git.libre-soc.org Git - mesa.git/log

mesa: Fix pack_uint_Z_FLOAT32()

Fixed pack_uint_Z_FLOAT32 by casting row data to float instead uint.
Remove code duplicate function pack_uint_Z_FLOAT32_X24S8.
Edited case in "_mesa_get_pack_uint_z_func".
Now it looks like "_mesa_get_pack_float_z_func".
Remove _mesa_problem call, which was added for debuging this issue.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91433
Signed-off-by: Illia Iorin <illia.iorin@globallogic.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>

intel: Introducing Whiskey Lake platform

Whiskey Lake uses the same gen graphics as Coffe Lake, including some
ids that were previously marked as reserved on Coffe Lake, but that
now are moved to WHL page.

This follows the ids and approach used on kernel's commit
b9be78531d27 ("drm/i915/whl: Introducing Whiskey Lake platform")
and commit c1c8f6fa731b ("drm/i915: Redefine some Whiskey Lake SKUs")

v2: Lionel noticed that GT{1,2,3} on kernel wasn't following
spec when looking to number of EUs, so kernel has been updated.

Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Anuj Phogat <anuj.phogat@gmail.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

st/va: use provided sizes and coords for vlVaGetImage

vlVaGetImage should respect the width, height, and coordinates x and y that
passed in. Therefore, pipe_box should be created with the passed in values
instead of surface width/height.

v2: add input size check, return error when size out of bounds
v3: fix the size check for vaimage
v4: add size adjustment for x and y coordinates

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Christian König <christian.koenig@amd.com>

radv: implement clear operations for R32G32B32

This fixes crashes for some CTS:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.*.linear_*_*
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.*.*_linear_*

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108113
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: disallow 3D images and mipmaps/layers for R32G32B32 linear formats

R32G32B32 are weird formats and we are only going to support
some basic operations for now.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add a workaround for a VGT hang with prim restart and strips

Otherwise, Yakuza and The Evil Within hang the GPU with DXVK.
This apparently only works on Polaris.

Suggested by Marek.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

glsl: remove redundant es_shader checks

The es check is already covered by the is_version() check.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

st/glsl_to_tgsi: initialise need_uarl in contructor

Found by coverity

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

glspirv: drop pointless assert (size_t is unsigned)

Found by coverity

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

radv: remove unsigned comparison against 0

The value is always >= 0 here.

Found by coverity

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

radv: remove dead code for master_fd close

We have never opened master_Fd at this point, so remove code to
close it.

Found by coverity.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

radv: don't pass shader key by copy

Coverity pointed out we were copying 168 bytes here unnecessarily.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

anv: add missing unlock in error path.

Not going to matter, but be consistent.

Found by coverity

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Fixes: caf41c78c (anv/allocator: Support softpin in the BO cache)

intel: Don't propagate conditional modifiers if a UD source is negated

This fixes a bug uncovered by my NIR integer division by constant
optimization series.

Fixes: 19f9cb72c8b "i965/fs: Add pass to propagate conditional..."
Fixes: 627f94b72e0 "i965/vec4: adding vec4_cmod_propagation..."
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

util: Add tests for fast integer division by constants

While I generally trust rediculousfish to have done his homework, we've
made some adjustments to suit the needs of mesa and it'd be good to
test those. Also, there's no better place than unit tests to clearly
document the different edge cases of the different methods.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

util: Add power-of-two divisor support to compute_fast_udiv_info

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

util: Generalize fast integer division to be variable bit-width

There's nothing inherently fixed-width in the code.  All that's required
to generalize it is to make everything internally 64-bit and pass
UINT_BITS in as a parameter to util_compute_fast_[us]div_info.  With
that, it can now handle 8, 16, 32, and 64-bit integer division by a
constant.

We also add support for division by 1 and by other powers of 2.  This is
useful if you want to divide by a uniform value in a shader where you
have the opportunity to adjust the uniform on the CPU before passing it
in.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

util: Add fast division helpers

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

util: import public domain code for integer division by a constant

Compilers can use this to generate optimal code for integer division
by a constant.

