/*
 * Copyright © 2008 Intel Corporation
 *
 * 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, sublicense,
 * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS 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:
 *    Eric Anholt <eric@anholt.net>
 *
 */

#include <drm/drmP.h>
#include <drm/i915_drm.h>
#include "i915_drv.h"
#include "i915_trace.h"
#include "intel_drv.h"
#include <linux/shmem_fs.h>
#include <linux/slab.h>
#include <linux/swap.h>
#include <linux/pci.h>
#include <linux/dma-buf.h>

static void i915_gem_object_flush_gtt_write_domain(struct drm_i915_gem_object *obj);
static void i915_gem_object_flush_cpu_write_domain(struct drm_i915_gem_object *obj);
static __must_check int i915_gem_object_bind_to_gtt(struct drm_i915_gem_object *obj,
                                                    unsigned alignment,
                                                    bool map_and_fenceable,
                                                    bool nonblocking);
static int i915_gem_phys_pwrite(struct drm_device *dev,
                                struct drm_i915_gem_object *obj,
                                struct drm_i915_gem_pwrite *args,
                                struct drm_file *file);

static void i915_gem_write_fence(struct drm_device *dev, int reg,
                                 struct drm_i915_gem_object *obj);
static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj,
                                         struct drm_i915_fence_reg *fence,
                                         bool enable);

static int i915_gem_inactive_shrink(struct shrinker *shrinker,
                                    struct shrink_control *sc);
static long i915_gem_purge(struct drm_i915_private *dev_priv, long target);
static void i915_gem_shrink_all(struct drm_i915_private *dev_priv);
static void i915_gem_object_truncate(struct drm_i915_gem_object *obj);

static inline void i915_gem_object_fence_lost(struct drm_i915_gem_object *obj)
{
        if (obj->tiling_mode)
                i915_gem_release_mmap(obj);

        /* As we do not have an associated fence register, we will force
         * a tiling change if we ever need to acquire one.
         */
        obj->fence_dirty = false;
        obj->fence_reg = I915_FENCE_REG_NONE;
}

/* some bookkeeping */
static void i915_gem_info_add_obj(struct drm_i915_private *dev_priv,
                                  size_t size)
{
        dev_priv->mm.object_count++;
        dev_priv->mm.object_memory += size;
}

static void i915_gem_info_remove_obj(struct drm_i915_private *dev_priv,
                                     size_t size)
{
        dev_priv->mm.object_count--;
        dev_priv->mm.object_memory -= size;
}

static int
i915_gem_wait_for_error(struct i915_gpu_error *error)
{
        int ret;

#define EXIT_COND (!i915_reset_in_progress(error) || \
                   i915_terminally_wedged(error))
        if (EXIT_COND)
                return 0;

        /*
         * Only wait 10 seconds for the gpu reset to complete to avoid hanging
         * userspace. If it takes that long something really bad is going on and
         * we should simply try to bail out and fail as gracefully as possible.
         */
        ret = wait_event_interruptible_timeout(error->reset_queue,
                                               EXIT_COND,
                                               10*HZ);
        if (ret == 0) {
                DRM_ERROR("Timed out waiting for the gpu reset to complete\n");
                return -EIO;
        } else if (ret < 0) {
                return ret;
        }
#undef EXIT_COND

        return 0;
}

int i915_mutex_lock_interruptible(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret;

        ret = i915_gem_wait_for_error(&dev_priv->gpu_error);
        if (ret)
                return ret;

        ret = mutex_lock_interruptible(&dev->struct_mutex);
        if (ret)
                return ret;

        WARN_ON(i915_verify_lists(dev));
        return 0;
}

static inline bool
i915_gem_object_is_inactive(struct drm_i915_gem_object *obj)
{
        return obj->gtt_space && !obj->active;
}

int
i915_gem_init_ioctl(struct drm_device *dev, void *data,
                    struct drm_file *file)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_init *args = data;

        if (drm_core_check_feature(dev, DRIVER_MODESET))
                return -ENODEV;

        if (args->gtt_start >= args->gtt_end ||
            (args->gtt_end | args->gtt_start) & (PAGE_SIZE - 1))
                return -EINVAL;

        /* GEM with user mode setting was never supported on ilk and later. */
        if (INTEL_INFO(dev)->gen >= 5)
                return -ENODEV;

        mutex_lock(&dev->struct_mutex);
        i915_gem_setup_global_gtt(dev, args->gtt_start, args->gtt_end,
                                  args->gtt_end);
        dev_priv->gtt.mappable_end = args->gtt_end;
        mutex_unlock(&dev->struct_mutex);

        return 0;
}

int
i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
                            struct drm_file *file)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_get_aperture *args = data;
        struct drm_i915_gem_object *obj;
        size_t pinned;

        pinned = 0;
        mutex_lock(&dev->struct_mutex);
        list_for_each_entry(obj, &dev_priv->mm.bound_list, gtt_list)
                if (obj->pin_count)
                        pinned += obj->gtt_space->size;
        mutex_unlock(&dev->struct_mutex);

        args->aper_size = dev_priv->gtt.total;
        args->aper_available_size = args->aper_size - pinned;

        return 0;
}

void *i915_gem_object_alloc(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        return kmem_cache_alloc(dev_priv->slab, GFP_KERNEL | __GFP_ZERO);
}

void i915_gem_object_free(struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
        kmem_cache_free(dev_priv->slab, obj);
}

static int
i915_gem_create(struct drm_file *file,
                struct drm_device *dev,
                uint64_t size,
                uint32_t *handle_p)
{
        struct drm_i915_gem_object *obj;
        int ret;
        u32 handle;

        size = roundup(size, PAGE_SIZE);
        if (size == 0)
                return -EINVAL;

        /* Allocate the new object */
        obj = i915_gem_alloc_object(dev, size);
        if (obj == NULL)
                return -ENOMEM;

        ret = drm_gem_handle_create(file, &obj->base, &handle);
        if (ret) {
                drm_gem_object_release(&obj->base);
                i915_gem_info_remove_obj(dev->dev_private, obj->base.size);
                i915_gem_object_free(obj);
                return ret;
        }

        /* drop reference from allocate - handle holds it now */
        drm_gem_object_unreference(&obj->base);
        trace_i915_gem_object_create(obj);

        *handle_p = handle;
        return 0;
}

int
i915_gem_dumb_create(struct drm_file *file,
                     struct drm_device *dev,
                     struct drm_mode_create_dumb *args)
{
        /* have to work out size/pitch and return them */
        args->pitch = ALIGN(args->width * ((args->bpp + 7) / 8), 64);
        args->size = args->pitch * args->height;
        return i915_gem_create(file, dev,
                               args->size, &args->handle);
}

int i915_gem_dumb_destroy(struct drm_file *file,
                          struct drm_device *dev,
                          uint32_t handle)
{
        return drm_gem_handle_delete(file, handle);
}

/**
 * Creates a new mm object and returns a handle to it.
 */
int
i915_gem_create_ioctl(struct drm_device *dev, void *data,
                      struct drm_file *file)
{
        struct drm_i915_gem_create *args = data;

        return i915_gem_create(file, dev,
                               args->size, &args->handle);
}

static inline int
__copy_to_user_swizzled(char __user *cpu_vaddr,
                        const char *gpu_vaddr, int gpu_offset,
                        int length)
{
        int ret, cpu_offset = 0;

        while (length > 0) {
                int cacheline_end = ALIGN(gpu_offset + 1, 64);
                int this_length = min(cacheline_end - gpu_offset, length);
                int swizzled_gpu_offset = gpu_offset ^ 64;

                ret = __copy_to_user(cpu_vaddr + cpu_offset,
                                     gpu_vaddr + swizzled_gpu_offset,
                                     this_length);
                if (ret)
                        return ret + length;

                cpu_offset += this_length;
                gpu_offset += this_length;
                length -= this_length;
        }

        return 0;
}

static inline int
__copy_from_user_swizzled(char *gpu_vaddr, int gpu_offset,
                          const char __user *cpu_vaddr,
                          int length)
{
        int ret, cpu_offset = 0;

        while (length > 0) {
                int cacheline_end = ALIGN(gpu_offset + 1, 64);
                int this_length = min(cacheline_end - gpu_offset, length);
                int swizzled_gpu_offset = gpu_offset ^ 64;

                ret = __copy_from_user(gpu_vaddr + swizzled_gpu_offset,
                                       cpu_vaddr + cpu_offset,
                                       this_length);
                if (ret)
                        return ret + length;

                cpu_offset += this_length;
                gpu_offset += this_length;
                length -= this_length;
        }

        return 0;
}

/* Per-page copy function for the shmem pread fastpath.
 * Flushes invalid cachelines before reading the target if
 * needs_clflush is set. */
static int
shmem_pread_fast(struct page *page, int shmem_page_offset, int page_length,
                 char __user *user_data,
                 bool page_do_bit17_swizzling, bool needs_clflush)
{
        char *vaddr;
        int ret;

        if (unlikely(page_do_bit17_swizzling))
                return -EINVAL;

        vaddr = kmap_atomic(page);
        if (needs_clflush)
                drm_clflush_virt_range(vaddr + shmem_page_offset,
                                       page_length);
        ret = __copy_to_user_inatomic(user_data,
                                      vaddr + shmem_page_offset,
                                      page_length);
        kunmap_atomic(vaddr);

        return ret ? -EFAULT : 0;
}

static void
shmem_clflush_swizzled_range(char *addr, unsigned long length,
                             bool swizzled)
{
        if (unlikely(swizzled)) {
                unsigned long start = (unsigned long) addr;
                unsigned long end = (unsigned long) addr + length;

                /* For swizzling simply ensure that we always flush both
                 * channels. Lame, but simple and it works. Swizzled
                 * pwrite/pread is far from a hotpath - current userspace
                 * doesn't use it at all. */
                start = round_down(start, 128);
                end = round_up(end, 128);

                drm_clflush_virt_range((void *)start, end - start);
        } else {
                drm_clflush_virt_range(addr, length);
        }

}

/* Only difference to the fast-path function is that this can handle bit17
 * and uses non-atomic copy and kmap functions. */
static int
shmem_pread_slow(struct page *page, int shmem_page_offset, int page_length,
                 char __user *user_data,
                 bool page_do_bit17_swizzling, bool needs_clflush)
{
        char *vaddr;
        int ret;

        vaddr = kmap(page);
        if (needs_clflush)
                shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
                                             page_length,
                                             page_do_bit17_swizzling);

        if (page_do_bit17_swizzling)
                ret = __copy_to_user_swizzled(user_data,
                                              vaddr, shmem_page_offset,
                                              page_length);
        else
                ret = __copy_to_user(user_data,
                                     vaddr + shmem_page_offset,
                                     page_length);
        kunmap(page);

        return ret ? - EFAULT : 0;
}

static int
i915_gem_shmem_pread(struct drm_device *dev,
                     struct drm_i915_gem_object *obj,
                     struct drm_i915_gem_pread *args,
                     struct drm_file *file)
{
        char __user *user_data;
        ssize_t remain;
        loff_t offset;
        int shmem_page_offset, page_length, ret = 0;
        int obj_do_bit17_swizzling, page_do_bit17_swizzling;
        int prefaulted = 0;
        int needs_clflush = 0;
        struct sg_page_iter sg_iter;

        user_data = to_user_ptr(args->data_ptr);
        remain = args->size;

        obj_do_bit17_swizzling = i915_gem_object_needs_bit17_swizzle(obj);

        if (!(obj->base.read_domains & I915_GEM_DOMAIN_CPU)) {
                /* If we're not in the cpu read domain, set ourself into the gtt
                 * read domain and manually flush cachelines (if required). This
                 * optimizes for the case when the gpu will dirty the data
                 * anyway again before the next pread happens. */
                if (obj->cache_level == I915_CACHE_NONE)
                        needs_clflush = 1;
                if (obj->gtt_space) {
                        ret = i915_gem_object_set_to_gtt_domain(obj, false);
                        if (ret)
                                return ret;
                }
        }

        ret = i915_gem_object_get_pages(obj);
        if (ret)
                return ret;

        i915_gem_object_pin_pages(obj);

        offset = args->offset;

        for_each_sg_page(obj->pages->sgl, &sg_iter, obj->pages->nents,
                         offset >> PAGE_SHIFT) {
                struct page *page = sg_page_iter_page(&sg_iter);

                if (remain <= 0)
                        break;

                /* Operation in this page
                 *
                 * shmem_page_offset = offset within page in shmem file
                 * page_length = bytes to copy for this page
                 */
                shmem_page_offset = offset_in_page(offset);
                page_length = remain;
                if ((shmem_page_offset + page_length) > PAGE_SIZE)
                        page_length = PAGE_SIZE - shmem_page_offset;

                page_do_bit17_swizzling = obj_do_bit17_swizzling &&
                        (page_to_phys(page) & (1 << 17)) != 0;

                ret = shmem_pread_fast(page, shmem_page_offset, page_length,
                                       user_data, page_do_bit17_swizzling,
                                       needs_clflush);
                if (ret == 0)
                        goto next_page;

                mutex_unlock(&dev->struct_mutex);

                if (!prefaulted) {
                        ret = fault_in_multipages_writeable(user_data, remain);
                        /* Userspace is tricking us, but we've already clobbered
                         * its pages with the prefault and promised to write the
                         * data up to the first fault. Hence ignore any errors
                         * and just continue. */
                        (void)ret;
                        prefaulted = 1;
                }

                ret = shmem_pread_slow(page, shmem_page_offset, page_length,
                                       user_data, page_do_bit17_swizzling,
                                       needs_clflush);

                mutex_lock(&dev->struct_mutex);

next_page:
                mark_page_accessed(page);

                if (ret)
                        goto out;

                remain -= page_length;
                user_data += page_length;
                offset += page_length;
        }

out:
        i915_gem_object_unpin_pages(obj);

        return ret;
}

/**
 * Reads data from the object referenced by handle.
 *
 * On error, the contents of *data are undefined.
 */
int
i915_gem_pread_ioctl(struct drm_device *dev, void *data,
                     struct drm_file *file)
{
        struct drm_i915_gem_pread *args = data;
        struct drm_i915_gem_object *obj;
        int ret = 0;

        if (args->size == 0)
                return 0;

        if (!access_ok(VERIFY_WRITE,
                       to_user_ptr(args->data_ptr),
                       args->size))
                return -EFAULT;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        /* Bounds check source.  */
        if (args->offset > obj->base.size ||
            args->size > obj->base.size - args->offset) {
                ret = -EINVAL;
                goto out;
        }

