meson: Use feature.require() and feature.allowed()

To reduce the size and complexity of checks. require() allows combining
auto and enabled checks(), so that something like
```meson
x = get_option('feature')
y = false
if x.enabled()
  if not condition
    error(...)
  endif
  y = condition
endif
```
can be rewritten as:
```meson
y = get_option('feature').require(condition, error_message : ...).allowed()
```
require checks the condition, then if the feature is required it emits
an error with the given message otherwise it returns a disabled feature.
allowed then returns whether the feature is not disabled, and returns
that (ie, .allowed() == not .disabled()). This is especially helpful for
longer more complex conditions

Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>

meson: fix intel requirements

Intel requires libpciaccess and an x86/x86_64 host, so if those
aren't found and it's enabled we need to error

Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>

amdgpu: add support for querying VM faults information

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

amdgpu: amdgpu_drm.h for new GPUVM fault ioctl

Based on agd5f/drm-next.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

ci: bump FreeBSD to 13.2

13.0 is no longer supported and causes a 404 when fetching the
image, see e.g. [1] for example failure.

[1]: https://gitlab.freedesktop.org/mesa/drm/-/jobs/45849458

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>

xf86drm: mark DRM_MAX_MINOR as deprecated

With the work in [1], libdrm users should no longer rely on the
minor numbering scheme we've used so far. Instead, they should use
drmGetDevices2().

[1]: https://lore.kernel.org/lkml/20230724211428.3831636-1-michal.winiarski@intel.com/

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>

modetest: print modifiers in hex as well

Print modifiers in hex in addtion to in strings returned by
drmGetFormatModifierName.  In some cases, hex numbers can be more easily
compared visually.

Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>

modetest: allocate and commit atomic request around set_property()

Currently the atomic request is only assigned after `set_property()` is
called, leaving `dev.req` in its uninitialized state causing
`drmModeAtomicAddProperty()` to return an error code, which is printed
as `"Success"` because `errno` is not set by `libdrm` (but it would have
been when non-atomic `drmModeObjectSetProperty()` called an IOCTL
immediately):

    sony-akatsuki-row ~ $ modetest -M msm -a -w 81:ACTIVE:0
    failed to set CRTC 81 property ACTIVE to 0: Success

Solve this by assigning a new atomic request object before calling
`set_property()`, when there are properties to set.  Likewise, commit
these properties after `set_property()` even if there is no other
operation (setting modes or planes) specified.

Furthermore `drmModeObjectSetProperty()` is implemented in terms of
`DRM_IOCTL()` which already returns `-errno` when `ioctl()` returns
`-1`, so we should instead pass `ret` to `strerror()` and get an
accurate error string out of `drmModeAtomicAddProperty()` too.

Fixes: 93220283 ("tests/modetest: Add atomic support")
Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>

modetest: allow using -r and -P

Since now -r sets the pipe struct and count like -s we can also
use -P with -r.

Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>

modetest: permit -r and -s to work together

Let's permit testing vsync with the default mode, this returns
back the pipe content and count when calling set_mode() so the
vsync test can also be used.

Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>

modetest: document why no blob is created for linear gamma LUT

As found and discussed in [MR 58] a blob is not created in the else arm
because adding the GAMMA_LUT property with a NULL/0 blob_id causes it
to be reset to a default linear / pass-thru gamma table.  The values
in the gamma_lut table might still be consumed in the legacy API path
below though, so it has to be initialized to a linear table.

[MR 58]: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/58#note_466972

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>

modetest: avoid erroring if there's no gamma legacy support

Let's follow the Rule of Silence. And while here,
document what's going on.

Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>

modetest: custom mode support

It is useful to be able to specify mode parameters manually. Add support
for setting user-supplied modes. This patch is based on the original
idea by Rohit and Jessica, but implemented from scratch.

Suggested-by: Rohith Iyer <quic_rohiiyer@quicinc.com>
Suggested-by: Jessica Zhang <quic_jesszhan@quicinc.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>

util: remove unused definitions of RED, GREEN, and BLUE

These are unused since commit edcef53685edf5fb ("modetest: Add test
pattern support for missing RGB formats").

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
v2:
  - Add Reviewed-by.

modetest: fix mode_vrefresh() for interlace/dblscan/vscan

mode_vrefresh() does not take into account interlaced, doublescan, and
multiscan modes, leading to incorrect refresh rates.

