Intel-GFX Archive on lore.kernel.org
 help / color / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Subject: [PATCH 04/29] drm/i915: Try harder to allocate an mmap_offset
Date: Sat, 11 Aug 2012 15:41:03 +0100
Message-ID: <1344696088-24760-5-git-send-email-chris@chris-wilson.co.uk> (raw)
In-Reply-To: <1344696088-24760-1-git-send-email-chris@chris-wilson.co.uk>

Given the persistence of an offset for the lifetime of an object, itis
easy to contemplate how the mmap space becomes badly fragmented to the
point that further allocations fail with ENOSPC. Our only recourse at
this point is to try to purge the objects to release some space and
reattempt the allocation.

References: https://bugs.freedesktop.org/show_bug.cgi?id=39552
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem.c |   50 ++++++++++++++++++++++++++++++++-------
 1 file changed, 41 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 3a7ac38..0e0fc1e 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1276,6 +1276,42 @@ i915_gem_get_unfenced_gtt_alignment(struct drm_device *dev,
 	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;
+
+	ret = drm_gem_create_mmap_offset(&obj->base);
+	if (ret != -ENOSPC)
+		return ret;
+
+	/* 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)
+		return ret;
+
+	i915_gem_shrink_all(dev_priv);
+	return drm_gem_create_mmap_offset(&obj->base);
+}
+
+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,
@@ -1307,11 +1343,9 @@ i915_gem_mmap_gtt(struct drm_file *file,
 		goto out;
 	}
 
-	if (!obj->base.map_list.map) {
-		ret = drm_gem_create_mmap_offset(&obj->base);
-		if (ret)
-			goto out;
-	}
+	ret = i915_gem_object_create_mmap_offset(obj);
+	if (ret)
+		goto out;
 
 	*offset = (u64)obj->base.map_list.hash.key << PAGE_SHIFT;
 
@@ -1360,8 +1394,7 @@ i915_gem_object_truncate(struct drm_i915_gem_object *obj)
 	inode = obj->base.filp->f_path.dentry->d_inode;
 	shmem_truncate_range(inode, 0, (loff_t)-1);
 
-	if (obj->base.map_list.map)
-		drm_gem_free_mmap_offset(&obj->base);
+	i915_gem_object_free_mmap_offset(obj);
 
 	obj->madv = __I915_MADV_PURGED;
 }
@@ -3614,8 +3647,7 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj)
 	}
 
 	i915_gem_object_put_pages_gtt(obj);
-	if (obj->base.map_list.map)
-		drm_gem_free_mmap_offset(&obj->base);
+	i915_gem_object_free_mmap_offset(obj);
 
 	drm_gem_object_release(&obj->base);
 	i915_gem_info_remove_obj(dev_priv, obj->base.size);
-- 
1.7.10.4

