freedreno: move ring_cache behind fd_bo_del()

So that it isn't bypassing normal refcnt'ing.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: fix spelling typo

Signed-off-by: Rob Clark <robclark@freedesktop.org>

tests/amdgpu: add unaligned VM test

Make a VM mapping which is as unaligned as possible.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

intel: annotate the intel genx helpers as private

They're used internally and never meant to be part of the API.
Add the drm_private notation, which should resolve that.

v2: (Rodrigo) Add missing include.
v3: (Rodrigo) Keep includes grouped per Eric suggestion.

Cc: Eric Engestrom <eric.engestrom@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Fixes: 4e81d4f9c9b ("intel: add generic functions to check PCI ID")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

gitlab-ci: use templates to deduplicate the build commands

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

intel: get gen once for gen >= 9

We don't need to call IS_GEN() for each gen >= 9: we can rather use the
new intel_is_genx() helper to iterate the pciids array once.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

intel: make gen9 use generic gen macro

The 2 PCI IDs that are used for the command line overrid mechanism
were left defined. The rest can be gone and then we just use the kernel
defines.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

intel: make gen10 use generic gen macro

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

intel: make gen11 use generic gen macro

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

intel: add generic functions to check PCI ID

This will allow platforms to reuse kernel IDs instead of manually
keeping them in sync. In most of the cases we only need to extend
IS_9XX().  Current platforms that fit this requirement can be ported
over to use this macro. Right now it's a nop since it doesn't have any
PCI ID added.

The i915_pciids.h header is in sync with kernel tree on
drm-tip 2018y-08m-20d-21h-41m-11s.

v2: - move to a separate .c so we can have the array in a single
      compilation unit
    - use a single array for all gens
    - add real functions to get or check gen by pciid
    - define our own pci device struct rather than inherit the one
      kernel uses: we can throw away most of the fields

v3: - add comment to keep ids sorted by gen
    - remove misleading comment about all gens

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

xf86drm: rename "real_path" to "pci_path"

"real_path" was getting confusing when there are other *paths in the
same functions.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

xf86drm: merge get_normal_pci_path() into get_real_pci_path()

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

Fix build with -std=c11

typeof() is a GNU extension that will only work when the compiler is passed
-std=gnu*.  __typeof__() works with -std=c*, however.
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

add gitlab-ci builds of libdrm

It currently does 4 builds: 2 using Meson and 2 using Autotools, 2 using
the latest dependencies on ArchLinux and 2 using very old dependencies
on Debian (including manually building libpciaccess to have the oldest
version supported, to make sure it keeps being supported).

All the build options are turned on for both Meson and Autotools.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Daniel Stone <daniels@collabora.com>

Add basic CONTRIBUTING file

I picked up a bunch of the pieces from wayland's version:

https://gitlab.freedesktop.org/wayland/wayland/blob/master/CONTRIBUTING.md

The weston one is fairly similar. Then I rather massively trimmed it
down since in reality libdrm is a bit a dumping ground with very few
real rules. The commit rights and CoC sections I've copied verbatim
from igt respectively drm-misc. Weston/Wayland only differ in their
pick of how many patches you need (10 instead of 5). I think for
libdrm this is supremely relevant, since most everyone will get their
commit rights by contributing already to the kernel or mesa and having
commit rights there already.

Anyway, I figured this is good to get the rules documented, even if
there's mostly not many rules.

Note: This references maintainers in a MAINTAINERS file, which needs
to be created first.

Note: With the gitlab migration the entire commit rights process is
still a bit up in the air. But gitlab commit rights and roles are
hierarchical, so we can do libdrm-only maintainer/commiter roles
("Owner" and "Developer" in gitlab-speak). This should avoid
conflating libdrm roles with mesa roles, useful for those pushing to
libdrm as primarily kernel contributors.

v2: Comments from Emil:
- Recommend subject prefix.
- Fix copypaste fumbles, this isn't igt/wayland ...

v3: Comments from Marek:
- libdrm moved to mesa, update the document. Atm the entire account
  request situation is entirely not clear for gitlab and mesa
  projects, so that's a bit up in the air. Also, should probably send
  an announcement to dri-devel@, which didn't happen.
- amd folks don't submit their patches to dri-devel, document that.
  Probably applies to other drivers too.

v4: Comments from Rob:
- Also include kernel/userspace in the commit counts criteria, due to
  libdrm's special role as a glue library.

v5: Summarize the irc discussion on gitlab roles in the commit message
a bit.

v6: Some grammer stuff from Eric E.

v7: Use --local in git config (Eric E.)

