tests/amdgpu: add atomic dma command to verify the secure buffer (v2)

DMA's atomic behavir is unlike GFX,If the comparing data is not
equal to destination data,
For GFX, loop again till gfx timeout(system hang).
For DMA, loop again till timer expired and then send interrupt.
So testcase can't use interrupt mechanism.
We take another way to verify. When the comparing data is not
equal to destination data, overwrite the source data to the destination
buffer. Otherwise, original destination data unchanged.
So if the bo_cpu data is overwritten, the result is passed.

Steps:
1. use linear write packet to write 0xdeadbeaf to secure buffer,
2. use atmoic packet and ATOMIC_CMPSWAP_RTN_32 opcode to compare
the cmp_data(0xdeadbeaf) to the written data which has been encrypted.

v2: add the case of (dest_data != cmp_data).

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add test to submit a gfx command with secure context

This patch is to test the command submission with secure context.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add atomic_mem cp_packet to verify the secure buffer

Secure buffer is only able to be read with trusted ip block.
So we need use GFX ip to read it back instead of CPU.
Steps:
1. use write_data packet to write 0xdeadbeaf to secure buffer,
2. use atmoic_mem packet and ATOMIC_CMPSWAP_RTN_32 opcode to compare
the cmp_data(0xdeadbeaf) to the written data which has been encrypted.
If the result is equal, then overwrite the src_data(0x12345678) to the
secure buffer and return directly. Otherwise loop again until gfx timeout
and the secure buffer data unchanged.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: expand secure param for exec_cs_helper (v2)

This patch expands secure param for amdgpu_test_exec_cs_helper_raw.
The flag is transfered to kernel with cs.

v2: squash in change from context to IB flag

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add device handle as input param for exec_cs_helper and write_linear_helper (v4)

This patch is to add add device handle as input param for exec_cs_helper
and write_linear_helper.

Because they are needed in security tests.

v2: fix typo that basic tests should be un-secure.
v3: refine the function implementation.
v4: remove amdgpu_cs_ctx_create3 calling.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: expand write linear helper for security (v3)

This patch expand write linear helper for security to submit the command
with secure context.

v2: refine the function implementation.
v3: remove amdgpu_cs_ctx_create3.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add secure buffer allocation test for invisible VRAM

This patch is to add secure buffer allocation test for invisible VRAM.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add secure buffer allocation test for system memory

This patch is to add secure buffer allocation test for system memory.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add security test suite (v2)

This patch is to add a new test suite to store security tests.
In Raven+ asics, it will support TMZ (trust memory zone), and it is
page-based protection feature.

v2: remove tests/amdgpu/Makefile.am and update to
tests/amdgpu/meson.build

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: sync up amdgpu_drm.h with latest from kernel

From drm-next:

commit c41219fda6e04255c44d37fd2c0d898c1c46abf1
Merge: e20bb857dea2 d96536f0fe69
Author: Dave Airlie <airlied@redhat.com>
Date:   Thu May 21 10:44:32 2020 +1000

    Merge tag 'drm-intel-next-fixes-2020-05-20' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

    Fix for TypeC power domain toggling on resets (Cc: stable).
    Two compile time warning fixes.

Signed-off-by: Dave Airlie <airlied@redhat.com>
    From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200520123227.GA21104@jlahtine-desk.ger.corp.intel.com
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: disable unsupported test cases for Arcturus

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: move arcturus asic check function to common place

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu/vcn: add Arcturus decode test support

Add Arcturus decode test support only

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: disable gfx engine basic test cases for Arcturus

Since Arcturus has no gfx pipeline(CPG), cases below is not suitable:
  - Command submission Test (GFX)
  - Command submission Test (Multi-Fence)
  - Sync dependency Test

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: create Active function for basic test suite

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

tests/amdgpu: add function to check Asic is Arcturus

Since Arcturus has no gfx engine, add function to blacklist gfx related test.

Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

amdgpu: add marketing names from 20.40

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

intel: add INTEL_DG1_IDS to the pciids list

This enables drm_intel_bufmgr on DG1 and allows us to pass dmabuf
import/export related tests with Piglit.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>

amdgpu: add device IDs for Raven, Picasso and Renoir

intel: sync i915_pciids.h with kernel

Add DG1 and clean-up VLV PCI IDs.

Align with kernel commits:
f2bde2546b81 ("drm/i915: Remove dubious Valleyview PCI IDs")
fd38cdb81105 ("drm/i915/dg1: Add DG1 PCI IDs")

Signed-off-by: Adam Miszczak <adam.miszczak@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

man: Update the bug URL to gitlab.freedesktop.org

Point users to the GitLab issue tracker instead of Bugzilla, which is no
longer used.

Signed-off-by: Jeremy Cline <jcline@redhat.com>

intel: Add PCI ID support to RKL platform

    Missing RKL PCI ID preventing below test cases to succeed on RKL Platform.
    igt@kms_frontbuffer_tracking
    igt@kms_draw-crc
    igt@kms_big_fb

Signed-off-by: sunil kumar dora sermsity <sunilx.kumar.dora.sermsity@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>

amdgpu: Add Device IDs for Embedded Raven2 platforms

Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Pavan Kumar Ramayanam <pavan.ramayanam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

intel: sync i915_pciids.h with kernel

Two new patches landed in kernel adding new PCI ids:
123f62de419f ("drm/i915/rkl: Add RKL platform info and PCI ids")
52797a8e8529 ("drm/i915/ehl: Add new PCI ids")

Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>

tests/amdgpu: clear the extension flag

This is workaround of firmware issue, and the change has no impact
on the legacy HW.

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>

tests/amdgpu: clear msg decode flag

It is not used for VCN from VCN1, but VCN3 use it
for other feature, so clear it, because we don't
use the feature for now

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>

tests/amdgpu: add VCN3.0 regs support

VCN3.0 has its own set of internal regs

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>

xf86drm.c: fix build failure

./xf86drm.c: In function 'drmNodeIsDRM':
../xf86drm.c:2825:7: error: "__FreeBSD__" is not defined [-Werror=undef]
 #elif __FreeBSD__
       ^
../xf86drm.c: In function 'drmGetMinorNameForFD':
../xf86drm.c:2938:7: error: "__FreeBSD__" is not defined [-Werror=undef]
 #elif __FreeBSD__
       ^
../xf86drm.c: In function 'drmParsePciBusInfo':
../xf86drm.c:3258:7: error: "__FreeBSD__" is not defined [-Werror=undef]
 #elif __FreeBSD__
       ^
../xf86drm.c: In function 'drmParsePciDeviceInfo':
../xf86drm.c:3427:7: error: "__FreeBSD__" is not defined [-Werror=undef]
 #elif __FreeBSD__
       ^
../xf86drm.c: In function 'drmGetDeviceNameFromFd2':
../xf86drm.c:4305:7: error: "__FreeBSD__" is not defined [-Werror=undef]
 #elif __FreeBSD__
       ^

Signed-off-by: Heiko Thiery <heiko.thiery@gmail.com>

core: use `O_RDONLY` instead of ambiguous `0` flag

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

amdgpu: add marketing names from 20.10

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Bump version to 2.4.102

Signed-off-by: Dave Airlie <airlied@redhat.com>

modetest: Add a new "-r" option to set a default mode

This option finds all connected connector and then sets its preferred
mode on it. If no preferred mode is available, first mode is used.

This option must be set w/o any mode or plane.

This allows for a quick test on all connected outputs.

Loosely based on the work by Ezequiel Garcia <ezequiel@collabora.com>

Changes since Ezequiel's work:
 - implement atomic codepath
 - set all connectors
 - pick correct crtc
 - don't set -r by default
 - nearly identical output in atomic and non-atomic codepaths

v2:
 - Use the crtc->crtc_id, instead of the plane's current crtc_id

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: reorder atomic path alike the non-atomic

Makes the code a tiny bit more symmetrical.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: don't error out of final AtomicCommit

The very final drmModeAtommicCommit tears down the existing mode/plane
setup. Following it we clean up other misc state laying around.

Chances are that it will not fail, but in the extremely unlikely case it
does, there's nothing one can do.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: factor out atomic pageflip test

Move the hunk of code into a function, making the overall flow easier to
follow and providing some symmetry to the non-atomic path.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: push pipe_resolve_connectors() to set_mode

The function is closely related to pipe_find_crtc_and_mode() so we might
as well keep them together.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: move pipe_resolve_connectors() further up

Move the function above set_mode, since we'll be using it from there as
of next commit.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: unify {,atomic_}set_mode()

Instead of duplicating the exact same code across the two functions,
fold them into one.

For some strange reason git diff may show atomic_clear_mode() as changed
The function in untouched, despite the misleading output.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: get the crtc_id from the pipe_arg

Makes the code shorter and easier to read.

Currently if the user has not set the crtc_id, we fetch the crtc yet do
not "bother" setting the id - do so.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: introduce and use get_crtc_by_id() and get_crtc_mask()

Let's make the code shorter, this avoid crashes (when drmModeGetCrtc()
fails) by using a couple of helpers. As get_resources() considers the
drmModeGetCrtc() fail non-fatal, we might as well handle it properly.

v2: Add a comment above the unreachable abort() (Eze)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: close the device on exit

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: remove drmMode{,Plane}Res

There's no point in keeping these around since we already fetch the
complete data set. Add respective count_ variables and greatly simplify
the existing code.

Extra brownie points for:
 - using the inverse order in free_resources()
 - don't memory leak the connector properties
 - free the properties themselves, instead of only the objects

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: add and use bo_fb_create() helper

Flesh out the bo_create + drmModeAddFB2 dance into a helper and use it.
Currently we're duplicating that in 4 places, many of which leaking et
al.

As a bonus point this highlights that the atomic_set_plane() seems tad
buggy. That'll be fixed with separate commit.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: set atomic cap, _only_ when needed

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: move basic args check before open()

Don't bother opening the device node, if the args combination is invalid

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: remove cursor/page_flipping_supported stubs

The two functions have been stubs for ages. The alluded generic ioctls
never came to be, assumingly because all new drivers support those.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

modetest: simplify "dump all" logic

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>

meson: require valgrind 3.10.0 to enable it with freedreno

Freedreno uses VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE that was
introduced in Valgrind 3.10.0

Raspbian Buster includes Valgrind 3.7.0, so when valgrind is installed
as freedreno is build by default the build becomes broken. So lets
require 3.10 to enable valgrind when freedreno is built.

v2: Keep the arguments listed in the same order (Emil Velikov)

Closes: https://gitlab.freedesktop.org/mesa/drm/-/issues/37
Signed-off-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

nouveau: sync up with nouveau_abi16.h and nouveau_drm.h

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>

drm mode : fix memory leak when freeing drmModePropertyPtr

Closes: https://gitlab.freedesktop.org/mesa/drm/-/issues/5
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmGetDeviceNameFromFd: Always return /dev/dri/ node for FreeBSD

Since we now always returns the /dev/dri/ node for
drmGet<nodetype>DeviceNameFromFd, be consistant with the names returned
in drmGetDeviceNameFromFd.

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>

libdrm: drmGetMinorNameFromFd: Fix FreeBSD variant

Fix the FreeBSD variant by getting the node type represented by fd to deduce
the target minor name.
We then return the full /dev/dri/<minorname><id> version.

Fix: #41
Fixes: 6818a50b12
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>

nouveau: fix compile error with -DDEBUG on newer gcc

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

libdrm: intel: add DRM_RDWR flag in drm_intel_bo_gem_export_to_prime

This is similar to b81d44d587d1706d5c7568e539340632a748782b: the
DRM_RDWR flag is needed for mmap to work.

Signed-off-by: Nicholas Bishop <nicholasbishop@gmail.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

intel: properly escape sed pattern for tests

The sed was incorrectly modifying e.g. "nicholasbishop" to
"nicholasbop". The updated pattern will only match `.sh` at the end of
the string.

