Mesa 26.2.7 Release Notes / 2026-11-11

Mesa 26.2.7 is a bug fix release which fixes bugs found since the 26.2.6 release.

Mesa 26.2.7 implements the OpenGL 4.6 API, but the version reported by glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) / glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used. Some drivers don’t support all the features required in OpenGL 4.6. OpenGL 4.6 is only available if requested at context creation. Compatibility contexts may report a lower version depending on each driver.

Mesa 26.2.7 implements the OpenCL 3.1 API, but the version reported by the CL_DEVICE_VERSION, CL_DEVICE_NUMERIC_VERSION and CL_DEVICE_OPENCL_C_ALL_VERSIONS clGetDeviceInfo queries depends on the particular driver being used.

Mesa 26.2.7 implements the Vulkan 1.4 API, but the version reported by the apiVersion property of the VkPhysicalDeviceProperties struct depends on the particular driver being used.

SHA checksums

TBD.

New features

• None

Bug fixes

• # va: H.264 encode spins forever in vaRenderPicture on some packed SPS headers (vl_rbsp misreads emulation prevention bytes)

• Asahi OpenCL compiler packing assertion failed

• Confidential issue #16794

• Fails to build on macOS < 10.12

• LLVM 23, riscv64, compile error gallium/auxiliary/gallivm/lp_bld_init_orc.cpp (API change?)

• Last Of Us Part 2 Remastered has glitches on hair strands - since Mesa 26.0.1

• RadeonSI syncrhonization/barrier issue with Clear+Dispatch (Repro included)

• V3D: Labwc fails to allocate GBM buffers wider than 4096 pixels unless non-MSAA texture size limit is enabled

• VC4: Kodi on Pi 3: DRM PRIME / V4L2 fails after Mesa 26.2.6 with avfilter_graph_config: Function not implemented (-38)

• Windows ARM memory-leak

• [ANV] Flickering black blocks in Project Wingman

• [OpenMW] Certain shaders cause visual corruption at the bottom of the screen when anti-aliasing is enabled, possibly Z-buffer related

• ac/radv/radeonsi: PS restricted to 2 of 6 CUs on Rembrandt 660M with fully harvested SA0 (~50% graphics perf loss)

• anv: compute shader writes to a storage image are partly lost after a render pass cleared the image earlier in the same command buffer on Intel LNL

• asahi: 64-bit multiply of a sign-extended sub-32-bit value produces wrong result

• llvmpipe/lavapipe: freed device memory is never released from the memfd

• radv: PixelInterlockOrderedEXT critical sections of consecutive draws in one subpass are not ordered (Radeon 780M, GFX11)

• radv: VK_EXT_descriptor_heap causes performance slow down in FSR 4.1 workloads

• rusticl/llvmpipe: heap corruption / SIGSEGV in `llvmpipe_set_global_binding` when kernels on two queues are released by one `clWaitForEvents`

• zink/ntv: pack_snorm_2x16 and unpack_snorm_2x16 unimplemented, aborts the process

Changes

Ali Milhim (1):

• zink: implement snorm/unorm pack and unpack ops in ntv

Benjamin Cheng (2):

• radv/video_enc: fix qp map va calculation

• radv/video_enc: fix VCN5 qp map buffer size

Birk Skyum (1):

• freedreno: Don’t double-close imported syncobj fd

Caio Oliveira (3):

• intel/mda: Preserve binary data in printraw

• nir/intel: Derive push-data divergence from its offset

• util: Initialize memstream outputs on Xbox

Calder Young (1):

• intel: Exclude helper lanes for optimized divergent atomics

Christian Gmeiner (2):

• llvmpipe: Track global bindings per context

• panvk: Correct maxUniformBufferRange for older archs

Christoph Pillmayer (4):

• pan/perf: Fix swapped block counts

• pan/ds: Add missing percent unit

• panvk/utrace: Fix payload assert

• pan/ds: Report correct gpu_clock_id

Dave Airlie (2):

• lavapipe: fix pipeline cache size return and checks

• gallivm: disable apx feature for now.

