diff mbox series

[2/3] s390/crypto: Rework on paes implementation

Message ID 20191113105523.8007-3-freude@linux.ibm.com (mailing list archive)
State Changes Requested
Delegated to: Herbert Xu
Headers show
Series provide paes selftests | expand

Commit Message

Harald Freudenberger Nov. 13, 2019, 10:55 a.m. UTC
There have been some findings during Eric Biggers rework of the
paes implementation which this patch tries to address:

A very minor finding within paes ctr where when the cpacf instruction
returns with only partially data en/decrytped the walk_done() was
mistakenly done with the all data counter.  Please note this can only
happen when the kmctr returns because the protected key became invalid
in the middle of the operation. And this is only with suspend and
resume on a system with different effective wrapping key.

Eric Biggers mentioned that the context struct within the tfm struct
may be shared among multiple kernel threads. So here now a rework
which uses a spinlock per context to protect the read and write of the
protected key blob value. The en/decrypt functions copy the protected
key(s) at the beginning into a param struct and do not work with the
protected key within the context any more. If the protected key in the
param struct becomes invalid, the key material is again converted to
protected key(s) and the context gets this update protected by the
spinlock. Race conditions are still possible and may result in writing
the very same protected key value more than once. So the spinlock
needs to make sure the protected key(s) within the context are
consistent updated.

The ctr page is now locked by a mutex instead of a spinlock. A similar
patch went into the aes_s390 code as a result of a complain "sleeping
function called from invalid context at ...algapi.h". See
commit 1c2c7029c008 ("s390/crypto: fix possible sleep during spinlock
aquired")' for more. The very same page is now allocated before
registration. As the selftest runs during registration the page needs
to be available before registration or the kernel would crash.

During testing with instrumented code another issue with the xts
en/decrypt function revealed. The retry cleared the running iv value
and thus let to wrong en/decrypted data.

Tested and verified with additional testcases via AF_ALG interface and
additional selftests within the kernel (which will be made available
as an additional patch to the crypto testmgr).

Reported-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
---
 arch/s390/crypto/paes_s390.c | 163 ++++++++++++++++++++++++++---------
 1 file changed, 120 insertions(+), 43 deletions(-)

Comments

Herbert Xu Nov. 22, 2019, 8:13 a.m. UTC | #1
On Wed, Nov 13, 2019 at 11:55:22AM +0100, Harald Freudenberger wrote:
>
> @@ -129,6 +128,7 @@ static int ecb_paes_init(struct crypto_skcipher *tfm)
>  	struct s390_paes_ctx *ctx = crypto_skcipher_ctx(tfm);
>  
>  	ctx->kb.key = NULL;
> +	spin_lock_init(&ctx->pk_lock);

This makes no sense.  The context is per-tfm, and each tfm should
have a single key at any time.  The synchronisation of setkey vs.
crypto operations is left to the user of the tfm, not the
implementor.

So why do you need this spin lock at all?

Cheers,
Harald Freudenberger Nov. 22, 2019, 9:54 a.m. UTC | #2
On 22.11.19 09:13, Herbert Xu wrote:
> On Wed, Nov 13, 2019 at 11:55:22AM +0100, Harald Freudenberger wrote:
>> @@ -129,6 +128,7 @@ static int ecb_paes_init(struct crypto_skcipher *tfm)
>>  	struct s390_paes_ctx *ctx = crypto_skcipher_ctx(tfm);
>>  
>>  	ctx->kb.key = NULL;
>> +	spin_lock_init(&ctx->pk_lock);
> This makes no sense.  The context is per-tfm, and each tfm should
> have a single key at any time.  The synchronisation of setkey vs.
> crypto operations is left to the user of the tfm, not the
> implementor.
>
> So why do you need this spin lock at all?
>
> Cheers,
The setkey() sets the base key material (usually a secure key) to an
tfm instance. From this key a 'protected key' (pkey) is derived which
may get invalid at any time and may need to get re-derived from the
base key material.
An tfm instance may be shared, so the context where the pkey is
stored into is also shared. So when a pkey gets invalid there is a need
to update the pkey value within the context struct. This update needs
to be done atomic as another thread may concurrently use this pkey
value. That's all what this spinlock does. Make sure read and write
operations on the pkey within the context are atomic.
It is still possible that two threads copy the pkey, try to use it, find out
that it is invalid and needs refresh, re-derive and both update the pkey
memory serialized by the spinlock. But this is no issue. The spinlock
makes sure the stored pkey is always a consistent pkey (which may
be valid or invalid but not corrupted).

