RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-27 00:27:45 +00:00
Code Issues Activity Actions

Kernel: Implement basic VirGL device

Browse source 
This commit flips VirtIOGPU back to using a Mutex for its operation
lock (instead of a spinlock). This is necessary for avoiding a few
system hangs when queuing actions on the driver from multiple
processes, which becomes much more of an issue when using VirGL from
multiple userspace process.

This does result in a few code paths where we inevitably have to grab
a mutex from inside a spinlock, the only way to fix both issues is to
move to issuing asynchronous virtio gpu commands.
This commit is contained in:
Sahan Fernando 2022-02-13 16:45:30 +11:00 • committed by Ali Mohammad Pur
parent 966989afe8
commit fd6a536c60
7 changed files with 503 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

97
Kernel/Graphics/VirtIOGPU/GPU3DDevice.cpp Normal file
View file

@ -0,0 +1,97 @@
/*
* Copyright (c) 2021, Sahan Fernando <sahan.h.fernando@gmail.com>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <Kernel/API/VirGL.h>
#include <Kernel/Graphics/GraphicsManagement.h>
#include <Kernel/Graphics/VirtIOGPU/Console.h>
#include <Kernel/Graphics/VirtIOGPU/GPU3DDevice.h>
#include <Kernel/Graphics/VirtIOGPU/GraphicsAdapter.h>
#include <Kernel/Graphics/VirtIOGPU/Protocol.h>
#include <Kernel/Random.h>
#include <LibC/sys/ioctl_numbers.h>
namespace Kernel::Graphics::VirtIOGPU {
GPU3DDevice::GPU3DDevice(GraphicsAdapter& graphics_adapter)
: CharacterDevice(28, 0)
, m_graphics_adapter(graphics_adapter)
{
m_kernel_context_id = m_graphics_adapter.create_context();
// Setup memory transfer region
auto region_result = MM.allocate_kernel_region(
NUM_TRANSFER_REGION_PAGES * PAGE_SIZE,
"VIRGL3D upload buffer",
Memory::Region::Access::ReadWrite,
AllocationStrategy::AllocateNow);
VERIFY(!region_result.is_error());
m_transfer_buffer_region = region_result.release_value();
}
ErrorOr<void> GPU3DDevice::ioctl(OpenFileDescription&, unsigned request, Userspace<void*> arg)
{
// TODO: We really should have ioctls for destroying resources as well
switch (request) {
case VIRGL_IOCTL_TRANSFER_DATA: {
auto user_transfer_descriptor = static_ptr_cast<VirGLTransferDescriptor const*>(arg);
auto transfer_descriptor = TRY(copy_typed_from_user(user_transfer_descriptor));
if (transfer_descriptor.direction == VIRGL_DATA_DIR_GUEST_TO_HOST) {
if (transfer_descriptor.offset_in_region + transfer_descriptor.num_bytes > NUM_TRANSFER_REGION_PAGES * PAGE_SIZE) {
return EOVERFLOW;
}
auto target = m_transfer_buffer_region->vaddr().offset(transfer_descriptor.offset_in_region).as_ptr();
return copy_from_user(target, transfer_descriptor.data, transfer_descriptor.num_bytes);
} else if (transfer_descriptor.direction == VIRGL_DATA_DIR_HOST_TO_GUEST) {
if (transfer_descriptor.offset_in_region + transfer_descriptor.num_bytes > NUM_TRANSFER_REGION_PAGES * PAGE_SIZE) {
return EOVERFLOW;
}
auto source = m_transfer_buffer_region->vaddr().offset(transfer_descriptor.offset_in_region).as_ptr();
return copy_to_user(transfer_descriptor.data, source, transfer_descriptor.num_bytes);
} else {
return EINVAL;
}
}
case VIRGL_IOCTL_SUBMIT_CMD: {
MutexLocker locker(m_graphics_adapter.operation_lock());
auto user_command_buffer = static_ptr_cast<VirGLCommandBuffer const*>(arg);
auto command_buffer = TRY(copy_typed_from_user(user_command_buffer));
m_graphics_adapter.submit_command_buffer(m_kernel_context_id, [&](Bytes buffer) {
auto num_bytes = command_buffer.num_elems * sizeof(u32);
VERIFY(num_bytes <= buffer.size());
MUST(copy_from_user(buffer.data(), command_buffer.data, num_bytes));
return num_bytes;
});
return {};
}
case VIRGL_IOCTL_CREATE_RESOURCE: {
auto user_spec = static_ptr_cast<VirGL3DResourceSpec const*>(arg);
VirGL3DResourceSpec spec = TRY(copy_typed_from_user(user_spec));
Protocol::Resource3DSpecification const resource_spec = {
.target = static_cast<Protocol::Gallium::PipeTextureTarget>(spec.target),
.format = spec.format,
.bind = spec.bind,
.width = spec.width,
.height = spec.height,
.depth = spec.depth,
.array_size = spec.array_size,
.last_level = spec.last_level,
.nr_samples = spec.nr_samples,
.flags = spec.flags
};
MutexLocker locker(m_graphics_adapter.operation_lock());
auto resource_id = m_graphics_adapter.create_3d_resource(resource_spec).value();
m_graphics_adapter.attach_resource_to_context(resource_id, m_kernel_context_id);
m_graphics_adapter.ensure_backing_storage(resource_id, *m_transfer_buffer_region, 0, NUM_TRANSFER_REGION_PAGES * PAGE_SIZE);
spec.created_resource_id = resource_id;
// FIXME: We should delete the resource we just created if we fail to copy the resource id out
return copy_to_user(static_ptr_cast<VirGL3DResourceSpec*>(arg), &spec);
}
}
return EINVAL;
}
}