CVE-2026-98395

Status: ReceivedUnscored—

Severity—Unscored

Exploitation probability0.20 %higher than 9% of CVEs

Active exploitationNonot in the CISA catalog

Public exploitNot knownno public code found

In the Linux kernel, the following vulnerability has been resolved:

powerpc/entry: Fix irq_soft_mask corruption on replayed interrupt exit

When __replay_soft_interrupts() replays a pending interrupt (e.g. PACA_IRQ_DEC -> timer_interrupt), it calls the handler directly with a synthetic pt_regs. The DEFINE_INTERRUPT_HANDLER_ASYNC wrapper around each handler calls arch_interrupt_async_exit_prepare() on the way out, which calls arch_interrupt_exit_prepare() -> local_irq_disable() -> arch_local_irq_disable(), which does:

This unconditionally overwrites irq_soft_mask with IRQS_DISABLED (0x1), stripping the IRQS_PMI_DISABLED (0x2) bit. The result is that irq_soft_mask is 0x1 instead of IRQS_ALL_DISABLED (0x3) when the handler returns to __replay_soft_interrupts().

Read full descriptionShow less

For a normally-taken interrupt this is harmless: the next interrupt always enters through arch_interrupt_enter_prepare() which unconditionally sets irq_soft_mask to IRQS_ALL_DISABLED. But during replay, next_interrupt() is called directly between replayed handlers without going back through arch_interrupt_enter_prepare(), so the stripped bit is never restored. next_interrupt() then fires a WARNING:

The warning was observed early in boot on a POWER10 pseries guest during kmem_cache_init_late(), where a spinlock release triggers interrupt replay that processes a pending timer interrupt.

The fix is to replace local_irq_disable() with hard_irq_disable().

hard_irq_disable() is the right primitive here for two reasons:

Technical details traces, logs and code from the original report

    irq_soft_mask_set(IRQS_DISABLED)   /* 0x1 */

    WARNING: arch/powerpc/kernel/irq_64.c:75
    WARN_ON(irq_soft_mask_return() != IRQS_ALL_DISABLED)

Debugger state confirming the bug:
  Before timer_interrupt(&regs):
    irq_soft_mask = 0x3 (IRQS_ALL_DISABLED)   correct
    irq_happened  = 0x41 (HARD_DIS|REPLAYING)  correct
  After timer_interrupt(&regs) returns:
    irq_soft_mask = 0x1 (IRQS_DISABLED)        WRONG - PMI bit stripped
    irq_happened  = 0x41                        unchanged

1. On PPC64 (hw_irq.h:301) it calls irq_soft_mask_set_return(IRQS_ALL_DISABLED),
   setting the soft mask to 0x3 (both IRQS_DISABLED and IRQS_PMI_DISABLED),
   which preserves the PMI bit and fixes the WARNING. The additional
   work it does (__hard_irq_disable(), PACA_IRQ_HARD_DIS |=) is
   redundant but safe since both are already set at this point in the
   exit path; the trace_hardirqs_off() inside is guarded by
   if (!arch_irqs_disabled_flags(flags)) so it will not double-fire.

2. On PPC32 (hw_irq.h:467) hard_irq_disable() maps to
   arch_local_irq_disable() -> __hard_irq_disable(), which clears
   MSR[EE] in hardware. This is exactly correct: PPC32 has no soft-mask
   PACA mechanism, so the hardware disable is the right way to satisfy
   irqentry_exit()'s requirement. This also fixes a build error on PPC32
   where irq_soft_mask_set() is only defined under CONFIG_PPC64:

       arch/powerpc/include/asm/entry-common.h:273: error: implicit
       declaration of function 'irq_soft_mask_set'

   Using hard_irq_disable() requires no #ifdef and is consistent with
   how the rest of the entry code (e.g. entry-common.h:463) handles the
   same PPC32/PPC64 split.

CVSS

NVD hasn't assigned a CVSS score to this CVE (common since the April 2026 policy change).

Exploitation probability (EPSS)

Probability of exploitation in the next 30 days: 0.20 %
Percentile among all scored CVEs: 9
Score date: 10/21/2026

EPSS (Exploit Prediction Scoring System, FIRST) estimates how likely a vulnerability is to be exploited in the wild within 30 days. It complements CVSS (impact) and CISA KEV (confirmed exploitation).

Affected technologies (1)

⚠ AI-inferred from the description — NVD hasn't analyzed this CVE yet, these aren't verified CPEs.