  parent reply index

Thread overview: 51+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-08-11 14:40 Stolen pages, with a little surprise Chris Wilson
2012-08-11 14:41 ` [PATCH 01/29] drm/i915: Track unbound pages Chris Wilson
2012-08-20  9:00   ` [PATCH 1/2] drm/i915: move functions around Daniel Vetter
2012-08-20  9:00     ` [PATCH 2/2] drm/i915: Track unbound pages Daniel Vetter
2012-08-20  9:23       ` [PATCH] Add some sanity checks to unbound tracking Chris Wilson
2012-08-20  9:36       ` [PATCH 2/2] drm/i915: Track unbound pages Chris Wilson
2012-08-20  9:42         ` Daniel Vetter
2012-08-11 14:41 ` [PATCH 02/29] drm/i915: Show (count, size) of purgeable objects in i915_gem_objects Chris Wilson
2012-08-20  9:04   ` Daniel Vetter
2012-08-20  9:17     ` Chris Wilson
2012-08-11 14:41 ` [PATCH 03/29] drm/i915: Show pin count in debugfs Chris Wilson
2012-08-11 14:41 ` Chris Wilson [this message]
2012-08-20  9:37   ` [PATCH 04/29] drm/i915: Try harder to allocate an mmap_offset Daniel Vetter
2012-08-20 11:31     ` Chris Wilson
2012-08-11 14:41 ` [PATCH 05/29] drm/i915: Only pwrite through the GTT if there is space in the aperture Chris Wilson
2012-08-11 14:41 ` [PATCH 06/29] drm/i915: Protect private gem objects from truncate (such as imported dmabuf) Chris Wilson
2012-08-11 14:41 ` [PATCH 07/29] drm/i915: Extract general object init routine Chris Wilson
2012-08-24  0:05   ` Daniel Vetter
2012-08-11 14:41 ` [PATCH 08/29] drm/i915: Introduce drm_i915_gem_object_ops Chris Wilson
2012-08-20 19:35   ` Daniel Vetter
2012-08-11 14:41 ` [PATCH 09/29] drm/i915: Pin backing pages whilst exporting through a dmabuf vmap Chris Wilson
2012-08-11 14:41 ` [PATCH 10/29] drm/i915: Pin backing pages for pwrite Chris Wilson
2012-08-11 14:41 ` [PATCH 11/29] drm/i915: Pin backing pages for pread Chris Wilson
2012-08-11 14:41 ` [PATCH 12/29] drm/i915: Replace the array of pages with a scatterlist Chris Wilson
2012-08-11 14:41 ` [PATCH 13/29] drm/i915: Convert the dmabuf object to use the new i915_gem_object_ops Chris Wilson
2012-08-11 14:41 ` [PATCH 14/29] drm: Introduce drm_mm_create_block() Chris Wilson
2012-08-11 14:41 ` [PATCH 15/29] drm/i915: Fix detection of stolen base for gen2 Chris Wilson
2012-08-11 14:41 ` [PATCH 16/29] drm/i915: Fix location of stolen memory register for SandyBridge+ Chris Wilson
2012-08-20 19:38   ` Daniel Vetter
2012-08-22 15:54     ` Chris Wilson
2012-08-11 14:41 ` [PATCH 17/29] drm/i915: Avoid clearing preallocated regions from the GTT Chris Wilson
2012-08-11 14:41 ` [PATCH 18/29] drm/i915: Delay allocation of stolen space for FBC Chris Wilson
2012-08-20 19:51   ` Daniel Vetter
2012-08-22 15:51     ` Chris Wilson
2012-08-11 14:41 ` [PATCH 19/29] drm/i915: Allow objects to be created with no backing pages, but stolen space Chris Wilson
2012-08-11 14:41 ` [PATCH 20/29] drm/i915: Differentiate between prime and stolen objects Chris Wilson
2012-08-11 14:41 ` [PATCH 21/29] drm/i915: Support readback of stolen objects upon error Chris Wilson
2012-08-11 14:41 ` [PATCH 22/29] drm/i915: Handle stolen objects in pwrite Chris Wilson
2012-08-20 19:56   ` Daniel Vetter
2012-08-22 15:47     ` Chris Wilson
2012-08-30 15:09     ` Chris Wilson
2012-08-11 14:41 ` [PATCH 23/29] drm/i915: Handle stolen objects for pread Chris Wilson
2012-08-11 14:41 ` [PATCH 24/29] drm/i915: Introduce i915_gem_object_create_stolen() Chris Wilson
2012-08-11 14:41 ` [PATCH 25/29] drm/i915: Allocate fbcon from stolen memory Chris Wilson
2012-08-11 14:41 ` [PATCH 26/29] drm/i915: Allocate ringbuffers " Chris Wilson
2012-08-11 14:41 ` [PATCH 27/29] drm/i915: Allocate overlay registers " Chris Wilson
2012-08-20 21:17   ` Daniel Vetter
2012-08-22 15:45     ` Chris Wilson
2012-08-22 16:26       ` Daniel Vetter
2012-08-11 14:41 ` [PATCH 28/29] drm/i915: Use a slab for object allocation Chris Wilson
2012-08-11 14:41 ` [PATCH 29/29] drm/i915: Introduce mapping of user pages into video memory (userptr) ioctl Chris Wilson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1344696088-24760-5-git-send-email-chris@chris-wilson.co.uk \
    --to=chris@chris-wilson.co.uk \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Intel-GFX Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/intel-gfx/0 intel-gfx/git/0.git
	git clone --mirror https://lore.kernel.org/intel-gfx/1 intel-gfx/git/1.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 intel-gfx intel-gfx/ https://lore.kernel.org/intel-gfx \
		intel-gfx@lists.freedesktop.org
	public-inbox-index intel-gfx

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.freedesktop.lists.intel-gfx


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git