All of lore.kernel.org
 help / color / mirror / Atom feed
From: Daniel Vetter <daniel.vetter-/w4YWyX8dFk@public.gmane.org>
To: DRI Development
	<dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>
Cc: Nouveau Dev
	<nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>,
	Intel Graphics Development
	<intel-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org>,
	Radeon Dev
	<xorg-driver-ati-go0+a7rfsptAfugRpC6u6w@public.gmane.org>,
	Daniel Vetter <daniel.vetter-/w4YWyX8dFk@public.gmane.org>
Subject: [PATCH 14/37] drm: only take the crtc lock for ->cursor_set
Date: Wed, 12 Dec 2012 14:06:54 +0100	[thread overview]
Message-ID: <1355317637-16742-15-git-send-email-daniel.vetter@ffwll.ch> (raw)
In-Reply-To: <1355317637-16742-1-git-send-email-daniel.vetter-/w4YWyX8dFk@public.gmane.org>

First convert ->cursor_set to only take the crtc lock, since that
seems to be the function with the least amount of state - the core
ioctl function doesn't check anything which can change at runtime, so
we don't have any object lifetime issues to contend.

The only thing which is important is that the driver's implementation
doesn't touch any state outside of that single crtc which is not yet
properly protected by other locking:

- ast: access the global ast->cache_kmap. Luckily we only have on crtc
  on this driver, so this is fine. Add a comment.

- gma500: calls gma_power_begin|and and psb_gtt_pin|unpin, both which
  have their own locking to protect their state. Everything else is
  crtc-local.

- i915: touches a bit of global gem state, all protected by the One
  Lock to Rule Them All (dev->struct_mutex).

- nouveau: Pre-nv50 is all nice, nv50+ uses the evo channels to queue
  up all display changes. And some of these channels are device
  global. But this is fine now since the previous patch introduced an
  evo channel mutex.

- radeon: Uses some indirect register access for cursor updates, but
  with the previous patches to protect these indirect 2-register
  access patterns with a spinlock, this should be fine now, too.

- vmwgfx: I have no idea how that works - update_cursor_position
  doesn't take any per-crtc argument and I haven't figured out any
  other place where this could be set in some form of a side-channel.
  But vmwgfx definitely has more than one crtc (or at least can
  register more than one), so I have no idea how this is supposed to
  not fail with the current code already. Hence take the easy way out
  and simply acquire all locks (which requires dropping the crtc lock
  the core acquired for us). That way it's not worse off for
  consistency than the old code.

Signed-off-by: Daniel Vetter <daniel.vetter-/w4YWyX8dFk@public.gmane.org>
---
 drivers/gpu/drm/ast/ast_drv.h       |    2 ++
 drivers/gpu/drm/drm_crtc.c          |    6 ++++--
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c |   32 ++++++++++++++++++++++++++------
 3 files changed, 32 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/ast/ast_drv.h b/drivers/gpu/drm/ast/ast_drv.h
index 5ccf984..5284292 100644
--- a/drivers/gpu/drm/ast/ast_drv.h
+++ b/drivers/gpu/drm/ast/ast_drv.h
@@ -98,6 +98,8 @@ struct ast_private {
 
 	struct drm_gem_object *cursor_cache;
 	uint64_t cursor_cache_gpu_addr;
+	/* Acces to this cache is protected by the crtc->mutex of the only crtc
+	 * we have. */
 	struct ttm_bo_kmap_obj cache_kmap;
 	int next_cursor;
 };
diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
index 91e8068..62b5002 100644
--- a/drivers/gpu/drm/drm_crtc.c
+++ b/drivers/gpu/drm/drm_crtc.c
@@ -2023,7 +2023,6 @@ int drm_mode_cursor_ioctl(struct drm_device *dev,
 	if (!req->flags || (~DRM_MODE_CURSOR_FLAGS & req->flags))
 		return -EINVAL;
 
-	drm_modeset_lock_all(dev);
 	obj = drm_mode_object_find(dev, req->crtc_id, DRM_MODE_OBJECT_CRTC);
 	if (!obj) {
 		DRM_DEBUG_KMS("Unknown CRTC ID %d\n", req->crtc_id);
@@ -2038,20 +2037,23 @@ int drm_mode_cursor_ioctl(struct drm_device *dev,
 			goto out;
 		}
 		/* Turns off the cursor if handle is 0 */
+		mutex_lock(&crtc->mutex);
 		ret = crtc->funcs->cursor_set(crtc, file_priv, req->handle,
 					      req->width, req->height);
+		mutex_unlock(&crtc->mutex);
 	}
 
 	if (req->flags & DRM_MODE_CURSOR_MOVE) {
 		if (crtc->funcs->cursor_move) {
+			drm_modeset_lock_all(dev);
 			ret = crtc->funcs->cursor_move(crtc, req->x, req->y);
+			drm_modeset_unlock_all(dev);
 		} else {
 			ret = -EFAULT;
 			goto out;
 		}
 	}
 out:
-	drm_modeset_unlock_all(dev);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
index 5474394..74b6734 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -180,16 +180,29 @@ int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
 	struct vmw_dma_buffer *dmabuf = NULL;
 	int ret;
 