Cc: Dave Airlie <airlied@gmail.com>
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Emil Velikov <emil.velikov@collabora.com>
Cc: Marek Olšák <marek.olsak@amd.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Rob Clark <robdclark@gmail.com> (v4)
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com> (v6)
Acked-by: Emil Velikov <emil.l.velikov@gmail.com> (v6)
Acked-by: Marek Olšák <marek.olsak@amd.com> (v5)
Acked-by: Dave Airlie <airlied@redhat.com>
References: https://gitlab.freedesktop.org/wayland/weston/blob/master/CONTRIBUTING.md
References: https://01.org/linuxgraphics/gfx-docs/maintainer-tools/drm-misc.html#commit-rights
References: https://cgit.freedesktop.org/drm/igt-gpu-tools/tree/CONTRIBUTING#n54
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

amdgpu: amdgpu_bo_inc_ref don't return dummy int

Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amdgpu: add amdgpu_bo_inc_ref() function.

For Pro OGL be able to work with upstream libdrm.

Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: add error return value for finding bo by cpu mapping (v2)

If nothing is found, error should be returned.

v2: udpate the error value different from parameter check

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

amdgpu-symbol-check: Add amdgpu_find_bo_by_cpu_mapping

Fixes: 4d454424e1f2 ("amdgpu: add a function to find bo by cpu mapping
                      (v2)"

xf86drm: fallback to normal path when realpath fails

Earlier commit reworked our sysfs handling to use realpath.
Sadly that backfired since the Firefox sandboxing mechanism rejects
that. Despite the files/folders being in the allowed list, of the
sandboxing mechanism.

Oddly enough, the Chromium sandboxing doesn't complain about any of
this.

Since there are no Firefox releases with the fix, add a temporary
solution which falls back to the original handling.

Sadly, this won't work for virgl.

v2: drop return type - function cannot return NULL (Eric)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107516
Fixes: a02900133b3 ("xf86drm: introduce a get_real_pci_path() helper")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

Bump to version 2.4.94

libdrm: add msm drm uapi header

msm_drm.h file Generated using make headers_install.

Generated from
tree - git://people.freedesktop.org/~airlied/linux
branch - drm-next
commit - 6d08b06e67cd117f6992c46611dfb4ce267cd71e

Remove freedreno/msm/msm_drm.h to maintain only
one copy of msm_drm.h and change freedreno Makefile
and meson.build file accordingly.

v2: Remove private freedreno/msm/msm_drm.h
v3: meson.build update
v3: README update (by anholt)

Signed-off-by: Tanmay Shah <tanmay@codeaurora.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

amdgpu: add a function to create amdgpu bo internally (v4)

a helper function to create and initialize amdgpu bo

v2: update error handling: add label and free bo
v3: update error handling: separate each error label
v4: update error handling and rebase

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amdgpu: free flink bo in bo import

Fix potential memory leak when handle flink bo in bo import.
Free the flink bo after bo import and in error handling.

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amdgpu: Eliminate void* arithmetic in amdgpu_find_bo_by_cpu_mapping

Arithmetic using void* pointers isn't defined by the C standard, only as
a GCC extension. Avoids compiler warnings:

../../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../../amdgpu/amdgpu_bo.c:554:48: warning: pointer of type ‘void *’ used in arithmetic [-Wpointer-arith]
   if (cpu >= bo->cpu_ptr && cpu < (bo->cpu_ptr + bo->alloc_size))
                                                ^
../../amdgpu/amdgpu_bo.c:561:23: warning: pointer of type ‘void *’ used in subtraction [-Wpointer-arith]
   *offset_in_bo = cpu - bo->cpu_ptr;
                       ^

v2: Use uintptr_t instead of char*, don't change function signature
    (Junwei Zhang)

Fixes: 4d454424e1f2 ("amdgpu: add a function to find bo by cpu mapping
                     (v2)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: Disable deadlock test suite for RV

disable deadlock test suite for RV

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: fix off by one in handle_table_insert

Stupid me, max_key must always be larger than key.

