| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20251120225554.GD24244@linuxonhyperv3.guj3yctzbm1etfxqx2vob5hsef.xx.internal.cloudapp.net>
Date: Thu, 20 Nov 2025 14:55:54 -0800
From: Praveen Paladugu <prapal@...ux.microsoft.com>
To: Stanislav Kinsburskii <skinsburskii@...ux.microsoft.com>
Cc: kys@...rosoft.com, haiyangz@...rosoft.com, wei.liu@...nel.org,
decui@...rosoft.com, tglx@...utronix.de, mingo@...hat.com,
linux-hyperv@...r.kernel.org, linux-kernel@...r.kernel.org,
bp@...en8.de, dave.hansen@...ux.intel.com, x86@...nel.org,
hpa@...or.com, arnd@...db.de, anbelski@...ux.microsoft.com,
easwar.hariharan@...ux.microsoft.com,
nunodasneves@...ux.microsoft.com
Subject: Re: [PATCH v5 3/3] hyperv: Cleanly shutdown root partition with MSHV
On Mon, Nov 17, 2025 at 03:45:36PM -0800, Stanislav Kinsburskii wrote:
> On Mon, Nov 17, 2025 at 03:08:18PM -0600, Praveen K Paladugu wrote:
> > When a root partition running on MSHV is powered off, the default
> > behavior is to write ACPI registers to power-off. However, this ACPI
> > write is intercepted by MSHV and will result in a Machine Check
> > Exception(MCE).
> >
> > The root partition eventually panics with a trace similar to:
> >
> > [ 81.306348] reboot: Power down
> > [ 81.314709] mce: [Hardware Error]: CPU 0: Machine Check Exception: 4 Bank 0: b2000000c0060001
> > [ 81.314711] mce: [Hardware Error]: TSC 3b8cb60a66 PPIN 11d98332458e4ea9
> > [ 81.314713] mce: [Hardware Error]: PROCESSOR 0:606a6 TIME 1759339405 SOCKET 0 APIC 0 microcode ffffffff
> > [ 81.314715] mce: [Hardware Error]: Run the above through 'mcelog --ascii'
> > [ 81.314716] mce: [Hardware Error]: Machine check: Processor context corrupt
> > [ 81.314717] Kernel panic - not syncing: Fatal machine check
> >
> > To correctly shutdown a root partition running on MSHV hypervisor, sleep
> > state information must be configured within the hypervsior. Later, the
> > HVCALL_ENTER_SLEEP_STATE hypercall should be invoked as the last step in
> > the shutdown sequence.
> >
> > The previous patch configures the sleep state information and this patch
> > invokes HVCALL_ENTER_SLEEP_STATE hypercall to cleanly shutdown the root
> > partition.
> >
> > Signed-off-by: Praveen K Paladugu <prapal@...ux.microsoft.com>
> > Co-developed-by: Anatol Belski <anbelski@...ux.microsoft.com>
> > Signed-off-by: Anatol Belski <anbelski@...ux.microsoft.com>
> > Reviewed-by: Easwar Hariharan <easwar.hariharan@...ux.microsoft.com>
> > ---
> > arch/x86/hyperv/hv_init.c | 2 ++
> > arch/x86/include/asm/mshyperv.h | 2 ++
> > drivers/hv/mshv_common.c | 18 ++++++++++++++++++
> > 3 files changed, 22 insertions(+)
> >
> > diff --git a/arch/x86/hyperv/hv_init.c b/arch/x86/hyperv/hv_init.c
> > index 645b52dd732e..24824534ff8d 100644
> > --- a/arch/x86/hyperv/hv_init.c
> > +++ b/arch/x86/hyperv/hv_init.c
> > @@ -34,6 +34,7 @@
> > #include <clocksource/hyperv_timer.h>
> > #include <linux/highmem.h>
> > #include <linux/export.h>
> > +#include <asm/reboot.h>
> >
> > void *hv_hypercall_pg;
> >
> > @@ -562,6 +563,7 @@ void __init hyperv_init(void)
> > * failures here.
> > */
> > hv_sleep_notifiers_register();
> > + machine_ops.power_off = hv_machine_power_off;
>
> It looks more natural to me to gather all the machine_ops hooks in one
> place (meaning in ms_hyperv_init_platform).
> It is better moving this assignment there and do the branching on the
> partition type in the power_off callback instead.
>
Moving machine_ops hooks assignment to ms_hyperv_init_platform sounds
reasonable to me. I have a concern about doing the branching within the
callback though.
This assignment overwrites the default of using ACPI for poweroff.
By moving the branching into hv_machine_power_off, it would look like:
hv_machine_power_off {
if not root {
fallback to default
} else {
Use mshv hypercall to poweroff
}
}
I would rather do something cleaner like:
if root {
machine_ops.power_off = hv_machine_power_off;
}
Praveen
> Thanks,
> Stanislav
>
>
> > } else {
> > hypercall_msr.guest_physical_address = vmalloc_to_pfn(hv_hypercall_pg);
> > wrmsrq(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
> > diff --git a/arch/x86/include/asm/mshyperv.h b/arch/x86/include/asm/mshyperv.h
> > index 166053df0484..4c22f3257368 100644
> > --- a/arch/x86/include/asm/mshyperv.h
> > +++ b/arch/x86/include/asm/mshyperv.h
> > @@ -183,9 +183,11 @@ void hv_apic_init(void);
> > void __init hv_init_spinlocks(void);
> > bool hv_vcpu_is_preempted(int vcpu);
> > void hv_sleep_notifiers_register(void);
> > +void hv_machine_power_off(void);
> > #else
> > static inline void hv_apic_init(void) {}
> > static inline void hv_sleep_notifiers_register(void) {};
> > +static inline void hv_machine_power_off(void) {};
> > #endif
> >
> > struct irq_domain *hv_create_pci_msi_domain(void);
> > diff --git a/drivers/hv/mshv_common.c b/drivers/hv/mshv_common.c
> > index ee733ba1575e..73505cbdc324 100644
> > --- a/drivers/hv/mshv_common.c
> > +++ b/drivers/hv/mshv_common.c
> > @@ -216,3 +216,21 @@ void hv_sleep_notifiers_register(void)
> > pr_err("%s: cannot register reboot notifier %d\n", __func__,
> > ret);
> > }
> > +
> > +/*
> > + * Power off the machine by entering S5 sleep state via Hyper-V hypercall.
> > + * This call does not return if successful.
> > + */
> > +void hv_machine_power_off(void)
> > +{
> > + unsigned long flags;
> > + struct hv_input_enter_sleep_state *in;
> > +
> > + local_irq_save(flags);
> > + in = *this_cpu_ptr(hyperv_pcpu_input_arg);
> > + in->sleep_state = HV_SLEEP_STATE_S5;
> > +
> > + (void)hv_do_hypercall(HVCALL_ENTER_SLEEP_STATE, in, NULL);
> > + local_irq_restore(flags);
> > +
> > +}
> > --
> > 2.51.0
Powered by blists - more mailing lists