Hope this makes it clear

Thanks
Herbert Xu Nov. 22, 2019, 10:42 a.m. UTC | #3
On Fri, Nov 22, 2019 at 10:54:50AM +0100, Harald Freudenberger wrote:
> The setkey() sets the base key material (usually a secure key) to an
> tfm instance. From this key a 'protected key' (pkey) is derived which
> may get invalid at any time and may need to get re-derived from the
> base key material.
> An tfm instance may be shared, so the context where the pkey is
> stored into is also shared. So when a pkey gets invalid there is a need
> to update the pkey value within the context struct. This update needs
> to be done atomic as another thread may concurrently use this pkey
> value. That's all what this spinlock does. Make sure read and write
> operations on the pkey within the context are atomic.
> It is still possible that two threads copy the pkey, try to use it, find out
> that it is invalid and needs refresh, re-derive and both update the pkey
> memory serialized by the spinlock. But this is no issue. The spinlock
> makes sure the stored pkey is always a consistent pkey (which may
> be valid or invalid but not corrupted).

OK.  Can you give me a bit more background info on how often
this is likely to happen? I mean it happened every time you
might as well not store the protected key in the tfm at all.

Thanks,
Harald Freudenberger Nov. 22, 2019, 1:38 p.m. UTC | #4
On 22.11.19 11:42, Herbert Xu wrote:
> On Fri, Nov 22, 2019 at 10:54:50AM +0100, Harald Freudenberger wrote:
>> The setkey() sets the base key material (usually a secure key) to an
>> tfm instance. From this key a 'protected key' (pkey) is derived which
>> may get invalid at any time and may need to get re-derived from the
>> base key material.
>> An tfm instance may be shared, so the context where the pkey is
>> stored into is also shared. So when a pkey gets invalid there is a need
>> to update the pkey value within the context struct. This update needs
>> to be done atomic as another thread may concurrently use this pkey
>> value. That's all what this spinlock does. Make sure read and write
>> operations on the pkey within the context are atomic.
>> It is still possible that two threads copy the pkey, try to use it, find out
>> that it is invalid and needs refresh, re-derive and both update the pkey
>> memory serialized by the spinlock. But this is no issue. The spinlock
>> makes sure the stored pkey is always a consistent pkey (which may
>> be valid or invalid but not corrupted).
> OK.  Can you give me a bit more background info on how often
> this is likely to happen? I mean it happened every time you
> might as well not store the protected key in the tfm at all.
>
> Thanks,
The pkey is in fact a encrypted key + a verification pattern for the
encrypted key used. It gets invalid when this encryption key changes.
The encryption key changes when the LPAR is re-activated so for
example on suspend/resume or an Linux running as kvm guest
gets relocated. So this happens very rarely.
Herbert Xu Nov. 22, 2019, 2:07 p.m. UTC | #5
On Fri, Nov 22, 2019 at 02:38:30PM +0100, Harald Freudenberger wrote:
>
> The pkey is in fact a encrypted key + a verification pattern for the
> encrypted key used. It gets invalid when this encryption key changes.
> The encryption key changes when the LPAR is re-activated so for
> example on suspend/resume or an Linux running as kvm guest
> gets relocated. So this happens very rarely.

