Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull s390 updates from Martin Schwidefsky:
 "The new features and main improvements in this merge for v4.9

   - Support for the UBSAN sanitizer

   - Set HAVE_EFFICIENT_UNALIGNED_ACCESS, it improves the code in some
     places

   - Improvements for the in-kernel fpu code, in particular the overhead
     for multiple consecutive in kernel fpu users is recuded

   - Add a SIMD implementation for the RAID6 gen and xor operations

   - Add RAID6 recovery based on the XC instruction

   - The PCI DMA flush logic has been improved to increase the speed of
     the map / unmap operations

   - The time synchronization code has seen some updates

  And bug fixes all over the place"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (48 commits)
  s390/con3270: fix insufficient space padding
  s390/con3270: fix use of uninitialised data
  MAINTAINERS: update DASD maintainer
  s390/cio: fix accidental interrupt enabling during resume
  s390/dasd: add missing \n to end of dev_err messages
  s390/config: Enable config options for Docker
  s390/dasd: make query host access interruptible
  s390/dasd: fix panic during offline processing
  s390/dasd: fix hanging offline processing
  s390/pci_dma: improve lazy flush for unmap
  s390/pci_dma: split dma_update_trans
  s390/pci_dma: improve map_sg
  s390/pci_dma: simplify dma address calculation
  s390/pci_dma: remove dma address range check
  iommu/s390: simplify registration of I/O address translation parameters
  s390: migrate exception table users off module.h and onto extable.h
  s390: export header for CLP ioctl
  s390/vmur: fix irq pointer dereference in int handler
  s390/dasd: add missing KOBJ_CHANGE event for unformatted devices
  s390: enable UBSAN
  ...

11 files changed, 293 insertions, 246 deletions

diff --git a/arch/s390/include/asm/cpacf.h b/arch/s390/include/asm/cpacf.h
index d28621de8e0b..2c680db7e5c1 100644
--- a/arch/s390/include/asm/cpacf.h
+++ b/arch/s390/include/asm/cpacf.h
@@ -28,67 +28,51 @@
 #define CPACF_PPNO		0xb93c		/* MSA5 */
 
 /*
- * Function codes for the KM (CIPHER MESSAGE)
- * instruction (0x80 is the decipher modifier bit)
+ * Decryption modifier bit
+ */
+#define CPACF_DECRYPT		0x80
+
+/*
+ * Function codes for the KM (CIPHER MESSAGE) instruction
  */
 #define CPACF_KM_QUERY		0x00
-#define CPACF_KM_DEA_ENC	0x01
-#define CPACF_KM_DEA_DEC	0x81
-#define CPACF_KM_TDEA_128_ENC	0x02
-#define CPACF_KM_TDEA_128_DEC	0x82
-#define CPACF_KM_TDEA_192_ENC	0x03
-#define CPACF_KM_TDEA_192_DEC	0x83
-#define CPACF_KM_AES_128_ENC	0x12
-#define CPACF_KM_AES_128_DEC	0x92
-#define CPACF_KM_AES_192_ENC	0x13
-#define CPACF_KM_AES_192_DEC	0x93
-#define CPACF_KM_AES_256_ENC	0x14
-#define CPACF_KM_AES_256_DEC	0x94
-#define CPACF_KM_XTS_128_ENC	0x32
-#define CPACF_KM_XTS_128_DEC	0xb2
-#define CPACF_KM_XTS_256_ENC	0x34
-#define CPACF_KM_XTS_256_DEC	0xb4
+#define CPACF_KM_DEA		0x01
+#define CPACF_KM_TDEA_128	0x02
+#define CPACF_KM_TDEA_192	0x03
+#define CPACF_KM_AES_128	0x12
+#define CPACF_KM_AES_192	0x13
+#define CPACF_KM_AES_256	0x14
+#define CPACF_KM_XTS_128	0x32
+#define CPACF_KM_XTS_256	0x34
 
 /*
  * Function codes for the KMC (CIPHER MESSAGE WITH CHAINING)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
 #define CPACF_KMC_QUERY		0x00
-#define CPACF_KMC_DEA_ENC	0x01
-#define CPACF_KMC_DEA_DEC	0x81
-#define CPACF_KMC_TDEA_128_ENC	0x02
-#define CPACF_KMC_TDEA_128_DEC	0x82
-#define CPACF_KMC_TDEA_192_ENC	0x03
-#define CPACF_KMC_TDEA_192_DEC	0x83
-#define CPACF_KMC_AES_128_ENC	0x12
-#define CPACF_KMC_AES_128_DEC	0x92
-#define CPACF_KMC_AES_192_ENC	0x13
-#define CPACF_KMC_AES_192_DEC	0x93
-#define CPACF_KMC_AES_256_ENC	0x14
-#define CPACF_KMC_AES_256_DEC	0x94
+#define CPACF_KMC_DEA		0x01
+#define CPACF_KMC_TDEA_128	0x02
+#define CPACF_KMC_TDEA_192	0x03
+#define CPACF_KMC_AES_128	0x12
+#define CPACF_KMC_AES_192	0x13
+#define CPACF_KMC_AES_256	0x14
 #define CPACF_KMC_PRNG		0x43
 
