Add queue id support to the user queue wait IOCTL
drm_amdgpu_userq_wait structure.
This is required to retrieve the wait user queue and maintain
the fence driver references in it so that the user queue in
the same context releases their reference to the fence drivers
at some point before queue destruction.
Otherwise, we would gather those references until we
don't have any more space left and crash.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This patch extends the user queue creation API to support:
1. Queue priority levels (normal low, low, normal high, high)
- High priority is restricted to admin-only
2. Secure queue flag for protected content access
The changes include:
- Adding a `flags` parameter to `amdgpu_create_userqueue()`
- Defining priority and security flags in `amdgpu_drm.h`
- Updating the `drm_amdgpu_userq_in` struct to replace padding with flags
- Documenting the new flags field
Related driver patches provided by Alex:
https://lists.freedesktop.org/archives/amd-gfx/2025-April/122782.htmlhttps://lists.freedesktop.org/archives/amd-gfx/2025-April/122780.htmlhttps://lists.freedesktop.org/archives/amd-gfx/2025-April/122786.html
Cc: Koenig, Christian <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse.Zhang <Jesse.zhang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This patch adds new IOCTL functions to support
userqueue create, remove, signal and wait etc.
v2:(Marek)
- Add csa support for SDMA queue.
- Addressed's review comments.
- Removed raw2/op2 ioctl.
- Added syncobj_timeline_handles in amdgpu_userq_wait IOCTL.
v3:(Yogesh)
- Rename timeline* objects as per UAPI review (Arvind).
v4: (Marek)
- Drop AMDGPU_USERQ_BO_WRITE as this should not be a global option
of the IOCTL, It should be option per buffer. Hence adding separate
array for read and write BO handles. (Arun)
- Modify num_fences to __u16, flags changed to __u16 and placed
the num_fences next to flags for optimal padding and size. (Arun)
v5:(Marek/Pierre-Eric)
- add more detail params description for signal and wait IOCTL calls.
- Remove the unused structure fields in signal and wait structs.
- Add separate array of read and write for BO handles. (Arun)
- Removes the unused flags parameter from the
amdgpu_create_userqueue IOCTL. (Arvind)
v6:(Pierre-Eric)
- Remove unused headers. (Arvind)
- Modify the function parameter names and struct
field names as per the review comments. (Arun)
v7:(Marek)
- Modify the structure field name and comments. (Arun)
- Rename vm_timeline_syncobj and add comment for
vm_timeline_point.
- Remove GDS buffer support from MQD. (Arvind)
v8:(Pierre-Eric)
- Modify the function parameter names.
- Added new function in amdgpu-symbols.txt (Arvind)
v9:(Marek)
- Use the drm signal/wait structure as the parameter. (Arun)
Cc: Deucher, Alexander <alexander.deucher@amd.com>
Cc: Koenig, Christian <christian.koenig@amd.com>
Cc: Sharma, Shashank <shashank.sharma@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
This patch adds UAPI interface changes for AMDGPU usermode
queues, semaphore and new AMDGPU GEM domain for doorbells.
Usermode queues allow a userspace process to create
and submit its graphics/compute/sdma work directly to the GPU.
v2:(Marek)
- Add csa support for SDMA queue.
- Rename UAPI objects and struct as per UAPI review. (Shashank)
v3:(Yogesh)
- Rename UAPI timeline* objects as per UAPI review. (Arvind)
v4: (Marek)
- Drop AMDGPU_USERQ_BO_WRITE as this should not be a global option
of the IOCTL, It should be option per buffer. Hence adding separate
array for read and write BO handles. (Arun)
- Modify num_fences to __u16, flags changed to __u16 and placed
the num_fences next to flags for optimal padding and size. (Arun)
- Fix 32-bit holes issue in sturct drm_amdgpu_gem_va as per
UAPI review (Arvind).
v5:(Marek/Pierre-Eric)
- add more detail params description for signal and wait IOCTL calls.
- Remove the unused structure fields in signal and wait structs.