        /* prime objects have no backing filp to GEM pread/pwrite
         * pages from.
         */
        if (!obj->base.filp) {
                ret = -EINVAL;
                goto out;
        }

        trace_i915_gem_object_pread(obj, args->offset, args->size);

        ret = i915_gem_shmem_pread(dev, obj, args, file);

out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/* This is the fast write path which cannot handle
 * page faults in the source data
 */

static inline int
fast_user_write(struct io_mapping *mapping,
                loff_t page_base, int page_offset,
                char __user *user_data,
                int length)
{
        void __iomem *vaddr_atomic;
        void *vaddr;
        unsigned long unwritten;

        vaddr_atomic = io_mapping_map_atomic_wc(mapping, page_base);
        /* We can use the cpu mem copy function because this is X86. */
        vaddr = (void __force*)vaddr_atomic + page_offset;
        unwritten = __copy_from_user_inatomic_nocache(vaddr,
                                                      user_data, length);
        io_mapping_unmap_atomic(vaddr_atomic);
        return unwritten;
}

/**
 * This is the fast pwrite path, where we copy the data directly from the
 * user into the GTT, uncached.
 */
static int
i915_gem_gtt_pwrite_fast(struct drm_device *dev,
                         struct drm_i915_gem_object *obj,
                         struct drm_i915_gem_pwrite *args,
                         struct drm_file *file)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        ssize_t remain;
        loff_t offset, page_base;
        char __user *user_data;
        int page_offset, page_length, ret;

        ret = i915_gem_object_pin(obj, 0, true, true);
        if (ret)
                goto out;

        ret = i915_gem_object_set_to_gtt_domain(obj, true);
        if (ret)
                goto out_unpin;

        ret = i915_gem_object_put_fence(obj);
        if (ret)
                goto out_unpin;

        user_data = to_user_ptr(args->data_ptr);
        remain = args->size;

        offset = obj->gtt_offset + args->offset;

        while (remain > 0) {
                /* Operation in this page
                 *
                 * page_base = page offset within aperture
                 * page_offset = offset within page
                 * page_length = bytes to copy for this page
                 */
                page_base = offset & PAGE_MASK;
                page_offset = offset_in_page(offset);
                page_length = remain;
                if ((page_offset + remain) > PAGE_SIZE)
                        page_length = PAGE_SIZE - page_offset;

                /* If we get a fault while copying data, then (presumably) our
                 * source page isn't available.  Return the error and we'll
                 * retry in the slow path.
                 */
                if (fast_user_write(dev_priv->gtt.mappable, page_base,
                                    page_offset, user_data, page_length)) {
                        ret = -EFAULT;
                        goto out_unpin;
                }

                remain -= page_length;
                user_data += page_length;
                offset += page_length;
        }

out_unpin:
        i915_gem_object_unpin(obj);
out:
        return ret;
}

/* Per-page copy function for the shmem pwrite fastpath.
 * Flushes invalid cachelines before writing to the target if
 * needs_clflush_before is set and flushes out any written cachelines after
 * writing if needs_clflush is set. */
static int
shmem_pwrite_fast(struct page *page, int shmem_page_offset, int page_length,
                  char __user *user_data,
                  bool page_do_bit17_swizzling,
                  bool needs_clflush_before,
                  bool needs_clflush_after)
{
        char *vaddr;
        int ret;

        if (unlikely(page_do_bit17_swizzling))
                return -EINVAL;

        vaddr = kmap_atomic(page);
        if (needs_clflush_before)
                drm_clflush_virt_range(vaddr + shmem_page_offset,
                                       page_length);
        ret = __copy_from_user_inatomic_nocache(vaddr + shmem_page_offset,
                                                user_data,
                                                page_length);
        if (needs_clflush_after)
                drm_clflush_virt_range(vaddr + shmem_page_offset,
                                       page_length);
        kunmap_atomic(vaddr);

        return ret ? -EFAULT : 0;
}

/* Only difference to the fast-path function is that this can handle bit17
 * and uses non-atomic copy and kmap functions. */
static int
shmem_pwrite_slow(struct page *page, int shmem_page_offset, int page_length,
                  char __user *user_data,
                  bool page_do_bit17_swizzling,
                  bool needs_clflush_before,
                  bool needs_clflush_after)
{
        char *vaddr;
        int ret;

        vaddr = kmap(page);
        if (unlikely(needs_clflush_before || page_do_bit17_swizzling))
                shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
                                             page_length,
                                             page_do_bit17_swizzling);
        if (page_do_bit17_swizzling)
                ret = __copy_from_user_swizzled(vaddr, shmem_page_offset,
                                                user_data,
                                                page_length);
        else
                ret = __copy_from_user(vaddr + shmem_page_offset,
                                       user_data,
                                       page_length);
        if (needs_clflush_after)
                shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
                                             page_length,
                                             page_do_bit17_swizzling);
        kunmap(page);

        return ret ? -EFAULT : 0;
}

static int
i915_gem_shmem_pwrite(struct drm_device *dev,
                      struct drm_i915_gem_object *obj,
                      struct drm_i915_gem_pwrite *args,
                      struct drm_file *file)
{
        ssize_t remain;
        loff_t offset;
        char __user *user_data;
        int shmem_page_offset, page_length, ret = 0;
        int obj_do_bit17_swizzling, page_do_bit17_swizzling;
        int hit_slowpath = 0;
        int needs_clflush_after = 0;
        int needs_clflush_before = 0;
        struct sg_page_iter sg_iter;

        user_data = to_user_ptr(args->data_ptr);
        remain = args->size;

        obj_do_bit17_swizzling = i915_gem_object_needs_bit17_swizzle(obj);

        if (obj->base.write_domain != I915_GEM_DOMAIN_CPU) {
                /* If we're not in the cpu write domain, set ourself into the gtt
                 * write domain and manually flush cachelines (if required). This
                 * optimizes for the case when the gpu will use the data
                 * right away and we therefore have to clflush anyway. */
                if (obj->cache_level == I915_CACHE_NONE)
                        needs_clflush_after = 1;
                if (obj->gtt_space) {
                        ret = i915_gem_object_set_to_gtt_domain(obj, true);
                        if (ret)
                                return ret;
                }
        }
        /* Same trick applies for invalidate partially written cachelines before
         * writing.  */
        if (!(obj->base.read_domains & I915_GEM_DOMAIN_CPU)
            && obj->cache_level == I915_CACHE_NONE)
                needs_clflush_before = 1;

        ret = i915_gem_object_get_pages(obj);
        if (ret)
                return ret;

        i915_gem_object_pin_pages(obj);

        offset = args->offset;
        obj->dirty = 1;

        for_each_sg_page(obj->pages->sgl, &sg_iter, obj->pages->nents,
                         offset >> PAGE_SHIFT) {
                struct page *page = sg_page_iter_page(&sg_iter);
                int partial_cacheline_write;

                if (remain <= 0)
                        break;

                /* Operation in this page
                 *
                 * shmem_page_offset = offset within page in shmem file
                 * page_length = bytes to copy for this page
                 */
                shmem_page_offset = offset_in_page(offset);

                page_length = remain;
                if ((shmem_page_offset + page_length) > PAGE_SIZE)
                        page_length = PAGE_SIZE - shmem_page_offset;

                /* If we don't overwrite a cacheline completely we need to be
                 * careful to have up-to-date data by first clflushing. Don't
                 * overcomplicate things and flush the entire patch. */
                partial_cacheline_write = needs_clflush_before &&
                        ((shmem_page_offset | page_length)
                                & (boot_cpu_data.x86_clflush_size - 1));

                page_do_bit17_swizzling = obj_do_bit17_swizzling &&
                        (page_to_phys(page) & (1 << 17)) != 0;

                ret = shmem_pwrite_fast(page, shmem_page_offset, page_length,
                                        user_data, page_do_bit17_swizzling,
                                        partial_cacheline_write,
                                        needs_clflush_after);
                if (ret == 0)
                        goto next_page;

                hit_slowpath = 1;
                mutex_unlock(&dev->struct_mutex);
                ret = shmem_pwrite_slow(page, shmem_page_offset, page_length,
                                        user_data, page_do_bit17_swizzling,
                                        partial_cacheline_write,
                                        needs_clflush_after);

                mutex_lock(&dev->struct_mutex);

next_page:
                set_page_dirty(page);
                mark_page_accessed(page);

                if (ret)
                        goto out;

                remain -= page_length;
                user_data += page_length;
                offset += page_length;
        }

out:
        i915_gem_object_unpin_pages(obj);

        if (hit_slowpath) {
                /*
                 * Fixup: Flush cpu caches in case we didn't flush the dirty
                 * cachelines in-line while writing and the object moved
                 * out of the cpu write domain while we've dropped the lock.
                 */
                if (!needs_clflush_after &&
                    obj->base.write_domain != I915_GEM_DOMAIN_CPU) {
                        i915_gem_clflush_object(obj);
                        i915_gem_chipset_flush(dev);
                }
        }

        if (needs_clflush_after)
                i915_gem_chipset_flush(dev);

        return ret;
}

/**
 * Writes data to the object referenced by handle.
 *
 * On error, the contents of the buffer that were to be modified are undefined.
 */
int
i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
                      struct drm_file *file)
{
        struct drm_i915_gem_pwrite *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        if (args->size == 0)
                return 0;

        if (!access_ok(VERIFY_READ,
                       to_user_ptr(args->data_ptr),
                       args->size))
                return -EFAULT;

        ret = fault_in_multipages_readable(to_user_ptr(args->data_ptr),
                                           args->size);
        if (ret)
                return -EFAULT;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        /* Bounds check destination. */
        if (args->offset > obj->base.size ||
            args->size > obj->base.size - args->offset) {
                ret = -EINVAL;
                goto out;
        }

        /* prime objects have no backing filp to GEM pread/pwrite
         * pages from.
         */
        if (!obj->base.filp) {
                ret = -EINVAL;
                goto out;
        }

        trace_i915_gem_object_pwrite(obj, args->offset, args->size);

        ret = -EFAULT;
        /* We can only do the GTT pwrite on untiled buffers, as otherwise
         * it would end up going through the fenced access, and we'll get
         * different detiling behavior between reading and writing.
         * pread/pwrite currently are reading and writing from the CPU
         * perspective, requiring manual detiling by the client.
         */
        if (obj->phys_obj) {
                ret = i915_gem_phys_pwrite(dev, obj, args, file);
                goto out;
        }

        if (obj->cache_level == I915_CACHE_NONE &&
            obj->tiling_mode == I915_TILING_NONE &&
            obj->base.write_domain != I915_GEM_DOMAIN_CPU) {
                ret = i915_gem_gtt_pwrite_fast(dev, obj, args, file);
                /* Note that the gtt paths might fail with non-page-backed user
                 * pointers (e.g. gtt mappings when moving data between
                 * textures). Fallback to the shmem path in that case. */
        }

        if (ret == -EFAULT || ret == -ENOSPC)
                ret = i915_gem_shmem_pwrite(dev, obj, args, file);

out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

int
i915_gem_check_wedge(struct i915_gpu_error *error,
                     bool interruptible)
{
        if (i915_reset_in_progress(error)) {
                /* Non-interruptible callers can't handle -EAGAIN, hence return
                 * -EIO unconditionally for these. */
                if (!interruptible)
                        return -EIO;

                /* Recovery complete, but the reset failed ... */
                if (i915_terminally_wedged(error))
                        return -EIO;

                return -EAGAIN;
        }

        return 0;
}

/*
 * Compare seqno against outstanding lazy request. Emit a request if they are
 * equal.
 */
static int
i915_gem_check_olr(struct intel_ring_buffer *ring, u32 seqno)
{
        int ret;

        BUG_ON(!mutex_is_locked(&ring->dev->struct_mutex));

        ret = 0;
        if (seqno == ring->outstanding_lazy_request)
                ret = i915_add_request(ring, NULL, NULL);

        return ret;
}

/**
 * __wait_seqno - wait until execution of seqno has finished
 * @ring: the ring expected to report seqno
 * @seqno: duh!
 * @reset_counter: reset sequence associated with the given seqno
 * @interruptible: do an interruptible wait (normally yes)
 * @timeout: in - how long to wait (NULL forever); out - how much time remaining
 *
 * Note: It is of utmost importance that the passed in seqno and reset_counter
 * values have been read by the caller in an smp safe manner. Where read-side
 * locks are involved, it is sufficient to read the reset_counter before
 * unlocking the lock that protects the seqno. For lockless tricks, the
 * reset_counter _must_ be read before, and an appropriate smp_rmb must be
 * inserted.
 *
 * Returns 0 if the seqno was found within the alloted time. Else returns the
 * errno with remaining time filled in timeout argument.
 */
static int __wait_seqno(struct intel_ring_buffer *ring, u32 seqno,
                        unsigned reset_counter,
                        bool interruptible, struct timespec *timeout)
{
        drm_i915_private_t *dev_priv = ring->dev->dev_private;
        struct timespec before, now, wait_time={1,0};
        unsigned long timeout_jiffies;
        long end;
        bool wait_forever = true;
        int ret;

        if (i915_seqno_passed(ring->get_seqno(ring, true), seqno))
                return 0;

        trace_i915_gem_request_wait_begin(ring, seqno);

        if (timeout != NULL) {
                wait_time = *timeout;
                wait_forever = false;

