Revert "crypto: tegra: Use separate buffer for each host1x command"

This reverts commit f6c3d49e92.

Reason for revert: Optimize implementation and align with upstream.

Bug 4883011

Change-Id: I32fe35daa30ddb5a272dbeee0c3316134e9ca55b
Signed-off-by: Brad Griffis <bgriffis@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3321697
(cherry picked from commit efeb1061bb)
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3322625
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>

drivers/crypto/tegra/tegra-se-aes.c

@@ -42,7 +42,6 @@ struct tegra_aes_reqctx {
 	u32 crypto_config;
 	u32 len;
 	u32 *iv;
-	u32 *cmdbuf;
 };
 
 struct tegra_aead_ctx {
@@ -70,7 +69,6 @@ struct tegra_aead_reqctx {
 	u32 key_id;
 	u32 iv[4];
 	u8 authdata[16];
-	u32 *cmdbuf;
 };
 
 struct tegra_cmac_ctx {
@@ -95,7 +93,6 @@ struct tegra_cmac_reqctx {
 	u32 key_id;
 	u32 *iv;
 	u32 result[CMAC_RESULT_REG_COUNT];
-	u32 *cmdbuf;
 };
 
 /* increment counter (128-bit int) */
@@ -216,7 +213,7 @@ static int tegra234_aes_cfg(u32 alg, bool encrypt)
 static unsigned int tegra_aes_prep_cmd(struct tegra_se *se, struct tegra_aes_reqctx *rctx)
 {
 	unsigned int data_count, res_bits, i = 0, j;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 	dma_addr_t addr = rctx->datbuf.addr;
 
 	data_count = rctx->len / AES_BLOCK_SIZE;
@@ -295,17 +292,16 @@ static int tegra_aes_do_one_req(struct crypto_engine *engine, void *areq)
 
 	/* Prepare the command and submit for execution */
 	cmdlen = tegra_aes_prep_cmd(se, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 
 	/* Copy the result */
 	tegra_aes_update_iv(req, ctx);
 	scatterwalk_map_and_copy(rctx->datbuf.buf, req->dst, 0, req->cryptlen, 1);
 
 	/* Free the buffer */
-	dma_free_coherent(se->dev, rctx->datbuf.size,
+	dma_free_coherent(ctx->se->dev, rctx->datbuf.size,
 			  rctx->datbuf.buf, rctx->datbuf.addr);
 
-	kfree(rctx->cmdbuf);
 	crypto_finalize_skcipher_request(se->engine, req, ret);
 
 	return 0;
@@ -463,13 +459,6 @@ static int tegra_aes_crypt(struct skcipher_request *req, bool encrypt)
 	if (!req->cryptlen)
 		return 0;
 
-	if (ctx->alg == SE_ALG_ECB)
-		req->iv = NULL;
-
-	rctx->cmdbuf = kzalloc(SE_MAX_CMDLEN, GFP_KERNEL);
-	if (!rctx->cmdbuf)
-		return -ENOMEM;
-
 	rctx->encrypt = encrypt;
 	rctx->config = tegra234_aes_cfg(ctx->alg, encrypt);
 	rctx->crypto_config = tegra234_aes_crypto_cfg(ctx->alg, encrypt);
@@ -608,7 +597,7 @@ static struct tegra_se_alg tegra_aes_algs[] = {
 static unsigned int tegra_gmac_prep_cmd(struct tegra_se *se, struct tegra_aead_reqctx *rctx)
 {
 	unsigned int data_count, res_bits, i = 0;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 
 	data_count = (rctx->assoclen / AES_BLOCK_SIZE);
 	res_bits = (rctx->assoclen % AES_BLOCK_SIZE) * 8;
@@ -646,7 +635,7 @@ static unsigned int tegra_gmac_prep_cmd(struct tegra_se *se, struct tegra_aead_r
 static unsigned int tegra_gcm_crypt_prep_cmd(struct tegra_se *se, struct tegra_aead_reqctx *rctx)
 {
 	unsigned int data_count, res_bits, i = 0, j;
-	u32 *cpuvaddr = rctx->cmdbuf, op;
+	u32 *cpuvaddr = se->cmdbuf->addr, op;
 
 	data_count = (rctx->cryptlen / AES_BLOCK_SIZE);
 	res_bits = (rctx->cryptlen % AES_BLOCK_SIZE) * 8;
@@ -766,7 +755,7 @@ static int tegra_gcm_do_gmac(struct tegra_aead_ctx *ctx, struct tegra_aead_reqct
 
 	cmdlen = tegra_gmac_prep_cmd(se, rctx);
 
