This reverts commit 48ba813701.
Thanks to Chris:
"drm_file->is_master is not synomous with having drm_file->master ==
drm_file->minor->master. This is because drm_file->master is the same
for all drm_files of the same generation and so when there is a master,
every drm_file believes itself to be the master. Confusion ensues and
things go pear shaped when one file is closed and there is no master
anymore."
Conflicts:
drivers/gpu/drm/drm_drv.c
drivers/gpu/drm/drm_stub.c
bunch of cleanups
* 'drm-next' of git://people.freedesktop.org/~dvdhrm/linux:
drm: mark drm_context support as legacy
drm: make sysfs device always available for minors
drm: make minor->index available early
drm: merge drm_drv.c into drm_ioctl.c
drm: move module initialization to drm_stub.c
drm: don't de-authenticate clients on master-close
drm: drop redundant drm_file->is_master
drm: extract legacy ctxbitmap flushing
This renames all drm-context helpers to drm_legacy_*() and moves the
internal definitions into the new drm_legacy.h header. This header is
local to DRM-core and drivers shouldn't access it.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Most of the new DRM management functions are nowadays in drm_stub.c. By
moving the core module initialization to drm_stub.c we can make several
global variables static and keep the stub-open helper local.
The core files now look like this:
drm_stub.c: Core management
drm_drv.c: Ioctl dispatcher
drm_ioctl.c: Actual ioctl backends
drm_fops.c: Char-dev file-operations
A follow-up patch will move what is left from drm_drv.c into drm_ioctl.c.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
If an active DRM-Master closes its device, we deauthenticate all clients
on that master. However, if an inactive DRM-Master closes its device, we
do nothing. This is quite inconsistent and breaks several scenarios:
1) If this was used as security mechanism, it fails horribly if a master
closes a device while VT switched away. Furthermore, none of the few
drivers using ->master_*() callbacks seems to require it, anyway.
2) If you spawn weston (or any other non-UMS compositor) in background
while another compositor is active, both will get assigned to the
same "drm_master" object. If the foreground compositor now exits, all
clients of both the foreground AND background compositor will be
de-authenticated leading to unexpected behavior.
Stop this non-sense and keep clients authenticated. We don't do this when
dropping DRM-Master (i.e., switching VTs) so don't do it on active-close
either!
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
The drm_file->is_master field is redundant as it's equivalent to:
drm_file->master && drm_file->master == drm_file->minor->master
1) "=>"
Whenever we set drm_file->is_master, we also set:
drm_file->minor->master = drm_file->master;
Whenever we clear drm_file->is_master, we also call:
drm_master_put(&drm_file->minor->master);
which implicitly clears it to NULL.
2) "<="
minor->master cannot be set if it is non-NULL. Therefore, it stays as
is unless a file drops it.
If minor->master is NULL, it is only set by places that also adjust
drm_file->is_master.
Therefore, we can safely drop is_master and replace it by an inline helper
that matches:
drm_file->master && drm_file->master == drm_file->minor->master
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
The ctxbitmap code is only used by legacy drivers so lets try to keep it
as separated as possible. Furthermore, the locking is non-obvious and
kinda weird with ctxlist_mutex *and* struct_mutex. Keeping all ctxbitmap
access in one file is much easier to review and makes drm_release() more
readable.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Linux doesn't run on i386, anymore. See:
commit d55c5a93db
Author: H. Peter Anvin <hpa@linux.intel.com>
Date: Wed Nov 28 11:50:24 2012 -0800
x86, 386 removal: Remove CONFIG_CMPXCHG
All 486+ CPUs support CMPXCHG, so remove the fallback 386 support
code.
Furthermore, as the commit-message states, all 486+ CPUs support the
CMPXCHG instruction and thus even legacy DRM can run fine.
Drop the now superfluous "x86 == 3" check.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
inode argument is no longer in use (cf. f4aede2e3).
Remove it.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The master management was previously protected by the drm_device::struct_mutex.
In order to avoid locking order violations in a reworked dropped master
security check in the vmwgfx driver, break it out into a separate master_mutex.
Locking order is master_mutex -> struct_mutex.
Also remove drm_master::blocked since it's not used.
v2: Add an inline comment about what drm_device::master_mutex is protecting.
v3: Remove unneeded struct_mutex locks. Fix error returns in
drm_setmaster_ioctl().
