diff --git a/arch/sparc/crypto/aes_asm.S b/arch/sparc/crypto/aes_asm.S
index 23f6cbb..1cda8aa 100644
--- a/arch/sparc/crypto/aes_asm.S
+++ b/arch/sparc/crypto/aes_asm.S
@@ -1024,7 +1024,11 @@
 	 add		%o2, 0x20, %o2
 	brlz,pt		%o3, 11f
 	 nop
-10:	ldx		[%o1 + 0x00], %g3
+10:	ldd		[%o0 + 0xd0], %f56
+	ldd		[%o0 + 0xd8], %f58
+	ldd		[%o0 + 0xe0], %f60
+	ldd		[%o0 + 0xe8], %f62
+	ldx		[%o1 + 0x00], %g3
 	ldx		[%o1 + 0x08], %g7
 	xor		%g1, %g3, %g3
 	xor		%g2, %g7, %g7
@@ -1128,9 +1132,9 @@
 	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len */
 	ldx		[%o0 - 0x10], %g1
 	subcc		%o3, 0x10, %o3
+	ldx		[%o0 - 0x08], %g2
 	be		10f
-	 ldx		[%o0 - 0x08], %g2
-	sub		%o0, 0xf0, %o0
+	 sub		%o0, 0xf0, %o0
 1:	ldx		[%o1 + 0x00], %g3
 	ldx		[%o1 + 0x08], %g7
 	ldx		[%o1 + 0x10], %o4
@@ -1154,7 +1158,11 @@
 	 add		%o2, 0x20, %o2
 	brlz,pt		%o3, 11f
 	 nop
-10:	ldx		[%o1 + 0x00], %g3
+10:	ldd		[%o0 + 0x18], %f56
+	ldd		[%o0 + 0x10], %f58
+	ldd		[%o0 + 0x08], %f60
+	ldd		[%o0 + 0x00], %f62
+	ldx		[%o1 + 0x00], %g3
 	ldx		[%o1 + 0x08], %g7
 	xor		%g1, %g3, %g3
 	xor		%g2, %g7, %g7
@@ -1511,11 +1519,11 @@
 	 add		%o2, 0x20, %o2
 	brlz,pt		%o3, 11f
 	 nop
-	ldd		[%o0 + 0xd0], %f56
+10:	ldd		[%o0 + 0xd0], %f56
 	ldd		[%o0 + 0xd8], %f58
 	ldd		[%o0 + 0xe0], %f60
 	ldd		[%o0 + 0xe8], %f62
-10:	xor		%g1, %g3, %o5
+	xor		%g1, %g3, %o5
 	MOVXTOD_O5_F0
 	xor		%g2, %g7, %o5
 	MOVXTOD_O5_F2
diff --git a/arch/sparc/crypto/aes_glue.c b/arch/sparc/crypto/aes_glue.c
index 3965d1d..503e6d9 100644
--- a/arch/sparc/crypto/aes_glue.c
+++ b/arch/sparc/crypto/aes_glue.c
@@ -222,6 +222,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	ctx->ops->load_encrypt_keys(&ctx->key[0]);
 	while ((nbytes = walk.nbytes)) {
@@ -251,6 +252,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	ctx->ops->load_decrypt_keys(&ctx->key[0]);
 	key_end = &ctx->key[ctx->expanded_key_length / sizeof(u64)];
@@ -280,6 +282,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	ctx->ops->load_encrypt_keys(&ctx->key[0]);
 	while ((nbytes = walk.nbytes)) {
@@ -309,6 +312,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	ctx->ops->load_decrypt_keys(&ctx->key[0]);
 	key_end = &ctx->key[ctx->expanded_key_length / sizeof(u64)];
@@ -329,6 +333,22 @@
 	return err;
 }
 
+static void ctr_crypt_final(struct crypto_sparc64_aes_ctx *ctx,
+			    struct blkcipher_walk *walk)
+{
+	u8 *ctrblk = walk->iv;
+	u64 keystream[AES_BLOCK_SIZE / sizeof(u64)];
+	u8 *src = walk->src.virt.addr;
+	u8 *dst = walk->dst.virt.addr;
+	unsigned int nbytes = walk->nbytes;
+
+	ctx->ops->ecb_encrypt(&ctx->key[0], (const u64 *)ctrblk,
+			      keystream, AES_BLOCK_SIZE);
+	crypto_xor((u8 *) keystream, src, nbytes);
+	memcpy(dst, keystream, nbytes);
+	crypto_inc(ctrblk, AES_BLOCK_SIZE);
+}
+
 static int ctr_crypt(struct blkcipher_desc *desc,
 		     struct scatterlist *dst, struct scatterlist *src,
 		     unsigned int nbytes)
@@ -338,10 +358,11 @@
 	int err;
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
-	err = blkcipher_walk_virt(desc, &walk);
+	err = blkcipher_walk_virt_block(desc, &walk, AES_BLOCK_SIZE);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	ctx->ops->load_encrypt_keys(&ctx->key[0]);
-	while ((nbytes = walk.nbytes)) {
+	while ((nbytes = walk.nbytes) >= AES_BLOCK_SIZE) {
 		unsigned int block_len = nbytes & AES_BLOCK_MASK;
 
