]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
[POWERPC] Move common PCI code out of pci_32/pci_64
authorKumar Gala <galak@kernel.crashing.org>
Wed, 27 Jun 2007 16:07:51 +0000 (11:07 -0500)
committerKumar Gala <galak@kernel.crashing.org>
Fri, 29 Jun 2007 06:58:43 +0000 (01:58 -0500)
Moved the low hanging fruit that was either identical or close
to it between ppc32 & ppc64 for PCI into pci-common.c

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
arch/powerpc/kernel/pci-common.c
arch/powerpc/kernel/pci_32.c
arch/powerpc/kernel/pci_64.c

index 3ca8cfb99dc2cd85cc6c60620c357603af5cc5f8..b518b880d2eb6c69df4b2aefefdcd7176039ce0b 100644 (file)
@@ -51,3 +51,317 @@ int pci_domain_nr(struct pci_bus *bus)
 }
 
 EXPORT_SYMBOL(pci_domain_nr);
+
+#ifdef CONFIG_PPC_OF
+static ssize_t pci_show_devspec(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct pci_dev *pdev;
+       struct device_node *np;
+
+       pdev = to_pci_dev (dev);
+       np = pci_device_to_OF_node(pdev);
+       if (np == NULL || np->full_name == NULL)
+               return 0;
+       return sprintf(buf, "%s", np->full_name);
+}
+static DEVICE_ATTR(devspec, S_IRUGO, pci_show_devspec, NULL);
+#endif /* CONFIG_PPC_OF */
+
+/* Add sysfs properties */
+void pcibios_add_platform_entries(struct pci_dev *pdev)
+{
+#ifdef CONFIG_PPC_OF
+       device_create_file(&pdev->dev, &dev_attr_devspec);
+#endif /* CONFIG_PPC_OF */
+}
+
+char __init *pcibios_setup(char *str)
+{
+       return str;
+}
+
+/*
+ * Reads the interrupt pin to determine if interrupt is use by card.
+ * If the interrupt is used, then gets the interrupt line from the
+ * openfirmware and sets it in the pci_dev and pci_config line.
+ */
+int pci_read_irq_line(struct pci_dev *pci_dev)
+{
+       struct of_irq oirq;
+       unsigned int virq;
+
+       DBG("Try to map irq for %s...\n", pci_name(pci_dev));
+
+#ifdef DEBUG
+       memset(&oirq, 0xff, sizeof(oirq));
+#endif
+       /* Try to get a mapping from the device-tree */
+       if (of_irq_map_pci(pci_dev, &oirq)) {
+               u8 line, pin;
+
+               /* If that fails, lets fallback to what is in the config
+                * space and map that through the default controller. We
+                * also set the type to level low since that's what PCI
+                * interrupts are. If your platform does differently, then
+                * either provide a proper interrupt tree or don't use this
+                * function.
+                */
+               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_PIN, &pin))
+                       return -1;
+               if (pin == 0)
+                       return -1;
+               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_LINE, &line) ||
+                   line == 0xff) {
+                       return -1;
+               }
+               DBG(" -> no map ! Using irq line %d from PCI config\n", line);
+
+               virq = irq_create_mapping(NULL, line);
+               if (virq != NO_IRQ)
+                       set_irq_type(virq, IRQ_TYPE_LEVEL_LOW);
+       } else {
+               DBG(" -> got one, spec %d cells (0x%08x 0x%08x...) on %s\n",
+                   oirq.size, oirq.specifier[0], oirq.specifier[1],
+                   oirq.controller->full_name);
+
+               virq = irq_create_of_mapping(oirq.controller, oirq.specifier,
+                                            oirq.size);
+       }
+       if(virq == NO_IRQ) {
+               DBG(" -> failed to map !\n");
+               return -1;
+       }
+
+       DBG(" -> mapped to linux irq %d\n", virq);
+
+       pci_dev->irq = virq;
+
+       return 0;
+}
+EXPORT_SYMBOL(pci_read_irq_line);
+
+/*
+ * Platform support for /proc/bus/pci/X/Y mmap()s,
+ * modelled on the sparc64 implementation by Dave Miller.
+ *  -- paulus.
+ */
+
+/*
+ * Adjust vm_pgoff of VMA such that it is the physical page offset
+ * corresponding to the 32-bit pci bus offset for DEV requested by the user.
