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: NDavid Herrmann <dh.herrmann@gmail.com>

We can significantly simplify this helper by using plain multiplication.
Note that we converted the minor-type to an enum earlier so this didn't
work before.

We also fix a minor range-bug here: the limit argument of idr_alloc() is
*exclusive*, not inclusive, so we should use 64 instead of 63 as offset.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

Properly name goto-labels, remove empty lines and use DRM_ERROR if
possible.
Signed-off-by: NDavid 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: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

No need to check for DEBUGFS, we already have dummy-fallbacks in our
headers.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

drm_get_minor() no longer allocates objects, and drm_unplug_minor() is now
the exact reverse of it. Rename it to _register/unregister() so their
name actually says what they do.

Furthermore, remove the direct minor-ptr and instead pass the minor-type.
This way we know the actual slot of the minor and can reset it if
required.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

_put/get() are used for ref-counting, which we clearly don't do here.
Rename it to _free() and also use the common drm_minor_* prefix.
Furthermore, avoid passing the minor directly but instead use the type
like the other functions do, this allows us to reset the slot.

We also drop the redundant call to drm_unplug_minor() as drm_minor_free()
is only used from paths were that has already be called.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Instead of waiting for device-registration, we now allocate minor-objects
during device allocation. The minors are not registered or assigned an ID.
This is still postponed to device-registration.

While at it, remove the superfluous output-parameter in drm_get_minor().

The reason for this early allocation is to make
dev->primary/control/render available atomically. So once the device is
alive, all of them are already set and we never have the situation where
one of them is set after another (they're either NULL or set, but never
changed). This will eventually allow us to reduce minor-ID allocation to
one base-ID instead of a single ID for each.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel 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: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Lets not trick ourselves into thinking "drm_device" objects are not
ref-counted. That's just utterly stupid. We manage "drm_minor" objects on
each drm-device and each minor can have an unlimited number of open
handles. Each of these handles has the drm_minor (and thus the drm_device)
as private-data in the file-handle. Therefore, we may not destroy
"drm_device" until all these handles are closed.

It is *not* possible to reset all these pointers atomically and restrict
access to them, and this is *not* how this is done! Instead, we use
ref-counts to make sure the object is valid and not freed.

Note that we currently use "dev->open_count" for that, which is *exactly*
the same as a reference-count, just open coded. So this patch doesn't
change any semantics on DRM devices (well, this patch just introduces the
ref-count, anyway. Follow-up patches will replace open_count by it).

Also note that generic VFS revoke support could allow us to drop this
ref-count again. We could then just synchronously disable any fops->xy()
calls. However, this is not the case, yet, and no such patches are
in sight (and I seriously question the idea of dropping the ref-cnt
again).
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

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: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With dev->anon_inode we have a global address_space ready for operation
right from the beginning. Therefore, there is no need to do a delayed
setup with TTM. Instead, set dev_mapping during initialization in
ttm_bo_device_init() and remove any "if (dev_mapping)" conditions.

Cc: Dave Airlie <airlied@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

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: NDavid Herrmann <dh.herrmann@gmail.com>

Our current DRM design uses a single address_space for all users of the
same DRM device. However, there is no way to create an anonymous
address_space without an underlying inode. Therefore, we wait for the
first ->open() callback on a registered char-dev and take-over the inode
of the char-dev. This worked well so far, but has several drawbacks:
 - We screw with FS internals and rely on some non-obvious invariants like
   inode->i_mapping being the same as inode->i_data for char-devs.
 - We don't have any address_space prior to the first ->open() from
   user-space. This leads to ugly fallback code and we cannot allocate
   global objects early.

As pointed out by Al-Viro, fs/anon_inode.c is *not* supposed to be used by
drivers for anonymous inode-allocation. Therefore, this patch follows the
proposed alternative solution and adds a pseudo filesystem mount-point to
DRM. We can then allocate private inodes including a private address_space
for each DRM device at initialization time.

Note that we could use:
  sysfs_get_inode(sysfs_mnt->mnt_sb, drm_device->dev->kobj.sd);
to get access to the underlying sysfs-inode of a "struct device" object.
However, most of this information is currently hidden and it's not clear
whether this address_space is suitable for driver access. Thus, unless
linux allows anonymous address_space objects or driver-core provides a
public inode per device, we're left with our own private internal mount
point.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

There is no need to initialize this variable, so drop it. Otherwise, the
compiler won't warn if we use it unintialized.
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

Lets make sure some basic expressions are always true:
  bpp != NULL
  width != NULL
  height != NULL
  stride = bpp * width < 2^32
  size = stride * height < 2^32
  PAGE_ALIGN(size) < 2^32

At least the udl driver doesn't check for multiplication-overflows, so
lets just make sure it will never happen. These checks allow drivers to do
any 32bit math without having to test for mult-overflows themselves.