        }

        timeout_jiffies = timespec_to_jiffies_timeout(&wait_time);

        if (WARN_ON(!ring->irq_get(ring)))
                return -ENODEV;

        /* Record current time in case interrupted by signal, or wedged * */
        getrawmonotonic(&before);

#define EXIT_COND \
        (i915_seqno_passed(ring->get_seqno(ring, false), seqno) || \
         i915_reset_in_progress(&dev_priv->gpu_error) || \
         reset_counter != atomic_read(&dev_priv->gpu_error.reset_counter))
        do {
                if (interruptible)
                        end = wait_event_interruptible_timeout(ring->irq_queue,
                                                               EXIT_COND,
                                                               timeout_jiffies);
                else
                        end = wait_event_timeout(ring->irq_queue, EXIT_COND,
                                                 timeout_jiffies);

                /* We need to check whether any gpu reset happened in between
                 * the caller grabbing the seqno and now ... */
                if (reset_counter != atomic_read(&dev_priv->gpu_error.reset_counter))
                        end = -EAGAIN;

                /* ... but upgrade the -EGAIN to an -EIO if the gpu is truely
                 * gone. */
                ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
                if (ret)
                        end = ret;
        } while (end == 0 && wait_forever);

        getrawmonotonic(&now);

        ring->irq_put(ring);
        trace_i915_gem_request_wait_end(ring, seqno);
#undef EXIT_COND

        if (timeout) {
                struct timespec sleep_time = timespec_sub(now, before);
                *timeout = timespec_sub(*timeout, sleep_time);
                if (!timespec_valid(timeout)) /* i.e. negative time remains */
                        set_normalized_timespec(timeout, 0, 0);
        }

        switch (end) {
        case -EIO:
        case -EAGAIN: /* Wedged */
        case -ERESTARTSYS: /* Signal */
                return (int)end;
        case 0: /* Timeout */
                return -ETIME;
        default: /* Completed */
                WARN_ON(end < 0); /* We're not aware of other errors */
                return 0;
        }
}

/**
 * Waits for a sequence number to be signaled, and cleans up the
 * request and object lists appropriately for that event.
 */
int
i915_wait_seqno(struct intel_ring_buffer *ring, uint32_t seqno)
{
        struct drm_device *dev = ring->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        bool interruptible = dev_priv->mm.interruptible;
        int ret;

        BUG_ON(!mutex_is_locked(&dev->struct_mutex));
        BUG_ON(seqno == 0);

        ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
        if (ret)
                return ret;

        ret = i915_gem_check_olr(ring, seqno);
        if (ret)
                return ret;

        return __wait_seqno(ring, seqno,
                            atomic_read(&dev_priv->gpu_error.reset_counter),
                            interruptible, NULL);
}

/**
 * Ensures that all rendering to the object has completed and the object is
 * safe to unbind from the GTT or access from the CPU.
 */
static __must_check int
i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
                               bool readonly)
{
        struct intel_ring_buffer *ring = obj->ring;
        u32 seqno;
        int ret;

        seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno;
        if (seqno == 0)
                return 0;

        ret = i915_wait_seqno(ring, seqno);
        if (ret)
                return ret;

        i915_gem_retire_requests_ring(ring);

        /* Manually manage the write flush as we may have not yet
         * retired the buffer.
         */
        if (obj->last_write_seqno &&
            i915_seqno_passed(seqno, obj->last_write_seqno)) {
                obj->last_write_seqno = 0;
                obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS;
        }

        return 0;
}

/* A nonblocking variant of the above wait. This is a highly dangerous routine
 * as the object state may change during this call.
 */
static __must_check int
i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
                                            bool readonly)
{
        struct drm_device *dev = obj->base.dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_ring_buffer *ring = obj->ring;
        unsigned reset_counter;
        u32 seqno;
        int ret;

        BUG_ON(!mutex_is_locked(&dev->struct_mutex));
        BUG_ON(!dev_priv->mm.interruptible);

        seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno;
        if (seqno == 0)
                return 0;

        ret = i915_gem_check_wedge(&dev_priv->gpu_error, true);
        if (ret)
                return ret;

        ret = i915_gem_check_olr(ring, seqno);
        if (ret)
                return ret;

        reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
        mutex_unlock(&dev->struct_mutex);
        ret = __wait_seqno(ring, seqno, reset_counter, true, NULL);
        mutex_lock(&dev->struct_mutex);

        i915_gem_retire_requests_ring(ring);

        /* Manually manage the write flush as we may have not yet
         * retired the buffer.
         */
        if (obj->last_write_seqno &&
            i915_seqno_passed(seqno, obj->last_write_seqno)) {
                obj->last_write_seqno = 0;
                obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS;
        }

        return ret;
}

/**
 * Called when user space prepares to use an object with the CPU, either
 * through the mmap ioctl's mapping or a GTT mapping.
 */
int
i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
                          struct drm_file *file)
{
        struct drm_i915_gem_set_domain *args = data;
        struct drm_i915_gem_object *obj;
        uint32_t read_domains = args->read_domains;
        uint32_t write_domain = args->write_domain;
        int ret;

        /* Only handle setting domains to types used by the CPU. */
        if (write_domain & I915_GEM_GPU_DOMAINS)
                return -EINVAL;

        if (read_domains & I915_GEM_GPU_DOMAINS)
                return -EINVAL;

        /* Having something in the write domain implies it's in the read
         * domain, and only that read domain.  Enforce that in the request.
         */
        if (write_domain != 0 && read_domains != write_domain)
                return -EINVAL;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        /* Try to flush the object off the GPU without holding the lock.
         * We will repeat the flush holding the lock in the normal manner
         * to catch cases where we are gazumped.
         */
        ret = i915_gem_object_wait_rendering__nonblocking(obj, !write_domain);
        if (ret)
                goto unref;

        if (read_domains & I915_GEM_DOMAIN_GTT) {
                ret = i915_gem_object_set_to_gtt_domain(obj, write_domain != 0);

                /* Silently promote "you're not bound, there was nothing to do"
                 * to success, since the client was just asking us to
                 * make sure everything was done.
                 */
                if (ret == -EINVAL)
                        ret = 0;
        } else {
                ret = i915_gem_object_set_to_cpu_domain(obj, write_domain != 0);
        }

unref:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/**
 * Called when user space has done writes to this buffer
 */
int
i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
                         struct drm_file *file)
{
        struct drm_i915_gem_sw_finish *args = data;
        struct drm_i915_gem_object *obj;
        int ret = 0;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        /* Pinned buffers may be scanout, so flush the cache */
        if (obj->pin_count)
                i915_gem_object_flush_cpu_write_domain(obj);

        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/**
 * Maps the contents of an object, returning the address it is mapped
 * into.
 *
 * While the mapping holds a reference on the contents of the object, it doesn't
 * imply a ref on the object itself.
 */
int
i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
                    struct drm_file *file)
{
        struct drm_i915_gem_mmap *args = data;
        struct drm_gem_object *obj;
        unsigned long addr;

        obj = drm_gem_object_lookup(dev, file, args->handle);
        if (obj == NULL)
                return -ENOENT;

        /* prime objects have no backing filp to GEM mmap
         * pages from.
         */
        if (!obj->filp) {
                drm_gem_object_unreference_unlocked(obj);
                return -EINVAL;
        }

        addr = vm_mmap(obj->filp, 0, args->size,
                       PROT_READ | PROT_WRITE, MAP_SHARED,
                       args->offset);
        drm_gem_object_unreference_unlocked(obj);
        if (IS_ERR((void *)addr))
                return addr;

        args->addr_ptr = (uint64_t) addr;

        return 0;
}

/**
 * i915_gem_fault - fault a page into the GTT
 * vma: VMA in question
 * vmf: fault info
 *
 * The fault handler is set up by drm_gem_mmap() when a object is GTT mapped
 * from userspace.  The fault handler takes care of binding the object to
 * the GTT (if needed), allocating and programming a fence register (again,
 * only if needed based on whether the old reg is still valid or the object
 * is tiled) and inserting a new PTE into the faulting process.
 *
 * Note that the faulting process may involve evicting existing objects
 * from the GTT and/or fence registers to make room.  So performance may
 * suffer if the GTT working set is large or there are few fence registers
 * left.
 */
int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
        struct drm_i915_gem_object *obj = to_intel_bo(vma->vm_private_data);
        struct drm_device *dev = obj->base.dev;
        drm_i915_private_t *dev_priv = dev->dev_private;
        pgoff_t page_offset;
        unsigned long pfn;
        int ret = 0;
        bool write = !!(vmf->flags & FAULT_FLAG_WRITE);

        /* We don't use vmf->pgoff since that has the fake offset */
        page_offset = ((unsigned long)vmf->virtual_address - vma->vm_start) >>
                PAGE_SHIFT;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                goto out;

        trace_i915_gem_object_fault(obj, page_offset, true, write);

        /* Access to snoopable pages through the GTT is incoherent. */
        if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(dev)) {
                ret = -EINVAL;
                goto unlock;
        }

        /* Now bind it into the GTT if needed */
        ret = i915_gem_object_pin(obj, 0, true, false);
        if (ret)
                goto unlock;

        ret = i915_gem_object_set_to_gtt_domain(obj, write);
        if (ret)
                goto unpin;

        ret = i915_gem_object_get_fence(obj);
        if (ret)
                goto unpin;

        obj->fault_mappable = true;

        pfn = ((dev_priv->gtt.mappable_base + obj->gtt_offset) >> PAGE_SHIFT) +
                page_offset;

        /* Finally, remap it using the new GTT offset */
        ret = vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);
unpin:
        i915_gem_object_unpin(obj);
unlock:
        mutex_unlock(&dev->struct_mutex);
out:
        switch (ret) {
        case -EIO:
                /* If this -EIO is due to a gpu hang, give the reset code a
                 * chance to clean up the mess. Otherwise return the proper
                 * SIGBUS. */
                if (i915_terminally_wedged(&dev_priv->gpu_error))
                        return VM_FAULT_SIGBUS;
        case -EAGAIN:
                /* Give the error handler a chance to run and move the
                 * objects off the GPU active list. Next time we service the
                 * fault, we should be able to transition the page into the
                 * GTT without touching the GPU (and so avoid further
                 * EIO/EGAIN). If the GPU is wedged, then there is no issue
                 * with coherency, just lost writes.
                 */
                set_need_resched();
        case 0:
        case -ERESTARTSYS:
        case -EINTR:
        case -EBUSY:
                /*
                 * EBUSY is ok: this just means that another thread
                 * already did the job.
                 */
                return VM_FAULT_NOPAGE;
        case -ENOMEM:
                return VM_FAULT_OOM;
        case -ENOSPC:
                return VM_FAULT_SIGBUS;
        default:
                WARN_ONCE(ret, "unhandled error in i915_gem_fault: %i\n", ret);
                return VM_FAULT_SIGBUS;
        }
}

/**
 * i915_gem_release_mmap - remove physical page mappings
 * @obj: obj in question
 *
 * Preserve the reservation of the mmapping with the DRM core code, but
 * relinquish ownership of the pages back to the system.
 *
 * It is vital that we remove the page mapping if we have mapped a tiled
 * object through the GTT and then lose the fence register due to
 * resource pressure. Similarly if the object has been moved out of the
 * aperture, than pages mapped into userspace must be revoked. Removing the
 * mapping will then trigger a page fault on the next user access, allowing
 * fixup by i915_gem_fault().
 */
void
i915_gem_release_mmap(struct drm_i915_gem_object *obj)
{
        if (!obj->fault_mappable)
                return;

        if (obj->base.dev->dev_mapping)
                unmap_mapping_range(obj->base.dev->dev_mapping,
                                    (loff_t)obj->base.map_list.hash.key<<PAGE_SHIFT,
                                    obj->base.size, 1);

        obj->fault_mappable = false;
}

uint32_t
i915_gem_get_gtt_size(struct drm_device *dev, uint32_t size, int tiling_mode)
{
        uint32_t gtt_size;

        if (INTEL_INFO(dev)->gen >= 4 ||
            tiling_mode == I915_TILING_NONE)
                return size;

        /* Previous chips need a power-of-two fence region when tiling */
        if (INTEL_INFO(dev)->gen == 3)
                gtt_size = 1024*1024;
        else
                gtt_size = 512*1024;

        while (gtt_size < size)
                gtt_size <<= 1;

        return gtt_size;
}

/**
 * i915_gem_get_gtt_alignment - return required GTT alignment for an object
 * @obj: object to check
 *
 * Return the required GTT alignment for an object, taking into account
 * potential fence register mapping.
 */
uint32_t
i915_gem_get_gtt_alignment(struct drm_device *dev, uint32_t size,
                           int tiling_mode, bool fenced)
{
        /*
         * Minimum alignment is 4k (GTT page size), but might be greater
         * if a fence register is needed for the object.
         */
        if (INTEL_INFO(dev)->gen >= 4 || (!fenced && IS_G33(dev)) ||
            tiling_mode == I915_TILING_NONE)
                return 4096;

        /*
         * Previous chips need to be aligned to the size of the smallest
         * fence register that can contain the object.
         */
        return i915_gem_get_gtt_size(dev, size, tiling_mode);
}

static int i915_gem_object_create_mmap_offset(struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
        int ret;

        if (obj->base.map_list.map)
                return 0;

        dev_priv->mm.shrinker_no_lock_stealing = true;

        ret = drm_gem_create_mmap_offset(&obj->base);
        if (ret != -ENOSPC)
                goto out;

