[PATCH 00/17] crypto: inside-secure - various improvements

[PATCH 00/17] crypto: inside-secure - various improvements

[Antoine Tenart]

Hi Herbert,

This series depends on the previous one, "crypto: inside-secure - set of
fixes"[1]. The series contains various improvements to the Inside Secure
SafeXcel driver, which are bundled in a single series to avoid any
conflict.

- Patches 1-4 are cosmetic fixes (typo, more comments and dead code removal).
- Patches 5-8 are small improvements, mostly to improve the invalidation
  path.
- Patches 9-15 improve the request processing in the driver, by moving
  the requests dequeing into a workqueue and the result handling into a
  threaded IRQ thread. Other small improvements are made in the same
  direction.
- Patches 16-17 introduce the support of another revision of this crypto
  engine, similar to the one already supported. This crypto engine can be
  found on Marvell boards. (The corresponding device tree patches will follow
  once this series lands in).

Thanks!
Antoine

[1] https://www.spinics.net/lists/linux-crypto/msg30313.html

Antoine Tenart (13):
  crypto: inside-secure - remove extra empty line
  crypto: inside-secure - fix typo in a comment
  crypto: inside-secure - remove useless memset
  crypto: inside-secure - comment the caching function
  crypto: inside-secure - move request dequeueing into a workqueue
  crypto: inside-secure - use threaded IRQs for result handling
  crypto: inside-secure - dequeue all requests at once
  crypto: inside-secure - increase the ring size
  crypto: inside-secure - acknowledge the result requests all at once
  crypto: inside-secure - handle more result requests when counter is
    full
  crypto: inside-secure - retry to proceed the request later on fail
  crypto: inside-secure - EIP97 support
  Documentation/bindings: crypto: document the SafeXcel EIP97 compatible

Ofer Heifetz (4):
  crypto: inside-secure - refrain from unneeded invalidations
  crypto: inside-secure - EBUSY is not an error on async request
  crypto: inside-secure - move cipher crypto mode to request context
  crypto: inside-secure - remove unused parameter in invalidate_cache

 .../bindings/crypto/inside-secure-safexcel.txt     |   3 +-
 drivers/crypto/inside-secure/safexcel.c            | 370 ++++++++++++++-------
 drivers/crypto/inside-secure/safexcel.h            | 173 +++++++---
 drivers/crypto/inside-secure/safexcel_cipher.c     |  53 +--
 drivers/crypto/inside-secure/safexcel_hash.c       |  69 ++--
 5 files changed, 449 insertions(+), 219 deletions(-)

-- 
2.14.3

[PATCH 01/17] crypto: inside-secure - remove extra empty line

[Antoine Tenart]

Cosmetic patch removing an extra empty line between header inclusions.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_hash.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index 0c5a5820b06e..d2d4551565fa 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -14,7 +14,6 @@
 #include <linux/dma-mapping.h>
 #include <linux/dmapool.h>
 
-
 #include "safexcel.h"
 
 struct safexcel_ahash_ctx {
-- 
2.14.3

[PATCH 02/17] crypto: inside-secure - fix typo in a comment

[Antoine Tenart]

Cosmetic patch fixing one typo in one of the driver's comments.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_hash.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index d2d4551565fa..d4afc070ff58 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -504,7 +504,7 @@ static int safexcel_ahash_cache(struct ahash_request *areq)
 		return areq->nbytes;
 	}
 
-	/* We could'nt cache all the data */
+	/* We couldn't cache all the data */
 	return -E2BIG;
 }
 
-- 
2.14.3

[PATCH 03/17] crypto: inside-secure - remove useless memset

[Antoine Tenart]

This patch removes an useless memset in the ahash_export function, as
the zeroed buffer will be entirely overridden the next line.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_hash.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index d4afc070ff58..ac922b91b2d7 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -624,7 +624,6 @@ static int safexcel_ahash_export(struct ahash_request *areq, void *out)
 	export->processed = req->processed;
 
 	memcpy(export->state, req->state, req->state_sz);
-	memset(export->cache, 0, crypto_ahash_blocksize(ahash));
 	memcpy(export->cache, req->cache, crypto_ahash_blocksize(ahash));
 
 	return 0;
-- 
2.14.3

[PATCH 04/17] crypto: inside-secure - comment the caching function

[Antoine Tenart]

Cosmetic patch adding a few comments to the ahash caching function to
understand easily what calculations are made in the functions; and how
the function is working.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_hash.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index ac922b91b2d7..e53e6b2331db 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -484,13 +484,23 @@ static int safexcel_ahash_exit_inv(struct crypto_tfm *tfm)
 	return 0;
 }
 
+/* safexcel_ahash_cache: cache data until at least one request can be sent to
+ * the engine, aka. when there is at least 1 block size in the pipe.
+ */
 static int safexcel_ahash_cache(struct ahash_request *areq)
 {
 	struct safexcel_ahash_req *req = ahash_request_ctx(areq);
 	struct crypto_ahash *ahash = crypto_ahash_reqtfm(areq);
 	int queued, cache_len;
 
+	/* cache_len: everyting accepted by the driver but not sent yet,
+	 * tot sz handled by update() - last req sz - tot sz handled by send()
+	 */
 	cache_len = req->len - areq->nbytes - req->processed;
+	/* queued: everything accepted by the driver which will be handled by
+	 * the next send() calls.
+	 * tot sz handled by update() - tot sz handled by send()
+	 */
 	queued = req->len - req->processed;
 
 	/*
-- 
2.14.3

[PATCH 05/17] crypto: inside-secure - refrain from unneeded invalidations

[Antoine Tenart]

From: Ofer Heifetz <oferh@marvell.com>

The check to know if an invalidation is needed (i.e. when the context
changes) is done even if the context does not exist yet. This happens
when first setting a key for ciphers and/or hmac operations.

This commits adds a check in the _setkey functions to only check if an
invalidation is needed when a context exists, as there is no need to
perform this check otherwise.

Signed-off-by: Ofer Heifetz <oferh@marvell.com>
[Antoine: commit message and added a comment and reworked one of the
checks]
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_cipher.c | 10 ++++++----
 drivers/crypto/inside-secure/safexcel_hash.c   | 24 ++++++++++++++++--------
 2 files changed, 22 insertions(+), 12 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel_cipher.c b/drivers/crypto/inside-secure/safexcel_cipher.c
index fcc0a606d748..794849b859a7 100644
--- a/drivers/crypto/inside-secure/safexcel_cipher.c
+++ b/drivers/crypto/inside-secure/safexcel_cipher.c
@@ -78,10 +78,12 @@ static int safexcel_aes_setkey(struct crypto_skcipher *ctfm, const u8 *key,
 		return ret;
 	}
 
-	for (i = 0; i < len / sizeof(u32); i++) {
-		if (ctx->key[i] != cpu_to_le32(aes.key_enc[i])) {
-			ctx->base.needs_inv = true;
-			break;
+	if (ctx->base.ctxr_dma) {
+		for (i = 0; i < len / sizeof(u32); i++) {
+			if (ctx->key[i] != cpu_to_le32(aes.key_enc[i])) {
+				ctx->base.needs_inv = true;
+				break;
+			}
 		}
 	}
 
diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index e53e6b2331db..d156637d6144 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -527,10 +527,16 @@ static int safexcel_ahash_enqueue(struct ahash_request *areq)
 
 	req->needs_inv = false;
 
-	if (req->processed && ctx->digest == CONTEXT_CONTROL_DIGEST_PRECOMPUTED)
-		ctx->base.needs_inv = safexcel_ahash_needs_inv_get(areq);
-
 	if (ctx->base.ctxr) {
+		if (!ctx->base.needs_inv && req->processed &&
+		    ctx->digest == CONTEXT_CONTROL_DIGEST_PRECOMPUTED)
+			/* We're still setting needs_inv here, even though it is
+			 * cleared right away, because the needs_inv flag can be
+			 * set in other functions and we want to keep the same
+			 * logic.
+			 */
+			ctx->base.needs_inv = safexcel_ahash_needs_inv_get(areq);
+
 		if (ctx->base.needs_inv) {
 			ctx->base.needs_inv = false;
 			req->needs_inv = true;
@@ -928,11 +934,13 @@ static int safexcel_hmac_sha1_setkey(struct crypto_ahash *tfm, const u8 *key,
 	if (ret)
 		return ret;
 
-	for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(u32); i++) {
-		if (ctx->ipad[i] != le32_to_cpu(istate.state[i]) ||
-		    ctx->opad[i] != le32_to_cpu(ostate.state[i])) {
-			ctx->base.needs_inv = true;
-			break;
+	if (ctx->base.ctxr) {
+		for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(u32); i++) {
+			if (ctx->ipad[i] != le32_to_cpu(istate.state[i]) ||
+			    ctx->opad[i] != le32_to_cpu(ostate.state[i])) {
+				ctx->base.needs_inv = true;
+				break;
+			}
 		}
 	}
 
-- 
2.14.3

[PATCH 06/17] crypto: inside-secure - EBUSY is not an error on async request

[Antoine Tenart]

From: Ofer Heifetz <oferh@marvell.com>

When initializing the IVs crypto_ahash_update() is called, which at some
point will call crypto_enqueue_request(). This function can return
-EBUSY when no resource is available and the request is queued. Since
this is a valid case, -EBUSY shouldn't be treated as an error.

Signed-off-by: Ofer Heifetz <oferh@marvell.com>
[Antoine: commit message]
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_hash.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index d156637d6144..4a4d95302a93 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -862,7 +862,7 @@ static int safexcel_hmac_init_iv(struct ahash_request *areq,
 	req->last_req = true;
 
 	ret = crypto_ahash_update(areq);
-	if (ret && ret != -EINPROGRESS)
+	if (ret && ret != -EINPROGRESS && ret != -EBUSY)
 		return ret;
 
 	wait_for_completion_interruptible(&result.completion);
-- 
2.14.3

[PATCH 07/17] crypto: inside-secure - move cipher crypto mode to request context

[Antoine Tenart]

From: Ofer Heifetz <oferh@marvell.com>

The cipher direction can be different for requests within the same
transformation context. This patch moves the direction flag from the
context to the request scope.

