このコンテンツは選択した言語では利用できません。
Chapter 4. Important changes to external kernel parameters
This chapter provides system administrators with a summary of significant changes in the kernel distributed with Red Hat Enterprise Linux 10.1. These changes could include, for example, added or updated proc entries, sysctl, and sysfs default values, boot parameters, kernel configuration options, or any noticeable behavior changes.
New kernel parameters
indirect_target_selection=
[X86, Intel] Mitigation control for the Indirect Target Selection (ITS) issue on Intel CPUs. Updated microcode is required for the IBPB fix.
Values:
on(default)- Enable mitigation.
off- Disable mitigation.
force- Force ITS bug state and deploy the default mitigation.
vmexit- Deploy mitigation only for guest/host isolation.
stuff- Use RSB-fill when retpoline is deployed; otherwise use the default mitigation.
See: Documentation/admin-guide/hw-vuln/indirect-target-selection.rst.
sdw_mclk_divider=
[SDW, Intel] Specify the MCLK divider for Intel SoundWire buses when the BIOS does not provide the clock rate properly.
thp_shmem=
[KNL, MM] Control the default huge page policy per size for the internal shmem mount.
Format:
<size>[KMG]<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy>
Policies: always, inherit, never, within_size, advise.
You can specify this multiple times to set policies for multiple THP sizes. See: Documentation/admin-guide/mm/transhuge.rst.
transparent_hugepage_shmem=
[KNL, MM] Control the huge page allocation policy for the internal shmem mount.
Values: always, within_size, advise, never, deny, force. See: Documentation/admin-guide/mm/transhuge.rst.
tsa=
[X86, AMD] Control mitigation for Transient Scheduler Attacks on AMD CPUs.
Values:
off- Disable the mitigation.
on(default)- Enable the mitigation.
user- Mitigate only user/kernel transitions.
vm- Mitigate only guest/host transitions.
For guidance, see vendor documentation for transient scheduler attacks.
Updated kernel parameters
init=
[KNL] Format: <full_path> Run the specified binary instead of /sbin/init as the system init process.
intremap=
[X86-64, Intel-IOMMU, EARLY]
Values:
on- Enable Interrupt Remapping (default).
off- Disable Interrupt Remapping.
kvm-arm.mode=
[KVM, ARM, EARLY] Select the Arm KVM virtualization mode.
Values:
nvhe- Standard nVHE-based mode without protected guests.
protected-
Support guests with state private from the host, using VHE or nVHE depending on hardware support. Disables kexec and hibernation on the host. To force nVHE on VHE hardware, add
arm64_sw.hvhe=0 id_aa64mmfr1.vh=0to the command line. nested- VHE-based mode with nested virtualization. Requires Armv8.4 hardware (FEAT_NV2). Experimental; use with extreme caution. Defaults to VHE or nVHE based on hardware support.
kvm-arm.vgic_v3_group0_trap=
[KVM, ARM, EARLY] Trap guest accesses to GICv3 group‑0 registers to the host for stricter isolation and debugging.
libata.force=
[SATA/ATA] Per‑port options:
- New:
external -
Mark the port as external (hotplug‑capable). Other options remain available (for example,
max_sec_lba48,[no]lpm,[no]setxfer).
nohz
[KNL] Disable the tick when a single task runs, and offload other kernel work such as RCU callbacks. Equivalent to nohz_full. A residual 1 Hz tick is offloaded to workqueues; affine these to housekeeping CPUs via the global workqueue CPU mask. See also rcu_nocbs= and isolcpus=.
pci=
[PCI] ACS configuration example updated to include a device selector: pci=config_acs=10x@pci:0:0:: Configure supported devices to enable P2P Request Redirect, disable Translation Blocking, and leave Source Validation unchanged along the specified device path.
pcie=
[PCIE] New system‑wide flag: notph:: Disable PCIe TLP Processing Hints support when CONFIG_PCIE_TPH is enabled.
pcie_aspm=
[PCIE] Forcibly enable or ignore PCIe Active State Power Management. Behavior unchanged; use alongside pcie= flags where needed.
preempt=
[KNL, Scheduler] Preemption control modes. New: lazy:: Scheduler‑controlled mode similar to full. The task gets one HZ tick to yield before the scheduler forces preemption. A preemption is counted when the task returns to user space.
print-fatal-signals=
[KNL] Enable debugging output for fatal signals.
skew_tick=
[KNL, EARLY] Offset the periodic timer tick per CPU to reduce contention in large systems and with CONFIG_MAXSMP.
slub_debug=
[MM] SLUB allocator debugging; slub_nomerge remains a legacy alias. See Documentation/mm/slub.rst.
spectre_v2=
[X86] Selecting a specific mitigation does not force‑enable user‑space mitigations. Selecting on enables kernel protections and mitigates user‑to‑user task attacks. Selecting off disables both.
tsc=
[X86] Disable clocksource stability checks for TSC.
Values:
reliable- Mark the TSC clocksource as reliable.
noirqtime- Do not use TSC for IRQ accounting.
unstable- Mark the TSC clocksource as unstable.
nowatchdog- Disable the clocksource watchdog.
recalibrate- Recalibrate against HPET or PM timer if TSC frequency came from MSR or CPUID(0x15); warn if the difference is more than 500 ppm.
watchdog- Use TSC as the watchdog clocksource on systems where TSC is trustworthy.
Note: An earlier tsc=nowatchdog suppresses watchdog. A later tsc=nowatchdog overrides it; the kernel logs any suppression or override.
transparent_hugepage_shmem=
[KNL, MM] Values: always, within_size, advise, never, deny, force. Controls the internal shmem mount policy. See Documentation/admin-guide/mm/transhuge.rst.
mitigations=
[Multi‑arch] Selecting off disables a set of kernel and user‑space mitigations. The equivalence list now includes indirect_target_selection=off on X86, in addition to existing entries such as kpti=0 on Arm64, gather_data_sampling=off, kvm.nx_huge_pages=off, l1tf=off, mds=off, and related X86 flags.
Updated sysctl parameters
timer_migration
When set to a non-zero value, the kernel attempts to migrate timers away from idle CPUs to help those CPUs remain in low-power states longer.
Default: 1 (enabled).