- Add separate array of read and write for BO handles. (Arun)
- Removes the unused flags parameter from the usermode queue UAPI structure
- Clarify comments on top of drm_amdgpu_userq_in
- Clarify comment for queue_id (in)
- Clarify comment for mqd
- Clarify comment for compute MQD size
- Clarify comment for queue_id (out)
- Adds padding variables in userqueue in/out structures. (Shashank)
v6:(Pierre-Eric)
- Modify the function parameter names and struct
field names as per the review comments. (Arun)
v7:(Marek)
- Modify the structure field name and comments. (Arun)
- Rename vm_timeline_syncobj and add comment for
vm_timeline_point.
- Remove GDS buffer support from MQD. (Arvind)
v8:(Pierre-Eric)
- Modify the function parameter names.
Cc: Koenig, Christian <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
This patch:
- adds a new subquery (AMDGPU_INFO_UQ_FW_AREAS) in AMDGPU_INFO_IOCTL
to get the size and alignment of shadow and csa objects from the
kernel. This information is required for a userqueue consumer (like
MESA/libdrm) to create the userqueue metadata objects properly.
- also adds supporting metadata structures and a high level wrapper
function (amdgpu_query_uq_metadata_info) to the query, to make it
easy to use.
The corresponding kernel changes for this UAPI extension can be found
in amd-gfx mailing list, link:
https://patchwork.freedesktop.org/patch/621390/?series=139715&rev=2
This patch adds support only for the GFX IP, and the other engines may
be supported in subsequent development.
V2: fix the build error due to exporting of helper function
V3: make an entry for amdgpu_query_uq_metadata_info in
amdgpu-symbols.txt
V4: Rename the subquery to AMDGPU_INFO_UQ_FW_AREAS (Marek, Pierre-Eric)
V5: Addressed review comments (Pierre-Eric):
- Fix the API comment to match the new IOCTL name
- remove the unused uq_metadata parameter
Cc: Marek Olsak <marek.olsak@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Arvind Yadav <arvind.yadav@amd.com>
Reviewed-by: Marek Olsak <marek.olsak@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Just as it is already handled in the caller, `drmGetVersion`.
I'm not sure what the offending driver is, but the Ubuntu incidents
seem to be coming from a dual Intel/Nvidia machine. And they show
it is `card1` so I'm guessing `nvidia-drm` is the offender.
Related: https://bugs.launchpad.net/bugs/2104352
Make modetest available to vendors on Android. libdrm_util and
libdrm_test_headers is also made available to vendors since these are
depenencies of modetest. This results in the module target
modetest.vendor being availble to vendor modules.
Signed-off-by: Rob Barnes <robbarnes at google.com>
Fixes the following building error:
external/libdrm/xf86drm.c:158:10: fatal error: 'generated_static_table_fourcc.h' file not found
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.
Change-Id: I1b0cac498ed63ebec6e8c03629bbf4a1b6a9618d
Reviewed-by: Mauro Rossi <issor.oruam@gmail.com>
Currently modetest supports only the YUV420 and YVU420 planar YCbCr
plane formats (aka YV12 and YU12). Extend the code to add support for
YUV422 / YVU422 and YUV444 / YVU444 plane formats.
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
In preparation to adding more planar YUV formats, introduce X, Y
subsampling ratios and use them to calculate plane offsets and buffer
size.
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Whenever `util_open()` is called to open a device for the first matching
module, it will skip devices for the `nvidia_drm` kernel module which
is not in the list. We could add this module for now, but keeping this
list of DRM modules up to date is cumbersome.
At the same time walking a list of modules and calling `drmOpen()` for
each of them is incredibly expensive (when the user doesn't explicitly
specify one with `-M`), as each each call opens every DRM node just
to see if they are associated to the requested module. And for no
good reason: all we want is the first `DRM_NODE_PRIMARY` (which is what
`drmOpen()` also returns) to use by default.
For example on the `"msm"` driver, which used to be the 9th in the
modules list, all nodes are opened for the 9th time before e.g.
`modetest` returns a useful result, which takes ages unless the user
painstakingly provides the module for the currently known device on
the cmdline.
