CVE-2024-42126

{"id": "CVE-2024-42126", "cveTags": [], "metrics": {}, "published": "2024-07-30T08:15:04.743", "references": [{"url": "https://git.kernel.org/stable/c/0db880fc865ffb522141ced4bfa66c12ab1fbb70", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}, {"url": "https://git.kernel.org/stable/c/0f37946c62c48a907625348cbc720a7a0c547d1e", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}, {"url": "https://git.kernel.org/stable/c/2c78c9411e685dbc9eac8c2845111b03501975b8", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}, {"url": "https://git.kernel.org/stable/c/8d3f83dfb23674540c827a8d65fba20aa300b252", "source": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}], "vulnStatus": "Awaiting Analysis", "descriptions": [{"lang": "en", "value": "In the Linux kernel, the following vulnerability has been resolved:\n\npowerpc: Avoid nmi_enter/nmi_exit in real mode interrupt.\n\nnmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel\ncrash when invoked during real mode interrupt handling (e.g. early HMI/MCE\ninterrupt handler) if percpu allocation comes from vmalloc area.\n\nEarly HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI()\nwrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when\npercpu allocation is from the embedded first chunk. However with\nCONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu\nallocation can come from the vmalloc area.\n\nWith kernel command line \"percpu_alloc=page\" we can force percpu allocation\nto come from vmalloc area and can see kernel crash in machine_check_early:\n\n[ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110\n[ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0\n[ 1.215719] --- interrupt: 200\n[ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable)\n[ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0\n[ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8\n\nFix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu\nfirst chunk is not embedded."}, {"lang": "es", "value": "En el kernel de Linux, se ha resuelto la siguiente vulnerabilidad: powerpc: Evite nmi_enter/nmi_exit en interrupci\u00f3n en modo real. nmi_enter()/nmi_exit() toca variables por CPU que pueden provocar un fallo del kernel cuando se invoca durante el manejo de interrupciones en modo real (por ejemplo, el controlador de interrupciones HMI/MCE temprano) si la asignaci\u00f3n de percpu proviene del \u00e1rea vmalloc. Los primeros controladores HMI/MCE se llaman a trav\u00e9s del contenedor DEFINE_INTERRUPT_HANDLER_NMI() que invoca llamadas nmi_enter/nmi_exit. No vemos ning\u00fan problema cuando la asignaci\u00f3n de percpu proviene del primer fragmento integrado. Sin embargo, con CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK habilitado, hay posibilidades de que la asignaci\u00f3n de percpu pueda provenir del \u00e1rea vmalloc. Con la l\u00ednea de comando del kernel \"percpu_alloc=page\" podemos forzar que la asignaci\u00f3n de percpu provenga del \u00e1rea vmalloc y podemos ver el fallo del kernel en machine_check_early: [1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 [1.215717] LR [c0000000000461a0] check_early+0xf0/ 0x2c0 [1.215719] --- interrupci\u00f3n: 200 [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (no confiable) [ 1.215722] [c000000fffd731b0] 000] 0x0 [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 Solucionar esto mediante evitando el uso de nmi_enter()/nmi_exit() en modo real si el primer fragmento de percpu no est\u00e1 incrustado."}], "lastModified": "2024-07-30T13:32:45.943", "sourceIdentifier": "416baaa9-dc9f-4396-8d5f-8c081fb06d67"}