Signed-off-by: Ofer Heifetz <oferh@marvell.com>
[Antoine: commit message]
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel_cipher.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel_cipher.c b/drivers/crypto/inside-secure/safexcel_cipher.c
index 794849b859a7..62382a7c4cbe 100644
--- a/drivers/crypto/inside-secure/safexcel_cipher.c
+++ b/drivers/crypto/inside-secure/safexcel_cipher.c
@@ -27,7 +27,6 @@ struct safexcel_cipher_ctx {
 	struct safexcel_context base;
 	struct safexcel_crypto_priv *priv;
 
-	enum safexcel_cipher_direction direction;
 	u32 mode;
 
 	__le32 key[8];
@@ -35,6 +34,7 @@ struct safexcel_cipher_ctx {
 };
 
 struct safexcel_cipher_req {
+	enum safexcel_cipher_direction direction;
 	bool needs_inv;
 };
 
@@ -97,12 +97,15 @@ static int safexcel_aes_setkey(struct crypto_skcipher *ctfm, const u8 *key,
 }
 
 static int safexcel_context_control(struct safexcel_cipher_ctx *ctx,
+				    struct crypto_async_request *async,
 				    struct safexcel_command_desc *cdesc)
 {
 	struct safexcel_crypto_priv *priv = ctx->priv;
+	struct skcipher_request *req = skcipher_request_cast(async);
+	struct safexcel_cipher_req *sreq = skcipher_request_ctx(req);
 	int ctrl_size;
 
-	if (ctx->direction == SAFEXCEL_ENCRYPT)
+	if (sreq->direction == SAFEXCEL_ENCRYPT)
 		cdesc->control_data.control0 |= CONTEXT_CONTROL_TYPE_CRYPTO_OUT;
 	else
 		cdesc->control_data.control0 |= CONTEXT_CONTROL_TYPE_CRYPTO_IN;
@@ -245,7 +248,7 @@ static int safexcel_aes_send(struct crypto_async_request *async,
 		n_cdesc++;
 
 		if (n_cdesc == 1) {
-			safexcel_context_control(ctx, cdesc);
+			safexcel_context_control(ctx, async, cdesc);
 			safexcel_cipher_token(ctx, async, cdesc, req->cryptlen);
 		}
 
@@ -469,7 +472,7 @@ static int safexcel_aes(struct skcipher_request *req,
 	int ret, ring;
 
 	sreq->needs_inv = false;
-	ctx->direction = dir;
+	sreq->direction = dir;
 	ctx->mode = mode;
 
 	if (ctx->base.ctxr) {
-- 
2.14.3

[PATCH 08/17] crypto: inside-secure - remove unused parameter in invalidate_cache

[Antoine Tenart]

From: Ofer Heifetz <oferh@marvell.com>

The SafeXcel context isn't used in the cache invalidation function. This
cosmetic patch removes it (as well as from the function prototype in the
header file and when the function is called).

Signed-off-by: Ofer Heifetz <oferh@marvell.com>
[Antoine: commit message]
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c        | 1 -
 drivers/crypto/inside-secure/safexcel.h        | 1 -
 drivers/crypto/inside-secure/safexcel_cipher.c | 2 +-
 drivers/crypto/inside-secure/safexcel_hash.c   | 2 +-
 4 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index 4bcef78a08aa..f250c3c1ab0f 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -540,7 +540,6 @@ void safexcel_inv_complete(struct crypto_async_request *req, int error)
 }
 
 int safexcel_invalidate_cache(struct crypto_async_request *async,
-			      struct safexcel_context *ctx,
 			      struct safexcel_crypto_priv *priv,
 			      dma_addr_t ctxr_dma, int ring,
 			      struct safexcel_request *request)
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index 304c5838c11a..d12c2b479a5e 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -539,7 +539,6 @@ void safexcel_free_context(struct safexcel_crypto_priv *priv,
 				  struct crypto_async_request *req,
 				  int result_sz);
 int safexcel_invalidate_cache(struct crypto_async_request *async,
-			      struct safexcel_context *ctx,
 			      struct safexcel_crypto_priv *priv,
 			      dma_addr_t ctxr_dma, int ring,
 			      struct safexcel_request *request);
diff --git a/drivers/crypto/inside-secure/safexcel_cipher.c b/drivers/crypto/inside-secure/safexcel_cipher.c
index 62382a7c4cbe..fe1d588d6a25 100644
--- a/drivers/crypto/inside-secure/safexcel_cipher.c
+++ b/drivers/crypto/inside-secure/safexcel_cipher.c
@@ -395,7 +395,7 @@ static int safexcel_cipher_send_inv(struct crypto_async_request *async,
 	struct safexcel_crypto_priv *priv = ctx->priv;
 	int ret;
 
-	ret = safexcel_invalidate_cache(async, &ctx->base, priv,
+	ret = safexcel_invalidate_cache(async, priv,
 					ctx->base.ctxr_dma, ring, request);
 	if (unlikely(ret))
 		return ret;
diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index 4a4d95302a93..d233f4a09712 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -417,7 +417,7 @@ static int safexcel_ahash_send_inv(struct crypto_async_request *async,
 	struct safexcel_ahash_ctx *ctx = crypto_ahash_ctx(crypto_ahash_reqtfm(areq));
 	int ret;
 
-	ret = safexcel_invalidate_cache(async, &ctx->base, ctx->priv,
+	ret = safexcel_invalidate_cache(async, ctx->priv,
 					ctx->base.ctxr_dma, ring, request);
 	if (unlikely(ret))
 		return ret;
-- 
2.14.3

[PATCH 09/17] crypto: inside-secure - move request dequeueing into a workqueue

[Antoine Tenart]

This patch moves the request dequeueing into a workqueue to improve the
coalescing of interrupts when sending requests to the engine; as the
engine is capable of having one single interrupt for n requests sent.
Using a workqueue allows to send more request at once.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c        | 29 ++++++++++++++------------
 drivers/crypto/inside-secure/safexcel.h        |  2 +-
 drivers/crypto/inside-secure/safexcel_cipher.c | 12 +++++------
 drivers/crypto/inside-secure/safexcel_hash.c   | 12 +++++------
 4 files changed, 29 insertions(+), 26 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index f250c3c1ab0f..f422af3eed2f 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -429,8 +429,6 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 	struct safexcel_request *request;
 	int ret, nreq = 0, cdesc = 0, rdesc = 0, commands, results;
 
-	priv->ring[ring].need_dequeue = false;
-
 	do {
 		spin_lock_bh(&priv->ring[ring].queue_lock);
 		backlog = crypto_get_backlog(&priv->ring[ring].queue);
@@ -445,8 +443,6 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 			spin_lock_bh(&priv->ring[ring].queue_lock);
 			crypto_enqueue_request(&priv->ring[ring].queue, req);
 			spin_unlock_bh(&priv->ring[ring].queue_lock);
-
-			priv->ring[ring].need_dequeue = true;
 			goto finalize;
 		}
 
@@ -455,7 +451,6 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 		if (ret) {
 			kfree(request);
 			req->complete(req, ret);
-			priv->ring[ring].need_dequeue = true;
 			goto finalize;
 		}
 
@@ -471,9 +466,7 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 	} while (nreq++ < EIP197_MAX_BATCH_SZ);
 
 finalize:
-	if (nreq == EIP197_MAX_BATCH_SZ)
-		priv->ring[ring].need_dequeue = true;
-	else if (!nreq)
+	if (!nreq)
 		return;
 
 	spin_lock_bh(&priv->ring[ring].lock);
@@ -628,13 +621,18 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 static void safexcel_handle_result_work(struct work_struct *work)
 {
 	struct safexcel_work_data *data =
-			container_of(work, struct safexcel_work_data, work);
+			container_of(work, struct safexcel_work_data, result_work);
 	struct safexcel_crypto_priv *priv = data->priv;
 
 	safexcel_handle_result_descriptor(priv, data->ring);
+}
+
+static void safexcel_dequeue_work(struct work_struct *work)
+{
+	struct safexcel_work_data *data =
+			container_of(work, struct safexcel_work_data, work);
 
-	if (priv->ring[data->ring].need_dequeue)
-		safexcel_dequeue(data->priv, data->ring);
+	safexcel_dequeue(data->priv, data->ring);
 }
 
 struct safexcel_ring_irq_data {
@@ -665,7 +663,10 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 			 */
 			dev_err(priv->dev, "RDR: fatal error.");
 		} else if (likely(stat & EIP197_xDR_THRESH)) {
-			queue_work(priv->ring[ring].workqueue, &priv->ring[ring].work_data.work);
+			queue_work(priv->ring[ring].workqueue,
+				   &priv->ring[ring].work_data.result_work);
+			queue_work(priv->ring[ring].workqueue,
+				   &priv->ring[ring].work_data.work);
 		}
 
 		/* ACK the interrupts */
@@ -846,7 +847,9 @@ static int safexcel_probe(struct platform_device *pdev)
 
 		priv->ring[i].work_data.priv = priv;
 		priv->ring[i].work_data.ring = i;
-		INIT_WORK(&priv->ring[i].work_data.work, safexcel_handle_result_work);
+		INIT_WORK(&priv->ring[i].work_data.result_work,
+			  safexcel_handle_result_work);
+		INIT_WORK(&priv->ring[i].work_data.work, safexcel_dequeue_work);
 
 		snprintf(wq_name, 9, "wq_ring%d", i);
 		priv->ring[i].workqueue = create_singlethread_workqueue(wq_name);
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index d12c2b479a5e..8e9c65183439 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -459,6 +459,7 @@ struct safexcel_config {
 
 struct safexcel_work_data {
 	struct work_struct work;
+	struct work_struct result_work;
 	struct safexcel_crypto_priv *priv;
 	int ring;
 };
@@ -489,7 +490,6 @@ struct safexcel_crypto_priv {
 		/* queue */
 		struct crypto_queue queue;
 		spinlock_t queue_lock;
-		bool need_dequeue;
 	} ring[EIP197_MAX_RINGS];
 };
 
diff --git a/drivers/crypto/inside-secure/safexcel_cipher.c b/drivers/crypto/inside-secure/safexcel_cipher.c
index fe1d588d6a25..0e5cc230e49a 100644
--- a/drivers/crypto/inside-secure/safexcel_cipher.c
+++ b/drivers/crypto/inside-secure/safexcel_cipher.c
@@ -358,8 +358,8 @@ static int safexcel_handle_inv_result(struct safexcel_crypto_priv *priv,
 	if (enq_ret != -EINPROGRESS)
 		*ret = enq_ret;
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	*should_complete = false;
 
@@ -448,8 +448,8 @@ static int safexcel_cipher_exit_inv(struct crypto_tfm *tfm)
 	crypto_enqueue_request(&priv->ring[ring].queue, &req->base);
 	spin_unlock_bh(&priv->ring[ring].queue_lock);
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	wait_for_completion_interruptible(&result.completion);
 
@@ -495,8 +495,8 @@ static int safexcel_aes(struct skcipher_request *req,
 	ret = crypto_enqueue_request(&priv->ring[ring].queue, &req->base);
 	spin_unlock_bh(&priv->ring[ring].queue_lock);
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	return ret;
 }
diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index d233f4a09712..2fb5bc6b6268 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -381,8 +381,8 @@ static int safexcel_handle_inv_result(struct safexcel_crypto_priv *priv,
 	if (enq_ret != -EINPROGRESS)
 		*ret = enq_ret;
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	*should_complete = false;
 
@@ -470,8 +470,8 @@ static int safexcel_ahash_exit_inv(struct crypto_tfm *tfm)
 	crypto_enqueue_request(&priv->ring[ring].queue, &req->base);
 	spin_unlock_bh(&priv->ring[ring].queue_lock);
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	wait_for_completion_interruptible(&result.completion);
 
@@ -556,8 +556,8 @@ static int safexcel_ahash_enqueue(struct ahash_request *areq)
 	ret = crypto_enqueue_request(&priv->ring[ring].queue, &areq->base);
 	spin_unlock_bh(&priv->ring[ring].queue_lock);
 
-	if (!priv->ring[ring].need_dequeue)
-		safexcel_dequeue(priv, ring);
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
 
 	return ret;
 }
-- 
2.14.3

[PATCH 10/17] crypto: inside-secure - use threaded IRQs for result handling

[Antoine Tenart]

This patch moves the result handling from an IRQ handler to a threaded
IRQ handler, to improve the number of complete requests being handled at
once.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c | 41 ++++++++++++++++++---------------
 drivers/crypto/inside-secure/safexcel.h |  1 -
 2 files changed, 22 insertions(+), 20 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index f422af3eed2f..aa4943cad55e 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -618,15 +618,6 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 	}
 }
 
