[S390] Use gmap translation for accessing guest memory

This patch removes kvm-s390 internal assumption of a linear mapping
of guest address space to user space. Previously, guest memory was
translated to user addresses using a fixed offset (gmsor). The new
code uses gmap_fault to resolve guest addresses.

Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/kvm_host.h

@@ -93,9 +93,7 @@ struct kvm_s390_sie_block {
 	__u32	scaol;			/* 0x0064 */
 	__u8	reserved68[4];		/* 0x0068 */
 	__u32	todpr;			/* 0x006c */
-	__u8	reserved70[16];		/* 0x0070 */
-	__u64	gmsor;			/* 0x0080 */
-	__u64	gmslm;			/* 0x0088 */
+	__u8	reserved70[32];		/* 0x0070 */
 	psw_t	gpsw;			/* 0x0090 */
 	__u64	gg14;			/* 0x00a0 */
 	__u64	gg15;			/* 0x00a8 */

arch/s390/kvm/gaccess.h

@@ -1,5 +1,5 @@
 /*
- * gaccess.h -  access guest memory
+ * access.h -  access guest memory
  *
  * Copyright IBM Corp. 2008,2009
  *
@@ -22,20 +22,13 @@ static inline void __user *__guestaddr_to_user(struct kvm_vcpu *vcpu,
 					       unsigned long guestaddr)
 {
 	unsigned long prefix  = vcpu->arch.sie_block->prefix;
-	unsigned long origin  = vcpu->arch.sie_block->gmsor;
-	unsigned long memsize = kvm_s390_vcpu_get_memsize(vcpu);
 
 	if (guestaddr < 2 * PAGE_SIZE)
 		guestaddr += prefix;
 	else if ((guestaddr >= prefix) && (guestaddr < prefix + 2 * PAGE_SIZE))
 		guestaddr -= prefix;
 
-	if (guestaddr > memsize)
-		return (void __user __force *) ERR_PTR(-EFAULT);
-
-	guestaddr += origin;
-
-	return (void __user *) guestaddr;
+	return (void __user *) gmap_fault(guestaddr, vcpu->arch.gmap);
 }
 
 static inline int get_guest_u64(struct kvm_vcpu *vcpu, unsigned long guestaddr,
@@ -141,11 +134,11 @@ static inline int put_guest_u8(struct kvm_vcpu *vcpu, unsigned long guestaddr,
 
 static inline int __copy_to_guest_slow(struct kvm_vcpu *vcpu,
 				       unsigned long guestdest,
-				       const void *from, unsigned long n)
+				       void *from, unsigned long n)
 {
 	int rc;
 	unsigned long i;
-	const u8 *data = from;
+	u8 *data = from;
 
 	for (i = 0; i < n; i++) {
 		rc = put_guest_u8(vcpu, guestdest++, *(data++));
@@ -155,12 +148,95 @@ static inline int __copy_to_guest_slow(struct kvm_vcpu *vcpu,
 	return 0;
 }
 
+static inline int __copy_to_guest_fast(struct kvm_vcpu *vcpu,
+				       unsigned long guestdest,
+				       void *from, unsigned long n)
+{
+	int r;
+	void __user *uptr;
+	unsigned long size;
+
+	if (guestdest + n < guestdest)
+		return -EFAULT;
+
+	/* simple case: all within one segment table entry? */
+	if ((guestdest & PMD_MASK) == ((guestdest+n) & PMD_MASK)) {
+		uptr = (void __user *) gmap_fault(guestdest, vcpu->arch.gmap);
+
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
+
+		r = copy_to_user(uptr, from, n);
+
+		if (r)
+			r = -EFAULT;
+
+		goto out;
+	}
+
+	/* copy first segment */
+	uptr = (void __user *)gmap_fault(guestdest, vcpu->arch.gmap);
+
+	if (IS_ERR((void __force *) uptr))
+		return PTR_ERR((void __force *) uptr);
+
+	size = PMD_SIZE - (guestdest & ~PMD_MASK);
+
+	r = copy_to_user(uptr, from, size);
+
+	if (r) {
+		r = -EFAULT;
+		goto out;
+	}
+	from += size;
+	n -= size;
+	guestdest += size;
+
+	/* copy full segments */
+	while (n >= PMD_SIZE) {
+		uptr = (void __user *)gmap_fault(guestdest, vcpu->arch.gmap);
+
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
+
+		r = copy_to_user(uptr, from, PMD_SIZE);
+
+		if (r) {
+			r = -EFAULT;
+			goto out;
+		}
+		from += PMD_SIZE;
+		n -= PMD_SIZE;
+		guestdest += PMD_SIZE;
+	}
+
+	/* copy the tail segment */
+	if (n) {
+		uptr = (void __user *)gmap_fault(guestdest, vcpu->arch.gmap);
+
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
+
+		r = copy_to_user(uptr, from, n);
+
+		if (r)
+			r = -EFAULT;
+	}
+out:
+	return r;
+}
+
+static inline int copy_to_guest_absolute(struct kvm_vcpu *vcpu,
+					 unsigned long guestdest,
+					 void *from, unsigned long n)
+{
+	return __copy_to_guest_fast(vcpu, guestdest, from, n);
+}
+
 static inline int copy_to_guest(struct kvm_vcpu *vcpu, unsigned long guestdest,
-				const void *from, unsigned long n)
+				void *from, unsigned long n)
 {
 	unsigned long prefix  = vcpu->arch.sie_block->prefix;
-	unsigned long origin  = vcpu->arch.sie_block->gmsor;
-	unsigned long memsize = kvm_s390_vcpu_get_memsize(vcpu);
 