+ *
+ * Basically, the user finds the base address for his device which he wishes
+ * to mmap.  They read the 32-bit value from the config space base register,
+ * add whatever PAGE_SIZE multiple offset they wish, and feed this into the
+ * offset parameter of mmap on /proc/bus/pci/XXX for that device.
+ *
+ * Returns negative error code on failure, zero on success.
+ */
+static struct resource *__pci_mmap_make_offset(struct pci_dev *dev,
+                                              resource_size_t *offset,
+                                              enum pci_mmap_state mmap_state)
+{
+       struct pci_controller *hose = pci_bus_to_host(dev->bus);
+       unsigned long io_offset = 0;
+       int i, res_bit;
+
+       if (hose == 0)
+               return NULL;            /* should never happen */
+
+       /* If memory, add on the PCI bridge address offset */
+       if (mmap_state == pci_mmap_mem) {
+#if 0 /* See comment in pci_resource_to_user() for why this is disabled */
+               *offset += hose->pci_mem_offset;
+#endif
+               res_bit = IORESOURCE_MEM;
+       } else {
+               io_offset = (unsigned long)hose->io_base_virt - _IO_BASE;
+               *offset += io_offset;
+               res_bit = IORESOURCE_IO;
+       }
+
+       /*
+        * Check that the offset requested corresponds to one of the
+        * resources of the device.
+        */
+       for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
+               struct resource *rp = &dev->resource[i];
+               int flags = rp->flags;
+
+               /* treat ROM as memory (should be already) */
+               if (i == PCI_ROM_RESOURCE)
+                       flags |= IORESOURCE_MEM;
+
+               /* Active and same type? */
+               if ((flags & res_bit) == 0)
+                       continue;
+
+               /* In the range of this resource? */
+               if (*offset < (rp->start & PAGE_MASK) || *offset > rp->end)
+                       continue;
+
+               /* found it! construct the final physical address */
+               if (mmap_state == pci_mmap_io)
+                       *offset += hose->io_base_phys - io_offset;
+               return rp;
+       }
+
+       return NULL;
+}
+
+/*
+ * Set vm_page_prot of VMA, as appropriate for this architecture, for a pci
+ * device mapping.
+ */
+static pgprot_t __pci_mmap_set_pgprot(struct pci_dev *dev, struct resource *rp,
+                                     pgprot_t protection,
+                                     enum pci_mmap_state mmap_state,
+                                     int write_combine)
+{
+       unsigned long prot = pgprot_val(protection);
+
+       /* Write combine is always 0 on non-memory space mappings. On
+        * memory space, if the user didn't pass 1, we check for a
+        * "prefetchable" resource. This is a bit hackish, but we use
+        * this to workaround the inability of /sysfs to provide a write
+        * combine bit
+        */
+       if (mmap_state != pci_mmap_mem)
+               write_combine = 0;
+       else if (write_combine == 0) {
+               if (rp->flags & IORESOURCE_PREFETCH)
+                       write_combine = 1;
+       }
+
+       /* XXX would be nice to have a way to ask for write-through */
+       prot |= _PAGE_NO_CACHE;
+       if (write_combine)
+               prot &= ~_PAGE_GUARDED;
+       else
+               prot |= _PAGE_GUARDED;
+
+       return __pgprot(prot);
+}
+
+/*
+ * This one is used by /dev/mem and fbdev who have no clue about the
+ * PCI device, it tries to find the PCI device first and calls the
+ * above routine
+ */
+pgprot_t pci_phys_mem_access_prot(struct file *file,
+                                 unsigned long pfn,
+                                 unsigned long size,
+                                 pgprot_t protection)
+{
+       struct pci_dev *pdev = NULL;
+       struct resource *found = NULL;
+       unsigned long prot = pgprot_val(protection);
+       unsigned long offset = pfn << PAGE_SHIFT;
+       int i;
+
+       if (page_is_ram(pfn))
+               return __pgprot(prot);
+
+       prot |= _PAGE_NO_CACHE | _PAGE_GUARDED;
+
+       for_each_pci_dev(pdev) {
+               for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
+                       struct resource *rp = &pdev->resource[i];
+                       int flags = rp->flags;
+
+                       /* Active and same type? */
+                       if ((flags & IORESOURCE_MEM) == 0)
+                               continue;
+                       /* In the range of this resource? */
+                       if (offset < (rp->start & PAGE_MASK) ||
+                           offset > rp->end)
+                               continue;
+                       found = rp;
+                       break;
+               }
+               if (found)
+                       break;
+       }
+       if (found) {
+               if (found->flags & IORESOURCE_PREFETCH)
+                       prot &= ~_PAGE_GUARDED;
+               pci_dev_put(pdev);
+       }
+
+       DBG("non-PCI map for %lx, prot: %lx\n", offset, prot);
+
+       return __pgprot(prot);
+}
+
+
+/*
+ * Perform the actual remap of the pages for a PCI device mapping, as
+ * appropriate for this architecture.  The region in the process to map
+ * is described by vm_start and vm_end members of VMA, the base physical
+ * address is found in vm_pgoff.
