#define I915_GEM_GPU_DOMAINS (~(I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT))
-static void
-i915_gem_object_set_to_gpu_domain(struct drm_gem_object *obj,
- uint32_t read_domains,
- uint32_t write_domain);
static void i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj);
static void i915_gem_object_flush_gtt_write_domain(struct drm_gem_object *obj);
static void i915_gem_object_flush_cpu_write_domain(struct drm_gem_object *obj);
static int i915_gem_object_wait_rendering(struct drm_gem_object *obj);
static int i915_gem_object_bind_to_gtt(struct drm_gem_object *obj,
unsigned alignment);
-static void i915_gem_object_get_fence_reg(struct drm_gem_object *obj);
+static int i915_gem_object_get_fence_reg(struct drm_gem_object *obj, bool write);
static void i915_gem_clear_fence_reg(struct drm_gem_object *obj);
static int i915_gem_evict_something(struct drm_device *dev);
static int i915_gem_phys_pwrite(struct drm_device *dev, struct drm_gem_object *obj,
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) >>
/* Need a new fence register? */
if (obj_priv->fence_reg == I915_FENCE_REG_NONE &&
- obj_priv->tiling_mode != I915_TILING_NONE)
- i915_gem_object_get_fence_reg(obj);
+ obj_priv->tiling_mode != I915_TILING_NONE) {
+ ret = i915_gem_object_get_fence_reg(obj, write);
+ if (ret) {
+ mutex_unlock(&dev->struct_mutex);
+ return VM_FAULT_SIGBUS;
+ }
+ }
pfn = ((dev->agp->base + obj_priv->gtt_offset) >> PAGE_SHIFT) +
page_offset;
case -EAGAIN:
return VM_FAULT_OOM;
case -EFAULT:
- case -EBUSY:
- DRM_ERROR("can't insert pfn?? fault or busy...\n");
return VM_FAULT_SIGBUS;
default:
return VM_FAULT_NOPAGE;
return ret;
}
+static void
+i915_gem_free_mmap_offset(struct drm_gem_object *obj)
+{
+ struct drm_device *dev = obj->dev;
+ struct drm_i915_gem_object *obj_priv = obj->driver_private;
+ struct drm_gem_mm *mm = dev->mm_private;
+ struct drm_map_list *list;
+
+ list = &obj->map_list;
+ drm_ht_remove_item(&mm->offset_hash, &list->hash);
+
+ if (list->file_offset_node) {
+ drm_mm_put_block(list->file_offset_node);
+ list->file_offset_node = NULL;
+ }
+
+ if (list->map) {
+ drm_free(list->map, sizeof(struct drm_map), DRM_MEM_DRIVER);
+ list->map = NULL;
+ }
+
+ obj_priv->mmap_offset = 0;
+}
+
/**
* i915_gem_get_gtt_alignment - return required GTT alignment for an object
* @obj: object to check
if (!obj_priv->mmap_offset) {
ret = i915_gem_create_mmap_offset(obj);
- if (ret)
+ if (ret) {
+ drm_gem_object_unreference(obj);
+ mutex_unlock(&dev->struct_mutex);
return ret;
+ }
}
args->offset = obj_priv->mmap_offset;
drm_i915_private_t *dev_priv = dev->dev_private;
uint32_t seqno;
+ if (!dev_priv->hw_status_page)
+ return;
+
seqno = i915_get_gem_seqno(dev);
while (!list_empty(&dev_priv->mm.request_list)) {
/**
* Unbinds an object from the GTT aperture.
