Re: [PATCH v7 4/5] firmware: imx: add driver for NXP EdgeLock Enclave

[Sascha Hauer]

On Wed, Sep 04, 2024 at 04:21:20PM +0530, Pankaj Gupta wrote:
> NXP hardware IP(s) for secure-enclaves like Edgelock Enclave(ELE),
> are embedded in the SoC to support the features like HSM, SHE & V2X,
> using message based communication interface.
> 
> The secure enclave FW communicates on a dedicated messaging unit(MU)
> based interface(s) with application core, where kernel is running.
> It exists on specific i.MX processors. e.g. i.MX8ULP, i.MX93.
> 
> This patch adds the driver for communication interface to secure-enclave,
> for exchanging messages with NXP secure enclave HW IP(s) like EdgeLock
> Enclave (ELE) from Kernel-space, used by kernel management layers like
> - DM-Crypt.
> 
> Signed-off-by: Pankaj Gupta <pankaj.gupta@xxxxxxx>
> ---
>  drivers/firmware/imx/Kconfig        |  12 +
>  drivers/firmware/imx/Makefile       |   2 +
>  drivers/firmware/imx/ele_base_msg.c | 286 ++++++++++++++++++++
>  drivers/firmware/imx/ele_base_msg.h |  95 +++++++
>  drivers/firmware/imx/ele_common.c   | 306 +++++++++++++++++++++
>  drivers/firmware/imx/ele_common.h   |  51 ++++
>  drivers/firmware/imx/se_ctrl.c      | 515 ++++++++++++++++++++++++++++++++++++
>  drivers/firmware/imx/se_ctrl.h      |  99 +++++++
>  include/linux/firmware/imx/se_api.h |  14 +
>  9 files changed, 1380 insertions(+)
> 
> diff --git a/drivers/firmware/imx/Kconfig b/drivers/firmware/imx/Kconfig
> index 183613f82a11..0f6877a24f0b 100644
> --- a/drivers/firmware/imx/Kconfig
> +++ b/drivers/firmware/imx/Kconfig
> @@ -22,3 +22,15 @@ config IMX_SCU
>  
>  	  This driver manages the IPC interface between host CPU and the
>  	  SCU firmware running on M4.
> +
> +config IMX_SEC_ENCLAVE
> +	tristate "i.MX Embedded Secure Enclave - EdgeLock Enclave Firmware driver."
> +	depends on IMX_MBOX && ARCH_MXC && ARM64
> +	default m if ARCH_MXC
> +
> +	help
> +	  It is possible to use APIs exposed by the iMX Secure Enclave HW IP called:
> +	  - EdgeLock Enclave Firmware (for i.MX8ULP, i.MX93),
> +	    like base, HSM, V2X & SHE using the SAB protocol via the shared Messaging
> +	    Unit. This driver exposes these interfaces via a set of file descriptors
> +	    allowing to configure shared memory, send and receive messages.
> diff --git a/drivers/firmware/imx/Makefile b/drivers/firmware/imx/Makefile
> index 8f9f04a513a8..aa9033e0e9e3 100644
> --- a/drivers/firmware/imx/Makefile
> +++ b/drivers/firmware/imx/Makefile
> @@ -1,3 +1,5 @@
>  # SPDX-License-Identifier: GPL-2.0
>  obj-$(CONFIG_IMX_DSP)		+= imx-dsp.o
>  obj-$(CONFIG_IMX_SCU)		+= imx-scu.o misc.o imx-scu-irq.o rm.o imx-scu-soc.o
> +sec_enclave-objs		= se_ctrl.o ele_common.o ele_base_msg.o
> +obj-${CONFIG_IMX_SEC_ENCLAVE}	+= sec_enclave.o
> diff --git a/drivers/firmware/imx/ele_base_msg.c b/drivers/firmware/imx/ele_base_msg.c
> new file mode 100644
> index 000000000000..e3e570a25e85
> --- /dev/null
> +++ b/drivers/firmware/imx/ele_base_msg.c
> @@ -0,0 +1,286 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +/*
> + * Copyright 2024 NXP
> + */
> +
> +#include <linux/types.h>
> +
> +#include <linux/completion.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/genalloc.h>
> +
> +#include "ele_base_msg.h"
> +#include "ele_common.h"
> +
> +int ele_get_info(struct device *dev, struct ele_dev_info *s_info)

I repeat once again:

The context pointer argument should be struct se_if_priv *.

Do not expect foreign code to pass in a struct device * here from which
you blindly expect that it's the right one.

> +int ele_fetch_soc_info(struct device *dev, u16 *soc_rev, u64 *serial_num)

Also here and all the other functions in this file.

> + *
> + * Header file for the EdgeLock Enclave Base API(s).
> + */
> +
> +#ifndef ELE_BASE_MSG_H
> +#define ELE_BASE_MSG_H
> +
> +#include <linux/device.h>
> +#include <linux/types.h>
> +
> +#define WORD_SZ				i4

Unused.

> +#define ELE_NONE_VAL			0x0
> +
> +#define ELE_GET_INFO_REQ		0xDA
> +#define ELE_GET_INFO_REQ_MSG_SZ		0x10
> +#define ELE_GET_INFO_RSP_MSG_SZ		0x08
> +
> +#define DEFAULT_IMX_SOC_VER		0xA000

Unused

> +#define SOC_VER_MASK			0xFFFF0000

Unused

> +int ele_msg_send(struct se_if_priv *priv,
> +		 void *tx_msg,
> +		 int tx_msg_sz)
> +{
> +	struct se_msg_hdr *header;
> +	int err;
> +
> +	header = tx_msg;
> +
> +	/*
> +	 * Check that the size passed as argument matches the size
> +	 * carried in the message.
> +	 */
> +	if (header->size << 2 != tx_msg_sz) {
> +		err = -EINVAL;
> +		dev_err(priv->dev,
> +			"User buf hdr: 0x%x, sz mismatced with input-sz (%d != %d).",
> +			*(u32 *)header,
> +			header->size << 2, tx_msg_sz);
> +		goto exit;
> +	}
> +	guard(mutex)(&priv->se_if_lock);

Drop this mutex. All it does is to protect mbox_send_message() which
already has its own locking.