Fix this, based on drm_mode_vrefresh() in Linux.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>

util: fix grey in YUV SMPTE patterns

The YUV SMPTE patterns use RGB 191/192/192 instead of 192/192/192 for
the grey color in the top color bar.

Change it to 192/192/192, to match the RGB SMPTE patterns.

Fixes: a94ee624292bff96 ("modetest: Add SMPTE test pattern")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
v2:
  - Add Reviewed-by.

modetest: add support for DRM_FORMAT_NV{24,42}

Add support for creating buffers using semi-planar YUV formats with
non-subsampled chroma planes.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
---
v2:
  - Add Reviewed-by.

util: add pattern support for DRM_FORMAT_NV{24,42}

Add support for drawing the SMPTE and tiles patterns in buffers using
semi-planar YUV formats with non-subsampled chroma planes.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
---
v2:
  - Add Reviewed-by.

util: add NV24 and NV42 frame buffer formats

Add the missing entries for semi-planar YUV formats with
non-subsampled chroma planes.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
---
v2:
  - Add Reviewed-by.

modetest: add support for writeback connector

Add writeback support to modetest with the below options:

- Passing in -a -c will now also show the writeback connector

- Dump the writeback output buffer to bitstream
  Usage: "./modetest -M msm -s <connector_id>:<widthxheight>
          -a -o <filepath>
          -P <plane_id>@<crtc_id>:<widthxheight>+0+0@RG24"

This currently supports single writeback connector.

Co-developed-by: Rohith Iyer <quic_rohiiyer@quicinc.com>
Signed-off-by: Jessica Zhang <quic_jesszhan@quicinc.com>
[DB: dropped custom mode support, fixed segfault]
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>

amdgpu: add marketing names from Adrenalin 23.9.1

amdgpu: add marketing names from Adrenalin 23.7.2

amdgpu: add marketing names from PRO Edition 23.Q3 W7000

amdgpu: add marketing names from amd-5.5.1 (23.10.1)

amdgpu: add marketing names from amd-5.4.6 (22.40.6)

amdgpu: Use PRI?64 to format uint64_t

On 32-bit:

    ../tests/amdgpu/amdgpu_stress.c: In function ‘alloc_bo’:
    ../tests/amdgpu/amdgpu_stress.c:178:49: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 4 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Allocated BO number %u at 0x%lx, domain 0x%x, size %lu\n",
                                                   ~~^
                                                   %llx
       num_buffers++, addr, domain, size);
                      ~~~~
    ../tests/amdgpu/amdgpu_stress.c:178:72: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 6 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Allocated BO number %u at 0x%lx, domain 0x%x, size %lu\n",
                                                                          ~~^
                                                                          %llu
       num_buffers++, addr, domain, size);
                                    ~~~~
    ../tests/amdgpu/amdgpu_stress.c: In function ‘submit_ib’:
    ../tests/amdgpu/amdgpu_stress.c:276:54: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 5 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
                                                        ~~^
                                                        %llx
       count, from, virtual[from], to, virtual[to], copied, delta / 1000);
                    ~~~~~~~~~~~~~
    ../tests/amdgpu/amdgpu_stress.c:276:65: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 7 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
                                                                   ~~^
                                                                   %llx
       count, from, virtual[from], to, virtual[to], copied, delta / 1000);
                                       ~~~~~~~~~~~
    ../tests/amdgpu/amdgpu_stress.c:276:70: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 8 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
                                                                        ~~^
                                                                        %llu
       count, from, virtual[from], to, virtual[to], copied, delta / 1000);
                                                    ~~~~~~
    ../tests/amdgpu/amdgpu_stress.c:276:85: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 9 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
      fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
                                                                                       ~~^
                                                                                       %llu
       count, from, virtual[from], to, virtual[to], copied, delta / 1000);
                                                            ~~~~~~~~~~~~
    ../tests/amdgpu/amdgpu_stress.c: In function ‘parse_size’:
    ../tests/amdgpu/amdgpu_stress.c:296:24: warning: format ‘%li’ expects argument of type ‘long int *’, but argument 3 has type ‘uint64_t *’ {aka ‘long long unsigned int *’} [-Wformat=]
      if (sscanf(optarg, "%li%1[kmgKMG]", &size, ext) < 1) {
                          ~~^             ~~~~~
                          %lli
    ../tests/amdgpu/amdgpu_stress.c: In function ‘main’:
    ../tests/amdgpu/amdgpu_stress.c:378:45: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 3 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
         fprintf(stderr, "Buffer size to small %lu\n", size);
                                               ~~^     ~~~~
                                               %llu

Fix this by using the proper "PRI?64" format specifiers.