-	return tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	return tegra_se_host1x_submit(se, cmdlen);
 }
 
 static int tegra_gcm_do_crypt(struct tegra_aead_ctx *ctx, struct tegra_aead_reqctx *rctx)
@@ -783,7 +772,7 @@ static int tegra_gcm_do_crypt(struct tegra_aead_ctx *ctx, struct tegra_aead_reqc
 
 	/* Prepare command and submit */
 	cmdlen = tegra_gcm_crypt_prep_cmd(se, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 	if (ret)
 		return ret;
 
@@ -797,7 +786,7 @@ static int tegra_gcm_do_crypt(struct tegra_aead_ctx *ctx, struct tegra_aead_reqc
 static int tegra_gcm_do_final(struct tegra_aead_ctx *ctx, struct tegra_aead_reqctx *rctx)
 {
 	struct tegra_se *se = ctx->se;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 	int cmdlen, ret, offset;
 
 	rctx->config = tegra234_aes_cfg(SE_ALG_GCM_FINAL, rctx->encrypt);
@@ -806,7 +795,7 @@ static int tegra_gcm_do_final(struct tegra_aead_ctx *ctx, struct tegra_aead_reqc
 
 	/* Prepare command and submit */
 	cmdlen = tegra_gcm_prep_final_cmd(se, cpuvaddr, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 	if (ret)
 		return ret;
 
@@ -851,7 +840,7 @@ static inline int tegra_ccm_check_iv(const u8 *iv)
 static unsigned int tegra_cbcmac_prep_cmd(struct tegra_se *se, struct tegra_aead_reqctx *rctx)
 {
 	unsigned int data_count, i = 0;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 
 	data_count = (rctx->inbuf.size / AES_BLOCK_SIZE) - 1;
 
@@ -884,7 +873,7 @@ static unsigned int tegra_cbcmac_prep_cmd(struct tegra_se *se, struct tegra_aead
 static unsigned int tegra_ctr_prep_cmd(struct tegra_se *se, struct tegra_aead_reqctx *rctx)
 {
 	unsigned int i = 0, j;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 
 	cpuvaddr[i++] = host1x_opcode_setpayload(SE_CRYPTO_CTR_REG_COUNT);
 	cpuvaddr[i++] = se_host1x_opcode_incr_w(se->hw->regs->linear_ctr);
@@ -934,7 +923,7 @@ static int tegra_ccm_do_cbcmac(struct tegra_aead_ctx *ctx, struct tegra_aead_req
 	/* Prepare command and submit */
 	cmdlen = tegra_cbcmac_prep_cmd(se, rctx);
 
-	return tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	return tegra_se_host1x_submit(se, cmdlen);
 }
 
 static int tegra_ccm_set_msg_len(u8 *block, unsigned int msglen, int csize)
@@ -1141,7 +1130,7 @@ static int tegra_ccm_do_ctr(struct tegra_aead_ctx *ctx, struct tegra_aead_reqctx
 
 	/* Prepare command and submit */
 	cmdlen = tegra_ctr_prep_cmd(se, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 	if (ret)
 		return ret;
 
@@ -1242,7 +1231,6 @@ outbuf_err:
 	dma_free_coherent(ctx->se->dev, rctx->inbuf.size,
 			  rctx->inbuf.buf, rctx->inbuf.addr);
 
-	kfree(rctx->cmdbuf);
 	crypto_finalize_aead_request(ctx->se->engine, req, ret);
 
 	return 0;
@@ -1316,7 +1304,6 @@ outbuf_err:
 			  rctx->inbuf.buf, rctx->inbuf.addr);
 
 	/* Finalize the request if there are no errors */
-	kfree(rctx->cmdbuf);
 	crypto_finalize_aead_request(ctx->se->engine, req, ret);
 
 	return 0;
@@ -1447,10 +1434,6 @@ static int tegra_aead_crypt(struct aead_request *req, bool encrypt)
 	struct tegra_aead_ctx *ctx = crypto_aead_ctx(tfm);
 	struct tegra_aead_reqctx *rctx = aead_request_ctx(req);
 