 /*
  * Function codes for the KMCTR (CIPHER MESSAGE WITH COUNTER)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
-#define CPACF_KMCTR_QUERY	 0x00
-#define CPACF_KMCTR_DEA_ENC	 0x01
-#define CPACF_KMCTR_DEA_DEC	 0x81
-#define CPACF_KMCTR_TDEA_128_ENC 0x02
-#define CPACF_KMCTR_TDEA_128_DEC 0x82
-#define CPACF_KMCTR_TDEA_192_ENC 0x03
-#define CPACF_KMCTR_TDEA_192_DEC 0x83
-#define CPACF_KMCTR_AES_128_ENC	 0x12
-#define CPACF_KMCTR_AES_128_DEC	 0x92
-#define CPACF_KMCTR_AES_192_ENC	 0x13
-#define CPACF_KMCTR_AES_192_DEC	 0x93
-#define CPACF_KMCTR_AES_256_ENC	 0x14
-#define CPACF_KMCTR_AES_256_DEC	 0x94
+#define CPACF_KMCTR_QUERY	0x00
+#define CPACF_KMCTR_DEA		0x01
+#define CPACF_KMCTR_TDEA_128	0x02
+#define CPACF_KMCTR_TDEA_192	0x03
+#define CPACF_KMCTR_AES_128	0x12
+#define CPACF_KMCTR_AES_192	0x13
+#define CPACF_KMCTR_AES_256	0x14
 
 /*
  * Function codes for the KIMD (COMPUTE INTERMEDIATE MESSAGE DIGEST)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
 #define CPACF_KIMD_QUERY	0x00
 #define CPACF_KIMD_SHA_1	0x01
@@ -98,7 +82,7 @@
 
 /*
  * Function codes for the KLMD (COMPUTE LAST MESSAGE DIGEST)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
 #define CPACF_KLMD_QUERY	0x00
 #define CPACF_KLMD_SHA_1	0x01
@@ -107,7 +91,7 @@
 
 /*
  * function codes for the KMAC (COMPUTE MESSAGE AUTHENTICATION CODE)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
 #define CPACF_KMAC_QUERY	0x00
 #define CPACF_KMAC_DEA		0x01
@@ -116,12 +100,14 @@
 
 /*
  * Function codes for the PPNO (PERFORM PSEUDORANDOM NUMBER OPERATION)
- * instruction (0x80 is the decipher modifier bit)
+ * instruction
  */
 #define CPACF_PPNO_QUERY		0x00
 #define CPACF_PPNO_SHA512_DRNG_GEN	0x03
 #define CPACF_PPNO_SHA512_DRNG_SEED	0x83
 
+typedef struct { unsigned char bytes[16]; } cpacf_mask_t;
+
 /**
  * cpacf_query() - check if a specific CPACF function is available
  * @opcode: the opcode of the crypto instruction
@@ -132,55 +118,66 @@
  *
  * Returns 1 if @func is available for @opcode, 0 otherwise
  */
-static inline void __cpacf_query(unsigned int opcode, unsigned char *status)
+static inline void __cpacf_query(unsigned int opcode, cpacf_mask_t *mask)
 {
-	typedef struct { unsigned char _[16]; } status_type;
 	register unsigned long r0 asm("0") = 0;	/* query function */
-	register unsigned long r1 asm("1") = (unsigned long) status;
+	register unsigned long r1 asm("1") = (unsigned long) mask;
 
 	asm volatile(
 		"	spm 0\n" /* pckmo doesn't change the cc */
 		/* Parameter registers are ignored, but may not be 0 */
 		"0:	.insn	rrf,%[opc] << 16,2,2,2,0\n"
 		"	brc	1,0b\n"	/* handle partial completion */
-		: "=m" (*(status_type *) status)
+		: "=m" (*mask)
 		: [fc] "d" (r0), [pba] "a" (r1), [opc] "i" (opcode)
 		: "cc");
 }
 
-static inline int cpacf_query(unsigned int opcode, unsigned int func)
+static inline int __cpacf_check_opcode(unsigned int opcode)
 {
-	unsigned char status[16];
-
 	switch (opcode) {
 	case CPACF_KMAC:
 	case CPACF_KM:
 	case CPACF_KMC:
 	case CPACF_KIMD:
 	case CPACF_KLMD:
-		if (!test_facility(17))	/* check for MSA */
-			return 0;
-		break;
+		return test_facility(17);	/* check for MSA */
 	case CPACF_PCKMO:
-		if (!test_facility(76))	/* check for MSA3 */
-			return 0;
-		break;
+		return test_facility(76);	/* check for MSA3 */
 	case CPACF_KMF:
 	case CPACF_KMO:
 	case CPACF_PCC:
 	case CPACF_KMCTR:
-		if (!test_facility(77))	/* check for MSA4 */
-			return 0;
-		break;
+		return test_facility(77);	/* check for MSA4 */
 	case CPACF_PPNO:
-		if (!test_facility(57))	/* check for MSA5 */
-			return 0;
-		break;
+		return test_facility(57);	/* check for MSA5 */
 	default:
 		BUG();
 	}
-	__cpacf_query(opcode, status);
-	return (status[func >> 3] & (0x80 >> (func & 7))) != 0;
+}
+
+static inline int cpacf_query(unsigned int opcode, cpacf_mask_t *mask)
+{
+	if (__cpacf_check_opcode(opcode)) {
+		__cpacf_query(opcode, mask);
+		return 1;
+	}
+	memset(mask, 0, sizeof(*mask));
+	return 0;
+}
+
+static inline int cpacf_test_func(cpacf_mask_t *mask, unsigned int func)
+{
+	return (mask->bytes[func >> 3] & (0x80 >> (func & 7))) != 0;
+}
+
+static inline int cpacf_query_func(unsigned int opcode, unsigned int func)
+{
+	cpacf_mask_t mask;
+
+	if (cpacf_query(opcode, &mask))
+		return cpacf_test_func(&mask, func);
+	return 0;
 }
 
