dma-buf: move to drivers/dma-buf
authorMaarten Lankhorst <maarten.lankhorst@canonical.com>
Tue, 1 Jul 2014 10:57:08 +0000 (12:57 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 8 Jul 2014 17:51:06 +0000 (10:51 -0700)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Sumit Semwal <sumit.semwal@linaro.org>
Acked-by: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Documentation/DocBook/device-drivers.tmpl
MAINTAINERS
drivers/Makefile
drivers/base/Makefile
drivers/base/dma-buf.c [deleted file]
drivers/base/reservation.c [deleted file]
drivers/dma-buf/Makefile [new file with mode: 0644]
drivers/dma-buf/dma-buf.c [new file with mode: 0644]
drivers/dma-buf/reservation.c [new file with mode: 0644]

index cc63f30de166501e37f4ec4a707dc3ad5a0c371e..ac61ebd9287574792512499641a1e31da3907db8 100644 (file)
@@ -128,8 +128,7 @@ X!Edrivers/base/interface.c
 !Edrivers/base/bus.c
      </sect1>
      <sect1><title>Device Drivers DMA Management</title>
-!Edrivers/base/dma-buf.c
-!Edrivers/base/reservation.c
+!Edrivers/dma-buf/dma-buf.c
 !Iinclude/linux/reservation.h
 !Edrivers/base/dma-coherent.c
 !Edrivers/base/dma-mapping.c
index 6813d0aa5ecf5c48868c6050532086b1c15f2b85..2eefee768d4692c6cb6802451145eee9d731268b 100644 (file)
@@ -2900,8 +2900,8 @@ S:        Maintained
 L:     linux-media@vger.kernel.org
 L:     dri-devel@lists.freedesktop.org
 L:     linaro-mm-sig@lists.linaro.org
-F:     drivers/base/dma-buf*
-F:     include/linux/dma-buf*
+F:     drivers/dma-buf/
+F:     include/linux/dma-buf* include/linux/reservation.h
 F:     Documentation/dma-buf-sharing.txt
 T:     git git://git.linaro.org/people/sumitsemwal/linux-dma-buf.git
 
index f98b50d8251d3849dfb871ca1fef595f0677bbc0..c00337be5351ebe8d0621053c3e7c4f28a75dcca 100644 (file)
@@ -61,6 +61,7 @@ obj-$(CONFIG_FB_INTEL)          += video/fbdev/intelfb/
 
 obj-$(CONFIG_PARPORT)          += parport/
 obj-y                          += base/ block/ misc/ mfd/ nfc/
+obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf/
 obj-$(CONFIG_NUBUS)            += nubus/
 obj-y                          += macintosh/
 obj-$(CONFIG_IDE)              += ide/
index 04b314e0fa516ad3fe89f47e9b75c9d6e541e8fa..4aab26ec029259377d0cd3a7ab7db9638b13f18c 100644 (file)
@@ -10,7 +10,6 @@ obj-$(CONFIG_DMA_CMA) += dma-contiguous.o
 obj-y                  += power/
 obj-$(CONFIG_HAS_DMA)  += dma-mapping.o
 obj-$(CONFIG_HAVE_GENERIC_DMA_COHERENT) += dma-coherent.o
-obj-$(CONFIG_DMA_SHARED_BUFFER) += dma-buf.o reservation.o
 obj-$(CONFIG_ISA)      += isa.o
 obj-$(CONFIG_FW_LOADER)        += firmware_class.o
 obj-$(CONFIG_NUMA)     += node.o
diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c
deleted file mode 100644 (file)
index 840c7fa..0000000
+++ /dev/null
@@ -1,743 +0,0 @@
-/*
- * Framework for buffer objects that can be shared across devices/subsystems.
- *
- * Copyright(C) 2011 Linaro Limited. All rights reserved.
- * Author: Sumit Semwal <sumit.semwal@ti.com>
- *
- * Many thanks to linaro-mm-sig list, and specially
- * Arnd Bergmann <arnd@arndb.de>, Rob Clark <rob@ti.com> and
- * Daniel Vetter <daniel@ffwll.ch> for their support in creation and
- * refining of this idea.
- *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * 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, see <http://www.gnu.org/licenses/>.
- */
-
-#include <linux/fs.h>
-#include <linux/slab.h>
-#include <linux/dma-buf.h>
-#include <linux/anon_inodes.h>
-#include <linux/export.h>
-#include <linux/debugfs.h>
-#include <linux/seq_file.h>
-
-static inline int is_dma_buf_file(struct file *);
-
-struct dma_buf_list {
-       struct list_head head;
-       struct mutex lock;
-};
-
-static struct dma_buf_list db_list;
-
-static int dma_buf_release(struct inode *inode, struct file *file)
-{
-       struct dma_buf *dmabuf;
-
-       if (!is_dma_buf_file(file))
-               return -EINVAL;
-
-       dmabuf = file->private_data;
-
-       BUG_ON(dmabuf->vmapping_counter);
-
-       dmabuf->ops->release(dmabuf);
-
-       mutex_lock(&db_list.lock);
-       list_del(&dmabuf->list_node);
-       mutex_unlock(&db_list.lock);
-
-       kfree(dmabuf);
-       return 0;
-}
-
-static int dma_buf_mmap_internal(struct file *file, struct vm_area_struct *vma)
-{
-       struct dma_buf *dmabuf;
-
-       if (!is_dma_buf_file(file))
-               return -EINVAL;
-
-       dmabuf = file->private_data;
-
-       /* check for overflowing the buffer's size */
-       if (vma->vm_pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) >
-           dmabuf->size >> PAGE_SHIFT)
-               return -EINVAL;
-
-       return dmabuf->ops->mmap(dmabuf, vma);
-}
-
-static loff_t dma_buf_llseek(struct file *file, loff_t offset, int whence)
-{
-       struct dma_buf *dmabuf;
-       loff_t base;
-
-       if (!is_dma_buf_file(file))
-               return -EBADF;
-
-       dmabuf = file->private_data;
-
-       /* only support discovering the end of the buffer,
-          but also allow SEEK_SET to maintain the idiomatic
-          SEEK_END(0), SEEK_CUR(0) pattern */
-       if (whence == SEEK_END)
-               base = dmabuf->size;
-       else if (whence == SEEK_SET)
-               base = 0;
-       else
-               return -EINVAL;
-
-       if (offset != 0)
-               return -EINVAL;
-
-       return base + offset;
-}
-
-static const struct file_operations dma_buf_fops = {
-       .release        = dma_buf_release,
-       .mmap           = dma_buf_mmap_internal,
-       .llseek         = dma_buf_llseek,
-};
-
-/*
- * is_dma_buf_file - Check if struct file* is associated with dma_buf
- */
-static inline int is_dma_buf_file(struct file *file)
-{
-       return file->f_op == &dma_buf_fops;
-}
-
-/**
- * dma_buf_export_named - Creates a new dma_buf, and associates an anon file
- * with this buffer, so it can be exported.
