{"id":"CVE-2026-98396","title":"In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc: Do not restore KUAP in arch_exit_to_user_mode_prepare()\n\nKUAP means kernel cannot touch user memory unless it explicitly is\nenabled","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc: Do not restore KUAP in arch_exit_to_user_mode_prepare()\n\nKUAP means kernel cannot touch user memory unless it explicitly is\nenabled. In the kernel it should st…","severity":"none","published":"2026-10-14","updated":"2026-10-14","sourceUpdated":"2026-10-14T14:17:50.530","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-98396","references":[{"url":"https://git.kernel.org/stable/c/c2549d749539487239475fbc8c614a1f9244d655","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"},{"url":"https://git.kernel.org/stable/c/c8e80c19c3b9c8db83a14fab00445a6064992442","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"}],"tags":["nvd"],"ingestedAt":"2026-10-14T15:36:13.978Z","slug":"CVE-2026-98396","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\npowerpc: Do not restore KUAP in arch_exit_to_user_mode_prepare()\n\nKUAP means kernel cannot touch user memory unless it explicitly is\nenabled. In the kernel it should stay AMR_KUAP_BLOCKED. While returning\nto userspace just before RFI, kernel should restore the user AMR value\nback.\n\nLooks like GENERIC_ENTRY might be treating arch_exit_to_user_mode_prepare()\nas the last architecture step before returning to userspace.\ncommit bee25f97ad24 (\"powerpc: Enable GENERIC_ENTRY feature\")\ntherefore called kuap_user_restore() from that hook. But on PowerPC that\nis too early. After irqentry_exit() / syscall_exit_to_user_mode() we\nstill run platform specific exit routines.\n\ne.g. code snippets showing both exception handling and system call\nhandling as the callers of function arch_exit_to_user_mode_prepare()\nwhich does kuap_user_restore(). The below path shows that calling\nkuap_user_restore() is too early when called from\narch_exit_to_user_mode_prepare().\n\nException handling in exceptions-64s.S\n=======================================\n\nbl\tCFUNC(do_page_fault)\n\t  ..DEFINE_INTERRUPT_HANDLER_ASYNC(do_page_fault)\n\t      arch_interrupt_async_enter_prepare(regs);\n\t      state = irqentry_enter(regs);\n\t      instrumentation_begin();\n\t      irq_enter_rcu();\n\t      handler(regs);\n\t      nap_adjust_return(regs);\n\t      irq_exit_rcu();\n\t      instrumentation_end();\n\t      arch_interrupt_async_exit_prepare(regs);\n\t      irqentry_exit(regs, state); \t\t   <<< too early\n\t        irqentry_exit_to_user_mode()\n\t\t  __exit_to_user_mode_prepare(regs, EXIT_TO_USER_MODE_WORK_IRQ);\n\t\t    arch_exit_to_user_mode_prepare(regs, ti_work);  <<< too early\nb\tinterrupt_return_srr\n\t\t.. bl\tCFUNC(interrupt_exit_user_prepare) <<< already calls kuap_user_restore\n\nprep_irq_for_enabled_exit() retry can run kernel code with IRQs on. So\nonly when that routine is fully finished is when the user KUAP should be\nfully restored which interrupt_exit_user_prepare() already takes care of\nbefore returning.\n\nSimilarly for system call handling in interrupt_64.S\n======================================================\n\n\tbl\tCFUNC(system_call_exception)\n\n.Lsyscall_exit:\n\taddi\tr4,r1,STACK_INT_FRAME_REGS\n\tli\tr5,0 /* !scv */\n\tbl\tCFUNC(syscall_exit_prepare)\n\t\t  .. kuap_assert_locked();\n\t\t     syscall_exit_to_user_mode(regs); <<< too early\n\t\t       syscall_exit_to_user_mode_prepare(regs);  <<< too early\n \t\t     kuap_user_restore(regs);         <<< already calls\n\nsyscall_exit_prepare(), which can enable IRQs, replay a pending\ninterrupt, and only then rfi. Those functions already restore KUAP\nimmediately before rfi.\n\nNote that if we restore the user AMR too early like in the current code\nas shown from the code snippets above, then we get the following warning\nwhen CONFIG_PPC_KUAP_DEBUG is enabled:\n  WARNING: arch/powerpc/include/asm/book3s/64/kup.h:293 at interrupt_exit_user_prepare+0x1a0/0x1c0\n  Hardware name: IBM pSeries (emulated by qemu) POWER10 (architected)\n  TRAP: 0700\n  LR: c00000000000d8d4 CTR: c0000000021fe500\n  MSR: <SF,EE,ME,IR,DR,RI,LE>  CR: 44000804  XER: 20040000\n  interrupt_exit_user_prepare+0x1a0/0x1c0\n  interrupt_return_srr_user+0x8/0x12c\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.","depth":"sunlit","depthScore":3,"depthScoreParts":{"impact":2.8,"likelihood":0,"exploitation":0,"ransomware":0},"changes":[]}