All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Subject: [PATCH 2/3] drm/i915: Only revoke the partial vma when releasing a fence register
Date: Fri, 23 Jun 2017 14:27:35 +0100	[thread overview]
Message-ID: <20170623132736.32655-2-chris@chris-wilson.co.uk> (raw)
In-Reply-To: <20170623132736.32655-1-chris@chris-wilson.co.uk>

As we may be using a partial vma for fence allocation, rather then
revoke the mmap of all vma on the object only revoke the overlapping
mmap for this fence.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_fence_reg.c |  4 ++--
 drivers/gpu/drm/i915/i915_vma.c           | 25 +++++++++++++++++++++++--
 drivers/gpu/drm/i915/i915_vma.h           |  1 +
 3 files changed, 26 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
index 52d5696f9e49..c5cd4dcc54ba 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
@@ -240,7 +240,7 @@ static int fence_update(struct drm_i915_fence_reg *fence,
 		/* Ensure that all userspace CPU access is completed before
 		 * stealing the fence.
 		 */
-		i915_gem_release_mmap(fence->vma->obj);
+		i915_vma_revoke_mmap(fence->vma);
 
 		fence->vma->fence = NULL;
 		fence->vma = NULL;
@@ -379,7 +379,7 @@ void i915_gem_revoke_fences(struct drm_i915_private *dev_priv)
 		struct drm_i915_fence_reg *fence = &dev_priv->fence_regs[i];
 
 		if (fence->vma)
-			i915_gem_release_mmap(fence->vma->obj);
+			i915_vma_revoke_mmap(fence->vma);
 	}
 }
 
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index fd671d493ca6..c4817ad7c965 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -642,6 +642,28 @@ static void __i915_vma_iounmap(struct i915_vma *vma)
 	vma->iomap = NULL;
 }
 
+void i915_vma_revoke_mmap(struct i915_vma *vma)
+{
+	struct drm_vma_offset_node *node = &vma->obj->base.vma_node;
+
+	lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+
+	if (!i915_vma_has_userfault(vma))
+		return;
+
+	GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma));
+	GEM_BUG_ON(list_empty(&vma->obj->userfault_link));
+
+	unmap_mapping_range(vma->vm->i915->drm.anon_inode->i_mapping,
+			    drm_vma_node_offset_addr(node) + vma->ggtt_view.partial.offset,
+			    vma->size,
+			    1);
+	vma->flags &= ~I915_VMA_USERFAULT;
+
+	if (vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL)
+		list_del_init(&vma->obj->userfault_link);
+}
+
 int i915_vma_unbind(struct i915_vma *vma)
 {
 	struct drm_i915_gem_object *obj = vma->obj;
@@ -705,8 +727,7 @@ int i915_vma_unbind(struct i915_vma *vma)
 			return ret;
 
 		/* Force a pagefault for domain tracking on next user access */
-		if (i915_vma_has_userfault(vma))
-			i915_gem_release_mmap(obj);
+		i915_vma_revoke_mmap(vma);
 
 		__i915_vma_iounmap(vma);
 		vma->flags &= ~I915_VMA_CAN_FENCE;
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 78afbac2bca8..218f9d48e7c6 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -247,6 +247,7 @@ bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level);
 bool i915_vma_misplaced(const struct i915_vma *vma,
 			u64 size, u64 alignment, u64 flags);
 void __i915_vma_set_map_and_fenceable(struct i915_vma *vma);
+void i915_vma_revoke_mmap(struct i915_vma *vma);
 int __must_check i915_vma_unbind(struct i915_vma *vma);
 void i915_vma_unlink_ctx(struct i915_vma *vma);
 void i915_vma_close(struct i915_vma *vma);
-- 
2.13.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  reply	other threads:[~2017-06-23 13:27 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-06-23 13:27 [PATCH 1/3] drm/i915: Track user GTT faulting per-vma Chris Wilson
2017-06-23 13:27 ` Chris Wilson [this message]
2017-06-23 13:27 ` [PATCH 3/3] drm/i915: Try a minimal attempt to insert the whole object for relocations Chris Wilson
2017-06-23 13:54 ` ✓ Fi.CI.BAT: success for series starting with [1/3] drm/i915: Track user GTT faulting per-vma Patchwork

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=20170623132736.32655-2-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
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.