Signed-off-by: Christian König <christian.koenig@amd.com>
Bugzilla: https://bugs.freedesktop.org/107552
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: Use uint32_t i in amdgpu_find_bo_by_cpu_mapping

The compiler points out that an int doesn't work as intended if
dev->bo_handles.max_key > INT_MAX:

../../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../../amdgpu/amdgpu_bo.c:550:16: warning: comparison of integer expressions of different signedness: ‘int’ and ‘uint32_t’ {aka ‘unsigned int’} [-Wsign-compare]
  for (i = 0; i < dev->bo_handles.max_key; i++) {
                ^
../../amdgpu/amdgpu_bo.c:558:8: warning: comparison of integer expressions of different signedness: ‘int’ and ‘uint32_t’ {aka ‘unsigned int’} [-Wsign-compare]
  if (i < dev->bo_handles.max_key) {
        ^

Fixes: 4d454424e1f2 ("amdgpu: add a function to find bo by cpu mapping
                     (v2)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

intel: Add a new CFL PCI ID.

One more CFL ID added to spec.

Align with kernel commit d0e062ebb3a4 ("drm/i915/cfl:
Add a new CFL PCI ID.")

v2: fix commit subject.

Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

freedreno: fix use-after-free with stateobj rb's

We could be dropping last reference in ->flush(), so clear the entry in
the parent rb's table to avoid deref'ing after free'd.

Also, ring_bo_del()'s use of ring_cache expects that it is dropping the
last reference.  So drop our ref to the stateobj's ring_bo first.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: don't leak stateobj rb refs

One stateobj can be emitted multiple times in a single cmdstream, but
only the first time is a cmd entry added to the parent.  Since it will
be only unref'd once after flush, we should only ref it the first time
it is emitted (ie. the time it is added to cmd table).

Signed-off-by: Rob Clark <robclark@freedesktop.org>

tests/amdgpu: add test for finding bo by CPU mapping

Add a test for API to query bo by CPU mapping

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amdgpu: add a function to find bo by cpu mapping (v2)

Userspace needs to know if the user memory is from BO or malloc.

v2: update mutex range and rebase

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amdgpu: add bo from user memory to handle table

When create bo from user memory, add it to handle table
for future query.

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

tests/util: Add support for stm module

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: Fix amdgpu build failure

Use the correct files to build libdrm_amdgpu.

Signed-of-by: Mike Lothian <mike@fireburn.co.uk>
Fixes: d6cb0ee408e ("amdgpu: remove the hash table implementation")
Cc: Christian König <christian.koenig@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

tests/modetest: Add atomic support

If "-a" option is set this make modetest use atomic API instead
of legacy API.

Test the frame rate ("-v") it does a loop and swap between two
framebuffer for each active planes.

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

amdgpu: always add all BOs to handle table

This way we can always find a BO structure by its handle.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: remove the hash table implementation

Not used any more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: use handle table for flink names

Instead of the hash use the handle table.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: use handle table for KMS handles

Instead of the hash use the handle table.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: add handle table implementation v2

The kernel handles are dense and the kernel always tries to use the
lowest free id. Use this to implement a more efficient handle table
by using a resizeable array instead of a hash.

v2: add handle_table_fini function, extra key checks,
    fix typo in function name

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

amdgpu: stop using the hash table for fd_tab

We have so few devices that just walking a linked list is probably
faster.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>

configure.ac: bump version to 2.4.93

freedreno/msm: "stateobj" support

Adds support for "state object" cmdstream buffers which can be
constructed once, and re-used many times.  This enables the use
for CP_SET_DRAW_STATE packets on newer hardware, to lower the
CPU overhead.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: slight reordering

Splitting code-motion out from next patch.  Once we add stateobj rb's
this loop could add new entries to bos table, so it needs to be before
we set req.bos/req.nr_bos.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: small cleanup

Make cheezy growable array thing less open-coded before adding more.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: add fd_ringbuffer_new_object()

Add new API for reusable "state objects" which can be re-used multiple
times.  Backend implementation for msm will follow.  (Probably not
needed to support this for any device that uses kgsl backend, since this
is mostly useful for a5xx+.)