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Acked-by: Daniel Vetter <daniel@ffwll.ch>
Add a drm_is_legacy() helper, constify argument to drm_is_render_client(),
and use / change helpers where appropriate.
v2: s/drm_is_legacy/drm_is_legacy_client/ and adapt to new code context.
v3: s/legacy_client/primary_client/
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Like for render-nodes, there is no point in maintaining the master concept
for control nodes, so set the struct drm_file::master pointer to NULL.
At the same time, make sure DRM_MASTER | DRM_CONTROL_ALLOW ioctls are always
allowed when called through the control node. Previously the caller also
needed to be master.
v2: Adapt to refactoring of ioctl permission check.
v3: Formatting of logical expression. Use drm_is_control_client() instead of
drm_is_control().
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
We used to protect minor-lookup and setup by the global drm lock. To
continue our attempts of dropping drm_global_mutex, this patch makes the
minor management independent of it. Furthermore, we make it all atomic and
switch to spin-locks instead of a mutex.
Now that minor-lookup is independent, we also move the
"drm_is_unplugged()" test into the minor-lookup path. There is no reason
to ever return a minor for unplugged objects, so keep that logic internal.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Whenever we access minor->device, we are in a minor->kdev->...->fops
callback so the minor->kdev pointer *must* be valid. Thus, simply use
minor->kdev->devt instead of minor->device and remove the redundant field.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Instead of accessing drm_minors_idr directly, this adds a small helper to
hide the internals. This will help us later to remove the drm_global_mutex
requirement for minor-lookup.
Furthermore, this also makes sure that minor->dev is always valid and
takes a reference-count to the device as long as the minor is used in an
open-file. This way, "struct file*"->private_data->dev is guaranteed to be
valid (which it has to, as we cannot reset it).
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The drm_open_helper() function is only used internally for drm_open() so
we can safely pass in the minor-object directly instead of the minor-id.
This way, we avoid the additional minor IDR lookup, which we already do
twice in drm_stub_open() and drm_open().
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
DRM drivers share a common address_space across all character-devices of a
single DRM device. This allows simple buffer eviction and mapping-control.
However, DRM core currently waits for the first ->open() on any char-dev
to mark the underlying inode as backing inode of the device. This delayed
initialization causes ugly conditions all over the place:
if (dev->dev_mapping)
do_sth();
To avoid delayed initialization and to stop reusing the inode of the
char-dev, we allocate an anonymous inode for each DRM device and reset
filp->f_mapping to it on ->open().
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
This is just used for a debugfs file, and we can easily reconstruct
this number by just walking the list twice. Which isn't really bad for
a debugfs file anyway.
So let's rip this out.
There's the other issue that the dev->vmalist itself is a bit useless,
since that can be reconstructed with all the memory mapping
information from proc. But remove that is a different topic entirely.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
It's racy, and it's only used in debugfs. There are simpler ways to
know whether something is going on (like looking at dmesg with full
debugging enabled). And they're all much more useful.
So let's just rip this out.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Now dev->ioctl_count tries to prevent the device from disappearing if
it's still in use. And if we'd actually need this code it would be
hopelessly racy and broken.
But luckily the vfs already takes care of this. So we can just rip it
out.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Pull drm updates from Dave Airlie:
"This is a combo of -next and some -fixes that came in in the
intervening time.
Highlights:
New drivers:
ARM Armada driver for Marvell Armada 510 SOCs
Intel:
Broadwell initial support under a default off switch,
Stereo/3D HDMI mode support
Valleyview improvements
Displayport improvements
Haswell fixes
initial mipi dsi panel support
CRC support for debugging
build with CONFIG_FB=n
Radeon:
enable DPM on a number of GPUs by default
secondary GPU powerdown support
enable HDMI audio by default
Hawaii support
Nouveau:
dynamic pm code infrastructure reworked, does nothing major yet
GK208 modesetting support
MSI fixes, on by default again
PMPEG improvements
pageflipping fixes
GMA500:
minnowboard SDVO support
VMware:
misc fixes
MSM:
prime, plane and rendernodes support
Tegra:
rearchitected to put the drm driver into the drm subsystem.