Signed-off-by: Nicholas Bishop <nicholasbishop@gmail.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

tests/amdgpu/vcn: add Renoir VCN2.0 decode support

Renoir is the same family as Raven, but it's with VCN2.0,
so it has to use VCN2.0 reg set

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-By: Thong Thai <thong.thai@amd.com>

libdrm: Handle usb_interface devices for usb parsing

Currently the code expects that the device found at
/sys/char/$maj:$min/device for USB devices is a "usb_device". However,
at least for some devices, such as for the udl driver, they are instead
a "usb_interface".

A usb_interface is a child of the usb_device we're interested in, so we
walk up one in the /sys path to get there.

For example, with a USB device I have, trimmed to show the relevant
information:
```
$ udevadm info /dev/dri/card1
P: /devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4/1-4:1.0/drm/card1
E: DEVTYPE=drm_minor
$ udevadm info /sys/devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4/1-4:1.0
E: DEVTYPE=usb_interface
E: DRIVER=udl
$ udevadm info /sys/devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4
E: DEVTYPE=usb_device
E: DRIVER=usb
E: BUSNUM=001
E: DEVNUM=009
```

Signed-off-by: Scott Anderson <scott@anderso.nz>

xf86drm: Check non-absolute path only for virtio based devices

This fixes bug in drmParseSubsystemType() that cases situation when
subsequent call to readlink() from get_subsystem_type() will result in
EACCESS.

Signed-off-by: Mikhail Golubev <mikhail.golubev@opensynergy.com>

tests/amdgpu: needs atomic_ops

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Acked-by: Alex Deucher alexander.deucher@amd.com
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

tests: install drmdevice

Just like the other tests (tools really) install drmdevice. It is a
simple tool which is useful for basic check/testing.

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

tests/nouveau/threaded: adapt ioctl signature for FreeBSD

FreeBSD also use (int, unsigned long int, ...) like GLIBC.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: Implement drmParsePciDeviceInfo for FreeBSD

The FreeBSD kernel expose a pseudo-device /dev/pci to obtain information
about present PCI device.
Uee the PCIOCGETCONF ioctl on this device to look up the desired device
information.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: get_pci_path is Linux only so add an ifdef

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: Add get_sysctl_pci_bus_info for FreeBSD

The FreeBSD drm driver expose a sysctl hw.dri.%d.busid which contain
the busid.
Use this sysctl to parse the busid information based on the major/minor
that allow us to implement FreeBSD support for drmParsePciBusInfo.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmCheckModesettingSupported: fix for FreeBSD

FreeBSD only support up to 10 GPUs not 16.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmCheckModesettingSupported: Fix for FreeBSD

Remove some useless busid rewritting.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmGetDeviceNameFromFd2: Add FreeBSD variant

The FreeBSD variant of drmGetDeviceNameFromFd can already handle
the different node type so just call it.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: Default to PCI for FreeBSD

FreeBSD have some support for DRM on !PCI device but no code is currently
upstream. Default to PCI for now.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmGetMinorNameForFD: Add FreeBSD variant

Use the FreeBSD variant function to obtain the minor name and the
device node.
Return the correct path based on where the node is (drm/ versus dri/).

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmGetDeviceNameFromFd: Add FreeBSD variant

Get the major/minor via fstat and after checking that this is a drm node
construct the full device node name using devname.
Note that we should be able to use fdevname to avoid calling fstat + devname
but for some reason it doesn't work on drm node (probably due to how the device
node are created in the linux compat code for drm on FreeBSD).

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmGetMinorType: Add FreeBSD version

Resolve the minor type based on the device node path.
The minor type is either in /dev/drm/X where X is the type or
in a Linux-compatible device node in /dev/dri/
This means we need the major number on FreeBSD so add it to the function
arguments.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

xf86drm: Remove ifdef for FreeBSD and DRM_MAJOR

FreeBSD devfs only provides on the fly generated major/minor.
The major number is irrelevant for FreeBSD so remove the special case.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: drmNodeIsDRM: Add FreeBSD variant

FreeBSD devfs have on the gly generated major minor so we cannot use them
to test if the device is a drm node.
Instead get the devfs node name and test if it is in a subdirectory "drm/"
or "dri/".
Historycally DRM device on FreeBSD are created in /dev/drm/ and link are
present in /dev/dri/ for compatibility reason.

Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

meson: don't detect <sys/sysctl.h> on Linux

The header is not required on Linux, and is in fact deprecated in glibc 2.30+

Reported-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Cc: Niclas Zeising <zeising@daemonic.se>
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Tested-by: Niclas Zeising <zeising@daemonc.se>

tests/amdgpu/vcn: add dummy feedback message

Latest VCN firmware has feedback header check.

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

Bump version to 2.4.101

amdgpu: increase cpu_map_count storage size

Mesa expects to be able to map the same buffer, without unmapping it.
This leads to problem on long-running program.

On the other hand, libdrm uses cpu_map_count as a refcount and expects
its value to decrease so it can unmap buffers.

The previoulsy proprosed fix (https://patchwork.freedesktop.org/patch/258005/)
stopped increased the counter when it went past INT_MAX.

This commit instead proposes to use a larger type to store cpu_map_count.
The outcome is the same: long running apps will not crash, only the
implementation differs.

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1423
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

intel: sync i915_pciids.h with kernel

Changes:
3882581753d1 ("drm/i915/tgl: Add new PCI IDs to TGL")

Signed-off-by: Swathi Dhanavanthri <swathi.dhanavanthri@intel.com>
Reviewed-by: Timo Aaltonen <timo.aaltonen@canonical.com>

Add DRM_MODE_CONNECTOR_WRITEBACK to xf86drmMode.h

xf86drmMode.h currently duplicates the connector-type definitions from
drm_mode.h. Add DRM_MODE_CONNECTOR_WRITEBACK, which is only visible
through a client cap, from drm_mode.h.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

tests: Only copy con->modes[0] if it exists

con->modes can be NULL. Fixes crash in that case.

Closes: https://gitlab.freedesktop.org/mesa/drm/issues/34
Reviewed-by: Daniel Stone <daniels@collabora.com>

xf86drm: fix subsystem type lookup for virtio mmio-based devices

Currently the code assumes that a virtio based device is always located
on the PCI bus.

Modify the parser to make it check the device's parent directory to
determine on which bus it is located.

Output for virtio-pci is the PCI bus.
Output for virtio-mmio is the Platform bus.

Signed-off-by: Vasyl Vavrychuk <vasyl.vavrychuk@opensynergy.com>
Signed-off-by: Mikhail Golubev <Mikhail.Golubev@opensynergy.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Tested-by: Gurchetan Singh <gurchetansingh@chromium.org>

xf86drm: generalize the device subsystem type parsing code

Move the code, which used to get the device subsystem type from a device
path in sysfs, to a separate function to be reusable.

Signed-off-by: Vasyl Vavrychuk <vasyl.vavrychuk@opensynergy.com>
Signed-off-by: Mikhail Golubev <Mikhail.Golubev@opensynergy.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Tested-by: Gurchetan Singh <gurchetansingh@chromium.org>

tests: drop redundant and slow `random` test

Suggested by Emil [1]:
> Feel free to drop the random test altogether. It's an old public API
> no active users (pretty ancient code uses it) and the in-tree users
> drmSL and drmHash already have respective tests.

This test takes minutes to run, while all the other tests combined take
barely more than a second.

Dropping it also helps the CI by avoiding random timeouts when `random`
takes more than the 4 minutes (!) we've allowed for it.

[1] https://gitlab.freedesktop.org/mesa/drm/merge_requests/26#note_390066

Suggested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

amdgpu: clean up the cs structure variable

This patch is to use generic variables as the input of amdgpu_cs_submit_raw2.
Because amdgpu_cs_submit_one won't handle IOCTL directly.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

amdgpu: remove the un-used chunk_array

This array won't be used.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

amdgpu: use amdgpu_cs_submit_raw2 in amdgpu_cs_submit

So far, amdgpu_cs_submit_raw2 is mainly used for upper layer (Mesa), however,
amdgpu_cs_submit is used for current all unit tests. Our intention is that the
unit tests can actually verify the API which is really used.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

amdgpu: use alloca for dependencies and sem_dependencies

Use alloca instead of malloc, then we don't need free them at the end of this
function.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

Add drmModeGetFB2

Add a wrapper around the getfb2 ioctl, which returns extended
framebuffer information mirroring addfb2, including multiple planes and
modifiers.

Changes since v7:
 - add new symbols to core-symbol.txt (Eric Engestrom)

Changes since v5:
 - style change

Changes since v4:
 - Set fb_id at init instead of memclear() and set (Eric Engestrom)

Changes since v3:
 - remove unnecessary null check in drmModeFreeFB2 (Daniel Stone)

Changes since v2:
 - getfb2 ioctl has been merged upstream
 - sync include/drm/drm.h in a seperate patch

Changes since v1:
 - functions should be drm_public
 - modifier should be 64 bits
 - update ioctl number

Signed-off-by: Juston Li <juston.li@intel.com>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

include/drm: sync up drm.h

a) delta: Adds DRM_IOCTL_MODE_GETFB2
  b) Generated using make headers_install
  c) Taken from drm-next-misc:
        commit 3ff4c24bdb1f494c217c80348f9db4896043ed81
        Author: Lyude Paul <lyude@redhat.com>
        Date:   Fri Jan 17 17:47:48 2020 -0500

        drm/dp_mst: Fix indenting in drm_dp_mst_topology_mgr_set_mst()

Signed-off-by: Juston Li <juston.li@intel.com>
Acked-by: Eric Engestrom <eric@engestrom.ch>

amdgpu: add new marketing names from 19.50

Add new marketing names.

Acked-by: Leo Liu <leo.liu@amd.com>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

intel: drm_intel_bo_gem_create_from_* on platforms w/o HW tiling

Platforms without a HW detiler doesn't support the get_tiling IOCTL.
Fix the drm_intel_bo_gem_create_from_* functions assuming the default
no-tiling, no-swizzling setting for the GEM buffer in this case.

v2:
- Add the missing gem handle IOCTL parameter. (Eric)

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>

meson: always define whether headers exist

Combined with -Wundef (added in 75758d2ccf & enforced in ba17673eed),
this provides absolute safety against #ifdef typos.

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

meson: add symbols check for core libdrm

All the libdrm_* submodules have symbols checks, no reason to keep core
libdrm wild.

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

modetest: Fix segmentation fault

When a mode is set with just a connector "-s foo",
we get a nasty segmentation fault. Fix it.

Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

libdrm: modetest: Allow selecting modes by index

Often there are many similar modes, which cannot be selected
via modetest due to its simple string matching.

This change adds a mode index in the display output, which can
then be used to specify a specific modeline to be set.

Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: Rob Clark <robdclark@chromium.org>
Cc: Bjorn Andersson <bjorn.andersson@linaro.org>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: John Stultz <john.stultz@linaro.org>
[emil: rebase]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

tests/amdgpu: Fix buffer overflow (v3)

This patch fixes the following warning:
-Wformat-overflow=

v2: Use the correct strlcat(3).
v3: Use strncat(3) and remove libbsd dependency.

Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

tests/amdgpu: Fix unused function warning (v2)

This patch fixes:
-Wunused-function

v2: Always enable amdgpu_ras_test().

Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

tests/amdgpu: Fix various warnings (v2)

This patch fixes the following warnings:
-Wformat=
-Wmaybe-uninitialized
-Wmisleading-indentation
-Wstringop-truncation
-Wunused-function
-Wunused-variable

It also removes forward declarations and moves
global functions to the bottom, keeping locals
at the top, in ras_tests.c.

v2: Fix compilation.

Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>

Revert "tests/amdgpu: Fix various warnings"

This reverts commit fb1634583f3ba22c67cad0df7022b6ac48a40c56.

Revert "tests/amdgpu: Fix unused function warning (v2)"

This reverts commit 4ff499cd85f86a7b5b28f3449de2fbad1b91c795.