Additionally, an unsigned division by a uniform that is constant but not
known at compile time can still be optimized by passing 2-4 division
factors to the shader as uniforms and executing one of the fast_udiv*
variants. The signed division algorithm doesn't have this capability.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

util: Add a simple big math library

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

meson: Don't allow building EGL on Windows or MacOS

Currently mesa only supports EGL on Unix like systems, cygwin, and
haiku. Meson should actually enforce this. This fixes the default build
on MacOS.

v2: - invert the condition, mark darwin and windows as not supported
instead of trying to mark what is supported.
v3: - add missing )
v3: - Update comment to reflect condition change in v2

CC: 18.2 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

glsl: ignore trailing whitespace when define redefined

The Nvidia/AMD binary drivers allow this, as does GCC.

This fixes shader compilation issues in the latest update of
No Mans Sky.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

intel/compiler: Don't handle fsign.sat

No shader-db or CI changes on any Intel platform.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir/algebraic: Simplify fsat of fsign

These allows us to not support fsign.sat in the Intel compiler backend,
and that will simplify some later changes.

No shader-db changes on any Intel platform.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir/algebraic: sign(x)*x*x is abs(x)*x

shader-db results:

All Gen7+ platforms had similar results. (Skylake shown)
total instructions in shared programs: 15106023 -> 15105981 (<.01%)
instructions in affected programs: 300 -> 258 (-14.00%)
helped: 6
HURT: 0
helped stats (abs) min: 7 max: 7 x̄: 7.00 x̃: 7
helped stats (rel) min: 14.00% max: 14.00% x̄: 14.00% x̃: 14.00%
95% mean confidence interval for instructions value: -7.00 -7.00
95% mean confidence interval for instructions %-change: -14.00% -14.00%
Instructions are helped.

total cycles in shared programs: 566050327 -> 566050075 (<.01%)
cycles in affected programs: 2826 -> 2574 (-8.92%)
helped: 6
HURT: 0
helped stats (abs) min: 40 max: 44 x̄: 42.00 x̃: 42
helped stats (rel) min: 8.89% max: 8.94% x̄: 8.92% x̃: 8.92%
95% mean confidence interval for cycles value: -44.30 -39.70
95% mean confidence interval for cycles %-change: -8.95% -8.88%
Cycles are helped.

No changes on Gen6 or earlier.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

nir: Add helper functions to get the instruction that generated a nir_src

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>

svga: change svga_destroy_shader_variant() to return void

svga_destroy_shader_variant() itself flushes and retries the command
if there's a failure. So no need for the callers to do it. Other
callers of the function were already ignoring the return value.

This also fixes a corner-case double-free reported by Coverity
(and reported by Dave Airlie).

Tested with various OpenGL apps.

Reviewed-by: Charmaine Lee <charmainel@vmware.com>

meson: Don't build glsl compiler tests unless OpenGL is enabled

Since there are no other users of the glsl compiler.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

meson: Only build gallium state tracker tests with shared_glapi

This has always been a requirement, it's just somehow been missed in the
meson build.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

meson: only build clapi tests when OpenGL is being built

Otherwise building just vulkan (among other things) will build these
tests, pull in a bunch of stuff they shouldn't, and potentially fail to
compile.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

nvc0: fix blitting red to srgb8_alpha

For some reason the 2d engine can't handle this. Red formats get special
treatment there, so perhaps related.

Fixes dEQP-GLES3 tests of the form:

dEQP-GLES3.functional.fbo.blit.conversion.r{8,16f,32f}_to_srgb8_alpha8

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Cc: mesa-stable@lists.freedesktop.org

nv50,nvc0: guard against zero-size blits

The current state tracker can generate these sometimes. Fixing this is
more involved, and due to some integer math we can generate
divisions-by-zero.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Cc: mesa-stable@lists.freedesktop.org

nv50,nvc0: mark RGBX_UINT formats as renderable

This helps st/mesa avoid some (apparently) buggy fallbacks. Specifically
the CopyTexSubImage fallback tries to read texture A as RGBA_FLOAT and
write back that data into the target format, which fails for integer
formats which have no appropriate logic to do the conversion.

Since integer formats don't blend, there's no harm in the fact that the
"A" component gets written anyways.