+ * The pci device structure is provided so that architectures may make mapping
+ * decisions on a per-device or per-bus basis.
+ *
+ * Returns a negative error code on failure, zero on success.
+ */
+int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
+                       enum pci_mmap_state mmap_state, int write_combine)
+{
+       resource_size_t offset = vma->vm_pgoff << PAGE_SHIFT;
+       struct resource *rp;
+       int ret;
+
+       rp = __pci_mmap_make_offset(dev, &offset, mmap_state);
+       if (rp == NULL)
+               return -EINVAL;
+
+       vma->vm_pgoff = offset >> PAGE_SHIFT;
+       vma->vm_page_prot = __pci_mmap_set_pgprot(dev, rp,
+                                                 vma->vm_page_prot,
+                                                 mmap_state, write_combine);
+
+       ret = remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
+                              vma->vm_end - vma->vm_start, vma->vm_page_prot);
+
+       return ret;
+}
+
+void pci_resource_to_user(const struct pci_dev *dev, int bar,
+                         const struct resource *rsrc,
+                         resource_size_t *start, resource_size_t *end)
+{
+       struct pci_controller *hose = pci_bus_to_host(dev->bus);
+       resource_size_t offset = 0;
+
+       if (hose == NULL)
+               return;
+
+       if (rsrc->flags & IORESOURCE_IO)
+               offset = (unsigned long)hose->io_base_virt - _IO_BASE;
+
+       /* We pass a fully fixed up address to userland for MMIO instead of
+        * a BAR value because X is lame and expects to be able to use that
+        * to pass to /dev/mem !
+        *
+        * That means that we'll have potentially 64 bits values where some
+        * userland apps only expect 32 (like X itself since it thinks only
+        * Sparc has 64 bits MMIO) but if we don't do that, we break it on
+        * 32 bits CHRPs :-(
+        *
+        * Hopefully, the sysfs insterface is immune to that gunk. Once X
+        * has been fixed (and the fix spread enough), we can re-enable the
+        * 2 lines below and pass down a BAR value to userland. In that case
+        * we'll also have to re-enable the matching code in
+        * __pci_mmap_make_offset().
+        *
+        * BenH.
+        */
+#if 0
+       else if (rsrc->flags & IORESOURCE_MEM)
+               offset = hose->pci_mem_offset;
+#endif
+
+       *start = rsrc->start - offset;
+       *end = rsrc->end - offset;
+}
index c81ffa2829772a3f7ca27fc82cc101577fce9848..3dd931ecce91ef8d24ef19245226d1b05f21e9a3 100644 (file)
@@ -1028,34 +1028,12 @@ pci_create_OF_bus_map(void)
        }
 }
 
-static ssize_t pci_show_devspec(struct device *dev, struct device_attribute *attr, char *buf)
-{
-       struct pci_dev *pdev;
-       struct device_node *np;
-
-       pdev = to_pci_dev (dev);
-       np = pci_device_to_OF_node(pdev);
-       if (np == NULL || np->full_name == NULL)
-               return 0;
-       return sprintf(buf, "%s", np->full_name);
-}
-static DEVICE_ATTR(devspec, S_IRUGO, pci_show_devspec, NULL);
-
 #else /* CONFIG_PPC_OF */
 void pcibios_make_OF_bus_map(void)
 {
 }
 #endif /* CONFIG_PPC_OF */
 
-/* Add sysfs properties */
-void pcibios_add_platform_entries(struct pci_dev *pdev)
-{
-#ifdef CONFIG_PPC_OF
-       device_create_file(&pdev->dev, &dev_attr_devspec);
-#endif /* CONFIG_PPC_OF */
-}
-
-
 #ifdef CONFIG_PPC_PMAC
 /*
  * This set of routines checks for PCI<->PCI bridges that have closed
@@ -1390,11 +1368,6 @@ void __init pcibios_fixup_bus(struct pci_bus *bus)
        }
 }
 
-char __init *pcibios_setup(char *str)
-{
-       return str;
-}
-
 /* the next one is stolen from the alpha port... */
 void __init
 pcibios_update_irq(struct pci_dev *dev, int irq)
@@ -1403,64 +1376,6 @@ pcibios_update_irq(struct pci_dev *dev, int irq)
        /* XXX FIXME - update OF device tree node interrupt property */
 }
 
-#ifdef CONFIG_PPC_MERGE
-/* XXX This is a copy of the ppc64 version. This is temporary until we start
- * merging the 2 PCI layers
- */
-/*
- * Reads the interrupt pin to determine if interrupt is use by card.