Signed-off-by: Rob Clark <robclark@freedesktop.org>

xf86drm: Fix error path in drmGetDevice2

In drmGetDevice2 when no local device is found or when
drm_device_has_rdev filters out all devices, *device might be left
uninitialized causing drmGetDevice2 to not return error - since
it's only returned when *device == NULL.

Above leads to crash in the firefox in system with amdgpu.

With this change firefox displays:

libGL error: MESA-LOADER: failed to retrieve device information
libGL error: unable to load driver: amdgpu_dri.so
libGL error: driver pointer missing
libGL error: failed to load driver: amdgpu
libGL error: MESA-LOADER: failed to retrieve device information
libGL error: unable to load driver: amdgpu_dri.so
libGL error: driver pointer missing
libGL error: failed to load driver: amdgpu

and doesn't crash.

Bugzilla: https://bugs.freedesktop.org/107384
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Mariusz Ceier <mceier+mesa-dev@gmail.com>

amdgpu: add amdgpu_bo_handle_type_kms_noimport

Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

drmdevice: print the correct host1x information

While fairly close, the host1x and platform are two separate things.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

drmdevice: convert the tabbed output into a tree

Making the output a little bit easier to parse by human beings.

v2: Add extra whitespace (Eric)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Eric Engestrom <eric@engestrom.ch> (v1)

tests/drmdevice: add a couple of printf headers

Add a few printf statements, which should make the output easier to
parse.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

tests/drmdevices: install alongside other utilities

It's mildly useful program, to ship it when the user wants the "tests"
installed. Obviously the "tests" in the name is a misnomer.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

xf86drm: Add drmDevice support for virtio_gpu

The GPU almost exclusively lives on the PCI bus, so we expose it as a
normal PCI one.

This allows all existing drmDevice users to work without any changes.

One could wonder why a separate typeset is not introduced, alike say
host1x. Unlike host1x the PCI/platform distinction for virtio provides
no extra information. Plus if needed we can add the separate set at a
later stage.

Here are a few 'features' that virtio seems to be missing:
 - provides extra information on top the plaform devices
 - supports a range of GPU devices
 - is considered hardware description (DT)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

xf86drm: introduce a get_real_pci_path() helper

Introduce a helper which gets the real sysfs path for the given pci
device.

In other words, instead opening the /sys/dev/char/*/device symlink, we
opt for the actual /sys/devices/pci*/*/

It folds three (nearly identical) snprintf's and paves the way of adding
extra devices (see next patch) a piece of pie.

v2: use a caller (on stack) provided real_path (Eric)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Eric Engestrom <eric@engestrom.ch> (v1)

xf86drm: Allocate drmDevicePtr's on stack

Currently we dynamically allocate 16 pointers and reallocate more as
needed.

Instead, allocate the maximum number (256) on stack - the number is
small enough and is unlikely to change in the foreseeable future.

This allows us to simplify the error handling and even shed a few bytes
off the final binary.

v2:
 - add a define & description behind the magic 256
 - report error to strerr and skip when over 256 device nodes

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch> (v1)

xf86drm: Fold drmDevice processing into process_device() helper

Don't duplicate the nearly identical code across the two call sites.
It improves legibility and the diff stat seems nice.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

xf86drm: introduce drm_device_has_rdev() helper

Currently we match the opened drmDevice fd with each drmDevice we
process.

Move that after all the devices are processed and folded, via the
drm_device_has_rdev(). This makes the code easier to follow and allows
us to unify the massive process loop across drmGetDevice2 and
drmGetDevices2. That in itself is coming with a later commit.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

xf86drm: drmGetDevice2: error out if the fd has unknown subsys

Currently one can open() any /dev node. If it's unknown
drmParseSubsystemType() will return an error.

