This week on xen-devel: 2 to 8 November 2026

Draft, not yet checked. The opening paragraph and the summaries are machine-written. Read each one against its thread, then delete this line.

This week, the list discussed speeding up interrupt delivery and improving address space isolation. Denis Mukhin posted a series adding support for AMD’s AVIC to bypass software emulation, and another series introducing an interrupt latency test to the Xen Test Framework. Jan Beulich reviewed patches for Address space isolation, which received a new command line option, and provided feedback on RISC-V virtual interrupt controller support.

664 messages in 98 threads went through the list from 2 to 8 November 2026. Each item below links to its full thread on lore.kernel.org.

Back from the archive

gnttab: follow-on to XSAs 500 and 501

7 patches · 2 messages from 2 people · first posted 96 days earlier · read the thread

Andrew Cooper suggests renaming a parameter in the grant table code (the part that manages memory shared between virtual machines) to req_grant_frames to match existing naming and add a clarifying comment. Jan Beulich notes that the req_ prefix was deliberately omitted earlier because it doesn’t fit all call sites, preferring a comment in the helper function instead.

Most discussed

x86: initial SVM AVIC support

28 patches · 56 messages from 6 people · read the thread

Denis Mukhin’s series adds support for AMD’s AVIC (Advanced Virtual Interrupt Controller) to speed up interrupt delivery by bypassing software emulation. Jan Beulich asks for documentation updates to be moved to the patch introducing the option and suggests using dashes instead of underscores. Ross Lagerwall questions whether a check for nested guests actually works since the code runs before the guest is active. Andrew Cooper points out that several patches rely on existing bugs in the CR8 handling logic.

ASI: HVM through X4

v6, 44 patches · 46 messages from 2 people · read the thread

Address space isolation (a security feature that keeps virtual machines from seeing each other’s memory) gets a new command line option in patch 06/44. Jan Beulich reviews patch 01/44 and adds a Reviewed-by tag.

RISC-V: virtual interrupt controller (vAPLIC/vIMSIC) support

v3, 39 patches · 41 messages from 3 people · first posted 32 days earlier · read the thread

Baptiste Le Duc asks for a clearer explanation of why the vCPU wake-up logic in patch 35/39 is sufficient to prevent missed interrupts, while Oleksii Kurochko agrees to adjust the wording in patch 17/39 to better reflect where MMIO emulation hooks in. Jan Beulich points out that volatile doesn’t express the memory clobber needed in patch 21/39, and Baptiste Le Duc suggests splitting a dense comment in patch 37/39 and moving a later patch earlier to fix a reserved encoding issue. Oleksii Kurochko and Jan Beulich also debate whether to strictly follow MISRA rules for integer types in patch 23/39, with Jan arguing that unsigned int is sufficient where it fits.

Add SMMUv3 Stage 1 Support for Xen guests

v5, 22 patches · 25 messages from 3 people · read the thread

Milan Djokic’s series adds emulated SMMUv3 support (the hardware that manages memory access for devices) to let guests handle Stage-1 translation. Robin Murphy notes that some newer MMU-700 chips don’t need the workaround in patch 07/22 and suggests disabling nested translation is too drastic. Jan Beulich says the new domctl command in patch 21/22 needs to be marked as Arm-only since it isn’t arch-independent.

x86: add interrupt latency test

15 patches · 16 messages from 1 person · read the thread

Denis Mukhin posted a 15-patch series to add an interrupt latency test to XTF, the Xen Test Framework (the hypervisor’s test suite).

x86: PVH Dom0 loading rework

v2, 14 patches · 15 messages from 1 person · read the thread

Jan Beulich posted the second revision of a series to let PVH Dom0 (the privileged virtual machine) load bigger initrd files without shrinking its memory. He asks for review on patch 10/14, which tries to map the initrd directly into Dom0.

xen/arm: scmi: introduce SCI SCMI SMC multi-agent support

v15, 6 patches · 14 messages from 2 people · read the thread

Oleksii Moisieiev posted a patch series that adds support for multiple SCMI agents (the part of the system that lets virtual machines talk to hardware management firmware) on ARM systems. Volodymyr Babchuk reviewed the driver code and approved it, then turned his attention to the documentation patch and flagged several wording and formatting issues. Oleksii Moisieiev responded by agreeing to fix the text and considering a stricter requirement for device permissions. Volodymyr Babchuk noted that the proposed device tree syntax needs adjustment because node names aren’t the same as labels.

