crypto: caam - Remove GFP_DMA and add DMA alignment padding

GFP_DMA does not guarantee that the returned memory is aligned
for DMA.  It should be removed where it is superfluous.

However, kmalloc may start returning DMA-unaligned memory in future
so fix this by adding the alignment by hand.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

drivers/crypto/caam/blob_gen.c

@@ -83,7 +83,7 @@ int caam_process_blob(struct caam_blob_priv *priv,
 		output_len = info->input_len - CAAM_BLOB_OVERHEAD;
 	}
 
-	desc = kzalloc(CAAM_BLOB_DESC_BYTES_MAX, GFP_KERNEL | GFP_DMA);
+	desc = kzalloc(CAAM_BLOB_DESC_BYTES_MAX, GFP_KERNEL);
 	if (!desc)
 		return -ENOMEM;
 

drivers/crypto/caam/caamalg.c

@@ -59,6 +59,8 @@
 #include <crypto/engine.h>
 #include <crypto/xts.h>
 #include <asm/unaligned.h>
+#include <linux/dma-mapping.h>
+#include <linux/kernel.h>
 
 /*
  * crypto alg
@@ -1379,8 +1381,7 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
 	sec4_sg_bytes = sec4_sg_len * sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kzalloc(sizeof(*edesc) + desc_bytes + sec4_sg_bytes,
-			GFP_DMA | flags);
+	edesc = kzalloc(sizeof(*edesc) + desc_bytes + sec4_sg_bytes, flags);
 	if (!edesc) {
 		caam_unmap(jrdev, req->src, req->dst, src_nents, dst_nents, 0,
 			   0, 0, 0);
@@ -1608,6 +1609,7 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req,
 	u8 *iv;
 	int ivsize = crypto_skcipher_ivsize(skcipher);
 	int dst_sg_idx, sec4_sg_ents, sec4_sg_bytes;
+	unsigned int aligned_size;
 
 	src_nents = sg_nents_for_len(req->src, req->cryptlen);
 	if (unlikely(src_nents < 0)) {
@@ -1681,15 +1683,18 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req,
 	/*
 	 * allocate space for base edesc and hw desc commands, link tables, IV
 	 */
-	edesc = kzalloc(sizeof(*edesc) + desc_bytes + sec4_sg_bytes + ivsize,
-			GFP_DMA | flags);
-	if (!edesc) {
+	aligned_size = ALIGN(ivsize, __alignof__(*edesc));
+	aligned_size += sizeof(*edesc) + desc_bytes + sec4_sg_bytes;
+	aligned_size = ALIGN(aligned_size, dma_get_cache_alignment());
+	iv = kzalloc(aligned_size, flags);
+	if (!iv) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
 		caam_unmap(jrdev, req->src, req->dst, src_nents, dst_nents, 0,
 			   0, 0, 0);
 		return ERR_PTR(-ENOMEM);
 	}
 
+	edesc = (void *)(iv + ALIGN(ivsize, __alignof__(*edesc)));
 	edesc->src_nents = src_nents;
 	edesc->dst_nents = dst_nents;
 	edesc->mapped_src_nents = mapped_src_nents;
@@ -1701,7 +1706,6 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req,
 
 	/* Make sure IV is located in a DMAable area */
 	if (ivsize) {
-		iv = (u8 *)edesc->sec4_sg + sec4_sg_bytes;
 		memcpy(iv, req->iv, ivsize);
 
 		iv_dma = dma_map_single(jrdev, iv, ivsize, DMA_BIDIRECTIONAL);

drivers/crypto/caam/caamalg_qi.c

@@ -20,6 +20,8 @@
 #include "caamalg_desc.h"
 #include <crypto/xts.h>
 #include <asm/unaligned.h>
+#include <linux/dma-mapping.h>
+#include <linux/kernel.h>
 
 /*
  * crypto alg
@@ -959,7 +961,7 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
 		return (struct aead_edesc *)drv_ctx;
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = qi_cache_alloc(GFP_DMA | flags);
+	edesc = qi_cache_alloc(flags);
 	if (unlikely(!edesc)) {
 		dev_err(qidev, "could not allocate extended descriptor\n");
 		return ERR_PTR(-ENOMEM);
@@ -1317,8 +1319,9 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req,
 		qm_sg_ents = 1 + pad_sg_nents(qm_sg_ents);
 
 	qm_sg_bytes = qm_sg_ents * sizeof(struct qm_sg_entry);
-	if (unlikely(offsetof(struct skcipher_edesc, sgt) + qm_sg_bytes +
-		     ivsize > CAAM_QI_MEMCACHE_SIZE)) {
+	if (unlikely(ALIGN(ivsize, __alignof__(*edesc)) +
+		     offsetof(struct skcipher_edesc, sgt) + qm_sg_bytes >
+		     CAAM_QI_MEMCACHE_SIZE)) {
 		dev_err(qidev, "No space for %d S/G entries and/or %dB IV\n",
 			qm_sg_ents, ivsize);
 		caam_unmap(qidev, req->src, req->dst, src_nents, dst_nents, 0,
@@ -1327,17 +1330,18 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req,
 	}
 