I see.  Is there any way of you finding out that the key has been
invalidated apart from trying out the crypto and having it fail?

Ideally you'd have a global counter that gets incremented everytime
an invalidation occurs.  You can then regenerate your key if its
generation counter differs from the current global counter.

Also when the crypto fails due to an invalid key you're currently
calling skcipher_walk_done with zero.  This is wrong as the done
function must be called with a positive value or an error.  In
some cases this can cause a crash in scatterwalk.

IOW you should just repeat the crypto operation after regenerating
the key rather than looping around again.

Cheers,
Harald Freudenberger Nov. 22, 2019, 2:45 p.m. UTC | #6
On 22.11.19 15:07, Herbert Xu wrote:
> On Fri, Nov 22, 2019 at 02:38:30PM +0100, Harald Freudenberger wrote:
>> The pkey is in fact a encrypted key + a verification pattern for the
>> encrypted key used. It gets invalid when this encryption key changes.
>> The encryption key changes when the LPAR is re-activated so for
>> example on suspend/resume or an Linux running as kvm guest
>> gets relocated. So this happens very rarely.
> I see.  Is there any way of you finding out that the key has been
> invalidated apart from trying out the crypto and having it fail?
No. By using the pkey for a crypto operation the hardware
checks the verification pattern and if there is a mismatch
it simple rejects the operation. Theoretically such an operation
can only partly complete and then a pkey could get invalid.
I have no way to check if the pkey is still valid before the
cpacf instruction call.
>
> Ideally you'd have a global counter that gets incremented everytime
> an invalidation occurs.  You can then regenerate your key if its
> generation counter differs from the current global counter.
>
> Also when the crypto fails due to an invalid key you're currently
> calling skcipher_walk_done with zero.  This is wrong as the done
> function must be called with a positive value or an error.  In
> some cases this can cause a crash in scatterwalk.
>
> IOW you should just repeat the crypto operation after regenerating
> the key rather than looping around again.
That's right. I'll try to rework the functions this way to
avoid calling skciper_walk_done with 0.

Thanks
>
> Cheers,
diff mbox series

Patch

diff --git a/arch/s390/crypto/paes_s390.c b/arch/s390/crypto/paes_s390.c
index c7119c617b6e..c1f1640a7c12 100644
--- a/arch/s390/crypto/paes_s390.c
+++ b/arch/s390/crypto/paes_s390.c
@@ -20,6 +20,7 @@ 
 #include <linux/module.h>
 #include <linux/cpufeature.h>
 #include <linux/init.h>
+#include <linux/mutex.h>
 #include <linux/spinlock.h>
 #include <crypto/internal/skcipher.h>
 #include <crypto/xts.h>
@@ -32,11 +33,11 @@ 
  * is called. As paes can handle different kinds of key blobs
  * and padding is also possible, the limits need to be generous.
  */
-#define PAES_MIN_KEYSIZE 64
+#define PAES_MIN_KEYSIZE 32
 #define PAES_MAX_KEYSIZE 256
 
 static u8 *ctrblk;
-static DEFINE_SPINLOCK(ctrblk_lock);
+static DEFINE_MUTEX(ctrblk_lock);
 
 static cpacf_mask_t km_functions, kmc_functions, kmctr_functions;
 
@@ -82,17 +83,19 @@  static inline void _free_kb_keybuf(struct key_blob *kb)
 struct s390_paes_ctx {
 	struct key_blob kb;
 	struct pkey_protkey pk;
+	spinlock_t pk_lock;
 	unsigned long fc;
 };
 
 struct s390_pxts_ctx {
 	struct key_blob kb[2];
 	struct pkey_protkey pk[2];
+	spinlock_t pk_lock;
 	unsigned long fc;
 };
 
-static inline int __paes_convert_key(struct key_blob *kb,
-				     struct pkey_protkey *pk)
+static inline int __paes_keyblob2pkey(struct key_blob *kb,
+				      struct pkey_protkey *pk)
 {
 	int i, ret;
 
