & CONFIG_IRQ_REQ)
-                       if (!(c->irq.Attributes & IRQ_FORCED_PULSE))
-                               c->Option |= COR_LEVEL_REQ;
-               pcmcia_write_cis_mem(s, 1, (base + CISREG_COR)>>1, 1, &c->Option);
+               if ((flags & CONF_ENABLE_IRQ) &&
+                       !(flags & CONF_ENABLE_PULSE_IRQ))
+                       option |= COR_LEVEL_REQ;
+               pcmcia_write_cis_mem(s, 1, (base + CISREG_COR)>>1, 1, &option);
                mdelay(40);
        }
-       if (req->Present & PRESENT_STATUS) {
-               c->Status = req->Status;
-               pcmcia_write_cis_mem(s, 1, (base + CISREG_CCSR)>>1, 1, &c->Status);
-       }
-       if (req->Present & PRESENT_PIN_REPLACE) {
-               c->Pin = req->Pin;
-               pcmcia_write_cis_mem(s, 1, (base + CISREG_PRR)>>1, 1, &c->Pin);
-       }
-       if (req->Present & PRESENT_EXT_STATUS) {
-               c->ExtStatus = req->ExtStatus;
-               pcmcia_write_cis_mem(s, 1, (base + CISREG_ESR)>>1, 1, &c->ExtStatus);
-       }
-       if (req->Present & PRESENT_IOBASE_0) {
-               u_char b = c->io.BasePort1 & 0xff;
+       if (p_dev->config_regs & PRESENT_STATUS)
+               pcmcia_write_cis_mem(s, 1, (base + CISREG_CCSR)>>1, 1, &status);
+
+       if (p_dev->config_regs & PRESENT_EXT_STATUS)
+               pcmcia_write_cis_mem(s, 1, (base + CISREG_ESR)>>1, 1,
+                                       &ext_status);
+
+       if (p_dev->config_regs & PRESENT_IOBASE_0) {
+               u8 b = c->io[0].start & 0xff;
                pcmcia_write_cis_mem(s, 1, (base + CISREG_IOBASE_0)>>1, 1, &b);
-               b = (c->io.BasePort1 >> 8) & 0xff;
+               b = (c->io[0].start >> 8) & 0xff;
                pcmcia_write_cis_mem(s, 1, (base + CISREG_IOBASE_1)>>1, 1, &b);
        }
-       if (req->Present & PRESENT_IOSIZE) {
-               u_char b = c->io.NumPorts1 + c->io.NumPorts2 - 1;
+       if (p_dev->config_regs & PRESENT_IOSIZE) {
+               u8 b = resource_size(&c->io[0]) + resource_size(&c->io[1]) - 1;
                pcmcia_write_cis_mem(s, 1, (base + CISREG_IOSIZE)>>1, 1, &b);
        }
 
        /* Configure I/O windows */
        if (c->state & CONFIG_IO_REQ) {
-               mutex_lock(&s->ops_mutex);
                iomap.speed = io_speed;
                for (i = 0; i < MAX_IO_WIN; i++)
                        if (s->io[i].res) {
@@ -623,80 +608,74 @@ int pcmcia_request_configuration(struct pcmcia_device *p_dev,
                                s->ops->set_io_map(s, &iomap);
                                s->io[i].Config++;
                        }
-               mutex_unlock(&s->ops_mutex);
        }
 
        c->state |= CONFIG_LOCKED;
        p_dev->_locked = 1;
+       mutex_unlock(&s->ops_mutex);
        return 0;
-} /* pcmcia_request_configuration */
-EXPORT_SYMBOL(pcmcia_request_configuration);
+} /* pcmcia_enable_device */
+EXPORT_SYMBOL(pcmcia_enable_device);
 
 
-/** pcmcia_request_io
+/**
+ * pcmcia_request_io() - attempt to reserve port ranges for PCMCIA devices
+ * @p_dev: the associated PCMCIA device
  *
- * Request_io() reserves ranges of port addresses for a socket.
- * I have not implemented range sharing or alias addressing.
+ * pcmcia_request_io() attempts to reserve the IO port ranges specified in
+ * &struct pcmcia_device @p_dev->resource[0] and @p_dev->resource[1]. The
+ * "start" value is the requested start of the IO port resource; "end"
+ * reflects the number of ports requested. The number of IO lines requested
+ * is specified in &struct pcmcia_device @p_dev->io_lines.
  */
-int pcmcia_request_io(struct pcmcia_device *p_dev, io_req_t *req)
+int pcmcia_request_io(struct pcmcia_device *p_dev)
 {
        struct pcmcia_socket *s = p_dev->socket;
-       config_t *c;
+       config_t *c = p_dev->function_config;
        int ret = -EINVAL;
 
        mutex_lock(&s->ops_mutex);
+       dev_dbg(&p_dev->dev, "pcmcia_request_io: %pR , %pR",
+               &c->io[0], &c->io[1]);
 
        if (!(s->state & SOCKET_PRESENT)) {
-               dev_dbg(&s->dev, "No card present\n");
+               dev_dbg(&p_dev->dev, "pcmcia_request_io: No card present\n");
                goto out;
        }
 
