GitLab

We can pretty easily handle merging the driver's driconf with the common
driverconf right there, rather than pushing that to each driver.

Part-of: <!6751>

This cuts even more code duplication.

Part-of: <!6751>

This gives me one place to document why it works this way.  This does make
the debug message a little less helpful ("etna" instead of "etnaviv" and
"vmwgfx" instead of "svga", but you should only be able to reach this when
doing something like trying the radeon/nouveau vdpau target on the wrong
DRM device for example.

Part-of: <!6751>

We can just reuse drm_helper.h, which has either the real code or the stub
for all pipe_screens based on the GALLIUM_* driver defines, and the
dynamic pipe loader's .c build will only define one GALLIUM_* driver
define.  The remaining stubs should get GCed by the linker.

Part-of: <!6751>

In 283ad859, radeonsi started using nir_find_var_copies.  However,
it was also calling nir_lower_var_copies in the optimization loop and
the two can end up fighting.  The simple solution is to wait to lower
copies until the end of the optimization loop.

Fixes: 283ad859
Closes: #3550Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6841>

the previous version of this leaked a reference to the streamout buffer here

thanks to deltragon on my blog for pointing this out!

Fixes: 37778fcd ("zink: implement transform feedback support to finish off opengl 3.0")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6457>

without VK_EXT_extended_dynamic_state, we need to pass this directly or
else only the first viewport will work
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6349>

each viewport must have a corresponding scissor, so we need to generate
an array of these to ensure our states match up
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6349>

this cap varies based on the shader stage, but it must be enabled any time this
output is written
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6349>

this is part of the pipeline state (there's an extension for setting it
dynamically but we don't supprot that atm
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6349>

this is just a simple case of connecting up the vertex state to the pipeline
state
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6270>

this is required by spec
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6263>

this is required by spec
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6263>

pipe_framebuffer_state::layers doesn't necessarily reflect the value we
need to use here, so we can use the util function to verify
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6263>

state->samples doesn't necessarily reflect the correct value here, so we need
to use the util function
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6263>

nr_samples is valid on the texture pipe_resource, not the surface
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <!6263>

This reverts commit 939ddf3f.

Intel has a separate pass for fusing FFMAs selectively.  We split
these flags in commit 1b72c31e and
the reasoning still stands.  The patch being reverted was just a
cleanup, so there should be no issue with reverting it.
Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6849>

Note: this also fixes vk_object_base_init() not being called for the wsi
fences, now that it goes through radv_CreateFence().
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <!6846>

Looks like this path is meant for RADV_FENCE_WINSYS type fences.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <!6846>

This allows to inline radeon_uses_secure_bos calls and reduce CPU overhead.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Subtitles are rendering with an upload through a staging texture.
So the sequence is:
  1. draw video (with a secure cs)
  2. copy staging texture to the real texture (via si_resource_copy_region) in
    a non-secure cs.
  3. draw video (with a secure cs)

Step 2 and 3 both generates a flush with RADEON_FLUSH_TOGGLE_SECURE_SUBMISSION.
These flushes are executed quite late: right before doing the draw/dispatch,
so maybe the issue here is the handling of dependencies.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

tess_rings must be encrypted when used in a secure job so this commit
introduces a tess_rings_tmz resource.

The cs_preamble_state doesn't contain the tess_rings address anymore since
it can change. The tess_rings related registers go in a separate preamble.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Instead of exposing a cs_set_secure() callback that always needs a call
to si_flush_gfx_cs before a switch, this commit introduces a new
flag to switch between secure and non-secure on submissions.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

This commit makes TMZ always allowed instead of being either off or forced-on
with AMD_DEBUG=tmz.

With this change:
- secure job can be used as soon as the application made a tmz allocation. Driver
  internal allocations are not enough to enable secure jobs (if tmz is supported
  and enabled by the kernel)
- AMD_DEBUG=tmz forces all scanout/depth/stencil buffers to be allocated as TMZ.
  This is useful to test app thats don't explicitely support protected content.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Tag allocations as driver internal.
Some of these allocations will need to be doubled to handle TMZ (one secure bo,
one normal bo) but these allocations shouldn't switch the winsys in "the app
is using TMZ".
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

This commit adds asserts verifying the following conditions when
using a secure job:
- fb textures are encrypted (both colors and depth/stencil buffers)
- all writeable bo are encrypted
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Fixes a hang with DCC.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6049>

Fences are now just a syncobj, which makes our life easier.

The next step will be to fill out ImportFenceFdKHR()/GetFenceFdKHR().
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <!6683>

Note: this means turnip requires kernel 5.8 (or older with syncobj patch).
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <!6683>

Avoid the extra indirect for this commonly used field.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <!6683>

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <!6683>

Without this, we end up throwing errors on code along these lines when
rendering using single-buffering:

GLint att;
glGetIntegerv(GL_READ_BUFFER, &att);
glGetFramebufferAttachmentParameteriv(GL_READ_FRAMEBUFFER, att, ...);

This is because we internally translate GL_BACK (which is what
glGetIntegerv returned) to GL_FRONT, which we don't handle in the
Desktop GL case. So let's start handling it.

This fixes the GLTF-GL33.gtf21.GL2FixedTests.buffer_color.blend_color
test for me.

Fixes: e6ca6e58 ("mesa: Handle pbuffers in desktop GL framebuffer attachment queries")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <!6815>