 		if (likely(block_len)) {
@@ -353,6 +374,10 @@
 		nbytes &= AES_BLOCK_SIZE - 1;
 		err = blkcipher_walk_done(desc, &walk, nbytes);
 	}
+	if (walk.nbytes) {
+		ctr_crypt_final(ctx, &walk);
+		err = blkcipher_walk_done(desc, &walk, 0);
+	}
 	fprs_write(0);
 	return err;
 }
@@ -418,7 +443,7 @@
 	.cra_driver_name	= "ctr-aes-sparc64",
 	.cra_priority		= SPARC_CR_OPCODE_PRIORITY,
 	.cra_flags		= CRYPTO_ALG_TYPE_BLKCIPHER,
-	.cra_blocksize		= AES_BLOCK_SIZE,
+	.cra_blocksize		= 1,
 	.cra_ctxsize		= sizeof(struct crypto_sparc64_aes_ctx),
 	.cra_alignmask		= 7,
 	.cra_type		= &crypto_blkcipher_type,
diff --git a/arch/sparc/crypto/camellia_glue.c b/arch/sparc/crypto/camellia_glue.c
index 62c89af..888f6260 100644
--- a/arch/sparc/crypto/camellia_glue.c
+++ b/arch/sparc/crypto/camellia_glue.c
@@ -98,6 +98,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	if (encrypt)
 		key = &ctx->encrypt_key[0];
@@ -160,6 +161,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	key = &ctx->encrypt_key[0];
 	camellia_sparc64_load_keys(key, ctx->key_len);
@@ -198,6 +200,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	key = &ctx->decrypt_key[0];
 	camellia_sparc64_load_keys(key, ctx->key_len);
diff --git a/arch/sparc/crypto/des_asm.S b/arch/sparc/crypto/des_asm.S
index 30b6e90..b5c8fc2 100644
--- a/arch/sparc/crypto/des_asm.S
+++ b/arch/sparc/crypto/des_asm.S
@@ -376,6 +376,7 @@
 1:	ldd	[%o1 + 0x00], %f60
 	DES3_LOOP_BODY(60)
 	std	%f60, [%o2 + 0x00]
+	add	%o1, 0x08, %o1
 	subcc	%o3, 0x08, %o3
 	bne,pt	%icc, 1b
 	 add	%o2, 0x08, %o2
diff --git a/arch/sparc/crypto/des_glue.c b/arch/sparc/crypto/des_glue.c
index 41524ce..3065bc6 100644
--- a/arch/sparc/crypto/des_glue.c
+++ b/arch/sparc/crypto/des_glue.c
@@ -100,6 +100,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	if (encrypt)
 		des_sparc64_load_keys(&ctx->encrypt_expkey[0]);
@@ -147,6 +148,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	des_sparc64_load_keys(&ctx->encrypt_expkey[0]);
 	while ((nbytes = walk.nbytes)) {
@@ -177,6 +179,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	des_sparc64_load_keys(&ctx->decrypt_expkey[0]);
 	while ((nbytes = walk.nbytes)) {
@@ -266,6 +269,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	if (encrypt)
 		K = &ctx->encrypt_expkey[0];
@@ -317,6 +321,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	K = &ctx->encrypt_expkey[0];
 	des3_ede_sparc64_load_keys(K);
@@ -352,6 +357,7 @@
 
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
+	desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
 
 	K = &ctx->decrypt_expkey[0];
 	des3_ede_sparc64_load_keys(K);
diff --git a/arch/sparc/include/asm/hugetlb.h b/arch/sparc/include/asm/hugetlb.h
index 8c5eed6..9661e9b 100644
--- a/arch/sparc/include/asm/hugetlb.h
+++ b/arch/sparc/include/asm/hugetlb.h
@@ -61,14 +61,20 @@
 static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
 					   unsigned long addr, pte_t *ptep)
 {
-	ptep_set_wrprotect(mm, addr, ptep);
+	pte_t old_pte = *ptep;
+	set_huge_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
 }
 
 static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma,
 					     unsigned long addr, pte_t *ptep,
 					     pte_t pte, int dirty)
 {
-	return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
+	int changed = !pte_same(*ptep, pte);
+	if (changed) {
+		set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+		flush_tlb_page(vma, addr);
+	}
+	return changed;
 }
 
 static inline pte_t huge_ptep_get(pte_t *ptep)
diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
index 95515f1..7870be0 100644
--- a/arch/sparc/include/asm/pgtable_64.h
+++ b/arch/sparc/include/asm/pgtable_64.h
@@ -617,6 +617,12 @@
 	return val;
 }
 
+#define pte_accessible pte_accessible
+static inline unsigned long pte_accessible(pte_t a)
+{
+	return pte_val(a) & _PAGE_VALID;
+}
+
 static inline unsigned long pte_special(pte_t pte)
 {
 	return pte_val(pte) & _PAGE_SPECIAL;
@@ -802,7 +808,7 @@
 	 * SUN4V NOTE: _PAGE_VALID is the same value in both the SUN4U
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
-	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
+	if (likely(mm != &init_mm) && pte_accessible(orig))
 		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }