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

powerpc: Avoid nmi_enter/nmi_exit in real mode interrupt.

nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel
crash when invoked during real mode interrupt handling (e.g. early HMI/MCE
interrupt handler) if percpu allocation comes from vmalloc area.

Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI()
wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when
percpu allocation is from the embedded first chunk. However with
CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu
allocation can come from the vmalloc area.

With kernel command line "percpu_alloc=page" we can force percpu allocation
to come from vmalloc area and can see kernel crash in machine_check_early:

[ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110
[ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0
[ 1.215719] --- interrupt: 200
[ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable)
[ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0
[ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8

Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu
first chunk is not embedded.

References

• https://nvd.nist.gov/vuln/detail/CVE-2024-42126
• https://git.kernel.org/stable/c/0db880fc865ffb522141ced4bfa66c12ab1fbb70
• https://git.kernel.org/stable/c/0f37946c62c48a907625348cbc720a7a0c547d1e
• https://git.kernel.org/stable/c/2c78c9411e685dbc9eac8c2845111b03501975b8
• https://git.kernel.org/stable/c/8d3f83dfb23674540c827a8d65fba20aa300b252
• https://git.kernel.org/stable/c/e2afb26615adf6c3ceaaa7732aa839bcd587a057
• https://git.kernel.org/stable/c/fb6675db04c4b79883373edc578d5df7bbc84848