 /**
@@ -194,7 +191,7 @@ static inline int cpacf_query(unsigned int opcode, unsigned int func)
  * Returns 0 for the query func, number of processed bytes for
  * encryption/decryption funcs
  */
-static inline int cpacf_km(long func, void *param,
+static inline int cpacf_km(unsigned long func, void *param,
 			   u8 *dest, const u8 *src, long src_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
@@ -224,7 +221,7 @@ static inline int cpacf_km(long func, void *param,
  * Returns 0 for the query func, number of processed bytes for
  * encryption/decryption funcs
  */
-static inline int cpacf_kmc(long func, void *param,
+static inline int cpacf_kmc(unsigned long func, void *param,
 			    u8 *dest, const u8 *src, long src_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
@@ -250,11 +247,9 @@ static inline int cpacf_kmc(long func, void *param,
  * @param: address of parameter block; see POP for details on each func
  * @src: address of source memory area
  * @src_len: length of src operand in bytes
- *
- * Returns 0 for the query func, number of processed bytes for digest funcs
  */
-static inline int cpacf_kimd(long func, void *param,
-			     const u8 *src, long src_len)
+static inline void cpacf_kimd(unsigned long func, void *param,
+			      const u8 *src, long src_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
 	register unsigned long r1 asm("1") = (unsigned long) param;
@@ -267,8 +262,6 @@ static inline int cpacf_kimd(long func, void *param,
 		: [src] "+a" (r2), [len] "+d" (r3)
 		: [fc] "d" (r0), [pba] "a" (r1), [opc] "i" (CPACF_KIMD)
 		: "cc", "memory");
-
-	return src_len - r3;
 }
 
 /**
@@ -277,11 +270,9 @@ static inline int cpacf_kimd(long func, void *param,
  * @param: address of parameter block; see POP for details on each func
  * @src: address of source memory area
  * @src_len: length of src operand in bytes
- *
- * Returns 0 for the query func, number of processed bytes for digest funcs
  */
-static inline int cpacf_klmd(long func, void *param,
-			     const u8 *src, long src_len)
+static inline void cpacf_klmd(unsigned long func, void *param,
+			      const u8 *src, long src_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
 	register unsigned long r1 asm("1") = (unsigned long) param;
@@ -294,8 +285,6 @@ static inline int cpacf_klmd(long func, void *param,
 		: [src] "+a" (r2), [len] "+d" (r3)
 		: [fc] "d" (r0), [pba] "a" (r1), [opc] "i" (CPACF_KLMD)
 		: "cc", "memory");
-
-	return src_len - r3;
 }
 
 /**
@@ -308,7 +297,7 @@ static inline int cpacf_klmd(long func, void *param,
  *
  * Returns 0 for the query func, number of processed bytes for digest funcs
  */
-static inline int cpacf_kmac(long func, void *param,
+static inline int cpacf_kmac(unsigned long func, void *param,
 			     const u8 *src, long src_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
@@ -338,7 +327,7 @@ static inline int cpacf_kmac(long func, void *param,
  * Returns 0 for the query func, number of processed bytes for
  * encryption/decryption funcs
  */
-static inline int cpacf_kmctr(long func, void *param, u8 *dest,
+static inline int cpacf_kmctr(unsigned long func, void *param, u8 *dest,
 			      const u8 *src, long src_len, u8 *counter)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
@@ -368,13 +357,10 @@ static inline int cpacf_kmctr(long func, void *param, u8 *dest,
  * @dest_len: size of destination memory area in bytes
  * @seed: address of seed data
  * @seed_len: size of seed data in bytes
- *
- * Returns 0 for the query func, number of random bytes stored in
- * dest buffer for generate function
  */
-static inline int cpacf_ppno(long func, void *param,
-			     u8 *dest, long dest_len,
-			     const u8 *seed, long seed_len)
+static inline void cpacf_ppno(unsigned long func, void *param,
+			      u8 *dest, long dest_len,
+			      const u8 *seed, long seed_len)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
 	register unsigned long r1 asm("1") = (unsigned long) param;
@@ -390,8 +376,6 @@ static inline int cpacf_ppno(long func, void *param,
 		: [fc] "d" (r0), [pba] "a" (r1),
 		  [seed] "a" (r4), [slen] "d" (r5), [opc] "i" (CPACF_PPNO)
 		: "cc", "memory");
-
-	return dest_len - r3;
 }
 
 /**
@@ -399,10 +383,8 @@ static inline int cpacf_ppno(long func, void *param,
  *		 instruction
  * @func: the function code passed to PCC; see CPACF_KM_xxx defines
  * @param: address of parameter block; see POP for details on each func
- *
- * Returns 0.
  */
-static inline int cpacf_pcc(long func, void *param)
+static inline void cpacf_pcc(unsigned long func, void *param)
 {
 	register unsigned long r0 asm("0") = (unsigned long) func;
 	register unsigned long r1 asm("1") = (unsigned long) param;
@@ -413,8 +395,6 @@ static inline int cpacf_pcc(long func, void *param)
 		:
 		: [fc] "d" (r0), [pba] "a" (r1), [opc] "i" (CPACF_PCC)
 		: "cc", "memory");
-
-	return 0;
 }
 