HDMI and gr2d support for tegra 114 SoC
QXL:
oops fix, and multi-head fixes
DRM core:
sysfs lifetime fixes
client capability ioctl
further cleanups to device midlayer
more vblank timestamp fixes"
* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (789 commits)
drm/nouveau: do not map evicted vram buffers in nouveau_bo_vma_add
drm/nvc0-/gr: shift wrapping bug in nvc0_grctx_generate_r406800
drm/nouveau/pwr: fix missing mutex unlock in a failure path
drm/nv40/therm: fix slowing down fan when pstate undefined
drm/nv11-: synchronise flips to vblank, unless async flip requested
drm/nvc0-: remove nasty fifo swmthd hack for flip completion method
drm/nv10-: we no longer need to create nvsw object on user channels
drm/nouveau: always queue flips relative to kernel channel activity
drm/nouveau: there is no need to reserve/fence the new fb when flipping
drm/nouveau: when bailing out of a pushbuf ioctl, do not remove previous fence
drm/nouveau: allow nouveau_fence_ref() to be a noop
drm/nvc8/mc: msi rearm is via the nvc0 method
drm/ttm: Fix vma page_prot bit manipulation
drm/vmwgfx: Fix a couple of compile / sparse warnings and errors
drm/vmwgfx: Resource evict fixes
drm/edid: compare actual vrefresh for all modes for quirks
drm: shmob_drm: Convert to clk_prepare/unprepare
drm/nouveau: fix 32-bit build
drm/i915/opregion: fix build error on CONFIG_ACPI=n
Revert "drm/radeon/audio: don't set speaker allocation on DCE4+"
...
When a second process opens the device and master transferrence is
complete, we walk the list of open devices and remove their
authentication. This also revokes our root privilege. Instead of simply
dropping the authentication, this patch reverts the authenticated state
back to its original value.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When there are unconsumed pending events, the events are
destroyed by calling destroy callback, but the events list
are remained, because there is no list_del().
It is possible that the page flip request is handled after
drm_events_release() is called and before drm_fb_release().
In this case a drm_pending_event is remained not freed.
So exynos driver checks again to remove it in its post
close routine. But the file_priv->event_list contains
undeleted ones, this can make oops for accessing invalid
memory.
Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The user of these counters was killed in
commit d79cdc8312
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Thu Aug 8 15:41:32 2013 +0200
drm: no-op out GET_STATS ioctl
so clean up the leftovers as well.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The only user of ctx_count is the via driver, and we can replace that
use with list_is_singular().
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Try to keep all functions that handle DRM file_operations in drm_fops.c
so internal helpers can be marked static later.
This makes the split between the 3 core files more obvious:
- drm_stub.c: DRM device allocation/destruction and management
- drm_fops.c: DRM file_operations (except for ioctl)
- drm_drv.c: Global DRM init + ioctl handling
Well, ioctl handling is still spread throughout hundreds of source files,
but at least the others are clearly defined this way.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This reverts commit 7c510133d9.
Well looks like not enough digging was done, libdrm_nouveau before 2.4.33
used contexts,
292da616fe1f936ca78a3fa8e1b1b19883e343b6 nouveau: pull in major libdrm rewrite
got rid of them,
Reported-by: Paul Zimmerman <Paul.Zimmerman@synopsys.com>
Reported-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Render nodes provide an API for userspace to use non-privileged GPU
commands without any running DRM-Master. It is useful for offscreen
rendering, GPGPU clients, and normal render clients which do not perform
modesetting.
Compared to legacy clients, render clients no longer need any
authentication to perform client ioctls. Instead, user-space controls
render/client access to GPUs via filesystem access-modes on the
render-node. Once a render-node was opened, a client has full access to
the client/render operations on the GPU. However, no modesetting or ioctls
that affect global state are allowed on render nodes.
To prevent privilege-escalation, drivers must explicitly state that they
support render nodes. They must mark their render-only ioctls as
DRM_RENDER_ALLOW so render clients can use them. Furthermore, they must
support clients without any attached master.
If filesystem access-modes are not enough for fine-grained access control
to render nodes (very unlikely, considering the versaitlity of FS-ACLs),
you may still fall-back to fd-passing from server to client (which allows
arbitrary access-control). However, note that revoking access is
currently impossible and unlikely to get implemented.
Note: Render clients no longer have any associated DRM-Master as they are
supposed to be independent of any server state. DRM core highly depends on
file_priv->master to be non-NULL for modesetting/ctx/etc. commands.
Therefore, drivers must be very careful to not require DRM-Master if they
support DRIVER_RENDER.
So far render-nodes are protected by "drm_rnodes". As long as this
module-parameter is not set to 1, a driver will not create render nodes.
This allows us to experiment with the API a bit before we stabilize it.
v2: drop insecure GEM_FLINK to force use of dmabuf
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The export dma-buf cache is semantically similar to an flink name. So
semantically it makes sense to treat it the same and remove the name
(i.e. the dma_buf pointer) and its references when the last gem handle
disappears.
Again we need to be careful, but double so: Not just could someone
race and export with a gem close ioctl (so we need to recheck
obj->handle_count again when assigning the new name), but multiple
exports can also race against each another. This is prevented by
holding the dev->object_name_lock across the entire section which
touches obj->dma_buf.
With the new scheme we also need to reinstate the obj->dma_buf link at
import time (in case the only reference userspace has held in-between
was through the dma-buf fd and not through any native gem handle). For
simplicity we don't check whether it's a native object but
unconditionally set up that link - with the new scheme of removing the
obj->dma_buf reference when the last handle disappears we can do that.
To make it clear that this is not just for exported buffers anymore
als rename it from export_dma_buf to dma_buf.
To make sure that now one can race a fd_to_handle or handle_to_fd with
gem_close we use the same tricks as in flink of extending the
dev->object_name_locking critical section. With this change we finally
have a guaranteed 1:1 relationship (at least for native objects)
between gem objects and dma-bufs, even accounting for races (which can
happen since the dma-buf itself holds a reference while in-flight).
This prevent igt/prime_self_import/export-vs-gem_close-race from
Oopsing the kernel. There is still a leak though since the per-file
priv dma-buf/handle cache handling is racy. That will be fixed in a
later patch.
v2: Remove the bogus dma_buf_put from the export_and_register_object
failure path if we've raced with the handle count dropping to 0.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We kzalloc this structure, and for real kms devices we should never
loose track of things really.
But ums/legacy drivers rely on the drm core to clean up a bit of cruft
between lastclose and firstopen (i.e. when X is being restarted), so
keep this around. But give it a clear drm_legacy_ prefix and
conditionalize the code on !DRIVER_MODESET.
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
It has way too much potential for driver writers to do stupid things
like delayed hw setup because the load sequence is somehow racy (e.g.
the imx driver in staging). So don't call it for modesetting drivers,
which reduces the complexity of the drm core -> driver interface a
notch.
v2: Don't forget to update DocBook.
v3: Go with Laurent's slightly more elaborate proposal for the DocBook
update. Add a few words on top of his diff to elaborate a bit on what
KMS drivers should and shouldn't do in lastclose. There was already a
paragraph present talking about restoring properties, I've simply
extended that one.
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
So I've stumbled over drm_fasync and wondered what it does. Digging
that up is quite a story.
First I've had to read up on what this does and ended up being rather
bewildered why peopled loved signals so much back in the days that
they've created SIGIO just for that ...
Then I wondered how this ever works, and what that strange "No-op."
comment right above it should mean. After all calling the core fasync
helper is pretty obviously not a noop. After reading through the
kernels FASYNC implementation I've noticed that signals are only sent
out to the processes attached with FASYNC by calling kill_fasync.
No merged drm driver has ever done that.
After more digging I've found out that the only driver that ever used
this is the so called GAMMA driver. I've frankly never heard of such a
gpu brand ever before. Now FASYNC seems to not have been the only bad
thing with that driver, since Dave Airlie removed it from the drm
driver with prejudice:
commit 1430163b4bbf7b00367ea1066c1c5fe85dbeefed
Author: Dave Airlie <airlied@linux.ie>
Date: Sun Aug 29 12:04:35 2004 +0000
Drop GAMMA DRM from a great height ...
Long story short, the drm fasync support seems to be doing absolutely
nothing. And the only user of it was never merged into the upstream
kernel. And we don't need any fops->fasync callback since the fcntl
implementation in the kernel already implements the noop case
correctly.
So stop this particular cargo-cult and rip it all out.
v2: Kill drm_fasync assignments in rcar (newly added) and imx drivers
(somehow I've missed that one in staging). Also drop the reference in
the drm DocBook. ARM compile-fail reported by Rob Clark.
v3: Move the removal of dev->buf_asnyc assignment in drm_setup to this
patch here.
v4: Actually git add ... tsk.
