Skip to content

Release Notes for 2026.2

Latest

Choose a tag to compare

@mkoppara-amd mkoppara-amd released this 17 Nov 04:04
· 5 commits to master since this release
xilinx_v2026.2

Changes for 2026.2

bsp

This release adds frequency-based timer tick configuration for AXI and SCU timers, including a new XTimer_SetIntervalHz() API and TTC callback, improving tick accuracy in the FreeRTOS SDT flow by avoiding millisecond truncation. It also fixes MicroBlaze FreeRTOS tick setup to prevent interrupts from firing before the ISR stack is initialized.

Enhancements

  • Adds frequency-based tick interval for AXI and SCU timers to avoid millisecond truncation in FreeRTOS SDT flow
  • Adds XTimer_SetIntervalHz() API and TTC callback to program timer rates accurately without millisecond truncation

Bug Fixes

  • Fixes MicroBlaze FreeRTOS tick setup to prevent interrupts before ISR stack is initialized

libmetal_xlnx_extension_v1_0:

Bug Fixes

  • Fixes MicroBlaze IRQ masking so disabling one interrupt no longer enables all sources

lwip220_v1_5:

Fixes multicast filter callback prototypes so AXI Ethernet and GEM (EMACPS) builds succeed when IGMP/MLD is enabled, and corrects RX frame length validation and jumbo pbuf chain handling on the AXI Ethernet FIFO receive path. Also resolves several build errors in TCP-disabled configurations, including TFTP examples and bridge FDB in raw API mode.

Bug Fixes

  • Corrects multicast filter callback prototypes so AXI Ethernet builds with IGMP/MLD enabled succeed.
  • Corrects multicast filter callback prototypes so GEM (EMACPS) builds with IGMP/MLD enabled succeed.
  • Validates RX frame length and handles jumbo pbuf chains correctly on AXI Ethernet FIFO receive path.
  • Fixes build errors when TCP is disabled by properly defining LWIP_TCP as off.
  • Removes unused TCP timers from TFTP examples, fixing build errors when TCP is disabled.
  • Fixes compile error in bridge FDB when timers are disabled in raw API mode.

CMakeLists.txt

Enhancements

  • Moves BSP CMake build output to shorter path, avoiding Windows long-path build failures.

amd_xspi

Features

  • Adds HyperBus HyperRAM polled and interrupt example applications for the AXI XSPI driver.
  • Adds new standalone driver for the AXI XSPI IP supporting standard, Dual, Quad, Octal, and HyperBus modes.

asu_ram_ecc

Bug Fixes

  • Fixes Yocto build failure for the ASU RAM ECC configuration driver.

audio_formatter

The audio formatter driver now exclusively supports the SDT (System Device Tree) flow, with legacy non-SDT code removed. This release also restores builds on GCC 14 and later toolchains and hardens channel handling by rejecting invalid ChannelId values that could previously cause out-of-range register accesses.

Enhancements

  • Audio formatter driver now supports only the SDT flow, dropping legacy non-SDT code and data files.

Bug Fixes

  • Fixes audio_formatter interrupt handler return types to restore builds on GCC 14+ toolchains.
  • Audio formatter now rejects invalid or raced ChannelId values, preventing out-of-range register accesses.

axicdma_v4_17:

The AXI CDMA examples now use allocated arrays for buffers and descriptors instead of hardcoded DDR addresses, improving portability. This release also fixes current BD lookup failures under non-identity address mapping and adds validation of simple transfer lengths against the per-instance hardware limit.

Enhancements

  • AXI CDMA examples now use allocated arrays instead of hardcoded DDR addresses for buffers and BDs.

Bug Fixes

  • Fixes AXI CDMA current BD lookup failing under non-identity address mapping.
  • AXI CDMA now validates simple transfer length against per-instance hardware limit.

axidma_v9_22:

Enhancements

  • AXI DMA examples now allocate buffer and BD space as global arrays instead of hardcoded DDR addresses.

axipcie_v3_6:

Bug Fixes

  • Fixes AXI PCIe global interrupt enable/disable so writes take effect correctly.

axivdma_v6_18:

Enhancements

  • AXI VDMA example now allocates buffers/BDs as global arrays instead of hardcoded DDR addresses, improving portability.

bram_v4_15:

Bug Fixes

  • Corrects BRAM interrupt status read so ECC error status is reported accurately.

canfd_v2_13:

Bug Fixes

  • Fixes CANFD driver build failure on versal_2ve_2vm when XilPM support is enabled.

clk_wiz_v1_12:

The clk_wiz driver adds new APIs for getting and setting clock frequencies, controlling clock input selection, and directly configuring MMCM M/D/O divisors for dynamic reconfiguration. Divisor calculation now validates the PFD frequency range to prevent unstable PLL configurations.

Enhancements

  • clk_wiz validates PFD frequency range during divisor calculation to prevent unstable PLL configurations.
  • clk_wiz adds APIs to get/set clock frequencies and control clock input selection.
  • clk_wiz adds APIs for direct MMCM M/D/O divisor configuration and reconfiguration.

csi

Bug Fixes

  • Fixes CSI virtual channel register offset so channels VC1-15 read correct VCxINF1R registers.

csi2tx

Enhancements

  • MIPI CSI-2 TX driver adds 16 virtual channel and 8 pixels-per-clock support.

csi2txss

This release adds support for the MIPI CSI-2 TX subsystem v3.0, including multi-virtual-channel operation for 4, 8, or 16 virtual channels. It also introduces runtime dynamic line rate configuration (400–4500 Mbps) for the CSI-2 TX PHY PLL and adds MIPI TX PHY sub-core support via new txphy-present and txphy-connected device-tree properties.

Features

  • Adds MIPI CSI-2 TX subsystem v3.0 support with multi-virtual-channel (4/8/16 VC) capability.
  • Supports 16 virtual channels and 8 pixels-per-clock operation.

Enhancements

  • Adds runtime dynamic line rate configuration (400-4500 Mbps) for the MIPI CSI-2 TX PHY PLL.
  • Adds MIPI TX PHY sub-core support with new txphy-present/connected device-tree properties.

clockps_v1_8:

Bug Fixes

  • Fixes clockps mux parent lookup to handle external/uninitialized parents, preventing bad node decoding and invalid function calls

csudma_v1_20:

Bug Fixes

  • Fixes CSU DMA stale-data reads on Cortex-A53 32-bit mode causing wrong SHA3 hashes and secure-boot failures.

dp12

This release hardens the DisplayPort driver against malformed EDID data and MST topology messages, fixing numerous out-of-bounds reads and writes, stack buffer overflows, and memory corruption that could occur during EDID parsing, MST topology discovery, and sideband message handling. It also prevents a divide-by-zero fault on hot-plug with malformed preferred timing and limits topology discovery recursion to avoid stack overflow.

Bug Fixes

  • Fixes out-of-bounds read in DisplayPort DisplayID data block parsing from malformed EDID.
  • Fixes out-of-bounds read when issuing GUIDs during MST topology discovery.
  • Bounds MST topology discovery recursion to the 15-hop DisplayPort RAD limit, preventing stack overflow.
  • Prevents divide-by-zero fault on DisplayPort hot-plug with malformed EDID preferred timing.
  • NACKs MST sideband replies when I2C/DPCD read handlers fail, preventing stack memory leak.
  • Fixes out-of-bounds write from unchecked port count in LINK_ADDRESS sideband reply.
  • Validates MST sideband message length and CRC before dispatching, preventing out-of-bounds reads.
  • Corrects remote I2C read chunk size and checks remote EDID read status to avoid stale data.
  • Bounds sideband reply copy to requested length, preventing stack buffer overflow on EDID/GUID reads.
  • Clamps MST payload timeslot allocation to table size, preventing out-of-bounds memory corruption.
  • Bounds MST topology node and sink tables to prevent out-of-bounds writes.

dp12rxss

Features

  • Adds RISC-V platform support to DisplayPort TX/RX subsystem examples for KCU105 and KC705 boards.

Enhancements

  • Converts DisplayPort TX subsystem to a clean SDT-only configuration flow, removing legacy BSP paths.

dp12txss

The DisplayPort TX/RX subsystem examples now support RISC-V platforms on the KCU105 and KC705 boards, and the TX subsystem driver has moved to a clean SDT-only configuration flow. This release also resolves several runtime issues, including crashes from NULL pointer dereferences during MST hot-plug and link bring-up, an out-of-bounds DPCD cache read that reported incorrect sink capabilities, and a fix that preserves YCbCr dynamic range and colorimetry so compliance sinks no longer blank YCbCr video. Additional fixes correct remote I2C/EDID read handling and prevent leaking uninitialized stack memory on AUX read failures.

Features

  • Adds RISC-V platform support to DisplayPort TX/RX subsystem examples for KCU105 and KC705 boards.

Enhancements

  • Converts DisplayPort TX subsystem to a clean SDT-only configuration flow, removing legacy BSP paths.

Bug Fixes

  • Preserves configured YCbCr dynamic range and colorimetry so compliance sinks no longer blank YCbCr video.
  • Prevents leaking uninitialized stack memory when an AUX read fails during link info report.
  • Fixes NULL pointer dereference during MST hot-plug when no downstream sinks are discovered.
  • Corrects remote I2C read chunk size and checks remote EDID read status to avoid stale data.
  • Fixes out-of-bounds DPCD cache read that reported bogus sink capability values.
  • Fixes crash during link bring-up from NULL dereference in VTC interlaced-mode check.

dp14

This release hardens the DisplayPort (dp14) driver against a range of memory-safety issues, fixing several out-of-bounds reads and writes and stack/buffer overflows in EDID parsing and MST topology discovery that could be triggered by malformed EDIDs or sideband messages. It also resolves a divide-by-zero fault on hot-plug with invalid EDID timing, validates MST sideband message length and CRC before dispatch, and corrects remote I2C/EDID reads to avoid returning stale data.

Bug Fixes

  • Fixes out-of-bounds read in DisplayID data block search within EDID buffer for DisplayPort.
  • Fixes out-of-bounds read when issuing GUID during MST topology discovery.
  • Prevents stack overflow and buffer overrun during MST topology discovery by enforcing the 15-hop RAD limit.
  • Avoids divide-by-zero fault on DisplayPort hot-plug with malformed EDID preferred timing.
  • NACKs MST sideband replies on I2C/DPCD read handler failure to prevent leaking firmware stack memory.
  • Fixes out-of-bounds write by clamping LINK_ADDRESS reply port count to array capacity.
  • Validates MST sideband message length and CRC before dispatch, preventing out-of-bounds read.
  • Fixes remote I2C read chunk size and checks remote EDID read status to avoid stale data.
  • Prevents stack buffer overflow by clamping MST sideband reply copy to requested length.
  • Fixes out-of-bounds write by clamping MST payload timeslot allocation to the 64-entry table.
  • Prevents out-of-bounds write by bounding MST topology node and sink tables to their capacity.

dp14rxss

Enhancements

  • DisplayPort RX subsystem now programs rx_dec_clk via the clocking wizard driver APIs instead of direct register writes.
  • Adds RISC-V support to DisplayPort 1.4 TX, RX, and pass-through example applications.

dp14txss

The DisplayPort 1.4 TX subsystem now supports RISC-V in its TX, RX, and pass-through example applications and has moved to an SDT-only build flow. This release also resolves several runtime issues, including a NULL pointer dereference during hot-plug with empty MST topologies, a crash in the VTC interlaced-mode check during link bring-up, remote I2C/EDID read errors that could return stale data, and debug reports printing uninitialized memory on failed AUX reads.

Enhancements

  • Adds RISC-V support to DisplayPort 1.4 TX, RX, and pass-through example applications.
  • Moves DisplayPort 1.4 TX subsystem to SDT-only flow, removing legacy non-SDT build artifacts.

Bug Fixes

  • Prevents DisplayPort TX debug reports from printing uninitialized stack memory on failed AUX reads.
  • Fixes NULL pointer dereference during hot-plug when an MST topology reports zero downstream sinks.
  • Fixes remote I2C read chunk size and checks remote EDID read status to avoid stale data.
  • Fixes crash in VTC interlaced-mode check when video mode is unresolved during link bring-up.

dp21_v2_7

The DisplayPort 2.1 driver now supports Display Stream Compression (DSC) with FEC, configurable via the xlnx,enable-dsc option, along with YUV420 VSC SDP transmission and reception for SST and up to 4 MST streams. This release also hardens MST topology and sideband message handling against numerous out-of-bounds accesses, stack overflows, and divide-by-zero faults from malformed EDIDs or hot-plug events, and improves link training clock recovery fallback.

Features

  • Adds DisplayPort 2.1 Display Stream Compression (DSC) and FEC support, configurable via xlnx,enable-dsc.
  • Adds YUV420 VSC SDP transmission support for up to 4 MST streams.
  • Adds YUV420 VSC SDP reception and decode for SST and MST streams.

Enhancements

  • Verifies INTRAHOP_AUX_REPLY_INDICATION is cleared before link training per DP2.1 CTS.
  • Adds VTC setup callback, improves EDID timing parsing, and adds video mode helpers.