@@ -106,22 +109,18 @@  static inline int __paes_convert_key(struct key_blob *kb,
 	return ret;
 }
 
-static int __paes_set_key(struct s390_paes_ctx *ctx)
+static inline int __paes_convert_key(struct s390_paes_ctx *ctx)
 {
-	unsigned long fc;
+	struct pkey_protkey pkey;
 
-	if (__paes_convert_key(&ctx->kb, &ctx->pk))
+	if (__paes_keyblob2pkey(&ctx->kb, &pkey))
 		return -EINVAL;
 
-	/* Pick the correct function code based on the protected key type */
-	fc = (ctx->pk.type == PKEY_KEYTYPE_AES_128) ? CPACF_KM_PAES_128 :
-		(ctx->pk.type == PKEY_KEYTYPE_AES_192) ? CPACF_KM_PAES_192 :
-		(ctx->pk.type == PKEY_KEYTYPE_AES_256) ? CPACF_KM_PAES_256 : 0;
+	spin_lock_bh(&ctx->pk_lock);
+	memcpy(&ctx->pk, &pkey, sizeof(pkey));
+	spin_unlock_bh(&ctx->pk_lock);
 
-	/* Check if the function code is available */
-	ctx->fc = (fc && cpacf_test_func(&km_functions, fc)) ? fc : 0;
-
-	return ctx->fc ? 0 : -EINVAL;
+	return 0;
 }
 
 static int ecb_paes_init(struct crypto_skcipher *tfm)
@@ -129,6 +128,7 @@  static int ecb_paes_init(struct crypto_skcipher *tfm)
 	struct s390_paes_ctx *ctx = crypto_skcipher_ctx(tfm);
 
 	ctx->kb.key = NULL;
+	spin_lock_init(&ctx->pk_lock);
 
 	return 0;
 }
@@ -140,6 +140,24 @@  static void ecb_paes_exit(struct crypto_skcipher *tfm)
 	_free_kb_keybuf(&ctx->kb);
 }
 
+static inline int __ecb_paes_set_key(struct s390_paes_ctx *ctx)
+{
+	unsigned long fc;
+
+	if (__paes_convert_key(ctx))
+		return -EINVAL;
+
+	/* Pick the correct function code based on the protected key type */
+	fc = (ctx->pk.type == PKEY_KEYTYPE_AES_128) ? CPACF_KM_PAES_128 :
+		(ctx->pk.type == PKEY_KEYTYPE_AES_192) ? CPACF_KM_PAES_192 :
+		(ctx->pk.type == PKEY_KEYTYPE_AES_256) ? CPACF_KM_PAES_256 : 0;
+
+	/* Check if the function code is available */
+	ctx->fc = (fc && cpacf_test_func(&km_functions, fc)) ? fc : 0;
+
+	return ctx->fc ? 0 : -EINVAL;
+}
+
 static int ecb_paes_set_key(struct crypto_skcipher *tfm, const u8 *in_key,
 			    unsigned int key_len)
 {
@@ -151,7 +169,7 @@  static int ecb_paes_set_key(struct crypto_skcipher *tfm, const u8 *in_key,
 	if (rc)
 		return rc;
 
-	if (__paes_set_key(ctx)) {
+	if (__ecb_paes_set_key(ctx)) {
 		crypto_skcipher_set_flags(tfm, CRYPTO_TFM_RES_BAD_KEY_LEN);
 		return -EINVAL;
 	}
@@ -165,18 +183,31 @@  static int ecb_paes_crypt(struct skcipher_request *req, unsigned long modifier)
 	struct skcipher_walk walk;
 	unsigned int nbytes, n, k;
 	int ret;
+	struct {
+		u8 key[MAXPROTKEYSIZE];
+	} param;
 
