x86/alternatives: Acquire pte lock with interrupts enabled

pte lock is never acquired in-IRQ context so it does not require interrupts
to be disabled. The lock is a regular spinlock which cannot be acquired
with interrupts disabled on RT.

RT complains about pte_lock() in __text_poke() because it's invoked after
disabling interrupts.

__text_poke() has to disable interrupts as use_temporary_mm() expects
interrupts to be off because it invokes switch_mm_irqs_off() and uses
per-CPU (current active mm) data.

Move the PTE lock handling outside the interrupt disabled region.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by; Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20200813105026.bvugytmsso6muljw@linutronix.de

arch/x86/kernel/alternative.c

@@ -874,8 +874,6 @@ static void *__text_poke(void *addr, const void *opcode, size_t len)
 	 */
 	BUG_ON(!pages[0] || (cross_page_boundary && !pages[1]));
 
-	local_irq_save(flags);
-
 	/*
 	 * Map the page without the global bit, as TLB flushing is done with
 	 * flush_tlb_mm_range(), which is intended for non-global PTEs.
@@ -892,6 +890,8 @@ static void *__text_poke(void *addr, const void *opcode, size_t len)
 	 */
 	VM_BUG_ON(!ptep);
 
+	local_irq_save(flags);
+
 	pte = mk_pte(pages[0], pgprot);
 	set_pte_at(poking_mm, poking_addr, ptep, pte);
 
@@ -941,8 +941,8 @@ static void *__text_poke(void *addr, const void *opcode, size_t len)
 	 */
 	BUG_ON(memcmp(addr, opcode, len));
 
-	pte_unmap_unlock(ptep, ptl);
 	local_irq_restore(flags);
+	pte_unmap_unlock(ptep, ptl);
 	return addr;
 }