Adam Dunlap, Cláudio Carvalho, Dionna Glaze, James Bottomley, Joerg Roedel, Jon Lange, Oliver Steffen, Peter Fang, Stefano Garzarella, Tyler Fanelli, Vasant Karasulli, Vijay Dhanraj
- Joerg Roedel reminded everyone to test the new branches of the Linux kernel and QEMU repositories.
- The changes were necessitated by the introduction of TPR usage in the SVSM.
- Peter Fang was asked to check and resubmit any changes that might have been accidentally removed from the
svsm-tdx
branch in the Linux kernel repository.
- Jon Lange discussed his draft PR for multiprocessor startup for the non-isolated native platform, which utilizes
INIT
andSIPI
. - He highlighted the similarities between the challenges in native multiprocessor startup and TDX startup, suggesting the possibility of sharing code between the two paths.
- Peter Fang expressed interest in the draft PR.
- Jon Lange also mentioned the upcoming IPI-based vCPU TLB-flush functionality.
- Jon Lange raised concerns about the ongoing challenge of addressing unsafe blocks in the codebase, particularly in the context of CI and clippy lints.
- Stefano Garzarella reported progress in reducing unsafe blocks in the vTPM module.
- Joerg Roedel agreed to help with reviewing and documenting unsafe blocks, particularly in the memory allocator, in January.
- A discussion ensued about setting a deadline for documenting all unsafe blocks, with the possibility of setting a goal of the end of February.
- Cláudio Carvalho initiated a discussion about the choice of EK template for generating the EK in the vTPM attestation protocol.
- James Bottomley advocated for using ECC P-256 due to its superior security and efficiency compared to RSA-2048.
- The possibility of supporting multiple EK pubs and configuring them at build time was discussed.
- James Bottomley suggested specifying the algorithm as part of the attestation report to address potential compatibility issues.
- Tyler Fanelli noted that Keylime has support for ECC P-256.
- Cláudio Carvalho inquired about the possibility of using the external UUID crate instead of the internal implementation.
- Stefano Garzarella mentioned the UUID dependency in the
igvmmeasure
tool. - Joerg Roedel agreed to investigate the external crate and compare it with the existing implementation.
- Peter Fang provided an update on the discussions with Elena regarding SVSM observability for TDX.
- He mentioned the need to convince Linux architects about the use case and gain their approval for adding the necessary infrastructure.
- Jon Lange offered to assist in convincing the relevant stakeholders.
- Adam Dunlap proposed a small feature to address the issue of variable memory requirements for the SVSM region.
- The idea involves having the hypervisor guess the required memory and pass the allocated amount to the SVSM via an IGVM command line parameter.
- Jon Lange suggested an alternative approach using device tree as an IGVM parameter to specify memory that is not exposed in the E820 map.
- Adam Dunlap agreed to explore this approach and potentially submit a PR.
- Jon Lange raised a concern about the potential need for a real heap in the SVSM memory manager to handle discontiguous memory regions.
- Joerg Roedel acknowledged this as planned work for the future.
- Vijay Dhanraj reminded Joerg Roedel about the user space support PR and requested a review.
- Joerg Roedel confirmed that it is on his list of things to review before the holidays.
- Joerg Roedel expressed gratitude for everyone's contributions and highlighted the achievements of the project in the past year, including:
- Merging VTPM
- Extending hypervisor support
- Basic support for TDX and other platforms
- Progress with user mode HVM support
- Nearing completion of IPI support
- Joining the Confidential Computing Consortium
- Joerg Roedel outlined the following goals for the next year:
- Completing persistent memory and attestation support
- Further improving hypervisor and multi-platform support
- Making progress on KVM support
- Moving closer to supporting para-virtualized workloads
- Establishing a Technical Steering Committee
- Making the first release of the project
- Joerg Roedel thanked everyone again and wished them happy holidays.
- The next meeting was scheduled for January 8th.