Fixes: d77ccdf3ba6f5a39 ("amdgpu: add amdgpu_stress utility v2")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
---
On Linux/amd64, the format strings in the resulting binary are
unchanged.

v3:
  - Add Reviewed-by,

v2:
  - Use PRI?64 to unbreak 64-bit build.

amdgpu: Fix pointer/integer mismatch warning

On 32-bit:

    ../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
    ../amdgpu/amdgpu_bo.c:554:13: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
           cpu < (void*)((uintptr_t)bo->cpu_ptr + bo->alloc_size))
                 ^

Indeed, as amdgpu_bo_info.alloc_size is "uint64_t", the sum is
always 64-bit, while "void *" can be 32-bit or 64-bit.

Fix this by casting bo->alloc_size to "size_t", which is either
32-bit or 64-bit, just like "void *".

Fixes: c6493f360e7529c2 ("amdgpu: Eliminate void* arithmetic in amdgpu_find_bo_by_cpu_mapping")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
---
v2:
  - Add Reviewed-by.

build: bump version to 2.4.116

Signed-off-by: Simon Ser <contact@emersion.fr>

nouveau: add interface to make buffer objects global

This is useful for when GEM handles are exported and may be shared
between multiple buffer objects without going through other libdrm
interfaces.

Signed-off-by: Dor Askayo <dor.askayo@gmail.com>
Reviewed-by: Karol Herbst <git@karolherbst.de>

xf86drm: use drm device name to identify drm node type

Currently drm node's minor range is used to identify node's type.
Since kernel drm uses node type name and minor to generate drm
device name, It will be more general to use drm device name to
identify drm node type.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Simon Ser <contact@emersion.fr>

xf86drm: update DRM_NODE_NAME_MAX supporting more nodes

Current DRM_NODE_NAME_MAX only can support up to 999 nodes,
Update to support up to 2^MINORBITS nodes.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Simon Ser <contact@emersion.fr>

xf86drm: add drmSyncobjEventfd

This is a wrapper for DRM_IOCTL_SYNCOBJ_EVENTFD.

Signed-off-by: Simon Ser <contact@emersion.fr>

Sync headers with drm-next

Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.

Generated using make headers_install.
Generated from drm-next branch commit 52920704df878050123dfeb469aa6ab8022547c1

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drm: drop control nodes implementation

Drop support for control nodes. The kernel never returns such
nodes. Stop trying to detect and handle them, and always return
an error when a caller tries to open them.

The header is left untouched to avoid breaking libdrm's API.

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drm: bump DRM_MAX_MINOR to 64

This is what the kernel uses (see drm_minor_alloc).

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drmMode: constify drmModeCrtcSetGamma

The data is never modified, so it should be const

Signed-off-by: Xaver Hugl <xaver.hugl@gmail.com>

tests/amdgpu/vcn: fix session buffer issue for vcn1-vcn3

issue:
   in vcn1-vcn3, session buffer was not truly added, it shows
   decoding creation commands cannot be sent multiple times.

problem:
   session buffer has to be the first buffer sending out.
   Otherwise, system could assume session buffer doesn't
   exist.

solution:
   move session buffer sending sequence to be the first one.

Reviewed-by: Veerabadhran Gopalakrishnan <Veerabadhran.Gopalakrishnan@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>

tests/amdgpu/vcn:update decoder unit test

update decoder unit test with session context buffer for VCN v1 to v4

v2: remove multiple checks for vcn4 (Ruijing Dong)

Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>

amdgpu: add an environment variable that overrides the context priority

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

tests/amdgpu/vcn: fix drm test failure

1. fixed an issue that drm test vcn3/4 encoding test
   could cause VCN engine stuck.
2. adding missing or errous encoding ib package members.

Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>

tests/amdgpu/vcn: change vbv_buffer name to input

It is input buffer instead of vbv_buffer.
Correct its name.

Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>

tests/amdgpu: Allow to exclude a test or a suite of tests

Add the command line argument -e s[.t] to exclude (disable) suite s, or to
exclude suite s test t.

This is useful for instance to run the Basic Suite, but disable the GPU reset
test, on the command line, like this:

    amdgpu_tests -s 1 -e 1.13

This option can be specified more than once on the command line, in order to
exclude more than one suite and/or suite and test combination from being run.

Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

tests/amdgpu: Add all 9 options to the help output

Add -s and -t to the help output, as well as sort
the options output alphabetically.

v1: Fix a spelling in the subject of this commit.

Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

tests/amdgpu: Fix Usage string

Fix the Usage: string on -h (help) in amdgpu_tests.c,
so brackets match, and remove mismatched angle brackets.

Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

amdgpu: add marketing names from amd-5.4.3 (22.40.3)

amdgpu: remove va::dev member

This is unused so drop it.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

test/amdgpu/hotunplug: add apu check for hotplug test

For apu, it is integrated with cpu.
So hotplug test should be unnecessary for it.

Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Flora Cui <flora.cui@amd.com>

meson: drop pthread-stubs dependency on BSDs

pthread-stubs >= 0.4 simply passes -pthread which is similar to what
dependency('threads') returns. And make it a private dependency
for subprojects even on Linux.

Reviewed-by: Emmanuel Vadot <manu@FreeBSD.org>

tests/util: Add vkms driver

Add an entry for the "vkms" driver, so that the test utilities work with
the vkms driver without passing the -M argument.

Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Yi Xie <yixie@google.com>

build: bump version to 2.4.115

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drm: add support for printing AMD GFX11 modifiers

See kernel commit 543036a2de71 ("drm/amd: Add GFX11 modifiers support
to AMDGPU (v3)").

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: add some additional marketing names

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: add marketing names from amd-5.4 (22.40)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

xf86drm: fix warning in drmGetFormatModifierNameFromVivante()

Fixes the following warning:

    ../xf86drm.c: In function ‘drmGetFormatModifierNameFromVivante’:
    ../xf86drm.c:614:14: warning: passing argument 1 of ‘asprintf’ from incompatible pointer type [-Wincompatible-pointer-types]
      614 |     asprintf(&mod_vivante, "%s%s%s", color_tiling, tile_status, compression);
          |              ^~~~~~~~~~~~
          |              |
          |              const char **
    In file included from ../xf86drm.c:34:
    /usr/include/stdio.h:396:40: note: expected ‘char ** restrict’ but argument is of type ‘const char **’
      396 | extern int asprintf (char **__restrict __ptr,
          |                      ~~~~~~~~~~~~~~~~~~^~~~~
    ../xf86drm.c:615:12: warning: return discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
      615 |     return mod_vivante;
          |            ^~~~~~~~~~~

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drm: Add support for decoding Vivante format modifiers

Allow applications to pretty-print Vivante format modifiers.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

drm_fourcc: sync drm_fourcc with latest drm-next kernel

Update drm_fourcc.h to include latest changes from drm-next branch.
This brings in sub-8bpp formats, AVUY and XVUY 8:8:8:8, and
Vivante tile-status and compression modifiers.

Generated using make headers_install.
Generated from drm-next branch commit 077bd80083ab

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

intel: Eliminate need to keep adding PCI IDs

The Intel-specific code in libdrm is continually updated with new PCI
IDs for each new platform so that we can recognize the IP version
properly.  However this is mostly a pointless exercise; none of the
Intel code in libdrm is conditional on IP versions above 8.  If we just
treat any future unrecognized Intel platforms as IP version 8, we should
get the same behavior without the need for continued PCI ID updates.

Note that the intel_decode tool probably _should_ have conditions on
newer IP versions, but it was last updated for gen8 and has been
bitrotting from gen9 onward.  This change won't make the tool behave any
more incorrectly than it already does today.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

tests/amdgpu/jpeg: enable unit test for jpeg 4

Enable decode unit test for jpeg4.

Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Sathishkumar S <sathishkumar.sundararaju@amd.com>

tests/amdgpu: use AMDGPU_TIMEOUT_INFINITE to query fence

We need to wait longer when running on emulator.

Signed-off-by: Lang Yu <Lang.Yu@amd.com>

build: bump version to 2.4.114

Signed-off-by: Simon Ser <contact@emersion.fr>

modetest: use dumb buffer helpers

Signed-off-by: Simon Ser <contact@emersion.fr>

modetest: use sized integers in struct bo

Use the same size types as the kernel.

Signed-off-by: Simon Ser <contact@emersion.fr>

modetest: drop unused offset field in struct bo

Signed-off-by: Simon Ser <contact@emersion.fr>

xf86drmMode: add helpers for dumb buffers

Up until now, DRM clients had to hand-roll their code to create,
destroy and map dumb buffers. This is slightly inconvenient,
a bit error-prone, and not easily discoverable.

Introduce wrappers for these operations, just like we have for
other KMS IOCTLs.