Bug Fixes

  • Aligns YUV420 VSC SDP transmission with VSYNC when VSC colorimetry is enabled.
  • Corrects assert macros to match function return types in DisplayPort driver.
  • Prevents out-of-bounds read when parsing DisplayID data blocks from malformed EDID.
  • Fixes out-of-bounds GUID table access during MST topology discovery.
  • Bounds MST topology recursion to the DisplayPort RAD limit, preventing stack overflow.
  • Avoids divide-by-zero fault on hot-plug when EDID preferred timing fields are zero.
  • Correctly NACKs sideband replies when I2C/DPCD read handlers fail, preventing memory leaks.
  • Clamps LINK_ADDRESS reply port count to prevent out-of-bounds write.
  • Validates sideband message length and CRC before dispatching to prevent out-of-bounds reads.
  • Fixes remote I2C read chunk size and validates EDID read status before use.
  • Fixes incorrect LinkCountTotal value in RX sideband reply constructors.
  • Bounds sideband reply copy to requested length, preventing stack buffer overflow.
  • Clamps MST payload timeslot allocation to table size, preventing memory corruption.
  • Bounds MST topology node and sink tables to prevent overflow writes.
  • Improves link training clock recovery fallback sequencing and poll interval.

dp21rxss_v1_7

The DP2.1 RX Subsystem driver now supports Display Stream Compression (DSC) with FEC, YUV420 VSC SDP reception and decode for SST and MST streams, and adds platform support for VCK190 and VEK385 in TX-only and pass-through applications. Enhancements include YUV420 and Static HDR SDP detection and forwarding across the pass-through, MST, and HDCP examples, RISC-V support for those examples, and automatic decode Clocking Wizard programming on link-rate changes. Key fixes correct the RX start and DP core initialization sequences to prevent unpredictable behavior, resolve MSA corruption from an incorrect MISC0 colorimetry rebuild, and cap the TX link rate to UHBR10 on VCK190/VEK385 when the RX cannot support higher UHBR rates.

Features

  • Adds VCK190 and VEK385 platform support for DP2.1 TX-only and pass-through applications.
  • Adds Display Stream Compression (DSC) and FEC support to the DP2.1 RX Subsystem driver.
  • Adds YUV420 VSC SDP reception and decode for SST and MST streams.

Enhancements

  • Adds YUV420 format detection and forwarding to the DP2.1 RX MST pass-through example.
  • Adds YUV420 format detection and forwarding to the DP2.1 RX HDCP pass-through example.
  • Adds Static HDR SDP forwarding through the DP2.1 RX HDCP pass-through path to preserve HDR metadata.
  • Adds YUV420 format detection and forwarding to the DP2.1 RX pass-through example.
  • Adds a dedicated RX-only linker script for the DP2.1 RX-only example.
  • Driver now programs the decode Clocking Wizard automatically on link-rate changes via Clocking Wizard APIs.
  • Adds RISC-V support to the DP2.1 RX pass-through, MST, and HDCP examples.
  • Adds Static HDR SDP detection and forwarding to the DP2.1 RX pass-through example.

Bug Fixes

  • Caps DP2.1 TX link rate to UHBR10 on VCK190/VEK385 for unsupported RX UHBR rates in pass-through.
  • Fixes incorrect MISC0 colorimetry rebuild that corrupted MSA information in pass-through mode.
  • Corrects DP RX start sequence in example so XDpRxSs_Start is called after PHY init.
  • Fixes DP core initialization sequence to avoid unpredictable RX behavior.

dp21txss_v2_8

The DP2.1 TX subsystem now supports the VCK190 and VEK385 platforms for TX-only and pass-through applications, and adds Display Stream Compression (DSC) with FEC as well as YUV420 VSC SDP transmission for up to four MST streams. Enhancements include driver-side Clocking Wizard programming on link-rate changes, an EDID-driven TX example that trains the link and configures the stream from sink capabilities, and new subsystem-level AUX read/write APIs for DPCD debugging. This release also resolves several runtime issues, including a hot-plug NULL pointer dereference when MST reports no sinks, a crash during interlaced-mode checks with unresolved video modes, incorrect HDCP authentication reporting, and remote I2C/EDID read handling.

Features

  • Adds VCK190 and VEK385 platform support for DP2.1 TX-only and pass-through applications.
  • Adds Display Stream Compression (DSC) and FEC support to the DP2.1 TX subsystem.
  • Adds YUV420 VSC SDP transmission support for up to 4 MST streams.

Enhancements

  • Programs link-derived Clocking Wizard from the driver on link-rate changes, with new callback hooks.
  • Reworks TX example to train link and configure stream from sink EDID instead of hardcoded modes.
  • Adds VTC setup callback, improves EDID timing parsing, and adds video mode helpers.
  • Adds subsystem-level AUX read/write APIs for DPCD debugging without direct core access.

Bug Fixes

  • Caps DP2.1 TX link rate to UHBR10 on VCK190/VEK385 for unsupported RX UHBR rates in pass-through.
  • Prevents disclosure of uninitialized stack memory in capability and link info debug reports.
  • Fixes NULL pointer dereference during hot-plug when MST topology reports no sinks.
  • Fixes remote I2C read chunk size and validates EDID read status before use.
  • Fixes crash in VTC interlaced-mode check when video mode is unresolved.
  • Prints MSA M/N video values in hex for correct display on RISC-V designs.
  • Reads LTTPR repeater capability first after HPD to meet DP CTS requirements.
  • Initializes authentication status to fail-safe default, fixing incorrect HDCP reporting.

dphy

Bug Fixes

  • Fixes DPHY data lane 4-7 status register access for 8-lane designs.

dppsu

Enhancements

  • DPPSU driver now uses SDT-only initialization and configuration flow, removing legacy non-SDT paths.

Bug Fixes

  • Fixes link failures in DPPSU interrupt and poll examples by providing missing platform/stream stub definitions.

dsi

The DSI driver now supports Ultra-Low Power State (ULPS) for reduced idle power and LPDT escape mode for sending DSI command and DCS packets in low-power mode. This release also fixes long packet transfers by sending the header before the payload, preventing command engine stalls.

Enhancements

  • Adds Ultra-Low Power State (ULPS) support to the DSI driver for reduced idle power.
  • Adds LPDT escape-mode support to send DSI command and DCS packets in low-power mode.

Bug Fixes

  • Fixes DSI long packet transfers by sending header before payload, preventing command engine stalls.

dsi2rx

The DSI2RX driver now detects lane-configuration and word-count corruption interrupts from malformed packet headers. This release also fixes incorrect interrupt enable/disable behavior for non-CRC interrupts and corrects the interrupt handler to properly report CRC/ECC errors and route callbacks.

Enhancements

  • Adds DSI2RX handling for lane-config and word-count corruption interrupts for malformed packet headers.

Bug Fixes

  • Fixes DSI2RX interrupt enable/disable incorrectly asserting when enabling non-CRC interrupts.
  • Fixes DSI2RX interrupt handler using wrong constants, causing missed CRC/ECC errors and misrouted callbacks.

dsirxss

Bug Fixes

  • Fixes DSI RX subsystem sub-core base address calculation for correct register access.
  • Fixes out-of-bounds write during DSI RX subsystem init on sub-core lookup miss.

dsitxss

The MIPI DSI TX Subsystem driver now supports Ultra-Low Power State (ULPS) and LPDT escape-mode transport, and migrates exclusively to the System Device Tree (SDT) flow (legacy flow removed). This release also fixes an out-of-bounds write during initialization when a sub-core base address lookup fails.

Enhancements

  • Adds ULPS (Ultra-Low Power State) support to the MIPI DSI TX Subsystem driver.
  • Adds LPDT escape-mode transport support to the MIPI DSI TX Subsystem driver.
  • DSI TX Subsystem driver now supports only the System Device Tree (SDT) flow; legacy flow removed.

Bug Fixes

  • Fixes out-of-bounds write during DSI TX init when sub-core base address lookup fails.

hdcp1x

The HDCP1x driver now uses the System Device Tree (SDT) flow exclusively, removing legacy non-SDT build paths. This release also fixes several KSV list buffer overflow issues on both the transmit and receive sides by bounding repeater device counts and validating repeater topology, preventing out-of-bounds writes and stack corruption.

Enhancements

  • Converts HDCP1x driver to SDT-only flow, removing legacy non-SDT code paths and build files.

Bug Fixes

  • Prevents HDCP1x RX KSV list buffer overflow by clamping repeater device count to 32.
  • Prevents HDCP1x TX KSV list overflow by bounding device count and using cached repeater status.
  • Fixes HDCP1x DP KSV FIFO reset size to prevent stack overflow into DPCD.
  • Validates HDCP1x TX repeater topology to prevent out-of-bounds writes during KSV list processing.

hdcp22_cipher

Enhancements

  • HDCP22 cipher driver now uses SDT-only configuration flow and base-address lookup.

hdcp22_cipher_dp

Enhancements

  • HDCP22 cipher DP driver converted to SDT-only configuration flow, dropping legacy BSP and DeviceId support.

hdcp22_common

Enhancements

  • Ensures sensitive RSA-CRT intermediate secrets are reliably wiped from stack memory under compiler optimization.
  • Clears temporary AES key and state buffers after use to remove residual key material from memory.

hdcp22_rng

Enhancements

  • HDCP22 RNG driver now uses SDT-only flow; legacy DeviceId configuration removed.

hdcp22_rx

This release hardens the HDCP 2.2 RX driver against several buffer overflows, including fixes to OAEP padding validation and decrypt output bounding, as well as protection against oversized HDMI DDC messages. It also resolves authentication reliability issues by handling transient empty downstream device counts in repeater topology and fixing a RepeaterAuth_Send_Ack race that could drop authentication in state C5.

Bug Fixes

  • Fixes HDCP 2.2 RX OAEP decode to reject malformed padding and prevent buffer overflow.
  • Prevents buffer overflow by bounding HDCP 2.2 RX OAEP decrypt output to Km buffer size.
  • Prevents receive buffer overflow from oversized HDMI DDC messages in HDCP 2.2 RX.
  • Allows empty downstream device count in HDCP 2.2 RX repeater topology during transient authentication.
  • Fixes HDCP 2.2 RX authentication drop caused by RepeaterAuth_Send_Ack race in state C5.

hdcp22_rx_dp

Bug Fixes

  • Fixes HDCP 2.2 DP receiver buffer overflow by validating OAEP-decoded Km length before copy.
  • Fixes HDCP 2.2 DP repeater authentication dropped when transmitter sends Ack early in state C5.

hdcp22_tx

Fixes a build failure on GCC 15 that blocked ZynqMP HDCP-enabled platforms. This release also hardens HDCP22 repeater authentication and SRM processing by preventing receive-buffer overflow and revocation-list corruption from oversized Receiver ID List messages and rejecting undersized SRMs before verification.

Bug Fixes

  • Fixes HDCP22 TX build failure on GCC 15 that blocked ZynqMP HDCP-enabled platforms.
  • Prevents receive-buffer overflow and revocation-list corruption from oversized Receiver ID List messages during HDCP22 repeater authentication.
  • Rejects undersized SRMs before verification, preventing out-of-bounds access during HDCP22 SRM processing.

i2srx

The I2S RX driver now resolves debug log ring buffer corruption under interrupt load and properly rejects out-of-range channel IDs to prevent register overwrites. The driver now supports only the System Device Tree flow.

Enhancements

  • Removes legacy non-SDT flow from I2S RX driver, retaining only the System Device Tree flow.

Bug Fixes

  • Fixes I2S RX debug log ring buffer corruption and bounds issues under interrupt load.
  • Fixes I2S RX channel mux to reject out-of-range channel IDs and prevent register overwrites.

i2stx

The I2S TX driver now requires the System Device Tree (SDT) flow and no longer supports legacy non-SDT builds. This release also hardens the driver by rejecting invalid channel IDs to avoid out-of-range register writes and fixing debug log corruption when the ring buffer fills from interrupt context.

Enhancements

  • I2S TX driver now supports only the System Device Tree (SDT) flow, removing legacy non-SDT build support.

Bug Fixes

  • Rejects invalid I2S TX channel IDs to prevent out-of-range register writes.
  • Prevents I2S TX debug log corruption by safely handling full ring buffer from interrupt context.

dcsub

The DC driver adds a new API to select the audio sample frequency and program channel status bits, along with a DC bypass SST mode that routes PL AVPG/VTC video directly to DP TX with the blender bypassed. Enhancements include configurable audio sample-rate selection in the DPDC example, DP video streams now driven from the run-time format and PPC configuration, PL IP base addresses taken from xparameters.h for custom address maps, and explicit failure on unsupported video timing modes. Several video regressions are fixed, including no-video issues in live, mixed, and nonlive functional modes, correct DC reset handling after DP link training, and accurate cursor positioning via wider blend coordinates.

Features

  • Added XDcSub_SetAudioSampleFrequency() API to select audio sample frequency and program channel status bits.
  • Added DC bypass SST mode routing PL AVPG/VTC video directly to DP TX with blender bypassed.

Enhancements

  • DPDC example now fails explicitly on unsupported video timing modes instead of silently defaulting to 1080p60.
  • Removed RGB888_GFX graphics format from the DC driver's supported video format list.
  • DPDC examples now take PL IP base addresses from xparameters.h, supporting custom address maps instead of hardcoded ones.
  • Enabled fixed-priority arbitration for audio SDP slots for deterministic delivery at higher sample rates.
  • DPDC example now applies the selected audio sample frequency instead of defaulting to 48 kHz.
  • Added audio sample frequency selection menu to the DPDC example supporting six sample rates.
  • DP video stream now configured from RunConfig format and PPC instead of hardcoded RGB 8-bpc single-pixel setup.
  • DPDC example gains full MST bring-up for up to 4 streams from 4 different AVTPG for different resolutions.

Bug Fixes

  • Fixed no-video regression by driving DP at single pixel mode in live and mixed functional modes.
  • Fixed no-video issue by releasing the DC block reset after DP link training regardless of boot CDO state.
  • Fixed cursor positioning by widening blend coordinates to u16 to match 14-bit hardware fields.
  • Fixed invalid DP PixModeSel in functional nonlive mode by initializing PPC and validating input.
  • Removed RGB888_GFX graphics format from the DC driver's supported video format list
  • Bypass clock wizard programming updated as per new design

i3c_v1_4:

Bug Fixes

  • Fixes I3C Hot Join dynamic address assignment by resetting FIFOs and controller state before DAA.

i3cpsx_v1_8:

The I3C driver now supports controller-mode Hot-Join and In-Band Interrupt (IBI) reception, along with new interrupt-driven slave send/receive capability and a loopback example application. This release also fixes FIFO handling for payloads that are not a multiple of 32-bit words, recovers targets left unresponsive after a receive FIFO overflow, and corrects slave transmit arming so master reads no longer return an address NACK.

Enhancements

  • Adds controller-mode Hot-Join support to the I3C driver.
  • Adds controller-mode In-Band Interrupt (IBI) reception support to the I3C driver.
  • Adds I3C slave interrupt-driven send/receive with new loopback example application.