 	/* allocate space for base edesc, link tables and IV */
-	edesc = qi_cache_alloc(GFP_DMA | flags);
-	if (unlikely(!edesc)) {
+	iv = qi_cache_alloc(flags);
+	if (unlikely(!iv)) {
 		dev_err(qidev, "could not allocate extended descriptor\n");
 		caam_unmap(qidev, req->src, req->dst, src_nents, dst_nents, 0,
 			   0, DMA_NONE, 0, 0);
 		return ERR_PTR(-ENOMEM);
 	}
 
+	edesc = (void *)(iv + ALIGN(ivsize, __alignof__(*edesc)));
+
 	/* Make sure IV is located in a DMAable area */
 	sg_table = &edesc->sgt[0];
-	iv = (u8 *)(sg_table + qm_sg_ents);
 	memcpy(iv, req->iv, ivsize);
 
 	iv_dma = dma_map_single(qidev, iv, ivsize, DMA_BIDIRECTIONAL);

drivers/crypto/caam/caamalg_qi2.c

@@ -16,7 +16,9 @@
 #include "caamalg_desc.h"
 #include "caamhash_desc.h"
 #include "dpseci-debugfs.h"
+#include <linux/dma-mapping.h>
 #include <linux/fsl/mc.h>
+#include <linux/kernel.h>
 #include <soc/fsl/dpaa2-io.h>
 #include <soc/fsl/dpaa2-fd.h>
 #include <crypto/xts.h>
@@ -370,7 +372,7 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
 	struct dpaa2_sg_entry *sg_table;
 
 	/* allocate space for base edesc, link tables and IV */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (unlikely(!edesc)) {
 		dev_err(dev, "could not allocate extended descriptor\n");
 		return ERR_PTR(-ENOMEM);
@@ -1189,7 +1191,7 @@ static struct skcipher_edesc *skcipher_edesc_alloc(struct skcipher_request *req)
 	}
 
 	/* allocate space for base edesc, link tables and IV */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (unlikely(!edesc)) {
 		dev_err(dev, "could not allocate extended descriptor\n");
 		caam_unmap(dev, req->src, req->dst, src_nents, dst_nents, 0,
@@ -3220,14 +3222,14 @@ static int hash_digest_key(struct caam_hash_ctx *ctx, u32 *keylen, u8 *key,
 	int ret = -ENOMEM;
 	struct dpaa2_fl_entry *in_fle, *out_fle;
 
-	req_ctx = kzalloc(sizeof(*req_ctx), GFP_KERNEL | GFP_DMA);
+	req_ctx = kzalloc(sizeof(*req_ctx), GFP_KERNEL);
 	if (!req_ctx)
 		return -ENOMEM;
 
 	in_fle = &req_ctx->fd_flt[1];
 	out_fle = &req_ctx->fd_flt[0];
 
-	flc = kzalloc(sizeof(*flc), GFP_KERNEL | GFP_DMA);
+	flc = kzalloc(sizeof(*flc), GFP_KERNEL);
 	if (!flc)
 		goto err_flc;
 
@@ -3316,7 +3318,13 @@ static int ahash_setkey(struct crypto_ahash *ahash, const u8 *key,
 	dev_dbg(ctx->dev, "keylen %d blocksize %d\n", keylen, blocksize);
 
 	if (keylen > blocksize) {
-		hashed_key = kmemdup(key, keylen, GFP_KERNEL | GFP_DMA);
+		unsigned int aligned_len =
+			ALIGN(keylen, dma_get_cache_alignment());
+
+		if (aligned_len < keylen)
+			return -EOVERFLOW;
+
+		hashed_key = kmemdup(key, aligned_len, GFP_KERNEL);
 		if (!hashed_key)
 			return -ENOMEM;
 		ret = hash_digest_key(ctx, &keylen, hashed_key, digestsize);
@@ -3560,7 +3568,7 @@ static int ahash_update_ctx(struct ahash_request *req)
 		}
 
 		/* allocate space for base edesc and link tables */
-		edesc = qi_cache_zalloc(GFP_DMA | flags);
+		edesc = qi_cache_zalloc(flags);
 		if (!edesc) {
 			dma_unmap_sg(ctx->dev, req->src, src_nents,
 				     DMA_TO_DEVICE);
@@ -3654,7 +3662,7 @@ static int ahash_final_ctx(struct ahash_request *req)
 	int ret;
 
