xtensa: userspace: save PS to thread struct later

This is simply done to conserve code space in the vector text
areas. These vector text areas are very small and we should
only put code that is absolutely necessary for interrupt and
exception entrance. The saving of PS into the thread struct
can be deferred a bit. So do that.

Signed-off-by: Daniel Leung <daniel.leung@intel.com>

Author: dcpleung

arch/xtensa/include/xtensa_asm2.inc.S

@@ -431,6 +431,29 @@ _xstack_returned_\@:
 	 */
 	s32i a2, a1, ___xtensa_irq_bsa_t_scratch_OFFSET
 
+#ifdef CONFIG_USERSPACE
+	/* When restoring context via xtensa_switch and
+	 * returning from non-nested interrupts, we will be
+	 * using the stashed PS value in the thread struct
+	 * instead of the one in the thread stack. Both of
+	 * these scenarios will have nested value of 0.
+	 * So when nested value is zero, we store the PS
+	 * value into thread struct.
+	 */
+	rsr.ZSR_CPU a0
+	l32i a2, a0, ___cpu_t_nested_OFFSET
+	bnez a2, _excint_skip_ps_save_to_thread
+
+	l32i a2, a0, ___cpu_t_current_OFFSET
+	s32i a3, a2, _thread_offset_to_return_ps
+
+_excint_skip_ps_save_to_thread:
+	/* DEF_EXCINT saved PS into A3 so we need to restore
+	 * A3 here before proceeding.
+	 */
+	l32i a3, a1, ___xtensa_irq_bsa_t_a3_OFFSET
+#endif
+
 	ODD_REG_SAVE a0, a1
 
 #if XCHAL_HAVE_FP && defined(CONFIG_CPU_HAS_FPU) && defined(CONFIG_FPU_SHARING)
@@ -740,22 +763,10 @@ _not_triple_fault:
 .endif
 
 #ifdef CONFIG_USERSPACE
-	/* When restoring context via xtensa_switch and
-	 * returning from non-nested interrupts, we will be
-	 * using the stashed PS value in the thread struct
-	 * instead of the one in the thread stack. Both of
-	 * these scenarios will have nested value of 0.
-	 * So when nested value is zero, we store the PS
-	 * value into thread struct.
+	/* Stash the PS into A3 so EXCINT_HANDLER can read this
+	 * and save it into thread struct if needed.
 	 */
-	rsr.ZSR_CPU a3
-	l32i a2, a3, ___cpu_t_nested_OFFSET
-	bnez a2, _excint_skip_ps_save_to_thread_\LVL
-
-	l32i a2, a3, ___cpu_t_current_OFFSET
-	s32i a0, a2, _thread_offset_to_return_ps
-
-_excint_skip_ps_save_to_thread_\LVL:
+	mov a3, a0
 #endif
 
 	rsr.epc\LVL a0