*/
-static int
+int
i915_gem_object_unbind(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
struct drm_i915_gem_object *obj_priv = obj->driver_private;
int regnum = obj_priv->fence_reg;
+ int tile_width;
uint32_t val;
uint32_t pitch_val;
if ((obj_priv->gtt_offset & ~I915_FENCE_START_MASK) ||
(obj_priv->gtt_offset & (obj->size - 1))) {
- WARN(1, "%s: object not 1M or size aligned\n", __func__);
+ WARN(1, "%s: object 0x%08x not 1M or size (0x%zx) aligned\n",
+ __func__, obj_priv->gtt_offset, obj->size);
return;
}
- if (obj_priv->tiling_mode == I915_TILING_Y && (IS_I945G(dev) ||
- IS_I945GM(dev) ||
- IS_G33(dev)))
- pitch_val = (obj_priv->stride / 128) - 1;
+ if (obj_priv->tiling_mode == I915_TILING_Y &&
+ HAS_128_BYTE_Y_TILING(dev))
+ tile_width = 128;
else
- pitch_val = (obj_priv->stride / 512) - 1;
+ tile_width = 512;
+
+ /* Note: pitch better be a power of two tile widths */
+ pitch_val = obj_priv->stride / tile_width;
+ pitch_val = ffs(pitch_val) - 1;
val = obj_priv->gtt_offset;
if (obj_priv->tiling_mode == I915_TILING_Y)
if ((obj_priv->gtt_offset & ~I915_FENCE_START_MASK) ||
(obj_priv->gtt_offset & (obj->size - 1))) {
- WARN(1, "%s: object not 1M or size aligned\n", __func__);
+ WARN(1, "%s: object 0x%08x not 1M or size aligned\n",
+ __func__, obj_priv->gtt_offset);
return;
}
/**
* i915_gem_object_get_fence_reg - set up a fence reg for an object
* @obj: object to map through a fence reg
+ * @write: object is about to be written
*
* 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.
* It then sets up the reg based on the object's properties: address, pitch
* and tiling format.
*/
-static void
-i915_gem_object_get_fence_reg(struct drm_gem_object *obj)
+static int
+i915_gem_object_get_fence_reg(struct drm_gem_object *obj, bool write)
{
struct drm_device *dev = obj->dev;
struct drm_i915_private *dev_priv = dev->dev_private;
WARN(1, "allocating a fence for non-tiled object?\n");
break;
case I915_TILING_X:
- WARN(obj_priv->stride & (512 - 1),
- "object is X tiled but has non-512B pitch\n");
+ if (!obj_priv->stride)
+ return -EINVAL;
+ WARN((obj_priv->stride & (512 - 1)),
+ "object 0x%08x is X tiled but has non-512B pitch\n",
+ obj_priv->gtt_offset);
break;
case I915_TILING_Y:
- WARN(obj_priv->stride & (128 - 1),
- "object is Y tiled but has non-128B pitch\n");
+ if (!obj_priv->stride)
+ return -EINVAL;
+ WARN((obj_priv->stride & (128 - 1)),
+ "object 0x%08x is Y tiled but has non-128B pitch\n",
+ obj_priv->gtt_offset);
break;
}
/* First try to find a free reg */
+try_again:
for (i = dev_priv->fence_reg_start; i < dev_priv->num_fence_regs; i++) {
reg = &dev_priv->fence_regs[i];
if (!reg->obj)
struct drm_i915_gem_object *old_obj_priv = NULL;
loff_t offset;
-try_again:
/* Could try to use LRU here instead... */
for (i = dev_priv->fence_reg_start;
i < dev_priv->num_fence_regs; i++) {
* objects to finish before trying again.
*/
if (i == dev_priv->num_fence_regs) {
- ret = i915_gem_object_wait_rendering(reg->obj);
+ ret = i915_gem_object_set_to_gtt_domain(reg->obj, 0);
if (ret) {
- WARN(ret, "wait_rendering failed: %d\n", ret);
- return;
+ WARN(ret != -ERESTARTSYS,
+ "switch to GTT domain failed: %d\n", ret);
+ return ret;
}
goto try_again;
}
i915_write_fence_reg(reg);
else
i830_write_fence_reg(reg);
+
+ return 0;
}
/**
if (dev_priv->mm.suspended)
return -EBUSY;
if (alignment == 0)
- alignment = PAGE_SIZE;
+ alignment = i915_gem_get_gtt_alignment(obj);
if (alignment & (PAGE_SIZE - 1)) {
DRM_ERROR("Invalid object alignment requested %u\n", alignment);
return -EINVAL;
* drm_agp_chipset_flush
*/
static void
-i915_gem_object_set_to_gpu_domain(struct drm_gem_object *obj,
- uint32_t read_domains,
- uint32_t write_domain)
+i915_gem_object_set_to_gpu_domain(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct drm_i915_gem_object *obj_priv = obj->driver_private;
uint32_t invalidate_domains = 0;
uint32_t flush_domains = 0;
- BUG_ON(read_domains & I915_GEM_DOMAIN_CPU);
- BUG_ON(write_domain == I915_GEM_DOMAIN_CPU);
+ BUG_ON(obj->pending_read_domains & I915_GEM_DOMAIN_CPU);
+ BUG_ON(obj->pending_write_domain == I915_GEM_DOMAIN_CPU);
#if WATCH_BUF
DRM_INFO("%s: object %p read %08x -> %08x write %08x -> %08x\n",
__func__, obj,
- obj->read_domains, read_domains,
- obj->write_domain, write_domain);
+ obj->read_domains, obj->pending_read_domains,
+ obj->write_domain, obj->pending_write_domain);
#endif
/*
* If the object isn't moving to a new write domain,
* let the object stay in multiple read domains
*/
- if (write_domain == 0)
- read_domains |= obj->read_domains;
+ if (obj->pending_write_domain == 0)
+ obj->pending_read_domains |= obj->read_domains;
else
obj_priv->dirty = 1;
* any read domains which differ from the old
* write domain
*/
- if (obj->write_domain && obj->write_domain != read_domains) {
+ if (obj->write_domain &&
+ obj->write_domain != obj->pending_read_domains) {
flush_domains |= obj->write_domain;
- invalidate_domains |= read_domains & ~obj->write_domain;
+ invalidate_domains |=
+ obj->pending_read_domains & ~obj->write_domain;
}
/*
* Invalidate any read caches which may have
* stale data. That is, any new read domains.