- * Also connect the allocator specific data and ops to the buffer.
- * Additionally, provide a name string for exporter; useful in debugging.
- *
- * @priv:      [in]    Attach private data of allocator to this buffer
- * @ops:       [in]    Attach allocator-defined dma buf ops to the new buffer.
- * @size:      [in]    Size of the buffer
- * @flags:     [in]    mode flags for the file.
- * @exp_name:  [in]    name of the exporting module - useful for debugging.
- *
- * Returns, on success, a newly created dma_buf object, which wraps the
- * supplied private data and operations for dma_buf_ops. On either missing
- * ops, or error in allocating struct dma_buf, will return negative error.
- *
- */
-struct dma_buf *dma_buf_export_named(void *priv, const struct dma_buf_ops *ops,
-                               size_t size, int flags, const char *exp_name)
-{
-       struct dma_buf *dmabuf;
-       struct file *file;
-
-       if (WARN_ON(!priv || !ops
-                         || !ops->map_dma_buf
-                         || !ops->unmap_dma_buf
-                         || !ops->release
-                         || !ops->kmap_atomic
-                         || !ops->kmap
-                         || !ops->mmap)) {
-               return ERR_PTR(-EINVAL);
-       }
-
-       dmabuf = kzalloc(sizeof(struct dma_buf), GFP_KERNEL);
-       if (dmabuf == NULL)
-               return ERR_PTR(-ENOMEM);
-
-       dmabuf->priv = priv;
-       dmabuf->ops = ops;
-       dmabuf->size = size;
-       dmabuf->exp_name = exp_name;
-
-       file = anon_inode_getfile("dmabuf", &dma_buf_fops, dmabuf, flags);
-       if (IS_ERR(file)) {
-               kfree(dmabuf);
-               return ERR_CAST(file);
-       }
-
-       file->f_mode |= FMODE_LSEEK;
-       dmabuf->file = file;
-
-       mutex_init(&dmabuf->lock);
-       INIT_LIST_HEAD(&dmabuf->attachments);
-
-       mutex_lock(&db_list.lock);
-       list_add(&dmabuf->list_node, &db_list.head);
-       mutex_unlock(&db_list.lock);
-
-       return dmabuf;
-}
-EXPORT_SYMBOL_GPL(dma_buf_export_named);
-
-
-/**
- * dma_buf_fd - returns a file descriptor for the given dma_buf
- * @dmabuf:    [in]    pointer to dma_buf for which fd is required.
- * @flags:      [in]    flags to give to fd
- *
- * On success, returns an associated 'fd'. Else, returns error.
- */
-int dma_buf_fd(struct dma_buf *dmabuf, int flags)
-{
-       int fd;
-
-       if (!dmabuf || !dmabuf->file)
-               return -EINVAL;
-
-       fd = get_unused_fd_flags(flags);
-       if (fd < 0)
-               return fd;
-
-       fd_install(fd, dmabuf->file);
-
-       return fd;
-}
-EXPORT_SYMBOL_GPL(dma_buf_fd);
-
-/**
- * dma_buf_get - returns the dma_buf structure related to an fd
- * @fd:        [in]    fd associated with the dma_buf to be returned
- *
- * On success, returns the dma_buf structure associated with an fd; uses
- * file's refcounting done by fget to increase refcount. returns ERR_PTR
- * otherwise.
- */
-struct dma_buf *dma_buf_get(int fd)
-{
-       struct file *file;
-
-       file = fget(fd);
-
-       if (!file)
-               return ERR_PTR(-EBADF);
-
-       if (!is_dma_buf_file(file)) {
-               fput(file);
-               return ERR_PTR(-EINVAL);
-       }
-
-       return file->private_data;
-}
-EXPORT_SYMBOL_GPL(dma_buf_get);
-
-/**
- * dma_buf_put - decreases refcount of the buffer
- * @dmabuf:    [in]    buffer to reduce refcount of
- *
- * Uses file's refcounting done implicitly by fput()
- */
-void dma_buf_put(struct dma_buf *dmabuf)
-{
-       if (WARN_ON(!dmabuf || !dmabuf->file))
-               return;
-
-       fput(dmabuf->file);
-}
-EXPORT_SYMBOL_GPL(dma_buf_put);
-
-/**
- * dma_buf_attach - Add the device to dma_buf's attachments list; optionally,
- * calls attach() of dma_buf_ops to allow device-specific attach functionality
- * @dmabuf:    [in]    buffer to attach device to.
- * @dev:       [in]    device to be attached.
- *
- * Returns struct dma_buf_attachment * for this attachment; returns ERR_PTR on
- * error.
- */
-struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf,
-                                         struct device *dev)
-{
-       struct dma_buf_attachment *attach;
-       int ret;
-
-       if (WARN_ON(!dmabuf || !dev))
-               return ERR_PTR(-EINVAL);
-
-       attach = kzalloc(sizeof(struct dma_buf_attachment), GFP_KERNEL);
-       if (attach == NULL)
-               return ERR_PTR(-ENOMEM);
-
-       attach->dev = dev;
-       attach->dmabuf = dmabuf;
-
-       mutex_lock(&dmabuf->lock);
-
-       if (dmabuf->ops->attach) {
-               ret = dmabuf->ops->attach(dmabuf, dev, attach);
-               if (ret)
-                       goto err_attach;
-       }
-       list_add(&attach->node, &dmabuf->attachments);
-
-       mutex_unlock(&dmabuf->lock);
-       return attach;
-
-err_attach:
-       kfree(attach);
-       mutex_unlock(&dmabuf->lock);
-       return ERR_PTR(ret);
-}
-EXPORT_SYMBOL_GPL(dma_buf_attach);
-
-/**
- * dma_buf_detach - Remove the given attachment from dmabuf's attachments list;
- * optionally calls detach() of dma_buf_ops for device-specific detach
- * @dmabuf:    [in]    buffer to detach from.
- * @attach:    [in]    attachment to be detached; is free'd after this call.
- *
- */
-void dma_buf_detach(struct dma_buf *dmabuf, struct dma_buf_attachment *attach)
-{
-       if (WARN_ON(!dmabuf || !attach))
-               return;
-
-       mutex_lock(&dmabuf->lock);
-       list_del(&attach->node);
-       if (dmabuf->ops->detach)
-               dmabuf->ops->detach(dmabuf, attach);
-
-       mutex_unlock(&dmabuf->lock);
-       kfree(attach);
-}
-EXPORT_SYMBOL_GPL(dma_buf_detach);
-
-/**
- * dma_buf_map_attachment - Returns the scatterlist table of the attachment;
- * mapped into _device_ address space. Is a wrapper for map_dma_buf() of the
- * dma_buf_ops.