-	rctx->cmdbuf = kzalloc(SE_MAX_CMDLEN, GFP_KERNEL);
-	if (!rctx->cmdbuf)
-		return -ENOMEM;
-
 	rctx->encrypt = encrypt;
 
 	return crypto_transfer_aead_request_to_engine(ctx->se->engine, req);
@@ -1482,7 +1465,7 @@ static int tegra_aead_setkey(struct crypto_aead *tfm,
 static unsigned int tegra_cmac_prep_cmd(struct tegra_se *se, struct tegra_cmac_reqctx *rctx)
 {
 	unsigned int data_count, res_bits = 0, i = 0, j;
-	u32 *cpuvaddr = rctx->cmdbuf, op;
+	u32 *cpuvaddr = se->cmdbuf->addr, op;
 
 	data_count = (rctx->datbuf.size / AES_BLOCK_SIZE);
 
@@ -1617,7 +1600,7 @@ static int tegra_cmac_do_update(struct ahash_request *req)
 
 	cmdlen = tegra_cmac_prep_cmd(se, rctx);
 
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 	/*
 	 * If this is not the final update, copy the intermediate results
 	 * from the registers so that it can be used in the next 'update'
@@ -1662,7 +1645,7 @@ static int tegra_cmac_do_final(struct ahash_request *req)
 
 	/* Prepare command and submit */
 	cmdlen = tegra_cmac_prep_cmd(se, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, cmdlen);
+	ret = tegra_se_host1x_submit(se, cmdlen);
 	if (ret)
 		goto out;
 
@@ -1680,8 +1663,6 @@ out:
 out_free:
 	dma_free_coherent(se->dev, crypto_ahash_blocksize(tfm) * 2,
 			  rctx->residue.buf, rctx->residue.addr);
-	kfree(rctx->cmdbuf);
-
 	return ret;
 }
 
@@ -1792,14 +1773,10 @@ static int tegra_cmac_init(struct ahash_request *req)
 	rctx->task = SHA_FIRST;
 	rctx->blk_size = crypto_ahash_blocksize(tfm);
 
-	rctx->cmdbuf = kzalloc(SE_MAX_CMDLEN, GFP_KERNEL);
-	if (!rctx->cmdbuf)
-		return -ENOMEM;
-
 	rctx->residue.buf = dma_alloc_coherent(se->dev, rctx->blk_size * 2,
 					       &rctx->residue.addr, GFP_KERNEL);
 	if (!rctx->residue.buf)
-		goto cmdbuf_free;
+		return -ENOMEM;
 
 	rctx->residue.size = 0;
 	rctx->datbuf.size = 0;
@@ -1809,11 +1786,6 @@ static int tegra_cmac_init(struct ahash_request *req)
 		writel(0, se->base + se->hw->regs->result + (i * 4));
 
 	return 0;
-
-cmdbuf_free:
-	kfree(rctx->cmdbuf);
-
-	return -ENOMEM;
 }
 
 static int tegra_cmac_setkey(struct crypto_ahash *tfm, const u8 *key,

drivers/crypto/tegra/tegra-se-hash.c

@@ -44,7 +44,6 @@ struct tegra_sha_reqctx {
 	unsigned int blk_size;
 	unsigned int task;
 	u32 key_id;
-	u32 *cmdbuf;
 	u32 result[HASH_RESULT_REG_COUNT];
 	struct ahash_request fallback_req;
 };
@@ -306,7 +305,7 @@ static int tegra_sha_do_update(struct ahash_request *req)
 	struct tegra_sha_ctx *ctx = crypto_ahash_ctx(crypto_ahash_reqtfm(req));
 	struct tegra_sha_reqctx *rctx = ahash_request_ctx(req);
 	unsigned int nblks, nresidue, size, ret;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = ctx->se->cmdbuf->addr;
 
 	nresidue = (req->nbytes + rctx->residue.size) % rctx->blk_size;
 	nblks = (req->nbytes + rctx->residue.size) / rctx->blk_size;
@@ -367,7 +366,7 @@ static int tegra_sha_do_update(struct ahash_request *req)
 
 	size = tegra_sha_prep_cmd(ctx->se, cpuvaddr, rctx);
 