        /* Badly fragmented mmap space? The only way we can recover
         * space is by destroying unwanted objects. We can't randomly release
         * mmap_offsets as userspace expects them to be persistent for the
         * lifetime of the objects. The closest we can is to release the
         * offsets on purgeable objects by truncating it and marking it purged,
         * which prevents userspace from ever using that object again.
         */
        i915_gem_purge(dev_priv, obj->base.size >> PAGE_SHIFT);
        ret = drm_gem_create_mmap_offset(&obj->base);
        if (ret != -ENOSPC)
                goto out;

        i915_gem_shrink_all(dev_priv);
        ret = drm_gem_create_mmap_offset(&obj->base);
out:
        dev_priv->mm.shrinker_no_lock_stealing = false;

        return ret;
}

static void i915_gem_object_free_mmap_offset(struct drm_i915_gem_object *obj)
{
        if (!obj->base.map_list.map)
                return;

        drm_gem_free_mmap_offset(&obj->base);
}

int
i915_gem_mmap_gtt(struct drm_file *file,
                  struct drm_device *dev,
                  uint32_t handle,
                  uint64_t *offset)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_object *obj;
        int ret;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        if (obj->base.size > dev_priv->gtt.mappable_end) {
                ret = -E2BIG;
                goto out;
        }

        if (obj->madv != I915_MADV_WILLNEED) {
                DRM_ERROR("Attempting to mmap a purgeable buffer\n");
                ret = -EINVAL;
                goto out;
        }

        ret = i915_gem_object_create_mmap_offset(obj);
        if (ret)
                goto out;

        *offset = (u64)obj->base.map_list.hash.key << PAGE_SHIFT;

out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/**
 * i915_gem_mmap_gtt_ioctl - prepare an object for GTT mmap'ing
 * @dev: DRM device
 * @data: GTT mapping ioctl data
 * @file: GEM object info
 *
 * Simply returns the fake offset to userspace so it can mmap it.
 * The mmap call will end up in drm_gem_mmap(), which will set things
 * up so we can get faults in the handler above.
 *
 * The fault handler will take care of binding the object into the GTT
 * (since it may have been evicted to make room for something), allocating
 * a fence register, and mapping the appropriate aperture address into
 * userspace.
 */
int
i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
                        struct drm_file *file)
{
        struct drm_i915_gem_mmap_gtt *args = data;

        return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset);
}

/* Immediately discard the backing storage */
static void
i915_gem_object_truncate(struct drm_i915_gem_object *obj)
{
        struct inode *inode;

        i915_gem_object_free_mmap_offset(obj);

        if (obj->base.filp == NULL)
                return;

        /* Our goal here is to return as much of the memory as
         * is possible back to the system as we are called from OOM.
         * To do this we must instruct the shmfs to drop all of its
         * backing pages, *now*.
         */
        inode = file_inode(obj->base.filp);
        shmem_truncate_range(inode, 0, (loff_t)-1);

        obj->madv = __I915_MADV_PURGED;
}

static inline int
i915_gem_object_is_purgeable(struct drm_i915_gem_object *obj)
{
        return obj->madv == I915_MADV_DONTNEED;
}

static void
i915_gem_object_put_pages_gtt(struct drm_i915_gem_object *obj)
{
        struct sg_page_iter sg_iter;
        int ret;

        BUG_ON(obj->madv == __I915_MADV_PURGED);

        ret = i915_gem_object_set_to_cpu_domain(obj, true);
        if (ret) {
                /* In the event of a disaster, abandon all caches and
                 * hope for the best.
                 */
                WARN_ON(ret != -EIO);
                i915_gem_clflush_object(obj);
                obj->base.read_domains = obj->base.write_domain = I915_GEM_DOMAIN_CPU;
        }

        if (i915_gem_object_needs_bit17_swizzle(obj))
                i915_gem_object_save_bit_17_swizzle(obj);

        if (obj->madv == I915_MADV_DONTNEED)
                obj->dirty = 0;

        for_each_sg_page(obj->pages->sgl, &sg_iter, obj->pages->nents, 0) {
                struct page *page = sg_page_iter_page(&sg_iter);

                if (obj->dirty)
                        set_page_dirty(page);

                if (obj->madv == I915_MADV_WILLNEED)
                        mark_page_accessed(page);

                page_cache_release(page);
        }
        obj->dirty = 0;

        sg_free_table(obj->pages);
        kfree(obj->pages);
}

int
i915_gem_object_put_pages(struct drm_i915_gem_object *obj)
{
        const struct drm_i915_gem_object_ops *ops = obj->ops;

        if (obj->pages == NULL)
                return 0;

        BUG_ON(obj->gtt_space);

        if (obj->pages_pin_count)
                return -EBUSY;

        /* ->put_pages might need to allocate memory for the bit17 swizzle
         * array, hence protect them from being reaped by removing them from gtt
         * lists early. */
        list_del(&obj->gtt_list);

        ops->put_pages(obj);
        obj->pages = NULL;

        if (i915_gem_object_is_purgeable(obj))
                i915_gem_object_truncate(obj);

        return 0;
}

static long
__i915_gem_shrink(struct drm_i915_private *dev_priv, long target,
                  bool purgeable_only)
{
        struct drm_i915_gem_object *obj, *next;
        long count = 0;

        list_for_each_entry_safe(obj, next,
                                 &dev_priv->mm.unbound_list,
                                 gtt_list) {
                if ((i915_gem_object_is_purgeable(obj) || !purgeable_only) &&
                    i915_gem_object_put_pages(obj) == 0) {
                        count += obj->base.size >> PAGE_SHIFT;
                        if (count >= target)
                                return count;
                }
        }

        list_for_each_entry_safe(obj, next,
                                 &dev_priv->mm.inactive_list,
                                 mm_list) {
                if ((i915_gem_object_is_purgeable(obj) || !purgeable_only) &&
                    i915_gem_object_unbind(obj) == 0 &&
                    i915_gem_object_put_pages(obj) == 0) {
                        count += obj->base.size >> PAGE_SHIFT;
                        if (count >= target)
                                return count;
                }
        }

        return count;
}

static long
i915_gem_purge(struct drm_i915_private *dev_priv, long target)
{
        return __i915_gem_shrink(dev_priv, target, true);
}

static void
i915_gem_shrink_all(struct drm_i915_private *dev_priv)
{
        struct drm_i915_gem_object *obj, *next;

        i915_gem_evict_everything(dev_priv->dev);

        list_for_each_entry_safe(obj, next, &dev_priv->mm.unbound_list, gtt_list)
                i915_gem_object_put_pages(obj);
}

static int
i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
        int page_count, i;
        struct address_space *mapping;
        struct sg_table *st;
        struct scatterlist *sg;
        struct sg_page_iter sg_iter;
        struct page *page;
        unsigned long last_pfn = 0;     /* suppress gcc warning */
        gfp_t gfp;

        /* Assert that the object is not currently in any GPU domain. As it
         * wasn't in the GTT, there shouldn't be any way it could have been in
         * a GPU cache
         */
        BUG_ON(obj->base.read_domains & I915_GEM_GPU_DOMAINS);
        BUG_ON(obj->base.write_domain & I915_GEM_GPU_DOMAINS);

        st = kmalloc(sizeof(*st), GFP_KERNEL);
        if (st == NULL)
                return -ENOMEM;

        page_count = obj->base.size / PAGE_SIZE;
        if (sg_alloc_table(st, page_count, GFP_KERNEL)) {
                sg_free_table(st);
                kfree(st);
                return -ENOMEM;
        }

        /* Get the list of pages out of our struct file.  They'll be pinned
         * at this point until we release them.
         *
         * Fail silently without starting the shrinker
         */
        mapping = file_inode(obj->base.filp)->i_mapping;
        gfp = mapping_gfp_mask(mapping);
        gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
        gfp &= ~(__GFP_IO | __GFP_WAIT);
        sg = st->sgl;
        st->nents = 0;
        for (i = 0; i < page_count; i++) {
                page = shmem_read_mapping_page_gfp(mapping, i, gfp);
                if (IS_ERR(page)) {
                        i915_gem_purge(dev_priv, page_count);
                        page = shmem_read_mapping_page_gfp(mapping, i, gfp);
                }
                if (IS_ERR(page)) {
                        /* We've tried hard to allocate the memory by reaping
                         * our own buffer, now let the real VM do its job and
                         * go down in flames if truly OOM.
                         */
                        gfp &= ~(__GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD);
                        gfp |= __GFP_IO | __GFP_WAIT;

                        i915_gem_shrink_all(dev_priv);
                        page = shmem_read_mapping_page_gfp(mapping, i, gfp);
                        if (IS_ERR(page))
                                goto err_pages;

                        gfp |= __GFP_NORETRY | __GFP_NOWARN | __GFP_NO_KSWAPD;
                        gfp &= ~(__GFP_IO | __GFP_WAIT);
                }
#ifdef CONFIG_SWIOTLB
                if (swiotlb_nr_tbl()) {
                        st->nents++;
                        sg_set_page(sg, page, PAGE_SIZE, 0);
                        sg = sg_next(sg);
                        continue;
                }
#endif
                if (!i || page_to_pfn(page) != last_pfn + 1) {
                        if (i)
                                sg = sg_next(sg);
                        st->nents++;
                        sg_set_page(sg, page, PAGE_SIZE, 0);
                } else {
                        sg->length += PAGE_SIZE;
                }
                last_pfn = page_to_pfn(page);
        }
#ifdef CONFIG_SWIOTLB
        if (!swiotlb_nr_tbl())
#endif
                sg_mark_end(sg);
        obj->pages = st;

        if (i915_gem_object_needs_bit17_swizzle(obj))
                i915_gem_object_do_bit_17_swizzle(obj);

        return 0;

err_pages:
        sg_mark_end(sg);
        for_each_sg_page(st->sgl, &sg_iter, st->nents, 0)
                page_cache_release(sg_page_iter_page(&sg_iter));
        sg_free_table(st);
        kfree(st);
        return PTR_ERR(page);
}

/* Ensure that the associated pages are gathered from the backing storage
 * and pinned into our object. i915_gem_object_get_pages() may be called
 * multiple times before they are released by a single call to
 * i915_gem_object_put_pages() - once the pages are no longer referenced
 * either as a result of memory pressure (reaping pages under the shrinker)
 * or as the object is itself released.
 */
int
i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
        const struct drm_i915_gem_object_ops *ops = obj->ops;
        int ret;

        if (obj->pages)
                return 0;

        if (obj->madv != I915_MADV_WILLNEED) {
                DRM_ERROR("Attempting to obtain a purgeable object\n");
                return -EINVAL;
        }

        BUG_ON(obj->pages_pin_count);

        ret = ops->get_pages(obj);
        if (ret)
                return ret;

        list_add_tail(&obj->gtt_list, &dev_priv->mm.unbound_list);
        return 0;
}

void
i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
                               struct intel_ring_buffer *ring)
{
        struct drm_device *dev = obj->base.dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        u32 seqno = intel_ring_get_seqno(ring);

        BUG_ON(ring == NULL);
        obj->ring = ring;

        /* Add a reference if we're newly entering the active list. */
        if (!obj->active) {
                drm_gem_object_reference(&obj->base);
                obj->active = 1;
        }

        /* Move from whatever list we were on to the tail of execution. */
        list_move_tail(&obj->mm_list, &dev_priv->mm.active_list);
        list_move_tail(&obj->ring_list, &ring->active_list);

        obj->last_read_seqno = seqno;

        if (obj->fenced_gpu_access) {
                obj->last_fenced_seqno = seqno;

                /* Bump MRU to take account of the delayed flush */
                if (obj->fence_reg != I915_FENCE_REG_NONE) {
                        struct drm_i915_fence_reg *reg;

                        reg = &dev_priv->fence_regs[obj->fence_reg];
                        list_move_tail(&reg->lru_list,
                                       &dev_priv->mm.fence_list);
                }
        }
}

static void
i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
{
        struct drm_device *dev = obj->base.dev;
        struct drm_i915_private *dev_priv = dev->dev_private;

        BUG_ON(obj->base.write_domain & ~I915_GEM_GPU_DOMAINS);
        BUG_ON(!obj->active);

        list_move_tail(&obj->mm_list, &dev_priv->mm.inactive_list);

        list_del_init(&obj->ring_list);
        obj->ring = NULL;

        obj->last_read_seqno = 0;
        obj->last_write_seqno = 0;
        obj->base.write_domain = 0;

        obj->last_fenced_seqno = 0;
        obj->fenced_gpu_access = false;

        obj->active = 0;
        drm_gem_object_unreference(&obj->base);

        WARN_ON(i915_verify_lists(dev));
}

static int
i915_gem_init_seqno(struct drm_device *dev, u32 seqno)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_ring_buffer *ring;
        int ret, i, j;

        /* Carefully retire all requests without writing to the rings */
        for_each_ring(ring, dev_priv, i) {
                ret = intel_ring_idle(ring);
                if (ret)
                        return ret;
        }
        i915_gem_retire_requests(dev);

        /* Finally reset hw state */
        for_each_ring(ring, dev_priv, i) {
                intel_ring_init_seqno(ring, seqno);

                for (j = 0; j < ARRAY_SIZE(ring->sync_seqno); j++)
                        ring->sync_seqno[j] = 0;
        }

        return 0;
}

int i915_gem_set_seqno(struct drm_device *dev, u32 seqno)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret;

        if (seqno == 0)
                return -EINVAL;

        /* HWS page needs to be set less than what we
         * will inject to ring
         */
        ret = i915_gem_init_seqno(dev, seqno - 1);
        if (ret)
                return ret;

        /* Carefully set the last_seqno value so that wrap
         * detection still works
         */
        dev_priv->next_seqno = seqno;
        dev_priv->last_seqno = seqno - 1;
        if (dev_priv->last_seqno == 0)
                dev_priv->last_seqno--;

        return 0;
}

int
i915_gem_get_seqno(struct drm_device *dev, u32 *seqno)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        /* reserve 0 for non-seqno */
        if (dev_priv->next_seqno == 0) {
                int ret = i915_gem_init_seqno(dev, 0);
                if (ret)
                        return ret;

                dev_priv->next_seqno = 1;
        }

        *seqno = dev_priv->last_seqno = dev_priv->next_seqno++;
        return 0;
}

int
i915_add_request(struct intel_ring_buffer *ring,
                 struct drm_file *file,
                 u32 *out_seqno)
{
        drm_i915_private_t *dev_priv = ring->dev->dev_private;
        struct drm_i915_gem_request *request;
        u32 request_ring_position;
        int was_empty;
        int ret;