Signed-off-by: Simon Ser <contact@emersion.fr>

amdgpu: silence uninitialized variable warning

The compiler isn't smart enough to tell that this can't happen:

    [30/74] Compiling C object amdgpu/libdrm_amdgpu.so.1.0.0.p/amdgpu_bo.c.o
    In file included from ../amdgpu/amdgpu_internal.h:32,
                     from ../amdgpu/amdgpu_bo.c:39:
    ../xf86atomic.h: In function ‘amdgpu_find_bo_by_cpu_mapping’:
    ../xf86atomic.h:47:54: warning: ‘bo’ may be used uninitialized [-Wmaybe-uninitialized]
       47 | # define atomic_inc(x) ((void) __sync_fetch_and_add (&(x)->atomic, 1))
          |                                                      ^
    ../amdgpu/amdgpu_bo.c:536:27: note: ‘bo’ was declared here
      536 |         struct amdgpu_bo *bo;
          |                           ^~

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: Add a default marketing name if none is found

Apparently quite a few apps use this API to get the GPU
name and end up with NULL as the GPU name.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu_ids: add MI marketing names

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu.ids: update to the latest marketing name

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu.ids: sort the file

So the list is in numeric order.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu.ids: use consistent formatting for RID

Use two digits for the revision id.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

xf86drm: handle DRM_FORMAT_BIG_ENDIAN in drmGetFormatName()

This bit can be added to a DRM format to indicate that it's
big endian instead of little endian.

Signed-off-by: Simon Ser <contact@emersion.fr>

tests/util: add imx-lcdif driver

This makes the test utilities work with the i.MX LCDIFv3 driver
without the necessity of using the -M argument.

Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>

build: automatically disable Intel if pciaccess is not found

Wire up the pciaccess dep to the intel option. This automatically
skips the dep if intel is explicitly disabled, fails if intel is
explicitly enabled and it's not found, and disables intel if it's
set to auto and the dep is not found.

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

intel: move declarations to top in drm_intel_gem_bo_unreference()

Fixes the following warning:

    [65/74] Compiling C object intel/libdrm_intel.so.1.0.0.p/intel_bufmgr_gem.c.o
    ../intel/intel_bufmgr_gem.c: In function â\80\98drm_intel_gem_bo_unreferenceâ\80\99:
    ../intel/intel_bufmgr_gem.c:1388:9: warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]
     1388 |         drm_intel_bufmgr_gem *bufmgr_gem =
          |         ^~~~~~~~~~~~~~~~~~~~

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

include/drm/i915_drm.h: Update from Linux v6.0-rc7

Generated from the Linux v6.0-rc7 tag with a sha1 of
f76349cf41451c5c42a99f18a9163377e4b364ff.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

include/drm/drm_fourcc.h: Update from Linux v6.0-rc7

Generated from the Linux v6.0-rc7 tag with a sha1 of
f76349cf41451c5c42a99f18a9163377e4b364ff.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

meson: fast-fail on unsupported OSes

It's not worth even attempting to configure anything on OSes where there
is no DRM to have a userspace library for.

This failure message can be useful in e.g. the case where libdrm is an
optional wrap fallback in another project.

Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

build: bump to version 2.4.113

Signed-off-by: Simon Ser <contact@emersion.fr>

tests/modetest: use drmGetFormatName()

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>

xf86drm: add drmGetFormatName()

Same as drmGetFormatModifierName() but for formats.

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>

intel: Hook up new platforms IDs

In commit 98794e2a0d ("lib: sync i915_pciids.h with kernel") we
resynchronized the PCI header with the kernel to bring in the
definitions for several new platforms.  But before those IDs will be
recognized, we still need to hook them up in the libdrm chipset code as
well.

References: https://gitlab.freedesktop.org/drm/intel/-/issues/5416
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

meson: auto-enable etnaviv on arm, arc, mips and loongarch architectures

There is a Vivante GC1000 gpu in LS2K1000 and LS7A1000.

LS7A1000 is a bridge chip made by Loongson corporation
which act as north and/or south bridge of loongson's
desktop and server level processor. It is equivalent
to RS780E or something like that. In fact, the company
use RS780E as bridge of LS3A3000 at its early stage,
but as RS780E is out of stock long long time ago, the
company have to made one by themself. More details can
be read from its user manual[1].

This bridge chip typically use with LS3A3000, LS3A4000
and LS3A5000.