CI: updates for XTF jobs

13 patches · 14 messages from 1 person · read the thread

Denis Mukhin posted a series of thirteen patches to overhaul the XTF test suite tooling (the framework that runs Xen tests) so it can orchestrate jobs on both QEMU and real hardware.

xen/gnttab: XSA-519/520 followup

6 patches · 13 messages from 3 people · read the thread

Andrew Cooper sends a six-part series to tighten up the grant table code (the part that manages memory shared between virtual machines) and clean up cache flushing functions. Jan Beulich asks why the grant table code shouldn’t use the same RCU locking logic as Arm’s platform operations and questions why that locking is kept at all, while he also notes that the new lock nesting in patch 5/6 is consistent but might be better avoided later. Michal Orzel reviews patch 6/6, and Andrew Cooper points out that speculatively mapping an unconfirmed memory frame is unsafe.

xen: address MISRA C Rule 11.8 violations

4 patches · 10 messages from 3 people · read the thread

Dmytro Prokopchuk posted a series to fix MISRA C Rule 11.8 violations, which stop casts from removing const or volatile qualifications from pointer types. Andrew Cooper objects to patch 4/4, arguing that freeing read-only memory is absurd and the real bug is livepatch storing heap allocations in const pointers. Jan Beulich notes that patches 2/4 and 3/4 are alternatives to earlier proposals and prefers his own variant for the latter. He also suggests the project needs a unified stance on whether freeing functions should take pointer-to-const before deviating any of them.

New series

• arm [PATCH v15 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support
• arm [PATCH] xen/arm: domain_build.c: Use hwdom instead of dom0 where applicable
• arm [PATCH] xen/arch/arm: Correcting comment typos across xen/arch/arm
• arm [PATCH] xen/arm: Preserve host GIC #interrupt-cells for Dom0
• arm [PATCH 1/4] xen/arm: guest_walk: refuse a block descriptor where the granule has none
• asi [PATCH v6 00/44] ASI: HVM through X4
• balloon [PATCH] xen/balloon: drop the device reference if registration fails
• build-id [PATCH 0/2] xen/build-id: Further cleanup
• ci [PATCH v3 0/6] CI: run the riscv64 smoke test with the QTB framework
• ci [PATCH v1 00/13] CI: updates for XTF jobs
• credit [PATCH] xen/credit: Uncount units dropped by csched_schedule()
• efi [PATCH] xen/efi: Handle efi_memmap with Xen relocating itself
• gnttab [PATCH] xen/gnttab: Shrink struct gnttab_copy_buf
• gnttab [PATCH 0/6] xen/gnttab: XSA-519/520 followup
• gnttab [PATCH] xen/gnttab: _cache_flush refuse an owner without a grant table
• lib [PATCH] lib: make memchr_inv() compliant with Misra rule 11.8
• libxl [PATCH] libxl: wait for QEMU exit before restarting a domain
• misra [PATCH v3] misra: deviate C library-style functions from Misra C:2012 rule 11.8
• other [PATCH v6] xen/arm, xen/common: Add Kconfig option to control Dom0 boot
• other [PATCH 0/4] xen: address MISRA C Rule 11.8 violations
• other [PATCH v4 0/3] “brk” allocator
• other [PATCH v5 00/22] Add SMMUv3 Stage 1 Support for Xen guests
• other [PATCH 0/4] libxl TPM emulator support
• pciback [PATCH] xen/pciback: drop the device reference if the export fails
• rfc RFC: Promote FLASK XSM Module to Supported in SUPPORT.md
• riscv [PATCH v1] xen/riscv: align header inclusion guards with CODING_STYLE
• tools [PATCH] tools/libs/guest: LZMA_OPTIONS_ERROR message cleanup
• tools [PATCH] tools/ocaml: Fix cleaning bindings test
• x86 [PATCH 2/2] x86/hvm: sync APs TSC with the BSP at creation only
• x86 [PATCH] x86/svm: Adjust vNMI handling to update the cleanbit
• x86 [PATCH] x86: Drop {nmi,mce}_pending defines
• x86 [PATCH] x86/pv: Improve code generation in async_exception_cleanup()
• x86 [PATCH] x86/mm: compile out __set_fixmap_x() when not needed
• x86 [PATCH] x86/MCE: reduce redundant code
• x86 [PATCH 0/2] x86/MCE: reduce redundant code
• x86 [PATCH] x86/cpuidle: allocate processor_powers[] dynamically
• x86 [PATCH] x86/cpufreq: allocate processor_pminfo[] dynamically
• x86 [PATCH] x86/nsvm: Fix L2 VINTR intercept merging when only Xen intercepts L1
• x86 [PATCH] x86/nsvm: Fix L2 VINTR intercept merging when only Xen intercepts L1
• x86 [PATCH] x86/platform: fix the MTRR rollback in XENPF_add_memtype
• x86 [PATCH v1 00/28] x86: initial SVM AVIC support
• x86 [PATCH] x86/EFI: fix padding of EDD strings
• x86 [PATCH v2 00/14] x86: PVH Dom0 loading rework
• x86 [PATCH] x86/nsvm: Fix L2 VINTR intercept merging when only Xen intercepts L1
• x86 [PATCH v2 0/4] x86/HVM: batch the flushes of grant unmaps
• x86 [PATCH] x86/bitops: Improve code generation for __scanbit()
• x86 [PATCH RFC] x86/hvm: Disallow modifying the APIC_ID register
• x86 [PATCH v2] x86/svm: Cleanup vintr_t type