 	if ((guestdest < 2 * PAGE_SIZE) && (guestdest + n > 2 * PAGE_SIZE))
 		goto slowpath;
@@ -177,15 +253,7 @@ static inline int copy_to_guest(struct kvm_vcpu *vcpu, unsigned long guestdest,
 	else if ((guestdest >= prefix) && (guestdest < prefix + 2 * PAGE_SIZE))
 		guestdest -= prefix;
 
-	if (guestdest + n > memsize)
-		return -EFAULT;
-
-	if (guestdest + n < guestdest)
-		return -EFAULT;
-
-	guestdest += origin;
-
-	return copy_to_user((void __user *) guestdest, from, n);
+	return __copy_to_guest_fast(vcpu, guestdest, from, n);
 slowpath:
 	return __copy_to_guest_slow(vcpu, guestdest, from, n);
 }
@@ -206,74 +274,113 @@ static inline int __copy_from_guest_slow(struct kvm_vcpu *vcpu, void *to,
 	return 0;
 }
 
-static inline int copy_from_guest(struct kvm_vcpu *vcpu, void *to,
-				  unsigned long guestsrc, unsigned long n)
+static inline int __copy_from_guest_fast(struct kvm_vcpu *vcpu, void *to,
+					 unsigned long guestsrc,
+					 unsigned long n)
 {
-	unsigned long prefix  = vcpu->arch.sie_block->prefix;
-	unsigned long origin  = vcpu->arch.sie_block->gmsor;
-	unsigned long memsize = kvm_s390_vcpu_get_memsize(vcpu);
+	int r;
+	void __user *uptr;
+	unsigned long size;
 
-	if ((guestsrc < 2 * PAGE_SIZE) && (guestsrc + n > 2 * PAGE_SIZE))
-		goto slowpath;
+	if (guestsrc + n < guestsrc)
+		return -EFAULT;
 
-	if ((guestsrc < prefix) && (guestsrc + n > prefix))
-		goto slowpath;
+	/* simple case: all within one segment table entry? */
+	if ((guestsrc & PMD_MASK) == ((guestsrc+n) & PMD_MASK)) {
+		uptr = (void __user *) gmap_fault(guestsrc, vcpu->arch.gmap);
 
-	if ((guestsrc < prefix + 2 * PAGE_SIZE)
-	    && (guestsrc + n > prefix + 2 * PAGE_SIZE))
-		goto slowpath;
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
 
-	if (guestsrc < 2 * PAGE_SIZE)
-		guestsrc += prefix;
-	else if ((guestsrc >= prefix) && (guestsrc < prefix + 2 * PAGE_SIZE))
-		guestsrc -= prefix;
+		r = copy_from_user(to, uptr, n);
 
-	if (guestsrc + n > memsize)
-		return -EFAULT;
+		if (r)
+			r = -EFAULT;
 
-	if (guestsrc + n < guestsrc)
-		return -EFAULT;
+		goto out;
+	}
 
-	guestsrc += origin;
+	/* copy first segment */
+	uptr = (void __user *)gmap_fault(guestsrc, vcpu->arch.gmap);
 