        /*
         * Emit any outstanding flushes - execbuf can fail to emit the flush
         * after having emitted the batchbuffer command. Hence we need to fix
         * things up similar to emitting the lazy request. The difference here
         * is that the flush _must_ happen before the next request, no matter
         * what.
         */
        ret = intel_ring_flush_all_caches(ring);
        if (ret)
                return ret;

        request = kmalloc(sizeof(*request), GFP_KERNEL);
        if (request == NULL)
                return -ENOMEM;


        /* Record the position of the start of the request so that
         * should we detect the updated seqno part-way through the
         * GPU processing the request, we never over-estimate the
         * position of the head.
         */
        request_ring_position = intel_ring_get_tail(ring);

        ret = ring->add_request(ring);
        if (ret) {
                kfree(request);
                return ret;
        }

        request->seqno = intel_ring_get_seqno(ring);
        request->ring = ring;
        request->tail = request_ring_position;
        request->emitted_jiffies = jiffies;
        was_empty = list_empty(&ring->request_list);
        list_add_tail(&request->list, &ring->request_list);
        request->file_priv = NULL;

        if (file) {
                struct drm_i915_file_private *file_priv = file->driver_priv;

                spin_lock(&file_priv->mm.lock);
                request->file_priv = file_priv;
                list_add_tail(&request->client_list,
                              &file_priv->mm.request_list);
                spin_unlock(&file_priv->mm.lock);
        }

        trace_i915_gem_request_add(ring, request->seqno);
        ring->outstanding_lazy_request = 0;

        if (!dev_priv->mm.suspended) {
                if (i915_enable_hangcheck) {
                        mod_timer(&dev_priv->gpu_error.hangcheck_timer,
                                  round_jiffies_up(jiffies + DRM_I915_HANGCHECK_JIFFIES));
                }
                if (was_empty) {
                        queue_delayed_work(dev_priv->wq,
                                           &dev_priv->mm.retire_work,
                                           round_jiffies_up_relative(HZ));
                        intel_mark_busy(dev_priv->dev);
                }
        }

        if (out_seqno)
                *out_seqno = request->seqno;
        return 0;
}

static inline void
i915_gem_request_remove_from_client(struct drm_i915_gem_request *request)
{
        struct drm_i915_file_private *file_priv = request->file_priv;

        if (!file_priv)
                return;

        spin_lock(&file_priv->mm.lock);
        if (request->file_priv) {
                list_del(&request->client_list);
                request->file_priv = NULL;
        }
        spin_unlock(&file_priv->mm.lock);
}

static void i915_gem_reset_ring_lists(struct drm_i915_private *dev_priv,
                                      struct intel_ring_buffer *ring)
{
        while (!list_empty(&ring->request_list)) {
                struct drm_i915_gem_request *request;

                request = list_first_entry(&ring->request_list,
                                           struct drm_i915_gem_request,
                                           list);

                list_del(&request->list);
                i915_gem_request_remove_from_client(request);
                kfree(request);
        }

        while (!list_empty(&ring->active_list)) {
                struct drm_i915_gem_object *obj;

                obj = list_first_entry(&ring->active_list,
                                       struct drm_i915_gem_object,
                                       ring_list);

                i915_gem_object_move_to_inactive(obj);
        }
}

void i915_gem_restore_fences(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int i;

        for (i = 0; i < dev_priv->num_fence_regs; i++) {
                struct drm_i915_fence_reg *reg = &dev_priv->fence_regs[i];
                i915_gem_write_fence(dev, i, reg->obj);
        }
}

void i915_gem_reset(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_object *obj;
        struct intel_ring_buffer *ring;
        int i;

        for_each_ring(ring, dev_priv, i)
                i915_gem_reset_ring_lists(dev_priv, ring);

        /* Move everything out of the GPU domains to ensure we do any
         * necessary invalidation upon reuse.
         */
        list_for_each_entry(obj,
                            &dev_priv->mm.inactive_list,
                            mm_list)
        {
                obj->base.read_domains &= ~I915_GEM_GPU_DOMAINS;
        }

        i915_gem_restore_fences(dev);
}

/**
 * This function clears the request list as sequence numbers are passed.
 */
void
i915_gem_retire_requests_ring(struct intel_ring_buffer *ring)
{
        uint32_t seqno;

        if (list_empty(&ring->request_list))
                return;

        WARN_ON(i915_verify_lists(ring->dev));

        seqno = ring->get_seqno(ring, true);

        while (!list_empty(&ring->request_list)) {
                struct drm_i915_gem_request *request;

                request = list_first_entry(&ring->request_list,
                                           struct drm_i915_gem_request,
                                           list);

                if (!i915_seqno_passed(seqno, request->seqno))
                        break;

                trace_i915_gem_request_retire(ring, request->seqno);
                /* We know the GPU must have read the request to have
                 * sent us the seqno + interrupt, so use the position
                 * of tail of the request to update the last known position
                 * of the GPU head.
                 */
                ring->last_retired_head = request->tail;

                list_del(&request->list);
                i915_gem_request_remove_from_client(request);
                kfree(request);
        }

        /* Move any buffers on the active list that are no longer referenced
         * by the ringbuffer to the flushing/inactive lists as appropriate.
         */
        while (!list_empty(&ring->active_list)) {
                struct drm_i915_gem_object *obj;

                obj = list_first_entry(&ring->active_list,
                                      struct drm_i915_gem_object,
                                      ring_list);

                if (!i915_seqno_passed(seqno, obj->last_read_seqno))
                        break;

                i915_gem_object_move_to_inactive(obj);
        }

        if (unlikely(ring->trace_irq_seqno &&
                     i915_seqno_passed(seqno, ring->trace_irq_seqno))) {
                ring->irq_put(ring);
                ring->trace_irq_seqno = 0;
        }

        WARN_ON(i915_verify_lists(ring->dev));
}

void
i915_gem_retire_requests(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        struct intel_ring_buffer *ring;
        int i;

        for_each_ring(ring, dev_priv, i)
                i915_gem_retire_requests_ring(ring);
}

static void
i915_gem_retire_work_handler(struct work_struct *work)
{
        drm_i915_private_t *dev_priv;
        struct drm_device *dev;
        struct intel_ring_buffer *ring;
        bool idle;
        int i;

        dev_priv = container_of(work, drm_i915_private_t,
                                mm.retire_work.work);
        dev = dev_priv->dev;

        /* Come back later if the device is busy... */
        if (!mutex_trylock(&dev->struct_mutex)) {
                queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work,
                                   round_jiffies_up_relative(HZ));
                return;
        }

        i915_gem_retire_requests(dev);

        /* Send a periodic flush down the ring so we don't hold onto GEM
         * objects indefinitely.
         */
        idle = true;
        for_each_ring(ring, dev_priv, i) {
                if (ring->gpu_caches_dirty)
                        i915_add_request(ring, NULL, NULL);

                idle &= list_empty(&ring->request_list);
        }

        if (!dev_priv->mm.suspended && !idle)
                queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work,
                                   round_jiffies_up_relative(HZ));
        if (idle)
                intel_mark_idle(dev);

        mutex_unlock(&dev->struct_mutex);
}

/**
 * Ensures that an object will eventually get non-busy by flushing any required
 * write domains, emitting any outstanding lazy request and retiring and
 * completed requests.
 */
static int
i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
{
        int ret;

        if (obj->active) {
                ret = i915_gem_check_olr(obj->ring, obj->last_read_seqno);
                if (ret)
                        return ret;

                i915_gem_retire_requests_ring(obj->ring);
        }

        return 0;
}

/**
 * i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
 * @DRM_IOCTL_ARGS: standard ioctl arguments
 *
 * Returns 0 if successful, else an error is returned with the remaining time in
 * the timeout parameter.
 *  -ETIME: object is still busy after timeout
 *  -ERESTARTSYS: signal interrupted the wait
 *  -ENONENT: object doesn't exist
 * Also possible, but rare:
 *  -EAGAIN: GPU wedged
 *  -ENOMEM: damn
 *  -ENODEV: Internal IRQ fail
 *  -E?: The add request failed
 *
 * The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any
 * non-zero timeout parameter the wait ioctl will wait for the given number of
 * nanoseconds on an object becoming unbusy. Since the wait itself does so
 * without holding struct_mutex the object may become re-busied before this
 * function completes. A similar but shorter * race condition exists in the busy
 * ioctl
 */
int
i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        struct drm_i915_gem_wait *args = data;
        struct drm_i915_gem_object *obj;
        struct intel_ring_buffer *ring = NULL;
        struct timespec timeout_stack, *timeout = NULL;
        unsigned reset_counter;
        u32 seqno = 0;
        int ret = 0;

        if (args->timeout_ns >= 0) {
                timeout_stack = ns_to_timespec(args->timeout_ns);
                timeout = &timeout_stack;
        }

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->bo_handle));
        if (&obj->base == NULL) {
                mutex_unlock(&dev->struct_mutex);
                return -ENOENT;
        }

        /* Need to make sure the object gets inactive eventually. */
        ret = i915_gem_object_flush_active(obj);
        if (ret)
                goto out;

        if (obj->active) {
                seqno = obj->last_read_seqno;
                ring = obj->ring;
        }

        if (seqno == 0)
                 goto out;

        /* Do this after OLR check to make sure we make forward progress polling
         * on this IOCTL with a 0 timeout (like busy ioctl)
         */
        if (!args->timeout_ns) {
                ret = -ETIME;
                goto out;
        }

        drm_gem_object_unreference(&obj->base);
        reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
        mutex_unlock(&dev->struct_mutex);

        ret = __wait_seqno(ring, seqno, reset_counter, true, timeout);
        if (timeout)
                args->timeout_ns = timespec_to_ns(timeout);
        return ret;

out:
        drm_gem_object_unreference(&obj->base);
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/**
 * i915_gem_object_sync - sync an object to a ring.
 *
 * @obj: object which may be in use on another ring.
 * @to: ring we wish to use the object on. May be NULL.
 *
 * This code is meant to abstract object synchronization with the GPU.
 * Calling with NULL implies synchronizing the object with the CPU
 * rather than a particular GPU ring.
 *
 * Returns 0 if successful, else propagates up the lower layer error.
 */
int
i915_gem_object_sync(struct drm_i915_gem_object *obj,
                     struct intel_ring_buffer *to)
{
        struct intel_ring_buffer *from = obj->ring;
        u32 seqno;
        int ret, idx;

        if (from == NULL || to == from)
                return 0;

        if (to == NULL || !i915_semaphore_is_enabled(obj->base.dev))
                return i915_gem_object_wait_rendering(obj, false);

        idx = intel_ring_sync_index(from, to);

        seqno = obj->last_read_seqno;
        if (seqno <= from->sync_seqno[idx])
                return 0;

        ret = i915_gem_check_olr(obj->ring, seqno);
        if (ret)
                return ret;

        ret = to->sync_to(to, from, seqno);
        if (!ret)
                /* We use last_read_seqno because sync_to()
                 * might have just caused seqno wrap under
                 * the radar.
                 */
                from->sync_seqno[idx] = obj->last_read_seqno;

        return ret;
}

static void i915_gem_object_finish_gtt(struct drm_i915_gem_object *obj)
{
        u32 old_write_domain, old_read_domains;

        /* Force a pagefault for domain tracking on next user access */
        i915_gem_release_mmap(obj);

        if ((obj->base.read_domains & I915_GEM_DOMAIN_GTT) == 0)
                return;

        /* Wait for any direct GTT access to complete */
        mb();

        old_read_domains = obj->base.read_domains;
        old_write_domain = obj->base.write_domain;

        obj->base.read_domains &= ~I915_GEM_DOMAIN_GTT;
        obj->base.write_domain &= ~I915_GEM_DOMAIN_GTT;

        trace_i915_gem_object_change_domain(obj,
                                            old_read_domains,
                                            old_write_domain);
}

/**
 * Unbinds an object from the GTT aperture.
 */
int
i915_gem_object_unbind(struct drm_i915_gem_object *obj)
{
        drm_i915_private_t *dev_priv = obj->base.dev->dev_private;
        int ret;

        if (obj->gtt_space == NULL)
                return 0;

        if (obj->pin_count)
                return -EBUSY;

        BUG_ON(obj->pages == NULL);

        ret = i915_gem_object_finish_gpu(obj);
        if (ret)
                return ret;
        /* Continue on if we fail due to EIO, the GPU is hung so we
         * should be safe and we need to cleanup or else we might
         * cause memory corruption through use-after-free.
         */

        i915_gem_object_finish_gtt(obj);

        /* release the fence reg _after_ flushing */
        ret = i915_gem_object_put_fence(obj);
        if (ret)
                return ret;

        trace_i915_gem_object_unbind(obj);

        if (obj->has_global_gtt_mapping)
                i915_gem_gtt_unbind_object(obj);
        if (obj->has_aliasing_ppgtt_mapping) {
                i915_ppgtt_unbind_object(dev_priv->mm.aliasing_ppgtt, obj);
                obj->has_aliasing_ppgtt_mapping = 0;
        }
        i915_gem_gtt_finish_object(obj);

        list_del(&obj->mm_list);
        list_move_tail(&obj->gtt_list, &dev_priv->mm.unbound_list);
        /* Avoid an unnecessary call to unbind on rebind. */
        obj->map_and_fenceable = true;

        drm_mm_put_block(obj->gtt_space);
        obj->gtt_space = NULL;
        obj->gtt_offset = 0;

        return 0;
}

int i915_gpu_idle(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        struct intel_ring_buffer *ring;
        int ret, i;