Bug Fixes

  • Recovers an I3C target left unresponsive after a receive FIFO overflow error.
  • Fixes I3C TX/RX FIFO handling for payloads not a multiple of 32-bit words.
  • Fixes I3C slave transmit arming so master reads no longer get address NACK.

intc_v3_23:

Bug Fixes

  • Fixes INTC cascade interrupt handling for multi-root topologies in the SDT flow.

mbox_v4_11:

Bug Fixes

  • Fixes mailbox interrupt example for non-FSL designs in SDT flow by adding required threshold/interrupt configuration.

mcdma_v1_15:

Bug Fixes

  • AXI MCDMA: validate channel ID in interrupt handlers to prevent out-of-bounds access
  • AXI MCDMA: propagate buffer-address errors so misaligned descriptors aren't queued with invalid addresses

mipi_tx_phy

Enhancements

  • MIPI TX PHY driver adds runtime dynamic line rate reconfiguration (400-4500 Mbps).

mmidp:

Enhancements

  • Add better support for MST (Multi-Stream Transport)
  • XMmiDp_MstModeEnable() now takes an Enable argument so MST can be turned off as well as on.
  • XMmiDp_MstActSeqEnable() performs a CCTL0 read-modify-write, preserving non-MST fields.
  • InitiateActSeq() replaces the CCTL0 self-clear poll with DPCD ACT_HANDLED, adding a 1 ms settle before polling.
  • Derives MST VIDEO_CONFIG5 TU from target symbol slots in 1/8-slot fixed point instead of the SST average path.
  • Assigns StartSlot from the final TimeSlots after distribution so each stream's slot follows the prior stream's allocation.
  • Dimensions MstStreamConfig[] by XMMIDP_STREAM_ID4 instead of the lane count.
  • Clears Doxygen 1.15 warnings in the CCTL0 setters, XMmiDp_SetCoreCtrl(), XMmiDp_SetMstVcpTable0..7(), and the SDP vertical control members.

Bug Fixes

  • MMI DisplayPort driver now validates pixel-mode selection to reject invalid PPC mappings.
  • Fixes XMMIDP_VIDEO_CONFIG5_AVG_BYTES_PER_TU_FRAC_MASK and INIT_THRESHOLD_MASK to match the correct [19:14] and [13:7] register bit fields.
  • Fixes link training starting with MST still enabled by clearing MST_EN, UP_REQ_EN, and CCTL0 ENABLE_MST_MODE before training, and gating MSTM_CTRL writes on DPCD rev >= 0x12.
  • Fixes ConfigureVideoController() interleaving VIDEO_CONFIG5 writes with MST payload setup by writing VIDEO_CONFIG5 for all streams before payload allocate and ACT.
  • Fixes XMmiDp_ComputeTimeSlot() retaining the previous StartSlot instead of clearing it on every recompute path.
  • Fixes WaitActHandledStatus() not clearing AuxReply before each read, which could misreport a failed or partial AUX read as ACT_HANDLED.
  • Fixes SetPixModeSel() accepting and storing invalid pixel-mode values by asserting on invalid PPC to PixModeSel mappings.

mipicsiss

Adds DSI TXSS display support with a Raspberry-Pi DSI panel in the SCU200 HPIO/XP5IO MIPI example designs, and introduces a runtime menu to switch between 1080p60 and 4Kp60 resolutions in the VEK385 D-PHY example. This release also corrects the CPHY active lane count to 3 in the VEK385 example.

Features

  • Adds DSI TXSS display support with Raspberry-Pi DSI panel in SCU200 HPIO/XP5IO MIPI example designs.

Enhancements

  • Adds runtime menu to select between 1080p60 and 4Kp60 resolutions in VEK385 D-PHY example.

Bug Fixes

  • Corrects CPHY active lane count to 3 in VEK385 example.

nandps_v2_12:

Bug Fixes

  • NAND PS driver now fails init when total block count exceeds driver limits, preventing buffer overflow.

ospipsv_v1_15:

The OSPI driver now uses pattern-based RX PHY tuning for more reliable read calibration, with updated examples demonstrating the new flow. This release also fixes several issues affecting real hardware, including a DAC address range check that could access the wrong memory, STIG write corruption for mixed-width and oversized transfers, flash reset on Versal NET, and calibration failures from an incorrect input clock on SpartanUP.

Enhancements

  • Updates OSPI examples to demonstrate the new pattern-based RX PHY tuning flow
  • Adds pattern-based RX PHY tuning for OSPI, improving read calibration over READ ID based tuning

Bug Fixes

  • Fixes OSPI DAC read/write address range check that could wrap and access wrong memory
  • Fixes flash reset on Versal NET by using GPIO reset so the flash returns to default state
  • Fixes OSPI STIG write path for mixed-width protocols and oversized transfers that corrupted data
  • Honors caller-supplied OSPI input clock on SpartanUP, preventing calibration failures and boot abort

pciepsu_v2_4:

Bug Fixes

  • Fixes PCIe PSU bus number overflow and unbounded recursion during device enumeration.

qspipsu_v1_24:

Adds support for Macronix mx25u25635f 256 Mb QSPI NOR flash to the QSPIPSU examples. This release also fixes a bus conflict on quad-mode commands, corrects address conversion and stripe-flag handling for MCP flash in parallel mode, and resolves build failures on ARMR5 and aarch64.

Enhancements

  • Adds support for Macronix mx25u25635f 256 Mb QSPI NOR flash in QSPIPSU examples.

Bug Fixes

  • Fixes address conversion for MCP flash and stripe flag handling in parallel mode in QSPIPSU examples.
  • Fixes QSPIPSU build failure on ARMR5 and aarch64 by defining XPS_SYS_CTRL_BASEADDR.
  • Fixes QSPIPSU bus conflict on quad-mode commands by queuing the full transfer before starting.

rfdc_v13_2:

This release fixes RFDC device initialization to accept device name strings containing either underscores or dashes, preventing libmetal open failures. It also addresses out-of-bounds memory access during clock distribution setup and multi-tile synchronization DTC calculations.

Bug Fixes

  • Fixes RFDC device initialization to accept both underscore and dash in device name strings, preventing libmetal open failures.
  • Adds range validation of source tile indices in XRFdc_SetClkDistribution to prevent out-of-bounds access.
  • Fixes out-of-bounds array write during RFDC multi-tile synchronization DTC code calculation.

sdiaud

Enhancements

  • SDI audio driver now exclusively uses the System Device Tree (SDT) flow, dropping legacy non-SDT support.

Bug Fixes

  • Prevents NULL callback dereference when handling SDI audio status-update interrupts.

spdif

Enhancements

  • SPDIF driver now exclusively supports the System Device Tree (SDT) flow, removing legacy non-SDT code.

Bug Fixes

  • Prevents divide-by-zero in SPDIF GetFs() when input is idle or invalid.

spi_v4_17:

Bug Fixes

  • Fixes SPI slave-mode buffer overflow by checking receive buffer space in RX read loops.
  • Rejects unaligned ByteCount in XSpi_Transfer to prevent buffer overrun with 16/32-bit data widths.

spips_v3_15:

Bug Fixes

  • Fixes SPI PS driver hangs by bounding hardware polling loops and properly releasing chip select on faults.

sysmonpsu_v3_4:

Enhancements

  • SysMonPSU supports PCAP status reads via SMC for EL1 non-secure/virtualized environments

tmr_manager_v1_10:

Bug Fixes

  • Fixes TMR Manager recovery handler register save/restore bugs and SDT Magic1 config copy.

tmrctr_v4_16:

Bug Fixes

  • Initializes timer/counter state fields during TmrCtr config init to ensure consistent behavior.

trngpsv_v1_9:

Bug Fixes

  • Fixes buffer overflow in TRNGPSV PTRNG mode with derivation function corrupting caller memory.
  • Corrects TRNGPSV DRNG seed-life off-by-one and keeps instance recoverable after reseed-required state.

trngpsx_v1_7:

Bug Fixes

  • TRNGPSX now correctly flags invalid-parameter failures instead of returning success.
  • TRNGPSX enforces DRNG seed life and reseeds before HRNG data collection per NIST SP 800-90A.

ttcps_v3_24:

Enhancements

  • Adds TTCPS driver build support for the Versal 2VE/2VM XilPM client.

Bug Fixes

  • Fixes out-of-bounds access in TTCPS when no matching base address is found, including SDT interrupt example.

ufspsxc_v1_3:

The UFS driver now caches the device capacity during initialization to avoid repeated descriptor queries. This release also validates LUN IDs against the device's LUN count to prevent out-of-range access and corrects an incomplete PHY high-speed calibration check.

Enhancements

  • Caches raw UFS device capacity during init to avoid repeated descriptor queries.

Bug Fixes

  • Validates UFS LUN IDs against device LUN count to prevent out-of-range access.
  • Fixes incomplete UFS PHY high-speed calibration check.

usb_v5_9:

Enhancements

  • Adds shared helpers for USB mass-storage command rejection and sense-data reporting in example applications.

Bug Fixes

  • USB mass-storage examples now reject out-of-range SCSI READ/WRITE commands instead of accessing invalid memory.

usbps_v2_12:

Enhancements

  • USBPS mass-storage example adds a shared CSW helper for consistent BOT command-status handling.

Bug Fixes

  • USBPS mass-storage example now rejects out-of-bounds SCSI READ/WRITE requests, preventing memory access beyond virtual flash.

usbpsu_v1_20:

Bug Fixes

  • Fixes USBPSU SDT build failure when hibernation is enabled.

v_frmbuf_rd

Enhancements

  • Frame buffer read example now selects correct DDR base address for SCU200 and KC705 boards.
  • HLS video example applications now distinguish MIG vs LPDDR memory controllers in MB_V designs.

v_frmbuf_wr

Enhancements

  • Frame buffer write example selects DDR base address per board (SCU200, KC705).
  • HLS video example applications now distinguish MIG vs LPDDR memory controllers in MB_V designs.

v_gamma_lut

Bug Fixes

  • Fixes Gamma LUT bounds validation to prevent out-of-range MMIO accesses from invalid offset/length.

v_hdmi_common

This release fixes several buffer overflow and out-of-bounds memory issues in the HDMI common library when parsing and handling HF-VSIF InfoFrame packets, improving robustness against malformed or oversized payloads.

Bug Fixes

  • Fixes HF-VSIF InfoFrame buffer overflow by validating payload lengths in HDMI common library.
  • Fixes buffer overflow during HF-VSIF ECC byte insertion in HDMI common library.
  • Fixes out-of-bounds disparity copy when parsing HF-VSIF packets in HDMI common library.

v_hdmiphy1

The HDMI PHY driver now supports GTHE4, GTYE4, GTYE5, and GTYP transceivers from a single runtime build, and raises the LR2 reference-clock limit to 340 MHz on GTYE5/GTYP to enable higher HDMI 2.1 pixel rates on Versal. This release also fixes an uninitialized value returned on failed DRP reads that could corrupt stack data.

Enhancements

  • Enables a single HDMI PHY driver build to support GTHE4, GTYE4, GTYE5 and GTYP transceivers at runtime.
  • Raises LR2 reference-clock upper bound to 340 MHz for GTYE5/GTYP, enabling higher HDMI 2.1 pixel rates on Versal.

Bug Fixes

  • Fixes uninitialized DRP read value returned on failed HDMI PHY reads, preventing stack residue corruption.

v_hdmirx1

Bug Fixes

  • Aligns HDMI RX1 version register offsets to updated IP core register map for correct access.
  • Fixes NULL pointer dereference in HDMI RX1 VTD interrupt handler for unsupported VIC values.

v_hdmirxss

Enhancements

  • Update HDMI RX example linker script to support RISC-V based BSP.

Bug Fixes

  • Correctly classify a 340 MHz sink as HDMI 1.4, preventing incorrect HDMI RX TX clock configuration.

v_hdmirxss1

The HDMI 2.1 Rx subsystem passthrough example now supports the SCU200 (MicroBlaze RISC-V) board and RISC-V processors, and adds HDCP repeater mode support.

Features

  • Adds SCU200 (MicroBlaze RISC-V) board support to the HDMI 2.1 passthrough example.

Enhancements

  • Enables RISC-V processor support in the HDMI RX passthrough example.
  • Adds HDCP repeater mode support to the HDMI 2.1 Rx subsystem passthrough example.

v_hdmitx

Bug Fixes

  • Fix out-of-bounds access in HDMI TX VSIF 3D metadata handling by bounding wire-supplied length.

v_hdmitx1

Bug Fixes

  • Aligns HDMI TX register offsets to updated IP core version register map for correct register access.

v_hdmitxss

Bug Fixes

  • Correctly classify HDMI 1.4 sinks advertising exactly 340 MHz TMDS clock in TX examples.

v_hdmitxss1

The HDMI 2.1 TX subsystem passthrough example now supports the SCU200 board and RISC-V (MicroBlaze RISC-V) processors, and adds HDCP repeater mode support.

Features

  • Adds SCU200 (MicroBlaze RISC-V) board support to the HDMI 2.1 passthrough example.

Enhancements

  • Adds RISC-V processor support to the HDMI TX passthrough example.
  • Adds HDCP repeater mode support to the HDMI 2.1 Tx subsystem passthrough example.

v_hscaler

Bug Fixes

  • Fixes potential buffer overflow in horizontal scaler by bounding phase array writes to max line width

v_letterbox

Bug Fixes

  • Fixes out-of-range color index read in V_Letterbox background color setting

v_mix

The Video Mixer driver now supports the v-mix 6.2 IP version and improves example portability by selecting the correct DDR base address per board (SCU200, KC705) and distinguishing MIG versus LPDDR memory controllers in MB_V designs. This release also fixes a layer array sizing issue so that a master layer plus 16 overlays no longer aliases the logo layer.

Enhancements

  • Mixer example selects DDR base address by board type for SCU200 and KC705 portability.
  • Adds device-tree binding support for v-mix 6.2 IP version.
  • HLS video example applications now distinguish MIG vs LPDDR memory controllers in MB_V designs.