Linux — Linux Kernel^AI

References

Show 2 references

Raw JSON (NVD)

Show

{
  "id": "CVE-2026-98395",
  "cveTags": [],
  "metrics": {},
  "affected": [
    {
      "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
      "affectedData": [
        {
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "product": "Linux",
          "versions": [
            {
              "status": "affected",
              "version": "334f3f6d7a1660fd65597bad9960bfa33a716d35",
              "lessThan": "95cd316a896f1c1aa4ee4a817ba3d4691103c81b",
              "versionType": "git"
            },
            {
              "status": "affected",
              "version": "334f3f6d7a1660fd65597bad9960bfa33a716d35",
              "lessThan": "63a7531ca31f9f097d9cc1cc3fe86ae683cdabdd",
              "versionType": "git"
            }
          ],
          "programFiles": [
            "arch/powerpc/include/asm/entry-common.h"
          ],
          "defaultStatus": "unaffected"
        },
        {
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "product": "Linux",
          "versions": [
            {
              "status": "affected",
              "version": "7.2"
            },
            {
              "status": "unaffected",
              "version": "0",
              "lessThan": "7.2",
              "versionType": "semver"
            },
            {
              "status": "unaffected",
              "version": "7.2.7",
              "versionType": "semver",
              "lessThanOrEqual": "7.2.*"
            },
            {
              "status": "unaffected",
              "version": "7.3",
              "versionType": "original_commit_for_fix",
              "lessThanOrEqual": "*"
            }
          ],
          "programFiles": [
            "arch/powerpc/include/asm/entry-common.h"
          ],
          "defaultStatus": "affected"
        }
      ]
    }
  ],
  "published": "2026-10-14T14:17:50.407",
  "references": [
    {
      "url": "https://git.kernel.org/stable/c/63a7531ca31f9f097d9cc1cc3fe86ae683cdabdd",
      "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"
    },
    {
      "url": "https://git.kernel.org/stable/c/95cd316a896f1c1aa4ee4a817ba3d4691103c81b",
      "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"
    }
  ],
  "vulnStatus": "Received",
  "descriptions": [
    {
      "lang": "en",
      "value": "In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc/entry: Fix irq_soft_mask corruption on replayed interrupt exit\n\nWhen __replay_soft_interrupts() replays a pending interrupt (e.g.\nPACA_IRQ_DEC -> timer_interrupt), it calls the handler directly with a\nsynthetic pt_regs. The DEFINE_INTERRUPT_HANDLER_ASYNC wrapper around\neach handler calls arch_interrupt_async_exit_prepare() on the way out,\nwhich calls arch_interrupt_exit_prepare() -> local_irq_disable() ->\narch_local_irq_disable(), which does:\n\n    irq_soft_mask_set(IRQS_DISABLED)   /* 0x1 */\n\nThis unconditionally overwrites irq_soft_mask with IRQS_DISABLED (0x1),\nstripping the IRQS_PMI_DISABLED (0x2) bit. The result is that\nirq_soft_mask is 0x1 instead of IRQS_ALL_DISABLED (0x3) when the\nhandler returns to __replay_soft_interrupts().\n\nFor a normally-taken interrupt this is harmless: the next interrupt\nalways enters through arch_interrupt_enter_prepare() which\nunconditionally sets irq_soft_mask to IRQS_ALL_DISABLED. But during\nreplay, next_interrupt() is called directly between replayed handlers\nwithout going back through arch_interrupt_enter_prepare(), so the\nstripped bit is never restored. next_interrupt() then fires a WARNING:\n\n    WARNING: arch/powerpc/kernel/irq_64.c:75\n    WARN_ON(irq_soft_mask_return() != IRQS_ALL_DISABLED)\n\nThe warning was observed early in boot on a POWER10 pseries guest\nduring kmem_cache_init_late(), where a spinlock release triggers\ninterrupt replay that processes a pending timer interrupt.\n\nDebugger state confirming the bug:\n  Before timer_interrupt(&regs):\n    irq_soft_mask = 0x3 (IRQS_ALL_DISABLED)   correct\n    irq_happened  = 0x41 (HARD_DIS|REPLAYING)  correct\n  After timer_interrupt(&regs) returns:\n    irq_soft_mask = 0x1 (IRQS_DISABLED)        WRONG - PMI bit stripped\n    irq_happened  = 0x41                        unchanged\n\nThe fix is to replace local_irq_disable() with hard_irq_disable().\n\nhard_irq_disable() is the right primitive here for two reasons:\n\n1. On PPC64 (hw_irq.h:301) it calls irq_soft_mask_set_return(IRQS_ALL_DISABLED),\n   setting the soft mask to 0x3 (both IRQS_DISABLED and IRQS_PMI_DISABLED),\n   which preserves the PMI bit and fixes the WARNING. The additional\n   work it does (__hard_irq_disable(), PACA_IRQ_HARD_DIS |=) is\n   redundant but safe since both are already set at this point in the\n   exit path; the trace_hardirqs_off() inside is guarded by\n   if (!arch_irqs_disabled_flags(flags)) so it will not double-fire.\n\n2. On PPC32 (hw_irq.h:467) hard_irq_disable() maps to\n   arch_local_irq_disable() -> __hard_irq_disable(), which clears\n   MSR[EE] in hardware. This is exactly correct: PPC32 has no soft-mask\n   PACA mechanism, so the hardware disable is the right way to satisfy\n   irqentry_exit()'s requirement. This also fixes a build error on PPC32\n   where irq_soft_mask_set() is only defined under CONFIG_PPC64:\n\n       arch/powerpc/include/asm/entry-common.h:273: error: implicit\n       declaration of function 'irq_soft_mask_set'\n\n   Using hard_irq_disable() requires no #ifdef and is consistent with\n   how the rest of the entry code (e.g. entry-common.h:463) handles the\n   same PPC32/PPC64 split."
    }
  ],
  "lastModified": "2026-10-19T10:18:42.880",
  "sourceIdentifier": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"
}