-static void safexcel_handle_result_work(struct work_struct *work)
-{
-	struct safexcel_work_data *data =
-			container_of(work, struct safexcel_work_data, result_work);
-	struct safexcel_crypto_priv *priv = data->priv;
-
-	safexcel_handle_result_descriptor(priv, data->ring);
-}
-
 static void safexcel_dequeue_work(struct work_struct *work)
 {
 	struct safexcel_work_data *data =
@@ -644,12 +635,12 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 {
 	struct safexcel_ring_irq_data *irq_data = data;
 	struct safexcel_crypto_priv *priv = irq_data->priv;
-	int ring = irq_data->ring;
+	int ring = irq_data->ring, rc = IRQ_NONE;
 	u32 status, stat;
 
 	status = readl(priv->base + EIP197_HIA_AIC_R_ENABLED_STAT(ring));
 	if (!status)
-		return IRQ_NONE;
+		return rc;
 
 	/* RDR interrupts */
 	if (status & EIP197_RDR_IRQ(ring)) {
@@ -663,10 +654,7 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 			 */
 			dev_err(priv->dev, "RDR: fatal error.");
 		} else if (likely(stat & EIP197_xDR_THRESH)) {
-			queue_work(priv->ring[ring].workqueue,
-				   &priv->ring[ring].work_data.result_work);
-			queue_work(priv->ring[ring].workqueue,
-				   &priv->ring[ring].work_data.work);
+			rc = IRQ_WAKE_THREAD;
 		}
 
 		/* ACK the interrupts */
@@ -677,11 +665,26 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 	/* ACK the interrupts */
 	writel(status, priv->base + EIP197_HIA_AIC_R_ACK(ring));
 
+	return rc;
+}
+
+static irqreturn_t safexcel_irq_ring_thread(int irq, void *data)
+{
+	struct safexcel_ring_irq_data *irq_data = data;
+	struct safexcel_crypto_priv *priv = irq_data->priv;
+	int ring = irq_data->ring;
+
+	safexcel_handle_result_descriptor(priv, ring);
+
+	queue_work(priv->ring[ring].workqueue,
+		   &priv->ring[ring].work_data.work);
+
 	return IRQ_HANDLED;
 }
 
 static int safexcel_request_ring_irq(struct platform_device *pdev, const char *name,
 				     irq_handler_t handler,
+				     irq_handler_t threaded_handler,
 				     struct safexcel_ring_irq_data *ring_irq_priv)
 {
 	int ret, irq = platform_get_irq_byname(pdev, name);
@@ -691,8 +694,9 @@ static int safexcel_request_ring_irq(struct platform_device *pdev, const char *n
 		return irq;
 	}
 
-	ret = devm_request_irq(&pdev->dev, irq, handler, 0,
-			       dev_name(&pdev->dev), ring_irq_priv);
+	ret = devm_request_threaded_irq(&pdev->dev, irq, handler,
+					threaded_handler, IRQF_ONESHOT,
+					dev_name(&pdev->dev), ring_irq_priv);
 	if (ret) {
 		dev_err(&pdev->dev, "unable to request IRQ %d\n", irq);
 		return ret;
@@ -839,6 +843,7 @@ static int safexcel_probe(struct platform_device *pdev)
 
 		snprintf(irq_name, 6, "ring%d", i);
 		irq = safexcel_request_ring_irq(pdev, irq_name, safexcel_irq_ring,
+						safexcel_irq_ring_thread,
 						ring_irq);
 		if (irq < 0) {
 			ret = irq;
@@ -847,8 +852,6 @@ static int safexcel_probe(struct platform_device *pdev)
 
 		priv->ring[i].work_data.priv = priv;
 		priv->ring[i].work_data.ring = i;
-		INIT_WORK(&priv->ring[i].work_data.result_work,
-			  safexcel_handle_result_work);
 		INIT_WORK(&priv->ring[i].work_data.work, safexcel_dequeue_work);
 
 		snprintf(wq_name, 9, "wq_ring%d", i);
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index 8e9c65183439..fffddefb0d9b 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -459,7 +459,6 @@ struct safexcel_config {
 
 struct safexcel_work_data {
 	struct work_struct work;
-	struct work_struct result_work;
 	struct safexcel_crypto_priv *priv;
 	int ring;
 };
-- 
2.14.3

[PATCH 11/17] crypto: inside-secure - dequeue all requests at once

[Antoine Tenart]

This patch updates the dequeueing logic to dequeue all requests at once.
Since we can have many requests in the queue, the interrupt coalescing
is kept so that the ring interrupt fires every EIP197_MAX_BATCH_SZ at
most.

To allow dequeueing all requests at once while still using reasonable
settings for the interrupt coalescing, the result handling function was
updated to setup the threshold interrupt when needed (i.e. when more
requests than EIP197_MAX_BATCH_SZ are in the queue). When using this
capability the ring is marked as busy so that the dequeue function
enqueue new requests without setting the threshold interrupt.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c | 60 ++++++++++++++++++++++++++-------
 drivers/crypto/inside-secure/safexcel.h |  8 +++++
 2 files changed, 56 insertions(+), 12 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index aa4943cad55e..db7ad9d3eeed 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -422,6 +422,23 @@ static int safexcel_hw_init(struct safexcel_crypto_priv *priv)
 	return 0;
 }
 
+/* Called with ring's lock taken */
+int safexcel_try_push_requests(struct safexcel_crypto_priv *priv, int ring,
+			       int reqs)
+{
+	int coal = min_t(int, reqs, EIP197_MAX_BATCH_SZ);
+
+	if (!coal)
+		return 0;
+
+	/* Configure when we want an interrupt */
+	writel(EIP197_HIA_RDR_THRESH_PKT_MODE |
+	       EIP197_HIA_RDR_THRESH_PROC_PKT(coal),
+	       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_THRESH);
+
+	return coal;
+}
+
 void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 {
 	struct crypto_async_request *req, *backlog;
@@ -429,7 +446,7 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 	struct safexcel_request *request;
 	int ret, nreq = 0, cdesc = 0, rdesc = 0, commands, results;
 
-	do {
+	while (true) {
 		spin_lock_bh(&priv->ring[ring].queue_lock);
 		backlog = crypto_get_backlog(&priv->ring[ring].queue);
 		req = crypto_dequeue_request(&priv->ring[ring].queue);
@@ -463,18 +480,24 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 
 		cdesc += commands;
 		rdesc += results;
-	} while (nreq++ < EIP197_MAX_BATCH_SZ);
+		nreq++;
+	}
 
 finalize:
 	if (!nreq)
 		return;
 
-	spin_lock_bh(&priv->ring[ring].lock);
+	spin_lock_bh(&priv->ring[ring].egress_lock);
 
-	/* Configure when we want an interrupt */
-	writel(EIP197_HIA_RDR_THRESH_PKT_MODE |
-	       EIP197_HIA_RDR_THRESH_PROC_PKT(nreq),
-	       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_THRESH);
+	if (!priv->ring[ring].busy) {
+		nreq -= safexcel_try_push_requests(priv, ring, nreq);
+		if (nreq)
+			priv->ring[ring].busy = true;
+	}
+
+	priv->ring[ring].requests_left += nreq;
+
+	spin_unlock_bh(&priv->ring[ring].egress_lock);
 
 	/* let the RDR know we have pending descriptors */
 	writel((rdesc * priv->config.rd_offset) << 2,
@@ -483,8 +506,6 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 	/* let the CDR know we have pending descriptors */
 	writel((cdesc * priv->config.cd_offset) << 2,
 	       priv->base + EIP197_HIA_CDR(ring) + EIP197_HIA_xDR_PREP_COUNT);
-
-	spin_unlock_bh(&priv->ring[ring].lock);
 }
 
 void safexcel_free_context(struct safexcel_crypto_priv *priv,
@@ -579,14 +600,14 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 {
 	struct safexcel_request *sreq;
 	struct safexcel_context *ctx;
-	int ret, i, nreq, ndesc = 0;
+	int ret, i, nreq, ndesc = 0, done;
 	bool should_complete;
 
 	nreq = readl(priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
 	nreq >>= 24;
 	nreq &= GENMASK(6, 0);
 	if (!nreq)
-		return;
+		goto requests_left;
 
 	for (i = 0; i < nreq; i++) {
 		spin_lock_bh(&priv->ring[ring].egress_lock);
@@ -601,7 +622,7 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 		if (ndesc < 0) {
 			kfree(sreq);
 			dev_err(priv->dev, "failed to handle result (%d)", ndesc);
-			return;
+			goto requests_left;
 		}
 
 		writel(EIP197_xDR_PROC_xD_PKT(1) |
@@ -616,6 +637,18 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 
 		kfree(sreq);
 	}
+
+requests_left:
+	spin_lock_bh(&priv->ring[ring].egress_lock);
+
+	done = safexcel_try_push_requests(priv, ring,
+					  priv->ring[ring].requests_left);
+
+	priv->ring[ring].requests_left -= done;
+	if (!done && !priv->ring[ring].requests_left)
+		priv->ring[ring].busy = false;
+
+	spin_unlock_bh(&priv->ring[ring].egress_lock);
 }
 
 static void safexcel_dequeue_work(struct work_struct *work)
@@ -861,6 +894,9 @@ static int safexcel_probe(struct platform_device *pdev)
 			goto err_clk;
 		}
 
+		priv->ring[i].requests_left = 0;
+		priv->ring[i].busy = false;
+
 		crypto_init_queue(&priv->ring[i].queue,
 				  EIP197_DEFAULT_RING_SIZE);
 
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index fffddefb0d9b..531e3e9d8384 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -489,6 +489,14 @@ struct safexcel_crypto_priv {
 		/* queue */
 		struct crypto_queue queue;
 		spinlock_t queue_lock;
+
+		/* Number of requests in the engine that needs the threshold
+		 * interrupt to be set up.
+		 */
+		int requests_left;
+
+		/* The ring is currently handling at least one request */
+		bool busy;
 	} ring[EIP197_MAX_RINGS];
 };
 
-- 
2.14.3

[PATCH 12/17] crypto: inside-secure - increase the ring size

[Antoine Tenart]

Increase the ring size to handle more requests in parallel, while
keeping the batch size (for interrupt coalescing) to its previous value.
The ring size and batch size are now unlinked.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index 531e3e9d8384..2a0ab6ce716a 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -19,11 +19,11 @@
 #define EIP197_HIA_VERSION_BE			0x35ca
 
 /* Static configuration */
-#define EIP197_DEFAULT_RING_SIZE		64
+#define EIP197_DEFAULT_RING_SIZE		400
 #define EIP197_MAX_TOKENS			5
 #define EIP197_MAX_RINGS			4
 #define EIP197_FETCH_COUNT			1
-#define EIP197_MAX_BATCH_SZ			EIP197_DEFAULT_RING_SIZE
+#define EIP197_MAX_BATCH_SZ			64
 
 #define EIP197_GFP_FLAGS(base)	((base).flags & CRYPTO_TFM_REQ_MAY_SLEEP ? \
 				 GFP_KERNEL : GFP_ATOMIC)
-- 
2.14.3

[PATCH 13/17] crypto: inside-secure - acknowledge the result requests all at once

[Antoine Tenart]

This patches moves the result request acknowledgment from a per request
process to acknowledging all the result requests handled at once.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index db7ad9d3eeed..36c9a72c47b1 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -600,7 +600,7 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 {
 	struct safexcel_request *sreq;
 	struct safexcel_context *ctx;
-	int ret, i, nreq, ndesc = 0, done;
+	int ret, i, nreq, ndesc = 0, tot_descs = 0, done;
 	bool should_complete;
 
 	nreq = readl(priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
@@ -622,13 +622,9 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 		if (ndesc < 0) {
 			kfree(sreq);
 			dev_err(priv->dev, "failed to handle result (%d)", ndesc);
-			goto requests_left;
+			goto acknowledge;
 		}
 
-		writel(EIP197_xDR_PROC_xD_PKT(1) |
-		       EIP197_xDR_PROC_xD_COUNT(ndesc * priv->config.rd_offset),
-		       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
-
 		if (should_complete) {
 			local_bh_disable();
 			sreq->req->complete(sreq->req, ret);
@@ -636,6 +632,14 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 		}
 
 		kfree(sreq);
+		tot_descs += ndesc;
+	}
+
+acknowledge:
+	if (i) {
+		writel(EIP197_xDR_PROC_xD_PKT(i) |
+		       EIP197_xDR_PROC_xD_COUNT(tot_descs * priv->config.rd_offset),
+		       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
 	}
 
 requests_left:
-- 
2.14.3

[PATCH 14/17] crypto: inside-secure - handle more result requests when counter is full

[Antoine Tenart]

This patch modifies the result handling logic to continue handling
results when the completed requests counter is full and not showing the
actual number of requests to handle.

Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c | 15 ++++++++++++---
 drivers/crypto/inside-secure/safexcel.h |  2 ++
 2 files changed, 14 insertions(+), 3 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index 36c9a72c47b1..8042922b4ed8 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -600,12 +600,15 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 {
 	struct safexcel_request *sreq;
 	struct safexcel_context *ctx;
-	int ret, i, nreq, ndesc = 0, tot_descs = 0, done;
+	int ret, i, nreq, ndesc, tot_descs, done;
 	bool should_complete;
 
+handle_results:
+	tot_descs = 0;
+
 	nreq = readl(priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
-	nreq >>= 24;
-	nreq &= GENMASK(6, 0);
+	nreq >>= EIP197_xDR_PROC_xD_PKT_OFFSET;
+	nreq &= EIP197_xDR_PROC_xD_PKT_MASK;
 	if (!nreq)
 		goto requests_left;
 
@@ -642,6 +645,12 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 		       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
 	}
 
+	/* If the number of requests overflowed the counter, try to proceed more
+	 * requests.
+	 */
+	if (nreq == EIP197_xDR_PROC_xD_PKT_MASK)
+		goto handle_results;
+
 requests_left:
 	spin_lock_bh(&priv->ring[ring].egress_lock);
 
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index 2a0ab6ce716a..0c47e792192d 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -117,6 +117,8 @@
 #define EIP197_xDR_PREP_CLR_COUNT		BIT(31)
 
 /* EIP197_HIA_xDR_PROC_COUNT */
+#define EIP197_xDR_PROC_xD_PKT_OFFSET		24
+#define EIP197_xDR_PROC_xD_PKT_MASK		GENMASK(6, 0)
 #define EIP197_xDR_PROC_xD_COUNT(n)		((n) << 2)
 #define EIP197_xDR_PROC_xD_PKT(n)		((n) << 24)
 #define EIP197_xDR_PROC_CLR_COUNT		BIT(31)
-- 
2.14.3

[PATCH 15/17] crypto: inside-secure - retry to proceed the request later on fail

[Antoine Tenart]

The dequeueing function was putting back a request in the crypto queue
on failure (when not enough resources are available) which is not
perfect as the request will be handled much later. This patch updates
this logic by keeping a reference on the failed request to try
proceeding it later when enough resources are available.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c | 32 +++++++++++++++++++++++---------
 drivers/crypto/inside-secure/safexcel.h |  6 ++++++
 2 files changed, 29 insertions(+), 9 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index 8042922b4ed8..4c7f205d83f0 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -446,29 +446,36 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 	struct safexcel_request *request;
 	int ret, nreq = 0, cdesc = 0, rdesc = 0, commands, results;
 
+	/* If a request wasn't properly dequeued because of a lack of resources,
+	 * proceeded it first,
+	 */
+	req = priv->ring[ring].req;
+	backlog = priv->ring[ring].backlog;
+	if (req)
+		goto handle_req;
+
 	while (true) {
 		spin_lock_bh(&priv->ring[ring].queue_lock);
 		backlog = crypto_get_backlog(&priv->ring[ring].queue);
 		req = crypto_dequeue_request(&priv->ring[ring].queue);
 		spin_unlock_bh(&priv->ring[ring].queue_lock);
 
-		if (!req)
+		if (!req) {
+			priv->ring[ring].req = NULL;
+			priv->ring[ring].backlog = NULL;
 			goto finalize;
+		}
 
+handle_req:
 		request = kzalloc(sizeof(*request), EIP197_GFP_FLAGS(*req));
-		if (!request) {
-			spin_lock_bh(&priv->ring[ring].queue_lock);
-			crypto_enqueue_request(&priv->ring[ring].queue, req);
-			spin_unlock_bh(&priv->ring[ring].queue_lock);
-			goto finalize;
-		}
+		if (!request)
+			goto request_failed;
 
 		ctx = crypto_tfm_ctx(req->tfm);
 		ret = ctx->send(req, ring, request, &commands, &results);
 		if (ret) {
 			kfree(request);
-			req->complete(req, ret);
-			goto finalize;
+			goto request_failed;
 		}
 
 		if (backlog)
@@ -483,6 +490,13 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 		nreq++;
 	}
 
+request_failed:
+	/* Not enough resources to handle all the requests. Bail out and save
+	 * the request and the backlog for the next dequeue call (per-ring).
+	 */
+	priv->ring[ring].req = req;
+	priv->ring[ring].backlog = backlog;
+
 finalize:
 	if (!nreq)
 		return;
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index 0c47e792192d..d4955abf873b 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -499,6 +499,12 @@ struct safexcel_crypto_priv {
 
 		/* The ring is currently handling at least one request */
 		bool busy;
+
+		/* Store for current requests when bailing out of the dequeueing
+		 * function when no enough resources are available.
+		 */
+		struct crypto_async_request *req;
+		struct crypto_async_request *backlog;
 	} ring[EIP197_MAX_RINGS];
 };
 
-- 
2.14.3

[PATCH 16/17] crypto: inside-secure - EIP97 support

[Antoine Tenart]

The Inside Secure SafeXcel driver was firstly designed to support the
EIP197 cryptographic engine which is an evolution (with much more
feature, better performances) of the EIP97 cryptographic engine. This
patch convert the Inside Secure SafeXcel driver to support both engines
(EIP97 + EIP197).

The main differences are the register offsets and the context
invalidation process which is EIP197 specific. This patch adds an
indirection on the register offsets and adds checks not to send any
invalidation request when driving the EIP97. A new compatible is added
as well to bind the driver from device trees.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 drivers/crypto/inside-secure/safexcel.c        | 212 +++++++++++++++----------
 drivers/crypto/inside-secure/safexcel.h        | 151 ++++++++++++------
 drivers/crypto/inside-secure/safexcel_cipher.c |  20 ++-
 drivers/crypto/inside-secure/safexcel_hash.c   |  19 ++-
 4 files changed, 264 insertions(+), 138 deletions(-)

diff --git a/drivers/crypto/inside-secure/safexcel.c b/drivers/crypto/inside-secure/safexcel.c
index 4c7f205d83f0..0d685be75241 100644
--- a/drivers/crypto/inside-secure/safexcel.c
+++ b/drivers/crypto/inside-secure/safexcel.c
@@ -108,10 +108,10 @@ static void eip197_write_firmware(struct safexcel_crypto_priv *priv,
 	writel(EIP197_PE_ICE_x_CTRL_SW_RESET |
 	       EIP197_PE_ICE_x_CTRL_CLR_ECC_CORR |
 	       EIP197_PE_ICE_x_CTRL_CLR_ECC_NON_CORR,
-	       priv->base + ctrl);
+	       EIP197_PE(priv) + ctrl);
 
 	/* Enable access to the program memory */
-	writel(prog_en, priv->base + EIP197_PE_ICE_RAM_CTRL);
+	writel(prog_en, EIP197_PE(priv) + EIP197_PE_ICE_RAM_CTRL);
 
 	/* Write the firmware */
 	for (i = 0; i < fw->size / sizeof(u32); i++)
@@ -119,12 +119,12 @@ static void eip197_write_firmware(struct safexcel_crypto_priv *priv,
 		       priv->base + EIP197_CLASSIFICATION_RAMS + i * sizeof(u32));
 
 	/* Disable access to the program memory */
-	writel(0, priv->base + EIP197_PE_ICE_RAM_CTRL);
+	writel(0, EIP197_PE(priv) + EIP197_PE_ICE_RAM_CTRL);
 
 	/* Release engine from reset */
-	val = readl(priv->base + ctrl);
+	val = readl(EIP197_PE(priv) + ctrl);
 	val &= ~EIP197_PE_ICE_x_CTRL_SW_RESET;
-	writel(val, priv->base + ctrl);
+	writel(val, EIP197_PE(priv) + ctrl);
 }
 
 static int eip197_load_firmwares(struct safexcel_crypto_priv *priv)
@@ -145,14 +145,14 @@ static int eip197_load_firmwares(struct safexcel_crypto_priv *priv)
 	 }
 
 	/* Clear the scratchpad memory */
-	val = readl(priv->base + EIP197_PE_ICE_SCRATCH_CTRL);
+	val = readl(EIP197_PE(priv) + EIP197_PE_ICE_SCRATCH_CTRL);
 	val |= EIP197_PE_ICE_SCRATCH_CTRL_CHANGE_TIMER |
 	       EIP197_PE_ICE_SCRATCH_CTRL_TIMER_EN |
 	       EIP197_PE_ICE_SCRATCH_CTRL_SCRATCH_ACCESS |
 	       EIP197_PE_ICE_SCRATCH_CTRL_CHANGE_ACCESS;
-	writel(val, priv->base + EIP197_PE_ICE_SCRATCH_CTRL);
+	writel(val, EIP197_PE(priv) + EIP197_PE_ICE_SCRATCH_CTRL);
 
-	memset(priv->base + EIP197_PE_ICE_SCRATCH_RAM, 0,
+	memset(EIP197_PE(priv) + EIP197_PE_ICE_SCRATCH_RAM, 0,
 	       EIP197_NUM_OF_SCRATCH_BLOCKS * sizeof(u32));
 
 	eip197_write_firmware(priv, fw[FW_IFPP], EIP197_PE_ICE_FPP_CTRL,
@@ -173,7 +173,7 @@ static int safexcel_hw_setup_cdesc_rings(struct safexcel_crypto_priv *priv)
 	u32 hdw, cd_size_rnd, val;
 	int i;
 
-	hdw = readl(priv->base + EIP197_HIA_OPTIONS);
+	hdw = readl(EIP197_HIA_AIC_G(priv) + EIP197_HIA_OPTIONS);
 	hdw &= GENMASK(27, 25);
 	hdw >>= 25;
 
@@ -182,26 +182,25 @@ static int safexcel_hw_setup_cdesc_rings(struct safexcel_crypto_priv *priv)
 	for (i = 0; i < priv->config.rings; i++) {
 		/* ring base address */
 		writel(lower_32_bits(priv->ring[i].cdr.base_dma),
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_RING_BASE_ADDR_LO);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_RING_BASE_ADDR_LO);
 		writel(upper_32_bits(priv->ring[i].cdr.base_dma),
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_RING_BASE_ADDR_HI);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_RING_BASE_ADDR_HI);
 
 		writel(EIP197_xDR_DESC_MODE_64BIT | (priv->config.cd_offset << 16) |
 		       priv->config.cd_size,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_DESC_SIZE);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_DESC_SIZE);
 		writel(((EIP197_FETCH_COUNT * (cd_size_rnd << hdw)) << 16) |
 		       (EIP197_FETCH_COUNT * priv->config.cd_offset),
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_CFG);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_CFG);
 