- * If the interrupt is used, then gets the interrupt line from the
- * openfirmware and sets it in the pci_dev and pci_config line.
- */
-int pci_read_irq_line(struct pci_dev *pci_dev)
-{
-       struct of_irq oirq;
-       unsigned int virq;
-
-       DBG("Try to map irq for %s...\n", pci_name(pci_dev));
-
-       /* Try to get a mapping from the device-tree */
-       if (of_irq_map_pci(pci_dev, &oirq)) {
-               u8 line, pin;
-
-               /* If that fails, lets fallback to what is in the config
-                * space and map that through the default controller. We
-                * also set the type to level low since that's what PCI
-                * interrupts are. If your platform does differently, then
-                * either provide a proper interrupt tree or don't use this
-                * function.
-                */
-               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_PIN, &pin))
-                       return -1;
-               if (pin == 0)
-                       return -1;
-               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_LINE, &line) ||
-                   line == 0xff) {
-                       return -1;
-               }
-               DBG(" -> no map ! Using irq line %d from PCI config\n", line);
-
-               virq = irq_create_mapping(NULL, line);
-               if (virq != NO_IRQ)
-                       set_irq_type(virq, IRQ_TYPE_LEVEL_LOW);
-       } else {
-               DBG(" -> got one, spec %d cells (0x%08x...) on %s\n",
-                   oirq.size, oirq.specifier[0], oirq.controller->full_name);
-
-               virq = irq_create_of_mapping(oirq.controller, oirq.specifier,
-                                            oirq.size);
-       }
-       if(virq == NO_IRQ) {
-               DBG(" -> failed to map !\n");
-               return -1;
-       }
-       pci_dev->irq = virq;
-
-       return 0;
-}
-EXPORT_SYMBOL(pci_read_irq_line);
-#endif /* CONFIG_PPC_MERGE */
-
 int pcibios_enable_device(struct pci_dev *dev, int mask)
 {
        u16 cmd, old_cmd;
@@ -1503,176 +1418,6 @@ pci_bus_to_hose(int bus)
        return NULL;
 }
 
-static struct resource *__pci_mmap_make_offset(struct pci_dev *dev,
-                                              resource_size_t *offset,
-                                              enum pci_mmap_state mmap_state)
-{
-       struct pci_controller *hose = pci_bus_to_host(dev->bus);
-       unsigned long io_offset = 0;
-       int i, res_bit;
-
-       if (hose == 0)
-               return NULL;            /* should never happen */
-
-       /* If memory, add on the PCI bridge address offset */
-       if (mmap_state == pci_mmap_mem) {
-#if 0 /* See comment in pci_resource_to_user() for why this is disabled */
-               *offset += hose->pci_mem_offset;
-#endif
-               res_bit = IORESOURCE_MEM;
-       } else {
-               io_offset = hose->io_base_virt - (void __iomem *)_IO_BASE;
-               *offset += io_offset;
-               res_bit = IORESOURCE_IO;
-       }
-
-       /*
-        * Check that the offset requested corresponds to one of the
-        * resources of the device.
-        */
-       for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
-               struct resource *rp = &dev->resource[i];
-               int flags = rp->flags;
-
-               /* treat ROM as memory (should be already) */
-               if (i == PCI_ROM_RESOURCE)
-                       flags |= IORESOURCE_MEM;
-
-               /* Active and same type? */
-               if ((flags & res_bit) == 0)
-                       continue;
-
-               /* In the range of this resource? */
-               if (*offset < (rp->start & PAGE_MASK) || *offset > rp->end)
-                       continue;
-
-               /* found it! construct the final physical address */
-               if (mmap_state == pci_mmap_io)
-                       *offset += hose->io_base_phys - io_offset;
-               return rp;
-       }
-
-       return NULL;
-}
-
-/*
- * Set vm_page_prot of VMA, as appropriate for this architecture, for a pci
- * device mapping.
- */
-static pgprot_t __pci_mmap_set_pgprot(struct pci_dev *dev, struct resource *rp,
-                                     pgprot_t protection,
-                                     enum pci_mmap_state mmap_state,
-                                     int write_combine)
-{
-       unsigned long prot = pgprot_val(protection);
-
-       /* Write combine is always 0 on non-memory space mappings. On
-        * memory space, if the user didn't pass 1, we check for a
-        * "prefetchable" resource. This is a bit hackish, but we use
-        * this to workaround the inability of /sysfs to provide a write
-        * combine bit
-        */
-       if (mmap_state != pci_mmap_mem)
-               write_combine = 0;
-       else if (write_combine == 0) {
-               if (rp->flags & IORESOURCE_PREFETCH)
-                       write_combine = 1;
-       }
-
-       /* XXX would be nice to have a way to ask for write-through */
-       prot |= _PAGE_NO_CACHE;
-       if (write_combine)
-               prot &= ~_PAGE_GUARDED;
-       else
-               prot |= _PAGE_GUARDED;
-
-       return __pgprot(prot);
-}
-
-/*
- * This one is used by /dev/mem and fbdev who have no clue about the
- * PCI device, it tries to find the PCI device first and calls the
- * above routine
- */
-pgprot_t pci_phys_mem_access_prot(struct file *file,
-                                 unsigned long pfn,
-                                 unsigned long size,
-                                 pgprot_t protection)
-{
-       struct pci_dev *pdev = NULL;
-       struct resource *found = NULL;
-       unsigned long prot = pgprot_val(protection);
-       unsigned long offset = pfn << PAGE_SHIFT;
-       int i;
-
-       if (page_is_ram(pfn))
-               return __pgprot(prot);
-
-       prot |= _PAGE_NO_CACHE | _PAGE_GUARDED;
-
-       for_each_pci_dev(pdev) {
-               for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
-                       struct resource *rp = &pdev->resource[i];
-                       int flags = rp->flags;
-
-                       /* Active and same type? */
-                       if ((flags & IORESOURCE_MEM) == 0)
-                               continue;
-                       /* In the range of this resource? */
-                       if (offset < (rp->start & PAGE_MASK) ||
-                           offset > rp->end)
-                               continue;
-                       found = rp;
-                       break;
-               }
-               if (found)
-                       break;
-       }
-       if (found) {
-               if (found->flags & IORESOURCE_PREFETCH)
-                       prot &= ~_PAGE_GUARDED;
-               pci_dev_put(pdev);
-       }
-
-       DBG("non-PCI map for %lx, prot: %lx\n", offset, prot);
-
-       return __pgprot(prot);
-}
-
-
-/*
- * Perform the actual remap of the pages for a PCI device mapping, as
- * appropriate for this architecture.  The region in the process to map
- * is described by vm_start and vm_end members of VMA, the base physical
- * address is found in vm_pgoff.
- * The pci device structure is provided so that architectures may make mapping
- * decisions on a per-device or per-bus basis.
- *
- * Returns a negative error code on failure, zero on success.
- */
-int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
-                       enum pci_mmap_state mmap_state,
-                       int write_combine)
-{
-       resource_size_t offset = vma->vm_pgoff << PAGE_SHIFT;
-       struct resource *rp;
-       int ret;
-
-       rp = __pci_mmap_make_offset(dev, &offset, mmap_state);
-       if (rp == NULL)
-               return -EINVAL;
-
-       vma->vm_pgoff = offset >> PAGE_SHIFT;
-       vma->vm_page_prot = __pci_mmap_set_pgprot(dev, rp,
-                                                 vma->vm_page_prot,
-                                                 mmap_state, write_combine);
-
-       ret = remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
-                              vma->vm_end - vma->vm_start, vma->vm_page_prot);
-
-       return ret;
-}
-
 /* Provide information on locations of various I/O regions in physical
  * memory.  Do this on a per-card basis so that we choose the right
  * root bridge.
@@ -1715,45 +1460,6 @@ long sys_pciconfig_iobase(long which, unsigned long bus, unsigned long devfn)
        return result;
 }
 
-void pci_resource_to_user(const struct pci_dev *dev, int bar,
-                         const struct resource *rsrc,
-                         resource_size_t *start, resource_size_t *end)
-{
-       struct pci_controller *hose = pci_bus_to_host(dev->bus);
-       resource_size_t offset = 0;
-
-       if (hose == NULL)
-               return;
-
-       if (rsrc->flags & IORESOURCE_IO)
-               offset = (unsigned long)hose->io_base_virt - _IO_BASE;
-
-       /* We pass a fully fixed up address to userland for MMIO instead of
-        * a BAR value because X is lame and expects to be able to use that
-        * to pass to /dev/mem !
-        *
-        * That means that we'll have potentially 64 bits values where some
-        * userland apps only expect 32 (like X itself since it thinks only
-        * Sparc has 64 bits MMIO) but if we don't do that, we break it on
-        * 32 bits CHRPs :-(
-        *
-        * Hopefully, the sysfs insterface is immune to that gunk. Once X
-        * has been fixed (and the fix spread enough), we can re-enable the
-        * 2 lines below and pass down a BAR value to userland. In that case
-        * we'll also have to re-enable the matching code in
-        * __pci_mmap_make_offset().
-        *
-        * BenH.
-        */
-#if 0
-       else if (rsrc->flags & IORESOURCE_MEM)
-               offset = hose->pci_mem_offset;
-#endif
-
-       *start = rsrc->start - offset;
-       *end = rsrc->end - offset;
-}
-
 unsigned long pci_address_to_pio(phys_addr_t address)
 {
        struct pci_controller* hose = hose_head;
index 57bdcd88f04a28c89a86253f1818a080f8978eab..384d2752fe60ad5952f34cadd635ae5e8e207610 100644 (file)
@@ -601,11 +601,6 @@ static int __init pcibios_init(void)
 
 subsys_initcall(pcibios_init);
 
-char __init *pcibios_setup(char *str)
-{
-       return str;
-}
-
 int pcibios_enable_device(struct pci_dev *dev, int mask)
 {
        u16 cmd, oldcmd;
@@ -647,211 +642,6 @@ int pci_proc_domain(struct pci_bus *bus)
        }
 }
 
-/*
- * Platform support for /proc/bus/pci/X/Y mmap()s,
- * modelled on the sparc64 implementation by Dave Miller.
- *  -- paulus.
- */
-
-/*
- * Adjust vm_pgoff of VMA such that it is the physical page offset
- * corresponding to the 32-bit pci bus offset for DEV requested by the user.
- *
- * Basically, the user finds the base address for his device which he wishes
- * to mmap.  They read the 32-bit value from the config space base register,
- * add whatever PAGE_SIZE multiple offset they wish, and feed this into the
- * offset parameter of mmap on /proc/bus/pci/XXX for that device.
- *
- * Returns negative error code on failure, zero on success.
- */
-static struct resource *__pci_mmap_make_offset(struct pci_dev *dev,
-                                              resource_size_t *offset,
-                                              enum pci_mmap_state mmap_state)
-{
-       struct pci_controller *hose = pci_bus_to_host(dev->bus);
-       unsigned long io_offset = 0;
-       int i, res_bit;
-
-       if (hose == 0)
-               return NULL;            /* should never happen */
-
-       /* If memory, add on the PCI bridge address offset */
-       if (mmap_state == pci_mmap_mem) {
-#if 0 /* See comment in pci_resource_to_user() for why this is disabled */
-               *offset += hose->pci_mem_offset;
-#endif
-               res_bit = IORESOURCE_MEM;
-       } else {
-               io_offset = (unsigned long)hose->io_base_virt - _IO_BASE;
-               *offset += io_offset;
-               res_bit = IORESOURCE_IO;
-       }
-
-       /*
-        * Check that the offset requested corresponds to one of the
-        * resources of the device.
-        */
-       for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
-               struct resource *rp = &dev->resource[i];
-               int flags = rp->flags;
-
-               /* treat ROM as memory (should be already) */
-               if (i == PCI_ROM_RESOURCE)
-                       flags |= IORESOURCE_MEM;
-
-               /* Active and same type? */
-               if ((flags & res_bit) == 0)
-                       continue;
-
-               /* In the range of this resource? */
-               if (*offset < (rp->start & PAGE_MASK) || *offset > rp->end)
-                       continue;
-
-               /* found it! construct the final physical address */
-               if (mmap_state == pci_mmap_io)
-                       *offset += hose->io_base_phys - io_offset;
-               return rp;
-       }
-
-       return NULL;
-}
-
-/*
- * Set vm_page_prot of VMA, as appropriate for this architecture, for a pci
- * device mapping.
- */
-static pgprot_t __pci_mmap_set_pgprot(struct pci_dev *dev, struct resource *rp,
-                                     pgprot_t protection,
-                                     enum pci_mmap_state mmap_state,
-                                     int write_combine)
-{
-       unsigned long prot = pgprot_val(protection);
-
-       /* Write combine is always 0 on non-memory space mappings. On
-        * memory space, if the user didn't pass 1, we check for a
-        * "prefetchable" resource. This is a bit hackish, but we use
-        * this to workaround the inability of /sysfs to provide a write
-        * combine bit
-        */
-       if (mmap_state != pci_mmap_mem)
-               write_combine = 0;
-       else if (write_combine == 0) {
-               if (rp->flags & IORESOURCE_PREFETCH)
-                       write_combine = 1;
-       }
-
-       /* XXX would be nice to have a way to ask for write-through */
-       prot |= _PAGE_NO_CACHE;
-       if (write_combine)
-               prot &= ~_PAGE_GUARDED;
-       else
-               prot |= _PAGE_GUARDED;
-
-       return __pgprot(prot);
-}
-
-/*
- * This one is used by /dev/mem and fbdev who have no clue about the
- * PCI device, it tries to find the PCI device first and calls the
- * above routine
- */
-pgprot_t pci_phys_mem_access_prot(struct file *file,
-                                 unsigned long pfn,
-                                 unsigned long size,
-                                 pgprot_t protection)
-{
-       struct pci_dev *pdev = NULL;
-       struct resource *found = NULL;
-       unsigned long prot = pgprot_val(protection);
-       unsigned long offset = pfn << PAGE_SHIFT;
-       int i;
-
-       if (page_is_ram(pfn))
-               return __pgprot(prot);
-
-       prot |= _PAGE_NO_CACHE | _PAGE_GUARDED;
-
-       for_each_pci_dev(pdev) {
-               for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
-                       struct resource *rp = &pdev->resource[i];
-                       int flags = rp->flags;
-
-                       /* Active and same type? */
-                       if ((flags & IORESOURCE_MEM) == 0)
-                               continue;
-                       /* In the range of this resource? */
-                       if (offset < (rp->start & PAGE_MASK) ||
-                           offset > rp->end)
-                               continue;
-                       found = rp;
-                       break;
-               }
-               if (found)
-                       break;
-       }
-       if (found) {
-               if (found->flags & IORESOURCE_PREFETCH)
-                       prot &= ~_PAGE_GUARDED;
-               pci_dev_put(pdev);
-       }
-
-       DBG("non-PCI map for %lx, prot: %lx\n", offset, prot);
-
-       return __pgprot(prot);
-}
-
-
-/*
- * Perform the actual remap of the pages for a PCI device mapping, as
- * appropriate for this architecture.  The region in the process to map
- * is described by vm_start and vm_end members of VMA, the base physical
- * address is found in vm_pgoff.
- * The pci device structure is provided so that architectures may make mapping
- * decisions on a per-device or per-bus basis.
- *
- * Returns a negative error code on failure, zero on success.
- */
-int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
-                       enum pci_mmap_state mmap_state, int write_combine)
-{
-       resource_size_t offset = vma->vm_pgoff << PAGE_SHIFT;
-       struct resource *rp;
-       int ret;
-
-       rp = __pci_mmap_make_offset(dev, &offset, mmap_state);
-       if (rp == NULL)
-               return -EINVAL;
-
-       vma->vm_pgoff = offset >> PAGE_SHIFT;
-       vma->vm_page_prot = __pci_mmap_set_pgprot(dev, rp,
-                                                 vma->vm_page_prot,
-                                                 mmap_state, write_combine);
-
-       ret = remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
-                              vma->vm_end - vma->vm_start, vma->vm_page_prot);
-
-       return ret;
-}
-
-static ssize_t pci_show_devspec(struct device *dev,
-               struct device_attribute *attr, char *buf)
-{
-       struct pci_dev *pdev;
-       struct device_node *np;
-
-       pdev = to_pci_dev (dev);
-       np = pci_device_to_OF_node(pdev);
-       if (np == NULL || np->full_name == NULL)
-               return 0;
-       return sprintf(buf, "%s", np->full_name);
-}
-static DEVICE_ATTR(devspec, S_IRUGO, pci_show_devspec, NULL);
-
-void pcibios_add_platform_entries(struct pci_dev *pdev)
-{
-       device_create_file(&pdev->dev, &dev_attr_devspec);
-}
-
 void __devinit pci_process_bridge_OF_ranges(struct pci_controller *hose,
                                            struct device_node *dev, int prim)
 {
@@ -1167,105 +957,6 @@ void __devinit pcibios_fixup_bus(struct pci_bus *bus)
 }
 EXPORT_SYMBOL(pcibios_fixup_bus);
 
-/*
- * Reads the interrupt pin to determine if interrupt is use by card.