 	ret = skcipher_walk_virt(&walk, req, false);
+	if (ret)
+		return ret;
+
+	spin_lock_bh(&ctx->pk_lock);
+	memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+	spin_unlock_bh(&ctx->pk_lock);
+
 	while ((nbytes = walk.nbytes) != 0) {
 		/* only use complete blocks */
 		n = nbytes & ~(AES_BLOCK_SIZE - 1);
-		k = cpacf_km(ctx->fc | modifier, ctx->pk.protkey,
+		k = cpacf_km(ctx->fc | modifier, &param,
 			     walk.dst.virt.addr, walk.src.virt.addr, n);
 		if (k)
 			ret = skcipher_walk_done(&walk, nbytes - k);
 		if (k < n) {
-			if (__paes_set_key(ctx) != 0)
+			if (__paes_convert_key(ctx))
 				return skcipher_walk_done(&walk, -EIO);
+			spin_lock_bh(&ctx->pk_lock);
+			memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+			spin_unlock_bh(&ctx->pk_lock);
 		}
 	}
 	return ret;
@@ -214,6 +245,7 @@  static int cbc_paes_init(struct crypto_skcipher *tfm)
 	struct s390_paes_ctx *ctx = crypto_skcipher_ctx(tfm);
 
 	ctx->kb.key = NULL;
+	spin_lock_init(&ctx->pk_lock);
 
 	return 0;
 }
@@ -225,11 +257,11 @@  static void cbc_paes_exit(struct crypto_skcipher *tfm)
 	_free_kb_keybuf(&ctx->kb);
 }
 
-static int __cbc_paes_set_key(struct s390_paes_ctx *ctx)
+static inline int __cbc_paes_set_key(struct s390_paes_ctx *ctx)
 {
 	unsigned long fc;
 
-	if (__paes_convert_key(&ctx->kb, &ctx->pk))
+	if (__paes_convert_key(ctx))
 		return -EINVAL;
 
 	/* Pick the correct function code based on the protected key type */
@@ -276,8 +308,12 @@  static int cbc_paes_crypt(struct skcipher_request *req, unsigned long modifier)
 	ret = skcipher_walk_virt(&walk, req, false);
 	if (ret)
 		return ret;
+
 	memcpy(param.iv, walk.iv, AES_BLOCK_SIZE);
+	spin_lock_bh(&ctx->pk_lock);
 	memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+	spin_unlock_bh(&ctx->pk_lock);
+
 	while ((nbytes = walk.nbytes) != 0) {
 		/* only use complete blocks */
 		n = nbytes & ~(AES_BLOCK_SIZE - 1);
@@ -288,9 +324,11 @@  static int cbc_paes_crypt(struct skcipher_request *req, unsigned long modifier)
 			ret = skcipher_walk_done(&walk, nbytes - k);
 		}
 		if (k < n) {
-			if (__cbc_paes_set_key(ctx) != 0)
+			if (__paes_convert_key(ctx))
 				return skcipher_walk_done(&walk, -EIO);
+			spin_lock_bh(&ctx->pk_lock);
 			memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+			spin_unlock_bh(&ctx->pk_lock);
 		}
 	}
 	return ret;
@@ -330,6 +368,7 @@  static int xts_paes_init(struct crypto_skcipher *tfm)
 
 	ctx->kb[0].key = NULL;
 	ctx->kb[1].key = NULL;
+	spin_lock_init(&ctx->pk_lock);
 
 	return 0;
 }
@@ -342,12 +381,27 @@  static void xts_paes_exit(struct crypto_skcipher *tfm)
 	_free_kb_keybuf(&ctx->kb[1]);
 }
 