 		/* Configure DMA tx control */
 		val = EIP197_HIA_xDR_CFG_WR_CACHE(WR_CACHE_3BITS);
 		val |= EIP197_HIA_xDR_CFG_RD_CACHE(RD_CACHE_3BITS);
-		writel(val,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_DMA_CFG);
+		writel(val, EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_DMA_CFG);
 
 		/* clear any pending interrupt */
 		writel(GENMASK(5, 0),
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_STAT);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_STAT);
 	}
 
 	return 0;
@@ -212,7 +211,7 @@ static int safexcel_hw_setup_rdesc_rings(struct safexcel_crypto_priv *priv)
 	u32 hdw, rd_size_rnd, val;
 	int i;
 
-	hdw = readl(priv->base + EIP197_HIA_OPTIONS);
+	hdw = readl(EIP197_HIA_AIC_G(priv) + EIP197_HIA_OPTIONS);
 	hdw &= GENMASK(27, 25);
 	hdw >>= 25;
 
@@ -221,33 +220,33 @@ static int safexcel_hw_setup_rdesc_rings(struct safexcel_crypto_priv *priv)
 	for (i = 0; i < priv->config.rings; i++) {
 		/* ring base address */
 		writel(lower_32_bits(priv->ring[i].rdr.base_dma),
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_RING_BASE_ADDR_LO);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_RING_BASE_ADDR_LO);
 		writel(upper_32_bits(priv->ring[i].rdr.base_dma),
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_RING_BASE_ADDR_HI);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_RING_BASE_ADDR_HI);
 
 		writel(EIP197_xDR_DESC_MODE_64BIT | (priv->config.rd_offset << 16) |
 		       priv->config.rd_size,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_DESC_SIZE);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_DESC_SIZE);
 
 		writel(((EIP197_FETCH_COUNT * (rd_size_rnd << hdw)) << 16) |
 		       (EIP197_FETCH_COUNT * priv->config.rd_offset),
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_CFG);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_CFG);
 
 		/* Configure DMA tx control */
 		val = EIP197_HIA_xDR_CFG_WR_CACHE(WR_CACHE_3BITS);
 		val |= EIP197_HIA_xDR_CFG_RD_CACHE(RD_CACHE_3BITS);
 		val |= EIP197_HIA_xDR_WR_RES_BUF | EIP197_HIA_xDR_WR_CTRL_BUG;
 		writel(val,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_DMA_CFG);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_DMA_CFG);
 
 		/* clear any pending interrupt */
 		writel(GENMASK(7, 0),
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_STAT);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_STAT);
 
 		/* enable ring interrupt */
-		val = readl(priv->base + EIP197_HIA_AIC_R_ENABLE_CTRL(i));
+		val = readl(EIP197_HIA_AIC_R(priv) + EIP197_HIA_AIC_R_ENABLE_CTRL(i));
 		val |= EIP197_RDR_IRQ(i);
-		writel(val, priv->base + EIP197_HIA_AIC_R_ENABLE_CTRL(i));
+		writel(val, EIP197_HIA_AIC_R(priv) + EIP197_HIA_AIC_R_ENABLE_CTRL(i));
 	}
 
 	return 0;
@@ -259,39 +258,40 @@ static int safexcel_hw_init(struct safexcel_crypto_priv *priv)
 	int i, ret;
 
 	/* Determine endianess and configure byte swap */
-	version = readl(priv->base + EIP197_HIA_VERSION);
-	val = readl(priv->base + EIP197_HIA_MST_CTRL);
+	version = readl(EIP197_HIA_AIC(priv) + EIP197_HIA_VERSION);
+	val = readl(EIP197_HIA_AIC(priv) + EIP197_HIA_MST_CTRL);
 
 	if ((version & 0xffff) == EIP197_HIA_VERSION_BE)
 		val |= EIP197_MST_CTRL_BYTE_SWAP;
 	else if (((version >> 16) & 0xffff) == EIP197_HIA_VERSION_LE)
 		val |= (EIP197_MST_CTRL_NO_BYTE_SWAP >> 24);
 
-	writel(val, priv->base + EIP197_HIA_MST_CTRL);
-
+	writel(val, EIP197_HIA_AIC(priv) + EIP197_HIA_MST_CTRL);
 
 	/* Configure wr/rd cache values */
 	writel(EIP197_MST_CTRL_RD_CACHE(RD_CACHE_4BITS) |
 	       EIP197_MST_CTRL_WD_CACHE(WR_CACHE_4BITS),
-	       priv->base + EIP197_MST_CTRL);
+	       EIP197_HIA_GEN_CFG(priv) + EIP197_MST_CTRL);
 
 	/* Interrupts reset */
 
 	/* Disable all global interrupts */
-	writel(0, priv->base + EIP197_HIA_AIC_G_ENABLE_CTRL);
+	writel(0, EIP197_HIA_AIC_G(priv) + EIP197_HIA_AIC_G_ENABLE_CTRL);
 
 	/* Clear any pending interrupt */
-	writel(GENMASK(31, 0), priv->base + EIP197_HIA_AIC_G_ACK);
+	writel(GENMASK(31, 0), EIP197_HIA_AIC_G(priv) + EIP197_HIA_AIC_G_ACK);
 
 	/* Data Fetch Engine configuration */
 
 	/* Reset all DFE threads */
 	writel(EIP197_DxE_THR_CTRL_RESET_PE,
-	       priv->base + EIP197_HIA_DFE_THR_CTRL);
+	       EIP197_HIA_DFE_THR(priv) + EIP197_HIA_DFE_THR_CTRL);
 
-	/* Reset HIA input interface arbiter */
-	writel(EIP197_HIA_RA_PE_CTRL_RESET,
-	       priv->base + EIP197_HIA_RA_PE_CTRL);
+	if (priv->version == EIP197) {
+		/* Reset HIA input interface arbiter */
+		writel(EIP197_HIA_RA_PE_CTRL_RESET,
+		       EIP197_HIA_AIC(priv) + EIP197_HIA_RA_PE_CTRL);
+	}
 
 	/* DMA transfer size to use */
 	val = EIP197_HIA_DFE_CFG_DIS_DEBUG;
@@ -299,29 +299,32 @@ static int safexcel_hw_init(struct safexcel_crypto_priv *priv)
 	val |= EIP197_HIA_DxE_CFG_MIN_CTRL_SIZE(5) | EIP197_HIA_DxE_CFG_MAX_CTRL_SIZE(7);
 	val |= EIP197_HIA_DxE_CFG_DATA_CACHE_CTRL(RD_CACHE_3BITS);
 	val |= EIP197_HIA_DxE_CFG_CTRL_CACHE_CTRL(RD_CACHE_3BITS);
-	writel(val, priv->base + EIP197_HIA_DFE_CFG);
+	writel(val, EIP197_HIA_DFE(priv) + EIP197_HIA_DFE_CFG);
 
 	/* Leave the DFE threads reset state */
-	writel(0, priv->base + EIP197_HIA_DFE_THR_CTRL);
+	writel(0, EIP197_HIA_DFE_THR(priv) + EIP197_HIA_DFE_THR_CTRL);
 
 	/* Configure the procesing engine thresholds */
 	writel(EIP197_PE_IN_xBUF_THRES_MIN(5) | EIP197_PE_IN_xBUF_THRES_MAX(9),
-	      priv->base + EIP197_PE_IN_DBUF_THRES);
+	       EIP197_PE(priv) + EIP197_PE_IN_DBUF_THRES);
 	writel(EIP197_PE_IN_xBUF_THRES_MIN(5) | EIP197_PE_IN_xBUF_THRES_MAX(7),
-	      priv->base + EIP197_PE_IN_TBUF_THRES);
+	       EIP197_PE(priv) + EIP197_PE_IN_TBUF_THRES);
 
-	/* enable HIA input interface arbiter and rings */
-	writel(EIP197_HIA_RA_PE_CTRL_EN | GENMASK(priv->config.rings - 1, 0),
-	       priv->base + EIP197_HIA_RA_PE_CTRL);
+	if (priv->version == EIP197) {
+		/* enable HIA input interface arbiter and rings */
+		writel(EIP197_HIA_RA_PE_CTRL_EN |
+		       GENMASK(priv->config.rings - 1, 0),
+		       EIP197_HIA_AIC(priv) + EIP197_HIA_RA_PE_CTRL);
+	}
 
 	/* Data Store Engine configuration */
 
 	/* Reset all DSE threads */
 	writel(EIP197_DxE_THR_CTRL_RESET_PE,
-	       priv->base + EIP197_HIA_DSE_THR_CTRL);
+	       EIP197_HIA_DSE_THR(priv) + EIP197_HIA_DSE_THR_CTRL);
 
 	/* Wait for all DSE threads to complete */
-	while ((readl(priv->base + EIP197_HIA_DSE_THR_STAT) &
+	while ((readl(EIP197_HIA_DSE_THR(priv) + EIP197_HIA_DSE_THR_STAT) &
 		GENMASK(15, 12)) != GENMASK(15, 12))
 		;
 
@@ -330,15 +333,19 @@ static int safexcel_hw_init(struct safexcel_crypto_priv *priv)
 	val |= EIP197_HIA_DxE_CFG_MIN_DATA_SIZE(7) | EIP197_HIA_DxE_CFG_MAX_DATA_SIZE(8);
 	val |= EIP197_HIA_DxE_CFG_DATA_CACHE_CTRL(WR_CACHE_3BITS);
 	val |= EIP197_HIA_DSE_CFG_ALLWAYS_BUFFERABLE;
-	val |= EIP197_HIA_DSE_CFG_EN_SINGLE_WR;
-	writel(val, priv->base + EIP197_HIA_DSE_CFG);
+	/* FIXME: instability issues can occur for EIP97 but disabling it impact
+	 * performances.
+	 */
+	if (priv->version == EIP197)
+		val |= EIP197_HIA_DSE_CFG_EN_SINGLE_WR;
+	writel(val, EIP197_HIA_DSE(priv) + EIP197_HIA_DSE_CFG);
 
 	/* Leave the DSE threads reset state */
-	writel(0, priv->base + EIP197_HIA_DSE_THR_CTRL);
+	writel(0, EIP197_HIA_DSE_THR(priv) + EIP197_HIA_DSE_THR_CTRL);
 
 	/* Configure the procesing engine thresholds */
 	writel(EIP197_PE_OUT_DBUF_THRES_MIN(7) | EIP197_PE_OUT_DBUF_THRES_MAX(8),
-	       priv->base + EIP197_PE_OUT_DBUF_THRES);
+	       EIP197_PE(priv) + EIP197_PE_OUT_DBUF_THRES);
 
 	/* Processing Engine configuration */
 
@@ -348,73 +355,75 @@ static int safexcel_hw_init(struct safexcel_crypto_priv *priv)
 	val |= EIP197_ALG_AES_ECB | EIP197_ALG_AES_CBC;
 	val |= EIP197_ALG_SHA1 | EIP197_ALG_HMAC_SHA1;
 	val |= EIP197_ALG_SHA2;
-	writel(val, priv->base + EIP197_PE_EIP96_FUNCTION_EN);
+	writel(val, EIP197_PE(priv) + EIP197_PE_EIP96_FUNCTION_EN);
 