 #endif	/* _ASM_S390_CPACF_H */
diff --git a/arch/s390/include/asm/facilities_src.h b/arch/s390/include/asm/facilities_src.h
index 4917728e5828..3b758f66e48b 100644
--- a/arch/s390/include/asm/facilities_src.h
+++ b/arch/s390/include/asm/facilities_src.h
@@ -55,4 +55,28 @@ static struct facility_def facility_defs[] = {
 			-1 /* END */
 		}
 	},
+	{
+		.name = "FACILITIES_KVM",
+		.bits = (int[]){
+			0,  /* N3 instructions */
+			1,  /* z/Arch mode installed */
+			2,  /* z/Arch mode active */
+			3,  /* DAT-enhancement */
+			4,  /* idte segment table */
+			5,  /* idte region table */
+			6,  /* ASN-and-LX reuse */
+			7,  /* stfle */
+			8,  /* enhanced-DAT 1 */
+			9,  /* sense-running-status */
+			10, /* conditional sske */
+			13, /* ipte-range */
+			14, /* nonquiescing key-setting */
+			73, /* transactional execution */
+			75, /* access-exception-fetch/store indication */
+			76, /* msa extension 3 */
+			77, /* msa extension 4 */
+			78, /* enhanced-DAT 2 */
+			-1  /* END */
+		}
+	},
 };
diff --git a/arch/s390/include/asm/fpu/api.h b/arch/s390/include/asm/fpu/api.h
index 6aba6fc406ad..02124d66bfb5 100644
--- a/arch/s390/include/asm/fpu/api.h
+++ b/arch/s390/include/asm/fpu/api.h
@@ -64,18 +64,18 @@ static inline int test_fp_ctl(u32 fpc)
 	return rc;
 }
 
-#define KERNEL_VXR_V0V7		1
-#define KERNEL_VXR_V8V15	2
-#define KERNEL_VXR_V16V23	4
-#define KERNEL_VXR_V24V31	8
-#define KERNEL_FPR		16
-#define KERNEL_FPC		256
+#define KERNEL_FPC		1
+#define KERNEL_VXR_V0V7		2
+#define KERNEL_VXR_V8V15	4
+#define KERNEL_VXR_V16V23	8
+#define KERNEL_VXR_V24V31	16
 
 #define KERNEL_VXR_LOW		(KERNEL_VXR_V0V7|KERNEL_VXR_V8V15)
 #define KERNEL_VXR_MID		(KERNEL_VXR_V8V15|KERNEL_VXR_V16V23)
 #define KERNEL_VXR_HIGH		(KERNEL_VXR_V16V23|KERNEL_VXR_V24V31)
 
-#define KERNEL_FPU_MASK		(KERNEL_VXR_LOW|KERNEL_VXR_HIGH|KERNEL_FPR)
+#define KERNEL_VXR		(KERNEL_VXR_LOW|KERNEL_VXR_HIGH)
+#define KERNEL_FPR		(KERNEL_FPC|KERNEL_VXR_V0V7)
 
 struct kernel_fpu;
 
@@ -87,18 +87,28 @@ struct kernel_fpu;
  * Prefer using the kernel_fpu_begin()/kernel_fpu_end() pair of functions.
  */
 void __kernel_fpu_begin(struct kernel_fpu *state, u32 flags);
-void __kernel_fpu_end(struct kernel_fpu *state);
+void __kernel_fpu_end(struct kernel_fpu *state, u32 flags);
 
 
 static inline void kernel_fpu_begin(struct kernel_fpu *state, u32 flags)
 {
 	preempt_disable();
-	__kernel_fpu_begin(state, flags);
+	state->mask = S390_lowcore.fpu_flags;
+	if (!test_cpu_flag(CIF_FPU))
+		/* Save user space FPU state and register contents */
+		save_fpu_regs();
+	else if (state->mask & flags)
+		/* Save FPU/vector register in-use by the kernel */
+		__kernel_fpu_begin(state, flags);
+	S390_lowcore.fpu_flags |= flags;
 }
 
-static inline void kernel_fpu_end(struct kernel_fpu *state)
+static inline void kernel_fpu_end(struct kernel_fpu *state, u32 flags)
 {
-	__kernel_fpu_end(state);
+	S390_lowcore.fpu_flags = state->mask;
+	if (state->mask & flags)
+		/* Restore FPU/vector register in-use by the kernel */
+		__kernel_fpu_end(state, flags);
 	preempt_enable();
 }
 
diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h
index d79ba7cf75b0..7b93b78f423c 100644
--- a/arch/s390/include/asm/lowcore.h
+++ b/arch/s390/include/asm/lowcore.h
@@ -129,7 +129,8 @@ struct lowcore {
 	__u8	pad_0x0390[0x0398-0x0390];	/* 0x0390 */
 	__u64	gmap;				/* 0x0398 */
 	__u32	spinlock_lockval;		/* 0x03a0 */
-	__u8	pad_0x03a0[0x0400-0x03a4];	/* 0x03a4 */
+	__u32	fpu_flags;			/* 0x03a4 */
+	__u8	pad_0x03a8[0x0400-0x03a8];	/* 0x03a8 */
 