- * @attach:    [in]    attachment whose scatterlist is to be returned
- * @direction: [in]    direction of DMA transfer
- *
- * Returns sg_table containing the scatterlist to be returned; returns ERR_PTR
- * on error.
- */
-struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
-                                       enum dma_data_direction direction)
-{
-       struct sg_table *sg_table = ERR_PTR(-EINVAL);
-
-       might_sleep();
-
-       if (WARN_ON(!attach || !attach->dmabuf))
-               return ERR_PTR(-EINVAL);
-
-       sg_table = attach->dmabuf->ops->map_dma_buf(attach, direction);
-       if (!sg_table)
-               sg_table = ERR_PTR(-ENOMEM);
-
-       return sg_table;
-}
-EXPORT_SYMBOL_GPL(dma_buf_map_attachment);
-
-/**
- * dma_buf_unmap_attachment - unmaps and decreases usecount of the buffer;might
- * deallocate the scatterlist associated. Is a wrapper for unmap_dma_buf() of
- * dma_buf_ops.
- * @attach:    [in]    attachment to unmap buffer from
- * @sg_table:  [in]    scatterlist info of the buffer to unmap
- * @direction:  [in]    direction of DMA transfer
- *
- */
-void dma_buf_unmap_attachment(struct dma_buf_attachment *attach,
-                               struct sg_table *sg_table,
-                               enum dma_data_direction direction)
-{
-       might_sleep();
-
-       if (WARN_ON(!attach || !attach->dmabuf || !sg_table))
-               return;
-
-       attach->dmabuf->ops->unmap_dma_buf(attach, sg_table,
-                                               direction);
-}
-EXPORT_SYMBOL_GPL(dma_buf_unmap_attachment);
-
-
-/**
- * dma_buf_begin_cpu_access - Must be called before accessing a dma_buf from the
- * cpu in the kernel context. Calls begin_cpu_access to allow exporter-specific
- * preparations. Coherency is only guaranteed in the specified range for the
- * specified access direction.
- * @dmabuf:    [in]    buffer to prepare cpu access for.
- * @start:     [in]    start of range for cpu access.
- * @len:       [in]    length of range for cpu access.
- * @direction: [in]    length of range for cpu access.
- *
- * Can return negative error values, returns 0 on success.
- */
-int dma_buf_begin_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,
-                            enum dma_data_direction direction)
-{
-       int ret = 0;
-
-       if (WARN_ON(!dmabuf))
-               return -EINVAL;
-
-       if (dmabuf->ops->begin_cpu_access)
-               ret = dmabuf->ops->begin_cpu_access(dmabuf, start, len, direction);
-
-       return ret;
-}
-EXPORT_SYMBOL_GPL(dma_buf_begin_cpu_access);
-
-/**
- * dma_buf_end_cpu_access - Must be called after accessing a dma_buf from the
- * cpu in the kernel context. Calls end_cpu_access to allow exporter-specific
- * actions. Coherency is only guaranteed in the specified range for the
- * specified access direction.
- * @dmabuf:    [in]    buffer to complete cpu access for.
- * @start:     [in]    start of range for cpu access.
- * @len:       [in]    length of range for cpu access.
- * @direction: [in]    length of range for cpu access.
- *
- * This call must always succeed.
- */
-void dma_buf_end_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,
-                           enum dma_data_direction direction)
-{
-       WARN_ON(!dmabuf);
-
-       if (dmabuf->ops->end_cpu_access)
-               dmabuf->ops->end_cpu_access(dmabuf, start, len, direction);
-}
-EXPORT_SYMBOL_GPL(dma_buf_end_cpu_access);
-
-/**
- * dma_buf_kmap_atomic - Map a page of the buffer object into kernel address
- * space. The same restrictions as for kmap_atomic and friends apply.
- * @dmabuf:    [in]    buffer to map page from.
- * @page_num:  [in]    page in PAGE_SIZE units to map.
- *
- * This call must always succeed, any necessary preparations that might fail
- * need to be done in begin_cpu_access.
- */
-void *dma_buf_kmap_atomic(struct dma_buf *dmabuf, unsigned long page_num)
-{
-       WARN_ON(!dmabuf);
-
-       return dmabuf->ops->kmap_atomic(dmabuf, page_num);
-}
-EXPORT_SYMBOL_GPL(dma_buf_kmap_atomic);
-
-/**
- * dma_buf_kunmap_atomic - Unmap a page obtained by dma_buf_kmap_atomic.
- * @dmabuf:    [in]    buffer to unmap page from.
- * @page_num:  [in]    page in PAGE_SIZE units to unmap.
- * @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap_atomic.
- *
- * This call must always succeed.
- */
-void dma_buf_kunmap_atomic(struct dma_buf *dmabuf, unsigned long page_num,
-                          void *vaddr)
-{
-       WARN_ON(!dmabuf);
-
-       if (dmabuf->ops->kunmap_atomic)
-               dmabuf->ops->kunmap_atomic(dmabuf, page_num, vaddr);
-}
-EXPORT_SYMBOL_GPL(dma_buf_kunmap_atomic);
-
-/**
- * dma_buf_kmap - Map a page of the buffer object into kernel address space. The
- * same restrictions as for kmap and friends apply.
- * @dmabuf:    [in]    buffer to map page from.
- * @page_num:  [in]    page in PAGE_SIZE units to map.
- *
- * This call must always succeed, any necessary preparations that might fail
- * need to be done in begin_cpu_access.
- */
-void *dma_buf_kmap(struct dma_buf *dmabuf, unsigned long page_num)
-{
-       WARN_ON(!dmabuf);
-
-       return dmabuf->ops->kmap(dmabuf, page_num);
-}
-EXPORT_SYMBOL_GPL(dma_buf_kmap);
-
-/**
- * dma_buf_kunmap - Unmap a page obtained by dma_buf_kmap.
- * @dmabuf:    [in]    buffer to unmap page from.
- * @page_num:  [in]    page in PAGE_SIZE units to unmap.
- * @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap.
- *
- * This call must always succeed.
- */
-void dma_buf_kunmap(struct dma_buf *dmabuf, unsigned long page_num,
-                   void *vaddr)
-{
-       WARN_ON(!dmabuf);
-
-       if (dmabuf->ops->kunmap)
-               dmabuf->ops->kunmap(dmabuf, page_num, vaddr);
-}
-EXPORT_SYMBOL_GPL(dma_buf_kunmap);
-
-
-/**
- * dma_buf_mmap - Setup up a userspace mmap with the given vma
- * @dmabuf:    [in]    buffer that should back the vma
- * @vma:       [in]    vma for the mmap
- * @pgoff:     [in]    offset in pages where this mmap should start within the
- *                     dma-buf buffer.
- *
- * This function adjusts the passed in vma so that it points at the file of the
- * dma_buf operation. It also adjusts the starting pgoff and does bounds
- * checking on the size of the vma. Then it calls the exporters mmap function to
- * set up the mapping.