 	/* Command Descriptor Rings prepare */
 	for (i = 0; i < priv->config.rings; i++) {
 		/* Clear interrupts for this ring */
 		writel(GENMASK(31, 0),
-		       priv->base + EIP197_HIA_AIC_R_ENABLE_CLR(i));
+		       EIP197_HIA_AIC_R(priv) + EIP197_HIA_AIC_R_ENABLE_CLR(i));
 
 		/* Disable external triggering */
-		writel(0, priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_CFG);
+		writel(0, EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_CFG);
 
 		/* Clear the pending prepared counter */
 		writel(EIP197_xDR_PREP_CLR_COUNT,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_PREP_COUNT);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_PREP_COUNT);
 
 		/* Clear the pending processed counter */
 		writel(EIP197_xDR_PROC_CLR_COUNT,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_PROC_COUNT);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_PROC_COUNT);
 
 		writel(0,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_PREP_PNTR);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_PREP_PNTR);
 		writel(0,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_PROC_PNTR);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_PROC_PNTR);
 
 		writel((EIP197_DEFAULT_RING_SIZE * priv->config.cd_offset) << 2,
-		       priv->base + EIP197_HIA_CDR(i) + EIP197_HIA_xDR_RING_SIZE);
+		       EIP197_HIA_CDR(priv, i) + EIP197_HIA_xDR_RING_SIZE);
 	}
 
 	/* Result Descriptor Ring prepare */
 	for (i = 0; i < priv->config.rings; i++) {
 		/* Disable external triggering*/
-		writel(0, priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_CFG);
+		writel(0, EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_CFG);
 
 		/* Clear the pending prepared counter */
 		writel(EIP197_xDR_PREP_CLR_COUNT,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_PREP_COUNT);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_PREP_COUNT);
 
 		/* Clear the pending processed counter */
 		writel(EIP197_xDR_PROC_CLR_COUNT,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_PROC_COUNT);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_PROC_COUNT);
 
 		writel(0,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_PREP_PNTR);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_PREP_PNTR);
 		writel(0,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_PROC_PNTR);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_PROC_PNTR);
 
 		/* Ring size */
 		writel((EIP197_DEFAULT_RING_SIZE * priv->config.rd_offset) << 2,
-		       priv->base + EIP197_HIA_RDR(i) + EIP197_HIA_xDR_RING_SIZE);
+		       EIP197_HIA_RDR(priv, i) + EIP197_HIA_xDR_RING_SIZE);
 	}
 
 	/* Enable command descriptor rings */
 	writel(EIP197_DxE_THR_CTRL_EN | GENMASK(priv->config.rings - 1, 0),
-	       priv->base + EIP197_HIA_DFE_THR_CTRL);
+	       EIP197_HIA_DFE_THR(priv) + EIP197_HIA_DFE_THR_CTRL);
 
 	/* Enable result descriptor rings */
 	writel(EIP197_DxE_THR_CTRL_EN | GENMASK(priv->config.rings - 1, 0),
-	       priv->base + EIP197_HIA_DSE_THR_CTRL);
+	       EIP197_HIA_DSE_THR(priv) + EIP197_HIA_DSE_THR_CTRL);
 
 	/* Clear any HIA interrupt */
-	writel(GENMASK(30, 20), priv->base + EIP197_HIA_AIC_G_ACK);
+	writel(GENMASK(30, 20), EIP197_HIA_AIC_G(priv) + EIP197_HIA_AIC_G_ACK);
 
-	eip197_trc_cache_init(priv);
+	if (priv->version == EIP197) {
+		eip197_trc_cache_init(priv);
 
-	ret = eip197_load_firmwares(priv);
-	if (ret)
-		return ret;
+		ret = eip197_load_firmwares(priv);
+		if (ret)
+			return ret;
+	}
 
 	safexcel_hw_setup_cdesc_rings(priv);
 	safexcel_hw_setup_rdesc_rings(priv);
@@ -434,7 +443,7 @@ int safexcel_try_push_requests(struct safexcel_crypto_priv *priv, int ring,
 	/* Configure when we want an interrupt */
 	writel(EIP197_HIA_RDR_THRESH_PKT_MODE |
 	       EIP197_HIA_RDR_THRESH_PROC_PKT(coal),
-	       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_THRESH);
+	       EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_THRESH);
 
 	return coal;
 }
@@ -515,11 +524,11 @@ void safexcel_dequeue(struct safexcel_crypto_priv *priv, int ring)
 
 	/* let the RDR know we have pending descriptors */
 	writel((rdesc * priv->config.rd_offset) << 2,
-	       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PREP_COUNT);
+	       EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_PREP_COUNT);
 
 	/* let the CDR know we have pending descriptors */
 	writel((cdesc * priv->config.cd_offset) << 2,
-	       priv->base + EIP197_HIA_CDR(ring) + EIP197_HIA_xDR_PREP_COUNT);
+	       EIP197_HIA_CDR(priv, ring) + EIP197_HIA_xDR_PREP_COUNT);
 }
 
 void safexcel_free_context(struct safexcel_crypto_priv *priv,
@@ -620,7 +629,7 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 handle_results:
 	tot_descs = 0;
 
-	nreq = readl(priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
+	nreq = readl(EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_PROC_COUNT);
 	nreq >>= EIP197_xDR_PROC_xD_PKT_OFFSET;
 	nreq &= EIP197_xDR_PROC_xD_PKT_MASK;
 	if (!nreq)
@@ -656,7 +665,7 @@ static inline void safexcel_handle_result_descriptor(struct safexcel_crypto_priv
 	if (i) {
 		writel(EIP197_xDR_PROC_xD_PKT(i) |
 		       EIP197_xDR_PROC_xD_COUNT(tot_descs * priv->config.rd_offset),
-		       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_PROC_COUNT);
+		       EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_PROC_COUNT);
 	}
 
 	/* If the number of requests overflowed the counter, try to proceed more
@@ -698,13 +707,13 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 	int ring = irq_data->ring, rc = IRQ_NONE;
 	u32 status, stat;
 
-	status = readl(priv->base + EIP197_HIA_AIC_R_ENABLED_STAT(ring));
+	status = readl(EIP197_HIA_AIC_R(priv) + EIP197_HIA_AIC_R_ENABLED_STAT(ring));
 	if (!status)
 		return rc;
 
 	/* RDR interrupts */
 	if (status & EIP197_RDR_IRQ(ring)) {
-		stat = readl(priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_STAT);
+		stat = readl(EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_STAT);
 
 		if (unlikely(stat & EIP197_xDR_ERR)) {
 			/*
@@ -719,11 +728,11 @@ static irqreturn_t safexcel_irq_ring(int irq, void *data)
 
 		/* ACK the interrupts */
 		writel(stat & 0xff,
-		       priv->base + EIP197_HIA_RDR(ring) + EIP197_HIA_xDR_STAT);
+		       EIP197_HIA_RDR(priv, ring) + EIP197_HIA_xDR_STAT);
 	}
 
 	/* ACK the interrupts */
-	writel(status, priv->base + EIP197_HIA_AIC_R_ACK(ring));
+	writel(status, EIP197_HIA_AIC_R(priv) + EIP197_HIA_AIC_R_ACK(ring));
 
 	return rc;
 }
@@ -819,11 +828,11 @@ static void safexcel_configure(struct safexcel_crypto_priv *priv)
 {
 	u32 val, mask;
 
-	val = readl(priv->base + EIP197_HIA_OPTIONS);
+	val = readl(EIP197_HIA_AIC_G(priv) + EIP197_HIA_OPTIONS);
 	val = (val & GENMASK(27, 25)) >> 25;
 	mask = BIT(val) - 1;
 
-	val = readl(priv->base + EIP197_HIA_OPTIONS);
+	val = readl(EIP197_HIA_AIC_G(priv) + EIP197_HIA_OPTIONS);
 	priv->config.rings = min_t(u32, val & GENMASK(3, 0), max_rings);
 
 	priv->config.cd_size = (sizeof(struct safexcel_command_desc) / sizeof(u32));
@@ -833,6 +842,35 @@ static void safexcel_configure(struct safexcel_crypto_priv *priv)
 	priv->config.rd_offset = (priv->config.rd_size + mask) & ~mask;
 }
 
+static void safexcel_init_register_offsets(struct safexcel_crypto_priv *priv)
+{
+	struct safexcel_register_offsets *offsets = &priv->offsets;
+
+	if (priv->version == EIP197) {
+		offsets->hia_aic	= EIP197_HIA_AIC_BASE;
+		offsets->hia_aic_g	= EIP197_HIA_AIC_G_BASE;
+		offsets->hia_aic_r	= EIP197_HIA_AIC_R_BASE;
+		offsets->hia_aic_xdr	= EIP197_HIA_AIC_xDR_BASE;
+		offsets->hia_dfe	= EIP197_HIA_DFE_BASE;
+		offsets->hia_dfe_thr	= EIP197_HIA_DFE_THR_BASE;
+		offsets->hia_dse	= EIP197_HIA_DSE_BASE;
+		offsets->hia_dse_thr	= EIP197_HIA_DSE_THR_BASE;
+		offsets->hia_gen_cfg	= EIP197_HIA_GEN_CFG_BASE;
+		offsets->pe		= EIP197_PE_BASE;
+	} else {
+		offsets->hia_aic	= EIP97_HIA_AIC_BASE;
+		offsets->hia_aic_g	= EIP97_HIA_AIC_G_BASE;
+		offsets->hia_aic_r	= EIP97_HIA_AIC_R_BASE;
+		offsets->hia_aic_xdr	= EIP97_HIA_AIC_xDR_BASE;
+		offsets->hia_dfe	= EIP97_HIA_DFE_BASE;
+		offsets->hia_dfe_thr	= EIP97_HIA_DFE_THR_BASE;
+		offsets->hia_dse	= EIP97_HIA_DSE_BASE;
+		offsets->hia_dse_thr	= EIP97_HIA_DSE_THR_BASE;
+		offsets->hia_gen_cfg	= EIP97_HIA_GEN_CFG_BASE;
+		offsets->pe		= EIP97_PE_BASE;
+	}
+}
+
 static int safexcel_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -845,6 +883,9 @@ static int safexcel_probe(struct platform_device *pdev)
 		return -ENOMEM;
 
 	priv->dev = dev;
+	priv->version = (enum safexcel_eip_version)of_device_get_match_data(dev);
+
+	safexcel_init_register_offsets(priv);
 
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	priv->base = devm_ioremap_resource(dev, res);
@@ -971,7 +1012,14 @@ static int safexcel_remove(struct platform_device *pdev)
 }
 
 static const struct of_device_id safexcel_of_match_table[] = {
-	{ .compatible = "inside-secure,safexcel-eip197" },
+	{
+		.compatible = "inside-secure,safexcel-eip97",
+		.data = (void *)EIP97,
+	},
+	{
+		.compatible = "inside-secure,safexcel-eip197",
+		.data = (void *)EIP197,
+	},
 	{},
 };
 
diff --git a/drivers/crypto/inside-secure/safexcel.h b/drivers/crypto/inside-secure/safexcel.h
index d4955abf873b..4e219c21608b 100644
--- a/drivers/crypto/inside-secure/safexcel.h
+++ b/drivers/crypto/inside-secure/safexcel.h
@@ -28,55 +28,94 @@
 #define EIP197_GFP_FLAGS(base)	((base).flags & CRYPTO_TFM_REQ_MAY_SLEEP ? \
 				 GFP_KERNEL : GFP_ATOMIC)
 