-	ret = tegra_se_host1x_submit(ctx->se, rctx->cmdbuf, size);
+	ret = tegra_se_host1x_submit(ctx->se, size);
 
 	/*
 	 * If this is not the final update, copy the intermediate results
@@ -389,7 +388,7 @@ static int tegra_sha_do_final(struct ahash_request *req)
 	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
 	struct tegra_sha_ctx *ctx = crypto_ahash_ctx(tfm);
 	struct tegra_se *se = ctx->se;
-	u32 *cpuvaddr = rctx->cmdbuf;
+	u32 *cpuvaddr = se->cmdbuf->addr;
 	int size, ret = 0;
 
 	rctx->datbuf.size = rctx->residue.size;
@@ -410,7 +409,7 @@ static int tegra_sha_do_final(struct ahash_request *req)
 	}
 
 	size = tegra_sha_prep_cmd(se, cpuvaddr, rctx);
-	ret = tegra_se_host1x_submit(se, rctx->cmdbuf, size);
+	ret = tegra_se_host1x_submit(se, size);
 	if (ret)
 		goto out;
 
@@ -426,8 +425,6 @@ out_free:
 			  rctx->residue.buf, rctx->residue.addr);
 	dma_free_coherent(se->dev, rctx->digest.size, rctx->digest.buf,
 			  rctx->digest.addr);
-	kfree(rctx->cmdbuf);
-
 	return ret;
 }
 
@@ -552,10 +549,6 @@ static int tegra_sha_init(struct ahash_request *req)
 	rctx->blk_size = crypto_ahash_blocksize(tfm);
 	rctx->digest.size = crypto_ahash_digestsize(tfm);
 
-	rctx->cmdbuf = kzalloc(SE_MAX_CMDLEN, GFP_KERNEL);
-	if (!rctx->cmdbuf)
-		return -ENOMEM;
-
 	rctx->digest.buf = dma_alloc_coherent(se->dev, rctx->digest.size,
 					      &rctx->digest.addr, GFP_KERNEL);
 	if (!rctx->digest.buf)
@@ -572,8 +565,6 @@ resbuf_fail:
 	dma_free_coherent(se->dev, rctx->digest.size, rctx->digest.buf,
 			  rctx->digest.addr);
 digbuf_fail:
-	kfree(rctx->cmdbuf);
-
 	return -ENOMEM;
 }
 

drivers/crypto/tegra/tegra-se-key.c

@@ -115,11 +115,11 @@ static int tegra_key_insert(struct tegra_se *se, const u8 *key,
 			    u32 keylen, u16 slot, u32 alg)
 {
 	const u32 *keyval = (u32 *)key;
-	u32 addr[100], size;
+	u32 *addr = se->cmdbuf->addr, size;
 
 	size = tegra_key_prep_ins_cmd(se, addr, keyval, keylen, slot, alg);
 
-	return tegra_se_host1x_submit(se, addr, size);
+	return tegra_se_host1x_submit(se, size);
 }
 
 void tegra_key_invalidate(struct tegra_se *se, u32 keyid, u32 alg)

drivers/crypto/tegra/tegra-se-main.c

@@ -141,10 +141,9 @@ static struct tegra_se_cmdbuf *tegra_se_host1x_bo_alloc(struct tegra_se *se, ssi
 	return cmdbuf;
 }
 
-int tegra_se_host1x_submit(struct tegra_se *se, u32 *cpuvaddr, u32 size)
+int tegra_se_host1x_submit(struct tegra_se *se, u32 size)
 {
 	struct host1x_job *job;
-	struct tegra_se_cmdbuf *cmdbuf;
 	int ret;
 
 	job = host1x_job_alloc(se->channel, 1, 0, true);
@@ -153,12 +152,6 @@ int tegra_se_host1x_submit(struct tegra_se *se, u32 *cpuvaddr, u32 size)
 		return -ENOMEM;
 	}
 
-	cmdbuf = tegra_se_host1x_bo_alloc(se, SZ_4K);
-	if (!cmdbuf)
-		goto job_put;
-
-	memcpy(cmdbuf->addr, cpuvaddr, size * 4);
-
 	job->syncpt = host1x_syncpt_get(se->syncpt);
 	job->syncpt_incrs = 1;
 	job->client = &se->client;
@@ -167,14 +160,14 @@ int tegra_se_host1x_submit(struct tegra_se *se, u32 *cpuvaddr, u32 size)
 	job->engine_fallback_streamid = se->stream_id;
 	job->engine_streamid_offset = SE_STREAM_ID;
 
-	cmdbuf->words = size;
+	se->cmdbuf->words = size;
 
-	host1x_job_add_gather(job, &cmdbuf->bo, size, 0);
+	host1x_job_add_gather(job, &se->cmdbuf->bo, size, 0);
 
 	ret = host1x_job_pin(job, se->dev);
 	if (ret) {
 		dev_err(se->dev, "failed to pin host1x job\n");
-		goto cmdbuf_put;
+		goto job_put;
 	}
 
 	ret = host1x_job_submit(job);
@@ -191,14 +184,10 @@ int tegra_se_host1x_submit(struct tegra_se *se, u32 *cpuvaddr, u32 size)
 	}
 
 	host1x_job_put(job);
-	tegra_se_cmdbuf_put(&cmdbuf->bo);
-
 	return 0;
 
 job_unpin:
 	host1x_job_unpin(job);
-cmdbuf_put:
-	tegra_se_cmdbuf_put(&cmdbuf->bo);
 job_put:
 	host1x_job_put(job);
 
@@ -225,14 +214,22 @@ static int tegra_se_client_init(struct host1x_client *client)
 
 	se->syncpt_id =  host1x_syncpt_id(se->syncpt);
 
+	se->cmdbuf = tegra_se_host1x_bo_alloc(se, SZ_4K);
+	if (!se->cmdbuf) {
+		ret = -ENOMEM;
+		goto syncpt_put;
+	}
+
 	ret = se->hw->init_alg(se);
 	if (ret) {
 		dev_err(se->dev, "failed to register algorithms\n");
-		goto syncpt_put;
+		goto cmdbuf_put;
 	}
 
 	return 0;
 
+cmdbuf_put:
+	tegra_se_cmdbuf_put(&se->cmdbuf->bo);
 syncpt_put:
 	host1x_syncpt_put(se->syncpt);
 channel_put:
@@ -246,6 +243,7 @@ static int tegra_se_client_deinit(struct host1x_client *client)
 	struct tegra_se *se = container_of(client, struct tegra_se, client);
 
 	se->hw->deinit_alg(se);
+	tegra_se_cmdbuf_put(&se->cmdbuf->bo);
 	host1x_syncpt_put(se->syncpt);
 	host1x_channel_put(se->channel);
 

drivers/crypto/tegra/tegra-se.h

@@ -23,7 +23,6 @@
 #define SE_OWNERSHIP_UID(x)				FIELD_GET(GENMASK(7, 0), x)
 #define TEGRA_GPSE_ID					3
 
-#define SE_MAX_CMDLEN					(100 * 4) /* max 100 commands of 4 bytes each */
 #define SE_STREAM_ID					0x90
 
 #define SE_SHA_CFG					0x4004
@@ -442,6 +441,7 @@ struct tegra_se {
 	const struct tegra_se_hw *hw;
 	struct host1x_client client;
 	struct host1x_channel *channel;
+	struct tegra_se_cmdbuf *cmdbuf;
 	struct crypto_engine *engine;
 	struct host1x_syncpt *syncpt;
 	struct device *dev;
@@ -524,7 +524,7 @@ void tegra_deinit_hash(struct tegra_se *se);
 int tegra_key_submit(struct tegra_se *se, const u8 *key,
 		     u32 keylen, u32 alg, u32 *keyid);
 void tegra_key_invalidate(struct tegra_se *se, u32 keyid, u32 alg);
-int tegra_se_host1x_submit(struct tegra_se *se, u32 *cpuvaddr, u32 size);
+int tegra_se_host1x_submit(struct tegra_se *se, u32 size);
 
 /* HOST1x OPCODES */
 static inline u32 host1x_opcode_setpayload(unsigned int payload)