This is very simply solved by calling `drmGetDevices(2)()`, which
iterates through all DRM nodes only once and allows us to immediately
find + `open()` the first device that has a PRIMARY node. A random
search for the error shows that this was also attempted in (a fork of?)
kmscube:
https://git.ti.com/cgit/glsdk/kmscube/commit/?id=456cabc661caac5c60729751d45efd668faa8e97
Finally we add a `drmIsKMS()` check to make sure we only include primary
nodes that actually support rendering, and also print the values from
`drmGetVersion()` on success to make it easier to identify the device.
In the future we could extrapolate this feature by letting query
commands like `modetest -c` list connectors for every device/module, not
just the first PRIMARY node that we found.
Bionic libc ships with `ioctl` that has two signatures, one with an
unsigned `request` parameter and one with a signed request parameter.
This leads to compilation failing due to `__typeof__(ioctl)` being used
by DRM which fails to resolve which overload to use, this has been fixed
by defining `BIONIC_IOCTL_NO_SIGNEDNESS_OVERLOAD` on Android.
Signed-off-by: Mark Collins <mark@igalia.com>
There is an excellent writeup explaining this requirement here:
https://gitlab.freedesktop.org/wayland/wayland/-/issues/175
In short, for mixed environments such as the Steam Runtime and other
container-like environments, choosing which libdrm to link into the
client's address space is a hard problem. If the runtime has a newer
libdrm than the host, then it should be preferred, because the client
may be using newly-added symbols. But if the host has a newer libdrm,
then that should be used, because drivers may be depending on those.
Bumping the DSO minor version is transparent to all users because apps
only link against the major version, e.g. DT_NEEDED libdrm.so.2; the
fact that libdrm.so.2 is a link to libdrm.so.2.122.0 is a detail known
only to the loader, but it does let a smart runtime make better
decisions.
Signed-off-by: Daniel Stone <daniels@collabora.com>
When the GPU feature is turned on in crosvm, these modules are added as
dependencies. Since crosvm is included in the virt APEX, add the APEX to
the apex_available properties of the modules to make them available in
the APEX.
To do so, crosvm and its dependencies have the apex_available property
set to "//apex_available:platform", "com.android.virt" to explicitly
acknowledge the joining.
drm_property_type_is function in xf86drmMode.h file can cause compiler
error because it performs unsigned to signed conversion. Some Android.mk
modules have been avoiding this by adding "-isystem external/libdrm"
flag, because warnings from system headers are suppressed.
This changes exported_include_dirs to export_system_include_dirs
to workaround the potential error with the same manner above.
... to make drm format header visible for host wayland server.
Adds -Wno-implicit-function-declaration for vasprintf() on
xf86drm.c:2965.
Adds -Wno-int-conversion for drm_mmap() on libdrm_macros.h:60.
Also, drive-by alphabetize cflags.
See build/soong/README.md for more information about Soong.
Removes BOARD_GPU_DRIVERS, which wasn't affecting anything, since none
of the HAVE_* macros are defined. Even if they were, we'd prefer to
compile all of them so that a single library can support multiple
boards.
Some platforms (eg. SunOS) explicitly need extra symbols in order to define
those functions. There're many files needing the __EXTENSIONS__ symbol,
so doing this on a global scale.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Synchronize drm_fourcc.h and drm_mode.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit 541b1b0a8fc235bca355921eb7f3f59a8efa3e9a
Signed-off-by: Simon Ser <contact@emersion.fr>
The <linux/fb.h> header only exists on Linux, but isn't used anymore,
so drop it entirely.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Several drivers still including <linux/stddef.h>, but not using anything
from it, thus breaking build on non-Linux platforms (eg. FreeBSD).
Since not needed at all, just stop including it.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
../xf86drm.c:4622:9: error: use of undeclared identifier 'written'; did you mean 'write'?
if (written + 1 > max_node_length)
^~~~~~~
write
../xf86drm.c:4624:21: error: use of undeclared identifier 'sbuf'; did you mean 'sbrk'?
if (stat(node, &sbuf))
^~~~
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
fix commit cc3c80c6("tests/amdgpu: refactor dispatch/draw test")
../../SOURCES/drm/tests/amdgpu/shader_code.h:113:2: error: initializer
element is not constant
ps_##_ps##_shader_patchinfo_code_size_gfx##_n, \
^
Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