Track that and bail as needed.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

amdgpu: make sure to set CLOEXEC on duplicated FDs

Otherwise we leak file descriptors into child processes.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

Revert "amdgpu: don't call add_handle_to_table for KMS BO exports"

This reverts commit fe0488aa13c35952b9f3f37ff2c74b6b858e8e73.

It caused messages like

 amdgpu 0000:23:00.0: bo 000000007dce0b3e va 0x0000101800-0x000010181f conflict with 0x0000101800-0x0000101820

in dmesg, and eventually a Xorg crash while running piglit.

Evidently, such BOs can actually be re-imported by other means than via
a KMS handle.

amdgpu: don't call add_handle_to_table for KMS BO exports

Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

intel: Introducing Amber Lake platform

Amber Lake uses the same gen graphics as Kaby Lake, including a id
that were previously marked as reserved on Kaby Lake, but that now is
moved to AML page.

So, let's just move it to AML macro that will feed into KBL macro
just to keep it better organized to make easier future code review
but it will be handled as a KBL.

This is a copy of merged i915's
commit e364672477a1 ("drm/i915/aml: Introducing Amber Lake platform")

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

intel: Introducing Whiskey Lake platform

Whiskey Lake uses the same gen graphics as Coffe Lake, including some
ids that were previously marked as reserved on Coffe Lake, but that
now are moved to WHL page.

So, let's just move them to WHL macros that will feed into CFL macro
just to keep it better organized to make easier future code review
but it will be handled as a CFL.

This is a copy of merged i915's
commit b9be78531d27 ("drm/i915/whl: Introducing Whiskey Lake platform")

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

freedreno: add user ptr to fd_ringbuffer

Something for users of fd_ringbuffer to use as they see fit.  (For now,
just so mesa can add some debugging state.)

Signed-off-by: Rob Clark <robclark@freedesktop.org>

amdgpu: Destroy fd_hash table when the last device is removed.

Fixes memory leak on module unload.
Analogous to mesa commit of the same name.
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu/util_hash_table: Add helper function to count the number of entries in hash table

Analogous to the mesa commit of the same name.
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: Take a lock before removing devices from fd_tab hash table.

Close the file descriptors under lock as well.
v2: close fds after removing from hash table

Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

Always pass O_CLOEXEC when opening DRM file descriptors

Reviewed-by: Christian König <christian.koenig@amd.com>

xf86drm: Be sure to closedir before return

removed in commit bb45ce4e3ac751315bfd7fbfd9e1425bf515ec0d

Adding it back as it is still needed in the case where we don't find a
match.

Signed-off-by: Kevin Strasser <kevin.strasser@intel.com>
Fixes: bb45ce4e3ac751315bfd "libdrm: Use readdir instead of readdir_r to
                             avoid build warnings"$
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

bump version for release

Signed-off-by: Rob Clark <robclark@freedesktop.org>

freedreno: add fd_pipe refcounting

In mesa/gallium, a pipe_fence can outlive the pipe_context it was
created from.  But to wait on the fence we need to know the submit-
queue (ie. the fd_pipe).

The most straightforward way to fix this is to add reference counting
to the fd_pipe and let the fence hold a reference to the pipe (rather
than hanging on to the context, which might have been destroyed before
the fence).

Signed-off-by: Rob Clark <robclark@freedesktop.org>

intel: add support for ICL 11

Add the PCI IDs and the basic code to enable ICL.  This is the current
PCI ID list in our documentation.

Kernel commit: d55cb4fa2cf0 ("drm/i915/icl: Add the ICL PCI IDs")

v2: Michel provided a fix to IS_9XX that was broken by rebase bot.
v3: Fix double definition of PCI IDs, update IDs according to bspec
    and keep them in the same order and rebase (Lucas)

Cc: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

amdgpu: Deinitialize vamgr_high{,_32}

Fixes memory leaks.

Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>

Intel: Add a Kaby Lake PCI ID

Based on kernel commit '672e314b21dc ("drm/i915/kbl: Add KBL GT2 sku")'

v2: name change M -> ULX, add enumeration in KBL ULX
v3: add entry to IS_KABYLAKE

Signed-off-by: Matt Atwood <matthew.s.atwood@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

amdgpu: enlarge the maximum number of cards supported

128 is the maximum number of cards that the kernel can support
at the moment.

Change-Id: I155b7b21635306d8ecc440b85fb8954501ab5599
Signed-off-by: Xiaojie Yuan <Xiaojie.Yuan@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

tests/amdgpu: add vce mv tests support and sets

Signed-off-by: James Zhu <James.Zhu@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>

libdrm: gralloc_handle.h: Fix build issue with Android

In trying to integrate the new gralloc_handle.h with the
drm_hwcomposer, I started seeing the following compilation
errors:

In file included from external/drm_hwcomposer/platformdrmgeneric.cpp:28:
external/libdrm/android/gralloc_handle.h:108:9: error: cannot initialize return object of type 'native_handle_t *' (aka 'native_handle *') with an lvalue of type 'struct gralloc_handle_t *'
        return handle;
               ^~~~~~
1 error generated.

This seems to be due to the gralloc_handle_create() definition
needs to return a native_handle_t * type, rather then a
gralloc_handle_t *, which is what the code actually returns.

After talking w/ Rob Herring, having the code return the
native handle should be the proper fix, so that is what
this patch changes.

Cc: Chih-Wei Huang <cwhuang@android-x86.org>
Cc: Stefan Schake <stschake@gmail.com>
Cc: Robert Foss <robert.foss@collabora.com>
Cc: Sean Paul <seanpaul@google.com>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Reviewed-by: Robert Foss <robert.foss@collabora.com>

headers: Update README

Nouveau has made a very deliberate choice to hide its actual kernel ABI
behind libdrm. i915 is no longer out of date.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>

headers: Sync with drm-next

Taken from the drm-next pull for 4.17-rc1 (694f54f680f7), and manually
reconciled:

  core:
    - Dropped DRM_MODE_TYPE_ALL and DRM_MODE_FLAG_ALL; these are purely
      internal details of the bits accepted by the currently running
      kernel, and can not be generally relied on by userspace
    - Add HDCP flags
    - Note CTM entry representation is sign-magnitude format, not
      two's-complement
  amdgpu:
    - Add QUERY_STATE2 context op
    - Add VCN firmware version query
  etnaviv:
    - Add more GPU feature flags
  i915:
    - Add caps, params and ioctls for PMU / perf-stream
    - Add support for explicit fencing
  nouveau:
    - Add TILE_COMP layout
  vc4:
    - Add perfmon ioctls
  virtgpu:
    - Add capset-fix param
  vmware:
    - Add handle-close ioctl and explicit-fencing support

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

headers: sync up amdgpu_drm.h with drm-next

Add sensor_info type
AMDGPU_INFO_SENSOR_STABLE_PSTATE_GFX_MCLK
AMDGPU_INFO_SENSOR_STABLE_PSTATE_GFX_SCLK

Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>

Revert "libdrm: intel/Android.mk: Filter libdrm_intel library requirements on x86/x86_64"

This reverts commit ed07718ae7bab596297abf210bb0c37c6dba58ed.

The commit added a guard since libpciaccess may be missing on some
setups. As of last commit there are no traces of the project, from
Android POV.

Hence, we can revert this workaround - which caused similar breakage to
the one it's trying to fix. This time in Mesa.

Cc: Rob Herring <rob.herring@linaro.org>
Acked-by: John Stultz <john.stultz@linaro.org>

intel: Do not use libpciaccess on Android

This patch makes the code not rely anymore on libpciaccess when compiled
for Android to eliminate ioperm() and iopl() syscalls required by that
library. As a side effect, the mappable aperture size is hardcoded to 64
MiB on Android, however nothing seems to rely on this value anyway, as
checked be grepping relevant code in drm_gralloc and Mesa.