 	/* allocate space for base edesc and link tables */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (!edesc)
 		return -ENOMEM;
 
@@ -3743,7 +3751,7 @@ static int ahash_finup_ctx(struct ahash_request *req)
 	}
 
 	/* allocate space for base edesc and link tables */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (!edesc) {
 		dma_unmap_sg(ctx->dev, req->src, src_nents, DMA_TO_DEVICE);
 		return -ENOMEM;
@@ -3836,7 +3844,7 @@ static int ahash_digest(struct ahash_request *req)
 	}
 
 	/* allocate space for base edesc and link tables */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (!edesc) {
 		dma_unmap_sg(ctx->dev, req->src, src_nents, DMA_TO_DEVICE);
 		return ret;
@@ -3913,7 +3921,7 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 	int ret = -ENOMEM;
 
 	/* allocate space for base edesc and link tables */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (!edesc)
 		return ret;
 
@@ -4012,7 +4020,7 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 		}
 
 		/* allocate space for base edesc and link tables */
-		edesc = qi_cache_zalloc(GFP_DMA | flags);
+		edesc = qi_cache_zalloc(flags);
 		if (!edesc) {
 			dma_unmap_sg(ctx->dev, req->src, src_nents,
 				     DMA_TO_DEVICE);
@@ -4125,7 +4133,7 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 	}
 
 	/* allocate space for base edesc and link tables */
-	edesc = qi_cache_zalloc(GFP_DMA | flags);
+	edesc = qi_cache_zalloc(flags);
 	if (!edesc) {
 		dma_unmap_sg(ctx->dev, req->src, src_nents, DMA_TO_DEVICE);
 		return ret;
@@ -4230,7 +4238,7 @@ static int ahash_update_first(struct ahash_request *req)
 		}
 
 		/* allocate space for base edesc and link tables */
-		edesc = qi_cache_zalloc(GFP_DMA | flags);
+		edesc = qi_cache_zalloc(flags);
 		if (!edesc) {
 			dma_unmap_sg(ctx->dev, req->src, src_nents,
 				     DMA_TO_DEVICE);
@@ -4926,6 +4934,7 @@ static int dpaa2_dpseci_congestion_setup(struct dpaa2_caam_priv *priv,
 {
 	struct dpseci_congestion_notification_cfg cong_notif_cfg = { 0 };
 	struct device *dev = priv->dev;
+	unsigned int alignmask;
 	int err;
 
 	/*
@@ -4936,13 +4945,14 @@ static int dpaa2_dpseci_congestion_setup(struct dpaa2_caam_priv *priv,
 	    !(priv->dpseci_attr.options & DPSECI_OPT_HAS_CG))
 		return 0;
 
-	priv->cscn_mem = kzalloc(DPAA2_CSCN_SIZE + DPAA2_CSCN_ALIGN,
-				 GFP_KERNEL | GFP_DMA);
+	alignmask = DPAA2_CSCN_ALIGN - 1;
+	alignmask |= dma_get_cache_alignment() - 1;
+	priv->cscn_mem = kzalloc(ALIGN(DPAA2_CSCN_SIZE, alignmask + 1),
+				 GFP_KERNEL);
 	if (!priv->cscn_mem)
 		return -ENOMEM;
 
-	priv->cscn_mem_aligned = PTR_ALIGN(priv->cscn_mem, DPAA2_CSCN_ALIGN);
-	priv->cscn_dma = dma_map_single(dev, priv->cscn_mem_aligned,
+	priv->cscn_dma = dma_map_single(dev, priv->cscn_mem,
 					DPAA2_CSCN_SIZE, DMA_FROM_DEVICE);
 	if (dma_mapping_error(dev, priv->cscn_dma)) {
 		dev_err(dev, "Error mapping CSCN memory area\n");
@@ -5174,7 +5184,7 @@ static int dpaa2_caam_probe(struct fsl_mc_device *dpseci_dev)
 	priv->domain = iommu_get_domain_for_dev(dev);
 
 	qi_cache = kmem_cache_create("dpaa2_caamqicache", CAAM_QI_MEMCACHE_SIZE,
-				     0, SLAB_CACHE_DMA, NULL);
+				     0, 0, NULL);
 	if (!qi_cache) {
 		dev_err(dev, "Can't allocate SEC cache\n");
 		return -ENOMEM;
@@ -5451,7 +5461,7 @@ int dpaa2_caam_enqueue(struct device *dev, struct caam_request *req)
 		dma_sync_single_for_cpu(priv->dev, priv->cscn_dma,
 					DPAA2_CSCN_SIZE,
 					DMA_FROM_DEVICE);
-		if (unlikely(dpaa2_cscn_state_congested(priv->cscn_mem_aligned))) {
+		if (unlikely(dpaa2_cscn_state_congested(priv->cscn_mem))) {
 			dev_dbg_ratelimited(dev, "Dropping request\n");
 			return -EBUSY;
 		}