+	/*
+	 * FIXME: Unclear whether there's any global state touched by the
+	 * cursor_set function, especially vmw_cursor_update_position looks
+	 * suspicious. For now take the easy route and reacquire all locks. We
+	 * can do this since the caller in the drm core doesn't check anything
+	 * which is protected by any looks.
+	 */
+	mutex_unlock(&crtc->mutex);
+	drm_modeset_lock_all(dev_priv->dev);
+
 	/* A lot of the code assumes this */
-	if (handle && (width != 64 || height != 64))
-		return -EINVAL;
+	if (handle && (width != 64 || height != 64)) {
+		ret = -EINVAL;
+		goto out;
+	}
 
 	if (handle) {
 		ret = vmw_user_lookup_handle(dev_priv, tfile,
 					     handle, &surface, &dmabuf);
 		if (ret) {
 			DRM_ERROR("failed to find surface or dmabuf: %i\n", ret);
-			return -EINVAL;
+			ret = -EINVAL;
+			goto out;
 		}
 	}
 
@@ -197,7 +210,8 @@ int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
 	if (surface && !surface->snooper.image) {
 		DRM_ERROR("surface not suitable for cursor\n");
 		vmw_surface_unreference(&surface);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out;
 	}
 
 	/* takedown old cursor */
@@ -225,14 +239,20 @@ int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
 					       du->hotspot_x, du->hotspot_y);
 	} else {
 		vmw_cursor_update_position(dev_priv, false, 0, 0);
-		return 0;
+		ret = 0;
+		goto out;
 	}
 
 	vmw_cursor_update_position(dev_priv, true,
 				   du->cursor_x + du->hotspot_x,
 				   du->cursor_y + du->hotspot_y);
 
-	return 0;
+	ret = 0;
+out:
+	drm_modeset_unlock_all(dev_priv->dev);
+	mutex_lock(&crtc->mutex);
+
+	return ret;
 }
 