        /* Flush everything onto the inactive list. */
        for_each_ring(ring, dev_priv, i) {
                ret = i915_switch_context(ring, NULL, DEFAULT_CONTEXT_ID);
                if (ret)
                        return ret;

                ret = intel_ring_idle(ring);
                if (ret)
                        return ret;
        }

        return 0;
}

static void i965_write_fence_reg(struct drm_device *dev, int reg,
                                 struct drm_i915_gem_object *obj)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        int fence_reg;
        int fence_pitch_shift;
        uint64_t val;

        if (INTEL_INFO(dev)->gen >= 6) {
                fence_reg = FENCE_REG_SANDYBRIDGE_0;
                fence_pitch_shift = SANDYBRIDGE_FENCE_PITCH_SHIFT;
        } else {
                fence_reg = FENCE_REG_965_0;
                fence_pitch_shift = I965_FENCE_PITCH_SHIFT;
        }

        if (obj) {
                u32 size = obj->gtt_space->size;

                val = (uint64_t)((obj->gtt_offset + size - 4096) &
                                 0xfffff000) << 32;
                val |= obj->gtt_offset & 0xfffff000;
                val |= (uint64_t)((obj->stride / 128) - 1) << fence_pitch_shift;
                if (obj->tiling_mode == I915_TILING_Y)
                        val |= 1 << I965_FENCE_TILING_Y_SHIFT;
                val |= I965_FENCE_REG_VALID;
        } else
                val = 0;

        fence_reg += reg * 8;
        I915_WRITE64(fence_reg, val);
        POSTING_READ(fence_reg);
}

static void i915_write_fence_reg(struct drm_device *dev, int reg,
                                 struct drm_i915_gem_object *obj)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        u32 val;

        if (obj) {
                u32 size = obj->gtt_space->size;
                int pitch_val;
                int tile_width;

                WARN((obj->gtt_offset & ~I915_FENCE_START_MASK) ||
                     (size & -size) != size ||
                     (obj->gtt_offset & (size - 1)),
                     "object 0x%08x [fenceable? %d] not 1M or pot-size (0x%08x) aligned\n",
                     obj->gtt_offset, obj->map_and_fenceable, size);

                if (obj->tiling_mode == I915_TILING_Y && HAS_128_BYTE_Y_TILING(dev))
                        tile_width = 128;
                else
                        tile_width = 512;

                /* Note: pitch better be a power of two tile widths */
                pitch_val = obj->stride / tile_width;
                pitch_val = ffs(pitch_val) - 1;

                val = obj->gtt_offset;
                if (obj->tiling_mode == I915_TILING_Y)
                        val |= 1 << I830_FENCE_TILING_Y_SHIFT;
                val |= I915_FENCE_SIZE_BITS(size);
                val |= pitch_val << I830_FENCE_PITCH_SHIFT;
                val |= I830_FENCE_REG_VALID;
        } else
                val = 0;

        if (reg < 8)
                reg = FENCE_REG_830_0 + reg * 4;
        else
                reg = FENCE_REG_945_8 + (reg - 8) * 4;

        I915_WRITE(reg, val);
        POSTING_READ(reg);
}

static void i830_write_fence_reg(struct drm_device *dev, int reg,
                                struct drm_i915_gem_object *obj)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        uint32_t val;

        if (obj) {
                u32 size = obj->gtt_space->size;
                uint32_t pitch_val;

                WARN((obj->gtt_offset & ~I830_FENCE_START_MASK) ||
                     (size & -size) != size ||
                     (obj->gtt_offset & (size - 1)),
                     "object 0x%08x not 512K or pot-size 0x%08x aligned\n",
                     obj->gtt_offset, size);

                pitch_val = obj->stride / 128;
                pitch_val = ffs(pitch_val) - 1;

                val = obj->gtt_offset;
                if (obj->tiling_mode == I915_TILING_Y)
                        val |= 1 << I830_FENCE_TILING_Y_SHIFT;
                val |= I830_FENCE_SIZE_BITS(size);
                val |= pitch_val << I830_FENCE_PITCH_SHIFT;
                val |= I830_FENCE_REG_VALID;
        } else
                val = 0;

        I915_WRITE(FENCE_REG_830_0 + reg * 4, val);
        POSTING_READ(FENCE_REG_830_0 + reg * 4);
}

inline static bool i915_gem_object_needs_mb(struct drm_i915_gem_object *obj)
{
        return obj && obj->base.read_domains & I915_GEM_DOMAIN_GTT;
}

static void i915_gem_write_fence(struct drm_device *dev, int reg,
                                 struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        /* Ensure that all CPU reads are completed before installing a fence
         * and all writes before removing the fence.
         */
        if (i915_gem_object_needs_mb(dev_priv->fence_regs[reg].obj))
                mb();

        switch (INTEL_INFO(dev)->gen) {
        case 7:
        case 6:
        case 5:
        case 4: i965_write_fence_reg(dev, reg, obj); break;
        case 3: i915_write_fence_reg(dev, reg, obj); break;
        case 2: i830_write_fence_reg(dev, reg, obj); break;
        default: BUG();
        }

        /* And similarly be paranoid that no direct access to this region
         * is reordered to before the fence is installed.
         */
        if (i915_gem_object_needs_mb(obj))
                mb();
}

static inline int fence_number(struct drm_i915_private *dev_priv,
                               struct drm_i915_fence_reg *fence)
{
        return fence - dev_priv->fence_regs;
}

static void i915_gem_write_fence__ipi(void *data)
{
        wbinvd();
}

static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj,
                                         struct drm_i915_fence_reg *fence,
                                         bool enable)
{
        struct drm_device *dev = obj->base.dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        int fence_reg = fence_number(dev_priv, fence);

        /* In order to fully serialize access to the fenced region and
         * the update to the fence register we need to take extreme
         * measures on SNB+. In theory, the write to the fence register
         * flushes all memory transactions before, and coupled with the
         * mb() placed around the register write we serialise all memory
         * operations with respect to the changes in the tiler. Yet, on
         * SNB+ we need to take a step further and emit an explicit wbinvd()
         * on each processor in order to manually flush all memory
         * transactions before updating the fence register.
         */
        if (HAS_LLC(obj->base.dev))
                on_each_cpu(i915_gem_write_fence__ipi, NULL, 1);
        i915_gem_write_fence(dev, fence_reg, enable ? obj : NULL);

        if (enable) {
                obj->fence_reg = fence_reg;
                fence->obj = obj;
                list_move_tail(&fence->lru_list, &dev_priv->mm.fence_list);
        } else {
                obj->fence_reg = I915_FENCE_REG_NONE;
                fence->obj = NULL;
                list_del_init(&fence->lru_list);
        }
}

static int
i915_gem_object_wait_fence(struct drm_i915_gem_object *obj)
{
        if (obj->last_fenced_seqno) {
                int ret = i915_wait_seqno(obj->ring, obj->last_fenced_seqno);
                if (ret)
                        return ret;

                obj->last_fenced_seqno = 0;
        }

        obj->fenced_gpu_access = false;
        return 0;
}

int
i915_gem_object_put_fence(struct drm_i915_gem_object *obj)
{
        struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
        struct drm_i915_fence_reg *fence;
        int ret;

        ret = i915_gem_object_wait_fence(obj);
        if (ret)
                return ret;

        if (obj->fence_reg == I915_FENCE_REG_NONE)
                return 0;

        fence = &dev_priv->fence_regs[obj->fence_reg];

        i915_gem_object_fence_lost(obj);
        i915_gem_object_update_fence(obj, fence, false);

        return 0;
}

static struct drm_i915_fence_reg *
i915_find_fence_reg(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_fence_reg *reg, *avail;
        int i;

        /* First try to find a free reg */
        avail = NULL;
        for (i = dev_priv->fence_reg_start; i < dev_priv->num_fence_regs; i++) {
                reg = &dev_priv->fence_regs[i];
                if (!reg->obj)
                        return reg;

                if (!reg->pin_count)
                        avail = reg;
        }

        if (avail == NULL)
                return NULL;

        /* None available, try to steal one or wait for a user to finish */
        list_for_each_entry(reg, &dev_priv->mm.fence_list, lru_list) {
                if (reg->pin_count)
                        continue;

                return reg;
        }

        return NULL;
}

/**
 * i915_gem_object_get_fence - set up fencing for an object
 * @obj: object to map through a fence reg
 *
 * When mapping objects through the GTT, userspace wants to be able to write
 * to them without having to worry about swizzling if the object is tiled.
 * This function walks the fence regs looking for a free one for @obj,
 * stealing one if it can't find any.
 *
 * It then sets up the reg based on the object's properties: address, pitch
 * and tiling format.
 *
 * For an untiled surface, this removes any existing fence.
 */
int
i915_gem_object_get_fence(struct drm_i915_gem_object *obj)
{
        struct drm_device *dev = obj->base.dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        bool enable = obj->tiling_mode != I915_TILING_NONE;
        struct drm_i915_fence_reg *reg;
        int ret;

        /* Have we updated the tiling parameters upon the object and so
         * will need to serialise the write to the associated fence register?
         */
        if (obj->fence_dirty) {
                ret = i915_gem_object_wait_fence(obj);
                if (ret)
                        return ret;
        }

        /* Just update our place in the LRU if our fence is getting reused. */
        if (obj->fence_reg != I915_FENCE_REG_NONE) {
                reg = &dev_priv->fence_regs[obj->fence_reg];
                if (!obj->fence_dirty) {
                        list_move_tail(&reg->lru_list,
                                       &dev_priv->mm.fence_list);
                        return 0;
                }
        } else if (enable) {
                reg = i915_find_fence_reg(dev);
                if (reg == NULL)
                        return -EDEADLK;

                if (reg->obj) {
                        struct drm_i915_gem_object *old = reg->obj;

                        ret = i915_gem_object_wait_fence(old);
                        if (ret)
                                return ret;

                        i915_gem_object_fence_lost(old);
                }
        } else
                return 0;

        i915_gem_object_update_fence(obj, reg, enable);
        obj->fence_dirty = false;

        return 0;
}

static bool i915_gem_valid_gtt_space(struct drm_device *dev,
                                     struct drm_mm_node *gtt_space,
                                     unsigned long cache_level)
{
        struct drm_mm_node *other;

        /* On non-LLC machines we have to be careful when putting differing
         * types of snoopable memory together to avoid the prefetcher
         * crossing memory domains and dying.
         */
        if (HAS_LLC(dev))
                return true;

        if (gtt_space == NULL)
                return true;

        if (list_empty(&gtt_space->node_list))
                return true;

        other = list_entry(gtt_space->node_list.prev, struct drm_mm_node, node_list);
        if (other->allocated && !other->hole_follows && other->color != cache_level)
                return false;

        other = list_entry(gtt_space->node_list.next, struct drm_mm_node, node_list);
        if (other->allocated && !gtt_space->hole_follows && other->color != cache_level)
                return false;

        return true;
}

static void i915_gem_verify_gtt(struct drm_device *dev)
{
#if WATCH_GTT
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_object *obj;
        int err = 0;

        list_for_each_entry(obj, &dev_priv->mm.gtt_list, gtt_list) {
                if (obj->gtt_space == NULL) {
                        printk(KERN_ERR "object found on GTT list with no space reserved\n");
                        err++;
                        continue;
                }

                if (obj->cache_level != obj->gtt_space->color) {
                        printk(KERN_ERR "object reserved space [%08lx, %08lx] with wrong color, cache_level=%x, color=%lx\n",
                               obj->gtt_space->start,
                               obj->gtt_space->start + obj->gtt_space->size,
                               obj->cache_level,
                               obj->gtt_space->color);
                        err++;
                        continue;
                }

                if (!i915_gem_valid_gtt_space(dev,
                                              obj->gtt_space,
                                              obj->cache_level)) {
                        printk(KERN_ERR "invalid GTT space found at [%08lx, %08lx] - color=%x\n",
                               obj->gtt_space->start,
                               obj->gtt_space->start + obj->gtt_space->size,
                               obj->cache_level);
                        err++;
                        continue;
                }
        }

        WARN_ON(err);
#endif
}

/**
 * Finds free space in the GTT aperture and binds the object there.
 */
static int
i915_gem_object_bind_to_gtt(struct drm_i915_gem_object *obj,
                            unsigned alignment,
                            bool map_and_fenceable,
                            bool nonblocking)
{
        struct drm_device *dev = obj->base.dev;
        drm_i915_private_t *dev_priv = dev->dev_private;
        struct drm_mm_node *node;
        u32 size, fence_size, fence_alignment, unfenced_alignment;
        bool mappable, fenceable;
        int ret;

        fence_size = i915_gem_get_gtt_size(dev,
                                           obj->base.size,
                                           obj->tiling_mode);
        fence_alignment = i915_gem_get_gtt_alignment(dev,
                                                     obj->base.size,
                                                     obj->tiling_mode, true);
        unfenced_alignment =
                i915_gem_get_gtt_alignment(dev,
                                                    obj->base.size,
                                                    obj->tiling_mode, false);

        if (alignment == 0)
                alignment = map_and_fenceable ? fence_alignment :
                                                unfenced_alignment;
        if (map_and_fenceable && alignment & (fence_alignment - 1)) {
                DRM_ERROR("Invalid object alignment requested %u\n", alignment);
                return -EINVAL;
        }

        size = map_and_fenceable ? fence_size : obj->base.size;

        /* If the object is bigger than the entire aperture, reject it early
         * before evicting everything in a vain attempt to find space.
         */
        if (obj->base.size >
            (map_and_fenceable ? dev_priv->gtt.mappable_end : dev_priv->gtt.total)) {
                DRM_ERROR("Attempting to bind an object larger than the aperture\n");
                return -E2BIG;
        }

        ret = i915_gem_object_get_pages(obj);
        if (ret)
                return ret;

        i915_gem_object_pin_pages(obj);

        node = kzalloc(sizeof(*node), GFP_KERNEL);
        if (node == NULL) {
                i915_gem_object_unpin_pages(obj);
                return -ENOMEM;
        }

 search_free:
        if (map_and_fenceable)
                ret = drm_mm_insert_node_in_range_generic(&dev_priv->mm.gtt_space, node,
                                                          size, alignment, obj->cache_level,
                                                          0, dev_priv->gtt.mappable_end);
        else
                ret = drm_mm_insert_node_generic(&dev_priv->mm.gtt_space, node,
                                                 size, alignment, obj->cache_level);
        if (ret) {
                ret = i915_gem_evict_something(dev, size, alignment,
                                               obj->cache_level,
                                               map_and_fenceable,
                                               nonblocking);
                if (ret == 0)
                        goto search_free;

                i915_gem_object_unpin_pages(obj);
                kfree(node);
                return ret;
        }
        if (WARN_ON(!i915_gem_valid_gtt_space(dev, node, obj->cache_level))) {
                i915_gem_object_unpin_pages(obj);
                drm_mm_put_block(node);
                return -EINVAL;