Bug Fixes

  • Fixes mixer layer array sizing so a master plus 16 overlays no longer aliases the logo layer.

v_multi_scaler

Enhancements

  • Adds 3-planar YUV444 and third-buffer support, plus non-SDT init APIs to V Multi Scaler driver.

Bug Fixes

  • Corrects stride calculation for tile formats and adds YUV10/YUV12 support in multi-scaler example.

v_scenechange

Enhancements

  • Scene change detector example now supports RISC-V and XGPIO for broader platform portability.

Bug Fixes

  • Fixes divide-by-zero in scene change IRQ by rejecting invalid layer dimensions.

v_sdirxss

Enhancements

  • SCU200 PICXO example now runs from LPDDR, supporting designs where it no longer fits in BRAM.
  • SDI RX subsystem driver now supports only the System Device Tree (SDT) flow; legacy non-SDT paths removed.

v_warp_init

Bug Fixes

  • Fixes Video Warp Init descriptor to zero unused fields, preventing residual stack data in DMA descriptor.

video_common_v4_20:

This release fixes several out-of-bounds reads and a buffer overflow in EDID and DisplayID parsing—affecting wide resolutions, oversized serial number strings, invalid DisplayID section lengths, and multiple extension blocks—improving robustness when handling data from display sinks. It also adds a VTC setup callback, improved EDID timing parsing, and new video mode helper functions.

Enhancements

  • Adds VTC setup callback, improves EDID timing parsing, and adds video mode helpers.

Bug Fixes

  • Fixes out-of-bounds read in video mode table scan for wide resolutions like 10240.
  • Fixes buffer overflow when parsing oversized DisplayID serial number strings from a sink.
  • Fixes out-of-bounds read when parsing DisplayID sections with invalid length fields.
  • Fixes incorrect EDID extension indexing causing out-of-bounds reads with multiple extension blocks.

visp_ss_v2_2:

This release adds image splitter/stitcher support that enables ISP resolutions above 4K through PL tiling, along with HDMI and DisplayPort live-output paths for displaying ISP output on such designs. New Cortex-R52 and RPU bare-metal platform support is included, together with an APU-side watchdog to detect RPU stalls, selectable low/high DDR ISP buffer pools, and a layered mailbox subsystem with optional sequence-counter and CRC-16 integrity checking. It also delivers extensive robustness fixes across the mailbox, sensor, buffer, and display paths—resolving out-of-bounds accesses, NULL dereferences, use-after-free, memory leaks, and buffer overflows, and correcting per-ISP buffer initialization for independent ISP streaming.

Features

  • Adds HDMI display support for >4K splitter/stitcher designs via stitched output routing.
  • Add image splitter/stitcher support enabling ISP resolutions above 4K via PL tiling
  • Add APU-side watchdog notifier to detect RPU stalls and halt video processing on expiry
  • Add DisplayPort live-output path to the visp example for displaying ISP output over DP
  • Add Cortex-R52 platform support with linker, MPU, tracing, and interrupt handling
  • Add RPU bare-metal example support with R52 linker and MPU configuration

Enhancements

  • Aligns APU baremetal mailbox cookie allocation with Linux driver for integrity correlation.
  • Adds startup latency measurement reporting time to first ShowBuffer call.
  • Add support for low- and high-memory ISP buffer pool placement with runtime address detection
  • Refactor mailbox subsystem into layered transport, integrity, interface, and command design
  • Add opt-in mailbox sequence counter and CRC-16 checksum integrity checking
  • Add switchable low/high DDR buffer pools and split large pool sizing
  • Improve multi-instance handling with per-instance indexing and skipping disabled instances
  • Make ISP buffer DDR region selectable between low and high memory via linker switch

Bug Fixes

  • Corrects per-ISP output buffer initialization and output-mode gating for independent ISP streaming.
  • Fixes LILO dump output so both memory and live output dumps are printed.
  • Uses actual FBWR buffer sizes for LILO dump commands to avoid truncation or over-read.
  • Fixes metadata stack leak by copying the structure correctly in ConfigMetadata.
  • Hardens osFread with argument validation and overflow checks to prevent bad reads.
  • Validates sensor device ID before array indexing to prevent out-of-bounds access.
  • Prevents NULL dereference and cleans up resources on sensor initialization failures.
  • Adds bounds validation for IBA config table access to prevent invalid lookups.
  • Fixes off-by-one hardware ID bounds checks in camera device APIs.
  • Disables lens shading correction for split ISP tiles to prevent tile corruption.
  • Prevents integer overflow and NULL allocation in RAW-to-RGBA conversion.
  • Fixes IDT 8T49N24x divider table overflow and broken clock range checks.
  • Skips HDMI sink capability derivation when EDID read or parse fails.
  • Clamps VMix HDMI layer coordinate index to prevent out-of-bounds window config.
  • Bounds TI TMDS1204 register table index to avoid out-of-bounds read on unknown revision.
  • Bounds Onsemi NB7NQ621M register table index to prevent out-of-bounds reads.
  • Fixes undersized stack array causing out-of-bounds access in display loop.
  • Rejects USERPTR buffer counts exceeding per-bufio slot limit to prevent buffer corruption.
  • Fixes out-of-bounds reads during AR0820 metadata scanning.
  • Validates buffer path before indexing to reject unrecognized path names safely.
  • Validates peer-controlled mailbox FIFO fields to prevent divide-by-zero and out-of-bounds access.
  • Validates mailbox payload size against buffer limit before copy to prevent overflow.
  • Clears stored sensor handle after unregister to prevent use-after-free.
  • Frees sensor context and validates pipe ID before publishing handle in create path.
  • Fix memory-pool slot leak and dangling handle on invalid pipeId in OX08B40 sensor init
  • Prevent null-pointer writes when sensor memory allocation fails during multi-sensor init
  • Validate sensor device ID bounds in OX03F10 before publishing handle or register access
  • Check allocation failure in osFread() to avoid DMA write to null address when loading raw images
  • Fix camera device init zero-init loop to cover full valid hardware ID range
  • Validate buffer count before packing address lists to prevent stack overflow in buffer pool creation
  • Stop leaking uninitialized caller memory onto mailbox in DPCC get requests
  • Stop leaking uninitialized caller memory onto mailbox in white-balance get requests
  • Stop leaking uninitialized caller memory onto mailbox in DG get requests
  • Stop leaking uninitialized caller memory onto mailbox in gamma-curve get requests
  • Stop leaking uninitialized caller memory onto mailbox in GTM get requests
  • Stop leaking uninitialized caller memory onto mailbox in histogram get requests
  • Validate hpId and port number across all six ISP instances to prevent config memory corruption

vphy

Enhancements

  • Adds extended 5-bit TX voltage swing (TXDIFFCTRL) support in the HDMI PHY driver.

Bug Fixes

  • Fixes HDMI PHY to correctly treat 340 Mcsc rate as HDMI 1.4 for TMDS clock ratio.

vprocss

The video processing subsystem now validates PIP/Zoom windows against input and output dimensions before programming VDMA, preventing invalid configurations, and fixes an unsigned integer overflow that could bypass window bounds checks. HLS video example applications also now distinguish between MIG and LPDDR memory controllers in MB_V designs.

Enhancements

  • HLS video example applications now distinguish MIG vs LPDDR memory controllers in MB_V designs.

Bug Fixes

  • Validate PIP/Zoom windows against input/output dimensions before programming VDMA, preventing invalid configurations.
  • Fix unsigned integer overflow in window size validation that could bypass bounds checks.

xdmapcie_v3_3:

Enhancements

  • Adds VEK385 board support with I2C init and GPIO-based PCIe reset sequence.

Bug Fixes

  • Corrects PCIe link-up status detection on Versal 2VE/2VM devices.

zdma_v1_23:

Bug Fixes

  • Fixes ZDMA scatter-gather descriptor count validation to prevent out-of-bounds writes and reject busy channels.

Findcommon.cmake

Enhancements

  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.

UserConfig.cmake

Enhancements

  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.

toolchainfiles

Enhancements

  • Unifies Cortex-R5/R52 assembly to use the armclang integrated assembler

standalone_v9_6:

This release introduces RISC-V M/U privilege mode-switching support, including a public API, monitor and guest example applications, and the associated BSP build integration. Enhancements add RV32D/RV64D double-precision floating-point support, RV32E compatibility, Versal Gen2 secure-register programming through PLM, and new client APIs for VID power state and crypto capability queries. Key fixes resolve ARMv8 cache invalidation and _sbrk heap bounds errors, out-of-bounds MPU region handling on Cortex-R52, cache-range and trap-handler infinite loops in MicroBlaze and RISC-V BSPs, and SDT multi-root INTC cascade interrupt handling.

Features

  • Adds RISC-V mode-switch monitor and guest example applications
  • Adds RISC-V M/U mode-switch entry, exit, and resume assembly paths
  • Implements RISC-V mode-switch control, syscall dispatch, and abort handling
  • Adds public API for RISC-V M/U privilege mode switching
  • Adds saved-session layout for RISC-V mode-switch resume
  • Adds DDR EDAC error-source tracing so PLM can report per-controller error snapshots for secure address decode.

Enhancements

  • Adds RV32D/RV64D double-precision floating-point BSP support with misaligned FP load/store emulation
  • Routes Versal Gen2 EL3 secure-register init through PLM for XPPU-safe programming
  • Detects hard-float ABI for Cortex-A9 BSPs and exposes it via bspconfig
  • Adds XPm_GetVidPowerState client API to read the PLM-selected VID rail mode.
  • Adds RISC-V guest U-mode BSP build support for mode switch
  • Adds build integration options for RISC-V mode-switch monitor and guest BSPs
  • Integrates U-mode ECALL and fault handling into shared RISC-V trap handler
  • Adds MPP and PMPCFG register definitions for RISC-V mode switch
  • XilSecure clients can query PLM server library version and compiled-in crypto capability bitmap.
  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.
  • Enables DAP device-state detection during HWRoT boot on Versal 2VP_P.
  • Adds RV32E compatibility to RISC-V startup and trap code
  • Moves BSP CMake build output to shorter path, avoiding Windows long-path build failures.
  • Enables PMC watchdog with reset action on versal_2vp_p for watchdog-triggered device reset.
  • Unifies Cortex-R5/R52 assembly to use the armclang integrated assembler

Bug Fixes

  • Fixes incorrect set-field bit offset in ARMv8 D-cache invalidate routine
  • Fixes xil_vprintf %.*s bounded-string formatting defect
  • Fixes RISC-V _sbrk heap-growth wrap-around that could return invalid memory
  • Fixes out-of-bounds heap growth in ARM _sbrk/sbrk bounds checking
  • Fixes RISC-V vless and Versal PLM build failures for mode switch
  • Fixes out-of-bounds MPU region index on partial range overlap for Cortex-R52
  • Fixes execution-level cmakedefine01 defines failing for non-empty variables
  • Fixes _sys_read/_sys_write return values for armclang semihosting
  • Enables PMU event export on Cortex-R5 so TCM ECC events reach the ISR
  • Fixes Xil_ICacheInvalidateRange on ARMv8 to invalidate the full range
  • Fixes out-of-bounds stack read in putnum by NUL-terminating buffer
  • Fixes INTC cascade interrupt handling for multi-root topologies in the SDT flow.
  • getentropy stub now returns ENOSYS so callers can detect missing entropy
  • Adds missing input validation to XilNvm eFuse and BBRAM client APIs, rejecting invalid parameters properly.
  • Adds missing msleep prototype to sleep header to fix compilation
  • Fixes MBV64 trap handler jump-table alignment causing infinite loop
  • Fixes cache range infinite loop in MicroBlaze and RISC-V BSPs

xilcert_v1_7:

This release adds OCP and certificate services—including DevIK/DevAK key management, attestation, and PCR measurement—along with X.509 certificate signature OID support for Versal 2VP Prime devices. It also fixes a stack buffer overrun that could occur when formatting the full PLM version string in certificates.

Features

  • Adds OCP and certificate services (DevIK/DevAK, attestation, PCR measurement) for Versal_2VP_P.

Enhancements

  • Adds X.509 certificate signature OID support for Versal 2VP Prime devices.

Bug Fixes

  • Fixes stack buffer overrun when formatting full PLM version string in certificates.

xilffs_v5_7:

Enhancements

  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.

Bug Fixes

  • Fixes out-of-bounds access and invalid LUN handling in xilffs disk I/O layer.
  • Fixes buffer overflow in UFS polled example on IAR/ICCARM builds.

xilloader_v2_5:

This release brings up new security and boot features for Versal 2VP and 2VP_P devices, including ECDSA P-521 HashBlock authentication, OCP and certificate services (DevIK/DevAK, attestation, PCR measurement), TPM support, JTAG configuration, and configurable glitch detectors for PMC, ASU, and LPD. DDRMC warm-boot support reuses stored calibration data to reduce boot time, and new client APIs let applications query library version, feature capabilities, and partition load addresses, alongside added support for Macronix 256Mbit QSPI flash. Numerous security hardening fixes address out-of-bounds accesses, buffer overruns, and memory disclosure from malformed PDIs, correct LMS/LMS-HSS signature verification, prevent crashes during authentication, and resolve APU ELF load failures and PDI corruption over SMAP.

Features

  • Adds ECDSA P-521 HashBlock authentication support on Versal 2VP_P.
  • Adds new API/IPI command to query the tlbin partition load address.
  • Adds DDRMC warm-boot support to reuse stored calibration data and reduce boot time.
  • Adds client APIs to query XilLoader library version and feature-capability bitmap.
  • Adds OCP and certificate services (DevIK/DevAK, attestation, PCR measurement) for Versal_2VP_P.
  • Adds IPI command and example to configure PMC, ASU, and LPD glitch detectors.
  • Adds TPM support for Versal_2VP, configurable via BSP settings.
  • Adds JTAG configuration example and support for Versal_2VP.
  • Adds JTAG configuration example and support for Versal_2VP_P.
  • Adds IPI command to configure a DDR address for subsystem image restart.

