Add crypto_engine support for AEAD algorithms, to make use of
the engine queue.
The requests, with backlog flag, will be listed into crypto-engine
queue and processed by CAAM when free.
If sending just the backlog request to crypto-engine, and non-blocking
directly to CAAM, the latter requests have a better chance to be
executed since JR has up to 1024 entries, more than the 10 entries
from crypto-engine.
Signed-off-by: Iuliana Prodan <[email protected]>
---
drivers/crypto/caam/caamalg.c | 107 ++++++++++++++++++++++++++++++------------
1 file changed, 77 insertions(+), 30 deletions(-)
diff --git a/drivers/crypto/caam/caamalg.c b/drivers/crypto/caam/caamalg.c
index 5aa01e3..03797f9 100644
--- a/drivers/crypto/caam/caamalg.c
+++ b/drivers/crypto/caam/caamalg.c
@@ -120,6 +120,10 @@ struct caam_skcipher_req_ctx {
struct skcipher_edesc *edesc;
};
+struct caam_aead_req_ctx {
+ struct aead_edesc *edesc;
+};
+
static int aead_null_set_sh_desc(struct crypto_aead *aead)
{
struct caam_ctx *ctx = crypto_aead_ctx(aead);
@@ -864,6 +868,7 @@ static int xts_skcipher_setkey(struct crypto_skcipher *skcipher, const u8 *key,
* @mapped_src_nents: number of segments in input h/w link table
* @mapped_dst_nents: number of segments in output h/w link table
* @sec4_sg_bytes: length of dma mapped sec4_sg space
+ * @bklog: stored to determine if the request needs backlog
* @sec4_sg_dma: bus physical mapped address of h/w link table
* @sec4_sg: pointer to h/w link table
* @hw_desc: the h/w job descriptor followed by any referenced link tables
@@ -874,6 +879,7 @@ struct aead_edesc {
int mapped_src_nents;
int mapped_dst_nents;
int sec4_sg_bytes;
+ bool bklog;
dma_addr_t sec4_sg_dma;
struct sec4_sg_entry *sec4_sg;
u32 hw_desc[];
@@ -953,12 +959,14 @@ static void aead_crypt_done(struct device *jrdev, u32 *desc, u32 err,
void *context)
{
struct aead_request *req = context;
+ struct caam_aead_req_ctx *rctx = aead_request_ctx(req);
+ struct caam_drv_private_jr *jrp = dev_get_drvdata(jrdev);
struct aead_edesc *edesc;
int ecode = 0;
dev_dbg(jrdev, "%s %d: err 0x%x\n", __func__, __LINE__, err);
- edesc = container_of(desc, struct aead_edesc, hw_desc[0]);
+ edesc = rctx->edesc;
if (err)
ecode = caam_jr_strstatus(jrdev, err);
@@ -967,7 +975,14 @@ static void aead_crypt_done(struct device *jrdev, u32 *desc, u32 err,
kfree(edesc);
- aead_request_complete(req, ecode);
+ /*
+ * If no backlog flag, the completion of the request is done
+ * by CAAM, not crypto engine.
+ */
+ if (!edesc->bklog)
+ aead_request_complete(req, ecode);
+ else
+ crypto_finalize_aead_request(jrp->engine, req, ecode);
}
static void skcipher_crypt_done(struct device *jrdev, u32 *desc, u32 err,
@@ -1262,6 +1277,7 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
struct crypto_aead *aead = crypto_aead_reqtfm(req);
struct caam_ctx *ctx = crypto_aead_ctx(aead);
struct device *jrdev = ctx->jrdev;
+ struct caam_aead_req_ctx *rctx = aead_request_ctx(req);
gfp_t flags = (req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP) ?
GFP_KERNEL : GFP_ATOMIC;
int src_nents, mapped_src_nents, dst_nents = 0, mapped_dst_nents = 0;
@@ -1362,6 +1378,9 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
edesc->mapped_dst_nents = mapped_dst_nents;
edesc->sec4_sg = (void *)edesc + sizeof(struct aead_edesc) +
desc_bytes;
+
+ rctx->edesc = edesc;
+
*all_contig_ptr = !(mapped_src_nents > 1);
sec4_sg_index = 0;
@@ -1392,6 +1411,33 @@ static struct aead_edesc *aead_edesc_alloc(struct aead_request *req,
return edesc;
}
+static int aead_enqueue_req(struct device *jrdev, struct aead_request *req)
+{
+ struct caam_drv_private_jr *jrpriv = dev_get_drvdata(jrdev);
+ struct caam_aead_req_ctx *rctx = aead_request_ctx(req);
+ struct aead_edesc *edesc = rctx->edesc;
+ u32 *desc = edesc->hw_desc;
+ int ret;
+
+ /*
+ * Only the backlog request are sent to crypto-engine since the others
+ * can be handled by CAAM, if free, especially since JR has up to 1024
+ * entries (more than the 10 entries from crypto-engine).