*/
- invalidate_domains |= read_domains & ~obj->read_domains;
+ invalidate_domains |= obj->pending_read_domains & ~obj->read_domains;
if ((flush_domains | invalidate_domains) & I915_GEM_DOMAIN_CPU) {
#if WATCH_BUF
DRM_INFO("%s: CPU domain flush %08x invalidate %08x\n",
i915_gem_clflush_object(obj);
}
- if ((write_domain | flush_domains) != 0)
- obj->write_domain = write_domain;
- obj->read_domains = read_domains;
+ /* The actual obj->write_domain will be updated with
+ * pending_write_domain after we emit the accumulated flush for all
+ * of our domain changes in execbuffers (which clears objects'
+ * write_domains). So if we have a current write domain that we
+ * aren't changing, set pending_write_domain to that.
+ */
+ if (flush_domains == 0 && obj->pending_write_domain == 0)
+ obj->pending_write_domain = obj->write_domain;
+ obj->read_domains = obj->pending_read_domains;
dev->invalidate_domains |= invalidate_domains;
dev->flush_domains |= flush_domains;
(int) reloc.offset,
reloc.read_domains,
reloc.write_domain);
+ drm_gem_object_unreference(target_obj);
+ i915_gem_object_unpin(obj);
return -EINVAL;
}
struct drm_i915_gem_exec_object *exec_list = NULL;
struct drm_gem_object **object_list = NULL;
struct drm_gem_object *batch_obj;
+ struct drm_i915_gem_object *obj_priv;
int ret, i, pinned = 0;
uint64_t exec_offset;
uint32_t seqno, flush_domains;
if (dev_priv->mm.wedged) {
DRM_ERROR("Execbuf while wedged\n");
mutex_unlock(&dev->struct_mutex);
- return -EIO;
+ ret = -EIO;
+ goto pre_mutex_err;
}
if (dev_priv->mm.suspended) {
DRM_ERROR("Execbuf while VT-switched.\n");
mutex_unlock(&dev->struct_mutex);
- return -EBUSY;
+ ret = -EBUSY;
+ goto pre_mutex_err;
}
/* Look up object handles */
ret = -EBADF;
goto err;
}
+
+ obj_priv = object_list[i]->driver_private;
+ if (obj_priv->in_execbuffer) {
+ DRM_ERROR("Object %p appears more than once in object list\n",
+ object_list[i]);
+ ret = -EBADF;
+ goto err;
+ }
+ obj_priv->in_execbuffer = true;
}
/* Pin and relocate */
struct drm_gem_object *obj = object_list[i];
/* Compute new gpu domains and update invalidate/flush */
- i915_gem_object_set_to_gpu_domain(obj,
- obj->pending_read_domains,
- obj->pending_write_domain);
+ i915_gem_object_set_to_gpu_domain(obj);
}
i915_verify_inactive(dev, __FILE__, __LINE__);
(void)i915_add_request(dev, dev->flush_domains);
}
+ for (i = 0; i < args->buffer_count; i++) {
+ struct drm_gem_object *obj = object_list[i];
+
+ obj->write_domain = obj->pending_write_domain;
+ }
+
i915_verify_inactive(dev, __FILE__, __LINE__);
#if WATCH_COHERENCY
i915_verify_inactive(dev, __FILE__, __LINE__);
- /* Copy the new buffer offsets back to the user's exec list. */
- ret = copy_to_user((struct drm_i915_relocation_entry __user *)
- (uintptr_t) args->buffers_ptr,
- exec_list,
- sizeof(*exec_list) * args->buffer_count);
- if (ret)
- DRM_ERROR("failed to copy %d exec entries "
- "back to user (%d)\n",
- args->buffer_count, ret);
err:
for (i = 0; i < pinned; i++)
i915_gem_object_unpin(object_list[i]);
- for (i = 0; i < args->buffer_count; i++)
+ for (i = 0; i < args->buffer_count; i++) {
+ if (object_list[i]) {
+ obj_priv = object_list[i]->driver_private;
