Now, only one instance of virgl_screen exists for a device
(/dev/dri/cardX), and it is shared by different frontends (eg GLX,
GBM, etc.). There is a problem with this, as follows:
/* Init GLX */
...
glXCreateContext(...);
...
/* GBM */
gbm_fd = open("/dev/dri/card0", O_RDWR);
dev = gbm_create_device(gbm_fd);
bo = gbm_bo_create(dev, ...);
plane_handle = gbm_bo_get_handle_for_plane(bo, ...);
drmPrimeHandleToFD(gbm_fd, handle.u32, flags, &plane_fd);
The above drmPrimeHandleToFD() call will fail with ENOENT.
The reason is that GBM and GLX share the same virgl_screen (file
descriptor), and it is not gbm_fd that is used to create gbm_bo,
but other fd (opened during GLX initialization). Since the scope
of prime handle is limited to drm_file, the above plane_handle is
invalid under gbm_fd.
By canceling the sharing of virgl_screen between different drm_files,
GBM can use the correct fd to create resources, thereby avoiding the
problem of invalid prime handle.
Signed-off-by: Feng Jiang <jiangfeng@kylinos.cn>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16738>
This allows the software scoreboarding pass, scheduler, and so on
to handle the individual instructions and handle them, rather than
trusting in the generator to do scoreboarding correctly when expanding
the virtual instruction to multiple actual instructions.
By using SHADER_OPCODE_READ_SR_REG, we also correctly handle the
software scoreboarding workaround when reading DMask/VMask.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17530>
The hwconfig api may change unexpectedly prior to public release of
new platforms. Also, public documentation of the hwconfig api
sometimes lags the release.
For these reasons, warnings about unhandled hwconfig keys are noisy,
likely to occur, and unhelpful to most users. This commit drops those
warnings, in favor of a separate internal process for tracking
hwconfig api changes.
Suggested-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17846>
Currently, res->maybe_busy is false by default. If wait immediately
after the resource is created, virgl_drm_resource_wait() will return
directly without checking the actual state of the kernel, which will
cause synchronization problems, such as:
On Guest:
pipe_buffer_create [mesa]
virgl_drm_winsys_resource_create
virtio_gpu_resource_create_ioctl [kernel]
virtio_gpu_fence_alloc
virtio_gpu_object_create
virtio_gpu_cmd_resource_create_3d
VIRTIO_GPU_CMD_RESOURCE_CREATE_3D
virtio_gpu_object_attach
virtio_gpu_cmd_resource_attach_backing
VIRTIO_GPU_CMD_RESOURCE_ATTACH_BACKING
resource_wait [mesa]
virgl_drm_resource_wait /* return directly without fence waiting */
pipe_buffer_map [mesa]
virgl_drm_resource_map
virtio_gpu_map_ioctl [kernel]
os_mmap
memcpy /* <== here */
On Host (with QEMU):
VIRTIO_GPU_CMD_RESOURCE_CREATE_3D
virgl_cmd_create_resource_3d [qemu]
virgl_renderer_resource_create [virglrenderer]
VIRTIO_GPU_CMD_RESOURCE_ATTACH_BACKING
virgl_resource_attach_backing [qemu]
virtio_gpu_create_mapping_iov
virgl_renderer_resource_attach_iov [virglrenderer]
virgl_resource_attach_iov
vrend_pipe_resource_attach_iov
vrend_write_to_iovec /* <== here */
virtio_gpu_cleanup_mapping_iov [qemu]
In the example above, there is a race condition between memcpy and
vrend_write_to_iovec.
Signed-off-by: Jiang Feng <jiangfeng@kylinos.cn>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17592>
When we don't want to communicate with minio, e.g. running
lava_job_submitter script locally, MINIO_RESULTS_UPLOAD should be unset.
But this variable is already set by generate-env script, so we need to
remove it from the /set-job-env-vars.sh to avoid declaring it in
unexpected scenarios.
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17645>
Is a phi source is an undef, there's no point in copying it or really
caring about it at all. We would just end up inserting a mov from an
undef to a register. Instead, treat phi sources which point to an undef
as if the phi source doesn't exist.
This also prevents them from being included in phi webs which should
reduce the overall interference seen in the shader. Currently, if two
phis share an undef, their phi webs are consdiered to interfere. By
ignoring undefs we can get rid of this false interference and reduce the
size of phi webs. Reducing the number of things being copied by the
parallel copy instructions should also free up the paralle copy
algorithm and reduce the over-all churn of movs.
Shader-db results on Haswell:
total instructions in shared programs: 8156608 -> 8155406 (-0.01%)
instructions in affected programs: 164838 -> 163636 (-0.73%)
Shader-db results on Skylake:
total instructions in shared programs: 18227370 -> 18227359 (<.01%)
instructions in affected programs: 519 -> 508 (-2.12%)
helped: 6
HURT: 0
Shader-db results on Tigerlake:
total instructions in shared programs: 21167987 -> 21168025 (<.01%)
instructions in affected programs: 23701 -> 23739 (0.16%)
helped: 21
HURT: 27
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16817>