-       if (!req)
-               goto out;
-
-       c = p_dev->function_config;
        if (c->state & CONFIG_LOCKED) {
-               dev_dbg(&s->dev, "Configuration is locked\n");
+               dev_dbg(&p_dev->dev, "Configuration is locked\n");
                goto out;
        }
        if (c->state & CONFIG_IO_REQ) {
-               dev_dbg(&s->dev, "IO already configured\n");
-               goto out;
-       }
-       if (req->Attributes1 & (IO_SHARED | IO_FORCE_ALIAS_ACCESS)) {
-               dev_dbg(&s->dev, "bad attribute setting for IO region 1\n");
-               goto out;
-       }
-       if ((req->NumPorts2 > 0) &&
-           (req->Attributes2 & (IO_SHARED | IO_FORCE_ALIAS_ACCESS))) {
-               dev_dbg(&s->dev, "bad attribute setting for IO region 2\n");
+               dev_dbg(&p_dev->dev, "IO already configured\n");
                goto out;
        }
 
-       dev_dbg(&s->dev, "trying to allocate resource 1\n");
-       ret = alloc_io_space(s, req->Attributes1, &req->BasePort1,
-                            req->NumPorts1, req->IOAddrLines);
-       if (ret) {
-               dev_dbg(&s->dev, "allocation of resource 1 failed\n");
+       ret = alloc_io_space(s, &c->io[0], p_dev->io_lines);
+       if (ret)
                goto out;
-       }
 
-       if (req->NumPorts2) {
-               dev_dbg(&s->dev, "trying to allocate resource 2\n");
-               ret = alloc_io_space(s, req->Attributes2, &req->BasePort2,
-                                    req->NumPorts2, req->IOAddrLines);
+       if (c->io[1].end) {
+               ret = alloc_io_space(s, &c->io[1], p_dev->io_lines);
                if (ret) {
-                       dev_dbg(&s->dev, "allocation of resource 2 failed\n");
-                       release_io_space(s, req->BasePort1, req->NumPorts1);
+                       struct resource tmp = c->io[0];
+                       /* release the previously allocated resource */
+                       release_io_space(s, &c->io[0]);
+                       /* but preserve the settings, for they worked... */
+                       c->io[0].end = resource_size(&tmp);
+                       c->io[0].start = tmp.start;
+                       c->io[0].flags = tmp.flags;
                        goto out;
                }
-       }
+       } else
+               c->io[1].start = 0;
 
-       c->io = *req;
        c->state |= CONFIG_IO_REQ;
        p_dev->_io = 1;
-       dev_dbg(&s->dev, "allocating resources succeeded: %d\n", ret);
 
+       dev_dbg(&p_dev->dev, "pcmcia_request_io succeeded: %pR , %pR",
+               &c->io[0], &c->io[1]);
 out:
        mutex_unlock(&s->ops_mutex);
 
@@ -705,195 +684,246 @@ out:
 EXPORT_SYMBOL(pcmcia_request_io);
 