- *
- * Can return negative error values, returns 0 on success.
- */
-int dma_buf_mmap(struct dma_buf *dmabuf, struct vm_area_struct *vma,
-                unsigned long pgoff)
-{
-       struct file *oldfile;
-       int ret;
-
-       if (WARN_ON(!dmabuf || !vma))
-               return -EINVAL;
-
-       /* check for offset overflow */
-       if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) < pgoff)
-               return -EOVERFLOW;
-
-       /* check for overflowing the buffer's size */
-       if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) >
-           dmabuf->size >> PAGE_SHIFT)
-               return -EINVAL;
-
-       /* readjust the vma */
-       get_file(dmabuf->file);
-       oldfile = vma->vm_file;
-       vma->vm_file = dmabuf->file;
-       vma->vm_pgoff = pgoff;
-
-       ret = dmabuf->ops->mmap(dmabuf, vma);
-       if (ret) {
-               /* restore old parameters on failure */
-               vma->vm_file = oldfile;
-               fput(dmabuf->file);
-       } else {
-               if (oldfile)
-                       fput(oldfile);
-       }
-       return ret;
-
-}
-EXPORT_SYMBOL_GPL(dma_buf_mmap);
-
-/**
- * dma_buf_vmap - Create virtual mapping for the buffer object into kernel
- * address space. Same restrictions as for vmap and friends apply.
- * @dmabuf:    [in]    buffer to vmap
- *
- * This call may fail due to lack of virtual mapping address space.
- * These calls are optional in drivers. The intended use for them
- * is for mapping objects linear in kernel space for high use objects.
- * Please attempt to use kmap/kunmap before thinking about these interfaces.
- *
- * Returns NULL on error.
- */
-void *dma_buf_vmap(struct dma_buf *dmabuf)
-{
-       void *ptr;
-
-       if (WARN_ON(!dmabuf))
-               return NULL;
-
-       if (!dmabuf->ops->vmap)
-               return NULL;
-
-       mutex_lock(&dmabuf->lock);
-       if (dmabuf->vmapping_counter) {
-               dmabuf->vmapping_counter++;
-               BUG_ON(!dmabuf->vmap_ptr);
-               ptr = dmabuf->vmap_ptr;
-               goto out_unlock;
-       }
-
-       BUG_ON(dmabuf->vmap_ptr);
-
-       ptr = dmabuf->ops->vmap(dmabuf);
-       if (WARN_ON_ONCE(IS_ERR(ptr)))
-               ptr = NULL;
-       if (!ptr)
-               goto out_unlock;
-
-       dmabuf->vmap_ptr = ptr;
-       dmabuf->vmapping_counter = 1;
-
-out_unlock:
-       mutex_unlock(&dmabuf->lock);
-       return ptr;
-}
-EXPORT_SYMBOL_GPL(dma_buf_vmap);
-
-/**
- * dma_buf_vunmap - Unmap a vmap obtained by dma_buf_vmap.
- * @dmabuf:    [in]    buffer to vunmap
- * @vaddr:     [in]    vmap to vunmap
- */
-void dma_buf_vunmap(struct dma_buf *dmabuf, void *vaddr)
-{
-       if (WARN_ON(!dmabuf))
-               return;
-
-       BUG_ON(!dmabuf->vmap_ptr);
-       BUG_ON(dmabuf->vmapping_counter == 0);
-       BUG_ON(dmabuf->vmap_ptr != vaddr);
-
-       mutex_lock(&dmabuf->lock);
-       if (--dmabuf->vmapping_counter == 0) {
-               if (dmabuf->ops->vunmap)
-                       dmabuf->ops->vunmap(dmabuf, vaddr);
-               dmabuf->vmap_ptr = NULL;
-       }
-       mutex_unlock(&dmabuf->lock);
-}
-EXPORT_SYMBOL_GPL(dma_buf_vunmap);
-
-#ifdef CONFIG_DEBUG_FS
-static int dma_buf_describe(struct seq_file *s)
-{
-       int ret;
-       struct dma_buf *buf_obj;
-       struct dma_buf_attachment *attach_obj;
-       int count = 0, attach_count;
-       size_t size = 0;
-
-       ret = mutex_lock_interruptible(&db_list.lock);
-
-       if (ret)
-               return ret;
-
-       seq_puts(s, "\nDma-buf Objects:\n");
-       seq_puts(s, "size\tflags\tmode\tcount\texp_name\n");
-
-       list_for_each_entry(buf_obj, &db_list.head, list_node) {
-               ret = mutex_lock_interruptible(&buf_obj->lock);
-
-               if (ret) {
-                       seq_puts(s,
-                                "\tERROR locking buffer object: skipping\n");
-                       continue;
-               }
-
-               seq_printf(s, "%08zu\t%08x\t%08x\t%08ld\t%s\n",
-                               buf_obj->size,
-                               buf_obj->file->f_flags, buf_obj->file->f_mode,
-                               (long)(buf_obj->file->f_count.counter),
-                               buf_obj->exp_name);
-
-               seq_puts(s, "\tAttached Devices:\n");
-               attach_count = 0;
-
-               list_for_each_entry(attach_obj, &buf_obj->attachments, node) {
-                       seq_puts(s, "\t");
-
-                       seq_printf(s, "%s\n", dev_name(attach_obj->dev));
-                       attach_count++;
-               }
-
-               seq_printf(s, "Total %d devices attached\n\n",
-                               attach_count);
-
-               count++;
-               size += buf_obj->size;
-               mutex_unlock(&buf_obj->lock);
-       }
-
-       seq_printf(s, "\nTotal %d objects, %zu bytes\n", count, size);
-
-       mutex_unlock(&db_list.lock);
-       return 0;
-}
-
-static int dma_buf_show(struct seq_file *s, void *unused)
-{
-       void (*func)(struct seq_file *) = s->private;
-       func(s);
-       return 0;
-}
-
-static int dma_buf_debug_open(struct inode *inode, struct file *file)
-{
-       return single_open(file, dma_buf_show, inode->i_private);
-}
-
-static const struct file_operations dma_buf_debug_fops = {
-       .open           = dma_buf_debug_open,
-       .read           = seq_read,
-       .llseek         = seq_lseek,
-       .release        = single_release,
-};
-
-static struct dentry *dma_buf_debugfs_dir;
-
-static int dma_buf_init_debugfs(void)
-{
-       int err = 0;
-       dma_buf_debugfs_dir = debugfs_create_dir("dma_buf", NULL);
-       if (IS_ERR(dma_buf_debugfs_dir)) {
-               err = PTR_ERR(dma_buf_debugfs_dir);
-               dma_buf_debugfs_dir = NULL;
-               return err;
-       }
-
-       err = dma_buf_debugfs_create_file("bufinfo", dma_buf_describe);
-
-       if (err)
-               pr_debug("dma_buf: debugfs: failed to create node bufinfo\n");
-
-       return err;
-}
-
-static void dma_buf_uninit_debugfs(void)
-{
-       if (dma_buf_debugfs_dir)
-               debugfs_remove_recursive(dma_buf_debugfs_dir);
-}
-
-int dma_buf_debugfs_create_file(const char *name,
-                               int (*write)(struct seq_file *))
-{
-       struct dentry *d;
-
-       d = debugfs_create_file(name, S_IRUGO, dma_buf_debugfs_dir,
-                       write, &dma_buf_debug_fops);
-
-       return PTR_ERR_OR_ZERO(d);
-}
-#else
-static inline int dma_buf_init_debugfs(void)
-{
-       return 0;
-}
-static inline void dma_buf_uninit_debugfs(void)
-{
-}
-#endif
-
-static int __init dma_buf_init(void)
-{
-       mutex_init(&db_list.lock);
-       INIT_LIST_HEAD(&db_list.head);
-       dma_buf_init_debugfs();
-       return 0;
-}
-subsys_initcall(dma_buf_init);
-
-static void __exit dma_buf_deinit(void)
-{
-       dma_buf_uninit_debugfs();
-}
-__exitcall(dma_buf_deinit);
diff --git a/drivers/base/reservation.c b/drivers/base/reservation.c
deleted file mode 100644 (file)
index a73fbf3..0000000
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Copyright (C) 2012-2013 Canonical Ltd
- *
- * Based on bo.c which bears the following copyright notice,
- * but is dual licensed:
- *
- * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-/*
- * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
- */
-
-#include <linux/reservation.h>
-#include <linux/export.h>
-
-DEFINE_WW_CLASS(reservation_ww_class);
-EXPORT_SYMBOL(reservation_ww_class);
diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
new file mode 100644 (file)
index 0000000..4a4f4c9
--- /dev/null
@@ -0,0 +1 @@
+obj-y := dma-buf.o reservation.o
diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
new file mode 100644 (file)
index 0000000..840c7fa
--- /dev/null
@@ -0,0 +1,743 @@
+/*
+ * Framework for buffer objects that can be shared across devices/subsystems.
+ *
+ * Copyright(C) 2011 Linaro Limited. All rights reserved.
+ * Author: Sumit Semwal <sumit.semwal@ti.com>
+ *
+ * Many thanks to linaro-mm-sig list, and specially
+ * Arnd Bergmann <arnd@arndb.de>, Rob Clark <rob@ti.com> and
+ * Daniel Vetter <daniel@ffwll.ch> for their support in creation and
+ * refining of this idea.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/dma-buf.h>
+#include <linux/anon_inodes.h>
+#include <linux/export.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+
+static inline int is_dma_buf_file(struct file *);
+
+struct dma_buf_list {
+       struct list_head head;
+       struct mutex lock;
+};
+
+static struct dma_buf_list db_list;
+
+static int dma_buf_release(struct inode *inode, struct file *file)
+{
+       struct dma_buf *dmabuf;
+
+       if (!is_dma_buf_file(file))
+               return -EINVAL;
+
+       dmabuf = file->private_data;
+
+       BUG_ON(dmabuf->vmapping_counter);
+
+       dmabuf->ops->release(dmabuf);
+
+       mutex_lock(&db_list.lock);
+       list_del(&dmabuf->list_node);
+       mutex_unlock(&db_list.lock);
+
+       kfree(dmabuf);
+       return 0;
+}
+
+static int dma_buf_mmap_internal(struct file *file, struct vm_area_struct *vma)
+{
+       struct dma_buf *dmabuf;
+
+       if (!is_dma_buf_file(file))
+               return -EINVAL;
+
+       dmabuf = file->private_data;
+
+       /* check for overflowing the buffer's size */
+       if (vma->vm_pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) >
+           dmabuf->size >> PAGE_SHIFT)
+               return -EINVAL;
+
+       return dmabuf->ops->mmap(dmabuf, vma);
+}
+
+static loff_t dma_buf_llseek(struct file *file, loff_t offset, int whence)
+{
+       struct dma_buf *dmabuf;
+       loff_t base;
+
+       if (!is_dma_buf_file(file))
+               return -EBADF;
+
+       dmabuf = file->private_data;
+
+       /* only support discovering the end of the buffer,
+          but also allow SEEK_SET to maintain the idiomatic
+          SEEK_END(0), SEEK_CUR(0) pattern */
+       if (whence == SEEK_END)
+               base = dmabuf->size;
+       else if (whence == SEEK_SET)
+               base = 0;
+       else
+               return -EINVAL;
+
+       if (offset != 0)
+               return -EINVAL;
+
+       return base + offset;
+}
+
+static const struct file_operations dma_buf_fops = {
+       .release        = dma_buf_release,
+       .mmap           = dma_buf_mmap_internal,
+       .llseek         = dma_buf_llseek,
+};
+
+/*
+ * is_dma_buf_file - Check if struct file* is associated with dma_buf
+ */
+static inline int is_dma_buf_file(struct file *file)
+{
+       return file->f_op == &dma_buf_fops;
+}
+
+/**
+ * dma_buf_export_named - Creates a new dma_buf, and associates an anon file
+ * with this buffer, so it can be exported.
+ * Also connect the allocator specific data and ops to the buffer.
+ * Additionally, provide a name string for exporter; useful in debugging.
+ *
+ * @priv:      [in]    Attach private data of allocator to this buffer
+ * @ops:       [in]    Attach allocator-defined dma buf ops to the new buffer.
+ * @size:      [in]    Size of the buffer
+ * @flags:     [in]    mode flags for the file.
+ * @exp_name:  [in]    name of the exporting module - useful for debugging.
+ *
+ * Returns, on success, a newly created dma_buf object, which wraps the
+ * supplied private data and operations for dma_buf_ops. On either missing
+ * ops, or error in allocating struct dma_buf, will return negative error.
+ *
+ */
+struct dma_buf *dma_buf_export_named(void *priv, const struct dma_buf_ops *ops,
+                               size_t size, int flags, const char *exp_name)
+{
+       struct dma_buf *dmabuf;
+       struct file *file;
+
+       if (WARN_ON(!priv || !ops
+                         || !ops->map_dma_buf
+                         || !ops->unmap_dma_buf
+                         || !ops->release
+                         || !ops->kmap_atomic
+                         || !ops->kmap
+                         || !ops->mmap)) {
+               return ERR_PTR(-EINVAL);
+       }
+
+       dmabuf = kzalloc(sizeof(struct dma_buf), GFP_KERNEL);
+       if (dmabuf == NULL)
+               return ERR_PTR(-ENOMEM);
+
+       dmabuf->priv = priv;
+       dmabuf->ops = ops;
+       dmabuf->size = size;
+       dmabuf->exp_name = exp_name;
+
+       file = anon_inode_getfile("dmabuf", &dma_buf_fops, dmabuf, flags);
+       if (IS_ERR(file)) {
+               kfree(dmabuf);
+               return ERR_CAST(file);
+       }
+
+       file->f_mode |= FMODE_LSEEK;
+       dmabuf->file = file;
+
+       mutex_init(&dmabuf->lock);
+       INIT_LIST_HEAD(&dmabuf->attachments);
+
+       mutex_lock(&db_list.lock);
+       list_add(&dmabuf->list_node, &db_list.head);
+       mutex_unlock(&db_list.lock);
+
+       return dmabuf;
+}
+EXPORT_SYMBOL_GPL(dma_buf_export_named);
+
+
+/**
+ * dma_buf_fd - returns a file descriptor for the given dma_buf
+ * @dmabuf:    [in]    pointer to dma_buf for which fd is required.