> +
> +	err = mbox_send_message(priv->tx_chan, tx_msg);
> +	if (err < 0) {
> +		dev_err(priv->dev, "Error: mbox_send_message failure.\n");
> +		return err;
> +	}
> +	err = tx_msg_sz;
> +
> +exit:
> +	return err;
> +}
> +
> +void se_if_rx_callback(struct mbox_client *mbox_cl, void *msg)
> +{
> +	struct se_clbk_handle *se_clbk_hdl;
> +	struct device *dev = mbox_cl->dev;
> +	struct se_msg_hdr *header;
> +	struct se_if_priv *priv;
> +	u32 rx_msg_sz;
> +
> +	priv = dev_get_drvdata(dev);
> +
> +	/* The function can be called with NULL msg */

You already identified this as a possible case...

> +	if (!msg) {
> +		dev_err(dev, "Message is invalid\n");

...so why print an error message here?

> +		return;
> +	}
> +
> +	header = msg;
> +	rx_msg_sz = header->size << 2;
> +
> +	/* Incoming command: wake up the receiver if any. */
> +	if (header->tag == priv->cmd_tag) {
> +		se_clbk_hdl = &priv->cmd_receiver_clbk_hdl;
> +		dev_dbg(dev,
> +			"Selecting cmd receiver for mesg header:0x%x.",
> +			*(u32 *) header);
> +
> +		/* Pre-allocated buffer of MAX_NVM_MSG_LEN
> +		 * as the NVM command are initiated by FW.
> +		 * Size is revealed as part of this call function.
> +		 */
> +		if (rx_msg_sz > MAX_NVM_MSG_LEN) {
> +			dev_err(dev,
> +				"CMD-RCVER NVM: hdr(0x%x) with different sz(%d != %d).\n",
> +				*(u32 *) header,
> +				rx_msg_sz, se_clbk_hdl->rx_msg_sz);
> +
> +			se_clbk_hdl->rx_msg_sz = MAX_NVM_MSG_LEN;
> +		}
> +		se_clbk_hdl->rx_msg_sz = rx_msg_sz;
> +
> +	} else if (header->tag == priv->rsp_tag) {
> +		se_clbk_hdl = &priv->waiting_rsp_clbk_hdl;
> +		dev_dbg(dev,
> +			"Selecting resp waiter for mesg header:0x%x.",
> +			*(u32 *) header);
> +
> +		if (rx_msg_sz != se_clbk_hdl->rx_msg_sz
> +				&& !exception_for_size(priv, header)) {
> +			dev_err(dev,
> +				"Rsp to CMD: hdr(0x%x) with different sz(%d != %d).\n",
> +				*(u32 *) header,
> +				rx_msg_sz, se_clbk_hdl->rx_msg_sz);
> +
> +			se_clbk_hdl->rx_msg_sz = min(rx_msg_sz, se_clbk_hdl->rx_msg_sz);
> +		}
> +	} else {
> +		dev_err(dev, "Failed to select a device for message: %.8x\n",
> +			*((u32 *) header));
> +		return;
> +	}
> +
> +	memcpy(se_clbk_hdl->rx_msg, msg, se_clbk_hdl->rx_msg_sz);
> +
> +	/* Allow user to read */
> +	atomic_inc(&se_clbk_hdl->pending_hdr);
> +
> +	wake_up_interruptible(&se_clbk_hdl->wq);

You are rebuilding a completion here, why not use a completion then?

> +#include <linux/mod_devicetable.h>
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/platform_device.h>
> +#include <linux/slab.h>
> +#include <linux/string.h>
> +#include <linux/sys_soc.h>
> +
> +#include "ele_base_msg.h"
> +#include "ele_common.h"
> +#include "se_ctrl.h"
> +
> +#define RESERVED_DMA_POOL		BIT(0)

Unused

> +static void se_load_firmware(const struct firmware *fw, void *context)
> +{
> +	struct se_if_priv *priv = context;
> +	const struct se_if_node_info *info = priv->info;
> +	phys_addr_t se_fw_phyaddr;
> +	u8 *se_fw_buf;
> +	int ret;
> +
> +	if (!fw) {
> +		if (priv->fw_fail > MAX_FW_LOAD_RETRIES)
> +			dev_dbg(priv->dev,
> +				 "External FW not found, using ROM FW.\n");
> +		else {
> +			/*add a bit delay to wait for firmware priv released */
> +			msleep(20);
> +
> +			/* Load firmware one more time if timeout */
> +			request_firmware_nowait(THIS_MODULE,
> +					FW_ACTION_UEVENT, priv->se_img_file_to_load,
> +					priv->dev, GFP_KERNEL, priv,
> +					se_load_firmware);
> +			priv->fw_fail++;
> +			dev_dbg(priv->dev, "Value of retries = 0x%x.\n",
> +				priv->fw_fail);
> +		}
> +
> +		return;
> +	}

Are you continuously trying to load the firmware here in the hope that
the rootfs is mounted before your retry counter exceeds?

Don't do this.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |