Merge branch 'io-mappings-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'io-mappings-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
io mapping: clean up #ifdefs
io mapping: improve documentation
i915: use io-mapping interfaces instead of a variety of mapping kludges
resources: add io-mapping functions to dynamically map large device apertures
x86: add iomap_atomic*()/iounmap_atomic() on 32-bit using fixmaps

Documentation/io-mapping.txt

···
+The io_mapping functions in linux/io-mapping.h provide an abstraction for
+efficiently mapping small regions of an I/O device to the CPU. The initial
+usage is to support the large graphics aperture on 32-bit processors where
+ioremap_wc cannot be used to statically map the entire aperture to the CPU
+as it would consume too much of the kernel address space.
+
+A mapping object is created during driver initialization using
+
+	struct io_mapping *io_mapping_create_wc(unsigned long base,
+						unsigned long size)
+
+		'base' is the bus address of the region to be made
+		mappable, while 'size' indicates how large a mapping region to
+		enable. Both are in bytes.
+
+		This _wc variant provides a mapping which may only be used
+		with the io_mapping_map_atomic_wc or io_mapping_map_wc.
+
+With this mapping object, individual pages can be mapped either atomically
+or not, depending on the necessary scheduling environment. Of course, atomic
+maps are more efficient:
+
+	void *io_mapping_map_atomic_wc(struct io_mapping *mapping,
+				       unsigned long offset)
+
+		'offset' is the offset within the defined mapping region.
+		Accessing addresses beyond the region specified in the
+		creation function yields undefined results. Using an offset
+		which is not page aligned yields an undefined result. The
+		return value points to a single page in CPU address space.
+
+		This _wc variant returns a write-combining map to the
+		page and may only be used with mappings created by
+		io_mapping_create_wc
+
+		Note that the task may not sleep while holding this page
+		mapped.
+
+	void io_mapping_unmap_atomic(void *vaddr)
+
+		'vaddr' must be the the value returned by the last
+		io_mapping_map_atomic_wc call. This unmaps the specified
+		page and allows the task to sleep once again.
+
+If you need to sleep while holding the lock, you can use the non-atomic
+variant, although they may be significantly slower.
+
+	void *io_mapping_map_wc(struct io_mapping *mapping,
+				unsigned long offset)
+
+		This works like io_mapping_map_atomic_wc except it allows
+		the task to sleep while holding the page mapped.
+
+	void io_mapping_unmap(void *vaddr)
+
+		This works like io_mapping_unmap_atomic, except it is used
+		for pages mapped with io_mapping_map_wc.
+
+At driver close time, the io_mapping object must be freed:
+
+	void io_mapping_free(struct io_mapping *mapping)
+
+Current Implementation:
+
+The initial implementation of these functions uses existing mapping
+mechanisms and so provides only an abstraction layer and no new
+functionality.
+
+On 64-bit processors, io_mapping_create_wc calls ioremap_wc for the whole
+range, creating a permanent kernel-visible mapping to the resource. The
+map_atomic and map functions add the requested offset to the base of the
+virtual address returned by ioremap_wc.
+
+On 32-bit processors with HIGHMEM defined, io_mapping_map_atomic_wc uses
+kmap_atomic_pfn to map the specified page in an atomic fashion;
+kmap_atomic_pfn isn't really supposed to be used with device pages, but it
+provides an efficient mapping for this usage.
+
+On 32-bit processors without HIGHMEM defined, io_mapping_map_atomic_wc and
+io_mapping_map_wc both use ioremap_wc, a terribly inefficient function which
+performs an IPI to inform all processors about the new mapping. This results
+in a significant performance penalty.

arch/x86/Kconfig

···
 endmenu
 
 
+config HAVE_ATOMIC_IOMAP
+	def_bool y
+	depends on X86_32
+
 source "net/Kconfig"
 
 source "drivers/Kconfig"

arch/x86/include/asm/fixmap.h

···
 
 extern int fixmaps_set;
 