 
-/** pcmcia_request_irq
+/**
+ * pcmcia_request_irq() - attempt to request a IRQ for a PCMCIA device
+ * @p_dev: the associated PCMCIA device
+ * @handler: IRQ handler to register
  *
- * Request_irq() reserves an irq for this client.
+ * pcmcia_request_irq() is a wrapper around request_irq() which allows
+ * the PCMCIA core to clean up the registration in pcmcia_disable_device().
+ * Drivers are free to use request_irq() directly, but then they need to
+ * call free_irq() themselfves, too. Also, only %IRQF_SHARED capable IRQ
+ * handlers are allowed.
+ */
+int __must_check pcmcia_request_irq(struct pcmcia_device *p_dev,
+                                   irq_handler_t handler)
+{
+       int ret;
+
+       if (!p_dev->irq)
+               return -EINVAL;
+
+       ret = request_irq(p_dev->irq, handler, IRQF_SHARED,
+                       p_dev->devname, p_dev->priv);
+       if (!ret)
+               p_dev->_irq = 1;
+
+       return ret;
+}
+EXPORT_SYMBOL(pcmcia_request_irq);
+
+
+/**
+ * pcmcia_request_exclusive_irq() - attempt to request an exclusive IRQ first
+ * @p_dev: the associated PCMCIA device
+ * @handler: IRQ handler to register
  *
- * Also, since Linux only reserves irq's when they are actually
- * hooked, we don't guarantee that an irq will still be available
- * when the configuration is locked.  Now that I think about it,
- * there might be a way to fix this using a dummy handler.
+ * pcmcia_request_exclusive_irq() is a wrapper around request_irq() which
+ * attempts first to request an exclusive IRQ. If it fails, it also accepts
+ * a shared IRQ, but prints out a warning. PCMCIA drivers should allow for
+ * IRQ sharing and either use request_irq directly (then they need to call
+ * free_irq() themselves, too), or the pcmcia_request_irq() function.
  */
+int __must_check
+__pcmcia_request_exclusive_irq(struct pcmcia_device *p_dev,
+                       irq_handler_t handler)
+{
+       int ret;
+
+       if (!p_dev->irq)
+               return -EINVAL;
+
+       ret = request_irq(p_dev->irq, handler, 0, p_dev->devname, p_dev->priv);
+       if (ret) {
+               ret = pcmcia_request_irq(p_dev, handler);
+               dev_printk(KERN_WARNING, &p_dev->dev, "pcmcia: "
+                       "request for exclusive IRQ could not be fulfilled.\n");
+               dev_printk(KERN_WARNING, &p_dev->dev, "pcmcia: the driver "
+                       "needs updating to supported shared IRQ lines.\n");
+       }
+       if (ret)
+               dev_printk(KERN_INFO, &p_dev->dev, "request_irq() failed\n");
+       else
+               p_dev->_irq = 1;
+
+       return ret;
+} /* pcmcia_request_exclusive_irq */
+EXPORT_SYMBOL(__pcmcia_request_exclusive_irq);
+
 
 #ifdef CONFIG_PCMCIA_PROBE
+
+/* mask of IRQs already reserved by other cards, we should avoid using them */
+static u8 pcmcia_used_irq[32];
+
 static irqreturn_t test_action(int cpl, void *dev_id)
 {
        return IRQ_NONE;
 }
-#endif
 
-int pcmcia_request_irq(struct pcmcia_device *p_dev, irq_req_t *req)
+/**
+ * pcmcia_setup_isa_irq() - determine whether an ISA IRQ can be used
+ * @p_dev - the associated PCMCIA device
+ *
+ * locking note: must be called with ops_mutex locked.
+ */
+static int pcmcia_setup_isa_irq(struct pcmcia_device *p_dev, int type)
 {
        struct pcmcia_socket *s = p_dev->socket;
-       config_t *c;
-       int ret = -EINVAL, irq = 0;
-       int type;
+       unsigned int try, irq;
+       u32 mask = s->irq_mask;
+       int ret = -ENODEV;
 
-       mutex_lock(&s->ops_mutex);
+       for (try = 0; try < 64; try++) {
+               irq = try % 32;
 
-       if (!(s->state & SOCKET_PRESENT)) {
-               dev_dbg(&s->dev, "No card present\n");
-               goto out;
-       }
-       c = p_dev->function_config;
-       if (c->state & CONFIG_LOCKED) {
-               dev_dbg(&s->dev, "Configuration is locked\n");
-               goto out;
-       }
-       if (c->state & CONFIG_IRQ_REQ) {
-               dev_dbg(&s->dev, "IRQ already configured\n");
-               goto out;
-       }
+               if (irq > NR_IRQS)
+                       continue;
 
-       /* Decide what type of interrupt we are registering */
-       type = 0;
-       if (s->functions > 1)           /* All of this ought to be handled higher up */
-               type = IRQF_SHARED;
-       else if (req->Attributes & IRQ_TYPE_DYNAMIC_SHARING)
-               type = IRQF_SHARED;
-       else
-               printk(KERN_WARNING "pcmcia: Driver needs updating to support IRQ sharing.\n");
+               /* marked as available by driver, not blocked by userspace? */
+               if (!((mask >> irq) & 1))
+                       continue;
 
-#ifdef CONFIG_PCMCIA_PROBE
+               /* avoid an IRQ which is already used by another PCMCIA card */
+               if ((try < 32) && pcmcia_used_irq[irq])
+                       continue;
 