-	return copy_from_user(to, (void __user *) guestsrc, n);
-slowpath:
-	return __copy_from_guest_slow(vcpu, to, guestsrc, n);
-}
+	if (IS_ERR((void __force *) uptr))
+		return PTR_ERR((void __force *) uptr);
 
-static inline int copy_to_guest_absolute(struct kvm_vcpu *vcpu,
-					 unsigned long guestdest,
-					 const void *from, unsigned long n)
-{
-	unsigned long origin  = vcpu->arch.sie_block->gmsor;
-	unsigned long memsize = kvm_s390_vcpu_get_memsize(vcpu);
+	size = PMD_SIZE - (guestsrc & ~PMD_MASK);
 
-	if (guestdest + n > memsize)
-		return -EFAULT;
+	r = copy_from_user(to, uptr, size);
 
-	if (guestdest + n < guestdest)
-		return -EFAULT;
+	if (r) {
+		r = -EFAULT;
+		goto out;
+	}
+	to += size;
+	n -= size;
+	guestsrc += size;
+
+	/* copy full segments */
+	while (n >= PMD_SIZE) {
+		uptr = (void __user *)gmap_fault(guestsrc, vcpu->arch.gmap);
+
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
+
+		r = copy_from_user(to, uptr, PMD_SIZE);
+
+		if (r) {
+			r = -EFAULT;
+			goto out;
+		}
+		to += PMD_SIZE;
+		n -= PMD_SIZE;
+		guestsrc += PMD_SIZE;
+	}
+
+	/* copy the tail segment */
+	if (n) {
+		uptr = (void __user *)gmap_fault(guestsrc, vcpu->arch.gmap);
 
-	guestdest += origin;
+		if (IS_ERR((void __force *) uptr))
+			return PTR_ERR((void __force *) uptr);
 
-	return copy_to_user((void __user *) guestdest, from, n);
+		r = copy_from_user(to, uptr, n);
+
+		if (r)
+			r = -EFAULT;
+	}
+out:
+	return r;
 }
 
 static inline int copy_from_guest_absolute(struct kvm_vcpu *vcpu, void *to,
 					   unsigned long guestsrc,
 					   unsigned long n)
 {
-	unsigned long origin  = vcpu->arch.sie_block->gmsor;
-	unsigned long memsize = kvm_s390_vcpu_get_memsize(vcpu);
+	return __copy_from_guest_fast(vcpu, to, guestsrc, n);
+}
 
-	if (guestsrc + n > memsize)
-		return -EFAULT;
+static inline int copy_from_guest(struct kvm_vcpu *vcpu, void *to,
+				  unsigned long guestsrc, unsigned long n)
+{
+	unsigned long prefix  = vcpu->arch.sie_block->prefix;
 
-	if (guestsrc + n < guestsrc)
-		return -EFAULT;
+	if ((guestsrc < 2 * PAGE_SIZE) && (guestsrc + n > 2 * PAGE_SIZE))
+		goto slowpath;
 
-	guestsrc += origin;
+	if ((guestsrc < prefix) && (guestsrc + n > prefix))
+		goto slowpath;
+
+	if ((guestsrc < prefix + 2 * PAGE_SIZE)
+	    && (guestsrc + n > prefix + 2 * PAGE_SIZE))
+		goto slowpath;
+
+	if (guestsrc < 2 * PAGE_SIZE)
+		guestsrc += prefix;
+	else if ((guestsrc >= prefix) && (guestsrc < prefix + 2 * PAGE_SIZE))
+		guestsrc -= prefix;
 
-	return copy_from_user(to, (void __user *) guestsrc, n);
+	return __copy_from_guest_fast(vcpu, to, guestsrc, n);
+slowpath:
+	return __copy_from_guest_slow(vcpu, to, guestsrc, n);
 }
 #endif

arch/s390/kvm/intercept.c

@@ -165,26 +165,30 @@ static int handle_validity(struct kvm_vcpu *vcpu)
 	int rc;
 