        }

        ret = i915_gem_gtt_prepare_object(obj);
        if (ret) {
                i915_gem_object_unpin_pages(obj);
                drm_mm_put_block(node);
                return ret;
        }

        list_move_tail(&obj->gtt_list, &dev_priv->mm.bound_list);
        list_add_tail(&obj->mm_list, &dev_priv->mm.inactive_list);

        obj->gtt_space = node;
        obj->gtt_offset = node->start;

        fenceable =
                node->size == fence_size &&
                (node->start & (fence_alignment - 1)) == 0;

        mappable =
                obj->gtt_offset + obj->base.size <= dev_priv->gtt.mappable_end;

        obj->map_and_fenceable = mappable && fenceable;

        i915_gem_object_unpin_pages(obj);
        trace_i915_gem_object_bind(obj, map_and_fenceable);
        i915_gem_verify_gtt(dev);
        return 0;
}

void
i915_gem_clflush_object(struct drm_i915_gem_object *obj)
{
        /* If we don't have a page list set up, then we're not pinned
         * to GPU, and we can ignore the cache flush because it'll happen
         * again at bind time.
         */
        if (obj->pages == NULL)
                return;

        /*
         * Stolen memory is always coherent with the GPU as it is explicitly
         * marked as wc by the system, or the system is cache-coherent.
         */
        if (obj->stolen)
                return;

        /* If the GPU is snooping the contents of the CPU cache,
         * we do not need to manually clear the CPU cache lines.  However,
         * the caches are only snooped when the render cache is
         * flushed/invalidated.  As we always have to emit invalidations
         * and flushes when moving into and out of the RENDER domain, correct
         * snooping behaviour occurs naturally as the result of our domain
         * tracking.
         */
        if (obj->cache_level != I915_CACHE_NONE)
                return;

        trace_i915_gem_object_clflush(obj);

        drm_clflush_sg(obj->pages);
}

/** Flushes the GTT write domain for the object if it's dirty. */
static void
i915_gem_object_flush_gtt_write_domain(struct drm_i915_gem_object *obj)
{
        uint32_t old_write_domain;

        if (obj->base.write_domain != I915_GEM_DOMAIN_GTT)
                return;

        /* No actual flushing is required for the GTT write domain.  Writes
         * to it immediately go to main memory as far as we know, so there's
         * no chipset flush.  It also doesn't land in render cache.
         *
         * However, we do have to enforce the order so that all writes through
         * the GTT land before any writes to the device, such as updates to
         * the GATT itself.
         */
        wmb();

        old_write_domain = obj->base.write_domain;
        obj->base.write_domain = 0;

        trace_i915_gem_object_change_domain(obj,
                                            obj->base.read_domains,
                                            old_write_domain);
}

/** Flushes the CPU write domain for the object if it's dirty. */
static void
i915_gem_object_flush_cpu_write_domain(struct drm_i915_gem_object *obj)
{
        uint32_t old_write_domain;

        if (obj->base.write_domain != I915_GEM_DOMAIN_CPU)
                return;

        i915_gem_clflush_object(obj);
        i915_gem_chipset_flush(obj->base.dev);
        old_write_domain = obj->base.write_domain;
        obj->base.write_domain = 0;

        trace_i915_gem_object_change_domain(obj,
                                            obj->base.read_domains,
                                            old_write_domain);
}

/**
 * Moves a single object to the GTT read, and possibly write domain.
 *
 * This function returns when the move is complete, including waiting on
 * flushes to occur.
 */
int
i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
{
        drm_i915_private_t *dev_priv = obj->base.dev->dev_private;
        uint32_t old_write_domain, old_read_domains;
        int ret;

        /* Not valid to be called on unbound objects. */
        if (obj->gtt_space == NULL)
                return -EINVAL;

        if (obj->base.write_domain == I915_GEM_DOMAIN_GTT)
                return 0;

        ret = i915_gem_object_wait_rendering(obj, !write);
        if (ret)
                return ret;

        i915_gem_object_flush_cpu_write_domain(obj);

        /* Serialise direct access to this object with the barriers for
         * coherent writes from the GPU, by effectively invalidating the
         * GTT domain upon first access.
         */
        if ((obj->base.read_domains & I915_GEM_DOMAIN_GTT) == 0)
                mb();

        old_write_domain = obj->base.write_domain;
        old_read_domains = obj->base.read_domains;

        /* It should now be out of any other write domains, and we can update
         * the domain values for our changes.
         */
        BUG_ON((obj->base.write_domain & ~I915_GEM_DOMAIN_GTT) != 0);
        obj->base.read_domains |= I915_GEM_DOMAIN_GTT;
        if (write) {
                obj->base.read_domains = I915_GEM_DOMAIN_GTT;
                obj->base.write_domain = I915_GEM_DOMAIN_GTT;
                obj->dirty = 1;
        }

        trace_i915_gem_object_change_domain(obj,
                                            old_read_domains,
                                            old_write_domain);

        /* And bump the LRU for this access */
        if (i915_gem_object_is_inactive(obj))
                list_move_tail(&obj->mm_list, &dev_priv->mm.inactive_list);

        return 0;
}

int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
                                    enum i915_cache_level cache_level)
{
        struct drm_device *dev = obj->base.dev;
        drm_i915_private_t *dev_priv = dev->dev_private;
        int ret;

        if (obj->cache_level == cache_level)
                return 0;

        if (obj->pin_count) {
                DRM_DEBUG("can not change the cache level of pinned objects\n");
                return -EBUSY;
        }

        if (!i915_gem_valid_gtt_space(dev, obj->gtt_space, cache_level)) {
                ret = i915_gem_object_unbind(obj);
                if (ret)
                        return ret;
        }

        if (obj->gtt_space) {
                ret = i915_gem_object_finish_gpu(obj);
                if (ret)
                        return ret;

                i915_gem_object_finish_gtt(obj);

                /* Before SandyBridge, you could not use tiling or fence
                 * registers with snooped memory, so relinquish any fences
                 * currently pointing to our region in the aperture.
                 */
                if (INTEL_INFO(dev)->gen < 6) {
                        ret = i915_gem_object_put_fence(obj);
                        if (ret)
                                return ret;
                }

                if (obj->has_global_gtt_mapping)
                        i915_gem_gtt_bind_object(obj, cache_level);
                if (obj->has_aliasing_ppgtt_mapping)
                        i915_ppgtt_bind_object(dev_priv->mm.aliasing_ppgtt,
                                               obj, cache_level);

                obj->gtt_space->color = cache_level;
        }

        if (cache_level == I915_CACHE_NONE) {
                u32 old_read_domains, old_write_domain;

                /* If we're coming from LLC cached, then we haven't
                 * actually been tracking whether the data is in the
                 * CPU cache or not, since we only allow one bit set
                 * in obj->write_domain and have been skipping the clflushes.
                 * Just set it to the CPU cache for now.
                 */
                WARN_ON(obj->base.write_domain & ~I915_GEM_DOMAIN_CPU);
                WARN_ON(obj->base.read_domains & ~I915_GEM_DOMAIN_CPU);

                old_read_domains = obj->base.read_domains;
                old_write_domain = obj->base.write_domain;

                obj->base.read_domains = I915_GEM_DOMAIN_CPU;
                obj->base.write_domain = I915_GEM_DOMAIN_CPU;

                trace_i915_gem_object_change_domain(obj,
                                                    old_read_domains,
                                                    old_write_domain);
        }

        obj->cache_level = cache_level;
        i915_gem_verify_gtt(dev);
        return 0;
}

int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
                               struct drm_file *file)
{
        struct drm_i915_gem_caching *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        args->caching = obj->cache_level != I915_CACHE_NONE;

        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
                               struct drm_file *file)
{
        struct drm_i915_gem_caching *args = data;
        struct drm_i915_gem_object *obj;
        enum i915_cache_level level;
        int ret;

        switch (args->caching) {
        case I915_CACHING_NONE:
                level = I915_CACHE_NONE;
                break;
        case I915_CACHING_CACHED:
                level = I915_CACHE_LLC;
                break;
        default:
                return -EINVAL;
        }

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        ret = i915_gem_object_set_cache_level(obj, level);

        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

/*
 * Prepare buffer for display plane (scanout, cursors, etc).
 * Can be called from an uninterruptible phase (modesetting) and allows
 * any flushes to be pipelined (for pageflips).
 */
int
i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
                                     u32 alignment,
                                     struct intel_ring_buffer *pipelined)
{
        u32 old_read_domains, old_write_domain;
        int ret;

        if (pipelined != obj->ring) {
                ret = i915_gem_object_sync(obj, pipelined);
                if (ret)
                        return ret;
        }

        /* The display engine is not coherent with the LLC cache on gen6.  As
         * a result, we make sure that the pinning that is about to occur is
         * done with uncached PTEs. This is lowest common denominator for all
         * chipsets.
         *
         * However for gen6+, we could do better by using the GFDT bit instead
         * of uncaching, which would allow us to flush all the LLC-cached data
         * with that bit in the PTE to main memory with just one PIPE_CONTROL.
         */
        ret = i915_gem_object_set_cache_level(obj, I915_CACHE_NONE);
        if (ret)
                return ret;

        /* As the user may map the buffer once pinned in the display plane
         * (e.g. libkms for the bootup splash), we have to ensure that we
         * always use map_and_fenceable for all scanout buffers.
         */
        ret = i915_gem_object_pin(obj, alignment, true, false);
        if (ret)
                return ret;

        i915_gem_object_flush_cpu_write_domain(obj);

        old_write_domain = obj->base.write_domain;
        old_read_domains = obj->base.read_domains;

        /* It should now be out of any other write domains, and we can update
         * the domain values for our changes.
         */
        obj->base.write_domain = 0;
        obj->base.read_domains |= I915_GEM_DOMAIN_GTT;

        trace_i915_gem_object_change_domain(obj,
                                            old_read_domains,
                                            old_write_domain);

        return 0;
}

int
i915_gem_object_finish_gpu(struct drm_i915_gem_object *obj)
{
        int ret;

        if ((obj->base.read_domains & I915_GEM_GPU_DOMAINS) == 0)
                return 0;

        ret = i915_gem_object_wait_rendering(obj, false);
        if (ret)
                return ret;

        /* Ensure that we invalidate the GPU's caches and TLBs. */
        obj->base.read_domains &= ~I915_GEM_GPU_DOMAINS;
        return 0;
}

/**
 * Moves a single object to the CPU read, and possibly write domain.
 *
 * This function returns when the move is complete, including waiting on
 * flushes to occur.
 */
int
i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
{
        uint32_t old_write_domain, old_read_domains;
        int ret;

        if (obj->base.write_domain == I915_GEM_DOMAIN_CPU)
                return 0;

        ret = i915_gem_object_wait_rendering(obj, !write);
        if (ret)
                return ret;

        i915_gem_object_flush_gtt_write_domain(obj);

        old_write_domain = obj->base.write_domain;
        old_read_domains = obj->base.read_domains;

        /* Flush the CPU cache if it's still invalid. */
        if ((obj->base.read_domains & I915_GEM_DOMAIN_CPU) == 0) {
                i915_gem_clflush_object(obj);

                obj->base.read_domains |= I915_GEM_DOMAIN_CPU;
        }

        /* It should now be out of any other write domains, and we can update
         * the domain values for our changes.
         */
        BUG_ON((obj->base.write_domain & ~I915_GEM_DOMAIN_CPU) != 0);