+ obj_priv->in_execbuffer = false;
+ }
drm_gem_object_unreference(object_list[i]);
+ }
mutex_unlock(&dev->struct_mutex);
+ if (!ret) {
+ /* Copy the new buffer offsets back to the user's exec list. */
+ ret = copy_to_user((struct drm_i915_relocation_entry __user *)
+ (uintptr_t) args->buffers_ptr,
+ exec_list,
+ sizeof(*exec_list) * args->buffer_count);
+ if (ret)
+ DRM_ERROR("failed to copy %d exec entries "
+ "back to user (%d)\n",
+ args->buffer_count, ret);
+ }
+
pre_mutex_err:
drm_free(object_list, sizeof(*object_list) * args->buffer_count,
DRM_MEM_DRIVER);
ret = i915_gem_object_bind_to_gtt(obj, alignment);
if (ret != 0) {
if (ret != -EBUSY && ret != -ERESTARTSYS)
- DRM_ERROR("Failure to bind: %d", ret);
+ DRM_ERROR("Failure to bind: %d\n", ret);
return ret;
}
+ /*
+ * Pre-965 chips need a fence register set up in order to
+ * properly handle tiled surfaces.
+ */
+ if (!IS_I965G(dev) &&
+ obj_priv->fence_reg == I915_FENCE_REG_NONE &&
+ obj_priv->tiling_mode != I915_TILING_NONE)
+ i915_gem_object_get_fence_reg(obj, true);
}
obj_priv->pin_count++;
if (obj_priv->pin_filp != NULL && obj_priv->pin_filp != file_priv) {
DRM_ERROR("Already pinned in i915_gem_pin_ioctl(): %d\n",
args->handle);
+ drm_gem_object_unreference(obj);
mutex_unlock(&dev->struct_mutex);
return -EINVAL;
}
return -EBADF;
}
+ /* Update the active list for the hardware's current position.
+ * Otherwise this only updates on a delayed timer or when irqs are
+ * actually unmasked, and our working set ends up being larger than
+ * required.
+ */
+ i915_gem_retire_requests(dev);
+
obj_priv = obj->driver_private;
/* Don't count being on the flushing list against the object being
* done. Otherwise, a buffer left on the flushing list but not getting
void i915_gem_free_object(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
- struct drm_gem_mm *mm = dev->mm_private;
- struct drm_map_list *list;
- struct drm_map *map;
struct drm_i915_gem_object *obj_priv = obj->driver_private;
while (obj_priv->pin_count > 0)
i915_gem_object_unbind(obj);
- list = &obj->map_list;
- drm_ht_remove_item(&mm->offset_hash, &list->hash);
-
- if (list->file_offset_node) {
- drm_mm_put_block(list->file_offset_node);
- list->file_offset_node = NULL;
- }
-
- map = list->map;
- if (map) {
- drm_free(map, sizeof(*map), DRM_MEM_DRIVER);
- list->map = NULL;
- }
+ i915_gem_free_mmap_offset(obj);
drm_free(obj_priv->page_cpu_valid, 1, DRM_MEM_DRIVER);
drm_free(obj->driver_private, 1, DRM_MEM_DRIVER);
return 0;
}
-static int
+int
i915_gem_idle(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
if (dev_priv->hw_status_page == NULL) {
DRM_ERROR("Failed to map status page.\n");
memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
+ i915_gem_object_unpin(obj);
drm_gem_object_unreference(obj);
return -EINVAL;
}
return 0;
}
+static void
+i915_gem_cleanup_hws(struct drm_device *dev)
+{
+ drm_i915_private_t *dev_priv = dev->dev_private;
+ struct drm_gem_object *obj;
+ struct drm_i915_gem_object *obj_priv;
+
+ if (dev_priv->hws_obj == NULL)
+ return;
+
+ obj = dev_priv->hws_obj;
+ obj_priv = obj->driver_private;
+
+ kunmap(obj_priv->page_list[0]);
+ i915_gem_object_unpin(obj);