 	/* Per cpu primary space access list */
 	__u32	paste[16];			/* 0x0400 */
diff --git a/arch/s390/include/asm/mmu.h b/arch/s390/include/asm/mmu.h
index 6d39329c894b..bea785d7f853 100644
--- a/arch/s390/include/asm/mmu.h
+++ b/arch/s390/include/asm/mmu.h
@@ -12,6 +12,7 @@ typedef struct {
 	struct list_head pgtable_list;
 	spinlock_t gmap_lock;
 	struct list_head gmap_list;
+	unsigned long gmap_asce;
 	unsigned long asce;
 	unsigned long asce_limit;
 	unsigned long vdso_base;
diff --git a/arch/s390/include/asm/mmu_context.h b/arch/s390/include/asm/mmu_context.h
index c6a088c91aee..515fea5a3fc4 100644
--- a/arch/s390/include/asm/mmu_context.h
+++ b/arch/s390/include/asm/mmu_context.h
@@ -21,6 +21,7 @@ static inline int init_new_context(struct task_struct *tsk,
 	INIT_LIST_HEAD(&mm->context.gmap_list);
 	cpumask_clear(&mm->context.cpu_attach_mask);
 	atomic_set(&mm->context.flush_count, 0);
+	mm->context.gmap_asce = 0;
 	mm->context.flush_mm = 0;
 #ifdef CONFIG_PGSTE
 	mm->context.alloc_pgste = page_table_allocate_pgste;
diff --git a/arch/s390/include/asm/pci.h b/arch/s390/include/asm/pci.h
index 0da91c4d30fd..6611f798d2be 100644
--- a/arch/s390/include/asm/pci.h
+++ b/arch/s390/include/asm/pci.h
@@ -11,6 +11,7 @@
 #include <asm-generic/pci.h>
 #include <asm/pci_clp.h>
 #include <asm/pci_debug.h>
+#include <asm/sclp.h>
 
 #define PCIBIOS_MIN_IO		0x1000
 #define PCIBIOS_MIN_MEM		0x10000000
@@ -117,6 +118,7 @@ struct zpci_dev {
 
 	spinlock_t	iommu_bitmap_lock;
 	unsigned long	*iommu_bitmap;
+	unsigned long	*lazy_bitmap;
 	unsigned long	iommu_size;
 	unsigned long	iommu_pages;
 	unsigned int	next_bit;
@@ -216,6 +218,9 @@ void zpci_debug_init_device(struct zpci_dev *, const char *);
 void zpci_debug_exit_device(struct zpci_dev *);
 void zpci_debug_info(struct zpci_dev *, struct seq_file *);
 
+/* Error reporting */
+int zpci_report_error(struct pci_dev *, struct zpci_report_error_header *);
+
 #ifdef CONFIG_NUMA
 
 /* Returns the node based on PCI bus */
diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index 72c7f60bfe83..0362cd5fa187 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -874,35 +874,31 @@ static inline pte_t pte_mkhuge(pte_t pte)
 }
 #endif
 
-static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
-{
-	unsigned long pto = (unsigned long) ptep;
-
-	/* Invalidation + global TLB flush for the pte */
-	asm volatile(
-		"	ipte	%2,%3"
-		: "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
-}
+#define IPTE_GLOBAL	0
+#define	IPTE_LOCAL	1
 
-static inline void __ptep_ipte_local(unsigned long address, pte_t *ptep)
+static inline void __ptep_ipte(unsigned long address, pte_t *ptep, int local)
 {
 	unsigned long pto = (unsigned long) ptep;
 
-	/* Invalidation + local TLB flush for the pte */
+	/* Invalidation + TLB flush for the pte */
 	asm volatile(
-		"	.insn rrf,0xb2210000,%2,%3,0,1"
-		: "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
+		"       .insn rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
+		: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
+		  [m4] "i" (local));
 }
 
-static inline void __ptep_ipte_range(unsigned long address, int nr, pte_t *ptep)
+static inline void __ptep_ipte_range(unsigned long address, int nr,
+				     pte_t *ptep, int local)
 {
 	unsigned long pto = (unsigned long) ptep;
 
-	/* Invalidate a range of ptes + global TLB flush of the ptes */
+	/* Invalidate a range of ptes + TLB flush of the ptes */
 	do {
 		asm volatile(
-			"	.insn rrf,0xb2210000,%2,%0,%1,0"
-			: "+a" (address), "+a" (nr) : "a" (pto) : "memory");
+			"       .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
+			: [r2] "+a" (address), [r3] "+a" (nr)
+			: [r1] "a" (pto), [m4] "i" (local) : "memory");
 	} while (nr != 255);
 }
 
@@ -1239,53 +1235,33 @@ static inline void __pmdp_csp(pmd_t *pmdp)
 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
 }
 
-static inline void __pmdp_idte(unsigned long address, pmd_t *pmdp)
-{
-	unsigned long sto;
-
-	sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
-	asm volatile(
-		"	.insn	rrf,0xb98e0000,%2,%3,0,0"
-		: "=m" (*pmdp)
-		: "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
-		: "cc" );
-}
-
-static inline void __pudp_idte(unsigned long address, pud_t *pudp)
-{
-	unsigned long r3o;
-
-	r3o = (unsigned long) pudp - pud_index(address) * sizeof(pud_t);
-	r3o |= _ASCE_TYPE_REGION3;
-	asm volatile(
-		"	.insn	rrf,0xb98e0000,%2,%3,0,0"
-		: "=m" (*pudp)
-		: "m" (*pudp), "a" (r3o), "a" ((address & PUD_MASK))
-		: "cc");
-}
+#define IDTE_GLOBAL	0
+#define IDTE_LOCAL	1
 
-static inline void __pmdp_idte_local(unsigned long address, pmd_t *pmdp)
+static inline void __pmdp_idte(unsigned long address, pmd_t *pmdp, int local)
 {
 	unsigned long sto;
 
 	sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
 	asm volatile(
-		"	.insn	rrf,0xb98e0000,%2,%3,0,1"
-		: "=m" (*pmdp)
-		: "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
+		"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
+		: "+m" (*pmdp)
+		: [r1] "a" (sto), [r2] "a" ((address & HPAGE_MASK)),
+		  [m4] "i" (local)
 		: "cc" );
 }
 