Fixes, among others:
https://www.khronos.org/registry/webgl/sdk/tests/conformance2/textures/canvas/tex-2d-rgb8ui-rgb_integer-unsigned_byte.html

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: mesa-stable@lists.freedesktop.org

radv: add missing meson c++ visibility arguments

Fixes: 6f3aee40f90d725653b6 "radv: using tls to store llvm related info
and speed up compiles (v10)"
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

gbm: Add GBM_FORMAT_ARGB1555 support

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/dri: Handle BGRA5551 format

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

freedreno/a5xx+a6xx: fix LRZ pitch alignment

Both RB_2D_DST_SIZE.PITCH (a6xx) and RB_MRT[n].PITCH (a5xx) need
alignment to 64.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: add LRZ support

As with a5xx, hidden behind FD_MESA_DEBUG=lrz due to being paranoid
about z-fighting issues with some games (in particular, this was
observed with 0ad on a5xx.. but I think the proper solution to enable
this by default is to figure out how to do driver specific driconf
options).

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: update generated headers

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: add helper for various CP_EVENT_WRITE

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: remove unused fxns

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/a6xx: remove fd6_shader_stateobj

Earlier gen's already got this cleanup, but a6xx was still off on a
branch then.

Signed-off-by: Rob Clark <robdclark@gmail.com>

glsl: fix array assignments of a swizzled vector

This happens in situations where we might do

  vec.wzyx[i] = ...

The swizzle would get effectively ignored because of the interaction
between how ir_assignment->set_lhs works and overwriting the write_mask.
There are two cases, one where i is a constant, and another where i is
variable. We have to be extra-careful in both cases.

Fixes the following WebGL test:

  https://www.khronos.org/registry/webgl/sdk/tests/conformance2/glsl3/vector-dynamic-indexing-swizzled-lvalue.html

And the new piglit tests:

  swizzled-writemask-indexing-nonconst.shader_test
  swizzled-writemask-indexing.shader_test

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: mesa-stable@lists.freedesktop.org

radv: tidy up radv_pipeline_init_multisample_state()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: always set PA_SC_MODE_CNTL_1.OUT_OF_ORDER_WATER_MARK

It has probably no effect without out of order rasterization
anyway.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: set DB_EQAA.INCOHERENT_EQAA_READS

My attempt was to set this field instead of duplicating one.

Fixes: 6cfa321c39 ("radv: add potential missing fields for DB_EQAA")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

i965: fallback RGBX to RGBA in glEGLImageTargetRenderbufferStorageOES

In the same fashion as is done for glEGLImageTextureTarget2D.

v2: share the fallback which sets baseformat and internalformat correctly
    which makes both of the tests pass (Tapani)

Fixes android.hardware.nativehardware.cts.AHardwareBufferNativeTests:

   #SingleLayer_ColorTest_GpuColorOutputCpuRead_R8G8B8X8_UNORM
   #SingleLayer_ColorTest_GpuColorOutputIsRenderable_R8G8B8X8_UNORM

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

glsl: do not attempt assignment if operand type not parsed correctly

v2: check types of both operands (Ian)

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108012

util/u_queue: add UTIL_QUEUE_INIT_SET_FULL_THREAD_AFFINITY

Initial version discussed with Rob Clark under a different patch name.
This approach leaves his driver unaffected.

radeonsi: fix a typo at CS_PARTIAL_FLUSH

harmless

ac: add ac_build_round

ac: correct PKT3_COPY_DATA definitions

ac: simplify LLVM alloca helpers

ac: define all address spaces properly

gallivm: Make it possible to disable some optimization shortcuts in release builds

For testing it is of interest that all tests of dEQP pass, e.g. to test
virglrenderer on a host only providing software rendering like in a CI.
Hence make it possible to disable certain optimizations that make tests fail.

While we are there also add some documentation to the flags to make it clear
that this is opt-out.

Setting the environment variable "GALLIVM_PERF=no_filter_hacks" can be used to make
the following tests pass in release mode:

  dEQP-GLES2.functional.texture.mipmap.2d.affine.*_linear_*
  dEQP-GLES2.functional.texture.mipmap.cube.generate.*
  dEQP-GLES2.functional.texture.vertex.2d.filtering.*_mipmap_linear_*
  dEQP-GLES2.functional.texture.vertex.2d.wrap.*

Related:
  https://bugs.freedesktop.org/show_bug.cgi?id=94957

v2: rename optimization disabling flag to 'safemath' and also move the
    nopt flag to the perf flags.

v3: rename flag "safemath" to "no_filter_hacks" since safemath is usually
    associated with floating point operations (Roland)

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

virgl: Pass resource size and transfer offsets

Pass the size of a resource when creating it so a backing can be kept in
the other side.