-static int __xts_paes_set_key(struct s390_pxts_ctx *ctx)
+static inline int __xts_paes_convert_key(struct s390_pxts_ctx *ctx)
+{
+	struct pkey_protkey pkey0, pkey1;
+
+	if (__paes_keyblob2pkey(&ctx->kb[0], &pkey0) ||
+	    __paes_keyblob2pkey(&ctx->kb[1], &pkey1))
+		return -EINVAL;
+
+	spin_lock_bh(&ctx->pk_lock);
+	memcpy(&ctx->pk[0], &pkey0, sizeof(pkey0));
+	memcpy(&ctx->pk[1], &pkey1, sizeof(pkey1));
+	spin_unlock_bh(&ctx->pk_lock);
+
+	return 0;
+}
+
+static inline int __xts_paes_set_key(struct s390_pxts_ctx *ctx)
 {
 	unsigned long fc;
 
-	if (__paes_convert_key(&ctx->kb[0], &ctx->pk[0]) ||
-	    __paes_convert_key(&ctx->kb[1], &ctx->pk[1]))
+	if (__xts_paes_convert_key(ctx))
 		return -EINVAL;
 
 	if (ctx->pk[0].type != ctx->pk[1].type)
@@ -425,15 +479,17 @@  static int xts_paes_crypt(struct skcipher_request *req, unsigned long modifier)
 	ret = skcipher_walk_virt(&walk, req, false);
 	if (ret)
 		return ret;
+
 	keylen = (ctx->pk[0].type == PKEY_KEYTYPE_AES_128) ? 48 : 64;
 	offset = (ctx->pk[0].type == PKEY_KEYTYPE_AES_128) ? 16 : 0;
-retry:
+
 	memset(&pcc_param, 0, sizeof(pcc_param));
 	memcpy(pcc_param.tweak, walk.iv, sizeof(pcc_param.tweak));
+	spin_lock_bh(&ctx->pk_lock);
 	memcpy(pcc_param.key + offset, ctx->pk[1].protkey, keylen);
-	cpacf_pcc(ctx->fc, pcc_param.key + offset);
-
 	memcpy(xts_param.key + offset, ctx->pk[0].protkey, keylen);
+	spin_unlock_bh(&ctx->pk_lock);
+	cpacf_pcc(ctx->fc, pcc_param.key + offset);
 	memcpy(xts_param.init, pcc_param.xts, 16);
 
 	while ((nbytes = walk.nbytes) != 0) {
@@ -444,11 +500,15 @@  static int xts_paes_crypt(struct skcipher_request *req, unsigned long modifier)
 		if (k)
 			ret = skcipher_walk_done(&walk, nbytes - k);
 		if (k < n) {
-			if (__xts_paes_set_key(ctx) != 0)
+			if (__xts_paes_convert_key(ctx))
 				return skcipher_walk_done(&walk, -EIO);
-			goto retry;
+			spin_lock_bh(&ctx->pk_lock);
+			memcpy(xts_param.key + offset,
+			       ctx->pk[0].protkey, keylen);
+			spin_unlock_bh(&ctx->pk_lock);
 		}
 	}
+
 	return ret;
 }
 
@@ -485,6 +545,7 @@  static int ctr_paes_init(struct crypto_skcipher *tfm)
 	struct s390_paes_ctx *ctx = crypto_skcipher_ctx(tfm);
 
 	ctx->kb.key = NULL;
+	spin_lock_init(&ctx->pk_lock);
 
 	return 0;
 }
@@ -496,11 +557,11 @@  static void ctr_paes_exit(struct crypto_skcipher *tfm)
 	_free_kb_keybuf(&ctx->kb);
 }
 
-static int __ctr_paes_set_key(struct s390_paes_ctx *ctx)
+static inline int __ctr_paes_set_key(struct s390_paes_ctx *ctx)
 {
 	unsigned long fc;
 
-	if (__paes_convert_key(&ctx->kb, &ctx->pk))
+	if (__paes_convert_key(ctx))
 		return -EINVAL;
 