-static inline void __pudp_idte_local(unsigned long address, pud_t *pudp)
+static inline void __pudp_idte(unsigned long address, pud_t *pudp, int local)
 {
 	unsigned long r3o;
 
 	r3o = (unsigned long) pudp - pud_index(address) * sizeof(pud_t);
 	r3o |= _ASCE_TYPE_REGION3;
 	asm volatile(
-		"	.insn	rrf,0xb98e0000,%2,%3,0,1"
-		: "=m" (*pudp)
-		: "m" (*pudp), "a" (r3o), "a" ((address & PUD_MASK))
+		"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
+		: "+m" (*pudp)
+		: [r1] "a" (r3o), [r2] "a" ((address & PUD_MASK)),
+		  [m4] "i" (local)
 		: "cc");
 }
 
diff --git a/arch/s390/include/asm/tlbflush.h b/arch/s390/include/asm/tlbflush.h
index 1a691ef740cf..39846100682a 100644
--- a/arch/s390/include/asm/tlbflush.h
+++ b/arch/s390/include/asm/tlbflush.h
@@ -26,17 +26,6 @@ static inline void __tlb_flush_idte(unsigned long asce)
 		: : "a" (2048), "a" (asce) : "cc");
 }
 
-/*
- * Flush TLB entries for a specific ASCE on the local CPU
- */
-static inline void __tlb_flush_idte_local(unsigned long asce)
-{
-	/* Local TLB flush for the mm */
-	asm volatile(
-		"	.insn	rrf,0xb98e0000,0,%0,%1,1"
-		: : "a" (2048), "a" (asce) : "cc");
-}
-
 #ifdef CONFIG_SMP
 void smp_ptlb_all(void);
 
@@ -65,35 +54,33 @@ static inline void __tlb_flush_full(struct mm_struct *mm)
 		/* Global TLB flush */
 		__tlb_flush_global();
 		/* Reset TLB flush mask */
-		if (MACHINE_HAS_TLB_LC)
-			cpumask_copy(mm_cpumask(mm),
-				     &mm->context.cpu_attach_mask);
+		cpumask_copy(mm_cpumask(mm), &mm->context.cpu_attach_mask);
 	}
 	atomic_dec(&mm->context.flush_count);
 	preempt_enable();
 }
 
-/*
- * Flush TLB entries for a specific ASCE on all CPUs. Should never be used
- * when more than one asce (e.g. gmap) ran on this mm.
- */
-static inline void __tlb_flush_asce(struct mm_struct *mm, unsigned long asce)
+static inline void __tlb_flush_mm(struct mm_struct *mm)
 {
+	unsigned long gmap_asce;
+
+	/*
+	 * If the machine has IDTE we prefer to do a per mm flush
+	 * on all cpus instead of doing a local flush if the mm
+	 * only ran on the local cpu.
+	 */
 	preempt_disable();
 	atomic_inc(&mm->context.flush_count);
-	if (MACHINE_HAS_TLB_LC &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		__tlb_flush_idte_local(asce);
+	gmap_asce = READ_ONCE(mm->context.gmap_asce);
+	if (MACHINE_HAS_IDTE && gmap_asce != -1UL) {
+		if (gmap_asce)
+			__tlb_flush_idte(gmap_asce);
+		__tlb_flush_idte(mm->context.asce);
 	} else {
-		if (MACHINE_HAS_IDTE)
-			__tlb_flush_idte(asce);
-		else
-			__tlb_flush_global();
-		/* Reset TLB flush mask */
-		if (MACHINE_HAS_TLB_LC)
-			cpumask_copy(mm_cpumask(mm),
-				     &mm->context.cpu_attach_mask);
+		__tlb_flush_full(mm);
 	}
+	/* Reset TLB flush mask */
+	cpumask_copy(mm_cpumask(mm), &mm->context.cpu_attach_mask);
 	atomic_dec(&mm->context.flush_count);
 	preempt_enable();
 }
@@ -112,36 +99,17 @@ static inline void __tlb_flush_kernel(void)
 /*
  * Flush TLB entries for a specific ASCE on all CPUs.
  */
-static inline void __tlb_flush_asce(struct mm_struct *mm, unsigned long asce)
+static inline void __tlb_flush_mm(struct mm_struct *mm)
 {
-	if (MACHINE_HAS_TLB_LC)
-		__tlb_flush_idte_local(asce);
-	else
-		__tlb_flush_local();
+	__tlb_flush_local();
 }
 
 static inline void __tlb_flush_kernel(void)
 {
-	if (MACHINE_HAS_TLB_LC)
-		__tlb_flush_idte_local(init_mm.context.asce);
-	else
-		__tlb_flush_local();
+	__tlb_flush_local();
 }
 #endif
 
-static inline void __tlb_flush_mm(struct mm_struct * mm)
-{
-	/*
-	 * If the machine has IDTE we prefer to do a per mm flush
-	 * on all cpus instead of doing a local flush if the mm
-	 * only ran on the local cpu.
-	 */
-	if (MACHINE_HAS_IDTE && list_empty(&mm->context.gmap_list))
-		__tlb_flush_asce(mm, mm->context.asce);
-	else
-		__tlb_flush_full(mm);
-}
-
 static inline void __tlb_flush_mm_lazy(struct mm_struct * mm)
 {
 	if (mm->context.flush_mm) {
diff --git a/arch/s390/include/asm/vx-insn.h b/arch/s390/include/asm/vx-insn.h
index 4a3135620f5e..49c24a2afce0 100644
--- a/arch/s390/include/asm/vx-insn.h
+++ b/arch/s390/include/asm/vx-insn.h
@@ -16,15 +16,13 @@
 