Also pass the required offset to transfer commands.

This moves vtest closer to how virtio-gpu works, making it more useful
for testing.

v2: - Use new messages for creation and transfers, as changing the
      behavior of the existing messages would be messy given that we don't
      want to break compatibility with older servers.

v3: - Use correct strides: The resource corresponding to the output display
      might have a differnt line stride then the IOVs, so when reading back
      to this resource take the resource stride and the the IOV stride
      into account.

v4: Fix transfer size calculation (Andrey Simiklit)

v5: Add comment about transfer size value in the PUT commend (Gurchetan).
    Add a comment about the size correction for transfers for reading and
    writing the resource. Fixing this by correctly evaluating the size
    upfront will need some work also  on the virglrenderer side.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> (v2)
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

virgl, vtest: Correct the transfer size calculation

The transfer size used in virglrenderer refers to uint32_t, so one
must add 3 and then divide by 4 instead of adding 3/4 which is a no-op
with integers.

Fixes: b3b82fe8ea virgl/vtest: add vtest driver
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

util: Make xmlconfig.c build on Solaris without d_type in dirent (v2)

v2: check for lstat() failing

Fixes: 04bdbbcab3c "xmlconfig: read more config files from drirc.d/"
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Roland Mainz <roland.mainz@nrubsig.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

radeonsi:optimizing SET_CONTEXT_REG for shaders vgt_vertex_reuse

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

radeonsi:optimizing SET_CONTEXT_REG for shaders Tessellation

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

radeonsi:optimizing SET_CONTEXT_REG for shaders PS

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

radeonsi:optimizing SET_CONTEXT_REG for shaders VS

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

radeonsi:optimizing SET_CONTEXT_REG for shaders GS

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

radeonsi: optimize and allow reg > 31 in radeon_opt_set_context_reg functions

reg_saved will have 64 bits, and (1 << reg) where reg > 31 has undefined
behavior. (1ull << reg) would be correct for 64 bits.

This commit shifts the other way in order to merge the conditions.

radeonsi: optimizing SET_CONTEXT_REG for shaders ES

Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

spirv: mark variables decorated with XfbBuffer as always active

Otherwise, they are removed during NIR linking or in some
lowering passes.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

docs: update calendar, add news and link release notes to 18.2.2

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>

docs: add sha256 checksums for 18.2.2

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit cb63a4e1144d9cd8feda3799c68a32a769417b5f)

docs: add release notes for 18.2.2

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
(cherry picked from commit abaeb79eb2c16d7abad06719f24d1e59ad775aa6)

nir/alu_to_scalar: Use ssa_for_alu_src in hand-rolled expansions

The ssa_for_alu_src helper will correctly handle swizzles and other
source modifiers for you. The expansions for unpack_half_2x16,
pack_uvec2_to_uint, and pack_uvec4_to_uint were all broken with regards
to swizzles. The brokenness of unpack_half_2x16 was causing rendering
errors in Rise of the Tomb Raider on Intel ever since c11833ab24dcba26
which added an extra copy propagation to the optimization pipeline and
caused us to start seeing swizzles where we hadn't seen any before.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107926
Fixes: 9ce901058f3d "nir: Add lowering of nir_op_unpack_half_2x16."
Fixes: 9b8786eba955 "nir: Add lowering support for packing opcodes."
Tested-by: Alex Smith <asmith@feralinteractive.com>
Tested-by: Józef Kucia <joseph.kucia@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glsl/linker: Check the subroutine associated functions names

>From Section 6.1.2 (Subroutines) of the GLSL 4.00 specification

    "A program will fail to compile or link if any shader
     or stage contains two or more functions with the same
     name if the name is associated with a subroutine type."

v2:
  - error out earlier (Tapani)
  - style fixes (Iago)

Fixes:
    * no-overloads.vert

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108109
Signed-off-by: Vadym Shovkoplias <vadym.shovkoplias@globallogic.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

virgl: Negotiate version with vtest server

Check if server supports version negotation by sending a PING_PROTOCOL_VERSION
message right before a dummy RESOURCE_BUSY_WAIT. If we don't get a reply
for the first, we know the server doesn't support it.