+ * @flags:      [in]    flags to give to fd
+ *
+ * On success, returns an associated 'fd'. Else, returns error.
+ */
+int dma_buf_fd(struct dma_buf *dmabuf, int flags)
+{
+       int fd;
+
+       if (!dmabuf || !dmabuf->file)
+               return -EINVAL;
+
+       fd = get_unused_fd_flags(flags);
+       if (fd < 0)
+               return fd;
+
+       fd_install(fd, dmabuf->file);
+
+       return fd;
+}
+EXPORT_SYMBOL_GPL(dma_buf_fd);
+
+/**
+ * dma_buf_get - returns the dma_buf structure related to an fd
+ * @fd:        [in]    fd associated with the dma_buf to be returned
+ *
+ * On success, returns the dma_buf structure associated with an fd; uses
+ * file's refcounting done by fget to increase refcount. returns ERR_PTR
+ * otherwise.
+ */
+struct dma_buf *dma_buf_get(int fd)
+{
+       struct file *file;
+
+       file = fget(fd);
+
+       if (!file)
+               return ERR_PTR(-EBADF);
+
+       if (!is_dma_buf_file(file)) {
+               fput(file);
+               return ERR_PTR(-EINVAL);
+       }
+
+       return file->private_data;
+}
+EXPORT_SYMBOL_GPL(dma_buf_get);
+
+/**
+ * dma_buf_put - decreases refcount of the buffer
+ * @dmabuf:    [in]    buffer to reduce refcount of
+ *
+ * Uses file's refcounting done implicitly by fput()
+ */
+void dma_buf_put(struct dma_buf *dmabuf)
+{
+       if (WARN_ON(!dmabuf || !dmabuf->file))
+               return;
+
+       fput(dmabuf->file);
+}
+EXPORT_SYMBOL_GPL(dma_buf_put);
+
+/**
+ * dma_buf_attach - Add the device to dma_buf's attachments list; optionally,
+ * calls attach() of dma_buf_ops to allow device-specific attach functionality
+ * @dmabuf:    [in]    buffer to attach device to.
+ * @dev:       [in]    device to be attached.
+ *
+ * Returns struct dma_buf_attachment * for this attachment; returns ERR_PTR on
+ * error.
+ */
+struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf,
+                                         struct device *dev)
+{
+       struct dma_buf_attachment *attach;
+       int ret;
+
+       if (WARN_ON(!dmabuf || !dev))
+               return ERR_PTR(-EINVAL);
+
+       attach = kzalloc(sizeof(struct dma_buf_attachment), GFP_KERNEL);
+       if (attach == NULL)
+               return ERR_PTR(-ENOMEM);
+
+       attach->dev = dev;
+       attach->dmabuf = dmabuf;
+
+       mutex_lock(&dmabuf->lock);
+
+       if (dmabuf->ops->attach) {
+               ret = dmabuf->ops->attach(dmabuf, dev, attach);
+               if (ret)
+                       goto err_attach;
+       }
+       list_add(&attach->node, &dmabuf->attachments);
+
+       mutex_unlock(&dmabuf->lock);
+       return attach;
+
+err_attach:
+       kfree(attach);
+       mutex_unlock(&dmabuf->lock);
+       return ERR_PTR(ret);
+}
+EXPORT_SYMBOL_GPL(dma_buf_attach);
+
+/**
+ * dma_buf_detach - Remove the given attachment from dmabuf's attachments list;
+ * optionally calls detach() of dma_buf_ops for device-specific detach
+ * @dmabuf:    [in]    buffer to detach from.
+ * @attach:    [in]    attachment to be detached; is free'd after this call.
+ *
+ */
+void dma_buf_detach(struct dma_buf *dmabuf, struct dma_buf_attachment *attach)
+{
+       if (WARN_ON(!dmabuf || !attach))
+               return;
+
+       mutex_lock(&dmabuf->lock);
+       list_del(&attach->node);
+       if (dmabuf->ops->detach)
+               dmabuf->ops->detach(dmabuf, attach);
+
+       mutex_unlock(&dmabuf->lock);
+       kfree(attach);
+}
+EXPORT_SYMBOL_GPL(dma_buf_detach);
+
+/**
+ * dma_buf_map_attachment - Returns the scatterlist table of the attachment;
+ * mapped into _device_ address space. Is a wrapper for map_dma_buf() of the
+ * dma_buf_ops.
+ * @attach:    [in]    attachment whose scatterlist is to be returned
+ * @direction: [in]    direction of DMA transfer
+ *
+ * Returns sg_table containing the scatterlist to be returned; returns ERR_PTR
+ * on error.
+ */
+struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
+                                       enum dma_data_direction direction)
+{
+       struct sg_table *sg_table = ERR_PTR(-EINVAL);
+
+       might_sleep();
+
+       if (WARN_ON(!attach || !attach->dmabuf))
+               return ERR_PTR(-EINVAL);
+
+       sg_table = attach->dmabuf->ops->map_dma_buf(attach, direction);
+       if (!sg_table)
+               sg_table = ERR_PTR(-ENOMEM);
+
+       return sg_table;
+}
+EXPORT_SYMBOL_GPL(dma_buf_map_attachment);
+
+/**
+ * dma_buf_unmap_attachment - unmaps and decreases usecount of the buffer;might
+ * deallocate the scatterlist associated. Is a wrapper for unmap_dma_buf() of
+ * dma_buf_ops.
+ * @attach:    [in]    attachment to unmap buffer from
+ * @sg_table:  [in]    scatterlist info of the buffer to unmap
+ * @direction:  [in]    direction of DMA transfer
+ *
+ */
+void dma_buf_unmap_attachment(struct dma_buf_attachment *attach,
+                               struct sg_table *sg_table,
+                               enum dma_data_direction direction)
+{
+       might_sleep();
+
+       if (WARN_ON(!attach || !attach->dmabuf || !sg_table))
+               return;
+
+       attach->dmabuf->ops->unmap_dma_buf(attach, sg_table,
+                                               direction);
+}
+EXPORT_SYMBOL_GPL(dma_buf_unmap_attachment);
+
+
+/**
+ * dma_buf_begin_cpu_access - Must be called before accessing a dma_buf from the
+ * cpu in the kernel context. Calls begin_cpu_access to allow exporter-specific
+ * preparations. Coherency is only guaranteed in the specified range for the
+ * specified access direction.
+ * @dmabuf:    [in]    buffer to prepare cpu access for.
+ * @start:     [in]    start of range for cpu access.
+ * @len:       [in]    length of range for cpu access.
+ * @direction: [in]    length of range for cpu access.
+ *
+ * Can return negative error values, returns 0 on success.
+ */
+int dma_buf_begin_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,
+                            enum dma_data_direction direction)
+{
+       int ret = 0;
+
+       if (WARN_ON(!dmabuf))
+               return -EINVAL;
+
+       if (dmabuf->ops->begin_cpu_access)
+               ret = dmabuf->ops->begin_cpu_access(dmabuf, start, len, direction);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(dma_buf_begin_cpu_access);
+
+/**
+ * dma_buf_end_cpu_access - Must be called after accessing a dma_buf from the
+ * cpu in the kernel context. Calls end_cpu_access to allow exporter-specific
+ * actions. Coherency is only guaranteed in the specified range for the
+ * specified access direction.
+ * @dmabuf:    [in]    buffer to complete cpu access for.
+ * @start:     [in]    start of range for cpu access.
+ * @len:       [in]    length of range for cpu access.
+ * @direction: [in]    length of range for cpu access.
+ *
+ * This call must always succeed.
+ */
+void dma_buf_end_cpu_access(struct dma_buf *dmabuf, size_t start, size_t len,
+                           enum dma_data_direction direction)
+{
+       WARN_ON(!dmabuf);
+
+       if (dmabuf->ops->end_cpu_access)
+               dmabuf->ops->end_cpu_access(dmabuf, start, len, direction);
+}
+EXPORT_SYMBOL_GPL(dma_buf_end_cpu_access);
+
+/**
+ * dma_buf_kmap_atomic - Map a page of the buffer object into kernel address
+ * space. The same restrictions as for kmap_atomic and friends apply.
+ * @dmabuf:    [in]    buffer to map page from.
+ * @page_num:  [in]    page in PAGE_SIZE units to map.
+ *
+ * This call must always succeed, any necessary preparations that might fail
+ * need to be done in begin_cpu_access.
+ */
+void *dma_buf_kmap_atomic(struct dma_buf *dmabuf, unsigned long page_num)
+{
+       WARN_ON(!dmabuf);
+
+       return dmabuf->ops->kmap_atomic(dmabuf, page_num);
+}
+EXPORT_SYMBOL_GPL(dma_buf_kmap_atomic);
+
+/**
+ * dma_buf_kunmap_atomic - Unmap a page obtained by dma_buf_kmap_atomic.
+ * @dmabuf:    [in]    buffer to unmap page from.
+ * @page_num:  [in]    page in PAGE_SIZE units to unmap.
+ * @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap_atomic.
+ *
+ * This call must always succeed.
+ */
+void dma_buf_kunmap_atomic(struct dma_buf *dmabuf, unsigned long page_num,
+                          void *vaddr)
+{
+       WARN_ON(!dmabuf);
+
+       if (dmabuf->ops->kunmap_atomic)
+               dmabuf->ops->kunmap_atomic(dmabuf, page_num, vaddr);
+}
+EXPORT_SYMBOL_GPL(dma_buf_kunmap_atomic);
+
+/**
+ * dma_buf_kmap - Map a page of the buffer object into kernel address space. The
+ * same restrictions as for kmap and friends apply.
+ * @dmabuf:    [in]    buffer to map page from.
+ * @page_num:  [in]    page in PAGE_SIZE units to map.
+ *
+ * This call must always succeed, any necessary preparations that might fail
+ * need to be done in begin_cpu_access.
+ */
+void *dma_buf_kmap(struct dma_buf *dmabuf, unsigned long page_num)
+{
+       WARN_ON(!dmabuf);
+
+       return dmabuf->ops->kmap(dmabuf, page_num);
+}
+EXPORT_SYMBOL_GPL(dma_buf_kmap);
+
+/**
+ * dma_buf_kunmap - Unmap a page obtained by dma_buf_kmap.
+ * @dmabuf:    [in]    buffer to unmap page from.
+ * @page_num:  [in]    page in PAGE_SIZE units to unmap.
+ * @vaddr:     [in]    kernel space pointer obtained from dma_buf_kmap.
+ *
+ * This call must always succeed.
+ */
+void dma_buf_kunmap(struct dma_buf *dmabuf, unsigned long page_num,
+                   void *vaddr)
+{
+       WARN_ON(!dmabuf);
+
+       if (dmabuf->ops->kunmap)
+               dmabuf->ops->kunmap(dmabuf, page_num, vaddr);
+}
+EXPORT_SYMBOL_GPL(dma_buf_kunmap);
+
+
+/**
+ * dma_buf_mmap - Setup up a userspace mmap with the given vma
+ * @dmabuf:    [in]    buffer that should back the vma
+ * @vma:       [in]    vma for the mmap
+ * @pgoff:     [in]    offset in pages where this mmap should start within the
+ *                     dma-buf buffer.
+ *
+ * This function adjusts the passed in vma so that it points at the file of the
+ * dma_buf operation. It also adjusts the starting pgoff and does bounds
+ * checking on the size of the vma. Then it calls the exporters mmap function to
+ * set up the mapping.
+ *
+ * Can return negative error values, returns 0 on success.
+ */
+int dma_buf_mmap(struct dma_buf *dmabuf, struct vm_area_struct *vma,
+                unsigned long pgoff)
+{
+       struct file *oldfile;
+       int ret;
+
+       if (WARN_ON(!dmabuf || !vma))
+               return -EINVAL;
+
+       /* check for offset overflow */
+       if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) < pgoff)
+               return -EOVERFLOW;
+
+       /* check for overflowing the buffer's size */
+       if (pgoff + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) >
+           dmabuf->size >> PAGE_SHIFT)
+               return -EINVAL;
+
+       /* readjust the vma */
+       get_file(dmabuf->file);
+       oldfile = vma->vm_file;
+       vma->vm_file = dmabuf->file;
+       vma->vm_pgoff = pgoff;
+
+       ret = dmabuf->ops->mmap(dmabuf, vma);
+       if (ret) {
+               /* restore old parameters on failure */
+               vma->vm_file = oldfile;
+               fput(dmabuf->file);
+       } else {
+               if (oldfile)
+                       fput(oldfile);
+       }
+       return ret;
+
+}
+EXPORT_SYMBOL_GPL(dma_buf_mmap);
+
+/**
+ * dma_buf_vmap - Create virtual mapping for the buffer object into kernel
+ * address space. Same restrictions as for vmap and friends apply.
+ * @dmabuf:    [in]    buffer to vmap
+ *
+ * This call may fail due to lack of virtual mapping address space.