The two divisions might hurt performance a bit, but dumb_create() is only
used for scanout-buffers, so that should be fine. We could use 64bit math
to avoid the divisions, but that may be slow on 32bit machines.. Or maybe
there should just be a "safe_mult32()" helper, which currently doesn't
exist (I think?).
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

All drivers currently need to clean up the vma-node manually. There is no
fancy logic involved so lets just clean it up unconditionally. The
vma-manager correctly catches multiple calls so we are fine.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

Remove double-whitespace and wrong indentation.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

Probably a typo.. we obviously need "(bpp + 7) / 8" instead of
"(bpp + 1) / 8". Unlikely to be hit in any sane code, but lets be safe.
Use DIV_ROUND_UP() to avoid the problem entirely and make the core more
readable.
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

We need to call dma_buf_end_cpu_access() in case a damage-request.
Unlikely, but might happen during device unplug.
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>

Just move all fields into radeon_cs_reloc, removing unused/duplicated fields.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Not needed any more.
Signed-off-by: NChristian König <christian.koenig@amd.com>

Not needed any more.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

No need to make it more complicated than necessary,
just allocate the page tables as normal BO and
flush whenever the address change.

v2: update comments and function name
v3: squash bug fixes, page directory and tables patch
v4: rebased on Mareks changes
Signed-off-by: NChristian König <christian.koenig@amd.com>

Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Both are complex enough on their own.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Signed-off-by: NChristian König <christian.koenig@amd.com>

Signed-off-by: NChristian König <christian.koenig@amd.com>

Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Userspace should set the first 4 bits of drm_radeon_cs_reloc::flags to
a number from 0 to 15. The higher the number, the higher the priority,
which means a buffer with a higher number will be validated sooner.

The old behavior is preserved: Buffers used for write are prioritized over
read-only buffers if the userspace doesn't set the number.

v2: add buffers to buckets directly, then concatenate them
v3: use a stable sort
Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

The statistics are:
- VRAM usage in bytes
- GTT usage in bytes
- number of bytes moved by TTM

The last one is actually a counter, so you need to sample it before and after
command submission and take the difference.

This is useful for finding performance bottlenecks. Userspace queries are
also added.

v2: use atomic64_t
Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

When passing buffers between processes, the receiving process needs to know
the original buffer domain, so that it doesn't accidentally move the buffer.

v2: reserve the buffer
Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Now that Christian fixed the performance problems with
the feedback buffer in mesa, we can enable variable UVD
clocks.  There are multiple UVD power states associated
with different types and numbers of streams.  This uses
the appropriate state based on that information rather
than always using the fastest UVD clocks which saves some
power.  One possible downside is that this may adversely
affect decode benchmarks since these power states target
specific playback requirements rather than maximum
performance.  If that becomes an issue, we can add a
sysfs attribute to force the max UVD state.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

We no longer need to take the ring lock while checking for
a gpu lockup, so just cleanup the code.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Use atomics and jiffies_64, so that we don't need to have the
ring mutex locked any more and avoid wrap arounds.

v2: fix some checkpatch warnings
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Implements an I2C-over-AUX I2C adapter on top of the generic drm_dp_aux
infrastructure. It extracts the retry logic from existing drivers, which
should help in porting those drivers to this new helper.
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>
---
Changes in v5:
- move comments partially to to header file
- keep MOT set between I2C messages
- return -EPROTO on short reads

Changes in v4:
- fix typo "bitrate" -> "bit rate"

Changes in v3:
- add back DRM_DEBUG_KMS and DRM_ERROR messages
- embed i2c_adapter within struct drm_dp_aux
- fix typo in comment

Add a helper to probe a DP link (read out the supported DPCD revision,
maximum rate, link count and capabilities) as well as power up the DP
link and configure it accordingly.
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>
---
Changes in v5:
- export helpers

Changes in v4:
- fix a couple of typos in comments as pointed out by Alex Deucher

Changes in v3:
- split into drm_dp_link_power_up() and drm_dp_link_configure()
- do not change sink state for DPCD versions earlier than 1.1
- sleep for 1-2 ms after setting local sink to D0 state
- read and write consecutive registers where possible
- read DPCD revision when link is probed
- remove duplicate kerneldoc

The function reads the link status (6 bytes starting at offset 0x202)
from the DPCD so that it can be conveniently passed to other DPCD
helpers.
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NThierry Reding <treding@nvidia.com>