diff mbox series

[V4,1/2] firmware: ti_sci: fix the secure_hdr in do_xfer

Message ID 20240130115847.860345-2-d-gole@ti.com
State Superseded
Delegated to: Tom Rini
Headers show
Series firmware: ti_sci: zero the secure_hdr in do_xfer | expand

Commit Message

Dhruva Gole Jan. 30, 2024, 11:58 a.m. UTC
The ti_sci driver in U-Boot has support for secure_msg as part of it's
do_xfer function. This let's U-boot send secure messages during boot up.

The protocol to send such secure messages is described as part of the
struct ti_sci_secure_msg_hdr. As part of this, there are 2 fields for
checksum and reserved that occupy the first 4 bytes of any secure
message. This is called as the secure_hdr.

As of now, the secure_hdr needs to be 0 init-ed before sending secure
messages. However the existing code was never putting the zero-inited vars
into the secure_buf, leading to possibility of the first 4 bytes of
secure_buf being possibly garbage.

Fix this by initialising the secure_hdr itself to the secure_buf
location, thus when we make secure_hdr members 0, it automatically ensures
the first 4 bytes of secure_buf are 0.

Fixes: 32cd25128bd849 ("firmware: Add basic support for TI System Control Interface (TI SCI)")
Cc: Nishanth Menon <nm@ti.com>
Cc: Andrew Davis <afd@ti.com>
Cc: Manorit Chawdhry <m-chawdhry@ti.com>
Signed-off-by: Dhruva Gole <d-gole@ti.com>
---
 drivers/firmware/ti_sci.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

Comments

Nishanth Menon Jan. 30, 2024, 12:29 p.m. UTC | #1
On 17:28-20240130, Dhruva Gole wrote:
> The ti_sci driver in U-Boot has support for secure_msg as part of it's
> do_xfer function. This let's U-boot send secure messages during boot up.
> 
> The protocol to send such secure messages is described as part of the
> struct ti_sci_secure_msg_hdr. As part of this, there are 2 fields for
> checksum and reserved that occupy the first 4 bytes of any secure
> message. This is called as the secure_hdr.
> 
> As of now, the secure_hdr needs to be 0 init-ed before sending secure
> messages. However the existing code was never putting the zero-inited vars
> into the secure_buf, leading to possibility of the first 4 bytes of
> secure_buf being possibly garbage.
> 
> Fix this by initialising the secure_hdr itself to the secure_buf
> location, thus when we make secure_hdr members 0, it automatically ensures
> the first 4 bytes of secure_buf are 0.
> 
> Fixes: 32cd25128bd849 ("firmware: Add basic support for TI System Control Interface (TI SCI)")
> Cc: Nishanth Menon <nm@ti.com>
> Cc: Andrew Davis <afd@ti.com>
> Cc: Manorit Chawdhry <m-chawdhry@ti.com>
> Signed-off-by: Dhruva Gole <d-gole@ti.com>
> ---
>  drivers/firmware/ti_sci.c | 12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/firmware/ti_sci.c b/drivers/firmware/ti_sci.c
> index 6e9f93e9a302..49d2696a6d09 100644
> --- a/drivers/firmware/ti_sci.c
> +++ b/drivers/firmware/ti_sci.c
> @@ -236,21 +236,21 @@ static int ti_sci_do_xfer(struct ti_sci_info *info,
>  {
>  	struct k3_sec_proxy_msg *msg = &xfer->tx_message;
>  	u8 secure_buf[info->desc->max_msg_size];
> -	struct ti_sci_secure_msg_hdr secure_hdr;
> +	struct ti_sci_secure_msg_hdr *secure_hdr = (struct ti_sci_secure_msg_hdr *)secure_buf;
>  	int ret;
>  
>  	if (info->is_secure) {
>  		/* ToDo: get checksum of the entire message */
> -		secure_hdr.checksum = 0;
> -		secure_hdr.reserved = 0;

I was thinking originally just adding memcpy(secure_buf, secure_hdr,
sizeof(secure_hdr)) would save all the churn.. but anyways.. we save
allocating secure_hdr struct.. not a big saving, but better code
anyways..

> -		memcpy(&secure_buf[sizeof(secure_hdr)], xfer->tx_message.buf,
> +		secure_hdr->checksum = 0;
> +		secure_hdr->reserved = 0;
> +		memcpy(&secure_buf[sizeof(struct ti_sci_secure_msg_hdr)], xfer->tx_message.buf,
>  		       xfer->tx_message.len);

here and below:

s/sizeof(var)/sizeof(*var) instead of sizeof(struct ... ) is probably
all the change we need? rather than converting it to sizeof(struct ..)?
same below. this would allow (theoretically), that the structure
name to change for secure_hdr and there would be less churn? not that it
matters here.. just a style thing..

Either way:

Reviewed-by: Nishanth Menon <nm@ti.com>

>  
>  		xfer->tx_message.buf = (u32 *)secure_buf;
> -		xfer->tx_message.len += sizeof(secure_hdr);
> +		xfer->tx_message.len += sizeof(struct ti_sci_secure_msg_hdr);
>  
>  		if (xfer->rx_len)
> -			xfer->rx_len += sizeof(secure_hdr);
> +			xfer->rx_len += sizeof(struct ti_sci_secure_msg_hdr);
>  	}
>  
>  	/* Send the message */
> -- 
> 2.34.1
>
diff mbox series

Patch

diff --git a/drivers/firmware/ti_sci.c b/drivers/firmware/ti_sci.c
index 6e9f93e9a302..49d2696a6d09 100644
--- a/drivers/firmware/ti_sci.c
+++ b/drivers/firmware/ti_sci.c
@@ -236,21 +236,21 @@  static int ti_sci_do_xfer(struct ti_sci_info *info,
 {
 	struct k3_sec_proxy_msg *msg = &xfer->tx_message;
 	u8 secure_buf[info->desc->max_msg_size];
-	struct ti_sci_secure_msg_hdr secure_hdr;
+	struct ti_sci_secure_msg_hdr *secure_hdr = (struct ti_sci_secure_msg_hdr *)secure_buf;
 	int ret;
 
 	if (info->is_secure) {
 		/* ToDo: get checksum of the entire message */
-		secure_hdr.checksum = 0;
-		secure_hdr.reserved = 0;
-		memcpy(&secure_buf[sizeof(secure_hdr)], xfer->tx_message.buf,
+		secure_hdr->checksum = 0;
+		secure_hdr->reserved = 0;
+		memcpy(&secure_buf[sizeof(struct ti_sci_secure_msg_hdr)], xfer->tx_message.buf,
 		       xfer->tx_message.len);
 
 		xfer->tx_message.buf = (u32 *)secure_buf;
-		xfer->tx_message.len += sizeof(secure_hdr);
+		xfer->tx_message.len += sizeof(struct ti_sci_secure_msg_hdr);
 
 		if (xfer->rx_len)
-			xfer->rx_len += sizeof(secure_hdr);
+			xfer->rx_len += sizeof(struct ti_sci_secure_msg_hdr);
 	}
 
 	/* Send the message */