-       if (s->irq.AssignedIRQ != 0) {
-               /* If the interrupt is already assigned, it must be the same */
-               irq = s->irq.AssignedIRQ;
-       } else {
-               int try;
-               u32 mask = s->irq_mask;
-               void *data = p_dev; /* something unique to this device */
-
-               for (try = 0; try < 64; try++) {
-                       irq = try % 32;
-
-                       /* marked as available by driver, and not blocked by userspace? */
-                       if (!((mask >> irq) & 1))
-                               continue;
-
-                       /* avoid an IRQ which is already used by a PCMCIA card */
-                       if ((try < 32) && pcmcia_used_irq[irq])
-                               continue;
-
-                       /* register the correct driver, if possible, of check whether
-                        * registering a dummy handle works, i.e. if the IRQ isn't
-                        * marked as used by the kernel resource management core */
-                       ret = request_irq(irq,
-                                         (req->Handler) ? req->Handler : test_action,
-                                         type,
-                                         p_dev->devname,
-                                         (req->Handler) ? p_dev->priv : data);
-                       if (!ret) {
-                               if (!req->Handler)
-                                       free_irq(irq, data);
-                               break;
-                       }
-               }
-       }
-#endif
-       /* only assign PCI irq if no IRQ already assigned */
-       if (ret && !s->irq.AssignedIRQ) {
-               if (!s->pci_irq) {
-                       dev_printk(KERN_INFO, &s->dev, "no IRQ found\n");
-                       goto out;
+               /* register the correct driver, if possible, to check whether
+                * registering a dummy handle works, i.e. if the IRQ isn't
+                * marked as used by the kernel resource management core */
+               ret = request_irq(irq, test_action, type, p_dev->devname,
+                                 p_dev);
+               if (!ret) {
+                       free_irq(irq, p_dev);
+                       p_dev->irq = s->pcmcia_irq = irq;
+                       pcmcia_used_irq[irq]++;
+                       break;
                }
-               type = IRQF_SHARED;
-               irq = s->pci_irq;
        }
 
-       if (ret && req->Handler) {
-               ret = request_irq(irq, req->Handler, type,
-                                 p_dev->devname, p_dev->priv);
-               if (ret) {
-                       dev_printk(KERN_INFO, &s->dev,
-                               "request_irq() failed\n");
-                       goto out;
-               }
-       }
+       return ret;
+}
 
-       /* Make sure the fact the request type was overridden is passed back */
-       if (type == IRQF_SHARED && !(req->Attributes & IRQ_TYPE_DYNAMIC_SHARING)) {
-               req->Attributes |= IRQ_TYPE_DYNAMIC_SHARING;
-               dev_printk(KERN_WARNING, &p_dev->dev, "pcmcia: "
-                       "request for exclusive IRQ could not be fulfilled.\n");
-               dev_printk(KERN_WARNING, &p_dev->dev, "pcmcia: the driver "
-                       "needs updating to supported shared IRQ lines.\n");
+void pcmcia_cleanup_irq(struct pcmcia_socket *s)
+{
+       pcmcia_used_irq[s->pcmcia_irq]--;
+       s->pcmcia_irq = 0;
+}
+
+#else /* CONFIG_PCMCIA_PROBE */
+
+static int pcmcia_setup_isa_irq(struct pcmcia_device *p_dev, int type)
+{
+       return -EINVAL;
+}
+
+void pcmcia_cleanup_irq(struct pcmcia_socket *s)
+{
+       s->pcmcia_irq = 0;
+       return;
+}
+
+#endif  /* CONFIG_PCMCIA_PROBE */
+
+
+/**
+ * pcmcia_setup_irq() - determine IRQ to be used for device
+ * @p_dev - the associated PCMCIA device
+ *
+ * locking note: must be called with ops_mutex locked.
+ */
+int pcmcia_setup_irq(struct pcmcia_device *p_dev)
+{
+       struct pcmcia_socket *s = p_dev->socket;
+
+       if (p_dev->irq)
+               return 0;
+
+       /* already assigned? */
+       if (s->pcmcia_irq) {
+               p_dev->irq = s->pcmcia_irq;
+               return 0;
        }
-       c->irq.Attributes = req->Attributes;
-       s->irq.AssignedIRQ = req->AssignedIRQ = irq;
-       s->irq.Config++;
 
-       c->state |= CONFIG_IRQ_REQ;
-       p_dev->_irq = 1;
+       /* prefer an exclusive ISA irq */
+       if (!pcmcia_setup_isa_irq(p_dev, 0))
+               return 0;
 