David Rosca (6):

• ac/vcn_dec: Fix division by zero in AV1 film grain params

• ac/vcn_dec: Clamp num_y/cb/cr_points and ar_coeff_lag in AV1 film grain params

• va: Unmap coded buffer on encode failure

• dri: Move driDestroyConfigs from glx/dri_common and use it

• glx: Fix driver_configs leak when creating display gpu screen

• egl/dri2: Fix driver_configs leak when creating display gpu screen

Dhruv Mark Collins (3):

• tu/autotune: Order dropping an RP history reference before its reaping

• tu/autotune: Remove latency tracking entries once their spikes are old

• tu/autotune: Take one RP history reference per RP decided

Dmitry Baryshkov (3):

• freedreno/ir3: wait for the indirect grid copy before loading it

• dri: round up the subsampled plane size on dma-buf import

• ir3: count half registers when deciding to double the thread size

Dom Cobley (1):

• v3d: don’t drop the last row in the sand30 blit

Duncan Brawley (2):

• pvr: Fix pipeline executable properties segfault

• pvr: Support pipeline executable properties VkPipelineCreateFlags2 flags

Eric Engestrom (5):

• docs: add sha sum for 26.2.6

• .pick_status.json: Update to db5503a26f76af5238bb6e706ba5daec2dbf45d2

• .pick_status.json: Mark 9c82a78f18dd60907e380dec60c46336eab64b25 as denominated

• .pick_status.json: Mark b492b11ecf5a72aa74f818b45c4930cd334474d8 as denominated

• .pick_status.json: Mark 988037f55378f077e9129b3df17729ea16337fd5 as denominated

Erik Faye-Lund (3):

• nir: use dst_access for the store

• panvk: restore zero-value for flush_id after copy_fbd

• panvk: add missing flush after copy_fbd

Fabricio Akio (1):

• freedreno/a5xx: Program the stream output buffer stride

Faith Ekstrand (2):

• nir/lower_multiview: Check variable modes before deleting

• panfrost,nir: Get rid of nir_downgrade_pls_vars()

Francisco Jerez (1):

• intel/gen: Fix possible corruption of instructions other than standard one/two-source layout during pre-compaction.

Georg Lehmann (4):

• aco/opt_postRA: check quad alignment of sgprs

• aco/opt_postRA: fix split_vector reassign with late kill operands

• nir/opt_load_store_vectorize: handle load_heap_descriptor

• nir/opt_large_constants: fix constant deduplication edge case

Icenowy Zheng (1):

• gallivm: adapt for JITLink in LLVM 23

Jeongik Cha (1):

• gfxstream: guard Goldfish sync fd collection in queue submit

Jesse Natalie (1):

• d3d12: Fix hash lookup for command signatures

Job Noorman (1):

• ir3: don’t emit alias.rt for relative const movs

Jose Maria Casanova Crespo (5):

• v3dv: pad the shader assembly BO for the QPU instruction prefetch

• v3d: pad shader uploads for the QPU instruction prefetch

• broadcom/compiler: align shared memory size to zero-initialization chunk

• vc4: compute the clip window after the predraw blits

• v3d/driconf: expose non-MSAA texture limits for labwc

Karmjit Mahil (1):

• tu: Skip zero sized inline uniform blocks

Karol Herbst (4):

• nir/load_libclc: do not compare sys_path against NULL

• nir/load_libclc: turn sys_path into an array

• nir/load_libclc: add llvm-23 libclc paths

• zink: cast last operands of SpvOpGroupNonUniformShuffle* ops to unsigned

Lionel Landwerlin (1):

• anv: fix PIPE_CONTROL with fragment->compute barrier on Xe2+

Louis-Francis Ratté-Boulianne (1):

• vulkan: initialize swapchain private mutex properly

Marek Olšák (3):

• ac/gpu_info: calculate min/max_good_cu_per_sa correctly

• radv: suspend conditional rendering in CmdResetQueryPool

• radv: suspend conditional rendering in radv_barrier

Mary Guillemard (4):

• agx: Skip signext optimisation in case of absolute

• hk: Properly handle pCounterBuffers being NUL

• hk: Stop adjusting the index buffer range

• libagx: Fix libagx_memcpy_to_aligned handling of unaligned start

Matt Arsenault (1):

• radv/llvm: keep the ring offsets in s[0:1] for GFX11+ compute shaders

Matt Turner (2):

• nir: don’t hardcode dedup_entry size in static_assert

• llvmpipe: hash cpu caps up to num_L3_caches, not 5 dwords

Mel Henning (1):

• nir: Fix incorrect is_ssa replacement

Mengyang Liu (1):

• aco: demote exec if condition is all-ones

Mike Blumenkrantz (18):

• st/mesa: don’t release the shared pixel map view after glDrawPixels

• st/mesa: add deref sources to the second pixel map lookup

• st/mesa: read the b and a pixel maps from the right channels

• st/mesa: use a new pixel map texture for each update

• zink: copy the spirv when compiling injected tcs

• mesa: free the trace dispatch table with the other tables

• util/tc: don’t full_copy an rp info that has ended

• util/tc: only set end_call for a blit that ends the info

• util/tc: advance the driver rp info before the first draw or clear

• util/tc: advance the driver rp info for indirect and vstate draws

• util/tc: move the driver rp info advance call list into a helper

• util/tc: write the resolve geometry before enqueueing the resolve

• util/tc: don’t merge a resolve into a different target

• util/tc: parse the rp info for mesh draws

• util/tc: don’t end the rp info without an end_call

• util/tc: reset the new rp info fence before linking it

• glsl: treat volatile as coherent when translating to nir

• zink: reset attachment storeOp when begin_rendering has no tc info

Mohamed Akram (1):