- * If the interrupt is used, then gets the interrupt line from the 
- * openfirmware and sets it in the pci_dev and pci_config line.
- */
-int pci_read_irq_line(struct pci_dev *pci_dev)
-{
-       struct of_irq oirq;
-       unsigned int virq;
-
-       DBG("Try to map irq for %s...\n", pci_name(pci_dev));
-
-#ifdef DEBUG
-       memset(&oirq, 0xff, sizeof(oirq));
-#endif
-       /* Try to get a mapping from the device-tree */
-       if (of_irq_map_pci(pci_dev, &oirq)) {
-               u8 line, pin;
-
-               /* If that fails, lets fallback to what is in the config
-                * space and map that through the default controller. We
-                * also set the type to level low since that's what PCI
-                * interrupts are. If your platform does differently, then
-                * either provide a proper interrupt tree or don't use this
-                * function.
-                */
-               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_PIN, &pin))
-                       return -1;
-               if (pin == 0)
-                       return -1;
-               if (pci_read_config_byte(pci_dev, PCI_INTERRUPT_LINE, &line) ||
-                   line == 0xff) {
-                       return -1;
-               }
-               DBG(" -> no map ! Using irq line %d from PCI config\n", line);
-
-               virq = irq_create_mapping(NULL, line);
-               if (virq != NO_IRQ)
-                       set_irq_type(virq, IRQ_TYPE_LEVEL_LOW);
-       } else {
-               DBG(" -> got one, spec %d cells (0x%08x 0x%08x...) on %s\n",
-                   oirq.size, oirq.specifier[0], oirq.specifier[1],
-                   oirq.controller->full_name);
-
-               virq = irq_create_of_mapping(oirq.controller, oirq.specifier,
-                                            oirq.size);
-       }
-       if(virq == NO_IRQ) {
-               DBG(" -> failed to map !\n");
-               return -1;
-       }
-
-       DBG(" -> mapped to linux irq %d\n", virq);
-
-       pci_dev->irq = virq;
-
-       return 0;
-}
-EXPORT_SYMBOL(pci_read_irq_line);
-
-void pci_resource_to_user(const struct pci_dev *dev, int bar,
-                         const struct resource *rsrc,
-                         resource_size_t *start, resource_size_t *end)
-{
-       struct pci_controller *hose = pci_bus_to_host(dev->bus);
-       resource_size_t offset = 0;
-
-       if (hose == NULL)
-               return;
-
-       if (rsrc->flags & IORESOURCE_IO)
-               offset = (unsigned long)hose->io_base_virt - _IO_BASE;
-
-       /* We pass a fully fixed up address to userland for MMIO instead of
-        * a BAR value because X is lame and expects to be able to use that
-        * to pass to /dev/mem !
-        *
-        * That means that we'll have potentially 64 bits values where some
-        * userland apps only expect 32 (like X itself since it thinks only
-        * Sparc has 64 bits MMIO) but if we don't do that, we break it on
-        * 32 bits CHRPs :-(
-        *
-        * Hopefully, the sysfs insterface is immune to that gunk. Once X
-        * has been fixed (and the fix spread enough), we can re-enable the
-        * 2 lines below and pass down a BAR value to userland. In that case
-        * we'll also have to re-enable the matching code in
-        * __pci_mmap_make_offset().
-        *
-        * BenH.
-        */
-#if 0
-       else if (rsrc->flags & IORESOURCE_MEM)
-               offset = hose->pci_mem_offset;
-#endif
-
-       *start = rsrc->start - offset;
-       *end = rsrc->end - offset;
-}
-
 struct pci_controller* pci_find_hose_for_OF_device(struct device_node* node)
 {
        if (!have_of)