 int vmw_du_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
-- 
1.7.10.4

  parent reply	other threads:[~2012-12-12 13:06 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-12-12 13:06 [PATCH 00/37] [RFC] revamped modeset locking Daniel Vetter
2012-12-12 13:06 ` [PATCH 01/37] drm: review locking rules in drm_crtc.c Daniel Vetter
2012-12-12 13:06 ` [PATCH 05/37] drm/i915: use drm_modeset_lock_all Daniel Vetter
2012-12-12 13:06 ` [PATCH 08/37] drm/shmobile: " Daniel Vetter
2012-12-12 13:06 ` [PATCH 10/37] drm: add per-crtc locks Daniel Vetter
2012-12-13 11:38   ` [Intel-gfx] " Ville Syrjälä
2012-12-13 11:54     ` Daniel Vetter
2012-12-13 14:25       ` Ville Syrjälä
2012-12-12 13:06 ` [PATCH 11/37] drm/radeon: add W|RREG32_IDX for MM_INDEX|DATA based mmio accesss Daniel Vetter
2012-12-12 13:07 ` [PATCH 25/37] drm: don't take modeset locks in getfb ioctl Daniel Vetter
2012-12-12 13:07 ` [PATCH 26/37] drm: fb refcounting for dirtyfb_ioctl Daniel Vetter
2012-12-12 13:07 ` [PATCH 31/37] drm/vmwgfx: add proper framebuffer refcounting Daniel Vetter
2012-12-12 13:07 ` [PATCH 33/37] drm/nouveau: try to protect nbo->pin_refcount Daniel Vetter
     [not found] ` <1355317637-16742-1-git-send-email-daniel.vetter-/w4YWyX8dFk@public.gmane.org>
2012-12-12 13:06   ` [PATCH 02/37] drm/doc: integrate drm_crtc.c kerneldoc Daniel Vetter
2012-12-12 13:06   ` [PATCH 03/37] drm: add drm_modeset_lock|unlock_all Daniel Vetter
2012-12-12 13:06   ` [PATCH 04/37] drm/i915: rework locking for intel_dpio|sbi_read|write Daniel Vetter
2012-12-12 20:54     ` [Intel-gfx] " Jesse Barnes
2012-12-12 22:00       ` Daniel Vetter
2012-12-12 22:05         ` Jesse Barnes
2012-12-12 13:06   ` [PATCH 06/37] drm/gma500: use drm_modeset_lock_all Daniel Vetter
2012-12-12 13:06   ` [PATCH 07/37] drm/ast: " Daniel Vetter
2012-12-12 13:06   ` [PATCH 09/37] drm/vmgfx: " Daniel Vetter
2012-12-12 13:06   ` [PATCH 12/37] drm/radeon: make indirect register access concurrency-safe Daniel Vetter
2012-12-12 13:06   ` [PATCH 13/37] drm/nouveau: protect evo_wait/evo_kick sections with a channel mutex Daniel Vetter
2012-12-12 13:06   ` Daniel Vetter [this message]
2012-12-12 13:06   ` [PATCH 15/37] drm: only take the crtc lock for ->cursor_move Daniel Vetter
2012-12-13 11:03     ` [PATCH] " Daniel Vetter
2012-12-12 13:06   ` [PATCH 16/37] drm/<drivers>: reorder framebuffer init sequence Daniel Vetter
2012-12-13 11:05     ` [PATCH 1/2] " Daniel Vetter
     [not found]       ` <1355396719-25286-1-git-send-email-daniel.vetter-/w4YWyX8dFk@public.gmane.org>
2012-12-13 11:05         ` [PATCH 2/2] drm/exynos: " Daniel Vetter
2012-12-13 12:26           ` Inki Dae
2012-12-13 12:43             ` Daniel Vetter
2012-12-13 15:16               ` Inki Dae
2012-12-13 16:38                 ` Daniel Vetter
2012-12-14  4:57                   ` Inki Dae
2012-12-14  8:40                     ` Daniel Vetter
2012-12-12 13:06   ` [PATCH 17/37] drm: revamp locking around fb creation/destruction Daniel Vetter
2012-12-12 13:06   ` [PATCH 18/37] drm: create drm_framebuffer_lookup Daniel Vetter
2012-12-12 13:06   ` [PATCH 19/37] drm/gma500: move fbcon restore to lastclose Daniel Vetter
2012-12-12 13:07   ` [PATCH 20/37] drm: revamp framebuffer cleanup interfaces Daniel Vetter
2012-12-12 13:07   ` [PATCH 21/37] drm: reference framebuffers which are on the idr Daniel Vetter
2012-12-12 13:07   ` [PATCH 22/37] drm: nest modeset locks within fpriv->fbs_lock Daniel Vetter
2012-12-12 13:07   ` [PATCH 23/37] drm/i915: fixup overlay stolen memory leak Daniel Vetter
2012-12-12 13:07   ` [PATCH 24/37] drm: push modeset_lock_all into ->fb_create driver callbacks Daniel Vetter
2012-12-12 13:07   ` [PATCH 27/37] drm: refcounting for sprite framebuffers Daniel Vetter
2012-12-12 13:07   ` [PATCH 28/37] drm: encapsulate crtc->set_config calls Daniel Vetter
2012-12-12 13:07   ` [PATCH 29/37] drm: refcounting for crtc framebuffers Daniel Vetter
2012-12-12 13:07   ` [PATCH 30/37] drm/i915: dump refcount into framebuffer debugfs file Daniel Vetter
2012-12-12 13:07   ` [PATCH 32/37] drm: optimize drm_framebuffer_remove Daniel Vetter
2012-12-12 13:07   ` [PATCH 34/37] drm/ttm: fix fence locking in ttm_buffer_object_transfer Daniel Vetter
2012-12-12 14:48     ` Jerome Glisse
     [not found]       ` <CAH3drwbL34+omj2S3ArOPgoYRQ3uzPEiRPFLn7YkP3cOMnPiZg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2012-12-12 15:42         ` Daniel Vetter
2012-12-13 11:06     ` [PATCH] allow shmob+imx drm drivers to be compiled Daniel Vetter
2012-12-13 11:18     ` Daniel Vetter
2012-12-13 11:26     ` [PATCH] drm/ttm: fix fence locking in ttm_buffer_object_transfer Daniel Vetter
2012-12-12 13:07   ` [PATCH 35/37] drm/radeon: fix fence locking in the pageflip callback Daniel Vetter
2012-12-12 13:07   ` [PATCH 36/37] drm: only grab the crtc lock for pageflips Daniel Vetter
2012-12-12 13:07   ` [PATCH 37/37] drm: don't hold crtc mutexes for connector ->detect callbacks Daniel Vetter
2012-12-12 14:18 ` [PATCH 00/37] [RFC] revamped modeset locking Daniel Vetter

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=1355317637-16742-15-git-send-email-daniel.vetter@ffwll.ch \
    --to=daniel.vetter-/w4ywyx8dfk@public.gmane.org \
    --cc=dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
    --cc=intel-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
    --cc=nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
    --cc=xorg-driver-ati-go0+a7rfsptAfugRpC6u6w@public.gmane.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.