• glx/apple: fix build on macOS < 10.12

Nick Hamilton (1):

• cso: Add quirk that stops adding samplers with border color to the cache

Olivia Lee (2):

• panvk: fix mark_all_access_non_uniform metadata preservation

• pan/va: fix out-of-bounds read propagating swizzles on pseudo instructions

Omar Rashwan (1):

• iris: fix double free in iris_bufmgr

Patrick Lerda (1):

• r600: fix shaders.indexing.tmp_array.vertexid related issue

Paulo Zanoni (2):

• vulkan: export vk_queue_stop_submit_thread()

• anv: extract and call anv_queue_drain() during anv_DestroyDevice()

Pavel Ondračka (6):

• r300: handle underflow and overflow when packing FP24 constants

• r300: use 32-bit immediate indices when applying an index bias

• r300: skip constant attributes when splitting a negative index bias

• r300: clamp sampler LOD bounds before converting to integers

• glsl: propagate invariance after lowering

• r300: skip output modifier absolute or negated inputs

Pierre-Eric Pelloux-Prayer (2):

• gallium/ddebug: implement resource_release

• radeonsi: fix debug dump of buffer descriptors

Rhys Perry (1):

• nir/uub: fix imul by constant with overflow

Ruben Solvang (1):

• llvmpipe: release freed memory from the allocation memfd

Rusi Genev (3):

• v3d: Use the initial sample mask for centroid interpolation

• v3d: Fix interpolateAtOffset without MSAA

• v3d: Fix centroid interpolation at offsets

Samuel Pitoiset (4):

• nir/opt_access: do not infer ACCESS_CAN_REORDER with shader interlock

• spirv: fix memory semantics for image loads/stores

• aco: make sure SSBO SMEM loads respect barriers

• nir/lower_memory_model: don’t skip readonly accesses

Tapani Pälli (4):

• anv: change assume_full_subgroups drirc detection

• anv/dric: set anv_assume_full_subgroups for Last Of Us 2

• anv: add anv_force_exact_glsl_fma drirc option

• anv/drirc: use anv_force_exact_glsl_fma for Project Wingman

Timothy Arceri (5):

• ac,radv: move the pipe-misaligned metadata check to common code

• radeonsi: invalidate L2 before shaders read pipe-misaligned HTILE

• radeonsi: decompress resident depth textures rendered to after becoming resident

• radeonsi: flush the IB on NGG -> legacy transitions on Navi21 too

• radeonsi: write back L2 after DCC fast clears if RBs and L2 aren’t coherent

Vinson Lee (1):

• kopper: fix unused and uninitialized warnings on macOS

Xianzhong Li (1):

• panfrost/csf: initialize flush_id to zero in the tiler OOM handler

Xueliang Zhang (1):

• radeonsi: dirty DB_Z_INFO when the sample count changes without Z/S

Yiwei Zhang (14):

• venus: push set template shall grab a ref on the desc set

• venus: refactor out a vn_descriptor_set_layout_fix_create_info helper

• venus: sanitize GetDescriptorSetLayoutSupport ignored fields

• venus: amend a missing state in async set alloc

• venus: properly fail mutable descriptor alloc

• venus: respect descriptor template entry stride for accel struct

• venus: fix missed cases in vn_descriptor_heap_get_descriptor_count

• venus: descriptor size cache to only query supported types

• venus: always chain dedicated reqs for buffer cache init

• venus: image format and reqs cache to hash sType of pNext chain

• venus: fill image mem reqs with cache locked

• vulkan/wsi/x11: use resolved present mode to calculate image count

• kraid/encode_v9: validate swizzle before src_info check

• kraid/hw_tests: fix multiply overflow errors

airidosas252 (1):

• freedreno/fdl: Use NUM_ELEMENTS for A8XX buffer descriptors

alexandre minette (1):

• freedreno/ir3: use implicit LOD zero for a3xx multisample fetches

faetalize (2):

• util/vl_vlc: allow removebits at the top of the bit buffer

• util/rbsp: check each byte once for emulation prevention

genie1997 (1):

• tu/kgsl: Force 8-channel macrotile mode on A635

hamb3r (1):

• dzn: Handle missing variable descriptor count layout support output

mozixun (1):

• zink: don’t charge a batch the whole slab for a suballocated resource

rowanG077 (1):

• agx: fix phi register shuffling order