If it does support it, we can query the max protocol version supported
by the server and fall back if needed.

v2: - Send a new message to negotiate the protocol version, checking if
      the server supports this message by immediately sending a busy wait
      message. (Dave Airlie)

v3: - Send a zero-arg command PING_PROTOCOL_VERSION so we actually keep
      compatibility with older servers. (Code by Dave Airlie)

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

intel: aubinator: Fix memory leaks

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/decoder: construct correct xml filename

construct correct gen xml filename when we try to load hardware xml
description from a given path

v2: remove temporary variable (Francesco Ansanelli)

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/decoder: Avoid freeing invalid pointer

v2: Free ctx.spec if error while reading genxml (Lionel Landwerlin)

v3: Handle case where genxml is empty (Lionel Landwerlin)

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/decoder: add gen_spec_init method

Initialize gen_spec instance properly when loading hardware xml
description from specifc directory to avoid segmentation fault.

v2: correct function definition (Lionel Landwerlin)

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

radv: fix resetting the pool for timestamp queries

Since the driver no longer uses the availability bit for
timestamp queries it shouldn't reset it. Instead, it should
reset the query values to UINT32_MAX. This fixes VM faults.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108164
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Józef Kucia <joseph.kucia@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

etnaviv: Use write combine instead of unached mappings for shader bo

The later are sensitive to unaligned accesses on arm64[1] and we don't
need an uncached mapping here.

[1]: https://lists.freedesktop.org/archives/etnaviv/2018-September/001956.html

Signed-off-by: Guido Günther <guido.gunther@puri.sm>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

drirc: add a workaround for ARMA 3

Cc: 18.2 <mesa-stable@lists.freedesktop.org>

anv/batch_chain: Don't start a new BO just for BATCH_BUFFER_START

Previously, we just went ahead and emitted MI_BATCH_BUFFER_START as
normal. If we are near enough to the end, this can cause us to start a
new BO just for the MI_BATCH_BUFFER_START which messes up chaining. We
always reserve enough space at the end for an MI_BATCH_BUFFER_START so
we can just increment cmd_buffer->batch.end prior to emitting the
command.

Fixes: a0b133286a3 "anv/batch_chain: Simplify secondary batch return..."
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107926
Tested-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

anv: Use separate MOCS settings for external BOs

On Broadwell and above, we have to use different MOCS settings to allow
the kernel to take over and disable caching when needed for external
buffers.  On Broadwell, this is especially important because the kernel
can't disable eLLC so we have to do it in userspace.  We very badly
don't want to do that on everything so we need separate MOCS for
external and internal BOs.

In order to do this, we add an anv-specific BO flag for "external" and
use that to distinguish between buffers which may be shared with other
processes and/or display and those which are entirely internal.  That,
together with an anv_mocs_for_bo helper lets us choose the right MOCS
settings for each BO use.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99507
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

meson: remove invalid "opencl" llvm component

Seeming copy/paste mistake from configure.ac which uses $2 for the
component and $3 for the fancy name printing.

Cc: Dylan Baker <dylan@pnwbakers.com>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

Revert "mesa: remove unnecessary 'sort by year' for the GL extensions"

This reverts commit 3d81e11b49366b5636b8524ba0f8c7076e3fdf34.

As reported by Federico, some games require the 'sort by year' since
they truncate the extensions which do not fit the fixed size string
array.

Seemingly I did not consider that, as the documentation (both Mesa and
Nvidia) mentions about program crashes ... which are worked around by
setting the env. variable.

This commit reinstates the workaround and enhances the documentation.