+/* Register base offsets */
+#define EIP197_HIA_AIC(priv)		((priv)->base + (priv)->offsets.hia_aic)
+#define EIP197_HIA_AIC_G(priv)		((priv)->base + (priv)->offsets.hia_aic_g)
+#define EIP197_HIA_AIC_R(priv)		((priv)->base + (priv)->offsets.hia_aic_r)
+#define EIP197_HIA_AIC_xDR(priv)	((priv)->base + (priv)->offsets.hia_aic_xdr)
+#define EIP197_HIA_DFE(priv)		((priv)->base + (priv)->offsets.hia_dfe)
+#define EIP197_HIA_DFE_THR(priv)	((priv)->base + (priv)->offsets.hia_dfe_thr)
+#define EIP197_HIA_DSE(priv)		((priv)->base + (priv)->offsets.hia_dse)
+#define EIP197_HIA_DSE_THR(priv)	((priv)->base + (priv)->offsets.hia_dse_thr)
+#define EIP197_HIA_GEN_CFG(priv)	((priv)->base + (priv)->offsets.hia_gen_cfg)
+#define EIP197_PE(priv)			((priv)->base + (priv)->offsets.pe)
+
+/* EIP197 base offsets */
+#define EIP197_HIA_AIC_BASE		0x90000
+#define EIP197_HIA_AIC_G_BASE		0x90000
+#define EIP197_HIA_AIC_R_BASE		0x90800
+#define EIP197_HIA_AIC_xDR_BASE		0x80000
+#define EIP197_HIA_DFE_BASE		0x8c000
+#define EIP197_HIA_DFE_THR_BASE		0x8c040
+#define EIP197_HIA_DSE_BASE		0x8d000
+#define EIP197_HIA_DSE_THR_BASE		0x8d040
+#define EIP197_HIA_GEN_CFG_BASE		0xf0000
+#define EIP197_PE_BASE			0xa0000
+
+/* EIP97 base offsets */
+#define EIP97_HIA_AIC_BASE		0x0
+#define EIP97_HIA_AIC_G_BASE		0x0
+#define EIP97_HIA_AIC_R_BASE		0x0
+#define EIP97_HIA_AIC_xDR_BASE		0x0
+#define EIP97_HIA_DFE_BASE		0xf000
+#define EIP97_HIA_DFE_THR_BASE		0xf200
+#define EIP97_HIA_DSE_BASE		0xf400
+#define EIP97_HIA_DSE_THR_BASE		0xf600
+#define EIP97_HIA_GEN_CFG_BASE		0x10000
+#define EIP97_PE_BASE			0x10000
+
 /* CDR/RDR register offsets */
-#define EIP197_HIA_xDR_OFF(r)			(0x80000 + (r) * 0x1000)
-#define EIP197_HIA_CDR(r)			(EIP197_HIA_xDR_OFF(r))
-#define EIP197_HIA_RDR(r)			(EIP197_HIA_xDR_OFF(r) + 0x800)
-#define EIP197_HIA_xDR_RING_BASE_ADDR_LO	0x0
-#define EIP197_HIA_xDR_RING_BASE_ADDR_HI	0x4
-#define EIP197_HIA_xDR_RING_SIZE		0x18
-#define EIP197_HIA_xDR_DESC_SIZE		0x1c
-#define EIP197_HIA_xDR_CFG			0x20
-#define EIP197_HIA_xDR_DMA_CFG			0x24
-#define EIP197_HIA_xDR_THRESH			0x28
-#define EIP197_HIA_xDR_PREP_COUNT		0x2c
-#define EIP197_HIA_xDR_PROC_COUNT		0x30
-#define EIP197_HIA_xDR_PREP_PNTR		0x34
-#define EIP197_HIA_xDR_PROC_PNTR		0x38
-#define EIP197_HIA_xDR_STAT			0x3c
+#define EIP197_HIA_xDR_OFF(priv, r)		(EIP197_HIA_AIC_xDR(priv) + (r) * 0x1000)
+#define EIP197_HIA_CDR(priv, r)			(EIP197_HIA_xDR_OFF(priv, r))
+#define EIP197_HIA_RDR(priv, r)			(EIP197_HIA_xDR_OFF(priv, r) + 0x800)
+#define EIP197_HIA_xDR_RING_BASE_ADDR_LO	0x0000
+#define EIP197_HIA_xDR_RING_BASE_ADDR_HI	0x0004
+#define EIP197_HIA_xDR_RING_SIZE		0x0018
+#define EIP197_HIA_xDR_DESC_SIZE		0x001c
+#define EIP197_HIA_xDR_CFG			0x0020
+#define EIP197_HIA_xDR_DMA_CFG			0x0024
+#define EIP197_HIA_xDR_THRESH			0x0028
+#define EIP197_HIA_xDR_PREP_COUNT		0x002c
+#define EIP197_HIA_xDR_PROC_COUNT		0x0030
+#define EIP197_HIA_xDR_PREP_PNTR		0x0034
+#define EIP197_HIA_xDR_PROC_PNTR		0x0038
+#define EIP197_HIA_xDR_STAT			0x003c
 
 /* register offsets */
-#define EIP197_HIA_DFE_CFG			0x8c000
-#define EIP197_HIA_DFE_THR_CTRL			0x8c040
-#define EIP197_HIA_DFE_THR_STAT			0x8c044
-#define EIP197_HIA_DSE_CFG			0x8d000
-#define EIP197_HIA_DSE_THR_CTRL			0x8d040
-#define EIP197_HIA_DSE_THR_STAT			0x8d044
-#define EIP197_HIA_RA_PE_CTRL			0x90010
-#define EIP197_HIA_RA_PE_STAT			0x90014
+#define EIP197_HIA_DFE_CFG			0x0000
+#define EIP197_HIA_DFE_THR_CTRL			0x0000
+#define EIP197_HIA_DFE_THR_STAT			0x0004
+#define EIP197_HIA_DSE_CFG			0x0000
+#define EIP197_HIA_DSE_THR_CTRL			0x0000
+#define EIP197_HIA_DSE_THR_STAT			0x0004
+#define EIP197_HIA_RA_PE_CTRL			0x0010
+#define EIP197_HIA_RA_PE_STAT			0x0014
 #define EIP197_HIA_AIC_R_OFF(r)			((r) * 0x1000)
-#define EIP197_HIA_AIC_R_ENABLE_CTRL(r)		(0x9e808 - EIP197_HIA_AIC_R_OFF(r))
-#define EIP197_HIA_AIC_R_ENABLED_STAT(r)	(0x9e810 - EIP197_HIA_AIC_R_OFF(r))
-#define EIP197_HIA_AIC_R_ACK(r)			(0x9e810 - EIP197_HIA_AIC_R_OFF(r))
-#define EIP197_HIA_AIC_R_ENABLE_CLR(r)		(0x9e814 - EIP197_HIA_AIC_R_OFF(r))
-#define EIP197_HIA_AIC_G_ENABLE_CTRL		0x9f808
-#define EIP197_HIA_AIC_G_ENABLED_STAT		0x9f810
-#define EIP197_HIA_AIC_G_ACK			0x9f810
-#define EIP197_HIA_MST_CTRL			0x9fff4
-#define EIP197_HIA_OPTIONS			0x9fff8
-#define EIP197_HIA_VERSION			0x9fffc
-#define EIP197_PE_IN_DBUF_THRES			0xa0000
-#define EIP197_PE_IN_TBUF_THRES			0xa0100
-#define EIP197_PE_ICE_SCRATCH_RAM		0xa0800
-#define EIP197_PE_ICE_PUE_CTRL			0xa0c80
-#define EIP197_PE_ICE_SCRATCH_CTRL		0xa0d04
-#define EIP197_PE_ICE_FPP_CTRL			0xa0d80
-#define EIP197_PE_ICE_RAM_CTRL			0xa0ff0
-#define EIP197_PE_EIP96_FUNCTION_EN		0xa1004
-#define EIP197_PE_EIP96_CONTEXT_CTRL		0xa1008
-#define EIP197_PE_EIP96_CONTEXT_STAT		0xa100c
-#define EIP197_PE_OUT_DBUF_THRES		0xa1c00
-#define EIP197_PE_OUT_TBUF_THRES		0xa1d00
+#define EIP197_HIA_AIC_R_ENABLE_CTRL(r)		(0xe008 - EIP197_HIA_AIC_R_OFF(r))
+#define EIP197_HIA_AIC_R_ENABLED_STAT(r)	(0xe010 - EIP197_HIA_AIC_R_OFF(r))
+#define EIP197_HIA_AIC_R_ACK(r)			(0xe010 - EIP197_HIA_AIC_R_OFF(r))
+#define EIP197_HIA_AIC_R_ENABLE_CLR(r)		(0xe014 - EIP197_HIA_AIC_R_OFF(r))
+#define EIP197_HIA_AIC_G_ENABLE_CTRL		0xf808
+#define EIP197_HIA_AIC_G_ENABLED_STAT		0xf810
+#define EIP197_HIA_AIC_G_ACK			0xf810
+#define EIP197_HIA_MST_CTRL			0xfff4
+#define EIP197_HIA_OPTIONS			0xfff8
+#define EIP197_HIA_VERSION			0xfffc
+#define EIP197_PE_IN_DBUF_THRES			0x0000
+#define EIP197_PE_IN_TBUF_THRES			0x0100
+#define EIP197_PE_ICE_SCRATCH_RAM		0x0800
+#define EIP197_PE_ICE_PUE_CTRL			0x0c80
+#define EIP197_PE_ICE_SCRATCH_CTRL		0x0d04
+#define EIP197_PE_ICE_FPP_CTRL			0x0d80
+#define EIP197_PE_ICE_RAM_CTRL			0x0ff0
+#define EIP197_PE_EIP96_FUNCTION_EN		0x1004
+#define EIP197_PE_EIP96_CONTEXT_CTRL		0x1008
+#define EIP197_PE_EIP96_CONTEXT_STAT		0x100c
+#define EIP197_PE_OUT_DBUF_THRES		0x1c00
+#define EIP197_PE_OUT_TBUF_THRES		0x1d00
+#define EIP197_MST_CTRL				0xfff4
+
+/* EIP197-specific registers, no indirection */
 #define EIP197_CLASSIFICATION_RAMS		0xe0000
 #define EIP197_TRC_CTRL				0xf0800
 #define EIP197_TRC_LASTRES			0xf0804
@@ -90,7 +129,6 @@
 #define EIP197_TRC_ECCDATASTAT			0xf083c
 #define EIP197_TRC_ECCDATA			0xf0840
 #define EIP197_CS_RAM_CTRL			0xf7ff0
-#define EIP197_MST_CTRL				0xffff4
 
 /* EIP197_HIA_xDR_DESC_SIZE */
 #define EIP197_xDR_DESC_MODE_64BIT		BIT(31)
@@ -465,12 +503,33 @@ struct safexcel_work_data {
 	int ring;
 };
 
+enum safexcel_eip_version {
+	EIP97,
+	EIP197,
+};
+
+struct safexcel_register_offsets {
+	u32 hia_aic;
+	u32 hia_aic_g;
+	u32 hia_aic_r;
+	u32 hia_aic_xdr;
+	u32 hia_dfe;
+	u32 hia_dfe_thr;
+	u32 hia_dse;
+	u32 hia_dse_thr;
+	u32 hia_gen_cfg;
+	u32 pe;
+};
+
 struct safexcel_crypto_priv {
 	void __iomem *base;
 	struct device *dev;
 	struct clk *clk;
 	struct safexcel_config config;
 