+extern pte_t *kmap_pte;
+extern pgprot_t kmap_prot;
+extern pte_t *pkmap_page_table;
+
 void __native_set_fixmap(enum fixed_addresses idx, pte_t pte);
 void native_set_fixmap(enum fixed_addresses idx,
 		       unsigned long phys, pgprot_t flags);

arch/x86/include/asm/fixmap_32.h

···
 #include <asm/acpi.h>
 #include <asm/apicdef.h>
 #include <asm/page.h>
-#ifdef CONFIG_HIGHMEM
 #include <linux/threads.h>
 #include <asm/kmap_types.h>
-#endif
 
 /*
  * Here we define all the compile-time 'special' virtual
···
 #ifdef CONFIG_X86_CYCLONE_TIMER
 	FIX_CYCLONE_TIMER, /*cyclone timer register*/
 #endif
-#ifdef CONFIG_HIGHMEM
 	FIX_KMAP_BEGIN,	/* reserved pte's for temporary kernel mappings */
 	FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1,
-#endif
 #ifdef CONFIG_PCI_MMCONFIG
 	FIX_PCIE_MCFG,
 #endif

arch/x86/include/asm/highmem.h

···
 #include <asm/kmap_types.h>
 #include <asm/tlbflush.h>
 #include <asm/paravirt.h>
+#include <asm/fixmap.h>
 
 /* declarations for highmem.c */
 extern unsigned long highstart_pfn, highend_pfn;
-
-extern pte_t *kmap_pte;
-extern pgprot_t kmap_prot;
-extern pte_t *pkmap_page_table;
 
 /*
  * Right now we initialize only a single pte table. It can be extended

arch/x86/mm/Makefile

···
 obj-y	:=  init_$(BITS).o fault.o ioremap.o extable.o pageattr.o mmap.o \
 	    pat.o pgtable.o gup.o
 
-obj-$(CONFIG_X86_32)		+= pgtable_32.o
+obj-$(CONFIG_X86_32)		+= pgtable_32.o iomap_32.o
 
 obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o
 obj-$(CONFIG_X86_PTDUMP)	+= dump_pagetables.o

arch/x86/mm/init_32.c

···
 	return 0;
 }
 
-#ifdef CONFIG_HIGHMEM
 pte_t *kmap_pte;
 pgprot_t kmap_prot;
 
···
 	kmap_prot = PAGE_KERNEL;
 }
 
+#ifdef CONFIG_HIGHMEM
 static void __init permanent_kmaps_init(pgd_t *pgd_base)
 {
 	unsigned long vaddr;
···
 #endif /* !CONFIG_NUMA */
 
 #else
-# define kmap_init()				do { } while (0)
 # define permanent_kmaps_init(pgd_base)		do { } while (0)
 # define set_highmem_pages_init()	do { } while (0)
 #endif /* CONFIG_HIGHMEM */

arch/x86/mm/iomap_32.c

···
+/*
+ * Copyright © 2008 Ingo Molnar
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
+ */
+
+#include <asm/iomap.h>
+#include <linux/module.h>
+
+/* Map 'pfn' using fixed map 'type' and protections 'prot'
+ */
+void *
+iomap_atomic_prot_pfn(unsigned long pfn, enum km_type type, pgprot_t prot)
+{
+	enum fixed_addresses idx;
+	unsigned long vaddr;
+
+	pagefault_disable();
+
+	idx = type + KM_TYPE_NR*smp_processor_id();
+	vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+	set_pte(kmap_pte-idx, pfn_pte(pfn, prot));
+	arch_flush_lazy_mmu_mode();
+
+	return (void*) vaddr;
+}
+EXPORT_SYMBOL_GPL(iomap_atomic_prot_pfn);
+
+void
+iounmap_atomic(void *kvaddr, enum km_type type)
+{
+	unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK;
+	enum fixed_addresses idx = type + KM_TYPE_NR*smp_processor_id();
+
+	/*
+	 * Force other mappings to Oops if they'll try to access this pte
+	 * without first remap it.  Keeping stale mappings around is a bad idea
+	 * also, in case the page changes cacheability attributes or becomes
+	 * a protected page in a hypervisor.
+	 */
+	if (vaddr == __fix_to_virt(FIX_KMAP_BEGIN+idx))
+		kpte_clear_flush(kmap_pte-idx, vaddr);
+
+	arch_flush_lazy_mmu_mode();
+	pagefault_enable();
+}
+EXPORT_SYMBOL_GPL(iounmap_atomic);

drivers/gpu/drm/i915/i915_drv.h

···
 #define _I915_DRV_H_
 
 #include "i915_reg.h"
+#include <linux/io-mapping.h>
 
 /* General customization:
  */
···
 
 	struct {
 		struct drm_mm gtt_space;
+
+		struct io_mapping *gtt_mapping;
 