-#ifdef CONFIG_PCMCIA_PROBE
-       pcmcia_used_irq[irq]++;
-#endif
+       /* but accept a shared ISA irq */
+       if (!pcmcia_setup_isa_irq(p_dev, IRQF_SHARED))
+               return 0;
 
-       ret = 0;
-out:
-       mutex_unlock(&s->ops_mutex);
-       return ret;
-} /* pcmcia_request_irq */
-EXPORT_SYMBOL(pcmcia_request_irq);
+       /* but use the PCI irq otherwise */
+       if (s->pci_irq) {
+               p_dev->irq = s->pcmcia_irq = s->pci_irq;
+               return 0;
+       }
+
+       return -EINVAL;
+}
 
 
-/** pcmcia_request_window
+/**
+ * pcmcia_request_window() - attempt to reserve iomem for PCMCIA devices
+ * @p_dev: the associated PCMCIA device
+ * @res: &struct resource pointing to p_dev->resource[2..5]
+ * @speed: access speed
  *
- * Request_window() establishes a mapping between card memory space
- * and system memory space.
+ * pcmcia_request_window() attepts to reserve an iomem ranges specified in
+ * &struct resource @res pointing to one of the entries in
+ * &struct pcmcia_device @p_dev->resource[2..5]. The "start" value is the
+ * requested start of the IO mem resource; "end" reflects the size
+ * requested.
  */
-int pcmcia_request_window(struct pcmcia_device *p_dev, win_req_t *req, window_handle_t *wh)
+int pcmcia_request_window(struct pcmcia_device *p_dev, struct resource *res,
+                       unsigned int speed)
 {
        struct pcmcia_socket *s = p_dev->socket;
        pccard_mem_map *win;
        u_long align;
        int w;
 
+       dev_dbg(&p_dev->dev, "request_window %pR %d\n", res, speed);
+
        if (!(s->state & SOCKET_PRESENT)) {
-               dev_dbg(&s->dev, "No card present\n");
+               dev_dbg(&p_dev->dev, "No card present\n");
                return -ENODEV;
        }
-       if (req->Attributes & (WIN_PAGED | WIN_SHARED)) {
-               dev_dbg(&s->dev, "bad attribute setting for iomem region\n");
-               return -EINVAL;
-       }
 
        /* Window size defaults to smallest available */
-       if (req->Size == 0)
-               req->Size = s->map_size;
-       align = (((s->features & SS_CAP_MEM_ALIGN) ||
-                 (req->Attributes & WIN_STRICT_ALIGN)) ?
-                req->Size : s->map_size);
-       if (req->Size & (s->map_size-1)) {
-               dev_dbg(&s->dev, "invalid map size\n");
+       if (res->end == 0)
+               res->end = s->map_size;
+       align = (s->features & SS_CAP_MEM_ALIGN) ? res->end : s->map_size;
+       if (res->end & (s->map_size-1)) {
+               dev_dbg(&p_dev->dev, "invalid map size\n");
                return -EINVAL;
        }
-       if ((req->Base && (s->features & SS_CAP_STATIC_MAP)) ||
-           (req->Base & (align-1))) {
-               dev_dbg(&s->dev, "invalid base address\n");
+       if ((res->start && (s->features & SS_CAP_STATIC_MAP)) ||
+           (res->start & (align-1))) {
+               dev_dbg(&p_dev->dev, "invalid base address\n");
                return -EINVAL;
        }
-       if (req->Base)
+       if (res->start)
                align = 0;
 
        /* Allocate system memory window */
+       mutex_lock(&s->ops_mutex);
        for (w = 0; w < MAX_WIN; w++)
                if (!(s->state & SOCKET_WIN_REQ(w)))
                        break;
        if (w == MAX_WIN) {
-               dev_dbg(&s->dev, "all windows are used already\n");
+               dev_dbg(&p_dev->dev, "all windows are used already\n");
+               mutex_unlock(&s->ops_mutex);
                return -EINVAL;
        }
 
-       mutex_lock(&s->ops_mutex);
        win = &s->win[w];
 
        if (!(s->features & SS_CAP_STATIC_MAP)) {
-               win->res = pcmcia_find_mem_region(req->Base, req->Size, align,
-                                                     (req->Attributes & WIN_MAP_BELOW_1MB), s);
+               win->res = pcmcia_find_mem_region(res->start, res->end, align,
+                                               0, s);
                if (!win->res) {
-                       dev_dbg(&s->dev, "allocating mem region failed\n");
+                       dev_dbg(&p_dev->dev, "allocating mem region failed\n");
                        mutex_unlock(&s->ops_mutex);
                        return -EINVAL;
                }
@@ -902,20 +932,12 @@ int pcmcia_request_window(struct pcmcia_device *p_dev, win_req_t *req, window_ha
 
