powerpc/32s: Move KUEP locking/unlocking in C

This can be done in C, do it.

Unrolling the loop gains approx. 15% performance.

From now on, prepare_transfer_to_handler() is only for
interrupts from kernel.

Signed-off-by: Christophe Leroy <[email protected]>
Signed-off-by: Michael Ellerman <[email protected]>
Link: https://lore.kernel.org/r/4eadd873927e9a73c3d1dfe2f9497353465514cf.1615552867.git.christophe.leroy@csgroup.eu

Author: chleroy

arch/powerpc/include/asm/book3s/32/kup.h

@@ -7,37 +7,6 @@
 
 #ifdef __ASSEMBLY__
 
-.macro kuep_update_sr	gpr1, gpr2		/* NEVER use r0 as gpr2 due to addis */
-101:	mtsrin	\gpr1, \gpr2
-	addi	\gpr1, \gpr1, 0x111		/* next VSID */
-	rlwinm	\gpr1, \gpr1, 0, 0xf0ffffff	/* clear VSID overflow */
-	addis	\gpr2, \gpr2, 0x1000		/* address of next segment */
-	bdnz	101b
-	isync
-.endm
-
-.macro kuep_lock	gpr1, gpr2
-#ifdef CONFIG_PPC_KUEP
-	li	\gpr1, NUM_USER_SEGMENTS
-	li	\gpr2, 0
-	mtctr	\gpr1
-	mfsrin	\gpr1, \gpr2
-	oris	\gpr1, \gpr1, SR_NX@h		/* set Nx */
-	kuep_update_sr \gpr1, \gpr2
-#endif
-.endm
-
-.macro kuep_unlock	gpr1, gpr2
-#ifdef CONFIG_PPC_KUEP
-	li	\gpr1, NUM_USER_SEGMENTS
-	li	\gpr2, 0
-	mtctr	\gpr1
-	mfsrin	\gpr1, \gpr2
-	rlwinm	\gpr1, \gpr1, 0, ~SR_NX		/* Clear Nx */
-	kuep_update_sr \gpr1, \gpr2
-#endif
-.endm
-
 #ifdef CONFIG_PPC_KUAP
 
 .macro kuap_update_sr	gpr1, gpr2, gpr3	/* NEVER use r0 as gpr2 due to addis */

arch/powerpc/include/asm/interrupt.h

@@ -33,8 +33,10 @@ static inline void interrupt_enter_prepare(struct pt_regs *regs, struct interrup
 	if (!arch_irq_disabled_regs(regs))
 		trace_hardirqs_off();
 
-	if (user_mode(regs))
+	if (user_mode(regs)) {
+		kuep_lock();
 		account_cpu_user_entry();
+	}
 #endif
 	/*
 	 * Book3E reconciles irq soft mask in asm
@@ -89,6 +91,8 @@ static inline void interrupt_exit_prepare(struct pt_regs *regs, struct interrupt
 	exception_exit(state->ctx_state);
 #endif
 
+	if (user_mode(regs))
+		kuep_unlock();
 	/*
 	 * Book3S exits to user via interrupt_exit_user_prepare(), which does
 	 * context tracking, which is a cleaner way to handle PREEMPT=y

arch/powerpc/include/asm/kup.h

@@ -55,6 +55,14 @@ void setup_kuep(bool disabled);
 static inline void setup_kuep(bool disabled) { }
 #endif /* CONFIG_PPC_KUEP */
 
+#if defined(CONFIG_PPC_KUEP) && defined(CONFIG_PPC_BOOK3S_32)
+void kuep_lock(void);
+void kuep_unlock(void);
+#else
+static inline void kuep_lock(void) { }
+static inline void kuep_unlock(void) { }
+#endif
+
 #ifdef CONFIG_PPC_KUAP
 void setup_kuap(bool disabled);
 #else

arch/powerpc/kernel/entry_32.S

@@ -51,16 +51,9 @@
 #if defined(CONFIG_PPC_BOOK3S_32) || defined(CONFIG_E500)
 	.globl	prepare_transfer_to_handler
 prepare_transfer_to_handler:
-	andi.	r12,r9,MSR_PR
 	addi	r12,r2,THREAD
-	beq	2f
-#ifdef CONFIG_PPC_BOOK3S_32
-	kuep_lock r11, r12
-#endif
-	blr
 