 /* Macros to generate vector instruction byte code */
 
-#define REG_NUM_INVALID	       255
-
 /* GR_NUM - Retrieve general-purpose register number
  *
  * @opd:	Operand to store register number
  * @r64:	String designation register in the format "%rN"
  */
 .macro	GR_NUM	opd gr
-	\opd = REG_NUM_INVALID
+	\opd = 255
 	.ifc \gr,%r0
 		\opd = 0
 	.endif
@@ -73,14 +71,11 @@
 	.ifc \gr,%r15
 		\opd = 15
 	.endif
-	.if \opd == REG_NUM_INVALID
-		.error "Invalid general-purpose register designation: \gr"
+	.if \opd == 255
+		\opd = \gr
 	.endif
 .endm
 
-/* VX_R() - Macro to encode the VX_NUM into the instruction */
-#define VX_R(v)		(v & 0x0F)
-
 /* VX_NUM - Retrieve vector register number
  *
  * @opd:	Operand to store register number
@@ -88,11 +83,10 @@
  *
  * The vector register number is used for as input number to the
  * instruction and, as well as, to compute the RXB field of the
- * instruction.  To encode the particular vector register number,
- * use the VX_R(v) macro to extract the instruction opcode.
+ * instruction.
  */
 .macro	VX_NUM	opd vxr
-	\opd = REG_NUM_INVALID
+	\opd = 255
 	.ifc \vxr,%v0
 		\opd = 0
 	.endif
@@ -189,8 +183,8 @@
 	.ifc \vxr,%v31
 		\opd = 31
 	.endif
-	.if \opd == REG_NUM_INVALID
-		.error "Invalid vector register designation: \vxr"
+	.if \opd == 255
+		\opd = \vxr
 	.endif
 .endm
 
@@ -251,7 +245,7 @@
 /* VECTOR GENERATE BYTE MASK */
 .macro	VGBM	vr imm2
 	VX_NUM	v1, \vr
-	.word	(0xE700 | (VX_R(v1) << 4))
+	.word	(0xE700 | ((v1&15) << 4))
 	.word	\imm2
 	MRXBOPC	0, 0x44, v1
 .endm
@@ -267,7 +261,7 @@
 	VX_NUM	v1, \v
 	GR_NUM	b2, "%r0"
 	GR_NUM	r3, \gr
-	.word	0xE700 | (VX_R(v1) << 4) | r3
+	.word	0xE700 | ((v1&15) << 4) | r3
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC	\m, 0x22, v1
 .endm
@@ -284,12 +278,21 @@
 	VLVG	\v, \gr, \index, 3
 .endm
 
+/* VECTOR LOAD REGISTER */
+.macro	VLR	v1, v2
+	VX_NUM	v1, \v1
+	VX_NUM	v2, \v2
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	0
+	MRXBOPC	0, 0x56, v1, v2
+.endm
+
 /* VECTOR LOAD */
 .macro	VL	v, disp, index="%r0", base
 	VX_NUM	v1, \v
 	GR_NUM	x2, \index
 	GR_NUM	b2, \base
-	.word	0xE700 | (VX_R(v1) << 4) | x2
+	.word	0xE700 | ((v1&15) << 4) | x2
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC 0, 0x06, v1
 .endm
@@ -299,7 +302,7 @@
 	VX_NUM	v1, \vr1
 	GR_NUM	x2, \index
 	GR_NUM	b2, \base
-	.word	0xE700 | (VX_R(v1) << 4) | x2
+	.word	0xE700 | ((v1&15) << 4) | x2
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC	\m3, \opc, v1
 .endm
@@ -319,7 +322,7 @@
 /* VECTOR LOAD ELEMENT IMMEDIATE */
 .macro	VLEIx	vr1, imm2, m3, opc
 	VX_NUM	v1, \vr1
-	.word	0xE700 | (VX_R(v1) << 4)
+	.word	0xE700 | ((v1&15) << 4)
 	.word	\imm2
 	MRXBOPC	\m3, \opc, v1
 .endm
@@ -341,7 +344,7 @@
 	GR_NUM	r1, \gr
 	GR_NUM	b2, \base
 	VX_NUM	v3, \vr
-	.word	0xE700 | (r1 << 4) | VX_R(v3)
+	.word	0xE700 | (r1 << 4) | (v3&15)
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC	\m, 0x21, v3
 .endm
@@ -363,7 +366,7 @@
 	VX_NUM	v1, \vfrom
 	VX_NUM	v3, \vto
 	GR_NUM	b2, \base	    /* Base register */
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v3)
+	.word	0xE700 | ((v1&15) << 4) | (v3&15)
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC	0, 0x36, v1, v3
 .endm
@@ -373,7 +376,7 @@
 	VX_NUM	v1, \vfrom
 	VX_NUM	v3, \vto
 	GR_NUM	b2, \base	    /* Base register */
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v3)
+	.word	0xE700 | ((v1&15) << 4) | (v3&15)
 	.word	(b2 << 12) | (\disp)
 	MRXBOPC	0, 0x3E, v1, v3
 .endm
@@ -384,16 +387,16 @@
 	VX_NUM	v2, \vr2
 	VX_NUM	v3, \vr3
 	VX_NUM	v4, \vr4
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
-	.word	(VX_R(v3) << 12)
-	MRXBOPC	VX_R(v4), 0x8C, v1, v2, v3, v4
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
+	MRXBOPC	(v4&15), 0x8C, v1, v2, v3, v4
 .endm
 
 /* VECTOR UNPACK LOGICAL LOW */
 .macro	VUPLL	vr1, vr2, m3
 	VX_NUM	v1, \vr1
 	VX_NUM	v2, \vr2
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
 	.word	0x0000
 	MRXBOPC	\m3, 0xD4, v1, v2
 .endm
@@ -410,13 +413,23 @@
 