        /* If we're writing through the CPU, then the GPU read domains will
         * need to be invalidated at next use.
         */
        if (write) {
                obj->base.read_domains = I915_GEM_DOMAIN_CPU;
                obj->base.write_domain = I915_GEM_DOMAIN_CPU;
        }

        trace_i915_gem_object_change_domain(obj,
                                            old_read_domains,
                                            old_write_domain);

        return 0;
}

/* Throttle our rendering by waiting until the ring has completed our requests
 * emitted over 20 msec ago.
 *
 * Note that if we were to use the current jiffies each time around the loop,
 * we wouldn't escape the function with any frames outstanding if the time to
 * render a frame was over 20ms.
 *
 * This should get us reasonable parallelism between CPU and GPU but also
 * relatively low latency when blocking on a particular request to finish.
 */
static int
i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_file_private *file_priv = file->driver_priv;
        unsigned long recent_enough = jiffies - msecs_to_jiffies(20);
        struct drm_i915_gem_request *request;
        struct intel_ring_buffer *ring = NULL;
        unsigned reset_counter;
        u32 seqno = 0;
        int ret;

        ret = i915_gem_wait_for_error(&dev_priv->gpu_error);
        if (ret)
                return ret;

        ret = i915_gem_check_wedge(&dev_priv->gpu_error, false);
        if (ret)
                return ret;

        spin_lock(&file_priv->mm.lock);
        list_for_each_entry(request, &file_priv->mm.request_list, client_list) {
                if (time_after_eq(request->emitted_jiffies, recent_enough))
                        break;

                ring = request->ring;
                seqno = request->seqno;
        }
        reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
        spin_unlock(&file_priv->mm.lock);

        if (seqno == 0)
                return 0;

        ret = __wait_seqno(ring, seqno, reset_counter, true, NULL);
        if (ret == 0)
                queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0);

        return ret;
}

int
i915_gem_object_pin(struct drm_i915_gem_object *obj,
                    uint32_t alignment,
                    bool map_and_fenceable,
                    bool nonblocking)
{
        int ret;

        if (WARN_ON(obj->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT))
                return -EBUSY;

        if (obj->gtt_space != NULL) {
                if ((alignment && obj->gtt_offset & (alignment - 1)) ||
                    (map_and_fenceable && !obj->map_and_fenceable)) {
                        WARN(obj->pin_count,
                             "bo is already pinned with incorrect alignment:"
                             " offset=%x, req.alignment=%x, req.map_and_fenceable=%d,"
                             " obj->map_and_fenceable=%d\n",
                             obj->gtt_offset, alignment,
                             map_and_fenceable,
                             obj->map_and_fenceable);
                        ret = i915_gem_object_unbind(obj);
                        if (ret)
                                return ret;
                }
        }

        if (obj->gtt_space == NULL) {
                struct drm_i915_private *dev_priv = obj->base.dev->dev_private;

                ret = i915_gem_object_bind_to_gtt(obj, alignment,
                                                  map_and_fenceable,
                                                  nonblocking);
                if (ret)
                        return ret;

                if (!dev_priv->mm.aliasing_ppgtt)
                        i915_gem_gtt_bind_object(obj, obj->cache_level);
        }

        if (!obj->has_global_gtt_mapping && map_and_fenceable)
                i915_gem_gtt_bind_object(obj, obj->cache_level);

        obj->pin_count++;
        obj->pin_mappable |= map_and_fenceable;

        return 0;
}

void
i915_gem_object_unpin(struct drm_i915_gem_object *obj)
{
        BUG_ON(obj->pin_count == 0);
        BUG_ON(obj->gtt_space == NULL);

        if (--obj->pin_count == 0)
                obj->pin_mappable = false;
}

int
i915_gem_pin_ioctl(struct drm_device *dev, void *data,
                   struct drm_file *file)
{
        struct drm_i915_gem_pin *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        if (obj->madv != I915_MADV_WILLNEED) {
                DRM_ERROR("Attempting to pin a purgeable buffer\n");
                ret = -EINVAL;
                goto out;
        }

        if (obj->pin_filp != NULL && obj->pin_filp != file) {
                DRM_ERROR("Already pinned in i915_gem_pin_ioctl(): %d\n",
                          args->handle);
                ret = -EINVAL;
                goto out;
        }

        if (obj->user_pin_count == 0) {
                ret = i915_gem_object_pin(obj, args->alignment, true, false);
                if (ret)
                        goto out;
        }

        obj->user_pin_count++;
        obj->pin_filp = file;

        /* XXX - flush the CPU caches for pinned objects
         * as the X server doesn't manage domains yet
         */
        i915_gem_object_flush_cpu_write_domain(obj);
        args->offset = obj->gtt_offset;
out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

int
i915_gem_unpin_ioctl(struct drm_device *dev, void *data,
                     struct drm_file *file)
{
        struct drm_i915_gem_pin *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        if (obj->pin_filp != file) {
                DRM_ERROR("Not pinned by caller in i915_gem_pin_ioctl(): %d\n",
                          args->handle);
                ret = -EINVAL;
                goto out;
        }
        obj->user_pin_count--;
        if (obj->user_pin_count == 0) {
                obj->pin_filp = NULL;
                i915_gem_object_unpin(obj);
        }

out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

int
i915_gem_busy_ioctl(struct drm_device *dev, void *data,
                    struct drm_file *file)
{
        struct drm_i915_gem_busy *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        /* Count all active objects as busy, even if they are currently not used
         * by the gpu. Users of this interface expect objects to eventually
         * become non-busy without any further actions, therefore emit any
         * necessary flushes here.
         */
        ret = i915_gem_object_flush_active(obj);

        args->busy = obj->active;
        if (obj->ring) {
                BUILD_BUG_ON(I915_NUM_RINGS > 16);
                args->busy |= intel_ring_flag(obj->ring) << 16;
        }

        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

int
i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
                        struct drm_file *file_priv)
{
        return i915_gem_ring_throttle(dev, file_priv);
}

int
i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
                       struct drm_file *file_priv)
{
        struct drm_i915_gem_madvise *args = data;
        struct drm_i915_gem_object *obj;
        int ret;

        switch (args->madv) {
        case I915_MADV_DONTNEED:
        case I915_MADV_WILLNEED:
            break;
        default:
            return -EINVAL;
        }

        ret = i915_mutex_lock_interruptible(dev);
        if (ret)
                return ret;

        obj = to_intel_bo(drm_gem_object_lookup(dev, file_priv, args->handle));
        if (&obj->base == NULL) {
                ret = -ENOENT;
                goto unlock;
        }

        if (obj->pin_count) {
                ret = -EINVAL;
                goto out;
        }

        if (obj->madv != __I915_MADV_PURGED)
                obj->madv = args->madv;

        /* if the object is no longer attached, discard its backing storage */
        if (i915_gem_object_is_purgeable(obj) && obj->pages == NULL)
                i915_gem_object_truncate(obj);

        args->retained = obj->madv != __I915_MADV_PURGED;

out:
        drm_gem_object_unreference(&obj->base);
unlock:
        mutex_unlock(&dev->struct_mutex);
        return ret;
}

void i915_gem_object_init(struct drm_i915_gem_object *obj,
                          const struct drm_i915_gem_object_ops *ops)
{
        INIT_LIST_HEAD(&obj->mm_list);
        INIT_LIST_HEAD(&obj->gtt_list);
        INIT_LIST_HEAD(&obj->ring_list);
        INIT_LIST_HEAD(&obj->exec_list);

        obj->ops = ops;

        obj->fence_reg = I915_FENCE_REG_NONE;
        obj->madv = I915_MADV_WILLNEED;
        /* Avoid an unnecessary call to unbind on the first bind. */
        obj->map_and_fenceable = true;

        i915_gem_info_add_obj(obj->base.dev->dev_private, obj->base.size);
}

static const struct drm_i915_gem_object_ops i915_gem_object_ops = {
        .get_pages = i915_gem_object_get_pages_gtt,
        .put_pages = i915_gem_object_put_pages_gtt,
};

struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
                                                  size_t size)
{
        struct drm_i915_gem_object *obj;
        struct address_space *mapping;
        gfp_t mask;

        obj = i915_gem_object_alloc(dev);
        if (obj == NULL)
                return NULL;

        if (drm_gem_object_init(dev, &obj->base, size) != 0) {
                i915_gem_object_free(obj);
                return NULL;
        }

        mask = GFP_HIGHUSER | __GFP_RECLAIMABLE;
        if (IS_CRESTLINE(dev) || IS_BROADWATER(dev)) {
                /* 965gm cannot relocate objects above 4GiB. */
                mask &= ~__GFP_HIGHMEM;
                mask |= __GFP_DMA32;
        }

        mapping = file_inode(obj->base.filp)->i_mapping;
        mapping_set_gfp_mask(mapping, mask);

        i915_gem_object_init(obj, &i915_gem_object_ops);

        obj->base.write_domain = I915_GEM_DOMAIN_CPU;
        obj->base.read_domains = I915_GEM_DOMAIN_CPU;

        if (HAS_LLC(dev)) {
                /* On some devices, we can have the GPU use the LLC (the CPU
                 * cache) for about a 10% performance improvement
                 * compared to uncached.  Graphics requests other than
                 * display scanout are coherent with the CPU in
                 * accessing this cache.  This means in this mode we
                 * don't need to clflush on the CPU side, and on the
                 * GPU side we only need to flush internal caches to
                 * get data visible to the CPU.
                 *
                 * However, we maintain the display planes as UC, and so
                 * need to rebind when first used as such.
                 */
                obj->cache_level = I915_CACHE_LLC;
        } else
                obj->cache_level = I915_CACHE_NONE;

        return obj;
}

int i915_gem_init_object(struct drm_gem_object *obj)
{
        BUG();

        return 0;
}

void i915_gem_free_object(struct drm_gem_object *gem_obj)
{
        struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
        struct drm_device *dev = obj->base.dev;
        drm_i915_private_t *dev_priv = dev->dev_private;

        trace_i915_gem_object_destroy(obj);

        if (obj->phys_obj)
                i915_gem_detach_phys_object(dev, obj);

        obj->pin_count = 0;
        if (WARN_ON(i915_gem_object_unbind(obj) == -ERESTARTSYS)) {
                bool was_interruptible;

                was_interruptible = dev_priv->mm.interruptible;
                dev_priv->mm.interruptible = false;

                WARN_ON(i915_gem_object_unbind(obj));

                dev_priv->mm.interruptible = was_interruptible;
        }

        obj->pages_pin_count = 0;
        i915_gem_object_put_pages(obj);
        i915_gem_object_free_mmap_offset(obj);
        i915_gem_object_release_stolen(obj);

        BUG_ON(obj->pages);

        if (obj->base.import_attach)
                drm_prime_gem_destroy(&obj->base, NULL);

        drm_gem_object_release(&obj->base);
        i915_gem_info_remove_obj(dev_priv, obj->base.size);

        kfree(obj->bit_17);
        i915_gem_object_free(obj);
}

int
i915_gem_idle(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        int ret;

        mutex_lock(&dev->struct_mutex);

        if (dev_priv->mm.suspended) {
                mutex_unlock(&dev->struct_mutex);
                return 0;
        }

        ret = i915_gpu_idle(dev);
        if (ret) {
                mutex_unlock(&dev->struct_mutex);
                return ret;
        }
        i915_gem_retire_requests(dev);

        /* Under UMS, be paranoid and evict. */
        if (!drm_core_check_feature(dev, DRIVER_MODESET))
                i915_gem_evict_everything(dev);

        /* Hack!  Don't let anybody do execbuf while we don't control the chip.
         * We need to replace this with a semaphore, or something.
         * And not confound mm.suspended!
         */
        dev_priv->mm.suspended = 1;
        del_timer_sync(&dev_priv->gpu_error.hangcheck_timer);

        i915_kernel_lost_context(dev);
        i915_gem_cleanup_ringbuffer(dev);

        mutex_unlock(&dev->struct_mutex);

        /* Cancel the retire work handler, which should be idle now. */
        cancel_delayed_work_sync(&dev_priv->mm.retire_work);

        return 0;
}

void i915_gem_l3_remap(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        u32 misccpctl;
        int i;

        if (!HAS_L3_GPU_CACHE(dev))
                return;

        if (!dev_priv->l3_parity.remap_info)
                return;

        misccpctl = I915_READ(GEN7_MISCCPCTL);
        I915_WRITE(GEN7_MISCCPCTL, misccpctl & ~GEN7_DOP_CLOCK_GATE_ENABLE);
        POSTING_READ(GEN7_MISCCPCTL);

        for (i = 0; i < GEN7_L3LOG_SIZE; i += 4) {
                u32 remap = I915_READ(GEN7_L3LOG_BASE + i);
                if (remap && remap != dev_priv->l3_parity.remap_info[i/4])
                        DRM_DEBUG("0x%x was already programmed to %x\n",
                                  GEN7_L3LOG_BASE + i, remap);
                if (remap && !dev_priv->l3_parity.remap_info[i/4])
                        DRM_DEBUG_DRIVER("Clearing remapped register\n");
                I915_WRITE(GEN7_L3LOG_BASE + i, dev_priv->l3_parity.remap_info[i/4]);
        }

        /* Make sure all the writes land before disabling dop clock gating */
        POSTING_READ(GEN7_L3LOG_BASE);

        I915_WRITE(GEN7_MISCCPCTL, misccpctl);
}

void i915_gem_init_swizzling(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;

        if (INTEL_INFO(dev)->gen < 5 ||
            dev_priv->mm.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_NONE)
                return;

        I915_WRITE(DISP_ARB_CTL, I915_READ(DISP_ARB_CTL) |
                                 DISP_TILE_SURFACE_SWIZZLING);

        if (IS_GEN5(dev))
                return;

        I915_WRITE(TILECTL, I915_READ(TILECTL) | TILECTL_SWZCTL);
        if (IS_GEN6(dev))
                I915_WRITE(ARB_MODE, _MASKED_BIT_ENABLE(ARB_MODE_SWIZZLE_SNB));
        else if (IS_GEN7(dev))
                I915_WRITE(ARB_MODE, _MASKED_BIT_ENABLE(ARB_MODE_SWIZZLE_IVB));
        else
                BUG();
}

static bool
intel_enable_blt(struct drm_device *dev)
{
        if (!HAS_BLT(dev))
                return false;

        /* The blitter was dysfunctional on early prototypes */
        if (IS_GEN6(dev) && dev->pdev->revision < 8) {
                DRM_INFO("BLT not supported on this pre-production hardware;"
                         " graphics performance will be degraded.\n");
                return false;
        }

        return true;
}

static int i915_gem_init_rings(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret;

        ret = intel_init_render_ring_buffer(dev);
        if (ret)
                return ret;

        if (HAS_BSD(dev)) {
                ret = intel_init_bsd_ring_buffer(dev);
                if (ret)
                        goto cleanup_render_ring;
        }

        if (intel_enable_blt(dev)) {
                ret = intel_init_blt_ring_buffer(dev);
                if (ret)
                        goto cleanup_bsd_ring;
        }

        ret = i915_gem_set_seqno(dev, ((u32)~0 - 0x1000));
        if (ret)
                goto cleanup_blt_ring;

        return 0;

cleanup_blt_ring:
        intel_cleanup_ring_buffer(&dev_priv->ring[BCS]);
cleanup_bsd_ring:
        intel_cleanup_ring_buffer(&dev_priv->ring[VCS]);
cleanup_render_ring:
        intel_cleanup_ring_buffer(&dev_priv->ring[RCS]);

        return ret;
}

int
i915_gem_init_hw(struct drm_device *dev)
{
        drm_i915_private_t *dev_priv = dev->dev_private;
        int ret;

        if (INTEL_INFO(dev)->gen < 6 && !intel_enable_gtt())
                return -EIO;