Enhancements

  • Adds PCR measurement from the hash block during image load on Versal 2VP_P.
  • Supports multiple algorithm bits in hybrid field for classical/PQC key pairing on Versal 2VP_P.
  • Improves DDRMC I2C handshake buffer validation with dedicated error codes and correct offset handling.
  • Adds glitch-redundancy hardening to the Auth JTAG A-HWRoT check.
  • Moves RSA PSS signature verification into XilSecure
  • Adds user-configurable PDI size input to the validate-PDI-auth example.
  • Adds support to bypass IDCODE checks via an IPI command flag.
  • Uses PDI optional data for OSPI RX PHY tuning with graceful DDR-to-SDR fallback.
  • Enables DAP device-state detection during HWRoT boot on Versal 2VP_P.
  • Restricts glitch detector IPI command to supported platforms only.
  • Aligns client-side authenticated PDI validation with server authentication on Versal_2VE_2VM.
  • Rejects reserved Image Store PDI ID range in Add/Remove IPI commands.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.
  • Enables support for different I2C masters for DDR5 across all platforms.
  • Logs the authentication algorithm name during PDI authentication.
  • Starts ASUFW keep-alive monitor on handoff to ASU firmware image.
  • Adds partition RevokeID verification to client-side Validate Auth PDI.
  • Defeatures Symmetric HWRoT boot mode on new devices.
  • Adds support for Macronix 256Mbit (mx25u256) QSPI flash parts.
  • Updates I2C device ID parsing for new DDRMC5_I2C_MASTER parameter format.
  • Adds input parameter validation to non-secure client APIs.
  • Adds timeout handling for OSPI 4-byte address mode polling to prevent indefinite hangs
  • Skips unnecessary SHA digest calculation for non-RSA/ECDSA authentication types

Bug Fixes

  • Fixes out-of-bounds access by validating NumBaseAddr in xilloader handshake command parsing.
  • Adds safety mitigations in CFI/CFRAME handling and defers Image Store PDI publish until DMA succeeds.
  • Enforces SPK ID revocation during key authentication on Versal 2VP_P, rejecting revoked boot images.
  • Fixes LMS/LMS-HSS signature length derivation to prevent incorrect verification of padded signatures.
  • Fixes NULL pointer use in LMS Auth JTAG that could crash during authentication.
  • Detects short SD/UFS filesystem reads to prevent copying partial or stale data.
  • Bounds HashBlock copy sizes to prevent buffer overrun from malformed PDIs.
  • Bounds DDRMC exchange offset to prevent out-of-aperture privileged accesses.
  • Fixes OSPI flag-status read overflowing a one-byte buffer onto PLM stack.
  • Clamps DFU GETSTATUS reply length to prevent disclosure of adjacent PLM memory over USB.
  • Validates image and partition counts to prevent out-of-bounds header access from malformed PDIs.
  • Runs Configuration Limiter lockdown before boot PDI load so lockdown completes earlier.
  • Resets I2C instance state after release to avoid reuse of stale driver state.
  • Unlocks LPD SLCR write protection so glitch-detector example register writes take effect.
  • Fixes false FIFO-drain warning by re-arming SBI only on SMAP PDI sources.
  • Fixes glitch-detector configuration to return early only on invalid-parameter errors.
  • Fixes PLM build failures for Versal_2VP_P by initializing module UpdateHandler fields.
  • Fixes APU ELF partition load failures on Versal_2VP_P.
  • Uses correct OSPI controller reset in DDR-to-SDR fallback on Versal 2VE/2VM.
  • Prevents PDI corruption when loading multiple PDIs over SMAP by re-arming sync detection.
  • Fix AES GMAC DMA byte-swap for Versal_2VP, Versal_2VP_P, and Versal_2VE_2VM parts.
  • Restores BootPdi fallback when Image Store PDI list is empty during restart.
  • Bounds image/partition header counts to reject malformed image header tables.
  • Ensure TRNG is initialized for client use on Versal 2VP regardless of PLM_OCP setting

xilnvm_v3_8:

XilNvm now supports second AES/user eFuse keys with independent write-lock and validation, adds APIs to query the PLM library version and feature capabilities, and enables the MicroBlaze RISC-V soft core as a client target on Versal_2VE_2VM. This release also fixes several eFuse read/write issues on 64-bit clients with 32-bit servers, corrects range and parameter validation to prevent buffer overflows and rejected invalid requests, and resolves stale-data reads caused by uncleared sticky status. Additional fixes address PLM build failures on Versal_2VP_P and strengthen input validation across the eFuse and BBRAM client APIs.

Enhancements

  • Enables D-cache by default for XilNvm client examples with matching cache maintenance on IPI-shared buffers.
  • Adds XilNvm APIs letting clients query the PLM library version and feature capability bitmap.
  • Adds support for second AES/user eFuse keys with independent write-lock, CRC, and validation hooks.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.
  • Removes XSysMonPsv driver dependency from PLM using direct register access to reduce binary size.

Bug Fixes

  • Fixes XilNvm eFuse reads into 64-bit buffers above 4GB being written to truncated addresses on 32-bit PMC.
  • Corrects XilNvm eFuse cache and user-fuse range checks, rejecting unaligned, zero-length, or empty requests.
  • Adds validation of user-fuse write parameters to prevent local buffer overflow in XilNvm IPI handler.
  • Fixes XilNvm user-fuse IPI write for 64-bit client with 32-bit server, avoiding pointer truncation.
  • Clears sticky RD_DONE status between eFuse row reads to avoid sampling stale data.
  • Fixes PLM build failures for Versal_2VP_P by initializing module UpdateHandler fields.
  • Fixes mislabeled security-control and PPK2 hash validation messages in XilNvm eFuse examples.
  • Adds missing input validation to XilNvm eFuse and BBRAM client APIs, rejecting invalid parameters properly.
  • Fixes pointer aliasing in 64-bit memory copy by splitting into direction-specific APIs.

xilocp_v1_8:

XilOCP now supports OCP services—including PCR measurement, DevIK/DevAK key management, and attestation—on the Versal 2VP and 2VP_P platforms, with new client-side APIs and examples covering hardware and software PCRs. New client APIs let applications query the library version and feature capabilities from the PLM, and the MicroBlaze RISC-V soft core is now a supported client target on Versal 2VE/2VM. This release also fixes several security-relevant issues, including corrected DICE measurement binding (which changes derived device identity on affected parts), rejection of invalid DICE CDI seeds and out-of-bounds attestation key requests, improved validation of CDI and hash transfer destinations, and corrected boot-mode and secure-config measurement handling.

Features

  • Adds OCP (PCR, key management, attestation) support for Versal 2VP_P.
  • Adds OCP and certificate services (DevIK/DevAK, attestation, PCR measurement) for Versal_2VP_P.

Enhancements

  • Adds XilOCP client-side support and examples for Versal 2VP platform.
  • Adds HW PCR and SW PCR client support and example for Versal 2VP_P platform.
  • Adds client APIs to query XilOCP library version and feature capability bitmap from the PLM server.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.

Bug Fixes

  • Fixes missing boot-mode disable bits in secure config measurement by reading from eFuse cache on Versal Net/2VE/2VM/2VP_P.
  • Validates public key offset in attestation buffer to reject out-of-bounds key wrap requests.
  • Fixes DICE measurements to bind full byte lengths; changes derived device identity on affected parts.
  • Guards extra PPK eFuse fields to prevent build failures.
  • Rejects all-zero DICE CDI seed instead of accepting it as valid.
  • Fixes PLM build failures for Versal_2VP_P by initializing module UpdateHandler fields.
  • Fixes DME challenge flow to correctly return signature failures to the caller.
  • Validates CDI and subsystem hash transfer destinations lie within ASU RAM before copying.
  • Corrects secure config IP disable mask and removes inapplicable eFuse fields on Versal 2VP.
  • Fixes memory aliasing issue by using address-specific memcpy APIs.

xilpdi_v1_14:

This release adds a new API to query the tlbin partition load address, improves OSPI RX PHY tuning via PDI optional data with automatic DDR-to-SDR fallback, and adds compatibility validation for in-place ASU firmware updates. Several robustness fixes harden PDI parsing against malformed or corrupt images, preventing PMC RAM buffer overruns, out-of-bounds header access, and infinite loops, and correctly locate optional data in ASU update PDIs.

Features

  • Adds new API/IPI command to query the tlbin partition load address.

Enhancements

  • Uses PDI optional data for OSPI RX PHY tuning with graceful DDR-to-SDR fallback.
  • Adds compatibility validation for in-place ASU firmware updates before image load.
  • Hardens XilPdi hash table access by returning a const pointer to prevent modification

Bug Fixes

  • Fixes XilPdi optional data length bound check to prevent PMC RAM buffer overrun
  • Validates image and partition counts to prevent out-of-bounds header access from malformed PDIs.
  • Correctly locates optional data in ASU update PDIs by accounting for SMAP prefix
  • Prevents infinite loop when parsing corrupt zero-length optional data headers

xilplmi_v2_5:

This release expands platform support for Versal 2VP, 2VP_P, and 2VE/2VM, adding OCP security services (PCR, key management, attestation, DevIK/DevAK), TPM, authenticated JTAG, and new glitch-detector, DDR EDAC error tracing, and FPD power-domain capabilities. DDRMC warm-boot support reuses stored calibration data to reduce boot time, and new in-place ASU firmware update compatibility checks, keep-alive health monitoring, and memory-reachability validation improve runtime robustness. Numerous fixes address in-place PLM update failures, SSIT IPI access permission enforcement and buffer bounds, IPI dispatch on non-contiguous channels, watchdog base-address selection, and several out-of-bounds and NULL-pointer hardening issues.

Features

  • Adds DDRMC warm-boot support to reuse stored calibration data and reduce boot time.
  • Adds OCP (PCR, key management, attestation) support for Versal 2VP_P.
  • Adds OCP and certificate services (DevIK/DevAK, attestation, PCR measurement) for Versal_2VP_P.
  • Adds IPI command and example to configure PMC, ASU, and LPD glitch detectors.
  • Adds DDR EDAC error-source tracing so PLM can report per-controller error snapshots for secure address decode.
  • Adds FPD power domain power-up/power-down support with GIC proxy wake arming and CMN flush.
  • Adds PMC-RAM memory-range reachability table letting runtime consumers validate region access.
  • Adds TPM support for Versal_2VP, configurable via BSP settings.

Enhancements

  • Adds validation and read-back checks to glitch detector client APIs for reliable configuration.
  • Enables subsystem runtime support by default in the 2VE/2VM make build flow.
  • Unifies PLM address-range validation under a single configurable option, off by default.
  • Decouples TPM init via callback to break circular dependency between xilplmi and xiltpm.
  • Enables authenticated JTAG (PLM_AUTH_JTAG) support on Versal 2VP.
  • Adds ISP XMPU interrupt handling with violation address and master ID logging on 2VE/2VM.
  • Adds HB_MON_4-7 error IDs for Versal 2VP and 2VP_P platforms.
  • Uses PDI optional data for OSPI RX PHY tuning with graceful DDR-to-SDR fallback.
  • Enables DAP device-state detection during HWRoT boot on Versal 2VP_P.
  • Adds optional slave error handlers for clock monitor and address decode errors.
  • Adds compatibility validation for in-place ASU firmware updates before image load.
  • Enables PMC watchdog with reset action on versal_2vp_p for watchdog-triggered device reset.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.
  • Enables support for different I2C masters for DDR5 across all platforms.
  • Adds periodic ASUFW keep-alive health monitor to detect stalled ASU firmware.
  • Adds input parameter validation to non-secure client APIs.
  • Removes XSysMonPsv driver dependency from PLM using direct register access to reduce binary size.
  • Adds timeout handling for OSPI 4-byte address mode polling to prevent indefinite hangs
  • Moves memory-address validation security hardening to callers for improved code size and redundancy control.
  • Moves memory address validation security hardening responsibility to callers to reduce code size.

Bug Fixes

  • Fixes in-place PLM update failure by skipping redundant link_nodes replay.
  • Prevents PDI boot failure on 2VP builds lacking DDR EDAC by accepting link_nodes as no-op.
  • Restores update-disable eFUSE check ordering so disabled devices report correct status.
  • Fixes SSIT forwarding to enforce IPI access permissions, preventing unauthorized secure API execution.
  • Bounds SSIT forwarded message and response buffers to prevent out-of-bounds reads.
  • Validates update optional data to prevent address wrap and out-of-range reads.
  • Adds SSIT stub symbols so prebuilt XilSEM links on Versal 2VP/2VP_P.
  • Fixes IPI interrupt dispatch for designs with non-contiguous or disabled IPI channels.
  • Fixes PLM build failures for Versal_2VP_P by initializing module UpdateHandler fields.
  • Fixes wrong PMC watchdog base address selection on platforms with multiple family selectors.
  • Gates EM link_nodes command behind config macro to fix undefined symbol in certain builds.
  • Aborts in-place ASU firmware update cleanly when ASUFW reports busy, allowing later retries.
  • Fixes false ASU not-alive error by using global control register for firmware presence detection.
  • Clears latched SSS configuration error bit on versal_2vp_p during PCR extension.
  • Fixes event logging client example buffer length units and incorrect cache invalidation.
  • Fixes PMC APB handler to clear PMC Analog Slave error interrupt status.
  • Bounds image/partition header counts to reject malformed image header tables.
  • Rejects out-of-range task priority to prevent out-of-bounds task queue access.
  • Adds NULL pointer check in module registration to prevent dereference.
  • Fixes out-of-bounds access in buffer move loop.
  • Fixes pointer aliasing in 64-bit memory copy by splitting into direction-specific APIs.
  • Fixes unaligned byte copy logic in memory set function.

xilpm_v7_1:

This release adds UCIe and CPM6/GTMPW housecleaning support on Versal, and improves Linux kexec reliability by re-requesting preallocated devices (USB, eMMC) to avoid clock/reset permission failures. A new XPm_GetVidPowerState API lets clients read the PLM-selected VID rail mode, and AIE array column partitioning now supports multiple device nodes owning distinct column ranges. This release also delivers extensive hardening and bug fixes across CDO command handling, DDRMC validation, clock and power-domain operations, and BISR repair, addressing numerous out-of-bounds accesses, NULL-pointer dereferences, and false errors that could affect runtime stability.