 /* Vector integer instructions */
 
+/* VECTOR AND */
+.macro	VN	vr1, vr2, vr3
+	VX_NUM	v1, \vr1
+	VX_NUM	v2, \vr2
+	VX_NUM	v3, \vr3
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
+	MRXBOPC	0, 0x68, v1, v2, v3
+.endm
+
 /* VECTOR EXCLUSIVE OR */
 .macro	VX	vr1, vr2, vr3
 	VX_NUM	v1, \vr1
 	VX_NUM	v2, \vr2
 	VX_NUM	v3, \vr3
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
-	.word	(VX_R(v3) << 12)
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
 	MRXBOPC	0, 0x6D, v1, v2, v3
 .endm
 
@@ -425,8 +438,8 @@
 	VX_NUM	v1, \vr1
 	VX_NUM	v2, \vr2
 	VX_NUM	v3, \vr3
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
-	.word	(VX_R(v3) << 12)
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
 	MRXBOPC	\m4, 0xB4, v1, v2, v3
 .endm
 .macro	VGFMB	vr1, vr2, vr3
@@ -448,9 +461,9 @@
 	VX_NUM	v2, \vr2
 	VX_NUM	v3, \vr3
 	VX_NUM	v4, \vr4
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
-	.word	(VX_R(v3) << 12) | (\m5 << 8)
-	MRXBOPC	VX_R(v4), 0xBC, v1, v2, v3, v4
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12) | (\m5 << 8)
+	MRXBOPC	(v4&15), 0xBC, v1, v2, v3, v4
 .endm
 .macro	VGFMAB	vr1, vr2, vr3, vr4
 	VGFMA	\vr1, \vr2, \vr3, \vr4, 0
@@ -470,11 +483,78 @@
 	VX_NUM	v1, \vr1
 	VX_NUM	v2, \vr2
 	VX_NUM	v3, \vr3
-	.word	0xE700 | (VX_R(v1) << 4) | VX_R(v2)
-	.word	(VX_R(v3) << 12)
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
 	MRXBOPC	0, 0x7D, v1, v2, v3
 .endm
 
+/* VECTOR REPLICATE IMMEDIATE */
+.macro	VREPI	vr1, imm2, m3
+	VX_NUM	v1, \vr1
+	.word	0xE700 | ((v1&15) << 4)
+	.word	\imm2
+	MRXBOPC	\m3, 0x45, v1
+.endm
+.macro	VREPIB	vr1, imm2
+	VREPI	\vr1, \imm2, 0
+.endm
+.macro	VREPIH	vr1, imm2
+	VREPI	\vr1, \imm2, 1
+.endm
+.macro	VREPIF	vr1, imm2
+	VREPI	\vr1, \imm2, 2
+.endm
+.macro	VREPIG	vr1, imm2
+	VREP	\vr1, \imm2, 3
+.endm
+
+/* VECTOR ADD */
+.macro	VA	vr1, vr2, vr3, m4
+	VX_NUM	v1, \vr1
+	VX_NUM	v2, \vr2
+	VX_NUM	v3, \vr3
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
+	MRXBOPC	\m4, 0xF3, v1, v2, v3
+.endm
+.macro	VAB	vr1, vr2, vr3
+	VA	\vr1, \vr2, \vr3, 0
+.endm
+.macro	VAH	vr1, vr2, vr3
+	VA	\vr1, \vr2, \vr3, 1
+.endm
+.macro	VAF	vr1, vr2, vr3
+	VA	\vr1, \vr2, \vr3, 2
+.endm
+.macro	VAG	vr1, vr2, vr3
+	VA	\vr1, \vr2, \vr3, 3
+.endm
+.macro	VAQ	vr1, vr2, vr3
+	VA	\vr1, \vr2, \vr3, 4
+.endm
+
+/* VECTOR ELEMENT SHIFT RIGHT ARITHMETIC */
+.macro	VESRAV	vr1, vr2, vr3, m4
+	VX_NUM	v1, \vr1
+	VX_NUM	v2, \vr2
+	VX_NUM	v3, \vr3
+	.word	0xE700 | ((v1&15) << 4) | (v2&15)
+	.word	((v3&15) << 12)
+	MRXBOPC \m4, 0x7A, v1, v2, v3
+.endm
+
+.macro	VESRAVB	vr1, vr2, vr3
+	VESRAV	\vr1, \vr2, \vr3, 0
+.endm
+.macro	VESRAVH	vr1, vr2, vr3
+	VESRAV	\vr1, \vr2, \vr3, 1
+.endm
+.macro	VESRAVF	vr1, vr2, vr3
+	VESRAV	\vr1, \vr2, \vr3, 2
+.endm
+.macro	VESRAVG	vr1, vr2, vr3
+	VESRAV	\vr1, \vr2, \vr3, 3
+.endm
 
 #endif	/* __ASSEMBLY__ */
 #endif	/* __ASM_S390_VX_INSN_H */
diff --git a/arch/s390/include/uapi/asm/Kbuild b/arch/s390/include/uapi/asm/Kbuild
index 08fe6dad9026..cc44b09c25fc 100644
--- a/arch/s390/include/uapi/asm/Kbuild
+++ b/arch/s390/include/uapi/asm/Kbuild
@@ -6,6 +6,7 @@ header-y += bitsperlong.h
 header-y += byteorder.h
 header-y += chpid.h
 header-y += chsc.h
+header-y += clp.h
 header-y += cmb.h
 header-y += dasd.h
 header-y += debug.h