Cc: Marek Olšák <maraeo@gmail.com>
Cc: Ian Romanick <idr@freedesktop.org>
Reported-by: Federico Dossena <info@fdossena.com>
Fixes: 3d81e11b493 ("mesa: remove unnecessary 'sort by year' for the GL
extensions")
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Tested-by: Federico Dossena <info@fdossena.com>

mesa: reorder and document the tokens in glheader.h

Split into different sections, document each one as well as strange
cases like GL_ATI_texture_compression_3dc.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

mesa: remove duplicate declarations from glheader.h

Remove all the desktop GL and GLX entries from the list.
Former are pulled by the gl.h and glext.h includes at the top while the
latter are no longer needed.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i965: reference __DRI_ATTRIB_SWAP_COPY token over the GLX one

Earlier commit updated the code to use the DRI tokens, yet forgot to
update the comment.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i915: reference __DRI_ATTRIB_SWAP_COPY token over the GLX one

Earlier commit updated the code to use the DRI tokens, yet forgot to
update the comment.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

dri/common: move the required GLX_* token definitions locally

Will allow us to remove even bigger hack elsewhere. But more
importantly, we should not be using _any_ GLX tokens in DRI.

Document the gory details about the current side-effects.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

dri/common: use __DRI_ATTRIB_SWAP* instances when describing db_modes

Somewhat recently Thomas Hellstrom added the respective DRI tokens
and updated the drivers. Update the documentation to match reality.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

egl/x11: remove eglSwap* surface check

Already handled further up in eglapi.c.

To make things a tiny bit strange, X11+DRI3 was doing the wrong thing by
returning EGL_FALSE (+ no error), while X11+DRI2 was returning EGL_TRUE.

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Eric Engestrom <eric.engestrom@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/surfaceless: remove eglSwap* stubs

The API validation in eglapi.c already returns if the surface type is
!window.

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Cc: Gurchetan Singh <gurchetansingh@chromium.org>
Cc: Chad Versace <chadversary@chromium.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

egl/drm: remove eglSwap* surface check

Already handled further up in eglapi.c

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

egl/android: remove eglSwap* surface check

Already handled further up in eglapi.c

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

egl: make eglSwapBuffers* a no-op for !window surfaces

Analogous to the previous commit - the spec says the function is a
no-op when a pbuffer or pixmap surface is used.

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

egl: make eglSwapInterval a no-op for !window surfaces

As the spec says, the function is a no-op when the surface is not a
window one.

That spec implies that EGL_TRUE should be returned in that case, yet
the ARM driver seems to return EGL_FALSE + EGL_BAD_SURFACE.

The Nvidia driver returns EGL_TRUE. We follow that behaviour until a
decision is made.

https://gitlab.khronos.org/egl/API/merge_requests/17

Cc: samiuddi <sami.uddin.mohammad@intel.com>
Cc: Erik Faye-Lund <kusmabite@gmail.com>
Cc: Tomasz Figa <tfiga@chromium.org>
Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

freedreno: add the a6xx sources to the Android build

Add the files otherwise things just won't build.
Haven't actually tested it, but it's a small step in the right
direction.

Fixes: de3b34df973 ("freedreno: Add a6xx backend")
Cc: Kristian H. Kristensen <hoegsberg@chromium.org>
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>

pipe-loader: add a dup() in pipe_loader_sw_probe_kms

The pipe_loader_release API closes the fd given, even if the pipe-loader
should _not_ take ownership of it.

With earlier commit we fixed pipe_loader_drm_probe_fd, and now with
cover the final piece.

Note that unlike the DRM case, here the caller _did_ forget to dup
before using it ... most likely leading to all sorts of fun.

Don't forget the close in the error path. Seems like the things are a
bit leaky/asymmetrical with the semi-recent config work. But we can shave
that yak another day ;-)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

pipe-loader: move dup(fd) within pipe_loader_drm_probe_fd

Currently pipe_loader_drm_probe_fd takes ownership of the fd given.
To match that, pipe_loader_release closes it.

Yet we have many instances which do not want the change of ownership,
and thus duplicate the fd before passing it to the pipe-loader.

Move the dup() within pipe-loader, explicitly document that and document
all the cases through the codebase.

A trivial git grep -2 pipe_loader_release makes things as obvious as it
gets ;-)

Cc: Leo Liu <leo.liu@amd.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Axel Davy <davyaxel0@gmail.com>
Cc: Patrick Rudolph <siro@das-labor.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Axel Davy <davyaxel0@gmail.com> (for nine)

st/nine: do not double-close the fd on teardown

As the newly introduced comment says:
The pipe loader takes ownership of the fd

Thus, there's no need to close it again.

Cc: Patrick Rudolph <siro@das-labor.org>
Cc: Axel Davy <davyaxel0@gmail.com>
Cc: mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <davyaxel0@gmail.com>