+ drm_gem_object_unreference(obj);
+ dev_priv->hws_obj = NULL;
+
+ memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
+ dev_priv->hw_status_page = NULL;
+
+ /* Write high address into HWS_PGA when disabling. */
+ I915_WRITE(HWS_PGA, 0x1ffff000);
+}
+
int
i915_gem_init_ringbuffer(struct drm_device *dev)
{
obj = drm_gem_object_alloc(dev, 128 * 1024);
if (obj == NULL) {
DRM_ERROR("Failed to allocate ringbuffer\n");
+ i915_gem_cleanup_hws(dev);
return -ENOMEM;
}
obj_priv = obj->driver_private;
ret = i915_gem_object_pin(obj, 4096);
if (ret != 0) {
drm_gem_object_unreference(obj);
+ i915_gem_cleanup_hws(dev);
return ret;
}
if (ring->map.handle == NULL) {
DRM_ERROR("Failed to map ringbuffer.\n");
memset(&dev_priv->ring, 0, sizeof(dev_priv->ring));
+ i915_gem_object_unpin(obj);
drm_gem_object_unreference(obj);
+ i915_gem_cleanup_hws(dev);
return -EINVAL;
}
ring->ring_obj = obj;
dev_priv->ring.ring_obj = NULL;
memset(&dev_priv->ring, 0, sizeof(dev_priv->ring));
- if (dev_priv->hws_obj != NULL) {
- struct drm_gem_object *obj = dev_priv->hws_obj;
- struct drm_i915_gem_object *obj_priv = obj->driver_private;
-
- kunmap(obj_priv->page_list[0]);
- i915_gem_object_unpin(obj);
- drm_gem_object_unreference(obj);
- dev_priv->hws_obj = NULL;
- memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
- dev_priv->hw_status_page = NULL;
-
- /* Write high address into HWS_PGA when disabling. */
- I915_WRITE(HWS_PGA, 0x1ffff000);
- }
+ i915_gem_cleanup_hws(dev);
}
int
dev_priv->mm.wedged = 0;
}
- dev_priv->mm.gtt_mapping = io_mapping_create_wc(dev->agp->base,
- dev->agp->agp_info.aper_size
- * 1024 * 1024);
-
mutex_lock(&dev->struct_mutex);
dev_priv->mm.suspended = 0;
i915_gem_leavevt_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
- drm_i915_private_t *dev_priv = dev->dev_private;
int ret;
if (drm_core_check_feature(dev, DRIVER_MODESET))
ret = i915_gem_idle(dev);
drm_irq_uninstall(dev);
- io_mapping_free(dev_priv->mm.gtt_mapping);
return ret;
}
{
int ret;
+ if (drm_core_check_feature(dev, DRIVER_MODESET))
+ return;
+
ret = i915_gem_idle(dev);
if (ret)
DRM_ERROR("failed to idle hardware: %d\n", ret);
/* Old X drivers will take 0-2 for front, back, depth buffers */
dev_priv->fence_reg_start = 3;
- if (IS_I965G(dev))
+ if (IS_I965G(dev) || IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev))
dev_priv->num_fence_regs = 16;
else
dev_priv->num_fence_regs = 8;
{
int i;
- for (i = 0; i < I915_MAX_PHYS_OBJECT; i++)
+ for (i = I915_GEM_PHYS_CURSOR_0; i <= I915_MAX_PHYS_OBJECT; i++)
i915_gem_free_phys_object(dev, i);
}
ret = i915_gem_init_phys_object(dev, id,
obj->size);
if (ret) {
- DRM_ERROR("failed to init phys object %d size: %d\n", id, obj->size);
+ DRM_ERROR("failed to init phys object %d size: %zu\n", id, obj->size);
goto out;
}
}
user_data = (char __user *) (uintptr_t) args->data_ptr;
obj_addr = obj_priv->phys_obj->handle->vaddr + args->offset;
- DRM_ERROR("obj_addr %p, %lld\n", obj_addr, args->size);
+ DRM_DEBUG("obj_addr %p, %lld\n", obj_addr, args->size);
ret = copy_from_user(obj_addr, user_data, args->size);
if (ret)
return -EFAULT;
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff