Re: [PATCH v3 1/5] crypto: aspeed: Add HACE hash driver

From: Christophe JAILLET
Date: Thu Jun 23 2022 - 03:13:56 EST


Le 21/06/2022 à 08:37, Neal Liu a écrit :
Hash and Crypto Engine (HACE) is designed to accelerate the
throughput of hash data digest, encryption, and decryption.

Basically, HACE can be divided into two independently engines
- Hash Engine and Crypto Engine. This patch aims to add HACE
hash engine driver for hash accelerator.

Signed-off-by: Neal Liu <neal_liu-SAlXDmAnmOAqDJ6do+/SaQ@xxxxxxxxxxxxxxxx>
Signed-off-by: Johnny Huang <johnny_huang-SAlXDmAnmOAqDJ6do+/SaQ@xxxxxxxxxxxxxxxx>
---

[...]

+++ b/drivers/crypto/aspeed/Kconfig
@@ -0,0 +1,23 @@
+config CRYPTO_DEV_ASPEED
+ tristate "Support for Aspeed cryptographic engine driver"
+ depends on ARCH_ASPEED
+ help
+ Hash and Crypto Engine (HACE) is designed to accelerate the
+ throughput of hash data digest, encryption and decryption.
+
+ Select y here to have support for the cryptographic driver
+ available on Aspeed SoC.
+
+config CRYPTO_DEV_ASPEED_HACE_HASH
+ bool "Enable ASPEED Hash & Crypto Engine (HACE) hash"

Nit: Sometimes you have ASPEED, sometimes you have Aspeed. (see a few lines above)

[...]

+static int aspeed_ahash_req_update(struct aspeed_hace_dev *hace_dev)
+{
+ struct aspeed_engine_hash *hash_engine = &hace_dev->hash_engine;
+ struct ahash_request *req = hash_engine->req;
+ struct aspeed_sham_reqctx *rctx = ahash_request_ctx(req);
+ aspeed_hace_fn_t resume;
+
+ AHASH_DBG(hace_dev, "\n");
+
+ if (hace_dev->version == AST2600_VERSION) {
+ rctx->cmd |= HASH_CMD_HASH_SRC_SG_CTRL;
+ resume = aspeed_ahash_update_resume_sg;
+
+ } else {
+ resume = aspeed_ahash_update_resume;
+ }
+
+ hash_engine->dma_prepare(hace_dev);

Apparently dma_prepare() can fail. Should there be some error handling here?

+
+ return aspeed_hace_ahash_trigger(hace_dev, resume);
+}
+

[...]

+static int aspeed_hace_probe(struct platform_device *pdev)
+{
+ const struct of_device_id *hace_dev_id;
+ struct aspeed_engine_hash *hash_engine;
+ struct aspeed_hace_dev *hace_dev;
+ struct resource *res;
+ int rc;
+
+ hace_dev = devm_kzalloc(&pdev->dev, sizeof(struct aspeed_hace_dev),
+ GFP_KERNEL);
+ if (!hace_dev)
+ return -ENOMEM;
+
+ hace_dev_id = of_match_device(aspeed_hace_of_matches, &pdev->dev);
+ if (!hace_dev_id) {
+ dev_err(&pdev->dev, "Failed to match hace dev id\n");
+ return -EINVAL;
+ }
+
+ hace_dev->dev = &pdev->dev;
+ hace_dev->version = (unsigned long)hace_dev_id->data;
+ hash_engine = &hace_dev->hash_engine;
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+
+ platform_set_drvdata(pdev, hace_dev);
+
+ /* Initialize crypto hardware engine structure for hash */
+ hace_dev->crypt_engine_hash = crypto_engine_alloc_init(hace_dev->dev,
+ true);

This returns NULL on error and crypto_engine_start() will crash in such a case.

+
+ rc = crypto_engine_start(hace_dev->crypt_engine_hash);
+ if (rc)
+ goto err_engine_hash_start;
+
+ tasklet_init(&hash_engine->done_task, aspeed_hace_hash_done_task,
+ (unsigned long)hace_dev);
+
+ hace_dev->regs = devm_ioremap_resource(&pdev->dev, res);
+ if (!hace_dev->regs) {
+ dev_err(&pdev->dev, "Failed to map resources\n");
+ return -ENOMEM;

I think that all direct returns from here to the end of the function should be "goto err_engine_hash_start;".

+ }
+
+ /* Get irq number and register it */
+ hace_dev->irq = platform_get_irq(pdev, 0);
+ if (!hace_dev->irq) {
+ dev_err(&pdev->dev, "Failed to get interrupt\n");
+ return -ENXIO;
+ }
+
+ rc = devm_request_irq(&pdev->dev, hace_dev->irq, aspeed_hace_irq, 0,
+ dev_name(&pdev->dev), hace_dev);
+ if (rc) {
+ dev_err(&pdev->dev, "Failed to request interrupt\n");
+ return rc;
+ }
+
+ /* Get clk and enable it */
+ hace_dev->clk = devm_clk_get(&pdev->dev, NULL);
+ if (IS_ERR(hace_dev->clk)) {
+ dev_err(&pdev->dev, "Failed to get clk\n");
+ return -ENODEV;
+ }
+
+ rc = clk_prepare_enable(hace_dev->clk);
+ if (rc) {
+ dev_err(&pdev->dev, "Failed to enable clock 0x%x\n", rc);
+ return rc;
+ }
+
+ /* Allocate DMA buffer for hash engine input used */
+ hash_engine->ahash_src_addr =
+ dma_alloc_coherent(&pdev->dev,
+ ASPEED_HASH_SRC_DMA_BUF_LEN,
+ &hash_engine->ahash_src_dma_addr,
+ GFP_KERNEL);

Most of the resources are devm_'ed. Does it make sense to use dmam_alloc_coherent() here to simplify the .remove function?

+ if (!hash_engine->ahash_src_addr) {
+ dev_err(&pdev->dev, "Failed to allocate dma buffer\n");
+ rc = -ENOMEM;
+ goto clk_exit;
+ }
+
+ aspeed_hace_register(hace_dev);
+
+ dev_info(&pdev->dev, "ASPEED Crypto Accelerator successfully registered\n");
+
Nit: Sometimes you have ASPEED, sometimes you have Aspeed.

+ return rc;
+
+clk_exit:
+ clk_disable_unprepare(hace_dev->clk);
+err_engine_hash_start:
+ crypto_engine_exit(hace_dev->crypt_engine_hash);
+
+ return rc;
+}
+

[...]

+MODULE_AUTHOR("Neal Liu <neal_liu-SAlXDmAnmOAqDJ6do+/SaQ@xxxxxxxxxxxxxxxx>");
+MODULE_DESCRIPTION("ASPEED HACE driver Crypto Accelerator");

Nit: Sometimes you have ASPEED, sometimes you have Aspeed.