+ */
+ if (req->base.flags & CRYPTO_TFM_REQ_MAY_BACKLOG)
+ ret = crypto_transfer_aead_request_to_engine(jrpriv->engine,
+ req);
+ else
+ ret = caam_jr_enqueue(jrdev, desc, aead_crypt_done, req);
+
+ if ((ret != -EINPROGRESS) && (ret != -EBUSY)) {
+ aead_unmap(jrdev, edesc, req);
+ kfree(rctx->edesc);
+ }
+
+ return ret;
+}
+
static inline int chachapoly_crypt(struct aead_request *req, bool encrypt)
{
struct aead_edesc *edesc;
@@ -1400,7 +1446,6 @@ static inline int chachapoly_crypt(struct aead_request *req, bool encrypt)
struct device *jrdev = ctx->jrdev;
bool all_contig;
u32 *desc;
- int ret;
edesc = aead_edesc_alloc(req, CHACHAPOLY_DESC_JOB_IO_LEN, &all_contig,
encrypt);
@@ -1414,13 +1459,7 @@ static inline int chachapoly_crypt(struct aead_request *req, bool encrypt)
DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc),
1);
- ret = caam_jr_enqueue(jrdev, desc, aead_crypt_done, req);
- if (ret != -EINPROGRESS) {
- aead_unmap(jrdev, edesc, req);
- kfree(edesc);
- }
-
- return ret;
+ return aead_enqueue_req(jrdev, req);
}
static int chachapoly_encrypt(struct aead_request *req)
@@ -1440,8 +1479,6 @@ static inline int aead_crypt(struct aead_request *req, bool encrypt)
struct caam_ctx *ctx = crypto_aead_ctx(aead);
struct device *jrdev = ctx->jrdev;
bool all_contig;
- u32 *desc;
- int ret = 0;
/* allocate extended descriptor */
edesc = aead_edesc_alloc(req, AUTHENC_DESC_JOB_IO_LEN,
@@ -1456,14 +1493,7 @@ static inline int aead_crypt(struct aead_request *req, bool encrypt)
DUMP_PREFIX_ADDRESS, 16, 4, edesc->hw_desc,
desc_bytes(edesc->hw_desc), 1);
- desc = edesc->hw_desc;
- ret = caam_jr_enqueue(jrdev, desc, aead_crypt_done, req);
- if (ret != -EINPROGRESS) {
- aead_unmap(jrdev, edesc, req);
- kfree(edesc);
- }
-
- return ret;
+ return aead_enqueue_req(jrdev, req);
}
static int aead_encrypt(struct aead_request *req)
@@ -1476,6 +1506,28 @@ static int aead_decrypt(struct aead_request *req)
return aead_crypt(req, false);
}
+static int aead_do_one_req(struct crypto_engine *engine, void *areq)
+{
+ struct aead_request *req = aead_request_cast(areq);
+ struct caam_ctx *ctx = crypto_aead_ctx(crypto_aead_reqtfm(req));
+ struct caam_aead_req_ctx *rctx = aead_request_ctx(req);
+ u32 *desc = rctx->edesc->hw_desc;
+ int ret;
+
+ rctx->edesc->bklog = true;
+
+ ret = caam_jr_enqueue(ctx->jrdev, desc, aead_crypt_done, req);
+
+ if (ret != -EINPROGRESS) {
+ aead_unmap(ctx->jrdev, rctx->edesc, req);
+ kfree(rctx->edesc);
+ } else {
+ ret = 0;
+ }
+
+ return ret;
+}
+
static inline int gcm_crypt(struct aead_request *req, bool encrypt)
{
struct aead_edesc *edesc;
@@ -1483,8 +1535,6 @@ static inline int gcm_crypt(struct aead_request *req, bool encrypt)
struct caam_ctx *ctx = crypto_aead_ctx(aead);
struct device *jrdev = ctx->jrdev;
bool all_contig;
- u32 *desc;
- int ret = 0;
/* allocate extended descriptor */
edesc = aead_edesc_alloc(req, GCM_DESC_JOB_IO_LEN, &all_contig,
@@ -1499,14 +1549,7 @@ static inline int gcm_crypt(struct aead_request *req, bool encrypt)
DUMP_PREFIX_ADDRESS, 16, 4, edesc->hw_desc,
desc_bytes(edesc->hw_desc), 1);
- desc = edesc->hw_desc;
- ret = caam_jr_enqueue(jrdev, desc, aead_crypt_done, req);
- if (ret != -EINPROGRESS) {
- aead_unmap(jrdev, edesc, req);
- kfree(edesc);
- }
-
- return ret;
+ return aead_enqueue_req(jrdev, req);
}
static int gcm_encrypt(struct aead_request *req)
@@ -3334,6 +3377,10 @@ static int caam_aead_init(struct crypto_aead *tfm)
container_of(alg, struct caam_aead_alg, aead);
struct caam_ctx *ctx = crypto_aead_ctx(tfm);
+ crypto_aead_set_reqsize(tfm, sizeof(struct caam_aead_req_ctx));
+
+ ctx->enginectx.op.do_one_request = aead_do_one_req;
+
return caam_init_common(ctx, &caam_alg->caam, !caam_alg->caam.nodkp);
}
--
2.1.0
On 2/12/2020 7:56 PM, Iuliana Prodan wrote:
> Add crypto_engine support for AEAD algorithms, to make use of
> the engine queue.
> The requests, with backlog flag, will be listed into crypto-engine
> queue and processed by CAAM when free.
> If sending just the backlog request to crypto-engine, and non-blocking
> directly to CAAM, the latter requests have a better chance to be
> executed since JR has up to 1024 entries, more than the 10 entries
> from crypto-engine.
>
> Signed-off-by: Iuliana Prodan <[email protected]>
Reviewed-by: Horia Geant? <[email protected]>
Thanks,
Horia