 	/* if from kernel, check interrupted DOZE/NAP mode */
-2:
 	kuap_save_and_lock r11, r12, r9, r5, r6
 	lwz	r12,TI_LOCAL_FLAGS(r2)
 	mtcrf	0x01,r12
@@ -86,9 +79,6 @@ _ASM_NOKPROBE_SYMBOL(prepare_transfer_to_handler)
 	.globl	transfer_to_syscall
 transfer_to_syscall:
 	SAVE_NVGPRS(r1)
-#ifdef CONFIG_PPC_BOOK3S_32
-	kuep_lock r11, r12
-#endif
 
 	/* Calling convention has r9 = orig r0, r10 = regs */
 	addi	r10,r1,STACK_FRAME_OVERHEAD
@@ -105,9 +95,6 @@ ret_from_syscall:
 	cmplwi	cr0,r5,0
 	bne-	2f
 #endif /* CONFIG_PPC_47x */
-#ifdef CONFIG_PPC_BOOK3S_32
-	kuep_unlock r5, r7
-#endif
 	kuap_check r2, r4
 	lwz	r4,_LINK(r1)
 	lwz	r5,_CCR(r1)
@@ -311,9 +298,6 @@ interrupt_return:
 	bne-	.Lrestore_nvgprs
 
 .Lfast_user_interrupt_return:
-#ifdef CONFIG_PPC_BOOK3S_32
-	kuep_unlock	r10, r11
-#endif
 	kuap_check r2, r4
 	lwz	r11,_NIP(r1)
 	lwz	r12,_MSR(r1)

arch/powerpc/kernel/head_32.h

@@ -133,7 +133,10 @@ _ASM_NOKPROBE_SYMBOL(\name\()_virt)
 
 .macro prepare_transfer_to_handler
 #ifdef CONFIG_PPC_BOOK3S_32
+	andi.	r12,r9,MSR_PR
+	bne	777f
 	bl	prepare_transfer_to_handler
+777:
 #endif
 .endm
 

arch/powerpc/kernel/head_booke.h

@@ -105,7 +105,10 @@ END_BTB_FLUSH_SECTION
 
 .macro prepare_transfer_to_handler
 #ifdef CONFIG_E500
+	andi.	r12,r9,MSR_PR
+	bne	777f
 	bl	prepare_transfer_to_handler
+777:
 #endif
 .endm
 

arch/powerpc/kernel/interrupt.c

@@ -33,6 +33,8 @@ notrace long system_call_exception(long r3, long r4, long r5,
 {
 	syscall_fn f;
 
+	kuep_lock();
+
 	regs->orig_gpr3 = r3;
 
 	if (IS_ENABLED(CONFIG_PPC_IRQ_SOFT_MASK_DEBUG))
@@ -354,6 +356,8 @@ notrace unsigned long syscall_exit_prepare(unsigned long r3,
 	 */
 	kuap_user_restore(regs);
 #endif
+	kuep_unlock();
+
 	return ret;
 }
 

arch/powerpc/mm/book3s32/Makefile

@@ -9,3 +9,4 @@ endif
 obj-y += mmu.o mmu_context.o
 obj-$(CONFIG_PPC_BOOK3S_603) += nohash_low.o
 obj-$(CONFIG_PPC_BOOK3S_604) += hash_low.o tlb.o
+obj-$(CONFIG_PPC_KUEP) += kuep.o

arch/powerpc/mm/book3s32/kuep.c

@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <asm/kup.h>
+#include <asm/reg.h>
+#include <asm/task_size_32.h>
+#include <asm/mmu.h>
+
+#define KUEP_UPDATE_TWO_USER_SEGMENTS(n) do {		\
+	if (TASK_SIZE > ((n) << 28))			\
+		mtsr(val1, (n) << 28);			\
+	if (TASK_SIZE > (((n) + 1) << 28))		\
+		mtsr(val2, ((n) + 1) << 28);		\
+	val1 = (val1 + 0x222) & 0xf0ffffff;		\
+	val2 = (val2 + 0x222) & 0xf0ffffff;		\
+} while (0)
+
+static __always_inline void kuep_update(u32 val)
+{
+	int val1 = val;
+	int val2 = (val + 0x111) & 0xf0ffffff;
+
+	KUEP_UPDATE_TWO_USER_SEGMENTS(0);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(2);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(4);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(6);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(8);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(10);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(12);
+	KUEP_UPDATE_TWO_USER_SEGMENTS(14);
+}
+
+void kuep_lock(void)
+{
+	kuep_update(mfsr(0) | SR_NX);
+}
+
+void kuep_unlock(void)
+{
+	kuep_update(mfsr(0) & ~SR_NX);
+}