        /* Configure the socket controller */
        win->map = w+1;
-       win->flags = 0;
-       win->speed = req->AccessSpeed;
-       if (req->Attributes & WIN_MEMORY_TYPE)
-               win->flags |= MAP_ATTRIB;
-       if (req->Attributes & WIN_ENABLE)
-               win->flags |= MAP_ACTIVE;
-       if (req->Attributes & WIN_DATA_WIDTH_16)
-               win->flags |= MAP_16BIT;
-       if (req->Attributes & WIN_USE_WAIT)
-               win->flags |= MAP_USE_WAIT;
+       win->flags = res->flags & WIN_FLAGS_MAP;
+       win->speed = speed;
        win->card_start = 0;
 
        if (s->ops->set_mem_map(s, win) != 0) {
-               dev_dbg(&s->dev, "failed to set memory mapping\n");
+               dev_dbg(&p_dev->dev, "failed to set memory mapping\n");
                mutex_unlock(&s->ops_mutex);
                return -EIO;
        }
@@ -923,252 +945,56 @@ int pcmcia_request_window(struct pcmcia_device *p_dev, win_req_t *req, window_ha
 
        /* Return window handle */
        if (s->features & SS_CAP_STATIC_MAP)
-               req->Base = win->static_start;
+               res->start = win->static_start;
        else
-               req->Base = win->res->start;
+               res->start = win->res->start;
+
+       /* convert to new-style resources */
+       res->end += res->start - 1;
+       res->flags &= ~WIN_FLAGS_REQ;
+       res->flags |= (win->map << 2) | IORESOURCE_MEM;
+       res->parent = win->res;
+       if (win->res)
+               request_resource(&iomem_resource, res);
+
+       dev_dbg(&p_dev->dev, "request_window results in %pR\n", res);
 
        mutex_unlock(&s->ops_mutex);
-       *wh = w + 1;
 
        return 0;
 } /* pcmcia_request_window */
 EXPORT_SYMBOL(pcmcia_request_window);
 
-void pcmcia_disable_device(struct pcmcia_device *p_dev)
-{
-       pcmcia_release_configuration(p_dev);
-       pcmcia_release_io(p_dev, &p_dev->io);
-       pcmcia_release_irq(p_dev, &p_dev->irq);
-       if (p_dev->win)
-               pcmcia_release_window(p_dev, p_dev->win);
-}
-EXPORT_SYMBOL(pcmcia_disable_device);
-
-
-struct pcmcia_cfg_mem {
-       struct pcmcia_device *p_dev;
-       void *priv_data;
-       int (*conf_check) (struct pcmcia_device *p_dev,
-                          cistpl_cftable_entry_t *cfg,
-                          cistpl_cftable_entry_t *dflt,
-                          unsigned int vcc,
-                          void *priv_data);
-       cisparse_t parse;
-       cistpl_cftable_entry_t dflt;
-};
-
-/**
- * pcmcia_do_loop_config() - internal helper for pcmcia_loop_config()
- *
- * pcmcia_do_loop_config() is the internal callback for the call from
- * pcmcia_loop_config() to pccard_loop_tuple(). Data is transferred
- * by a struct pcmcia_cfg_mem.
- */
-static int pcmcia_do_loop_config(tuple_t *tuple, cisparse_t *parse, void *priv)
-{
-       cistpl_cftable_entry_t *cfg = &parse->cftable_entry;
-       struct pcmcia_cfg_mem *cfg_mem = priv;
-
-       /* default values */
-       cfg_mem->p_dev->conf.ConfigIndex = cfg->index;
-       if (cfg->flags & CISTPL_CFTABLE_DEFAULT)
-               cfg_mem->dflt = *cfg;
-
-       return cfg_mem->conf_check(cfg_mem->p_dev, cfg, &cfg_mem->dflt,
-                                  cfg_mem->p_dev->socket->socket.Vcc,
-                                  cfg_mem->priv_data);
-}
 