Cc: Rob Herring <rob.herring@linaro.org>
Signed-off-by: Tomasz Figa <tfiga@google.com>
[Emil Velikov: rebase against master. add missing __func__, Eric]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: John Stultz <john.stultz@linaro.org>

xf86drmMode: merge successive mutually-exclusive #ifs

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: Use readdir instead of readdir_r to avoid build warnings

Building libdrm under AOSP, we see the following build warning:
external/libdrm/xf86drm.c:2861:12: warning: 'readdir_r' is deprecated: readdir_r is deprecated; use readdir instead [-Wdeprecated-declarations]
    while (readdir_r(sysdir, pent, &ent) == 0 && ent != NULL) {
           ^

Building on Linux with glibc produces the same warning.
Thus, this patch replaces readdir_r with readdir.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102031
Cc: Robert Foss <robert.foss@collabora.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Stefan Schake <stschake@gmail.com>
Cc: John Stultz <john.stultz@linaro.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
[Emil Velikov: remove unused variables, Eric]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

android: Add missing include exports

They were set for the static library but not the shared variant.

Signed-off-by: Stefan Schake <stschake@gmail.com>
Acked-by: John Stultz <john.stultz@linaro.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

omap: add Android build support

Add Android.mk file to build libdrm_omap library.

Signed-off-by: Gowtham Tammana <g-tammana@ti.com>
Signed-off-by: Andrew F. Davis <afd@ti.com>

libdrm: amdgpu: Adding DRM_RDWR flag in amdgpu_bo_export

Currently while exporting prime handle to fd read write access is
not granted. mmap fails because of this. mmap was not supported on
prime initially.
Here is link to related discussion
https://lists.freedesktop.org/archives/dri-devel/2017-February/131840.html

Adding the DRM_RDWR flag in amdgpu_bo_export to support mmap.

Signed-off-by: Satyajit <satyajit.sahu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>

meson: drop unnecessary variable

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

meson: move line to allow using `config` earlier

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

meson: drop unneeded dependency to libudev

libdrm only needed libudev for a few days 3 years ago,
between fde4969176822fe54197 and its revert 5b0e76f143887c4ec7db.

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

meson,configure: include config.h automatically

This will prevent any more missing `#include "config.h"` bug, at the
cost of having to recompile some files that didn't need to be when
changing build options.

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

meson: replace `if(compiles) have=true` with `have=compiles`

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

meson,configure: always define UDEV

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

meson,configure: always define HAVE_VISIBILITY

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

meson,configure: always define HAVE_OPEN_MEMSTREAM

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

tests: fix memory leak issue

Fixed memory leak issue to drmModeRes and drmModePlaneRes objects.

These objects were allocated by drmModeGetResources and
drmModeGetPlaneResources functions but not freed properly.

So this patch frees them by calling drmModeFreeResources
drmModeFreePlaneResources functions at failure case.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

libdrm: intel/Android.mk: Filter libdrm_intel library requirements on x86/x86_64

When building AOSP after updating libdrm project to the
freedesktop/master branch, I've seen the following build errors:

external/libdrm/intel/Android.mk: error: libdrm_intel
(SHARED_LIBRARIES android-arm64) missing libpciaccess
(SHARED_LIBRARIES android-arm64) You can set
ALLOW_MISSING_DEPENDENCIES=true in your environment if this is
intentional, but that may defer real problems until later in the
build.

Using ALLOW_MISSING_DEPENDENCIES=true when building allows
things to function properly, but is not ideal.

So basically, while I'm not including the libdrm_intel package
into the build, just the fact that the Android.mk file references
libpciaccess which isn't a repo included in AOSP causes the build
failure.

So it seems we need some sort of conditional filter in the
Android.mk to skip over it if we're not building for intel.

Cc: Chad Versace <chad.versace@linux.intel.com>
Cc: Marissa Wall <marissaw@google.com>
Cc: Sean Paul <seanpaul@google.com>
Cc: Dan Willemsen <dwillemsen@google.com>
Cc: Tomasz Figa <tfiga@google.com>
Cc: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Rob Herring <robh@kernel.org>

tests/exynos: remove dead condition

There is already condition checking input values between 2 and 4096
so condition checking 0 is always false. Remove the dead condition.

Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>