APPLIED: [SRU][Q/N/J:linux-azure][PATCH 0/2] [Azure] AMDv4 VMs may get no timer interrupts or suffer from timer interrupt storms

John Cabaj john.cabaj at canonical.com
Fri Jan 23 15:58:44 UTC 2026


On 1/20/26 1:15 PM, John Cabaj wrote:
> BugLink: https://bugs.launchpad.net/bugs/2137674
> 
> SRU Justification
> 
> [Impact]
> 
> * Microsoft requested patches to work around intermittent deployment failures on AMDv4 VM sizes due to issues with large values of Hyper-V timer
> 
> [Fix]
> 
> * Questing and Noble:
> - cherry-picks of the following SAUCE patches from Microsoft (https://github.com/dcui/linux/tree/decui/Ubuntu-azure-6.8-6.8.0-1043.49_22.04.1-V2):
>    - 28b18cb: "UBUNTU: SAUCE: clocksource: hyper-v: do not use an insanely big TSC in hv_read_tsc_page_tsc()"
>    - f796530: "UBUNTU: SAUCE: clocksource: hyper-v: use the APIC timer if the Hyper-V timer is unreliable on some CPUs"
> * Jammy:
> - cherry-pick of the following SAUCE patch from Microsoft (https://github.com/dcui/linux/tree/decui/Ubuntu-azure-6.8-6.8.0-1043.49_22.04.1-V2):
>    - 28b18cb: "UBUNTU: SAUCE: clocksource: hyper-v: do not use an insanely big TSC in hv_read_tsc_page_tsc()"
> - backport of the following SAUCE patch from Microsoft (https://github.com/dcui/linux/tree/decui/Ubuntu-azure-6.8-6.8.0-1043.49_22.04.1-V2):
>    - f796530: "UBUNTU: SAUCE: clocksource: hyper-v: use the APIC timer if the Hyper-V timer is unreliable on some CPUs"
> - cherry-pick of the following upstream patches
>    - fd0c6b0: "x86/vdso: Fix gettimeofday masking"
>    - 4b52223: "clocksource: hyper-v: Provide noinstr sched_clock()"
> - backports of the following upstream patches
>    - 9dfb6bf: "clocksource: hyper-v: Adjust hv_read_tsc_page_tsc() to avoid special casing U64_MAX"
>    - 47a2a14: "clocksource: hyper-v: Introduce a pointer to TSC page"
> 
> [Test Plan]
> 
> * Compile tested
> * Microsoft tested
> 
> [Regression potential]
> 
> * Upstream changes are largely localized to time and clock source functions, and have been upstream for some time. Regression likliehood is slim.
> * Possible regressions could result in unexpected behavior of Hyper-V driver.
> 
> [Other info]
> 
> * SF#00422328
> 
> Dexuan Cui (2):
>    UBUNTU: SAUCE: clocksource: hyper-v: do not use an insanely big TSC in
>      hv_read_tsc_page_tsc()
>    UBUNTU: SAUCE: clocksource: hyper-v: use the APIC timer if the Hyper-V
>      timer is unreliable on some CPUs
> 
>   arch/x86/kernel/cpu/mshyperv.c     | 43 ++++++++++++++++++++++++++++++
>   drivers/clocksource/hyperv_timer.c |  4 ++-
>   include/clocksource/hyperv_timer.h |  4 +++
>   3 files changed, 50 insertions(+), 1 deletion(-)
> 

Applied to Questing, Noble, and Jammy linux-azure master-next branches.


Thanks,
j-c





More information about the kernel-team mailing list