 	vcpu->stat.exit_validity++;
-	if ((viwhy == 0x37) && (vcpu->arch.sie_block->prefix
-		<= kvm_s390_vcpu_get_memsize(vcpu) - 2*PAGE_SIZE)) {
-		rc = fault_in_pages_writeable((char __user *)
-			 vcpu->arch.sie_block->gmsor +
-			 vcpu->arch.sie_block->prefix,
-			 2*PAGE_SIZE);
+	if (viwhy == 0x37) {
+		vmaddr = gmap_fault(vcpu->arch.sie_block->prefix,
+				    vcpu->arch.gmap);
+		if (IS_ERR_VALUE(vmaddr)) {
+			rc = -EOPNOTSUPP;
+			goto out;
+		}
+		rc = fault_in_pages_writeable((char __user *) vmaddr,
+			 PAGE_SIZE);
 		if (rc) {
 			/* user will receive sigsegv, exit to user */
 			rc = -EOPNOTSUPP;
 			goto out;
 		}
-		vmaddr = gmap_fault(vcpu->arch.sie_block->prefix,
+		vmaddr = gmap_fault(vcpu->arch.sie_block->prefix + PAGE_SIZE,
 				    vcpu->arch.gmap);
 		if (IS_ERR_VALUE(vmaddr)) {
 			rc = -EOPNOTSUPP;
 			goto out;
 		}
-		vmaddr = gmap_fault(vcpu->arch.sie_block->prefix + PAGE_SIZE,
-				    vcpu->arch.gmap);
-		if (IS_ERR_VALUE(vmaddr)) {
+		rc = fault_in_pages_writeable((char __user *) vmaddr,
+			 PAGE_SIZE);
+		if (rc) {
+			/* user will receive sigsegv, exit to user */
 			rc = -EOPNOTSUPP;
 			goto out;
 		}

arch/s390/kvm/kvm-s390.c

@@ -549,7 +549,7 @@ rerun_vcpu:
 	return rc;
 }
 
-static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, const void *from,
+static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, void *from,
 		       unsigned long n, int prefix)
 {
 	if (prefix)
@@ -566,7 +566,7 @@ static int __guestcopy(struct kvm_vcpu *vcpu, u64 guestdest, const void *from,
  */
 int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
 {
-	const unsigned char archmode = 1;
+	unsigned char archmode = 1;
 	int prefix;
 
 	if (addr == KVM_S390_STORE_STATUS_NOADDR) {

arch/s390/kvm/kvm-s390.h

@@ -58,31 +58,8 @@ int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
 int kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code);
 int kvm_s390_inject_sigp_stop(struct kvm_vcpu *vcpu, int action);
 
-static inline long kvm_s390_vcpu_get_memsize(struct kvm_vcpu *vcpu)
-{
-	return vcpu->arch.sie_block->gmslm
-		- vcpu->arch.sie_block->gmsor
-		- VIRTIODESCSPACE + 1ul;
-}
-
 static inline void kvm_s390_vcpu_set_mem(struct kvm_vcpu *vcpu)
 {
-	int idx;
-	struct kvm_memory_slot *mem;
-	struct kvm_memslots *memslots;
-
-	idx = srcu_read_lock(&vcpu->kvm->srcu);
-	memslots = kvm_memslots(vcpu->kvm);
-
-	mem = &memslots->memslots[0];
-
-	vcpu->arch.sie_block->gmsor = mem->userspace_addr;
-	vcpu->arch.sie_block->gmslm =
-		mem->userspace_addr +
-		(mem->npages << PAGE_SHIFT) +
-		VIRTIODESCSPACE - 1ul;
-
-	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 }
 
 /* implemented in priv.c */

arch/s390/kvm/sigp.c

@@ -189,10 +189,8 @@ static int __sigp_set_prefix(struct kvm_vcpu *vcpu, u16 cpu_addr, u32 address,
 
 	/* make sure that the new value is valid memory */
 	address = address & 0x7fffe000u;
-	if ((copy_from_user(&tmp, (void __user *)
-		(address + vcpu->arch.sie_block->gmsor) , 1)) ||
-	   (copy_from_user(&tmp, (void __user *)(address +
-			vcpu->arch.sie_block->gmsor + PAGE_SIZE), 1))) {
+	if (copy_from_guest_absolute(vcpu, &tmp, address, 1) ||
+	   copy_from_guest_absolute(vcpu, &tmp, address + PAGE_SIZE, 1)) {
 		*reg |= SIGP_STAT_INVALID_PARAMETER;
 		return 1; /* invalid parameter */
 	}