 /**
- * pcmcia_loop_config() - loop over configuration options
- * @p_dev:     the struct pcmcia_device which we need to loop for.
- * @conf_check:        function to call for each configuration option.
- *             It gets passed the struct pcmcia_device, the CIS data
- *             describing the configuration option, and private data
- *             being passed to pcmcia_loop_config()
- * @priv_data: private data to be passed to the conf_check function.
+ * pcmcia_disable_device() - disable and clean up a PCMCIA device
+ * @p_dev: the associated PCMCIA device
  *
- * pcmcia_loop_config() loops over all configuration options, and calls
- * the driver-specific conf_check() for each one, checking whether
- * it is a valid one. Returns 0 on success or errorcode otherwise.
+ * pcmcia_disable_device() is the driver-callable counterpart to
+ * pcmcia_enable_device(): If a PCMCIA device is no longer used,
+ * drivers are expected to clean up and disable the device by calling
+ * this function. Any I/O ranges (iomem and ioports) will be released,
+ * the Vpp voltage will be set to 0, and IRQs will no longer be
+ * generated -- at least if there is no other card function (of
+ * multifunction devices) being used.
  */
-int pcmcia_loop_config(struct pcmcia_device *p_dev,
-                      int      (*conf_check)   (struct pcmcia_device *p_dev,
-                                                cistpl_cftable_entry_t *cfg,
-                                                cistpl_cftable_entry_t *dflt,
-                                                unsigned int vcc,
-                                                void *priv_data),
-                      void *priv_data)
-{
-       struct pcmcia_cfg_mem *cfg_mem;
-       int ret;
-
-       cfg_mem = kzalloc(sizeof(struct pcmcia_cfg_mem), GFP_KERNEL);
-       if (cfg_mem == NULL)
-               return -ENOMEM;
-
-       cfg_mem->p_dev = p_dev;
-       cfg_mem->conf_check = conf_check;
-       cfg_mem->priv_data = priv_data;
-
-       ret = pccard_loop_tuple(p_dev->socket, p_dev->func,
-                               CISTPL_CFTABLE_ENTRY, &cfg_mem->parse,
-                               cfg_mem, pcmcia_do_loop_config);
-
-       kfree(cfg_mem);
-       return ret;
-}
-EXPORT_SYMBOL(pcmcia_loop_config);
-
-
-struct pcmcia_loop_mem {
-       struct pcmcia_device *p_dev;
-       void *priv_data;
-       int (*loop_tuple) (struct pcmcia_device *p_dev,
-                          tuple_t *tuple,
-                          void *priv_data);
-};
-
-/**
- * pcmcia_do_loop_tuple() - internal helper for pcmcia_loop_config()
- *
- * pcmcia_do_loop_tuple() is the internal callback for the call from
- * pcmcia_loop_tuple() to pccard_loop_tuple(). Data is transferred
- * by a struct pcmcia_cfg_mem.
- */
-static int pcmcia_do_loop_tuple(tuple_t *tuple, cisparse_t *parse, void *priv)
-{
-       struct pcmcia_loop_mem *loop = priv;
-
-       return loop->loop_tuple(loop->p_dev, tuple, loop->priv_data);
-};
-
-/**
- * pcmcia_loop_tuple() - loop over tuples in the CIS
- * @p_dev:     the struct pcmcia_device which we need to loop for.
- * @code:      which CIS code shall we look for?
- * @priv_data: private data to be passed to the loop_tuple function.
- * @loop_tuple:        function to call for each CIS entry of type @function. IT
- *             gets passed the raw tuple and @priv_data.
- *
- * pcmcia_loop_tuple() loops over all CIS entries of type @function, and
- * calls the @loop_tuple function for each entry. If the call to @loop_tuple
- * returns 0, the loop exits. Returns 0 on success or errorcode otherwise.
- */
-int pcmcia_loop_tuple(struct pcmcia_device *p_dev, cisdata_t code,
-                     int (*loop_tuple) (struct pcmcia_device *p_dev,
-                                        tuple_t *tuple,
-                                        void *priv_data),
-                     void *priv_data)
-{
-       struct pcmcia_loop_mem loop = {
-               .p_dev = p_dev,
-               .loop_tuple = loop_tuple,
-               .priv_data = priv_data};
-
-       return pccard_loop_tuple(p_dev->socket, p_dev->func, code, NULL,
-                                &loop, pcmcia_do_loop_tuple);
-}
-EXPORT_SYMBOL(pcmcia_loop_tuple);
-
-
-struct pcmcia_loop_get {
-       size_t len;
-       cisdata_t **buf;
-};
-
-/**
- * pcmcia_do_get_tuple() - internal helper for pcmcia_get_tuple()
- *
- * pcmcia_do_get_tuple() is the internal callback for the call from
- * pcmcia_get_tuple() to pcmcia_loop_tuple(). As we're only interested in
- * the first tuple, return 0 unconditionally. Create a memory buffer large
- * enough to hold the content of the tuple, and fill it with the tuple data.
- * The caller is responsible to free the buffer.
- */
-static int pcmcia_do_get_tuple(struct pcmcia_device *p_dev, tuple_t *tuple,
-                              void *priv)
-{
-       struct pcmcia_loop_get *get = priv;
-
-       *get->buf = kzalloc(tuple->TupleDataLen, GFP_KERNEL);
-       if (*get->buf) {
-               get->len = tuple->TupleDataLen;
-               memcpy(*get->buf, tuple->TupleData, tuple->TupleDataLen);
-       } else
-               dev_dbg(&p_dev->dev, "do_get_tuple: out of memory\n");
-       return 0;
-}
-
-/**
- * pcmcia_get_tuple() - get first tuple from CIS
- * @p_dev:     the struct pcmcia_device which we need to loop for.
- * @code:      which CIS code shall we look for?
- * @buf:        pointer to store the buffer to.
- *
- * pcmcia_get_tuple() gets the content of the first CIS entry of type @code.
- * It returns the buffer length (or zero). The caller is responsible to free
- * the buffer passed in @buf.
- */
-size_t pcmcia_get_tuple(struct pcmcia_device *p_dev, cisdata_t code,
-                       unsigned char **buf)
-{
-       struct pcmcia_loop_get get = {
-               .len = 0,
-               .buf = buf,
-       };
-
-       *get.buf = NULL;
-       pcmcia_loop_tuple(p_dev, code, pcmcia_do_get_tuple, &get);
-
-       return get.len;
-}
-EXPORT_SYMBOL(pcmcia_get_tuple);
-
-
-/**
- * pcmcia_do_get_mac() - internal helper for pcmcia_get_mac_from_cis()
- *
- * pcmcia_do_get_mac() is the internal callback for the call from
- * pcmcia_get_mac_from_cis() to pcmcia_loop_tuple(). We check whether the
- * tuple contains a proper LAN_NODE_ID of length 6, and copy the data
- * to struct net_device->dev_addr[i].
- */
-static int pcmcia_do_get_mac(struct pcmcia_device *p_dev, tuple_t *tuple,
-                            void *priv)
+void pcmcia_disable_device(struct pcmcia_device *p_dev)
 {
-       struct net_device *dev = priv;
        int i;
 