+ * These calls are optional in drivers. The intended use for them
+ * is for mapping objects linear in kernel space for high use objects.
+ * Please attempt to use kmap/kunmap before thinking about these interfaces.
+ *
+ * Returns NULL on error.
+ */
+void *dma_buf_vmap(struct dma_buf *dmabuf)
+{
+       void *ptr;
+
+       if (WARN_ON(!dmabuf))
+               return NULL;
+
+       if (!dmabuf->ops->vmap)
+               return NULL;
+
+       mutex_lock(&dmabuf->lock);
+       if (dmabuf->vmapping_counter) {
+               dmabuf->vmapping_counter++;
+               BUG_ON(!dmabuf->vmap_ptr);
+               ptr = dmabuf->vmap_ptr;
+               goto out_unlock;
+       }
+
+       BUG_ON(dmabuf->vmap_ptr);
+
+       ptr = dmabuf->ops->vmap(dmabuf);
+       if (WARN_ON_ONCE(IS_ERR(ptr)))
+               ptr = NULL;
+       if (!ptr)
+               goto out_unlock;
+
+       dmabuf->vmap_ptr = ptr;
+       dmabuf->vmapping_counter = 1;
+
+out_unlock:
+       mutex_unlock(&dmabuf->lock);
+       return ptr;
+}
+EXPORT_SYMBOL_GPL(dma_buf_vmap);
+
+/**
+ * dma_buf_vunmap - Unmap a vmap obtained by dma_buf_vmap.
+ * @dmabuf:    [in]    buffer to vunmap
+ * @vaddr:     [in]    vmap to vunmap
+ */
+void dma_buf_vunmap(struct dma_buf *dmabuf, void *vaddr)
+{
+       if (WARN_ON(!dmabuf))
+               return;
+
+       BUG_ON(!dmabuf->vmap_ptr);
+       BUG_ON(dmabuf->vmapping_counter == 0);
+       BUG_ON(dmabuf->vmap_ptr != vaddr);
+
+       mutex_lock(&dmabuf->lock);
+       if (--dmabuf->vmapping_counter == 0) {
+               if (dmabuf->ops->vunmap)
+                       dmabuf->ops->vunmap(dmabuf, vaddr);
+               dmabuf->vmap_ptr = NULL;
+       }
+       mutex_unlock(&dmabuf->lock);
+}
+EXPORT_SYMBOL_GPL(dma_buf_vunmap);
+
+#ifdef CONFIG_DEBUG_FS
+static int dma_buf_describe(struct seq_file *s)
+{
+       int ret;
+       struct dma_buf *buf_obj;
+       struct dma_buf_attachment *attach_obj;
+       int count = 0, attach_count;
+       size_t size = 0;
+
+       ret = mutex_lock_interruptible(&db_list.lock);
+
+       if (ret)
+               return ret;
+
+       seq_puts(s, "\nDma-buf Objects:\n");
+       seq_puts(s, "size\tflags\tmode\tcount\texp_name\n");
+
+       list_for_each_entry(buf_obj, &db_list.head, list_node) {
+               ret = mutex_lock_interruptible(&buf_obj->lock);
+
+               if (ret) {
+                       seq_puts(s,
+                                "\tERROR locking buffer object: skipping\n");
+                       continue;
+               }
+
+               seq_printf(s, "%08zu\t%08x\t%08x\t%08ld\t%s\n",
+                               buf_obj->size,
+                               buf_obj->file->f_flags, buf_obj->file->f_mode,
+                               (long)(buf_obj->file->f_count.counter),
+                               buf_obj->exp_name);
+
+               seq_puts(s, "\tAttached Devices:\n");
+               attach_count = 0;
+
+               list_for_each_entry(attach_obj, &buf_obj->attachments, node) {
+                       seq_puts(s, "\t");
+
+                       seq_printf(s, "%s\n", dev_name(attach_obj->dev));
+                       attach_count++;
+               }
+
+               seq_printf(s, "Total %d devices attached\n\n",
+                               attach_count);
+
+               count++;
+               size += buf_obj->size;
+               mutex_unlock(&buf_obj->lock);
+       }
+
+       seq_printf(s, "\nTotal %d objects, %zu bytes\n", count, size);
+
+       mutex_unlock(&db_list.lock);
+       return 0;
+}
+
+static int dma_buf_show(struct seq_file *s, void *unused)
+{
+       void (*func)(struct seq_file *) = s->private;
+       func(s);
+       return 0;
+}
+
+static int dma_buf_debug_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, dma_buf_show, inode->i_private);
+}
+
+static const struct file_operations dma_buf_debug_fops = {
+       .open           = dma_buf_debug_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+static struct dentry *dma_buf_debugfs_dir;
+
+static int dma_buf_init_debugfs(void)
+{
+       int err = 0;
+       dma_buf_debugfs_dir = debugfs_create_dir("dma_buf", NULL);
+       if (IS_ERR(dma_buf_debugfs_dir)) {
+               err = PTR_ERR(dma_buf_debugfs_dir);
+               dma_buf_debugfs_dir = NULL;
+               return err;
+       }
+
+       err = dma_buf_debugfs_create_file("bufinfo", dma_buf_describe);
+
+       if (err)
+               pr_debug("dma_buf: debugfs: failed to create node bufinfo\n");
+
+       return err;
+}
+
+static void dma_buf_uninit_debugfs(void)
+{
+       if (dma_buf_debugfs_dir)
+               debugfs_remove_recursive(dma_buf_debugfs_dir);
+}
+
+int dma_buf_debugfs_create_file(const char *name,
+                               int (*write)(struct seq_file *))
+{
+       struct dentry *d;
+
+       d = debugfs_create_file(name, S_IRUGO, dma_buf_debugfs_dir,
+                       write, &dma_buf_debug_fops);
+
+       return PTR_ERR_OR_ZERO(d);
+}
+#else
+static inline int dma_buf_init_debugfs(void)
+{
+       return 0;
+}
+static inline void dma_buf_uninit_debugfs(void)
+{
+}
+#endif
+
+static int __init dma_buf_init(void)
+{
+       mutex_init(&db_list.lock);
+       INIT_LIST_HEAD(&db_list.head);
+       dma_buf_init_debugfs();
+       return 0;
+}
+subsys_initcall(dma_buf_init);
+
+static void __exit dma_buf_deinit(void)
+{
+       dma_buf_uninit_debugfs();
+}
+__exitcall(dma_buf_deinit);
diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
new file mode 100644 (file)
index 0000000..a73fbf3
--- /dev/null
@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2012-2013 Canonical Ltd
+ *
+ * Based on bo.c which bears the following copyright notice,
+ * but is dual licensed:
+ *
+ * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include <linux/reservation.h>
+#include <linux/export.h>
+
+DEFINE_WW_CLASS(reservation_ww_class);
+EXPORT_SYMBOL(reservation_ww_class);