Features

  • Adds UCIe AP_MAIN and SP_MAIN scan-clear and mem-clear pre-config sequences on Versal.
  • Adds CPM6 and GTMPW BISR repair support during housecleaning on Versal.

Enhancements

  • Re-requests preallocated devices (USB, eMMC) during Linux kexec to avoid clock/reset permission failures.
  • Adds bulk preallocated device re-request support for Linux kexec in xilpm_ng.
  • Makes CPM6 handling device-agnostic via a single feature macro and runtime node detection.
  • Adds XPm_GetVidPowerState client API to read the PLM-selected VID rail mode.
  • Adds device-specific BISR eFuse cache offsets for xc2vp3202.
  • Adds housecleaning skip support for UCIe pre-configuration on Versal.
  • Adds housecleaning skip checks for BFR-B BISR, MBIST, and scan clear operations on Versal.
  • Adds AIE array column partition bounds enforcement, enabling multiple device nodes to own distinct column ranges.
  • Adds NoC scan clear and DDRMC crypto housecleaning workaround support for Versal.
  • Allows boards to bypass PL scan clear completion check while preserving the trigger sequence.
  • Extends NPD secure efuse transfer CPM6 SSC bypass workaround to the XC2VP3202 device.
  • Adds UCIE node type, index, and PM device ID definitions.
  • Removes XSysMonPsv driver dependency from PLM using direct register access to reduce binary size.
  • Moves memory address validation security hardening responsibility to callers to reduce code size.

Bug Fixes

  • Fixes spurious APU/RPU BISR triggering on some boards by setting BISR bitmask only on actual repair.
  • Corrects DDRMC lookup for DFX reconfigurable partitions, preventing use of wrong memory controller device.
  • Handles PLL and output clocks without a power domain, avoiding NULL pointer access.
  • Validates GIC proxy wake routing and supports eight groups on Versal Net to prevent register overruns.
  • Validates power-domain parent types and allocations to prevent out-of-bounds access from malformed nodes.
  • Rejects abort-suspend requests for cores not owned by the calling subsystem.
  • Validates DDRMC region count before use to prevent corruption from out-of-range values.
  • Fixes exclusive/time-shared device sharing scan that silently skipped later-registered subsystems.
  • Rebuilds DDRMC validation list after in-place PLM update so DDR addresses stay valid.
  • Skips zero-size DDRMC regions to prevent invalid memory-access validation.
  • Fixes out-of-bounds read in XPm_DevIoctl2 for short payload buffers.
  • Corrects HBMon software error mask mapping so recovery actions target the right monitors.
  • Zero-initializes PSM IPI payloads to avoid leaking stale stack data into IPI buffers.
  • Zero-initializes Versal PSM IPI payloads to prevent stale stack data in IPI buffers.
  • Validates isolation record operand lengths to prevent reads past the CDO buffer.
  • Clears spurious NOC2 XBR4X2 scan-clear errors during boot to prevent false EAM errors.
  • Rejects out-of-range index in clock mux-source query, closing an arbitrary PLM memory read.
  • Guards against NULL subsystem and power node during Versal Net direct power-down.
  • Rejects non-output clocks in GetMaxDivisor to prevent memory corruption from invalid clock IDs.
  • Bounds I2C mode command payload length to prevent buffer overflow from crafted CDO commands.
  • Validates PSM-supplied event address against PSM RAM window before writing to it.
  • Rejects out-of-range index in pin group query, closing an arbitrary PLM memory read.
  • Adds bounds check on RPU cpuid in pm_get_master to prevent out-of-bounds table access.
  • Retains NULL terminator when setting RPU processor name to prevent buffer over-read.
  • Validates reset type before indexing reset ops, preventing out-of-bounds call from crafted CDO.
  • Validates power domain node subclass in ForceHouseClean to prevent out-of-bounds indirect calls.
  • Fixes AIE OPS column range validation and restores legacy AIE node compatibility on Versal Gen2.
  • Avoids false NPI PCSR write failures on QEMU/COSIM by skipping read-back verification.
  • Prevents memory corruption from oversized regulator I2C payloads in PM_ADD_NODE commands.
  • Rejects out-of-range Function values to prevent out-of-bounds access during power domain init.
  • Prevents out-of-bounds reads from underflowed base-address count in PM_ADD_NODE power commands.
  • Validates PMBus IC_DEVICE_ID length to prevent stack buffer overflow during regulator identification.
  • Validates PL device node type in XPm_NocClockEnable to prevent memory corruption.
  • Bounds FPx BISR repair tag size to prevent out-of-bounds eFuse-driven writes.
  • Validates power node type before AIE domain downcast to prevent memory corruption.
  • Bounds Laguna repair frame loop to prevent stack corruption from malformed eFuse data.
  • Fixes BISR HardBlock/Laguna stub return types so repair failures are correctly detected.
  • Rejects truncated PM_INIT_NODE payloads to prevent out-of-bounds PMC RAM reads on Versal.
  • Rejects truncated PM_INIT_NODE payloads to prevent out-of-bounds PMC RAM reads on Versal Net.
  • Validates isolation dependency count to prevent PM node corruption from crafted PM_ADD_NODE commands.
  • Reads FPD secure base address only when supplied, avoiding out-of-bounds payload read on Versal.
  • Reads FPD secure base address only when supplied, avoiding out-of-bounds payload read on Versal Net.
  • Removes unused UCIE_AP_SHIM device nodes.
  • Fixes PLM build failures for Versal_2VP_P by initializing module UpdateHandler fields.
  • Prevents secure-lockdown hangs on XC2VP3602 by skipping problematic GTMPW/CPM6 houseclean steps.
  • Prevents boot errors on HW_EMU/COSIM designs lacking an AIE power domain.
  • Corrects PCIe reset node ID ordering on Versal Net so IDs match indices.
  • Adds PCIe resets to the permitted global reset list, fixing PCIe core reset failures.
  • Ignores false scan_mem_clear PASS on XCVR1602/XCVR1652 SDFEC columns due to hold timing violations.
  • Fixes uninitialized pointer read when an unexpected RPU ClusterId is encountered in XPm_SetPrimaryProc.
  • Fixes RPU TCM boot failure after a prior boot from DDR by always configuring the vector table boot address.
  • Enables DDR address validation on SSIT slave SLRs for security-critical NVM, eFuse, and PUF IPI commands.
  • Re-applies RPU core BISR SRAM repair data on hot-swap power-up to prevent corruption.
  • Re-applies APU core BISR SRAM repair data on hot-swap power-up to prevent corruption.

xilpm_ng_v1_1:

This release expands power-management support on Versal Gen2 with FPD power domain power-up/power-down (including a new option to power down FPD after suspend-to-RAM), DDR EDAC error tracing, VID-adjusted rail power-state queries, and bulk device re-request handling for Linux kexec. It resolves several reliability issues affecting Linux warm reboots and APU/RPU restart, including repeated power interrupts and reboot hangs on specific APU cores, FPD powering off during subsystem restart, GIC reset handling, and RPU TCM boot failures, and adds UART wake-from-suspend support. Numerous hardening fixes bound CDO command and query payloads to prevent out-of-bounds reads and PM node corruption, and correct PCIe reset handling on Versal Net.

Features

  • Adds XPM_QID_GET_VID_POWER_STATE query to report the VID-adjusted rail power state.
  • Adds DDR EDAC error-source tracing so PLM can report per-controller error snapshots for secure address decode.
  • Adds FPD power domain power-up/power-down support with GIC proxy wake arming and CMN flush.
  • Adds PMC-RAM memory-range reachability table letting runtime consumers validate region access.

Enhancements

  • Adds XILPM_NG_FPD_OFF option to power down FPD after suspend-to-RAM on Versal Gen2.
  • Adds bulk preallocated device re-request support for Linux kexec in xilpm_ng.
  • Unifies PLM address-range validation under a single configurable option, off by default.
  • Rebalances memory pools to support additional DDR regions on Versal Gen2.
  • Uses the topology-supplied GIC proxy wakeup mask for RTC instead of a hardcoded override.
  • Manages AMS SSC bypass and FPD resets correctly across FPD power cycle during suspend.
  • Hardens AIE zeroization by confirming the trigger register was accepted before polling for completion.
  • Adds AIE array column partition bounds enforcement, enabling multiple device nodes to own distinct column ranges.
  • Adds bulk release of all peripheral devices owned by a subsystem in one request.
  • Adds ability to unregister all of a subsystem's event notifiers in one request.
  • Adds UCIE node type, index, and PM device ID definitions.
  • Removes XSysMonPsv driver dependency from PLM using direct register access to reduce binary size.
  • Moves memory-address validation security hardening to callers for improved code size and redundancy control.

Bug Fixes

  • Fixes repeated power interrupts and Linux warm reboot hang on APU cluster 0 cores 2/3 (Versal 2VE/2VM).
  • Fixes APU GIC reset on subsystem restart so Linux reboots cleanly on any APU core.
  • Keeps FPD powered on across forced subsystem power-down, preventing PLM hang during APU restart.
  • Validates power-domain parent types and allocations to prevent out-of-bounds access from malformed nodes.
  • PMC requests the domain-control GPIO so FPD/PLD rail switching keeps working after subsystems release it.
  • Isolates FPD-MMI interface before powering down FPD on Versal Gen2.
  • Rejects abort-suspend requests for cores not owned by the calling subsystem.
  • Registers UART GIC proxy wake handlers so UART interrupts can wake the APU from suspend.
  • Validates DDRMC region count and command length to prevent out-of-bounds reads and stale links.
  • Fixes HBMon software error mask so monitors 4-7 map to correct error table slots.
  • Validates isolation record operand lengths to prevent reads beyond the CDO command buffer.
  • Clears spurious NOC2 XBR4X2 scan-clear errors during boot to prevent false EAM errors.
  • Prevents FPD from powering off during subsystem restart on Versal Gen2.
  • Bounds mux-source query Index to prevent unauthenticated arbitrary-offset reads of PLM memory.
  • Rejects non-output clocks in GetMaxDivisor to avoid invalid topology member access.
  • Bounds CDO-supplied I2C mode command payload to prevent memory corruption of adjacent nodes.
  • Bounds pin-group query Index to prevent unauthenticated relative reads of PLM memory.
  • Keeps SoC-critical clocks like NPI_REF enabled so device release no longer stalls NPI access.
  • Fixes AIE OPS column range validation and restores legacy AIE node compatibility on Versal Gen2.
  • Rejects PM_INIT_NODE commands with truncated payloads to prevent out-of-bounds PMC RAM reads.
  • Validates regulator I2C command length to prevent overwriting adjacent PM node data.
  • Bounds PM_ADD_NODE power base-address count to prevent out-of-bounds payload reads.
  • Rejects out-of-range isolation dependency counts to prevent PM node corruption.
  • Increases Versal Gen2 topology pool capacity to prevent DDR mapping allocation failures.
  • Fixes out-of-bounds read of FPD secure base address when topology supplies a single entry.
  • Removes unused UCIE_AP_SHIM device nodes.
  • Fixes address validation denying memory access for built-in PMC and ASU subsystems.
  • Prevents shared cluster reset from disrupting a running sibling RPU core during split-mode TCM boot.
  • Gates PM_GET_EVENT_STATUS on link-nodes support to fix builds on leaner Versal Gen2 backends.
  • Fixes boot-only build failure by guarding runtime-only GIC proxy include in FPD power cycle code.
  • Corrects PCIe reset node ID ordering on Versal Net so IDs match indices.
  • Adds PCIe resets to the permitted global reset list, fixing PCIe core reset failures.
  • Restores external linkage of XilSEM callback registration API for SEM-enabled PLM builds.
  • Restores original callback signatures to fix incompatible function-pointer defects.
  • Fixes GPIO handler not firing when LPD MIO is used as PCIe reset source.
  • Fixes RPU TCM boot failure after a prior boot from DDR by always configuring the vector table boot address.

xilpuf_v2_8:

XilPuf adds a new client example for Versal 2VP_P supporting PUF registration/regeneration and red key encryption, along with new client APIs to query PLM library version and feature capabilities and support for the MicroBlaze RISC-V soft core on Versal_2VE_2VM. This release also enables the D-cache by default for client examples with proper cache maintenance on IPI buffers. Key fixes include validating the PUF operation during registration to close an IPI access-rule bypass, preventing a NULL pointer dereference in the server, correcting DME key index 3 eFuse programming on Versal Net, and resolving Spartan boot failures caused by an omitted GVF bit in the shutter value.

Features

  • Adds new XilPuf client example for Versal 2VP_P supporting PUF registration/regeneration and red key encryption.

Enhancements

  • Enables D-cache by default for XilPuf client examples with proper cache maintenance on IPI buffers.
  • Adds XilPuf client APIs to query PLM library version and feature capability bitmap.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.
  • Add input validation and redundancy hardening across LMS/HSS, ECC keypair, and PUF paths.

Bug Fixes

  • Validates PufOperation during registration to prevent bypassing access-rule checks via IPI.
  • Maps mailbox shared memory into a separate section in the XilPuf enc/dec example.
  • Validates command pointer before use in XilPuf server to prevent NULL dereference.
  • Fixes DME key index 3 eFuse programming in Versal Net XilPuf example.
  • Includes GVF bit in printed shutter value to avoid boot failures in Spartan XilPuf examples.
  • Fixes pointer aliasing in 64-bit memory copy by splitting into direction-specific APIs.
  • Adds NULL and length validation to XilPuf client communication APIs.

xilsecure_v5_8:

XilSecure adds an RSA PSS client-server interface for signature verification and introduces SHAKE256 variable-length XOF hashing on Versal_2VP_P and Versal_2VE_2VM, along with new queries for PLM library version, crypto capability, and NIST-compliance information. This release also hardens security across LMS/HSS, ECC, ML-DSA, and PUF paths against glitch and forgery attacks, and adds broader input validation throughout the client and server APIs, with MicroBlaze RISC-V now supported as a client target on Versal_2VE_2VM. Key fixes resolve encrypted boot failures with eFUSE black keys, OAEP decode buffer overflows, AES GMAC byte-swap and zeroization issues and several NULL-pointer and uninitialized-data defects.