-       if (tuple->TupleData[0] != CISTPL_FUNCE_LAN_NODE_ID)
-               return -EINVAL;
-       if (tuple->TupleDataLen < ETH_ALEN + 2) {
-               dev_warn(&p_dev->dev, "Invalid CIS tuple length for "
-                       "LAN_NODE_ID\n");
-               return -EINVAL;
-       }
+       dev_dbg(&p_dev->dev, "disabling device\n");
 
-       if (tuple->TupleData[1] != ETH_ALEN) {
-               dev_warn(&p_dev->dev, "Invalid header for LAN_NODE_ID\n");
-               return -EINVAL;
+       for (i = 0; i < MAX_WIN; i++) {
+               struct resource *res = p_dev->resource[MAX_IO_WIN + i];
+               if (res->flags & WIN_FLAGS_REQ)
+                       pcmcia_release_window(p_dev, res);
        }
-       for (i = 0; i < 6; i++)
-               dev->dev_addr[i] = tuple->TupleData[i+2];
-       return 0;
-}
 
-/**
- * pcmcia_get_mac_from_cis() - read out MAC address from CISTPL_FUNCE
- * @p_dev:     the struct pcmcia_device for which we want the address.
- * @dev:       a properly prepared struct net_device to store the info to.
- *
- * pcmcia_get_mac_from_cis() reads out the hardware MAC address from
- * CISTPL_FUNCE and stores it into struct net_device *dev->dev_addr which
- * must be set up properly by the driver (see examples!).
- */
-int pcmcia_get_mac_from_cis(struct pcmcia_device *p_dev, struct net_device *dev)
-{
-       return pcmcia_loop_tuple(p_dev, CISTPL_FUNCE, pcmcia_do_get_mac, dev);
+       pcmcia_release_configuration(p_dev);
+       pcmcia_release_io(p_dev);
+       if (p_dev->_irq) {
+               free_irq(p_dev->irq, p_dev->priv);
+               p_dev->_irq = 0;
+       }
 }
-EXPORT_SYMBOL(pcmcia_get_mac_from_cis);
-
+EXPORT_SYMBOL(pcmcia_disable_device);