LS3A3000 is 4 core 1.45gHz mips64r2 compatible cpu.
LS3A4000 is 4 core 1.8gHz mips64r5 compatible cpu.
LS3A5000 is 4 core 2.5gHz loongarch cpu, the company
acclaim that loongarch a new archtecture with its
instruction set is released[2].

LS2K1000 is a double core 1.0Ghz mips64r2 compatible SoC[3].

we need to enable it to test and developing driver on above
listed archtecture.

[1] https://loongson.github.io/LoongArch-Documentation/Loongson-7A1000-usermanual-EN.html
[2] https://loongson.github.io/LoongArch-Documentation/Loongson-3A5000-usermanual-EN.html
[3] https://wiki.debian.org/InstallingDebianOn/Lemote/Loongson2K1000

Signed-off-by: Sui Jingfeng <15330273260@189.cn>
[Eric: rebase over meson changes, add ARM & ARC architectures, and drop
"experimental" from the description]
Signed-off-by: Eric Engestrom <eric@engestrom.ch>

intel: Avoid aliasing violation

../intel/test_decode.c: In function ‘compare_batch’:
../intel/test_decode.c:109:39: error: dereferencing type-punned pointer might break strict-aliasing rules [-Werror=strict-aliasing]
  109 |         out = open_memstream((char **)&ptr, &size);
      |                                       ^~~~
cc1: some warnings being treated as errors

The fix is simple: just declare `ptr` as a `char *` to begin with.

atomic: fix atomic_add_unless() fallback's return value

According to the kernel documentation:
  Returns non-zero if @v was not @u, and zero otherwise.

Fixes: 63fc571863aa64683400 ("atomic: add atomic_add_unless()")
Closes: https://gitlab.freedesktop.org/mesa/drm/issues/17
Signed-off-by: David Shao <davshao@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
[Eric: fix its callers to maintain current behaviour]
Signed-off-by: Eric Engestrom <eric@engestrom.ch>

Remove unused 3rd parameter to open(2)

The 3rd parameter is only used with the O_CREAT flag

Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>

meson: simplify some more build options by using features

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Simon Ser <contact@emersion.fr>

meson: convert auto combos into proper features

Allows users to easily enable everything (eg. packagers), or select just
the drivers they want with something like:
    -D auto-features=disabled -D amdgpu=enabled

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Simon Ser <contact@emersion.fr>

ci: drop dead script since e722ba9f67

Noticed-by: Michel Daenzer <michel@daenzer.net>

meson: fix value of `auto` for a bunch of drivers

You can't have an error if your driver is requested by you're missing
a dep, but then happily build that driver without the dep in `auto`.

Signed-off-by: Eric Engestrom <eric@engestrom.ch>

ci: bump images tags to take !255 into effect

Reviewed-by: Simon Ser <contact@emersion.fr>

ci/freedesktop: bump python version of docutils package as 3.8 no longer exists

ci: fix the tested meson version
Reviewed-by: Simon Ser <contact@emersion.fr>

ci: remove system meson before installing the pip one
Reviewed-by: Simon Ser <contact@emersion.fr>

build: make tests optional

Building the project as a meson subproject, meson inherits the warning level
from the parent project. Making the tests optional bypasses that issue and
reduces build time.

Signed-off-by: Simon Zeni <simon@bl4ckb0ne.ca>
Reviewed-by: Simon Ser <contact@emersion.fr>

tests/amdgpu: add sdma slow copy linear hang test

Issue slow copy linear for sdma to trigger SDMA hang test.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>

tests/amdgpu: add sdma corrupted header hang test

Issue corrupted header for sdma to trigger SDMA hang test.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>

tests/amdgpu: add dispatch/draw test for gfx11

Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>

tests/amdgpu: refactor dispatch/draw test

Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>

tests/amdgpu: skip gfx CE subtest in gfx11

CE is not available in gfx11

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Likun Gao <Likun.Gao@amd.com>

lib: sync i915_pciids.h with kernel

This synchronizes with kernel commit 7835303982d1 ("drm/i915/mtl: Add
MeteorLake PCI IDs") to bring in the missing PCI IDs for several recent
platforms.

These days adding PCI IDs to libdrm doesn't really matter for real-world
system usage.  However there are still a few driver testing situations
where they're needed (such as the IGT dma-buf tests that still rely on
libdrm's bufmgr code).  At some point we should probably break that
final IGT dependency on libdrm so that these PCI ID resyncs won't be
necessary anymore, but that hasn't happened yet.

References: https://gitlab.freedesktop.org/drm/intel/-/issues/5416
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>