Features

  • Adds RSA PSS client-server interface for signature verification
  • XilSecure clients can query PLM server XilSecure library version and compiled-in crypto capability bitmap.

Enhancements

  • Adds input parameter validation across XilSecure client and server APIs
  • Deprecate XSecure_AesKekDecrypt client API; it now returns an error and is removed on Versal_2VP.
  • Adds MicroBlaze RISC-V soft core as a valid client target on Versal_2VE_2VM.
  • SHA IPI handler now routes SHAKE256 finish through variable-length XOF output.
  • Add input validation and redundancy hardening across LMS/HSS, ECC keypair, and PUF paths.
  • Allow arbitrary-length (unaligned) AES AAD on Versal_2VP_P.
  • Add crypto algorithm version and NIST compliance query support on Versal_2VP_P.
  • Hardens LMS signature verification against glitch-induced early-exit attacks during Merkle tree traversal.

Bug Fixes

  • Validates AES header lengths against image size and limits failure zeroization to decrypted bytes
  • Fixes SHA IPI rejecting valid zero-size buffer flows under address range validation
  • Fixes ML-DSA zeroize access while core is held in reset
  • Zeroizes derived HKDF key left in stack memory after function exit
  • Fixes duplicate AES key table symbol link error
  • Resets SHA engine when chain-mode start fails after release
  • Avoids unneeded AES soft reset on KEK decrypt parameter errors and validates key source
  • Fixes ML-DSA verification forgery risk by snapshotting signature commitment
  • Adds missing NULL checks before pointer dereferences in IPI handlers and ECDSA path
  • Prevents stale KAT pass status when mask update is skipped
  • Elliptic key generation no longer returns uninitialized stack data on public key failure.
  • AES GMAC AAD state no longer set prematurely when DMA transfer fails.
  • Fix OAEP decode buffer overflow by validating decoded message against caller buffer size in key unwrap.
  • Fix encrypted boot failure with eFUSE black key by correcting KEK IV byte-swap handling.
  • AES key zeroize no longer masks failed key write or GCM tag mismatch as success.
  • Fix AES GMAC DMA byte-swap for Versal_2VP, Versal_2VP_P, and Versal_2VE_2VM parts.
  • Fix const output buffers in client examples that are written by PLM over IPI.
  • SHA3 client validates Size to prevent aliasing control flags and resetting the SHA engine mid-stream.
  • Fix ML-DSA signature mismatch by using correct memory copy for register aperture instead of PMC DMA.
  • XSecure_HmacFinal now returns correct error code on parameter validation failure.
  • Fix memory aliasing issue by using address-specific copy APIs across XilSecure server.
  • Fix key unwrap example failure by adding proper cache invalidate and flush.

Notes

  • Removes deprecated AES KEK decrypt server IPI handler. Customers shall avoid using the API XSecure_AesKekDecrypt().

xilsem_v1_14:

  • Added client support for SMC call
  • Added support to capture SW ECC calibration error details
  • Added support to restore NPI descriptor golden SHA after error injection.
  • Updated XilSEM library for GCC 15.3.0 tool chain

Enhancements

  • XilSEM captures detailed SW ECC calibration error locations and restores NPI golden SHA after error injection.
  • Adds client support and examples to decode SW ECC calibration error details for multiple ECC/CRC locations.
  • Adds XilPM SGI event notification validation to XilSEM CRAM and NPI examples for EL1 non-secure SMC flow.

Bug Fixes

  • Fixes NPI golden SHA restoration after error injection so injected SHA errors clear correctly.

xilsfl_v1_4:

The xilsfl library now supports Advanced Sector Protection (ASP) for sector lock and password management, and adds support for several new QSPI and OSPI flash devices, including Winbond W25Q/W25H 1.8V, Macronix MX25U6432F and OSPI parts, and GigaDevice GD25LR512MF. This release also fixes flash reset on Versal NET by using GPIO reset, corrects stale message-field handling that caused unpredictable read and OSPI transfer behavior, and resolves DDR-PHY protocol selection and byte-count issues for register transfers.

Features

  • Adds Advanced Sector Protection (ASP) support to xilsfl for sector lock and password management.

Enhancements

  • Adds support for Winbond W25Q/W25H series 1.8V QSPI NOR flash devices to xilsfl.
  • Adds support for Macronix MX25U6432F 64Mb QSPI flash to xilsfl.
  • Adds support for GigaDevice GD25LR512MF 512Mb QSPI flash to xilsfl.
  • Adds support for Macronix mx66um2g45g and mx66uw2g345g OSPI flash devices to xilsfl.

Bug Fixes

  • Fixes flash reset on Versal NET by using GPIO reset so the flash returns to default state
  • Fixes OSPI example to erase all sectors covered by the write, preventing data check failures.
  • Fixes stale/uninitialized flash message fields causing unpredictable read and OSPI transfer behavior.
  • Corrects ASP password-length validation and distinguishes unsupported devices from malformed passwords.
  • Fixes DDR-PHY protocol selection and byte-count handling for flash register read/write transfers.

xilskey_v8_0:

Bug Fixes

  • Fixes XilSKey CRC calculation underflow when key string is empty by validating length.
  • Corrects obfuscation status check in XilSKey eFUSE example to use proper status bits.

xiltimer_v2_5:

XilTimer adds a new frequency-based tick interval and the XTimer_SetIntervalHz() API for AXI, SCU, and TTC timers, improving timer rate accuracy by avoiding millisecond truncation in the FreeRTOS SDT flow, and enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors. This release also fixes MicroBlaze FreeRTOS tick setup to prevent interrupts from firing before the ISR stack is initialized.

Enhancements

  • Adds frequency-based tick interval for AXI and SCU timers to avoid millisecond truncation in FreeRTOS SDT flow
  • Adds XTimer_SetIntervalHz() API and TTC callback to program timer rates accurately without millisecond truncation
  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.

Bug Fixes

  • Fixes MicroBlaze FreeRTOS tick setup to prevent interrupts before ISR stack is initialized

xiltpm_v1_3:

The xiltpm library now supports TPM on Versal_2VP, configurable through BSP settings. This release also includes several reliability fixes, including corrected TPM response length validation to prevent PCR read buffer issues, resolved PCR buffer overflow/underflow, added polling timeouts to avoid PLM stalls, and fixed data loss when receiving responses larger than 255 bytes.

Features

  • Adds TPM support for Versal_2VP, configurable via BSP settings.

Bug Fixes

  • Corrects TPM response length validation in PCR reads and transfers to prevent buffer issues.
  • Fixes PCR response buffer overflow/underflow and adds timeouts to TPM polling to prevent PLM stalls.
  • Fixes memory aliasing issue in TPM by using address-specific copy APIs.
  • Fixes data loss when receiving TPM responses larger than 255 bytes.

asufw:

Bug Fixes

  • Reject key vault allocation outside lower DDR
  • Fix key manager subsystem validation issues
  • Fix code spell, doxygen, Coverity, MISRA and code complexity violations
  • Generate cryptographic root material on demand instead of retaining it from boot
  • Add error code when wrapped key is used with keywrap disabled
  • Clear keys on AES operation failure
  • Implement epoch based key expiry in key manager
  • Add resource availability check to RSA key pair generation task
  • Add overlay support to load selected crypto code and data from DDR into IMEM and DMEM at runtime
  • Add periodic keep-alive counter for PLM to monitor ASUFW health
  • Store device key pairs to ASU key vault instead of RAM
  • Reject cross-module key IDs when resolving keys from vault
  • Add ASUFW firmware version and runtime feature capability reporting for crypto modules
  • Enhance RSA and AES operations in key manager for public exponent and curve attributes
  • Add basic constraints extension field in X509 CSR generation
  • Add USER eFUSE key provisioning support
  • Add support to generate user X509 certificate from vault held keys
  • Require vault key to hold every requested use case
  • Add support for extended hash output of SHAKE256
  • Reject in-place ASU update when crypto operations are active
  • Preserve pending PLM events across in-place ASU update
  • Add subsystem based user address range validation for client buffers
  • Export versioned data structure metadata for in-place ASU update
  • Zeroize data remnants after authentication and decryption operations
  • Add user-addressable key vaults and key slots
  • Add PQC foundation for memory, overlay and module info
  • Add ML-DSA and ML-KEM firmware support
  • Add ML-DSA and ML-KEM key vault support
  • Update module versions and add client compatibility check
  • Add Key-XOR support for modified key generation in key manager
  • Add PL extension driver to stream data to crypto blocks in PL
  • Introduce SLH-DSA crypto module with keygen, sign and verify support
  • Protect UDE key using PUF KEK
  • Add UDE provisioning support
  • Skip key size validation for vault stored keys in AES, ECC, RSA and KeyWrap APIs
  • Fix AES context switch save and restore across modes
  • Add ASU exportable vault and ECC key pair generation for provisioning
  • Fix ECDSA ephemeral nonce handling for Brainpool curves
  • Omit parameters field of ECDSA signature AlgorithmIdentifier in X509 generation
  • Fix PL extension encrypt returning plaintext
  • Add authenticated encrypted key vault export and import support
  • Fix client buffer validation for vault sourced keys and IVs
  • Reject non-zero length for vault sourced key or IV in client interface
  • Use RSA core ECC APIs for OCP, provisioning and vault export or import paths
  • Separate RSA and ECC overlays
  • Read actual ASU clock frequency for PIT and performance timers
  • Remove SCA leakage from ECC endianness byte swaps
  • Enable AES operations using eFuse User Key 1
  • Fix incorrect serial number generation in X509 certificate generation
  • Add Device Provisioning Certificate support
  • Reject debug log buffer placed in PMC or PPU RAM
  • Zeroize sensitive data on KeyManager RSA and ECC keygen failure
  • Add authenticated encrypted provisioning support
  • Bound requested KDF output length to prevent wraparound success
  • Relocate default overlay DDR base to 0x7FF00000
  • Fix OCP UDE response handling to avoid dereferencing unvalidated client response address
  • Make xasufw hardware definitions platform specific for Versal_2VE_2VM
  • Add cooperative yield for SLH-DSA keygen and sign
  • Add missing redundancy and input validation checks in PL extension, TRNG, X509 and LMS paths
  • Add build configuration for RSA and ECC engines
  • Store ECDH shared secret into key vault when shared secret address is zero
  • Return key lookup errors instead of a NULL pointer and enforce key expiry on lookup
  • Fix queue stall behind a request waiting for DMA completion
  • Fix HMAC state after failure and validate OCP certificate buffers
  • Advance keep-alive counter during long hardware waits
  • Resume SLH-DSA sign and verify across SHA message DMA
  • Add STL framework support for ASUFW
  • Allow SLH-DSA sign and key pair generation from all IPI channels
  • Unify key manager validation and add length parameters where needed
  • Support red and black USER eFUSE key 1
  • Fix hang on unaligned intermediate AES update

lwip_echo_server

Bug Fixes

  • Adds periodic Zynq-7000 GEM Rx-lockup recovery for lwIP applications in the SDT flow.

lwip_tcp_perf_client

Bug Fixes

  • Adds periodic Zynq-7000 GEM Rx-lockup recovery for lwIP applications in the SDT flow.

lwip_tcp_perf_server

Bug Fixes

  • Adds periodic Zynq-7000 GEM Rx-lockup recovery for lwIP applications in the SDT flow.

lwip_udp_perf_client

Bug Fixes

  • Adds periodic Zynq-7000 GEM Rx-lockup recovery for lwIP applications in the SDT flow.

lwip_udp_perf_server

Bug Fixes

  • Adds periodic Zynq-7000 GEM Rx-lockup recovery for lwIP applications in the SDT flow.

memory_tests

Bug Fixes

  • Fixes memory test on Versal PL soft MicroBlaze/RISC-V CPUs to avoid 32-bit DDR address truncation.

openamp_sdt

This release fixes OpenAMP demo memory carveouts so Versal and ZynqMP FreeRTOS demos build correctly and Linux boots, and aligns RPU0 carveouts with the Yocto device tree so remoteproc can load the firmware ELF. It also prevents Linux from claiming the TTC timers reserved for the RPU0 demo on Versal and corrects RPU domain CPU masks and remote timer selections. In addition, OpenAMP default domain YAMLs are now resolved from the template repo, reusing existing BSP YAMLs.

Enhancements

  • Resolves OpenAMP default domain YAMLs from the template repo and reuses existing BSP YAMLs.

Bug Fixes

  • Fixes OpenAMP demo memory carveouts so Versal/ZynqMP FreeRTOS demos build and Linux boots.
  • Prevents Linux from claiming TTC0/TTC3 timers reserved for the RPU0 OpenAMP demo on Versal.
  • Aligns OpenAMP RPU0 carveouts with Yocto device tree so remoteproc loads the firmware ELF.
  • Corrects RPU domain CPU masks and remote timer selections for OpenAMP demos.

srec_spi_bootloader

Bug Fixes

  • Fixes compile failure in SREC SPI bootloader when building for PowerPC targets.

versal_plm

This release expands security and platform services with a new RSA PSS signature-verification interface, OCP and certificate services (DevIK/DevAK, attestation, and PCR measurement) for Versal 2VP_P, and new IPI commands to configure glitch detectors and manage subsystem image restart. Enhancements include default subsystem runtime support in the 2VE/2VM build flow, PMC watchdog with reset action on 2VP_P, compatibility validation for in-place ASU firmware updates, and reduced binary size by removing the XSysMonPsv dependency. Key fixes address a buffer overflow in secure SSIT communication, DDR mapping allocation failures on Versal Gen2, PLM link and infinite-loop issues, and improved validation of update optional data to prevent out-of-range reads.