Bugs and security

BUG: Xen BUG at mce_amd.c:100 (mc_ec2type) on Zen 3; host left wedged, no reboot

2 messages from 1 person · read the thread

Jan Beulich points out that a pointer value in the machine check code (the part that handles hardware errors) is being rotated by 16 bits, which explains why the values look like garbage. He asks if the reporter is using the XEN_MC_msrinject hypercall (a way for virtual machines to inject hardware errors) to rule out external interference. He then suggests removing an intpose_lookup call from the mca_rdmsr function to simplify how the system reads model-specific registers. Finally, he sends a patch that folds two separate reads into one to help debug the underlying issue further.

Review trailers sent this week

Linux and QEMU side

Threads copied to a Linux list (vger) or qemu-devel: Xen drivers, KVM and device models rather than the hypervisor.

• (qemu) [PATCH v8 00/18] hw/usb: Add a usbredir server transport — 50 messages
• (qemu) [PULL 00/25] Tracing patches — 28 messages
• (qemu) [PULL v2 00/25] Tracing patches — 28 messages
• (linux) [PATCH net-next v3 00/15] net: skbuff: follow-ups to the BUG_ON() removal — 24 messages
• (linux) [PATCH v11 00/23] KVM: x86: Cleaning up the KVM clock mess, part 2 — 24 messages
• (linux) [PATCH 00/14] mm: convert further simple mmap hooks to mmap_prepare — 22 messages
• (qemu) [PATCH 00/15] hw: expand DEFINE_MACHINE to explicit TypeInfo registration — 17 messages
• (linux) [PATCH net-next v4 00/15] net: skbuff: follow-ups to the BUG_ON() removal — 16 messages
• (linux) [PATCH v3 0/9] Add __must_init and require explicit kobj_type::release — 10 messages
• (linux) [PATCH] x86/CPU/AMD: Don’t print reset reason in VMs — 5 messages
• (linux) [PATCH 0/2] Re-raise irq_work lost during Xen suspend — 4 messages
• (linux) [PATCH] x86/paravirt: Add ioremap_verify_range hook and implement it for Xen PV — 2 messages
• (linux) [PATCH v2 0/9] Add __must_init and require explicit kobj_type::release — 2 messages
• (linux) [PATCH v6 00/51] x86: Try to wrangle PV clocks vs. TSC — 2 messages
• (linux) [PATCH] x86/xen: Fix lazy MMU pause count wrap-around — 1 message
• (linux) [PATCH] xen/grant-table: annotate on-stack work in gnttab_unmap_refs_sync() — 1 message
• (linux) [PATCH] xen/grant-table: annotate on-stack work in gnttab_unmap_refs_sync() — 1 message
• (linux) [PATCH 41/47] ARM: xen: Move CPUHP_AP_ARM_XEN_STARTING post stop_machine() — 1 message
• (linux) [PATCH] xen/events: replace __get_free_pages() with kmalloc() — 1 message

Counts