 	/* Pick the correct function code based on the protected key type */
@@ -556,45 +617,61 @@  static int ctr_paes_crypt(struct skcipher_request *req)
 	struct skcipher_walk walk;
 	unsigned int nbytes, n, k;
 	int ret, locked;
-
-	locked = spin_trylock(&ctrblk_lock);
+	struct {
+		u8 key[MAXPROTKEYSIZE];
+	} param;
 
 	ret = skcipher_walk_virt(&walk, req, false);
+	if (ret)
+		return ret;
+
+	spin_lock_bh(&ctx->pk_lock);
+	memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+	spin_unlock_bh(&ctx->pk_lock);
+
+	locked = mutex_trylock(&ctrblk_lock);
+
 	while ((nbytes = walk.nbytes) >= AES_BLOCK_SIZE) {
 		n = AES_BLOCK_SIZE;
 		if (nbytes >= 2*AES_BLOCK_SIZE && locked)
 			n = __ctrblk_init(ctrblk, walk.iv, nbytes);
 		ctrptr = (n > AES_BLOCK_SIZE) ? ctrblk : walk.iv;
-		k = cpacf_kmctr(ctx->fc, ctx->pk.protkey, walk.dst.virt.addr,
+		k = cpacf_kmctr(ctx->fc, &param, walk.dst.virt.addr,
 				walk.src.virt.addr, n, ctrptr);
 		if (k) {
 			if (ctrptr == ctrblk)
 				memcpy(walk.iv, ctrptr + k - AES_BLOCK_SIZE,
 				       AES_BLOCK_SIZE);
 			crypto_inc(walk.iv, AES_BLOCK_SIZE);
-			ret = skcipher_walk_done(&walk, nbytes - n);
+			ret = skcipher_walk_done(&walk, nbytes - k);
 		}
 		if (k < n) {
-			if (__ctr_paes_set_key(ctx) != 0) {
+			if (__paes_convert_key(ctx)) {
 				if (locked)
-					spin_unlock(&ctrblk_lock);
+					mutex_unlock(&ctrblk_lock);
 				return skcipher_walk_done(&walk, -EIO);
 			}
+			spin_lock_bh(&ctx->pk_lock);
+			memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+			spin_unlock_bh(&ctx->pk_lock);
 		}
 	}
 	if (locked)
-		spin_unlock(&ctrblk_lock);
+		mutex_unlock(&ctrblk_lock);
 	/*
 	 * final block may be < AES_BLOCK_SIZE, copy only nbytes
 	 */
 	if (nbytes) {
 		while (1) {
-			if (cpacf_kmctr(ctx->fc, ctx->pk.protkey, buf,
+			if (cpacf_kmctr(ctx->fc, &param, buf,
 					walk.src.virt.addr, AES_BLOCK_SIZE,
 					walk.iv) == AES_BLOCK_SIZE)
 				break;
-			if (__ctr_paes_set_key(ctx) != 0)
+			if (__paes_convert_key(ctx))
 				return skcipher_walk_done(&walk, -EIO);
+			spin_lock_bh(&ctx->pk_lock);
+			memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
+			spin_unlock_bh(&ctx->pk_lock);
 		}
 		memcpy(walk.dst.virt.addr, buf, nbytes);
 		crypto_inc(walk.iv, AES_BLOCK_SIZE);
@@ -674,14 +751,14 @@  static int __init paes_s390_init(void)
 	if (cpacf_test_func(&kmctr_functions, CPACF_KMCTR_PAES_128) ||
 	    cpacf_test_func(&kmctr_functions, CPACF_KMCTR_PAES_192) ||
 	    cpacf_test_func(&kmctr_functions, CPACF_KMCTR_PAES_256)) {
-		ret = crypto_register_skcipher(&ctr_paes_alg);
-		if (ret)
-			goto out_err;
 		ctrblk = (u8 *) __get_free_page(GFP_KERNEL);
 		if (!ctrblk) {
 			ret = -ENOMEM;
 			goto out_err;
 		}
+		ret = crypto_register_skcipher(&ctr_paes_alg);
+		if (ret)
+			goto out_err;
 	}
 
 	return 0;