CVE-2024-56585

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> LoongArch: Fix sleeping in atomic context for PREEMPT_RT<br /> <br /> Commit bab1c299f3945ffe79 ("LoongArch: Fix sleeping in atomic context in<br /> setup_tlb_handler()") changes the gfp flag from GFP_KERNEL to GFP_ATOMIC<br /> for alloc_pages_node(). However, for PREEMPT_RT kernels we can still get<br /> a "sleeping in atomic context" error:<br /> <br /> [ 0.372259] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48<br /> [ 0.372266] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 0, name: swapper/1<br /> [ 0.372268] preempt_count: 1, expected: 0<br /> [ 0.372270] RCU nest depth: 1, expected: 1<br /> [ 0.372272] 3 locks held by swapper/1/0:<br /> [ 0.372274] #0: 900000000c9f5e60 (&amp;pcp-&gt;lock){+.+.}-{3:3}, at: get_page_from_freelist+0x524/0x1c60<br /> [ 0.372294] #1: 90000000087013b8 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x50/0x140<br /> [ 0.372305] #2: 900000047fffd388 (&amp;zone-&gt;lock){+.+.}-{3:3}, at: __rmqueue_pcplist+0x30c/0xea0<br /> [ 0.372314] irq event stamp: 0<br /> [ 0.372316] hardirqs last enabled at (0): [] 0x0<br /> [ 0.372322] hardirqs last disabled at (0): [] copy_process+0x9c0/0x26e0<br /> [ 0.372329] softirqs last enabled at (0): [] copy_process+0x9c0/0x26e0<br /> [ 0.372335] softirqs last disabled at (0): [] 0x0<br /> [ 0.372341] CPU: 1 UID: 0 PID: 0 Comm: swapper/1 Not tainted 6.12.0-rc7+ #1891<br /> [ 0.372346] Hardware name: Loongson Loongson-3A5000-7A1000-1w-CRB/Loongson-LS3A5000-7A1000-1w-CRB, BIOS vUDK2018-LoongArch-V2.0.0-prebeta9 10/21/2022<br /> [ 0.372349] Stack : 0000000000000089 9000000005a0db9c 90000000071519c8 9000000100388000<br /> [ 0.372486] 900000010038b890 0000000000000000 900000010038b898 9000000007e53788<br /> [ 0.372492] 900000000815bcc8 900000000815bcc0 900000010038b700 0000000000000001<br /> [ 0.372498] 0000000000000001 4b031894b9d6b725 00000000055ec000 9000000100338fc0<br /> [ 0.372503] 00000000000000c4 0000000000000001 000000000000002d 0000000000000003<br /> [ 0.372509] 0000000000000030 0000000000000003 00000000055ec000 0000000000000003<br /> [ 0.372515] 900000000806d000 9000000007e53788 00000000000000b0 0000000000000004<br /> [ 0.372521] 0000000000000000 0000000000000000 900000000c9f5f10 0000000000000000<br /> [ 0.372526] 90000000076f12d8 9000000007e53788 9000000005924778 0000000000000000<br /> [ 0.372532] 00000000000000b0 0000000000000004 0000000000000000 0000000000070000<br /> [ 0.372537] ...<br /> [ 0.372540] Call Trace:<br /> [ 0.372542] [] show_stack+0x38/0x180<br /> [ 0.372548] [] dump_stack_lvl+0x94/0xe4<br /> [ 0.372555] [] __might_resched+0x1a0/0x260<br /> [ 0.372561] [] rt_spin_lock+0x4c/0x140<br /> [ 0.372565] [] __rmqueue_pcplist+0x308/0xea0<br /> [ 0.372570] [] get_page_from_freelist+0x564/0x1c60<br /> [ 0.372575] [] __alloc_pages_noprof+0x218/0x1820<br /> [ 0.372580] [] tlb_init+0x1ac/0x298<br /> [ 0.372585] [] per_cpu_trap_init+0x114/0x140<br /> [ 0.372589] [] cpu_probe+0x4e4/0xa60<br /> [ 0.372592] [] start_secondary+0x34/0xc0<br /> [ 0.372599] [] smpboot_entry+0x64/0x6c<br /> <br /> This is because in PREEMPT_RT kernels normal spinlocks are replaced by<br /> rt spinlocks and rt_spin_lock() will cause sleeping. Fix it by disabling<br /> NUMA optimization completely for PREEMPT_RT kernels.