Skip to content

Latest commit

 

History

History
111 lines (80 loc) · 3.3 KB

microcode_patching.md

File metadata and controls

111 lines (80 loc) · 3.3 KB
Raw

Microcode patching on PC Engines apu2/3/4/5

With v4.8.0.7 release we have implemented an experimental feature of microcode patching. Inspired by community we decided to add such feature to the PC Engines firmware.

The first implementation did not work as expected unfortunately. The procedure patched only BSP core leaving APs unpatched and AGESA was overwriting the patch with its own microcode in one of initialization phases.

Thus we have redesigned the feature to overwrite the microcode patch.

User guide

To build the firmware from scratch and add microcode patch follow the steps:

  1. Clone the pce-fw-builder

  2. Pull or build docker container:

    docker pull pcengines/pce-fw-builder

    or for legacy:

    docker pull pcengines/pce-fw-builder-legacy

  3. Build v4.8.0.7 image (v4.0.22 for legacy):

    ./build.sh release v4.8.0.7 {apu2|apu3|apu4|apu5}

  4. Download the microcode patch and place it in release/coreboot which is relative to cloned pce-fw-builder directory.

  5. Make changes to menuconfig:

    ./build.sh dev-build $PWD/release/coreboot {apu2|apu3|apu4|apu5} menuconfig

    In the Chipset submenu find Include CPU microcode in CBFS and choose Add microcode patch for AMD fam16h (EXPERIMENTAL) option. Then in the Chipset submenu fill the Microcode binary path and filename field with cpu00730F01_ver07030106_2018-02-09_88EDFAA0.bin which was downloaded in previous step. If the binary has been renamed, please fill the renamed binary here. Path is relative to coreboot root directory. When finished, save the config file.

  6. Build the image again:

    ./build.sh dev-build $PWD/release/coreboot {apu2|apu3|apu4|apu5} CPUS=$(nproc)

  7. Flash the new image. The firmware image can be found in release/coreboot/build which is relative to cloned pce-fw-builder directory.

Summary

By default the microcode patch level is 0x07030105 according to dmesg reports on Linux system. So v4.8.0.7 release binary will show:

microcode: CPU0: patch_level=0x07030105
microcode: CPU1: patch_level=0x07030105
microcode: CPU2: patch_level=0x07030105
microcode: CPU3: patch_level=0x07030105

After successfully building and flashing the image, dmesg should show:

microcode: CPU0: patch_level=0x07030106
microcode: CPU1: patch_level=0x07030106
microcode: CPU2: patch_level=0x07030106
microcode: CPU3: patch_level=0x07030106

Also when checking the vulnerability status with spectre-meltdown-checker one can notice that following fields have changed:

With microcode patch:

* PRED_CMD MSR is available:  YES
* CPU indicates IBPB capability:  YES  (IBPB_SUPPORT feature bit)
...
* IBPB enabled and active:  YES

Without microcode patch:

* PRED_CMD MSR is available:  NO
* CPU indicates IBPB capability:  NO
...
* IBPB enabled and active:  NO

Load msr kernel module before launching the script.