LoongArch: Fix sleeping in atomic context for PREEMPT_RT

Commit bab1c29 ("LoongArch: Fix sleeping in atomic context in
setup_tlb_handler()") changes the gfp flag from GFP_KERNEL to GFP_ATOMIC
for alloc_pages_node(). However, for PREEMPT_RT kernels we can still get
a "sleeping in atomic context" error:

[    0.372259] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48
[    0.372266] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 0, name: swapper/1
[    0.372268] preempt_count: 1, expected: 0
[    0.372270] RCU nest depth: 1, expected: 1
[    0.372272] 3 locks held by swapper/1/0:
[    0.372274]  #0: 900000000c9f5e60 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x524/0x1c60
[    0.372294]  Rust-for-Linux#1: 90000000087013b8 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x50/0x140
[    0.372305]  Rust-for-Linux#2: 900000047fffd388 (&zone->lock){+.+.}-{3:3}, at: __rmqueue_pcplist+0x30c/0xea0
[    0.372314] irq event stamp: 0
[    0.372316] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
[    0.372322] hardirqs last disabled at (0): [<9000000005947320>] copy_process+0x9c0/0x26e0
[    0.372329] softirqs last  enabled at (0): [<9000000005947320>] copy_process+0x9c0/0x26e0
[    0.372335] softirqs last disabled at (0): [<0000000000000000>] 0x0
[    0.372341] CPU: 1 UID: 0 PID: 0 Comm: swapper/1 Not tainted 6.12.0-rc7+ #1891
[    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
[    0.372349] Stack : 0000000000000089 9000000005a0db9c 90000000071519c8 9000000100388000
[    0.372486]         900000010038b890 0000000000000000 900000010038b898 9000000007e53788
[    0.372492]         900000000815bcc8 900000000815bcc0 900000010038b700 0000000000000001
[    0.372498]         0000000000000001 4b031894b9d6b725 00000000055ec000 9000000100338fc0
[    0.372503]         00000000000000c4 0000000000000001 000000000000002d 0000000000000003
[    0.372509]         0000000000000030 0000000000000003 00000000055ec000 0000000000000003
[    0.372515]         900000000806d000 9000000007e53788 00000000000000b0 0000000000000004
[    0.372521]         0000000000000000 0000000000000000 900000000c9f5f10 0000000000000000
[    0.372526]         90000000076f12d8 9000000007e53788 9000000005924778 0000000000000000
[    0.372532]         00000000000000b0 0000000000000004 0000000000000000 0000000000070000
[    0.372537]         ...
[    0.372540] Call Trace:
[    0.372542] [<9000000005924778>] show_stack+0x38/0x180
[    0.372548] [<90000000071519c4>] dump_stack_lvl+0x94/0xe4
[    0.372555] [<900000000599b880>] __might_resched+0x1a0/0x260
[    0.372561] [<90000000071675cc>] rt_spin_lock+0x4c/0x140
[    0.372565] [<9000000005cbb768>] __rmqueue_pcplist+0x308/0xea0
[    0.372570] [<9000000005cbed84>] get_page_from_freelist+0x564/0x1c60
[    0.372575] [<9000000005cc0d98>] __alloc_pages_noprof+0x218/0x1820
[    0.372580] [<900000000593b36c>] tlb_init+0x1ac/0x298
[    0.372585] [<9000000005924b74>] per_cpu_trap_init+0x114/0x140
[    0.372589] [<9000000005921964>] cpu_probe+0x4e4/0xa60
[    0.372592] [<9000000005934874>] start_secondary+0x34/0xc0
[    0.372599] [<900000000715615c>] smpboot_entry+0x64/0x6c

This is because in PREEMPT_RT kernels normal spinlocks are replaced by
rt spinlocks and rt_spin_lock() will cause sleeping. Fix it by disabling
NUMA optimization completely for PREEMPT_RT kernels.

Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>

arch/loongarch/mm/tlb.c

@@ -289,7 +289,7 @@ static void setup_tlb_handler(int cpu)
 		/* Avoid lockdep warning */
 		rcutree_report_cpu_starting(cpu);
 
-#ifdef CONFIG_NUMA
+#if defined(CONFIG_NUMA) && !defined(CONFIG_PREEMPT_RT)
 		vec_sz = sizeof(exception_handlers);
 
 		if (pcpu_handlers[cpu])