drivers/crypto/caam/caamalg_qi2.h

@@ -7,13 +7,14 @@
 #ifndef _CAAMALG_QI2_H_
 #define _CAAMALG_QI2_H_
 
+#include <crypto/internal/skcipher.h>
+#include <linux/compiler_attributes.h>
 #include <soc/fsl/dpaa2-io.h>
 #include <soc/fsl/dpaa2-fd.h>
 #include <linux/threads.h>
 #include <linux/netdevice.h>
 #include "dpseci.h"
 #include "desc_constr.h"
-#include <crypto/skcipher.h>
 
 #define DPAA2_CAAM_STORE_SIZE	16
 /* NAPI weight *must* be a multiple of the store size. */
@@ -36,8 +37,6 @@
  * @tx_queue_attr: array of Tx queue attributes
  * @cscn_mem: pointer to memory region containing the congestion SCN
  *	it's size is larger than to accommodate alignment
- * @cscn_mem_aligned: pointer to congestion SCN; it is computed as
- *	PTR_ALIGN(cscn_mem, DPAA2_CSCN_ALIGN)
  * @cscn_dma: dma address used by the QMAN to write CSCN messages
  * @dev: device associated with the DPSECI object
  * @mc_io: pointer to MC portal's I/O object
@@ -58,7 +57,6 @@ struct dpaa2_caam_priv {
 
 	/* congestion */
 	void *cscn_mem;
-	void *cscn_mem_aligned;
 	dma_addr_t cscn_dma;
 
 	struct device *dev;
@@ -158,7 +156,7 @@ struct ahash_edesc {
 struct caam_flc {
 	u32 flc[16];
 	u32 sh_desc[MAX_SDLEN];
-} ____cacheline_aligned;
+} __aligned(CRYPTO_DMA_ALIGN);
 
 enum optype {
 	ENCRYPT = 0,
@@ -180,7 +178,7 @@ enum optype {
  * @edesc: extended descriptor; points to one of {skcipher,aead}_edesc
  */
 struct caam_request {
-	struct dpaa2_fl_entry fd_flt[2];
+	struct dpaa2_fl_entry fd_flt[2] __aligned(CRYPTO_DMA_ALIGN);
 	dma_addr_t fd_flt_dma;
 	struct caam_flc *flc;
 	dma_addr_t flc_dma;

drivers/crypto/caam/caamhash.c

@@ -66,6 +66,8 @@
 #include "key_gen.h"
 #include "caamhash_desc.h"
 #include <crypto/engine.h>
+#include <linux/dma-mapping.h>
+#include <linux/kernel.h>
 
 #define CAAM_CRA_PRIORITY		3000
 
@@ -365,7 +367,7 @@ static int hash_digest_key(struct caam_hash_ctx *ctx, u32 *keylen, u8 *key,
 	dma_addr_t key_dma;
 	int ret;
 
-	desc = kmalloc(CAAM_CMD_SZ * 8 + CAAM_PTR_SZ * 2, GFP_KERNEL | GFP_DMA);
+	desc = kmalloc(CAAM_CMD_SZ * 8 + CAAM_PTR_SZ * 2, GFP_KERNEL);
 	if (!desc) {
 		dev_err(jrdev, "unable to allocate key input memory\n");
 		return -ENOMEM;
@@ -432,7 +434,13 @@ static int ahash_setkey(struct crypto_ahash *ahash,
 	dev_dbg(jrdev, "keylen %d\n", keylen);
 
 	if (keylen > blocksize) {
-		hashed_key = kmemdup(key, keylen, GFP_KERNEL | GFP_DMA);
+		unsigned int aligned_len =
+			ALIGN(keylen, dma_get_cache_alignment());
+
+		if (aligned_len < keylen)
+			return -EOVERFLOW;
+
+		hashed_key = kmemdup(key, keylen, GFP_KERNEL);
 		if (!hashed_key)
 			return -ENOMEM;
 		ret = hash_digest_key(ctx, &keylen, hashed_key, digestsize);
@@ -702,7 +710,7 @@ static struct ahash_edesc *ahash_edesc_alloc(struct ahash_request *req,
 	struct ahash_edesc *edesc;
 	unsigned int sg_size = sg_num * sizeof(struct sec4_sg_entry);
 
-	edesc = kzalloc(sizeof(*edesc) + sg_size, GFP_DMA | flags);
+	edesc = kzalloc(sizeof(*edesc) + sg_size, flags);
 	if (!edesc) {
 		dev_err(ctx->jrdev, "could not allocate extended descriptor\n");
 		return NULL;