 		/**
 		 * List of objects currently involved in rendering from the

drivers/gpu/drm/i915/i915_gem.c

···
 	return 0;
 }
 
-/*
- * Try to write quickly with an atomic kmap. Return true on success.
- *
- * If this fails (which includes a partial write), we'll redo the whole
- * thing with the slow version.
- *
- * This is a workaround for the low performance of iounmap (approximate
- * 10% cpu cost on normal 3D workloads).  kmap_atomic on HIGHMEM kernels
- * happens to let us map card memory without taking IPIs.  When the vmap
- * rework lands we should be able to dump this hack.
+/* This is the fast write path which cannot handle
+ * page faults in the source data
  */
-static inline int fast_user_write(unsigned long pfn, char __user *user_data,
-				  int l, int o)
-{
-#ifdef CONFIG_HIGHMEM
-	unsigned long unwritten;
-	char *vaddr_atomic;
 
-	vaddr_atomic = kmap_atomic_pfn(pfn, KM_USER0);
-#if WATCH_PWRITE
-	DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n",
-		 i, o, l, pfn, vaddr_atomic);
-#endif
-	unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + o, user_data, l);
-	kunmap_atomic(vaddr_atomic, KM_USER0);
-	return !unwritten;
-#else
+static inline int
+fast_user_write(struct io_mapping *mapping,
+		loff_t page_base, int page_offset,
+		char __user *user_data,
+		int length)
+{
+	char *vaddr_atomic;
+	unsigned long unwritten;
+
+	vaddr_atomic = io_mapping_map_atomic_wc(mapping, page_base);
+	unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + page_offset,
+						      user_data, length);
+	io_mapping_unmap_atomic(vaddr_atomic);
+	if (unwritten)
+		return -EFAULT;
 	return 0;
-#endif
+}
+
+/* Here's the write path which can sleep for
+ * page faults
+ */
+
+static inline int
+slow_user_write(struct io_mapping *mapping,
+		loff_t page_base, int page_offset,
+		char __user *user_data,
+		int length)
+{
+	char __iomem *vaddr;
+	unsigned long unwritten;
+
+	vaddr = io_mapping_map_wc(mapping, page_base);
+	if (vaddr == NULL)
+		return -EFAULT;
+	unwritten = __copy_from_user(vaddr + page_offset,
+				     user_data, length);
+	io_mapping_unmap(vaddr);
+	if (unwritten)
+		return -EFAULT;
+	return 0;
 }
 
 static int
···
 		    struct drm_file *file_priv)
 {
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	ssize_t remain;
-	loff_t offset;
+	loff_t offset, page_base;
 	char __user *user_data;
-	int ret = 0;
+	int page_offset, page_length;
+	int ret;
 
 	user_data = (char __user *) (uintptr_t) args->data_ptr;
 	remain = args->size;
···
 	obj_priv->dirty = 1;
 
 	while (remain > 0) {
-		unsigned long pfn;
-		int i, o, l;
-
 		/* Operation in this page
 		 *
-		 * i = page number
-		 * o = offset within page
-		 * l = bytes to copy
+		 * page_base = page offset within aperture
+		 * page_offset = offset within page
+		 * page_length = bytes to copy for this page
 		 */
-		i = offset >> PAGE_SHIFT;
-		o = offset & (PAGE_SIZE-1);
-		l = remain;
-		if ((o + l) > PAGE_SIZE)
-			l = PAGE_SIZE - o;
+		page_base = (offset & ~(PAGE_SIZE-1));
+		page_offset = offset & (PAGE_SIZE-1);
+		page_length = remain;
+		if ((page_offset + remain) > PAGE_SIZE)
+			page_length = PAGE_SIZE - page_offset;
 
-		pfn = (dev->agp->base >> PAGE_SHIFT) + i;
+		ret = fast_user_write (dev_priv->mm.gtt_mapping, page_base,
+				       page_offset, user_data, page_length);
 
-		if (!fast_user_write(pfn, user_data, l, o)) {
-			unsigned long unwritten;
-			char __iomem *vaddr;
-
-			vaddr = ioremap_wc(pfn << PAGE_SHIFT, PAGE_SIZE);
-#if WATCH_PWRITE
-			DRM_INFO("pwrite slow i %d o %d l %d "
-				 "pfn %ld vaddr %p\n",
-				 i, o, l, pfn, vaddr);
-#endif
-			if (vaddr == NULL) {
-				ret = -EFAULT;
+		/* If we get a fault while copying data, then (presumably) our
+		 * source page isn't available. In this case, use the
+		 * non-atomic function
+		 */
+		if (ret) {
+			ret = slow_user_write (dev_priv->mm.gtt_mapping,
+					       page_base, page_offset,
+					       user_data, page_length);
+			if (ret)
 				goto fail;
-			}
-			unwritten = __copy_from_user(vaddr + o, user_data, l);
-#if WATCH_PWRITE
-			DRM_INFO("unwritten %ld\n", unwritten);
-#endif
-			iounmap(vaddr);
-			if (unwritten) {
-				ret = -EFAULT;
-				goto fail;
-			}
 		}
 
-		remain -= l;
-		user_data += l;
-		offset += l;
+		remain -= page_length;
+		user_data += page_length;
+		offset += page_length;
 	}
-#if WATCH_PWRITE && 1
-	i915_gem_clflush_object(obj);
-	i915_gem_dump_object(obj, args->offset + args->size, __func__, ~0);
-	i915_gem_clflush_object(obj);
-#endif
 
 fail:
 	i915_gem_object_unpin(obj);
···
 				 struct drm_i915_gem_exec_object *entry)
 {
 	struct drm_device *dev = obj->dev;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_relocation_entry reloc;
 	struct drm_i915_gem_relocation_entry __user *relocs;
 	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 	int i, ret;
-	uint32_t last_reloc_offset = -1;
-	void __iomem *reloc_page = NULL;
+	void __iomem *reloc_page;
 
 	/* Choose the GTT offset for our buffer and put it there. */
 	ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment);
···
 		 * perform.
 		 */
 		reloc_offset = obj_priv->gtt_offset + reloc.offset;
-		if (reloc_page == NULL ||
-		    (last_reloc_offset & ~(PAGE_SIZE - 1)) !=
-		    (reloc_offset & ~(PAGE_SIZE - 1))) {
-			if (reloc_page != NULL)
-				iounmap(reloc_page);
-
-			reloc_page = ioremap_wc(dev->agp->base +
-						(reloc_offset &
-						 ~(PAGE_SIZE - 1)),
-						PAGE_SIZE);
-			last_reloc_offset = reloc_offset;
-			if (reloc_page == NULL) {
-				drm_gem_object_unreference(target_obj);
-				i915_gem_object_unpin(obj);
-				return -ENOMEM;
-			}
-		}
-
+		reloc_page = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
+						      (reloc_offset &
+						       ~(PAGE_SIZE - 1)));
 		reloc_entry = (uint32_t __iomem *)(reloc_page +
-					   (reloc_offset & (PAGE_SIZE - 1)));
+						   (reloc_offset & (PAGE_SIZE - 1)));
 		reloc_val = target_obj_priv->gtt_offset + reloc.delta;
 
 #if WATCH_BUF
···
 			  readl(reloc_entry), reloc_val);
 #endif
 		writel(reloc_val, reloc_entry);
+		io_mapping_unmap_atomic(reloc_page);
 
 		/* Write the updated presumed offset for this entry back out
 		 * to the user.
···
 
 		drm_gem_object_unreference(target_obj);
 	}
-
-	if (reloc_page != NULL)
-		iounmap(reloc_page);
 
 #if WATCH_BUF
 	if (0)
···
 	if (ret != 0)
 		return ret;
 
+	dev_priv->mm.gtt_mapping = io_mapping_create_wc(dev->agp->base,
+							dev->agp->agp_info.aper_size
+							* 1024 * 1024);
+
 	mutex_lock(&dev->struct_mutex);
 	BUG_ON(!list_empty(&dev_priv->mm.active_list));
 	BUG_ON(!list_empty(&dev_priv->mm.flushing_list));
···
 i915_gem_leavevt_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv)
 {
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	int ret;
 
 	ret = i915_gem_idle(dev);
 	drm_irq_uninstall(dev);
 
+	io_mapping_free(dev_priv->mm.gtt_mapping);
 	return ret;
 }
 

include/asm-x86/iomap.h

···
+/*
+ * Copyright © 2008 Ingo Molnar
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/uaccess.h>
+#include <asm/cacheflush.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+
+void *
+iomap_atomic_prot_pfn(unsigned long pfn, enum km_type type, pgprot_t prot);
+
+void
+iounmap_atomic(void *kvaddr, enum km_type type);

include/linux/io-mapping.h

···
+/*
+ * Copyright © 2008 Keith Packard <[e-mail]>
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+
+#ifndef _LINUX_IO_MAPPING_H
+#define _LINUX_IO_MAPPING_H
+
+#include <linux/types.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/iomap.h>
+
+/*
+ * The io_mapping mechanism provides an abstraction for mapping
+ * individual pages from an io device to the CPU in an efficient fashion.
+ *
+ * See Documentation/io_mapping.txt
+ */
+
+/* this struct isn't actually defined anywhere */
+struct io_mapping;
+
+#ifdef CONFIG_HAVE_ATOMIC_IOMAP
+
+/*
+ * For small address space machines, mapping large objects
+ * into the kernel virtual space isn't practical. Where
+ * available, use fixmap support to dynamically map pages
+ * of the object at run time.
+ */
+
+static inline struct io_mapping *
+io_mapping_create_wc(unsigned long base, unsigned long size)
+{
+	return (struct io_mapping *) base;
+}
+
+static inline void
+io_mapping_free(struct io_mapping *mapping)
+{
+}
+
+/* Atomic map/unmap */
+static inline void *
+io_mapping_map_atomic_wc(struct io_mapping *mapping, unsigned long offset)
+{
+	offset += (unsigned long) mapping;
+	return iomap_atomic_prot_pfn(offset >> PAGE_SHIFT, KM_USER0,
+				     __pgprot(__PAGE_KERNEL_WC));
+}
+
+static inline void
+io_mapping_unmap_atomic(void *vaddr)
+{
+	iounmap_atomic(vaddr, KM_USER0);
+}
+
+static inline void *
+io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
+{
+	offset += (unsigned long) mapping;
+	return ioremap_wc(offset, PAGE_SIZE);
+}
+
+static inline void
+io_mapping_unmap(void *vaddr)
+{
+	iounmap(vaddr);
+}
+
+#else
+
+/* Create the io_mapping object*/
+static inline struct io_mapping *
+io_mapping_create_wc(unsigned long base, unsigned long size)
+{
+	return (struct io_mapping *) ioremap_wc(base, size);
+}
+
+static inline void
+io_mapping_free(struct io_mapping *mapping)
+{
+	iounmap(mapping);
+}
+
+/* Atomic map/unmap */
+static inline void *
+io_mapping_map_atomic_wc(struct io_mapping *mapping, unsigned long offset)
+{
+	return ((char *) mapping) + offset;
+}
+
+static inline void
+io_mapping_unmap_atomic(void *vaddr)
+{
+}
+
+/* Non-atomic map/unmap */
+static inline void *
+io_mapping_map_wc(struct io_mapping *mapping, unsigned long offset)
+{
+	return ((char *) mapping) + offset;
+}
+
+static inline void
+io_mapping_unmap(void *vaddr)
+{
+}
+
+#endif /* HAVE_ATOMIC_IOMAP */
+
+#endif /* _LINUX_IO_MAPPING_H */