Features

  • Adds RSA PSS client-server interface for signature verification
  • Adds OCP and certificate services (DevIK/DevAK, attestation, PCR measurement) for Versal_2VP_P.
  • Adds IPI command and example to configure PMC, ASU, and LPD glitch detectors.
  • Adds PMC-RAM memory-range reachability table letting runtime consumers validate region access.
  • Adds IPI command to configure a DDR address for subsystem image restart.

Enhancements

  • Enables subsystem runtime support by default in the 2VE/2VM make build flow.
  • Removes PLM-to-ASU key transfer API on Versal 2VP_P
  • Removes deprecated AES KEK decrypt server IPI handler
  • Rebalances memory pools to support additional DDR regions on Versal Gen2.
  • Decouples TPM init via callback to break circular dependency between xilplmi and xiltpm.
  • Adds get_tl_bin_load_addr IPI command support on Versal 2VE/2VM PLM.
  • Adds compatibility validation for in-place ASU firmware updates before image load.
  • Enables PMC watchdog with reset action on versal_2vp_p for watchdog-triggered device reset.
  • Resumes ASUFW keep-alive monitoring after in-place PLM update.
  • Removes XSysMonPsv driver dependency from PLM using direct register access to reduce binary size.

Bug Fixes

  • Fixes buffer overflow in secure SSIT communication by bounding AES-GCM message length on slave SLR.
  • Validates update optional data to prevent address wrap and out-of-range reads.
  • Runs Configuration Limiter lockdown before boot PDI load so lockdown completes earlier.
  • Increases Versal Gen2 topology pool capacity to prevent DDR mapping allocation failures.
  • Prevents XilPM pool overlap in Versal 2VE/2VM PLM builds with TPM by optimizing BSS layout.
  • Fixes PLM link failure for memory-table init on versal_gen2_tr build.
  • Fixes PLM infinite loop when compatibility check encounters zero-length optional-data entry.

versal_psmfw

This release improves Versal Net PSM firmware reliability by correcting BISR SRAM repair handling, including re-applying APU and RPU core repair data on hot-swap power-up to prevent corruption and fixing Core 3 repair through its power dependency on Core 2. It also hardens the firmware against integer overflow and out-of-bounds reads by validating proc buffer bounds and CDO command lengths, and now always displays BISR repair failure messages on the console for easier field diagnosis.

Enhancements

  • Always display BISR repair failure messages on console for easier field diagnosis on Versal Net.

Bug Fixes

  • Prevent integer overflow in PSM firmware proc buffer bounds check on Versal Net.
  • Validate CDO command length to prevent out-of-bounds reads in Versal Net PSM firmware.
  • Re-applies RPU core BISR SRAM repair data on hot-swap power-up to prevent corruption.
  • Fix Core 3 BISR repair by handling its power dependency on sibling Core 2 on Versal Net.
  • Re-applies APU core BISR SRAM repair data on hot-swap power-up to prevent corruption.

vid_voltage_monitor

Features

  • New VEK385 bare-metal RPU app for external rail voltage monitoring via VID power states and MPQ79500 monitor.

zynq_fsbl

The Zynq FSBL now performs stronger validation during boot, rejecting invalid image types, out-of-range load addresses, and oversized partition or RSA sizes to prevent memory overflows. It also aborts on SD and QSPI read errors or short reads, preventing the parsing of stale or out-of-bounds boot image data.

Bug Fixes

  • Zynq FSBL now aborts on SD read errors or short reads, preventing parsing of stale boot image data.
  • Zynq FSBL rejects partitions with invalid image types and validates load addresses for all non-PL partitions.
  • Zynq FSBL validates partition lengths, DDR bounds, and RSA sizes to prevent overflow during boot image transfers.
  • Zynq FSBL prevents QSPI read buffer overflows by validating flash read ranges and propagating transfer failures.

zynqmp_dram_test

Enhancements

  • Adds ECC byte-lane eye measurement support to ZynqMP DRAM test for 2/4/8-lane DDR configurations.

zynqmp_fsbl

The ZynqMP FSBL now tracks per-core APU usage so PMUFW can correctly identify cores without loaded partitions, and only builds the IIC driver when an IIC instance is present. This release fixes several security and reliability issues, including integer overflow and out-of-bounds access in USB DFU handling, TPM response buffer overflows and data loss with responses larger than 255 bytes, and a lockstep-mode issue that could cause PMUFW to power off an in-use RPU core.

Enhancements

  • Tracks per-core APU usage in FSBL so PMUFW can identify cores without loaded partitions.
  • Removes hard IIC driver dependency in FSBL board config, building it only when an IIC instance is present.

Bug Fixes

  • Fixes integer overflow in ZynqMP FSBL USB DFU bounds check allowing out-of-range addresses.
  • Prevents TPM response buffer overflow in ZynqMP FSBL by validating response lengths against buffer capacity.
  • Fixes lockstep mode so both RPU cores are marked in-use, preventing PMUFW from powering one off.
  • Fixes USB DFU_GETSTATUS memory leak by clamping reply length to buffer size in ZynqMP FSBL.
  • Fixes out-of-bounds read in ZynqMP FSBL DDR HIF address calculation.
  • Ensures SYSMON tamper alarms are restored on all FSBL protection-config error paths.
  • Fixes data loss when receiving TPM responses larger than 255 bytes.

zynqmp_pmufw

The PMU firmware now supports automatically starting the external watchdog at boot and applying interval changes at runtime, plus an optional build-time feature to power down unused APUs and the FPD after FSBL handoff. This release also fixes suspend/restart features being disabled in SDT builds, corrects boot-type detection via MIO26 sampling, and hardens the secure AES handler against concurrent shared-memory tampering, along with several reset-handler and register-permission fixes.

Enhancements

  • Adds optional build-time feature to power down unused APUs and FPD after FSBL handoff.
  • Starts external watchdog kick automatically at boot in runtime mode to prevent premature expiry.

Bug Fixes

  • Fixes PMUFW I2C and GPIO reset handlers using incompatible function pointer types.
  • Fixes suspend/restart features staying disabled in SDT builds by ordering DDR/QSPI device aliases correctly.
  • Corrects argument validation in PMUFW OCM error injection hook to use valid OCM index.
  • Fixes boot-type detection by correctly asserting GPIO reset after sampling MIO26.
  • Hardens PM_SECURE_AES handler against TOCTOU attacks from concurrent shared-memory modification.
  • Corrects PMU_GLOBAL_PWR_STATE register access permissions to read-only.
  • Applies external watchdog kick interval changes at runtime and reports failures to caller.

xilasu

This release significantly expands post-quantum and secure provisioning capabilities, adding SLH-DSA, ML-DSA, and ML-KEM signature and key-encapsulation APIs with key-vault storage, along with authenticated encrypted provisioning (AEPI), Device Provisioning Certificate (DPC) support, and UDE/USER eFUSE AES key provisioning with PUF-KEK encryption. It introduces a PL extension driver for streaming data to custom crypto blocks in the PL fabric, enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors, and adds runtime reporting of firmware version and per-module feature capabilities. Numerous fixes improve runtime reliability, including resolving an AES hang on unaligned AAD/payload updates, hardening Key Manager and client input validation, correcting context-free and buffer-validation issues, and ensuring examples properly propagate server-side error status.

Features

  • Adds authenticated encrypted provisioning (AEPI) for secure wrapping and execution of provisioning records.
  • Adds Device Provisioning Certificate (DPC) support.
  • Adds UDE key provisioning: generate, PUF-KEK encrypt, program to eFUSEs, with self-signed certificate support.
  • Adds SLH-DSA post-quantum signature support with keygen, sign, and verify APIs.
  • Adds PL extension driver to stream data to custom crypto blocks in the PL fabric.
  • Adds ML-DSA and ML-KEM key vault storage and generation support.
  • Adds client ML-DSA and ML-KEM post-quantum crypto APIs and examples.
  • Adds USER eFUSE AES key provisioning with multiple key sources and PUF-KEK encryption.

Enhancements

  • Adds support for red and black forms of USER eFUSE key 1 in AES provisioning.
  • Unifies Key Manager API validation and adds explicit length parameters for client inputs.
  • Allows SLH-DSA sign and key-pair generation from all IPI channels, not just secure ones.
  • Adds STL (Software Test Library) framework support to ASU firmware.
  • Enables SLH-DSA sign/verify to resume across large SHA message DMA transfers.
  • Adds key-vault storage of the ECDH shared secret when no output address is given.
  • Adds build options to enable or disable the RSA and ECC engines for devices lacking them.
  • Adds cooperative yielding during long SLH-DSA keygen/sign and pairwise consistency testing.
  • Moves provisioning key-pair generation/deletion from Key Manager to the provision module and removes client support.
  • Returns UDE certificate size via the response buffer instead of a DMA copy.
  • Enables AES operations using eFuse User Key 1 via the vault key path.
  • Adds a client interface to read key-vault export result size and modified flag.
  • Adds ASU-owned exportable vault and ECC key-pair generation for wrapped-key provisioning.
  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.
  • Adds key-XOR support to combine two AES key splits into a vault slot.
  • Adds ASUFW firmware version compatibility check during client initialization.
  • Adds memory and module-info foundation for ML-DSA/ML-KEM post-quantum support.
  • Allows applications to select key vault ID and key slot when storing or generating keys.
  • Adds SHAKE256 extendable-output (XOF) support for arbitrary-length hash output.
  • Adds example for issuing user X.509 certificates by key ID without exposing private keys.
  • Adds OCP command to generate user X.509 certificates from vaulted ECC keys.
  • Refines AES/RSA/ECC key object structures; adds public exponent and curve type fields.
  • Exposes ASUFW firmware version and per-module feature capabilities at runtime.
  • Adds RTCA keep-alive counter address so PLM can verify ASUFW is running.
  • Stores device identity/attestation key pairs in the ASU key vault instead of RAM.
  • Adds code/data overlay support to reduce ASUFW memory footprint; enables LMS by default.
  • Adds epoch-based key expiry so expired keys are rejected by the key manager.

Bug Fixes

  • Add validation checks to reject Edwards curve in ECIES operations
  • Fix ECIES curve validation check in CMakeLists
  • Minor enhancements to xilasu
  • Fix code spell, doxygen, Coverity, MISRA and code complexity violations
  • Fix build failure when ECC P448 or P25519 is enabled with HMAC disabled and ECIES auto-enabled
  • Implement epoch based key expiry in key manager
  • Propagate server error status in client examples
  • Add RTCA keep-alive counter address in client interface
  • Add ASUFW firmware version and runtime feature capability reporting for crypto modules
  • Update client interface for AES and RSA key manager enhancements
  • Add example for KeyManager module
  • Add client interface updates for USER eFUSE key provisioning and eFuse User Key 1
  • Add example for user X509 certificate generation
  • Add user-addressable key vaults and key slots
  • Fix RSA keygen size selection to be mutually exclusive
  • Add support for extended hash output of SHAKE256
  • Add client interface and examples for ML-DSA and ML-KEM
  • Add liboqs headers for ML-DSA and ML-KEM
  • Update module versions and add client compatibility check
  • Add client interface for Key-XOR based key generation
  • Add client interface for SLH-DSA operations
  • Add support for xilasu client on PL MicroBlaze and MicroBlaze V
  • Add client interface updates for UDE provisioning and Device Provisioning Certificate
  • Skip key size validation for vault stored keys in AES, ECC, RSA and KeyWrap APIs
  • Add client interface for key vault export result
  • Reject non-zero length for vault sourced key or IV in client interface
  • Enable ECC P-384 curve when OCP is enabled
  • Free client context only when it belongs to the request
  • Add authenticated encrypted provisioning support
  • Harden key manager and client input validation for FMEA findings
  • Unify key manager validation and add length parameters where needed

iicps_v3_24:

  • Refactor high-complexity APIs into helper functions
  • Restore HOLD release order and remove redundant Error check
  • Initialize Is10BitAddr in XIicPs_CfgInitialize

pyesw

pyesw now supports uC/OS-III BSPs and applications in the SDT workflow, and enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors. BSP builds use a shorter, configurable output path to avoid Windows long-path failures, support selecting individual component sources, and improve OpenAMP default domain YAML handling. This release also fixes backward compatibility when re-running or regenerating a BSP whose build directory already exists.

Features

  • Adds uC/OS-III BSP and application support to the pyesw SDT workflow.

Enhancements

  • Enables the XilASU client on PL MicroBlaze and MicroBlaze-V processors.
  • Moves BSP CMake build output to shorter path, avoiding Windows long-path build failures.
  • Adds support for picking partial/individual component sources in the BSP workflow.
  • Adds configurable BSP CMake build directory to avoid Windows path-length build failures.
  • Resolves OpenAMP default domain YAMLs from the template repo and reuses existing BSP YAMLs.

Bug Fixes

  • Restores backward compatibility when BSP build directory already exists during re-run or regen.

Following components were updated in this release with documentation/minor enhancements:

axiethernet_v5_20
axipmon_v6_15
axis_switch_v1_9
can_v3_13
canps_v3_13
cframe_v2_1
coresightps_dcc_v1_14
ddrcpsu_v1_9
ddrps_v1_3
ddrpsv_v1_9
dmaps_v2_13
emaclite_v4_14
emacps_v3_25
emc_v4_7
gpio_v4_14
gpiops_v3_16
iic_v3_16
iomodule_v2_21
ipipsu_v2_20
llfifo_v6_2
mutex_v4_12
nandpsu_v1_17
qspips_v3_16
resetps_v1_9
rtcpsu_v1_18
scugic_v5_8
scutimer_v2_9
scuwdt_v2_8
sd_fec_v1_6
sdps_v4_7
sysmon_v8_1
sysmonpsv_v5_4
tft_v6_9
tmr_inject_v1_9
trafgen_v4_10
uartlite_v3_14
uartns550_v3_15
uartps_v3_19
uartpsv_v1_15
wdtps_v3_10
wdttb_v5_13
xxvethernet_v1_14
xilflash_v4_14
xilfpga_v6_11
xilmailbox_v1_14
libmetal_v2_11