Cc: Dave Airlie <airlied@linux.ie>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Rob Clark <robdclark@gmail.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
So after a lot of digging around in git histories it looks like this
has only ever be used by dri1 render clients. Hence we can fully
disable the entire thing for modesetting drivers and so greatly reduce
the attack surface for potential exploits (or at least tools like
trinity ...).
Also add the drm_legacy prefix for functions which are called from
common code. To further reduce the impact on common code also extract
all the ctx release handling into a function (instead of only
releasing individual handles) and make ctxbitmap_cleanup return void -
it can never fail.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
And hide the checks a bit better. This was already disallowed for
modesetting drivers, so no functinal change here.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Only ever re-cleared in drm_setup, otherwise completely unused.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@gmail.com>
Only ever assigned in the context code for real, with no readers
anywhere. Remove it.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@gmail.com>
No one ever waits on this waitqueue, so the wake_up call is wasted.
Remove it all.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@gmail.com>
There are missing parts to handle error in drm_open_helper().
The priv->minor, assigned by idr_find() which can return NULL,
should be checked whether it is NULL or not before referencing it.
put_pid(), drm_gem_release(), and drm_prime_destory_file_private()
should be called when error happens after their pair functions are
called. If an error occurs after executing dev->driver->open()
which allocates driver specific per-file private data, then the
private data should be released.
Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Reviewed-by: Chris Wilson <chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
If first drm_open fails, the error-handling path will
incorrectly restore inode's mapping to NULL. This can
cause the crash later on. Fix by separately storing
away mapping pointers that drm_open can touch and
restore each from its own respective variable if the
call fails.
Fixes: https://bugzilla.novell.com/show_bug.cgi?id=807850
(thanks to Michal Hocko for investigating investigating and
finding the root cause of the bug)
Reference:
http://lists.freedesktop.org/archives/dri-devel/2013-March/036564.html
v2: Use one variable to store file and inode mapping
since they are the same at the function entry.
Fix spelling mistakes in commit message.
v3: Add reference to the original bug report.
Reported-by: Marco Munderloh <munderl@tnt.uni-hannover.de>
Tested-by: Marco Munderloh <munderl@tnt.uni-hannover.de>
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
Well, at least step 1. The goal here is that framebuffer objects can
survive outside of the mode_config lock, with just a reference held
as protection. The first step to get there is to introduce a special
fb_lock which protects fb lookup, creation and destruction, to make
them appear atomic.
This new fb_lock can nest within the mode_config lock. But the idea is
(once the reference counting part is completed) that we only quickly
take that fb_lock to lookup a framebuffer and grab a reference,
without any other locks involved.
vmwgfx is the only driver which does framebuffer lookups itself, also
wrap those calls to drm_mode_object_find with the new lock.
Also protect the fb_list walking in i915 and omapdrm with the new lock.
As a slight complication there's also the list of user-created fbs
attached to the file private. The problem now is that at fclose() time
we need to walk that list, eventually do a modeset call to remove the
fb from active usage (and are required to be able to take the
mode_config lock), but in the end we need to grab the new fb_lock to
remove the fb from the list. The easiest solution is to add another
mutex to protect this per-file list.
Currently that new fbs_lock nests within the modeset locks and so
appears redudant. But later patches will switch around this sequence
so that taking the modeset locks in the fb destruction path is
optional in the fastpath. Ultimately the goal is that addfb and rmfb
do not require the mode_config lock, since otherwise they have the
potential to introduce stalls in the pageflip sequence of a compositor
(if the compositor e.g. switches to a fullscreen client or if it
enables a plane). But that requires a few more steps and hoops to jump
through.
Note that framebuffer creation/destruction is now double-protected -
once by the fb_lock and in parts by the idr_lock. The later would be
unnecessariy if framebuffers would have their own idr allocator. But
that's material for another patch (series).
v2: Properly initialize the fb->filp_head list in _init, otherwise the
newly added WARN to check whether the fb isn't on a fpriv list any
more will fail for driver-private objects.
v3: Fixup two error-case unlock bugs spotted by Richard Wilbur.
Reviewed-by: Rob Clark <rob@ti.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Some drivers (specifically vmwgfx) look at dev_mapping
in their open hook, so we have to set dev->dev_mapping
earlier in the process.
Reference:
http://lists.freedesktop.org/archives/dri-devel/2012-October/029420.html
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reported-by: Thomas Hellstrom <thellstrom@vmware.com>
Cc: stable@vger.kernel.org
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Dave Airlie