+	enum safexcel_eip_version version;
+	struct safexcel_register_offsets offsets;
+
 	/* context DMA pool */
 	struct dma_pool *context_pool;
 
diff --git a/drivers/crypto/inside-secure/safexcel_cipher.c b/drivers/crypto/inside-secure/safexcel_cipher.c
index 0e5cc230e49a..63a8768ed2ae 100644
--- a/drivers/crypto/inside-secure/safexcel_cipher.c
+++ b/drivers/crypto/inside-secure/safexcel_cipher.c
@@ -69,6 +69,7 @@ static int safexcel_aes_setkey(struct crypto_skcipher *ctfm, const u8 *key,
 {
 	struct crypto_tfm *tfm = crypto_skcipher_tfm(ctfm);
 	struct safexcel_cipher_ctx *ctx = crypto_tfm_ctx(tfm);
+	struct safexcel_crypto_priv *priv = ctx->priv;
 	struct crypto_aes_ctx aes;
 	int ret, i;
 
@@ -78,7 +79,7 @@ static int safexcel_aes_setkey(struct crypto_skcipher *ctfm, const u8 *key,
 		return ret;
 	}
 
-	if (ctx->base.ctxr_dma) {
+	if (priv->version == EIP197 && ctx->base.ctxr_dma) {
 		for (i = 0; i < len / sizeof(u32); i++) {
 			if (ctx->key[i] != cpu_to_le32(aes.key_enc[i])) {
 				ctx->base.needs_inv = true;
@@ -411,9 +412,13 @@ static int safexcel_send(struct crypto_async_request *async,
 			 int *commands, int *results)
 {
 	struct skcipher_request *req = skcipher_request_cast(async);
+	struct safexcel_cipher_ctx *ctx = crypto_tfm_ctx(req->base.tfm);
 	struct safexcel_cipher_req *sreq = skcipher_request_ctx(req);
+	struct safexcel_crypto_priv *priv = ctx->priv;
 	int ret;
 
+	BUG_ON(priv->version == EIP97 && sreq->needs_inv);
+
 	if (sreq->needs_inv)
 		ret = safexcel_cipher_send_inv(async, ring, request,
 					       commands, results);
@@ -476,7 +481,7 @@ static int safexcel_aes(struct skcipher_request *req,
 	ctx->mode = mode;
 
 	if (ctx->base.ctxr) {
-		if (ctx->base.needs_inv) {
+		if (priv->version == EIP197 && ctx->base.needs_inv) {
 			sreq->needs_inv = true;
 			ctx->base.needs_inv = false;
 		}
@@ -544,9 +549,14 @@ static void safexcel_skcipher_cra_exit(struct crypto_tfm *tfm)
 
 	memzero_explicit(ctx->base.ctxr->data, 8 * sizeof(u32));
 
-	ret = safexcel_cipher_exit_inv(tfm);
-	if (ret)
-		dev_warn(priv->dev, "cipher: invalidation error %d\n", ret);
+	if (priv->version == EIP197) {
+		ret = safexcel_cipher_exit_inv(tfm);
+		if (ret)
+			dev_warn(priv->dev, "cipher: invalidation error %d\n", ret);
+	} else {
+		dma_pool_free(priv->context_pool, ctx->base.ctxr,
+			      ctx->base.ctxr_dma);
+	}
 }
 
 struct safexcel_alg_template safexcel_alg_ecb_aes = {
diff --git a/drivers/crypto/inside-secure/safexcel_hash.c b/drivers/crypto/inside-secure/safexcel_hash.c
index 2fb5bc6b6268..d94614afc53d 100644
--- a/drivers/crypto/inside-secure/safexcel_hash.c
+++ b/drivers/crypto/inside-secure/safexcel_hash.c
@@ -397,6 +397,8 @@ static int safexcel_handle_result(struct safexcel_crypto_priv *priv, int ring,
 	struct safexcel_ahash_req *req = ahash_request_ctx(areq);
 	int err;
 
+	BUG_ON(priv->version == EIP97 && req->needs_inv);
+
 	if (req->needs_inv) {
 		req->needs_inv = false;
 		err = safexcel_handle_inv_result(priv, ring, async,
@@ -528,7 +530,8 @@ static int safexcel_ahash_enqueue(struct ahash_request *areq)
 	req->needs_inv = false;
 
 	if (ctx->base.ctxr) {
-		if (!ctx->base.needs_inv && req->processed &&
+		if (priv->version == EIP197 &&
+		    !ctx->base.needs_inv && req->processed &&
 		    ctx->digest == CONTEXT_CONTROL_DIGEST_PRECOMPUTED)
 			/* We're still setting needs_inv here, even though it is
 			 * cleared right away, because the needs_inv flag can be
@@ -721,9 +724,14 @@ static void safexcel_ahash_cra_exit(struct crypto_tfm *tfm)
 	if (!ctx->base.ctxr)
 		return;
 
-	ret = safexcel_ahash_exit_inv(tfm);
-	if (ret)
-		dev_warn(priv->dev, "hash: invalidation error %d\n", ret);
+	if (priv->version == EIP197) {
+		ret = safexcel_ahash_exit_inv(tfm);
+		if (ret)
+			dev_warn(priv->dev, "hash: invalidation error %d\n", ret);
+	} else {
+		dma_pool_free(priv->context_pool, ctx->base.ctxr,
+			      ctx->base.ctxr_dma);
+	}
 }
 
 struct safexcel_alg_template safexcel_alg_sha1 = {
@@ -927,6 +935,7 @@ static int safexcel_hmac_sha1_setkey(struct crypto_ahash *tfm, const u8 *key,
 				     unsigned int keylen)
 {
 	struct safexcel_ahash_ctx *ctx = crypto_tfm_ctx(crypto_ahash_tfm(tfm));
+	struct safexcel_crypto_priv *priv = ctx->priv;
 	struct safexcel_ahash_export_state istate, ostate;
 	int ret, i;
 
@@ -934,7 +943,7 @@ static int safexcel_hmac_sha1_setkey(struct crypto_ahash *tfm, const u8 *key,
 	if (ret)
 		return ret;
 
-	if (ctx->base.ctxr) {
+	if (priv->version == EIP197 && ctx->base.ctxr) {
 		for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(u32); i++) {
 			if (ctx->ipad[i] != le32_to_cpu(istate.state[i]) ||
 			    ctx->opad[i] != le32_to_cpu(ostate.state[i])) {
-- 
2.14.3

[PATCH 17/17] Documentation/bindings: crypto: document the SafeXcel EIP97 compatible

[Antoine Tenart]

This patch adds the SafeXcel EIP97 compatible to the Inside Secure
device tree bindings documentation.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
---
 Documentation/devicetree/bindings/crypto/inside-secure-safexcel.txt | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/crypto/inside-secure-safexcel.txt b/Documentation/devicetree/bindings/crypto/inside-secure-safexcel.txt
index fbc07d12322f..30c3ce6b502e 100644
--- a/Documentation/devicetree/bindings/crypto/inside-secure-safexcel.txt
+++ b/Documentation/devicetree/bindings/crypto/inside-secure-safexcel.txt
@@ -1,7 +1,8 @@
 Inside Secure SafeXcel cryptographic engine
 
 Required properties:
-- compatible: Should be "inside-secure,safexcel-eip197".
+- compatible: Should be "inside-secure,safexcel-eip197" or
+              "inside-secure,safexcel-eip97".
 - reg: Base physical address of the engine and length of memory mapped region.
 - interrupts: Interrupt numbers for the rings and engine.
 - interrupt-names: Should be "ring0", "ring1", "ring2", "ring3", "eip", "mem".
-- 
2.14.3

Re: [PATCH 00/17] crypto: inside-secure - various improvements

[Herbert Xu]

On Thu, Dec 14, 2017 at 03:26:42PM +0100, Antoine Tenart wrote:
> Hi Herbert,
> 
> This series depends on the previous one, "crypto: inside-secure - set of
> fixes"[1]. The series contains various improvements to the Inside Secure
> SafeXcel driver, which are bundled in a single series to avoid any
> conflict.
> 
> - Patches 1-4 are cosmetic fixes (typo, more comments and dead code removal).
> - Patches 5-8 are small improvements, mostly to improve the invalidation
>   path.
> - Patches 9-15 improve the request processing in the driver, by moving
>   the requests dequeing into a workqueue and the result handling into a
>   threaded IRQ thread. Other small improvements are made in the same
>   direction.
> - Patches 16-17 introduce the support of another revision of this crypto
>   engine, similar to the one already supported. This crypto engine can be
>   found on Marvell boards. (The corresponding device tree patches will follow
>   once this series lands in).

Patches 1-16 applied.  If patch 17 needs to go through the crypto
tree please let me know.

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH 00/17] crypto: inside-secure - various improvements

[Antoine Tenart]

Hi Herbert,

On Fri, Dec 22, 2017 at 08:11:54PM +1100, Herbert Xu wrote:
> On Thu, Dec 14, 2017 at 03:26:42PM +0100, Antoine Tenart wrote:
> > 
> > This series depends on the previous one, "crypto: inside-secure - set of
> > fixes"[1]. The series contains various improvements to the Inside Secure
> > SafeXcel driver, which are bundled in a single series to avoid any
> > conflict.
> > 
> > - Patches 1-4 are cosmetic fixes (typo, more comments and dead code removal).
> > - Patches 5-8 are small improvements, mostly to improve the invalidation
> >   path.
> > - Patches 9-15 improve the request processing in the driver, by moving
> >   the requests dequeing into a workqueue and the result handling into a
> >   threaded IRQ thread. Other small improvements are made in the same
> >   direction.
> > - Patches 16-17 introduce the support of another revision of this crypto
> >   engine, similar to the one already supported. This crypto engine can be
> >   found on Marvell boards. (The corresponding device tree patches will follow
> >   once this series lands in).
> 
> Patches 1-16 applied.  If patch 17 needs to go through the crypto
> tree please let me know.

Thanks! Yes patch 17 should got through the crypto tree as well.

Thanks,
Antoine

-- 
Antoine Ténart, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: [PATCH 00/17] crypto: inside-secure - various improvements

[Antoine Tenart]

Hi Herbert,

On Fri, Dec 22, 2017 at 03:28:38PM +0100, Antoine Tenart wrote:
> On Fri, Dec 22, 2017 at 08:11:54PM +1100, Herbert Xu wrote:
> > 
> > Patches 1-16 applied.  If patch 17 needs to go through the crypto
> > tree please let me know.
> 
> Thanks! Yes patch 17 should got through the crypto tree as well.

I'm not sure you took patch 17 in your tree. Could you check?

Thanks!
Antoine

-- 
Antoine Ténart, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: [PATCH 00/17] crypto: inside-secure - various improvements

[Herbert Xu]

On Fri, Jan 12, 2018 at 02:02:51PM +0100, Antoine Tenart wrote:
> Hi Herbert,
> 
> On Fri, Dec 22, 2017 at 03:28:38PM +0100, Antoine Tenart wrote:
> > On Fri, Dec 22, 2017 at 08:11:54PM +1100, Herbert Xu wrote:
> > > 
> > > Patches 1-16 applied.  If patch 17 needs to go through the crypto
> > > tree please let me know.
> > 
> > Thanks! Yes patch 17 should got through the crypto tree as well.
> 
> I'm not sure you took patch 17 in your tree. Could you check?

Thanks for the reminder.  I have added it now.
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt